Christian Marangi says:
====================
net: mdio-ipq4019: fix wrong default MDC rate
This was a long journey to arrive and discover this problem.
To not waste too much char, there is a race problem with PHY and driver
probe. This was observed with Aquantia PHY firmware loading.
With some hacks the race problem was workarounded but an interesting
thing was notice. It took more than a minute for the firmware to load
via MDIO.
This was strange as the same operation was done by UBoot in at max 5
second and the same data was loaded.
A similar problem was observed on a mtk board that also had an
Aquantia PHY where the load was very slow. It was notice that the cause
was the MDIO bus running at a very low speed and the firmware
was missing a property (present in mtk sdk) that set the right frequency
to the MDIO bus.
It was fun to find that THE VERY SAME PROBLEM is present on IPQ in a
different form. The MDIO apply internally a division to the feed clock
resulting in the bus running at 390KHz instead of 6.25Mhz.
Searching around the web for some documentation and some include and
analyzing the uboot codeflow resulted in the divider being set wrongly
at /256 instead of /16 as the value was actually never set.
Applying the value restore the original load time for the Aquantia PHY.
This series mainly handle this by adding support for the "clock-frequency"
property.
Changes v3:
- Add Reviewed-by tag
- Fix english grammar error in comment
- Drop DTS patch
Changes v2:
- Use DIV_ROUND_UP
- Introduce logic to chose a default value for 802.3 spec 2.5MHz
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
The IPQ4019 MDIO internally divide the clock feed by AHB based on the
MDIO_MODE reg. On reset or power up, the default value for the
divider is 0xff that reflect the divider set to /256.
This makes the MDC run at a very low rate, that is, considering AHB is
always fixed to 100Mhz, a value of 390KHz.
This hasn't have been a problem as MDIO wasn't used for time sensitive
operation, it is now that on IPQ807x is usually mounted with PHY that
requires MDIO to load their firmware (example Aquantia PHY).
To handle this problem and permit to set the correct designed MDC
frequency for the SoC add support for the standard "clock-frequency"
property for the MDIO node.
The divider supports value from /1 to /256 and the common value are to
set it to /16 to reflect 6.25Mhz or to /8 on newer platform to reflect
12.5Mhz.
To scan if the requested rate is supported by the divider, loop with
each supported divider and stop when the requested rate match the final
rate with the current divider. An error is returned if the rate doesn't
match any value.
On MDIO reset, the divider is restored to the requested value to prevent
any kind of downclocking caused by the divider reverting to a default
value.
To follow 802.3 spec of 2.5MHz of default value, if divider is set at
/256 and "clock-frequency" is not set in DT, assume nobody set the
divider and try to find the closest MDC rate to 2.5MHz. (in the case of
AHB set to 100MHz, it's 1.5625MHz)
While at is also document other bits of the MDIO_MODE reg to have a
clear idea of what is actually applied there.
Documentation of some BITs is skipped as they are marked as reserved and
their usage is not clear (RES 11:9 GENPHY 16:13 RES1 19:17)
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Document support for clock-frequency and add details on why this
property is needed and what values are supported.
From internal documentation, while other values are supported, the
correct function of the MDIO bus is not assured hence add only the
suggested supported values to the property enum.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Commit 56e58d6c8a ("net: stmmac: Implement Safety Features in
XGMAC core") checks and reports safety errors, but leaves the
Data Path Parity Errors for each channel in DMA unhandled at all, lead to
a storm of interrupt.
Fix it by checking and clearing the DMA_DPP_Interrupt_Status register.
Fixes: 56e58d6c8a ("net: stmmac: Implement Safety Features in XGMAC core")
Signed-off-by: Furong Xu <0x1207@gmail.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
CAP_NET_ADMIN allows to configure network interfaces, not CAP_SYS_ADMIN
which only allows to call unshare(2). Without this change, running
network tests as a non-root user but with all capabilities would fail at
the setup_loopback() step with "RTNETLINK answers: Operation not
permitted".
The issue is only visible when running tests with non-root users (i.e.
only relying on ambient capabilities). Indeed, when configuring the
network interface, the "ip" command is called, which may lead to the
special handling of capabilities for the root user by execve(2). If
root is the caller, then the inherited, permitted and effective
capabilities are all reset, which then includes CAP_NET_ADMIN. However,
if a non-root user is the caller, then ambient capabilities are masked
by the inherited ones, which were explicitly dropped.
To make execution deterministic whatever users are running the tests,
set the noroot secure bit for each test, and set the inheritable and
ambient capabilities to CAP_NET_ADMIN, the only capability that may be
required after an execve(2).
Factor out _effective_cap() into _change_cap(), and use it to manage
ambient capabilities with the new set_ambient_cap() and
clear_ambient_cap() helpers.
This makes it possible to run all Landlock tests with check-linux.sh
from https://github.com/landlock-lsm/landlock-test-tools
Cc: Konstantin Meskhidze <konstantin.meskhidze@huawei.com>
Fixes: a549d055a2 ("selftests/landlock: Add network tests")
Link: https://lore.kernel.org/r/20240125153230.3817165-2-mic@digikod.net
[mic: Make sure SECBIT_NOROOT_LOCKED is set]
Signed-off-by: Mickaël Salaün <mic@digikod.net>
Commit 8015bee0bf ("drm/i915/display: Add framework to add parameters
specific to display") added the file intel_display_debugfs_params.c,
which calls the functions "debugfs_create_{bool, ulong, str}" -- all of
which are defined in <linux/debugfs.h>. The missing inclusion of this
header file is breaking the ChromeOS build -- add an explicit include
to fix that.
Signed-off-by: Paz Zcharya <pazz@chromium.org>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240131204658.795278-1-pazz@chromium.org
The commit 2ad56efa80 ("powerpc/iommu: Setup a default domain and
remove set_platform_dma_ops") refactored the code removing the
set_platform_dma_ops(). It missed out the table group
release_ownership() call which would have got called otherwise
during the guest shutdown via vfio_group_detach_container(). On
PPC64, this particular call actually sets up the 32-bit TCE table,
and enables the 64-bit DMA bypass etc. Now after guest shutdown,
the subsequent host driver (e.g megaraid-sas) probe post unbind
from vfio-pci fails like,
megaraid_sas 0031:01:00.0: Warning: IOMMU dma not supported: mask 0x7fffffffffffffff, table unavailable
megaraid_sas 0031:01:00.0: Warning: IOMMU dma not supported: mask 0xffffffff, table unavailable
megaraid_sas 0031:01:00.0: Failed to set DMA mask
megaraid_sas 0031:01:00.0: Failed from megasas_init_fw 6539
The patch brings back the call to table_group release_ownership()
call when switching back to PLATFORM domain from BLOCKED, while
also separates the domain_ops for both.
Fixes: 2ad56efa80 ("powerpc/iommu: Setup a default domain and remove set_platform_dma_ops")
Signed-off-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Reviewed-by: Jason Gunthorpe <jgg@nvidia.com>
Link: https://lore.kernel.org/r/170628173462.3742.18330000394415935845.stgit@ltcd48-lp2.aus.stglab.ibm.com
Signed-off-by: Joerg Roedel <jroedel@suse.de>
Timur pointed this out before, and it just slipped my mind,
but this might help some things work better, around pcie power
management.
Fixes: 8d55b0a940 ("nouveau/gsp: add some basic registry entries.")
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/576336/
Add some (hopefully useful) hints to tips.txt
Also some minor corrections.
Would probably good to make it a reviewer rule that if generally useful
options are added the patch must add an example to tips.txt
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/r/20240131021352.151440-1-ak@linux.intel.com
The original test report was too complicated to read with information
that not really useful. This new update simplify the report which should
largely improve the readibility.
Signed-off-by: Weilin Wang <weilin.wang@intel.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Cc: Caleb Biggers <caleb.biggers@intel.com>
Cc: Perry Taylor <perry.taylor@intel.com>
Cc: Samantha Alt <samantha.alt@intel.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/r/20240130180907.639729-1-weilin.wang@intel.com
data->id has been initialized at line 2362, remove duplicate initialization.
Fixes: 3ad31d8a0d ("perf evsel: Centralize perf_sample initialization")
Signed-off-by: Yang Jihong <yangjihong1@huawei.com>
Reviewed-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Reviewed-by: Ian Rogers <irogers@google.com>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/r/20240127025756.4041808-1-yangjihong1@huawei.com
Since get_states() assumes the existence of libtraceevent, so move
to where it should belong, i.e, util/trace-event-parse.c, and also
rename it to parse_task_states().
Leave evsel_getstate() untouched as it fits well in the evsel
category.
Also make some necessary tweaks for python support, and get it
verified with: perf test python.
Signed-off-by: Ze Gao <zegao@tencent.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Link: https://lore.kernel.org/r/20240123070210.1669843-2-zegao@tencent.com
Since ext4_map_blocks() can recognize a delayed allocated only extent,
make ext4_set_iomap() can also recognize it, and remove the useless
separate check in ext4_iomap_begin_report().
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20240127015825.1608160-7-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Add a new map flag EXT4_MAP_DELAYED to indicate the mapping range is a
delayed allocated only (not unwritten) one, and making
ext4_map_blocks() can distinguish it, no longer mixing it with holes.
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20240127015825.1608160-6-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
In order to cache hole extents in the extent status tree and keep the
hole length as long as possible, re-add a hole entry to the cache just
after punching a hole.
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20240127015825.1608160-5-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Alex Elder says:
====================
net: ipa: simplify TX power handling
In order to deliver a packet to the IPA hardware, we must ensure
it is powered. We request power by calling pm_runtime_get(), and
its return value tells us the power state. We can't block in
ipa_start_xmit(), so if power isn't enabled we prevent further
transmit attempts by calling netif_stop_queue(). Power will
eventually become enabled, at which point we call netif_wake_queue()
to allow the transmit to be retried. When it does, the power should
be enabled, so the packet delivery can proceed.
The logic that handles this is convoluted, and was put in place
to address a race condition pointed out by Jakub Kicinski during
review. The fix addressed that race, as well as another one that
was found while investigating it:
b8e36e13ea ("net: ipa: fix TX queue race")
I have wanted to simplify this code ever since, and I'm pleased to
report that this series implements a much better solution that
avoids both race conditions.
The first race occurs between the ->ndo_start_xmit thread and the
runtime resume thread. If we find power is not enabled when
requested in ipa_start_xmit(), we stop queueing. But there's a
chance the runtime resume will enable queuing just before that,
leaving queueing stopped forever. A flag is used to ensure that
does not occur.
A second flag is used to prevent NETDEV_TX_BUSY from being returned
repeatedly during the small window between enabling queueing and
finishing power resume. This can happen if resume was underway when
pm_runtime_get() was called and completes immediately afterward.
This condition only exists because of the use of the first flag.
The fix is to disable transmit for *every* call to ipa_start_xmit(),
disabling it *before* calling pm_runtime_get(). This leaves three
cases:
- If the return value indicates power is not active (or is in
transition), queueing remains disabled--thereby avoiding
the race between disabling it and a concurrent power thread
enabling it.
- If pm_runtime_get() returns an error, we drop the packet and
re-enable queueing.
- Finally, if the hardware is powered, we re-enable queueing
before delivering the packet to the hardware.
So the first race is avoided. And as a result, the second condition
will not occur.
The first patch adds pointers to the TX and RX IPA endpoints in the
netdev private data. The second patch has netif_stop_queue() be
called for all packets; if pm_runtime_get() indicates power is
enabled (or an error), netif_wake_queue() is called to enable it
again. The third and fourth patches get rid of the STARTED and
STOPPED IPA power flags, as well as the power spinlock, because they
are no longer needed. The last three patches just eliminate some
trivial functions, open-coding them instead.
====================
Link: https://lore.kernel.org/r/20240130192305.250915-1-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
All ipa_power_modem_queue_wake() does is call netif_wake_queue()
on the modem netdev. There is no need to wrap that call in a
trivial function (and certainly not one defined in "ipa_power.c").
So get rid of ipa_power_modem_queue_wake(), and replace its one
caller with a direct call to netif_wake_queue(). Determine the
netdev pointer to use from the private TX endpoint's netdev pointer.
Signed-off-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20240130192305.250915-8-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
All ipa_power_modem_queue_active() does now is call netif_wake_queue().
Just call netif_wake_queue() in the two places it's needed, and get
rid of ipa_power_modem_queue_active().
Signed-off-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20240130192305.250915-7-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
All ipa_power_modem_queue_stop() does now is call netif_stop_queue().
Just call netif_stop_queue() in the one place it's needed, and get
rid of ipa_power_modem_queue_stop().
Signed-off-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20240130192305.250915-6-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Currently the STOPPED IPA power flag is used to indicate that the
transmit queue has been stopped. Previously this was used to avoid
setting the STARTED flag unless the queue had already been stopped.
It meant transmit queuing would be enabled on resume if it was
stopped by the transmit path--and if so, it ensured it only got
enabled once.
We only stop the transmit queue in the transmit path. The STARTED
flag has been removed, and it causes no real harm to enable
transmits when they're already enabled. So we can get rid of
the STOPPED flag and call netif_wake_queue() unconditionally.
This makes the IPA power spinlock unnecessary, so it can be removed
as well.
Signed-off-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20240130192305.250915-5-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
A transmit on the modem netdev can only complete if the IPA hardware
is powered. Currently, if a transmit request arrives when the
hardware was not powered, further transmits are be stopped to allow
power-up to complete. Once power-up completes, transmits are once
again enabled.
Runtime resume can complete at the same time a transmit request is
being handled, and previously there was a race between stopping and
restarting transmits. The STARTED flag was used to ensure the
stop request in the transmit path was skipped if the start request
in the runtime resume path had already occurred.
Now, the queue is *always* stopped in the transmit path, *before*
determining whether power is ACTIVE. If power is found to already
be active (or if the socket buffer is gets dropped), transmit is
re-enabled. Otherwise it will (always) be enabled after runtime
resume completes.
The race between transmit and runtime resume no longer exists, so
there is no longer any need to maintain the STARTED flag.
Signed-off-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20240130192305.250915-4-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
There are a number of flags used in the IPA driver to attempt to
manage race conditions that can occur between runtime resume and
netdev transmit. If we disable TX before requesting power, we can
avoid these races entirely, simplifying things considerably.
This patch implements the main change, disabling transmit always in
the net_device->ndo_start_xmit() callback, then re-enabling it again
whenever we find power is active (or when we drop the skb).
The patches that follow will refactor the "old" code to the point
that most of it can be eliminated.
Signed-off-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20240130192305.250915-3-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Rather than repeatedly looking up the endpoints in the name map,
save the modem TX and RX endpoint pointers in the netdev private
area.
Signed-off-by: Alex Elder <elder@linaro.org>
Link: https://lore.kernel.org/r/20240130192305.250915-2-elder@linaro.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
After commit 936e4d49ec ("Input: atkbd - skip ATKBD_CMD_GETID in
translated mode") not only the getid command is skipped, but also
the de-activating of the keyboard at the end of atkbd_probe(), potentially
re-introducing the problem fixed by commit be2d7e4233 ("Input: atkbd -
fix multi-byte scancode handling on reconnect").
Make sure multi-byte scancode handling on reconnect is still handled
correctly by not skipping the atkbd_deactivate() call.
Fixes: 936e4d49ec ("Input: atkbd - skip ATKBD_CMD_GETID in translated mode")
Tested-by: Paul Menzel <pmenzel@molgen.mpg.de>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://lore.kernel.org/r/20240126160724.13278-3-hdegoede@redhat.com
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
In ext4_map_blocks(), if we can't find a range of mapping in the
extents cache, we are calling ext4_ext_map_blocks() to search the real
path and ext4_ext_determine_hole() to determine the hole range. But if
the querying range was partially or completely overlaped by a delalloc
extent, we can't find it in the real extent path, so the returned hole
length could be incorrect.
Fortunately, ext4_ext_put_gap_in_cache() have already handle delalloc
extent, but it searches start from the expanded hole_start, doesn't
start from the querying range, so the delalloc extent found could not be
the one that overlaped the querying range, plus, it also didn't adjust
the hole length. Let's just remove ext4_ext_put_gap_in_cache(), handle
delalloc and insert adjusted hole extent in ext4_ext_determine_hole().
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Suggested-by: Jan Kara <jack@suse.cz>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20240127015825.1608160-4-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
ext4_da_map_blocks() only hold i_data_sem in shared mode and i_rwsem
when inserting delalloc extents, it could be raced by another querying
path of ext4_map_blocks() without i_rwsem, .e.g buffered read path.
Suppose we buffered read a file containing just a hole, and without any
cached extents tree, then it is raced by another delayed buffered write
to the same area or the near area belongs to the same hole, and the new
delalloc extent could be overwritten to a hole extent.
pread() pwrite()
filemap_read_folio()
ext4_mpage_readpages()
ext4_map_blocks()
down_read(i_data_sem)
ext4_ext_determine_hole()
//find hole
ext4_ext_put_gap_in_cache()
ext4_es_find_extent_range()
//no delalloc extent
ext4_da_map_blocks()
down_read(i_data_sem)
ext4_insert_delayed_block()
//insert delalloc extent
ext4_es_insert_extent()
//overwrite delalloc extent to hole
This race could lead to inconsistent delalloc extents tree and
incorrect reserved space counter. Fix this by converting to hold
i_data_sem in exclusive mode when adding a new delalloc extent in
ext4_da_map_blocks().
Cc: stable@vger.kernel.org
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Suggested-by: Jan Kara <jack@suse.cz>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20240127015825.1608160-3-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Refactor and cleanup ext4_da_map_blocks(), reduce some unnecessary
parameters and branches, no logic changes.
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20240127015825.1608160-2-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
- Only allow a single user-fence per exec / bind.
The reason for this clarification fix is a limitation in the implementation
which can be lifted moving forward, if needed.
Driver Changes:
- A crash fix
- A fix for an assert due to missing mem_acces ref
- Only allow a single user-fence per exec / bind.
- Some sparse warning fixes
- Two fixes for compilation failures on various odd
combinations of gcc / arch pointed out on LKML.
- Fix a fragile partial allocation pointed out on LKML.
Cross-driver Change:
- A sysfs ABI documentation warning fix
This also touches i915 and is acked by i915 maintainers.
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQRskUM7w1oG5rx2IZO4FpNVCsYGvwUCZbt/9gAKCRC4FpNVCsYG
v5ioAP4wD+RIZ9iJ9DOG/UoQEieOje0JFjjdrwtPlUmt/kpBJQD/fvsArL0y/o3Z
cWahxhlpkqunNSG8r7lhcLMlnrK16gw=
=4Dmq
-----END PGP SIGNATURE-----
Merge tag 'drm-xe-fixes-2024-02-01' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-fixes
UAPI Changes:
- Only allow a single user-fence per exec / bind.
The reason for this clarification fix is a limitation in the implementation
which can be lifted moving forward, if needed.
Driver Changes:
- A crash fix
- A fix for an assert due to missing mem_acces ref
- Only allow a single user-fence per exec / bind.
- Some sparse warning fixes
- Two fixes for compilation failures on various odd
combinations of gcc / arch pointed out on LKML.
- Fix a fragile partial allocation pointed out on LKML.
Cross-driver Change:
- A sysfs ABI documentation warning fix
This also touches i915 and is acked by i915 maintainers.
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Thomas Hellstrom <thomas.hellstrom@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/ZbuCYdMDVK-kAWC5@fedora
virtio to set the segment size of the virtio_gpu device.
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQRcEzekXsqa64kGDp7j7w1vZxhRxQUCZbtpxwAKCRDj7w1vZxhR
xd2kAP4pu7ebCuq39M65T2HZKtzj5CXC0ZNXUn66j0eNUincuQEAkXL6OyMvWC/I
0WjTnRfVrMdiIykcFJPb8ZqtvfSLNwM=
=9pOf
-----END PGP SIGNATURE-----
Merge tag 'drm-misc-fixes-2024-02-01' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes
A quiet week: one fix for CMA dma-buf pages accounting, and one to
virtio to set the segment size of the virtio_gpu device.
Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <mripard@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/obnofpccz73c3uiqfyipxmjta5fgm4cle55dmtnissgtgxfgv7@22o7kb62efri
This reverts commit abe2023b4c.
Changing the locking order means that scheduler/msm_job_run() can race
with the recovery kthread worker, with the result that the GPU gets an
extra runpm get when we are trying to power it off. Leaving the GPU in
an unrecovered state.
I'll need to come up with a different scheme for appeasing lockdep.
Signed-off-by: Rob Clark <robdclark@chromium.org>
Patchwork: https://patchwork.freedesktop.org/patch/573835/
The spin lock is not necessary here as it can be replaced with
memory barrier which should be better performance-wise.
When Merkle tree block size differs from page size, in
is_hash_block_verified() two things are modified during check - a
bitmap and PG_checked flag of the page.
Each bit in the bitmap represent verification status of the Merkle
tree blocks. PG_checked flag tells if page was just re-instantiated
or was in pagecache. Both of this states are shared between
verification threads. Page which was re-instantiated can not have
already verified blocks (bit set in bitmap).
The spin lock was used to allow only one thread to modify both of
these states and keep order of operations. The only requirement here
is that PG_Checked is set strictly after bitmap is updated.
This way other threads which see that PG_Checked=1 (page cached)
knows that bitmap is up-to-date. Otherwise, if PG_Checked is set
before bitmap is cleared, other threads can see bit=1 and therefore
will not perform verification of that Merkle tree block.
However, there's still the case when one thread is setting a bit in
verify_data_block() and other thread is clearing it in
is_hash_block_verified(). This can happen if two threads get to
!PageChecked branch and one of the threads is rescheduled before
resetting the bitmap. This is fine as at worst blocks are
re-verified in each thread.
Signed-off-by: Andrey Albershteyn <aalbersh@redhat.com>
[ebiggers: improved the comment and removed the 'verified' variable]
Link: https://lore.kernel.org/r/20240201052813.68380-1-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
During the testing of Gnome on Qualcomm Robotics platform screen
corruption has been observed. Lowering GPU's highest_bank_bit from 14 to
13 seems to fix the screen corruption.
Note, the MDSS and DPU drivers use HBB=1 (which maps to the
highest_bank_bit = 14). So this change merely works around the UBWC
swizzling issue on this platform until the real cause is found.
Fixes: e7fc9398e6 ("drm/msm/a6xx: Add A610 support")
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Patchwork: https://patchwork.freedesktop.org/patch/573838/
Signed-off-by: Rob Clark <robdclark@chromium.org>
Since commit 79e2cf2e7a ("drm/gem: Take reservation lock for vmap/vunmap
operations"), the resv lock is already held in the prime vmap path, so
don't try to grab it again.
v2: This applies to vunmap path as well
v3: Fix fixes commit
Fixes: 79e2cf2e7a ("drm/gem: Take reservation lock for vmap/vunmap operations")
Signed-off-by: Rob Clark <robdclark@chromium.org>
Acked-by: Christian König <christian.koenig@amd.com>
Patchwork: https://patchwork.freedesktop.org/patch/576642/
It turns out it was never just gcc-11 that was broken. Apparently it
just happens to work on x86-64 with other gcc versions.
On arm64, I see warnings with gcc version 13.2.1, and the kernel test
robot reports the same problem on s390 with gcc 13.2.0.
Admittedly it seems to be just the new Xe drm driver, but this is
keeping me from doing my normal arm64 build testing. So it gets
reverted until somebody figures out what causes the problem (and why it
doesn't show on x86-64, which is what makes me suspect it was never just
about gcc-11, and more about just random happenstance).
This also changes the Kconfig naming a bit - just make the "disable this
for GCC" conditional be one simple Kconfig entry, and we can put the gcc
version dependencies in that entry once we figure out what the correct
rules are.
The version dependency _may_ still end up being "gcc version larger than
11" if the issue is purely in the Xe driver, but even if that ends up
the case, let's make that all part of the "GCC_NO_STRINGOP_OVERFLOW"
logic.
For now, we just disable it for all gcc versions while the exact cause
is unknown.
Link: https://lore.kernel.org/all/202401161031.hjGJHMiJ-lkp@intel.com/T/
Cc: Gustavo A. R. Silva <gustavoars@kernel.org>
Cc: Kees Cook <keescook@chromium.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use the device lifecycle managed register function. This helps prevent
mistakes like unregistering out of order in cleanup functions and
forgetting to unregister on error paths.
Signed-off-by: Andrew Davis <afd@ti.com>
Link: https://lore.kernel.org/r/20240129190246.73067-5-afd@ti.com
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>
Use the device lifecycle managed register function. This helps prevent
mistakes like unregistering out of order in cleanup functions and
forgetting to unregister on error paths.
Signed-off-by: Andrew Davis <afd@ti.com>
Link: https://lore.kernel.org/r/20240129190246.73067-4-afd@ti.com
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>
Use the device lifecycle managed register function. This helps prevent
mistakes like unregistering out of order in cleanup functions and
forgetting to unregister on error paths.
Signed-off-by: Andrew Davis <afd@ti.com>
Link: https://lore.kernel.org/r/20240129190246.73067-3-afd@ti.com
Signed-off-by: Sebastian Reichel <sebastian.reichel@collabora.com>