[PATCH 13/19 v2] xtensa/mm/fault.c: Port OOM changes to do_page_fault

[PATCH 13/19 v2] xtensa/mm/fault.c: Port OOM changes to do_page_fault

[Kautuk Consul]

Commit d065bd810b6deb67d4897a14bfe21f8eb526ba99
(mm: retry page fault when blocking on disk transfer) and
commit 37b23e0525d393d48a7d59f870b3bc061a30ccdb
(x86,mm: make pagefault killable)

The above commits introduced changes into the x86 pagefault handler
for making the page fault handler retryable as well as killable.

These changes reduce the mmap_sem hold time, which is crucial
during OOM killer invocation.

Port these changes to xtensa.

Signed-off-by: Kautuk Consul <consul.kautuk@gmail.com>
---
 arch/xtensa/mm/fault.c |   29 ++++++++++++++++++++++++-----
 1 files changed, 24 insertions(+), 5 deletions(-)

diff --git a/arch/xtensa/mm/fault.c b/arch/xtensa/mm/fault.c
index b17885a..5a74c53 100644
--- a/arch/xtensa/mm/fault.c
+++ b/arch/xtensa/mm/fault.c
@@ -44,6 +44,7 @@ void do_page_fault(struct pt_regs *regs)
 
 	int is_write, is_exec;
 	int fault;
+	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
 
 	info.si_code = SEGV_MAPERR;
 
@@ -71,6 +72,7 @@ void do_page_fault(struct pt_regs *regs)
 	       address, exccause, regs->pc, is_write? "w":"", is_exec? "x":"");
 #endif
 
+retry:
 	down_read(&mm->mmap_sem);
 	vma = find_vma(mm, address);
 
@@ -93,6 +95,7 @@ good_area:
 	if (is_write) {
 		if (!(vma->vm_flags & VM_WRITE))
 			goto bad_area;
+		flags |= FAULT_FLAG_WRITE;
 	} else if (is_exec) {
 		if (!(vma->vm_flags & VM_EXEC))
 			goto bad_area;
@@ -104,7 +107,11 @@ good_area:
 	 * make sure we exit gracefully rather than endlessly redo
 	 * the fault.
 	 */
-	fault = handle_mm_fault(mm, vma, address, is_write ? FAULT_FLAG_WRITE : 0);
+	fault = handle_mm_fault(mm, vma, address, flags);
+
+	if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
+		return;
+
 	if (unlikely(fault & VM_FAULT_ERROR)) {
 		if (fault & VM_FAULT_OOM)
 			goto out_of_memory;
@@ -112,10 +119,22 @@ good_area:
 			goto do_sigbus;
 		BUG();
 	}
-	if (fault & VM_FAULT_MAJOR)
-		current->maj_flt++;
-	else
-		current->min_flt++;
+	if (flags & FAULT_FLAG_ALLOW_RETRY) {
+		if (fault & VM_FAULT_MAJOR)
+			current->maj_flt++;
+		else
+			current->min_flt++;
+		if (fault & VM_FAULT_RETRY) {
+			flags &= ~FAULT_FLAG_ALLOW_RETRY;
+
+			 /* No need to up_read(&mm->mmap_sem) as we would
+			 * have already released it in __lock_page_or_retry
+			 * in mm/filemap.c.
+			 */
+
+			goto retry;
+		}
+	}
 
 	up_read(&mm->mmap_sem);
 	return;
-- 
1.7.5.4

Re: [PATCH 13/19 v2] xtensa/mm/fault.c: Port OOM changes to do_page_fault

[Kautuk Consul]

Hi Andrew,

Can you please accept this patch into linux-next ?

I have Chris Zankel's approval as below..

Thanks,
Kautuk.


On Wed, Jun 13, 2012 at 3:05 AM, Chris Zankel <chris@zankel.net> wrote:
> Hi Kautuk,
>
> Patch looks fine, so:
>
> Acked-by: Chris Zankel <chris@zankel.net>
>
> Thanks,
> -Chris
>
>
> On 6/12/12 8:54 AM, Kautuk Consul wrote:
>>
>> Hi,
>>
>> Could you please review this patch ?
>>
>> Many archs have accepted this patch, so I think maybe this is okay. :)
>>
>> Thanks,
>> Kautuk.
>>
>> On Sat, Mar 31, 2012 at 5:35 PM, Kautuk Consul<consul.kautuk@gmail.com>
>>  wrote:
>>>
>>> Commit d065bd810b6deb67d4897a14bfe21f8eb526ba99
>>> (mm: retry page fault when blocking on disk transfer) and
>>> commit 37b23e0525d393d48a7d59f870b3bc061a30ccdb
>>> (x86,mm: make pagefault killable)
>>>
>>> The above commits introduced changes into the x86 pagefault handler
>>> for making the page fault handler retryable as well as killable.
>>>
>>> These changes reduce the mmap_sem hold time, which is crucial
>>> during OOM killer invocation.
>>>
>>> Port these changes to xtensa.
>>>
>>> Signed-off-by: Kautuk Consul<consul.kautuk@gmail.com>
>>> ---
>>>  arch/xtensa/mm/fault.c |   29 ++++++++++++++++++++++++-----
>>>  1 files changed, 24 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/arch/xtensa/mm/fault.c b/arch/xtensa/mm/fault.c
>>> index b17885a..5a74c53 100644
>>> --- a/arch/xtensa/mm/fault.c
>>> +++ b/arch/xtensa/mm/fault.c
>>> @@ -44,6 +44,7 @@ void do_page_fault(struct pt_regs *regs)
>>>
>>>        int is_write, is_exec;
>>>        int fault;
>>> +       unsigned int flags = FAULT_FLAG_ALLOW_RETRY |
>>> FAULT_FLAG_KILLABLE;
>>>
>>>        info.si_code = SEGV_MAPERR;
>>>
>>> @@ -71,6 +72,7 @@ void do_page_fault(struct pt_regs *regs)
>>>               address, exccause, regs->pc, is_write? "w":"", is_exec?
>>> "x":"");
>>>  #endif
>>>
>>> +retry:
>>>        down_read(&mm->mmap_sem);
>>>        vma = find_vma(mm, address);
>>>
>>> @@ -93,6 +95,7 @@ good_area:
>>>        if (is_write) {
>>>                if (!(vma->vm_flags&  VM_WRITE))
>>>
>>>                        goto bad_area;
>>> +               flags |= FAULT_FLAG_WRITE;
>>>        } else if (is_exec) {
>>>                if (!(vma->vm_flags&  VM_EXEC))
>>>
>>>                        goto bad_area;
>>> @@ -104,7 +107,11 @@ good_area:
>>>         * make sure we exit gracefully rather than endlessly redo
>>>         * the fault.
>>>         */
>>> -       fault = handle_mm_fault(mm, vma, address, is_write ?
>>> FAULT_FLAG_WRITE : 0);
>>> +       fault = handle_mm_fault(mm, vma, address, flags);
>>> +
>>> +       if ((fault&  VM_FAULT_RETRY)&&  fatal_signal_pending(current))
>>>
>>> +               return;
>>> +
>>>        if (unlikely(fault&  VM_FAULT_ERROR)) {
>>>                if (fault&  VM_FAULT_OOM)
>>>
>>>                        goto out_of_memory;
>>> @@ -112,10 +119,22 @@ good_area:
>>>                        goto do_sigbus;
>>>                BUG();
>>>        }
>>> -       if (fault&  VM_FAULT_MAJOR)
>>>
>>> -               current->maj_flt++;
>>> -       else
>>> -               current->min_flt++;
>>> +       if (flags&  FAULT_FLAG_ALLOW_RETRY) {
>>> +               if (fault&  VM_FAULT_MAJOR)
>>>
>>> +                       current->maj_flt++;
>>> +               else
>>> +                       current->min_flt++;
>>> +               if (fault&  VM_FAULT_RETRY) {
>>> +                       flags&= ~FAULT_FLAG_ALLOW_RETRY;
>>>
>>> +
>>> +                        /* No need to up_read(&mm->mmap_sem) as we would
>>> +                        * have already released it in
>>> __lock_page_or_retry
>>> +                        * in mm/filemap.c.
>>> +                        */
>>> +
>>> +                       goto retry;
>>> +               }
>>> +       }
>>>
>>>        up_read(&mm->mmap_sem);
>>>        return;
>>> --
>>> 1.7.5.4
>>>
>