From: Daniel Vetter <daniel@ffwll.ch>
To: Jason Ekstrand <jason@jlekstrand.net>
Cc: intel-gfx@lists.freedesktop.org, dri-devel@lists.freedesktop.org
Subject: Re: [Intel-gfx] [PATCH 06/21] drm/i915: Implement SINGLE_TIMELINE with a syncobj (v3)
Date: Tue, 27 Apr 2021 11:55:06 +0200 [thread overview]
Message-ID: <YIffemMS/eiXLZ2Q@phenom.ffwll.local> (raw)
In-Reply-To: <20210423223131.879208-7-jason@jlekstrand.net>
On Fri, Apr 23, 2021 at 05:31:16PM -0500, Jason Ekstrand wrote:
> This API is entirely unnecessary and I'd love to get rid of it. If
> userspace wants a single timeline across multiple contexts, they can
> either use implicit synchronization or a syncobj, both of which existed
> at the time this feature landed. The justification given at the time
> was that it would help GL drivers which are inherently single-timeline.
> However, neither of our GL drivers actually wanted the feature. i965
> was already in maintenance mode at the time and iris uses syncobj for
> everything.
>
> Unfortunately, as much as I'd love to get rid of it, it is used by the
> media driver so we can't do that. We can, however, do the next-best
> thing which is to embed a syncobj in the context and do exactly what
> we'd expect from userspace internally. This isn't an entirely identical
> implementation because it's no longer atomic if userspace races with
> itself by calling execbuffer2 twice simultaneously from different
> threads. It won't crash in that case; it just doesn't guarantee any
> ordering between those two submits.
>
> Moving SINGLE_TIMELINE to a syncobj emulation has a couple of technical
> advantages beyond mere annoyance. One is that intel_timeline is no
> longer an api-visible object and can remain entirely an implementation
> detail. This may be advantageous as we make scheduler changes going
> forward. Second is that, together with deleting the CLONE_CONTEXT API,
> we should now have a 1:1 mapping between intel_context and
> intel_timeline which may help us reduce locking.
>
> v2 (Jason Ekstrand):
> - Update the comment on i915_gem_context::syncobj to mention that it's
> an emulation and the possible race if userspace calls execbuffer2
> twice on the same context concurrently.
> - Wrap the checks for eb.gem_context->syncobj in unlikely()
> - Drop the dma_fence reference
> - Improved commit message
>
> v3 (Jason Ekstrand):
> - Move the dma_fence_put() to before the error exit
>
> Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
> Cc: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
I'm assuming that igt coverage is good enough. Otoh if CI didn't catch
that racing execbuf are now unsynced maybe it wasn't good enough, but
whatever :-)
-Daniel
> ---
> drivers/gpu/drm/i915/gem/i915_gem_context.c | 49 +++++--------------
> .../gpu/drm/i915/gem/i915_gem_context_types.h | 14 +++++-
> .../gpu/drm/i915/gem/i915_gem_execbuffer.c | 16 ++++++
> 3 files changed, 40 insertions(+), 39 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> index 2c2fefa912805..a72c9b256723b 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
> @@ -67,6 +67,8 @@
> #include <linux/log2.h>
> #include <linux/nospec.h>
>
> +#include <drm/drm_syncobj.h>
> +
> #include "gt/gen6_ppgtt.h"
> #include "gt/intel_context.h"
> #include "gt/intel_context_param.h"
> @@ -225,10 +227,6 @@ static void intel_context_set_gem(struct intel_context *ce,
> ce->vm = vm;
> }
>
> - GEM_BUG_ON(ce->timeline);
> - if (ctx->timeline)
> - ce->timeline = intel_timeline_get(ctx->timeline);
> -
> if (ctx->sched.priority >= I915_PRIORITY_NORMAL &&
> intel_engine_has_timeslices(ce->engine))
> __set_bit(CONTEXT_USE_SEMAPHORES, &ce->flags);
> @@ -351,9 +349,6 @@ void i915_gem_context_release(struct kref *ref)
> mutex_destroy(&ctx->engines_mutex);
> mutex_destroy(&ctx->lut_mutex);
>
> - if (ctx->timeline)
> - intel_timeline_put(ctx->timeline);
> -
> put_pid(ctx->pid);
> mutex_destroy(&ctx->mutex);
>
> @@ -570,6 +565,9 @@ static void context_close(struct i915_gem_context *ctx)
> if (vm)
> i915_vm_close(vm);
>
> + if (ctx->syncobj)
> + drm_syncobj_put(ctx->syncobj);
> +
> ctx->file_priv = ERR_PTR(-EBADF);
>
> /*
> @@ -765,33 +763,11 @@ static void __assign_ppgtt(struct i915_gem_context *ctx,
> i915_vm_close(vm);
> }
>
> -static void __set_timeline(struct intel_timeline **dst,
> - struct intel_timeline *src)
> -{
> - struct intel_timeline *old = *dst;
> -
> - *dst = src ? intel_timeline_get(src) : NULL;
> -
> - if (old)
> - intel_timeline_put(old);
> -}
> -
> -static void __apply_timeline(struct intel_context *ce, void *timeline)
> -{
> - __set_timeline(&ce->timeline, timeline);
> -}
> -
> -static void __assign_timeline(struct i915_gem_context *ctx,
> - struct intel_timeline *timeline)
> -{
> - __set_timeline(&ctx->timeline, timeline);
> - context_apply_all(ctx, __apply_timeline, timeline);
> -}
> -
> static struct i915_gem_context *
> i915_gem_create_context(struct drm_i915_private *i915, unsigned int flags)
> {
> struct i915_gem_context *ctx;
> + int ret;
>
> if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE &&
> !HAS_EXECLISTS(i915))
> @@ -820,16 +796,13 @@ i915_gem_create_context(struct drm_i915_private *i915, unsigned int flags)
> }
>
> if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE) {
> - struct intel_timeline *timeline;
> -
> - timeline = intel_timeline_create(&i915->gt);
> - if (IS_ERR(timeline)) {
> + ret = drm_syncobj_create(&ctx->syncobj,
> + DRM_SYNCOBJ_CREATE_SIGNALED,
> + NULL);
> + if (ret) {
> context_close(ctx);
> - return ERR_CAST(timeline);
> + return ERR_PTR(ret);
> }
> -
> - __assign_timeline(ctx, timeline);
> - intel_timeline_put(timeline);
> }
>
> trace_i915_context_create(ctx);
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
> index 676592e27e7d2..df76767f0c41b 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
> @@ -83,7 +83,19 @@ struct i915_gem_context {
> struct i915_gem_engines __rcu *engines;
> struct mutex engines_mutex; /* guards writes to engines */
>
> - struct intel_timeline *timeline;
> + /**
> + * @syncobj: Shared timeline syncobj
> + *
> + * When the SHARED_TIMELINE flag is set on context creation, we
> + * emulate a single timeline across all engines using this syncobj.
> + * For every execbuffer2 call, this syncobj is used as both an in-
> + * and out-fence. Unlike the real intel_timeline, this doesn't
> + * provide perfect atomic in-order guarantees if the client races
> + * with itself by calling execbuffer2 twice concurrently. However,
> + * if userspace races with itself, that's not likely to yield well-
> + * defined results anyway so we choose to not care.
> + */
> + struct drm_syncobj *syncobj;
>
> /**
> * @vm: unique address space (GTT)
> diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
> index b812f313422a9..d640bba6ad9ab 100644
> --- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
> +++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
> @@ -3460,6 +3460,16 @@ i915_gem_do_execbuffer(struct drm_device *dev,
> goto err_vma;
> }
>
> + if (unlikely(eb.gem_context->syncobj)) {
> + struct dma_fence *fence;
> +
> + fence = drm_syncobj_fence_get(eb.gem_context->syncobj);
> + err = i915_request_await_dma_fence(eb.request, fence);
> + dma_fence_put(fence);
> + if (err)
> + goto err_ext;
> + }
> +
> if (in_fence) {
> if (args->flags & I915_EXEC_FENCE_SUBMIT)
> err = i915_request_await_execution(eb.request,
> @@ -3517,6 +3527,12 @@ i915_gem_do_execbuffer(struct drm_device *dev,
> fput(out_fence->file);
> }
> }
> +
> + if (unlikely(eb.gem_context->syncobj)) {
> + drm_syncobj_replace_fence(eb.gem_context->syncobj,
> + &eb.request->fence);
> + }
> +
> i915_request_put(eb.request);
>
> err_vma:
> --
> 2.31.1
>
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/intel-gfx
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel
next prev parent reply other threads:[~2021-04-27 9:55 UTC|newest]
Thread overview: 110+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-04-23 22:31 [PATCH 00/21] drm/i915/gem: ioctl clean-ups Jason Ekstrand
2021-04-23 22:31 ` [PATCH 01/21] drm/i915: Drop I915_CONTEXT_PARAM_RINGSIZE Jason Ekstrand
2021-04-27 9:32 ` Daniel Vetter
2021-04-28 3:33 ` Jason Ekstrand
2021-04-23 22:31 ` [PATCH 02/21] drm/i915: Drop I915_CONTEXT_PARAM_NO_ZEROMAP Jason Ekstrand
2021-04-27 9:38 ` [Intel-gfx] " Daniel Vetter
2021-04-23 22:31 ` [PATCH 03/21] drm/i915/gem: Set the watchdog timeout directly in intel_context_set_gem Jason Ekstrand
2021-04-27 9:42 ` Daniel Vetter
2021-04-28 15:55 ` [Intel-gfx] " Tvrtko Ursulin
2021-04-28 17:24 ` Jason Ekstrand
2021-04-29 8:04 ` Tvrtko Ursulin
2021-04-29 14:54 ` Jason Ekstrand
2021-04-29 17:12 ` Daniel Vetter
2021-04-29 17:13 ` Daniel Vetter
2021-04-29 18:41 ` Jason Ekstrand
2021-04-30 11:18 ` Tvrtko Ursulin
2021-04-30 15:35 ` Jason Ekstrand
2021-04-23 22:31 ` [PATCH 04/21] drm/i915/gem: Return void from context_apply_all Jason Ekstrand
2021-04-27 9:42 ` Daniel Vetter
2021-04-23 22:31 ` [PATCH 05/21] drm/i915: Drop the CONTEXT_CLONE API Jason Ekstrand
2021-04-27 9:49 ` [Intel-gfx] " Daniel Vetter
2021-04-28 17:38 ` Jason Ekstrand
2021-04-28 15:59 ` Tvrtko Ursulin
2021-04-23 22:31 ` [PATCH 06/21] drm/i915: Implement SINGLE_TIMELINE with a syncobj (v3) Jason Ekstrand
2021-04-27 9:55 ` Daniel Vetter [this message]
2021-04-28 15:49 ` [Intel-gfx] " Tvrtko Ursulin
2021-04-28 17:26 ` Jason Ekstrand
2021-04-29 8:06 ` Tvrtko Ursulin
2021-04-29 12:08 ` Daniel Vetter
2021-04-29 14:47 ` Jason Ekstrand
2021-04-23 22:31 ` [PATCH 07/21] drm/i915: Drop getparam support for I915_CONTEXT_PARAM_ENGINES Jason Ekstrand
2021-04-27 9:58 ` Daniel Vetter
2021-04-23 22:31 ` [PATCH 08/21] drm/i915/gem: Disallow bonding of virtual engines Jason Ekstrand
2021-04-26 23:43 ` [PATCH 08/20] drm/i915/gem: Disallow bonding of virtual engines (v2) Jason Ekstrand
2021-04-27 13:58 ` [Intel-gfx] " Daniel Vetter
2021-04-27 13:51 ` [PATCH 08/21] drm/i915/gem: Disallow bonding of virtual engines Jason Ekstrand
2021-04-28 10:13 ` Daniel Vetter
2021-04-28 17:18 ` Jason Ekstrand
2021-04-28 17:18 ` [Intel-gfx] " Matthew Brost
2021-04-28 17:46 ` Jason Ekstrand
2021-04-28 17:55 ` Matthew Brost
2021-04-28 18:17 ` Jason Ekstrand
2021-04-29 12:14 ` Daniel Vetter
2021-04-30 4:03 ` Matthew Brost
2021-04-30 10:11 ` Daniel Vetter
2021-05-01 17:17 ` Matthew Brost
2021-05-04 7:36 ` Daniel Vetter
2021-04-28 18:58 ` Jason Ekstrand
2021-04-29 12:16 ` Daniel Vetter
2021-04-29 16:02 ` Jason Ekstrand
2021-04-29 17:14 ` Daniel Vetter
2021-04-28 15:51 ` [Intel-gfx] " Tvrtko Ursulin
2021-04-29 12:24 ` Daniel Vetter
2021-04-29 12:54 ` Tvrtko Ursulin
2021-04-29 15:41 ` Jason Ekstrand
2021-04-23 22:31 ` [PATCH 09/21] drm/i915/gem: Disallow creating contexts with too many engines Jason Ekstrand
2021-04-28 10:16 ` [Intel-gfx] " Daniel Vetter
2021-04-28 10:42 ` Tvrtko Ursulin
2021-04-28 14:02 ` Daniel Vetter
2021-04-28 14:26 ` Tvrtko Ursulin
2021-04-28 17:09 ` Jason Ekstrand
2021-04-29 8:01 ` Tvrtko Ursulin
2021-04-29 19:16 ` Jason Ekstrand
2021-04-30 11:40 ` Tvrtko Ursulin
2021-04-30 15:54 ` Jason Ekstrand
2021-04-23 22:31 ` [PATCH 10/21] drm/i915/request: Remove the hook from await_execution Jason Ekstrand
2021-04-26 23:44 ` Jason Ekstrand
2021-04-23 22:31 ` [PATCH 11/21] drm/i915: Stop manually RCU banging in reset_stats_ioctl Jason Ekstrand
2021-04-28 10:27 ` Daniel Vetter
2021-04-28 18:22 ` Jason Ekstrand
2021-04-29 12:22 ` Daniel Vetter
2021-04-23 22:31 ` [PATCH 12/21] drm/i915/gem: Add a separate validate_priority helper Jason Ekstrand
2021-04-28 14:37 ` Daniel Vetter
2021-04-23 22:31 ` [PATCH 13/21] drm/i915/gem: Add an intermediate proto_context struct Jason Ekstrand
2021-04-29 13:02 ` [Intel-gfx] " Daniel Vetter
2021-04-29 16:44 ` Jason Ekstrand
2021-04-23 22:31 ` [PATCH 14/21] drm/i915/gem: Return an error ptr from context_lookup Jason Ekstrand
2021-04-29 13:27 ` [Intel-gfx] " Daniel Vetter
2021-04-29 15:29 ` Jason Ekstrand
2021-04-29 17:16 ` Daniel Vetter
2021-04-23 22:31 ` [PATCH 15/21] drm/i915/gt: Drop i915_address_space::file Jason Ekstrand
2021-04-29 12:37 ` Daniel Vetter
2021-04-29 15:26 ` Jason Ekstrand
2021-04-23 22:31 ` [PATCH 16/21] drm/i915/gem: Delay context creation Jason Ekstrand
2021-04-24 3:21 ` [Intel-gfx] " kernel test robot
2021-04-24 3:24 ` kernel test robot
2021-04-29 15:51 ` Daniel Vetter
2021-04-29 18:16 ` Jason Ekstrand
2021-04-29 18:56 ` Daniel Vetter
2021-04-29 19:01 ` Jason Ekstrand
2021-04-29 19:07 ` Daniel Vetter
2021-04-29 21:35 ` Jason Ekstrand
2021-04-30 6:53 ` Daniel Vetter
2021-04-30 11:58 ` Tvrtko Ursulin
2021-04-30 12:30 ` Daniel Vetter
2021-04-30 12:44 ` Tvrtko Ursulin
2021-04-30 13:07 ` Daniel Vetter
2021-04-30 13:15 ` Tvrtko Ursulin
2021-04-30 16:27 ` Jason Ekstrand
2021-04-30 16:33 ` Daniel Vetter
2021-04-30 16:57 ` Jason Ekstrand
2021-04-30 17:08 ` Daniel Vetter
2021-04-23 22:31 ` [PATCH 17/21] drm/i915/gem: Don't allow changing the VM on running contexts Jason Ekstrand
2021-04-23 22:31 ` [PATCH 18/21] drm/i915/gem: Don't allow changing the engine set " Jason Ekstrand
2021-04-29 17:21 ` Daniel Vetter
2021-04-23 22:31 ` [PATCH 19/21] drm/i915/selftests: Take a VM in kernel_context() Jason Ekstrand
2021-04-23 22:31 ` [PATCH 20/21] i915/gem/selftests: Assign the VM at context creation in igt_shared_ctx_exec Jason Ekstrand
2021-04-29 17:19 ` [Intel-gfx] " Daniel Vetter
2021-04-23 22:31 ` [PATCH 21/21] drm/i915/gem: Roll all of context creation together Jason Ekstrand
2021-04-29 17:25 ` Daniel Vetter
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=YIffemMS/eiXLZ2Q@phenom.ffwll.local \
--to=daniel@ffwll.ch \
--cc=dri-devel@lists.freedesktop.org \
--cc=intel-gfx@lists.freedesktop.org \
--cc=jason@jlekstrand.net \
/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;
as well as URLs for NNTP newsgroup(s).