Some BO's might be mapped onto physical memory chunkwise and on demand,
like Panfrost's tiler heap. In this case, even though the
drm_gem_shmem_object page array might already be allocated, only a very
small fraction of the BO is currently backed by system memory, but
drm_show_memory_stats will then proceed to add its entire virtual size to
the file's total resident size regardless.
This led to very unrealistic RSS sizes being reckoned for Panfrost, where
said tiler heap buffer is initially allocated with a virtual size of 128
MiB, but only a small part of it will eventually be backed by system memory
after successive GPU page faults.
Provide a new DRM object generic function that would allow drivers to
return a more accurate RSS and purgeable sizes for their BOs.
Signed-off-by: Adrián Larumbe <adrian.larumbe@collabora.com>
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Reviewed-by: Steven Price <steven.price@arm.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230929181616.2769345-5-adrian.larumbe@collabora.com
Rename struct drm_gpuva_manager to struct drm_gpuvm including
corresponding functions. This way the GPUVA manager's structures align
very well with the documentation of VM_BIND [1] and VM_BIND locking [2].
It also provides a better foundation for the naming of data structures
and functions introduced for implementing a common dma-resv per GPU-VM
including tracking of external and evicted objects in subsequent
patches.
[1] Documentation/gpu/drm-vm-bind-async.rst
[2] Documentation/gpu/drm-vm-bind-locking.rst
Cc: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Cc: Matthew Brost <matthew.brost@intel.com>
Acked-by: Dave Airlie <airlied@redhat.com>
Acked-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Danilo Krummrich <dakr@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230920144343.64830-2-dakr@redhat.com
Once a BO is attached with slicing configuration that BO can only be used
for that particular setting. With this new feature user can detach slicing
configuration off an already sliced BO and attach new slicing configuration
using QAIC_ATTACH_SLICE_BO.
This will support BO recycling.
detach_slice_bo() detaches slicing configuration from a BO. This new
helper function can also be used in release_dbc() as we are doing the
exact same thing.
Signed-off-by: Pranjal Ramajor Asha Kanojiya <quic_pkanojiy@quicinc.com>
Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com>
[jhugo: add documentation for new ioctl]
Signed-off-by: Jeffrey Hugo <quic_jhugo@quicinc.com>
Reviewed-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230901172247.11410-8-quic_jhugo@quicinc.com
->size field in struct qaic_bo stores user requested buffer size for
allocate path or size of the dmabuf(PRIME). Now for allocate path driver
allocates a BO of size which is PAGE_SIZE aligned, this size is already
stored in base BO structure (struct drm_gem_object).
So difference is ->size of struct qaic_bo stores the raw value coming from
user and ->size in struct drm_gem_object stores the PAGE_SZIE aligned size.
Do not use ->size from struct qaic_bo for any validation or operation
instead use ->size from struct drm_gem_object since we already have
allocated that much memory then why not use it. Only validate if user
is trying to use more then the BO size. This make the driver more flexible.
After this change ->size field of struct qaic_bo becomes redundant. Remove
it.
Signed-off-by: Pranjal Ramajor Asha Kanojiya <quic_pkanojiy@quicinc.com>
Reviewed-by: Jeffrey Hugo <quic_jhugo@quicinc.com>
Signed-off-by: Jeffrey Hugo <quic_jhugo@quicinc.com>
Reviewed-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230901172247.11410-2-quic_jhugo@quicinc.com
With the typical model where the display server opens the file descriptor
and then hands it over to the client(*), we were showing stale data in
debugfs.
Fix it by updating the drm_file->pid on ioctl access from a different
process.
The field is also made RCU protected to allow for lockless readers. Update
side is protected with dev->filelist_mutex.
Before:
$ cat /sys/kernel/debug/dri/0/clients
command pid dev master a uid magic
Xorg 2344 0 y y 0 0
Xorg 2344 0 n y 0 2
Xorg 2344 0 n y 0 3
Xorg 2344 0 n y 0 4
After:
$ cat /sys/kernel/debug/dri/0/clients
command tgid dev master a uid magic
Xorg 830 0 y y 0 0
xfce4-session 880 0 n y 0 1
xfwm4 943 0 n y 0 2
neverball 1095 0 n y 0 3
*)
More detailed and historically accurate description of various handover
implementation kindly provided by Emil Velikov:
"""
The traditional model, the server was the orchestrator managing the
primary device node. From the fd, to the master status and
authentication. But looking at the fd alone, this has varied across
the years.
IIRC in the DRI1 days, Xorg (libdrm really) would have a list of open
fd(s) and reuse those whenever needed, DRI2 the client was responsible
for open() themselves and with DRI3 the fd was passed to the client.
Around the inception of DRI3 and systemd-logind, the latter became
another possible orchestrator. Whereby Xorg and Wayland compositors
could ask it for the fd. For various reasons (hysterical and genuine
ones) Xorg has a fallback path going the open(), whereas Wayland
compositors are moving to solely relying on logind... some never had
fallback even.
Over the past few years, more projects have emerged which provide
functionality similar (be that on API level, Dbus, or otherwise) to
systemd-logind.
"""
v2:
* Fixed typo in commit text and added a fine historical explanation
from Emil.
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: "Christian König" <christian.koenig@amd.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Acked-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Tested-by: Rob Clark <robdclark@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230621094824.2348732-1-tvrtko.ursulin@linux.intel.com
Signed-off-by: Christian König <christian.koenig@amd.com>
When building without CONFIG_DEBUG_FS:
drivers/gpu/drm/tegra/dc.c:1757:59: error: too many arguments to function call, expected 3, have 4
1757 | drm_debugfs_remove_files(dc->debugfs_files, count, root, minor);
| ~~~~~~~~~~~~~~~~~~~~~~~~ ^~~~~
include/drm/drm_debugfs.h:162:19: note: 'drm_debugfs_remove_files' declared here
162 | static inline int drm_debugfs_remove_files(const struct drm_info_list *files,
| ^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
163 | int count, struct drm_minor *minor)
| ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1 error generated.
Update the stub to include the root parameter.
Fixes: 8e455145d8 ("drm/debugfs: rework drm_debugfs_create_files implementation v2")
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230913-fix-drm_debugfs_remove_files-stub-v1-1-6b952ac559f3@kernel.org
Signed-off-by: Christian König <christian.koenig@amd.com>
Problem statement: The current method roundup_power_of_two()
to allocate contiguous address triggers -ENOSPC in some cases
even though we have enough free spaces and so to help with
that we introduce a try harder mechanism.
In case of -ENOSPC, the new try harder mechanism rounddown the
original size to power of 2 and iterating over the round down
sized freelist blocks to allocate the required size traversing
RHS and LHS.
As part of the above new method implementation we moved
contiguous/alignment size computation part and trim function
to the drm buddy file.
v2: Modify the alloc_range() function to return total allocated size
on -ENOSPC err and traverse RHS/LHS to allocate the required
size (Matthew).
Signed-off-by: Arunpravin Paneer Selvam <Arunpravin.PaneerSelvam@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230909160902.15644-1-Arunpravin.PaneerSelvam@amd.com
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
[Why]
Today, the allocation/deallocation steps and status is a bit unclear.
For instance, payload->vc_start_slot = -1 stands for "the failure of
updating DPCD payload ID table" and can also represent as "payload is not
allocated yet". These two cases should be handled differently and hence
better to distinguish them for better understanding.
[How]
Define enumeration - ALLOCATION_LOCAL, ALLOCATION_DFP and ALLOCATION_REMOTE
to distinguish different allocation status. Adjust the code to handle
different status accordingly for better understanding the sequence of
payload allocation and payload removement.
For payload creation, the procedure should look like this:
DRM part 1:
* step 1 - update sw mst mgr variables to add a new payload
* step 2 - add payload at immediate DFP DPCD payload table
Driver:
* Add new payload in HW and sync up with DFP by sending ACT
DRM Part 2:
* Send ALLOCATE_PAYLOAD sideband message to allocate bandwidth along the
virtual channel.
And as for payload removement, the procedure should look like this:
DRM part 1:
* step 1 - Send ALLOCATE_PAYLOAD sideband message to release bandwidth
along the virtual channel
* step 2 - Clear payload allocation at immediate DFP DPCD payload table
Driver:
* Remove the payload in HW and sync up with DFP by sending ACT
DRM part 2:
* update sw mst mgr variables to remove the payload
Note that it's fine to fail when communicate with the branch device
connected at immediate downstrean-facing port, but updating variables of
SW mst mgr and HW configuration should be conducted anyway. That's because
it's under commit_tail and we need to complete the HW programming.
Changes since v1:
* Remove the set but not use variable 'old_payload' in function
'nv50_msto_prepare'. Catched by kernel test robot <lkp@intel.com>
Signed-off-by: Wayne Lin <Wayne.Lin@amd.com>
Signed-off-by: Lyude Paul <lyude@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230807025639.1612361-3-Wayne.Lin@amd.com
The mutex was completely pointless in the first place since any
parallel adding of files to this list would result in random
behavior since the list is filled and consumed multiple times.
Completely drop that approach and just create the files directly but
return -ENODEV while opening the file when the minors are not
registered yet.
v2: rebase on debugfs directory rework, limit access before minors are
registered.
Signed-off-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230829110115.3442-5-christian.koenig@amd.com
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Instead of the per minor directories only create a single debugfs
directory for the whole device directly when the device is initialized.
For DRM devices each minor gets a symlink to the per device directory
for now until we can be sure that this isn't useful any more in any way.
Accel devices create only the per device directory and also drops the mid
layer callback to create driver specific files.
v2: cleanup accel component as well
v3: fix typo when debugfs is disabled
v4: call drm_debugfs_dev_fini() during release as well,
some kerneldoc typos fixed
v5: rebased and one more kerneldoc fix
Signed-off-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230829110115.3442-4-christian.koenig@amd.com
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Currently, NO_PREFETCH is passed implicitly through
drm_nouveau_gem_pushbuf_push::length and drm_nouveau_exec_push::va_len.
Since this is a direct representation of how the HW is programmed it
isn't really future proof for a uAPI. Hence, fix this up for the new
uAPI and split up the va_len field of struct drm_nouveau_exec_push,
such that we keep 32bit for va_len and 32bit for flags.
For drm_nouveau_gem_pushbuf_push::length at least provide
NOUVEAU_GEM_PUSHBUF_NO_PREFETCH to indicate the bit shift.
While at it, fix up nv50_dma_push() as well, such that the caller
doesn't need to encode the NO_PREFETCH flag into the length parameter.
Signed-off-by: Danilo Krummrich <dakr@redhat.com>
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230823181746.3446-1-dakr@redhat.com
This commit provides the interfaces for the new UAPI motivated by the
Vulkan API. It allows user mode drivers (UMDs) to:
1) Initialize a GPU virtual address (VA) space via the new
DRM_IOCTL_NOUVEAU_VM_INIT ioctl. UMDs can provide a kernel reserved
VA area.
2) Bind and unbind GPU VA space mappings via the new
DRM_IOCTL_NOUVEAU_VM_BIND ioctl.
3) Execute push buffers with the new DRM_IOCTL_NOUVEAU_EXEC ioctl.
Both, DRM_IOCTL_NOUVEAU_VM_BIND and DRM_IOCTL_NOUVEAU_EXEC support
asynchronous processing with DRM syncobjs as synchronization mechanism.
The default DRM_IOCTL_NOUVEAU_VM_BIND is synchronous processing,
DRM_IOCTL_NOUVEAU_EXEC supports asynchronous processing only.
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Reviewed-by: Dave Airlie <airlied@redhat.com>
Co-developed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Danilo Krummrich <dakr@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230804182406.5222-4-dakr@redhat.com
nouveau > 10 years ago had a plan for new multiplexer inside a multiplexer
API using nvif. It never fully reached fruition, fast forward 10 years,
and the new vulkan driver is avoiding libdrm and calling ioctls, and
these 3 ioctls, getparam, channel alloc + free don't seem to be things
we'd want to use nvif for.
Undeprecate and put them into the uapi header so we can just copy it
into mesa later.
v2: use uapi types.
Reviewed-by: Faith Ekstrand <faith.ekstrand@collabora.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Danilo Krummrich <dakr@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230804182406.5222-3-dakr@redhat.com
When no custom lock is set to protect a GEMs GPUVA list, lockdep checks
should fall back to the GEM objects dma-resv lock. With the current
implementation we're setting the lock_dep_map of the GEM objects 'resv'
pointer (in case no custom lock_dep_map is set yet) on
drm_gem_private_object_init().
However, the GEM objects 'resv' pointer might still change after
drm_gem_private_object_init() is called, e.g. through
ttm_bo_init_reserved(). This can result in the wrong lock being tracked.
To fix this, call dma_resv_held() directly from
drm_gem_gpuva_assert_lock_held() and fall back to the GEMs lock_dep_map
pointer only if an actual custom lock is set.
Fixes: e6303f323b ("drm: manager to keep track of GPUs VA mappings")
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Danilo Krummrich <dakr@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230804182406.5222-2-dakr@redhat.com
DRM bridges are not visible to the userspace and it may not be
immediately clear if the chain is somehow constructed incorrectly. I
have had two separate instances of a bridge driver failing to do a
drm_bridge_attach() call, resulting in the bridge connector not being
part of the chain. In some situations this doesn't seem to cause issues,
but it will if DRM_BRIDGE_ATTACH_NO_CONNECTOR flag is used.
Add a debugfs file to print the bridge chains. For me, on this TI AM62
based platform, I get the following output:
encoder[39]
bridge[0] type: 0, ops: 0x0
bridge[1] type: 0, ops: 0x0, OF: /bus@f0000/i2c@20000000/dsi@e:toshiba,tc358778
bridge[2] type: 0, ops: 0x3, OF: /bus@f0000/i2c@20010000/hdmi@48:lontium,lt8912b
bridge[3] type: 11, ops: 0x7, OF: /hdmi-connector:hdmi-connector
Tested-by: Alexander Stein <alexander.stein@ew.tq-group.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Acked-by: Neil Armstrong <neil.armstrong@linaro.org>
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20230802-drm-bridge-chain-debugfs-v4-1-7e3ae3d137c0@ideasonboard.com
These days, it's fairly common to see panels that have touchscreens
attached to them. The panel and the touchscreen can somewhat be
thought of as totally separate devices and, historically, this is how
Linux has treated them. However, treating them as separate isn't
necessarily the best way to model the two devices, it was just that
there was no better way. Specifically, there is little practical
reason to have the touchscreen powered on when the panel is turned
off, but if we model the devices separately we have no way to keep the
two devices' power states in sync with each other.
The issue described above makes it sound as if the problem here is
just about efficiency. We're wasting power keeping the touchscreen
powered up when the screen is off. While that's true, the problem can
go deeper. Specifically, hardware designers see that there's no reason
to have the touchscreen on while the screen is off and then build
hardware assuming that software would never turn the touchscreen on
while the screen is off.
In the very simplest case of hardware designs like this, the
touchscreen and the panel share some power rails. In most cases, this
turns out not to be terrible and is, again, just a little less
efficient. Specifically if we tell Linux that the touchscreen and the
panel are using the same rails then Linux will keep the rails on when
_either_ device is turned on. That ends to work OK-ish, but now if you
turn the panel off not only will the touchscreen remain powered, but
the power rails for the panel itself won't be switched off, burning
extra power.
The above two inefficiencies are _extra_ minor when you consider the
fact that laptops rarely spend much time with the screen off. The main
use case would be when an external screen (and presumably a power
supply) is attached.
Unfortunately, it gets worse from here. On sc7180-trogdor-homestar,
for instance, the display's TCON (timing controller) sometimes crashes
if you don't power cycle it whenever you stop and restart the video
stream (like during a modeset). The touchscreen keeping the power
rails on causes real problems. One proposal in the homestar timeframe
was to move the touchscreen to an always-on rail, dedicating the main
power rail to the panel. That caused _different_ problems as talked
about in commit 557e05fa9f ("HID: i2c-hid: goodix: Stop tying the
reset line to the regulator"). The end result of all of this was to
add an extra regulator to the board, increasing cost.
Recently, Cong Yang posted a patch [1] where things are even worse.
The panel and touch controller on that system seem even more
intimately tied together and really can't be thought of separately.
To address this issue, let's start allowing devices to register
themselves as "panel followers". These devices will get called after a
panel has been powered on and before a panel is powered off. This
makes the panel the primary device in charge of the power state, which
matches how userspace uses it.
The panel follower API should be fairly straightforward to use. The
current code assumes that panel followers are using device tree and
have a "panel" property pointing to the panel to follow. More
flexibility and non-DT implementations could be added as needed.
Right now, panel followers can follow the prepare/unprepare functions.
There could be arguments made that, instead, they should follow
enable/disable. I've chosen prepare/unprepare for now since those
functions are guaranteed to power up/power down the panel and it seems
better to start the process earlier.
A bit of explaining about why this is a roll-your-own API instead of
using something more standard:
1. In standard APIs in Linux, parent devices are automatically powered
on when a child needs power. Applying that here, it would mean that
we'd force the panel on any time someone was listening to the
touchscreen. That, unfortunately, would have broken homestar's need
(if we hadn't changed the hardware, as per above) where the panel
absolutely needs to be able to power cycle itself. While one could
argue that homestar is broken hardware and we shouldn't have the
API do backflips for it, _officially_ the eDP timing guidelines
agree with homestar's needs and the panel power sequencing diagrams
show power going off. It's nice to be able to support this.
2. We could, conceibably, try to add a new flag to device_link causing
the parent to be in charge of power. Then we could at least use
normal pm_runtime APIs. This sounds great, except that we run into
problems with initial probe. As talked about in the later patch
("HID: i2c-hid: Support being a panel follower") the initial power
on of a panel follower might need to do things (like add
sub-devices) that aren't allowed in a runtime_resume function.
The above complexities explain why this API isn't using common
functions. That being said, this patch is very small and
self-contained, so if someone was later able to adapt it to using more
common APIs while solving the above issues then that could happen in
the future.
[1] https://lore.kernel.org/r/20230519032316.3464732-1-yangcong5@huaqin.corp-partner.google.com
Reviewed-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20230727101636.v4.3.Icd5f96342d2242051c754364f4bee13ef2b986d4@changeid
In a whole pile of panel drivers, we have code to make the
prepare/unprepare/enable/disable callbacks behave as no-ops if they've
already been called. It's silly to have this code duplicated
everywhere. Add it to the core instead so that we can eventually
delete it from all the drivers. Note: to get some idea of the
duplicated code, try:
git grep 'if.*>prepared' -- drivers/gpu/drm/panel
git grep 'if.*>enabled' -- drivers/gpu/drm/panel
NOTE: arguably, the right thing to do here is actually to skip this
patch and simply remove all the extra checks from the individual
drivers. Perhaps the checks were needed at some point in time in the
past but maybe they no longer are? Certainly as we continue
transitioning over to "panel_bridge" then we expect there to be much
less variety in how these calls are made. When we're called as part of
the bridge chain, things should be pretty simple. In fact, there was
some discussion in the past about these checks [1], including a
discussion about whether the checks were needed and whether the calls
ought to be refcounted. At the time, I decided not to mess with it
because it felt too risky.
Looking closer at it now, I'm fairly certain that nothing in the
existing codebase is expecting these calls to be refcounted. The only
real question is whether someone is already doing something to ensure
prepare()/unprepare() match and enabled()/disable() match. I would say
that, even if there is something else ensuring that things match,
there's enough complexity that adding an extra bool and an extra
double-check here is a good idea. Let's add a drm_warn() to let people
know that it's considered a minor error to take advantage of
drm_panel's double-checking but we'll still make things work fine.
We'll also add an entry to the official DRM todo list to remove the
now pointless check from the panels after this patch lands and,
eventually, fixup anyone who is triggering the new warning.
[1] https://lore.kernel.org/r/20210416153909.v4.27.I502f2a92ddd36c3d28d014dd75e170c2d405a0a5@changeid
Acked-by: Neil Armstrong <neil.armstrong@linaro.org>
Reviewed-by: Maxime Ripard <mripard@kernel.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20230727101636.v4.2.I59b417d4c29151cc2eff053369ec4822b606f375@changeid
