Merge branch 'for-linus' of git://git390.marist.edu/pub/scm/linux-2.6

* 'for-linus' of git://git390.marist.edu/pub/scm/linux-2.6: (72 commits)
  [S390] 3215/3270 console: remove wrong comment
  [S390] dasd: remove BKL from extended error reporting code
  [S390] vmlogrdr: remove BKL
  [S390] vmur: remove BKL
  [S390] zcrypt: remove BKL
  [S390] 3270: remove BKL
  [S390] vmwatchdog: remove lock_kernel() from open() function
  [S390] monwriter: remove lock_kernel() from open() function
  [S390] monreader: remove lock_kernel() from open() function
  [S390] s390: remove unused nfsd #includes
  [S390] ftrace: build ftrace.o when CONFIG_FTRACE_SYSCALLS is set for s390
  [S390] etr/stp: put correct per cpu variable
  [S390] tty3270: move keyboard compat ioctls
  [S390] sclp: improve servicability setting
  [S390] s390: use change recording override for kernel mapping
  [S390] MAINTAINERS: Add s390 drivers block
  [S390] use generic sockios.h header file
  [S390] use generic termbits.h header file
  [S390] smp: remove unused typedef and defines
  [S390] cmm: free pages on hibernate.
  ...

MAINTAINERS

@@ -3139,6 +3139,7 @@ S:	Supported
 F:	Documentation/s390/kvm.txt
 F:	arch/s390/include/asm/kvm*
 F:	arch/s390/kvm/
+F:	drivers/s390/kvm/
 
 KEXEC
 M:	Eric Biederman <ebiederm@xmission.com>
@@ -4553,6 +4554,7 @@ L:	linux-s390@vger.kernel.org
 W:	http://www.ibm.com/developerworks/linux/linux390/
 S:	Supported
 F:	arch/s390/
+F:	drivers/s390/
 
 S390 NETWORK DRIVERS
 M:	Ursula Braun <ursula.braun@de.ibm.com>
@@ -4568,6 +4570,7 @@ M:	Felix Beck <felix.beck@de.ibm.com>
 M:	Ralph Wuerthner <ralph.wuerthner@de.ibm.com>
 M:	linux390@de.ibm.com
 L:	linux-s390@vger.kernel.org
+W:	http://www.ibm.com/developerworks/linux/linux390/
 S:	Supported
 F:	drivers/s390/crypto/
 

arch/s390/Kconfig

@@ -220,23 +220,8 @@ config AUDIT_ARCH
 	bool
 	default y
 
-config S390_SWITCH_AMODE
-	bool "Switch kernel/user addressing modes"
-	help
-	  This option allows to switch the addressing modes of kernel and user
-	  space. The kernel parameter switch_amode=on will enable this feature,
-	  default is disabled. Enabling this (via kernel parameter) on machines
-	  earlier than IBM System z9-109 EC/BC will reduce system performance.
-
-	  Note that this option will also be selected by selecting the execute
-	  protection option below. Enabling the execute protection via the
-	  noexec kernel parameter will also switch the addressing modes,
-	  independent of the switch_amode kernel parameter.
-
-
 config S390_EXEC_PROTECT
 	bool "Data execute protection"
-	select S390_SWITCH_AMODE
 	help
 	  This option allows to enable a buffer overflow protection for user
 	  space programs and it also selects the addressing mode option above.

arch/s390/defconfig

@@ -185,7 +185,6 @@ CONFIG_HOTPLUG_CPU=y
 CONFIG_COMPAT=y
 CONFIG_SYSVIPC_COMPAT=y
 CONFIG_AUDIT_ARCH=y
-CONFIG_S390_SWITCH_AMODE=y
 CONFIG_S390_EXEC_PROTECT=y
 
 #

arch/s390/include/asm/atomic.h

@@ -21,7 +21,7 @@
 #if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ > 2)
 
 #define __CS_LOOP(ptr, op_val, op_string) ({				\
-	typeof(ptr->counter) old_val, new_val;				\
+	int old_val, new_val;						\
 	asm volatile(							\
 		"	l	%0,%2\n"				\
 		"0:	lr	%1,%0\n"				\
@@ -38,7 +38,7 @@
 #else /* __GNUC__ */
 
 #define __CS_LOOP(ptr, op_val, op_string) ({				\
-	typeof(ptr->counter) old_val, new_val;				\
+	int old_val, new_val;						\
 	asm volatile(							\
 		"	l	%0,0(%3)\n"				\
 		"0:	lr	%1,%0\n"				\
@@ -143,7 +143,7 @@ static inline int atomic_add_unless(atomic_t *v, int a, int u)
 #if __GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ > 2)
 
 #define __CSG_LOOP(ptr, op_val, op_string) ({				\
-	typeof(ptr->counter) old_val, new_val;				\
+	long long old_val, new_val;					\
 	asm volatile(							\
 		"	lg	%0,%2\n"				\
 		"0:	lgr	%1,%0\n"				\
@@ -160,7 +160,7 @@ static inline int atomic_add_unless(atomic_t *v, int a, int u)
 #else /* __GNUC__ */
 
 #define __CSG_LOOP(ptr, op_val, op_string) ({				\
-	typeof(ptr->counter) old_val, new_val;				\
+	long long old_val, new_val;					\
 	asm volatile(							\
 		"	lg	%0,0(%3)\n"				\
 		"0:	lgr	%1,%0\n"				\

arch/s390/include/asm/ccwdev.h

@@ -142,6 +142,8 @@ struct ccw1;
 extern int ccw_device_set_options_mask(struct ccw_device *, unsigned long);
 extern int ccw_device_set_options(struct ccw_device *, unsigned long);
 extern void ccw_device_clear_options(struct ccw_device *, unsigned long);
+int ccw_device_is_pathgroup(struct ccw_device *cdev);
+int ccw_device_is_multipath(struct ccw_device *cdev);
 
 /* Allow for i/o completion notification after primary interrupt status. */
 #define CCWDEV_EARLY_NOTIFICATION	0x0001
@@ -151,6 +153,8 @@ extern void ccw_device_clear_options(struct ccw_device *, unsigned long);
 #define CCWDEV_DO_PATHGROUP             0x0004
 /* Allow forced onlining of boxed devices. */
 #define CCWDEV_ALLOW_FORCE              0x0008
+/* Try to use multipath mode. */
+#define CCWDEV_DO_MULTIPATH		0x0010
 
 extern int ccw_device_start(struct ccw_device *, struct ccw1 *,
 			    unsigned long, __u8, unsigned long);

arch/s390/include/asm/mmu_context.h

@@ -36,7 +36,7 @@ static inline int init_new_context(struct task_struct *tsk,
 		mm->context.has_pgste = 1;
 		mm->context.alloc_pgste = 1;
 	} else {
-		mm->context.noexec = s390_noexec;
+		mm->context.noexec = (user_mode == SECONDARY_SPACE_MODE);
 		mm->context.has_pgste = 0;
 		mm->context.alloc_pgste = 0;
 	}
@@ -58,7 +58,7 @@ static inline void update_mm(struct mm_struct *mm, struct task_struct *tsk)
 	pgd_t *pgd = mm->pgd;
 
 	S390_lowcore.user_asce = mm->context.asce_bits | __pa(pgd);
-	if (switch_amode) {
+	if (user_mode != HOME_SPACE_MODE) {
 		/* Load primary space page table origin. */
 		pgd = mm->context.noexec ? get_shadow_table(pgd) : pgd;
 		S390_lowcore.user_exec_asce = mm->context.asce_bits | __pa(pgd);

arch/s390/include/asm/pgalloc.h

@@ -143,7 +143,8 @@ static inline pgd_t *pgd_alloc(struct mm_struct *mm)
 	spin_lock_init(&mm->context.list_lock);
 	INIT_LIST_HEAD(&mm->context.crst_list);
 	INIT_LIST_HEAD(&mm->context.pgtable_list);
-	return (pgd_t *) crst_table_alloc(mm, s390_noexec);
+	return (pgd_t *)
+		crst_table_alloc(mm, user_mode == SECONDARY_SPACE_MODE);
 }
 #define pgd_free(mm, pgd) crst_table_free(mm, (unsigned long *) pgd)
 

arch/s390/include/asm/pgtable.h

@@ -169,12 +169,13 @@ extern unsigned long VMALLOC_START;
  * STL Segment-Table-Length:  Segment-table length (STL+1*16 entries -> up to 2048)
  *
  * A 64 bit pagetable entry of S390 has following format:
- * |                     PFRA                         |0IP0|  OS  |
+ * |			 PFRA			      |0IPC|  OS  |
  * 0000000000111111111122222222223333333333444444444455555555556666
  * 0123456789012345678901234567890123456789012345678901234567890123
  *
  * I Page-Invalid Bit:    Page is not available for address-translation
  * P Page-Protection Bit: Store access not possible for page
+ * C Change-bit override: HW is not required to set change bit
  *
  * A 64 bit segmenttable entry of S390 has following format:
  * |        P-table origin                              |      TT
@@ -218,6 +219,7 @@ extern unsigned long VMALLOC_START;
  */
 
 /* Hardware bits in the page table entry */
+#define _PAGE_CO	0x100		/* HW Change-bit override */
 #define _PAGE_RO	0x200		/* HW read-only bit  */
 #define _PAGE_INVALID	0x400		/* HW invalid bit    */
 

arch/s390/include/asm/setup.h

@@ -49,17 +49,12 @@ extern unsigned long memory_end;
 
 void detect_memory_layout(struct mem_chunk chunk[]);
 
-#ifdef CONFIG_S390_SWITCH_AMODE
-extern unsigned int switch_amode;
-#else
-#define switch_amode	(0)
-#endif
-
-#ifdef CONFIG_S390_EXEC_PROTECT
-extern unsigned int s390_noexec;
-#else
-#define s390_noexec	(0)
-#endif
+#define PRIMARY_SPACE_MODE	0
+#define ACCESS_REGISTER_MODE	1
+#define SECONDARY_SPACE_MODE	2
+#define HOME_SPACE_MODE		3
+
+extern unsigned int user_mode;
 
 /*
  * Machine features detected in head.S

arch/s390/include/asm/smp.h

@@ -1,57 +1,22 @@
 /*
- *  include/asm-s390/smp.h
- *
- *  S390 version
- *    Copyright (C) 1999 IBM Deutschland Entwicklung GmbH, IBM Corporation
- *    Author(s): Denis Joseph Barrow (djbarrow@de.ibm.com,barrow_dj@yahoo.com),
- *               Martin Schwidefsky (schwidefsky@de.ibm.com)
- *               Heiko Carstens (heiko.carstens@de.ibm.com)
+ *    Copyright IBM Corp. 1999,2009
+ *    Author(s): Denis Joseph Barrow,
+ *		 Martin Schwidefsky <schwidefsky@de.ibm.com>,
+ *		 Heiko Carstens <heiko.carstens@de.ibm.com>,
  */
 #ifndef __ASM_SMP_H
 #define __ASM_SMP_H
 
-#include <linux/threads.h>
-#include <linux/cpumask.h>
-#include <linux/bitops.h>
+#ifdef CONFIG_SMP
 
-#if defined(__KERNEL__) && defined(CONFIG_SMP) && !defined(__ASSEMBLY__)
-
-#include <asm/lowcore.h>
-#include <asm/sigp.h>
-#include <asm/ptrace.h>
 #include <asm/system.h>
-
-/*
-  s390 specific smp.c headers
- */
-typedef struct
-{
-	int        intresting;
-	sigp_ccode ccode; 
-	__u32      status;
-	__u16      cpu;
-} sigp_info;
+#include <asm/sigp.h>
 
 extern void machine_restart_smp(char *);
 extern void machine_halt_smp(void);
 extern void machine_power_off_smp(void);
 
-#define NO_PROC_ID		0xFF		/* No processor magic marker */
-
-/*
- *	This magic constant controls our willingness to transfer
- *	a process across CPUs. Such a transfer incurs misses on the L1
- *	cache, and on a P6 or P5 with multiple L2 caches L2 hits. My
- *	gut feeling is this will vary by board in value. For a board
- *	with separate L2 cache it probably depends also on the RSS, and
- *	for a board with shared L2 cache it ought to decay fast as other
- *	processes are run.
- */
- 
-#define PROC_CHANGE_PENALTY	20		/* Schedule penalty */
-
 #define raw_smp_processor_id()	(S390_lowcore.cpu_nr)
-#define cpu_logical_map(cpu)	(cpu)
 
 extern int __cpu_disable (void);
 extern void __cpu_die (unsigned int cpu);
@@ -64,7 +29,9 @@ extern int smp_cpu_polarization[];
 extern void arch_send_call_function_single_ipi(int cpu);
 extern void arch_send_call_function_ipi_mask(const struct cpumask *mask);
 
-#endif
+extern union save_area *zfcpdump_save_areas[NR_CPUS + 1];
+
+#endif /* CONFIG_SMP */
 
 #ifdef CONFIG_HOTPLUG_CPU
 extern int smp_rescan_cpus(void);
@@ -72,5 +39,4 @@ extern int smp_rescan_cpus(void);
 static inline int smp_rescan_cpus(void) { return 0; }
 #endif
 
-extern union save_area *zfcpdump_save_areas[NR_CPUS + 1];
-#endif
+#endif /* __ASM_SMP_H */

arch/s390/include/asm/sockios.h

@@ -1,21 +1,6 @@
-/*
- *  include/asm-s390/sockios.h
- *
- *  S390 version
- *
- *  Derived from "include/asm-i386/sockios.h"
- */
+#ifndef _ASM_S390_SOCKIOS_H
+#define _ASM_S390_SOCKIOS_H
 
-#ifndef __ARCH_S390_SOCKIOS__
-#define __ARCH_S390_SOCKIOS__
-
-/* Socket-level I/O control calls. */
-#define FIOSETOWN 	0x8901
-#define SIOCSPGRP	0x8902
-#define FIOGETOWN	0x8903
-#define SIOCGPGRP	0x8904
-#define SIOCATMARK	0x8905
-#define SIOCGSTAMP	0x8906		/* Get stamp (timeval) */
-#define SIOCGSTAMPNS	0x8907		/* Get stamp (timespec) */
+#include <asm-generic/sockios.h>
 
 #endif

arch/s390/include/asm/termbits.h

@@ -1,206 +1,6 @@
-/*
- *  include/asm-s390/termbits.h
- *
- *  S390 version
- *
- *  Derived from "include/asm-i386/termbits.h"
- */
+#ifndef _ASM_S390_TERMBITS_H
+#define _ASM_S390_TERMBITS_H
 
-#ifndef __ARCH_S390_TERMBITS_H__
-#define __ARCH_S390_TERMBITS_H__
-
-#include <linux/posix_types.h>
-
-typedef unsigned char	cc_t;
-typedef unsigned int	speed_t;
-typedef unsigned int	tcflag_t;
-
-#define NCCS 19
-struct termios {
-	tcflag_t c_iflag;		/* input mode flags */
-	tcflag_t c_oflag;		/* output mode flags */
-	tcflag_t c_cflag;		/* control mode flags */
-	tcflag_t c_lflag;		/* local mode flags */
-	cc_t c_line;			/* line discipline */
-	cc_t c_cc[NCCS];		/* control characters */
-};
-
-struct termios2 {
-	tcflag_t c_iflag;		/* input mode flags */
-	tcflag_t c_oflag;		/* output mode flags */
-	tcflag_t c_cflag;		/* control mode flags */
-	tcflag_t c_lflag;		/* local mode flags */
-	cc_t c_line;			/* line discipline */
-	cc_t c_cc[NCCS];		/* control characters */
-	speed_t c_ispeed;		/* input speed */
-	speed_t c_ospeed;		/* output speed */
-};
-
-struct ktermios {
-	tcflag_t c_iflag;		/* input mode flags */
-	tcflag_t c_oflag;		/* output mode flags */
-	tcflag_t c_cflag;		/* control mode flags */
-	tcflag_t c_lflag;		/* local mode flags */
-	cc_t c_line;			/* line discipline */
-	cc_t c_cc[NCCS];		/* control characters */
-	speed_t c_ispeed;		/* input speed */
-	speed_t c_ospeed;		/* output speed */
-};
-
-/* c_cc characters */
-#define VINTR 0
-#define VQUIT 1
-#define VERASE 2
-#define VKILL 3
-#define VEOF 4
-#define VTIME 5
-#define VMIN 6
-#define VSWTC 7
-#define VSTART 8
-#define VSTOP 9
-#define VSUSP 10
-#define VEOL 11
-#define VREPRINT 12
-#define VDISCARD 13
-#define VWERASE 14
-#define VLNEXT 15
-#define VEOL2 16
-
-/* c_iflag bits */
-#define IGNBRK	0000001
-#define BRKINT	0000002
-#define IGNPAR	0000004
-#define PARMRK	0000010
-#define INPCK	0000020
-#define ISTRIP	0000040
-#define INLCR	0000100
-#define IGNCR	0000200
-#define ICRNL	0000400
-#define IUCLC	0001000
-#define IXON	0002000
-#define IXANY	0004000
-#define IXOFF	0010000
-#define IMAXBEL	0020000
-#define IUTF8	0040000
-
-/* c_oflag bits */
-#define OPOST	0000001
-#define OLCUC	0000002
-#define ONLCR	0000004
-#define OCRNL	0000010
-#define ONOCR	0000020
-#define ONLRET	0000040
-#define OFILL	0000100
-#define OFDEL	0000200
-#define NLDLY	0000400
-#define   NL0	0000000
-#define   NL1	0000400
-#define CRDLY	0003000
-#define   CR0	0000000
-#define   CR1	0001000
-#define   CR2	0002000
-#define   CR3	0003000
-#define TABDLY	0014000
-#define   TAB0	0000000
-#define   TAB1	0004000
-#define   TAB2	0010000
-#define   TAB3	0014000
-#define   XTABS	0014000
-#define BSDLY	0020000
-#define   BS0	0000000
-#define   BS1	0020000
-#define VTDLY	0040000
-#define   VT0	0000000
-#define   VT1	0040000
-#define FFDLY	0100000
-#define   FF0	0000000
-#define   FF1	0100000
-
-/* c_cflag bit meaning */
-#define CBAUD	0010017
-#define  B0	0000000		/* hang up */
-#define  B50	0000001
-#define  B75	0000002
-#define  B110	0000003
-#define  B134	0000004
-#define  B150	0000005
-#define  B200	0000006
-#define  B300	0000007
-#define  B600	0000010
-#define  B1200	0000011
-#define  B1800	0000012
-#define  B2400	0000013
-#define  B4800	0000014
-#define  B9600	0000015
-#define  B19200	0000016
-#define  B38400	0000017
-#define EXTA B19200
-#define EXTB B38400
-#define CSIZE	0000060
-#define   CS5	0000000
-#define   CS6	0000020
-#define   CS7	0000040
-#define   CS8	0000060
-#define CSTOPB	0000100
-#define CREAD	0000200
-#define PARENB	0000400
-#define PARODD	0001000
-#define HUPCL	0002000
-#define CLOCAL	0004000
-#define CBAUDEX 0010000
-#define  BOTHER  0010000
-#define  B57600  0010001
-#define  B115200 0010002
-#define  B230400 0010003
-#define  B460800 0010004
-#define   B500000 0010005
-#define   B576000 0010006
-#define   B921600 0010007
-#define  B1000000 0010010
-#define  B1152000 0010011
-#define  B1500000 0010012
-#define  B2000000 0010013
-#define  B2500000 0010014
-#define  B3000000 0010015
-#define  B3500000 0010016
-#define  B4000000 0010017
-#define CIBAUD	  002003600000	/* input baud rate */
-#define CMSPAR	  010000000000		/* mark or space (stick) parity */
-#define CRTSCTS	  020000000000		/* flow control */
-
-#define IBSHIFT	  16		/* Shift from CBAUD to CIBAUD */
-
-/* c_lflag bits */
-#define ISIG	0000001
-#define ICANON	0000002
-#define XCASE	0000004
-#define ECHO	0000010
-#define ECHOE	0000020
-#define ECHOK	0000040
-#define ECHONL	0000100
-#define NOFLSH	0000200
-#define TOSTOP	0000400
-#define ECHOCTL	0001000
-#define ECHOPRT	0002000
-#define ECHOKE	0004000
-#define FLUSHO	0010000
-#define PENDIN	0040000
-#define IEXTEN	0100000
-
-/* tcflow() and TCXONC use these */
-#define	TCOOFF		0
-#define	TCOON		1
-#define	TCIOFF		2
-#define	TCION		3
-
-/* tcflush() and TCFLSH use these */
-#define	TCIFLUSH	0
-#define	TCOFLUSH	1
-#define	TCIOFLUSH	2
-
-/* tcsetattr uses these */
-#define	TCSANOW		0
-#define	TCSADRAIN	1
-#define	TCSAFLUSH	2
+#include <asm-generic/termbits.h>
 
 #endif

arch/s390/include/asm/todclk.h

@@ -1,23 +0,0 @@
-/*
- * File...........: linux/include/asm/todclk.h
- * Author(s)......: Holger Smolinski <Holger.Smolinski@de.ibm.com>
- * Bugreports.to..: <Linux390@de.ibm.com>
- * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999,2000
- *
- * History of changes (starts July 2000)
- */
-
-#ifndef __ASM_TODCLK_H
-#define __ASM_TODCLK_H
-
-#ifdef __KERNEL__
-
-#define TOD_uSEC (0x1000ULL)
-#define TOD_mSEC (1000 * TOD_uSEC)
-#define TOD_SEC (1000 * TOD_mSEC)
-#define TOD_MIN (60 * TOD_SEC)
-#define TOD_HOUR (60 * TOD_MIN)
-
-#endif
-
-#endif

arch/s390/include/asm/uaccess.h

@@ -93,6 +93,8 @@ extern struct uaccess_ops uaccess_mvcos;
 extern struct uaccess_ops uaccess_mvcos_switch;
 extern struct uaccess_ops uaccess_pt;
 
+extern int __handle_fault(unsigned long, unsigned long, int);
+
 static inline int __put_user_fn(size_t size, void __user *ptr, void *x)
 {
 	size = uaccess.copy_to_user_small(size, ptr, x);

arch/s390/kernel/Makefile

@@ -44,6 +44,7 @@ obj-$(CONFIG_KPROBES)		+= kprobes.o
 obj-$(CONFIG_FUNCTION_TRACER)	+= $(if $(CONFIG_64BIT),mcount64.o,mcount.o)
 obj-$(CONFIG_DYNAMIC_FTRACE)	+= ftrace.o
 obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += ftrace.o
+obj-$(CONFIG_FTRACE_SYSCALLS)  += ftrace.o
 
 # Kexec part
 S390_KEXEC_OBJS := machine_kexec.o crash.o

arch/s390/kernel/compat_linux.c

@@ -31,14 +31,8 @@
 #include <linux/shm.h>
 #include <linux/slab.h>
 #include <linux/uio.h>
-#include <linux/nfs_fs.h>
 #include <linux/quota.h>
 #include <linux/module.h>
-#include <linux/sunrpc/svc.h>
-#include <linux/nfsd/nfsd.h>
-#include <linux/nfsd/cache.h>
-#include <linux/nfsd/xdr.h>
-#include <linux/nfsd/syscall.h>
 #include <linux/poll.h>
 #include <linux/personality.h>
 #include <linux/stat.h>

arch/s390/kernel/compat_linux.h

@@ -4,10 +4,6 @@
 #include <linux/compat.h>
 #include <linux/socket.h>
 #include <linux/syscalls.h>
-#include <linux/nfs_fs.h>
-#include <linux/sunrpc/svc.h>
-#include <linux/nfsd/nfsd.h>
-#include <linux/nfsd/export.h>
 
 /* Macro that masks the high order bit of an 32 bit pointer and converts it*/
 /*       to a 64 bit pointer */

arch/s390/kernel/head64.S

@@ -83,6 +83,8 @@ startup_continue:
 	slr	%r0,%r0 		# set cpuid to zero
 	sigp	%r1,%r0,0x12		# switch to esame mode
 	sam64				# switch to 64 bit mode
+	llgfr	%r13,%r13		# clear high-order half of base reg
+	lmh	%r0,%r15,.Lzero64-.LPG1(%r13)	# clear high-order half
 	lctlg	%c0,%c15,.Lctl-.LPG1(%r13)	# load control registers
 	lg	%r12,.Lparmaddr-.LPG1(%r13)	# pointer to parameter area
 					# move IPL device to lowcore
@@ -127,6 +129,7 @@ startup_continue:
 .L4malign:.quad 0xffffffffffc00000
 .Lscan2g:.quad	0x80000000 + 0x20000 - 8	# 2GB + 128K - 8
 .Lnop:	.long	0x07000700
+.Lzero64:.fill	16,4,0x0
 #ifdef CONFIG_ZFCPDUMP
 .Lcurrent_cpu:
 	.long 0x0

arch/s390/kernel/setup.c

@@ -305,9 +305,8 @@ static int __init early_parse_mem(char *p)
 }
 early_param("mem", early_parse_mem);
 
-#ifdef CONFIG_S390_SWITCH_AMODE
-unsigned int switch_amode = 0;
-EXPORT_SYMBOL_GPL(switch_amode);
+unsigned int user_mode = HOME_SPACE_MODE;
+EXPORT_SYMBOL_GPL(user_mode);
 
 static int set_amode_and_uaccess(unsigned long user_amode,
 				 unsigned long user32_amode)
@@ -340,23 +339,29 @@ static int set_amode_and_uaccess(unsigned long user_amode,
  */
 static int __init early_parse_switch_amode(char *p)
 {
-	switch_amode = 1;
+	if (user_mode != SECONDARY_SPACE_MODE)
+		user_mode = PRIMARY_SPACE_MODE;
 	return 0;
 }
 early_param("switch_amode", early_parse_switch_amode);
 
-#else /* CONFIG_S390_SWITCH_AMODE */
-static inline int set_amode_and_uaccess(unsigned long user_amode,
-					unsigned long user32_amode)
+static int __init early_parse_user_mode(char *p)
 {
+	if (p && strcmp(p, "primary") == 0)
+		user_mode = PRIMARY_SPACE_MODE;
+#ifdef CONFIG_S390_EXEC_PROTECT
+	else if (p && strcmp(p, "secondary") == 0)
+		user_mode = SECONDARY_SPACE_MODE;
+#endif
+	else if (!p || strcmp(p, "home") == 0)
+		user_mode = HOME_SPACE_MODE;
+	else
+		return 1;
 	return 0;
 }
-#endif /* CONFIG_S390_SWITCH_AMODE */
+early_param("user_mode", early_parse_user_mode);
 
 #ifdef CONFIG_S390_EXEC_PROTECT
-unsigned int s390_noexec = 0;
-EXPORT_SYMBOL_GPL(s390_noexec);
-
 /*
  * Enable execute protection?
  */
@@ -364,8 +369,7 @@ static int __init early_parse_noexec(char *p)
 {
 	if (!strncmp(p, "off", 3))
 		return 0;
-	switch_amode = 1;
-	s390_noexec = 1;
+	user_mode = SECONDARY_SPACE_MODE;
 	return 0;
 }
 early_param("noexec", early_parse_noexec);
@@ -373,7 +377,7 @@ early_param("noexec", early_parse_noexec);
 
 static void setup_addressing_mode(void)
 {
-	if (s390_noexec) {
+	if (user_mode == SECONDARY_SPACE_MODE) {
 		if (set_amode_and_uaccess(PSW_ASC_SECONDARY,
 					  PSW32_ASC_SECONDARY))
 			pr_info("Execute protection active, "
@@ -381,7 +385,7 @@ static void setup_addressing_mode(void)
 		else
 			pr_info("Execute protection active, "
 				"mvcos not available\n");
-	} else if (switch_amode) {
+	} else if (user_mode == PRIMARY_SPACE_MODE) {
 		if (set_amode_and_uaccess(PSW_ASC_PRIMARY, PSW32_ASC_PRIMARY))
 			pr_info("Address spaces switched, "
 				"mvcos available\n");
@@ -411,7 +415,7 @@ setup_lowcore(void)
 	lc->restart_psw.mask = PSW_BASE_BITS | PSW_DEFAULT_KEY;
 	lc->restart_psw.addr =
 		PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
-	if (switch_amode)
+	if (user_mode != HOME_SPACE_MODE)
 		lc->restart_psw.mask |= PSW_ASC_HOME;
 	lc->external_new_psw.mask = psw_kernel_bits;
 	lc->external_new_psw.addr =

arch/s390/kernel/time.c

@@ -335,7 +335,7 @@ int get_sync_clock(unsigned long long *clock)
 	sw0 = atomic_read(sw_ptr);
 	*clock = get_clock();
 	sw1 = atomic_read(sw_ptr);
-	put_cpu_var(clock_sync_sync);
+	put_cpu_var(clock_sync_word);
 	if (sw0 == sw1 && (sw0 & 0x80000000U))
 		/* Success: time is in sync. */
 		return 0;
@@ -385,7 +385,7 @@ static inline int check_sync_clock(void)
 
 	sw_ptr = &get_cpu_var(clock_sync_word);
 	rc = (atomic_read(sw_ptr) & 0x80000000U) != 0;
-	put_cpu_var(clock_sync_sync);
+	put_cpu_var(clock_sync_word);
 	return rc;
 }
 

arch/s390/kernel/vdso.c

@@ -86,7 +86,8 @@ static void vdso_init_data(struct vdso_data *vd)
 	unsigned int facility_list;
 
 	facility_list = stfl();
-	vd->ectg_available = switch_amode && (facility_list & 1);
+	vd->ectg_available =
+		user_mode != HOME_SPACE_MODE && (facility_list & 1);
 }
 
 #ifdef CONFIG_64BIT
@@ -114,7 +115,7 @@ int vdso_alloc_per_cpu(int cpu, struct _lowcore *lowcore)
 
 	lowcore->vdso_per_cpu_data = __LC_PASTE;
 
-	if (!switch_amode || !vdso_enabled)
+	if (user_mode == HOME_SPACE_MODE || !vdso_enabled)
 		return 0;
 
 	segment_table = __get_free_pages(GFP_KERNEL, SEGMENT_ORDER);
@@ -160,7 +161,7 @@ void vdso_free_per_cpu(int cpu, struct _lowcore *lowcore)
 	unsigned long segment_table, page_table, page_frame;
 	u32 *psal, *aste;
 
-	if (!switch_amode || !vdso_enabled)
+	if (user_mode == HOME_SPACE_MODE || !vdso_enabled)
 		return;
 
 	psal = (u32 *)(addr_t) lowcore->paste[4];
@@ -184,7 +185,7 @@ static void __vdso_init_cr5(void *dummy)
 
 static void vdso_init_cr5(void)
 {
-	if (switch_amode && vdso_enabled)
+	if (user_mode != HOME_SPACE_MODE && vdso_enabled)
 		on_each_cpu(__vdso_init_cr5, NULL, 1);
 }
 #endif /* CONFIG_64BIT */

arch/s390/kvm/Kconfig

@@ -20,7 +20,6 @@ config KVM
 	depends on HAVE_KVM && EXPERIMENTAL
 	select PREEMPT_NOTIFIERS
 	select ANON_INODES
-	select S390_SWITCH_AMODE
 	---help---
 	  Support hosting paravirtualized guest machines using the SIE
 	  virtualization capability on the mainframe. This should work

arch/s390/lib/uaccess_mvcos.c

@@ -162,7 +162,6 @@ static size_t clear_user_mvcos(size_t size, void __user *to)
 	return size;
 }
 
-#ifdef CONFIG_S390_SWITCH_AMODE
 static size_t strnlen_user_mvcos(size_t count, const char __user *src)
 {
 	char buf[256];
@@ -200,7 +199,6 @@ static size_t strncpy_from_user_mvcos(size_t count, const char __user *src,
 	} while ((len_str == len) && (done < count));
 	return done;
 }
-#endif /* CONFIG_S390_SWITCH_AMODE */
 
 struct uaccess_ops uaccess_mvcos = {
 	.copy_from_user = copy_from_user_mvcos_check,
@@ -215,7 +213,6 @@ struct uaccess_ops uaccess_mvcos = {
 	.futex_atomic_cmpxchg = futex_atomic_cmpxchg_std,
 };
 
-#ifdef CONFIG_S390_SWITCH_AMODE
 struct uaccess_ops uaccess_mvcos_switch = {
 	.copy_from_user = copy_from_user_mvcos,
 	.copy_from_user_small = copy_from_user_mvcos,
@@ -228,4 +225,3 @@ struct uaccess_ops uaccess_mvcos_switch = {
 	.futex_atomic_op = futex_atomic_op_pt,
 	.futex_atomic_cmpxchg = futex_atomic_cmpxchg_pt,
 };
-#endif

arch/s390/lib/uaccess_pt.c

@@ -23,86 +23,21 @@ static inline pte_t *follow_table(struct mm_struct *mm, unsigned long addr)
 
 	pgd = pgd_offset(mm, addr);
 	if (pgd_none(*pgd) || unlikely(pgd_bad(*pgd)))
-		return NULL;
+		return (pte_t *) 0x3a;
 
 	pud = pud_offset(pgd, addr);
 	if (pud_none(*pud) || unlikely(pud_bad(*pud)))
-		return NULL;
+		return (pte_t *) 0x3b;
 
 	pmd = pmd_offset(pud, addr);
 	if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
-		return NULL;
+		return (pte_t *) 0x10;
 
 	return pte_offset_map(pmd, addr);
 }
 
-static int __handle_fault(struct mm_struct *mm, unsigned long address,
-			  int write_access)
-{
-	struct vm_area_struct *vma;
-	int ret = -EFAULT;
-	int fault;
-
-	if (in_atomic())
-		return ret;
-	down_read(&mm->mmap_sem);
-	vma = find_vma(mm, address);
-	if (unlikely(!vma))
-		goto out;
-	if (unlikely(vma->vm_start > address)) {
-		if (!(vma->vm_flags & VM_GROWSDOWN))
-			goto out;
-		if (expand_stack(vma, address))
-			goto out;
-	}
-
-	if (!write_access) {
-		/* page not present, check vm flags */
-		if (!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE)))
-			goto out;
-	} else {
-		if (!(vma->vm_flags & VM_WRITE))
-			goto out;
-	}
-
-survive:
-	fault = handle_mm_fault(mm, vma, address, write_access ? FAULT_FLAG_WRITE : 0);
-	if (unlikely(fault & VM_FAULT_ERROR)) {
-		if (fault & VM_FAULT_OOM)
-			goto out_of_memory;
-		else if (fault & VM_FAULT_SIGBUS)
-			goto out_sigbus;
-		BUG();
-	}
-	if (fault & VM_FAULT_MAJOR)
-		current->maj_flt++;
-	else
-		current->min_flt++;
-	ret = 0;
-out:
-	up_read(&mm->mmap_sem);
-	return ret;
-
-out_of_memory:
-	up_read(&mm->mmap_sem);
-	if (is_global_init(current)) {
-		yield();
-		down_read(&mm->mmap_sem);
-		goto survive;
-	}
-	printk("VM: killing process %s\n", current->comm);
-	return ret;
-
-out_sigbus:
-	up_read(&mm->mmap_sem);
-	current->thread.prot_addr = address;
-	current->thread.trap_no = 0x11;
-	force_sig(SIGBUS, current);
-	return ret;
-}
-
-static size_t __user_copy_pt(unsigned long uaddr, void *kptr,
-			     size_t n, int write_user)
+static __always_inline size_t __user_copy_pt(unsigned long uaddr, void *kptr,
+					     size_t n, int write_user)
 {
 	struct mm_struct *mm = current->mm;
 	unsigned long offset, pfn, done, size;
@@ -114,12 +49,17 @@ retry:
 	spin_lock(&mm->page_table_lock);
 	do {
 		pte = follow_table(mm, uaddr);
-		if (!pte || !pte_present(*pte) ||
-		    (write_user && !pte_write(*pte)))
+		if ((unsigned long) pte < 0x1000)
 			goto fault;
+		if (!pte_present(*pte)) {
+			pte = (pte_t *) 0x11;
+			goto fault;
+		} else if (write_user && !pte_write(*pte)) {
+			pte = (pte_t *) 0x04;
+			goto fault;
+		}
 
 		pfn = pte_pfn(*pte);
-
 		offset = uaddr & (PAGE_SIZE - 1);
 		size = min(n - done, PAGE_SIZE - offset);
 		if (write_user) {
@@ -137,7 +77,7 @@ retry:
 	return n - done;
 fault:
 	spin_unlock(&mm->page_table_lock);
-	if (__handle_fault(mm, uaddr, write_user))
+	if (__handle_fault(uaddr, (unsigned long) pte, write_user))
 		return n - done;
 	goto retry;
 }
@@ -146,30 +86,31 @@ fault:
  * Do DAT for user address by page table walk, return kernel address.
  * This function needs to be called with current->mm->page_table_lock held.
  */
-static unsigned long __dat_user_addr(unsigned long uaddr)
+static __always_inline unsigned long __dat_user_addr(unsigned long uaddr)
 {
 	struct mm_struct *mm = current->mm;
-	unsigned long pfn, ret;
+	unsigned long pfn;
 	pte_t *pte;
 	int rc;
 
-	ret = 0;
 retry:
 	pte = follow_table(mm, uaddr);
-	if (!pte || !pte_present(*pte))
+	if ((unsigned long) pte < 0x1000)
 		goto fault;
+	if (!pte_present(*pte)) {
+		pte = (pte_t *) 0x11;
+		goto fault;
+	}
 
 	pfn = pte_pfn(*pte);
-	ret = (pfn << PAGE_SHIFT) + (uaddr & (PAGE_SIZE - 1));
-out:
-	return ret;
+	return (pfn << PAGE_SHIFT) + (uaddr & (PAGE_SIZE - 1));
 fault:
 	spin_unlock(&mm->page_table_lock);
-	rc = __handle_fault(mm, uaddr, 0);
+	rc = __handle_fault(uaddr, (unsigned long) pte, 0);
 	spin_lock(&mm->page_table_lock);
-	if (rc)
-		goto out;
-	goto retry;
+	if (!rc)
+		goto retry;
+	return 0;
 }
 
 size_t copy_from_user_pt(size_t n, const void __user *from, void *to)
@@ -234,8 +175,12 @@ retry:
 	spin_lock(&mm->page_table_lock);
 	do {
 		pte = follow_table(mm, uaddr);
-		if (!pte || !pte_present(*pte))
+		if ((unsigned long) pte < 0x1000)
+			goto fault;
+		if (!pte_present(*pte)) {
+			pte = (pte_t *) 0x11;
 			goto fault;
+		}
 
 		pfn = pte_pfn(*pte);
 		offset = uaddr & (PAGE_SIZE-1);
@@ -249,9 +194,8 @@ retry:
 	return done + 1;
 fault:
 	spin_unlock(&mm->page_table_lock);
-	if (__handle_fault(mm, uaddr, 0)) {
+	if (__handle_fault(uaddr, (unsigned long) pte, 0))
 		return 0;
-	}
 	goto retry;
 }
 
@@ -284,7 +228,7 @@ static size_t copy_in_user_pt(size_t n, void __user *to,
 {
 	struct mm_struct *mm = current->mm;
 	unsigned long offset_from, offset_to, offset_max, pfn_from, pfn_to,
-		      uaddr, done, size;
+		      uaddr, done, size, error_code;
 	unsigned long uaddr_from = (unsigned long) from;
 	unsigned long uaddr_to = (unsigned long) to;
 	pte_t *pte_from, *pte_to;
@@ -298,17 +242,28 @@ static size_t copy_in_user_pt(size_t n, void __user *to,
 retry:
 	spin_lock(&mm->page_table_lock);
 	do {
+		write_user = 0;
+		uaddr = uaddr_from;
 		pte_from = follow_table(mm, uaddr_from);
-		if (!pte_from || !pte_present(*pte_from)) {
-			uaddr = uaddr_from;
-			write_user = 0;
+		error_code = (unsigned long) pte_from;
+		if (error_code < 0x1000)
+			goto fault;
+		if (!pte_present(*pte_from)) {
+			error_code = 0x11;
 			goto fault;
 		}
 
+		write_user = 1;
+		uaddr = uaddr_to;
 		pte_to = follow_table(mm, uaddr_to);
-		if (!pte_to || !pte_present(*pte_to) || !pte_write(*pte_to)) {
-			uaddr = uaddr_to;
-			write_user = 1;
+		error_code = (unsigned long) pte_to;
+		if (error_code < 0x1000)
+			goto fault;
+		if (!pte_present(*pte_to)) {
+			error_code = 0x11;
+			goto fault;
+		} else if (!pte_write(*pte_to)) {
+			error_code = 0x04;
 			goto fault;
 		}
 
@@ -329,7 +284,7 @@ retry:
 	return n - done;
 fault:
 	spin_unlock(&mm->page_table_lock);
-	if (__handle_fault(mm, uaddr, write_user))
+	if (__handle_fault(uaddr, error_code, write_user))
 		return n - done;
 	goto retry;
 }

arch/s390/mm/cmm.c

@@ -18,6 +18,7 @@
 #include <linux/swap.h>
 #include <linux/kthread.h>
 #include <linux/oom.h>
+#include <linux/suspend.h>
 
 #include <asm/pgalloc.h>
 #include <asm/uaccess.h>
@@ -44,6 +45,7 @@ static volatile long cmm_pages_target;
 static volatile long cmm_timed_pages_target;
 static long cmm_timeout_pages;
 static long cmm_timeout_seconds;
+static int cmm_suspended;
 
 static struct cmm_page_array *cmm_page_list;
 static struct cmm_page_array *cmm_timed_page_list;
@@ -147,9 +149,9 @@ cmm_thread(void *dummy)
 
 	while (1) {
 		rc = wait_event_interruptible(cmm_thread_wait,
-			(cmm_pages != cmm_pages_target ||
-			 cmm_timed_pages != cmm_timed_pages_target ||
-			 kthread_should_stop()));
+			(!cmm_suspended && (cmm_pages != cmm_pages_target ||
+			 cmm_timed_pages != cmm_timed_pages_target)) ||
+			 kthread_should_stop());
 		if (kthread_should_stop() || rc == -ERESTARTSYS) {
 			cmm_pages_target = cmm_pages;
 			cmm_timed_pages_target = cmm_timed_pages;
@@ -410,6 +412,38 @@ cmm_smsg_target(char *from, char *msg)
 
 static struct ctl_table_header *cmm_sysctl_header;
 
+static int cmm_suspend(void)
+{
+	cmm_suspended = 1;
+	cmm_free_pages(cmm_pages, &cmm_pages, &cmm_page_list);
+	cmm_free_pages(cmm_timed_pages, &cmm_timed_pages, &cmm_timed_page_list);
+	return 0;
+}
+
+static int cmm_resume(void)
+{
+	cmm_suspended = 0;
+	cmm_kick_thread();
+	return 0;
+}
+
+static int cmm_power_event(struct notifier_block *this,
+			   unsigned long event, void *ptr)
+{
+	switch (event) {
+	case PM_POST_HIBERNATION:
+		return cmm_resume();
+	case PM_HIBERNATION_PREPARE:
+		return cmm_suspend();
+	default:
+		return NOTIFY_DONE;
+	}
+}
+
+static struct notifier_block cmm_power_notifier = {
+	.notifier_call = cmm_power_event,
+};
+
 static int
 cmm_init (void)
 {
@@ -418,7 +452,7 @@ cmm_init (void)
 #ifdef CONFIG_CMM_PROC
 	cmm_sysctl_header = register_sysctl_table(cmm_dir_table);
 	if (!cmm_sysctl_header)
-		goto out;
+		goto out_sysctl;
 #endif
 #ifdef CONFIG_CMM_IUCV
 	rc = smsg_register_callback(SMSG_PREFIX, cmm_smsg_target);
@@ -428,17 +462,21 @@ cmm_init (void)
 	rc = register_oom_notifier(&cmm_oom_nb);
 	if (rc < 0)
 		goto out_oom_notify;
+	rc = register_pm_notifier(&cmm_power_notifier);
+	if (rc)
+		goto out_pm;
 	init_waitqueue_head(&cmm_thread_wait);
 	init_timer(&cmm_timer);
 	cmm_thread_ptr = kthread_run(cmm_thread, NULL, "cmmthread");
 	rc = IS_ERR(cmm_thread_ptr) ? PTR_ERR(cmm_thread_ptr) : 0;
-	if (!rc)
-		goto out;
-	/*
-	 * kthread_create failed. undo all the stuff from above again.
-	 */
-	unregister_oom_notifier(&cmm_oom_nb);
+	if (rc)
+		goto out_kthread;
+	return 0;
 
+out_kthread:
+	unregister_pm_notifier(&cmm_power_notifier);
+out_pm:
+	unregister_oom_notifier(&cmm_oom_nb);
 out_oom_notify:
 #ifdef CONFIG_CMM_IUCV
 	smsg_unregister_callback(SMSG_PREFIX, cmm_smsg_target);
@@ -446,8 +484,8 @@ out_smsg:
 #endif
 #ifdef CONFIG_CMM_PROC
 	unregister_sysctl_table(cmm_sysctl_header);
+out_sysctl:
 #endif
-out:
 	return rc;
 }
 
@@ -455,6 +493,7 @@ static void
 cmm_exit(void)
 {
 	kthread_stop(cmm_thread_ptr);
+	unregister_pm_notifier(&cmm_power_notifier);
 	unregister_oom_notifier(&cmm_oom_nb);
 	cmm_free_pages(cmm_pages, &cmm_pages, &cmm_page_list);
 	cmm_free_pages(cmm_timed_pages, &cmm_timed_pages, &cmm_timed_page_list);

arch/s390/mm/fault.c

@@ -34,16 +34,15 @@
 #include <asm/pgtable.h>
 #include <asm/s390_ext.h>
 #include <asm/mmu_context.h>
+#include <asm/compat.h>
 #include "../kernel/entry.h"
 
 #ifndef CONFIG_64BIT
 #define __FAIL_ADDR_MASK 0x7ffff000
-#define __FIXUP_MASK 0x7fffffff
 #define __SUBCODE_MASK 0x0200
 #define __PF_RES_FIELD 0ULL
 #else /* CONFIG_64BIT */
 #define __FAIL_ADDR_MASK -4096L
-#define __FIXUP_MASK ~0L
 #define __SUBCODE_MASK 0x0600
 #define __PF_RES_FIELD 0x8000000000000000ULL
 #endif /* CONFIG_64BIT */
@@ -52,11 +51,15 @@
 extern int sysctl_userprocess_debug;
 #endif
 
-#ifdef CONFIG_KPROBES
-static inline int notify_page_fault(struct pt_regs *regs, long err)
+#define VM_FAULT_BADCONTEXT	0x010000
+#define VM_FAULT_BADMAP		0x020000
+#define VM_FAULT_BADACCESS	0x040000
+
+static inline int notify_page_fault(struct pt_regs *regs)
 {
 	int ret = 0;
 
+#ifdef CONFIG_KPROBES
 	/* kprobe_running() needs smp_processor_id() */
 	if (!user_mode(regs)) {
 		preempt_disable();
@@ -64,15 +67,9 @@ static inline int notify_page_fault(struct pt_regs *regs, long err)
 			ret = 1;
 		preempt_enable();
 	}
-
+#endif
 	return ret;
 }
-#else
-static inline int notify_page_fault(struct pt_regs *regs, long err)
-{
-	return 0;
-}
-#endif
 
 
 /*
@@ -100,57 +97,50 @@ void bust_spinlocks(int yes)
 
 /*
  * Returns the address space associated with the fault.
- * Returns 0 for kernel space, 1 for user space and
- * 2 for code execution in user space with noexec=on.
+ * Returns 0 for kernel space and 1 for user space.
  */
-static inline int check_space(struct task_struct *tsk)
+static inline int user_space_fault(unsigned long trans_exc_code)
 {
 	/*
-	 * The lowest two bits of S390_lowcore.trans_exc_code
-	 * indicate which paging table was used.
+	 * The lowest two bits of the translation exception
+	 * identification indicate which paging table was used.
 	 */
-	int desc = S390_lowcore.trans_exc_code & 3;
-
-	if (desc == 3)	/* Home Segment Table Descriptor */
-		return switch_amode == 0;
-	if (desc == 2)	/* Secondary Segment Table Descriptor */
-		return tsk->thread.mm_segment.ar4;
-#ifdef CONFIG_S390_SWITCH_AMODE
-	if (unlikely(desc == 1)) { /* STD determined via access register */
-		/* %a0 always indicates primary space. */
-		if (S390_lowcore.exc_access_id != 0) {
-			save_access_regs(tsk->thread.acrs);
-			/*
-			 * An alet of 0 indicates primary space.
-			 * An alet of 1 indicates secondary space.
-			 * Any other alet values generate an
-			 * alen-translation exception.
-			 */
-			if (tsk->thread.acrs[S390_lowcore.exc_access_id])
-				return tsk->thread.mm_segment.ar4;
-		}
-	}
-#endif
-	/* Primary Segment Table Descriptor */
-	return switch_amode << s390_noexec;
+	trans_exc_code &= 3;
+	if (trans_exc_code == 2)
+		/* Access via secondary space, set_fs setting decides */
+		return current->thread.mm_segment.ar4;
+	if (user_mode == HOME_SPACE_MODE)
+		/* User space if the access has been done via home space. */
+		return trans_exc_code == 3;
+	/*
+	 * If the user space is not the home space the kernel runs in home
+	 * space. Access via secondary space has already been covered,
+	 * access via primary space or access register is from user space
+	 * and access via home space is from the kernel.
+	 */
+	return trans_exc_code != 3;
 }
 
 /*
  * Send SIGSEGV to task.  This is an external routine
  * to keep the stack usage of do_page_fault small.
  */
-static void do_sigsegv(struct pt_regs *regs, unsigned long error_code,
-		       int si_code, unsigned long address)
+static noinline void do_sigsegv(struct pt_regs *regs, long int_code,
+				int si_code, unsigned long trans_exc_code)
 {
 	struct siginfo si;
+	unsigned long address;
 
+	address = trans_exc_code & __FAIL_ADDR_MASK;
+	current->thread.prot_addr = address;
+	current->thread.trap_no = int_code;
 #if defined(CONFIG_SYSCTL) || defined(CONFIG_PROCESS_DEBUG)
 #if defined(CONFIG_SYSCTL)
 	if (sysctl_userprocess_debug)
 #endif
 	{
 		printk("User process fault: interruption code 0x%lX\n",
-		       error_code);
+		       int_code);
 		printk("failing address: %lX\n", address);
 		show_regs(regs);
 	}
@@ -161,13 +151,14 @@ static void do_sigsegv(struct pt_regs *regs, unsigned long error_code,
 	force_sig_info(SIGSEGV, &si, current);
 }
 
-static void do_no_context(struct pt_regs *regs, unsigned long error_code,
-			  unsigned long address)
+static noinline void do_no_context(struct pt_regs *regs, long int_code,
+				   unsigned long trans_exc_code)
 {
 	const struct exception_table_entry *fixup;
+	unsigned long address;
 
 	/* Are we prepared to handle this kernel fault?  */
-	fixup = search_exception_tables(regs->psw.addr & __FIXUP_MASK);
+	fixup = search_exception_tables(regs->psw.addr & PSW_ADDR_INSN);
 	if (fixup) {
 		regs->psw.addr = fixup->fixup | PSW_ADDR_AMODE;
 		return;
@@ -177,129 +168,149 @@ static void do_no_context(struct pt_regs *regs, unsigned long error_code,
 	 * Oops. The kernel tried to access some bad page. We'll have to
 	 * terminate things with extreme prejudice.
 	 */
-	if (check_space(current) == 0)
+	address = trans_exc_code & __FAIL_ADDR_MASK;
+	if (!user_space_fault(trans_exc_code))
 		printk(KERN_ALERT "Unable to handle kernel pointer dereference"
 		       " at virtual kernel address %p\n", (void *)address);
 	else
 		printk(KERN_ALERT "Unable to handle kernel paging request"
 		       " at virtual user address %p\n", (void *)address);
 
-	die("Oops", regs, error_code);
+	die("Oops", regs, int_code);
 	do_exit(SIGKILL);
 }
 
-static void do_low_address(struct pt_regs *regs, unsigned long error_code)
+static noinline void do_low_address(struct pt_regs *regs, long int_code,
+				    unsigned long trans_exc_code)
 {
 	/* Low-address protection hit in kernel mode means
 	   NULL pointer write access in kernel mode.  */
 	if (regs->psw.mask & PSW_MASK_PSTATE) {
 		/* Low-address protection hit in user mode 'cannot happen'. */
-		die ("Low-address protection", regs, error_code);
+		die ("Low-address protection", regs, int_code);
 		do_exit(SIGKILL);
 	}
 
-	do_no_context(regs, error_code, 0);
+	do_no_context(regs, int_code, trans_exc_code);
 }
 
-static void do_sigbus(struct pt_regs *regs, unsigned long error_code,
-		      unsigned long address)
+static noinline void do_sigbus(struct pt_regs *regs, long int_code,
+			       unsigned long trans_exc_code)
 {
 	struct task_struct *tsk = current;
-	struct mm_struct *mm = tsk->mm;
 
-	up_read(&mm->mmap_sem);
 	/*
 	 * Send a sigbus, regardless of whether we were in kernel
 	 * or user mode.
 	 */
-	tsk->thread.prot_addr = address;
-	tsk->thread.trap_no = error_code;
+	tsk->thread.prot_addr = trans_exc_code & __FAIL_ADDR_MASK;
+	tsk->thread.trap_no = int_code;
 	force_sig(SIGBUS, tsk);
-
-	/* Kernel mode? Handle exceptions or die */
-	if (!(regs->psw.mask & PSW_MASK_PSTATE))
-		do_no_context(regs, error_code, address);
 }
 
 #ifdef CONFIG_S390_EXEC_PROTECT
-static int signal_return(struct mm_struct *mm, struct pt_regs *regs,
-			 unsigned long address, unsigned long error_code)
+static noinline int signal_return(struct pt_regs *regs, long int_code,
+				  unsigned long trans_exc_code)
 {
 	u16 instruction;
 	int rc;
-#ifdef CONFIG_COMPAT
-	int compat;
-#endif
 
-	pagefault_disable();
 	rc = __get_user(instruction, (u16 __user *) regs->psw.addr);
-	pagefault_enable();
-	if (rc)
-		return -EFAULT;
 
-	up_read(&mm->mmap_sem);
-	clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
-#ifdef CONFIG_COMPAT
-	compat = is_compat_task();
-	if (compat && instruction == 0x0a77)
-		sys32_sigreturn();
-	else if (compat && instruction == 0x0aad)
-		sys32_rt_sigreturn();
-	else
-#endif
-	if (instruction == 0x0a77)
-		sys_sigreturn();
-	else if (instruction == 0x0aad)
-		sys_rt_sigreturn();
-	else {
-		current->thread.prot_addr = address;
-		current->thread.trap_no = error_code;
-		do_sigsegv(regs, error_code, SEGV_MAPERR, address);
-	}
+	if (!rc && instruction == 0x0a77) {
+		clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
+		if (is_compat_task())
+			sys32_sigreturn();
+		else
+			sys_sigreturn();
+	} else if (!rc && instruction == 0x0aad) {
+		clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
+		if (is_compat_task())
+			sys32_rt_sigreturn();
+		else
+			sys_rt_sigreturn();
+	} else
+		do_sigsegv(regs, int_code, SEGV_MAPERR, trans_exc_code);
 	return 0;
 }
 #endif /* CONFIG_S390_EXEC_PROTECT */
 
+static noinline void do_fault_error(struct pt_regs *regs, long int_code,
+				    unsigned long trans_exc_code, int fault)
+{
+	int si_code;
+
+	switch (fault) {
+	case VM_FAULT_BADACCESS:
+#ifdef CONFIG_S390_EXEC_PROTECT
+		if ((regs->psw.mask & PSW_MASK_ASC) == PSW_ASC_SECONDARY &&
+		    (trans_exc_code & 3) == 0) {
+			signal_return(regs, int_code, trans_exc_code);
+			break;
+		}
+#endif /* CONFIG_S390_EXEC_PROTECT */
+	case VM_FAULT_BADMAP:
+		/* Bad memory access. Check if it is kernel or user space. */
+		if (regs->psw.mask & PSW_MASK_PSTATE) {
+			/* User mode accesses just cause a SIGSEGV */
+			si_code = (fault == VM_FAULT_BADMAP) ?
+				SEGV_MAPERR : SEGV_ACCERR;
+			do_sigsegv(regs, int_code, si_code, trans_exc_code);
+			return;
+		}
+	case VM_FAULT_BADCONTEXT:
+		do_no_context(regs, int_code, trans_exc_code);
+		break;
+	default: /* fault & VM_FAULT_ERROR */
+		if (fault & VM_FAULT_OOM)
+			pagefault_out_of_memory();
+		else if (fault & VM_FAULT_SIGBUS) {
+			do_sigbus(regs, int_code, trans_exc_code);
+			/* Kernel mode? Handle exceptions or die */
+			if (!(regs->psw.mask & PSW_MASK_PSTATE))
+				do_no_context(regs, int_code, trans_exc_code);
+		} else
+			BUG();
+		break;
+	}
+}
+
 /*
  * This routine handles page faults.  It determines the address,
  * and the problem, and then passes it off to one of the appropriate
  * routines.
  *
- * error_code:
+ * interruption code (int_code):
  *   04       Protection           ->  Write-Protection  (suprression)
  *   10       Segment translation  ->  Not present       (nullification)
  *   11       Page translation     ->  Not present       (nullification)
  *   3b       Region third trans.  ->  Not present       (nullification)
  */
-static inline void
-do_exception(struct pt_regs *regs, unsigned long error_code, int write)
+static inline int do_exception(struct pt_regs *regs, int access,
+			       unsigned long trans_exc_code)
 {
 	struct task_struct *tsk;
 	struct mm_struct *mm;
 	struct vm_area_struct *vma;
 	unsigned long address;
-	int space;
-	int si_code;
 	int fault;
 
-	if (notify_page_fault(regs, error_code))
-		return;
+	if (notify_page_fault(regs))
+		return 0;
 
 	tsk = current;
 	mm = tsk->mm;
 
-	/* get the failing address and the affected space */
-	address = S390_lowcore.trans_exc_code & __FAIL_ADDR_MASK;
-	space = check_space(tsk);
-
 	/*
 	 * Verify that the fault happened in user space, that
 	 * we are not in an interrupt and that there is a 
 	 * user context.
 	 */
-	if (unlikely(space == 0 || in_atomic() || !mm))
-		goto no_context;
+	fault = VM_FAULT_BADCONTEXT;
+	if (unlikely(!user_space_fault(trans_exc_code) || in_atomic() || !mm))
+		goto out;
 
+	address = trans_exc_code & __FAIL_ADDR_MASK;
 	/*
 	 * When we get here, the fault happened in the current
 	 * task's user address space, so we can switch on the
@@ -309,42 +320,26 @@ do_exception(struct pt_regs *regs, unsigned long error_code, int write)
 	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address);
 	down_read(&mm->mmap_sem);
 
-	si_code = SEGV_MAPERR;
+	fault = VM_FAULT_BADMAP;
 	vma = find_vma(mm, address);
 	if (!vma)
-		goto bad_area;
-
-#ifdef CONFIG_S390_EXEC_PROTECT
-	if (unlikely((space == 2) && !(vma->vm_flags & VM_EXEC)))
-		if (!signal_return(mm, regs, address, error_code))
-			/*
-			 * signal_return() has done an up_read(&mm->mmap_sem)
-			 * if it returns 0.
-			 */
-			return;
-#endif
+		goto out_up;
 
-	if (vma->vm_start <= address)
-		goto good_area;
-	if (!(vma->vm_flags & VM_GROWSDOWN))
-		goto bad_area;
-	if (expand_stack(vma, address))
-		goto bad_area;
-/*
- * Ok, we have a good vm_area for this memory access, so
- * we can handle it..
- */
-good_area:
-	si_code = SEGV_ACCERR;
-	if (!write) {
-		/* page not present, check vm flags */
-		if (!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE)))
-			goto bad_area;
-	} else {
-		if (!(vma->vm_flags & VM_WRITE))
-			goto bad_area;
+	if (unlikely(vma->vm_start > address)) {
+		if (!(vma->vm_flags & VM_GROWSDOWN))
+			goto out_up;
+		if (expand_stack(vma, address))
+			goto out_up;
 	}
 
+	/*
+	 * Ok, we have a good vm_area for this memory access, so
+	 * we can handle it..
+	 */
+	fault = VM_FAULT_BADACCESS;
+	if (unlikely(!(vma->vm_flags & access)))
+		goto out_up;
+
 	if (is_vm_hugetlb_page(vma))
 		address &= HPAGE_MASK;
 	/*
@@ -352,18 +347,11 @@ good_area:
 	 * make sure we exit gracefully rather than endlessly redo
 	 * the fault.
 	 */
-	fault = handle_mm_fault(mm, vma, address, write ? FAULT_FLAG_WRITE : 0);
-	if (unlikely(fault & VM_FAULT_ERROR)) {
-		if (fault & VM_FAULT_OOM) {
-			up_read(&mm->mmap_sem);
-			pagefault_out_of_memory();
-			return;
-		} else if (fault & VM_FAULT_SIGBUS) {
-			do_sigbus(regs, error_code, address);
-			return;
-		}
-		BUG();
-	}
+	fault = handle_mm_fault(mm, vma, address,
+				(access == VM_WRITE) ? FAULT_FLAG_WRITE : 0);
+	if (unlikely(fault & VM_FAULT_ERROR))
+		goto out_up;
+
 	if (fault & VM_FAULT_MAJOR) {
 		tsk->maj_flt++;
 		perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0,
@@ -373,74 +361,69 @@ good_area:
 		perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0,
 				     regs, address);
 	}
-        up_read(&mm->mmap_sem);
 	/*
 	 * The instruction that caused the program check will
 	 * be repeated. Don't signal single step via SIGTRAP.
 	 */
 	clear_tsk_thread_flag(tsk, TIF_SINGLE_STEP);
-        return;
-
-/*
- * Something tried to access memory that isn't in our memory map..
- * Fix it, but check if it's kernel or user first..
- */
-bad_area:
+	fault = 0;
+out_up:
 	up_read(&mm->mmap_sem);
-
-	/* User mode accesses just cause a SIGSEGV */
-	if (regs->psw.mask & PSW_MASK_PSTATE) {
-		tsk->thread.prot_addr = address;
-		tsk->thread.trap_no = error_code;
-		do_sigsegv(regs, error_code, si_code, address);
-		return;
-	}
-
-no_context:
-	do_no_context(regs, error_code, address);
+out:
+	return fault;
 }
 
-void __kprobes do_protection_exception(struct pt_regs *regs,
-				       long error_code)
+void __kprobes do_protection_exception(struct pt_regs *regs, long int_code)
 {
+	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
+	int fault;
+
 	/* Protection exception is supressing, decrement psw address. */
-	regs->psw.addr -= (error_code >> 16);
+	regs->psw.addr -= (int_code >> 16);
 	/*
 	 * Check for low-address protection.  This needs to be treated
 	 * as a special case because the translation exception code
 	 * field is not guaranteed to contain valid data in this case.
 	 */
-	if (unlikely(!(S390_lowcore.trans_exc_code & 4))) {
-		do_low_address(regs, error_code);
+	if (unlikely(!(trans_exc_code & 4))) {
+		do_low_address(regs, int_code, trans_exc_code);
 		return;
 	}
-	do_exception(regs, 4, 1);
+	fault = do_exception(regs, VM_WRITE, trans_exc_code);
+	if (unlikely(fault))
+		do_fault_error(regs, 4, trans_exc_code, fault);
 }
 
-void __kprobes do_dat_exception(struct pt_regs *regs, long error_code)
+void __kprobes do_dat_exception(struct pt_regs *regs, long int_code)
 {
-	do_exception(regs, error_code & 0xff, 0);
+	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
+	int access, fault;
+
+	access = VM_READ | VM_EXEC | VM_WRITE;
+#ifdef CONFIG_S390_EXEC_PROTECT
+	if ((regs->psw.mask & PSW_MASK_ASC) == PSW_ASC_SECONDARY &&
+	    (trans_exc_code & 3) == 0)
+		access = VM_EXEC;
+#endif
+	fault = do_exception(regs, access, trans_exc_code);
+	if (unlikely(fault))
+		do_fault_error(regs, int_code & 255, trans_exc_code, fault);
 }
 
 #ifdef CONFIG_64BIT
-void __kprobes do_asce_exception(struct pt_regs *regs, unsigned long error_code)
+void __kprobes do_asce_exception(struct pt_regs *regs, long int_code)
 {
-	struct mm_struct *mm;
+	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
+	struct mm_struct *mm = current->mm;
 	struct vm_area_struct *vma;
-	unsigned long address;
-	int space;
-
-	mm = current->mm;
-	address = S390_lowcore.trans_exc_code & __FAIL_ADDR_MASK;
-	space = check_space(current);
 
-	if (unlikely(space == 0 || in_atomic() || !mm))
+	if (unlikely(!user_space_fault(trans_exc_code) || in_atomic() || !mm))
 		goto no_context;
 
 	local_irq_enable();
 
 	down_read(&mm->mmap_sem);
-	vma = find_vma(mm, address);
+	vma = find_vma(mm, trans_exc_code & __FAIL_ADDR_MASK);
 	up_read(&mm->mmap_sem);
 
 	if (vma) {
@@ -450,17 +433,38 @@ void __kprobes do_asce_exception(struct pt_regs *regs, unsigned long error_code)
 
 	/* User mode accesses just cause a SIGSEGV */
 	if (regs->psw.mask & PSW_MASK_PSTATE) {
-		current->thread.prot_addr = address;
-		current->thread.trap_no = error_code;
-		do_sigsegv(regs, error_code, SEGV_MAPERR, address);
+		do_sigsegv(regs, int_code, SEGV_MAPERR, trans_exc_code);
 		return;
 	}
 
 no_context:
-	do_no_context(regs, error_code, address);
+	do_no_context(regs, int_code, trans_exc_code);
 }
 #endif
 
+int __handle_fault(unsigned long uaddr, unsigned long int_code, int write_user)
+{
+	struct pt_regs regs;
+	int access, fault;
+
+	regs.psw.mask = psw_kernel_bits;
+	if (!irqs_disabled())
+		regs.psw.mask |= PSW_MASK_IO | PSW_MASK_EXT;
+	regs.psw.addr = (unsigned long) __builtin_return_address(0);
+	regs.psw.addr |= PSW_ADDR_AMODE;
+	uaddr &= PAGE_MASK;
+	access = write_user ? VM_WRITE : VM_READ;
+	fault = do_exception(&regs, access, uaddr | 2);
+	if (unlikely(fault)) {
+		if (fault & VM_FAULT_OOM) {
+			pagefault_out_of_memory();
+			fault = 0;
+		} else if (fault & VM_FAULT_SIGBUS)
+			do_sigbus(&regs, int_code, uaddr);
+	}
+	return fault ? -EFAULT : 0;
+}
+
 #ifdef CONFIG_PFAULT 
 /*
  * 'pfault' pseudo page faults routines.
@@ -522,7 +526,7 @@ void pfault_fini(void)
 		: : "a" (&refbk), "m" (refbk) : "cc");
 }
 
-static void pfault_interrupt(__u16 error_code)
+static void pfault_interrupt(__u16 int_code)
 {
 	struct task_struct *tsk;
 	__u16 subcode;

arch/s390/mm/pgtable.c

@@ -269,7 +269,7 @@ int s390_enable_sie(void)
 	struct mm_struct *mm, *old_mm;
 
 	/* Do we have switched amode? If no, we cannot do sie */
-	if (!switch_amode)
+	if (user_mode == HOME_SPACE_MODE)
 		return -EINVAL;
 
 	/* Do we have pgstes? if yes, we are done */

arch/s390/mm/vmem.c

@@ -70,8 +70,12 @@ static pte_t __ref *vmem_pte_alloc(void)
 		pte = alloc_bootmem(PTRS_PER_PTE * sizeof(pte_t));
 	if (!pte)
 		return NULL;
-	clear_table((unsigned long *) pte, _PAGE_TYPE_EMPTY,
-		    PTRS_PER_PTE * sizeof(pte_t));
+	if (MACHINE_HAS_HPAGE)
+		clear_table((unsigned long *) pte, _PAGE_TYPE_EMPTY | _PAGE_CO,
+			    PTRS_PER_PTE * sizeof(pte_t));
+	else
+		clear_table((unsigned long *) pte, _PAGE_TYPE_EMPTY,
+			    PTRS_PER_PTE * sizeof(pte_t));
 	return pte;
 }
 
@@ -112,7 +116,8 @@ static int vmem_add_mem(unsigned long start, unsigned long size, int ro)
 		if (MACHINE_HAS_HPAGE && !(address & ~HPAGE_MASK) &&
 		    (address + HPAGE_SIZE <= start + size) &&
 		    (address >= HPAGE_SIZE)) {
-			pte_val(pte) |= _SEGMENT_ENTRY_LARGE;
+			pte_val(pte) |= _SEGMENT_ENTRY_LARGE |
+					_SEGMENT_ENTRY_CO;
 			pmd_val(*pm_dir) = pte_val(pte);
 			address += HPAGE_SIZE - PAGE_SIZE;
 			continue;

drivers/s390/block/dasd.c

@@ -24,7 +24,6 @@
 #include <asm/ccwdev.h>
 #include <asm/ebcdic.h>
 #include <asm/idals.h>
-#include <asm/todclk.h>
 #include <asm/itcw.h>
 
 /* This is ugly... */
@@ -64,6 +63,7 @@ static void do_restore_device(struct work_struct *);
 static void dasd_return_cqr_cb(struct dasd_ccw_req *, void *);
 static void dasd_device_timeout(unsigned long);
 static void dasd_block_timeout(unsigned long);
+static void __dasd_process_erp(struct dasd_device *, struct dasd_ccw_req *);
 
 /*
  * SECTION: Operations on the device structure.
@@ -960,7 +960,7 @@ static void dasd_device_timeout(unsigned long ptr)
 	device = (struct dasd_device *) ptr;
 	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
 	/* re-activate request queue */
-        device->stopped &= ~DASD_STOPPED_PENDING;
+	dasd_device_remove_stop_bits(device, DASD_STOPPED_PENDING);
 	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
 	dasd_schedule_device_bh(device);
 }
@@ -994,10 +994,9 @@ static void dasd_handle_killed_request(struct ccw_device *cdev,
 		return;
 	cqr = (struct dasd_ccw_req *) intparm;
 	if (cqr->status != DASD_CQR_IN_IO) {
-		DBF_EVENT(DBF_DEBUG,
-			"invalid status in handle_killed_request: "
-			"bus_id %s, status %02x",
-			dev_name(&cdev->dev), cqr->status);
+		DBF_EVENT_DEVID(DBF_DEBUG, cdev,
+				"invalid status in handle_killed_request: "
+				"%02x", cqr->status);
 		return;
 	}
 
@@ -1023,7 +1022,7 @@ void dasd_generic_handle_state_change(struct dasd_device *device)
 	/* First of all start sense subsystem status request. */
 	dasd_eer_snss(device);
 
-	device->stopped &= ~DASD_STOPPED_PENDING;
+	dasd_device_remove_stop_bits(device, DASD_STOPPED_PENDING);
 	dasd_schedule_device_bh(device);
 	if (device->block)
 		dasd_schedule_block_bh(device->block);
@@ -1045,12 +1044,13 @@ void dasd_int_handler(struct ccw_device *cdev, unsigned long intparm,
 		case -EIO:
 			break;
 		case -ETIMEDOUT:
-			DBF_EVENT(DBF_WARNING, "%s(%s): request timed out\n",
-			       __func__, dev_name(&cdev->dev));
+			DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s: "
+					"request timed out\n", __func__);
 			break;
 		default:
-			DBF_EVENT(DBF_WARNING, "%s(%s): unknown error %ld\n",
-			       __func__, dev_name(&cdev->dev), PTR_ERR(irb));
+			DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s: "
+					"unknown error %ld\n", __func__,
+					PTR_ERR(irb));
 		}
 		dasd_handle_killed_request(cdev, intparm);
 		return;
@@ -1405,6 +1405,20 @@ void dasd_schedule_device_bh(struct dasd_device *device)
 	tasklet_hi_schedule(&device->tasklet);
 }
 
+void dasd_device_set_stop_bits(struct dasd_device *device, int bits)
+{
+	device->stopped |= bits;
+}
+EXPORT_SYMBOL_GPL(dasd_device_set_stop_bits);
+
+void dasd_device_remove_stop_bits(struct dasd_device *device, int bits)
+{
+	device->stopped &= ~bits;
+	if (!device->stopped)
+		wake_up(&generic_waitq);
+}
+EXPORT_SYMBOL_GPL(dasd_device_remove_stop_bits);
+
 /*
  * Queue a request to the head of the device ccw_queue.
  * Start the I/O if possible.
@@ -1465,58 +1479,135 @@ static inline int _wait_for_wakeup(struct dasd_ccw_req *cqr)
 }
 
 /*
- * Queue a request to the tail of the device ccw_queue and wait for
- * it's completion.
+ * checks if error recovery is necessary, returns 1 if yes, 0 otherwise.
  */
-int dasd_sleep_on(struct dasd_ccw_req *cqr)
+static int __dasd_sleep_on_erp(struct dasd_ccw_req *cqr)
 {
 	struct dasd_device *device;
-	int rc;
+	dasd_erp_fn_t erp_fn;
 
+	if (cqr->status == DASD_CQR_FILLED)
+		return 0;
 	device = cqr->startdev;
+	if (test_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags)) {
+		if (cqr->status == DASD_CQR_TERMINATED) {
+			device->discipline->handle_terminated_request(cqr);
+			return 1;
+		}
+		if (cqr->status == DASD_CQR_NEED_ERP) {
+			erp_fn = device->discipline->erp_action(cqr);
+			erp_fn(cqr);
+			return 1;
+		}
+		if (cqr->status == DASD_CQR_FAILED)
+			dasd_log_sense(cqr, &cqr->irb);
+		if (cqr->refers) {
+			__dasd_process_erp(device, cqr);
+			return 1;
+		}
+	}
+	return 0;
+}
 
-	cqr->callback = dasd_wakeup_cb;
-	cqr->callback_data = (void *) &generic_waitq;
-	dasd_add_request_tail(cqr);
-	wait_event(generic_waitq, _wait_for_wakeup(cqr));
+static int __dasd_sleep_on_loop_condition(struct dasd_ccw_req *cqr)
+{
+	if (test_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags)) {
+		if (cqr->refers) /* erp is not done yet */
+			return 1;
+		return ((cqr->status != DASD_CQR_DONE) &&
+			(cqr->status != DASD_CQR_FAILED));
+	} else
+		return (cqr->status == DASD_CQR_FILLED);
+}
 
-	if (cqr->status == DASD_CQR_DONE)
+static int _dasd_sleep_on(struct dasd_ccw_req *maincqr, int interruptible)
+{
+	struct dasd_device *device;
+	int rc;
+	struct list_head ccw_queue;
+	struct dasd_ccw_req *cqr;
+
+	INIT_LIST_HEAD(&ccw_queue);
+	maincqr->status = DASD_CQR_FILLED;
+	device = maincqr->startdev;
+	list_add(&maincqr->blocklist, &ccw_queue);
+	for (cqr = maincqr;  __dasd_sleep_on_loop_condition(cqr);
+	     cqr = list_first_entry(&ccw_queue,
+				    struct dasd_ccw_req, blocklist)) {
+
+		if (__dasd_sleep_on_erp(cqr))
+			continue;
+		if (cqr->status != DASD_CQR_FILLED) /* could be failed */
+			continue;
+
+		/* Non-temporary stop condition will trigger fail fast */
+		if (device->stopped & ~DASD_STOPPED_PENDING &&
+		    test_bit(DASD_CQR_FLAGS_FAILFAST, &cqr->flags) &&
+		    (!dasd_eer_enabled(device))) {
+			cqr->status = DASD_CQR_FAILED;
+			continue;
+		}
+
+		/* Don't try to start requests if device is stopped */
+		if (interruptible) {
+			rc = wait_event_interruptible(
+				generic_waitq, !(device->stopped));
+			if (rc == -ERESTARTSYS) {
+				cqr->status = DASD_CQR_FAILED;
+				maincqr->intrc = rc;
+				continue;
+			}
+		} else
+			wait_event(generic_waitq, !(device->stopped));
+
+		cqr->callback = dasd_wakeup_cb;
+		cqr->callback_data = (void *) &generic_waitq;
+		dasd_add_request_tail(cqr);
+		if (interruptible) {
+			rc = wait_event_interruptible(
+				generic_waitq, _wait_for_wakeup(cqr));
+			if (rc == -ERESTARTSYS) {
+				dasd_cancel_req(cqr);
+				/* wait (non-interruptible) for final status */
+				wait_event(generic_waitq,
+					   _wait_for_wakeup(cqr));
+				cqr->status = DASD_CQR_FAILED;
+				maincqr->intrc = rc;
+				continue;
+			}
+		} else
+			wait_event(generic_waitq, _wait_for_wakeup(cqr));
+	}
+
+	maincqr->endclk = get_clock();
+	if ((maincqr->status != DASD_CQR_DONE) &&
+	    (maincqr->intrc != -ERESTARTSYS))
+		dasd_log_sense(maincqr, &maincqr->irb);
+	if (maincqr->status == DASD_CQR_DONE)
 		rc = 0;
-	else if (cqr->intrc)
-		rc = cqr->intrc;
+	else if (maincqr->intrc)
+		rc = maincqr->intrc;
 	else
 		rc = -EIO;
 	return rc;
 }
 
+/*
+ * Queue a request to the tail of the device ccw_queue and wait for
+ * it's completion.
+ */
+int dasd_sleep_on(struct dasd_ccw_req *cqr)
+{
+	return _dasd_sleep_on(cqr, 0);
+}
+
 /*
  * Queue a request to the tail of the device ccw_queue and wait
  * interruptible for it's completion.
  */
 int dasd_sleep_on_interruptible(struct dasd_ccw_req *cqr)
 {
-	struct dasd_device *device;
-	int rc;
-
-	device = cqr->startdev;
-	cqr->callback = dasd_wakeup_cb;
-	cqr->callback_data = (void *) &generic_waitq;
-	dasd_add_request_tail(cqr);
-	rc = wait_event_interruptible(generic_waitq, _wait_for_wakeup(cqr));
-	if (rc == -ERESTARTSYS) {
-		dasd_cancel_req(cqr);
-		/* wait (non-interruptible) for final status */
-		wait_event(generic_waitq, _wait_for_wakeup(cqr));
-		cqr->intrc = rc;
-	}
-
-	if (cqr->status == DASD_CQR_DONE)
-		rc = 0;
-	else if (cqr->intrc)
-		rc = cqr->intrc;
-	else
-		rc = -EIO;
-	return rc;
+	return _dasd_sleep_on(cqr, 1);
 }
 
 /*
@@ -1630,7 +1721,7 @@ static void dasd_block_timeout(unsigned long ptr)
 	block = (struct dasd_block *) ptr;
 	spin_lock_irqsave(get_ccwdev_lock(block->base->cdev), flags);
 	/* re-activate request queue */
-	block->base->stopped &= ~DASD_STOPPED_PENDING;
+	dasd_device_remove_stop_bits(block->base, DASD_STOPPED_PENDING);
 	spin_unlock_irqrestore(get_ccwdev_lock(block->base->cdev), flags);
 	dasd_schedule_block_bh(block);
 }
@@ -1657,11 +1748,10 @@ void dasd_block_clear_timer(struct dasd_block *block)
 /*
  * Process finished error recovery ccw.
  */
-static inline void __dasd_block_process_erp(struct dasd_block *block,
-					    struct dasd_ccw_req *cqr)
+static void __dasd_process_erp(struct dasd_device *device,
+			       struct dasd_ccw_req *cqr)
 {
 	dasd_erp_fn_t erp_fn;
-	struct dasd_device *device = block->base;
 
 	if (cqr->status == DASD_CQR_DONE)
 		DBF_DEV_EVENT(DBF_NOTICE, device, "%s", "ERP successful");
@@ -1725,9 +1815,12 @@ static void __dasd_process_request_queue(struct dasd_block *block)
 				 */
 				if (!list_empty(&block->ccw_queue))
 					break;
-				spin_lock_irqsave(get_ccwdev_lock(basedev->cdev), flags);
-				basedev->stopped |= DASD_STOPPED_PENDING;
-				spin_unlock_irqrestore(get_ccwdev_lock(basedev->cdev), flags);
+				spin_lock_irqsave(
+					get_ccwdev_lock(basedev->cdev), flags);
+				dasd_device_set_stop_bits(basedev,
+							  DASD_STOPPED_PENDING);
+				spin_unlock_irqrestore(
+					get_ccwdev_lock(basedev->cdev), flags);
 				dasd_block_set_timer(block, HZ/2);
 				break;
 			}
@@ -1813,7 +1906,7 @@ restart:
 			cqr->status = DASD_CQR_FILLED;
 			cqr->retries = 255;
 			spin_lock_irqsave(get_ccwdev_lock(base->cdev), flags);
-			base->stopped |= DASD_STOPPED_QUIESCE;
+			dasd_device_set_stop_bits(base, DASD_STOPPED_QUIESCE);
 			spin_unlock_irqrestore(get_ccwdev_lock(base->cdev),
 					       flags);
 			goto restart;
@@ -1821,7 +1914,7 @@ restart:
 
 		/* Process finished ERP request. */
 		if (cqr->refers) {
-			__dasd_block_process_erp(block, cqr);
+			__dasd_process_erp(base, cqr);
 			goto restart;
 		}
 
@@ -1952,7 +2045,7 @@ restart_cb:
 		/* Process finished ERP request. */
 		if (cqr->refers) {
 			spin_lock_bh(&block->queue_lock);
-			__dasd_block_process_erp(block, cqr);
+			__dasd_process_erp(block->base, cqr);
 			spin_unlock_bh(&block->queue_lock);
 			/* restart list_for_xx loop since dasd_process_erp
 			 * might remove multiple elements */
@@ -2208,18 +2301,11 @@ int dasd_generic_probe(struct ccw_device *cdev,
 {
 	int ret;
 
-	ret = ccw_device_set_options(cdev, CCWDEV_DO_PATHGROUP);
-	if (ret) {
-		DBF_EVENT(DBF_WARNING,
-		       "dasd_generic_probe: could not set ccw-device options "
-		       "for %s\n", dev_name(&cdev->dev));
-		return ret;
-	}
 	ret = dasd_add_sysfs_files(cdev);
 	if (ret) {
-		DBF_EVENT(DBF_WARNING,
-		       "dasd_generic_probe: could not add sysfs entries "
-		       "for %s\n", dev_name(&cdev->dev));
+		DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s",
+				"dasd_generic_probe: could not add "
+				"sysfs entries");
 		return ret;
 	}
 	cdev->handler = &dasd_int_handler;
@@ -2418,16 +2504,16 @@ int dasd_generic_notify(struct ccw_device *cdev, int event)
 				cqr->status = DASD_CQR_QUEUED;
 				cqr->retries++;
 			}
-		device->stopped |= DASD_STOPPED_DC_WAIT;
+		dasd_device_set_stop_bits(device, DASD_STOPPED_DC_WAIT);
 		dasd_device_clear_timer(device);
 		dasd_schedule_device_bh(device);
 		ret = 1;
 		break;
 	case CIO_OPER:
 		/* FIXME: add a sanity check. */
-		device->stopped &= ~DASD_STOPPED_DC_WAIT;
+		dasd_device_remove_stop_bits(device, DASD_STOPPED_DC_WAIT);
 		if (device->stopped & DASD_UNRESUMED_PM) {
-			device->stopped &= ~DASD_UNRESUMED_PM;
+			dasd_device_remove_stop_bits(device, DASD_UNRESUMED_PM);
 			dasd_restore_device(device);
 			ret = 1;
 			break;
@@ -2452,7 +2538,7 @@ int dasd_generic_pm_freeze(struct ccw_device *cdev)
 	if (IS_ERR(device))
 		return PTR_ERR(device);
 	/* disallow new I/O  */
-	device->stopped |= DASD_STOPPED_PM;
+	dasd_device_set_stop_bits(device, DASD_STOPPED_PM);
 	/* clear active requests */
 	INIT_LIST_HEAD(&freeze_queue);
 	spin_lock_irq(get_ccwdev_lock(cdev));
@@ -2504,14 +2590,18 @@ int dasd_generic_restore_device(struct ccw_device *cdev)
 		return PTR_ERR(device);
 
 	/* allow new IO again */
-	device->stopped &= ~DASD_STOPPED_PM;
-	device->stopped &= ~DASD_UNRESUMED_PM;
+	dasd_device_remove_stop_bits(device,
+				     (DASD_STOPPED_PM | DASD_UNRESUMED_PM));
 
 	dasd_schedule_device_bh(device);
 
-	if (device->discipline->restore)
+	/*
+	 * call discipline restore function
+	 * if device is stopped do nothing e.g. for disconnected devices
+	 */
+	if (device->discipline->restore && !(device->stopped))
 		rc = device->discipline->restore(device);
-	if (rc)
+	if (rc || device->stopped)
 		/*
 		 * if the resume failed for the DASD we put it in
 		 * an UNRESUMED stop state
@@ -2561,8 +2651,7 @@ static struct dasd_ccw_req *dasd_generic_build_rdc(struct dasd_device *device,
 	cqr->startdev = device;
 	cqr->memdev = device;
 	cqr->expires = 10*HZ;
-	clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags);
-	cqr->retries = 2;
+	cqr->retries = 256;
 	cqr->buildclk = get_clock();
 	cqr->status = DASD_CQR_FILLED;
 	return cqr;

drivers/s390/block/dasd_3990_erp.c

@@ -12,7 +12,6 @@
 #include <linux/timer.h>
 #include <linux/slab.h>
 #include <asm/idals.h>
-#include <asm/todclk.h>
 
 #define PRINTK_HEADER "dasd_erp(3990): "
 
@@ -70,8 +69,7 @@ dasd_3990_erp_cleanup(struct dasd_ccw_req * erp, char final_status)
  *   processing until the started timer has expired or an related
  *   interrupt was received.
  */
-static void
-dasd_3990_erp_block_queue(struct dasd_ccw_req * erp, int expires)
+static void dasd_3990_erp_block_queue(struct dasd_ccw_req *erp, int expires)
 {
 
 	struct dasd_device *device = erp->startdev;
@@ -81,10 +79,13 @@ dasd_3990_erp_block_queue(struct dasd_ccw_req * erp, int expires)
 		    "blocking request queue for %is", expires/HZ);
 
 	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
-	device->stopped |= DASD_STOPPED_PENDING;
+	dasd_device_set_stop_bits(device, DASD_STOPPED_PENDING);
 	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
 	erp->status = DASD_CQR_FILLED;
-	dasd_block_set_timer(device->block, expires);
+	if (erp->block)
+		dasd_block_set_timer(erp->block, expires);
+	else
+		dasd_device_set_timer(device, expires);
 }
 
 /*
@@ -243,9 +244,13 @@ dasd_3990_erp_DCTL(struct dasd_ccw_req * erp, char modifier)
  * DESCRIPTION
  *   Setup ERP to do the ERP action 1 (see Reference manual).
  *   Repeat the operation on a different channel path.
- *   If all alternate paths have been tried, the request is posted with a
- *   permanent error.
- *   Note: duplex handling is not implemented (yet).
+ *   As deviation from the recommended recovery action, we reset the path mask
+ *   after we have tried each path and go through all paths a second time.
+ *   This will cover situations where only one path at a time is actually down,
+ *   but all paths fail and recover just with the same sequence and timing as
+ *   we try to use them (flapping links).
+ *   If all alternate paths have been tried twice, the request is posted with
+ *   a permanent error.
  *
  *  PARAMETER
  *   erp		pointer to the current ERP
@@ -254,17 +259,25 @@ dasd_3990_erp_DCTL(struct dasd_ccw_req * erp, char modifier)
  *   erp		pointer to the ERP
  *
  */
-static struct dasd_ccw_req *
-dasd_3990_erp_action_1(struct dasd_ccw_req * erp)
+static struct dasd_ccw_req *dasd_3990_erp_action_1_sec(struct dasd_ccw_req *erp)
 {
+	erp->function = dasd_3990_erp_action_1_sec;
+	dasd_3990_erp_alternate_path(erp);
+	return erp;
+}
 
+static struct dasd_ccw_req *dasd_3990_erp_action_1(struct dasd_ccw_req *erp)
+{
 	erp->function = dasd_3990_erp_action_1;
-
 	dasd_3990_erp_alternate_path(erp);
-
+	if (erp->status == DASD_CQR_FAILED) {
+		erp->status = DASD_CQR_FILLED;
+		erp->retries = 10;
+		erp->lpm = LPM_ANYPATH;
+		erp->function = dasd_3990_erp_action_1_sec;
+	}
 	return erp;
-
-}				/* end dasd_3990_erp_action_1 */
+}				/* end dasd_3990_erp_action_1(b) */
 
 /*
  * DASD_3990_ERP_ACTION_4
@@ -2295,6 +2308,7 @@ static struct dasd_ccw_req *dasd_3990_erp_add_erp(struct dasd_ccw_req *cqr)
 		return cqr;
 	}
 
+	ccw = cqr->cpaddr;
 	if (cqr->cpmode == 1) {
 		/* make a shallow copy of the original tcw but set new tsb */
 		erp->cpmode = 1;
@@ -2303,6 +2317,9 @@ static struct dasd_ccw_req *dasd_3990_erp_add_erp(struct dasd_ccw_req *cqr)
 		tsb = (struct tsb *) &tcw[1];
 		*tcw = *((struct tcw *)cqr->cpaddr);
 		tcw->tsb = (long)tsb;
+	} else if (ccw->cmd_code == DASD_ECKD_CCW_PSF) {
+		/* PSF cannot be chained from NOOP/TIC */
+		erp->cpaddr = cqr->cpaddr;
 	} else {
 		/* initialize request with default TIC to current ERP/CQR */
 		ccw = erp->cpaddr;
@@ -2487,6 +2504,8 @@ dasd_3990_erp_further_erp(struct dasd_ccw_req *erp)
 
 		erp = dasd_3990_erp_action_1(erp);
 
+	} else if (erp->function == dasd_3990_erp_action_1_sec) {
+		erp = dasd_3990_erp_action_1_sec(erp);
 	} else if (erp->function == dasd_3990_erp_action_5) {
 
 		/* retries have not been successful */

drivers/s390/block/dasd_alias.c

@@ -152,6 +152,7 @@ static struct alias_lcu *_allocate_lcu(struct dasd_uid *uid)
 	INIT_WORK(&lcu->suc_data.worker, summary_unit_check_handling_work);
 	INIT_DELAYED_WORK(&lcu->ruac_data.dwork, lcu_update_work);
 	spin_lock_init(&lcu->lock);
+	init_completion(&lcu->lcu_setup);
 	return lcu;
 
 out_err4:
@@ -239,6 +240,67 @@ int dasd_alias_make_device_known_to_lcu(struct dasd_device *device)
 	return is_lcu_known;
 }
 
+/*
+ * The first device to be registered on an LCU will have to do
+ * some additional setup steps to configure that LCU on the
+ * storage server. All further devices should wait with their
+ * initialization until the first device is done.
+ * To synchronize this work, the first device will call
+ * dasd_alias_lcu_setup_complete when it is done, and all
+ * other devices will wait for it with dasd_alias_wait_for_lcu_setup.
+ */
+void dasd_alias_lcu_setup_complete(struct dasd_device *device)
+{
+	struct dasd_eckd_private *private;
+	unsigned long flags;
+	struct alias_server *server;
+	struct alias_lcu *lcu;
+	struct dasd_uid *uid;
+
+	private = (struct dasd_eckd_private *) device->private;
+	uid = &private->uid;
+	lcu = NULL;
+	spin_lock_irqsave(&aliastree.lock, flags);
+	server = _find_server(uid);
+	if (server)
+		lcu = _find_lcu(server, uid);
+	spin_unlock_irqrestore(&aliastree.lock, flags);
+	if (!lcu) {
+		DBF_EVENT_DEVID(DBF_ERR, device->cdev,
+				"could not find lcu for %04x %02x",
+				uid->ssid, uid->real_unit_addr);
+		WARN_ON(1);
+		return;
+	}
+	complete_all(&lcu->lcu_setup);
+}
+
+void dasd_alias_wait_for_lcu_setup(struct dasd_device *device)
+{
+	struct dasd_eckd_private *private;
+	unsigned long flags;
+	struct alias_server *server;
+	struct alias_lcu *lcu;
+	struct dasd_uid *uid;
+
+	private = (struct dasd_eckd_private *) device->private;
+	uid = &private->uid;
+	lcu = NULL;
+	spin_lock_irqsave(&aliastree.lock, flags);
+	server = _find_server(uid);
+	if (server)
+		lcu = _find_lcu(server, uid);
+	spin_unlock_irqrestore(&aliastree.lock, flags);
+	if (!lcu) {
+		DBF_EVENT_DEVID(DBF_ERR, device->cdev,
+				"could not find lcu for %04x %02x",
+				uid->ssid, uid->real_unit_addr);
+		WARN_ON(1);
+		return;
+	}
+	wait_for_completion(&lcu->lcu_setup);
+}
+
 /*
  * This function removes a device from the scope of alias management.
  * The complicated part is to make sure that it is not in use by
@@ -755,11 +817,11 @@ static void __stop_device_on_lcu(struct dasd_device *device,
 {
 	/* If pos == device then device is already locked! */
 	if (pos == device) {
-		pos->stopped |= DASD_STOPPED_SU;
+		dasd_device_set_stop_bits(pos, DASD_STOPPED_SU);
 		return;
 	}
 	spin_lock(get_ccwdev_lock(pos->cdev));
-	pos->stopped |= DASD_STOPPED_SU;
+	dasd_device_set_stop_bits(pos, DASD_STOPPED_SU);
 	spin_unlock(get_ccwdev_lock(pos->cdev));
 }
 
@@ -793,26 +855,26 @@ static void _unstop_all_devices_on_lcu(struct alias_lcu *lcu)
 
 	list_for_each_entry(device, &lcu->active_devices, alias_list) {
 		spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
-		device->stopped &= ~DASD_STOPPED_SU;
+		dasd_device_remove_stop_bits(device, DASD_STOPPED_SU);
 		spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
 	}
 
 	list_for_each_entry(device, &lcu->inactive_devices, alias_list) {
 		spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
-		device->stopped &= ~DASD_STOPPED_SU;
+		dasd_device_remove_stop_bits(device, DASD_STOPPED_SU);
 		spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
 	}
 
 	list_for_each_entry(pavgroup, &lcu->grouplist, group) {
 		list_for_each_entry(device, &pavgroup->baselist, alias_list) {
 			spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
-			device->stopped &= ~DASD_STOPPED_SU;
+			dasd_device_remove_stop_bits(device, DASD_STOPPED_SU);
 			spin_unlock_irqrestore(get_ccwdev_lock(device->cdev),
 					       flags);
 		}
 		list_for_each_entry(device, &pavgroup->aliaslist, alias_list) {
 			spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
-			device->stopped &= ~DASD_STOPPED_SU;
+			dasd_device_remove_stop_bits(device, DASD_STOPPED_SU);
 			spin_unlock_irqrestore(get_ccwdev_lock(device->cdev),
 					       flags);
 		}
@@ -836,7 +898,8 @@ static void summary_unit_check_handling_work(struct work_struct *work)
 
 	/* 2. reset summary unit check */
 	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);
-	device->stopped &= ~(DASD_STOPPED_SU | DASD_STOPPED_PENDING);
+	dasd_device_remove_stop_bits(device,
+				     (DASD_STOPPED_SU | DASD_STOPPED_PENDING));
 	spin_unlock_irqrestore(get_ccwdev_lock(device->cdev), flags);
 	reset_summary_unit_check(lcu, device, suc_data->reason);
 

drivers/s390/block/dasd_diag.c

@@ -24,7 +24,6 @@
 #include <asm/ebcdic.h>
 #include <asm/io.h>
 #include <asm/s390_ext.h>
-#include <asm/todclk.h>
 #include <asm/vtoc.h>
 #include <asm/diag.h>
 
@@ -145,6 +144,15 @@ dasd_diag_erp(struct dasd_device *device)
 
 	mdsk_term_io(device);
 	rc = mdsk_init_io(device, device->block->bp_block, 0, NULL);
+	if (rc == 4) {
+		if (!(device->features & DASD_FEATURE_READONLY)) {
+			dev_warn(&device->cdev->dev,
+				 "The access mode of a DIAG device changed"
+				 " to read-only");
+			device->features |= DASD_FEATURE_READONLY;
+		}
+		rc = 0;
+	}
 	if (rc)
 		dev_warn(&device->cdev->dev, "DIAG ERP failed with "
 			    "rc=%d\n", rc);
@@ -433,16 +441,20 @@ dasd_diag_check_device(struct dasd_device *device)
 	for (sb = 512; sb < bsize; sb = sb << 1)
 		block->s2b_shift++;
 	rc = mdsk_init_io(device, block->bp_block, 0, NULL);
-	if (rc) {
+	if (rc && (rc != 4)) {
 		dev_warn(&device->cdev->dev, "DIAG initialization "
 			"failed with rc=%d\n", rc);
 		rc = -EIO;
 	} else {
+		if (rc == 4)
+			device->features |= DASD_FEATURE_READONLY;
 		dev_info(&device->cdev->dev,
-			 "New DASD with %ld byte/block, total size %ld KB\n",
+			 "New DASD with %ld byte/block, total size %ld KB%s\n",
 			 (unsigned long) block->bp_block,
 			 (unsigned long) (block->blocks <<
-					  block->s2b_shift) >> 1);
+					  block->s2b_shift) >> 1,
+			 (rc == 4) ? ", read-only device" : "");
+		rc = 0;
 	}
 out_label:
 	free_page((long) label);

drivers/s390/block/dasd_eckd.c

@@ -24,7 +24,6 @@
 #include <asm/idals.h>
 #include <asm/ebcdic.h>
 #include <asm/io.h>
-#include <asm/todclk.h>
 #include <asm/uaccess.h>
 #include <asm/cio.h>
 #include <asm/ccwdev.h>
@@ -78,6 +77,11 @@ MODULE_DEVICE_TABLE(ccw, dasd_eckd_ids);
 
 static struct ccw_driver dasd_eckd_driver; /* see below */
 
+#define INIT_CQR_OK 0
+#define INIT_CQR_UNFORMATTED 1
+#define INIT_CQR_ERROR 2
+
+
 /* initial attempt at a probe function. this can be simplified once
  * the other detection code is gone */
 static int
@@ -86,11 +90,12 @@ dasd_eckd_probe (struct ccw_device *cdev)
 	int ret;
 
 	/* set ECKD specific ccw-device options */
-	ret = ccw_device_set_options(cdev, CCWDEV_ALLOW_FORCE);
+	ret = ccw_device_set_options(cdev, CCWDEV_ALLOW_FORCE |
+				     CCWDEV_DO_PATHGROUP | CCWDEV_DO_MULTIPATH);
 	if (ret) {
-		DBF_EVENT(DBF_WARNING,
-		       "dasd_eckd_probe: could not set ccw-device options "
-		       "for %s\n", dev_name(&cdev->dev));
+		DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s",
+				"dasd_eckd_probe: could not set "
+				"ccw-device options");
 		return ret;
 	}
 	ret = dasd_generic_probe(cdev, &dasd_eckd_discipline);
@@ -749,8 +754,7 @@ static struct dasd_ccw_req *dasd_eckd_build_rcd_lpm(struct dasd_device *device,
 	cqr->block = NULL;
 	cqr->expires = 10*HZ;
 	cqr->lpm = lpm;
-	clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags);
-	cqr->retries = 2;
+	cqr->retries = 256;
 	cqr->buildclk = get_clock();
 	cqr->status = DASD_CQR_FILLED;
 	return cqr;
@@ -885,16 +889,15 @@ static int dasd_eckd_read_conf(struct dasd_device *device)
 			rc = dasd_eckd_read_conf_lpm(device, &conf_data,
 						     &conf_len, lpm);
 			if (rc && rc != -EOPNOTSUPP) {	/* -EOPNOTSUPP is ok */
-				DBF_EVENT(DBF_WARNING,
+				DBF_EVENT_DEVID(DBF_WARNING, device->cdev,
 					  "Read configuration data returned "
-					  "error %d for device: %s", rc,
-					  dev_name(&device->cdev->dev));
+					  "error %d", rc);
 				return rc;
 			}
 			if (conf_data == NULL) {
-				DBF_EVENT(DBF_WARNING, "No configuration "
-					  "data retrieved for device: %s",
-					  dev_name(&device->cdev->dev));
+				DBF_EVENT_DEVID(DBF_WARNING, device->cdev, "%s",
+						"No configuration data "
+						"retrieved");
 				continue;	/* no error */
 			}
 			/* save first valid configuration data */
@@ -941,16 +944,14 @@ static int dasd_eckd_read_features(struct dasd_device *device)
 				    sizeof(struct dasd_rssd_features)),
 				   device);
 	if (IS_ERR(cqr)) {
-		DBF_EVENT(DBF_WARNING, "Could not allocate initialization "
-			  "request for device: %s",
-			  dev_name(&device->cdev->dev));
+		DBF_EVENT_DEVID(DBF_WARNING, device->cdev, "%s", "Could not "
+				"allocate initialization request");
 		return PTR_ERR(cqr);
 	}
 	cqr->startdev = device;
 	cqr->memdev = device;
 	cqr->block = NULL;
-	clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags);
-	cqr->retries = 5;
+	cqr->retries = 256;
 	cqr->expires = 10 * HZ;
 
 	/* Prepare for Read Subsystem Data */
@@ -1012,9 +1013,9 @@ static struct dasd_ccw_req *dasd_eckd_build_psf_ssc(struct dasd_device *device,
 	}
 	psf_ssc_data = (struct dasd_psf_ssc_data *)cqr->data;
 	psf_ssc_data->order = PSF_ORDER_SSC;
-	psf_ssc_data->suborder = 0x40;
+	psf_ssc_data->suborder = 0xc0;
 	if (enable_pav) {
-		psf_ssc_data->suborder |= 0x88;
+		psf_ssc_data->suborder |= 0x08;
 		psf_ssc_data->reserved[0] = 0x88;
 	}
 	ccw = cqr->cpaddr;
@@ -1025,6 +1026,7 @@ static struct dasd_ccw_req *dasd_eckd_build_psf_ssc(struct dasd_device *device,
 	cqr->startdev = device;
 	cqr->memdev = device;
 	cqr->block = NULL;
+	cqr->retries = 256;
 	cqr->expires = 10*HZ;
 	cqr->buildclk = get_clock();
 	cqr->status = DASD_CQR_FILLED;
@@ -1057,7 +1059,7 @@ dasd_eckd_psf_ssc(struct dasd_device *device, int enable_pav)
 /*
  * Valide storage server of current device.
  */
-static int dasd_eckd_validate_server(struct dasd_device *device)
+static void dasd_eckd_validate_server(struct dasd_device *device)
 {
 	int rc;
 	struct dasd_eckd_private *private;
@@ -1068,15 +1070,12 @@ static int dasd_eckd_validate_server(struct dasd_device *device)
 	else
 		enable_pav = 1;
 	rc = dasd_eckd_psf_ssc(device, enable_pav);
+
 	/* may be requested feature is not available on server,
 	 * therefore just report error and go ahead */
 	private = (struct dasd_eckd_private *) device->private;
-	DBF_EVENT(DBF_WARNING, "PSF-SSC on storage subsystem %s.%s.%04x "
-		  "returned rc=%d for device: %s",
-		  private->uid.vendor, private->uid.serial,
-		  private->uid.ssid, rc, dev_name(&device->cdev->dev));
-	/* RE-Read Configuration Data */
-	return dasd_eckd_read_conf(device);
+	DBF_EVENT_DEVID(DBF_WARNING, device->cdev, "PSF-SSC for SSID %04x "
+			"returned rc=%d", private->uid.ssid, rc);
 }
 
 /*
@@ -1090,6 +1089,15 @@ dasd_eckd_check_characteristics(struct dasd_device *device)
 	struct dasd_block *block;
 	int is_known, rc;
 
+	if (!ccw_device_is_pathgroup(device->cdev)) {
+		dev_warn(&device->cdev->dev,
+			 "A channel path group could not be established\n");
+		return -EIO;
+	}
+	if (!ccw_device_is_multipath(device->cdev)) {
+		dev_info(&device->cdev->dev,
+			 "The DASD is not operating in multipath mode\n");
+	}
 	private = (struct dasd_eckd_private *) device->private;
 	if (!private) {
 		private = kzalloc(sizeof(*private), GFP_KERNEL | GFP_DMA);
@@ -1123,9 +1131,9 @@ dasd_eckd_check_characteristics(struct dasd_device *device)
 	if (private->uid.type == UA_BASE_DEVICE) {
 		block = dasd_alloc_block();
 		if (IS_ERR(block)) {
-			DBF_EVENT(DBF_WARNING, "could not allocate dasd "
-				  "block structure for device: %s",
-				  dev_name(&device->cdev->dev));
+			DBF_EVENT_DEVID(DBF_WARNING, device->cdev, "%s",
+					"could not allocate dasd "
+					"block structure");
 			rc = PTR_ERR(block);
 			goto out_err1;
 		}
@@ -1139,12 +1147,21 @@ dasd_eckd_check_characteristics(struct dasd_device *device)
 		rc = is_known;
 		goto out_err2;
 	}
+	/*
+	 * dasd_eckd_vaildate_server is done on the first device that
+	 * is found for an LCU. All later other devices have to wait
+	 * for it, so they will read the correct feature codes.
+	 */
 	if (!is_known) {
-		/* new lcu found */
-		rc = dasd_eckd_validate_server(device); /* will switch pav on */
-		if (rc)
-			goto out_err3;
-	}
+		dasd_eckd_validate_server(device);
+		dasd_alias_lcu_setup_complete(device);
+	} else
+		dasd_alias_wait_for_lcu_setup(device);
+
+	/* device may report different configuration data after LCU setup */
+	rc = dasd_eckd_read_conf(device);
+	if (rc)
+		goto out_err3;
 
 	/* Read Feature Codes */
 	dasd_eckd_read_features(device);
@@ -1153,9 +1170,8 @@ dasd_eckd_check_characteristics(struct dasd_device *device)
 	rc = dasd_generic_read_dev_chars(device, DASD_ECKD_MAGIC,
 					 &private->rdc_data, 64);
 	if (rc) {
-		DBF_EVENT(DBF_WARNING,
-			  "Read device characteristics failed, rc=%d for "
-			  "device: %s", rc, dev_name(&device->cdev->dev));
+		DBF_EVENT_DEVID(DBF_WARNING, device->cdev,
+				"Read device characteristic failed, rc=%d", rc);
 		goto out_err3;
 	}
 	/* find the vaild cylinder size */
@@ -1256,12 +1272,29 @@ dasd_eckd_analysis_ccw(struct dasd_device *device)
 	cqr->block = NULL;
 	cqr->startdev = device;
 	cqr->memdev = device;
-	cqr->retries = 0;
+	cqr->retries = 255;
 	cqr->buildclk = get_clock();
 	cqr->status = DASD_CQR_FILLED;
 	return cqr;
 }
 
+/* differentiate between 'no record found' and any other error */
+static int dasd_eckd_analysis_evaluation(struct dasd_ccw_req *init_cqr)
+{
+	char *sense;
+	if (init_cqr->status == DASD_CQR_DONE)
+		return INIT_CQR_OK;
+	else if (init_cqr->status == DASD_CQR_NEED_ERP ||
+		 init_cqr->status == DASD_CQR_FAILED) {
+		sense = dasd_get_sense(&init_cqr->irb);
+		if (sense && (sense[1] & SNS1_NO_REC_FOUND))
+			return INIT_CQR_UNFORMATTED;
+		else
+			return INIT_CQR_ERROR;
+	} else
+		return INIT_CQR_ERROR;
+}
+
 /*
  * This is the callback function for the init_analysis cqr. It saves
  * the status of the initial analysis ccw before it frees it and kicks
@@ -1269,21 +1302,20 @@ dasd_eckd_analysis_ccw(struct dasd_device *device)
  * dasd_eckd_do_analysis again (if the devices has not been marked
  * for deletion in the meantime).
  */
-static void
-dasd_eckd_analysis_callback(struct dasd_ccw_req *init_cqr, void *data)
+static void dasd_eckd_analysis_callback(struct dasd_ccw_req *init_cqr,
+					void *data)
 {
 	struct dasd_eckd_private *private;
 	struct dasd_device *device;
 
 	device = init_cqr->startdev;
 	private = (struct dasd_eckd_private *) device->private;
-	private->init_cqr_status = init_cqr->status;
+	private->init_cqr_status = dasd_eckd_analysis_evaluation(init_cqr);
 	dasd_sfree_request(init_cqr, device);
 	dasd_kick_device(device);
 }
 
-static int
-dasd_eckd_start_analysis(struct dasd_block *block)
+static int dasd_eckd_start_analysis(struct dasd_block *block)
 {
 	struct dasd_eckd_private *private;
 	struct dasd_ccw_req *init_cqr;
@@ -1295,27 +1327,44 @@ dasd_eckd_start_analysis(struct dasd_block *block)
 	init_cqr->callback = dasd_eckd_analysis_callback;
 	init_cqr->callback_data = NULL;
 	init_cqr->expires = 5*HZ;
+	/* first try without ERP, so we can later handle unformatted
+	 * devices as special case
+	 */
+	clear_bit(DASD_CQR_FLAGS_USE_ERP, &init_cqr->flags);
+	init_cqr->retries = 0;
 	dasd_add_request_head(init_cqr);
 	return -EAGAIN;
 }
 
-static int
-dasd_eckd_end_analysis(struct dasd_block *block)
+static int dasd_eckd_end_analysis(struct dasd_block *block)
 {
 	struct dasd_device *device;
 	struct dasd_eckd_private *private;
 	struct eckd_count *count_area;
 	unsigned int sb, blk_per_trk;
 	int status, i;
+	struct dasd_ccw_req *init_cqr;
 
 	device = block->base;
 	private = (struct dasd_eckd_private *) device->private;
 	status = private->init_cqr_status;
 	private->init_cqr_status = -1;
-	if (status != DASD_CQR_DONE) {
-		dev_warn(&device->cdev->dev,
-			    "The DASD is not formatted\n");
+	if (status == INIT_CQR_ERROR) {
+		/* try again, this time with full ERP */
+		init_cqr = dasd_eckd_analysis_ccw(device);
+		dasd_sleep_on(init_cqr);
+		status = dasd_eckd_analysis_evaluation(init_cqr);
+		dasd_sfree_request(init_cqr, device);
+	}
+
+	if (status == INIT_CQR_UNFORMATTED) {
+		dev_warn(&device->cdev->dev, "The DASD is not formatted\n");
 		return -EMEDIUMTYPE;
+	} else if (status == INIT_CQR_ERROR) {
+		dev_err(&device->cdev->dev,
+			"Detecting the DASD disk layout failed because "
+			"of an I/O error\n");
+		return -EIO;
 	}
 
 	private->uses_cdl = 1;
@@ -1607,8 +1656,7 @@ dasd_eckd_format_device(struct dasd_device * device,
 	}
 	fcp->startdev = device;
 	fcp->memdev = device;
-	clear_bit(DASD_CQR_FLAGS_USE_ERP, &fcp->flags);
-	fcp->retries = 5;	/* set retry counter to enable default ERP */
+	fcp->retries = 256;
 	fcp->buildclk = get_clock();
 	fcp->status = DASD_CQR_FILLED;
 	return fcp;
@@ -2690,6 +2738,7 @@ dasd_eckd_performance(struct dasd_device *device, void __user *argp)
 	cqr->startdev = device;
 	cqr->memdev = device;
 	cqr->retries = 0;
+	clear_bit(DASD_CQR_FLAGS_USE_ERP, &cqr->flags);
 	cqr->expires = 10 * HZ;
 
 	/* Prepare for Read Subsystem Data */
@@ -3240,11 +3289,15 @@ int dasd_eckd_restore_device(struct dasd_device *device)
 	if (is_known < 0)
 		return is_known;
 	if (!is_known) {
-		/* new lcu found */
-		rc = dasd_eckd_validate_server(device); /* will switch pav on */
-		if (rc)
-			goto out_err;
-	}
+		dasd_eckd_validate_server(device);
+		dasd_alias_lcu_setup_complete(device);
+	} else
+		dasd_alias_wait_for_lcu_setup(device);
+
+	/* RE-Read Configuration Data */
+	rc = dasd_eckd_read_conf(device);
+	if (rc)
+		goto out_err;
 
 	/* Read Feature Codes */
 	dasd_eckd_read_features(device);
@@ -3253,9 +3306,8 @@ int dasd_eckd_restore_device(struct dasd_device *device)
 	rc = dasd_generic_read_dev_chars(device, DASD_ECKD_MAGIC,
 					 &temp_rdc_data, 64);
 	if (rc) {
-		DBF_EVENT(DBF_WARNING,
-			  "Read device characteristics failed, rc=%d for "
-			  "device: %s", rc, dev_name(&device->cdev->dev));
+		DBF_EVENT_DEVID(DBF_WARNING, device->cdev,
+				"Read device characteristic failed, rc=%d", rc);
 		goto out_err;
 	}
 	spin_lock_irqsave(get_ccwdev_lock(device->cdev), flags);

drivers/s390/block/dasd_eckd.h

@@ -414,6 +414,7 @@ struct alias_lcu {
 	struct summary_unit_check_work_data suc_data;
 	struct read_uac_work_data ruac_data;
 	struct dasd_ccw_req *rsu_cqr;
+	struct completion lcu_setup;
 };
 
 struct alias_pav_group {
@@ -460,5 +461,6 @@ int dasd_alias_remove_device(struct dasd_device *);
 struct dasd_device *dasd_alias_get_start_dev(struct dasd_device *);
 void dasd_alias_handle_summary_unit_check(struct dasd_device *, struct irb *);
 void dasd_eckd_reset_ccw_to_base_io(struct dasd_ccw_req *);
-
+void dasd_alias_lcu_setup_complete(struct dasd_device *);
+void dasd_alias_wait_for_lcu_setup(struct dasd_device *);
 #endif				/* DASD_ECKD_H */

drivers/s390/block/dasd_eer.c

@@ -536,7 +536,6 @@ static int dasd_eer_open(struct inode *inp, struct file *filp)
 	eerb = kzalloc(sizeof(struct eerbuffer), GFP_KERNEL);
 	if (!eerb)
 		return -ENOMEM;
-	lock_kernel();
 	eerb->buffer_page_count = eer_pages;
 	if (eerb->buffer_page_count < 1 ||
 	    eerb->buffer_page_count > INT_MAX / PAGE_SIZE) {
@@ -544,7 +543,6 @@ static int dasd_eer_open(struct inode *inp, struct file *filp)
 		DBF_EVENT(DBF_WARNING, "can't open device since module "
 			"parameter eer_pages is smaller than 1 or"
 			" bigger than %d", (int)(INT_MAX / PAGE_SIZE));
-		unlock_kernel();
 		return -EINVAL;
 	}
 	eerb->buffersize = eerb->buffer_page_count * PAGE_SIZE;
@@ -552,14 +550,12 @@ static int dasd_eer_open(struct inode *inp, struct file *filp)
 			       GFP_KERNEL);
         if (!eerb->buffer) {
 		kfree(eerb);
-		unlock_kernel();
                 return -ENOMEM;
 	}
 	if (dasd_eer_allocate_buffer_pages(eerb->buffer,
 					   eerb->buffer_page_count)) {
 		kfree(eerb->buffer);
 		kfree(eerb);
-		unlock_kernel();
 		return -ENOMEM;
 	}
 	filp->private_data = eerb;
@@ -567,7 +563,6 @@ static int dasd_eer_open(struct inode *inp, struct file *filp)
 	list_add(&eerb->list, &bufferlist);
 	spin_unlock_irqrestore(&bufferlock, flags);
 
-	unlock_kernel();
 	return nonseekable_open(inp,filp);
 }
 

drivers/s390/block/dasd_fba.c

@@ -20,7 +20,6 @@
 #include <asm/idals.h>
 #include <asm/ebcdic.h>
 #include <asm/io.h>
-#include <asm/todclk.h>
 #include <asm/ccwdev.h>
 
 #include "dasd_int.h"
@@ -141,9 +140,8 @@ dasd_fba_check_characteristics(struct dasd_device *device)
 	}
 	block = dasd_alloc_block();
 	if (IS_ERR(block)) {
-		DBF_EVENT(DBF_WARNING, "could not allocate dasd block "
-			  "structure for device: %s",
-			  dev_name(&device->cdev->dev));
+		DBF_EVENT_DEVID(DBF_WARNING, cdev, "%s", "could not allocate "
+				"dasd block structure");
 		device->private = NULL;
 		kfree(private);
 		return PTR_ERR(block);
@@ -155,9 +153,8 @@ dasd_fba_check_characteristics(struct dasd_device *device)
 	rc = dasd_generic_read_dev_chars(device, DASD_FBA_MAGIC,
 					 &private->rdc_data, 32);
 	if (rc) {
-		DBF_EVENT(DBF_WARNING, "Read device characteristics returned "
-			  "error %d for device: %s",
-			  rc, dev_name(&device->cdev->dev));
+		DBF_EVENT_DEVID(DBF_WARNING, cdev, "Read device "
+				"characteristics returned error %d", rc);
 		device->block = NULL;
 		dasd_free_block(block);
 		device->private = NULL;

drivers/s390/block/dasd_int.h

@@ -108,6 +108,16 @@ do { \
 			    d_data); \
 } while(0)
 
+#define DBF_EVENT_DEVID(d_level, d_cdev, d_str, d_data...)	\
+do { \
+	struct ccw_dev_id __dev_id;			\
+	ccw_device_get_id(d_cdev, &__dev_id);		\
+	debug_sprintf_event(dasd_debug_area,		\
+			    d_level,					\
+			    "0.%x.%04x " d_str "\n",			\
+			    __dev_id.ssid, __dev_id.devno, d_data);	\
+} while (0)
+
 #define DBF_EXC(d_level, d_str, d_data...)\
 do { \
 	debug_sprintf_exception(dasd_debug_area, \
@@ -595,6 +605,9 @@ int dasd_generic_restore_device(struct ccw_device *);
 int dasd_generic_read_dev_chars(struct dasd_device *, int, void *, int);
 char *dasd_get_sense(struct irb *);
 
+void dasd_device_set_stop_bits(struct dasd_device *, int);
+void dasd_device_remove_stop_bits(struct dasd_device *, int);
+
 /* externals in dasd_devmap.c */
 extern int dasd_max_devindex;
 extern int dasd_probeonly;

drivers/s390/block/dasd_ioctl.c

@@ -101,7 +101,7 @@ static int dasd_ioctl_quiesce(struct dasd_block *block)
 	pr_info("%s: The DASD has been put in the quiesce "
 		"state\n", dev_name(&base->cdev->dev));
 	spin_lock_irqsave(get_ccwdev_lock(base->cdev), flags);
-	base->stopped |= DASD_STOPPED_QUIESCE;
+	dasd_device_set_stop_bits(base, DASD_STOPPED_QUIESCE);
 	spin_unlock_irqrestore(get_ccwdev_lock(base->cdev), flags);
 	return 0;
 }
@@ -122,7 +122,7 @@ static int dasd_ioctl_resume(struct dasd_block *block)
 	pr_info("%s: I/O operations have been resumed "
 		"on the DASD\n", dev_name(&base->cdev->dev));
 	spin_lock_irqsave(get_ccwdev_lock(base->cdev), flags);
-	base->stopped &= ~DASD_STOPPED_QUIESCE;
+	dasd_device_remove_stop_bits(base, DASD_STOPPED_QUIESCE);
 	spin_unlock_irqrestore(get_ccwdev_lock(base->cdev), flags);
 
 	dasd_schedule_block_bh(block);

drivers/s390/char/con3215.c

@@ -857,7 +857,6 @@ static struct console con3215 = {
 
 /*
  * 3215 console initialization code called from console_init().
- * NOTE: This is called before kmalloc is available.
  */
 static int __init con3215_init(void)
 {

drivers/s390/char/con3270.c

@@ -572,7 +572,6 @@ static struct console con3270 = {
 
 /*
  * 3270 console initialization code called from console_init().
- * NOTE: This is called before kmalloc is available.
  */
 static int __init
 con3270_init(void)

drivers/s390/char/fs3270.c

@@ -38,6 +38,8 @@ struct fs3270 {
 	size_t rdbuf_size;		/* size of data returned by RDBUF */
 };
 
+static DEFINE_MUTEX(fs3270_mutex);
+
 static void
 fs3270_wake_up(struct raw3270_request *rq, void *data)
 {
@@ -328,7 +330,7 @@ fs3270_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 	if (!fp)
 		return -ENODEV;
 	rc = 0;
-	lock_kernel();
+	mutex_lock(&fs3270_mutex);
 	switch (cmd) {
 	case TUBICMD:
 		fp->read_command = arg;
@@ -354,7 +356,7 @@ fs3270_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
 			rc = -EFAULT;
 		break;
 	}
-	unlock_kernel();
+	mutex_unlock(&fs3270_mutex);
 	return rc;
 }
 
@@ -437,7 +439,7 @@ fs3270_open(struct inode *inode, struct file *filp)
 		minor = tty->index + RAW3270_FIRSTMINOR;
 		tty_kref_put(tty);
 	}
-	lock_kernel();
+	mutex_lock(&fs3270_mutex);
 	/* Check if some other program is already using fullscreen mode. */
 	fp = (struct fs3270 *) raw3270_find_view(&fs3270_fn, minor);
 	if (!IS_ERR(fp)) {
@@ -478,7 +480,7 @@ fs3270_open(struct inode *inode, struct file *filp)
 	}
 	filp->private_data = fp;
 out:
-	unlock_kernel();
+	mutex_unlock(&fs3270_mutex);
 	return rc;
 }
 

drivers/s390/char/monreader.c

@@ -12,7 +12,6 @@
 #include <linux/module.h>
 #include <linux/moduleparam.h>
 #include <linux/init.h>
-#include <linux/smp_lock.h>
 #include <linux/errno.h>
 #include <linux/types.h>
 #include <linux/kernel.h>
@@ -283,7 +282,6 @@ static int mon_open(struct inode *inode, struct file *filp)
 	/*
 	 * only one user allowed
 	 */
-	lock_kernel();
 	rc = -EBUSY;
 	if (test_and_set_bit(MON_IN_USE, &mon_in_use))
 		goto out;
@@ -321,7 +319,6 @@ static int mon_open(struct inode *inode, struct file *filp)
 	}
 	filp->private_data = monpriv;
 	dev_set_drvdata(monreader_device, monpriv);
-	unlock_kernel();
 	return nonseekable_open(inode, filp);
 
 out_path:
@@ -331,7 +328,6 @@ out_priv:
 out_use:
 	clear_bit(MON_IN_USE, &mon_in_use);
 out:
-	unlock_kernel();
 	return rc;
 }
 
@@ -607,6 +603,10 @@ static int __init mon_init(void)
 	}
 	dcss_mkname(mon_dcss_name, &user_data_connect[8]);
 
+	/*
+	 * misc_register() has to be the last action in module_init(), because
+	 * file operations will be available right after this.
+	 */
 	rc = misc_register(&mon_dev);
 	if (rc < 0 )
 		goto out;

drivers/s390/char/monwriter.c

@@ -13,7 +13,6 @@
 #include <linux/moduleparam.h>
 #include <linux/init.h>
 #include <linux/errno.h>
-#include <linux/smp_lock.h>
 #include <linux/types.h>
 #include <linux/kernel.h>
 #include <linux/miscdevice.h>
@@ -185,13 +184,11 @@ static int monwrite_open(struct inode *inode, struct file *filp)
 	monpriv = kzalloc(sizeof(struct mon_private), GFP_KERNEL);
 	if (!monpriv)
 		return -ENOMEM;
-	lock_kernel();
 	INIT_LIST_HEAD(&monpriv->list);
 	monpriv->hdr_to_read = sizeof(monpriv->hdr);
 	mutex_init(&monpriv->thread_mutex);
 	filp->private_data = monpriv;
 	list_add_tail(&monpriv->priv_list, &mon_priv_list);
-	unlock_kernel();
 	return nonseekable_open(inode, filp);
 }
 
@@ -364,6 +361,10 @@ static int __init mon_init(void)
 		goto out_driver;
 	}
 
+	/*
+	 * misc_register() has to be the last action in module_init(), because
+	 * file operations will be available right after this.
+	 */
 	rc = misc_register(&mon_dev);
 	if (rc)
 		goto out_device;

drivers/s390/char/sclp_cmd.c

@@ -84,6 +84,7 @@ static void __init sclp_read_info_early(void)
 		do {
 			memset(sccb, 0, sizeof(*sccb));
 			sccb->header.length = sizeof(*sccb);
+			sccb->header.function_code = 0x80;
 			sccb->header.control_mask[2] = 0x80;
 			rc = sclp_cmd_sync_early(commands[i], sccb);
 		} while (rc == -EBUSY);

drivers/s390/char/tape.h

@@ -212,6 +212,9 @@ struct tape_device {
 	struct tape_class_device *	nt;
 	struct tape_class_device *	rt;
 
+	/* Device mutex to serialize tape commands. */
+	struct mutex			mutex;
+
 	/* Device discipline information. */
 	struct tape_discipline *	discipline;
 	void *				discdata;
@@ -292,9 +295,9 @@ extern int tape_generic_pm_suspend(struct ccw_device *);
 extern int tape_generic_probe(struct ccw_device *);
 extern void tape_generic_remove(struct ccw_device *);
 
-extern struct tape_device *tape_get_device(int devindex);
-extern struct tape_device *tape_get_device_reference(struct tape_device *);
-extern struct tape_device *tape_put_device(struct tape_device *);
+extern struct tape_device *tape_find_device(int devindex);
+extern struct tape_device *tape_get_device(struct tape_device *);
+extern void tape_put_device(struct tape_device *);
 
 /* Externals from tape_char.c */
 extern int tapechar_init(void);

drivers/s390/char/tape_34xx.c

@@ -113,16 +113,16 @@ tape_34xx_work_handler(struct work_struct *work)
 {
 	struct tape_34xx_work *p =
 		container_of(work, struct tape_34xx_work, work);
+	struct tape_device *device = p->device;
 
 	switch(p->op) {
 		case TO_MSEN:
-			tape_34xx_medium_sense(p->device);
+			tape_34xx_medium_sense(device);
 			break;
 		default:
 			DBF_EVENT(3, "T34XX: internal error: unknown work\n");
 	}
-
-	p->device = tape_put_device(p->device);
+	tape_put_device(device);
 	kfree(p);
 }
 
@@ -136,7 +136,7 @@ tape_34xx_schedule_work(struct tape_device *device, enum tape_op op)
 
 	INIT_WORK(&p->work, tape_34xx_work_handler);
 
-	p->device = tape_get_device_reference(device);
+	p->device = tape_get_device(device);
 	p->op     = op;
 
 	schedule_work(&p->work);

drivers/s390/char/tape_3590.c

@@ -608,7 +608,7 @@ tape_3590_schedule_work(struct tape_device *device, enum tape_op op)
 
 	INIT_WORK(&p->work, tape_3590_work_handler);
 
-	p->device = tape_get_device_reference(device);
+	p->device = tape_get_device(device);
 	p->op = op;
 
 	schedule_work(&p->work);

drivers/s390/char/tape_block.c

@@ -54,7 +54,7 @@ static const struct block_device_operations tapeblock_fops = {
 	.owner		 = THIS_MODULE,
 	.open		 = tapeblock_open,
 	.release	 = tapeblock_release,
-	.locked_ioctl           = tapeblock_ioctl,
+	.ioctl		 = tapeblock_ioctl,
 	.media_changed   = tapeblock_medium_changed,
 	.revalidate_disk = tapeblock_revalidate_disk,
 };
@@ -239,7 +239,7 @@ tapeblock_setup_device(struct tape_device * device)
 	disk->major = tapeblock_major;
 	disk->first_minor = device->first_minor;
 	disk->fops = &tapeblock_fops;
-	disk->private_data = tape_get_device_reference(device);
+	disk->private_data = tape_get_device(device);
 	disk->queue = blkdat->request_queue;
 	set_capacity(disk, 0);
 	sprintf(disk->disk_name, "btibm%d",
@@ -247,11 +247,11 @@ tapeblock_setup_device(struct tape_device * device)
 
 	blkdat->disk = disk;
 	blkdat->medium_changed = 1;
-	blkdat->request_queue->queuedata = tape_get_device_reference(device);
+	blkdat->request_queue->queuedata = tape_get_device(device);
 
 	add_disk(disk);
 
-	tape_get_device_reference(device);
+	tape_get_device(device);
 	INIT_WORK(&blkdat->requeue_task, tapeblock_requeue);
 
 	return 0;
@@ -274,13 +274,14 @@ tapeblock_cleanup_device(struct tape_device *device)
 	}
 
 	del_gendisk(device->blk_data.disk);
-	device->blk_data.disk->private_data =
-		tape_put_device(device->blk_data.disk->private_data);
+	device->blk_data.disk->private_data = NULL;
+	tape_put_device(device);
 	put_disk(device->blk_data.disk);
 
 	device->blk_data.disk = NULL;
 cleanup_queue:
-	device->blk_data.request_queue->queuedata = tape_put_device(device);
+	device->blk_data.request_queue->queuedata = NULL;
+	tape_put_device(device);
 
 	blk_cleanup_queue(device->blk_data.request_queue);
 	device->blk_data.request_queue = NULL;
@@ -363,7 +364,7 @@ tapeblock_open(struct block_device *bdev, fmode_t mode)
 	struct tape_device *	device;
 	int			rc;
 
-	device = tape_get_device_reference(disk->private_data);
+	device = tape_get_device(disk->private_data);
 
 	if (device->required_tapemarks) {
 		DBF_EVENT(2, "TBLOCK: missing tapemarks\n");

drivers/s390/char/tape_char.c

@@ -33,8 +33,7 @@ static ssize_t tapechar_read(struct file *, char __user *, size_t, loff_t *);
 static ssize_t tapechar_write(struct file *, const char __user *, size_t, loff_t *);
 static int tapechar_open(struct inode *,struct file *);
 static int tapechar_release(struct inode *,struct file *);
-static int tapechar_ioctl(struct inode *, struct file *, unsigned int,
-			  unsigned long);
+static long tapechar_ioctl(struct file *, unsigned int, unsigned long);
 static long tapechar_compat_ioctl(struct file *, unsigned int,
 			  unsigned long);
 
@@ -43,7 +42,7 @@ static const struct file_operations tape_fops =
 	.owner = THIS_MODULE,
 	.read = tapechar_read,
 	.write = tapechar_write,
-	.ioctl = tapechar_ioctl,
+	.unlocked_ioctl = tapechar_ioctl,
 	.compat_ioctl = tapechar_compat_ioctl,
 	.open = tapechar_open,
 	.release = tapechar_release,
@@ -170,7 +169,6 @@ tapechar_read(struct file *filp, char __user *data, size_t count, loff_t *ppos)
 	if (rc == 0) {
 		rc = block_size - request->rescnt;
 		DBF_EVENT(6, "TCHAR:rbytes:  %x\n", rc);
-		filp->f_pos += rc;
 		/* Copy data from idal buffer to user space. */
 		if (idal_buffer_to_user(device->char_data.idal_buf,
 					data, rc) != 0)
@@ -238,7 +236,6 @@ tapechar_write(struct file *filp, const char __user *data, size_t count, loff_t
 			break;
 		DBF_EVENT(6, "TCHAR:wbytes: %lx\n",
 			  block_size - request->rescnt);
-		filp->f_pos += block_size - request->rescnt;
 		written += block_size - request->rescnt;
 		if (request->rescnt != 0)
 			break;
@@ -286,26 +283,20 @@ tapechar_open (struct inode *inode, struct file *filp)
 	if (imajor(filp->f_path.dentry->d_inode) != tapechar_major)
 		return -ENODEV;
 
-	lock_kernel();
 	minor = iminor(filp->f_path.dentry->d_inode);
-	device = tape_get_device(minor / TAPE_MINORS_PER_DEV);
+	device = tape_find_device(minor / TAPE_MINORS_PER_DEV);
 	if (IS_ERR(device)) {
-		DBF_EVENT(3, "TCHAR:open: tape_get_device() failed\n");
-		rc = PTR_ERR(device);
-		goto out;
+		DBF_EVENT(3, "TCHAR:open: tape_find_device() failed\n");
+		return PTR_ERR(device);
 	}
 
-
 	rc = tape_open(device);
 	if (rc == 0) {
 		filp->private_data = device;
-		rc = nonseekable_open(inode, filp);
-	}
-	else
+		nonseekable_open(inode, filp);
+	} else
 		tape_put_device(device);
 
-out:
-	unlock_kernel();
 	return rc;
 }
 
@@ -342,7 +333,8 @@ tapechar_release(struct inode *inode, struct file *filp)
 		device->char_data.idal_buf = NULL;
 	}
 	tape_release(device);
-	filp->private_data = tape_put_device(device);
+	filp->private_data = NULL;
+	tape_put_device(device);
 
 	return 0;
 }
@@ -351,16 +343,11 @@ tapechar_release(struct inode *inode, struct file *filp)
  * Tape device io controls.
  */
 static int
-tapechar_ioctl(struct inode *inp, struct file *filp,
-	       unsigned int no, unsigned long data)
+__tapechar_ioctl(struct tape_device *device,
+		 unsigned int no, unsigned long data)
 {
-	struct tape_device *device;
 	int rc;
 
-	DBF_EVENT(6, "TCHAR:ioct\n");
-
-	device = (struct tape_device *) filp->private_data;
-
 	if (no == MTIOCTOP) {
 		struct mtop op;
 
@@ -452,6 +439,21 @@ tapechar_ioctl(struct inode *inp, struct file *filp,
 	return device->discipline->ioctl_fn(device, no, data);
 }
 
+static long
+tapechar_ioctl(struct file *filp, unsigned int no, unsigned long data)
+{
+	struct tape_device *device;
+	long rc;
+
+	DBF_EVENT(6, "TCHAR:ioct\n");
+
+	device = (struct tape_device *) filp->private_data;
+	mutex_lock(&device->mutex);
+	rc = __tapechar_ioctl(device, no, data);
+	mutex_unlock(&device->mutex);
+	return rc;
+}
+
 static long
 tapechar_compat_ioctl(struct file *filp, unsigned int no, unsigned long data)
 {
@@ -459,9 +461,9 @@ tapechar_compat_ioctl(struct file *filp, unsigned int no, unsigned long data)
 	int rval = -ENOIOCTLCMD;
 
 	if (device->discipline->ioctl_fn) {
-		lock_kernel();
+		mutex_lock(&device->mutex);
 		rval = device->discipline->ioctl_fn(device, no, data);
-		unlock_kernel();
+		mutex_unlock(&device->mutex);
 		if (rval == -EINVAL)
 			rval = -ENOIOCTLCMD;
 	}

drivers/s390/char/tape_core.c

@@ -492,6 +492,7 @@ tape_alloc_device(void)
 		kfree(device);
 		return ERR_PTR(-ENOMEM);
 	}
+	mutex_init(&device->mutex);
 	INIT_LIST_HEAD(&device->req_queue);
 	INIT_LIST_HEAD(&device->node);
 	init_waitqueue_head(&device->state_change_wq);
@@ -511,11 +512,12 @@ tape_alloc_device(void)
  * increment the reference count.
  */
 struct tape_device *
-tape_get_device_reference(struct tape_device *device)
+tape_get_device(struct tape_device *device)
 {
-	DBF_EVENT(4, "tape_get_device_reference(%p) = %i\n", device,
-		atomic_inc_return(&device->ref_count));
+	int count;
 
+	count = atomic_inc_return(&device->ref_count);
+	DBF_EVENT(4, "tape_get_device(%p) = %i\n", device, count);
 	return device;
 }
 
@@ -525,32 +527,25 @@ tape_get_device_reference(struct tape_device *device)
  * The function returns a NULL pointer to be used by the caller
  * for clearing reference pointers.
  */
-struct tape_device *
+void
 tape_put_device(struct tape_device *device)
 {
-	int remain;
+	int count;
 
-	remain = atomic_dec_return(&device->ref_count);
-	if (remain > 0) {
-		DBF_EVENT(4, "tape_put_device(%p) -> %i\n", device, remain);
-	} else {
-		if (remain < 0) {
-			DBF_EVENT(4, "put device without reference\n");
-		} else {
-			DBF_EVENT(4, "tape_free_device(%p)\n", device);
-			kfree(device->modeset_byte);
-			kfree(device);
-		}
+	count = atomic_dec_return(&device->ref_count);
+	DBF_EVENT(4, "tape_put_device(%p) -> %i\n", device, count);
+	BUG_ON(count < 0);
+	if (count == 0) {
+		kfree(device->modeset_byte);
+		kfree(device);
 	}
-
-	return NULL;			
 }
 
 /*
  * Find tape device by a device index.
  */
 struct tape_device *
-tape_get_device(int devindex)
+tape_find_device(int devindex)
 {
 	struct tape_device *device, *tmp;
 
@@ -558,7 +553,7 @@ tape_get_device(int devindex)
 	read_lock(&tape_device_lock);
 	list_for_each_entry(tmp, &tape_device_list, node) {
 		if (tmp->first_minor / TAPE_MINORS_PER_DEV == devindex) {
-			device = tape_get_device_reference(tmp);
+			device = tape_get_device(tmp);
 			break;
 		}
 	}
@@ -579,7 +574,8 @@ tape_generic_probe(struct ccw_device *cdev)
 	device = tape_alloc_device();
 	if (IS_ERR(device))
 		return -ENODEV;
-	ccw_device_set_options(cdev, CCWDEV_DO_PATHGROUP);
+	ccw_device_set_options(cdev, CCWDEV_DO_PATHGROUP |
+				     CCWDEV_DO_MULTIPATH);
 	ret = sysfs_create_group(&cdev->dev.kobj, &tape_attr_group);
 	if (ret) {
 		tape_put_device(device);
@@ -606,7 +602,8 @@ __tape_discard_requests(struct tape_device *device)
 		list_del(&request->list);
 
 		/* Decrease ref_count for removed request. */
-		request->device = tape_put_device(device);
+		request->device = NULL;
+		tape_put_device(device);
 		request->rc = -EIO;
 		if (request->callback != NULL)
 			request->callback(request, request->callback_data);
@@ -664,9 +661,11 @@ tape_generic_remove(struct ccw_device *cdev)
 			tape_cleanup_device(device);
 	}
 
-	if (!dev_get_drvdata(&cdev->dev)) {
+	device = dev_get_drvdata(&cdev->dev);
+	if (device) {
 		sysfs_remove_group(&cdev->dev.kobj, &tape_attr_group);
-		dev_set_drvdata(&cdev->dev, tape_put_device(dev_get_drvdata(&cdev->dev)));
+		dev_set_drvdata(&cdev->dev, NULL);
+		tape_put_device(device);
 	}
 }
 
@@ -721,9 +720,8 @@ tape_free_request (struct tape_request * request)
 {
 	DBF_LH(6, "Free request %p\n", request);
 
-	if (request->device != NULL) {
-		request->device = tape_put_device(request->device);
-	}
+	if (request->device)
+		tape_put_device(request->device);
 	kfree(request->cpdata);
 	kfree(request->cpaddr);
 	kfree(request);
@@ -838,7 +836,8 @@ static void tape_long_busy_timeout(unsigned long data)
 	BUG_ON(request->status != TAPE_REQUEST_LONG_BUSY);
 	DBF_LH(6, "%08x: Long busy timeout.\n", device->cdev_id);
 	__tape_start_next_request(device);
-	device->lb_timeout.data = (unsigned long) tape_put_device(device);
+	device->lb_timeout.data = 0UL;
+	tape_put_device(device);
 	spin_unlock_irq(get_ccwdev_lock(device->cdev));
 }
 
@@ -918,7 +917,7 @@ __tape_start_request(struct tape_device *device, struct tape_request *request)
 	}
 
 	/* Increase use count of device for the added request. */
-	request->device = tape_get_device_reference(device);
+	request->device = tape_get_device(device);
 
 	if (list_empty(&device->req_queue)) {
 		/* No other requests are on the queue. Start this one. */
@@ -1117,8 +1116,8 @@ __tape_do_irq (struct ccw_device *cdev, unsigned long intparm, struct irb *irb)
 		if (req->status == TAPE_REQUEST_LONG_BUSY) {
 			DBF_EVENT(3, "(%08x): del timer\n", device->cdev_id);
 			if (del_timer(&device->lb_timeout)) {
-				device->lb_timeout.data = (unsigned long)
-					tape_put_device(device);
+				device->lb_timeout.data = 0UL;
+				tape_put_device(device);
 				__tape_start_next_request(device);
 			}
 			return;
@@ -1173,7 +1172,7 @@ __tape_do_irq (struct ccw_device *cdev, unsigned long intparm, struct irb *irb)
 			break;
 		case TAPE_IO_LONG_BUSY:
 			device->lb_timeout.data =
-				(unsigned long)tape_get_device_reference(device);
+				(unsigned long) tape_get_device(device);
 			device->lb_timeout.expires = jiffies +
 				LONG_BUSY_TIMEOUT * HZ;
 			DBF_EVENT(3, "(%08x): add timer\n", device->cdev_id);
@@ -1326,7 +1325,7 @@ EXPORT_SYMBOL(tape_generic_online);
 EXPORT_SYMBOL(tape_generic_offline);
 EXPORT_SYMBOL(tape_generic_pm_suspend);
 EXPORT_SYMBOL(tape_put_device);
-EXPORT_SYMBOL(tape_get_device_reference);
+EXPORT_SYMBOL(tape_get_device);
 EXPORT_SYMBOL(tape_state_verbose);
 EXPORT_SYMBOL(tape_op_verbose);
 EXPORT_SYMBOL(tape_state_set);

drivers/s390/char/tape_proc.c

@@ -45,7 +45,7 @@ static int tape_proc_show(struct seq_file *m, void *v)
 		seq_printf(m, "TapeNo\tBusID      CuType/Model\t"
 			"DevType/Model\tBlkSize\tState\tOp\tMedState\n");
 	}
-	device = tape_get_device(n);
+	device = tape_find_device(n);
 	if (IS_ERR(device))
 		return 0;
 	spin_lock_irq(get_ccwdev_lock(device->cdev));

drivers/s390/char/tty3270.c

@@ -19,6 +19,7 @@
 
 #include <linux/slab.h>
 #include <linux/bootmem.h>
+#include <linux/compat.h>
 
 #include <asm/ccwdev.h>
 #include <asm/cio.h>
@@ -1731,6 +1732,22 @@ tty3270_ioctl(struct tty_struct *tty, struct file *file,
 	return kbd_ioctl(tp->kbd, file, cmd, arg);
 }
 
+#ifdef CONFIG_COMPAT
+static long
+tty3270_compat_ioctl(struct tty_struct *tty, struct file *file,
+	      unsigned int cmd, unsigned long arg)
+{
+	struct tty3270 *tp;
+
+	tp = tty->driver_data;
+	if (!tp)
+		return -ENODEV;
+	if (tty->flags & (1 << TTY_IO_ERROR))
+		return -EIO;
+	return kbd_ioctl(tp->kbd, file, cmd, (unsigned long)compat_ptr(arg));
+}
+#endif
+
 static const struct tty_operations tty3270_ops = {
 	.open = tty3270_open,
 	.close = tty3270_close,
@@ -1745,6 +1762,9 @@ static const struct tty_operations tty3270_ops = {
 	.hangup = tty3270_hangup,
 	.wait_until_sent = tty3270_wait_until_sent,
 	.ioctl = tty3270_ioctl,
+#ifdef CONFIG_COMPAT
+	.compat_ioctl = tty3270_compat_ioctl,
+#endif
 	.set_termios = tty3270_set_termios
 };
 

drivers/s390/char/vmlogrdr.c

@@ -312,11 +312,9 @@ static int vmlogrdr_open (struct inode *inode, struct file *filp)
 		return -ENOSYS;
 
 	/* Besure this device hasn't already been opened */
-	lock_kernel();
 	spin_lock_bh(&logptr->priv_lock);
 	if (logptr->dev_in_use)	{
 		spin_unlock_bh(&logptr->priv_lock);
-		unlock_kernel();
 		return -EBUSY;
 	}
 	logptr->dev_in_use = 1;
@@ -360,9 +358,8 @@ static int vmlogrdr_open (struct inode *inode, struct file *filp)
 		   || (logptr->iucv_path_severed));
 	if (logptr->iucv_path_severed)
 		goto out_record;
- 	ret = nonseekable_open(inode, filp);
-	unlock_kernel();
-	return ret;
+	nonseekable_open(inode, filp);
+	return 0;
 
 out_record:
 	if (logptr->autorecording)
@@ -372,7 +369,6 @@ out_path:
 	logptr->path = NULL;
 out_dev:
 	logptr->dev_in_use = 0;
-	unlock_kernel();
 	return -EIO;
 }
 

drivers/s390/char/vmur.c

@@ -695,7 +695,6 @@ static int ur_open(struct inode *inode, struct file *file)
 
 	if (accmode == O_RDWR)
 		return -EACCES;
-	lock_kernel();
 	/*
 	 * We treat the minor number as the devno of the ur device
 	 * to find in the driver tree.
@@ -749,7 +748,6 @@ static int ur_open(struct inode *inode, struct file *file)
 		goto fail_urfile_free;
 	urf->file_reclen = rc;
 	file->private_data = urf;
-	unlock_kernel();
 	return 0;
 
 fail_urfile_free:
@@ -761,7 +759,6 @@ fail_unlock:
 fail_put:
 	urdev_put(urd);
 out:
-	unlock_kernel();
 	return rc;
 }
 

drivers/s390/char/vmwatchdog.c

@@ -19,7 +19,6 @@
 #include <linux/moduleparam.h>
 #include <linux/suspend.h>
 #include <linux/watchdog.h>
-#include <linux/smp_lock.h>
 
 #include <asm/ebcdic.h>
 #include <asm/io.h>
@@ -49,6 +48,8 @@ static unsigned int vmwdt_interval = 60;
 static unsigned long vmwdt_is_open;
 static int vmwdt_expect_close;
 
+static DEFINE_MUTEX(vmwdt_mutex);
+
 #define VMWDT_OPEN	0	/* devnode is open or suspend in progress */
 #define VMWDT_RUNNING	1	/* The watchdog is armed */
 
@@ -133,15 +134,11 @@ static int __init vmwdt_probe(void)
 static int vmwdt_open(struct inode *i, struct file *f)
 {
 	int ret;
-	lock_kernel();
-	if (test_and_set_bit(VMWDT_OPEN, &vmwdt_is_open)) {
-		unlock_kernel();
+	if (test_and_set_bit(VMWDT_OPEN, &vmwdt_is_open))
 		return -EBUSY;
-	}
 	ret = vmwdt_keepalive();
 	if (ret)
 		clear_bit(VMWDT_OPEN, &vmwdt_is_open);
-	unlock_kernel();
 	return ret ? ret : nonseekable_open(i, f);
 }
 
@@ -160,8 +157,7 @@ static struct watchdog_info vmwdt_info = {
 	.identity = "z/VM Watchdog Timer",
 };
 
-static int vmwdt_ioctl(struct inode *i, struct file *f,
-			  unsigned int cmd, unsigned long arg)
+static int __vmwdt_ioctl(unsigned int cmd, unsigned long arg)
 {
 	switch (cmd) {
 	case WDIOC_GETSUPPORT:
@@ -205,10 +201,19 @@ static int vmwdt_ioctl(struct inode *i, struct file *f,
 	case WDIOC_KEEPALIVE:
 		return vmwdt_keepalive();
 	}
-
 	return -EINVAL;
 }
 
+static long vmwdt_ioctl(struct file *f, unsigned int cmd, unsigned long arg)
+{
+	int rc;
+
+	mutex_lock(&vmwdt_mutex);
+	rc = __vmwdt_ioctl(cmd, arg);
+	mutex_unlock(&vmwdt_mutex);
+	return (long) rc;
+}
+
 static ssize_t vmwdt_write(struct file *f, const char __user *buf,
 				size_t count, loff_t *ppos)
 {
@@ -288,7 +293,7 @@ static struct notifier_block vmwdt_power_notifier = {
 static const struct file_operations vmwdt_fops = {
 	.open    = &vmwdt_open,
 	.release = &vmwdt_close,
-	.ioctl   = &vmwdt_ioctl,
+	.unlocked_ioctl = &vmwdt_ioctl,
 	.write   = &vmwdt_write,
 	.owner   = THIS_MODULE,
 };
@@ -309,6 +314,10 @@ static int __init vmwdt_init(void)
 	ret = register_pm_notifier(&vmwdt_power_notifier);
 	if (ret)
 		return ret;
+	/*
+	 * misc_register() has to be the last action in module_init(), because
+	 * file operations will be available right after this.
+	 */
 	ret = misc_register(&vmwdt_dev);
 	if (ret) {
 		unregister_pm_notifier(&vmwdt_power_notifier);

drivers/s390/cio/Makefile

@@ -3,7 +3,7 @@
 #
 
 obj-y += airq.o blacklist.o chsc.o cio.o css.o chp.o idset.o isc.o \
-	fcx.o itcw.o crw.o
+	fcx.o itcw.o crw.o ccwreq.o
 ccw_device-objs += device.o device_fsm.o device_ops.o
 ccw_device-objs += device_id.o device_pgid.o device_status.o
 obj-y += ccw_device.o cmf.o

drivers/s390/cio/ccwreq.c

@@ -0,0 +1,328 @@
+/*
+ *  Handling of internal CCW device requests.
+ *
+ *    Copyright IBM Corp. 2009
+ *    Author(s): Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
+ */
+
+#include <linux/types.h>
+#include <linux/err.h>
+#include <asm/ccwdev.h>
+#include <asm/cio.h>
+
+#include "io_sch.h"
+#include "cio.h"
+#include "device.h"
+#include "cio_debug.h"
+
+/**
+ * lpm_adjust - adjust path mask
+ * @lpm: path mask to adjust
+ * @mask: mask of available paths
+ *
+ * Shift @lpm right until @lpm and @mask have at least one bit in common or
+ * until @lpm is zero. Return the resulting lpm.
+ */
+int lpm_adjust(int lpm, int mask)
+{
+	while (lpm && ((lpm & mask) == 0))
+		lpm >>= 1;
+	return lpm;
+}
+
+/*
+ * Adjust path mask to use next path and reset retry count. Return resulting
+ * path mask.
+ */
+static u16 ccwreq_next_path(struct ccw_device *cdev)
+{
+	struct ccw_request *req = &cdev->private->req;
+
+	req->retries	= req->maxretries;
+	req->mask	= lpm_adjust(req->mask >>= 1, req->lpm);
+
+	return req->mask;
+}
+
+/*
+ * Clean up device state and report to callback.
+ */
+static void ccwreq_stop(struct ccw_device *cdev, int rc)
+{
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+
+	if (req->done)
+		return;
+	req->done = 1;
+	ccw_device_set_timeout(cdev, 0);
+	memset(&cdev->private->irb, 0, sizeof(struct irb));
+	sch->lpm = sch->schib.pmcw.pam;
+	if (rc && rc != -ENODEV && req->drc)
+		rc = req->drc;
+	req->callback(cdev, req->data, rc);
+}
+
+/*
+ * (Re-)Start the operation until retries and paths are exhausted.
+ */
+static void ccwreq_do(struct ccw_device *cdev)
+{
+	struct ccw_request *req = &cdev->private->req;
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw1 *cp = req->cp;
+	int rc = -EACCES;
+
+	while (req->mask) {
+		if (req->retries-- == 0) {
+			/* Retries exhausted, try next path. */
+			ccwreq_next_path(cdev);
+			continue;
+		}
+		/* Perform start function. */
+		sch->lpm = 0xff;
+		memset(&cdev->private->irb, 0, sizeof(struct irb));
+		rc = cio_start(sch, cp, (u8) req->mask);
+		if (rc == 0) {
+			/* I/O started successfully. */
+			ccw_device_set_timeout(cdev, req->timeout);
+			return;
+		}
+		if (rc == -ENODEV) {
+			/* Permanent device error. */
+			break;
+		}
+		if (rc == -EACCES) {
+			/* Permant path error. */
+			ccwreq_next_path(cdev);
+			continue;
+		}
+		/* Temporary improper status. */
+		rc = cio_clear(sch);
+		if (rc)
+			break;
+		return;
+	}
+	ccwreq_stop(cdev, rc);
+}
+
+/**
+ * ccw_request_start - perform I/O request
+ * @cdev: ccw device
+ *
+ * Perform the I/O request specified by cdev->req.
+ */
+void ccw_request_start(struct ccw_device *cdev)
+{
+	struct ccw_request *req = &cdev->private->req;
+
+	/* Try all paths twice to counter link flapping. */
+	req->mask	= 0x8080;
+	req->retries	= req->maxretries;
+	req->mask	= lpm_adjust(req->mask, req->lpm);
+	req->drc	= 0;
+	req->done	= 0;
+	req->cancel	= 0;
+	if (!req->mask)
+		goto out_nopath;
+	ccwreq_do(cdev);
+	return;
+
+out_nopath:
+	ccwreq_stop(cdev, -EACCES);
+}
+
+/**
+ * ccw_request_cancel - cancel running I/O request
+ * @cdev: ccw device
+ *
+ * Cancel the I/O request specified by cdev->req. Return non-zero if request
+ * has already finished, zero otherwise.
+ */
+int ccw_request_cancel(struct ccw_device *cdev)
+{
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+	int rc;
+
+	if (req->done)
+		return 1;
+	req->cancel = 1;
+	rc = cio_clear(sch);
+	if (rc)
+		ccwreq_stop(cdev, rc);
+	return 0;
+}
+
+/*
+ * Return the status of the internal I/O started on the specified ccw device.
+ * Perform BASIC SENSE if required.
+ */
+static enum io_status ccwreq_status(struct ccw_device *cdev, struct irb *lcirb)
+{
+	struct irb *irb = &cdev->private->irb;
+	struct cmd_scsw *scsw = &irb->scsw.cmd;
+
+	/* Perform BASIC SENSE if needed. */
+	if (ccw_device_accumulate_and_sense(cdev, lcirb))
+		return IO_RUNNING;
+	/* Check for halt/clear interrupt. */
+	if (scsw->fctl & (SCSW_FCTL_HALT_FUNC | SCSW_FCTL_CLEAR_FUNC))
+		return IO_KILLED;
+	/* Check for path error. */
+	if (scsw->cc == 3 || scsw->pno)
+		return IO_PATH_ERROR;
+	/* Handle BASIC SENSE data. */
+	if (irb->esw.esw0.erw.cons) {
+		CIO_TRACE_EVENT(2, "sensedata");
+		CIO_HEX_EVENT(2, &cdev->private->dev_id,
+			      sizeof(struct ccw_dev_id));
+		CIO_HEX_EVENT(2, &cdev->private->irb.ecw, SENSE_MAX_COUNT);
+		/* Check for command reject. */
+		if (irb->ecw[0] & SNS0_CMD_REJECT)
+			return IO_REJECTED;
+		/* Assume that unexpected SENSE data implies an error. */
+		return IO_STATUS_ERROR;
+	}
+	/* Check for channel errors. */
+	if (scsw->cstat != 0)
+		return IO_STATUS_ERROR;
+	/* Check for device errors. */
+	if (scsw->dstat & ~(DEV_STAT_CHN_END | DEV_STAT_DEV_END))
+		return IO_STATUS_ERROR;
+	/* Check for final state. */
+	if (!(scsw->dstat & DEV_STAT_DEV_END))
+		return IO_RUNNING;
+	/* Check for other improper status. */
+	if (scsw->cc == 1 && (scsw->stctl & SCSW_STCTL_ALERT_STATUS))
+		return IO_STATUS_ERROR;
+	return IO_DONE;
+}
+
+/*
+ * Log ccw request status.
+ */
+static void ccwreq_log_status(struct ccw_device *cdev, enum io_status status)
+{
+	struct ccw_request *req = &cdev->private->req;
+	struct {
+		struct ccw_dev_id dev_id;
+		u16 retries;
+		u8 lpm;
+		u8 status;
+	}  __attribute__ ((packed)) data;
+	data.dev_id	= cdev->private->dev_id;
+	data.retries	= req->retries;
+	data.lpm	= (u8) req->mask;
+	data.status	= (u8) status;
+	CIO_TRACE_EVENT(2, "reqstat");
+	CIO_HEX_EVENT(2, &data, sizeof(data));
+}
+
+/**
+ * ccw_request_handler - interrupt handler for I/O request procedure.
+ * @cdev: ccw device
+ *
+ * Handle interrupt during I/O request procedure.
+ */
+void ccw_request_handler(struct ccw_device *cdev)
+{
+	struct ccw_request *req = &cdev->private->req;
+	struct irb *irb = (struct irb *) __LC_IRB;
+	enum io_status status;
+	int rc = -EOPNOTSUPP;
+
+	/* Check status of I/O request. */
+	status = ccwreq_status(cdev, irb);
+	if (req->filter)
+		status = req->filter(cdev, req->data, irb, status);
+	if (status != IO_RUNNING)
+		ccw_device_set_timeout(cdev, 0);
+	if (status != IO_DONE && status != IO_RUNNING)
+		ccwreq_log_status(cdev, status);
+	switch (status) {
+	case IO_DONE:
+		break;
+	case IO_RUNNING:
+		return;
+	case IO_REJECTED:
+		goto err;
+	case IO_PATH_ERROR:
+		goto out_next_path;
+	case IO_STATUS_ERROR:
+		goto out_restart;
+	case IO_KILLED:
+		/* Check if request was cancelled on purpose. */
+		if (req->cancel) {
+			rc = -EIO;
+			goto err;
+		}
+		goto out_restart;
+	}
+	/* Check back with request initiator. */
+	if (!req->check)
+		goto out;
+	switch (req->check(cdev, req->data)) {
+	case 0:
+		break;
+	case -EAGAIN:
+		goto out_restart;
+	case -EACCES:
+		goto out_next_path;
+	default:
+		goto err;
+	}
+out:
+	ccwreq_stop(cdev, 0);
+	return;
+
+out_next_path:
+	/* Try next path and restart I/O. */
+	if (!ccwreq_next_path(cdev)) {
+		rc = -EACCES;
+		goto err;
+	}
+out_restart:
+	/* Restart. */
+	ccwreq_do(cdev);
+	return;
+err:
+	ccwreq_stop(cdev, rc);
+}
+
+
+/**
+ * ccw_request_timeout - timeout handler for I/O request procedure
+ * @cdev: ccw device
+ *
+ * Handle timeout during I/O request procedure.
+ */
+void ccw_request_timeout(struct ccw_device *cdev)
+{
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+	int rc;
+
+	if (!ccwreq_next_path(cdev)) {
+		/* set the final return code for this request */
+		req->drc = -ETIME;
+	}
+	rc = cio_clear(sch);
+	if (rc)
+		goto err;
+	return;
+
+err:
+	ccwreq_stop(cdev, rc);
+}
+
+/**
+ * ccw_request_notoper - notoper handler for I/O request procedure
+ * @cdev: ccw device
+ *
+ * Handle timeout during I/O request procedure.
+ */
+void ccw_request_notoper(struct ccw_device *cdev)
+{
+	ccwreq_stop(cdev, -ENODEV);
+}

drivers/s390/cio/cio.h

@@ -68,6 +68,11 @@ struct schib {
 	__u8 mda[4];		 /* model dependent area */
 } __attribute__ ((packed,aligned(4)));
 
+enum sch_todo {
+	SCH_TODO_NOTHING,
+	SCH_TODO_UNREG,
+};
+
 /* subchannel data structure used by I/O subroutines */
 struct subchannel {
 	struct subchannel_id schid;
@@ -95,7 +100,8 @@ struct subchannel {
 	struct device dev;	/* entry in device tree */
 	struct css_driver *driver;
 	void *private; /* private per subchannel type data */
-	struct work_struct work;
+	enum sch_todo todo;
+	struct work_struct todo_work;
 	struct schib_config config;
 } __attribute__ ((aligned(8)));
 

drivers/s390/cio/css.c

@@ -133,6 +133,8 @@ out:
 	return rc;
 }
 
+static void css_sch_todo(struct work_struct *work);
+
 static struct subchannel *
 css_alloc_subchannel(struct subchannel_id schid)
 {
@@ -147,6 +149,7 @@ css_alloc_subchannel(struct subchannel_id schid)
 		kfree(sch);
 		return ERR_PTR(ret);
 	}
+	INIT_WORK(&sch->todo_work, css_sch_todo);
 	return sch;
 }
 
@@ -190,6 +193,51 @@ void css_sch_device_unregister(struct subchannel *sch)
 }
 EXPORT_SYMBOL_GPL(css_sch_device_unregister);
 
+static void css_sch_todo(struct work_struct *work)
+{
+	struct subchannel *sch;
+	enum sch_todo todo;
+
+	sch = container_of(work, struct subchannel, todo_work);
+	/* Find out todo. */
+	spin_lock_irq(sch->lock);
+	todo = sch->todo;
+	CIO_MSG_EVENT(4, "sch_todo: sch=0.%x.%04x, todo=%d\n", sch->schid.ssid,
+		      sch->schid.sch_no, todo);
+	sch->todo = SCH_TODO_NOTHING;
+	spin_unlock_irq(sch->lock);
+	/* Perform todo. */
+	if (todo == SCH_TODO_UNREG)
+		css_sch_device_unregister(sch);
+	/* Release workqueue ref. */
+	put_device(&sch->dev);
+}
+
+/**
+ * css_sched_sch_todo - schedule a subchannel operation
+ * @sch: subchannel
+ * @todo: todo
+ *
+ * Schedule the operation identified by @todo to be performed on the slow path
+ * workqueue. Do nothing if another operation with higher priority is already
+ * scheduled. Needs to be called with subchannel lock held.
+ */
+void css_sched_sch_todo(struct subchannel *sch, enum sch_todo todo)
+{
+	CIO_MSG_EVENT(4, "sch_todo: sched sch=0.%x.%04x todo=%d\n",
+		      sch->schid.ssid, sch->schid.sch_no, todo);
+	if (sch->todo >= todo)
+		return;
+	/* Get workqueue ref. */
+	if (!get_device(&sch->dev))
+		return;
+	sch->todo = todo;
+	if (!queue_work(slow_path_wq, &sch->todo_work)) {
+		/* Already queued, release workqueue ref. */
+		put_device(&sch->dev);
+	}
+}
+
 static void ssd_from_pmcw(struct chsc_ssd_info *ssd, struct pmcw *pmcw)
 {
 	int i;
@@ -376,8 +424,8 @@ static int css_evaluate_new_subchannel(struct subchannel_id schid, int slow)
 		/* Unusable - ignore. */
 		return 0;
 	}
-	CIO_MSG_EVENT(4, "Evaluating schid 0.%x.%04x, event %d, unknown, "
-			 "slow path.\n", schid.ssid, schid.sch_no, CIO_OPER);
+	CIO_MSG_EVENT(4, "event: sch 0.%x.%04x, new\n", schid.ssid,
+		      schid.sch_no);
 
 	return css_probe_device(schid);
 }
@@ -394,6 +442,10 @@ static int css_evaluate_known_subchannel(struct subchannel *sch, int slow)
 				"Got subchannel machine check but "
 				"no sch_event handler provided.\n");
 	}
+	if (ret != 0 && ret != -EAGAIN) {
+		CIO_MSG_EVENT(2, "eval: sch 0.%x.%04x, rc=%d\n",
+			      sch->schid.ssid, sch->schid.sch_no, ret);
+	}
 	return ret;
 }
 
@@ -684,6 +736,7 @@ static int __init setup_css(int nr)
 	css->pseudo_subchannel->dev.parent = &css->device;
 	css->pseudo_subchannel->dev.release = css_subchannel_release;
 	dev_set_name(&css->pseudo_subchannel->dev, "defunct");
+	mutex_init(&css->pseudo_subchannel->reg_mutex);
 	ret = cio_create_sch_lock(css->pseudo_subchannel);
 	if (ret) {
 		kfree(css->pseudo_subchannel);

drivers/s390/cio/css.h

@@ -11,6 +11,8 @@
 #include <asm/chpid.h>
 #include <asm/schid.h>
 
+#include "cio.h"
+
 /*
  * path grouping stuff
  */
@@ -151,4 +153,5 @@ int css_sch_is_valid(struct schib *);
 
 extern struct workqueue_struct *slow_path_wq;
 void css_wait_for_slow_path(void);
+void css_sched_sch_todo(struct subchannel *sch, enum sch_todo todo);
 #endif

drivers/s390/cio/device.c

@@ -7,6 +7,10 @@
  *		 Cornelia Huck (cornelia.huck@de.ibm.com)
  *		 Martin Schwidefsky (schwidefsky@de.ibm.com)
  */
+
+#define KMSG_COMPONENT "cio"
+#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
+
 #include <linux/module.h>
 #include <linux/init.h>
 #include <linux/spinlock.h>
@@ -299,53 +303,18 @@ int ccw_device_is_orphan(struct ccw_device *cdev)
 
 static void ccw_device_unregister(struct ccw_device *cdev)
 {
-	if (test_and_clear_bit(1, &cdev->private->registered)) {
+	if (device_is_registered(&cdev->dev)) {
+		/* Undo device_add(). */
 		device_del(&cdev->dev);
+	}
+	if (cdev->private->flags.initialized) {
+		cdev->private->flags.initialized = 0;
 		/* Release reference from device_initialize(). */
 		put_device(&cdev->dev);
 	}
 }
 
-static void ccw_device_remove_orphan_cb(struct work_struct *work)
-{
-	struct ccw_device_private *priv;
-	struct ccw_device *cdev;
-
-	priv = container_of(work, struct ccw_device_private, kick_work);
-	cdev = priv->cdev;
-	ccw_device_unregister(cdev);
-	/* Release cdev reference for workqueue processing. */
-	put_device(&cdev->dev);
-}
-
-static void
-ccw_device_remove_disconnected(struct ccw_device *cdev)
-{
-	unsigned long flags;
-
-	/*
-	 * Forced offline in disconnected state means
-	 * 'throw away device'.
-	 */
-	if (ccw_device_is_orphan(cdev)) {
-		/*
-		 * Deregister ccw device.
-		 * Unfortunately, we cannot do this directly from the
-		 * attribute method.
-		 */
-		/* Get cdev reference for workqueue processing. */
-		if (!get_device(&cdev->dev))
-			return;
-		spin_lock_irqsave(cdev->ccwlock, flags);
-		cdev->private->state = DEV_STATE_NOT_OPER;
-		spin_unlock_irqrestore(cdev->ccwlock, flags);
-		PREPARE_WORK(&cdev->private->kick_work,
-				ccw_device_remove_orphan_cb);
-		queue_work(slow_path_wq, &cdev->private->kick_work);
-	} else
-		/* Deregister subchannel, which will kill the ccw device. */
-		ccw_device_schedule_sch_unregister(cdev);
-}
+static void io_subchannel_quiesce(struct subchannel *);
 
 /**
  * ccw_device_set_offline() - disable a ccw device for I/O
@@ -360,7 +329,8 @@ ccw_device_remove_disconnected(struct ccw_device *cdev)
  */
 int ccw_device_set_offline(struct ccw_device *cdev)
 {
-	int ret;
+	struct subchannel *sch;
+	int ret, state;
 
 	if (!cdev)
 		return -ENODEV;
@@ -374,6 +344,7 @@ int ccw_device_set_offline(struct ccw_device *cdev)
 	}
 	cdev->online = 0;
 	spin_lock_irq(cdev->ccwlock);
+	sch = to_subchannel(cdev->dev.parent);
 	/* Wait until a final state or DISCONNECTED is reached */
 	while (!dev_fsm_final_state(cdev) &&
 	       cdev->private->state != DEV_STATE_DISCONNECTED) {
@@ -382,20 +353,37 @@ int ccw_device_set_offline(struct ccw_device *cdev)
 			   cdev->private->state == DEV_STATE_DISCONNECTED));
 		spin_lock_irq(cdev->ccwlock);
 	}
-	ret = ccw_device_offline(cdev);
-	if (ret)
-		goto error;
+	do {
+		ret = ccw_device_offline(cdev);
+		if (!ret)
+			break;
+		CIO_MSG_EVENT(0, "ccw_device_offline returned %d, device "
+			      "0.%x.%04x\n", ret, cdev->private->dev_id.ssid,
+			      cdev->private->dev_id.devno);
+		if (ret != -EBUSY)
+			goto error;
+		state = cdev->private->state;
+		spin_unlock_irq(cdev->ccwlock);
+		io_subchannel_quiesce(sch);
+		spin_lock_irq(cdev->ccwlock);
+		cdev->private->state = state;
+	} while (ret == -EBUSY);
 	spin_unlock_irq(cdev->ccwlock);
 	wait_event(cdev->private->wait_q, (dev_fsm_final_state(cdev) ||
 		   cdev->private->state == DEV_STATE_DISCONNECTED));
+	/* Inform the user if set offline failed. */
+	if (cdev->private->state == DEV_STATE_BOXED) {
+		pr_warning("%s: The device entered boxed state while "
+			   "being set offline\n", dev_name(&cdev->dev));
+	} else if (cdev->private->state == DEV_STATE_NOT_OPER) {
+		pr_warning("%s: The device stopped operating while "
+			   "being set offline\n", dev_name(&cdev->dev));
+	}
 	/* Give up reference from ccw_device_set_online(). */
 	put_device(&cdev->dev);
 	return 0;
 
 error:
-	CIO_MSG_EVENT(0, "ccw_device_offline returned %d, device 0.%x.%04x\n",
-		      ret, cdev->private->dev_id.ssid,
-		      cdev->private->dev_id.devno);
 	cdev->private->state = DEV_STATE_OFFLINE;
 	dev_fsm_event(cdev, DEV_EVENT_NOTOPER);
 	spin_unlock_irq(cdev->ccwlock);
@@ -448,6 +436,16 @@ int ccw_device_set_online(struct ccw_device *cdev)
 	if ((cdev->private->state != DEV_STATE_ONLINE) &&
 	    (cdev->private->state != DEV_STATE_W4SENSE)) {
 		spin_unlock_irq(cdev->ccwlock);
+		/* Inform the user that set online failed. */
+		if (cdev->private->state == DEV_STATE_BOXED) {
+			pr_warning("%s: Setting the device online failed "
+				   "because it is boxed\n",
+				   dev_name(&cdev->dev));
+		} else if (cdev->private->state == DEV_STATE_NOT_OPER) {
+			pr_warning("%s: Setting the device online failed "
+				   "because it is not operational\n",
+				   dev_name(&cdev->dev));
+		}
 		/* Give up online reference since onlining failed. */
 		put_device(&cdev->dev);
 		return -ENODEV;
@@ -494,27 +492,22 @@ error:
 
 static int online_store_handle_offline(struct ccw_device *cdev)
 {
-	if (cdev->private->state == DEV_STATE_DISCONNECTED)
-		ccw_device_remove_disconnected(cdev);
-	else if (cdev->online && cdev->drv && cdev->drv->set_offline)
+	if (cdev->private->state == DEV_STATE_DISCONNECTED) {
+		spin_lock_irq(cdev->ccwlock);
+		ccw_device_sched_todo(cdev, CDEV_TODO_UNREG_EVAL);
+		spin_unlock_irq(cdev->ccwlock);
+	} else if (cdev->online && cdev->drv && cdev->drv->set_offline)
 		return ccw_device_set_offline(cdev);
 	return 0;
 }
 
 static int online_store_recog_and_online(struct ccw_device *cdev)
 {
-	int ret;
-
 	/* Do device recognition, if needed. */
 	if (cdev->private->state == DEV_STATE_BOXED) {
-		ret = ccw_device_recognition(cdev);
-		if (ret) {
-			CIO_MSG_EVENT(0, "Couldn't start recognition "
-				      "for device 0.%x.%04x (ret=%d)\n",
-				      cdev->private->dev_id.ssid,
-				      cdev->private->dev_id.devno, ret);
-			return ret;
-		}
+		spin_lock_irq(cdev->ccwlock);
+		ccw_device_recognition(cdev);
+		spin_unlock_irq(cdev->ccwlock);
 		wait_event(cdev->private->wait_q,
 			   cdev->private->flags.recog_done);
 		if (cdev->private->state != DEV_STATE_OFFLINE)
@@ -553,11 +546,10 @@ static ssize_t online_store (struct device *dev, struct device_attribute *attr,
 	int force, ret;
 	unsigned long i;
 
-	if ((cdev->private->state != DEV_STATE_OFFLINE &&
-	     cdev->private->state != DEV_STATE_ONLINE &&
-	     cdev->private->state != DEV_STATE_BOXED &&
-	     cdev->private->state != DEV_STATE_DISCONNECTED) ||
-	    atomic_cmpxchg(&cdev->private->onoff, 0, 1) != 0)
+	if (!dev_fsm_final_state(cdev) &&
+	    cdev->private->state != DEV_STATE_DISCONNECTED)
+		return -EAGAIN;
+	if (atomic_cmpxchg(&cdev->private->onoff, 0, 1) != 0)
 		return -EAGAIN;
 
 	if (cdev->drv && !try_module_get(cdev->drv->owner)) {
@@ -665,81 +657,31 @@ static int ccw_device_register(struct ccw_device *cdev)
 			   cdev->private->dev_id.devno);
 	if (ret)
 		return ret;
-	ret = device_add(dev);
-	if (ret)
-		return ret;
-
-	set_bit(1, &cdev->private->registered);
-	return ret;
+	return device_add(dev);
 }
 
-struct match_data {
-	struct ccw_dev_id dev_id;
-	struct ccw_device * sibling;
-};
-
-static int
-match_devno(struct device * dev, void * data)
-{
-	struct match_data * d = data;
-	struct ccw_device * cdev;
-
-	cdev = to_ccwdev(dev);
-	if ((cdev->private->state == DEV_STATE_DISCONNECTED) &&
-	    !ccw_device_is_orphan(cdev) &&
-	    ccw_dev_id_is_equal(&cdev->private->dev_id, &d->dev_id) &&
-	    (cdev != d->sibling))
-		return 1;
-	return 0;
-}
-
-static struct ccw_device * get_disc_ccwdev_by_dev_id(struct ccw_dev_id *dev_id,
-						     struct ccw_device *sibling)
-{
-	struct device *dev;
-	struct match_data data;
-
-	data.dev_id = *dev_id;
-	data.sibling = sibling;
-	dev = bus_find_device(&ccw_bus_type, NULL, &data, match_devno);
-
-	return dev ? to_ccwdev(dev) : NULL;
-}
-
-static int match_orphan(struct device *dev, void *data)
+static int match_dev_id(struct device *dev, void *data)
 {
-	struct ccw_dev_id *dev_id;
-	struct ccw_device *cdev;
+	struct ccw_device *cdev = to_ccwdev(dev);
+	struct ccw_dev_id *dev_id = data;
 
-	dev_id = data;
-	cdev = to_ccwdev(dev);
 	return ccw_dev_id_is_equal(&cdev->private->dev_id, dev_id);
 }
 
-static struct ccw_device *
-get_orphaned_ccwdev_by_dev_id(struct channel_subsystem *css,
-			      struct ccw_dev_id *dev_id)
+static struct ccw_device *get_ccwdev_by_dev_id(struct ccw_dev_id *dev_id)
 {
 	struct device *dev;
 
-	dev = device_find_child(&css->pseudo_subchannel->dev, dev_id,
-				match_orphan);
+	dev = bus_find_device(&ccw_bus_type, NULL, dev_id, match_dev_id);
 
 	return dev ? to_ccwdev(dev) : NULL;
 }
 
-void ccw_device_do_unbind_bind(struct work_struct *work)
+static void ccw_device_do_unbind_bind(struct ccw_device *cdev)
 {
-	struct ccw_device_private *priv;
-	struct ccw_device *cdev;
-	struct subchannel *sch;
 	int ret;
 
-	priv = container_of(work, struct ccw_device_private, kick_work);
-	cdev = priv->cdev;
-	sch = to_subchannel(cdev->dev.parent);
-
-	if (test_bit(1, &cdev->private->registered)) {
+	if (device_is_registered(&cdev->dev)) {
 		device_release_driver(&cdev->dev);
 		ret = device_attach(&cdev->dev);
 		WARN_ON(ret == -ENODEV);
@@ -773,6 +715,8 @@ static struct ccw_device * io_subchannel_allocate_dev(struct subchannel *sch)
 	return ERR_PTR(-ENOMEM);
 }
 
+static void ccw_device_todo(struct work_struct *work);
+
 static int io_subchannel_initialize_dev(struct subchannel *sch,
 					struct ccw_device *cdev)
 {
@@ -780,7 +724,7 @@ static int io_subchannel_initialize_dev(struct subchannel *sch,
 	atomic_set(&cdev->private->onoff, 0);
 	cdev->dev.parent = &sch->dev;
 	cdev->dev.release = ccw_device_release;
-	INIT_WORK(&cdev->private->kick_work, NULL);
+	INIT_WORK(&cdev->private->todo_work, ccw_device_todo);
 	cdev->dev.groups = ccwdev_attr_groups;
 	/* Do first half of device_register. */
 	device_initialize(&cdev->dev);
@@ -789,6 +733,7 @@ static int io_subchannel_initialize_dev(struct subchannel *sch,
 		put_device(&cdev->dev);
 		return -ENODEV;
 	}
+	cdev->private->flags.initialized = 1;
 	return 0;
 }
 
@@ -806,76 +751,7 @@ static struct ccw_device * io_subchannel_create_ccwdev(struct subchannel *sch)
 	return cdev;
 }
 
-static int io_subchannel_recog(struct ccw_device *, struct subchannel *);
-
-static void sch_attach_device(struct subchannel *sch,
-			      struct ccw_device *cdev)
-{
-	css_update_ssd_info(sch);
-	spin_lock_irq(sch->lock);
-	sch_set_cdev(sch, cdev);
-	cdev->private->schid = sch->schid;
-	cdev->ccwlock = sch->lock;
-	ccw_device_trigger_reprobe(cdev);
-	spin_unlock_irq(sch->lock);
-}
-
-static void sch_attach_disconnected_device(struct subchannel *sch,
-					   struct ccw_device *cdev)
-{
-	struct subchannel *other_sch;
-	int ret;
-
-	/* Get reference for new parent. */
-	if (!get_device(&sch->dev))
-		return;
-	other_sch = to_subchannel(cdev->dev.parent);
-	/* Note: device_move() changes cdev->dev.parent */
-	ret = device_move(&cdev->dev, &sch->dev, DPM_ORDER_PARENT_BEFORE_DEV);
-	if (ret) {
-		CIO_MSG_EVENT(0, "Moving disconnected device 0.%x.%04x failed "
-			      "(ret=%d)!\n", cdev->private->dev_id.ssid,
-			      cdev->private->dev_id.devno, ret);
-		/* Put reference for new parent. */
-		put_device(&sch->dev);
-		return;
-	}
-	sch_set_cdev(other_sch, NULL);
-	/* No need to keep a subchannel without ccw device around. */
-	css_sch_device_unregister(other_sch);
-	sch_attach_device(sch, cdev);
-	/* Put reference for old parent. */
-	put_device(&other_sch->dev);
-}
-
-static void sch_attach_orphaned_device(struct subchannel *sch,
-				       struct ccw_device *cdev)
-{
-	int ret;
-	struct subchannel *pseudo_sch;
-
-	/* Get reference for new parent. */
-	if (!get_device(&sch->dev))
-		return;
-	pseudo_sch = to_subchannel(cdev->dev.parent);
-	/*
-	 * Try to move the ccw device to its new subchannel.
-	 * Note: device_move() changes cdev->dev.parent
-	 */
-	ret = device_move(&cdev->dev, &sch->dev, DPM_ORDER_PARENT_BEFORE_DEV);
-	if (ret) {
-		CIO_MSG_EVENT(0, "Moving device 0.%x.%04x from orphanage "
-			      "failed (ret=%d)!\n",
-			      cdev->private->dev_id.ssid,
-			      cdev->private->dev_id.devno, ret);
-		/* Put reference for new parent. */
-		put_device(&sch->dev);
-		return;
-	}
-	sch_attach_device(sch, cdev);
-	/* Put reference on pseudo subchannel. */
-	put_device(&pseudo_sch->dev);
-}
+static void io_subchannel_recog(struct ccw_device *, struct subchannel *);
 
 static void sch_create_and_recog_new_device(struct subchannel *sch)
 {
@@ -888,100 +764,19 @@ static void sch_create_and_recog_new_device(struct subchannel *sch)
 		css_sch_device_unregister(sch);
 		return;
 	}
-	spin_lock_irq(sch->lock);
-	sch_set_cdev(sch, cdev);
-	spin_unlock_irq(sch->lock);
 	/* Start recognition for the new ccw device. */
-	if (io_subchannel_recog(cdev, sch)) {
-		spin_lock_irq(sch->lock);
-		sch_set_cdev(sch, NULL);
-		spin_unlock_irq(sch->lock);
-		css_sch_device_unregister(sch);
-		/* Put reference from io_subchannel_create_ccwdev(). */
-		put_device(&sch->dev);
-		/* Give up initial reference. */
-		put_device(&cdev->dev);
-	}
-}
-
-
-void ccw_device_move_to_orphanage(struct work_struct *work)
-{
-	struct ccw_device_private *priv;
-	struct ccw_device *cdev;
-	struct ccw_device *replacing_cdev;
-	struct subchannel *sch;
-	int ret;
-	struct channel_subsystem *css;
-	struct ccw_dev_id dev_id;
-
-	priv = container_of(work, struct ccw_device_private, kick_work);
-	cdev = priv->cdev;
-	sch = to_subchannel(cdev->dev.parent);
-	css = to_css(sch->dev.parent);
-	dev_id.devno = sch->schib.pmcw.dev;
-	dev_id.ssid = sch->schid.ssid;
-
-	/* Increase refcount for pseudo subchannel. */
-	get_device(&css->pseudo_subchannel->dev);
-	/*
-	 * Move the orphaned ccw device to the orphanage so the replacing
-	 * ccw device can take its place on the subchannel.
-	 * Note: device_move() changes cdev->dev.parent
-	 */
-	ret = device_move(&cdev->dev, &css->pseudo_subchannel->dev,
-		DPM_ORDER_NONE);
-	if (ret) {
-		CIO_MSG_EVENT(0, "Moving device 0.%x.%04x to orphanage failed "
-			      "(ret=%d)!\n", cdev->private->dev_id.ssid,
-			      cdev->private->dev_id.devno, ret);
-		/* Decrease refcount for pseudo subchannel again. */
-		put_device(&css->pseudo_subchannel->dev);
-		return;
-	}
-	cdev->ccwlock = css->pseudo_subchannel->lock;
-	/*
-	 * Search for the replacing ccw device
-	 * - among the disconnected devices
-	 * - in the orphanage
-	 */
-	replacing_cdev = get_disc_ccwdev_by_dev_id(&dev_id, cdev);
-	if (replacing_cdev) {
-		sch_attach_disconnected_device(sch, replacing_cdev);
-		/* Release reference from get_disc_ccwdev_by_dev_id() */
-		put_device(&replacing_cdev->dev);
-		/* Release reference of subchannel from old cdev. */
-		put_device(&sch->dev);
-		return;
-	}
-	replacing_cdev = get_orphaned_ccwdev_by_dev_id(css, &dev_id);
-	if (replacing_cdev) {
-		sch_attach_orphaned_device(sch, replacing_cdev);
-		/* Release reference from get_orphaned_ccwdev_by_dev_id() */
-		put_device(&replacing_cdev->dev);
-		/* Release reference of subchannel from old cdev. */
-		put_device(&sch->dev);
-		return;
-	}
-	sch_create_and_recog_new_device(sch);
-	/* Release reference of subchannel from old cdev. */
-	put_device(&sch->dev);
+	io_subchannel_recog(cdev, sch);
 }
 
 /*
  * Register recognized device.
  */
-static void
-io_subchannel_register(struct work_struct *work)
+static void io_subchannel_register(struct ccw_device *cdev)
 {
-	struct ccw_device_private *priv;
-	struct ccw_device *cdev;
 	struct subchannel *sch;
 	int ret;
 	unsigned long flags;
 
-	priv = container_of(work, struct ccw_device_private, kick_work);
-	cdev = priv->cdev;
 	sch = to_subchannel(cdev->dev.parent);
 	/*
 	 * Check if subchannel is still registered. It may have become
@@ -1033,41 +828,23 @@ out:
 	cdev->private->flags.recog_done = 1;
 	wake_up(&cdev->private->wait_q);
 out_err:
-	/* Release reference for workqueue processing. */
-	put_device(&cdev->dev);
 	if (atomic_dec_and_test(&ccw_device_init_count))
 		wake_up(&ccw_device_init_wq);
 }
 
-static void ccw_device_call_sch_unregister(struct work_struct *work)
+static void ccw_device_call_sch_unregister(struct ccw_device *cdev)
 {
-	struct ccw_device_private *priv;
-	struct ccw_device *cdev;
 	struct subchannel *sch;
 
-	priv = container_of(work, struct ccw_device_private, kick_work);
-	cdev = priv->cdev;
 	/* Get subchannel reference for local processing. */
 	if (!get_device(cdev->dev.parent))
 		return;
 	sch = to_subchannel(cdev->dev.parent);
 	css_sch_device_unregister(sch);
-	/* Release cdev reference for workqueue processing.*/
-	put_device(&cdev->dev);
 	/* Release subchannel reference for local processing. */
 	put_device(&sch->dev);
 }
 
-void ccw_device_schedule_sch_unregister(struct ccw_device *cdev)
-{
-	/* Get cdev reference for workqueue processing. */
-	if (!get_device(&cdev->dev))
-		return;
-	PREPARE_WORK(&cdev->private->kick_work,
-		     ccw_device_call_sch_unregister);
-	queue_work(slow_path_wq, &cdev->private->kick_work);
-}
-
 /*
  * subchannel recognition done. Called from the state machine.
  */
@@ -1083,7 +860,8 @@ io_subchannel_recog_done(struct ccw_device *cdev)
 		/* Device did not respond in time. */
 	case DEV_STATE_NOT_OPER:
 		cdev->private->flags.recog_done = 1;
-		ccw_device_schedule_sch_unregister(cdev);
+		/* Remove device found not operational. */
+		ccw_device_sched_todo(cdev, CDEV_TODO_UNREG);
 		if (atomic_dec_and_test(&ccw_device_init_count))
 			wake_up(&ccw_device_init_wq);
 		break;
@@ -1092,22 +870,15 @@ io_subchannel_recog_done(struct ccw_device *cdev)
 		 * We can't register the device in interrupt context so
 		 * we schedule a work item.
 		 */
-		if (!get_device(&cdev->dev))
-			break;
-		PREPARE_WORK(&cdev->private->kick_work,
-			     io_subchannel_register);
-		queue_work(slow_path_wq, &cdev->private->kick_work);
+		ccw_device_sched_todo(cdev, CDEV_TODO_REGISTER);
 		break;
 	}
 }
 
-static int
-io_subchannel_recog(struct ccw_device *cdev, struct subchannel *sch)
+static void io_subchannel_recog(struct ccw_device *cdev, struct subchannel *sch)
 {
-	int rc;
 	struct ccw_device_private *priv;
 
-	sch_set_cdev(sch, cdev);
 	cdev->ccwlock = sch->lock;
 
 	/* Init private data. */
@@ -1125,62 +896,81 @@ io_subchannel_recog(struct ccw_device *cdev, struct subchannel *sch)
 
 	/* Start async. device sensing. */
 	spin_lock_irq(sch->lock);
-	rc = ccw_device_recognition(cdev);
+	sch_set_cdev(sch, cdev);
+	ccw_device_recognition(cdev);
 	spin_unlock_irq(sch->lock);
-	if (rc) {
-		if (atomic_dec_and_test(&ccw_device_init_count))
-			wake_up(&ccw_device_init_wq);
-	}
-	return rc;
 }
 
-static void ccw_device_move_to_sch(struct work_struct *work)
+static int ccw_device_move_to_sch(struct ccw_device *cdev,
+				  struct subchannel *sch)
 {
-	struct ccw_device_private *priv;
-	int rc;
-	struct subchannel *sch;
-	struct ccw_device *cdev;
-	struct subchannel *former_parent;
+	struct subchannel *old_sch;
+	int rc, old_enabled = 0;
 
-	priv = container_of(work, struct ccw_device_private, kick_work);
-	sch = priv->sch;
-	cdev = priv->cdev;
-	former_parent = to_subchannel(cdev->dev.parent);
-	/* Get reference for new parent. */
+	old_sch = to_subchannel(cdev->dev.parent);
+	/* Obtain child reference for new parent. */
 	if (!get_device(&sch->dev))
-		return;
+		return -ENODEV;
+
+	if (!sch_is_pseudo_sch(old_sch)) {
+		spin_lock_irq(old_sch->lock);
+		old_enabled = old_sch->schib.pmcw.ena;
+		rc = 0;
+		if (old_enabled)
+			rc = cio_disable_subchannel(old_sch);
+		spin_unlock_irq(old_sch->lock);
+		if (rc == -EBUSY) {
+			/* Release child reference for new parent. */
+			put_device(&sch->dev);
+			return rc;
+		}
+	}
+
 	mutex_lock(&sch->reg_mutex);
-	/*
-	 * Try to move the ccw device to its new subchannel.
-	 * Note: device_move() changes cdev->dev.parent
-	 */
 	rc = device_move(&cdev->dev, &sch->dev, DPM_ORDER_PARENT_BEFORE_DEV);
 	mutex_unlock(&sch->reg_mutex);
 	if (rc) {
-		CIO_MSG_EVENT(0, "Moving device 0.%x.%04x to subchannel "
-			      "0.%x.%04x failed (ret=%d)!\n",
+		CIO_MSG_EVENT(0, "device_move(0.%x.%04x,0.%x.%04x)=%d\n",
 			      cdev->private->dev_id.ssid,
 			      cdev->private->dev_id.devno, sch->schid.ssid,
-			      sch->schid.sch_no, rc);
-		css_sch_device_unregister(sch);
-		/* Put reference for new parent again. */
+			      sch->schib.pmcw.dev, rc);
+		if (old_enabled) {
+			/* Try to reenable the old subchannel. */
+			spin_lock_irq(old_sch->lock);
+			cio_enable_subchannel(old_sch, (u32)(addr_t)old_sch);
+			spin_unlock_irq(old_sch->lock);
+		}
+		/* Release child reference for new parent. */
 		put_device(&sch->dev);
-		goto out;
+		return rc;
 	}
-	if (!sch_is_pseudo_sch(former_parent)) {
-		spin_lock_irq(former_parent->lock);
-		sch_set_cdev(former_parent, NULL);
-		spin_unlock_irq(former_parent->lock);
-		css_sch_device_unregister(former_parent);
-		/* Reset intparm to zeroes. */
-		former_parent->config.intparm = 0;
-		cio_commit_config(former_parent);
+	/* Clean up old subchannel. */
+	if (!sch_is_pseudo_sch(old_sch)) {
+		spin_lock_irq(old_sch->lock);
+		sch_set_cdev(old_sch, NULL);
+		spin_unlock_irq(old_sch->lock);
+		css_schedule_eval(old_sch->schid);
 	}
-	sch_attach_device(sch, cdev);
-out:
-	/* Put reference for old parent. */
-	put_device(&former_parent->dev);
-	put_device(&cdev->dev);
+	/* Release child reference for old parent. */
+	put_device(&old_sch->dev);
+	/* Initialize new subchannel. */
+	spin_lock_irq(sch->lock);
+	cdev->private->schid = sch->schid;
+	cdev->ccwlock = sch->lock;
+	if (!sch_is_pseudo_sch(sch))
+		sch_set_cdev(sch, cdev);
+	spin_unlock_irq(sch->lock);
+	if (!sch_is_pseudo_sch(sch))
+		css_update_ssd_info(sch);
+	return 0;
+}
+
+static int ccw_device_move_to_orph(struct ccw_device *cdev)
+{
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct channel_subsystem *css = to_css(sch->dev.parent);
+
+	return ccw_device_move_to_sch(cdev, css->pseudo_subchannel);
 }
 
 static void io_subchannel_irq(struct subchannel *sch)
@@ -1199,9 +989,6 @@ void io_subchannel_init_config(struct subchannel *sch)
 {
 	memset(&sch->config, 0, sizeof(sch->config));
 	sch->config.csense = 1;
-	/* Use subchannel mp mode when there is more than 1 installed CHPID. */
-	if ((sch->schib.pmcw.pim & (sch->schib.pmcw.pim - 1)) != 0)
-		sch->config.mp = 1;
 }
 
 static void io_subchannel_init_fields(struct subchannel *sch)
@@ -1222,23 +1009,6 @@ static void io_subchannel_init_fields(struct subchannel *sch)
 	io_subchannel_init_config(sch);
 }
 
-static void io_subchannel_do_unreg(struct work_struct *work)
-{
-	struct subchannel *sch;
-
-	sch = container_of(work, struct subchannel, work);
-	css_sch_device_unregister(sch);
-	put_device(&sch->dev);
-}
-
-/* Schedule unregister if we have no cdev. */
-static void io_subchannel_schedule_removal(struct subchannel *sch)
-{
-	get_device(&sch->dev);
-	INIT_WORK(&sch->work, io_subchannel_do_unreg);
-	queue_work(slow_path_wq, &sch->work);
-}
-
 /*
  * Note: We always return 0 so that we bind to the device even on error.
  * This is needed so that our remove function is called on unregister.
@@ -1247,8 +1017,6 @@ static int io_subchannel_probe(struct subchannel *sch)
 {
 	struct ccw_device *cdev;
 	int rc;
-	unsigned long flags;
-	struct ccw_dev_id dev_id;
 
 	if (cio_is_console(sch->schid)) {
 		rc = sysfs_create_group(&sch->dev.kobj,
@@ -1268,6 +1036,7 @@ static int io_subchannel_probe(struct subchannel *sch)
 		cdev = sch_get_cdev(sch);
 		cdev->dev.groups = ccwdev_attr_groups;
 		device_initialize(&cdev->dev);
+		cdev->private->flags.initialized = 1;
 		ccw_device_register(cdev);
 		/*
 		 * Check if the device is already online. If it is
@@ -1292,44 +1061,14 @@ static int io_subchannel_probe(struct subchannel *sch)
 	sch->private = kzalloc(sizeof(struct io_subchannel_private),
 			       GFP_KERNEL | GFP_DMA);
 	if (!sch->private)
-		goto out_err;
-	/*
-	 * First check if a fitting device may be found amongst the
-	 * disconnected devices or in the orphanage.
-	 */
-	dev_id.devno = sch->schib.pmcw.dev;
-	dev_id.ssid = sch->schid.ssid;
-	cdev = get_disc_ccwdev_by_dev_id(&dev_id, NULL);
-	if (!cdev)
-		cdev = get_orphaned_ccwdev_by_dev_id(to_css(sch->dev.parent),
-						     &dev_id);
-	if (cdev) {
-		/*
-		 * Schedule moving the device until when we have a registered
-		 * subchannel to move to and succeed the probe. We can
-		 * unregister later again, when the probe is through.
-		 */
-		cdev->private->sch = sch;
-		PREPARE_WORK(&cdev->private->kick_work,
-			     ccw_device_move_to_sch);
-		queue_work(slow_path_wq, &cdev->private->kick_work);
-		return 0;
-	}
-	cdev = io_subchannel_create_ccwdev(sch);
-	if (IS_ERR(cdev))
-		goto out_err;
-	rc = io_subchannel_recog(cdev, sch);
-	if (rc) {
-		spin_lock_irqsave(sch->lock, flags);
-		io_subchannel_recog_done(cdev);
-		spin_unlock_irqrestore(sch->lock, flags);
-	}
+		goto out_schedule;
+	css_schedule_eval(sch->schid);
 	return 0;
-out_err:
-	kfree(sch->private);
-	sysfs_remove_group(&sch->dev.kobj, &io_subchannel_attr_group);
+
 out_schedule:
-	io_subchannel_schedule_removal(sch);
+	spin_lock_irq(sch->lock);
+	css_sched_sch_todo(sch, SCH_TODO_UNREG);
+	spin_unlock_irq(sch->lock);
 	return 0;
 }
 
@@ -1337,32 +1076,23 @@ static int
 io_subchannel_remove (struct subchannel *sch)
 {
 	struct ccw_device *cdev;
-	unsigned long flags;
 
 	cdev = sch_get_cdev(sch);
 	if (!cdev)
-		return 0;
+		goto out_free;
+	io_subchannel_quiesce(sch);
 	/* Set ccw device to not operational and drop reference. */
-	spin_lock_irqsave(cdev->ccwlock, flags);
+	spin_lock_irq(cdev->ccwlock);
 	sch_set_cdev(sch, NULL);
 	cdev->private->state = DEV_STATE_NOT_OPER;
-	spin_unlock_irqrestore(cdev->ccwlock, flags);
+	spin_unlock_irq(cdev->ccwlock);
 	ccw_device_unregister(cdev);
+out_free:
 	kfree(sch->private);
 	sysfs_remove_group(&sch->dev.kobj, &io_subchannel_attr_group);
 	return 0;
 }
 
-static int io_subchannel_notify(struct subchannel *sch, int event)
-{
-	struct ccw_device *cdev;
-
-	cdev = sch_get_cdev(sch);
-	if (!cdev)
-		return 0;
-	return ccw_device_notify(cdev, event);
-}
-
 static void io_subchannel_verify(struct subchannel *sch)
 {
 	struct ccw_device *cdev;
@@ -1372,36 +1102,6 @@ static void io_subchannel_verify(struct subchannel *sch)
 		dev_fsm_event(cdev, DEV_EVENT_VERIFY);
 }
 
-static int check_for_io_on_path(struct subchannel *sch, int mask)
-{
-	if (cio_update_schib(sch))
-		return 0;
-	if (scsw_actl(&sch->schib.scsw) && sch->schib.pmcw.lpum == mask)
-		return 1;
-	return 0;
-}
-
-static void terminate_internal_io(struct subchannel *sch,
-				  struct ccw_device *cdev)
-{
-	if (cio_clear(sch)) {
-		/* Recheck device in case clear failed. */
-		sch->lpm = 0;
-		if (cdev->online)
-			dev_fsm_event(cdev, DEV_EVENT_VERIFY);
-		else
-			css_schedule_eval(sch->schid);
-		return;
-	}
-	cdev->private->state = DEV_STATE_CLEAR_VERIFY;
-	/* Request retry of internal operation. */
-	cdev->private->flags.intretry = 1;
-	/* Call handler. */
-	if (cdev->handler)
-		cdev->handler(cdev, cdev->private->intparm,
-			      ERR_PTR(-EIO));
-}
-
 static void io_subchannel_terminate_path(struct subchannel *sch, u8 mask)
 {
 	struct ccw_device *cdev;
@@ -1409,18 +1109,24 @@ static void io_subchannel_terminate_path(struct subchannel *sch, u8 mask)
 	cdev = sch_get_cdev(sch);
 	if (!cdev)
 		return;
-	if (check_for_io_on_path(sch, mask)) {
-		if (cdev->private->state == DEV_STATE_ONLINE)
-			ccw_device_kill_io(cdev);
-		else {
-			terminate_internal_io(sch, cdev);
-			/* Re-start path verification. */
-			dev_fsm_event(cdev, DEV_EVENT_VERIFY);
-		}
-	} else
-		/* trigger path verification. */
-		dev_fsm_event(cdev, DEV_EVENT_VERIFY);
+	if (cio_update_schib(sch))
+		goto err;
+	/* Check for I/O on path. */
+	if (scsw_actl(&sch->schib.scsw) == 0 || sch->schib.pmcw.lpum != mask)
+		goto out;
+	if (cdev->private->state == DEV_STATE_ONLINE) {
+		ccw_device_kill_io(cdev);
+		goto out;
+	}
+	if (cio_clear(sch))
+		goto err;
+out:
+	/* Trigger path verification. */
+	dev_fsm_event(cdev, DEV_EVENT_VERIFY);
+	return;
 
+err:
+	dev_fsm_event(cdev, DEV_EVENT_NOTOPER);
 }
 
 static int io_subchannel_chp_event(struct subchannel *sch,
@@ -1457,46 +1163,41 @@ static int io_subchannel_chp_event(struct subchannel *sch,
 	return 0;
 }
 
-static void
-io_subchannel_shutdown(struct subchannel *sch)
+static void io_subchannel_quiesce(struct subchannel *sch)
 {
 	struct ccw_device *cdev;
 	int ret;
 
+	spin_lock_irq(sch->lock);
 	cdev = sch_get_cdev(sch);
-
 	if (cio_is_console(sch->schid))
-		return;
+		goto out_unlock;
 	if (!sch->schib.pmcw.ena)
-		/* Nothing to do. */
-		return;
+		goto out_unlock;
 	ret = cio_disable_subchannel(sch);
 	if (ret != -EBUSY)
-		/* Subchannel is disabled, we're done. */
-		return;
-	cdev->private->state = DEV_STATE_QUIESCE;
+		goto out_unlock;
 	if (cdev->handler)
-		cdev->handler(cdev, cdev->private->intparm,
-			      ERR_PTR(-EIO));
-	ret = ccw_device_cancel_halt_clear(cdev);
-	if (ret == -EBUSY) {
-		ccw_device_set_timeout(cdev, HZ/10);
-		wait_event(cdev->private->wait_q, dev_fsm_final_state(cdev));
+		cdev->handler(cdev, cdev->private->intparm, ERR_PTR(-EIO));
+	while (ret == -EBUSY) {
+		cdev->private->state = DEV_STATE_QUIESCE;
+		ret = ccw_device_cancel_halt_clear(cdev);
+		if (ret == -EBUSY) {
+			ccw_device_set_timeout(cdev, HZ/10);
+			spin_unlock_irq(sch->lock);
+			wait_event(cdev->private->wait_q,
+				   cdev->private->state != DEV_STATE_QUIESCE);
+			spin_lock_irq(sch->lock);
+		}
+		ret = cio_disable_subchannel(sch);
 	}
-	cio_disable_subchannel(sch);
+out_unlock:
+	spin_unlock_irq(sch->lock);
 }
 
-static int io_subchannel_get_status(struct subchannel *sch)
+static void io_subchannel_shutdown(struct subchannel *sch)
 {
-	struct schib schib;
-
-	if (stsch(sch->schid, &schib) || !schib.pmcw.dnv)
-		return CIO_GONE;
-	if (sch->schib.pmcw.dnv && (schib.pmcw.dev != sch->schib.pmcw.dev))
-		return CIO_REVALIDATE;
-	if (!sch->lpm)
-		return CIO_NO_PATH;
-	return CIO_OPER;
+	io_subchannel_quiesce(sch);
 }
 
 static int device_is_disconnected(struct ccw_device *cdev)
@@ -1575,20 +1276,16 @@ static void ccw_device_schedule_recovery(void)
 static int purge_fn(struct device *dev, void *data)
 {
 	struct ccw_device *cdev = to_ccwdev(dev);
-	struct ccw_device_private *priv = cdev->private;
-	int unreg;
+	struct ccw_dev_id *id = &cdev->private->dev_id;
 
 	spin_lock_irq(cdev->ccwlock);
-	unreg = is_blacklisted(priv->dev_id.ssid, priv->dev_id.devno) &&
-		(priv->state == DEV_STATE_OFFLINE);
+	if (is_blacklisted(id->ssid, id->devno) &&
+	    (cdev->private->state == DEV_STATE_OFFLINE)) {
+		CIO_MSG_EVENT(3, "ccw: purging 0.%x.%04x\n", id->ssid,
+			      id->devno);
+		ccw_device_sched_todo(cdev, CDEV_TODO_UNREG);
+	}
 	spin_unlock_irq(cdev->ccwlock);
-	if (!unreg)
-		goto out;
-	CIO_MSG_EVENT(3, "ccw: purging 0.%x.%04x\n", priv->dev_id.ssid,
-		      priv->dev_id.devno);
-	ccw_device_schedule_sch_unregister(cdev);
-
-out:
 	/* Abort loop in case of pending signal. */
 	if (signal_pending(current))
 		return -EINTR;
@@ -1630,91 +1327,169 @@ void ccw_device_set_notoper(struct ccw_device *cdev)
 	cdev->private->state = DEV_STATE_NOT_OPER;
 }
 
-static int io_subchannel_sch_event(struct subchannel *sch, int slow)
+enum io_sch_action {
+	IO_SCH_UNREG,
+	IO_SCH_ORPH_UNREG,
+	IO_SCH_ATTACH,
+	IO_SCH_UNREG_ATTACH,
+	IO_SCH_ORPH_ATTACH,
+	IO_SCH_REPROBE,
+	IO_SCH_VERIFY,
+	IO_SCH_DISC,
+	IO_SCH_NOP,
+};
+
+static enum io_sch_action sch_get_action(struct subchannel *sch)
+{
+	struct ccw_device *cdev;
+
+	cdev = sch_get_cdev(sch);
+	if (cio_update_schib(sch)) {
+		/* Not operational. */
+		if (!cdev)
+			return IO_SCH_UNREG;
+		if (!ccw_device_notify(cdev, CIO_GONE))
+			return IO_SCH_UNREG;
+		return IO_SCH_ORPH_UNREG;
+	}
+	/* Operational. */
+	if (!cdev)
+		return IO_SCH_ATTACH;
+	if (sch->schib.pmcw.dev != cdev->private->dev_id.devno) {
+		if (!ccw_device_notify(cdev, CIO_GONE))
+			return IO_SCH_UNREG_ATTACH;
+		return IO_SCH_ORPH_ATTACH;
+	}
+	if ((sch->schib.pmcw.pam & sch->opm) == 0) {
+		if (!ccw_device_notify(cdev, CIO_NO_PATH))
+			return IO_SCH_UNREG;
+		return IO_SCH_DISC;
+	}
+	if (device_is_disconnected(cdev))
+		return IO_SCH_REPROBE;
+	if (cdev->online)
+		return IO_SCH_VERIFY;
+	return IO_SCH_NOP;
+}
+
+/**
+ * io_subchannel_sch_event - process subchannel event
+ * @sch: subchannel
+ * @process: non-zero if function is called in process context
+ *
+ * An unspecified event occurred for this subchannel. Adjust data according
+ * to the current operational state of the subchannel and device. Return
+ * zero when the event has been handled sufficiently or -EAGAIN when this
+ * function should be called again in process context.
+ */
+static int io_subchannel_sch_event(struct subchannel *sch, int process)
 {
-	int event, ret, disc;
 	unsigned long flags;
-	enum { NONE, UNREGISTER, UNREGISTER_PROBE, REPROBE, DISC } action;
 	struct ccw_device *cdev;
+	struct ccw_dev_id dev_id;
+	enum io_sch_action action;
+	int rc = -EAGAIN;
 
 	spin_lock_irqsave(sch->lock, flags);
+	if (!device_is_registered(&sch->dev))
+		goto out_unlock;
+	if (work_pending(&sch->todo_work))
+		goto out_unlock;
 	cdev = sch_get_cdev(sch);
-	disc = device_is_disconnected(cdev);
-	if (disc && slow) {
-		/* Disconnected devices are evaluated directly only.*/
-		spin_unlock_irqrestore(sch->lock, flags);
-		return 0;
-	}
-	/* No interrupt after machine check - kill pending timers. */
-	if (cdev)
-		ccw_device_set_timeout(cdev, 0);
-	if (!disc && !slow) {
-		/* Non-disconnected devices are evaluated on the slow path. */
-		spin_unlock_irqrestore(sch->lock, flags);
-		return -EAGAIN;
+	if (cdev && work_pending(&cdev->private->todo_work))
+		goto out_unlock;
+	action = sch_get_action(sch);
+	CIO_MSG_EVENT(2, "event: sch 0.%x.%04x, process=%d, action=%d\n",
+		      sch->schid.ssid, sch->schid.sch_no, process,
+		      action);
+	/* Perform immediate actions while holding the lock. */
+	switch (action) {
+	case IO_SCH_REPROBE:
+		/* Trigger device recognition. */
+		ccw_device_trigger_reprobe(cdev);
+		rc = 0;
+		goto out_unlock;
+	case IO_SCH_VERIFY:
+		/* Trigger path verification. */
+		io_subchannel_verify(sch);
+		rc = 0;
+		goto out_unlock;
+	case IO_SCH_DISC:
+		ccw_device_set_disconnected(cdev);
+		rc = 0;
+		goto out_unlock;
+	case IO_SCH_ORPH_UNREG:
+	case IO_SCH_ORPH_ATTACH:
+		ccw_device_set_disconnected(cdev);
+		break;
+	case IO_SCH_UNREG_ATTACH:
+	case IO_SCH_UNREG:
+		if (cdev)
+			ccw_device_set_notoper(cdev);
+		break;
+	case IO_SCH_NOP:
+		rc = 0;
+		goto out_unlock;
+	default:
+		break;
 	}
-	event = io_subchannel_get_status(sch);
-	CIO_MSG_EVENT(4, "Evaluating schid 0.%x.%04x, event %d, %s, %s path.\n",
-		      sch->schid.ssid, sch->schid.sch_no, event,
-		      disc ? "disconnected" : "normal",
-		      slow ? "slow" : "fast");
-	/* Analyze subchannel status. */
-	action = NONE;
-	switch (event) {
-	case CIO_NO_PATH:
-		if (disc) {
-			/* Check if paths have become available. */
-			action = REPROBE;
-			break;
-		}
-		/* fall through */
-	case CIO_GONE:
-		/* Ask driver what to do with device. */
-		if (io_subchannel_notify(sch, event))
-			action = DISC;
-		else
-			action = UNREGISTER;
+	spin_unlock_irqrestore(sch->lock, flags);
+	/* All other actions require process context. */
+	if (!process)
+		goto out;
+	/* Handle attached ccw device. */
+	switch (action) {
+	case IO_SCH_ORPH_UNREG:
+	case IO_SCH_ORPH_ATTACH:
+		/* Move ccw device to orphanage. */
+		rc = ccw_device_move_to_orph(cdev);
+		if (rc)
+			goto out;
 		break;
-	case CIO_REVALIDATE:
-		/* Device will be removed, so no notify necessary. */
-		if (disc)
-			/* Reprobe because immediate unregister might block. */
-			action = REPROBE;
-		else
-			action = UNREGISTER_PROBE;
+	case IO_SCH_UNREG_ATTACH:
+		/* Unregister ccw device. */
+		ccw_device_unregister(cdev);
 		break;
-	case CIO_OPER:
-		if (disc)
-			/* Get device operational again. */
-			action = REPROBE;
+	default:
 		break;
 	}
-	/* Perform action. */
-	ret = 0;
+	/* Handle subchannel. */
 	switch (action) {
-	case UNREGISTER:
-	case UNREGISTER_PROBE:
-		ccw_device_set_notoper(cdev);
-		/* Unregister device (will use subchannel lock). */
-		spin_unlock_irqrestore(sch->lock, flags);
+	case IO_SCH_ORPH_UNREG:
+	case IO_SCH_UNREG:
 		css_sch_device_unregister(sch);
-		spin_lock_irqsave(sch->lock, flags);
 		break;
-	case REPROBE:
+	case IO_SCH_ORPH_ATTACH:
+	case IO_SCH_UNREG_ATTACH:
+	case IO_SCH_ATTACH:
+		dev_id.ssid = sch->schid.ssid;
+		dev_id.devno = sch->schib.pmcw.dev;
+		cdev = get_ccwdev_by_dev_id(&dev_id);
+		if (!cdev) {
+			sch_create_and_recog_new_device(sch);
+			break;
+		}
+		rc = ccw_device_move_to_sch(cdev, sch);
+		if (rc) {
+			/* Release reference from get_ccwdev_by_dev_id() */
+			put_device(&cdev->dev);
+			goto out;
+		}
+		spin_lock_irqsave(sch->lock, flags);
 		ccw_device_trigger_reprobe(cdev);
-		break;
-	case DISC:
-		ccw_device_set_disconnected(cdev);
+		spin_unlock_irqrestore(sch->lock, flags);
+		/* Release reference from get_ccwdev_by_dev_id() */
+		put_device(&cdev->dev);
 		break;
 	default:
 		break;
 	}
-	spin_unlock_irqrestore(sch->lock, flags);
-	/* Probe if necessary. */
-	if (action == UNREGISTER_PROBE)
-		ret = css_probe_device(sch->schid);
+	return 0;
 
-	return ret;
+out_unlock:
+	spin_unlock_irqrestore(sch->lock, flags);
+out:
+	return rc;
 }
 
 #ifdef CONFIG_CCW_CONSOLE
@@ -1744,10 +1519,7 @@ static int ccw_device_console_enable(struct ccw_device *cdev,
 	sch->driver = &io_subchannel_driver;
 	/* Initialize the ccw_device structure. */
 	cdev->dev.parent= &sch->dev;
-	rc = io_subchannel_recog(cdev, sch);
-	if (rc)
-		return rc;
-
+	io_subchannel_recog(cdev, sch);
 	/* Now wait for the async. recognition to come to an end. */
 	spin_lock_irq(cdev->ccwlock);
 	while (!dev_fsm_final_state(cdev))
@@ -1763,7 +1535,7 @@ static int ccw_device_console_enable(struct ccw_device *cdev,
 	rc = 0;
 out_unlock:
 	spin_unlock_irq(cdev->ccwlock);
-	return 0;
+	return rc;
 }
 
 struct ccw_device *
@@ -1919,7 +1691,7 @@ static int ccw_device_pm_prepare(struct device *dev)
 {
 	struct ccw_device *cdev = to_ccwdev(dev);
 
-	if (work_pending(&cdev->private->kick_work))
+	if (work_pending(&cdev->private->todo_work))
 		return -EAGAIN;
 	/* Fail while device is being set online/offline. */
 	if (atomic_read(&cdev->private->onoff))
@@ -2005,7 +1777,6 @@ static int ccw_device_pm_thaw(struct device *dev)
 static void __ccw_device_pm_restore(struct ccw_device *cdev)
 {
 	struct subchannel *sch = to_subchannel(cdev->dev.parent);
-	int ret;
 
 	if (cio_is_console(sch->schid))
 		goto out;
@@ -2015,22 +1786,10 @@ static void __ccw_device_pm_restore(struct ccw_device *cdev)
 	 */
 	spin_lock_irq(sch->lock);
 	cdev->private->flags.resuming = 1;
-	ret = ccw_device_recognition(cdev);
+	ccw_device_recognition(cdev);
 	spin_unlock_irq(sch->lock);
-	if (ret) {
-		CIO_MSG_EVENT(0, "Couldn't start recognition for device "
-			      "0.%x.%04x (ret=%d)\n",
-			      cdev->private->dev_id.ssid,
-			      cdev->private->dev_id.devno, ret);
-		spin_lock_irq(sch->lock);
-		cdev->private->state = DEV_STATE_DISCONNECTED;
-		spin_unlock_irq(sch->lock);
-		/* notify driver after the resume cb */
-		goto out;
-	}
 	wait_event(cdev->private->wait_q, dev_fsm_final_state(cdev) ||
 		   cdev->private->state == DEV_STATE_DISCONNECTED);
-
 out:
 	cdev->private->flags.resuming = 0;
 }
@@ -2040,7 +1799,7 @@ static int resume_handle_boxed(struct ccw_device *cdev)
 	cdev->private->state = DEV_STATE_BOXED;
 	if (ccw_device_notify(cdev, CIO_BOXED))
 		return 0;
-	ccw_device_schedule_sch_unregister(cdev);
+	ccw_device_sched_todo(cdev, CDEV_TODO_UNREG);
 	return -ENODEV;
 }
 
@@ -2049,7 +1808,7 @@ static int resume_handle_disc(struct ccw_device *cdev)
 	cdev->private->state = DEV_STATE_DISCONNECTED;
 	if (ccw_device_notify(cdev, CIO_GONE))
 		return 0;
-	ccw_device_schedule_sch_unregister(cdev);
+	ccw_device_sched_todo(cdev, CDEV_TODO_UNREG);
 	return -ENODEV;
 }
 
@@ -2094,9 +1853,7 @@ static int ccw_device_pm_restore(struct device *dev)
 	/* check if the device type has changed */
 	if (!ccw_device_test_sense_data(cdev)) {
 		ccw_device_update_sense_data(cdev);
-		PREPARE_WORK(&cdev->private->kick_work,
-			     ccw_device_do_unbind_bind);
-		queue_work(ccw_device_work, &cdev->private->kick_work);
+		ccw_device_sched_todo(cdev, CDEV_TODO_REBIND);
 		ret = -ENODEV;
 		goto out_unlock;
 	}
@@ -2140,7 +1897,7 @@ out_disc_unlock:
 	goto out_restore;
 
 out_unreg_unlock:
-	ccw_device_schedule_sch_unregister(cdev);
+	ccw_device_sched_todo(cdev, CDEV_TODO_UNREG_EVAL);
 	ret = -ENODEV;
 out_unlock:
 	spin_unlock_irq(sch->lock);
@@ -2205,6 +1962,77 @@ ccw_device_get_subchannel_id(struct ccw_device *cdev)
 	return sch->schid;
 }
 
+static void ccw_device_todo(struct work_struct *work)
+{
+	struct ccw_device_private *priv;
+	struct ccw_device *cdev;
+	struct subchannel *sch;
+	enum cdev_todo todo;
+
+	priv = container_of(work, struct ccw_device_private, todo_work);
+	cdev = priv->cdev;
+	sch = to_subchannel(cdev->dev.parent);
+	/* Find out todo. */
+	spin_lock_irq(cdev->ccwlock);
+	todo = priv->todo;
+	priv->todo = CDEV_TODO_NOTHING;
+	CIO_MSG_EVENT(4, "cdev_todo: cdev=0.%x.%04x todo=%d\n",
+		      priv->dev_id.ssid, priv->dev_id.devno, todo);
+	spin_unlock_irq(cdev->ccwlock);
+	/* Perform todo. */
+	switch (todo) {
+	case CDEV_TODO_ENABLE_CMF:
+		cmf_reenable(cdev);
+		break;
+	case CDEV_TODO_REBIND:
+		ccw_device_do_unbind_bind(cdev);
+		break;
+	case CDEV_TODO_REGISTER:
+		io_subchannel_register(cdev);
+		break;
+	case CDEV_TODO_UNREG_EVAL:
+		if (!sch_is_pseudo_sch(sch))
+			css_schedule_eval(sch->schid);
+		/* fall-through */
+	case CDEV_TODO_UNREG:
+		if (sch_is_pseudo_sch(sch))
+			ccw_device_unregister(cdev);
+		else
+			ccw_device_call_sch_unregister(cdev);
+		break;
+	default:
+		break;
+	}
+	/* Release workqueue ref. */
+	put_device(&cdev->dev);
+}
+
+/**
+ * ccw_device_sched_todo - schedule ccw device operation
+ * @cdev: ccw device
+ * @todo: todo
+ *
+ * Schedule the operation identified by @todo to be performed on the slow path
+ * workqueue. Do nothing if another operation with higher priority is already
+ * scheduled. Needs to be called with ccwdev lock held.
+ */
+void ccw_device_sched_todo(struct ccw_device *cdev, enum cdev_todo todo)
+{
+	CIO_MSG_EVENT(4, "cdev_todo: sched cdev=0.%x.%04x todo=%d\n",
+		      cdev->private->dev_id.ssid, cdev->private->dev_id.devno,
+		      todo);
+	if (cdev->private->todo >= todo)
+		return;
+	cdev->private->todo = todo;
+	/* Get workqueue ref. */
+	if (!get_device(&cdev->dev))
+		return;
+	if (!queue_work(slow_path_wq, &cdev->private->todo_work)) {
+		/* Already queued, release workqueue ref. */
+		put_device(&cdev->dev);
+	}
+}
+
 MODULE_LICENSE("GPL");
 EXPORT_SYMBOL(ccw_device_set_online);
 EXPORT_SYMBOL(ccw_device_set_offline);

drivers/s390/cio/device.h

@@ -21,7 +21,6 @@ enum dev_state {
 	DEV_STATE_DISBAND_PGID,
 	DEV_STATE_BOXED,
 	/* states to wait for i/o completion before doing something */
-	DEV_STATE_CLEAR_VERIFY,
 	DEV_STATE_TIMEOUT_KILL,
 	DEV_STATE_QUIESCE,
 	/* special states for devices gone not operational */
@@ -29,6 +28,7 @@ enum dev_state {
 	DEV_STATE_DISCONNECTED_SENSE_ID,
 	DEV_STATE_CMFCHANGE,
 	DEV_STATE_CMFUPDATE,
+	DEV_STATE_STEAL_LOCK,
 	/* last element! */
 	NR_DEV_STATES
 };
@@ -81,17 +81,16 @@ void io_subchannel_init_config(struct subchannel *sch);
 
 int ccw_device_cancel_halt_clear(struct ccw_device *);
 
-void ccw_device_do_unbind_bind(struct work_struct *);
-void ccw_device_move_to_orphanage(struct work_struct *);
 int ccw_device_is_orphan(struct ccw_device *);
 
-int ccw_device_recognition(struct ccw_device *);
+void ccw_device_recognition(struct ccw_device *);
 int ccw_device_online(struct ccw_device *);
 int ccw_device_offline(struct ccw_device *);
 void ccw_device_update_sense_data(struct ccw_device *);
 int ccw_device_test_sense_data(struct ccw_device *);
 void ccw_device_schedule_sch_unregister(struct ccw_device *);
 int ccw_purge_blacklisted(void);
+void ccw_device_sched_todo(struct ccw_device *cdev, enum cdev_todo todo);
 
 /* Function prototypes for device status and basic sense stuff. */
 void ccw_device_accumulate_irb(struct ccw_device *, struct irb *);
@@ -99,24 +98,28 @@ void ccw_device_accumulate_basic_sense(struct ccw_device *, struct irb *);
 int ccw_device_accumulate_and_sense(struct ccw_device *, struct irb *);
 int ccw_device_do_sense(struct ccw_device *, struct irb *);
 
+/* Function prototype for internal request handling. */
+int lpm_adjust(int lpm, int mask);
+void ccw_request_start(struct ccw_device *);
+int ccw_request_cancel(struct ccw_device *cdev);
+void ccw_request_handler(struct ccw_device *cdev);
+void ccw_request_timeout(struct ccw_device *cdev);
+void ccw_request_notoper(struct ccw_device *cdev);
+
 /* Function prototypes for sense id stuff. */
 void ccw_device_sense_id_start(struct ccw_device *);
-void ccw_device_sense_id_irq(struct ccw_device *, enum dev_event);
 void ccw_device_sense_id_done(struct ccw_device *, int);
 
 /* Function prototypes for path grouping stuff. */
-void ccw_device_sense_pgid_start(struct ccw_device *);
-void ccw_device_sense_pgid_irq(struct ccw_device *, enum dev_event);
-void ccw_device_sense_pgid_done(struct ccw_device *, int);
-
 void ccw_device_verify_start(struct ccw_device *);
-void ccw_device_verify_irq(struct ccw_device *, enum dev_event);
 void ccw_device_verify_done(struct ccw_device *, int);
 
 void ccw_device_disband_start(struct ccw_device *);
-void ccw_device_disband_irq(struct ccw_device *, enum dev_event);
 void ccw_device_disband_done(struct ccw_device *, int);
 
+void ccw_device_stlck_start(struct ccw_device *, void *, void *, void *);
+void ccw_device_stlck_done(struct ccw_device *, void *, int);
+
 int ccw_device_call_handler(struct ccw_device *);
 
 int ccw_device_stlck(struct ccw_device *);

drivers/s390/cio/device_fsm.c

@@ -229,8 +229,8 @@ ccw_device_recog_done(struct ccw_device *cdev, int state)
 
 	sch = to_subchannel(cdev->dev.parent);
 
-	ccw_device_set_timeout(cdev, 0);
-	cio_disable_subchannel(sch);
+	if (cio_disable_subchannel(sch))
+		state = DEV_STATE_NOT_OPER;
 	/*
 	 * Now that we tried recognition, we have performed device selection
 	 * through ssch() and the path information is up to date.
@@ -263,22 +263,10 @@ ccw_device_recog_done(struct ccw_device *cdev, int state)
 	}
 	switch (state) {
 	case DEV_STATE_NOT_OPER:
-		CIO_MSG_EVENT(2, "SenseID : unknown device %04x on "
-			      "subchannel 0.%x.%04x\n",
-			      cdev->private->dev_id.devno,
-			      sch->schid.ssid, sch->schid.sch_no);
 		break;
 	case DEV_STATE_OFFLINE:
 		if (!cdev->online) {
 			ccw_device_update_sense_data(cdev);
-			/* Issue device info message. */
-			CIO_MSG_EVENT(4, "SenseID : device 0.%x.%04x reports: "
-				      "CU  Type/Mod = %04X/%02X, Dev Type/Mod "
-				      "= %04X/%02X\n",
-				      cdev->private->dev_id.ssid,
-				      cdev->private->dev_id.devno,
-				      cdev->id.cu_type, cdev->id.cu_model,
-				      cdev->id.dev_type, cdev->id.dev_model);
 			break;
 		}
 		cdev->private->state = DEV_STATE_OFFLINE;
@@ -289,16 +277,10 @@ ccw_device_recog_done(struct ccw_device *cdev, int state)
 			wake_up(&cdev->private->wait_q);
 		} else {
 			ccw_device_update_sense_data(cdev);
-			PREPARE_WORK(&cdev->private->kick_work,
-				     ccw_device_do_unbind_bind);
-			queue_work(ccw_device_work, &cdev->private->kick_work);
+			ccw_device_sched_todo(cdev, CDEV_TODO_REBIND);
 		}
 		return;
 	case DEV_STATE_BOXED:
-		CIO_MSG_EVENT(0, "SenseID : boxed device %04x on "
-			      " subchannel 0.%x.%04x\n",
-			      cdev->private->dev_id.devno,
-			      sch->schid.ssid, sch->schid.sch_no);
 		if (cdev->id.cu_type != 0) { /* device was recognized before */
 			cdev->private->flags.recog_done = 1;
 			cdev->private->state = DEV_STATE_BOXED;
@@ -343,28 +325,16 @@ int ccw_device_notify(struct ccw_device *cdev, int event)
 	return cdev->drv->notify ? cdev->drv->notify(cdev, event) : 0;
 }
 
-static void cmf_reenable_delayed(struct work_struct *work)
-{
-	struct ccw_device_private *priv;
-	struct ccw_device *cdev;
-
-	priv = container_of(work, struct ccw_device_private, kick_work);
-	cdev = priv->cdev;
-	cmf_reenable(cdev);
-}
-
 static void ccw_device_oper_notify(struct ccw_device *cdev)
 {
 	if (ccw_device_notify(cdev, CIO_OPER)) {
 		/* Reenable channel measurements, if needed. */
-		PREPARE_WORK(&cdev->private->kick_work, cmf_reenable_delayed);
-		queue_work(ccw_device_work, &cdev->private->kick_work);
+		ccw_device_sched_todo(cdev, CDEV_TODO_ENABLE_CMF);
 		return;
 	}
 	/* Driver doesn't want device back. */
 	ccw_device_set_notoper(cdev);
-	PREPARE_WORK(&cdev->private->kick_work, ccw_device_do_unbind_bind);
-	queue_work(ccw_device_work, &cdev->private->kick_work);
+	ccw_device_sched_todo(cdev, CDEV_TODO_REBIND);
 }
 
 /*
@@ -392,14 +362,14 @@ ccw_device_done(struct ccw_device *cdev, int state)
 		CIO_MSG_EVENT(0, "Boxed device %04x on subchannel %04x\n",
 			      cdev->private->dev_id.devno, sch->schid.sch_no);
 		if (cdev->online && !ccw_device_notify(cdev, CIO_BOXED))
-			ccw_device_schedule_sch_unregister(cdev);
+			ccw_device_sched_todo(cdev, CDEV_TODO_UNREG);
 		cdev->private->flags.donotify = 0;
 		break;
 	case DEV_STATE_NOT_OPER:
 		CIO_MSG_EVENT(0, "Device %04x gone on subchannel %04x\n",
 			      cdev->private->dev_id.devno, sch->schid.sch_no);
 		if (!ccw_device_notify(cdev, CIO_GONE))
-			ccw_device_schedule_sch_unregister(cdev);
+			ccw_device_sched_todo(cdev, CDEV_TODO_UNREG);
 		else
 			ccw_device_set_disconnected(cdev);
 		cdev->private->flags.donotify = 0;
@@ -409,7 +379,7 @@ ccw_device_done(struct ccw_device *cdev, int state)
 			      "%04x\n", cdev->private->dev_id.devno,
 			      sch->schid.sch_no);
 		if (!ccw_device_notify(cdev, CIO_NO_PATH))
-			ccw_device_schedule_sch_unregister(cdev);
+			ccw_device_sched_todo(cdev, CDEV_TODO_UNREG);
 		else
 			ccw_device_set_disconnected(cdev);
 		cdev->private->flags.donotify = 0;
@@ -425,107 +395,12 @@ ccw_device_done(struct ccw_device *cdev, int state)
 	wake_up(&cdev->private->wait_q);
 }
 
-static int cmp_pgid(struct pgid *p1, struct pgid *p2)
-{
-	char *c1;
-	char *c2;
-
-	c1 = (char *)p1;
-	c2 = (char *)p2;
-
-	return memcmp(c1 + 1, c2 + 1, sizeof(struct pgid) - 1);
-}
-
-static void __ccw_device_get_common_pgid(struct ccw_device *cdev)
-{
-	int i;
-	int last;
-
-	last = 0;
-	for (i = 0; i < 8; i++) {
-		if (cdev->private->pgid[i].inf.ps.state1 == SNID_STATE1_RESET)
-			/* No PGID yet */
-			continue;
-		if (cdev->private->pgid[last].inf.ps.state1 ==
-		    SNID_STATE1_RESET) {
-			/* First non-zero PGID */
-			last = i;
-			continue;
-		}
-		if (cmp_pgid(&cdev->private->pgid[i],
-			     &cdev->private->pgid[last]) == 0)
-			/* Non-conflicting PGIDs */
-			continue;
-
-		/* PGID mismatch, can't pathgroup. */
-		CIO_MSG_EVENT(0, "SNID - pgid mismatch for device "
-			      "0.%x.%04x, can't pathgroup\n",
-			      cdev->private->dev_id.ssid,
-			      cdev->private->dev_id.devno);
-		cdev->private->options.pgroup = 0;
-		return;
-	}
-	if (cdev->private->pgid[last].inf.ps.state1 ==
-	    SNID_STATE1_RESET)
-		/* No previous pgid found */
-		memcpy(&cdev->private->pgid[0],
-		       &channel_subsystems[0]->global_pgid,
-		       sizeof(struct pgid));
-	else
-		/* Use existing pgid */
-		memcpy(&cdev->private->pgid[0], &cdev->private->pgid[last],
-		       sizeof(struct pgid));
-}
-
-/*
- * Function called from device_pgid.c after sense path ground has completed.
- */
-void
-ccw_device_sense_pgid_done(struct ccw_device *cdev, int err)
-{
-	struct subchannel *sch;
-
-	sch = to_subchannel(cdev->dev.parent);
-	switch (err) {
-	case -EOPNOTSUPP: /* path grouping not supported, use nop instead. */
-		cdev->private->options.pgroup = 0;
-		break;
-	case 0: /* success */
-	case -EACCES: /* partial success, some paths not operational */
-		/* Check if all pgids are equal or 0. */
-		__ccw_device_get_common_pgid(cdev);
-		break;
-	case -ETIME:		/* Sense path group id stopped by timeout. */
-	case -EUSERS:		/* device is reserved for someone else. */
-		ccw_device_done(cdev, DEV_STATE_BOXED);
-		return;
-	default:
-		ccw_device_done(cdev, DEV_STATE_NOT_OPER);
-		return;
-	}
-	/* Start Path Group verification. */
-	cdev->private->state = DEV_STATE_VERIFY;
-	cdev->private->flags.doverify = 0;
-	ccw_device_verify_start(cdev);
-}
-
 /*
  * Start device recognition.
  */
-int
-ccw_device_recognition(struct ccw_device *cdev)
+void ccw_device_recognition(struct ccw_device *cdev)
 {
-	struct subchannel *sch;
-	int ret;
-
-	sch = to_subchannel(cdev->dev.parent);
-	ret = cio_enable_subchannel(sch, (u32)(addr_t)sch);
-	if (ret != 0)
-		/* Couldn't enable the subchannel for i/o. Sick device. */
-		return ret;
-
-	/* After 60s the device recognition is considered to have failed. */
-	ccw_device_set_timeout(cdev, 60*HZ);
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
 
 	/*
 	 * We used to start here with a sense pgid to find out whether a device
@@ -537,32 +412,33 @@ ccw_device_recognition(struct ccw_device *cdev)
 	 */
 	cdev->private->flags.recog_done = 0;
 	cdev->private->state = DEV_STATE_SENSE_ID;
+	if (cio_enable_subchannel(sch, (u32) (addr_t) sch)) {
+		ccw_device_recog_done(cdev, DEV_STATE_NOT_OPER);
+		return;
+	}
 	ccw_device_sense_id_start(cdev);
-	return 0;
 }
 
 /*
- * Handle timeout in device recognition.
+ * Handle events for states that use the ccw request infrastructure.
  */
-static void
-ccw_device_recog_timeout(struct ccw_device *cdev, enum dev_event dev_event)
+static void ccw_device_request_event(struct ccw_device *cdev, enum dev_event e)
 {
-	int ret;
-
-	ret = ccw_device_cancel_halt_clear(cdev);
-	switch (ret) {
-	case 0:
-		ccw_device_recog_done(cdev, DEV_STATE_BOXED);
+	switch (e) {
+	case DEV_EVENT_NOTOPER:
+		ccw_request_notoper(cdev);
 		break;
-	case -ENODEV:
-		ccw_device_recog_done(cdev, DEV_STATE_NOT_OPER);
+	case DEV_EVENT_INTERRUPT:
+		ccw_request_handler(cdev);
+		break;
+	case DEV_EVENT_TIMEOUT:
+		ccw_request_timeout(cdev);
 		break;
 	default:
-		ccw_device_set_timeout(cdev, 3*HZ);
+		break;
 	}
 }
 
-
 void
 ccw_device_verify_done(struct ccw_device *cdev, int err)
 {
@@ -571,21 +447,18 @@ ccw_device_verify_done(struct ccw_device *cdev, int err)
 	sch = to_subchannel(cdev->dev.parent);
 	/* Update schib - pom may have changed. */
 	if (cio_update_schib(sch)) {
-		cdev->private->flags.donotify = 0;
-		ccw_device_done(cdev, DEV_STATE_NOT_OPER);
-		return;
+		err = -ENODEV;
+		goto callback;
 	}
 	/* Update lpm with verified path mask. */
 	sch->lpm = sch->vpm;
 	/* Repeat path verification? */
 	if (cdev->private->flags.doverify) {
-		cdev->private->flags.doverify = 0;
 		ccw_device_verify_start(cdev);
 		return;
 	}
+callback:
 	switch (err) {
-	case -EOPNOTSUPP: /* path grouping not supported, just set online. */
-		cdev->private->options.pgroup = 0;
 	case 0:
 		ccw_device_done(cdev, DEV_STATE_ONLINE);
 		/* Deliver fake irb to device driver, if needed. */
@@ -604,18 +477,20 @@ ccw_device_verify_done(struct ccw_device *cdev, int err)
 		}
 		break;
 	case -ETIME:
+	case -EUSERS:
 		/* Reset oper notify indication after verify error. */
 		cdev->private->flags.donotify = 0;
 		ccw_device_done(cdev, DEV_STATE_BOXED);
 		break;
+	case -EACCES:
+		/* Reset oper notify indication after verify error. */
+		cdev->private->flags.donotify = 0;
+		ccw_device_done(cdev, DEV_STATE_DISCONNECTED);
+		break;
 	default:
 		/* Reset oper notify indication after verify error. */
 		cdev->private->flags.donotify = 0;
-		if (cdev->online) {
-			ccw_device_set_timeout(cdev, 0);
-			dev_fsm_event(cdev, DEV_EVENT_NOTOPER);
-		} else
-			ccw_device_done(cdev, DEV_STATE_NOT_OPER);
+		ccw_device_done(cdev, DEV_STATE_NOT_OPER);
 		break;
 	}
 }
@@ -640,17 +515,9 @@ ccw_device_online(struct ccw_device *cdev)
 			dev_fsm_event(cdev, DEV_EVENT_NOTOPER);
 		return ret;
 	}
-	/* Do we want to do path grouping? */
-	if (!cdev->private->options.pgroup) {
-		/* Start initial path verification. */
-		cdev->private->state = DEV_STATE_VERIFY;
-		cdev->private->flags.doverify = 0;
-		ccw_device_verify_start(cdev);
-		return 0;
-	}
-	/* Do a SensePGID first. */
-	cdev->private->state = DEV_STATE_SENSE_PGID;
-	ccw_device_sense_pgid_start(cdev);
+	/* Start initial path verification. */
+	cdev->private->state = DEV_STATE_VERIFY;
+	ccw_device_verify_start(cdev);
 	return 0;
 }
 
@@ -666,7 +533,6 @@ ccw_device_disband_done(struct ccw_device *cdev, int err)
 		break;
 	default:
 		cdev->private->flags.donotify = 0;
-		dev_fsm_event(cdev, DEV_EVENT_NOTOPER);
 		ccw_device_done(cdev, DEV_STATE_NOT_OPER);
 		break;
 	}
@@ -703,7 +569,7 @@ ccw_device_offline(struct ccw_device *cdev)
 	if (cdev->private->state != DEV_STATE_ONLINE)
 		return -EINVAL;
 	/* Are we doing path grouping? */
-	if (!cdev->private->options.pgroup) {
+	if (!cdev->private->flags.pgroup) {
 		/* No, set state offline immediately. */
 		ccw_device_done(cdev, DEV_STATE_OFFLINE);
 		return 0;
@@ -714,36 +580,6 @@ ccw_device_offline(struct ccw_device *cdev)
 	return 0;
 }
 
-/*
- * Handle timeout in device online/offline process.
- */
-static void
-ccw_device_onoff_timeout(struct ccw_device *cdev, enum dev_event dev_event)
-{
-	int ret;
-
-	ret = ccw_device_cancel_halt_clear(cdev);
-	switch (ret) {
-	case 0:
-		ccw_device_done(cdev, DEV_STATE_BOXED);
-		break;
-	case -ENODEV:
-		ccw_device_done(cdev, DEV_STATE_NOT_OPER);
-		break;
-	default:
-		ccw_device_set_timeout(cdev, 3*HZ);
-	}
-}
-
-/*
- * Handle not oper event in device recognition.
- */
-static void
-ccw_device_recog_notoper(struct ccw_device *cdev, enum dev_event dev_event)
-{
-	ccw_device_recog_done(cdev, DEV_STATE_NOT_OPER);
-}
-
 /*
  * Handle not operational event in non-special state.
  */
@@ -751,7 +587,7 @@ static void ccw_device_generic_notoper(struct ccw_device *cdev,
 				       enum dev_event dev_event)
 {
 	if (!ccw_device_notify(cdev, CIO_GONE))
-		ccw_device_schedule_sch_unregister(cdev);
+		ccw_device_sched_todo(cdev, CDEV_TODO_UNREG);
 	else
 		ccw_device_set_disconnected(cdev);
 }
@@ -802,10 +638,26 @@ ccw_device_online_verify(struct ccw_device *cdev, enum dev_event dev_event)
 	}
 	/* Device is idle, we can do the path verification. */
 	cdev->private->state = DEV_STATE_VERIFY;
-	cdev->private->flags.doverify = 0;
 	ccw_device_verify_start(cdev);
 }
 
+/*
+ * Handle path verification event in boxed state.
+ */
+static void ccw_device_boxed_verify(struct ccw_device *cdev,
+				    enum dev_event dev_event)
+{
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+
+	if (cdev->online) {
+		if (cio_enable_subchannel(sch, (u32) (addr_t) sch))
+			ccw_device_done(cdev, DEV_STATE_NOT_OPER);
+		else
+			ccw_device_online_verify(cdev, dev_event);
+	} else
+		css_schedule_eval(sch->schid);
+}
+
 /*
  * Got an interrupt for a normal io (state online).
  */
@@ -904,12 +756,6 @@ ccw_device_w4sense(struct ccw_device *cdev, enum dev_event dev_event)
 	 */
 	if (scsw_fctl(&irb->scsw) &
 	    (SCSW_FCTL_CLEAR_FUNC | SCSW_FCTL_HALT_FUNC)) {
-		/* Retry Basic Sense if requested. */
-		if (cdev->private->flags.intretry) {
-			cdev->private->flags.intretry = 0;
-			ccw_device_do_sense(cdev, irb);
-			return;
-		}
 		cdev->private->flags.dosense = 0;
 		memset(&cdev->private->irb, 0, sizeof(struct irb));
 		ccw_device_accumulate_irb(cdev, irb);
@@ -932,21 +778,6 @@ call_handler:
 		ccw_device_online_verify(cdev, 0);
 }
 
-static void
-ccw_device_clear_verify(struct ccw_device *cdev, enum dev_event dev_event)
-{
-	struct irb *irb;
-
-	irb = (struct irb *) __LC_IRB;
-	/* Accumulate status. We don't do basic sense. */
-	ccw_device_accumulate_irb(cdev, irb);
-	/* Remember to clear irb to avoid residuals. */
-	memset(&cdev->private->irb, 0, sizeof(struct irb));
-	/* Try to start delayed device verification. */
-	ccw_device_online_verify(cdev, 0);
-	/* Note: Don't call handler for cio initiated clear! */
-}
-
 static void
 ccw_device_killing_irq(struct ccw_device *cdev, enum dev_event dev_event)
 {
@@ -1003,32 +834,6 @@ ccw_device_delay_verify(struct ccw_device *cdev, enum dev_event dev_event)
 	cdev->private->flags.doverify = 1;
 }
 
-static void
-ccw_device_stlck_done(struct ccw_device *cdev, enum dev_event dev_event)
-{
-	struct irb *irb;
-
-	switch (dev_event) {
-	case DEV_EVENT_INTERRUPT:
-		irb = (struct irb *) __LC_IRB;
-		/* Check for unsolicited interrupt. */
-		if ((scsw_stctl(&irb->scsw) ==
-		     (SCSW_STCTL_STATUS_PEND | SCSW_STCTL_ALERT_STATUS)) &&
-		    (!scsw_cc(&irb->scsw)))
-			/* FIXME: we should restart stlck here, but this
-			 * is extremely unlikely ... */
-			goto out_wakeup;
-
-		ccw_device_accumulate_irb(cdev, irb);
-		/* We don't care about basic sense etc. */
-		break;
-	default: /* timeout */
-		break;
-	}
-out_wakeup:
-	wake_up(&cdev->private->wait_q);
-}
-
 static void
 ccw_device_start_id(struct ccw_device *cdev, enum dev_event dev_event)
 {
@@ -1038,10 +843,6 @@ ccw_device_start_id(struct ccw_device *cdev, enum dev_event dev_event)
 	if (cio_enable_subchannel(sch, (u32)(addr_t)sch) != 0)
 		/* Couldn't enable the subchannel for i/o. Sick device. */
 		return;
-
-	/* After 60s the device recognition is considered to have failed. */
-	ccw_device_set_timeout(cdev, 60*HZ);
-
 	cdev->private->state = DEV_STATE_DISCONNECTED_SENSE_ID;
 	ccw_device_sense_id_start(cdev);
 }
@@ -1072,22 +873,20 @@ void ccw_device_trigger_reprobe(struct ccw_device *cdev)
 
 	/* We should also udate ssd info, but this has to wait. */
 	/* Check if this is another device which appeared on the same sch. */
-	if (sch->schib.pmcw.dev != cdev->private->dev_id.devno) {
-		PREPARE_WORK(&cdev->private->kick_work,
-			     ccw_device_move_to_orphanage);
-		queue_work(slow_path_wq, &cdev->private->kick_work);
-	} else
+	if (sch->schib.pmcw.dev != cdev->private->dev_id.devno)
+		css_schedule_eval(sch->schid);
+	else
 		ccw_device_start_id(cdev, 0);
 }
 
-static void
-ccw_device_offline_irq(struct ccw_device *cdev, enum dev_event dev_event)
+static void ccw_device_disabled_irq(struct ccw_device *cdev,
+				    enum dev_event dev_event)
 {
 	struct subchannel *sch;
 
 	sch = to_subchannel(cdev->dev.parent);
 	/*
-	 * An interrupt in state offline means a previous disable was not
+	 * An interrupt in a disabled state means a previous disable was not
 	 * successful - should not happen, but we try to disable again.
 	 */
 	cio_disable_subchannel(sch);
@@ -1113,10 +912,7 @@ static void
 ccw_device_quiesce_done(struct ccw_device *cdev, enum dev_event dev_event)
 {
 	ccw_device_set_timeout(cdev, 0);
-	if (dev_event == DEV_EVENT_NOTOPER)
-		cdev->private->state = DEV_STATE_NOT_OPER;
-	else
-		cdev->private->state = DEV_STATE_OFFLINE;
+	cdev->private->state = DEV_STATE_NOT_OPER;
 	wake_up(&cdev->private->wait_q);
 }
 
@@ -1126,17 +922,11 @@ ccw_device_quiesce_timeout(struct ccw_device *cdev, enum dev_event dev_event)
 	int ret;
 
 	ret = ccw_device_cancel_halt_clear(cdev);
-	switch (ret) {
-	case 0:
-		cdev->private->state = DEV_STATE_OFFLINE;
-		wake_up(&cdev->private->wait_q);
-		break;
-	case -ENODEV:
+	if (ret == -EBUSY) {
+		ccw_device_set_timeout(cdev, HZ/10);
+	} else {
 		cdev->private->state = DEV_STATE_NOT_OPER;
 		wake_up(&cdev->private->wait_q);
-		break;
-	default:
-		ccw_device_set_timeout(cdev, HZ/10);
 	}
 }
 
@@ -1149,51 +939,38 @@ ccw_device_nop(struct ccw_device *cdev, enum dev_event dev_event)
 {
 }
 
-/*
- * Bug operation action. 
- */
-static void
-ccw_device_bug(struct ccw_device *cdev, enum dev_event dev_event)
-{
-	CIO_MSG_EVENT(0, "Internal state [%i][%i] not handled for device "
-		      "0.%x.%04x\n", cdev->private->state, dev_event,
-		      cdev->private->dev_id.ssid,
-		      cdev->private->dev_id.devno);
-	BUG();
-}
-
 /*
  * device statemachine
  */
 fsm_func_t *dev_jumptable[NR_DEV_STATES][NR_DEV_EVENTS] = {
 	[DEV_STATE_NOT_OPER] = {
 		[DEV_EVENT_NOTOPER]	= ccw_device_nop,
-		[DEV_EVENT_INTERRUPT]	= ccw_device_bug,
+		[DEV_EVENT_INTERRUPT]	= ccw_device_disabled_irq,
 		[DEV_EVENT_TIMEOUT]	= ccw_device_nop,
 		[DEV_EVENT_VERIFY]	= ccw_device_nop,
 	},
 	[DEV_STATE_SENSE_PGID] = {
-		[DEV_EVENT_NOTOPER]	= ccw_device_generic_notoper,
-		[DEV_EVENT_INTERRUPT]	= ccw_device_sense_pgid_irq,
-		[DEV_EVENT_TIMEOUT]	= ccw_device_onoff_timeout,
+		[DEV_EVENT_NOTOPER]	= ccw_device_request_event,
+		[DEV_EVENT_INTERRUPT]	= ccw_device_request_event,
+		[DEV_EVENT_TIMEOUT]	= ccw_device_request_event,
 		[DEV_EVENT_VERIFY]	= ccw_device_nop,
 	},
 	[DEV_STATE_SENSE_ID] = {
-		[DEV_EVENT_NOTOPER]	= ccw_device_recog_notoper,
-		[DEV_EVENT_INTERRUPT]	= ccw_device_sense_id_irq,
-		[DEV_EVENT_TIMEOUT]	= ccw_device_recog_timeout,
+		[DEV_EVENT_NOTOPER]	= ccw_device_request_event,
+		[DEV_EVENT_INTERRUPT]	= ccw_device_request_event,
+		[DEV_EVENT_TIMEOUT]	= ccw_device_request_event,
 		[DEV_EVENT_VERIFY]	= ccw_device_nop,
 	},
 	[DEV_STATE_OFFLINE] = {
 		[DEV_EVENT_NOTOPER]	= ccw_device_generic_notoper,
-		[DEV_EVENT_INTERRUPT]	= ccw_device_offline_irq,
+		[DEV_EVENT_INTERRUPT]	= ccw_device_disabled_irq,
 		[DEV_EVENT_TIMEOUT]	= ccw_device_nop,
 		[DEV_EVENT_VERIFY]	= ccw_device_offline_verify,
 	},
 	[DEV_STATE_VERIFY] = {
-		[DEV_EVENT_NOTOPER]	= ccw_device_generic_notoper,
-		[DEV_EVENT_INTERRUPT]	= ccw_device_verify_irq,
-		[DEV_EVENT_TIMEOUT]	= ccw_device_onoff_timeout,
+		[DEV_EVENT_NOTOPER]	= ccw_device_request_event,
+		[DEV_EVENT_INTERRUPT]	= ccw_device_request_event,
+		[DEV_EVENT_TIMEOUT]	= ccw_device_request_event,
 		[DEV_EVENT_VERIFY]	= ccw_device_delay_verify,
 	},
 	[DEV_STATE_ONLINE] = {
@@ -1209,24 +986,18 @@ fsm_func_t *dev_jumptable[NR_DEV_STATES][NR_DEV_EVENTS] = {
 		[DEV_EVENT_VERIFY]	= ccw_device_online_verify,
 	},
 	[DEV_STATE_DISBAND_PGID] = {
-		[DEV_EVENT_NOTOPER]	= ccw_device_generic_notoper,
-		[DEV_EVENT_INTERRUPT]	= ccw_device_disband_irq,
-		[DEV_EVENT_TIMEOUT]	= ccw_device_onoff_timeout,
+		[DEV_EVENT_NOTOPER]	= ccw_device_request_event,
+		[DEV_EVENT_INTERRUPT]	= ccw_device_request_event,
+		[DEV_EVENT_TIMEOUT]	= ccw_device_request_event,
 		[DEV_EVENT_VERIFY]	= ccw_device_nop,
 	},
 	[DEV_STATE_BOXED] = {
 		[DEV_EVENT_NOTOPER]	= ccw_device_generic_notoper,
-		[DEV_EVENT_INTERRUPT]	= ccw_device_stlck_done,
-		[DEV_EVENT_TIMEOUT]	= ccw_device_stlck_done,
-		[DEV_EVENT_VERIFY]	= ccw_device_nop,
-	},
-	/* states to wait for i/o completion before doing something */
-	[DEV_STATE_CLEAR_VERIFY] = {
-		[DEV_EVENT_NOTOPER]	= ccw_device_generic_notoper,
-		[DEV_EVENT_INTERRUPT]   = ccw_device_clear_verify,
+		[DEV_EVENT_INTERRUPT]	= ccw_device_nop,
 		[DEV_EVENT_TIMEOUT]	= ccw_device_nop,
-		[DEV_EVENT_VERIFY]	= ccw_device_nop,
+		[DEV_EVENT_VERIFY]	= ccw_device_boxed_verify,
 	},
+	/* states to wait for i/o completion before doing something */
 	[DEV_STATE_TIMEOUT_KILL] = {
 		[DEV_EVENT_NOTOPER]	= ccw_device_generic_notoper,
 		[DEV_EVENT_INTERRUPT]	= ccw_device_killing_irq,
@@ -1243,13 +1014,13 @@ fsm_func_t *dev_jumptable[NR_DEV_STATES][NR_DEV_EVENTS] = {
 	[DEV_STATE_DISCONNECTED] = {
 		[DEV_EVENT_NOTOPER]	= ccw_device_nop,
 		[DEV_EVENT_INTERRUPT]	= ccw_device_start_id,
-		[DEV_EVENT_TIMEOUT]	= ccw_device_bug,
+		[DEV_EVENT_TIMEOUT]	= ccw_device_nop,
 		[DEV_EVENT_VERIFY]	= ccw_device_start_id,
 	},
 	[DEV_STATE_DISCONNECTED_SENSE_ID] = {
-		[DEV_EVENT_NOTOPER]	= ccw_device_recog_notoper,
-		[DEV_EVENT_INTERRUPT]	= ccw_device_sense_id_irq,
-		[DEV_EVENT_TIMEOUT]	= ccw_device_recog_timeout,
+		[DEV_EVENT_NOTOPER]	= ccw_device_request_event,
+		[DEV_EVENT_INTERRUPT]	= ccw_device_request_event,
+		[DEV_EVENT_TIMEOUT]	= ccw_device_request_event,
 		[DEV_EVENT_VERIFY]	= ccw_device_nop,
 	},
 	[DEV_STATE_CMFCHANGE] = {
@@ -1264,6 +1035,12 @@ fsm_func_t *dev_jumptable[NR_DEV_STATES][NR_DEV_EVENTS] = {
 		[DEV_EVENT_TIMEOUT]	= ccw_device_update_cmfblock,
 		[DEV_EVENT_VERIFY]	= ccw_device_update_cmfblock,
 	},
+	[DEV_STATE_STEAL_LOCK] = {
+		[DEV_EVENT_NOTOPER]	= ccw_device_request_event,
+		[DEV_EVENT_INTERRUPT]	= ccw_device_request_event,
+		[DEV_EVENT_TIMEOUT]	= ccw_device_request_event,
+		[DEV_EVENT_VERIFY]	= ccw_device_nop,
+	},
 };
 
 EXPORT_SYMBOL_GPL(ccw_device_set_timeout);

drivers/s390/cio/device_id.c

@@ -1,40 +1,39 @@
 /*
- * drivers/s390/cio/device_id.c
+ *  CCW device SENSE ID I/O handling.
  *
- *    Copyright (C) 2002 IBM Deutschland Entwicklung GmbH,
- *			 IBM Corporation
- *    Author(s): Cornelia Huck (cornelia.huck@de.ibm.com)
- *		 Martin Schwidefsky (schwidefsky@de.ibm.com)
- *
- * Sense ID functions.
+ *    Copyright IBM Corp. 2002,2009
+ *    Author(s): Cornelia Huck <cornelia.huck@de.ibm.com>
+ *		 Martin Schwidefsky <schwidefsky@de.ibm.com>
+ *		 Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
  */
 
-#include <linux/module.h>
-#include <linux/init.h>
 #include <linux/kernel.h>
-
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/errno.h>
 #include <asm/ccwdev.h>
-#include <asm/delay.h>
+#include <asm/setup.h>
 #include <asm/cio.h>
-#include <asm/lowcore.h>
 #include <asm/diag.h>
 
 #include "cio.h"
 #include "cio_debug.h"
-#include "css.h"
 #include "device.h"
-#include "ioasm.h"
 #include "io_sch.h"
 
+#define SENSE_ID_RETRIES	256
+#define SENSE_ID_TIMEOUT	(10 * HZ)
+#define SENSE_ID_MIN_LEN	4
+#define SENSE_ID_BASIC_LEN	7
+
 /**
- * vm_vdev_to_cu_type - Convert vm virtual device into control unit type
- *			for certain devices.
- * @class: virtual device class
- * @type: virtual device type
+ * diag210_to_senseid - convert diag 0x210 data to sense id information
+ * @senseid: sense id
+ * @diag: diag 0x210 data
  *
- * Returns control unit type if a match was made or %0xffff otherwise.
+ * Return 0 on success, non-zero otherwise.
  */
-static int vm_vdev_to_cu_type(int class, int type)
+static int diag210_to_senseid(struct senseid *senseid, struct diag210 *diag)
 {
 	static struct {
 		int class, type, cu_type;
@@ -71,253 +70,153 @@ static int vm_vdev_to_cu_type(int class, int type)
 	};
 	int i;
 
-	for (i = 0; i < ARRAY_SIZE(vm_devices); i++)
-		if (class == vm_devices[i].class && type == vm_devices[i].type)
-			return vm_devices[i].cu_type;
+	/* Special case for osa devices. */
+	if (diag->vrdcvcla == 0x02 && diag->vrdcvtyp == 0x20) {
+		senseid->cu_type = 0x3088;
+		senseid->cu_model = 0x60;
+		senseid->reserved = 0xff;
+		return 0;
+	}
+	for (i = 0; i < ARRAY_SIZE(vm_devices); i++) {
+		if (diag->vrdcvcla == vm_devices[i].class &&
+		    diag->vrdcvtyp == vm_devices[i].type) {
+			senseid->cu_type = vm_devices[i].cu_type;
+			senseid->reserved = 0xff;
+			return 0;
+		}
+	}
 
-	return 0xffff;
+	return -ENODEV;
 }
 
 /**
- * diag_get_dev_info - retrieve device information via DIAG X'210'
- * @devno: device number
- * @ps: pointer to sense ID data area
+ * diag_get_dev_info - retrieve device information via diag 0x210
+ * @cdev: ccw device
  *
  * Returns zero on success, non-zero otherwise.
  */
-static int diag_get_dev_info(u16 devno, struct senseid *ps)
+static int diag210_get_dev_info(struct ccw_device *cdev)
 {
+	struct ccw_dev_id *dev_id = &cdev->private->dev_id;
+	struct senseid *senseid = &cdev->private->senseid;
 	struct diag210 diag_data;
-	int ccode;
-
-	CIO_TRACE_EVENT (4, "VMvdinf");
-
-	diag_data = (struct diag210) {
-		.vrdcdvno = devno,
-		.vrdclen = sizeof (diag_data),
-	};
-
-	ccode = diag210 (&diag_data);
-	if ((ccode == 0) || (ccode == 2)) {
-		ps->reserved = 0xff;
-
-		/* Special case for osa devices. */
-		if (diag_data.vrdcvcla == 0x02 && diag_data.vrdcvtyp == 0x20) {
-			ps->cu_type = 0x3088;
-			ps->cu_model = 0x60;
-			return 0;
-		}
-		ps->cu_type = vm_vdev_to_cu_type(diag_data.vrdcvcla,
-						diag_data.vrdcvtyp);
-		if (ps->cu_type != 0xffff)
-			return 0;
-	}
-
-	CIO_MSG_EVENT(0, "DIAG X'210' for device %04X returned (cc = %d):"
-		      "vdev class : %02X, vdev type : %04X \n ...  "
-		      "rdev class : %02X, rdev type : %04X, "
-		      "rdev model: %02X\n",
-		      devno, ccode,
-		      diag_data.vrdcvcla, diag_data.vrdcvtyp,
-		      diag_data.vrdcrccl, diag_data.vrdccrty,
-		      diag_data.vrdccrmd);
-
+	int rc;
+
+	if (dev_id->ssid != 0)
+		return -ENODEV;
+	memset(&diag_data, 0, sizeof(diag_data));
+	diag_data.vrdcdvno	= dev_id->devno;
+	diag_data.vrdclen	= sizeof(diag_data);
+	rc = diag210(&diag_data);
+	CIO_TRACE_EVENT(4, "diag210");
+	CIO_HEX_EVENT(4, &rc, sizeof(rc));
+	CIO_HEX_EVENT(4, &diag_data, sizeof(diag_data));
+	if (rc != 0 && rc != 2)
+		goto err_failed;
+	if (diag210_to_senseid(senseid, &diag_data))
+		goto err_unknown;
+	return 0;
+
+err_unknown:
+	CIO_MSG_EVENT(0, "snsid: device 0.%x.%04x: unknown diag210 data\n",
+		      dev_id->ssid, dev_id->devno);
+	return -ENODEV;
+err_failed:
+	CIO_MSG_EVENT(0, "snsid: device 0.%x.%04x: diag210 failed (rc=%d)\n",
+		      dev_id->ssid, dev_id->devno, rc);
 	return -ENODEV;
 }
 
 /*
- * Start Sense ID helper function.
- * Try to obtain the 'control unit'/'device type' information
- *  associated with the subchannel.
+ * Initialize SENSE ID data.
  */
-static int
-__ccw_device_sense_id_start(struct ccw_device *cdev)
-{
-	struct subchannel *sch;
-	struct ccw1 *ccw;
-	int ret;
-
-	sch = to_subchannel(cdev->dev.parent);
-	/* Setup sense channel program. */
-	ccw = cdev->private->iccws;
-	ccw->cmd_code = CCW_CMD_SENSE_ID;
-	ccw->cda = (__u32) __pa (&cdev->private->senseid);
-	ccw->count = sizeof (struct senseid);
-	ccw->flags = CCW_FLAG_SLI;
-
-	/* Reset device status. */
-	memset(&cdev->private->irb, 0, sizeof(struct irb));
-
-	/* Try on every path. */
-	ret = -ENODEV;
-	while (cdev->private->imask != 0) {
-		cdev->private->senseid.cu_type = 0xFFFF;
-		if ((sch->opm & cdev->private->imask) != 0 &&
-		    cdev->private->iretry > 0) {
-			cdev->private->iretry--;
-			/* Reset internal retry indication. */
-			cdev->private->flags.intretry = 0;
-			ret = cio_start (sch, cdev->private->iccws,
-					 cdev->private->imask);
-			/* ret is 0, -EBUSY, -EACCES or -ENODEV */
-			if (ret != -EACCES)
-				return ret;
-		}
-		cdev->private->imask >>= 1;
-		cdev->private->iretry = 5;
-	}
-	return ret;
-}
-
-void
-ccw_device_sense_id_start(struct ccw_device *cdev)
+static void snsid_init(struct ccw_device *cdev)
 {
-	int ret;
-
-	memset (&cdev->private->senseid, 0, sizeof (struct senseid));
-	cdev->private->imask = 0x80;
-	cdev->private->iretry = 5;
-	ret = __ccw_device_sense_id_start(cdev);
-	if (ret && ret != -EBUSY)
-		ccw_device_sense_id_done(cdev, ret);
+	cdev->private->flags.esid = 0;
+	memset(&cdev->private->senseid, 0, sizeof(cdev->private->senseid));
+	cdev->private->senseid.cu_type = 0xffff;
 }
 
 /*
- * Called from interrupt context to check if a valid answer
- * to Sense ID was received.
+ * Check for complete SENSE ID data.
  */
-static int
-ccw_device_check_sense_id(struct ccw_device *cdev)
+static int snsid_check(struct ccw_device *cdev, void *data)
 {
-	struct subchannel *sch;
-	struct irb *irb;
-
-	sch = to_subchannel(cdev->dev.parent);
-	irb = &cdev->private->irb;
-
-	/* Check the error cases. */
-	if (irb->scsw.cmd.fctl & (SCSW_FCTL_HALT_FUNC | SCSW_FCTL_CLEAR_FUNC)) {
-		/* Retry Sense ID if requested. */
-		if (cdev->private->flags.intretry) {
-			cdev->private->flags.intretry = 0;
-			return -EAGAIN;
-		}
-		return -ETIME;
-	}
-	if (irb->esw.esw0.erw.cons && (irb->ecw[0] & SNS0_CMD_REJECT)) {
-		/*
-		 * if the device doesn't support the SenseID
-		 *  command further retries wouldn't help ...
-		 * NB: We don't check here for intervention required like we
-		 *     did before, because tape devices with no tape inserted
-		 *     may present this status *in conjunction with* the
-		 *     sense id information. So, for intervention required,
-		 *     we use the "whack it until it talks" strategy...
-		 */
-		CIO_MSG_EVENT(0, "SenseID : device %04x on Subchannel "
-			      "0.%x.%04x reports cmd reject\n",
-			      cdev->private->dev_id.devno, sch->schid.ssid,
-			      sch->schid.sch_no);
+	struct cmd_scsw *scsw = &cdev->private->irb.scsw.cmd;
+	int len = sizeof(struct senseid) - scsw->count;
+
+	/* Check for incomplete SENSE ID data. */
+	if (len < SENSE_ID_MIN_LEN)
+		goto out_restart;
+	if (cdev->private->senseid.cu_type == 0xffff)
+		goto out_restart;
+	/* Check for incompatible SENSE ID data. */
+	if (cdev->private->senseid.reserved != 0xff)
 		return -EOPNOTSUPP;
-	}
-	if (irb->esw.esw0.erw.cons) {
-		CIO_MSG_EVENT(2, "SenseID : UC on dev 0.%x.%04x, "
-			      "lpum %02X, cnt %02d, sns :"
-			      " %02X%02X%02X%02X %02X%02X%02X%02X ...\n",
-			      cdev->private->dev_id.ssid,
-			      cdev->private->dev_id.devno,
-			      irb->esw.esw0.sublog.lpum,
-			      irb->esw.esw0.erw.scnt,
-			      irb->ecw[0], irb->ecw[1],
-			      irb->ecw[2], irb->ecw[3],
-			      irb->ecw[4], irb->ecw[5],
-			      irb->ecw[6], irb->ecw[7]);
-		return -EAGAIN;
-	}
-	if (irb->scsw.cmd.cc == 3) {
-		u8 lpm;
+	/* Check for extended-identification information. */
+	if (len > SENSE_ID_BASIC_LEN)
+		cdev->private->flags.esid = 1;
+	return 0;
 
-		lpm = to_io_private(sch)->orb.cmd.lpm;
-		if ((lpm & sch->schib.pmcw.pim & sch->schib.pmcw.pam) != 0)
-			CIO_MSG_EVENT(4, "SenseID : path %02X for device %04x "
-				      "on subchannel 0.%x.%04x is "
-				      "'not operational'\n", lpm,
-				      cdev->private->dev_id.devno,
-				      sch->schid.ssid, sch->schid.sch_no);
-		return -EACCES;
-	}
-
-	/* Did we get a proper answer ? */
-	if (irb->scsw.cmd.cc == 0 && cdev->private->senseid.cu_type != 0xFFFF &&
-	    cdev->private->senseid.reserved == 0xFF) {
-		if (irb->scsw.cmd.count < sizeof(struct senseid) - 8)
-			cdev->private->flags.esid = 1;
-		return 0; /* Success */
-	}
-
-	/* Hmm, whatever happened, try again. */
-	CIO_MSG_EVENT(2, "SenseID : start_IO() for device %04x on "
-		      "subchannel 0.%x.%04x returns status %02X%02X\n",
-		      cdev->private->dev_id.devno, sch->schid.ssid,
-		      sch->schid.sch_no,
-		      irb->scsw.cmd.dstat, irb->scsw.cmd.cstat);
+out_restart:
+	snsid_init(cdev);
 	return -EAGAIN;
 }
 
 /*
- * Got interrupt for Sense ID.
+ * Process SENSE ID request result.
  */
-void
-ccw_device_sense_id_irq(struct ccw_device *cdev, enum dev_event dev_event)
+static void snsid_callback(struct ccw_device *cdev, void *data, int rc)
 {
-	struct subchannel *sch;
-	struct irb *irb;
-	int ret;
-
-	sch = to_subchannel(cdev->dev.parent);
-	irb = (struct irb *) __LC_IRB;
-	/* Retry sense id, if needed. */
-	if (irb->scsw.cmd.stctl ==
-	    (SCSW_STCTL_STATUS_PEND | SCSW_STCTL_ALERT_STATUS)) {
-		if ((irb->scsw.cmd.cc == 1) || !irb->scsw.cmd.actl) {
-			ret = __ccw_device_sense_id_start(cdev);
-			if (ret && ret != -EBUSY)
-				ccw_device_sense_id_done(cdev, ret);
+	struct ccw_dev_id *id = &cdev->private->dev_id;
+	struct senseid *senseid = &cdev->private->senseid;
+	int vm = 0;
+
+	if (rc && MACHINE_IS_VM) {
+		/* Try diag 0x210 fallback on z/VM. */
+		snsid_init(cdev);
+		if (diag210_get_dev_info(cdev) == 0) {
+			rc = 0;
+			vm = 1;
 		}
-		return;
 	}
-	if (ccw_device_accumulate_and_sense(cdev, irb) != 0)
-		return;
-	ret = ccw_device_check_sense_id(cdev);
-	memset(&cdev->private->irb, 0, sizeof(struct irb));
-	switch (ret) {
-	/* 0, -ETIME, -EOPNOTSUPP, -EAGAIN or -EACCES */
-	case 0:			/* Sense id succeeded. */
-	case -ETIME:		/* Sense id stopped by timeout. */
-		ccw_device_sense_id_done(cdev, ret);
-		break;
-	case -EACCES:		/* channel is not operational. */
-		sch->lpm &= ~cdev->private->imask;
-		cdev->private->imask >>= 1;
-		cdev->private->iretry = 5;
-		/* fall through. */
-	case -EAGAIN:		/* try again. */
-		ret = __ccw_device_sense_id_start(cdev);
-		if (ret == 0 || ret == -EBUSY)
-			break;
-		/* fall through. */
-	default:		/* Sense ID failed. Try asking VM. */
-		if (MACHINE_IS_VM)
-			ret = diag_get_dev_info(cdev->private->dev_id.devno,
-						&cdev->private->senseid);
-		else
-			/*
-			 * If we can't couldn't identify the device type we
-			 *  consider the device "not operational".
-			 */
-			ret = -ENODEV;
+	CIO_MSG_EVENT(2, "snsid: device 0.%x.%04x: rc=%d %04x/%02x "
+		      "%04x/%02x%s\n", id->ssid, id->devno, rc,
+		      senseid->cu_type, senseid->cu_model, senseid->dev_type,
+		      senseid->dev_model, vm ? " (diag210)" : "");
+	ccw_device_sense_id_done(cdev, rc);
+}
 
-		ccw_device_sense_id_done(cdev, ret);
-		break;
-	}
+/**
+ * ccw_device_sense_id_start - perform SENSE ID
+ * @cdev: ccw device
+ *
+ * Execute a SENSE ID channel program on @cdev to update its sense id
+ * information. When finished, call ccw_device_sense_id_done with a
+ * return code specifying the result.
+ */
+void ccw_device_sense_id_start(struct ccw_device *cdev)
+{
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+	struct ccw1 *cp = cdev->private->iccws;
+
+	CIO_TRACE_EVENT(4, "snsid");
+	CIO_HEX_EVENT(4, &cdev->private->dev_id, sizeof(cdev->private->dev_id));
+	/* Data setup. */
+	snsid_init(cdev);
+	/* Channel program setup. */
+	cp->cmd_code	= CCW_CMD_SENSE_ID;
+	cp->cda		= (u32) (addr_t) &cdev->private->senseid;
+	cp->count	= sizeof(struct senseid);
+	cp->flags	= CCW_FLAG_SLI;
+	/* Request setup. */
+	memset(req, 0, sizeof(*req));
+	req->cp		= cp;
+	req->timeout	= SENSE_ID_TIMEOUT;
+	req->maxretries	= SENSE_ID_RETRIES;
+	req->lpm	= sch->schib.pmcw.pam & sch->opm;
+	req->check	= snsid_check;
+	req->callback	= snsid_callback;
+	ccw_request_start(cdev);
 }

drivers/s390/cio/device_ops.c

@@ -11,6 +11,7 @@
 #include <linux/list.h>
 #include <linux/device.h>
 #include <linux/delay.h>
+#include <linux/completion.h>
 
 #include <asm/ccwdev.h>
 #include <asm/idals.h>
@@ -46,6 +47,7 @@ int ccw_device_set_options_mask(struct ccw_device *cdev, unsigned long flags)
 	cdev->private->options.repall = (flags & CCWDEV_REPORT_ALL) != 0;
 	cdev->private->options.pgroup = (flags & CCWDEV_DO_PATHGROUP) != 0;
 	cdev->private->options.force = (flags & CCWDEV_ALLOW_FORCE) != 0;
+	cdev->private->options.mpath = (flags & CCWDEV_DO_MULTIPATH) != 0;
 	return 0;
 }
 
@@ -74,6 +76,7 @@ int ccw_device_set_options(struct ccw_device *cdev, unsigned long flags)
 	cdev->private->options.repall |= (flags & CCWDEV_REPORT_ALL) != 0;
 	cdev->private->options.pgroup |= (flags & CCWDEV_DO_PATHGROUP) != 0;
 	cdev->private->options.force |= (flags & CCWDEV_ALLOW_FORCE) != 0;
+	cdev->private->options.mpath |= (flags & CCWDEV_DO_MULTIPATH) != 0;
 	return 0;
 }
 
@@ -90,8 +93,33 @@ void ccw_device_clear_options(struct ccw_device *cdev, unsigned long flags)
 	cdev->private->options.repall &= (flags & CCWDEV_REPORT_ALL) == 0;
 	cdev->private->options.pgroup &= (flags & CCWDEV_DO_PATHGROUP) == 0;
 	cdev->private->options.force &= (flags & CCWDEV_ALLOW_FORCE) == 0;
+	cdev->private->options.mpath &= (flags & CCWDEV_DO_MULTIPATH) == 0;
 }
 
+/**
+ * ccw_device_is_pathgroup - determine if paths to this device are grouped
+ * @cdev: ccw device
+ *
+ * Return non-zero if there is a path group, zero otherwise.
+ */
+int ccw_device_is_pathgroup(struct ccw_device *cdev)
+{
+	return cdev->private->flags.pgroup;
+}
+EXPORT_SYMBOL(ccw_device_is_pathgroup);
+
+/**
+ * ccw_device_is_multipath - determine if device is operating in multipath mode
+ * @cdev: ccw device
+ *
+ * Return non-zero if device is operating in multipath mode, zero otherwise.
+ */
+int ccw_device_is_multipath(struct ccw_device *cdev)
+{
+	return cdev->private->flags.mpath;
+}
+EXPORT_SYMBOL(ccw_device_is_multipath);
+
 /**
  * ccw_device_clear() - terminate I/O request processing
  * @cdev: target ccw device
@@ -167,8 +195,7 @@ int ccw_device_start_key(struct ccw_device *cdev, struct ccw1 *cpa,
 		return -EINVAL;
 	if (cdev->private->state == DEV_STATE_NOT_OPER)
 		return -ENODEV;
-	if (cdev->private->state == DEV_STATE_VERIFY ||
-	    cdev->private->state == DEV_STATE_CLEAR_VERIFY) {
+	if (cdev->private->state == DEV_STATE_VERIFY) {
 		/* Remember to fake irb when finished. */
 		if (!cdev->private->flags.fake_irb) {
 			cdev->private->flags.fake_irb = 1;
@@ -478,74 +505,65 @@ __u8 ccw_device_get_path_mask(struct ccw_device *cdev)
 	return sch->lpm;
 }
 
-/*
- * Try to break the lock on a boxed device.
- */
-int
-ccw_device_stlck(struct ccw_device *cdev)
-{
-	void *buf, *buf2;
-	unsigned long flags;
-	struct subchannel *sch;
-	int ret;
+struct stlck_data {
+	struct completion done;
+	int rc;
+};
 
-	if (!cdev)
-		return -ENODEV;
+void ccw_device_stlck_done(struct ccw_device *cdev, void *data, int rc)
+{
+	struct stlck_data *sdata = data;
 
-	if (cdev->drv && !cdev->private->options.force)
-		return -EINVAL;
+	sdata->rc = rc;
+	complete(&sdata->done);
+}
 
-	sch = to_subchannel(cdev->dev.parent);
-	
-	CIO_TRACE_EVENT(2, "stl lock");
-	CIO_TRACE_EVENT(2, dev_name(&cdev->dev));
+/*
+ * Perform unconditional reserve + release.
+ */
+int ccw_device_stlck(struct ccw_device *cdev)
+{
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct stlck_data data;
+	u8 *buffer;
+	int rc;
 
-	buf = kmalloc(32*sizeof(char), GFP_DMA|GFP_KERNEL);
-	if (!buf)
-		return -ENOMEM;
-	buf2 = kmalloc(32*sizeof(char), GFP_DMA|GFP_KERNEL);
-	if (!buf2) {
-		kfree(buf);
-		return -ENOMEM;
+	/* Check if steal lock operation is valid for this device. */
+	if (cdev->drv) {
+		if (!cdev->private->options.force)
+			return -EINVAL;
 	}
-	spin_lock_irqsave(sch->lock, flags);
-	ret = cio_enable_subchannel(sch, (u32)(addr_t)sch);
-	if (ret)
-		goto out_unlock;
-	/*
-	 * Setup ccw. We chain an unconditional reserve and a release so we
-	 * only break the lock.
-	 */
-	cdev->private->iccws[0].cmd_code = CCW_CMD_STLCK;
-	cdev->private->iccws[0].cda = (__u32) __pa(buf);
-	cdev->private->iccws[0].count = 32;
-	cdev->private->iccws[0].flags = CCW_FLAG_CC;
-	cdev->private->iccws[1].cmd_code = CCW_CMD_RELEASE;
-	cdev->private->iccws[1].cda = (__u32) __pa(buf2);
-	cdev->private->iccws[1].count = 32;
-	cdev->private->iccws[1].flags = 0;
-	ret = cio_start(sch, cdev->private->iccws, 0);
-	if (ret) {
-		cio_disable_subchannel(sch); //FIXME: return code?
+	buffer = kzalloc(64, GFP_DMA | GFP_KERNEL);
+	if (!buffer)
+		return -ENOMEM;
+	init_completion(&data.done);
+	data.rc = -EIO;
+	spin_lock_irq(sch->lock);
+	rc = cio_enable_subchannel(sch, (u32) (addr_t) sch);
+	if (rc)
 		goto out_unlock;
+	/* Perform operation. */
+	cdev->private->state = DEV_STATE_STEAL_LOCK,
+	ccw_device_stlck_start(cdev, &data, &buffer[0], &buffer[32]);
+	spin_unlock_irq(sch->lock);
+	/* Wait for operation to finish. */
+	if (wait_for_completion_interruptible(&data.done)) {
+		/* Got a signal. */
+		spin_lock_irq(sch->lock);
+		ccw_request_cancel(cdev);
+		spin_unlock_irq(sch->lock);
+		wait_for_completion(&data.done);
 	}
-	cdev->private->irb.scsw.cmd.actl |= SCSW_ACTL_START_PEND;
-	spin_unlock_irqrestore(sch->lock, flags);
-	wait_event(cdev->private->wait_q,
-		   cdev->private->irb.scsw.cmd.actl == 0);
-	spin_lock_irqsave(sch->lock, flags);
-	cio_disable_subchannel(sch); //FIXME: return code?
-	if ((cdev->private->irb.scsw.cmd.dstat !=
-	     (DEV_STAT_CHN_END|DEV_STAT_DEV_END)) ||
-	    (cdev->private->irb.scsw.cmd.cstat != 0))
-		ret = -EIO;
-	/* Clear irb. */
-	memset(&cdev->private->irb, 0, sizeof(struct irb));
+	rc = data.rc;
+	/* Check results. */
+	spin_lock_irq(sch->lock);
+	cio_disable_subchannel(sch);
+	cdev->private->state = DEV_STATE_BOXED;
 out_unlock:
-	kfree(buf);
-	kfree(buf2);
-	spin_unlock_irqrestore(sch->lock, flags);
-	return ret;
+	spin_unlock_irq(sch->lock);
+	kfree(buffer);
+
+	return rc;
 }
 
 void *ccw_device_get_chp_desc(struct ccw_device *cdev, int chp_no)

drivers/s390/cio/device_pgid.c

@@ -1,594 +1,561 @@
 /*
- * drivers/s390/cio/device_pgid.c
+ *  CCW device PGID and path verification I/O handling.
  *
- *    Copyright (C) 2002 IBM Deutschland Entwicklung GmbH,
- *			 IBM Corporation
- *    Author(s): Cornelia Huck (cornelia.huck@de.ibm.com)
- *		 Martin Schwidefsky (schwidefsky@de.ibm.com)
- *
- * Path Group ID functions.
+ *    Copyright IBM Corp. 2002,2009
+ *    Author(s): Cornelia Huck <cornelia.huck@de.ibm.com>
+ *		 Martin Schwidefsky <schwidefsky@de.ibm.com>
+ *		 Peter Oberparleiter <peter.oberparleiter@de.ibm.com>
  */
 
-#include <linux/module.h>
-#include <linux/init.h>
-
+#include <linux/kernel.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/errno.h>
+#include <linux/bitops.h>
 #include <asm/ccwdev.h>
 #include <asm/cio.h>
-#include <asm/delay.h>
-#include <asm/lowcore.h>
 
 #include "cio.h"
 #include "cio_debug.h"
-#include "css.h"
 #include "device.h"
-#include "ioasm.h"
 #include "io_sch.h"
 
+#define PGID_RETRIES	256
+#define PGID_TIMEOUT	(10 * HZ)
+
 /*
- * Helper function called from interrupt context to decide whether an
- * operation should be tried again.
+ * Process path verification data and report result.
  */
-static int __ccw_device_should_retry(union scsw *scsw)
+static void verify_done(struct ccw_device *cdev, int rc)
 {
-	/* CC is only valid if start function bit is set. */
-	if ((scsw->cmd.fctl & SCSW_FCTL_START_FUNC) && scsw->cmd.cc == 1)
-		return 1;
-	/* No more activity. For sense and set PGID we stubbornly try again. */
-	if (!scsw->cmd.actl)
-		return 1;
-	return 0;
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_dev_id *id = &cdev->private->dev_id;
+	int mpath = cdev->private->flags.mpath;
+	int pgroup = cdev->private->flags.pgroup;
+
+	if (rc)
+		goto out;
+	/* Ensure consistent multipathing state at device and channel. */
+	if (sch->config.mp != mpath) {
+		sch->config.mp = mpath;
+		rc = cio_commit_config(sch);
+	}
+out:
+	CIO_MSG_EVENT(2, "vrfy: device 0.%x.%04x: rc=%d pgroup=%d mpath=%d "
+			 "vpm=%02x\n", id->ssid, id->devno, rc, pgroup, mpath,
+			 sch->vpm);
+	ccw_device_verify_done(cdev, rc);
 }
 
 /*
- * Start Sense Path Group ID helper function. Used in ccw_device_recog
- * and ccw_device_sense_pgid.
+ * Create channel program to perform a NOOP.
  */
-static int
-__ccw_device_sense_pgid_start(struct ccw_device *cdev)
+static void nop_build_cp(struct ccw_device *cdev)
 {
-	struct subchannel *sch;
-	struct ccw1 *ccw;
-	int ret;
-	int i;
-
-	sch = to_subchannel(cdev->dev.parent);
-	/* Return if we already checked on all paths. */
-	if (cdev->private->imask == 0)
-		return (sch->lpm == 0) ? -ENODEV : -EACCES;
-	i = 8 - ffs(cdev->private->imask);
-
-	/* Setup sense path group id channel program. */
-	ccw = cdev->private->iccws;
-	ccw->cmd_code = CCW_CMD_SENSE_PGID;
-	ccw->count = sizeof (struct pgid);
-	ccw->flags = CCW_FLAG_SLI;
-
-	/* Reset device status. */
-	memset(&cdev->private->irb, 0, sizeof(struct irb));
-	/* Try on every path. */
-	ret = -ENODEV;
-	while (cdev->private->imask != 0) {
-		/* Try every path multiple times. */
-		ccw->cda = (__u32) __pa (&cdev->private->pgid[i]);
-		if (cdev->private->iretry > 0) {
-			cdev->private->iretry--;
-			/* Reset internal retry indication. */
-			cdev->private->flags.intretry = 0;
-			ret = cio_start (sch, cdev->private->iccws, 
-					 cdev->private->imask);
-			/* ret is 0, -EBUSY, -EACCES or -ENODEV */
-			if (ret != -EACCES)
-				return ret;
-			CIO_MSG_EVENT(3, "SNID - Device %04x on Subchannel "
-				      "0.%x.%04x, lpm %02X, became 'not "
-				      "operational'\n",
-				      cdev->private->dev_id.devno,
-				      sch->schid.ssid,
-				      sch->schid.sch_no, cdev->private->imask);
-
-		}
-		cdev->private->imask >>= 1;
-		cdev->private->iretry = 5;
-		i++;
-	}
-
-	return ret;
+	struct ccw_request *req = &cdev->private->req;
+	struct ccw1 *cp = cdev->private->iccws;
+
+	cp->cmd_code	= CCW_CMD_NOOP;
+	cp->cda		= 0;
+	cp->count	= 0;
+	cp->flags	= CCW_FLAG_SLI;
+	req->cp		= cp;
 }
 
-void
-ccw_device_sense_pgid_start(struct ccw_device *cdev)
+/*
+ * Perform NOOP on a single path.
+ */
+static void nop_do(struct ccw_device *cdev)
 {
-	int ret;
-
-	/* Set a timeout of 60s */
-	ccw_device_set_timeout(cdev, 60*HZ);
-
-	cdev->private->state = DEV_STATE_SENSE_PGID;
-	cdev->private->imask = 0x80;
-	cdev->private->iretry = 5;
-	memset (&cdev->private->pgid, 0, sizeof (cdev->private->pgid));
-	ret = __ccw_device_sense_pgid_start(cdev);
-	if (ret && ret != -EBUSY)
-		ccw_device_sense_pgid_done(cdev, ret);
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+
+	/* Adjust lpm. */
+	req->lpm = lpm_adjust(req->lpm, sch->schib.pmcw.pam & sch->opm);
+	if (!req->lpm)
+		goto out_nopath;
+	nop_build_cp(cdev);
+	ccw_request_start(cdev);
+	return;
+
+out_nopath:
+	verify_done(cdev, sch->vpm ? 0 : -EACCES);
 }
 
 /*
- * Called from interrupt context to check if a valid answer
- * to Sense Path Group ID was received.
+ * Adjust NOOP I/O status.
  */
-static int
-__ccw_device_check_sense_pgid(struct ccw_device *cdev)
+static enum io_status nop_filter(struct ccw_device *cdev, void *data,
+				 struct irb *irb, enum io_status status)
 {
-	struct subchannel *sch;
-	struct irb *irb;
-	int i;
-
-	sch = to_subchannel(cdev->dev.parent);
-	irb = &cdev->private->irb;
-	if (irb->scsw.cmd.fctl & (SCSW_FCTL_HALT_FUNC | SCSW_FCTL_CLEAR_FUNC)) {
-		/* Retry Sense PGID if requested. */
-		if (cdev->private->flags.intretry) {
-			cdev->private->flags.intretry = 0;
-			return -EAGAIN;
-		}
-		return -ETIME;
-	}
-	if (irb->esw.esw0.erw.cons &&
-	    (irb->ecw[0]&(SNS0_CMD_REJECT|SNS0_INTERVENTION_REQ))) {
-		/*
-		 * If the device doesn't support the Sense Path Group ID
-		 *  command further retries wouldn't help ...
-		 */
-		return -EOPNOTSUPP;
-	}
-	if (irb->esw.esw0.erw.cons) {
-		CIO_MSG_EVENT(2, "SNID - device 0.%x.%04x, unit check, "
-			      "lpum %02X, cnt %02d, sns : "
-			      "%02X%02X%02X%02X %02X%02X%02X%02X ...\n",
-			      cdev->private->dev_id.ssid,
-			      cdev->private->dev_id.devno,
-			      irb->esw.esw0.sublog.lpum,
-			      irb->esw.esw0.erw.scnt,
-			      irb->ecw[0], irb->ecw[1],
-			      irb->ecw[2], irb->ecw[3],
-			      irb->ecw[4], irb->ecw[5],
-			      irb->ecw[6], irb->ecw[7]);
-		return -EAGAIN;
-	}
-	if (irb->scsw.cmd.cc == 3) {
-		u8 lpm;
-
-		lpm = to_io_private(sch)->orb.cmd.lpm;
-		CIO_MSG_EVENT(3, "SNID - Device %04x on Subchannel 0.%x.%04x,"
-			      " lpm %02X, became 'not operational'\n",
-			      cdev->private->dev_id.devno, sch->schid.ssid,
-			      sch->schid.sch_no, lpm);
-		return -EACCES;
-	}
-	i = 8 - ffs(cdev->private->imask);
-	if (cdev->private->pgid[i].inf.ps.state2 == SNID_STATE2_RESVD_ELSE) {
-		CIO_MSG_EVENT(2, "SNID - Device %04x on Subchannel 0.%x.%04x "
-			      "is reserved by someone else\n",
-			      cdev->private->dev_id.devno, sch->schid.ssid,
-			      sch->schid.sch_no);
-		return -EUSERS;
-	}
-	return 0;
+	/* Only subchannel status might indicate a path error. */
+	if (status == IO_STATUS_ERROR && irb->scsw.cmd.cstat == 0)
+		return IO_DONE;
+	return status;
 }
 
 /*
- * Got interrupt for Sense Path Group ID.
+ * Process NOOP request result for a single path.
  */
-void
-ccw_device_sense_pgid_irq(struct ccw_device *cdev, enum dev_event dev_event)
+static void nop_callback(struct ccw_device *cdev, void *data, int rc)
 {
-	struct subchannel *sch;
-	struct irb *irb;
-	int ret;
-
-	irb = (struct irb *) __LC_IRB;
-
-	if (irb->scsw.cmd.stctl ==
-	    (SCSW_STCTL_STATUS_PEND | SCSW_STCTL_ALERT_STATUS)) {
-		if (__ccw_device_should_retry(&irb->scsw)) {
-			ret = __ccw_device_sense_pgid_start(cdev);
-			if (ret && ret != -EBUSY)
-				ccw_device_sense_pgid_done(cdev, ret);
-		}
-		return;
-	}
-	if (ccw_device_accumulate_and_sense(cdev, irb) != 0)
-		return;
-	sch = to_subchannel(cdev->dev.parent);
-	ret = __ccw_device_check_sense_pgid(cdev);
-	memset(&cdev->private->irb, 0, sizeof(struct irb));
-	switch (ret) {
-	/* 0, -ETIME, -EOPNOTSUPP, -EAGAIN, -EACCES or -EUSERS */
-	case -EOPNOTSUPP:	/* Sense Path Group ID not supported */
-		ccw_device_sense_pgid_done(cdev, -EOPNOTSUPP);
-		break;
-	case -ETIME:		/* Sense path group id stopped by timeout. */
-		ccw_device_sense_pgid_done(cdev, -ETIME);
-		break;
-	case -EACCES:		/* channel is not operational. */
-		sch->lpm &= ~cdev->private->imask;
-		/* Fall through. */
-	case 0:			/* Sense Path Group ID successful. */
-		cdev->private->imask >>= 1;
-		cdev->private->iretry = 5;
-		/* Fall through. */
-	case -EAGAIN:		/* Try again. */
-		ret = __ccw_device_sense_pgid_start(cdev);
-		if (ret != 0 && ret != -EBUSY)
-			ccw_device_sense_pgid_done(cdev, ret);
-		break;
-	case -EUSERS:		/* device is reserved for someone else. */
-		ccw_device_sense_pgid_done(cdev, -EUSERS);
-		break;
-	}
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+
+	if (rc == 0)
+		sch->vpm |= req->lpm;
+	else if (rc != -EACCES)
+		goto err;
+	req->lpm >>= 1;
+	nop_do(cdev);
+	return;
+
+err:
+	verify_done(cdev, rc);
 }
 
 /*
- * Path Group ID helper function.
+ * Create channel program to perform SET PGID on a single path.
  */
-static int
-__ccw_device_do_pgid(struct ccw_device *cdev, __u8 func)
+static void spid_build_cp(struct ccw_device *cdev, u8 fn)
 {
-	struct subchannel *sch;
-	struct ccw1 *ccw;
-	int ret;
-
-	sch = to_subchannel(cdev->dev.parent);
-
-	/* Setup sense path group id channel program. */
-	cdev->private->pgid[0].inf.fc = func;
-	ccw = cdev->private->iccws;
-	if (cdev->private->flags.pgid_single)
-		cdev->private->pgid[0].inf.fc |= SPID_FUNC_SINGLE_PATH;
-	else
-		cdev->private->pgid[0].inf.fc |= SPID_FUNC_MULTI_PATH;
-	ccw->cmd_code = CCW_CMD_SET_PGID;
-	ccw->cda = (__u32) __pa (&cdev->private->pgid[0]);
-	ccw->count = sizeof (struct pgid);
-	ccw->flags = CCW_FLAG_SLI;
-
-	/* Reset device status. */
-	memset(&cdev->private->irb, 0, sizeof(struct irb));
-
-	/* Try multiple times. */
-	ret = -EACCES;
-	if (cdev->private->iretry > 0) {
-		cdev->private->iretry--;
-		/* Reset internal retry indication. */
-		cdev->private->flags.intretry = 0;
-		ret = cio_start (sch, cdev->private->iccws,
-				 cdev->private->imask);
-		/* We expect an interrupt in case of success or busy
-		 * indication. */
-		if ((ret == 0) || (ret == -EBUSY))
-			return ret;
-	}
-	/* PGID command failed on this path. */
-	CIO_MSG_EVENT(3, "SPID - Device %04x on Subchannel "
-		      "0.%x.%04x, lpm %02X, became 'not operational'\n",
-		      cdev->private->dev_id.devno, sch->schid.ssid,
-		      sch->schid.sch_no, cdev->private->imask);
-	return ret;
+	struct ccw_request *req = &cdev->private->req;
+	struct ccw1 *cp = cdev->private->iccws;
+	int i = 8 - ffs(req->lpm);
+	struct pgid *pgid = &cdev->private->pgid[i];
+
+	pgid->inf.fc	= fn;
+	cp->cmd_code	= CCW_CMD_SET_PGID;
+	cp->cda		= (u32) (addr_t) pgid;
+	cp->count	= sizeof(*pgid);
+	cp->flags	= CCW_FLAG_SLI;
+	req->cp		= cp;
 }
 
 /*
- * Helper function to send a nop ccw down a path.
+ * Perform establish/resign SET PGID on a single path.
  */
-static int __ccw_device_do_nop(struct ccw_device *cdev)
+static void spid_do(struct ccw_device *cdev)
 {
-	struct subchannel *sch;
-	struct ccw1 *ccw;
-	int ret;
-
-	sch = to_subchannel(cdev->dev.parent);
-
-	/* Setup nop channel program. */
-	ccw = cdev->private->iccws;
-	ccw->cmd_code = CCW_CMD_NOOP;
-	ccw->cda = 0;
-	ccw->count = 0;
-	ccw->flags = CCW_FLAG_SLI;
-
-	/* Reset device status. */
-	memset(&cdev->private->irb, 0, sizeof(struct irb));
-
-	/* Try multiple times. */
-	ret = -EACCES;
-	if (cdev->private->iretry > 0) {
-		cdev->private->iretry--;
-		/* Reset internal retry indication. */
-		cdev->private->flags.intretry = 0;
-		ret = cio_start (sch, cdev->private->iccws,
-				 cdev->private->imask);
-		/* We expect an interrupt in case of success or busy
-		 * indication. */
-		if ((ret == 0) || (ret == -EBUSY))
-			return ret;
-	}
-	/* nop command failed on this path. */
-	CIO_MSG_EVENT(3, "NOP - Device %04x on Subchannel "
-		      "0.%x.%04x, lpm %02X, became 'not operational'\n",
-		      cdev->private->dev_id.devno, sch->schid.ssid,
-		      sch->schid.sch_no, cdev->private->imask);
-	return ret;
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+	u8 fn;
+
+	/* Use next available path that is not already in correct state. */
+	req->lpm = lpm_adjust(req->lpm, sch->schib.pmcw.pam & ~sch->vpm);
+	if (!req->lpm)
+		goto out_nopath;
+	/* Channel program setup. */
+	if (req->lpm & sch->opm)
+		fn = SPID_FUNC_ESTABLISH;
+	else
+		fn = SPID_FUNC_RESIGN;
+	if (cdev->private->flags.mpath)
+		fn |= SPID_FUNC_MULTI_PATH;
+	spid_build_cp(cdev, fn);
+	ccw_request_start(cdev);
+	return;
+
+out_nopath:
+	verify_done(cdev, sch->vpm ? 0 : -EACCES);
 }
 
+static void verify_start(struct ccw_device *cdev);
 
 /*
- * Called from interrupt context to check if a valid answer
- * to Set Path Group ID was received.
+ * Process SET PGID request result for a single path.
  */
-static int
-__ccw_device_check_pgid(struct ccw_device *cdev)
+static void spid_callback(struct ccw_device *cdev, void *data, int rc)
 {
-	struct subchannel *sch;
-	struct irb *irb;
-
-	sch = to_subchannel(cdev->dev.parent);
-	irb = &cdev->private->irb;
-	if (irb->scsw.cmd.fctl & (SCSW_FCTL_HALT_FUNC | SCSW_FCTL_CLEAR_FUNC)) {
-		/* Retry Set PGID if requested. */
-		if (cdev->private->flags.intretry) {
-			cdev->private->flags.intretry = 0;
-			return -EAGAIN;
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+
+	switch (rc) {
+	case 0:
+		sch->vpm |= req->lpm & sch->opm;
+		break;
+	case -EACCES:
+		break;
+	case -EOPNOTSUPP:
+		if (cdev->private->flags.mpath) {
+			/* Try without multipathing. */
+			cdev->private->flags.mpath = 0;
+			goto out_restart;
 		}
-		return -ETIME;
+		/* Try without pathgrouping. */
+		cdev->private->flags.pgroup = 0;
+		goto out_restart;
+	default:
+		goto err;
 	}
-	if (irb->esw.esw0.erw.cons) {
-		if (irb->ecw[0] & SNS0_CMD_REJECT)
-			return -EOPNOTSUPP;
-		/* Hmm, whatever happened, try again. */
-		CIO_MSG_EVENT(2, "SPID - device 0.%x.%04x, unit check, "
-			      "cnt %02d, "
-			      "sns : %02X%02X%02X%02X %02X%02X%02X%02X ...\n",
-			      cdev->private->dev_id.ssid,
-			      cdev->private->dev_id.devno,
-			      irb->esw.esw0.erw.scnt,
-			      irb->ecw[0], irb->ecw[1],
-			      irb->ecw[2], irb->ecw[3],
-			      irb->ecw[4], irb->ecw[5],
-			      irb->ecw[6], irb->ecw[7]);
-		return -EAGAIN;
-	}
-	if (irb->scsw.cmd.cc == 3) {
-		CIO_MSG_EVENT(3, "SPID - Device %04x on Subchannel 0.%x.%04x,"
-			      " lpm %02X, became 'not operational'\n",
-			      cdev->private->dev_id.devno, sch->schid.ssid,
-			      sch->schid.sch_no, cdev->private->imask);
-		return -EACCES;
-	}
-	return 0;
+	req->lpm >>= 1;
+	spid_do(cdev);
+	return;
+
+out_restart:
+	verify_start(cdev);
+	return;
+err:
+	verify_done(cdev, rc);
+}
+
+static void spid_start(struct ccw_device *cdev)
+{
+	struct ccw_request *req = &cdev->private->req;
+
+	/* Initialize request data. */
+	memset(req, 0, sizeof(*req));
+	req->timeout	= PGID_TIMEOUT;
+	req->maxretries	= PGID_RETRIES;
+	req->lpm	= 0x80;
+	req->callback	= spid_callback;
+	spid_do(cdev);
+}
+
+static int pgid_cmp(struct pgid *p1, struct pgid *p2)
+{
+	return memcmp((char *) p1 + 1, (char *) p2 + 1,
+		      sizeof(struct pgid) - 1);
 }
 
 /*
- * Called from interrupt context to check the path status after a nop has
- * been send.
+ * Determine pathgroup state from PGID data.
  */
-static int __ccw_device_check_nop(struct ccw_device *cdev)
+static void pgid_analyze(struct ccw_device *cdev, struct pgid **p,
+			 int *mismatch, int *reserved, int *reset)
 {
-	struct subchannel *sch;
-	struct irb *irb;
-
-	sch = to_subchannel(cdev->dev.parent);
-	irb = &cdev->private->irb;
-	if (irb->scsw.cmd.fctl & (SCSW_FCTL_HALT_FUNC | SCSW_FCTL_CLEAR_FUNC)) {
-		/* Retry NOP if requested. */
-		if (cdev->private->flags.intretry) {
-			cdev->private->flags.intretry = 0;
-			return -EAGAIN;
+	struct pgid *pgid = &cdev->private->pgid[0];
+	struct pgid *first = NULL;
+	int lpm;
+	int i;
+
+	*mismatch = 0;
+	*reserved = 0;
+	*reset = 0;
+	for (i = 0, lpm = 0x80; i < 8; i++, pgid++, lpm >>= 1) {
+		if ((cdev->private->pgid_valid_mask & lpm) == 0)
+			continue;
+		if (pgid->inf.ps.state2 == SNID_STATE2_RESVD_ELSE)
+			*reserved = 1;
+		if (pgid->inf.ps.state1 == SNID_STATE1_RESET) {
+			/* A PGID was reset. */
+			*reset = 1;
+			continue;
 		}
-		return -ETIME;
-	}
-	if (irb->scsw.cmd.cc == 3) {
-		CIO_MSG_EVENT(3, "NOP - Device %04x on Subchannel 0.%x.%04x,"
-			      " lpm %02X, became 'not operational'\n",
-			      cdev->private->dev_id.devno, sch->schid.ssid,
-			      sch->schid.sch_no, cdev->private->imask);
-		return -EACCES;
+		if (!first) {
+			first = pgid;
+			continue;
+		}
+		if (pgid_cmp(pgid, first) != 0)
+			*mismatch = 1;
 	}
-	return 0;
+	if (!first)
+		first = &channel_subsystems[0]->global_pgid;
+	*p = first;
 }
 
-static void
-__ccw_device_verify_start(struct ccw_device *cdev)
+static u8 pgid_to_vpm(struct ccw_device *cdev)
 {
-	struct subchannel *sch;
-	__u8 func;
-	int ret;
-
-	sch = to_subchannel(cdev->dev.parent);
-	/* Repeat for all paths. */
-	for (; cdev->private->imask; cdev->private->imask >>= 1,
-				     cdev->private->iretry = 5) {
-		if ((cdev->private->imask & sch->schib.pmcw.pam) == 0)
-			/* Path not available, try next. */
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct pgid *pgid;
+	int i;
+	int lpm;
+	u8 vpm = 0;
+
+	/* Set VPM bits for paths which are already in the target state. */
+	for (i = 0; i < 8; i++) {
+		lpm = 0x80 >> i;
+		if ((cdev->private->pgid_valid_mask & lpm) == 0)
 			continue;
-		if (cdev->private->options.pgroup) {
-			if (sch->opm & cdev->private->imask)
-				func = SPID_FUNC_ESTABLISH;
-			else
-				func = SPID_FUNC_RESIGN;
-			ret = __ccw_device_do_pgid(cdev, func);
-		} else
-			ret = __ccw_device_do_nop(cdev);
-		/* We expect an interrupt in case of success or busy
-		 * indication. */
-		if (ret == 0 || ret == -EBUSY)
-			return;
-		/* Permanent path failure, try next. */
+		pgid = &cdev->private->pgid[i];
+		if (sch->opm & lpm) {
+			if (pgid->inf.ps.state1 != SNID_STATE1_GROUPED)
+				continue;
+		} else {
+			if (pgid->inf.ps.state1 != SNID_STATE1_UNGROUPED)
+				continue;
+		}
+		if (cdev->private->flags.mpath) {
+			if (pgid->inf.ps.state3 != SNID_STATE3_MULTI_PATH)
+				continue;
+		} else {
+			if (pgid->inf.ps.state3 != SNID_STATE3_SINGLE_PATH)
+				continue;
+		}
+		vpm |= lpm;
 	}
-	/* Done with all paths. */
-	ccw_device_verify_done(cdev, (sch->vpm != 0) ? 0 : -ENODEV);
+
+	return vpm;
 }
-		
-/*
- * Got interrupt for Set Path Group ID.
- */
-void
-ccw_device_verify_irq(struct ccw_device *cdev, enum dev_event dev_event)
+
+static void pgid_fill(struct ccw_device *cdev, struct pgid *pgid)
 {
-	struct subchannel *sch;
-	struct irb *irb;
-	int ret;
+	int i;
 
-	irb = (struct irb *) __LC_IRB;
+	for (i = 0; i < 8; i++)
+		memcpy(&cdev->private->pgid[i], pgid, sizeof(struct pgid));
+}
 
-	if (irb->scsw.cmd.stctl ==
-	    (SCSW_STCTL_STATUS_PEND | SCSW_STCTL_ALERT_STATUS)) {
-		if (__ccw_device_should_retry(&irb->scsw))
-			__ccw_device_verify_start(cdev);
-		return;
+/*
+ * Process SENSE PGID data and report result.
+ */
+static void snid_done(struct ccw_device *cdev, int rc)
+{
+	struct ccw_dev_id *id = &cdev->private->dev_id;
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct pgid *pgid;
+	int mismatch = 0;
+	int reserved = 0;
+	int reset = 0;
+
+	if (rc)
+		goto out;
+	pgid_analyze(cdev, &pgid, &mismatch, &reserved, &reset);
+	if (reserved)
+		rc = -EUSERS;
+	else if (mismatch)
+		rc = -EOPNOTSUPP;
+	else {
+		sch->vpm = pgid_to_vpm(cdev);
+		pgid_fill(cdev, pgid);
 	}
-	if (ccw_device_accumulate_and_sense(cdev, irb) != 0)
-		return;
-	sch = to_subchannel(cdev->dev.parent);
-	if (cdev->private->options.pgroup)
-		ret = __ccw_device_check_pgid(cdev);
-	else
-		ret = __ccw_device_check_nop(cdev);
-	memset(&cdev->private->irb, 0, sizeof(struct irb));
-
-	switch (ret) {
-	/* 0, -ETIME, -EAGAIN, -EOPNOTSUPP or -EACCES */
+out:
+	CIO_MSG_EVENT(2, "snid: device 0.%x.%04x: rc=%d pvm=%02x vpm=%02x "
+		      "mism=%d rsvd=%d reset=%d\n", id->ssid, id->devno, rc,
+		      cdev->private->pgid_valid_mask, sch->vpm, mismatch,
+		      reserved, reset);
+	switch (rc) {
 	case 0:
-		/* Path verification ccw finished successfully, update lpm. */
-		sch->vpm |= sch->opm & cdev->private->imask;
-		/* Go on with next path. */
-		cdev->private->imask >>= 1;
-		cdev->private->iretry = 5;
-		__ccw_device_verify_start(cdev);
+		/* Anything left to do? */
+		if (sch->vpm == sch->schib.pmcw.pam) {
+			verify_done(cdev, sch->vpm == 0 ? -EACCES : 0);
+			return;
+		}
+		/* Perform path-grouping. */
+		spid_start(cdev);
 		break;
 	case -EOPNOTSUPP:
-		/*
-		 * One of those strange devices which claim to be able
-		 * to do multipathing but not for Set Path Group ID.
-		 */
-		if (cdev->private->flags.pgid_single)
-			cdev->private->options.pgroup = 0;
-		else
-			cdev->private->flags.pgid_single = 1;
-		/* Retry */
-		sch->vpm = 0;
-		cdev->private->imask = 0x80;
-		cdev->private->iretry = 5;
-		/* fall through. */
-	case -EAGAIN:		/* Try again. */
-		__ccw_device_verify_start(cdev);
-		break;
-	case -ETIME:		/* Set path group id stopped by timeout. */
-		ccw_device_verify_done(cdev, -ETIME);
-		break;
-	case -EACCES:		/* channel is not operational. */
-		cdev->private->imask >>= 1;
-		cdev->private->iretry = 5;
-		__ccw_device_verify_start(cdev);
+		/* Path-grouping not supported. */
+		cdev->private->flags.pgroup = 0;
+		cdev->private->flags.mpath = 0;
+		verify_start(cdev);
 		break;
+	default:
+		verify_done(cdev, rc);
 	}
 }
 
-void
-ccw_device_verify_start(struct ccw_device *cdev)
+/*
+ * Create channel program to perform a SENSE PGID on a single path.
+ */
+static void snid_build_cp(struct ccw_device *cdev)
+{
+	struct ccw_request *req = &cdev->private->req;
+	struct ccw1 *cp = cdev->private->iccws;
+	int i = 8 - ffs(req->lpm);
+
+	/* Channel program setup. */
+	cp->cmd_code	= CCW_CMD_SENSE_PGID;
+	cp->cda		= (u32) (addr_t) &cdev->private->pgid[i];
+	cp->count	= sizeof(struct pgid);
+	cp->flags	= CCW_FLAG_SLI;
+	req->cp		= cp;
+}
+
+/*
+ * Perform SENSE PGID on a single path.
+ */
+static void snid_do(struct ccw_device *cdev)
 {
 	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+
+	/* Adjust lpm if paths are not set in pam. */
+	req->lpm = lpm_adjust(req->lpm, sch->schib.pmcw.pam);
+	if (!req->lpm)
+		goto out_nopath;
+	snid_build_cp(cdev);
+	ccw_request_start(cdev);
+	return;
+
+out_nopath:
+	snid_done(cdev, cdev->private->pgid_valid_mask ? 0 : -EACCES);
+}
 
-	cdev->private->flags.pgid_single = 0;
-	cdev->private->imask = 0x80;
-	cdev->private->iretry = 5;
+/*
+ * Process SENSE PGID request result for single path.
+ */
+static void snid_callback(struct ccw_device *cdev, void *data, int rc)
+{
+	struct ccw_request *req = &cdev->private->req;
+
+	if (rc == 0)
+		cdev->private->pgid_valid_mask |= req->lpm;
+	else if (rc != -EACCES)
+		goto err;
+	req->lpm >>= 1;
+	snid_do(cdev);
+	return;
+
+err:
+	snid_done(cdev, rc);
+}
 
-	/* Start with empty vpm. */
-	sch->vpm = 0;
+/*
+ * Perform path verification.
+ */
+static void verify_start(struct ccw_device *cdev)
+{
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+	struct ccw_dev_id *devid = &cdev->private->dev_id;
 
-	/* Get current pam. */
-	if (cio_update_schib(sch)) {
-		ccw_device_verify_done(cdev, -ENODEV);
-		return;
+	sch->vpm = 0;
+	/* Initialize request data. */
+	memset(req, 0, sizeof(*req));
+	req->timeout	= PGID_TIMEOUT;
+	req->maxretries	= PGID_RETRIES;
+	req->lpm	= 0x80;
+	if (cdev->private->flags.pgroup) {
+		CIO_TRACE_EVENT(4, "snid");
+		CIO_HEX_EVENT(4, devid, sizeof(*devid));
+		req->callback	= snid_callback;
+		snid_do(cdev);
+	} else {
+		CIO_TRACE_EVENT(4, "nop");
+		CIO_HEX_EVENT(4, devid, sizeof(*devid));
+		req->filter	= nop_filter;
+		req->callback	= nop_callback;
+		nop_do(cdev);
 	}
-	/* After 60s path verification is considered to have failed. */
-	ccw_device_set_timeout(cdev, 60*HZ);
-	__ccw_device_verify_start(cdev);
 }
 
-static void
-__ccw_device_disband_start(struct ccw_device *cdev)
+/**
+ * ccw_device_verify_start - perform path verification
+ * @cdev: ccw device
+ *
+ * Perform an I/O on each available channel path to @cdev to determine which
+ * paths are operational. The resulting path mask is stored in sch->vpm.
+ * If device options specify pathgrouping, establish a pathgroup for the
+ * operational paths. When finished, call ccw_device_verify_done with a
+ * return code specifying the result.
+ */
+void ccw_device_verify_start(struct ccw_device *cdev)
 {
-	struct subchannel *sch;
-	int ret;
-
-	sch = to_subchannel(cdev->dev.parent);
-	while (cdev->private->imask != 0) {
-		if (sch->lpm & cdev->private->imask) {
-			ret = __ccw_device_do_pgid(cdev, SPID_FUNC_DISBAND);
-			if (ret == 0)
-				return;
-		}
-		cdev->private->iretry = 5;
-		cdev->private->imask >>= 1;
-	}
-	ccw_device_disband_done(cdev, (sch->lpm != 0) ? 0 : -ENODEV);
+	CIO_TRACE_EVENT(4, "vrfy");
+	CIO_HEX_EVENT(4, &cdev->private->dev_id, sizeof(cdev->private->dev_id));
+	/* Initialize PGID data. */
+	memset(cdev->private->pgid, 0, sizeof(cdev->private->pgid));
+	cdev->private->pgid_valid_mask = 0;
+	/*
+	 * Initialize pathgroup and multipath state with target values.
+	 * They may change in the course of path verification.
+	 */
+	cdev->private->flags.pgroup = cdev->private->options.pgroup;
+	cdev->private->flags.mpath = cdev->private->options.mpath;
+	cdev->private->flags.doverify = 0;
+	verify_start(cdev);
 }
 
 /*
- * Got interrupt for Unset Path Group ID.
+ * Process disband SET PGID request result.
  */
-void
-ccw_device_disband_irq(struct ccw_device *cdev, enum dev_event dev_event)
+static void disband_callback(struct ccw_device *cdev, void *data, int rc)
 {
-	struct subchannel *sch;
-	struct irb *irb;
-	int ret;
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_dev_id *id = &cdev->private->dev_id;
+
+	if (rc)
+		goto out;
+	/* Ensure consistent multipathing state at device and channel. */
+	cdev->private->flags.mpath = 0;
+	if (sch->config.mp) {
+		sch->config.mp = 0;
+		rc = cio_commit_config(sch);
+	}
+out:
+	CIO_MSG_EVENT(0, "disb: device 0.%x.%04x: rc=%d\n", id->ssid, id->devno,
+		      rc);
+	ccw_device_disband_done(cdev, rc);
+}
 
-	irb = (struct irb *) __LC_IRB;
+/**
+ * ccw_device_disband_start - disband pathgroup
+ * @cdev: ccw device
+ *
+ * Execute a SET PGID channel program on @cdev to disband a previously
+ * established pathgroup. When finished, call ccw_device_disband_done with
+ * a return code specifying the result.
+ */
+void ccw_device_disband_start(struct ccw_device *cdev)
+{
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+	u8 fn;
+
+	CIO_TRACE_EVENT(4, "disb");
+	CIO_HEX_EVENT(4, &cdev->private->dev_id, sizeof(cdev->private->dev_id));
+	/* Request setup. */
+	memset(req, 0, sizeof(*req));
+	req->timeout	= PGID_TIMEOUT;
+	req->maxretries	= PGID_RETRIES;
+	req->lpm	= sch->schib.pmcw.pam & sch->opm;
+	req->callback	= disband_callback;
+	fn = SPID_FUNC_DISBAND;
+	if (cdev->private->flags.mpath)
+		fn |= SPID_FUNC_MULTI_PATH;
+	spid_build_cp(cdev, fn);
+	ccw_request_start(cdev);
+}
 
-	if (irb->scsw.cmd.stctl ==
-	    (SCSW_STCTL_STATUS_PEND | SCSW_STCTL_ALERT_STATUS)) {
-		if (__ccw_device_should_retry(&irb->scsw))
-			__ccw_device_disband_start(cdev);
-		return;
-	}
-	if (ccw_device_accumulate_and_sense(cdev, irb) != 0)
-		return;
-	sch = to_subchannel(cdev->dev.parent);
-	ret = __ccw_device_check_pgid(cdev);
-	memset(&cdev->private->irb, 0, sizeof(struct irb));
-	switch (ret) {
-	/* 0, -ETIME, -EAGAIN, -EOPNOTSUPP or -EACCES */
-	case 0:			/* disband successful. */
-		ccw_device_disband_done(cdev, ret);
-		break;
-	case -EOPNOTSUPP:
-		/*
-		 * One of those strange devices which claim to be able
-		 * to do multipathing but not for Unset Path Group ID.
-		 */
-		cdev->private->flags.pgid_single = 1;
-		/* fall through. */
-	case -EAGAIN:		/* Try again. */
-		__ccw_device_disband_start(cdev);
-		break;
-	case -ETIME:		/* Set path group id stopped by timeout. */
-		ccw_device_disband_done(cdev, -ETIME);
-		break;
-	case -EACCES:		/* channel is not operational. */
-		cdev->private->imask >>= 1;
-		cdev->private->iretry = 5;
-		__ccw_device_disband_start(cdev);
-		break;
-	}
+static void stlck_build_cp(struct ccw_device *cdev, void *buf1, void *buf2)
+{
+	struct ccw_request *req = &cdev->private->req;
+	struct ccw1 *cp = cdev->private->iccws;
+
+	cp[0].cmd_code = CCW_CMD_STLCK;
+	cp[0].cda = (u32) (addr_t) buf1;
+	cp[0].count = 32;
+	cp[0].flags = CCW_FLAG_CC;
+	cp[1].cmd_code = CCW_CMD_RELEASE;
+	cp[1].cda = (u32) (addr_t) buf2;
+	cp[1].count = 32;
+	cp[1].flags = 0;
+	req->cp = cp;
 }
 
-void
-ccw_device_disband_start(struct ccw_device *cdev)
+static void stlck_callback(struct ccw_device *cdev, void *data, int rc)
 {
-	/* After 60s disbanding is considered to have failed. */
-	ccw_device_set_timeout(cdev, 60*HZ);
+	ccw_device_stlck_done(cdev, data, rc);
+}
 
-	cdev->private->flags.pgid_single = 0;
-	cdev->private->iretry = 5;
-	cdev->private->imask = 0x80;
-	__ccw_device_disband_start(cdev);
+/**
+ * ccw_device_stlck_start - perform unconditional release
+ * @cdev: ccw device
+ * @data: data pointer to be passed to ccw_device_stlck_done
+ * @buf1: data pointer used in channel program
+ * @buf2: data pointer used in channel program
+ *
+ * Execute a channel program on @cdev to release an existing PGID reservation.
+ * When finished, call ccw_device_stlck_done with a return code specifying the
+ * result.
+ */
+void ccw_device_stlck_start(struct ccw_device *cdev, void *data, void *buf1,
+			    void *buf2)
+{
+	struct subchannel *sch = to_subchannel(cdev->dev.parent);
+	struct ccw_request *req = &cdev->private->req;
+
+	CIO_TRACE_EVENT(4, "stlck");
+	CIO_HEX_EVENT(4, &cdev->private->dev_id, sizeof(cdev->private->dev_id));
+	/* Request setup. */
+	memset(req, 0, sizeof(*req));
+	req->timeout	= PGID_TIMEOUT;
+	req->maxretries	= PGID_RETRIES;
+	req->lpm	= sch->schib.pmcw.pam & sch->opm;
+	req->data	= data;
+	req->callback	= stlck_callback;
+	stlck_build_cp(cdev, buf1, buf2);
+	ccw_request_start(cdev);
 }
+

drivers/s390/cio/device_status.c

@@ -336,9 +336,6 @@ ccw_device_do_sense(struct ccw_device *cdev, struct irb *irb)
 	sense_ccw->count = SENSE_MAX_COUNT;
 	sense_ccw->flags = CCW_FLAG_SLI;
 
-	/* Reset internal retry indication. */
-	cdev->private->flags.intretry = 0;
-
 	rc = cio_start(sch, sense_ccw, 0xff);
 	if (rc == -ENODEV || rc == -EACCES)
 		dev_fsm_event(cdev, DEV_EVENT_VERIFY);

drivers/s390/cio/io_sch.h

@@ -1,7 +1,10 @@
 #ifndef S390_IO_SCH_H
 #define S390_IO_SCH_H
 
+#include <linux/types.h>
 #include <asm/schid.h>
+#include <asm/ccwdev.h>
+#include "css.h"
 
 /*
  * command-mode operation request block
@@ -67,6 +70,52 @@ struct io_subchannel_private {
 
 #define MAX_CIWS 8
 
+/*
+ * Possible status values for a CCW request's I/O.
+ */
+enum io_status {
+	IO_DONE,
+	IO_RUNNING,
+	IO_STATUS_ERROR,
+	IO_PATH_ERROR,
+	IO_REJECTED,
+	IO_KILLED
+};
+
+/**
+ * ccw_request - Internal CCW request.
+ * @cp: channel program to start
+ * @timeout: maximum allowable time in jiffies between start I/O and interrupt
+ * @maxretries: number of retries per I/O operation and path
+ * @lpm: mask of paths to use
+ * @check: optional callback that determines if results are final
+ * @filter: optional callback to adjust request status based on IRB data
+ * @callback: final callback
+ * @data: user-defined pointer passed to all callbacks
+ * @mask: current path mask
+ * @retries: current number of retries
+ * @drc: delayed return code
+ * @cancel: non-zero if request was cancelled
+ * @done: non-zero if request was finished
+ */
+struct ccw_request {
+	struct ccw1 *cp;
+	unsigned long timeout;
+	u16 maxretries;
+	u8 lpm;
+	int (*check)(struct ccw_device *, void *);
+	enum io_status (*filter)(struct ccw_device *, void *, struct irb *,
+				 enum io_status);
+	void (*callback)(struct ccw_device *, void *, int);
+	void *data;
+	/* These fields are used internally. */
+	u16 mask;
+	u16 retries;
+	int drc;
+	int cancel:1;
+	int done:1;
+} __attribute__((packed));
+
 /*
  * sense-id response buffer layout
  */
@@ -82,32 +131,43 @@ struct senseid {
 	struct ciw ciw[MAX_CIWS];	/* variable # of CIWs */
 }  __attribute__ ((packed, aligned(4)));
 
+enum cdev_todo {
+	CDEV_TODO_NOTHING,
+	CDEV_TODO_ENABLE_CMF,
+	CDEV_TODO_REBIND,
+	CDEV_TODO_REGISTER,
+	CDEV_TODO_UNREG,
+	CDEV_TODO_UNREG_EVAL,
+};
+
 struct ccw_device_private {
 	struct ccw_device *cdev;
 	struct subchannel *sch;
 	int state;		/* device state */
 	atomic_t onoff;
-	unsigned long registered;
 	struct ccw_dev_id dev_id;	/* device id */
 	struct subchannel_id schid;	/* subchannel number */
-	u8 imask;		/* lpm mask for SNID/SID/SPGID */
-	int iretry;		/* retry counter SNID/SID/SPGID */
+	struct ccw_request req;		/* internal I/O request */
+	int iretry;
+	u8 pgid_valid_mask;		/* mask of valid PGIDs */
 	struct {
 		unsigned int fast:1;	/* post with "channel end" */
 		unsigned int repall:1;	/* report every interrupt status */
 		unsigned int pgroup:1;	/* do path grouping */
 		unsigned int force:1;	/* allow forced online */
+		unsigned int mpath:1;	/* do multipathing */
 	} __attribute__ ((packed)) options;
 	struct {
-		unsigned int pgid_single:1; /* use single path for Set PGID */
 		unsigned int esid:1;	    /* Ext. SenseID supported by HW */
 		unsigned int dosense:1;	    /* delayed SENSE required */
 		unsigned int doverify:1;    /* delayed path verification */
 		unsigned int donotify:1;    /* call notify function */
 		unsigned int recog_done:1;  /* dev. recog. complete */
 		unsigned int fake_irb:1;    /* deliver faked irb */
-		unsigned int intretry:1;    /* retry internal operation */
 		unsigned int resuming:1;    /* recognition while resume */
+		unsigned int pgroup:1;	    /* pathgroup is set up */
+		unsigned int mpath:1;	    /* multipathing is set up */
+		unsigned int initialized:1; /* set if initial reference held */
 	} __attribute__((packed)) flags;
 	unsigned long intparm;	/* user interruption parameter */
 	struct qdio_irq *qdio_data;
@@ -115,7 +175,8 @@ struct ccw_device_private {
 	struct senseid senseid;	/* SenseID info */
 	struct pgid pgid[8];	/* path group IDs per chpid*/
 	struct ccw1 iccws[2];	/* ccws for SNID/SID/SPGID commands */
-	struct work_struct kick_work;
+	struct work_struct todo_work;
+	enum cdev_todo todo;
 	wait_queue_head_t wait_q;
 	struct timer_list timer;
 	void *cmb;			/* measurement information */

drivers/s390/crypto/ap_bus.c

@@ -102,6 +102,7 @@ static atomic_t ap_poll_requests = ATOMIC_INIT(0);
 static DECLARE_WAIT_QUEUE_HEAD(ap_poll_wait);
 static struct task_struct *ap_poll_kthread = NULL;
 static DEFINE_MUTEX(ap_poll_thread_mutex);
+static DEFINE_SPINLOCK(ap_poll_timer_lock);
 static void *ap_interrupt_indicator;
 static struct hrtimer ap_poll_timer;
 /* In LPAR poll with 4kHz frequency. Poll every 250000 nanoseconds.
@@ -282,6 +283,7 @@ static int ap_queue_enable_interruption(ap_qid_t qid, void *ind)
  * @psmid: The program supplied message identifier
  * @msg: The message text
  * @length: The message length
+ * @special: Special Bit
  *
  * Returns AP queue status structure.
  * Condition code 1 on NQAP can't happen because the L bit is 1.
@@ -289,7 +291,8 @@ static int ap_queue_enable_interruption(ap_qid_t qid, void *ind)
  * because a segment boundary was reached. The NQAP is repeated.
  */
 static inline struct ap_queue_status
-__ap_send(ap_qid_t qid, unsigned long long psmid, void *msg, size_t length)
+__ap_send(ap_qid_t qid, unsigned long long psmid, void *msg, size_t length,
+	  unsigned int special)
 {
 	typedef struct { char _[length]; } msgblock;
 	register unsigned long reg0 asm ("0") = qid | 0x40000000UL;
@@ -299,6 +302,9 @@ __ap_send(ap_qid_t qid, unsigned long long psmid, void *msg, size_t length)
 	register unsigned long reg4 asm ("4") = (unsigned int) (psmid >> 32);
 	register unsigned long reg5 asm ("5") = (unsigned int) psmid;
 
+	if (special == 1)
+		reg0 |= 0x400000UL;
+
 	asm volatile (
 		"0: .long 0xb2ad0042\n"		/* DQAP */
 		"   brc   2,0b"
@@ -312,13 +318,15 @@ int ap_send(ap_qid_t qid, unsigned long long psmid, void *msg, size_t length)
 {
 	struct ap_queue_status status;
 
-	status = __ap_send(qid, psmid, msg, length);
+	status = __ap_send(qid, psmid, msg, length, 0);
 	switch (status.response_code) {
 	case AP_RESPONSE_NORMAL:
 		return 0;
 	case AP_RESPONSE_Q_FULL:
 	case AP_RESPONSE_RESET_IN_PROGRESS:
 		return -EBUSY;
+	case AP_RESPONSE_REQ_FAC_NOT_INST:
+		return -EINVAL;
 	default:	/* Device is gone. */
 		return -ENODEV;
 	}
@@ -1008,7 +1016,7 @@ static int ap_probe_device_type(struct ap_device *ap_dev)
 	}
 
 	status = __ap_send(ap_dev->qid, 0x0102030405060708ULL,
-			   msg, sizeof(msg));
+			   msg, sizeof(msg), 0);
 	if (status.response_code != AP_RESPONSE_NORMAL) {
 		rc = -ENODEV;
 		goto out_free;
@@ -1163,16 +1171,19 @@ ap_config_timeout(unsigned long ptr)
 static inline void ap_schedule_poll_timer(void)
 {
 	ktime_t hr_time;
+
+	spin_lock_bh(&ap_poll_timer_lock);
 	if (ap_using_interrupts() || ap_suspend_flag)
-		return;
+		goto out;
 	if (hrtimer_is_queued(&ap_poll_timer))
-		return;
+		goto out;
 	if (ktime_to_ns(hrtimer_expires_remaining(&ap_poll_timer)) <= 0) {
 		hr_time = ktime_set(0, poll_timeout);
 		hrtimer_forward_now(&ap_poll_timer, hr_time);
 		hrtimer_restart(&ap_poll_timer);
 	}
-	return;
+out:
+	spin_unlock_bh(&ap_poll_timer_lock);
 }
 
 /**
@@ -1243,7 +1254,7 @@ static int ap_poll_write(struct ap_device *ap_dev, unsigned long *flags)
 	/* Start the next request on the queue. */
 	ap_msg = list_entry(ap_dev->requestq.next, struct ap_message, list);
 	status = __ap_send(ap_dev->qid, ap_msg->psmid,
-			   ap_msg->message, ap_msg->length);
+			   ap_msg->message, ap_msg->length, ap_msg->special);
 	switch (status.response_code) {
 	case AP_RESPONSE_NORMAL:
 		atomic_inc(&ap_poll_requests);
@@ -1261,6 +1272,7 @@ static int ap_poll_write(struct ap_device *ap_dev, unsigned long *flags)
 		*flags |= 2;
 		break;
 	case AP_RESPONSE_MESSAGE_TOO_BIG:
+	case AP_RESPONSE_REQ_FAC_NOT_INST:
 		return -EINVAL;
 	default:
 		return -ENODEV;
@@ -1302,7 +1314,8 @@ static int __ap_queue_message(struct ap_device *ap_dev, struct ap_message *ap_ms
 	if (list_empty(&ap_dev->requestq) &&
 	    ap_dev->queue_count < ap_dev->queue_depth) {
 		status = __ap_send(ap_dev->qid, ap_msg->psmid,
-				   ap_msg->message, ap_msg->length);
+				   ap_msg->message, ap_msg->length,
+				   ap_msg->special);
 		switch (status.response_code) {
 		case AP_RESPONSE_NORMAL:
 			list_add_tail(&ap_msg->list, &ap_dev->pendingq);
@@ -1317,6 +1330,7 @@ static int __ap_queue_message(struct ap_device *ap_dev, struct ap_message *ap_ms
 			ap_dev->requestq_count++;
 			ap_dev->total_request_count++;
 			return -EBUSY;
+		case AP_RESPONSE_REQ_FAC_NOT_INST:
 		case AP_RESPONSE_MESSAGE_TOO_BIG:
 			ap_dev->drv->receive(ap_dev, ap_msg, ERR_PTR(-EINVAL));
 			return -EINVAL;
@@ -1658,6 +1672,7 @@ int __init ap_module_init(void)
 	 */
 	if (MACHINE_IS_VM)
 		poll_timeout = 1500000;
+	spin_lock_init(&ap_poll_timer_lock);
 	hrtimer_init(&ap_poll_timer, CLOCK_MONOTONIC, HRTIMER_MODE_ABS);
 	ap_poll_timer.function = ap_poll_timeout;
 

drivers/s390/crypto/ap_bus.h

@@ -87,6 +87,7 @@ struct ap_queue_status {
 #define AP_RESPONSE_INDEX_TOO_BIG	0x11
 #define AP_RESPONSE_NO_FIRST_PART	0x13
 #define AP_RESPONSE_MESSAGE_TOO_BIG	0x15
+#define AP_RESPONSE_REQ_FAC_NOT_INST	0x16
 
 /*
  * Known device types
@@ -96,8 +97,8 @@ struct ap_queue_status {
 #define AP_DEVICE_TYPE_PCIXCC	5
 #define AP_DEVICE_TYPE_CEX2A	6
 #define AP_DEVICE_TYPE_CEX2C	7
-#define AP_DEVICE_TYPE_CEX2A2	8
-#define AP_DEVICE_TYPE_CEX2C2	9
+#define AP_DEVICE_TYPE_CEX3A	8
+#define AP_DEVICE_TYPE_CEX3C	9
 
 /*
  * AP reset flag states
@@ -161,12 +162,25 @@ struct ap_message {
 	size_t length;			/* Message length. */
 
 	void *private;			/* ap driver private pointer. */
+	unsigned int special:1;		/* Used for special commands. */
 };
 
 #define AP_DEVICE(dt)					\
 	.dev_type=(dt),					\
 	.match_flags=AP_DEVICE_ID_MATCH_DEVICE_TYPE,
 
+/**
+ * ap_init_message() - Initialize ap_message.
+ * Initialize a message before using. Otherwise this might result in
+ * unexpected behaviour.
+ */
+static inline void ap_init_message(struct ap_message *ap_msg)
+{
+	ap_msg->psmid = 0;
+	ap_msg->length = 0;
+	ap_msg->special = 0;
+}
+
 /*
  * Note: don't use ap_send/ap_recv after using ap_queue_message
  * for the first time. Otherwise the ap message queue will get

drivers/s390/crypto/zcrypt_api.c

@@ -299,9 +299,7 @@ static ssize_t zcrypt_write(struct file *filp, const char __user *buf,
  */
 static int zcrypt_open(struct inode *inode, struct file *filp)
 {
-	lock_kernel();
 	atomic_inc(&zcrypt_open_count);
-	unlock_kernel();
 	return 0;
 }
 
@@ -1009,6 +1007,10 @@ static int zcrypt_status_read(char *resp_buff, char **start, off_t offset,
 		       zcrypt_count_type(ZCRYPT_CEX2C));
 	len += sprintf(resp_buff + len, "CEX2A count: %d\n",
 		       zcrypt_count_type(ZCRYPT_CEX2A));
+	len += sprintf(resp_buff + len, "CEX3C count: %d\n",
+		       zcrypt_count_type(ZCRYPT_CEX3C));
+	len += sprintf(resp_buff + len, "CEX3A count: %d\n",
+		       zcrypt_count_type(ZCRYPT_CEX3A));
 	len += sprintf(resp_buff + len, "requestq count: %d\n",
 		       zcrypt_requestq_count());
 	len += sprintf(resp_buff + len, "pendingq count: %d\n",
@@ -1017,7 +1019,7 @@ static int zcrypt_status_read(char *resp_buff, char **start, off_t offset,
 		       atomic_read(&zcrypt_open_count));
 	zcrypt_status_mask(workarea);
 	len += sprinthx("Online devices: 1=PCICA 2=PCICC 3=PCIXCC(MCL2) "
-			"4=PCIXCC(MCL3) 5=CEX2C 6=CEX2A",
+			"4=PCIXCC(MCL3) 5=CEX2C 6=CEX2A 7=CEX3C 8=CEX3A",
 			resp_buff+len, workarea, AP_DEVICES);
 	zcrypt_qdepth_mask(workarea);
 	len += sprinthx("Waiting work element counts",
@@ -1095,8 +1097,9 @@ static int zcrypt_status_write(struct file *file, const char __user *buffer,
 		 * '0' for no device, '1' for PCICA, '2' for PCICC,
 		 * '3' for PCIXCC_MCL2, '4' for PCIXCC_MCL3,
 		 * '5' for CEX2C and '6' for CEX2A'
+		 * '7' for CEX3C and '8' for CEX3A
 		 */
-		if (*ptr >= '0' && *ptr <= '6')
+		if (*ptr >= '0' && *ptr <= '8')
 			j++;
 		else if (*ptr == 'd' || *ptr == 'D')
 			zcrypt_disable_card(j++);

drivers/s390/crypto/zcrypt_api.h

@@ -71,6 +71,8 @@ struct ica_z90_status {
 #define ZCRYPT_PCIXCC_MCL3	4
 #define ZCRYPT_CEX2C		5
 #define ZCRYPT_CEX2A		6
+#define ZCRYPT_CEX3C		7
+#define ZCRYPT_CEX3A		8
 
 /**
  * Large random numbers are pulled in 4096 byte chunks from the crypto cards

drivers/s390/crypto/zcrypt_cex2a.c

@@ -39,17 +39,24 @@
 
 #define CEX2A_MIN_MOD_SIZE	  1	/*    8 bits	*/
 #define CEX2A_MAX_MOD_SIZE	256	/* 2048 bits	*/
+#define CEX3A_MIN_MOD_SIZE	CEX2A_MIN_MOD_SIZE
+#define CEX3A_MAX_MOD_SIZE	CEX2A_MAX_MOD_SIZE
 
 #define CEX2A_SPEED_RATING	970
+#define CEX3A_SPEED_RATING	900 /* Fixme: Needs finetuning */
 
 #define CEX2A_MAX_MESSAGE_SIZE	0x390	/* sizeof(struct type50_crb2_msg)    */
 #define CEX2A_MAX_RESPONSE_SIZE 0x110	/* max outputdatalength + type80_hdr */
 
+#define CEX3A_MAX_MESSAGE_SIZE	CEX2A_MAX_MESSAGE_SIZE
+#define CEX3A_MAX_RESPONSE_SIZE	CEX2A_MAX_RESPONSE_SIZE
+
 #define CEX2A_CLEANUP_TIME	(15*HZ)
+#define CEX3A_CLEANUP_TIME	CEX2A_CLEANUP_TIME
 
 static struct ap_device_id zcrypt_cex2a_ids[] = {
 	{ AP_DEVICE(AP_DEVICE_TYPE_CEX2A) },
-	{ AP_DEVICE(AP_DEVICE_TYPE_CEX2A2) },
+	{ AP_DEVICE(AP_DEVICE_TYPE_CEX3A) },
 	{ /* end of list */ },
 };
 
@@ -298,6 +305,7 @@ static long zcrypt_cex2a_modexpo(struct zcrypt_device *zdev,
 	struct completion work;
 	int rc;
 
+	ap_init_message(&ap_msg);
 	ap_msg.message = kmalloc(CEX2A_MAX_MESSAGE_SIZE, GFP_KERNEL);
 	if (!ap_msg.message)
 		return -ENOMEM;
@@ -335,6 +343,7 @@ static long zcrypt_cex2a_modexpo_crt(struct zcrypt_device *zdev,
 	struct completion work;
 	int rc;
 
+	ap_init_message(&ap_msg);
 	ap_msg.message = kmalloc(CEX2A_MAX_MESSAGE_SIZE, GFP_KERNEL);
 	if (!ap_msg.message)
 		return -ENOMEM;
@@ -373,31 +382,45 @@ static struct zcrypt_ops zcrypt_cex2a_ops = {
  */
 static int zcrypt_cex2a_probe(struct ap_device *ap_dev)
 {
-	struct zcrypt_device *zdev;
-	int rc;
-
-	zdev = zcrypt_device_alloc(CEX2A_MAX_RESPONSE_SIZE);
-	if (!zdev)
-		return -ENOMEM;
-	zdev->ap_dev = ap_dev;
-	zdev->ops = &zcrypt_cex2a_ops;
-	zdev->online = 1;
-	zdev->user_space_type = ZCRYPT_CEX2A;
-	zdev->type_string = "CEX2A";
-	zdev->min_mod_size = CEX2A_MIN_MOD_SIZE;
-	zdev->max_mod_size = CEX2A_MAX_MOD_SIZE;
-	zdev->short_crt = 1;
-	zdev->speed_rating = CEX2A_SPEED_RATING;
-	ap_dev->reply = &zdev->reply;
-	ap_dev->private = zdev;
-	rc = zcrypt_device_register(zdev);
-	if (rc)
-		goto out_free;
-	return 0;
-
-out_free:
-	ap_dev->private = NULL;
-	zcrypt_device_free(zdev);
+	struct zcrypt_device *zdev = NULL;
+	int rc = 0;
+
+	switch (ap_dev->device_type) {
+	case AP_DEVICE_TYPE_CEX2A:
+		zdev = zcrypt_device_alloc(CEX2A_MAX_RESPONSE_SIZE);
+		if (!zdev)
+			return -ENOMEM;
+		zdev->user_space_type = ZCRYPT_CEX2A;
+		zdev->type_string = "CEX2A";
+		zdev->min_mod_size = CEX2A_MIN_MOD_SIZE;
+		zdev->max_mod_size = CEX2A_MAX_MOD_SIZE;
+		zdev->short_crt = 1;
+		zdev->speed_rating = CEX2A_SPEED_RATING;
+		break;
+	case AP_DEVICE_TYPE_CEX3A:
+		zdev = zcrypt_device_alloc(CEX3A_MAX_RESPONSE_SIZE);
+		if (!zdev)
+			return -ENOMEM;
+		zdev->user_space_type = ZCRYPT_CEX3A;
+		zdev->type_string = "CEX3A";
+		zdev->min_mod_size = CEX3A_MIN_MOD_SIZE;
+		zdev->max_mod_size = CEX3A_MAX_MOD_SIZE;
+		zdev->short_crt = 1;
+		zdev->speed_rating = CEX3A_SPEED_RATING;
+		break;
+	}
+	if (zdev != NULL) {
+		zdev->ap_dev = ap_dev;
+		zdev->ops = &zcrypt_cex2a_ops;
+		zdev->online = 1;
+		ap_dev->reply = &zdev->reply;
+		ap_dev->private = zdev;
+		rc = zcrypt_device_register(zdev);
+	}
+	if (rc) {
+		ap_dev->private = NULL;
+		zcrypt_device_free(zdev);
+	}
 	return rc;
 }
 

drivers/s390/crypto/zcrypt_pcica.c

@@ -281,6 +281,7 @@ static long zcrypt_pcica_modexpo(struct zcrypt_device *zdev,
 	struct completion work;
 	int rc;
 
+	ap_init_message(&ap_msg);
 	ap_msg.message = kmalloc(PCICA_MAX_MESSAGE_SIZE, GFP_KERNEL);
 	if (!ap_msg.message)
 		return -ENOMEM;
@@ -318,6 +319,7 @@ static long zcrypt_pcica_modexpo_crt(struct zcrypt_device *zdev,
 	struct completion work;
 	int rc;
 
+	ap_init_message(&ap_msg);
 	ap_msg.message = kmalloc(PCICA_MAX_MESSAGE_SIZE, GFP_KERNEL);
 	if (!ap_msg.message)
 		return -ENOMEM;

drivers/s390/crypto/zcrypt_pcicc.c

@@ -483,6 +483,7 @@ static long zcrypt_pcicc_modexpo(struct zcrypt_device *zdev,
 	struct completion work;
 	int rc;
 
+	ap_init_message(&ap_msg);
 	ap_msg.message = (void *) get_zeroed_page(GFP_KERNEL);
 	if (!ap_msg.message)
 		return -ENOMEM;
@@ -521,6 +522,7 @@ static long zcrypt_pcicc_modexpo_crt(struct zcrypt_device *zdev,
 	struct completion work;
 	int rc;
 
+	ap_init_message(&ap_msg);
 	ap_msg.message = (void *) get_zeroed_page(GFP_KERNEL);
 	if (!ap_msg.message)
 		return -ENOMEM;

drivers/s390/crypto/zcrypt_pcixcc.c

@@ -43,10 +43,13 @@
 #define PCIXCC_MIN_MOD_SIZE	 16	/*  128 bits	*/
 #define PCIXCC_MIN_MOD_SIZE_OLD	 64	/*  512 bits	*/
 #define PCIXCC_MAX_MOD_SIZE	256	/* 2048 bits	*/
+#define CEX3C_MIN_MOD_SIZE	PCIXCC_MIN_MOD_SIZE
+#define CEX3C_MAX_MOD_SIZE	PCIXCC_MAX_MOD_SIZE
 
-#define PCIXCC_MCL2_SPEED_RATING	7870	/* FIXME: needs finetuning */
+#define PCIXCC_MCL2_SPEED_RATING	7870
 #define PCIXCC_MCL3_SPEED_RATING	7870
-#define CEX2C_SPEED_RATING		8540
+#define CEX2C_SPEED_RATING		7000
+#define CEX3C_SPEED_RATING		6500	/* FIXME: needs finetuning */
 
 #define PCIXCC_MAX_ICA_MESSAGE_SIZE 0x77c  /* max size type6 v2 crt message */
 #define PCIXCC_MAX_ICA_RESPONSE_SIZE 0x77c /* max size type86 v2 reply	    */
@@ -72,7 +75,7 @@ struct response_type {
 static struct ap_device_id zcrypt_pcixcc_ids[] = {
 	{ AP_DEVICE(AP_DEVICE_TYPE_PCIXCC) },
 	{ AP_DEVICE(AP_DEVICE_TYPE_CEX2C) },
-	{ AP_DEVICE(AP_DEVICE_TYPE_CEX2C2) },
+	{ AP_DEVICE(AP_DEVICE_TYPE_CEX3C) },
 	{ /* end of list */ },
 };
 
@@ -326,6 +329,11 @@ static int XCRB_msg_to_type6CPRB_msgX(struct zcrypt_device *zdev,
 	function_code = ((unsigned char *)&msg->cprbx) + msg->cprbx.cprb_len;
 	memcpy(msg->hdr.function_code, function_code, sizeof(msg->hdr.function_code));
 
+	if (memcmp(function_code, "US", 2) == 0)
+		ap_msg->special = 1;
+	else
+		ap_msg->special = 0;
+
 	/* copy data block */
 	if (xcRB->request_data_length &&
 	    copy_from_user(req_data, xcRB->request_data_address,
@@ -688,6 +696,7 @@ static long zcrypt_pcixcc_modexpo(struct zcrypt_device *zdev,
 	};
 	int rc;
 
+	ap_init_message(&ap_msg);
 	ap_msg.message = (void *) get_zeroed_page(GFP_KERNEL);
 	if (!ap_msg.message)
 		return -ENOMEM;
@@ -727,6 +736,7 @@ static long zcrypt_pcixcc_modexpo_crt(struct zcrypt_device *zdev,
 	};
 	int rc;
 
+	ap_init_message(&ap_msg);
 	ap_msg.message = (void *) get_zeroed_page(GFP_KERNEL);
 	if (!ap_msg.message)
 		return -ENOMEM;
@@ -766,6 +776,7 @@ static long zcrypt_pcixcc_send_cprb(struct zcrypt_device *zdev,
 	};
 	int rc;
 
+	ap_init_message(&ap_msg);
 	ap_msg.message = kmalloc(PCIXCC_MAX_XCRB_MESSAGE_SIZE, GFP_KERNEL);
 	if (!ap_msg.message)
 		return -ENOMEM;
@@ -805,6 +816,7 @@ static long zcrypt_pcixcc_rng(struct zcrypt_device *zdev,
 	};
 	int rc;
 
+	ap_init_message(&ap_msg);
 	ap_msg.message = kmalloc(PCIXCC_MAX_XCRB_MESSAGE_SIZE, GFP_KERNEL);
 	if (!ap_msg.message)
 		return -ENOMEM;
@@ -972,6 +984,7 @@ static int zcrypt_pcixcc_rng_supported(struct ap_device *ap_dev)
 	} __attribute__((packed)) *reply;
 	int rc, i;
 
+	ap_init_message(&ap_msg);
 	ap_msg.message = (void *) get_zeroed_page(GFP_KERNEL);
 	if (!ap_msg.message)
 		return -ENOMEM;
@@ -1016,14 +1029,15 @@ out_free:
 static int zcrypt_pcixcc_probe(struct ap_device *ap_dev)
 {
 	struct zcrypt_device *zdev;
-	int rc;
+	int rc = 0;
 
 	zdev = zcrypt_device_alloc(PCIXCC_MAX_RESPONSE_SIZE);
 	if (!zdev)
 		return -ENOMEM;
 	zdev->ap_dev = ap_dev;
 	zdev->online = 1;
-	if (ap_dev->device_type == AP_DEVICE_TYPE_PCIXCC) {
+	switch (ap_dev->device_type) {
+	case AP_DEVICE_TYPE_PCIXCC:
 		rc = zcrypt_pcixcc_mcl(ap_dev);
 		if (rc < 0) {
 			zcrypt_device_free(zdev);
@@ -1041,13 +1055,25 @@ static int zcrypt_pcixcc_probe(struct ap_device *ap_dev)
 			zdev->min_mod_size = PCIXCC_MIN_MOD_SIZE;
 			zdev->max_mod_size = PCIXCC_MAX_MOD_SIZE;
 		}
-	} else {
+		break;
+	case AP_DEVICE_TYPE_CEX2C:
 		zdev->user_space_type = ZCRYPT_CEX2C;
 		zdev->type_string = "CEX2C";
 		zdev->speed_rating = CEX2C_SPEED_RATING;
 		zdev->min_mod_size = PCIXCC_MIN_MOD_SIZE;
 		zdev->max_mod_size = PCIXCC_MAX_MOD_SIZE;
+		break;
+	case AP_DEVICE_TYPE_CEX3C:
+		zdev->user_space_type = ZCRYPT_CEX3C;
+		zdev->type_string = "CEX3C";
+		zdev->speed_rating = CEX3C_SPEED_RATING;
+		zdev->min_mod_size = CEX3C_MIN_MOD_SIZE;
+		zdev->max_mod_size = CEX3C_MAX_MOD_SIZE;
+		break;
+	default:
+		goto out_free;
 	}
+
 	rc = zcrypt_pcixcc_rng_supported(ap_dev);
 	if (rc < 0) {
 		zcrypt_device_free(zdev);