[PATCH] Fix PMU + soft interrupt disable bug

[PATCH] Fix PMU + soft interrupt disable bug

[Anton Blanchard]

Since the PMU is an NMI now, it can come at any time we are only soft
disabled. We must hard disable around the two places we allow the kernel
stack SLB and r1 to go out of sync. Otherwise the PMU exception can
force a kernel stack SLB into another slot.

Signed-off-by: Anton Blanchard <anton@samba.org>
---

Index: kernel/arch/powerpc/mm/slb.c
===================================================================
--- kernel.orig/arch/powerpc/mm/slb.c	2007-11-05 11:22:03.000000000 -0600
+++ kernel/arch/powerpc/mm/slb.c	2007-11-05 18:09:45.000000000 -0600
@@ -113,6 +113,12 @@ void slb_flush_and_rebolt(void)
 		slb_shadow_update(get_paca()->kstack, lflags, 2);
 	}
 
+	/*
+	 * We can't take a PMU exception in the following code, so hard
+	 * disable interrupts.
+	 */
+	hard_irq_disable();
+
 	/* We need to do this all in asm, so we're sure we don't touch
 	 * the stack between the slbia and rebolting it. */
 	asm volatile("isync\n"
Index: kernel/arch/powerpc/kernel/process.c
===================================================================
--- kernel.orig/arch/powerpc/kernel/process.c	2007-11-05 18:10:44.000000000 -0600
+++ kernel/arch/powerpc/kernel/process.c	2007-11-05 18:10:46.000000000 -0600
@@ -331,6 +331,12 @@ struct task_struct *__switch_to(struct t
 	account_process_vtime(current);
 	calculate_steal_time();
 
+	/*
+	 * We can't take a PMU exception inside _switch() since there is a
+	 * window where the kernel stack SLB and the kernel stack are out
+	 * of sync. Hard disable here.
+	 */
+	hard_irq_disable();
 	last = _switch(old_thread, new_thread);
 
 	local_irq_restore(flags);

Re: [PATCH] Fix PMU + soft interrupt disable bug

[Benjamin Herrenschmidt]

On Sun, 2008-03-16 at 23:27 -0500, Anton Blanchard wrote:
> Since the PMU is an NMI now, it can come at any time we are only soft
> disabled. We must hard disable around the two places we allow the kernel
> stack SLB and r1 to go out of sync. Otherwise the PMU exception can
> force a kernel stack SLB into another slot.
> 
> Signed-off-by: Anton Blanchard <anton@samba.org>

Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
---

> ---
> 
> Index: kernel/arch/powerpc/mm/slb.c
> ===================================================================
> --- kernel.orig/arch/powerpc/mm/slb.c	2007-11-05 11:22:03.000000000 -0600
> +++ kernel/arch/powerpc/mm/slb.c	2007-11-05 18:09:45.000000000 -0600
> @@ -113,6 +113,12 @@ void slb_flush_and_rebolt(void)
>  		slb_shadow_update(get_paca()->kstack, lflags, 2);
>  	}
>  
> +	/*
> +	 * We can't take a PMU exception in the following code, so hard
> +	 * disable interrupts.
> +	 */
> +	hard_irq_disable();
> +
>  	/* We need to do this all in asm, so we're sure we don't touch
>  	 * the stack between the slbia and rebolting it. */
>  	asm volatile("isync\n"
> Index: kernel/arch/powerpc/kernel/process.c
> ===================================================================
> --- kernel.orig/arch/powerpc/kernel/process.c	2007-11-05 18:10:44.000000000 -0600
> +++ kernel/arch/powerpc/kernel/process.c	2007-11-05 18:10:46.000000000 -0600
> @@ -331,6 +331,12 @@ struct task_struct *__switch_to(struct t
>  	account_process_vtime(current);
>  	calculate_steal_time();
>  
> +	/*
> +	 * We can't take a PMU exception inside _switch() since there is a
> +	 * window where the kernel stack SLB and the kernel stack are out
> +	 * of sync. Hard disable here.
> +	 */
> +	hard_irq_disable();
>  	last = _switch(old_thread, new_thread);
>  
>  	local_irq_restore(flags);
> 
> 
> _______________________________________________
> Linuxppc-dev mailing list
> Linuxppc-dev@ozlabs.org
> https://ozlabs.org/mailman/listinfo/linuxppc-dev

Re: [PATCH] Fix PMU + soft interrupt disable bug

[Olof Johansson]

On Sun, Mar 16, 2008 at 11:27:09PM -0500, Anton Blanchard wrote:
> 
> Since the PMU is an NMI now, it can come at any time we are only soft
> disabled. We must hard disable around the two places we allow the kernel
> stack SLB and r1 to go out of sync. Otherwise the PMU exception can
> force a kernel stack SLB into another slot.
> 
> Signed-off-by: Anton Blanchard <anton@samba.org>

Acked-by: Olof Johansson <olof@lixom.net>

This fixes some problems we've seen with multi-slb hits when running
oprofile.