Merge branches 'sh/hwblk', 'sh/cpuidle' and 'sh/stable-updates'

arch/sh/boards/mach-se/7724/setup.c

@@ -19,6 +19,7 @@
 #include <linux/smc91x.h>
 #include <linux/gpio.h>
 #include <linux/input.h>
+#include <linux/usb/r8a66597.h>
 #include <video/sh_mobile_lcdc.h>
 #include <media/sh_mobile_ceu.h>
 #include <asm/io.h>
@@ -302,6 +303,34 @@ static struct platform_device sh_eth_device = {
 	.resource = sh_eth_resources,
 };
 
+static struct r8a66597_platdata sh7724_usb0_host_data = {
+};
+
+static struct resource sh7724_usb0_host_resources[] = {
+	[0] = {
+		.start	= 0xa4d80000,
+		.end	= 0xa4d800ff,
+		.flags	= IORESOURCE_MEM,
+	},
+	[1] = {
+		.start	= 65,
+		.end	= 65,
+		.flags	= IORESOURCE_IRQ | IRQF_TRIGGER_LOW,
+	},
+};
+
+static struct platform_device sh7724_usb0_host_device = {
+	.name		= "r8a66597_hcd",
+	.id		= 0,
+	.dev = {
+		.dma_mask		= NULL,         /*  not use dma */
+		.coherent_dma_mask	= 0xffffffff,
+		.platform_data		= &sh7724_usb0_host_data,
+	},
+	.num_resources	= ARRAY_SIZE(sh7724_usb0_host_resources),
+	.resource	= sh7724_usb0_host_resources,
+};
+
 static struct platform_device *ms7724se_devices[] __initdata = {
 	&heartbeat_device,
 	&smc91x_eth_device,
@@ -311,6 +340,7 @@ static struct platform_device *ms7724se_devices[] __initdata = {
 	&ceu1_device,
 	&keysc_device,
 	&sh_eth_device,
+	&sh7724_usb0_host_device,
 };
 
 #define EEPROM_OP   0xBA206000
@@ -364,6 +394,7 @@ static void __init sh_eth_init(void)
 #define SW4140    0xBA201000
 #define FPGA_OUT  0xBA200400
 #define PORT_HIZA 0xA4050158
+#define PORT_MSELCRB 0xA4050182
 
 #define SW41_A    0x0100
 #define SW41_B    0x0200
@@ -373,6 +404,7 @@ static void __init sh_eth_init(void)
 #define SW41_F    0x2000
 #define SW41_G    0x4000
 #define SW41_H    0x8000
+
 static int __init devices_setup(void)
 {
 	u16 sw = ctrl_inw(SW4140); /* select camera, monitor */
@@ -385,6 +417,12 @@ static int __init devices_setup(void)
 		    (1 << 14)), /* RMII */
 		  FPGA_OUT);
 
+	/* turn on USB clocks, use external clock */
+	ctrl_outw((ctrl_inw(PORT_MSELCRB) & ~0xc000) | 0x8000, PORT_MSELCRB);
+
+	/* enable USB0 port */
+	ctrl_outw(0x0600, 0xa40501d4);
+
 	/* enable IRQ 0,1,2 */
 	gpio_request(GPIO_FN_INTC_IRQ0, NULL);
 	gpio_request(GPIO_FN_INTC_IRQ1, NULL);

arch/sh/include/asm/hwblk.h

@@ -0,0 +1,61 @@
+#ifndef __ASM_SH_HWBLK_H
+#define __ASM_SH_HWBLK_H
+
+#include <asm/clock.h>
+#include <asm/io.h>
+
+#define HWBLK_AREA_FLAG_PARENT (1 << 0) /* valid parent */
+
+#define HWBLK_AREA(_flags, _parent)		\
+{						\
+	.flags = _flags,			\
+	.parent = _parent,			\
+}
+
+struct hwblk_area {
+	unsigned long cnt;
+	unsigned char parent;
+	unsigned char flags;
+};
+
+#define HWBLK(_mstp, _bit, _area)		\
+{						\
+	.mstp = (void __iomem *)_mstp,		\
+	.bit = _bit,				\
+	.area = _area,				\
+}
+
+struct hwblk {
+	void __iomem *mstp;
+	unsigned char bit;
+	unsigned char area;
+	unsigned long cnt;
+};
+
+struct hwblk_info {
+	struct hwblk_area *areas;
+	int nr_areas;
+	struct hwblk *hwblks;
+	int nr_hwblks;
+};
+
+/* Should be defined by processor-specific code */
+int arch_hwblk_init(void);
+int arch_hwblk_sleep_mode(void);
+
+int hwblk_register(struct hwblk_info *info);
+int hwblk_init(void);
+
+/* allow clocks to enable and disable hardware blocks */
+#define SH_HWBLK_CLK(_name, _id, _parent, _hwblk, _flags)	\
+{							\
+	.name		= _name,			\
+	.id		= _id,				\
+	.parent		= _parent,			\
+	.arch_flags	= _hwblk,			\
+	.flags		= _flags,			\
+}
+
+int sh_hwblk_clk_register(struct clk *clks, int nr);
+
+#endif /* __ASM_SH_HWBLK_H */

arch/sh/include/asm/suspend.h

@@ -10,6 +10,15 @@ struct swsusp_arch_regs {
 	struct pt_regs user_regs;
 	unsigned long bank1_regs[8];
 };
+
+void sh_mobile_call_standby(unsigned long mode);
+
+#ifdef CONFIG_CPU_IDLE
+void sh_mobile_setup_cpuidle(void);
+#else
+static inline void sh_mobile_setup_cpuidle(void) {}
+#endif
+
 #endif
 
 /* flags passed to assembly suspend code */

arch/sh/include/cpu-sh4/cpu/sh7722.h

@@ -221,4 +221,18 @@ enum {
 	GPIO_FN_KEYOUT3, GPIO_FN_KEYOUT4_IN6, GPIO_FN_KEYOUT5_IN5,
 };
 
+enum {
+	HWBLK_UNKNOWN = 0,
+	HWBLK_TLB, HWBLK_IC, HWBLK_OC, HWBLK_URAM, HWBLK_XYMEM,
+	HWBLK_INTC, HWBLK_DMAC, HWBLK_SHYWAY, HWBLK_HUDI,
+	HWBLK_UBC, HWBLK_TMU, HWBLK_CMT, HWBLK_RWDT, HWBLK_FLCTL,
+	HWBLK_SCIF0, HWBLK_SCIF1, HWBLK_SCIF2, HWBLK_SIO,
+	HWBLK_SIOF0, HWBLK_SIOF1, HWBLK_IIC, HWBLK_RTC,
+	HWBLK_TPU, HWBLK_IRDA, HWBLK_SDHI, HWBLK_SIM, HWBLK_KEYSC,
+	HWBLK_TSIF, HWBLK_USBF, HWBLK_2DG, HWBLK_SIU, HWBLK_VOU,
+	HWBLK_JPU, HWBLK_BEU, HWBLK_CEU, HWBLK_VEU, HWBLK_VPU,
+	HWBLK_LCDC,
+	HWBLK_NR,
+};
+
 #endif /* __ASM_SH7722_H__ */

arch/sh/kernel/cpu/Makefile

@@ -19,4 +19,4 @@ obj-$(CONFIG_UBC_WAKEUP)	+= ubc.o
 obj-$(CONFIG_SH_ADC)		+= adc.o
 obj-$(CONFIG_SH_CLK_CPG)	+= clock-cpg.o
 
-obj-y	+= irq/ init.o clock.o
+obj-y	+= irq/ init.o clock.o hwblk.o

arch/sh/kernel/cpu/hwblk.c

@@ -0,0 +1,130 @@
+#include <linux/clk.h>
+#include <linux/compiler.h>
+#include <linux/slab.h>
+#include <linux/io.h>
+#include <linux/spinlock.h>
+#include <asm/suspend.h>
+#include <asm/hwblk.h>
+#include <asm/clock.h>
+
+static DEFINE_SPINLOCK(hwblk_lock);
+
+static void hwblk_area_inc(struct hwblk_info *info, int area)
+{
+	struct hwblk_area *hap = info->areas + area;
+
+	hap->cnt++;
+	if (hap->cnt == 1)
+		if (hap->flags & HWBLK_AREA_FLAG_PARENT)
+			hwblk_area_inc(info, hap->parent);
+}
+
+static void hwblk_area_dec(struct hwblk_info *info, int area)
+{
+	struct hwblk_area *hap = info->areas + area;
+
+	if (hap->cnt == 1)
+		if (hap->flags & HWBLK_AREA_FLAG_PARENT)
+			hwblk_area_dec(info, hap->parent);
+	hap->cnt--;
+}
+
+static void hwblk_enable(struct hwblk_info *info, int hwblk)
+{
+	struct hwblk *hp = info->hwblks + hwblk;
+	unsigned long tmp;
+	unsigned long flags;
+
+	spin_lock_irqsave(&hwblk_lock, flags);
+
+	hp->cnt++;
+	if (hp->cnt == 1) {
+		hwblk_area_inc(info, hp->area);
+
+		tmp = __raw_readl(hp->mstp);
+		tmp &= ~(1 << hp->bit);
+		__raw_writel(tmp, hp->mstp);
+	}
+
+	spin_unlock_irqrestore(&hwblk_lock, flags);
+}
+
+static void hwblk_disable(struct hwblk_info *info, int hwblk)
+{
+	struct hwblk *hp = info->hwblks + hwblk;
+	unsigned long tmp;
+	unsigned long flags;
+
+	spin_lock_irqsave(&hwblk_lock, flags);
+
+	if (hp->cnt == 1) {
+		hwblk_area_dec(info, hp->area);
+
+		tmp = __raw_readl(hp->mstp);
+		tmp |= 1 << hp->bit;
+		__raw_writel(tmp, hp->mstp);
+	}
+	hp->cnt--;
+
+	spin_unlock_irqrestore(&hwblk_lock, flags);
+}
+
+static struct hwblk_info *hwblk_info;
+
+int __init hwblk_register(struct hwblk_info *info)
+{
+	hwblk_info = info;
+	return 0;
+}
+
+int __init __weak arch_hwblk_init(void)
+{
+	return 0;
+}
+
+int __weak arch_hwblk_sleep_mode(void)
+{
+	return SUSP_SH_SLEEP;
+}
+
+int __init hwblk_init(void)
+{
+	return arch_hwblk_init();
+}
+
+/* allow clocks to enable and disable hardware blocks */
+static int sh_hwblk_clk_enable(struct clk *clk)
+{
+	if (!hwblk_info)
+		return -ENOENT;
+
+	hwblk_enable(hwblk_info, clk->arch_flags);
+	return 0;
+}
+
+static void sh_hwblk_clk_disable(struct clk *clk)
+{
+	if (hwblk_info)
+		hwblk_disable(hwblk_info, clk->arch_flags);
+}
+
+static struct clk_ops sh_hwblk_clk_ops = {
+	.enable		= sh_hwblk_clk_enable,
+	.disable	= sh_hwblk_clk_disable,
+	.recalc		= followparent_recalc,
+};
+
+int __init sh_hwblk_clk_register(struct clk *clks, int nr)
+{
+	struct clk *clkp;
+	int ret = 0;
+	int k;
+
+	for (k = 0; !ret && (k < nr); k++) {
+		clkp = clks + k;
+		clkp->ops = &sh_hwblk_clk_ops;
+		ret |= clk_register(clkp);
+	}
+
+	return ret;
+}

arch/sh/kernel/cpu/sh4a/Makefile

@@ -25,7 +25,7 @@ clock-$(CONFIG_CPU_SUBTYPE_SH7780)	:= clock-sh7780.o
 clock-$(CONFIG_CPU_SUBTYPE_SH7785)	:= clock-sh7785.o
 clock-$(CONFIG_CPU_SUBTYPE_SH7786)	:= clock-sh7786.o
 clock-$(CONFIG_CPU_SUBTYPE_SH7343)	:= clock-sh7343.o
-clock-$(CONFIG_CPU_SUBTYPE_SH7722)	:= clock-sh7722.o
+clock-$(CONFIG_CPU_SUBTYPE_SH7722)	:= clock-sh7722.o hwblk-sh7722.o
 clock-$(CONFIG_CPU_SUBTYPE_SH7723)	:= clock-sh7723.o
 clock-$(CONFIG_CPU_SUBTYPE_SH7724)	:= clock-sh7724.o
 clock-$(CONFIG_CPU_SUBTYPE_SH7366)	:= clock-sh7366.o

arch/sh/kernel/cpu/sh4a/clock-sh7722.c

@@ -22,6 +22,8 @@
 #include <linux/kernel.h>
 #include <linux/io.h>
 #include <asm/clock.h>
+#include <asm/hwblk.h>
+#include <cpu/sh7722.h>
 
 /* SH7722 registers */
 #define FRQCR		0xa4150000
@@ -140,35 +142,37 @@ struct clk div6_clks[] = {
 	SH_CLK_DIV6("video_clk", &pll_clk, VCLKCR, 0),
 };
 
-#define MSTP(_str, _parent, _reg, _bit, _flags) \
-  SH_CLK_MSTP32(_str, -1, _parent, _reg, _bit, _flags)
+#define R_CLK &r_clk
+#define P_CLK &div4_clks[DIV4_P]
+#define B_CLK &div4_clks[DIV4_B]
+#define U_CLK &div4_clks[DIV4_U]
 
 static struct clk mstp_clks[] = {
-	MSTP("uram0", &div4_clks[DIV4_U], MSTPCR0, 28, CLK_ENABLE_ON_INIT),
-	MSTP("xymem0", &div4_clks[DIV4_B], MSTPCR0, 26, CLK_ENABLE_ON_INIT),
-	MSTP("tmu0", &div4_clks[DIV4_P], MSTPCR0, 15, 0),
-	MSTP("cmt0", &r_clk, MSTPCR0, 14, 0),
-	MSTP("rwdt0", &r_clk, MSTPCR0, 13, 0),
-	MSTP("flctl0", &div4_clks[DIV4_P], MSTPCR0, 10, 0),
-	MSTP("scif0", &div4_clks[DIV4_P], MSTPCR0, 7, 0),
-	MSTP("scif1", &div4_clks[DIV4_P], MSTPCR0, 6, 0),
-	MSTP("scif2", &div4_clks[DIV4_P], MSTPCR0, 5, 0),
-
-	MSTP("i2c0", &div4_clks[DIV4_P], MSTPCR1, 9, 0),
-	MSTP("rtc0", &r_clk, MSTPCR1, 8, 0),
-
-	MSTP("sdhi0", &div4_clks[DIV4_P], MSTPCR2, 18, 0),
-	MSTP("keysc0", &r_clk, MSTPCR2, 14, 0),
-	MSTP("usbf0", &div4_clks[DIV4_P], MSTPCR2, 11, 0),
-	MSTP("2dg0", &div4_clks[DIV4_B], MSTPCR2, 9, 0),
-	MSTP("siu0", &div4_clks[DIV4_B], MSTPCR2, 8, 0),
-	MSTP("vou0", &div4_clks[DIV4_B], MSTPCR2, 5, 0),
-	MSTP("jpu0", &div4_clks[DIV4_B], MSTPCR2, 6, CLK_ENABLE_ON_INIT),
-	MSTP("beu0", &div4_clks[DIV4_B], MSTPCR2, 4, 0),
-	MSTP("ceu0", &div4_clks[DIV4_B], MSTPCR2, 3, 0),
-	MSTP("veu0", &div4_clks[DIV4_B], MSTPCR2, 2, CLK_ENABLE_ON_INIT),
-	MSTP("vpu0", &div4_clks[DIV4_B], MSTPCR2, 1, CLK_ENABLE_ON_INIT),
-	MSTP("lcdc0", &div4_clks[DIV4_B], MSTPCR2, 0, 0),
+	SH_HWBLK_CLK("uram0", -1, U_CLK, HWBLK_URAM, CLK_ENABLE_ON_INIT),
+	SH_HWBLK_CLK("xymem0", -1, B_CLK, HWBLK_XYMEM, CLK_ENABLE_ON_INIT),
+	SH_HWBLK_CLK("tmu0", -1, P_CLK, HWBLK_TMU, 0),
+	SH_HWBLK_CLK("cmt0", -1, R_CLK, HWBLK_CMT, 0),
+	SH_HWBLK_CLK("rwdt0", -1, R_CLK, HWBLK_RWDT, 0),
+	SH_HWBLK_CLK("flctl0", -1, P_CLK, HWBLK_FLCTL, 0),
+	SH_HWBLK_CLK("scif0", -1, P_CLK, HWBLK_SCIF0, 0),
+	SH_HWBLK_CLK("scif1", -1, P_CLK, HWBLK_SCIF1, 0),
+	SH_HWBLK_CLK("scif2", -1, P_CLK, HWBLK_SCIF2, 0),
+
+	SH_HWBLK_CLK("i2c0", -1, P_CLK, HWBLK_IIC, 0),
+	SH_HWBLK_CLK("rtc0", -1, R_CLK, HWBLK_RTC, 0),
+
+	SH_HWBLK_CLK("sdhi0", -1, P_CLK, HWBLK_SDHI, 0),
+	SH_HWBLK_CLK("keysc0", -1, R_CLK, HWBLK_KEYSC, 0),
+	SH_HWBLK_CLK("usbf0", -1, P_CLK, HWBLK_USBF, 0),
+	SH_HWBLK_CLK("2dg0", -1, B_CLK, HWBLK_2DG, 0),
+	SH_HWBLK_CLK("siu0", -1, B_CLK, HWBLK_SIU, 0),
+	SH_HWBLK_CLK("vou0", -1, B_CLK, HWBLK_VOU, 0),
+	SH_HWBLK_CLK("jpu0", -1, B_CLK, HWBLK_JPU, CLK_ENABLE_ON_INIT),
+	SH_HWBLK_CLK("beu0", -1, B_CLK, HWBLK_BEU, 0),
+	SH_HWBLK_CLK("ceu0", -1, B_CLK, HWBLK_CEU, 0),
+	SH_HWBLK_CLK("veu0", -1, B_CLK, HWBLK_VEU, CLK_ENABLE_ON_INIT),
+	SH_HWBLK_CLK("vpu0", -1, B_CLK, HWBLK_VPU, CLK_ENABLE_ON_INIT),
+	SH_HWBLK_CLK("lcdc0", -1, P_CLK, HWBLK_LCDC, 0),
 };
 
 int __init arch_clk_init(void)
@@ -191,7 +195,7 @@ int __init arch_clk_init(void)
 		ret = sh_clk_div6_register(div6_clks, ARRAY_SIZE(div6_clks));
 
 	if (!ret)
-		ret = sh_clk_mstp32_register(mstp_clks, ARRAY_SIZE(mstp_clks));
+		ret = sh_hwblk_clk_register(mstp_clks, ARRAY_SIZE(mstp_clks));
 
 	return ret;
 }

arch/sh/kernel/cpu/sh4a/hwblk-sh7722.c

@@ -0,0 +1,106 @@
+/*
+ * arch/sh/kernel/cpu/sh4a/hwblk-sh7722.c
+ *
+ * SH7722 hardware block support
+ *
+ * Copyright (C) 2009 Magnus Damm
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/io.h>
+#include <asm/suspend.h>
+#include <asm/hwblk.h>
+#include <cpu/sh7722.h>
+
+/* SH7722 registers */
+#define MSTPCR0		0xa4150030
+#define MSTPCR1		0xa4150034
+#define MSTPCR2		0xa4150038
+
+/* SH7722 Power Domains */
+enum { CORE_AREA, SUB_AREA, CORE_AREA_BM };
+static struct hwblk_area sh7722_hwblk_area[] = {
+	[CORE_AREA] = HWBLK_AREA(0, 0),
+	[CORE_AREA_BM] = HWBLK_AREA(HWBLK_AREA_FLAG_PARENT, CORE_AREA),
+	[SUB_AREA] = HWBLK_AREA(0, 0),
+};
+
+/* Table mapping HWBLK to Module Stop Bit and Power Domain */
+static struct hwblk sh7722_hwblk[HWBLK_NR] = {
+	[HWBLK_TLB] = HWBLK(MSTPCR0, 31, CORE_AREA),
+	[HWBLK_IC] = HWBLK(MSTPCR0, 30, CORE_AREA),
+	[HWBLK_OC] = HWBLK(MSTPCR0, 29, CORE_AREA),
+	[HWBLK_URAM] = HWBLK(MSTPCR0, 28, CORE_AREA),
+	[HWBLK_XYMEM] = HWBLK(MSTPCR0, 26, CORE_AREA),
+	[HWBLK_INTC] = HWBLK(MSTPCR0, 22, CORE_AREA),
+	[HWBLK_DMAC] = HWBLK(MSTPCR0, 21, CORE_AREA_BM),
+	[HWBLK_SHYWAY] = HWBLK(MSTPCR0, 20, CORE_AREA),
+	[HWBLK_HUDI] = HWBLK(MSTPCR0, 19, CORE_AREA),
+	[HWBLK_UBC] = HWBLK(MSTPCR0, 17, CORE_AREA),
+	[HWBLK_TMU] = HWBLK(MSTPCR0, 15, CORE_AREA),
+	[HWBLK_CMT] = HWBLK(MSTPCR0, 14, SUB_AREA),
+	[HWBLK_RWDT] = HWBLK(MSTPCR0, 13, SUB_AREA),
+	[HWBLK_FLCTL] = HWBLK(MSTPCR0, 10, CORE_AREA),
+	[HWBLK_SCIF0] = HWBLK(MSTPCR0, 7, CORE_AREA),
+	[HWBLK_SCIF1] = HWBLK(MSTPCR0, 6, CORE_AREA),
+	[HWBLK_SCIF2] = HWBLK(MSTPCR0, 5, CORE_AREA),
+	[HWBLK_SIO] = HWBLK(MSTPCR0, 3, CORE_AREA),
+	[HWBLK_SIOF0] = HWBLK(MSTPCR0, 2, CORE_AREA),
+	[HWBLK_SIOF1] = HWBLK(MSTPCR0, 1, CORE_AREA),
+
+	[HWBLK_IIC] = HWBLK(MSTPCR1, 9, CORE_AREA),
+	[HWBLK_RTC] = HWBLK(MSTPCR1, 8, SUB_AREA),
+
+	[HWBLK_TPU] = HWBLK(MSTPCR2, 25, CORE_AREA),
+	[HWBLK_IRDA] = HWBLK(MSTPCR2, 24, CORE_AREA),
+	[HWBLK_SDHI] = HWBLK(MSTPCR2, 18, CORE_AREA),
+	[HWBLK_SIM] = HWBLK(MSTPCR2, 16, CORE_AREA),
+	[HWBLK_KEYSC] = HWBLK(MSTPCR2, 14, SUB_AREA),
+	[HWBLK_TSIF] = HWBLK(MSTPCR2, 13, SUB_AREA),
+	[HWBLK_USBF] = HWBLK(MSTPCR2, 11, CORE_AREA),
+	[HWBLK_2DG] = HWBLK(MSTPCR2, 9, CORE_AREA_BM),
+	[HWBLK_SIU] = HWBLK(MSTPCR2, 8, CORE_AREA),
+	[HWBLK_JPU] = HWBLK(MSTPCR2, 6, CORE_AREA_BM),
+	[HWBLK_VOU] = HWBLK(MSTPCR2, 5, CORE_AREA_BM),
+	[HWBLK_BEU] = HWBLK(MSTPCR2, 4, CORE_AREA_BM),
+	[HWBLK_CEU] = HWBLK(MSTPCR2, 3, CORE_AREA_BM),
+	[HWBLK_VEU] = HWBLK(MSTPCR2, 2, CORE_AREA_BM),
+	[HWBLK_VPU] = HWBLK(MSTPCR2, 1, CORE_AREA_BM),
+	[HWBLK_LCDC] = HWBLK(MSTPCR2, 0, CORE_AREA_BM),
+};
+
+static struct hwblk_info sh7722_hwblk_info = {
+	.areas = sh7722_hwblk_area,
+	.nr_areas = ARRAY_SIZE(sh7722_hwblk_area),
+	.hwblks = sh7722_hwblk,
+	.nr_hwblks = ARRAY_SIZE(sh7722_hwblk),
+};
+
+int arch_hwblk_sleep_mode(void)
+{
+	if (!sh7722_hwblk_area[CORE_AREA].cnt)
+		return SUSP_SH_STANDBY | SUSP_SH_SF;
+
+	if (!sh7722_hwblk_area[CORE_AREA_BM].cnt)
+		return SUSP_SH_SLEEP | SUSP_SH_SF;
+
+	return SUSP_SH_SLEEP;
+}
+
+int __init arch_hwblk_init(void)
+{
+	return hwblk_register(&sh7722_hwblk_info);
+}

arch/sh/kernel/cpu/shmobile/Makefile

@@ -4,3 +4,4 @@
 
 # Power Management & Sleep mode
 obj-$(CONFIG_PM)	+= pm.o sleep.o
+obj-$(CONFIG_CPU_IDLE)	+= cpuidle.o

arch/sh/kernel/cpu/shmobile/cpuidle.c

@@ -0,0 +1,102 @@
+/*
+ * arch/sh/kernel/cpu/shmobile/cpuidle.c
+ *
+ * Cpuidle support code for SuperH Mobile
+ *
+ *  Copyright (C) 2009 Magnus Damm
+ *
+ * This file is subject to the terms and conditions of the GNU General Public
+ * License.  See the file "COPYING" in the main directory of this archive
+ * for more details.
+ */
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/io.h>
+#include <linux/suspend.h>
+#include <linux/cpuidle.h>
+#include <asm/suspend.h>
+#include <asm/uaccess.h>
+#include <asm/hwblk.h>
+
+static unsigned long cpuidle_mode[] = {
+	SUSP_SH_SLEEP, /* regular sleep mode */
+	SUSP_SH_SLEEP | SUSP_SH_SF, /* sleep mode + self refresh */
+};
+
+static int cpuidle_sleep_enter(struct cpuidle_device *dev,
+			       struct cpuidle_state *state)
+{
+	unsigned long allowed_mode = arch_hwblk_sleep_mode();
+	ktime_t before, after;
+	int requested_state = state - &dev->states[0];
+	int allowed_state;
+	int k;
+
+	/* convert allowed mode to allowed state */
+	for (k = ARRAY_SIZE(cpuidle_mode) - 1; k > 0; k--)
+		if (cpuidle_mode[k] == allowed_mode)
+			break;
+
+	allowed_state = k;
+
+	/* take the following into account for sleep mode selection:
+	 * - allowed_state: best mode allowed by hardware (clock deps)
+	 * - requested_state: best mode allowed by software (latencies)
+	 */
+	k = min_t(int, allowed_state, requested_state);
+
+	dev->last_state = &dev->states[k];
+	before = ktime_get();
+	sh_mobile_call_standby(cpuidle_mode[k]);
+	after = ktime_get();
+	return ktime_to_ns(ktime_sub(after, before)) >> 10;
+}
+
+static struct cpuidle_device cpuidle_dev;
+static struct cpuidle_driver cpuidle_driver = {
+	.name =		"sh_idle",
+	.owner =	THIS_MODULE,
+};
+
+void sh_mobile_setup_cpuidle(void)
+{
+	struct cpuidle_device *dev = &cpuidle_dev;
+	struct cpuidle_state *state;
+	int i;
+
+	cpuidle_register_driver(&cpuidle_driver);
+
+	for (i = 0; i < CPUIDLE_STATE_MAX; i++) {
+		dev->states[i].name[0] = '\0';
+		dev->states[i].desc[0] = '\0';
+	}
+
+	i = CPUIDLE_DRIVER_STATE_START;
+
+	state = &dev->states[i++];
+	snprintf(state->name, CPUIDLE_NAME_LEN, "C0");
+	strncpy(state->desc, "SuperH Sleep Mode", CPUIDLE_DESC_LEN);
+	state->exit_latency = 1;
+	state->target_residency = 1 * 2;
+	state->power_usage = 3;
+	state->flags = 0;
+	state->flags |= CPUIDLE_FLAG_SHALLOW;
+	state->flags |= CPUIDLE_FLAG_TIME_VALID;
+	state->enter = cpuidle_sleep_enter;
+
+	dev->safe_state = state;
+
+	state = &dev->states[i++];
+	snprintf(state->name, CPUIDLE_NAME_LEN, "C1");
+	strncpy(state->desc, "SuperH Sleep Mode [SF]", CPUIDLE_DESC_LEN);
+	state->exit_latency = 100;
+	state->target_residency = 1 * 2;
+	state->power_usage = 1;
+	state->flags = 0;
+	state->flags |= CPUIDLE_FLAG_TIME_VALID;
+	state->enter = cpuidle_sleep_enter;
+
+	dev->state_count = i;
+
+	cpuidle_register_device(dev);
+}

arch/sh/kernel/cpu/shmobile/pm.c

@@ -1,5 +1,5 @@
 /*
- * arch/sh/kernel/cpu/sh4a/pm-sh_mobile.c
+ * arch/sh/kernel/cpu/shmobile/pm.c
  *
  * Power management support code for SuperH Mobile
  *
@@ -32,20 +32,17 @@
  *
  * R-standby mode is unsupported, but will be added in the future
  * U-standby mode is low priority since it needs bootloader hacks
- *
- * All modes should be tied in with cpuidle. But before that can
- * happen we need to keep track of enabled hardware blocks so we
- * can avoid entering sleep modes that stop clocks to hardware
- * blocks that are in use even though the cpu core is idle.
  */
 
+#define ILRAM_BASE 0xe5200000
+
 extern const unsigned char sh_mobile_standby[];
 extern const unsigned int sh_mobile_standby_size;
 
-static void sh_mobile_call_standby(unsigned long mode)
+void sh_mobile_call_standby(unsigned long mode)
 {
 	extern void *vbr_base;
-	void *onchip_mem = (void *)0xe5200000; /* ILRAM */
+	void *onchip_mem = (void *)ILRAM_BASE;
 	void (*standby_onchip_mem)(unsigned long) = onchip_mem;
 
 	/* Note: Wake up from sleep may generate exceptions!
@@ -55,11 +52,6 @@ static void sh_mobile_call_standby(unsigned long mode)
 	if (mode & SUSP_SH_SF)
 		asm volatile("ldc %0, vbr" : : "r" (onchip_mem) : "memory");
 
-	/* Copy the assembly snippet to the otherwise ununsed ILRAM */
-	memcpy(onchip_mem, sh_mobile_standby, sh_mobile_standby_size);
-	wmb();
-	ctrl_barrier();
-
 	/* Let assembly snippet in on-chip memory handle the rest */
 	standby_onchip_mem(mode);
 
@@ -85,7 +77,15 @@ static struct platform_suspend_ops sh_pm_ops = {
 
 static int __init sh_pm_init(void)
 {
+	void *onchip_mem = (void *)ILRAM_BASE;
+
+	/* Copy the assembly snippet to the otherwise ununsed ILRAM */
+	memcpy(onchip_mem, sh_mobile_standby, sh_mobile_standby_size);
+	wmb();
+	ctrl_barrier();
+
 	suspend_set_ops(&sh_pm_ops);
+	sh_mobile_setup_cpuidle();
 	return 0;
 }
 

arch/sh/kernel/time.c

@@ -21,6 +21,7 @@
 #include <linux/smp.h>
 #include <linux/rtc.h>
 #include <asm/clock.h>
+#include <asm/hwblk.h>
 #include <asm/rtc.h>
 
 /* Dummy RTC ops */
@@ -96,6 +97,7 @@ void __init time_init(void)
 	if (board_time_init)
 		board_time_init();
 
+	hwblk_init();
 	clk_init();
 
 	rtc_sh_get_time(&xtime);

arch/x86/Kconfig

@@ -1913,25 +1913,14 @@ config DMAR_DEFAULT_ON
 	  recommended you say N here while the DMAR code remains
 	  experimental.
 
-config DMAR_GFX_WA
-	def_bool y
-	prompt "Support for Graphics workaround"
-	depends on DMAR
-	---help---
-	  Current Graphics drivers tend to use physical address
-	  for DMA and avoid using DMA APIs. Setting this config
-	  option permits the IOMMU driver to set a unity map for
-	  all the OS-visible memory. Hence the driver can continue
-	  to use physical addresses for DMA.
-
 config DMAR_FLOPPY_WA
 	def_bool y
 	depends on DMAR
 	---help---
-	  Floppy disk drivers are know to bypass DMA API calls
+	  Floppy disk drivers are known to bypass DMA API calls
 	  thereby failing to work when IOMMU is enabled. This
 	  workaround will setup a 1:1 mapping for the first
-	  16M to make floppy (an ISA device) work.
+	  16MiB to make floppy (an ISA device) work.
 
 config INTR_REMAP
 	bool "Support for Interrupt Remapping (EXPERIMENTAL)"

arch/x86/kernel/pci-dma.c

@@ -211,11 +211,11 @@ static __init int iommu_setup(char *p)
 #ifdef CONFIG_SWIOTLB
 		if (!strncmp(p, "soft", 4))
 			swiotlb = 1;
+#endif
 		if (!strncmp(p, "pt", 2)) {
 			iommu_pass_through = 1;
 			return 1;
 		}
-#endif
 
 		gart_parse_options(p);
 

drivers/pci/intel-iommu.c

@@ -56,14 +56,32 @@
 #define MAX_AGAW_WIDTH 64
 
 #define DOMAIN_MAX_ADDR(gaw) ((((u64)1) << gaw) - 1)
+#define DOMAIN_MAX_PFN(gaw)  ((((u64)1) << (gaw-VTD_PAGE_SHIFT)) - 1)
 
 #define IOVA_PFN(addr)		((addr) >> PAGE_SHIFT)
 #define DMA_32BIT_PFN		IOVA_PFN(DMA_BIT_MASK(32))
 #define DMA_64BIT_PFN		IOVA_PFN(DMA_BIT_MASK(64))
 
-#ifndef PHYSICAL_PAGE_MASK
-#define PHYSICAL_PAGE_MASK PAGE_MASK
-#endif
+
+/* VT-d pages must always be _smaller_ than MM pages. Otherwise things
+   are never going to work. */
+static inline unsigned long dma_to_mm_pfn(unsigned long dma_pfn)
+{
+	return dma_pfn >> (PAGE_SHIFT - VTD_PAGE_SHIFT);
+}
+
+static inline unsigned long mm_to_dma_pfn(unsigned long mm_pfn)
+{
+	return mm_pfn << (PAGE_SHIFT - VTD_PAGE_SHIFT);
+}
+static inline unsigned long page_to_dma_pfn(struct page *pg)
+{
+	return mm_to_dma_pfn(page_to_pfn(pg));
+}
+static inline unsigned long virt_to_dma_pfn(void *p)
+{
+	return page_to_dma_pfn(virt_to_page(p));
+}
 
 /* global iommu list, set NULL for ignored DMAR units */
 static struct intel_iommu **g_iommus;
@@ -204,12 +222,17 @@ static inline void dma_set_pte_prot(struct dma_pte *pte, unsigned long prot)
 
 static inline u64 dma_pte_addr(struct dma_pte *pte)
 {
-	return (pte->val & VTD_PAGE_MASK);
+#ifdef CONFIG_64BIT
+	return pte->val & VTD_PAGE_MASK;
+#else
+	/* Must have a full atomic 64-bit read */
+	return  __cmpxchg64(pte, 0ULL, 0ULL) & VTD_PAGE_MASK;
+#endif
 }
 
-static inline void dma_set_pte_addr(struct dma_pte *pte, u64 addr)
+static inline void dma_set_pte_pfn(struct dma_pte *pte, unsigned long pfn)
 {
-	pte->val |= (addr & VTD_PAGE_MASK);
+	pte->val |= (uint64_t)pfn << VTD_PAGE_SHIFT;
 }
 
 static inline bool dma_pte_present(struct dma_pte *pte)
@@ -217,6 +240,11 @@ static inline bool dma_pte_present(struct dma_pte *pte)
 	return (pte->val & 3) != 0;
 }
 
+static inline int first_pte_in_page(struct dma_pte *pte)
+{
+	return !((unsigned long)pte & ~VTD_PAGE_MASK);
+}
+
 /*
  * This domain is a statically identity mapping domain.
  *	1. This domain creats a static 1:1 mapping to all usable memory.
@@ -244,7 +272,6 @@ struct dmar_domain {
 	struct iova_domain iovad;	/* iova's that belong to this domain */
 
 	struct dma_pte	*pgd;		/* virtual address */
-	spinlock_t	mapping_lock;	/* page table lock */
 	int		gaw;		/* max guest address width */
 
 	/* adjusted guest address width, 0 is level 2 30-bit */
@@ -648,80 +675,78 @@ static inline int width_to_agaw(int width)
 
 static inline unsigned int level_to_offset_bits(int level)
 {
-	return (12 + (level - 1) * LEVEL_STRIDE);
+	return (level - 1) * LEVEL_STRIDE;
 }
 
-static inline int address_level_offset(u64 addr, int level)
+static inline int pfn_level_offset(unsigned long pfn, int level)
 {
-	return ((addr >> level_to_offset_bits(level)) & LEVEL_MASK);
+	return (pfn >> level_to_offset_bits(level)) & LEVEL_MASK;
 }
 
-static inline u64 level_mask(int level)
+static inline unsigned long level_mask(int level)
 {
-	return ((u64)-1 << level_to_offset_bits(level));
+	return -1UL << level_to_offset_bits(level);
 }
 
-static inline u64 level_size(int level)
+static inline unsigned long level_size(int level)
 {
-	return ((u64)1 << level_to_offset_bits(level));
+	return 1UL << level_to_offset_bits(level);
 }
 
-static inline u64 align_to_level(u64 addr, int level)
+static inline unsigned long align_to_level(unsigned long pfn, int level)
 {
-	return ((addr + level_size(level) - 1) & level_mask(level));
+	return (pfn + level_size(level) - 1) & level_mask(level);
 }
 
-static struct dma_pte * addr_to_dma_pte(struct dmar_domain *domain, u64 addr)
+static struct dma_pte *pfn_to_dma_pte(struct dmar_domain *domain,
+				      unsigned long pfn)
 {
-	int addr_width = agaw_to_width(domain->agaw);
+	int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
 	struct dma_pte *parent, *pte = NULL;
 	int level = agaw_to_level(domain->agaw);
 	int offset;
-	unsigned long flags;
 
 	BUG_ON(!domain->pgd);
-
-	addr &= (((u64)1) << addr_width) - 1;
+	BUG_ON(addr_width < BITS_PER_LONG && pfn >> addr_width);
 	parent = domain->pgd;
 
-	spin_lock_irqsave(&domain->mapping_lock, flags);
 	while (level > 0) {
 		void *tmp_page;
 
-		offset = address_level_offset(addr, level);
+		offset = pfn_level_offset(pfn, level);
 		pte = &parent[offset];
 		if (level == 1)
 			break;
 
 		if (!dma_pte_present(pte)) {
+			uint64_t pteval;
+
 			tmp_page = alloc_pgtable_page();
 
-			if (!tmp_page) {
-				spin_unlock_irqrestore(&domain->mapping_lock,
-					flags);
+			if (!tmp_page)
 				return NULL;
+
+			domain_flush_cache(domain, tmp_page, VTD_PAGE_SIZE);
+			pteval = (virt_to_dma_pfn(tmp_page) << VTD_PAGE_SHIFT) | DMA_PTE_READ | DMA_PTE_WRITE;
+			if (cmpxchg64(&pte->val, 0ULL, pteval)) {
+				/* Someone else set it while we were thinking; use theirs. */
+				free_pgtable_page(tmp_page);
+			} else {
+				dma_pte_addr(pte);
+				domain_flush_cache(domain, pte, sizeof(*pte));
 			}
-			domain_flush_cache(domain, tmp_page, PAGE_SIZE);
-			dma_set_pte_addr(pte, virt_to_phys(tmp_page));
-			/*
-			 * high level table always sets r/w, last level page
-			 * table control read/write
-			 */
-			dma_set_pte_readable(pte);
-			dma_set_pte_writable(pte);
-			domain_flush_cache(domain, pte, sizeof(*pte));
 		}
 		parent = phys_to_virt(dma_pte_addr(pte));
 		level--;
 	}
 
-	spin_unlock_irqrestore(&domain->mapping_lock, flags);
 	return pte;
 }
 
 /* return address's pte at specific level */
-static struct dma_pte *dma_addr_level_pte(struct dmar_domain *domain, u64 addr,
-		int level)
+static struct dma_pte *dma_pfn_level_pte(struct dmar_domain *domain,
+					 unsigned long pfn,
+					 int level)
 {
 	struct dma_pte *parent, *pte = NULL;
 	int total = agaw_to_level(domain->agaw);
@@ -729,7 +754,7 @@ static struct dma_pte *dma_addr_level_pte(struct dmar_domain *domain, u64 addr,
 
 	parent = domain->pgd;
 	while (level <= total) {
-		offset = address_level_offset(addr, total);
+		offset = pfn_level_offset(pfn, total);
 		pte = &parent[offset];
 		if (level == total)
 			return pte;
@@ -742,74 +767,82 @@ static struct dma_pte *dma_addr_level_pte(struct dmar_domain *domain, u64 addr,
 	return NULL;
 }
 
-/* clear one page's page table */
-static void dma_pte_clear_one(struct dmar_domain *domain, u64 addr)
-{
-	struct dma_pte *pte = NULL;
-
-	/* get last level pte */
-	pte = dma_addr_level_pte(domain, addr, 1);
-
-	if (pte) {
-		dma_clear_pte(pte);
-		domain_flush_cache(domain, pte, sizeof(*pte));
-	}
-}
-
 /* clear last level pte, a tlb flush should be followed */
-static void dma_pte_clear_range(struct dmar_domain *domain, u64 start, u64 end)
+static void dma_pte_clear_range(struct dmar_domain *domain,
+				unsigned long start_pfn,
+				unsigned long last_pfn)
 {
-	int addr_width = agaw_to_width(domain->agaw);
-	int npages;
+	int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
+	struct dma_pte *first_pte, *pte;
+
+	BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width);
+	BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width);
 
-	start &= (((u64)1) << addr_width) - 1;
-	end &= (((u64)1) << addr_width) - 1;
-	/* in case it's partial page */
-	start &= PAGE_MASK;
-	end = PAGE_ALIGN(end);
-	npages = (end - start) / VTD_PAGE_SIZE;
+	/* we don't need lock here; nobody else touches the iova range */
+	while (start_pfn <= last_pfn) {
+		first_pte = pte = dma_pfn_level_pte(domain, start_pfn, 1);
+		if (!pte) {
+			start_pfn = align_to_level(start_pfn + 1, 2);
+			continue;
+		}
+		do { 
+			dma_clear_pte(pte);
+			start_pfn++;
+			pte++;
+		} while (start_pfn <= last_pfn && !first_pte_in_page(pte));
 
-	/* we don't need lock here, nobody else touches the iova range */
-	while (npages--) {
-		dma_pte_clear_one(domain, start);
-		start += VTD_PAGE_SIZE;
+		domain_flush_cache(domain, first_pte,
+				   (void *)pte - (void *)first_pte);
 	}
 }
 
 /* free page table pages. last level pte should already be cleared */
 static void dma_pte_free_pagetable(struct dmar_domain *domain,
-	u64 start, u64 end)
+				   unsigned long start_pfn,
+				   unsigned long last_pfn)
 {
-	int addr_width = agaw_to_width(domain->agaw);
-	struct dma_pte *pte;
+	int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
+	struct dma_pte *first_pte, *pte;
 	int total = agaw_to_level(domain->agaw);
 	int level;
-	u64 tmp;
+	unsigned long tmp;
 
-	start &= (((u64)1) << addr_width) - 1;
-	end &= (((u64)1) << addr_width) - 1;
+	BUG_ON(addr_width < BITS_PER_LONG && start_pfn >> addr_width);
+	BUG_ON(addr_width < BITS_PER_LONG && last_pfn >> addr_width);
 
-	/* we don't need lock here, nobody else touches the iova range */
+	/* We don't need lock here; nobody else touches the iova range */
 	level = 2;
 	while (level <= total) {
-		tmp = align_to_level(start, level);
-		if (tmp >= end || (tmp + level_size(level) > end))
+		tmp = align_to_level(start_pfn, level);
+
+		/* If we can't even clear one PTE at this level, we're done */
+		if (tmp + level_size(level) - 1 > last_pfn)
 			return;
 
-		while (tmp < end) {
-			pte = dma_addr_level_pte(domain, tmp, level);
-			if (pte) {
-				free_pgtable_page(
-					phys_to_virt(dma_pte_addr(pte)));
-				dma_clear_pte(pte);
-				domain_flush_cache(domain, pte, sizeof(*pte));
+		while (tmp + level_size(level) - 1 <= last_pfn) {
+			first_pte = pte = dma_pfn_level_pte(domain, tmp, level);
+			if (!pte) {
+				tmp = align_to_level(tmp + 1, level + 1);
+				continue;
 			}
-			tmp += level_size(level);
+			do {
+				if (dma_pte_present(pte)) {
+					free_pgtable_page(phys_to_virt(dma_pte_addr(pte)));
+					dma_clear_pte(pte);
+				}
+				pte++;
+				tmp += level_size(level);
+			} while (!first_pte_in_page(pte) &&
+				 tmp + level_size(level) - 1 <= last_pfn);
+
+			domain_flush_cache(domain, first_pte,
+					   (void *)pte - (void *)first_pte);
+			
 		}
 		level++;
 	}
 	/* free pgd */
-	if (start == 0 && end >= ((((u64)1) << addr_width) - 1)) {
+	if (start_pfn == 0 && last_pfn == DOMAIN_MAX_PFN(domain->gaw)) {
 		free_pgtable_page(domain->pgd);
 		domain->pgd = NULL;
 	}
@@ -1035,11 +1068,11 @@ static void iommu_flush_dev_iotlb(struct dmar_domain *domain,
 }
 
 static void iommu_flush_iotlb_psi(struct intel_iommu *iommu, u16 did,
-				  u64 addr, unsigned int pages)
+				  unsigned long pfn, unsigned int pages)
 {
 	unsigned int mask = ilog2(__roundup_pow_of_two(pages));
+	uint64_t addr = (uint64_t)pfn << VTD_PAGE_SHIFT;
 
-	BUG_ON(addr & (~VTD_PAGE_MASK));
 	BUG_ON(pages == 0);
 
 	/*
@@ -1054,7 +1087,12 @@ static void iommu_flush_iotlb_psi(struct intel_iommu *iommu, u16 did,
 	else
 		iommu->flush.flush_iotlb(iommu, did, addr, mask,
 						DMA_TLB_PSI_FLUSH);
-	if (did)
+
+	/*
+	 * In caching mode, domain ID 0 is reserved for non-present to present
+	 * mapping flush. Device IOTLB doesn't need to be flushed in this case.
+	 */
+	if (!cap_caching_mode(iommu->cap) || did)
 		iommu_flush_dev_iotlb(iommu->domains[did], addr, mask);
 }
 
@@ -1279,7 +1317,6 @@ static void dmar_init_reserved_ranges(void)
 	struct pci_dev *pdev = NULL;
 	struct iova *iova;
 	int i;
-	u64 addr, size;
 
 	init_iova_domain(&reserved_iova_list, DMA_32BIT_PFN);
 
@@ -1302,12 +1339,9 @@ static void dmar_init_reserved_ranges(void)
 			r = &pdev->resource[i];
 			if (!r->flags || !(r->flags & IORESOURCE_MEM))
 				continue;
-			addr = r->start;
-			addr &= PHYSICAL_PAGE_MASK;
-			size = r->end - addr;
-			size = PAGE_ALIGN(size);
-			iova = reserve_iova(&reserved_iova_list, IOVA_PFN(addr),
-				IOVA_PFN(size + addr) - 1);
+			iova = reserve_iova(&reserved_iova_list,
+					    IOVA_PFN(r->start),
+					    IOVA_PFN(r->end));
 			if (!iova)
 				printk(KERN_ERR "Reserve iova failed\n");
 		}
@@ -1341,7 +1375,6 @@ static int domain_init(struct dmar_domain *domain, int guest_width)
 	unsigned long sagaw;
 
 	init_iova_domain(&domain->iovad, DMA_32BIT_PFN);
-	spin_lock_init(&domain->mapping_lock);
 	spin_lock_init(&domain->iommu_lock);
 
 	domain_reserve_special_ranges(domain);
@@ -1388,7 +1421,6 @@ static void domain_exit(struct dmar_domain *domain)
 {
 	struct dmar_drhd_unit *drhd;
 	struct intel_iommu *iommu;
-	u64 end;
 
 	/* Domain 0 is reserved, so dont process it */
 	if (!domain)
@@ -1397,14 +1429,12 @@ static void domain_exit(struct dmar_domain *domain)
 	domain_remove_dev_info(domain);
 	/* destroy iovas */
 	put_iova_domain(&domain->iovad);
-	end = DOMAIN_MAX_ADDR(domain->gaw);
-	end = end & (~PAGE_MASK);
 
 	/* clear ptes */
-	dma_pte_clear_range(domain, 0, end);
+	dma_pte_clear_range(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
 
 	/* free page tables */
-	dma_pte_free_pagetable(domain, 0, end);
+	dma_pte_free_pagetable(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
 
 	for_each_active_iommu(iommu, drhd)
 		if (test_bit(iommu->seq_id, &domain->iommu_bmp))
@@ -1618,42 +1648,86 @@ static int domain_context_mapped(struct pci_dev *pdev)
 					     tmp->devfn);
 }
 
-static int
-domain_page_mapping(struct dmar_domain *domain, dma_addr_t iova,
-			u64 hpa, size_t size, int prot)
+static int __domain_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
+			    struct scatterlist *sg, unsigned long phys_pfn,
+			    unsigned long nr_pages, int prot)
 {
-	u64 start_pfn, end_pfn;
-	struct dma_pte *pte;
-	int index;
-	int addr_width = agaw_to_width(domain->agaw);
+	struct dma_pte *first_pte = NULL, *pte = NULL;
+	phys_addr_t uninitialized_var(pteval);
+	int addr_width = agaw_to_width(domain->agaw) - VTD_PAGE_SHIFT;
+	unsigned long sg_res;
 
-	hpa &= (((u64)1) << addr_width) - 1;
+	BUG_ON(addr_width < BITS_PER_LONG && (iov_pfn + nr_pages - 1) >> addr_width);
 
 	if ((prot & (DMA_PTE_READ|DMA_PTE_WRITE)) == 0)
 		return -EINVAL;
-	iova &= PAGE_MASK;
-	start_pfn = ((u64)hpa) >> VTD_PAGE_SHIFT;
-	end_pfn = (VTD_PAGE_ALIGN(((u64)hpa) + size)) >> VTD_PAGE_SHIFT;
-	index = 0;
-	while (start_pfn < end_pfn) {
-		pte = addr_to_dma_pte(domain, iova + VTD_PAGE_SIZE * index);
-		if (!pte)
-			return -ENOMEM;
+
+	prot &= DMA_PTE_READ | DMA_PTE_WRITE | DMA_PTE_SNP;
+
+	if (sg)
+		sg_res = 0;
+	else {
+		sg_res = nr_pages + 1;
+		pteval = ((phys_addr_t)phys_pfn << VTD_PAGE_SHIFT) | prot;
+	}
+
+	while (nr_pages--) {
+		uint64_t tmp;
+
+		if (!sg_res) {
+			sg_res = (sg->offset + sg->length + VTD_PAGE_SIZE - 1) >> VTD_PAGE_SHIFT;
+			sg->dma_address = ((dma_addr_t)iov_pfn << VTD_PAGE_SHIFT) + sg->offset;
+			sg->dma_length = sg->length;
+			pteval = page_to_phys(sg_page(sg)) | prot;
+		}
+		if (!pte) {
+			first_pte = pte = pfn_to_dma_pte(domain, iov_pfn);
+			if (!pte)
+				return -ENOMEM;
+		}
 		/* We don't need lock here, nobody else
 		 * touches the iova range
 		 */
-		BUG_ON(dma_pte_addr(pte));
-		dma_set_pte_addr(pte, start_pfn << VTD_PAGE_SHIFT);
-		dma_set_pte_prot(pte, prot);
-		if (prot & DMA_PTE_SNP)
-			dma_set_pte_snp(pte);
-		domain_flush_cache(domain, pte, sizeof(*pte));
-		start_pfn++;
-		index++;
+		tmp = cmpxchg64_local(&pte->val, 0ULL, pteval);
+		if (tmp) {
+			static int dumps = 5;
+			printk(KERN_CRIT "ERROR: DMA PTE for vPFN 0x%lx already set (to %llx not %llx)\n",
+			       iov_pfn, tmp, (unsigned long long)pteval);
+			if (dumps) {
+				dumps--;
+				debug_dma_dump_mappings(NULL);
+			}
+			WARN_ON(1);
+		}
+		pte++;
+		if (!nr_pages || first_pte_in_page(pte)) {
+			domain_flush_cache(domain, first_pte,
+					   (void *)pte - (void *)first_pte);
+			pte = NULL;
+		}
+		iov_pfn++;
+		pteval += VTD_PAGE_SIZE;
+		sg_res--;
+		if (!sg_res)
+			sg = sg_next(sg);
 	}
 	return 0;
 }
 
+static inline int domain_sg_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
+				    struct scatterlist *sg, unsigned long nr_pages,
+				    int prot)
+{
+	return __domain_mapping(domain, iov_pfn, sg, 0, nr_pages, prot);
+}
+
+static inline int domain_pfn_mapping(struct dmar_domain *domain, unsigned long iov_pfn,
+				     unsigned long phys_pfn, unsigned long nr_pages,
+				     int prot)
+{
+	return __domain_mapping(domain, iov_pfn, NULL, phys_pfn, nr_pages, prot);
+}
+
 static void iommu_detach_dev(struct intel_iommu *iommu, u8 bus, u8 devfn)
 {
 	if (!iommu)
@@ -1844,58 +1918,61 @@ error:
 
 static int iommu_identity_mapping;
 
+static int iommu_domain_identity_map(struct dmar_domain *domain,
+				     unsigned long long start,
+				     unsigned long long end)
+{
+	unsigned long first_vpfn = start >> VTD_PAGE_SHIFT;
+	unsigned long last_vpfn = end >> VTD_PAGE_SHIFT;
+
+	if (!reserve_iova(&domain->iovad, dma_to_mm_pfn(first_vpfn),
+			  dma_to_mm_pfn(last_vpfn))) {
+		printk(KERN_ERR "IOMMU: reserve iova failed\n");
+		return -ENOMEM;
+	}
+
+	pr_debug("Mapping reserved region %llx-%llx for domain %d\n",
+		 start, end, domain->id);
+	/*
+	 * RMRR range might have overlap with physical memory range,
+	 * clear it first
+	 */
+	dma_pte_clear_range(domain, first_vpfn, last_vpfn);
+
+	return domain_pfn_mapping(domain, first_vpfn, first_vpfn,
+				  last_vpfn - first_vpfn + 1,
+				  DMA_PTE_READ|DMA_PTE_WRITE);
+}
+
 static int iommu_prepare_identity_map(struct pci_dev *pdev,
 				      unsigned long long start,
 				      unsigned long long end)
 {
 	struct dmar_domain *domain;
-	unsigned long size;
-	unsigned long long base;
 	int ret;
 
 	printk(KERN_INFO
-		"IOMMU: Setting identity map for device %s [0x%Lx - 0x%Lx]\n",
-		pci_name(pdev), start, end);
-	if (iommu_identity_mapping)
-		domain = si_domain;
-	else
-		/* page table init */
-		domain = get_domain_for_dev(pdev, DEFAULT_DOMAIN_ADDRESS_WIDTH);
+	       "IOMMU: Setting identity map for device %s [0x%Lx - 0x%Lx]\n",
+	       pci_name(pdev), start, end);
+
+	domain = get_domain_for_dev(pdev, DEFAULT_DOMAIN_ADDRESS_WIDTH);
 	if (!domain)
 		return -ENOMEM;
 
-	/* The address might not be aligned */
-	base = start & PAGE_MASK;
-	size = end - base;
-	size = PAGE_ALIGN(size);
-	if (!reserve_iova(&domain->iovad, IOVA_PFN(base),
-			IOVA_PFN(base + size) - 1)) {
-		printk(KERN_ERR "IOMMU: reserve iova failed\n");
-		ret = -ENOMEM;
-		goto error;
-	}
-
-	pr_debug("Mapping reserved region %lx@%llx for %s\n",
-		size, base, pci_name(pdev));
-	/*
-	 * RMRR range might have overlap with physical memory range,
-	 * clear it first
-	 */
-	dma_pte_clear_range(domain, base, base + size);
-
-	ret = domain_page_mapping(domain, base, base, size,
-		DMA_PTE_READ|DMA_PTE_WRITE);
+	ret = iommu_domain_identity_map(domain, start, end);
 	if (ret)
 		goto error;
 
 	/* context entry init */
 	ret = domain_context_mapping(domain, pdev, CONTEXT_TT_MULTI_LEVEL);
-	if (!ret)
-		return 0;
-error:
+	if (ret)
+		goto error;
+
+	return 0;
+
+ error:
 	domain_exit(domain);
 	return ret;
-
 }
 
 static inline int iommu_prepare_rmrr_dev(struct dmar_rmrr_unit *rmrr,
@@ -1907,64 +1984,6 @@ static inline int iommu_prepare_rmrr_dev(struct dmar_rmrr_unit *rmrr,
 		rmrr->end_address + 1);
 }
 
-struct iommu_prepare_data {
-	struct pci_dev *pdev;
-	int ret;
-};
-
-static int __init iommu_prepare_work_fn(unsigned long start_pfn,
-					 unsigned long end_pfn, void *datax)
-{
-	struct iommu_prepare_data *data;
-
-	data = (struct iommu_prepare_data *)datax;
-
-	data->ret = iommu_prepare_identity_map(data->pdev,
-				start_pfn<<PAGE_SHIFT, end_pfn<<PAGE_SHIFT);
-	return data->ret;
-
-}
-
-static int __init iommu_prepare_with_active_regions(struct pci_dev *pdev)
-{
-	int nid;
-	struct iommu_prepare_data data;
-
-	data.pdev = pdev;
-	data.ret = 0;
-
-	for_each_online_node(nid) {
-		work_with_active_regions(nid, iommu_prepare_work_fn, &data);
-		if (data.ret)
-			return data.ret;
-	}
-	return data.ret;
-}
-
-#ifdef CONFIG_DMAR_GFX_WA
-static void __init iommu_prepare_gfx_mapping(void)
-{
-	struct pci_dev *pdev = NULL;
-	int ret;
-
-	for_each_pci_dev(pdev) {
-		if (pdev->dev.archdata.iommu == DUMMY_DEVICE_DOMAIN_INFO ||
-				!IS_GFX_DEVICE(pdev))
-			continue;
-		printk(KERN_INFO "IOMMU: gfx device %s 1-1 mapping\n",
-			pci_name(pdev));
-		ret = iommu_prepare_with_active_regions(pdev);
-		if (ret)
-			printk(KERN_ERR "IOMMU: mapping reserved region failed\n");
-	}
-}
-#else /* !CONFIG_DMAR_GFX_WA */
-static inline void iommu_prepare_gfx_mapping(void)
-{
-	return;
-}
-#endif
-
 #ifdef CONFIG_DMAR_FLOPPY_WA
 static inline void iommu_prepare_isa(void)
 {
@@ -1975,12 +1994,12 @@ static inline void iommu_prepare_isa(void)
 	if (!pdev)
 		return;
 
-	printk(KERN_INFO "IOMMU: Prepare 0-16M unity mapping for LPC\n");
+	printk(KERN_INFO "IOMMU: Prepare 0-16MiB unity mapping for LPC\n");
 	ret = iommu_prepare_identity_map(pdev, 0, 16*1024*1024);
 
 	if (ret)
-		printk(KERN_ERR "IOMMU: Failed to create 0-64M identity map, "
-			"floppy might not work\n");
+		printk(KERN_ERR "IOMMU: Failed to create 0-16MiB identity map; "
+		       "floppy might not work\n");
 
 }
 #else
@@ -2008,16 +2027,30 @@ static int __init init_context_pass_through(void)
 }
 
 static int md_domain_init(struct dmar_domain *domain, int guest_width);
+
+static int __init si_domain_work_fn(unsigned long start_pfn,
+				    unsigned long end_pfn, void *datax)
+{
+	int *ret = datax;
+
+	*ret = iommu_domain_identity_map(si_domain,
+					 (uint64_t)start_pfn << PAGE_SHIFT,
+					 (uint64_t)end_pfn << PAGE_SHIFT);
+	return *ret;
+
+}
+
 static int si_domain_init(void)
 {
 	struct dmar_drhd_unit *drhd;
 	struct intel_iommu *iommu;
-	int ret = 0;
+	int nid, ret = 0;
 
 	si_domain = alloc_domain();
 	if (!si_domain)
 		return -EFAULT;
 
+	pr_debug("Identity mapping domain is domain %d\n", si_domain->id);
 
 	for_each_active_iommu(iommu, drhd) {
 		ret = iommu_attach_domain(si_domain, iommu);
@@ -2034,6 +2067,12 @@ static int si_domain_init(void)
 
 	si_domain->flags = DOMAIN_FLAG_STATIC_IDENTITY;
 
+	for_each_online_node(nid) {
+		work_with_active_regions(nid, si_domain_work_fn, &ret);
+		if (ret)
+			return ret;
+	}
+
 	return 0;
 }
 
@@ -2087,13 +2126,14 @@ static int iommu_prepare_static_identity_mapping(void)
 	if (ret)
 		return -EFAULT;
 
-	printk(KERN_INFO "IOMMU: Setting identity map:\n");
 	for_each_pci_dev(pdev) {
-		ret = iommu_prepare_with_active_regions(pdev);
-		if (ret) {
-			printk(KERN_INFO "1:1 mapping to one domain failed.\n");
-			return -EFAULT;
-		}
+		printk(KERN_INFO "IOMMU: identity mapping for device %s\n",
+		       pci_name(pdev));
+
+		ret = domain_context_mapping(si_domain, pdev,
+					     CONTEXT_TT_MULTI_LEVEL);
+		if (ret)
+			return ret;
 		ret = domain_add_dev_info(si_domain, pdev);
 		if (ret)
 			return ret;
@@ -2284,8 +2324,6 @@ int __init init_dmars(void)
 			}
 		}
 
-		iommu_prepare_gfx_mapping();
-
 		iommu_prepare_isa();
 	}
 
@@ -2330,50 +2368,40 @@ error:
 	return ret;
 }
 
-static inline u64 aligned_size(u64 host_addr, size_t size)
-{
-	u64 addr;
-	addr = (host_addr & (~PAGE_MASK)) + size;
-	return PAGE_ALIGN(addr);
-}
-
-struct iova *
-iommu_alloc_iova(struct dmar_domain *domain, size_t size, u64 end)
+static inline unsigned long aligned_nrpages(unsigned long host_addr,
+					    size_t size)
 {
-	struct iova *piova;
+	host_addr &= ~PAGE_MASK;
+	host_addr += size + PAGE_SIZE - 1;
 
-	/* Make sure it's in range */
-	end = min_t(u64, DOMAIN_MAX_ADDR(domain->gaw), end);
-	if (!size || (IOVA_START_ADDR + size > end))
-		return NULL;
-
-	piova = alloc_iova(&domain->iovad,
-			size >> PAGE_SHIFT, IOVA_PFN(end), 1);
-	return piova;
+	return host_addr >> VTD_PAGE_SHIFT;
 }
 
-static struct iova *
-__intel_alloc_iova(struct device *dev, struct dmar_domain *domain,
-		   size_t size, u64 dma_mask)
+static struct iova *intel_alloc_iova(struct device *dev,
+				     struct dmar_domain *domain,
+				     unsigned long nrpages, uint64_t dma_mask)
 {
 	struct pci_dev *pdev = to_pci_dev(dev);
 	struct iova *iova = NULL;
 
-	if (dma_mask <= DMA_BIT_MASK(32) || dmar_forcedac)
-		iova = iommu_alloc_iova(domain, size, dma_mask);
-	else {
+	/* Restrict dma_mask to the width that the iommu can handle */
+	dma_mask = min_t(uint64_t, DOMAIN_MAX_ADDR(domain->gaw), dma_mask);
+
+	if (!dmar_forcedac && dma_mask > DMA_BIT_MASK(32)) {
 		/*
 		 * First try to allocate an io virtual address in
 		 * DMA_BIT_MASK(32) and if that fails then try allocating
 		 * from higher range
 		 */
-		iova = iommu_alloc_iova(domain, size, DMA_BIT_MASK(32));
-		if (!iova)
-			iova = iommu_alloc_iova(domain, size, dma_mask);
-	}
-
-	if (!iova) {
-		printk(KERN_ERR"Allocating iova for %s failed", pci_name(pdev));
+		iova = alloc_iova(&domain->iovad, nrpages,
+				  IOVA_PFN(DMA_BIT_MASK(32)), 1);
+		if (iova)
+			return iova;
+	}
+	iova = alloc_iova(&domain->iovad, nrpages, IOVA_PFN(dma_mask), 1);
+	if (unlikely(!iova)) {
+		printk(KERN_ERR "Allocating %ld-page iova for %s failed",
+		       nrpages, pci_name(pdev));
 		return NULL;
 	}
 
@@ -2476,14 +2504,12 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
 		return 0;
 
 	iommu = domain_get_iommu(domain);
-	size = aligned_size((u64)paddr, size);
+	size = aligned_nrpages(paddr, size);
 
-	iova = __intel_alloc_iova(hwdev, domain, size, pdev->dma_mask);
+	iova = intel_alloc_iova(hwdev, domain, size, pdev->dma_mask);
 	if (!iova)
 		goto error;
 
-	start_paddr = (phys_addr_t)iova->pfn_lo << PAGE_SHIFT;
-
 	/*
 	 * Check if DMAR supports zero-length reads on write only
 	 * mappings..
@@ -2499,20 +2525,20 @@ static dma_addr_t __intel_map_single(struct device *hwdev, phys_addr_t paddr,
 	 * might have two guest_addr mapping to the same host paddr, but this
 	 * is not a big problem
 	 */
-	ret = domain_page_mapping(domain, start_paddr,
-				  ((u64)paddr) & PHYSICAL_PAGE_MASK,
-				  size, prot);
+	ret = domain_pfn_mapping(domain, mm_to_dma_pfn(iova->pfn_lo),
+				 paddr >> VTD_PAGE_SHIFT, size, prot);
 	if (ret)
 		goto error;
 
 	/* it's a non-present to present mapping. Only flush if caching mode */
 	if (cap_caching_mode(iommu->cap))
-		iommu_flush_iotlb_psi(iommu, 0, start_paddr,
-				      size >> VTD_PAGE_SHIFT);
+		iommu_flush_iotlb_psi(iommu, 0, mm_to_dma_pfn(iova->pfn_lo), size);
 	else
 		iommu_flush_write_buffer(iommu);
 
-	return start_paddr + ((u64)paddr & (~PAGE_MASK));
+	start_paddr = (phys_addr_t)iova->pfn_lo << PAGE_SHIFT;
+	start_paddr += paddr & ~PAGE_MASK;
+	return start_paddr;
 
 error:
 	if (iova)
@@ -2605,7 +2631,7 @@ static void intel_unmap_page(struct device *dev, dma_addr_t dev_addr,
 {
 	struct pci_dev *pdev = to_pci_dev(dev);
 	struct dmar_domain *domain;
-	unsigned long start_addr;
+	unsigned long start_pfn, last_pfn;
 	struct iova *iova;
 	struct intel_iommu *iommu;
 
@@ -2618,22 +2644,25 @@ static void intel_unmap_page(struct device *dev, dma_addr_t dev_addr,
 	iommu = domain_get_iommu(domain);
 
 	iova = find_iova(&domain->iovad, IOVA_PFN(dev_addr));
-	if (!iova)
+	if (WARN_ONCE(!iova, "Driver unmaps unmatched page at PFN %llx\n",
+		      (unsigned long long)dev_addr))
 		return;
 
-	start_addr = iova->pfn_lo << PAGE_SHIFT;
-	size = aligned_size((u64)dev_addr, size);
+	start_pfn = mm_to_dma_pfn(iova->pfn_lo);
+	last_pfn = mm_to_dma_pfn(iova->pfn_hi + 1) - 1;
 
-	pr_debug("Device %s unmapping: %zx@%llx\n",
-		pci_name(pdev), size, (unsigned long long)start_addr);
+	pr_debug("Device %s unmapping: pfn %lx-%lx\n",
+		 pci_name(pdev), start_pfn, last_pfn);
 
 	/*  clear the whole page */
-	dma_pte_clear_range(domain, start_addr, start_addr + size);
+	dma_pte_clear_range(domain, start_pfn, last_pfn);
+
 	/* free page tables */
-	dma_pte_free_pagetable(domain, start_addr, start_addr + size);
+	dma_pte_free_pagetable(domain, start_pfn, last_pfn);
+
 	if (intel_iommu_strict) {
-		iommu_flush_iotlb_psi(iommu, domain->id, start_addr,
-				      size >> VTD_PAGE_SHIFT);
+		iommu_flush_iotlb_psi(iommu, domain->id, start_pfn,
+				      last_pfn - start_pfn + 1);
 		/* free iova */
 		__free_iova(&domain->iovad, iova);
 	} else {
@@ -2691,14 +2720,10 @@ static void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist,
 			   int nelems, enum dma_data_direction dir,
 			   struct dma_attrs *attrs)
 {
-	int i;
 	struct pci_dev *pdev = to_pci_dev(hwdev);
 	struct dmar_domain *domain;
-	unsigned long start_addr;
+	unsigned long start_pfn, last_pfn;
 	struct iova *iova;
-	size_t size = 0;
-	phys_addr_t addr;
-	struct scatterlist *sg;
 	struct intel_iommu *iommu;
 
 	if (iommu_no_mapping(pdev))
@@ -2710,22 +2735,21 @@ static void intel_unmap_sg(struct device *hwdev, struct scatterlist *sglist,
 	iommu = domain_get_iommu(domain);
 
 	iova = find_iova(&domain->iovad, IOVA_PFN(sglist[0].dma_address));
-	if (!iova)
+	if (WARN_ONCE(!iova, "Driver unmaps unmatched sglist at PFN %llx\n",
+		      (unsigned long long)sglist[0].dma_address))
 		return;
-	for_each_sg(sglist, sg, nelems, i) {
-		addr = page_to_phys(sg_page(sg)) + sg->offset;
-		size += aligned_size((u64)addr, sg->length);
-	}
 
-	start_addr = iova->pfn_lo << PAGE_SHIFT;
+	start_pfn = mm_to_dma_pfn(iova->pfn_lo);
+	last_pfn = mm_to_dma_pfn(iova->pfn_hi + 1) - 1;
 
 	/*  clear the whole page */
-	dma_pte_clear_range(domain, start_addr, start_addr + size);
+	dma_pte_clear_range(domain, start_pfn, last_pfn);
+
 	/* free page tables */
-	dma_pte_free_pagetable(domain, start_addr, start_addr + size);
+	dma_pte_free_pagetable(domain, start_pfn, last_pfn);
 
-	iommu_flush_iotlb_psi(iommu, domain->id, start_addr,
-			      size >> VTD_PAGE_SHIFT);
+	iommu_flush_iotlb_psi(iommu, domain->id, start_pfn,
+			      (last_pfn - start_pfn + 1));
 
 	/* free iova */
 	__free_iova(&domain->iovad, iova);
@@ -2748,17 +2772,16 @@ static int intel_nontranslate_map_sg(struct device *hddev,
 static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int nelems,
 			enum dma_data_direction dir, struct dma_attrs *attrs)
 {
-	phys_addr_t addr;
 	int i;
 	struct pci_dev *pdev = to_pci_dev(hwdev);
 	struct dmar_domain *domain;
 	size_t size = 0;
 	int prot = 0;
-	size_t offset = 0;
+	size_t offset_pfn = 0;
 	struct iova *iova = NULL;
 	int ret;
 	struct scatterlist *sg;
-	unsigned long start_addr;
+	unsigned long start_vpfn;
 	struct intel_iommu *iommu;
 
 	BUG_ON(dir == DMA_NONE);
@@ -2771,12 +2794,10 @@ static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int ne
 
 	iommu = domain_get_iommu(domain);
 
-	for_each_sg(sglist, sg, nelems, i) {
-		addr = page_to_phys(sg_page(sg)) + sg->offset;
-		size += aligned_size((u64)addr, sg->length);
-	}
+	for_each_sg(sglist, sg, nelems, i)
+		size += aligned_nrpages(sg->offset, sg->length);
 
-	iova = __intel_alloc_iova(hwdev, domain, size, pdev->dma_mask);
+	iova = intel_alloc_iova(hwdev, domain, size, pdev->dma_mask);
 	if (!iova) {
 		sglist->dma_length = 0;
 		return 0;
@@ -2792,35 +2813,24 @@ static int intel_map_sg(struct device *hwdev, struct scatterlist *sglist, int ne
 	if (dir == DMA_FROM_DEVICE || dir == DMA_BIDIRECTIONAL)
 		prot |= DMA_PTE_WRITE;
 
-	start_addr = iova->pfn_lo << PAGE_SHIFT;
-	offset = 0;
-	for_each_sg(sglist, sg, nelems, i) {
-		addr = page_to_phys(sg_page(sg)) + sg->offset;
-		size = aligned_size((u64)addr, sg->length);
-		ret = domain_page_mapping(domain, start_addr + offset,
-					  ((u64)addr) & PHYSICAL_PAGE_MASK,
-					  size, prot);
-		if (ret) {
-			/*  clear the page */
-			dma_pte_clear_range(domain, start_addr,
-				  start_addr + offset);
-			/* free page tables */
-			dma_pte_free_pagetable(domain, start_addr,
-				  start_addr + offset);
-			/* free iova */
-			__free_iova(&domain->iovad, iova);
-			return 0;
-		}
-		sg->dma_address = start_addr + offset +
-				((u64)addr & (~PAGE_MASK));
-		sg->dma_length = sg->length;
-		offset += size;
+	start_vpfn = mm_to_dma_pfn(iova->pfn_lo);
+
+	ret = domain_sg_mapping(domain, start_vpfn, sglist, mm_to_dma_pfn(size), prot);
+	if (unlikely(ret)) {
+		/*  clear the page */
+		dma_pte_clear_range(domain, start_vpfn,
+				    start_vpfn + size - 1);
+		/* free page tables */
+		dma_pte_free_pagetable(domain, start_vpfn,
+				       start_vpfn + size - 1);
+		/* free iova */
+		__free_iova(&domain->iovad, iova);
+		return 0;
 	}
 
 	/* it's a non-present to present mapping. Only flush if caching mode */
 	if (cap_caching_mode(iommu->cap))
-		iommu_flush_iotlb_psi(iommu, 0, start_addr,
-				      offset >> VTD_PAGE_SHIFT);
+		iommu_flush_iotlb_psi(iommu, 0, start_vpfn, offset_pfn);
 	else
 		iommu_flush_write_buffer(iommu);
 
@@ -3325,7 +3335,6 @@ static int md_domain_init(struct dmar_domain *domain, int guest_width)
 	int adjust_width;
 
 	init_iova_domain(&domain->iovad, DMA_32BIT_PFN);
-	spin_lock_init(&domain->mapping_lock);
 	spin_lock_init(&domain->iommu_lock);
 
 	domain_reserve_special_ranges(domain);
@@ -3379,8 +3388,6 @@ static void iommu_free_vm_domain(struct dmar_domain *domain)
 
 static void vm_domain_exit(struct dmar_domain *domain)
 {
-	u64 end;
-
 	/* Domain 0 is reserved, so dont process it */
 	if (!domain)
 		return;
@@ -3388,14 +3395,12 @@ static void vm_domain_exit(struct dmar_domain *domain)
 	vm_domain_remove_all_dev_info(domain);
 	/* destroy iovas */
 	put_iova_domain(&domain->iovad);
-	end = DOMAIN_MAX_ADDR(domain->gaw);
-	end = end & (~VTD_PAGE_MASK);
 
 	/* clear ptes */
-	dma_pte_clear_range(domain, 0, end);
+	dma_pte_clear_range(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
 
 	/* free page tables */
-	dma_pte_free_pagetable(domain, 0, end);
+	dma_pte_free_pagetable(domain, 0, DOMAIN_MAX_PFN(domain->gaw));
 
 	iommu_free_vm_domain(domain);
 	free_domain_mem(domain);
@@ -3504,7 +3509,7 @@ static int intel_iommu_map_range(struct iommu_domain *domain,
 	if ((iommu_prot & IOMMU_CACHE) && dmar_domain->iommu_snooping)
 		prot |= DMA_PTE_SNP;
 
-	max_addr = (iova & VTD_PAGE_MASK) + VTD_PAGE_ALIGN(size);
+	max_addr = iova + size;
 	if (dmar_domain->max_addr < max_addr) {
 		int min_agaw;
 		u64 end;
@@ -3522,8 +3527,11 @@ static int intel_iommu_map_range(struct iommu_domain *domain,
 		}
 		dmar_domain->max_addr = max_addr;
 	}
-
-	ret = domain_page_mapping(dmar_domain, iova, hpa, size, prot);
+	/* Round up size to next multiple of PAGE_SIZE, if it and
+	   the low bits of hpa would take us onto the next page */
+	size = aligned_nrpages(hpa, size);
+	ret = domain_pfn_mapping(dmar_domain, iova >> VTD_PAGE_SHIFT,
+				 hpa >> VTD_PAGE_SHIFT, size, prot);
 	return ret;
 }
 
@@ -3531,15 +3539,12 @@ static void intel_iommu_unmap_range(struct iommu_domain *domain,
 				    unsigned long iova, size_t size)
 {
 	struct dmar_domain *dmar_domain = domain->priv;
-	dma_addr_t base;
 
-	/* The address might not be aligned */
-	base = iova & VTD_PAGE_MASK;
-	size = VTD_PAGE_ALIGN(size);
-	dma_pte_clear_range(dmar_domain, base, base + size);
+	dma_pte_clear_range(dmar_domain, iova >> VTD_PAGE_SHIFT,
+			    (iova + size - 1) >> VTD_PAGE_SHIFT);
 
-	if (dmar_domain->max_addr == base + size)
-		dmar_domain->max_addr = base;
+	if (dmar_domain->max_addr == iova + size)
+		dmar_domain->max_addr = iova;
 }
 
 static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
@@ -3549,7 +3554,7 @@ static phys_addr_t intel_iommu_iova_to_phys(struct iommu_domain *domain,
 	struct dma_pte *pte;
 	u64 phys = 0;
 
-	pte = addr_to_dma_pte(dmar_domain, iova);
+	pte = pfn_to_dma_pte(dmar_domain, iova >> VTD_PAGE_SHIFT);
 	if (pte)
 		phys = dma_pte_addr(pte);
 

drivers/scsi/cxgb3i/cxgb3i_iscsi.c

@@ -13,6 +13,7 @@
 
 #include <linux/inet.h>
 #include <linux/crypto.h>
+#include <linux/if_vlan.h>
 #include <net/dst.h>
 #include <net/tcp.h>
 #include <scsi/scsi_cmnd.h>
@@ -184,6 +185,9 @@ static struct cxgb3i_hba *cxgb3i_hba_find_by_netdev(struct net_device *ndev)
 	struct cxgb3i_adapter *snic;
 	int i;
 
+	if (ndev->priv_flags & IFF_802_1Q_VLAN)
+		ndev = vlan_dev_real_dev(ndev);
+
 	read_lock(&cxgb3i_snic_rwlock);
 	list_for_each_entry(snic, &cxgb3i_snic_list, list_head) {
 		for (i = 0; i < snic->hba_cnt; i++) {

drivers/scsi/fnic/fnic_main.c

@@ -473,16 +473,16 @@ static int __devinit fnic_probe(struct pci_dev *pdev,
 	 * limitation for the device.  Try 40-bit first, and
 	 * fail to 32-bit.
 	 */
-	err = pci_set_dma_mask(pdev, DMA_40BIT_MASK);
+	err = pci_set_dma_mask(pdev, DMA_BIT_MASK(40));
 	if (err) {
-		err = pci_set_dma_mask(pdev, DMA_32BIT_MASK);
+		err = pci_set_dma_mask(pdev, DMA_BIT_MASK(32));
 		if (err) {
 			shost_printk(KERN_ERR, fnic->lport->host,
 				     "No usable DMA configuration "
 				     "aborting\n");
 			goto err_out_release_regions;
 		}
-		err = pci_set_consistent_dma_mask(pdev, DMA_32BIT_MASK);
+		err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(32));
 		if (err) {
 			shost_printk(KERN_ERR, fnic->lport->host,
 				     "Unable to obtain 32-bit DMA "
@@ -490,7 +490,7 @@ static int __devinit fnic_probe(struct pci_dev *pdev,
 			goto err_out_release_regions;
 		}
 	} else {
-		err = pci_set_consistent_dma_mask(pdev, DMA_40BIT_MASK);
+		err = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(40));
 		if (err) {
 			shost_printk(KERN_ERR, fnic->lport->host,
 				     "Unable to obtain 40-bit DMA "

drivers/scsi/fnic/fnic_scsi.c

@@ -245,7 +245,7 @@ static inline int fnic_queue_wq_copy_desc(struct fnic *fnic,
 					  struct vnic_wq_copy *wq,
 					  struct fnic_io_req *io_req,
 					  struct scsi_cmnd *sc,
-					  u32 sg_count)
+					  int sg_count)
 {
 	struct scatterlist *sg;
 	struct fc_rport *rport = starget_to_rport(scsi_target(sc->device));
@@ -260,9 +260,6 @@ static inline int fnic_queue_wq_copy_desc(struct fnic *fnic,
 	char msg[2];
 
 	if (sg_count) {
-		BUG_ON(sg_count < 0);
-		BUG_ON(sg_count > FNIC_MAX_SG_DESC_CNT);
-
 		/* For each SGE, create a device desc entry */
 		desc = io_req->sgl_list;
 		for_each_sg(scsi_sglist(sc), sg, sg_count, i) {
@@ -344,7 +341,7 @@ int fnic_queuecommand(struct scsi_cmnd *sc, void (*done)(struct scsi_cmnd *))
 	struct fnic *fnic;
 	struct vnic_wq_copy *wq;
 	int ret;
-	u32 sg_count;
+	int sg_count;
 	unsigned long flags;
 	unsigned long ptr;
 

drivers/scsi/ibmvscsi/ibmvscsi.c

@@ -1095,9 +1095,14 @@ static void adapter_info_rsp(struct srp_event_struct *evt_struct)
 				MAX_INDIRECT_BUFS);
 			hostdata->host->sg_tablesize = MAX_INDIRECT_BUFS;
 		}
+
+		if (hostdata->madapter_info.os_type == 3) {
+			enable_fast_fail(hostdata);
+			return;
+		}
 	}
 
-	enable_fast_fail(hostdata);
+	send_srp_login(hostdata);
 }
 
 /**

drivers/scsi/scsi_transport_fc.c

@@ -3670,13 +3670,14 @@ static void
 fc_bsg_goose_queue(struct fc_rport *rport)
 {
 	int flagset;
+	unsigned long flags;
 
 	if (!rport->rqst_q)
 		return;
 
 	get_device(&rport->dev);
 
-	spin_lock(rport->rqst_q->queue_lock);
+	spin_lock_irqsave(rport->rqst_q->queue_lock, flags);
 	flagset = test_bit(QUEUE_FLAG_REENTER, &rport->rqst_q->queue_flags) &&
 		  !test_bit(QUEUE_FLAG_REENTER, &rport->rqst_q->queue_flags);
 	if (flagset)
@@ -3684,7 +3685,7 @@ fc_bsg_goose_queue(struct fc_rport *rport)
 	__blk_run_queue(rport->rqst_q);
 	if (flagset)
 		queue_flag_clear(QUEUE_FLAG_REENTER, rport->rqst_q);
-	spin_unlock(rport->rqst_q->queue_lock);
+	spin_unlock_irqrestore(rport->rqst_q->queue_lock, flags);
 
 	put_device(&rport->dev);
 }

drivers/scsi/zalon.c

@@ -134,7 +134,7 @@ zalon_probe(struct parisc_device *dev)
 
 	host = ncr_attach(&zalon7xx_template, unit, &device);
 	if (!host)
-		goto fail;
+		return -ENODEV;
 
 	if (request_irq(dev->irq, ncr53c8xx_intr, IRQF_SHARED, "zalon", host)) {
 	  dev_printk(KERN_ERR, &dev->dev, "irq problem with %d, detaching\n ",

drivers/serial/sh-sci.c

@@ -707,12 +707,13 @@ static irqreturn_t sci_br_interrupt(int irq, void *ptr)
 
 static irqreturn_t sci_mpxed_interrupt(int irq, void *ptr)
 {
-	unsigned short ssr_status, scr_status;
+	unsigned short ssr_status, scr_status, err_enabled;
 	struct uart_port *port = ptr;
 	irqreturn_t ret = IRQ_NONE;
 
 	ssr_status = sci_in(port, SCxSR);
 	scr_status = sci_in(port, SCSCR);
+	err_enabled = scr_status & (SCI_CTRL_FLAGS_REIE | SCI_CTRL_FLAGS_RIE);
 
 	/* Tx Interrupt */
 	if ((ssr_status & 0x0020) && (scr_status & SCI_CTRL_FLAGS_TIE))
@@ -721,10 +722,10 @@ static irqreturn_t sci_mpxed_interrupt(int irq, void *ptr)
 	if ((ssr_status & 0x0002) && (scr_status & SCI_CTRL_FLAGS_RIE))
 		ret = sci_rx_interrupt(irq, ptr);
 	/* Error Interrupt */
-	if ((ssr_status & 0x0080) && (scr_status & SCI_CTRL_FLAGS_REIE))
+	if ((ssr_status & 0x0080) && err_enabled)
 		ret = sci_er_interrupt(irq, ptr);
 	/* Break Interrupt */
-	if ((ssr_status & 0x0010) && (scr_status & SCI_CTRL_FLAGS_REIE))
+	if ((ssr_status & 0x0010) && err_enabled)
 		ret = sci_br_interrupt(irq, ptr);
 
 	return ret;

drivers/usb/host/Kconfig

@@ -337,10 +337,10 @@ config USB_R8A66597_HCD
 
 config SUPERH_ON_CHIP_R8A66597
 	boolean "Enable SuperH on-chip R8A66597 USB"
-	depends on USB_R8A66597_HCD && (CPU_SUBTYPE_SH7366 || CPU_SUBTYPE_SH7723)
+	depends on USB_R8A66597_HCD && (CPU_SUBTYPE_SH7366 || CPU_SUBTYPE_SH7723 || CPU_SUBTYPE_SH7724)
 	help
 	   This driver enables support for the on-chip R8A66597 in the
-	   SH7366 and SH7723 processors.
+	   SH7366, SH7723 and SH7724 processors.
 
 config USB_WHCI_HCD
 	tristate "Wireless USB Host Controller Interface (WHCI) driver (EXPERIMENTAL)"

fs/btrfs/async-thread.c

@@ -299,8 +299,8 @@ int btrfs_start_workers(struct btrfs_workers *workers, int num_workers)
 					   "btrfs-%s-%d", workers->name,
 					   workers->num_workers + i);
 		if (IS_ERR(worker->task)) {
-			kfree(worker);
 			ret = PTR_ERR(worker->task);
+			kfree(worker);
 			goto fail;
 		}
 

fs/btrfs/ctree.h

@@ -2074,8 +2074,7 @@ static inline int btrfs_insert_empty_item(struct btrfs_trans_handle *trans,
 int btrfs_next_leaf(struct btrfs_root *root, struct btrfs_path *path);
 int btrfs_prev_leaf(struct btrfs_root *root, struct btrfs_path *path);
 int btrfs_leaf_free_space(struct btrfs_root *root, struct extent_buffer *leaf);
-int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
-			*root);
+int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref);
 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root,
 			struct extent_buffer *node,

fs/btrfs/extent-tree.c

@@ -990,15 +990,13 @@ static inline int extent_ref_type(u64 parent, u64 owner)
 	return type;
 }
 
-static int find_next_key(struct btrfs_path *path, struct btrfs_key *key)
+static int find_next_key(struct btrfs_path *path, int level,
+			 struct btrfs_key *key)
 
 {
-	int level;
-	BUG_ON(!path->keep_locks);
-	for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
+	for (; level < BTRFS_MAX_LEVEL; level++) {
 		if (!path->nodes[level])
 			break;
-		btrfs_assert_tree_locked(path->nodes[level]);
 		if (path->slots[level] + 1 >=
 		    btrfs_header_nritems(path->nodes[level]))
 			continue;
@@ -1158,7 +1156,8 @@ int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
 		 * For simplicity, we just do not add new inline back
 		 * ref if there is any kind of item for this block
 		 */
-		if (find_next_key(path, &key) == 0 && key.objectid == bytenr &&
+		if (find_next_key(path, 0, &key) == 0 &&
+		    key.objectid == bytenr &&
 		    key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
 			err = -EAGAIN;
 			goto out;
@@ -2697,7 +2696,7 @@ again:
 
 		printk(KERN_ERR "no space left, need %llu, %llu delalloc bytes"
 		       ", %llu bytes_used, %llu bytes_reserved, "
-		       "%llu bytes_pinned, %llu bytes_readonly, %llu may use"
+		       "%llu bytes_pinned, %llu bytes_readonly, %llu may use "
 		       "%llu total\n", (unsigned long long)bytes,
 		       (unsigned long long)data_sinfo->bytes_delalloc,
 		       (unsigned long long)data_sinfo->bytes_used,
@@ -4128,6 +4127,7 @@ struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
 	return buf;
 }
 
+#if 0
 int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root, struct extent_buffer *leaf)
 {
@@ -4171,8 +4171,6 @@ int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans,
 	return 0;
 }
 
-#if 0
-
 static noinline int cache_drop_leaf_ref(struct btrfs_trans_handle *trans,
 					struct btrfs_root *root,
 					struct btrfs_leaf_ref *ref)
@@ -4553,262 +4551,471 @@ out:
 }
 #endif
 
+struct walk_control {
+	u64 refs[BTRFS_MAX_LEVEL];
+	u64 flags[BTRFS_MAX_LEVEL];
+	struct btrfs_key update_progress;
+	int stage;
+	int level;
+	int shared_level;
+	int update_ref;
+	int keep_locks;
+};
+
+#define DROP_REFERENCE	1
+#define UPDATE_BACKREF	2
+
 /*
- * helper function for drop_subtree, this function is similar to
- * walk_down_tree. The main difference is that it checks reference
- * counts while tree blocks are locked.
+ * hepler to process tree block while walking down the tree.
+ *
+ * when wc->stage == DROP_REFERENCE, this function checks
+ * reference count of the block. if the block is shared and
+ * we need update back refs for the subtree rooted at the
+ * block, this function changes wc->stage to UPDATE_BACKREF
+ *
+ * when wc->stage == UPDATE_BACKREF, this function updates
+ * back refs for pointers in the block.
+ *
+ * NOTE: return value 1 means we should stop walking down.
  */
-static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
+static noinline int walk_down_proc(struct btrfs_trans_handle *trans,
 				   struct btrfs_root *root,
-				   struct btrfs_path *path, int *level)
+				   struct btrfs_path *path,
+				   struct walk_control *wc)
 {
-	struct extent_buffer *next;
-	struct extent_buffer *cur;
-	struct extent_buffer *parent;
-	u64 bytenr;
-	u64 ptr_gen;
-	u64 refs;
-	u64 flags;
-	u32 blocksize;
+	int level = wc->level;
+	struct extent_buffer *eb = path->nodes[level];
+	struct btrfs_key key;
+	u64 flag = BTRFS_BLOCK_FLAG_FULL_BACKREF;
 	int ret;
 
-	cur = path->nodes[*level];
-	ret = btrfs_lookup_extent_info(trans, root, cur->start, cur->len,
-				       &refs, &flags);
-	BUG_ON(ret);
-	if (refs > 1)
-		goto out;
+	if (wc->stage == UPDATE_BACKREF &&
+	    btrfs_header_owner(eb) != root->root_key.objectid)
+		return 1;
 
-	BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
+	/*
+	 * when reference count of tree block is 1, it won't increase
+	 * again. once full backref flag is set, we never clear it.
+	 */
+	if ((wc->stage == DROP_REFERENCE && wc->refs[level] != 1) ||
+	    (wc->stage == UPDATE_BACKREF && !(wc->flags[level] & flag))) {
+		BUG_ON(!path->locks[level]);
+		ret = btrfs_lookup_extent_info(trans, root,
+					       eb->start, eb->len,
+					       &wc->refs[level],
+					       &wc->flags[level]);
+		BUG_ON(ret);
+		BUG_ON(wc->refs[level] == 0);
+	}
 
-	while (*level >= 0) {
-		cur = path->nodes[*level];
-		if (*level == 0) {
-			ret = btrfs_drop_leaf_ref(trans, root, cur);
-			BUG_ON(ret);
-			clean_tree_block(trans, root, cur);
-			break;
-		}
-		if (path->slots[*level] >= btrfs_header_nritems(cur)) {
-			clean_tree_block(trans, root, cur);
-			break;
+	if (wc->stage == DROP_REFERENCE &&
+	    wc->update_ref && wc->refs[level] > 1) {
+		BUG_ON(eb == root->node);
+		BUG_ON(path->slots[level] > 0);
+		if (level == 0)
+			btrfs_item_key_to_cpu(eb, &key, path->slots[level]);
+		else
+			btrfs_node_key_to_cpu(eb, &key, path->slots[level]);
+		if (btrfs_header_owner(eb) == root->root_key.objectid &&
+		    btrfs_comp_cpu_keys(&key, &wc->update_progress) >= 0) {
+			wc->stage = UPDATE_BACKREF;
+			wc->shared_level = level;
 		}
+	}
 
-		bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
-		blocksize = btrfs_level_size(root, *level - 1);
-		ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
+	if (wc->stage == DROP_REFERENCE) {
+		if (wc->refs[level] > 1)
+			return 1;
 
-		next = read_tree_block(root, bytenr, blocksize, ptr_gen);
-		btrfs_tree_lock(next);
-		btrfs_set_lock_blocking(next);
+		if (path->locks[level] && !wc->keep_locks) {
+			btrfs_tree_unlock(eb);
+			path->locks[level] = 0;
+		}
+		return 0;
+	}
 
-		ret = btrfs_lookup_extent_info(trans, root, bytenr, blocksize,
-					       &refs, &flags);
+	/* wc->stage == UPDATE_BACKREF */
+	if (!(wc->flags[level] & flag)) {
+		BUG_ON(!path->locks[level]);
+		ret = btrfs_inc_ref(trans, root, eb, 1);
 		BUG_ON(ret);
-		if (refs > 1) {
-			parent = path->nodes[*level];
-			ret = btrfs_free_extent(trans, root, bytenr,
-						blocksize, parent->start,
-						btrfs_header_owner(parent),
-						*level - 1, 0);
+		ret = btrfs_dec_ref(trans, root, eb, 0);
+		BUG_ON(ret);
+		ret = btrfs_set_disk_extent_flags(trans, root, eb->start,
+						  eb->len, flag, 0);
+		BUG_ON(ret);
+		wc->flags[level] |= flag;
+	}
+
+	/*
+	 * the block is shared by multiple trees, so it's not good to
+	 * keep the tree lock
+	 */
+	if (path->locks[level] && level > 0) {
+		btrfs_tree_unlock(eb);
+		path->locks[level] = 0;
+	}
+	return 0;
+}
+
+/*
+ * hepler to process tree block while walking up the tree.
+ *
+ * when wc->stage == DROP_REFERENCE, this function drops
+ * reference count on the block.
+ *
+ * when wc->stage == UPDATE_BACKREF, this function changes
+ * wc->stage back to DROP_REFERENCE if we changed wc->stage
+ * to UPDATE_BACKREF previously while processing the block.
+ *
+ * NOTE: return value 1 means we should stop walking up.
+ */
+static noinline int walk_up_proc(struct btrfs_trans_handle *trans,
+				 struct btrfs_root *root,
+				 struct btrfs_path *path,
+				 struct walk_control *wc)
+{
+	int ret = 0;
+	int level = wc->level;
+	struct extent_buffer *eb = path->nodes[level];
+	u64 parent = 0;
+
+	if (wc->stage == UPDATE_BACKREF) {
+		BUG_ON(wc->shared_level < level);
+		if (level < wc->shared_level)
+			goto out;
+
+		BUG_ON(wc->refs[level] <= 1);
+		ret = find_next_key(path, level + 1, &wc->update_progress);
+		if (ret > 0)
+			wc->update_ref = 0;
+
+		wc->stage = DROP_REFERENCE;
+		wc->shared_level = -1;
+		path->slots[level] = 0;
+
+		/*
+		 * check reference count again if the block isn't locked.
+		 * we should start walking down the tree again if reference
+		 * count is one.
+		 */
+		if (!path->locks[level]) {
+			BUG_ON(level == 0);
+			btrfs_tree_lock(eb);
+			btrfs_set_lock_blocking(eb);
+			path->locks[level] = 1;
+
+			ret = btrfs_lookup_extent_info(trans, root,
+						       eb->start, eb->len,
+						       &wc->refs[level],
+						       &wc->flags[level]);
 			BUG_ON(ret);
-			path->slots[*level]++;
-			btrfs_tree_unlock(next);
-			free_extent_buffer(next);
-			continue;
+			BUG_ON(wc->refs[level] == 0);
+			if (wc->refs[level] == 1) {
+				btrfs_tree_unlock(eb);
+				path->locks[level] = 0;
+				return 1;
+			}
+		} else {
+			BUG_ON(level != 0);
 		}
+	}
 
-		BUG_ON(!(flags & BTRFS_BLOCK_FLAG_FULL_BACKREF));
+	/* wc->stage == DROP_REFERENCE */
+	BUG_ON(wc->refs[level] > 1 && !path->locks[level]);
 
-		*level = btrfs_header_level(next);
-		path->nodes[*level] = next;
-		path->slots[*level] = 0;
-		path->locks[*level] = 1;
-		cond_resched();
+	if (wc->refs[level] == 1) {
+		if (level == 0) {
+			if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
+				ret = btrfs_dec_ref(trans, root, eb, 1);
+			else
+				ret = btrfs_dec_ref(trans, root, eb, 0);
+			BUG_ON(ret);
+		}
+		/* make block locked assertion in clean_tree_block happy */
+		if (!path->locks[level] &&
+		    btrfs_header_generation(eb) == trans->transid) {
+			btrfs_tree_lock(eb);
+			btrfs_set_lock_blocking(eb);
+			path->locks[level] = 1;
+		}
+		clean_tree_block(trans, root, eb);
+	}
+
+	if (eb == root->node) {
+		if (wc->flags[level] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
+			parent = eb->start;
+		else
+			BUG_ON(root->root_key.objectid !=
+			       btrfs_header_owner(eb));
+	} else {
+		if (wc->flags[level + 1] & BTRFS_BLOCK_FLAG_FULL_BACKREF)
+			parent = path->nodes[level + 1]->start;
+		else
+			BUG_ON(root->root_key.objectid !=
+			       btrfs_header_owner(path->nodes[level + 1]));
 	}
-out:
-	if (path->nodes[*level] == root->node)
-		parent = path->nodes[*level];
-	else
-		parent = path->nodes[*level + 1];
-	bytenr = path->nodes[*level]->start;
-	blocksize = path->nodes[*level]->len;
 
-	ret = btrfs_free_extent(trans, root, bytenr, blocksize, parent->start,
-				btrfs_header_owner(parent), *level, 0);
+	ret = btrfs_free_extent(trans, root, eb->start, eb->len, parent,
+				root->root_key.objectid, level, 0);
 	BUG_ON(ret);
+out:
+	wc->refs[level] = 0;
+	wc->flags[level] = 0;
+	return ret;
+}
+
+static noinline int walk_down_tree(struct btrfs_trans_handle *trans,
+				   struct btrfs_root *root,
+				   struct btrfs_path *path,
+				   struct walk_control *wc)
+{
+	struct extent_buffer *next;
+	struct extent_buffer *cur;
+	u64 bytenr;
+	u64 ptr_gen;
+	u32 blocksize;
+	int level = wc->level;
+	int ret;
+
+	while (level >= 0) {
+		cur = path->nodes[level];
+		BUG_ON(path->slots[level] >= btrfs_header_nritems(cur));
 
-	if (path->locks[*level]) {
-		btrfs_tree_unlock(path->nodes[*level]);
-		path->locks[*level] = 0;
+		ret = walk_down_proc(trans, root, path, wc);
+		if (ret > 0)
+			break;
+
+		if (level == 0)
+			break;
+
+		bytenr = btrfs_node_blockptr(cur, path->slots[level]);
+		blocksize = btrfs_level_size(root, level - 1);
+		ptr_gen = btrfs_node_ptr_generation(cur, path->slots[level]);
+
+		next = read_tree_block(root, bytenr, blocksize, ptr_gen);
+		btrfs_tree_lock(next);
+		btrfs_set_lock_blocking(next);
+
+		level--;
+		BUG_ON(level != btrfs_header_level(next));
+		path->nodes[level] = next;
+		path->slots[level] = 0;
+		path->locks[level] = 1;
+		wc->level = level;
 	}
-	free_extent_buffer(path->nodes[*level]);
-	path->nodes[*level] = NULL;
-	*level += 1;
-	cond_resched();
 	return 0;
 }
 
-/*
- * helper for dropping snapshots.  This walks back up the tree in the path
- * to find the first node higher up where we haven't yet gone through
- * all the slots
- */
 static noinline int walk_up_tree(struct btrfs_trans_handle *trans,
 				 struct btrfs_root *root,
 				 struct btrfs_path *path,
-				 int *level, int max_level)
+				 struct walk_control *wc, int max_level)
 {
-	struct btrfs_root_item *root_item = &root->root_item;
-	int i;
-	int slot;
+	int level = wc->level;
 	int ret;
 
-	for (i = *level; i < max_level && path->nodes[i]; i++) {
-		slot = path->slots[i];
-		if (slot + 1 < btrfs_header_nritems(path->nodes[i])) {
-			/*
-			 * there is more work to do in this level.
-			 * Update the drop_progress marker to reflect
-			 * the work we've done so far, and then bump
-			 * the slot number
-			 */
-			path->slots[i]++;
-			WARN_ON(*level == 0);
-			if (max_level == BTRFS_MAX_LEVEL) {
-				btrfs_node_key(path->nodes[i],
-					       &root_item->drop_progress,
-					       path->slots[i]);
-				root_item->drop_level = i;
-			}
-			*level = i;
+	path->slots[level] = btrfs_header_nritems(path->nodes[level]);
+	while (level < max_level && path->nodes[level]) {
+		wc->level = level;
+		if (path->slots[level] + 1 <
+		    btrfs_header_nritems(path->nodes[level])) {
+			path->slots[level]++;
 			return 0;
 		} else {
-			struct extent_buffer *parent;
-
-			/*
-			 * this whole node is done, free our reference
-			 * on it and go up one level
-			 */
-			if (path->nodes[*level] == root->node)
-				parent = path->nodes[*level];
-			else
-				parent = path->nodes[*level + 1];
+			ret = walk_up_proc(trans, root, path, wc);
+			if (ret > 0)
+				return 0;
 
-			clean_tree_block(trans, root, path->nodes[i]);
-			ret = btrfs_free_extent(trans, root,
-						path->nodes[i]->start,
-						path->nodes[i]->len,
-						parent->start,
-						btrfs_header_owner(parent),
-						*level, 0);
-			BUG_ON(ret);
-			if (path->locks[*level]) {
-				btrfs_tree_unlock(path->nodes[i]);
-				path->locks[i] = 0;
+			if (path->locks[level]) {
+				btrfs_tree_unlock(path->nodes[level]);
+				path->locks[level] = 0;
 			}
-			free_extent_buffer(path->nodes[i]);
-			path->nodes[i] = NULL;
-			*level = i + 1;
+			free_extent_buffer(path->nodes[level]);
+			path->nodes[level] = NULL;
+			level++;
 		}
 	}
 	return 1;
 }
 
 /*
- * drop the reference count on the tree rooted at 'snap'.  This traverses
- * the tree freeing any blocks that have a ref count of zero after being
- * decremented.
+ * drop a subvolume tree.
+ *
+ * this function traverses the tree freeing any blocks that only
+ * referenced by the tree.
+ *
+ * when a shared tree block is found. this function decreases its
+ * reference count by one. if update_ref is true, this function
+ * also make sure backrefs for the shared block and all lower level
+ * blocks are properly updated.
  */
-int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
-			*root)
+int btrfs_drop_snapshot(struct btrfs_root *root, int update_ref)
 {
-	int ret = 0;
-	int wret;
-	int level;
 	struct btrfs_path *path;
-	int update_count;
+	struct btrfs_trans_handle *trans;
+	struct btrfs_root *tree_root = root->fs_info->tree_root;
 	struct btrfs_root_item *root_item = &root->root_item;
+	struct walk_control *wc;
+	struct btrfs_key key;
+	int err = 0;
+	int ret;
+	int level;
 
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
 
-	level = btrfs_header_level(root->node);
+	wc = kzalloc(sizeof(*wc), GFP_NOFS);
+	BUG_ON(!wc);
+
+	trans = btrfs_start_transaction(tree_root, 1);
+
 	if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
+		level = btrfs_header_level(root->node);
 		path->nodes[level] = btrfs_lock_root_node(root);
 		btrfs_set_lock_blocking(path->nodes[level]);
 		path->slots[level] = 0;
 		path->locks[level] = 1;
+		memset(&wc->update_progress, 0,
+		       sizeof(wc->update_progress));
 	} else {
-		struct btrfs_key key;
-		struct btrfs_disk_key found_key;
-		struct extent_buffer *node;
-
 		btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
+		memcpy(&wc->update_progress, &key,
+		       sizeof(wc->update_progress));
+
 		level = root_item->drop_level;
+		BUG_ON(level == 0);
 		path->lowest_level = level;
-		wret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
-		if (wret < 0) {
-			ret = wret;
+		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+		path->lowest_level = 0;
+		if (ret < 0) {
+			err = ret;
 			goto out;
 		}
-		node = path->nodes[level];
-		btrfs_node_key(node, &found_key, path->slots[level]);
-		WARN_ON(memcmp(&found_key, &root_item->drop_progress,
-			       sizeof(found_key)));
+		btrfs_node_key_to_cpu(path->nodes[level], &key,
+				      path->slots[level]);
+		WARN_ON(memcmp(&key, &wc->update_progress, sizeof(key)));
+
 		/*
 		 * unlock our path, this is safe because only this
 		 * function is allowed to delete this snapshot
 		 */
 		btrfs_unlock_up_safe(path, 0);
+
+		level = btrfs_header_level(root->node);
+		while (1) {
+			btrfs_tree_lock(path->nodes[level]);
+			btrfs_set_lock_blocking(path->nodes[level]);
+
+			ret = btrfs_lookup_extent_info(trans, root,
+						path->nodes[level]->start,
+						path->nodes[level]->len,
+						&wc->refs[level],
+						&wc->flags[level]);
+			BUG_ON(ret);
+			BUG_ON(wc->refs[level] == 0);
+
+			if (level == root_item->drop_level)
+				break;
+
+			btrfs_tree_unlock(path->nodes[level]);
+			WARN_ON(wc->refs[level] != 1);
+			level--;
+		}
 	}
+
+	wc->level = level;
+	wc->shared_level = -1;
+	wc->stage = DROP_REFERENCE;
+	wc->update_ref = update_ref;
+	wc->keep_locks = 0;
+
 	while (1) {
-		unsigned long update;
-		wret = walk_down_tree(trans, root, path, &level);
-		if (wret > 0)
+		ret = walk_down_tree(trans, root, path, wc);
+		if (ret < 0) {
+			err = ret;
 			break;
-		if (wret < 0)
-			ret = wret;
+		}
 
-		wret = walk_up_tree(trans, root, path, &level,
-				    BTRFS_MAX_LEVEL);
-		if (wret > 0)
+		ret = walk_up_tree(trans, root, path, wc, BTRFS_MAX_LEVEL);
+		if (ret < 0) {
+			err = ret;
 			break;
-		if (wret < 0)
-			ret = wret;
-		if (trans->transaction->in_commit ||
-		    trans->transaction->delayed_refs.flushing) {
-			ret = -EAGAIN;
+		}
+
+		if (ret > 0) {
+			BUG_ON(wc->stage != DROP_REFERENCE);
 			break;
 		}
-		for (update_count = 0; update_count < 16; update_count++) {
+
+		if (wc->stage == DROP_REFERENCE) {
+			level = wc->level;
+			btrfs_node_key(path->nodes[level],
+				       &root_item->drop_progress,
+				       path->slots[level]);
+			root_item->drop_level = level;
+		}
+
+		BUG_ON(wc->level == 0);
+		if (trans->transaction->in_commit ||
+		    trans->transaction->delayed_refs.flushing) {
+			ret = btrfs_update_root(trans, tree_root,
+						&root->root_key,
+						root_item);
+			BUG_ON(ret);
+
+			btrfs_end_transaction(trans, tree_root);
+			trans = btrfs_start_transaction(tree_root, 1);
+		} else {
+			unsigned long update;
 			update = trans->delayed_ref_updates;
 			trans->delayed_ref_updates = 0;
 			if (update)
-				btrfs_run_delayed_refs(trans, root, update);
-			else
-				break;
+				btrfs_run_delayed_refs(trans, tree_root,
+						       update);
 		}
 	}
+	btrfs_release_path(root, path);
+	BUG_ON(err);
+
+	ret = btrfs_del_root(trans, tree_root, &root->root_key);
+	BUG_ON(ret);
+
+	free_extent_buffer(root->node);
+	free_extent_buffer(root->commit_root);
+	kfree(root);
 out:
+	btrfs_end_transaction(trans, tree_root);
+	kfree(wc);
 	btrfs_free_path(path);
-	return ret;
+	return err;
 }
 
+/*
+ * drop subtree rooted at tree block 'node'.
+ *
+ * NOTE: this function will unlock and release tree block 'node'
+ */
 int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
 			struct btrfs_root *root,
 			struct extent_buffer *node,
 			struct extent_buffer *parent)
 {
 	struct btrfs_path *path;
+	struct walk_control *wc;
 	int level;
 	int parent_level;
 	int ret = 0;
 	int wret;
 
+	BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
+
 	path = btrfs_alloc_path();
 	BUG_ON(!path);
 
+	wc = kzalloc(sizeof(*wc), GFP_NOFS);
+	BUG_ON(!wc);
+
 	btrfs_assert_tree_locked(parent);
 	parent_level = btrfs_header_level(parent);
 	extent_buffer_get(parent);
@@ -4817,24 +5024,33 @@ int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
 
 	btrfs_assert_tree_locked(node);
 	level = btrfs_header_level(node);
-	extent_buffer_get(node);
 	path->nodes[level] = node;
 	path->slots[level] = 0;
+	path->locks[level] = 1;
+
+	wc->refs[parent_level] = 1;
+	wc->flags[parent_level] = BTRFS_BLOCK_FLAG_FULL_BACKREF;
+	wc->level = level;
+	wc->shared_level = -1;
+	wc->stage = DROP_REFERENCE;
+	wc->update_ref = 0;
+	wc->keep_locks = 1;
 
 	while (1) {
-		wret = walk_down_tree(trans, root, path, &level);
-		if (wret < 0)
+		wret = walk_down_tree(trans, root, path, wc);
+		if (wret < 0) {
 			ret = wret;
-		if (wret != 0)
 			break;
+		}
 
-		wret = walk_up_tree(trans, root, path, &level, parent_level);
+		wret = walk_up_tree(trans, root, path, wc, parent_level);
 		if (wret < 0)
 			ret = wret;
 		if (wret != 0)
 			break;
 	}
 
+	kfree(wc);
 	btrfs_free_path(path);
 	return ret;
 }

fs/btrfs/file.c

@@ -151,7 +151,10 @@ static noinline int dirty_and_release_pages(struct btrfs_trans_handle *trans,
 	}
 	if (end_pos > isize) {
 		i_size_write(inode, end_pos);
-		btrfs_update_inode(trans, root, inode);
+		/* we've only changed i_size in ram, and we haven't updated
+		 * the disk i_size.  There is no need to log the inode
+		 * at this time.
+		 */
 	}
 	err = btrfs_end_transaction(trans, root);
 out_unlock:

fs/btrfs/inode.c

@@ -3580,12 +3580,6 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 		owner = 1;
 	BTRFS_I(inode)->block_group =
 			btrfs_find_block_group(root, 0, alloc_hint, owner);
-	if ((mode & S_IFREG)) {
-		if (btrfs_test_opt(root, NODATASUM))
-			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
-		if (btrfs_test_opt(root, NODATACOW))
-			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
-	}
 
 	key[0].objectid = objectid;
 	btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
@@ -3640,6 +3634,13 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
 
 	btrfs_inherit_iflags(inode, dir);
 
+	if ((mode & S_IFREG)) {
+		if (btrfs_test_opt(root, NODATASUM))
+			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
+		if (btrfs_test_opt(root, NODATACOW))
+			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
+	}
+
 	insert_inode_hash(inode);
 	inode_tree_add(inode);
 	return inode;
@@ -5082,6 +5083,7 @@ static long btrfs_fallocate(struct inode *inode, int mode,
 	u64 mask = BTRFS_I(inode)->root->sectorsize - 1;
 	struct extent_map *em;
 	struct btrfs_trans_handle *trans;
+	struct btrfs_root *root;
 	int ret;
 
 	alloc_start = offset & ~mask;
@@ -5100,6 +5102,13 @@ static long btrfs_fallocate(struct inode *inode, int mode,
 			goto out;
 	}
 
+	root = BTRFS_I(inode)->root;
+
+	ret = btrfs_check_data_free_space(root, inode,
+					  alloc_end - alloc_start);
+	if (ret)
+		goto out;
+
 	locked_end = alloc_end - 1;
 	while (1) {
 		struct btrfs_ordered_extent *ordered;
@@ -5107,7 +5116,7 @@ static long btrfs_fallocate(struct inode *inode, int mode,
 		trans = btrfs_start_transaction(BTRFS_I(inode)->root, 1);
 		if (!trans) {
 			ret = -EIO;
-			goto out;
+			goto out_free;
 		}
 
 		/* the extent lock is ordered inside the running
@@ -5168,6 +5177,8 @@ static long btrfs_fallocate(struct inode *inode, int mode,
 		      GFP_NOFS);
 
 	btrfs_end_transaction(trans, BTRFS_I(inode)->root);
+out_free:
+	btrfs_free_reserved_data_space(root, inode, alloc_end - alloc_start);
 out:
 	mutex_unlock(&inode->i_mutex);
 	return ret;

fs/btrfs/ioctl.c

@@ -1028,7 +1028,8 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
 						struct btrfs_file_extent_item);
 			comp = btrfs_file_extent_compression(leaf, extent);
 			type = btrfs_file_extent_type(leaf, extent);
-			if (type == BTRFS_FILE_EXTENT_REG) {
+			if (type == BTRFS_FILE_EXTENT_REG ||
+			    type == BTRFS_FILE_EXTENT_PREALLOC) {
 				disko = btrfs_file_extent_disk_bytenr(leaf,
 								      extent);
 				diskl = btrfs_file_extent_disk_num_bytes(leaf,
@@ -1051,7 +1052,8 @@ static long btrfs_ioctl_clone(struct file *file, unsigned long srcfd,
 			new_key.objectid = inode->i_ino;
 			new_key.offset = key.offset + destoff - off;
 
-			if (type == BTRFS_FILE_EXTENT_REG) {
+			if (type == BTRFS_FILE_EXTENT_REG ||
+			    type == BTRFS_FILE_EXTENT_PREALLOC) {
 				ret = btrfs_insert_empty_item(trans, root, path,
 							      &new_key, size);
 				if (ret)

fs/btrfs/relocation.c

@@ -1788,7 +1788,7 @@ static void merge_func(struct btrfs_work *work)
 		btrfs_end_transaction(trans, root);
 	}
 
-	btrfs_drop_dead_root(reloc_root);
+	btrfs_drop_snapshot(reloc_root, 0);
 
 	if (atomic_dec_and_test(async->num_pending))
 		complete(async->done);
@@ -2075,9 +2075,6 @@ static int do_relocation(struct btrfs_trans_handle *trans,
 
 			ret = btrfs_drop_subtree(trans, root, eb, upper->eb);
 			BUG_ON(ret);
-
-			btrfs_tree_unlock(eb);
-			free_extent_buffer(eb);
 		}
 		if (!lowest) {
 			btrfs_tree_unlock(upper->eb);

fs/btrfs/transaction.c

@@ -593,6 +593,7 @@ int btrfs_defrag_root(struct btrfs_root *root, int cacheonly)
 	return 0;
 }
 
+#if 0
 /*
  * when dropping snapshots, we generate a ton of delayed refs, and it makes
  * sense not to join the transaction while it is trying to flush the current
@@ -681,6 +682,7 @@ int btrfs_drop_dead_root(struct btrfs_root *root)
 	btrfs_btree_balance_dirty(tree_root, nr);
 	return ret;
 }
+#endif
 
 /*
  * new snapshots need to be created at a very specific time in the
@@ -1081,7 +1083,7 @@ int btrfs_clean_old_snapshots(struct btrfs_root *root)
 	while (!list_empty(&list)) {
 		root = list_entry(list.next, struct btrfs_root, root_list);
 		list_del_init(&root->root_list);
-		btrfs_drop_dead_root(root);
+		btrfs_drop_snapshot(root, 0);
 	}
 	return 0;
 }

fs/notify/inotify/inotify_user.c

@@ -296,12 +296,15 @@ static int inotify_fasync(int fd, struct file *file, int on)
 static int inotify_release(struct inode *ignored, struct file *file)
 {
 	struct fsnotify_group *group = file->private_data;
+	struct user_struct *user = group->inotify_data.user;
 
 	fsnotify_clear_marks_by_group(group);
 
 	/* free this group, matching get was inotify_init->fsnotify_obtain_group */
 	fsnotify_put_group(group);
 
+	atomic_dec(&user->inotify_devs);
+
 	return 0;
 }