On the -d revision of the bcm2712 the pinctrl differs from the c0
revision. The driver already supports both and distinguishes the two
with the compatible string.
Update the compatible string and reg length to reflect the different
pinctrl.
Signed-off-by: Gregor Herburger <gregor.herburger@linutronix.de>
Link: https://lore.kernel.org/r/20260226-raspi-dts-updates-v1-5-60832d20ff04@linutronix.de
Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
The bcm2712 found in the Raspberry Pi 5 has a PiSP Image Signal
Processor back end image processor.
Add the relevant node to the devicetree.
Signed-off-by: Gregor Herburger <gregor.herburger@linutronix.de>
Link: https://lore.kernel.org/r/20260226-raspi-dts-updates-v1-3-60832d20ff04@linutronix.de
Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
The RaspberryPi 5 has 2 PiSP Camera front end controller on the RP1
chipset.
Add the relevant nodes to the devicetree.
Signed-off-by: Gregor Herburger <gregor.herburger@linutronix.de>
Link: https://lore.kernel.org/r/20260226-raspi-dts-updates-v1-2-60832d20ff04@linutronix.de
Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
They are always required and instead of duplicating a definition in each
dts file, place it in dtsi with labels and work based on that.
Also changed each bridge@ to pcie@ to get extra dtc static analysis.
Fixed bridge numbers as a result.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://lore.kernel.org/r/20260302000736.592422-1-rosenp@gmail.com
Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
Expose HW constant value in a shared header, to be used by core/EN
drivers.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260309093435.1850724-10-tariqt@nvidia.com
Reviewed-by: Dragos Tatulea <dtatulea@nvidia.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Downstream patches will introduce SW-only LAG (e.g. shared_fdb without
HW LAG). In this mode the firmware cannot create the LAG demux table,
but vport demuxing is still required.
Move LAG demux flow-table ownership to the LAG layer and introduce APIs
to init/cleanup the demux table and add/delete per-vport rules. Adjust
the RDMA driver to use the new APIs.
In this mode, the LAG layer will create a flow group that matches vport
metadata. Vports that are not native to the LAG master eswitch add the
demux rule during IB representor load and remove it on unload.
The demux rule forward traffic from said vports to their native eswitch
manager via a new dest type - MLX5_FLOW_DESTINATION_TYPE_VHCA_RX.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260309093435.1850724-9-tariqt@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Introduce MLX5_FLOW_DESTINATION_TYPE_VHCA_RX as a new flow steering
destination type.
Wire the new destination through flow steering command setup by mapping
it to MLX5_IFC_FLOW_DESTINATION_TYPE_VHCA_RX and passing the vhca id,
extend forward-destination validation to accept it, and teach the flow
steering tracepoint formatter to print rx_vhca_id.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260309093435.1850724-8-tariqt@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Introduce mlx5_lag_get_dev_seq() which returns a device's sequence
number within the LAG: master is always 0, remaining devices numbered
sequentially. This provides a stable index for peer flow tracking and
vport ordering without depending on native_port_num.
Replace mlx5_get_dev_index() usage in en_tc.c (peer flow array
indexing) and ib_rep.c (vport index ordering) with the new API.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260309093435.1850724-7-tariqt@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Replace the fixed-size array peer_miss_rules[MLX5_MAX_PORTS], indexed
by physical function index, with an xarray indexed by vhca_id.
This decouples peer_miss_rules from mlx5_get_dev_index(), removing the
dependency on a PF-derived index and the arbitrary MLX5_MAX_PORTS bounds
check. Using vhca_id as the key simplifies insertion/removal logic and
scales better across multi-peer topologies.
Initialize and destroy the xarray alongside the existing esw->paired
xarray in mlx5_esw_offloads_devcom_init/cleanup respectively.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260309093435.1850724-6-tariqt@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Replace the use of mlx5_get_dev_index() for xarray indexing with
xa_alloc() to dynamically allocate indices. This decouples the LAG
xarray index from the physical device index.
Update mlx5_ldev_add_netdev/remove_mdev to find entries by dev pointer
and replace mlx5_lag_get_dev_index_by_seq(ldev, MLX5_LAG_P1) calls with
mlx5_lag_get_master_idx() where appropriate.
No functional changes intended
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260309093435.1850724-5-tariqt@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Replace the fixed-size array with a dynamic xarray.
This commit changes:
- Adds mlx5_lag_pf() helper for consistent xarray access
- Converts all direct pf[] accesses to use the new helper/macro
- Dynamically allocates lag_func entries via xa_store/xa_load
No functional changes intended. This prepares the LAG infrastructure
for future flexibility in device indexing.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260309093435.1850724-4-tariqt@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Update the mlx5 IFC headers with newly defined capability and
command-layout bits:
- Add silent_mode_query and rename silent_mode to silent_mode_set cap
fields.
- Add forward_vhca_rx and MLX5_IFC_FLOW_DESTINATION_TYPE_VHCA_RX.
- Expose silent mode fields in the L2 table query command structures.
Update the SD support check to use the new capability name
(silent_mode_set) to match the updated IFC definition.
Signed-off-by: Shay Drory <shayd@nvidia.com>
Reviewed-by: Mark Bloch <mbloch@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260309093435.1850724-3-tariqt@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Add hardware interface definitions for shared headroom pool (SHP) in
port buffer management:
- shp_pbmc_pbsr_support: capability bit in PCAM enhanced features
indicating device support for shared headroom pool in PBMC/PBSR.
- shared_headroom_pool: buffer entry in PBMC register (pbmc_reg_bits)
for the shared headroom pool configuration, reusing the bufferx
layout; reduce trailing reserved region accordingly.
Signed-off-by: Alexei Lazar <alazar@nvidia.com>
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Link: https://patch.msgid.link/20260309093435.1850724-2-tariqt@nvidia.com
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Commits 0ad5bc1ce4 ("drm/v3d: fix up register addresses for V3D 7.x")
and 6fd9487147 ("drm/v3d: add brcm,2712-v3d as a compatible V3D device")
added driver support for V3D on BCM2712, but the corresponding device
tree node is still missing.
Add the V3D device tree node to the BCM2712 DTS.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Stefan Wahren <wahrenst@gmx.net>
Link: https://lore.kernel.org/r/20260114120610.82531-1-mcanal@igalia.com
Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com>
Capture the output of the patch command to detect when a patch applies
with fuzz or line offsets.
If such "fuzz" is detected during the validation phase, warn the user
and display the details. This helps identify input patches that may
need refreshing against the target source tree.
Ensure that internal patch operations (such as those in refresh_patch or
during the final build phase) can still run quietly.
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Acked-by: Song Liu <song@kernel.org>
Link: https://patch.msgid.link/20260310203751.1479229-13-joe.lawrence@redhat.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Improve the readability of klp-build output by implementing a basic
color scheme. When the standard output and error are connected to a
terminal, highlight status messages in bold and warning/error prefixes
in yellow/red.
Acked-by: Song Liu <song@kernel.org>
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Link: https://patch.msgid.link/20260310203751.1479229-12-joe.lawrence@redhat.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Provide more context for common klp-build failure modes. Clarify which
user-provided patch is unsupported or failed to apply, and explicitly
identify which kernel build (original or patched) failed.
Acked-by: Song Liu <song@kernel.org>
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Link: https://patch.msgid.link/20260310203751.1479229-11-joe.lawrence@redhat.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
If .config hasn't been synced with auto.conf, any recent changes to
CONFIG_LOCALVERSION* may not get reflected in the kernel version name.
Use "make syncconfig" to force them to sync, and "make -s kernelrelease"
to get the version instead of having to construct it manually.
Fixes: 24ebfcd65a ("livepatch/klp-build: Introduce klp-build script for generating livepatch modules")
Closes: https://lore.kernel.org/20260217160645.3434685-10-joe.lawrence@redhat.com
Reported-by: Joe Lawrence <joe.lawrence@redhat.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Acked-by: Song Liu <song@kernel.org>
Link: https://patch.msgid.link/20260310203751.1479229-10-joe.lawrence@redhat.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Update SHORT_CIRCUIT behavior to better handle patch validation and
argument processing in later klp-build steps.
Perform patch validation for both step 1 (building original kernel) and
step 2 (building patched kernel) to ensure patches are verified before
any compilation occurs.
Additionally, allow the user to omit input patches when skipping past
step 2.
Acked-by: Song Liu <song@kernel.org>
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Link: https://patch.msgid.link/20260310203751.1479229-9-joe.lawrence@redhat.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Fix or suppress the following shellcheck warnings:
In klp-build line 57:
command grep "$@" || true
^--^ SC2317 (info): Command appears to be unreachable. Check usage (or ignore if invoked indirectly).
Fix the following warning:
In klp-build line 565:
local file_dir="$(dirname "$file")"
^------^ SC2034 (warning): file_dir appears unused. Verify use (or export if used externally).
Acked-by: Song Liu <song@kernel.org>
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Link: https://patch.msgid.link/20260310203751.1479229-8-joe.lawrence@redhat.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Add a standalone Makefile with a 'check' target that runs static code
analysis (shellcheck) on the klp-build script(s). This is intended
strictly as a development aid.
Acked-by: Song Liu <song@kernel.org>
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Link: https://patch.msgid.link/20260310203751.1479229-7-joe.lawrence@redhat.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Provide a custom grep() function to catch direct usage of the command.
Bare grep calls are generally incompatible with pipefail and
errexit behavior (where a failed match causes the script to exit).
Developers can still call grep via command grep if that behavior is
explicitly desired.
Acked-by: Song Liu <song@kernel.org>
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Link: https://patch.msgid.link/20260310203751.1479229-6-joe.lawrence@redhat.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
The klp-build script is currently very strict with input patches,
requiring them to apply cleanly via `git apply --recount`. This
prevents the use of patches with minor contextual fuzz relative to the
target kernel sources.
To allow users to reuse a patch across similar kernel streams, switch to
using GNU patch and patchutils for intermediate patch manipulation.
Update the logic for applying, reverting, and regenerating patches:
- Use 'patch -p1' for better handling of context fuzz.
- Use 'recountdiff' to update line counts after FIX_PATCH_LINES.
- Drop git_refresh() and related git-specific logic.
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Acked-by: Song Liu <song@kernel.org>
Link: https://patch.msgid.link/20260310203751.1479229-5-joe.lawrence@redhat.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
The klp-build script prepares a clean patch by populating two temporary
directories ('a' and 'b') with source files and diffing the result.
However, this process fails when a patch introduces a new source file,
as the script attempts to copy files that do not yet exist in the
original source tree. Likewise, it fails when a patch removes a source
file and the script attempts to copy a file that no longer exists.
Refactor the file-gathering logic to distinguish between original input
files and patched output files:
- Split get_patch_files() into get_patch_input_files() and
get_patch_output_files() to identify which files exist before and
after patch application.
- Filter out "/dev/null" from both to handle file creation/deletion.
- Update refresh_patch() to only copy existing input files to the 'a'
directory and the resulting output files to the 'b' directory.
Acked-by: Song Liu <song@kernel.org>
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Link: https://patch.msgid.link/20260310203751.1479229-4-joe.lawrence@redhat.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
The elf_create_file() function fails with EINVAL when the build directory
path is long enough to truncate the "XXXXXX" suffix in the 256-byte
tmp_name buffer.
Simplify the code to remove the unnecessary dirname()/basename() split
and concatenation. Instead, allocate the exact number of bytes needed for
the path.
Acked-by: Song Liu <song@kernel.org>
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Link: https://patch.msgid.link/20260310203751.1479229-3-joe.lawrence@redhat.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Commit 356e4b2f5b ("objtool: Fix data alignment in elf_add_data()")
corrected the alignment of data within a section (honoring the section's
sh_addralign). Apply the same alignment when klp-diff mode clones a
symbol, adjusting the new symbol's offset for the output section's
sh_addralign.
Fixes: dd590d4d57 ("objtool/klp: Introduce klp diff subcommand for diffing object files")
Signed-off-by: Joe Lawrence <joe.lawrence@redhat.com>
Link: https://patch.msgid.link/20260310203751.1479229-2-joe.lawrence@redhat.com
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
The class_create() call has been deprecated in favor of class_register()
as the driver core now allows for a struct class to be in read-only
memory. Replace vfio->class with a const struct class and drop the
class_create() call.
Compile tested and found no errors/warns in dmesg after enabling
VFIO_GROUP.
Link: https://lore.kernel.org/all/2023040244-duffel-pushpin-f738@gregkh/
Suggested-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Jori Koolstra <jkoolstra@xs4all.nl>
Link: https://lore.kernel.org/r/20260306190628.259203-1-jkoolstra@xs4all.nl
Signed-off-by: Alex Williamson <alex@shazbot.org>
There are two special cases in the idle loop that are handled
inconsistently even though they are analogous.
The first one is when a cpuidle driver is absent and the default CPU
idle time power management implemented by the architecture code is used.
In that case, the scheduler tick is stopped every time before invoking
default_idle_call().
The second one is when a cpuidle driver is present, but there is only
one idle state in its table. In that case, the scheduler tick is never
stopped at all.
Since each of these approaches has its drawbacks, reconcile them with
the help of one simple heuristic. Namely, stop the tick if the CPU has
been woken up by it in the previous iteration of the idle loop, or let
it tick otherwise.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Christian Loehle <christian.loehle@arm.com>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Reviewed-by: Qais Yousef <qyousef@layalina.io>
Reviewed-by: Aboorva Devarajan <aboorvad@linux.ibm.com>
Fixes: ed98c34919 ("sched: idle: Do not stop the tick before cpuidle_idle_call()")
[ rjw: Added Fixes tag, changelog edits ]
Link: https://patch.msgid.link/4741364.LvFx2qVVIh@rafael.j.wysocki
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
All are singletons - please see the changelogs for details.
-----BEGIN PGP SIGNATURE-----
iHUEABYKAB0WIQTTMBEPP41GrTpTJgfdBJ7gKXxAjgUCabhW5gAKCRDdBJ7gKXxA
jiy+AQDq0x5J94Ezavp0//wqpQ3gwCbo+E8cOE4qb3ig20gBFAEA2hr8qvWYeGCV
r+FhX6PAmY6fxL8e/akqR3h7fR7iPAk=
=g8d7
-----END PGP SIGNATURE-----
Merge tag 'mm-hotfixes-stable-2026-03-16-12-15' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
Pull misc fixes from Andrew Morton:
"6 hotfixes. 4 are cc:stable. 3 are for MM.
All are singletons - please see the changelogs for details"
* tag 'mm-hotfixes-stable-2026-03-16-12-15' of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm:
MAINTAINERS: update email address for Ignat Korchagin
mm/huge_memory: fix early failure try_to_migrate() when split huge pmd for shared THP
mm/rmap: fix incorrect pte restoration for lazyfree folios
mm/huge_memory: fix use of NULL folio in move_pages_huge_pmd()
build_bug.h: correct function parameters names in kernel-doc
crash_dump: don't log dm-crypt key bytes in read_key_from_user_keying
Charles Keepax <ckeepax@opensource.cirrus.com> says:
Normally the SDCA jack detection controls will be named after the GE
widget that represents the grouping of everything in the topology
controlled by the jack selection. However, in the case that the jack
selection only controls a single widget the control will be named after
the SU widget that implements that. It is rather confusing to have the
jack detection controls change naming scheme between devices. Add a new
widget type, similar to mixer widgets, to force use of the control name
rather than falling back to the widget names.
Most of the time a GE Entity will control many SU Entities and thus the
generated control will use the control name, since the control is
shared. However, in rare cases the GE Entity will only control a single
SU Entity. In this case the control is only linked to a single SU widget
and will thus inherit the name from that. This leads to the naming for
the Selected Mode control to be unpredictable. Update to use the new
named control muxes to always force the use of the control name.
Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://patch.msgid.link/20260303155308.138989-3-ckeepax@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
There is already a version of the mixer widget that forces use of the
specified control name, rather than factoring in the widget name. Add
the same feature for mux widgets.
Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://patch.msgid.link/20260303155308.138989-2-ckeepax@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
The WCD934x audio outputs (earpiece, headphone, speaker) share two power
supply converters, a buck and a flyback, managed by reference counters
(buck_users, flyback_users) in the Class-H controller.
The early return in wcd_clsh_ctrl_set_state() when nstate == ctrl->state
prevented _wcd_clsh_ctrl_set_state() from being called when switching
between outputs sharing the same state value. As a result, the buck and
flyback reference counters were never decremented on disable, leaving the
converters active and their counters out of sync with the actual hardware
state.
This caused audible distortion on the earpiece output and spurious MBHC
over-current protection interrupts on HPHL/HPHR during output switching.
Remove the early return so that CLSH_REQ_ENABLE and CLSH_REQ_DISABLE are
always dispatched, keeping the buck and flyback reference counters
consistent on every state transition.
Signed-off-by: Cédric Bellegarde <cedric.bellegarde@adishatz.org>
Link: https://patch.msgid.link/20260304141006.280894-1-cedric.bellegarde@adishatz.org
Signed-off-by: Mark Brown <broonie@kernel.org>
The controller reference count is not decremented on deregistration if
the controller has been allocated using devm_spi_alloc_host/target().
Amend the kernel-doc for devm_spi_register_controller() and
spi_unregister_controller() so that it reflects this (more recent)
behaviour.
Fixes: 5e844cc37a ("spi: Introduce device-managed SPI controller allocation")
Signed-off-by: Johan Hovold <johan@kernel.org>
Link: https://patch.msgid.link/20260312151817.32100-6-johan@kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
The controller reference count is not decremented on registration
failures (and has not been for a very long time) contrary to what the
outdated kernel-doc says.
Drop the entire sentence about return values which are now documented in
the Return section.
Fixes: b885244eb2 ("[PATCH] spi: add spi_driver to SPI framework")
Signed-off-by: Johan Hovold <johan@kernel.org>
Link: https://patch.msgid.link/20260312151817.32100-5-johan@kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
The controller per-cpu statistics is not allocated until after the
controller has been registered with driver core, which leaves a window
where accessing the sysfs attributes can trigger a NULL-pointer
dereference.
Fix this by moving the statistics allocation to controller allocation
while tying its lifetime to that of the controller (rather than using
implicit devres).
Fixes: 6598b91b5a ("spi: spi.c: Convert statistics to per-cpu u64_stats_t")
Cc: stable@vger.kernel.org # 6.0
Cc: David Jander <david@protonic.nl>
Signed-off-by: Johan Hovold <johan@kernel.org>
Link: https://patch.msgid.link/20260312151817.32100-3-johan@kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Make sure to deregister from driver core also in the unlikely event that
per-cpu statistics allocation fails during controller registration to
avoid use-after-free (of driver resources) and unclocked register
accesses.
Fixes: 6598b91b5a ("spi: spi.c: Convert statistics to per-cpu u64_stats_t")
Cc: stable@vger.kernel.org # 6.0
Cc: David Jander <david@protonic.nl>
Signed-off-by: Johan Hovold <johan@kernel.org>
Link: https://patch.msgid.link/20260312151817.32100-2-johan@kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Emil Tsalapatis says:
====================
bpf: Relax 8 frame limitation for global subprogs
The BPF verifier currently limits the maximum runtime call stack to
8 frames. Larger BPF programs like sched-ext schedulers routinely
fail verification because they exceed this limit, even as they use
very little actual stack space for each frame.
Relax the verifier to permit call stacks > 8 frames deep when the
call stacks include global subprogs. The old 8 stack frame limit now
only applies to call stacks composed entirely of static function calls.
This works because global functions are each verified in isolation, so
the verifier does not need to cross-reference verification state across
the function call boundary, which has been the reason for limiting the
call stack size in the first place.
This patch does not change the verification time limit of 8 stack
frames. Static functions that are inlined for verification purposes
still only go 8 frames deep to avoid changing the verifier's internal
data structures used for verification. These data structures only
support holding information on up to 8 stack frames.
This patch also does not adjust the actual maximum stack size of 512.
CHANGELOG
=========
v5 -> v6 (https://lore.kernel.org/bpf/20260311182831.91219-1-emil@etsalapatis.com/)
- Make bpf_subprog_call_depth_info internal to verifier.c (Alexei)
v4 -> v5 (https://lore.kernel.org/bpf/20260309204430.201219-1-emil@etsalapatis.com/)
- Move depth tracking state to verifier (Eduard) and free it after verification (Alexei)
- Fix selftest patch title and formatting errors (Yonghong)
v3 -> v4 (https://lore.kernel.org/bpf/20260303043106.406099-1-emil@etsalapatis.com/)
- Factor out temp call depth tracking info into its own struct (Eduard)
- Bring depth calculation loop in line with the other instances (Mykyta)
- Add comment on why selftest call stack is 16 bytes/frame (Eduard)
- Rename "cidx" to "caller" for clarity (Mykyta, Eduard)
v2 -> v3 (https://lore.kernel.org/bpf/20260210213606.475415-1-emil@etsalapatis.com/)
- Change logic to remove arbitrary limit on call depth (Eduard)
- Add additional selftests (Eduard)
v1 -> v2 (https://lore.kernel.org/bpf/20260202233716.835638-1-emil@etsalapatis.com)
- Adjust patch to only increase the runtime stack depth, leaving the
verification-time stack depth unchanged (Alexei)
Signed-off-by: Emil Tsalapatis <emil@etsalapatis.com>
====================
Link: https://patch.msgid.link/20260316161225.128011-1-emil@etsalapatis.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Add tests that demonstrate the verifier support for deep call stacks
while still enforcing maximum stack size limits.
Acked-by: Eduard Zingerman <eddyz87@gmail.com>
Signed-off-by: Emil Tsalapatis <emil@etsalapatis.com>
Link: https://lore.kernel.org/r/20260316161225.128011-3-emil@etsalapatis.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
The BPF verifier currently enforces a call stack depth of 8 frames,
regardless of the actual stack space consumption of those frames. The
limit is necessary for static call stacks, because the bookkeeping data
structures used by the verifier when stepping into static functions
during verification only support 8 stack frames. However, this
limitation only matters for static stack frames: Global subprogs are
verified by themselves and do not require limiting the call depth.
Relax this limitation to only apply to static stack frames. Verification
now only fails when there is a sequence of 8 calls to non-global
subprogs. Calling into a global subprog resets the counter. This allows
deeper call stacks, provided all frames still fit in the stack.
The change does not increase the maximum size of the call stack, only
the maximum number of frames we can place in it.
Also change the progs/test_global_func3.c selftest to use static
functions, since with the new patch it would otherwise unexpectedly
pass verification.
Acked-by: Mykyta Yatsenko <yatsenko@meta.com>
Acked-by: Eduard Zingerman <eddyz87@gmail.com>
Signed-off-by: Emil Tsalapatis <emil@etsalapatis.com>
Link: https://lore.kernel.org/r/20260316161225.128011-2-emil@etsalapatis.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
