[PATCH] drm/ttm: handle OOM in ttm_tt_swapout

[PATCH] drm/ttm: handle OOM in ttm_tt_swapout

[Maarten Maathuis]

- Without this change I get a general protection fault.
- Also use PTR_ERR where applicable.

Signed-off-by: Maarten Maathuis <madman2003@gmail.com>
---
 drivers/gpu/drm/ttm/ttm_tt.c |   18 +++++++++++-------
 1 files changed, 11 insertions(+), 7 deletions(-)

diff --git a/drivers/gpu/drm/ttm/ttm_tt.c b/drivers/gpu/drm/ttm/ttm_tt.c
index 9c2b1cc..5b5c320 100644
--- a/drivers/gpu/drm/ttm/ttm_tt.c
+++ b/drivers/gpu/drm/ttm/ttm_tt.c
@@ -467,7 +467,7 @@ static int ttm_tt_swapin(struct ttm_tt *ttm)
 	void *from_virtual;
 	void *to_virtual;
 	int i;
-	int ret;
+	int ret = -ENOMEM;
 
 	if (ttm->page_flags & TTM_PAGE_FLAG_USER) {
 		ret = ttm_tt_set_user(ttm, ttm->tsk, ttm->start,
@@ -486,8 +486,10 @@ static int ttm_tt_swapin(struct ttm_tt *ttm)
 
 	for (i = 0; i < ttm->num_pages; ++i) {
 		from_page = read_mapping_page(swap_space, i, NULL);
-		if (IS_ERR(from_page))
+		if (IS_ERR(from_page)) {
+			ret = PTR_ERR(from_page);
 			goto out_err;
+		}
 		to_page = __ttm_tt_get_page(ttm, i);
 		if (unlikely(to_page == NULL))
 			goto out_err;
@@ -510,7 +512,7 @@ static int ttm_tt_swapin(struct ttm_tt *ttm)
 	return 0;
 out_err:
 	ttm_tt_free_alloced_pages(ttm);
-	return -ENOMEM;
+	return ret;
 }
 
 int ttm_tt_swapout(struct ttm_tt *ttm, struct file *persistant_swap_storage)
@@ -522,6 +524,7 @@ int ttm_tt_swapout(struct ttm_tt *ttm, struct file *persistant_swap_storage)
 	void *from_virtual;
 	void *to_virtual;
 	int i;
+	int ret = -ENOMEM;
 
 	BUG_ON(ttm->state != tt_unbound && ttm->state != tt_unpopulated);
 	BUG_ON(ttm->caching_state != tt_cached);
@@ -544,7 +547,7 @@ int ttm_tt_swapout(struct ttm_tt *ttm, struct file *persistant_swap_storage)
 						0);
 		if (unlikely(IS_ERR(swap_storage))) {
 			printk(KERN_ERR "Failed allocating swap storage.\n");
-			return -ENOMEM;
+			return PTR_ERR(swap_storage);
 		}
 	} else
 		swap_storage = persistant_swap_storage;
@@ -556,9 +559,10 @@ int ttm_tt_swapout(struct ttm_tt *ttm, struct file *persistant_swap_storage)
 		if (unlikely(from_page == NULL))
 			continue;
 		to_page = read_mapping_page(swap_space, i, NULL);
-		if (unlikely(to_page == NULL))
+		if (unlikely(IS_ERR(to_page))) {
+			ret = PTR_ERR(to_page);
 			goto out_err;
-
+		}
 		preempt_disable();
 		from_virtual = kmap_atomic(from_page, KM_USER0);
 		to_virtual = kmap_atomic(to_page, KM_USER1);
@@ -582,5 +586,5 @@ out_err:
 	if (!persistant_swap_storage)
 		fput(swap_storage);
 
-	return -ENOMEM;
+	return ret;
 }
-- 
1.7.0


------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

Re: [PATCH] drm/ttm: handle OOM in ttm_tt_swapout

[Dave Airlie]

On Sat, Feb 20, 2010 at 12:22 PM, Maarten Maathuis <madman2003@gmail.com> wrote:
> - Without this change I get a general protection fault.
> - Also use PTR_ERR where applicable.

I just want to make sure I understand, but really the only bit of this
patch that matters is:

> @@ -556,9 +559,10 @@ int ttm_tt_swapout(struct ttm_tt *ttm, struct file *persistant_swap_storage)
>                if (unlikely(from_page == NULL))
>                        continue;
>                to_page = read_mapping_page(swap_space, i, NULL);
> -               if (unlikely(to_page == NULL))
> +               if (unlikely(IS_ERR(to_page))) {

^^ these two lines where we are testing for NULL but should be
checking for an error?

> +                       ret = PTR_ERR(to_page);
>                        goto out_err;
> -
> +               }

If that is true and the rest is just nice cleanups then I'm okay with it,.

Reviewed-by: Dave Airlie <airlied@redhat.com>

I'll need Thomas's ack on this also.

Dave.

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

Re: [PATCH] drm/ttm: handle OOM in ttm_tt_swapout

[Maarten Maathuis]

On Tue, Feb 23, 2010 at 6:59 AM, Dave Airlie <airlied@gmail.com> wrote:
> On Sat, Feb 20, 2010 at 12:22 PM, Maarten Maathuis <madman2003@gmail.com> wrote:
>> - Without this change I get a general protection fault.
>> - Also use PTR_ERR where applicable.
>
> I just want to make sure I understand, but really the only bit of this
> patch that matters is:
>
>> @@ -556,9 +559,10 @@ int ttm_tt_swapout(struct ttm_tt *ttm, struct file *persistant_swap_storage)
>>                if (unlikely(from_page == NULL))
>>                        continue;
>>                to_page = read_mapping_page(swap_space, i, NULL);
>> -               if (unlikely(to_page == NULL))
>> +               if (unlikely(IS_ERR(to_page))) {
>
> ^^ these two lines where we are testing for NULL but should be
> checking for an error?

Yes, that's the critical bit. As you can see we were already doing
this for ttm_tt_swapin.

>
>> +                       ret = PTR_ERR(to_page);
>>                        goto out_err;
>> -
>> +               }
>
> If that is true and the rest is just nice cleanups then I'm okay with it,.
>
> Reviewed-by: Dave Airlie <airlied@redhat.com>
>
> I'll need Thomas's ack on this also.
>
> Dave.

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

Re: [PATCH] drm/ttm: handle OOM in ttm_tt_swapout

[Thomas Hellstrom]

Maarten Maathuis wrote:
> On Tue, Feb 23, 2010 at 6:59 AM, Dave Airlie <airlied@gmail.com> wrote:
>   
>> On Sat, Feb 20, 2010 at 12:22 PM, Maarten Maathuis <madman2003@gmail.com> wrote:
>>     
>>> - Without this change I get a general protection fault.
>>> - Also use PTR_ERR where applicable.
>>>       
>> I just want to make sure I understand, but really the only bit of this
>> patch that matters is:
>>
>>     
>>> @@ -556,9 +559,10 @@ int ttm_tt_swapout(struct ttm_tt *ttm, struct file *persistant_swap_storage)
>>>                if (unlikely(from_page == NULL))
>>>                        continue;
>>>                to_page = read_mapping_page(swap_space, i, NULL);
>>> -               if (unlikely(to_page == NULL))
>>> +               if (unlikely(IS_ERR(to_page))) {
>>>       
>> ^^ these two lines where we are testing for NULL but should be
>> checking for an error?
>>     
>
> Yes, that's the critical bit. As you can see we were already doing
> this for ttm_tt_swapin.
>
>   
>>> +                       ret = PTR_ERR(to_page);
>>>                        goto out_err;
>>> -
>>> +               }
>>>       
>> If that is true and the rest is just nice cleanups then I'm okay with it,.
>>
>> Reviewed-by: Dave Airlie <airlied@redhat.com>
>>
>> I'll need Thomas's ack on this also.
>>
>> Dave.
>>     
Acked-by: Thomas Hellstrom <thellstrom@vmware.com>

/Thomas


------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--