diff mbox series

[06/34] drm/i915: convert put_page() to put_user_page*()

Message ID 20190802022005.5117-7-jhubbard@nvidia.com (mailing list archive)
State New, archived
Headers show
Series put_user_pages(): miscellaneous call sites | expand

Commit Message

john.hubbard@gmail.com Aug. 2, 2019, 2:19 a.m. UTC
From: John Hubbard <jhubbard@nvidia.com>

For pages that were retained via get_user_pages*(), release those pages
via the new put_user_page*() routines, instead of via put_page() or
release_pages().

This is part a tree-wide conversion, as described in commit fc1d8e7cca2d
("mm: introduce put_user_page*(), placeholder versions").

Note that this effectively changes the code's behavior in
i915_gem_userptr_put_pages(): it now calls set_page_dirty_lock(),
instead of set_page_dirty(). This is probably more accurate.

As Christophe Hellwig put it, "set_page_dirty() is only safe if we are
dealing with a file backed page where we have reference on the inode it
hangs off." [1]

[1] https://lore.kernel.org/r/20190723153640.GB720@lst.de

Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Cc: David Airlie <airlied@linux.ie>
Cc: intel-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
---
 drivers/gpu/drm/i915/gem/i915_gem_userptr.c | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

Comments

Joonas Lahtinen Aug. 2, 2019, 9:19 a.m. UTC | #1
Quoting john.hubbard@gmail.com (2019-08-02 05:19:37)
> From: John Hubbard <jhubbard@nvidia.com>
> 
> For pages that were retained via get_user_pages*(), release those pages
> via the new put_user_page*() routines, instead of via put_page() or
> release_pages().
> 
> This is part a tree-wide conversion, as described in commit fc1d8e7cca2d
> ("mm: introduce put_user_page*(), placeholder versions").
> 
> Note that this effectively changes the code's behavior in
> i915_gem_userptr_put_pages(): it now calls set_page_dirty_lock(),
> instead of set_page_dirty(). This is probably more accurate.

We've already fixed this in drm-tip where the current code uses
set_page_dirty_lock().

This would conflict with our tree. Rodrigo is handling
drm-intel-next for 5.4, so you guys want to coordinate how
to merge.

Regards, Joonas
John Hubbard Aug. 2, 2019, 6:48 p.m. UTC | #2
On 8/2/19 2:19 AM, Joonas Lahtinen wrote:
> Quoting john.hubbard@gmail.com (2019-08-02 05:19:37)
>> From: John Hubbard <jhubbard@nvidia.com>
>>
>> For pages that were retained via get_user_pages*(), release those pages
>> via the new put_user_page*() routines, instead of via put_page() or
>> release_pages().
>>
>> This is part a tree-wide conversion, as described in commit fc1d8e7cca2d
>> ("mm: introduce put_user_page*(), placeholder versions").
>>
>> Note that this effectively changes the code's behavior in
>> i915_gem_userptr_put_pages(): it now calls set_page_dirty_lock(),
>> instead of set_page_dirty(). This is probably more accurate.
> 
> We've already fixed this in drm-tip where the current code uses
> set_page_dirty_lock().
> 
> This would conflict with our tree. Rodrigo is handling
> drm-intel-next for 5.4, so you guys want to coordinate how
> to merge.
> 

Hi Joonas, Rodrigo,

First of all, I apologize for the API breakage: put_user_pages_dirty_lock()
has an additional "dirty" parameter.

In order to deal with the merge problem, I'll drop this patch from my series,
and I'd recommend that the drm-intel-next take the following approach:

1) For now, s/put_page/put_user_page/ in i915_gem_userptr_put_pages(),
and fix up the set_page_dirty() --> set_page_dirty_lock() issue, like this
(based against linux.git):

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c 
b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
index 528b61678334..94721cc0093b 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
@@ -664,10 +664,10 @@ i915_gem_userptr_put_pages(struct drm_i915_gem_object *obj,

         for_each_sgt_page(page, sgt_iter, pages) {
                 if (obj->mm.dirty)
-                       set_page_dirty(page);
+                       set_page_dirty_lock(page);

                 mark_page_accessed(page);
-               put_page(page);
+               put_user_page(page);
         }
         obj->mm.dirty = false;


That will leave you with your original set_page_dirty_lock() calls
and everything works properly.

2) Next cycle, move to the new put_user_pages_dirty_lock().

thanks,
John Hubbard Aug. 3, 2019, 8:03 p.m. UTC | #3
On 8/2/19 11:48 AM, John Hubbard wrote:
> On 8/2/19 2:19 AM, Joonas Lahtinen wrote:
>> Quoting john.hubbard@gmail.com (2019-08-02 05:19:37)
>>> From: John Hubbard <jhubbard@nvidia.com>
...
> In order to deal with the merge problem, I'll drop this patch from my series,
> and I'd recommend that the drm-intel-next take the following approach:

Actually, I just pulled the latest linux.git, and there are a few changes:

> 
> 1) For now, s/put_page/put_user_page/ in i915_gem_userptr_put_pages(),
> and fix up the set_page_dirty() --> set_page_dirty_lock() issue, like this
> (based against linux.git):
> 
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
> index 528b61678334..94721cc0093b 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
> @@ -664,10 +664,10 @@ i915_gem_userptr_put_pages(struct drm_i915_gem_object *obj,
> 
>         for_each_sgt_page(page, sgt_iter, pages) {
>                 if (obj->mm.dirty)
> -                       set_page_dirty(page);
> +                       set_page_dirty_lock(page);

I see you've already applied this fix to your tree, in linux.git already.

> 
>                 mark_page_accessed(page);
> -               put_page(page);
> +               put_user_page(page);

But this conversion still needs doing. So I'll repost a patch that only does 
this (plus the other call sites). 

That can go in via either your tree, or Andrew's -mm tree, without generating
any conflicts.

thanks,
diff mbox series

Patch

diff --git a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
index 528b61678334..c18008d3cc2a 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
@@ -527,7 +527,7 @@  __i915_gem_userptr_get_pages_worker(struct work_struct *_work)
 	}
 	mutex_unlock(&obj->mm.lock);
 
-	release_pages(pvec, pinned);
+	put_user_pages(pvec, pinned);
 	kvfree(pvec);
 
 	i915_gem_object_put(obj);
@@ -640,7 +640,7 @@  static int i915_gem_userptr_get_pages(struct drm_i915_gem_object *obj)
 		__i915_gem_userptr_set_active(obj, true);
 
 	if (IS_ERR(pages))
-		release_pages(pvec, pinned);
+		put_user_pages(pvec, pinned);
 	kvfree(pvec);
 
 	return PTR_ERR_OR_ZERO(pages);
@@ -663,11 +663,8 @@  i915_gem_userptr_put_pages(struct drm_i915_gem_object *obj,
 	i915_gem_gtt_finish_pages(obj, pages);
 
 	for_each_sgt_page(page, sgt_iter, pages) {
-		if (obj->mm.dirty)
-			set_page_dirty(page);
-
 		mark_page_accessed(page);
-		put_page(page);
+		put_user_pages_dirty_lock(&page, 1, obj->mm.dirty);
 	}
 	obj->mm.dirty = false;