fbdev: Disable sysfb device registration when removing conflicting FBs
The platform devices registered by sysfb match with firmware-based DRM or
fbdev drivers, that are used to have early graphics using a framebuffer
provided by the system firmware.
DRM or fbdev drivers later are probed and remove conflicting framebuffers,
leading to these platform devices for generic drivers to be unregistered.
But the current solution has a race, since the sysfb_init() function could
be called after a DRM or fbdev driver is probed and request to unregister
the devices for drivers with conflicting framebuffes.
To prevent this, disable any future sysfb platform device registration by
calling sysfb_disable(), if a driver requests to remove the conflicting
framebuffers.
drm/mgag200: Include <linux/vmalloc.h> for G200 BIOS code
After moving the vmalloc() call to another file, the rsp include
statement needs to be moved as well. Resolves a build warning on
parisc.
drivers/gpu/drm/mgag200/mgag200_g200.c: In function
'mgag200_g200_init_refclk':
drivers/gpu/drm/mgag200/mgag200_g200.c:120:16: error: implicit
declaration of function 'vmalloc'; did you mean 'kvmalloc'?
[-Werror=implicit-function-declaration]
Miaoqian Lin [Thu, 2 Jun 2022 10:42:22 +0000 (14:42 +0400)]
drm/virtio: Fix NULL vs IS_ERR checking in virtio_gpu_object_shmem_init
Since drm_prime_pages_to_sg() function return error pointers.
The drm_gem_shmem_get_sg_table() function returns error pointers too.
Using IS_ERR() to check the return value to fix this.
Vivek Kasireddy [Fri, 20 May 2022 20:52:35 +0000 (13:52 -0700)]
udmabuf: Set the DMA mask for the udmabuf device (v2)
If the DMA mask is not set explicitly, the following warning occurs
when the userspace tries to access the dma-buf via the CPU as
reported by syzbot here:
Xiaomeng Tong [Sun, 27 Mar 2022 05:09:45 +0000 (13:09 +0800)]
virtio-gpu: fix a missing check to avoid NULL dereference
'cache_ent' could be set NULL inside virtio_gpu_cmd_get_capset()
and it will lead to a NULL dereference by a lately use of it
(i.e., ptr = cache_ent->caps_cache). Fix it with a NULL check.
initmem_freed was removed in v2.1.124, and the underlying issue was
fixed for good in commit a65b279d3ed11df3 ("video/logo: prevent use of
logos after they have been freed").
Daniel Thompson [Wed, 8 Jun 2022 13:58:21 +0000 (14:58 +0100)]
drm/cma-helper: Describe what a "contiguous chunk" actually means
Since it's inception in 2012 it has been understood that the DRM GEM CMA
helpers do not depend on CMA as the backend allocator. In fact the first
bug fix to ensure the cma-helpers work correctly with an IOMMU backend
appeared in 2014. However currently the documentation for
drm_gem_cma_create() talks about "a contiguous chunk of memory" without
making clear which address space it will be a contiguous part of.
Additionally the CMA introduction is actively misleading because it only
contemplates the CMA backend.
This matters because when the device accesses the bus through an IOMMU
(and don't use the CMA backend) then the allocated memory is contiguous
only in the IOVA space. This is a significant difference compared to the
CMA backend and the behaviour can be a surprise even to someone who does
a reasonable level of code browsing (but doesn't find all the relevant
function pointers ;-) ).
Jason Ekstrand [Wed, 8 Jun 2022 15:21:42 +0000 (10:21 -0500)]
dma-buf: Add an API for importing sync files (v10)
This patch is analogous to the previous sync file export patch in that
it allows you to import a sync_file into a dma-buf. Unlike the previous
patch, however, this does add genuinely new functionality to dma-buf.
Without this, the only way to attach a sync_file to a dma-buf is to
submit a batch to your driver of choice which waits on the sync_file and
claims to write to the dma-buf. Even if said batch is a no-op, a submit
is typically way more overhead than just attaching a fence. A submit
may also imply extra synchronization with other work because it happens
on a hardware queue.
In the Vulkan world, this is useful for dealing with the out-fence from
vkQueuePresent. Current Linux window-systems (X11, Wayland, etc.) all
rely on dma-buf implicit sync. Since Vulkan is an explicit sync API, we
get a set of fences (VkSemaphores) in vkQueuePresent and have to stash
those as an exclusive (write) fence on the dma-buf. We handle it in
Mesa today with the above mentioned dummy submit trick. This ioctl
would allow us to set it directly without the dummy submit.
This may also open up possibilities for GPU drivers to move away from
implicit sync for their kernel driver uAPI and instead provide sync
files and rely on dma-buf import/export for communicating with other
implicit sync clients.
We make the explicit choice here to only allow setting RW fences which
translates to an exclusive fence on the dma_resv. There's no use for
read-only fences for communicating with other implicit sync userspace
and any such attempts are likely to be racy at best. When we got to
insert the RW fence, the actual fence we set as the new exclusive fence
is a combination of the sync_file provided by the user and all the other
fences on the dma_resv. This ensures that the newly added exclusive
fence will never signal before the old one would have and ensures that
we don't break any dma_resv contracts. We require userspace to specify
RW in the flags for symmetry with the export ioctl and in case we ever
want to support read fences in the future.
There is one downside here that's worth documenting: If two clients
writing to the same dma-buf using this API race with each other, their
actions on the dma-buf may happen in parallel or in an undefined order.
Both with and without this API, the pattern is the same: Collect all
the fences on dma-buf, submit work which depends on said fences, and
then set a new exclusive (write) fence on the dma-buf which depends on
said work. The difference is that, when it's all handled by the GPU
driver's submit ioctl, the three operations happen atomically under the
dma_resv lock. If two userspace submits race, one will happen before
the other. You aren't guaranteed which but you are guaranteed that
they're strictly ordered. If userspace manages the fences itself, then
these three operations happen separately and the two render operations
may happen genuinely in parallel or get interleaved. However, this is a
case of userspace racing with itself. As long as we ensure userspace
can't back the kernel into a corner, it should be fine.
v2 (Jason Ekstrand):
- Use a wrapper dma_fence_array of all fences including the new one
when importing an exclusive fence.
v3 (Jason Ekstrand):
- Lock around setting shared fences as well as exclusive
- Mark SIGNAL_SYNC_FILE as a read-write ioctl.
- Initialize ret to 0 in dma_buf_wait_sync_file
v4 (Jason Ekstrand):
- Use the new dma_resv_get_singleton helper
v5 (Jason Ekstrand):
- Rename the IOCTLs to import/export rather than wait/signal
- Drop the WRITE flag and always get/set the exclusive fence
v6 (Jason Ekstrand):
- Split import and export into separate patches
- New commit message
v7 (Daniel Vetter):
- Fix the uapi header to use the right struct in the ioctl
- Use a separate dma_buf_import_sync_file struct
- Add kerneldoc for dma_buf_import_sync_file
v8 (Jason Ekstrand):
- Rebase on Christian König's fence rework
v9 (Daniel Vetter):
- Fix -EINVAL checks for the flags parameter
- Add documentation about read/write fences
- Add documentation about the expected usage of import/export and
specifically call out the possible userspace race.
v10 (Simon Ser):
- Fix a typo in the docs
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net> Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com> Signed-off-by: Jason Ekstrand <jason.ekstrand@collabora.com> Reviewed-by: Christian König <christian.koenig@amd.com> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Sumit Semwal <sumit.semwal@linaro.org> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Signed-off-by: Simon Ser <contact@emersion.fr> Link: https://patchwork.freedesktop.org/patch/msgid/20220608152142.14495-3-jason@jlekstrand.net
Jason Ekstrand [Wed, 8 Jun 2022 15:21:41 +0000 (10:21 -0500)]
dma-buf: Add an API for exporting sync files (v14)
Modern userspace APIs like Vulkan are built on an explicit
synchronization model. This doesn't always play nicely with the
implicit synchronization used in the kernel and assumed by X11 and
Wayland. The client -> compositor half of the synchronization isn't too
bad, at least on intel, because we can control whether or not i915
synchronizes on the buffer and whether or not it's considered written.
The harder part is the compositor -> client synchronization when we get
the buffer back from the compositor. We're required to be able to
provide the client with a VkSemaphore and VkFence representing the point
in time where the window system (compositor and/or display) finished
using the buffer. With current APIs, it's very hard to do this in such
a way that we don't get confused by the Vulkan driver's access of the
buffer. In particular, once we tell the kernel that we're rendering to
the buffer again, any CPU waits on the buffer or GPU dependencies will
wait on some of the client rendering and not just the compositor.
This new IOCTL solves this problem by allowing us to get a snapshot of
the implicit synchronization state of a given dma-buf in the form of a
sync file. It's effectively the same as a poll() or I915_GEM_WAIT only,
instead of CPU waiting directly, it encapsulates the wait operation, at
the current moment in time, in a sync_file so we can check/wait on it
later. As long as the Vulkan driver does the sync_file export from the
dma-buf before we re-introduce it for rendering, it will only contain
fences from the compositor or display. This allows to accurately turn
it into a VkFence or VkSemaphore without any over-synchronization.
By making this an ioctl on the dma-buf itself, it allows this new
functionality to be used in an entirely driver-agnostic way without
having access to a DRM fd. This makes it ideal for use in driver-generic
code in Mesa or in a client such as a compositor where the DRM fd may be
hard to reach.
v2 (Jason Ekstrand):
- Use a wrapper dma_fence_array of all fences including the new one
when importing an exclusive fence.
v3 (Jason Ekstrand):
- Lock around setting shared fences as well as exclusive
- Mark SIGNAL_SYNC_FILE as a read-write ioctl.
- Initialize ret to 0 in dma_buf_wait_sync_file
v4 (Jason Ekstrand):
- Use the new dma_resv_get_singleton helper
v5 (Jason Ekstrand):
- Rename the IOCTLs to import/export rather than wait/signal
- Drop the WRITE flag and always get/set the exclusive fence
v6 (Jason Ekstrand):
- Drop the sync_file import as it was all-around sketchy and not nearly
as useful as import.
- Re-introduce READ/WRITE flag support for export
- Rework the commit message
v7 (Jason Ekstrand):
- Require at least one sync flag
- Fix a refcounting bug: dma_resv_get_excl() doesn't take a reference
- Use _rcu helpers since we're accessing the dma_resv read-only
v8 (Jason Ekstrand):
- Return -ENOMEM if the sync_file_create fails
- Predicate support on IS_ENABLED(CONFIG_SYNC_FILE)
v9 (Jason Ekstrand):
- Add documentation for the new ioctl
v10 (Jason Ekstrand):
- Go back to dma_buf_sync_file as the ioctl struct name
v11 (Daniel Vetter):
- Go back to dma_buf_export_sync_file as the ioctl struct name
- Better kerneldoc describing what the read/write flags do
v12 (Christian König):
- Document why we chose to make it an ioctl on dma-buf
v13 (Jason Ekstrand):
- Rebase on Christian König's fence rework
v14 (Daniel Vetter & Christian König):
- Use dma_rev_usage_rw to get the properly inverted usage to pass to
dma_resv_get_singleton()
- Clean up the sync_file and fd if copy_to_user() fails
Signed-off-by: Jason Ekstrand <jason@jlekstrand.net> Signed-off-by: Jason Ekstrand <jason.ekstrand@intel.com> Signed-off-by: Jason Ekstrand <jason.ekstrand@collabora.com> Acked-by: Simon Ser <contact@emersion.fr> Reviewed-by: Christian König <christian.koenig@amd.com> Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch> Cc: Sumit Semwal <sumit.semwal@linaro.org> Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Signed-off-by: Simon Ser <contact@emersion.fr> Link: https://patchwork.freedesktop.org/patch/msgid/20220608152142.14495-2-jason@jlekstrand.net
Systems with AST graphics can have multiple output; typically VGA
plus some other port. Record detected output chips in a bitmask and
initialize each output on its own.
Assume a VGA output by default and use SIL164 and DP501 if available.
For ASTDP assume that it can run in parallel with VGA.
Tested on AST2100.
v3:
* define a macro for each BIT(ast_tx_chip) (Patrik)
v2:
* make VGA/SIL164/DP501 mutually exclusive
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de> Reviewed-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com> Fixes: 02e861909549 ("drm/ast: Initialize encoder and connector for VGA in helper function") Cc: Thomas Zimmermann <tzimmermann@suse.de> Cc: Javier Martinez Canillas <javierm@redhat.com> Cc: Dave Airlie <airlied@redhat.com> Cc: dri-devel@lists.freedesktop.org Link: https://patchwork.freedesktop.org/patch/msgid/20220607092008.22123-2-tzimmermann@suse.de
Bjorn Andersson [Wed, 1 Jun 2022 23:38:18 +0000 (16:38 -0700)]
drm/bridge: lt9611uxc: Cancel only driver's work
During device remove care needs to be taken that no work is pending
before it removes the underlying DRM bridge etc, but this can be done on
the specific work rather than waiting for the flush of the system-wide
workqueue.
Miaoqian Lin [Wed, 1 Jun 2022 03:39:27 +0000 (07:39 +0400)]
drm/meson: encoder_hdmi: Fix refcount leak in meson_encoder_hdmi_init
of_graph_get_remote_node() returns remote device nodepointer with
refcount incremented, we should use of_node_put() on it when done.
Add missing of_node_put() to avoid refcount leak.
Miaoqian Lin [Wed, 1 Jun 2022 03:39:26 +0000 (07:39 +0400)]
drm/meson: encoder_cvbs: Fix refcount leak in meson_encoder_cvbs_init
of_graph_get_remote_node() returns remote device nodepointer with
refcount incremented, we should use of_node_put() on it when done.
Add missing of_node_put() to avoid refcount leak.
Fixes: 5d4774fcb39f ("drm/meson: encoder_cvbs: switch to bridge with ATTACH_NO_CONNECTOR") Signed-off-by: Miaoqian Lin <linmq006@gmail.com> Reviewed-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> Link: https://patchwork.freedesktop.org/patch/msgid/20220601033927.47814-2-linmq006@gmail.com
Set new vidrst flag in device info for models that synchronize with
external sources (i.e., BMCs). In modesetting, set the corresponding
bits from the device-info flag.
drm/mgag200: Store maximum resolution and memory bandwidth in device info
The maximum resolution and memory bandwidth are model-specific limits.
Both are used during display-mode validation. Store the values in struct
mgag200_device_info and simplify the validation code.
drm/mgag200: Store HW_BUG_NO_STARTADD flag in device info
Flag devices with broken handling of the startadd field in
struct mgag200_device_info, instead of PCI driver data. This
reduces the driver data to a simple type constant.
Rework mgag200_regs_init() and mgag200_mm_init() into device preinit
and init functions. The preinit function, mgag200_device_preinit(),
requests and maps a device's I/O and video memory. The init function,
mgag200_device_init() initializes the state of struct mga_device.
Splitting the initialization between the two functions is necessary
to perform per-model operations between the two calls, such as reading
the unique revision ID on G200SEs.
drm/mgag200: Move PCI-option setup into model-specific code
Split the PCI code into a single call for each model. G200 and G200SE
each contain a dedicated helper with additional instructions. Noteably,
the G200ER has no code for PCI setup.
In a later patch, the magic numbers should be replaced by descriptive
constants.
drm/mgag200: Remove special case for G200SE with <2 MiB
Remove old test for 32-bit vs 16-bit colors. Prefer 24-bit color depth
on all devices. 32-bit color depth doesn't exist, it should have always
been 24-bit.
G200SE with less than 2 MiB of video memory have defaulted to 16-bit
color depth, as the original revision of the G200SE had only 1.75 MiB
of video memory. Using 16-bit colors enabled XGA resolution. But we
now already limit these devices to VGA resolutions as the memory-bandwith
test assumes 32-bit pixel size. So drop the special case from color-depth
selection.
drm/probe-helper: Default to 640x480 if no EDID on DP
If we're unable to read the EDID for a display because it's corrupt /
bogus / invalid then we'll add a set of standard modes for the
display. Since we have no true information about the connected
display, these modes are essentially guesses but better than nothing.
At the moment, none of the modes returned is marked as preferred, but
the modes are sorted such that the higher resolution modes are listed
first.
When userspace sees these modes presented by the kernel it needs to
figure out which one to pick. At least one userspace, ChromeOS [1]
seems to use the rules (which seem pretty reasonable):
1. Try to pick the first mode marked as preferred.
2. Try to pick the mode which matches the first detailed timing
descriptor in the EDID.
3. If no modes were marked as preferred then pick the first mode.
Unfortunately, userspace's rules combined with what the kernel is
doing causes us to fail section 4.2.2.6 (EDID Corruption Detection) of
the DP 1.4a Link CTS. That test case says that, while it's OK to allow
some implementation-specific fall-back modes if the EDID is bad that
userspace should _default_ to 640x480.
Let's fix this by marking 640x480 as default for DP in the no-EDID
case.
NOTES:
- In the discussion around v3 of this patch [2] there was talk about
solving this in userspace and I even implemented a patch that would
have solved this for ChromeOS, but then the discussion turned back
to solving this in the kernel.
- Also in the discussion of v3 [2] it was requested to limit this
change to just DP since folks were worried that it would break some
subtle corner case on VGA or HDMI.
Timur Tabi [Wed, 11 May 2022 16:37:16 +0000 (11:37 -0500)]
drm/nouveau: fix another off-by-one in nvbios_addr
This check determines whether a given address is part of
image 0 or image 1. Image 1 starts at offset image0_size,
so that address should be included.
Fixes: a5fc52e4bf667 ("drm/nouveau/bios: guard against out-of-bounds accesses to image") Cc: <stable@vger.kernel.org> # v4.8+ Signed-off-by: Timur Tabi <ttabi@nvidia.com> Reviewed-by: Karol Herbst <kherbst@redhat.com> Signed-off-by: Lyude Paul <lyude@redhat.com> Link: https://patchwork.freedesktop.org/patch/msgid/20220511163716.3520591-1-ttabi@nvidia.com
Douglas Anderson [Tue, 10 May 2022 19:29:44 +0000 (12:29 -0700)]
drm/bridge: parade-ps8640: Handle DP AUX more properly
While it works, for the most part, to assume that the panel has
finished probing when devm_of_dp_aux_populate_ep_devices() returns,
it's a bit fragile. This is talked about at length in commit 0f27f5316c50 ("drm/bridge: ti-sn65dsi86: Promote the AUX channel to
its own sub-dev").
When reviewing the ps8640 code, I managed to convince myself that it
was OK not to worry about it there and that maybe it wasn't really
_that_ fragile. However, it turns out that it really is. Simply
hardcoding panel_edp_probe() to return -EPROBE_DEFER was enough to put
the boot process into an infinite loop. I believe this manages to trip
the same issues that we used to trip with the main MSM code where
something about our actions trigger Linux to re-probe previously
deferred devices right away and each time we try again we re-trigger
Linux to re-probe.
Let's fix this using the callback introduced in the patch ("drm/dp:
Callbacks to make it easier for drivers to use DP AUX bus properly").
When using the new callback, we have to be a little careful. The
probe_done() callback is no longer always called in the context of
our probe routine. That means we can't rely on being able to return
-EPROBE_DEFER from it. We re-jigger the order of things a bit to
account for that.
With this change, the device still boots (though obviously the panel
doesn't come up) if I force panel-edp to always return
-EPROBE_DEFER. If I fake it and make the panel probe exactly once it
also works.
Douglas Anderson [Tue, 10 May 2022 19:29:43 +0000 (12:29 -0700)]
drm/bridge: Add devm_drm_bridge_add()
This adds a devm managed version of drm_bridge_add(). Like other
"devm" function listed in drm_bridge.h, this function takes an
explicit "dev" to use for the lifetime management. A few notes:
* In general we have a "struct device" for bridges that makes a good
candidate for where the lifetime matches exactly what we want.
* The "bridge->dev->dev" device appears to be the encoder
device. That's not the right device to use for lifetime management.
Douglas Anderson [Tue, 10 May 2022 19:29:42 +0000 (12:29 -0700)]
drm/dp: Add callbacks to make using DP AUX bus properly easier
As talked about in this patch in the kerneldoc of
of_dp_aux_populate_ep_device() and also in the past in commit 0f27f5316c50 ("drm/bridge: ti-sn65dsi86: Promote the AUX channel to
its own sub-dev"), it can be difficult for eDP controller drivers to
know when the panel has finished probing when they're using
of_dp_aux_populate_ep_devices().
The ti-sn65dsi86 driver managed to solve this because it was already
broken up into a bunch of sub-drivers. That means we could solve the
problem there by adding a new sub-driver to get the panel. We could
use the traditional -EPROBE_DEFER retry mechansim to handle the case
where the panel hadn't probed yet.
In parade-ps8640 we didn't really solve this. The code just expects
the panel to be ready right away. While reviewing the code originally
I had managed to convince myself it was fine to just expect the panel
right away, but additional testing has shown that not to be the
case. We could fix parade-ps8640 like we did ti-sn65dsi86 but it's
pretty cumbersome (since we're not already broken into multiple
drivers) and requires a bunch of boilerplate code.
After discussion [1] it seems like the best solution for most people
is:
- Accept that there's always at most one device that will probe as a
result of the DP AUX bus (it may have sub-devices, but there will be
one device _directly_ probed).
- When that device finishes probing, we can just have a call back.
This patch implements that idea. We'll now take a callback as an
argument to the populate function. To make this easier to land in
pieces, we'll make wrappers for the old functions. The functions with
the new name (which make it clear that we only have one child) will
take the callback and the functions with the old name will temporarily
wrap.
drm/gma500: Read EDID from the correct i2c adapter
Someone made the mistake to try reading EDID from the backlight i2c
adapter. This has been wrong for a very long time but since we read out
the modes correctly on init and don't hotplug lvds it has been working
anyway. Correct this by using connector->ddc instead of
encoder->i2c_bus. Both PSB and CDV are affected but this bug.
drm/gma500: Make oaktrail lvds use ddc adapter from drm_connector
We're moving all uses of ddc_bus to drm_connector where they belong.
The initialization of the gma_i2c_chan for Oaktrail is a bit backwards
so it required improvements. Also cleanup the error handling in
oaktrail_lvds_init(). Since this is the last user of
gma_encoder->ddc_bus we can remove it.
drm/gma500: Make psb lvds use ddc adapter from drm_connector
We're moving all uses of ddc_bus to drm_connector where they belong.
Also cleanup the error handling in psb_intel_lvds_init() and remove
unused ddc_bus in psb_intel_lvds_priv.
drm/gma500: Make cdv hdmi use ddc adapter from drm_connector
We're moving all uses of ddc_bus from gma_encoder to drm_connector where
they belong. Also, cleanup the error handling in cdv_hdmi_init()
and remove unused i2c pointer in mid_intel_hdmi_priv.
drm/gma500: Make cdv lvds use ddc adapter from drm_connector
We're moving all uses of ddc_bus to drm_connector where they belong.
Also, add missing call to destroy ddc bus when destroying the connector
and cleanup the error handling in cdv_intel_lvds_init().
drm/ssd130x: Only define a SPI device ID table when built as a module
The kernel test robot reports a compile warning due the ssd130x_spi_table
variable being defined but not used. This happen when ssd130x-spi driver
is built-in instead of being built as a module, i.e:
CC drivers/gpu/drm/solomon/ssd130x-spi.o
AR drivers/base/firmware_loader/built-in.a
AR drivers/base/built-in.a
CC kernel/trace/trace.o
drivers/gpu/drm/solomon/ssd130x-spi.c:155:35: warning: ‘ssd130x_spi_table’ defined but not used [-Wunused-const-variable=]
155 | static const struct spi_device_id ssd130x_spi_table[] = {
| ^~~~~~~~~~~~~~~~~
The driver shouldn't need a SPI device ID table and only have an OF device
ID table, but the former is needed to workaround an issue in the SPI core.
This always reports a MODALIAS of the form "spi:<device>" even for devices
registered through Device Trees.
But the table is only needed when the driver built as a module to populate
the .ko alias info. It's not needed when the driver is built-in the kernel.
Saurabh Sengar [Sat, 21 May 2022 14:23:39 +0000 (07:23 -0700)]
drm/hyperv : Removing the restruction of VRAM allocation with PCI bar size
There were two different approaches getting used in this driver to
allocate vram:
1. VRAM allocation from PCI region for Gen1
2. VRAM alloaction from MMIO region for Gen2
First approach limilts the vram to PCI BAR size, which is 64 MB in most
legacy systems. This limits the maximum resolution to be restricted to
64 MB size, and with recent conclusion on fbdev issue its concluded to have
similar allocation strategy for both Gen1 and Gen2. This patch unifies
the Gen1 and Gen2 vram allocation strategy.
Christian König [Mon, 25 Apr 2022 12:22:12 +0000 (14:22 +0200)]
dma-buf: generalize dma_fence unwrap & merging v3
Introduce a dma_fence_unwrap_merge() macro which allows to unwrap fences
which potentially can be containers as well and then merge them back
together into a flat dma_fence_array.
v2: rename the function, add some more comments about how the wrapper is
used, move filtering of signaled fences into the unwrap iterator,
add complex selftest which covers more cases.
v3: fix signaled fence filtering once more
Yunhao Tian [Tue, 10 May 2022 03:02:19 +0000 (11:02 +0800)]
drm/mipi-dbi: align max_chunk to 2 in spi_transfer
In __spi_validate, there's a validation that no partial transfers
are accepted (xfer->len % w_size must be zero). When
max_chunk is not a multiple of bpw (e.g. max_chunk = 65535,
bpw = 16), the transfer will be rejected.
This patch aligns max_chunk to 2 bytes (the maximum value of bpw is 16),
so that no partial transfer will occur.
Add the features, issues, and GPU ID for Mali-G57, a first-generation
Valhall GPU. Other first- and second-generation Valhall GPUs should be
similar.
v2: Split out issue list for r0p0 from newer Natt GPUs, as TTRX_3485 was
fixed in r0p1. Unfortunately, MT8192 has a r0p0, so we do need to handle
TTRX_3485.
L2_MMU_CONFIG is an implementation-defined register. Different Mali GPUs
define slightly different MAX_READS and MAX_WRITES fields, which
throttle outstanding reads and writes when set to non-zero values. When
left as zero, reads and writes are not throttled.
Both kbase and panfrost always zero these registers. Per discussion with
Steven Price, there are two reasons these quirks may be used:
1. Simulating slower memory subsystems. This use case is only of
interest to system-on-chip designers; it is not relevant to mainline.
2. Working around broken memory subsystems. Hopefully we never see this
case in mainline. If we do, we'll need to set this register based on
an SoC-compatible, rather than generally matching on the GPU model.
To the best of our knowledge, these fields are zero at reset, so the
write is not necessary. Let's remove the write to aid porting to new
Mali GPUs, which have different layouts for the L2_MMU_CONFIG register.
Add the HW_FEATURE_CLEAN_ONLY_SAFE bit based on kbase. When I actually
tried to port the logic from kbase, trivial jobs raised Data Invalid
Faults, so this may depend on other coherency details. It's still useful
to have the bit to record the feature bit when adding new models.
TTRX_3485 requires the infamous "dummy job" workaround. I have this
workaround implemented in a local branch, but I have not yet hit a case
that requires it so I cannot test whether the implementation is correct.
In the mean time, add the quirk bit so we can document which platforms
may need it in the future.
From the kernel's perspective, (pre-CSF, "Job Manager") Valhall is more
or less compatible with Bifrost, although they differ to userspace. Add
a compatible for Valhall to the existing Bifrost bindings documentation.
As the first SoC with a Valhall GPU receiving mainline support, add a
specific compatible for the MediaTek MT8192, which instantiates a
Mali-G57.
v2: Change compatible to arm,mali-valhall-jm (Daniel Stone).
Jonathan Liu [Mon, 23 May 2022 13:01:44 +0000 (23:01 +1000)]
drm: bridge: icn6211: Adjust clock phase using SYS_CTRL_1
The code from [1] sets SYS_CTRL_1 to different values depending on the
desired clock phase (0, 1/4, 1/2 or 3/4). A clock phase of 0 aligns the
positive edge of the clock with the pixel data while other values delay
the clock by a fraction of the clock period. A clock phase of 1/2 aligns
the negative edge of the clock with the pixel data.
The driver currently hard codes SYS_CTRL_1 to 0x88 which corresponds to
aligning the positive edge of the clock with the pixel data. This won't
work correctly for panels that require aligning the negative edge of the
clock with the pixel data.
Adjust the clock phase to 0 if DRM_BUS_FLAG_PIXDATA_DRIVE_POSEDGE is
present in bus_flags, otherwise adjust the clock phase to 1/2 as
appropriate for DRM_BUS_FLAG_PIXDATA_DRIVE_NEGEDGE.
Rob Herring [Wed, 25 May 2022 20:56:26 +0000 (15:56 -0500)]
dt-bindings: display: ingenic,jz4780-hdmi: Drop undocumented 'ddc-i2c-bus'
While 'ddc-i2c-bus' is a common property, it should be in a connector
node rather than the HDMI bridge node as the I2C bus goes to a
connector and not the HDMI block. Drop it from the example.
Fabio Estevam [Wed, 25 May 2022 21:53:16 +0000 (18:53 -0300)]
drm: bridge: adv7511: Move CEC definitions to adv7511_cec.c
ADV7511_REG_CEC_RX_FRAME_HDR[] and ADV7511_REG_CEC_RX_FRAME_LEN[]
are only used inside adv7511_cec.c.
Move their definitions to this file to avoid the following build
warnings when CONFIG_DRM_I2C_ADV7511_CEC is not selected:
drivers/gpu/drm/bridge/adv7511/adv7511.h:229:17: warning: 'ADV7511_REG_CEC_RX_FRAME_HDR' defined but not used [-Wunused-const-variable=]
drivers/gpu/drm/bridge/adv7511/adv7511.h:235:17: warning: 'ADV7511_REG_CEC_RX_FRAME_LEN' defined but not used [-Wunused-const-variable=]
Reported-by: kernel test robot <lkp@intel.com> Fixes: c47378d86879 ("drm: bridge: adv7511: use non-legacy mode for CEC RX") Signed-off-by: Fabio Estevam <festevam@gmail.com> Reviewed-by: Alvin Šipraga <alsi@bang-olufsen.dk> Signed-off-by: Robert Foss <robert.foss@linaro.org> Link: https://patchwork.freedesktop.org/patch/msgid/20220525215316.1133057-1-festevam@gmail.com
Since commit 49dba410a730 ("drm/vkms: Bugfix racing hrtimer vblank
handle") the work is scheduled at vkms_vblank_simulate() and since
commit 45fdc1e0993c ("drm/vkms: flush crc workers earlier in commit
flow") the work is flushed at vkms_atomic_commit_tail(). Update function
commment to reflect that.
Gao Chao [Tue, 24 May 2022 02:45:51 +0000 (10:45 +0800)]
drm/panel: Fix build error when CONFIG_DRM_PANEL_SAMSUNG_ATNA33XC20=y && CONFIG_DRM_DISPLAY_HELPER=m
If CONFIG_DRM_PANEL_SAMSUNG_ATNA33XC20=y && CONFIG_DRM_DISPLAY_HELPER=m,
bulding fails:
drivers/gpu/drm/panel/panel-samsung-atna33xc20.o: In function `atana33xc20_probe':
panel-samsung-atna33xc20.c:(.text+0x744): undefined reference to
`drm_panel_dp_aux_backlight'
make: *** [vmlinux] Error 1
Let CONFIG_DRM_PANEL_SAMSUNG_ATNA33XC20 select DRM_DISPLAY_DP_HELPER and
CONFIG_DRM_DISPLAY_HELPER to fix this error.
Fixes: 76db0fa2d68f ("drm/panel: atna33xc20: Introduce the Samsung ATNA33XC20 panel") Reported-by: Hulk Robot <hulkci@huawei.com> Signed-off-by: Gao Chao <gaochao49@huawei.com> Reviewed-by: Douglas Anderson <dianders@chromium.org> Signed-off-by: Douglas Anderson <dianders@chromium.org> Link: https://patchwork.freedesktop.org/patch/msgid/20220524024551.539-1-gaochao49@huawei.com
Mark Menzynski [Mon, 23 May 2022 11:35:41 +0000 (13:35 +0200)]
drm/nouveau: clear output poll workers before nouveau_fbcon_destroy()
Resources needed for output poll workers are destroyed in
nouveau_fbcon_fini() before output poll workers are cleared in
nouveau_display_fini(). This means there is a time between fbcon_fini()
and display_fini(), where if output poll happens, it crashes.
This patch introduces another output poll clearing before fbcon
resources are destroyed.
BUG: KASAN: use-after-free in
__drm_fb_helper_initial_config_and_unlock.cold+0x1f3/0x291
[drm_kms_helper]
Cc: Ben Skeggs <bskeggs@redhat.com> Cc: Karol Herbst <kherbst@redhat.com> Cc: Lyude Paul <lyude@redhat.com> Cc: David Airlie <airlied@linux.ie> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: dri-devel@lists.freedesktop.org Cc: nouveau@lists.freedesktop.org Cc: linux-kernel@vger.kernel.org Signed-off-by: Mark Menzynski <mmenzyns@redhat.com> Reviewed-by: Lyude Paul <lyude@redhat.com> Signed-off-by: Lyude Paul <lyude@redhat.com> Link: https://patchwork.freedesktop.org/patch/msgid/20220523113541.10562-1-mmenzyns@redhat.com
This patch depends on the patches just aplied to the media tree, and will
not build without them, which leaves drm-misc-next in a broken state.
Let's revert the two latter patches until rc1 has been branched,
and rc1 has been backmerged into drm-misc-next.
This patch depends on the patches just aplied to the media tree, and will
not build without them, which leaves drm-misc-next in a broken state.
Let's revert the two latter patches until rc1 has been branched,
and rc1 has been backmerged into drm-misc-next.
drm/st7735r: Fix module autoloading for Okaya RH128128T
The SPI core always reports a "MODALIAS=spi:<foo>", even if the device was
registered via OF. This means that the st7735r.ko module won't autoload if
a DT has a node with a compatible "okaya,rh128128t" string.
In that case, kmod expects a "MODALIAS=of:N*T*Cokaya,rh128128t" uevent but
instead will get a "MODALIAS=spi:rh128128t", which is not present in the
list of aliases:
John Stultz [Wed, 11 May 2022 01:26:12 +0000 (01:26 +0000)]
drm/bridge: lt9611: Use both bits for HDMI sensing
In commit b38eafa7adae ("lontium-lt9611: check a different
register bit for HDMI sensing"), the bit flag used to detect
HDMI cable connect was switched from BIT(2) to BIT(0) to improve
compatibility with some monitors that didn't seem to set BIT(2).
However, with that change, I've seen occasional issues where the
detection failed, because BIT(2) was set, but not BIT(0).
Unfortunately, as I understand it, the bits and their function
was never clearly documented. So lets instead check both
(BIT(2) | BIT(0)) when checking the register.
Cc: Yongqin Liu <yongqin.liu@linaro.org> Cc: Amit Pundir <amit.pundir@linaro.org> Cc: Peter Collingbourne <pcc@google.com> Cc: Vinod Koul <vkoul@kernel.org> Cc: Bjorn Andersson <bjorn.andersson@linaro.org> Cc: Robert Foss <robert.foss@linaro.org> Cc: kernel-team@android.com Fixes: b38eafa7adae ("lontium-lt9611: check a different register bit for HDMI sensing") Signed-off-by: John Stultz <jstultz@google.com> Reviewed-by: Robert Foss <robert.foss@linaro.org> Signed-off-by: Robert Foss <robert.foss@linaro.org> Link: https://patchwork.freedesktop.org/patch/msgid/20220511012612.3297577-2-jstultz@google.com
Marek Vasut [Thu, 19 May 2022 11:23:37 +0000 (13:23 +0200)]
drm/bridge: anx7625: Add missing of_node_put for endpoint
Add of_node_put call on the endpoint node after it is not needed.
Signed-off-by: Marek Vasut <marex@denx.de> Cc: Alex Deucher <alexander.deucher@amd.com> Cc: Javier Martinez Canillas <javierm@redhat.com> Cc: Lyude Paul <lyude@redhat.com> Cc: Thomas Zimmermann <tzimmermann@suse.de> Cc: Xin Ji <xji@analogixsemi.com> Reviewed-by: Robert Foss <robert.foss@linaro.org> Signed-off-by: Robert Foss <robert.foss@linaro.org> Link: https://patchwork.freedesktop.org/patch/msgid/20220519112337.62198-1-marex@denx.de
Marek Vasut [Wed, 18 May 2022 23:38:44 +0000 (01:38 +0200)]
drm/bridge: ti-sn65dsi83: Handle dsi_lanes == 0 as invalid
Handle empty data-lanes = < >; property, which translates to
dsi_lanes = 0 as invalid.
Fixes: 454c13700cba3 ("drm/bridge: ti-sn65dsi83: Add TI SN65DSI83 and SN65DSI84 driver") Signed-off-by: Marek Vasut <marex@denx.de> Cc: Jonas Karlman <jonas@kwiboo.se> Cc: Laurent Pinchart <Laurent.pinchart@ideasonboard.com> Cc: Lucas Stach <l.stach@pengutronix.de> Cc: Marek Vasut <marex@denx.de> Cc: Maxime Ripard <maxime@cerno.tech> Cc: Neil Armstrong <narmstrong@baylibre.com> Cc: Robert Foss <robert.foss@linaro.org> Cc: Sam Ravnborg <sam@ravnborg.org> Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com> Reviewed-by: Lucas Stach <l.stach@pengutronix.de> Signed-off-by: Robert Foss <robert.foss@linaro.org> Link: https://patchwork.freedesktop.org/patch/msgid/20220518233844.248504-1-marex@denx.de
drm/i915/display/debug: Expose crtc current bpc via debugfs
This new debugfs will expose the currently using bpc by crtc.
It is very useful for verifying whether we enter the correct
output color depth from IGT.
This patch will also add the connector's max supported bpc to
"i915_display_info" debugfs.
drm/r128: Fix undefined behavior due to shift overflowing the constant
Fix:
drivers/gpu/drm/r128/r128_cce.c: In function ‘r128_do_init_cce’:
drivers/gpu/drm/r128/r128_cce.c:417:2: error: case label does not reduce to an integer constant
case R128_PM4_64BM_64VCBM_64INDBM:
^~~~
drivers/gpu/drm/r128/r128_cce.c:418:2: error: case label does not reduce to an integer constant
case R128_PM4_64PIO_64VCPIO_64INDPIO:
^~~~
See https://lore.kernel.org/r/YkwQ6%2BtIH8GQpuct@zn.tnic for the gory
details as to why it triggers with older gccs only.
Reviewed-by: Randy Dunlap <rdunlap@infradead.org> Signed-off-by: Borislav Petkov <bp@suse.de> Cc: David Airlie <airlied@linux.ie> Cc: Daniel Vetter <daniel@ffwll.ch> Cc: Alex Deucher <alexander.deucher@amd.com> Cc: dri-devel@lists.freedesktop.org Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch> Link: https://patchwork.freedesktop.org/patch/msgid/20220405151517.29753-5-bp@alien8.de
Marek Vasut [Fri, 20 May 2022 12:15:43 +0000 (14:15 +0200)]
drm/bridge: tc358767: Make sure Refclk clock are enabled
The Refclk may be supplied by SoC clock output instead of crystal
oscillator, make sure the clock are enabled before any other action
is performed with the bridge chip, otherwise it may either fail to
operate at all, or miss reset GPIO toggle.
Reviewed-by: Lucas Stach <l.stach@pengutronix.de> Fixes: 16bb6b52adee4 ("drm/bridge: tc358767: Add DPI to eDP bridge driver") Signed-off-by: Marek Vasut <marex@denx.de> Cc: Jonas Karlman <jonas@kwiboo.se> Cc: Laurent Pinchart <Laurent.pinchart@ideasonboard.com> Cc: Lucas Stach <l.stach@pengutronix.de> Cc: Marek Vasut <marex@denx.de> Cc: Maxime Ripard <maxime@cerno.tech> Cc: Neil Armstrong <narmstrong@baylibre.com> Cc: Robert Foss <robert.foss@linaro.org> Cc: Sam Ravnborg <sam@ravnborg.org> Reviewed-by: Maxime Ripard <maxime@cerno.tech> Link: https://patchwork.freedesktop.org/patch/msgid/20220520121543.11550-1-marex@denx.de
Douglas Anderson [Tue, 10 May 2022 19:29:41 +0000 (12:29 -0700)]
drm/dp: Export symbol / kerneldoc fixes for DP AUX bus
While working on the DP AUX bus code I found a few small things that
should be fixed. Namely the non-devm version of
of_dp_aux_populate_ep_devices() was missing an export. There was also
an extra blank line in a kerneldoc and a kerneldoc that incorrectly
documented a return value. Fix these.
drm: Document the power requirements for DP AUX transfers
When doing DP AUX transfers there are two actors that need to be
powered in order for the DP AUX transfer to work: the DP source and
the DP sink. Commit 26ff68b5ec8c ("drm: Mention the power state
requirement on side-channel operations") added some documentation
saying that the DP source is required to power itself up (if needed)
to do AUX transfers. However, that commit doesn't talk anything about
the DP sink.
For full fledged DP the sink isn't really a problem. It's expected
that if an external DP monitor isn't plugged in that attempting to do
AUX transfers won't work. It's also expected that if a DP monitor is
plugged in (and thus asserting HPD) then AUX transfers will work.
When we're looking at eDP, however, things are less obvious. Let's add
some documentation about expectations. Here's what we'll say:
1. We don't expect the DP AUX transfer function to power on an eDP
panel. If an eDP panel is physically connected but powered off then it
makes sense for the transfer to fail.
2. We'll document that the official way to power on a panel is via the
bridge chain, specifically by making sure that the panel's prepare
function has been called (which is called by
panel_bridge_pre_enable()). It's already specified in the kernel doc
of drm_panel_prepare() that this is the way to power the panel on and
also that after this call "it is possible to communicate with any
integrated circuitry via a command bus."
3. We'll also document that for code running in the panel driver
itself that it is legal for the panel driver to power itself up
however it wants (it doesn't need to officially call
drm_panel_pre_enable()) and then it can do AUX bus transfers. This is
currently the way that edp-panel works when it's running atop the DP
AUX bus.
NOTE: there was much discussion of all of this in response to v1 [1]
of this patch. A summary of that is:
* With the Intel i195 driver, apparently eDP panels do get powered
up. We won't forbid this but it is expected that code that wants to
run on a variety of platforms should ensure that the drm_panel's
prepare() function has been called.
* There is at least a reasonable amount of agreement that the
transfer() functions itself shouldn't be responsible for powering
the panel. It's proposed that if we need the DP AUX dev nodes to be
robust for eDP that the code handling the DP AUX dev nodes could
handle powering the panel by ensuring that the panel's prepare()
call was made. Potentially drm_dp_aux_dev_get_by_minor() could be a
good place to do this. This is left as a future exercise. Until
that's fixed the DP AUX dev nodes for eDP are probably best just
used for debugging.
* If a panel could be in PSR and DP AUX via the dev node needs to be
reliable then we need to be able to pull the panel out of PSR. On
i915 this is also apparently handled as part of the transfer()
function.
Douglas Anderson [Wed, 11 May 2022 22:58:08 +0000 (15:58 -0700)]
drm/probe-helper: For DP, add 640x480 if all other modes are bad
As per Displayport spec section 5.2.1.2 ("Video Timing Format") says
that all detachable sinks shall support 640x480 @60Hz as a fail safe
mode.
A DP compliance test expected us to utilize the above fact when all
modes it presented to the DP source were not achievable. It presented
only modes that would be achievable with more lanes and/or higher
speeds than we had available and expected that when we couldn't do
that then we'd fall back to 640x480 even though it didn't advertise
this size.
In order to pass the compliance test (and also support any users who
might fall into a similar situation with their display), we need to
add 640x480 into the list of modes. However, we don't want to add
640x480 all the time. Despite the fact that the DP spec says all sinks
_shall support_ 640x480, they're not guaranteed to support it
_well_. Continuing to read the spec you can see that the display is
not required to really treat 640x480 equal to all the other modes. It
doesn't need to scale or anything--just display the pixels somehow for
failsafe purposes. It should also be noted that it's not hard to find
a display hooked up via DisplayPort that _doesn't_ support 640x480 at
all. The HP ZR30w screen I'm sitting in front of has a native DP port
and doesn't work at 640x480. I also plugged in a tiny 800x480 HDMI
display via a DP to HDMI adapter and that screen definitely doesn't
support 640x480.
As a compromise solution, let's only add the 640x480 mode if:
* We're on DP.
* All other modes have been pruned.
This acknowledges that 640x480 might not be the best mode to use but,
since sinks are _supposed_ to support it, we will at least fall back
to it if there's nothing else.
Note that we _don't_ add higher resolution modes like 1024x768 in this
case. We only add those modes for a failed EDID read where we have no
idea what's going on. In the case where we've pruned all modes then
instead we only want 640x480 which is the only defined "Fail Safe"
resolution.
This patch originated in response to Kuogee Hsieh's patch [1].
Douglas Anderson [Wed, 11 May 2022 22:58:07 +0000 (15:58 -0700)]
drm/probe-helper: Add helper for drm_helper_probe_single_connector_modes()
The drm_helper_probe_single_connector_modes() is a bit long. Let's
break a chunk off to update and validate modes. This helps avoid one
goto and also will allow us to more easily call the helper a second
time in a future patch without adding looping or another goto.
This change is intended to be a no-op change--just code movement.
drm/gem-vram: Share code between GEM VRAM's _{prepare, cleanup}_fb()
The error-recovery code in drm_gem_vram_plane_helper_prepare_fb() is of
the same pattern as drm_gem_vram_plane_helper_cleanup_fb(). Implement
both of them using an internal helper. No functional changes.
drm/gem: Ignore color planes that are unused by framebuffer format
Only handle color planes that exist in a framebuffer's color format.
Ignore non-existing planes.
So far, several helpers assumed that all 4 planes are available and
silently ignored non-existing planes. This lead to subtil bugs with
uninitialized data in instances of struct iosys_map. [1]
drm/gem: Share code between drm_gem_fb_{begin,end}_cpu_access()
The error-recovery code in drm_gem_fb_begin() is of the same pattern
as drm_gem_fb_end(). Implement both of them using an internal helper.
No functional changes.
v2:
* print additional information in error message (Javier)
* fix commit description (Javier)
Dongjin Kim [Mon, 16 May 2022 07:22:45 +0000 (07:22 +0000)]
drm/meson: add YUV422 output support
Support YUV422 output from the Amlogic Meson SoC VPU to the HDMI
controller. Without this YUV422 format out of the HDMI encoder
leads to using the dw-hdmi YUV444 to YUV422 color conversion which
gives wrong colors and a green line on the left edge of the screen.
Signed-off-by: Dongjin Kim <tobetter@gmail.com> Signed-off-by: Christian Hewitt <christianshewitt@gmail.com> Tested-by: Furkan Kardame <f.kardame@manjaro.org> Acked-by: Neil Armstrong <narmstrong@baylibre.com> Tested-by: Dan Johansen <strit@manjaro.org> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> Link: https://patchwork.freedesktop.org/patch/msgid/20220516072245.10745-1-christianshewitt@gmail.com
Miaoqian Lin [Wed, 11 May 2022 05:40:51 +0000 (09:40 +0400)]
drm/meson: Fix refcount leak in meson_encoder_hdmi_init
of_find_device_by_node() takes reference, we should use put_device()
to release it when not need anymore.
Add missing put_device() in error path to avoid refcount
leak.
Fixes: bcf5446630b8 ("drm/meson: encoder_hdmi: switch to bridge DRM_BRIDGE_ATTACH_NO_CONNECTOR") Signed-off-by: Miaoqian Lin <linmq006@gmail.com> Reviewed-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com> Reviewed-by: Neil Armstrong <narmstrong@baylibre.com> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> Link: https://patchwork.freedesktop.org/patch/msgid/20220511054052.51981-1-linmq006@gmail.com
projects / kernel.git / log