x86/asm/entry: Replace this_cpu_sp0() with current_top_of_stack() and fix it on x86_32

I broke 32-bit kernels.  The implementation of sp0 was correct
as far as I can tell, but sp0 was much weirder on x86_32 than I
realized.  It has the following issues:

 - Init's sp0 is inconsistent with everything else's: non-init tasks
   are offset by 8 bytes.  (I have no idea why, and the comment is unhelpful.)

 - vm86 does crazy things to sp0.

Fix it up by replacing this_cpu_sp0() with
current_top_of_stack() and using a new percpu variable to track
the top of the stack on x86_32.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 75182b1632a8 ("x86/asm/entry: Switch all C consumers of kernel_stack to this_cpu_sp0()")
Link: http://lkml.kernel.org/r/d09dbe270883433776e0cbee3c7079433349e96d.1425692936.git.luto@amacapital.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>

arch/x86/include/asm/processor.h

@@ -284,6 +284,10 @@ struct tss_struct {
 
 DECLARE_PER_CPU_SHARED_ALIGNED(struct tss_struct, cpu_tss);
 
+#ifdef CONFIG_X86_32
+DECLARE_PER_CPU(unsigned long, cpu_current_top_of_stack);
+#endif
+
 /*
  * Save the original ist values for checking stack pointers during debugging
  */
@@ -564,9 +568,14 @@ static inline void native_swapgs(void)
 #endif
 }
 
-static inline unsigned long this_cpu_sp0(void)
+static inline unsigned long current_top_of_stack(void)
 {
+#ifdef CONFIG_X86_64
 	return this_cpu_read_stable(cpu_tss.x86_tss.sp0);
+#else
+	/* sp0 on x86_32 is special in and around vm86 mode. */
+	return this_cpu_read_stable(cpu_current_top_of_stack);
+#endif
 }
 
 #ifdef CONFIG_PARAVIRT

arch/x86/include/asm/thread_info.h

@@ -158,9 +158,7 @@ DECLARE_PER_CPU(unsigned long, kernel_stack);
 
 static inline struct thread_info *current_thread_info(void)
 {
-	struct thread_info *ti;
-	ti = (void *)(this_cpu_sp0() - THREAD_SIZE);
-	return ti;
+	return (struct thread_info *)(current_top_of_stack() - THREAD_SIZE);
 }
 
 static inline unsigned long current_stack_pointer(void)

arch/x86/kernel/cpu/common.c

@@ -1130,8 +1130,8 @@ DEFINE_PER_CPU_FIRST(union irq_stack_union,
 		     irq_stack_union) __aligned(PAGE_SIZE) __visible;
 
 /*
- * The following four percpu variables are hot.  Align current_task to
- * cacheline size such that all four fall in the same cacheline.
+ * The following percpu variables are hot.  Align current_task to
+ * cacheline size such that they fall in the same cacheline.
  */
 DEFINE_PER_CPU(struct task_struct *, current_task) ____cacheline_aligned =
 	&init_task;
@@ -1226,6 +1226,15 @@ DEFINE_PER_CPU(int, __preempt_count) = INIT_PREEMPT_COUNT;
 EXPORT_PER_CPU_SYMBOL(__preempt_count);
 DEFINE_PER_CPU(struct task_struct *, fpu_owner_task);
 
+/*
+ * On x86_32, vm86 modifies tss.sp0, so sp0 isn't a reliable way to find
+ * the top of the kernel stack.  Use an extra percpu variable to track the
+ * top of the kernel stack directly.
+ */
+DEFINE_PER_CPU(unsigned long, cpu_current_top_of_stack) =
+	(unsigned long)&init_thread_union + THREAD_SIZE;
+EXPORT_PER_CPU_SYMBOL(cpu_current_top_of_stack);
+
 #ifdef CONFIG_CC_STACKPROTECTOR
 DEFINE_PER_CPU_ALIGNED(struct stack_canary, stack_canary);
 #endif

arch/x86/kernel/process_32.c

@@ -306,13 +306,16 @@ __switch_to(struct task_struct *prev_p, struct task_struct *next_p)
 	arch_end_context_switch(next_p);
 
 	/*
-	 * Reload esp0.  This changes current_thread_info().
+	 * Reload esp0, kernel_stack, and current_top_of_stack.  This changes
+	 * current_thread_info().
 	 */
 	load_sp0(tss, next);
-
 	this_cpu_write(kernel_stack,
-		  (unsigned long)task_stack_page(next_p) +
-		  THREAD_SIZE - KERNEL_STACK_OFFSET);
+		       (unsigned long)task_stack_page(next_p) +
+		       THREAD_SIZE - KERNEL_STACK_OFFSET);
+	this_cpu_write(cpu_current_top_of_stack,
+		       (unsigned long)task_stack_page(next_p) +
+		       THREAD_SIZE);
 
 	/*
 	 * Restore %gs if needed (which is common)

arch/x86/kernel/smpboot.c

@@ -806,6 +806,8 @@ static int do_boot_cpu(int apicid, int cpu, struct task_struct *idle)
 #ifdef CONFIG_X86_32
 	/* Stack for startup_32 can be just as for start_secondary onwards */
 	irq_ctx_init(cpu);
+	per_cpu(cpu_current_top_of_stack, cpu) =
+		(unsigned long)task_stack_page(idle) + THREAD_SIZE;
 #else
 	clear_tsk_thread_flag(idle, TIF_FORK);
 	initial_gs = per_cpu_offset(cpu);

arch/x86/kernel/traps.c

@@ -174,8 +174,8 @@ void ist_begin_non_atomic(struct pt_regs *regs)
 	 * will catch asm bugs and any attempt to use ist_preempt_enable
 	 * from double_fault.
 	 */
-	BUG_ON((unsigned long)(this_cpu_sp0() - current_stack_pointer()) >=
-	       THREAD_SIZE);
+	BUG_ON((unsigned long)(current_top_of_stack() -
+			       current_stack_pointer()) >= THREAD_SIZE);
 
 	preempt_count_sub(HARDIRQ_OFFSET);
 }