[PATCH] sched-fix-scheduling-latencies-in-mttrc reenables interrupts

[PATCH] sched-fix-scheduling-latencies-in-mttrc reenables interrupts

[Bernard Blackham]

[-- Attachment #1: Type: text/plain, Size: 443 bytes --]

Hi Ingo & Andrew,

The patch sched-fix-scheduling-latencies-in-mttr in recent -mm
kernels has the bad side-effect of re-enabling interrupts even if
they were disabled. This caused bugs in Software Suspend 2 which
reenabled MTRRs whilst interrupts were already disabled.

Attached is a replacement patch which uses spin_lock_irqsave instead
of spin_lock_irq.

Kind regards,

Bernard.

-- 
 Bernard Blackham <bernard at blackham dot com dot au>

[-- Attachment #2: sched-fix-scheduling-latencies-in-mttrc-good.patch --]
[-- Type: text/plain, Size: 1599 bytes --]

--- linux-2.6.10/arch/i386/kernel/cpu/mtrr/generic.c.orig	2004-10-19 05:54:32.000000000 +0800
+++ linux-2.6.10/arch/i386/kernel/cpu/mtrr/generic.c	2004-12-26 18:20:17.000000000 +0800
@@ -232,6 +232,7 @@ static unsigned long set_mtrr_state(u32 
 static unsigned long cr4 = 0;
 static u32 deftype_lo, deftype_hi;
 static spinlock_t set_atomicity_lock = SPIN_LOCK_UNLOCKED;
+static unsigned long sal_flags;
 
 static void prepare_set(void)
 {
@@ -240,11 +241,14 @@ static void prepare_set(void)
 	/*  Note that this is not ideal, since the cache is only flushed/disabled
 	   for this CPU while the MTRRs are changed, but changing this requires
 	   more invasive changes to the way the kernel boots  */
-	spin_lock(&set_atomicity_lock);
+	/*
+	 * Since we are disabling the cache dont allow any interrupts - they
+	 * would run extremely slow and would only increase the pain:
+	 */
+	spin_lock_irqsave(&set_atomicity_lock, sal_flags);
 
 	/*  Enter the no-fill (CD=1, NW=0) cache mode and flush caches. */
 	cr0 = read_cr0() | 0x40000000;	/* set CD flag */
-	wbinvd();
 	write_cr0(cr0);
 	wbinvd();
 
@@ -266,8 +270,7 @@ static void prepare_set(void)
 
 static void post_set(void)
 {
-	/*  Flush caches and TLBs  */
-	wbinvd();
+	/*  Flush TLBs (no need to flush caches - they are disabled)  */
 	__flush_tlb();
 
 	/* Intel (P6) standard MTRRs */
@@ -279,7 +282,7 @@ static void post_set(void)
 	/*  Restore value of CR4  */
 	if ( cpu_has_pge )
 		write_cr4(cr4);
-	spin_unlock(&set_atomicity_lock);
+	spin_unlock_irqrestore(&set_atomicity_lock, sal_flags);
 }
 
 static void generic_set_all(void)

Re: [PATCH] sched-fix-scheduling-latencies-in-mttrc reenables interrupts

[Andrew Morton]

Bernard Blackham <bernard@blackham.com.au> wrote:
>
> The patch sched-fix-scheduling-latencies-in-mttr in recent -mm
>  kernels has the bad side-effect of re-enabling interrupts even if
>  they were disabled. This caused bugs in Software Suspend 2 which
>  reenabled MTRRs whilst interrupts were already disabled.

OK.

>  Attached is a replacement patch which uses spin_lock_irqsave instead
>  of spin_lock_irq.

>  +static unsigned long sal_flags;
>   
>   static void prepare_set(void)
>   {
>  @@ -240,11 +241,14 @@ static void prepare_set(void)
>   	/*  Note that this is not ideal, since the cache is only flushed/disabled
>   	   for this CPU while the MTRRs are changed, but changing this requires
>   	   more invasive changes to the way the kernel boots  */
>  -	spin_lock(&set_atomicity_lock);
>  +	/*
>  +	 * Since we are disabling the cache dont allow any interrupts - they
>  +	 * would run extremely slow and would only increase the pain:
>  +	 */
>  +	spin_lock_irqsave(&set_atomicity_lock, sal_flags);

Not OK.  That global `sal_flags' is not protected by the spinlock because
spin_lock_irqsave() saves the flags before taking the lock.

How about this?


diff -puN arch/i386/kernel/cpu/mtrr/generic.c~sched-fix-scheduling-latencies-in-mttrc-reenables-interrupts arch/i386/kernel/cpu/mtrr/generic.c
--- 25/arch/i386/kernel/cpu/mtrr/generic.c~sched-fix-scheduling-latencies-in-mttrc-reenables-interrupts	2004-12-27 01:31:41.718983736 -0800
+++ 25-akpm/arch/i386/kernel/cpu/mtrr/generic.c	2004-12-27 01:34:15.180654016 -0800
@@ -233,6 +233,13 @@ static unsigned long cr4 = 0;
 static u32 deftype_lo, deftype_hi;
 static spinlock_t set_atomicity_lock = SPIN_LOCK_UNLOCKED;
 
+/*
+ * Since we are disabling the cache don't allow any interrupts - they
+ * would run extremely slow and would only increase the pain.  The caller must
+ * ensure that local interrupts are disabled and are reenabled after post_set()
+ * has been called.
+ */
+
 static void prepare_set(void)
 {
 	unsigned long cr0;
@@ -240,11 +247,8 @@ static void prepare_set(void)
 	/*  Note that this is not ideal, since the cache is only flushed/disabled
 	   for this CPU while the MTRRs are changed, but changing this requires
 	   more invasive changes to the way the kernel boots  */
-	/*
-	 * Since we are disabling the cache dont allow any interrupts - they
-	 * would run extremely slow and would only increase the pain:
-	 */
-	spin_lock_irq(&set_atomicity_lock);
+
+	spin_lock(&set_atomicity_lock);
 
 	/*  Enter the no-fill (CD=1, NW=0) cache mode and flush caches. */
 	cr0 = read_cr0() | 0x40000000;	/* set CD flag */
@@ -281,19 +285,22 @@ static void post_set(void)
 	/*  Restore value of CR4  */
 	if ( cpu_has_pge )
 		write_cr4(cr4);
-	spin_unlock_irq(&set_atomicity_lock);
+	spin_unlock(&set_atomicity_lock);
 }
 
 static void generic_set_all(void)
 {
 	unsigned long mask, count;
+	unsigned long flags;
 
+	local_irq_save(flags);
 	prepare_set();
 
 	/* Actually set the state */
 	mask = set_mtrr_state(deftype_lo,deftype_hi);
 
 	post_set();
+	local_irq_restore(flags);
 
 	/*  Use the atomic bitops to update the global mask  */
 	for (count = 0; count < sizeof mask * 8; ++count) {
@@ -316,6 +323,9 @@ static void generic_set_mtrr(unsigned in
     [RETURNS] Nothing.
 */
 {
+	unsigned long flags;
+
+	local_irq_save(flags);
 	prepare_set();
 
 	if (size == 0) {
@@ -330,6 +340,7 @@ static void generic_set_mtrr(unsigned in
 	}
 
 	post_set();
+	local_irq_restore(flags);
 }
 
 int generic_validate_add_page(unsigned long base, unsigned long size, unsigned int type)
_

Re: [PATCH] sched-fix-scheduling-latencies-in-mttrc reenables interrupts

[Bernard Blackham]

On Mon, Dec 27, 2004 at 01:36:53AM -0800, Andrew Morton wrote:
> Not OK.  That global `sal_flags' is not protected by the spinlock
> because spin_lock_irqsave() saves the flags before taking the
> lock.

Ahh, d'oh.

> How about this?
> 
> diff -puN arch/i386/kernel/cpu/mtrr/generic.c~sched-fix-scheduling-latencies-in-mttrc-reenables-interrupts arch/i386/kernel/cpu/mtrr/generic.c

Won't object here.

Thanks,

Bernard.

-- 
 Bernard Blackham <bernard at blackham dot com dot au>