Abstract the complexity of iterating over all the fences
in a dma_resv object.
The new loop handles the whole RCU and retry dance and
returns only fences where we can be sure we grabbed the
right one.
v2: fix accessing the shared fences while they might be freed,
improve kerneldoc, rename _cursor to _iter, add
dma_resv_iter_is_exclusive, add dma_resv_iter_begin/end
v3: restructor the code, move rcu_read_lock()/unlock() into the
iterator, add dma_resv_iter_is_restarted()
v4: fix NULL deref when no explicit fence exists, drop superflous
rcu_read_lock()/unlock() calls.
v5: fix typos in the documentation
v6: fix coding error when excl fence is NULL
v7: one more logic fix
v8: fix index check in dma_resv_iter_is_exclusive()
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> (v7)
Link: https://patchwork.freedesktop.org/patch/msgid/20211005113742.1101-2-christian.koenig@amd.com
Daniel pointed me towards this function and there are multiple obvious problems
in the implementation.
First of all the retry loop is not working as intended. In general the retry
makes only sense if you grab the reference first and then check the sequence
values.
Then we should always also wait for the exclusive fence.
It's also good practice to keep the reference around when installing callbacks
to fences you don't own.
And last the whole implementation was unnecessary complex and rather hard to
understand which could lead to probably unexpected behavior of the IOCTL.
Fix all this by reworking the implementation from scratch. Dropping the
whole RCU approach and taking the lock instead.
Only mildly tested and needs a thoughtful review of the code.
Pushing through drm-misc-next to avoid merge conflicts and give the code
another round of testing.
v2: fix the reference counting as well
v3: keep the excl fence handling as is for stable
v4: back to testing all fences, drop RCU
v5: handle in and out separately
v6: add missing clear of events
v7: change coding style as suggested by Michel, drop unused variables
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Tested-by: Michel Dänzer <mdaenzer@redhat.com>
CC: stable@vger.kernel.org
Link: https://patchwork.freedesktop.org/patch/msgid/20210720131110.88512-1-christian.koenig@amd.com
Specifically document the new/clarified rules around how the shared
fences do not have any ordering requirements against the exclusive
fence.
But also document all the things a bit better, given how central
struct dma_resv to dynamic buffer management the docs have been very
inadequat.
- Lots more links to other pieces of the puzzle. Unfortunately
ttm_buffer_object has no docs, so no links :-(
- Explain/complain a bit about dma_resv_locking_ctx(). I still don't
like that one, but fixing the ttm call chains is going to be
horrible. Plus we want to plug in real slowpath locking when we do
that anyway.
- Main part of the patch is some actual docs for struct dma_resv.
Overall I think we still have a lot of bad naming in this area (e.g.
dma_resv.fence is singular, but contains the multiple shared fences),
but I think that's more indicative of how the semantics and rules are
just not great.
Another thing that's real awkard is how chaining exclusive fences
right now means direct dma_resv.exclusive_fence pointer access with an
rcu_assign_pointer. Not so great either.
v2:
- Fix a pile of typos (Matt, Jason)
- Hammer it in that breaking the rules leads to use-after-free issues
around dma-buf sharing (Christian)
Reviewed-by: Christian König <christian.koenig@amd.com>
Cc: Jason Ekstrand <jason@jlekstrand.net>
Cc: Matthew Auld <matthew.auld@intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Cc: Sumit Semwal <sumit.semwal@linaro.org>
Cc: "Christian König" <christian.koenig@amd.com>
Cc: linux-media@vger.kernel.org
Cc: linaro-mm-sig@lists.linaro.org
Link: https://patchwork.freedesktop.org/patch/msgid/20210805104705.862416-21-daniel.vetter@ffwll.ch
It is expected from the clients to follow the below steps on an imported
dmabuf fd:
a) dmabuf = dma_buf_get(fd) // Get the dmabuf from fd
b) dma_buf_attach(dmabuf); // Clients attach to the dmabuf
o Here the kernel does some slab allocations, say for
dma_buf_attachment and may be some other slab allocation in the
dmabuf->ops->attach().
c) Client may need to do dma_buf_map_attachment().
d) Accordingly dma_buf_unmap_attachment() should be called.
e) dma_buf_detach () // Clients detach to the dmabuf.
o Here the slab allocations made in b) are freed.
f) dma_buf_put(dmabuf) // Can free the dmabuf if it is the last
reference.
Now say an erroneous client failed at step c) above thus it directly
called dma_buf_put(), step f) above. Considering that it may be the last
reference to the dmabuf, buffer will be freed with pending attachments
left to the dmabuf which can show up as the 'memory leak'. This should
at least be reported as the WARN().
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Reviewed-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/1627043468-16381-1-git-send-email-charante@codeaurora.org
Signed-off-by: Christian König <christian.koenig@amd.com>
drm-misc-next for v5.15-rc1:
UAPI Changes:
- Remove sysfs stats for dma-buf attachments, as it causes a performance regression.
Previous merge is not in a rc kernel yet, so no userspace regression possible.
Cross-subsystem Changes:
- Sanitize user input in kyro's viewport ioctl.
- Use refcount_t in fb_info->count
- Assorted fixes to dma-buf.
- Extend x86 efifb handling to all archs.
- Fix neofb divide by 0.
- Document corpro,gm7123 bridge dt bindings.
Core Changes:
- Slightly rework drm master handling.
- Cleanup vgaarb handling.
- Assorted fixes.
Driver Changes:
- Add support for ws2401 panel.
- Assorted fixes to stm, ast, bochs.
- Demidlayer ingenic irq.
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/2d0d2fe8-01fc-e216-c3fd-38db9e69944e@linux.intel.com
drm-misc-next for v5.15:
UAPI Changes:
Cross-subsystem Changes:
- udmabuf: Add support for mapping hugepages
- Add dma-buf stats to sysfs.
- Assorted fixes to fbdev/omap2.
- dma-buf: Document DMA_BUF_IOCTL_SYNC
- Improve dma-buf non-dynamic exporter expectations better.
- Add module parameters for dma-buf size and list limit.
- Add HDMI codec support to vc4, to replace vc4's own codec.
- Document dma-buf implicit fencing rules.
- dma_resv_test_signaled test_all handling.
Core Changes:
- Extract i915's eDP backlight code into DRM helpers.
- Assorted docbook updates.
- Rework drm_dp_aux documentation.
- Add support for the DP aux bus.
- Shrink dma-fence-chain slightly.
- Add alloc/free helpers for dma-fence-chain.
- Assorted fixes to TTM., drm/of, bridge
- drm_gem_plane_helper_prepare/cleanup_fb is now the default for gem drivers.
- Small fix for scheduler completion.
- Remove use of drm_device.irq_enabled.
- Print the driver name to dmesg when registering framebuffer.
- Export drm/gem's shadow plane handling, and use it in vkms.
- Assorted small fixes.
Driver Changes:
- Add eDP backlight to nouveau.
- Assorted fixes and cleanups to nouveau, panfrost, vmwgfx, anx7625,
amdgpu, gma500, radeon, mgag200, vgem, vc4, vkms, omapdrm.
- Add support for Samsung DB7430, Samsung ATNA33XC20, EDT ETMV570G2DHU,
EDT ETM0350G0DH6, Innolux EJ030NA panels.
- Fix some simple pannels missing bus_format and connector types.
- Add mks-guest-stats instrumentation support to vmwgfx.
- Merge i915-ttm topic branch.
- Make s6e63m0 panel use Mipi-DBI helpers.
- Add detect() supoprt for AST.
- Use interrupts for hotplug on vc4.
- vmwgfx is now moved to drm-misc-next, as sroland is no longer a maintainer for now.
- vmwgfx now uses copies of vmware's internal device headers.
- Slowly convert ti-sn65dsi83 over to atomic.
- Rework amdgpu dma-resv handling.
- Fix virtio fencing for planes.
- Ensure amdgpu can always evict to SYSTEM.
- Many drivers fixed for implicit fencing rules.
- Set default prepare/cleanup fb for tiny, vram and simple helpers too.
- Rework panfrost gpu reset and related serialization.
- Update VKMS todo list.
- Make bochs a tiny gpu driver, and use vram helper.
- Use linux irq interfaces instead of drm_irq in some drivers.
- Add support for Raspberry Pi Pico to GUD.
Signed-off-by: Dave Airlie <airlied@redhat.com>
# gpg: Signature made Fri 16 Jul 2021 21:06:04 AEST
# gpg: using RSA key B97BD6A80CAC4981091AE547FE558C72A67013C3
# gpg: Good signature from "Maarten Lankhorst <maarten.lankhorst@linux.intel.com>" [expired]
# gpg: aka "Maarten Lankhorst <maarten@debian.org>" [expired]
# gpg: aka "Maarten Lankhorst <maarten.lankhorst@canonical.com>" [expired]
# gpg: Note: This key has expired!
# Primary key fingerprint: B97B D6A8 0CAC 4981 091A E547 FE55 8C72 A670 13C3
From: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/444811c3-cbec-e9d5-9a6b-9632eda7962a@linux.intel.com
The DMA-BUF attachment statistics form a subset of the DMA-BUF
sysfs statistics that recently merged to the drm-misc tree. They are not
UABI yet since they have not merged to the upstream Linux kernel.
Since there has been a reported a performance regression due to the
overhead of sysfs directory creation/teardown during
dma_buf_attach()/dma_buf_detach(), this patch deletes the DMA-BUF
attachment statistics from sysfs.
Fixes: bdb8d06dfe ("dmabuf: Add the capability to expose DMA-BUF stats in sysfs")
Signed-off-by: Hridya Valsaraju <hridya@google.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20210713040742.2680135-1-hridya@google.com
Signed-off-by: Christian König <christian.koenig@amd.com>
Default list_limit and size_limit_mb are not big enough to cover all
possible use cases. For example, list_limit could be well over its default,
1024 if only one or several pages are chained in all individual list entries
when creating dmabuf backed by >4MB buffer. list_limit and size_limit_mb are
now defined as module parameters so that those can be optionally configured
by root with proper values to remove these constraints.
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Dongwon Kim <dongwon.kim@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20210611212107.9876-1-dongwon.kim@intel.com
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Christian and me realized we have a pretty massive disconnect about
different interpretations of what dma_resv is used for by different
drivers. The discussion is much, much bigger than this change here,
but this is an important one:
Non-dynamic exporters must guarantee that the memory they return is
ready for use. They cannot expect importers to wait for the exclusive
fence. Only dynamic importers are required to obey the dma_resv fences
strictly (and more patches are needed to define exactly what this
means).
Christian has patches to update nouvea, radeon and amdgpu. The only
other driver using both ttm and supporting dma-buf export is qxl,
which only uses synchronous ttm_bo_move.
v2: To hammer this in document that dynamic importers _must_ wait for
the exclusive fence after having called dma_buf_map_attachment.
Reviewed-by: Christian König <christian.koenig@amd.com>
Cc: Christian König <ckoenig.leichtzumerken@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210621151758.2347474-1-daniel.vetter@ffwll.ch
Overview
========
The patch adds DMA-BUF statistics to /sys/kernel/dmabuf/buffers. It
allows statistics to be enabled for each DMA-BUF in sysfs by enabling
the config CONFIG_DMABUF_SYSFS_STATS.
The following stats will be exposed by the interface:
/sys/kernel/dmabuf/buffers/<inode_number>/exporter_name
/sys/kernel/dmabuf/buffers/<inode_number>/size
/sys/kernel/dmabuf/buffers/<inode_number>/attachments/<attach_uid>/device
/sys/kernel/dmabuf/buffers/<inode_number>/attachments/<attach_uid>/map_counter
The inode_number is unique for each DMA-BUF and was added earlier [1]
in order to allow userspace to track DMA-BUF usage across different
processes.
Use Cases
=========
The interface provides a way to gather DMA-BUF per-buffer statistics
from production devices. These statistics will be used to derive DMA-BUF
per-exporter stats and per-device usage stats for Android Bug reports.
The corresponding userspace changes can be found at [2].
Telemetry tools will also capture this information(along with other
memory metrics) periodically as well as on important events like a
foreground app kill (which might have been triggered by Low Memory
Killer). It will also contribute to provide a snapshot of the system
memory usage on other events such as OOM kills and Application Not
Responding events.
Background
==========
Currently, there are two existing interfaces that provide information
about DMA-BUFs.
1) /sys/kernel/debug/dma_buf/bufinfo
debugfs is however unsuitable to be mounted in production systems and
cannot be considered as an alternative to the sysfs interface being
proposed.
2) proc/<pid>/fdinfo/<fd>
The proc/<pid>/fdinfo/<fd> files expose information about DMA-BUF fds.
However, the existing procfs interfaces can only provide information
about the buffers for which processes hold fds or have the buffers
mmapped into their address space. Since the procfs interfaces alone
cannot provide a full picture of all DMA-BUFs in the system, there is
the need for an alternate interface to provide this information on
production systems.
The patch contains the following major improvements over v1:
1) Each attachment is represented by its own directory to allow creating
a symlink to the importing device and to also provide room for future
expansion.
2) The number of distinct mappings of each attachment is exposed in a
separate file.
3) The per-buffer statistics are now in /sys/kernel/dmabuf/buffers
inorder to make the interface expandable in future.
All of the improvements above are based on suggestions/feedback from
Daniel Vetter and Christian König.
A shell script that can be run on a classic Linux environment to read
out the DMA-BUF statistics can be found at [3](suggested by John
Stultz).
[1]: https://lore.kernel.org/patchwork/patch/1088791/
[2]: https://android-review.googlesource.com/q/topic:%22dmabuf-sysfs%22+(status:open%20OR%20status:merged)
[3]: https://android-review.googlesource.com/c/platform/system/memory/libmeminfo/+/1549734
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Hridya Valsaraju <hridya@google.com>
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://patchwork.freedesktop.org/patch/msgid/20210603214758.2955251-1-hridya@google.com
If the VMM's (Qemu) memory backend is backed up by memfd + Hugepages
(hugetlbfs and not THP), we have to first find the hugepage(s) where
the Guest allocations are located and then extract the regular 4k
sized subpages from them.
v2: Ensure that the subpage and hugepage offsets are calculated correctly
when the range of subpage allocations cuts across multiple hugepages.
v3: Instead of repeatedly looking up the hugepage for each subpage,
only do it when the subpage allocation crosses over into a different
hugepage. (suggested by Gerd and DW)
v4: Fix the following warning identified by checkpatch:
CHECK:OPEN_ENDED_LINE: Lines should not end with a '('
Cc: Gerd Hoffmann <kraxel@redhat.com>
Signed-off-by: Vivek Kasireddy <vivek.kasireddy@intel.com>
Signed-off-by: Dongwon Kim <dongwon.kim@intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20210609182915.592743-1-vivek.kasireddy@intel.com
[ kraxel: one more checkpatch format tweak ]
Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
Allocate a new private stub fence in drm_syncobj_assign_null_handle,
instead of using a static stub fence.
When userspace creates a fence with DRM_SYNCOBJ_CREATE_SIGNALED or when
userspace signals a fence via DRM_IOCTL_SYNCOBJ_SIGNAL, the timestamp
obtained when the fence is exported and queried with SYNC_IOC_FILE_INFO
should match when the fence's status was changed from the perspective of
userspace, which is during the respective ioctl.
When a static stub fence started being used in by these ioctls, this
behavior changed. Instead, the timestamp returned by SYNC_IOC_FILE_INFO
became the first time anything used the static stub fence, which has no
meaning to userspace.
Signed-off-by: David Stevens <stevensd@chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20210408095428.3983055-1-stevensd@google.com
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Some drivers have hardware capability to get the precise HW timestamp
of certain events based on which the fences are triggered. The delta
between the event HW timestamp & current HW reference timestamp can
be used to calculate the timestamp in kernel's CLOCK_MONOTONIC time
domain. This allows it to set accurate timestamp factoring out any
software and IRQ latencies. Add a timestamp variant of fence signal
function, dma_fence_signal_timestamp to allow drivers to update the
precise timestamp for fences.
Changes in v2:
- Add a new fence signal variant instead of modifying fence struct
Changes in v3:
- Add timestamp domain information to commit-text and
dma_fence_signal_timestamp documentation
Signed-off-by: Veera Sundaram Sankaran <veeras@codeaurora.org>
Reviewed-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
[sumits: minor parenthesis alignment]
Link: https://patchwork.freedesktop.org/patch/msgid/1610757107-11892-1-git-send-email-veeras@codeaurora.org
(cherry picked from commit 5a164ac4db)
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
