From: Chris Wilson <chris@chris-wilson.co.uk> To: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: intel-gfx@lists.freedesktop.org, stable@vger.kernel.org Subject: Re: [Intel-gfx] [PATCH v3 2/3] drm/i915: Fix userptr deadlock with aliased GTT mmappings Date: Thu, 10 Sep 2015 10:51:30 +0100 [thread overview] Message-ID: <20150910095130.GR32324@nuc-i3427.alporthouse.com> (raw) In-Reply-To: <55F150EE.1080503@linux.intel.com> On Thu, Sep 10, 2015 at 10:44:14AM +0100, Tvrtko Ursulin wrote: > > On 09/09/2015 04:03 PM, Chris Wilson wrote: > >On Wed, Sep 09, 2015 at 02:56:16PM +0100, Tvrtko Ursulin wrote: > >> > >>Hi, > >> > >>On 08/10/2015 09:51 AM, Chris Wilson wrote: > >>>+out: > >>> drm_free_large(pvec); > >>> return ret; > >>>+ > >>>+err: > >>>+ /* No pages here, no need for the mmu-notifier to wake us */ > >>>+ __i915_gem_userptr_set_active(obj, false); > >>>+err_active: > >>>+ release_pages(pvec, pinned, 0); > >>>+ goto out; > >>> } > >> > >>I don't like the goto dance. Would something like the below be clearer? > > > >We can condense it if we use a bool active and then feed everything > >through the single exit path: > > > > active = false; > > if (pinned < 0) > > ret = pinned, pinned = 0; > > else if (pinned < num_pages) > > ret = __i915_gem_userptr_get_pages_queue(obj, &active); > > else > > ret = __i915_gem_userptr_set_pages(obj, pvec, num_pages); > > if (ret) { > > __i915_gem_userptr_set_active(obj, active); > > release_pages(pvec, pinned, 0); > > } > > drm_free_large(pvec); > > return ret; > > > >Not happy with _queue. I guess i915_gem_userptr_get_pages_via_worker() > >is better. Or i915_gem_userptr_get_pages_deferred(). > > Looks much better on a glance. If release_pages with pinned = 0 is OK. It does a loop over for(int i = 0; i < pinned; i++); so it is fine as it was. > For the queueue/via_worker/deferred maybe _schedule_get_pages_worker? I want to keep the i915_gem_object_get_pages prefix (haven't yet broken that pattern, so no reason to start now), but _schedule seems reasonable. -Chris -- Chris Wilson, Intel Open Source Technology Centre
WARNING: multiple messages have this Message-ID (diff)
From: Chris Wilson <chris@chris-wilson.co.uk> To: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com> Cc: intel-gfx@lists.freedesktop.org, stable@vger.kernel.org Subject: Re: [PATCH v3 2/3] drm/i915: Fix userptr deadlock with aliased GTT mmappings Date: Thu, 10 Sep 2015 10:51:30 +0100 [thread overview] Message-ID: <20150910095130.GR32324@nuc-i3427.alporthouse.com> (raw) In-Reply-To: <55F150EE.1080503@linux.intel.com> On Thu, Sep 10, 2015 at 10:44:14AM +0100, Tvrtko Ursulin wrote: > > On 09/09/2015 04:03 PM, Chris Wilson wrote: > >On Wed, Sep 09, 2015 at 02:56:16PM +0100, Tvrtko Ursulin wrote: > >> > >>Hi, > >> > >>On 08/10/2015 09:51 AM, Chris Wilson wrote: > >>>+out: > >>> drm_free_large(pvec); > >>> return ret; > >>>+ > >>>+err: > >>>+ /* No pages here, no need for the mmu-notifier to wake us */ > >>>+ __i915_gem_userptr_set_active(obj, false); > >>>+err_active: > >>>+ release_pages(pvec, pinned, 0); > >>>+ goto out; > >>> } > >> > >>I don't like the goto dance. Would something like the below be clearer? > > > >We can condense it if we use a bool active and then feed everything > >through the single exit path: > > > > active = false; > > if (pinned < 0) > > ret = pinned, pinned = 0; > > else if (pinned < num_pages) > > ret = __i915_gem_userptr_get_pages_queue(obj, &active); > > else > > ret = __i915_gem_userptr_set_pages(obj, pvec, num_pages); > > if (ret) { > > __i915_gem_userptr_set_active(obj, active); > > release_pages(pvec, pinned, 0); > > } > > drm_free_large(pvec); > > return ret; > > > >Not happy with _queue. I guess i915_gem_userptr_get_pages_via_worker() > >is better. Or i915_gem_userptr_get_pages_deferred(). > > Looks much better on a glance. If release_pages with pinned = 0 is OK. It does a loop over for(int i = 0; i < pinned; i++); so it is fine as it was. > For the queueue/via_worker/deferred maybe _schedule_get_pages_worker? I want to keep the i915_gem_object_get_pages prefix (haven't yet broken that pattern, so no reason to start now), but _schedule seems reasonable. -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
next prev parent reply other threads:[~2015-09-10 9:51 UTC|newest] Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top 2015-08-10 8:51 [PATCH v3 1/3] drm/i915: Only update the current userptr worker Chris Wilson 2015-08-10 8:51 ` [PATCH v3 2/3] drm/i915: Fix userptr deadlock with aliased GTT mmappings Chris Wilson 2015-09-09 13:56 ` [Intel-gfx] " Tvrtko Ursulin 2015-09-09 15:03 ` Chris Wilson 2015-09-09 15:03 ` Chris Wilson 2015-09-10 9:44 ` [Intel-gfx] " Tvrtko Ursulin 2015-09-10 9:51 ` Chris Wilson [this message] 2015-09-10 9:51 ` Chris Wilson 2015-08-10 8:51 ` [PATCH v3 3/3] drm/i915: Use a task to cancel the userptr on invalidate_range Chris Wilson 2015-09-09 14:45 ` Tvrtko Ursulin 2015-09-09 15:08 ` Chris Wilson 2015-09-09 15:20 ` Tvrtko Ursulin 2015-09-09 15:42 ` Chris Wilson 2015-09-10 9:50 ` Tvrtko Ursulin 2015-09-09 10:39 ` [PATCH v3 1/3] drm/i915: Only update the current userptr worker Tvrtko Ursulin 2015-09-09 10:44 ` Chris Wilson
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=20150910095130.GR32324@nuc-i3427.alporthouse.com \ --to=chris@chris-wilson.co.uk \ --cc=intel-gfx@lists.freedesktop.org \ --cc=stable@vger.kernel.org \ --cc=tvrtko.ursulin@linux.intel.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: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.