2006-11-08 00:22:03

by Jeremy Fitzhardinge

Subject: [PATCH] Fix kunmap_atomic's use of kpte_clear_flush()

kunmap_atomic() will call kpte_clear_flush with vaddr/ptep arguments
which don't correspond if the vaddr is just a normal lowmem address
(ie, not in the KMAP area). This patch makes sure that the pte is
only cleared if kmap area was actually used for the mapping.

Signed-off-by: Jeremy Fitzhardinge <[email protected]>
Signed-off-by: Rusty Russell <[email protected]>
Cc: Zachary Amsden <[email protected]>

===================================================================
--- a/arch/i386/mm/highmem.c
+++ b/arch/i386/mm/highmem.c
@@ -56,22 +56,20 @@ void kunmap_atomic(void *kvaddr, enum km
unsigned long vaddr = (unsigned long) kvaddr & PAGE_MASK;
enum fixed_addresses idx = type + KM_TYPE_NR*smp_processor_id();

-#ifdef CONFIG_DEBUG_HIGHMEM
- if (vaddr >= PAGE_OFFSET && vaddr < (unsigned long)high_memory) {
- pagefault_enable();
- return;
- }
-
- if (vaddr != __fix_to_virt(FIX_KMAP_BEGIN+idx))
- BUG();
-#endif
/*
* Force other mappings to Oops if they'll try to access this pte
* without first remap it. Keeping stale mappings around is a bad idea
* also, in case the page changes cacheability attributes or becomes
* a protected page in a hypervisor.
*/
- kpte_clear_flush(kmap_pte-idx, vaddr);
+ if (vaddr == __fix_to_virt(FIX_KMAP_BEGIN+idx))
+ kpte_clear_flush(kmap_pte-idx, vaddr);
+ else {
+#ifdef CONFIG_DEBUG_HIGHMEM
+ BUG_ON(vaddr < PAGE_OFFSET);
+ BUG_ON(vaddr >= (unsigned long)high_memory);
+#endif
+ }

pagefault_enable();
}

2006-11-08 00:42:55

by Rusty Russell

Subject: Re: [PATCH] Fix kunmap_atomic's use of kpte_clear_flush()

On Tue, 2006-11-07 at 16:22 -0800, Jeremy Fitzhardinge wrote:
> kunmap_atomic() will call kpte_clear_flush with vaddr/ptep arguments
> which don't correspond if the vaddr is just a normal lowmem address
> (ie, not in the KMAP area). This patch makes sure that the pte is
> only cleared if kmap area was actually used for the mapping.

Or in other words, if kmap_atomic() does nothing, kunmap_atomic() should
do nothing.

Rusty.

2006-11-08 00:42:13

by Zachary Amsden

Subject: Re: [PATCH] Fix kunmap_atomic's use of kpte_clear_flush()

Jeremy Fitzhardinge wrote:
> kunmap_atomic() will call kpte_clear_flush with vaddr/ptep arguments
> which don't correspond if the vaddr is just a normal lowmem address
> (ie, not in the KMAP area). This patch makes sure that the pte is
> only cleared if kmap area was actually used for the mapping.
>
> Signed-off-by: Jeremy Fitzhardinge <[email protected]>
> Signed-off-by: Rusty Russell <[email protected]>
> Cc: Zachary Amsden <[email protected]>

Ack.

2006-11-08 01:10:52

by Jeremy Fitzhardinge

Subject: Re: [PATCH] Fix kunmap_atomic's use of kpte_clear_flush()

Rusty Russell wrote:
> On Tue, 2006-11-07 at 16:22 -0800, Jeremy Fitzhardinge wrote:
>
>> kunmap_atomic() will call kpte_clear_flush with vaddr/ptep arguments
>> which don't correspond if the vaddr is just a normal lowmem address
>> (ie, not in the KMAP area). This patch makes sure that the pte is
>> only cleared if kmap area was actually used for the mapping.
>>
>
> Or in other words, if kmap_atomic() does nothing, kunmap_atomic() should
> do nothing.
>

Sure. I guess there's no particular reason why kmap_atomic() couldn't
always map, but I guess avoid the pagetable updates is worthwhile.

J