public inbox for intel-gfx@lists.freedesktop.org
 help / color / mirror / Atom feed
From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
To: Matthew Auld <matthew.william.auld@gmail.com>
Cc: Intel Graphics Development <intel-gfx@lists.freedesktop.org>,
	ML dri-devel <dri-devel@lists.freedesktop.org>
Subject: Re: [Intel-gfx] [PATCH v2 07/16] drm/i915: Take trylock during eviction, v2.
Date: Wed, 8 Dec 2021 14:28:48 +0100	[thread overview]
Message-ID: <8c012291-4fdc-42df-3cbf-dd0b9147038d@linux.intel.com> (raw)
In-Reply-To: <CAM0jSHO-rLxHLoM+mDtuQ-syUx4d_SzV5c_wpW87YTvu1X+AVg@mail.gmail.com>

On 07-12-2021 12:01, Matthew Auld wrote:
> On Mon, 29 Nov 2021 at 13:57, Maarten Lankhorst
> <maarten.lankhorst@linux.intel.com> wrote:
>> Now that freeing objects takes the object lock when destroying the
>> backing pages, we can confidently take the object lock even for dead
>> objects.
> That looks to be a future patch, at least with non-TTM backend? Does
> something need to be re-ordered in the series?
>
>> Use this fact to take the object lock in the shrinker, without requiring
>> a reference to the object, so all calls to unbind take the object lock.
> Hmm, the previous patch was talking about "we don't evict when
> refcount = 0", but this looks to be doing something else?

Yeah, a careful inspection of eviction code does allow evicting when refcount = 0 using trylocks, iff we take the object lock during destruction., in order to free all remaining backing.

However, you can't do a blocking lock for dead objects, because VM lock is the inner lock in the hierarchy. Lockdep will complain (rightfully).

To free dead objects to reclaim memory, you would need to call drain_freed_objects(), which will deadlock if you hold either vm lock or resv locks.

>> This is the last step to requiring the object lock for vma_unbind.
>>
>> Changes since v1:
>> - No longer require the refcount, as every freed object now holds the lock
>>   when unbinding VMA's.
>>
>> Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
>> ---
>>  drivers/gpu/drm/i915/gem/i915_gem_shrinker.c |  6 ++++
>>  drivers/gpu/drm/i915/i915_gem_evict.c        | 34 +++++++++++++++++---
>>  2 files changed, 35 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c b/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
>> index eebff4735781..ad2123369e0d 100644
>> --- a/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
>> +++ b/drivers/gpu/drm/i915/gem/i915_gem_shrinker.c
>> @@ -405,12 +405,18 @@ i915_gem_shrinker_vmap(struct notifier_block *nb, unsigned long event, void *ptr
>>         list_for_each_entry_safe(vma, next,
>>                                  &i915->ggtt.vm.bound_list, vm_link) {
>>                 unsigned long count = vma->node.size >> PAGE_SHIFT;
>> +               struct drm_i915_gem_object *obj = vma->obj;
>>
>>                 if (!vma->iomap || i915_vma_is_active(vma))
>>                         continue;
>>
>> +               if (!i915_gem_object_trylock(obj))
>> +                       continue;
>> +
>>                 if (__i915_vma_unbind(vma) == 0)
>>                         freed_pages += count;
>> +
>> +               i915_gem_object_unlock(obj);
>>         }
>>         mutex_unlock(&i915->ggtt.vm.mutex);
>>
>> diff --git a/drivers/gpu/drm/i915/i915_gem_evict.c b/drivers/gpu/drm/i915/i915_gem_evict.c
>> index 2b73ddb11c66..286efa462eca 100644
>> --- a/drivers/gpu/drm/i915/i915_gem_evict.c
>> +++ b/drivers/gpu/drm/i915/i915_gem_evict.c
>> @@ -58,6 +58,9 @@ mark_free(struct drm_mm_scan *scan,
>>         if (i915_vma_is_pinned(vma))
>>                 return false;
>>
>> +       if (!i915_gem_object_trylock(vma->obj))
>> +               return false;
>> +
>>         list_add(&vma->evict_link, unwind);
>>         return drm_mm_scan_add_block(scan, &vma->node);
>>  }
>> @@ -178,6 +181,7 @@ i915_gem_evict_something(struct i915_address_space *vm,
>>         list_for_each_entry_safe(vma, next, &eviction_list, evict_link) {
>>                 ret = drm_mm_scan_remove_block(&scan, &vma->node);
>>                 BUG_ON(ret);
>> +               i915_gem_object_unlock(vma->obj);
>>         }
>>
>>         /*
>> @@ -222,10 +226,12 @@ i915_gem_evict_something(struct i915_address_space *vm,
>>          * of any of our objects, thus corrupting the list).
>>          */
>>         list_for_each_entry_safe(vma, next, &eviction_list, evict_link) {
>> -               if (drm_mm_scan_remove_block(&scan, &vma->node))
>> +               if (drm_mm_scan_remove_block(&scan, &vma->node)) {
>>                         __i915_vma_pin(vma);
>> -               else
>> +               } else {
>>                         list_del(&vma->evict_link);
>> +                       i915_gem_object_unlock(vma->obj);
>> +               }
>>         }
>>
>>         /* Unbinding will emit any required flushes */
>> @@ -234,16 +240,22 @@ i915_gem_evict_something(struct i915_address_space *vm,
>>                 __i915_vma_unpin(vma);
>>                 if (ret == 0)
>>                         ret = __i915_vma_unbind(vma);
>> +
>> +               i915_gem_object_unlock(vma->obj);
>>         }
>>
>>         while (ret == 0 && (node = drm_mm_scan_color_evict(&scan))) {
>>                 vma = container_of(node, struct i915_vma, node);
>>
>> +
>>                 /* If we find any non-objects (!vma), we cannot evict them */
>> -               if (vma->node.color != I915_COLOR_UNEVICTABLE)
>> +               if (vma->node.color != I915_COLOR_UNEVICTABLE &&
>> +                   i915_gem_object_trylock(vma->obj)) {
>>                         ret = __i915_vma_unbind(vma);
>> -               else
>> -                       ret = -ENOSPC; /* XXX search failed, try again? */
>> +                       i915_gem_object_unlock(vma->obj);
>> +               } else {
>> +                       ret = -ENOSPC;
>> +               }
>>         }
>>
>>         return ret;
>> @@ -333,6 +345,11 @@ int i915_gem_evict_for_node(struct i915_address_space *vm,
>>                         break;
>>                 }
>>
>> +               if (!i915_gem_object_trylock(vma->obj)) {
>> +                       ret = -ENOSPC;
>> +                       break;
>> +               }
>> +
>>                 /*
>>                  * Never show fear in the face of dragons!
>>                  *
>> @@ -350,6 +367,8 @@ int i915_gem_evict_for_node(struct i915_address_space *vm,
>>                 __i915_vma_unpin(vma);
>>                 if (ret == 0)
>>                         ret = __i915_vma_unbind(vma);
>> +
>> +               i915_gem_object_unlock(vma->obj);
>>         }
>>
>>         return ret;
>> @@ -393,6 +412,9 @@ int i915_gem_evict_vm(struct i915_address_space *vm)
>>                         if (i915_vma_is_pinned(vma))
>>                                 continue;
>>
>> +                       if (!i915_gem_object_trylock(vma->obj))
>> +                               continue;
>> +
>>                         __i915_vma_pin(vma);
>>                         list_add(&vma->evict_link, &eviction_list);
>>                 }
>> @@ -406,6 +428,8 @@ int i915_gem_evict_vm(struct i915_address_space *vm)
>>                                 ret = __i915_vma_unbind(vma);
>>                         if (ret != -EINTR) /* "Get me out of here!" */
>>                                 ret = 0;
>> +
>> +                       i915_gem_object_unlock(vma->obj);
>>                 }
>>         } while (ret == 0);
>>
>> --
>> 2.34.0
>>


  reply	other threads:[~2021-12-08 13:29 UTC|newest]

Thread overview: 59+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-11-29 13:47 [Intel-gfx] [PATCH v2 00/16] drm/i915: Remove short term pins from execbuf Maarten Lankhorst
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 01/16] drm/i915: Remove unused bits of i915_vma/active api Maarten Lankhorst
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 02/16] drm/i915: Change shrink ordering to use locking around unbinding Maarten Lankhorst
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 03/16] drm/i915: Remove pages_mutex and intel_gtt->vma_ops.set/clear_pages members, v2 Maarten Lankhorst
2021-12-06 13:13   ` Matthew Auld
2021-12-06 15:18     ` Maarten Lankhorst
2021-12-06 17:00       ` Matthew Auld
2021-12-07 18:15         ` Daniel Vetter
2021-12-06 17:10   ` Matthew Auld
2021-12-07 10:06     ` Maarten Lankhorst
2021-12-07 10:45       ` Matthew Auld
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 04/16] drm/i915: Take object lock in i915_ggtt_pin if ww is not set Maarten Lankhorst
2021-12-06 13:18   ` Matthew Auld
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 05/16] drm/i915: Force ww lock for i915_gem_object_ggtt_pin_ww Maarten Lankhorst
2021-11-30  9:20   ` [Intel-gfx] [PATCH] drm/i915: Force ww lock for i915_gem_object_ggtt_pin_ww, v2 Maarten Lankhorst
2021-12-01 15:07     ` Matthew Auld
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 06/16] drm/i915: Ensure gem_contexts selftests work with unbind changes Maarten Lankhorst
2021-12-07 10:44   ` Matthew Auld
2021-12-08 13:20     ` Maarten Lankhorst
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 07/16] drm/i915: Take trylock during eviction, v2 Maarten Lankhorst
2021-12-07 11:01   ` Matthew Auld
2021-12-08 13:28     ` Maarten Lankhorst [this message]
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 08/16] drm/i915: Pass trylock context to callers Maarten Lankhorst
2021-12-07 14:26   ` Matthew Auld
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 09/16] drm/i915: Ensure i915_vma tests do not get -ENOSPC with the locking changes Maarten Lankhorst
2021-12-08 11:49   ` Matthew Auld
2021-12-08 12:01     ` Matthew Auld
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 10/16] drm/i915: Make i915_gem_evict_vm work correctly for already locked objects Maarten Lankhorst
2021-12-08 12:07   ` Matthew Auld
2021-12-08 13:34     ` Maarten Lankhorst
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 11/16] drm/i915: Call i915_gem_evict_vm in vm_fault_gtt to prevent new ENOSPC errors Maarten Lankhorst
2021-12-09 12:17   ` Matthew Auld
2021-12-09 12:59     ` Maarten Lankhorst
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 12/16] drm/i915: Add i915_vma_unbind_unlocked, and take obj lock for i915_vma_unbind Maarten Lankhorst
2021-12-09 13:05   ` Matthew Auld
2021-12-09 13:25     ` Maarten Lankhorst
2021-12-09 13:40       ` Matthew Auld
2021-12-09 13:45         ` Maarten Lankhorst
2021-12-09 14:27           ` Matthew Auld
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 13/16] drm/i915: Require object lock when freeing pages during destruction Maarten Lankhorst
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 14/16] drm/i915: Remove assert_object_held_shared Maarten Lankhorst
2021-12-09 13:07   ` Matthew Auld
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 15/16] drm/i915: Remove support for unlocked i915_vma unbind Maarten Lankhorst
2021-11-29 13:47 ` [Intel-gfx] [PATCH v2 16/16] drm/i915: Remove short-term pins from execbuf, v5 Maarten Lankhorst
2021-12-09 16:22   ` Matthew Auld
2021-11-29 15:32 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for drm/i915: Remove short term pins from execbuf Patchwork
2021-11-29 15:33 ` [Intel-gfx] ✗ Fi.CI.SPARSE: " Patchwork
2021-11-29 15:37 ` [Intel-gfx] ✗ Fi.CI.DOCS: " Patchwork
2021-11-29 16:11 ` [Intel-gfx] ✗ Fi.CI.BAT: failure " Patchwork
2021-11-30  8:54 ` [Intel-gfx] [PATCH v2 00/16] " Tvrtko Ursulin
2021-11-30 11:17   ` Maarten Lankhorst
2021-11-30 18:38     ` Tvrtko Ursulin
2021-12-01 11:15       ` Maarten Lankhorst
2021-12-01 13:11         ` Tvrtko Ursulin
2021-11-30 11:18 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for drm/i915: Remove short term pins from execbuf. (rev2) Patchwork
2021-11-30 11:19 ` [Intel-gfx] ✗ Fi.CI.SPARSE: " Patchwork
2021-11-30 11:23 ` [Intel-gfx] ✗ Fi.CI.DOCS: " Patchwork
2021-11-30 11:49 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork
2021-11-30 14:51 ` [Intel-gfx] ✗ Fi.CI.IGT: failure " Patchwork

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=8c012291-4fdc-42df-3cbf-dd0b9147038d@linux.intel.com \
    --to=maarten.lankhorst@linux.intel.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=matthew.william.auld@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox