Revert lima fdinfo patchset
Subsystems (3)
cover Revert lima fdinfo patchset
From: Qiang Yu <[email protected]>
Upstream has reverted the dependant commitdf622729ddbf ("drm/scheduler: track GPU active time per entity""), but this patchset has been pushed to drm-misc-next which still has that commit. To avoid other branch build fail after merge drm-misc-next, just revert the lima patchset on drm-misc-next too.
Qiang Yu (3):
Revert "drm/lima: add show_fdinfo for drm usage stats" Revert "drm/lima: allocate unique id per drm_file" Revert "drm/lima: add usage counting method to ctx_mgr"
drivers/gpu/drm/lima/lima_ctx.c | 30 +-------------------- drivers/gpu/drm/lima/lima_ctx.h | 3 --- drivers/gpu/drm/lima/lima_device.h | 3 --- drivers/gpu/drm/lima/lima_drv.c | 43 +----------------------------- drivers/gpu/drm/lima/lima_drv.h | 1 - 5 files changed, 2 insertions(+), 78 deletions(-)
Signature · 1 line
2.25.1
PATCH 1/3 Revert "drm/lima: add show_fdinfo for drm usage stats"
From: Qiang Yu <[email protected]>
This reverts commit 4a66f3da99dcb4dcbd28544110636b50adfb0f0d.
This is due to the depend commit has been reverted on upstream:baad10973fdb ("Revert "drm/scheduler: track GPU active time per entity"")
--- drivers/gpu/drm/lima/lima_drv.c | 31 +------------------------------ 1 file changed, 1 insertion(+), 30 deletions(-)
drivers/gpu/drm/lima/lima_drv.c +1 -30
@@ -261,36 +261,7 @@ static const struct drm_ioctl_desc lima_drm_driver_ioctls[] = { DRM_IOCTL_DEF_DRV(LIMA_CTX_FREE, lima_ioctl_ctx_free, DRM_RENDER_ALLOW), }; -static void lima_drm_driver_show_fdinfo (struct seq_file *m,struct file *filp) -{ -struct drm_file *file = filp->private_data; -struct drm_device *dev = file->minor->dev; -struct lima_device *ldev =to_lima_dev (dev); -struct lima_drm_priv *priv = file->driver_priv; -struct lima_ctx_mgr *ctx_mgr = &priv->ctx_mgr; -u64 usage[lima_pipe_num]; - -lima_ctx_mgr_usage (ctx_mgr, usage); - -/* -* For a description of the text output format used here, see -* Documentation/gpu/drm-usage-stats.rst. -*/ -seq_printf (m,"drm-driver:\t%s\n" , dev->driver->name); -seq_printf (m,"drm-client-id:\t%u\n" , priv->id); -for (int i =0 ; i < lima_pipe_num; i++) { -struct lima_sched_pipe *pipe = &ldev->pipe[i]; -struct drm_gpu_scheduler *sched = &pipe->base; - -seq_printf (m,"drm-engine-%s:\t%llu ns\n" , sched->name, usage[i]); - } -} - -static const struct file_operations lima_drm_driver_fops = { - .owner =THIS_MODULE , -DRM_GEM_FOPS , - .show_fdinfo = lima_drm_driver_show_fdinfo, -}; +DEFINE_DRM_GEM_FOPS (lima_drm_driver_fops);/* * Changelog:
Signature · 1 line
2.25.1
Re: [PATCH 1/3] Revert "drm/lima: add show_fdinfo for drm usage stats"
Quoted 8 lines · On Tue, Apr 04, 2023 at 08:25:59AM +0800, [email protected] wrote:
From: Qiang Yu <[email protected]>
This reverts commit
4a66f3da99dcb4dcbd28544110636b50adfb0f0d.This is due to the depend commit has been reverted on upstream:
baad10973fdb("Revert "drm/scheduler: track GPU active time per entity"")Signed-off-by Qiang Yu <[email protected]>
A bit an aside, but it feels like a lot more of the fdinfo scheduler code should be common, and not just the minimal timekeeping? Just a thought for the next round.
-Daniel
Quoted 49 lines
--- drivers/gpu/drm/lima/lima_drv.c | 31 +------------------------------ 1 file changed, 1 insertion(+), 30 deletions(-)drivers/gpu/drm/lima/lima_drv.c +1
-30@@ -261,36 +261,7 @@ static const struct drm_ioctl_desc lima_drm_driver_ioctls[] = { DRM_IOCTL_DEF_DRV(LIMA_CTX_FREE, lima_ioctl_ctx_free, DRM_RENDER_ALLOW), }; -static void lima_drm_driver_show_fdinfo (struct seq_file *m,struct file *filp) -{ -struct drm_file *file = filp->private_data; -struct drm_device *dev = file->minor->dev; -struct lima_device *ldev =to_lima_dev (dev); -struct lima_drm_priv *priv = file->driver_priv; -struct lima_ctx_mgr *ctx_mgr = &priv->ctx_mgr; -u64 usage[lima_pipe_num]; - -lima_ctx_mgr_usage (ctx_mgr, usage); - -/* -* For a description of the text output format used here, see -* Documentation/gpu/drm-usage-stats.rst. -*/ -seq_printf (m,"drm-driver:\t%s\n" , dev->driver->name); -seq_printf (m,"drm-client-id:\t%u\n" , priv->id); -for (int i =0 ; i < lima_pipe_num; i++) { -struct lima_sched_pipe *pipe = &ldev->pipe[i]; -struct drm_gpu_scheduler *sched = &pipe->base; - -seq_printf (m,"drm-engine-%s:\t%llu ns\n" , sched->name, usage[i]); - } -} - -static const struct file_operations lima_drm_driver_fops = { - .owner =THIS_MODULE , -DRM_GEM_FOPS , - .show_fdinfo = lima_drm_driver_show_fdinfo, -}; +DEFINE_DRM_GEM_FOPS (lima_drm_driver_fops);/* * Changelog: Signature · 1 line
2.25.1
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch
PATCH 2/3 Revert "drm/lima: allocate unique id per drm_file"
In reply to [email protected]From: Qiang Yu <[email protected]>
This reverts commit 87767de835edf527b879a363d518c33da68adb81.
This is due to the depend commit has been reverted on upstream:baad10973fdb ("Revert "drm/scheduler: track GPU active time per entity"")
--- drivers/gpu/drm/lima/lima_device.h | 3 --- drivers/gpu/drm/lima/lima_drv.c | 12 ------------ drivers/gpu/drm/lima/lima_drv.h | 1 - 3 files changed, 16 deletions(-)
drivers/gpu/drm/lima/lima_device.h +0 -3
@@ -106,9 +106,6 @@ struct lima_device {struct lima_dump_head dump;struct list_head error_task_list;struct mutex error_task_list_lock; - -struct xarray active_contexts; -u32 next_context_id; }; static inline struct lima_device *
drivers/gpu/drm/lima/lima_drv.c +0 -12
@@ -218,11 +218,6 @@ static int lima_drm_driver_open(struct drm_device *dev, struct drm_file *file)if (!priv)return -ENOMEM ; - err =xa_alloc_cyclic (&ldev->active_contexts, &priv->id, priv, - xa_limit_32b, &ldev->next_context_id,GFP_KERNEL ); -if (err <0 ) - goto err_out0; - priv->vm =lima_vm_create (ldev);if (!priv->vm) { err = -ENOMEM ;
@@ -242,9 +237,6 @@ static int lima_drm_driver_open(struct drm_device *dev, struct drm_file *file)static void lima_drm_driver_postclose (struct drm_device *dev,struct drm_file *file) {struct lima_drm_priv *priv = file->driver_priv; -struct lima_device *ldev =to_lima_dev (dev); - -xa_erase (&ldev->active_contexts, priv->id);lima_ctx_mgr_fini (&priv->ctx_mgr);lima_vm_put (priv->vm);
@@ -396,8 +388,6 @@ static int lima_pdev_probe(struct platform_device *pdev) ldev->dev = &pdev->dev; ldev->id = (enum lima_gpu_id )of_device_get_match_data (&pdev->dev); -xa_init_flags (&ldev->active_contexts,XA_FLAGS_ALLOC ); -platform_set_drvdata (pdev, ldev);/* Allocate and initialize the DRM device. */
@@ -456,8 +446,6 @@ static int lima_pdev_remove(struct platform_device *pdev)struct lima_device *ldev =platform_get_drvdata (pdev);struct drm_device *ddev = ldev->ddev; -xa_destroy (&ldev->active_contexts); -sysfs_remove_bin_file (&ldev->dev->kobj, &lima_error_state_attr);drm_dev_unregister (ddev);
drivers/gpu/drm/lima/lima_drv.h +0 -1
@@ -20,7 +20,6 @@ struct lima_sched_task;struct drm_lima_gem_submit_bo ;struct lima_drm_priv { -int id;struct lima_vm *vm;struct lima_ctx_mgr ctx_mgr; };
Signature · 1 line
2.25.1
PATCH 3/3 Revert "drm/lima: add usage counting method to ctx_mgr"
In reply to [email protected]From: Qiang Yu <[email protected]>
This reverts commit bccafec957a5c4b22ac29e53a39e82d0a0008348.
This is due to the depend commit has been reverted on upstream:baad10973fdb ("Revert "drm/scheduler: track GPU active time per entity"")
--- drivers/gpu/drm/lima/lima_ctx.c | 30 +----------------------------- drivers/gpu/drm/lima/lima_ctx.h | 3 --- 2 files changed, 1 insertion(+), 32 deletions(-)
drivers/gpu/drm/lima/lima_ctx.c +1 -29
@@ -15,7 +15,6 @@ int lima_ctx_create(struct lima_device *dev, struct lima_ctx_mgr *mgr, u32 *id)if (!ctx)return -ENOMEM ; ctx->dev = dev; - ctx->mgr = mgr;kref_init (&ctx->refcnt);for (i =0 ; i < lima_pipe_num; i++) {
@@ -43,17 +42,10 @@ int lima_ctx_create(struct lima_device *dev, struct lima_ctx_mgr *mgr, u32 *id)static void lima_ctx_do_release (struct kref *ref) { struct lima_ctx *ctx = container_of(ref, struct lima_ctx, refcnt); -struct lima_ctx_mgr *mgr = ctx->mgr;int i; -for (i =0 ; i < lima_pipe_num; i++) { -struct lima_sched_context *context = &ctx->context[i]; -struct drm_sched_entity *entity = &context->base; - - mgr->elapsed_ns[i] += entity->elapsed_ns; - +for (i =0 ; i < lima_pipe_num; i++)lima_sched_context_fini (ctx->dev->pipe + i, ctx->context + i); - }kfree (ctx); }
@@ -107,23 +99,3 @@ void lima_ctx_mgr_fini(struct lima_ctx_mgr *mgr)xa_destroy (&mgr->handles);mutex_destroy (&mgr->lock); } - -void lima_ctx_mgr_usage (struct lima_ctx_mgr *mgr,u64 usage[lima_pipe_num]) -{ -struct lima_ctx *ctx; -unsigned long id; - -for (int i =0 ; i < lima_pipe_num; i++) - usage[i] = mgr->elapsed_ns[i]; - -mutex_lock (&mgr->lock); -xa_for_each (&mgr->handles, id, ctx) { -for (int i =0 ; i < lima_pipe_num; i++) { -struct lima_sched_context *context = &ctx->context[i]; -struct drm_sched_entity *entity = &context->base; - - usage[i] += entity->elapsed_ns; - } - } -mutex_unlock (&mgr->lock); -}
drivers/gpu/drm/lima/lima_ctx.h +0 -3
@@ -12,7 +12,6 @@struct lima_ctx {struct kref refcnt;struct lima_device *dev; -struct lima_ctx_mgr *mgr;struct lima_sched_context context[lima_pipe_num];atomic_t guilty;
@@ -24,7 +23,6 @@ struct lima_ctx {struct lima_ctx_mgr {struct mutex lock;struct xarray handles; -u64 elapsed_ns[lima_pipe_num]; };int lima_ctx_create (struct lima_device *dev,struct lima_ctx_mgr *mgr,u32 *id);
@@ -33,6 +31,5 @@ struct lima_ctx *lima_ctx_get(struct lima_ctx_mgr *mgr, u32 id);void lima_ctx_put (struct lima_ctx *ctx);void lima_ctx_mgr_init (struct lima_ctx_mgr *mgr);void lima_ctx_mgr_fini (struct lima_ctx_mgr *mgr); -void lima_ctx_mgr_usage (struct lima_ctx_mgr *mgr,u64 usage[lima_pipe_num]);#endif
Signature · 1 line
2.25.1
Quoted 9 lines · On Tue, 4 Apr 2023 at 08:13, <[email protected]> wrote:
From: Qiang Yu <[email protected]>
Upstream has reverted the dependant commit
df622729ddbf("drm/scheduler: track GPU active time per entity""), but this patchset has been pushed to drm-misc-next which still has that commit. To avoid other branch build fail after merge drm-misc-next, just revert the lima patchset on drm-misc-next too.
The bug/revert is unfortunate, although we better keep the build clean or Linus will go bananas ^_^
Fwiw for the series:
HTH
-Emil
Quoted 11 lines · Emil Velikov
Quoted 9 lines · On Tue, 4 Apr 2023 at 08:13, <[email protected]> wrote:
From: Qiang Yu <[email protected]>
Upstream has reverted the dependant commit
df622729ddbf("drm/scheduler: track GPU active time per entity""), but this patchset has been pushed to drm-misc-next which still has that commit. To avoid other branch build fail after merge drm-misc-next, just revert the lima patchset on drm-misc-next too.
Quoted 5 lines
The bug/revert is unfortunate, although we better keep the build clean or Linus will go bananas ^_^
Fwiw for the series:
Acked-by Emil Velikov <[email protected]>
Can you (eitehr of you really) please push asap and make sure this doesn't miss the last drm-misc-next pull (-rc6 is this w/e)? Otherwise we'll have a bit a mess.
-Daniel
Quoted 3 lines
HTH
-Emil
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch
Applied to drm-misc-next, sorry for the trouble.
Regards,
Qiang
Quoted 32 lines · Daniel Vetter
Quoted 11 lines · Emil Velikov
Quoted 9 lines · On Tue, 4 Apr 2023 at 08:13, <[email protected]> wrote:
From: Qiang Yu <[email protected]>
Upstream has reverted the dependant commit
df622729ddbf("drm/scheduler: track GPU active time per entity""), but this patchset has been pushed to drm-misc-next which still has that commit. To avoid other branch build fail after merge drm-misc-next, just revert the lima patchset on drm-misc-next too.Quoted 5 lines
The bug/revert is unfortunate, although we better keep the build clean or Linus will go bananas ^_^
Fwiw for the series:
Acked-by Emil Velikov <[email protected]>Can you (eitehr of you really) please push asap and make sure this doesn't miss the last drm-misc-next pull (-rc6 is this w/e)? Otherwise we'll have a bit a mess.
-DanielQuoted 3 lines
HTH
-Emil--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
Quoted 1 line · Qiang Yu
Applied to drm-misc-next, sorry for the trouble.
No worries, I already complained to Lucas/etnaviv people that the sched revert should have been at least posted/discussed on dri-devel. Imo this isn't on you.
-Daniel
Quoted 37 lines
Regards,
QiangQuoted 32 lines · Daniel Vetter
Quoted 11 lines · Emil Velikov
Quoted 9 lines · On Tue, 4 Apr 2023 at 08:13, <[email protected]> wrote:
From: Qiang Yu <[email protected]>
Upstream has reverted the dependant commit
df622729ddbf("drm/scheduler: track GPU active time per entity""), but this patchset has been pushed to drm-misc-next which still has that commit. To avoid other branch build fail after merge drm-misc-next, just revert the lima patchset on drm-misc-next too.Quoted 5 lines
The bug/revert is unfortunate, although we better keep the build clean or Linus will go bananas ^_^
Fwiw for the series:
Acked-by Emil Velikov <[email protected]>Can you (eitehr of you really) please push asap and make sure this doesn't miss the last drm-misc-next pull (-rc6 is this w/e)? Otherwise we'll have a bit a mess.
-DanielQuoted 3 lines
HTH
-Emil--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch