From: "Christian König" <christian.koenig@amd.com>
To: Dave Airlie <airlied@gmail.com>, dri-devel@lists.freedesktop.org
Cc: sroland@vmware.com, linux-graphics-maintainer@vmware.com,
bskeggs@redhat.com, kraxel@redhat.com
Subject: Re: [PATCH 05/59] drm/ttm/amdgpu: consolidate ttm reserve paths
Date: Tue, 4 Aug 2020 12:33:32 +0200 [thread overview]
Message-ID: <8bc9e384-e0ff-ddd0-63c6-87e5843378e4@amd.com> (raw)
In-Reply-To: <20200804025632.3868079-6-airlied@gmail.com>
Am 04.08.20 um 04:55 schrieb Dave Airlie:
> From: Dave Airlie <airlied@redhat.com>
>
> Drop the WARN_ON and consolidate the two paths into one.
>
> Use the consolidate slowpath in the execbuf utils code.
>
> Signed-off-by: Dave Airlie <airlied@redhat.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_object.h | 2 +-
> drivers/gpu/drm/ttm/ttm_execbuf_util.c | 12 +--
> include/drm/ttm/ttm_bo_driver.h | 91 ++++------------------
> 3 files changed, 20 insertions(+), 85 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
> index afa5189dba7d..e01e8903741e 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
> @@ -160,7 +160,7 @@ static inline int amdgpu_bo_reserve(struct amdgpu_bo *bo, bool no_intr)
> struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev);
> int r;
>
> - r = __ttm_bo_reserve(&bo->tbo, !no_intr, false, NULL);
> + r = ttm_bo_reserve(&bo->tbo, !no_intr, false, NULL);
> if (unlikely(r != 0)) {
> if (r != -ERESTARTSYS)
> dev_err(adev->dev, "%p reserve failed\n", bo);
> diff --git a/drivers/gpu/drm/ttm/ttm_execbuf_util.c b/drivers/gpu/drm/ttm/ttm_execbuf_util.c
> index 1797f04c0534..8a8f1a6a83a6 100644
> --- a/drivers/gpu/drm/ttm/ttm_execbuf_util.c
> +++ b/drivers/gpu/drm/ttm/ttm_execbuf_util.c
> @@ -93,7 +93,7 @@ int ttm_eu_reserve_buffers(struct ww_acquire_ctx *ticket,
> list_for_each_entry(entry, list, head) {
> struct ttm_buffer_object *bo = entry->bo;
>
> - ret = __ttm_bo_reserve(bo, intr, (ticket == NULL), ticket);
> + ret = ttm_bo_reserve(bo, intr, (ticket == NULL), ticket);
> if (ret == -EALREADY && dups) {
> struct ttm_validate_buffer *safe = entry;
> entry = list_prev_entry(entry, head);
> @@ -119,13 +119,7 @@ int ttm_eu_reserve_buffers(struct ww_acquire_ctx *ticket,
> ttm_eu_backoff_reservation_reverse(list, entry);
>
> if (ret == -EDEADLK) {
> - if (intr) {
> - ret = dma_resv_lock_slow_interruptible(bo->base.resv,
> - ticket);
> - } else {
> - dma_resv_lock_slow(bo->base.resv, ticket);
> - ret = 0;
> - }
> + ret = ttm_bo_reserve_slowpath(bo, intr, ticket);
> }
>
> if (!ret && entry->num_shared)
> @@ -133,8 +127,6 @@ int ttm_eu_reserve_buffers(struct ww_acquire_ctx *ticket,
> entry->num_shared);
>
> if (unlikely(ret != 0)) {
> - if (ret == -EINTR)
> - ret = -ERESTARTSYS;
> if (ticket) {
> ww_acquire_done(ticket);
> ww_acquire_fini(ticket);
> diff --git a/include/drm/ttm/ttm_bo_driver.h b/include/drm/ttm/ttm_bo_driver.h
> index 09211ecbf84f..c20fef4da1d3 100644
> --- a/include/drm/ttm/ttm_bo_driver.h
> +++ b/include/drm/ttm/ttm_bo_driver.h
> @@ -588,29 +588,30 @@ int ttm_mem_io_lock(struct ttm_mem_type_manager *man, bool interruptible);
> void ttm_mem_io_unlock(struct ttm_mem_type_manager *man);
>
> /**
> - * __ttm_bo_reserve:
> + * ttm_bo_reserve:
> *
> * @bo: A pointer to a struct ttm_buffer_object.
> * @interruptible: Sleep interruptible if waiting.
> * @no_wait: Don't sleep while trying to reserve, rather return -EBUSY.
> * @ticket: ticket used to acquire the ww_mutex.
> *
> - * Will not remove reserved buffers from the lru lists.
> - * Otherwise identical to ttm_bo_reserve.
> + * Locks a buffer object for validation. (Or prevents other processes from
> + * locking it for validation), while taking a number of measures to prevent
> + * deadlocks.
> *
> * Returns:
> * -EDEADLK: The reservation may cause a deadlock.
> * Release all buffer reservations, wait for @bo to become unreserved and
> - * try again. (only if use_sequence == 1).
> + * try again.
> * -ERESTARTSYS: A wait for the buffer to become unreserved was interrupted by
> * a signal. Release all buffer reservations and return to user-space.
> * -EBUSY: The function needed to sleep, but @no_wait was true
> * -EALREADY: Bo already reserved using @ticket. This error code will only
> * be returned if @use_ticket is set to true.
> */
> -static inline int __ttm_bo_reserve(struct ttm_buffer_object *bo,
> - bool interruptible, bool no_wait,
> - struct ww_acquire_ctx *ticket)
> +static inline int ttm_bo_reserve(struct ttm_buffer_object *bo,
> + bool interruptible, bool no_wait,
> + struct ww_acquire_ctx *ticket)
> {
> int ret = 0;
>
> @@ -632,59 +633,6 @@ static inline int __ttm_bo_reserve(struct ttm_buffer_object *bo,
> return ret;
> }
>
> -/**
> - * ttm_bo_reserve:
> - *
> - * @bo: A pointer to a struct ttm_buffer_object.
> - * @interruptible: Sleep interruptible if waiting.
> - * @no_wait: Don't sleep while trying to reserve, rather return -EBUSY.
> - * @ticket: ticket used to acquire the ww_mutex.
> - *
> - * Locks a buffer object for validation. (Or prevents other processes from
> - * locking it for validation) and removes it from lru lists, while taking
> - * a number of measures to prevent deadlocks.
> - *
> - * Deadlocks may occur when two processes try to reserve multiple buffers in
> - * different order, either by will or as a result of a buffer being evicted
> - * to make room for a buffer already reserved. (Buffers are reserved before
> - * they are evicted). The following algorithm prevents such deadlocks from
> - * occurring:
> - * Processes attempting to reserve multiple buffers other than for eviction,
> - * (typically execbuf), should first obtain a unique 32-bit
> - * validation sequence number,
> - * and call this function with @use_ticket == 1 and @ticket->stamp == the unique
> - * sequence number. If upon call of this function, the buffer object is already
> - * reserved, the validation sequence is checked against the validation
> - * sequence of the process currently reserving the buffer,
> - * and if the current validation sequence is greater than that of the process
> - * holding the reservation, the function returns -EDEADLK. Otherwise it sleeps
> - * waiting for the buffer to become unreserved, after which it retries
> - * reserving.
> - * The caller should, when receiving an -EDEADLK error
> - * release all its buffer reservations, wait for @bo to become unreserved, and
> - * then rerun the validation with the same validation sequence. This procedure
> - * will always guarantee that the process with the lowest validation sequence
> - * will eventually succeed, preventing both deadlocks and starvation.
> - *
> - * Returns:
> - * -EDEADLK: The reservation may cause a deadlock.
> - * Release all buffer reservations, wait for @bo to become unreserved and
> - * try again. (only if use_sequence == 1).
> - * -ERESTARTSYS: A wait for the buffer to become unreserved was interrupted by
> - * a signal. Release all buffer reservations and return to user-space.
> - * -EBUSY: The function needed to sleep, but @no_wait was true
> - * -EALREADY: Bo already reserved using @ticket. This error code will only
> - * be returned if @use_ticket is set to true.
> - */
> -static inline int ttm_bo_reserve(struct ttm_buffer_object *bo,
> - bool interruptible, bool no_wait,
> - struct ww_acquire_ctx *ticket)
> -{
> - WARN_ON(!kref_read(&bo->kref));
> -
> - return __ttm_bo_reserve(bo, interruptible, no_wait, ticket);
> -}
> -
> /**
> * ttm_bo_reserve_slowpath:
> * @bo: A pointer to a struct ttm_buffer_object.
> @@ -699,20 +647,15 @@ static inline int ttm_bo_reserve_slowpath(struct ttm_buffer_object *bo,
> bool interruptible,
> struct ww_acquire_ctx *ticket)
> {
> - int ret = 0;
> -
> - WARN_ON(!kref_read(&bo->kref));
> -
> - if (interruptible)
> - ret = dma_resv_lock_slow_interruptible(bo->base.resv,
> - ticket);
> - else
> - dma_resv_lock_slow(bo->base.resv, ticket);
> -
> - if (ret == -EINTR)
> - ret = -ERESTARTSYS;
> -
> - return ret;
> + if (interruptible) {
> + int ret = dma_resv_lock_slow_interruptible(bo->base.resv,
> + ticket);
> + if (ret == -EINTR)
> + ret = -ERESTARTSYS;
> + return ret;
> + }
> + dma_resv_lock_slow(bo->base.resv, ticket);
> + return 0;
> }
>
> /**
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel
next prev parent reply other threads:[~2020-08-04 10:33 UTC|newest]
Thread overview: 136+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-08-04 2:55 [00/59] ttm misc cleanups, mem refactoring, rename objects. (v2) Dave Airlie
2020-08-04 2:55 ` [PATCH 01/59] drm/vmwgfx: consolidate ttm object creation and populate Dave Airlie
2020-08-05 7:56 ` daniel
2020-08-04 2:55 ` [PATCH 02/59] drm/vmwgfx: drop bo map/unmap dma functions Dave Airlie
2020-08-05 7:59 ` daniel
2020-08-04 2:55 ` [PATCH 03/59] nouveau: use ttm populate mapping functions. (v2) Dave Airlie
2020-08-05 5:32 ` Ben Skeggs
2020-08-04 2:55 ` [PATCH 04/59] qxl/ttm: drop the unusued no wait flag to reserve function Dave Airlie
2020-08-04 10:46 ` Gerd Hoffmann
2020-08-04 2:55 ` [PATCH 05/59] drm/ttm/amdgpu: consolidate ttm reserve paths Dave Airlie
2020-08-04 10:33 ` Christian König [this message]
2020-08-04 2:55 ` [PATCH 06/59] drm/ttm: use a helper for unlocked moves to the lru tail Dave Airlie
2020-08-04 10:34 ` Christian König
2020-08-05 5:32 ` Ben Skeggs
2020-08-04 2:55 ` [PATCH 07/59] drm/vram-helper: remove populate/unpopulate Dave Airlie
2020-08-04 6:54 ` Thomas Zimmermann
2020-08-04 2:55 ` [PATCH 08/59] drm/ttm: export memory type debug entrypoint Dave Airlie
2020-08-04 10:35 ` Christian König
2020-08-05 5:34 ` Ben Skeggs
2020-08-04 2:55 ` [PATCH 09/59] drm/nouveau/ttm: don't fill in blank ttm debug callback Dave Airlie
2020-08-05 5:34 ` Ben Skeggs
2020-08-04 2:55 ` [PATCH 10/59] drm/vmwgfx/gmrid: don't provide pointless " Dave Airlie
2020-08-04 2:55 ` [PATCH 11/59] drm/qxl/ttm: call ttm manager debug (v2) Dave Airlie
2020-08-04 10:48 ` Gerd Hoffmann
2020-08-04 2:55 ` [PATCH 12/59] drm/vram-helper: call the ttm manager debug function Dave Airlie
2020-08-04 6:55 ` Thomas Zimmermann
2020-08-04 10:48 ` Gerd Hoffmann
2020-08-04 2:55 ` [PATCH 13/59] drm/ttm: split the mm manager init code (v2) Dave Airlie
2020-08-04 11:07 ` Christian König
2020-08-04 13:08 ` Christian König
2020-08-04 11:10 ` Christian König
2020-08-04 2:55 ` [PATCH 14/59] drm/ttm: provide a driver-led init path for range mm manager. (v2) Dave Airlie
2020-08-04 2:55 ` [PATCH 15/59] drm/amdgpu/ttm: init managers from the driver side Dave Airlie
2020-08-04 11:15 ` Christian König
2020-08-04 2:55 ` [PATCH 16/59] drm/radeon: use new ttm man init path Dave Airlie
2020-08-04 11:15 ` Christian König
2020-08-04 2:55 ` [PATCH 17/59] drm/qxl/ttm: use new init path for manager Dave Airlie
2020-08-04 10:49 ` Gerd Hoffmann
2020-08-04 2:55 ` [PATCH 18/59] drm/vram_helper: use new ttm manager init function Dave Airlie
2020-08-04 6:58 ` Thomas Zimmermann
2020-08-04 10:49 ` Gerd Hoffmann
2020-08-04 2:55 ` [PATCH 19/59] drm/nouveau: use new memory manager init paths Dave Airlie
2020-08-05 5:40 ` Ben Skeggs
2020-08-04 2:55 ` [PATCH 20/59] drm/vmwgfx/ttm: convert vram mm init to new code paths Dave Airlie
2020-08-05 8:57 ` daniel
2020-08-04 2:55 ` [PATCH 21/59] drm/vmwgfx/ttm: switch gmrid allocator to new init paths Dave Airlie
2020-08-05 9:00 ` daniel
2020-08-04 2:55 ` [PATCH 22/59] drm/ttm: convert system manager init to new code Dave Airlie
2020-08-05 5:40 ` Ben Skeggs
2020-08-04 2:55 ` [PATCH 23/59] drm/ttm: purge old manager init path Dave Airlie
2020-08-05 5:41 ` Ben Skeggs
2020-08-04 2:55 ` [PATCH 24/59] drm/ttm: pass man around instead of mem_type in some places Dave Airlie
2020-08-05 5:42 ` Ben Skeggs
2020-08-04 2:55 ` [PATCH 25/59] drm/ttm: make some inline helper functions for cleanup paths. (v2) Dave Airlie
2020-08-04 11:18 ` Christian König
2020-08-05 5:42 ` Ben Skeggs
2020-08-04 2:55 ` [PATCH 26/59] drm/ttm: start allowing drivers to use new takedown path (v2) Dave Airlie
2020-08-04 11:20 ` Christian König
2020-08-05 5:43 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 27/59] drm/amdgpu/ttm: use new takedown path Dave Airlie
2020-08-04 2:56 ` [PATCH 28/59] drm/vmwgfx: takedown vram manager Dave Airlie
2020-08-05 9:19 ` daniel
2020-08-04 2:56 ` [PATCH 29/59] drm/vram_helper: call explicit mm takedown Dave Airlie
2020-08-04 6:59 ` Thomas Zimmermann
2020-08-04 2:56 ` [PATCH 30/59] drm/nouveau: use new cleanup paths Dave Airlie
2020-08-05 5:44 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 31/59] drm/radeon/ttm: use new takedown paths Dave Airlie
2020-08-04 2:56 ` [PATCH 32/59] drm/qxl/ttm: use new takedown path Dave Airlie
2020-08-04 10:50 ` Gerd Hoffmann
2020-08-04 2:56 ` [PATCH 33/59] drm/vmwgfx: fix gmrid takedown paths to new interface Dave Airlie
2020-08-05 9:21 ` daniel
2020-08-04 2:56 ` [PATCH 34/59] drm/ttm: remove range manager legacy takedown path Dave Airlie
2020-08-05 5:45 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 35/59] drm/ttm: make TTM responsible for cleaning system only Dave Airlie
2020-08-05 5:46 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 36/59] drm/ttm: add wrapper to get manager from bdev Dave Airlie
2020-08-04 11:25 ` Christian König
2020-08-05 5:47 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 37/59] drm/amdgfx/ttm: use wrapper to get ttm memory managers Dave Airlie
2020-08-04 11:26 ` Christian König
2020-08-04 2:56 ` [PATCH 38/59] drm/vram-helper: use wrapper to access " Dave Airlie
2020-08-04 7:00 ` Thomas Zimmermann
2020-08-04 2:56 ` [PATCH 39/59] drm/nouveau/ttm: " Dave Airlie
2020-08-05 5:48 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 40/59] drm/qxl/ttm: use wrapper to access memory manager Dave Airlie
2020-08-04 10:50 ` Gerd Hoffmann
2020-08-04 2:56 ` [PATCH 41/59] drm/radeon/ttm: " Dave Airlie
2020-08-04 11:29 ` Christian König
2020-08-04 2:56 ` [PATCH 42/59] drm/vmwgfx/ttm: " Dave Airlie
2020-08-05 9:22 ` daniel
2020-08-04 2:56 ` [PATCH 43/59] drm/ttm: rename manager variable to make sure wrapper is used Dave Airlie
2020-08-04 11:29 ` Christian König
2020-08-05 5:49 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 44/59] drm/ttm: allow drivers to provide their own manager subclasses Dave Airlie
2020-08-04 11:30 ` Christian König
2020-08-05 5:49 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 45/59] drm/amdgpu/ttm: use bo manager subclassing for vram/gtt mgrs Dave Airlie
2020-08-04 11:32 ` Christian König
2020-08-04 2:56 ` [PATCH 46/59] drm/ttm: make ttm_range_man_init/takedown take type + args Dave Airlie
2020-08-04 11:35 ` Christian König
2020-08-05 5:51 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 47/59] drm/ttm: move range manager to subclassed driver allocation Dave Airlie
2020-08-05 5:52 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 48/59] drm/vmwgfx/ttm: move thp to driver managed Dave Airlie
2020-08-05 9:24 ` daniel
2020-08-04 2:56 ` [PATCH 49/59] drm/vmwgfx/gmrid: convert to driver controlled allocation Dave Airlie
2020-08-05 9:26 ` daniel
2020-08-04 2:56 ` [PATCH 50/59] drm/nouveau/ttm: move to driver allocated manager Dave Airlie
2020-08-05 5:53 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 51/59] drm/ttm: drop priv pointer in memory manager Dave Airlie
2020-08-05 5:54 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 52/59] drm/amdgpu/ttm: remove man->bdev references Dave Airlie
2020-08-04 2:56 ` [PATCH 53/59] drm/ttm: drop man->bdev link Dave Airlie
2020-08-05 5:54 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 54/59] drm/ttm: drop list of memory managers from device. (v2) Dave Airlie
2020-08-04 11:37 ` Christian König
2020-08-05 5:55 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 55/59] drm/ttm: drop type manager has_type Dave Airlie
2020-08-04 11:37 ` Christian König
2020-08-05 5:55 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 56/59] drm/ttm: add a wrapper for checking if manager is in use Dave Airlie
2020-08-04 11:38 ` Christian König
2020-08-05 5:56 ` Ben Skeggs
2020-08-05 9:04 ` daniel
2020-08-04 2:56 ` [PATCH 57/59] drm/ttm: rename bo manager to range manager Dave Airlie
2020-08-04 11:40 ` Christian König
2020-08-05 5:56 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 58/59] drm/ttm: rename ttm_mem_type_manager -> ttm_resource_manager Dave Airlie
2020-08-04 11:41 ` Christian König
2020-08-05 5:57 ` Ben Skeggs
2020-08-04 2:56 ` [PATCH 59/59] drm/ttm: rename ttm_mem_reg to ttm_resource Dave Airlie
2020-08-04 11:41 ` Christian König
2020-08-05 5:58 ` Ben Skeggs
2020-08-04 3:01 ` [00/59] ttm misc cleanups, mem refactoring, rename objects. (v2) Dave Airlie
2020-08-05 11:55 ` Christian König
2020-08-05 11:59 ` Christian König
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=8bc9e384-e0ff-ddd0-63c6-87e5843378e4@amd.com \
--to=christian.koenig@amd.com \
--cc=airlied@gmail.com \
--cc=bskeggs@redhat.com \
--cc=dri-devel@lists.freedesktop.org \
--cc=kraxel@redhat.com \
--cc=linux-graphics-maintainer@vmware.com \
--cc=sroland@vmware.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;
as well as URLs for NNTP newsgroup(s).