Incorrectly handled transaction restarts can be a source of heisenbugs;
add a mode where we randomly inject them to shake them out.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
UVD and VCN were split into separate dpm helpers in commit
ff69bba05f ("drm/amd/pm: add inst to dpm_set_powergating_by_smu")
as such, there is no need to include UVD in the is_vcn variable since
UVD and VCN are handled by separate dpm helpers now. Fix the check.
Fixes: ff69bba05f ("drm/amd/pm: add inst to dpm_set_powergating_by_smu")
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/3959
Link: https://lists.freedesktop.org/archives/amd-gfx/2025-February/119827.html
Reviewed-by: Lijo Lazar <lijo.lazar@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: Boyuan Zhang <boyuan.zhang@amd.com>
The way to assign NPE (network processing engines) shifted
during device tree design and an erroneous entry was left
behind in a disabled node. Fix it up.
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Link: https://lore.kernel.org/20250210-ixp4xx-dts-v1-1-6b752d745e04@linaro.org
This is the idiomatic way that opcodes should setup their async data,
so that it's always valid inside ->issue() without issue needing to
do that.
Fixes: f31ecf671d ("io_uring: add IORING_OP_WAITID support")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Since this is deep in the architecture, and the code is
called nested into other deep management code, this really
needs to be a raw spinlock. Convert it.
Link: https://patch.msgid.link/20250110125550.32479-8-johannes@sipsolutions.net
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
This is needed because at least in time-travel the code
can be called directly from the deep architecture and
IRQ handling code.
Link: https://patch.msgid.link/20250110125550.32479-7-johannes@sipsolutions.net
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
This code can be called deep in the IRQ handling, for
example, and then cannot normally use kmalloc(). Have
its own pre-allocated memory and use from there instead
so this doesn't occur. Only in the (very rare) case of
memcpy_toio() we'd still need to allocate memory.
Link: https://patch.msgid.link/20250110125550.32479-6-johannes@sipsolutions.net
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
The stub execution uses the somewhat new close_range and execveat
syscalls. Of these two, the execveat call is essential, but the
close_range call is more about stub process hygiene rather than safety
(and its result is ignored).
Replace both calls with a raw syscall as older machines might not have a
recent enough kernel for close_range (with CLOSE_RANGE_CLOEXEC) or a
libc that does not yet expose both of the syscalls.
Fixes: 32e8eaf263 ("um: use execveat to create userspace MMs")
Reported-by: Glenn Washburn <development@efficientek.com>
Closes: https://lore.kernel.org/20250108022404.05e0de1e@crass-HP-ZBook-15-G2
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Link: https://patch.msgid.link/20250113094107.674738-1-benjamin@sipsolutions.net
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
The stack needs to be properly aligned so 16 byte memory accesses on the
stack are correct. This was broken when introducing the dynamic math
register sizing as the rounding was not moved appropriately.
Fixes: 3f17fed214 ("um: switch to regset API and depend on XSTATE")
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Link: https://patch.msgid.link/20250107133509.265576-1-benjamin@sipsolutions.net
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
The init_task instance of struct task_struct is statically allocated and
does not contain the dynamic area for the userspace FP registers. As
such, limit the copy to the valid area of init_task and fill the rest
with zero.
Note that the FP state is only needed for userspace, and as such it is
entirely reasonable for init_task to not contain it.
Reported-by: Brian Norris <briannorris@chromium.org>
Closes: https://lore.kernel.org/Z1ySXmjZm-xOqk90@google.com
Fixes: 3f17fed214 ("um: switch to regset API and depend on XSTATE")
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Link: https://patch.msgid.link/20241217202745.1402932-3-benjamin@sipsolutions.net
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
It was reported that qemu may not enable the XSTATE CPU extension, which
is a requirement after commit 3f17fed214 ("um: switch to regset API
and depend on XSTATE"). Add a fallback to use FXSAVE (FP registers on
x86_64 and XFP on i386) which is just a shorter version of the same
data. The only difference is that the XSTATE magic should not be set in
the signal frame.
Note that this still drops support for the older i386 FP register layout
as supporting this would require more backward compatibility to build a
correct signal frame.
Fixes: 3f17fed214 ("um: switch to regset API and depend on XSTATE")
Reported-by: SeongJae Park <sj@kernel.org>
Closes: https://lore.kernel.org/r/20241203070218.240797-1-sj@kernel.org
Tested-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Benjamin Berg <benjamin.berg@intel.com>
Link: https://patch.msgid.link/20241204074827.1582917-1-benjamin@sipsolutions.net
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Richard Weinberger <richard@nod.at>
Any uring_cmd always has async data allocated now, there's no reason to
check and clear a cached copy of the SQE.
Fixes: d10f19dff5 ("io_uring/uring_cmd: switch to always allocating async data")
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Starting with Rust 1.86.0 (to be released 2025-04-03), Clippy will have
a new lint, `doc_overindented_list_items` [1], which catches cases of
overindented list items.
The lint has been added by Yutaro Ohno, based on feedback from the kernel
[2] on a patch that fixed a similar case -- commit 0c5928dead ("rust:
block: fix formatting in GenDisk doc").
Clippy reports a few cases in the kernel, apart from the one already
fixed in the commit above. One is this one:
error: doc list item overindented
--> rust/kernel/rbtree.rs:1152:5
|
1152 | /// null, it is a pointer to the root of the [`RBTree`].
| ^^^^ help: try using ` ` (2 spaces)
|
= help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#doc_overindented_list_items
= note: `-D clippy::doc-overindented-list-items` implied by `-D warnings`
= help: to override `-D warnings` add `#[allow(clippy::doc_overindented_list_items)]`
Thus clean it up.
Cc: Yutaro Ohno <yutaro.ono.418@gmail.com>
Cc: stable@vger.kernel.org # Needed in 6.12.y and 6.13.y only (Rust is pinned in older LTSs).
Fixes: a335e95914 ("rust: rbtree: add `RBTree::entry`")
Link: https://github.com/rust-lang/rust-clippy/pull/13711 [1]
Link: https://github.com/rust-lang/rust-clippy/issues/13601 [2]
Reviewed-by: Alice Ryhl <aliceryhl@google.com>
Reviewed-by: Yutaro Ohno <yutaro.ono.418@gmail.com>
Link: https://lore.kernel.org/r/20250206232022.599998-1-ojeda@kernel.org
[ There are a few other cases, so updated message. - Miguel ]
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Starting with Rust 1.85.0 (currently in beta, to be released 2025-02-20),
under some kernel configurations with `CONFIG_RUST_DEBUG_ASSERTIONS=y`,
one may trigger a new `objtool` warning:
rust/kernel.o: warning: objtool: _R...securityNtB2_11SecurityCtx8as_bytes()
falls through to next function _R...core3ops4drop4Drop4drop()
due to a call to the `noreturn` symbol:
core::panicking::assert_failed::<usize, usize>
Thus add it to the list so that `objtool` knows it is actually `noreturn`.
Do so matching with `strstr` since it is a generic.
See commit 56d680dd23 ("objtool/rust: list `noreturn` Rust functions")
for more details.
Cc: stable@vger.kernel.org # Needed in 6.12.y and 6.13.y only (Rust is pinned in older LTSs).
Fixes: 56d680dd23 ("objtool/rust: list `noreturn` Rust functions")
Reviewed-by: Gary Guo <gary@garyguo.net>
Link: https://lore.kernel.org/r/20250112143951.751139-1-ojeda@kernel.org
[ Updated Cc: stable@ to include 6.13.y. - Miguel ]
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
TDA998x is the HDMI bridge driver, incorporating drm_connector and
optional drm_encoder (created via the component bind API by the TICLDC
and HDLCD drivers). Thus it should be residing together with the other
DRM bridge drivers under drivers/gpu/drm/bridge/.
Acked-by: Neil Armstrong <neil.armstrong@linaro.org>
Acked-by: Liviu Dudau <liviu.dudau@arm.com>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250113-drm-move-tda998x-v3-3-214e0682a5e4@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Move the driver for NXP TDA9950 / CEC part of TDA998x together to
drivers/media/i2c, close to other CEC drivers. Specify 'default
DRM_I2C_NXP_TDA998X' in order to simplify migration from old config
files as the Kconfig name has been changed to follow media/cec style.
Acked-by: Hans Verkuil <hverkuil@xs4all.nl>
Link: https://patchwork.freedesktop.org/patch/msgid/20250113-drm-move-tda998x-v3-2-214e0682a5e4@linaro.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
As reported in the below link, it seems older versions of gcc cannot
determine that the howmany variable is known for all callers. Include
a test so that newer compilers can enforce this sanity check and older
compilers can still work. Add __always_inline attribute to give the
compiler an even better chance to know the inputs.
Link: https://lore.kernel.org/r/20250212185337.293023-1-alex.williamson@redhat.com
Fixes: 4453f36086 ("PCI: Batch BAR sizing operations")
Reported-by: Oleg Nesterov <oleg@redhat.com>
Link: https://lore.kernel.org/all/20250209154512.GA18688@redhat.com
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Oleg Nesterov <oleg@redhat.com>
Tested-by: Mitchell Augustin <mitchell.augustin@canonical.com>
Extract a private header and convert the prime_numbers self-test to a
KUnit test. I considered parameterizing the test using
`KUNIT_CASE_PARAM` but didn't see how it was possible since the test
logic is entangled with the test parameter generation logic.
Signed-off-by: Tamir Duberstein <tamird@gmail.com>
Link: https://lore.kernel.org/r/20250208-prime_numbers-kunit-convert-v5-2-b0cb82ae7c7d@gmail.com
Signed-off-by: Kees Cook <kees@kernel.org>
Add a KUnit test suite for the gcd() function.
This test suite verifies the correctness of gcd() across various
scenarios, including edge cases.
Signed-off-by: Yu-Chun Lin <eleanor15x@gmail.com>
Reviewed-by: Kuan-Wei Chiu <visitorckw@gmail.com>
Link: https://lore.kernel.org/r/20250203075400.3431330-1-eleanor15x@gmail.com
Signed-off-by: Kees Cook <kees@kernel.org>
I want to be sure that ipv6-specific code
is not compiled in kernel binaries
if ipv6 is not configured.
[1] was getting rid of "unused variable" warning, but,
with that, it also mandated compilation of a handful ipv6-
specific functions in ipv4-only kernel configurations:
smk_ipv6_localhost, smack_ipv6host_label, smk_ipv6_check.
Their compiled bodies are likely to be removed by compiler
from the resulting binary, but, to be on the safe side,
I remove them from the compiler view.
[1]
Fixes: 00720f0e7f ("smack: avoid unused 'sip' variable warning")
Signed-off-by: Konstantin Andreev <andreev@swemel.ru>
Signed-off-by: Casey Schaufler <casey@schaufler-ca.com>
5eff57fa9f ("io_uring/uring_cmd: defer SQE copying until it's needed")
moved the unconditional memcpy() of the uring_cmd SQE to async_data
to 2 cases when the request goes async:
- If REQ_F_FORCE_ASYNC is set to force the initial issue to go async
- If ->uring_cmd() returns -EAGAIN in the initial non-blocking issue
Unlike the REQ_F_FORCE_ASYNC case, in the EAGAIN case, io_uring_cmd()
copies the SQE to async_data but neglects to update the io_uring_cmd's
sqe field to point to async_data. As a result, sqe still points to the
slot in the userspace-mapped SQ. At the end of io_submit_sqes(), the
kernel advances the SQ head index, allowing userspace to reuse the slot
for a new SQE. If userspace reuses the slot before the io_uring worker
reissues the original SQE, the io_uring_cmd's SQE will be corrupted.
Introduce a helper io_uring_cmd_cache_sqes() to copy the original SQE to
the io_uring_cmd's async_data and point sqe there. Use it for both the
REQ_F_FORCE_ASYNC and EAGAIN cases. This ensures the uring_cmd doesn't
read from the SQ slot after it has been returned to userspace.
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Fixes: 5eff57fa9f ("io_uring/uring_cmd: defer SQE copying until it's needed")
Link: https://lore.kernel.org/r/20250212204546.3751645-3-csander@purestorage.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
eaf72f7b41 ("io_uring/uring_cmd: cleanup struct io_uring_cmd_data
layout") removed most of the places assuming struct io_uring_cmd_data
has sqes as its first field. However, the EAGAIN case in io_uring_cmd()
still compares ioucmd->sqe to the struct io_uring_cmd_data pointer using
a void * cast. Since fa3595523d ("io_uring: get rid of alloc cache
init_once handling"), sqes is no longer io_uring_cmd_data's first field.
As a result, the pointers will always compare unequal and memcpy() may
be called with the same source and destination.
Replace the incorrect void * cast with the address of the sqes field.
Signed-off-by: Caleb Sander Mateos <csander@purestorage.com>
Fixes: eaf72f7b41 ("io_uring/uring_cmd: cleanup struct io_uring_cmd_data layout")
Link: https://lore.kernel.org/r/20250212204546.3751645-2-csander@purestorage.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The 2k1000 reference boards do not have a Rohm DAC on them as far as I
can tell, and they certainly do not have a dh2228fv, as this device does
not actually exist! Remove the dac nodes from the devicetrees as it is
not acceptable to pretend to have a device on a board in order to bind
the spidev driver in Linux.
Signed-off-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Link: https://lore.kernel.org/r/20240717-preacher-sandal-2aeffa322b9f@spud
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Bindings expect GPIO hog names to end with 'hog' suffix, so correct it
to fix dtbs_check warning:
ep7211-edb7211.dtb: lcden: $nodename:0: 'lcden' does not match '^.+-hog(-[0-9]+)?$'
Link: https://lore.kernel.org/r/20250115211747.194191-1-krzysztof.kozlowski@linaro.org
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Nodes should be sorted by name but it is also nice to have same class of
devices together, so rename both PMU nodes (A53 and M2) to use "pmu"
prefix, instead of suffix.
Link: https://lore.kernel.org/r/20241222145257.31451-1-krzysztof.kozlowski@linaro.org
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
crc32c_arch() is affected by
https://github.com/llvm/llvm-project/issues/20571 where clang
unnecessarily spills the inputs to "rm"-constrained operands to the
stack. Replace "rm" with ASM_INPUT_RM which partially works around this
by expanding to "r" when the compiler is clang. This results in better
code generation with clang, though still not optimal.
Link: https://lore.kernel.org/r/20250210210741.471725-1-ebiggers@kernel.org
Signed-off-by: Eric Biggers <ebiggers@google.com>
Rename kvm_handle_hva_range() to kvm_age_hva_range(),
kvm_handle_hva_range_no_flush() to kvm_age_hva_range_no_flush(), and
__kvm_handle_hva_range() to kvm_handle_hva_range(), as
kvm_age_hva_range() will get more aging-specific functionality.
Suggested-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: James Houghton <jthoughton@google.com>
Link: https://lore.kernel.org/r/20250204004038.1680123-2-jthoughton@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Since we don't do mailbox updates the push send bit
should alwyas clear by the time the delay vblank fires
and the flip completes. Check for that to make sure we
haven't screwed up the sequencing/vblank evasion/etc.
On the DSB path we should be able to guarantee this
since we don't have to deal with any scheduler latencies
and whatnot. I suppose unexpected DMA/memory latencies
might be the only thing that might trip us up here.
For the MMIO path we do always have a non-zero chance
that vblank evasion fails (since we can't really guarantee
anything about the scheduling behaviour). That could trip
up this check, but that seems fine since we already print
errors for other types of vblank evasion failures.
Should the CPU vblank evasion actually fail, then the push
send bit can still be set when the next commit happens. But
both the DSB and MMIO paths should handle that situation
gracefully.
v2: Only check once instead of polling for two scanlines
since we should now be guaranteed to be past the
delayed vblank.
Also check in the MMIO path for good measure
v3: Skip the push send check when VRR is disabled.
With joiner the secondary pipe's DSBs doen't have access
to the transcoder registers, and so doing this check
there triggers a reponse timeout error on the DSB. VRR
is not currently allowed when using joiner, so this will
prevent the bogus register access.
Reviewed-by: Ankit Nautiyal <ankit.k.nautiyal@intel.com>
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20250210160711.24010-1-ville.syrjala@linux.intel.com
When registering an encrypted memory region for SEV-MEM/SEV-ES guests,
pin the pages with FOLL_TERM so that the pages are migrated out of
MIGRATE_CMA/ZONE_MOVABLE. Failure to do so violates the CMA/MOVABLE
mechanisms and can result in fragmentation due to unmovable pages, e.g.
can make CMA allocations fail.
Signed-off-by: Ge Yang <yangge1116@126.com>
Reviewed-by: Tom Lendacky <thomas.lendacky@amd.com>
Acked-by: David Hildenbrand <david@redhat.com>
Link: https://lore.kernel.org/r/1739241423-14326-1-git-send-email-yangge1116@126.com
[sean: massage changelog, make @flags an unsigned int]
Signed-off-by: Sean Christopherson <seanjc@google.com>
Oleksij Rempel says:
====================
Use PHYlib for reset randomization and adjustable polling
This patch set tackles a DP83TG720 reset lock issue and improves PHY
polling. Rather than adding a separate polling worker to randomize PHY
resets, I chose to extend the PHYlib framework - which already handles
most of the needed functionality - with adjustable polling. This
approach not only addresses the DP83TG720-specific problem (where
synchronized resets can lock the link) but also lays the groundwork for
optimizing PHY stats polling across all PHY drivers. With generic PHY
stats coming in, we can adjust the polling interval based on hardware
characteristics, such as using longer intervals for PHYs with stable HW
counters or shorter ones for high-speed links prone to counter
overflows.
Patch version changes are tracked in separate patches.
====================
Link: https://patch.msgid.link/20250210082358.200751-1-o.rempel@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Address the limitations of the DP83TG720 PHY, which cannot reliably
detect or report a stable link state. To handle this, the PHY must be
periodically reset when the link is down. However, synchronized reset
intervals between the PHY and its link partner can result in a deadlock,
preventing the link from re-establishing.
This change introduces a randomized polling interval when the link is
down to desynchronize resets between link partners.
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20250210082358.200751-3-o.rempel@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Introduce the `phy_get_next_update_time` function to allow PHY drivers
to dynamically determine the time (in jiffies) until the next state
update event. This enables more flexible and adaptive polling intervals
based on the link state or other conditions.
Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20250210082358.200751-2-o.rempel@pengutronix.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Tariq Toukan says:
====================
mlx5 misc
Patches 1-3 by William reduce the memory consumption for representors to
achieve better scalability.
Patches 4-5 by Akiva expose ICM memory consumption per function.
Patches 6-8 expose helpful information on RSS resources in devlink RX
reporter diagnose.
Patches 9-10 are simple enhancements by Alex Lazar.
====================
Link: https://patch.msgid.link/20250209101716.112774-1-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In XDP scenarios, fragmented packets can occur if the MTU is larger
than the page size, even when the packet size fits within the linear
part.
If XDP multi-buffer support is disabled, the fragmented part won't be
handled in the TX flow, leading to packet drops.
Since XDP multi-buffer support is always available, this commit removes
the conditional check for enabling it.
This ensures that XDP multi-buffer support is always enabled,
regardless of the `is_xdp_mb` parameter, and guarantees the handling of
fragmented packets in such scenarios.
Signed-off-by: Alexei Lazar <alazar@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20250209101716.112774-16-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Current loopback test validation ignores non-linear SKB case in
the SKB access, which can lead to failures in scenarios such as
when HW GRO is enabled.
Linearize the SKB so both cases will be handled.
Signed-off-by: Alexei Lazar <alazar@nvidia.com>
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20250209101716.112774-15-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Move rx reporter RQs diagnose from mlx5e_rx_reporter_diagnose() to a
dedicated function. This change is a preparation for the following
series which extends diagnose output for the rx reporter. While at it,
also pass a mlx5e_priv pointer to
mlx5e_rx_reporter_diagnose_common_config() as this is the argument the
latter actually needs.
Signed-off-by: Amir Tzin <amirtz@nvidia.com>
Reviewed-by: Aya Levin <ayal@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20250209101716.112774-12-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
ICM is a portion of the host's memory assigned to a function by the OS
through requests made by the NIC's firmware.
PF ICM consumption can be accessed directly, while VF/SF ICM consumption
can be accessed through their representors in switchdev mode.
The value is exposed to the user in granularity of 4KB through the vnic
health reporter as follows:
$ devlink health diagnose pci/0000:08:00.0 reporter vnic
vNIC env counters:
total_error_queues: 0 send_queue_priority_update_flow: 0
comp_eq_overrun: 0 async_eq_overrun: 0 cq_overrun: 0
invalid_command: 0 quota_exceeded_command: 0
nic_receive_steering_discard: 0 icm_consumption: 1032
Signed-off-by: Akiva Goldberger <agoldberger@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20250209101716.112774-11-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Rename mlx5_esw_query_vport_vhca_id to mlx5_vport_get_vhca_id and move
it to vport file. Also, add function declaration to mlx5_core header
file. This better represents the function's usage and allows for it to
be called from other parts of the mlx5_core driver.
Signed-off-by: Akiva Goldberger <agoldberger@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20250209101716.112774-10-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
By default, the mq netdev creates a pfifo_fast qdisc. On a
system with 16 core, the pfifo_fast with 3 bands consumes
16 * 3 * 8 (size of pointer) * 1024 (default tx queue len)
= 393KB. The patch sets the tx qlen to representor default
value, 128 (1<<MLX5E_REP_PARAMS_DEF_LOG_SQ_SIZE), which
consumes 16 * 3 * 8 * 128 = 49KB, saving 344KB for each
representor at ECPF.
Signed-off-by: William Tu <witu@nvidia.com>
Reviewed-by: Daniel Jurgens <danielj@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Link: https://patch.msgid.link/20250209101716.112774-9-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
By experiments, a single queue representor netdev consumes kernel
memory around 2.8MB, and 1.8MB out of the 2.8MB is due to page
pool for the RXQ. Scaling to a thousand representors consumes 2.8GB,
which becomes a memory pressure issue for embedded devices such as
BlueField-2 16GB / BlueField-3 32GB memory.
Since representor netdevs mostly handles miss traffic, and ideally,
most of the traffic will be offloaded, reduce the default non-uplink
rep netdev's RXQ default depth from 1024 to 256 if mdev is ecpf eswitch
manager. This saves around 1MB of memory per regular RQ,
(1024 - 256) * 2KB, allocated from page pool.
With rxq depth of 256, the netlink page pool tool reports
$./tools/net/ynl/cli.py --spec Documentation/netlink/specs/netdev.yaml \
--dump page-pool-get
{'id': 277,
'ifindex': 9,
'inflight': 128,
'inflight-mem': 786432,
'napi-id': 775}]
This is due to mtu 1500 + headroom consumes half pages, so 256 rxq
entries consumes around 128 pages (thus create a page pool with
size 128), shown above at inflight.
Note that each netdev has multiple types of RQs, including
Regular RQ, XSK, PTP, Drop, Trap RQ. Since non-uplink representor
only supports regular rq, this patch only changes the regular RQ's
default depth.
Signed-off-by: William Tu <witu@nvidia.com>
Reviewed-by: Bodong Wang <bodong@nvidia.com>
Reviewed-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Link: https://patch.msgid.link/20250209101716.112774-8-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
For the ECPF and representors, reduce the max MPWRQ size from 256KB (18)
to 128KB (17). This prepares the later patch for saving representor
memory.
With Striding RQ, there is a minimum of 4 MPWQEs. So with 128KB of max
MPWRQ size, the minimal memory is 4 * 128KB = 512KB. When creating page
pool, consider 1500 mtu, the minimal page pool size will be 512KB/4KB =
128 pages = 256 rx ring entries (2 entries per page).
Before this patch, setting RX ringsize (ethtool -G rx) to 256 causes
driver to allocate page pool size more than it needs due to max MPWRQ
is 256KB (18). Ex: 4 * 256KB = 1MB, 1MB/4KB = 256 pages, but actually
128 pages is good enough. Reducing the max MPWRQ to 128KB fixes the
limitation.
Signed-off-by: William Tu <witu@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
Link: https://patch.msgid.link/20250209101716.112774-7-tariqt@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>