Add support for workload submission trace points.
Co-developed-by: Donald Robson <donald.robson@imgtec.com>
Signed-off-by: Donald Robson <donald.robson@imgtec.com>
Signed-off-by: Alexandru Dadu <alexandru.dadu@imgtec.com>
Reviewed-by: Matt Coster <matt.coster@imgtec.com>
Link: https://patch.msgid.link/20260513-b4-pvr-trace-points-v1-1-81222d1a4c99@imgtec.com
Signed-off-by: Matt Coster <matt.coster@imgtec.com>
arm_smmu_write_context_bank() assumes it is being called with RPM
active, but it turns out that is not guaranteed in the path from
qcom_adreno_smmu_set_ttbr0_cfg(), so it's possible for the register
writes to get lost when configuring the context bank while the GPU is
idle, leading to page faults later.
Add the RPM calls here to make sure the SMMU is active before we touch
it.
Signed-off-by: Anna Maniscalco <anna.maniscalco2000@gmail.com>
Reviewed-by: Rob Clark <rob.clark@oss.qualcomm.com>
Reviewed-by: Robin Murphy <robin.murphy@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
Update FW initialised state shared resource access with READ/WRITE_ONCE
to prevent following:
- Non-atomic access to variable in multi-thread/CPU case.
- Merge successive loads from the same variable.
Signed-off-by: Brajesh Gupta <brajesh.gupta@imgtec.com>
Reviewed-by: Matt Coster <matt.coster@imgtec.com>
Link: https://patch.msgid.link/20260519-b4-context_reset-v2-4-931018a7131d@imgtec.com
Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Rename the variable from 'booted' to 'initialised' to align with the
driver's expectations and state, indicating that FW boot initialisation
is complete and the firmware is fully initialised.
Signed-off-by: Brajesh Gupta <brajesh.gupta@imgtec.com>
Reviewed-by: Matt Coster <matt.coster@imgtec.com>
Link: https://patch.msgid.link/20260519-b4-context_reset-v2-3-931018a7131d@imgtec.com
Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Verify the job’s fence in the timeout handler; if the firmware has since
signaled completion, then report NO HANG.
Signed-off-by: Brajesh Gupta <brajesh.gupta@imgtec.com>
Reviewed-by: Matt Coster <matt.coster@imgtec.com>
Link: https://patch.msgid.link/20260519-b4-context_reset-v2-2-931018a7131d@imgtec.com
Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Initialise the context ID for the FW common context correctly by moving
the context allocation earlier.
Signed-off-by: Brajesh Gupta <brajesh.gupta@imgtec.com>
Reviewed-by: Matt Coster <matt.coster@imgtec.com>
Link: https://patch.msgid.link/20260519-b4-context_reset-v2-1-931018a7131d@imgtec.com
Signed-off-by: Matt Coster <matt.coster@imgtec.com>
Recent optimizations of sd->shared assignment moved to allocating a
single instance of per-CPU sched_domain_shared objects per s_data.
Recent optimizations to select_idle_capacity() moved the sd->shared
assignments to "sd_asym" domain when ASYM_CPUCAPACITY is detected but
cache-aware scheduling mandates the presence of "sd_llc_shared" to
compute and cache per-LLC statistics.
Use an "alloc_flags" union in sched_domain_shared to claim a
sched_domain_shared object per sched_domain. Allocation starts searching
for an available / matching sched_domain_shared instance from the first
CPU of sched_domain_span(sd) (sd can be sd_llc, or sd_asym). If the
shared object is claimed by another domain, the instance corresponding
to next CPU in the domain span is explored until a matching / available
instance is found.
In case of a single CPU in sched_domain_span(), the domain will be
degenerated and a temporary overlap of ->shared objects across different
domains is acceptable.
"alloc_flags" forms a union with "nr_idle_scan" and the stale flags are
left as is when the sd->shared is published. The expectation is for the
first load balancing instance to correct the value just like the current
behavior, except the initial value is no longer 0.
Originally-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: K Prateek Nayak <kprateek.nayak@amd.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Andrea Righi <arighi@nvidia.com>
The retry count was initially set to 10, which with the 5 second
timeouts give 55 seconds to fail a message.
The IPMI spec specifies the 5 second timeout, and it specifies retries
for KCS but does not specify a number. 55 seconds is a long time, so
reduce retries to 2. It matches the default in the BT state machine.
This is 15 seconds, then, which is still a long time, but more
reasonable.
Signed-off-by: Corey Minyard <corey@minyard.net>
This set of tests require userspace memory to map objects, so
run them only if this that memory is available.
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-by: Sebastian Brzezinka <sebastian.brzezinka@intel.com>
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Reviewed-by: Janusz Krzysztofik <janusz.krzysztofik@linux.intel.com>
Reviewed-by: Michał Grzelak <michal.grzelak@intel.com>
Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com>
Link: https://lore.kernel.org/r/20260508080214.1979686-3-krzysztof.karas@intel.com
Migration testing in i915 assumes current task's address space
to allocate new userspace mapping and uses it without
registering real user for that address space in mm_struct.
On single NUMA node setups PCI probe executes in the same
context as userspace process calling the test (i915_selftest
from IGT), but when multiple nodes are available, the PCI code
puts probe into a kernel workqueue. This switches execution to
a kworker, which does not have its own address space in
userspace and must borrow such memory from another process, so
"current->active_mm" is unknown at the start of the test.
It was observed that mm->mm_users would occasionally be 0
or drop to 0 during the test due to short delay between
scheduling and executing work in forked process, which reaped
userspace mappings, further leading to failures upon reading
from userland memory.
Prevent this by adding a PID parameter to a trusted task, so its
mm_struct may be used if needed.
Closes: https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/14204
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-by: Michał Grzelak <michal.grzelak@intel.com>
Reviewed-by: Andi Shyti <andi.shyti@linux.intel.com>
Reviewed-by: Sebastian Brzezinka <sebastian.brzezinka@intel.com>
Reviewed-by: Janusz Krzysztofik <janusz.krzysztofik@linux.intel.com>
Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com>
Link: https://lore.kernel.org/r/20260508080214.1979686-2-krzysztof.karas@intel.com
init_irq_stacks() and init_irq_scs() may fail when arch_alloc_vmap_stack
or scs_alloc return NULL. Return -ENOMEM from both and call panic() once
from init_IRQ(), covering per-CPU IRQ stacks and shadow IRQ stacks
consistently.
Signed-off-by: Osama Abdelkader <osama.abdelkader@gmail.com>
Signed-off-by: Will Deacon <will@kernel.org>
Under nosmp (maxcpus=0), arm64 never brings up secondary CPUs.
smp_prepare_cpus() already treats this as a UP-mandated boot and returns
before marking secondary CPUs present. However, smp_init_cpus() may still
enumerate firmware-described secondary CPUs and mark them possible before
that point.
Avoid marking secondary CPUs possible when nosmp/maxcpus=0 is in effect,
so that cpu_possible_mask reflects the nosmp boot policy for this boot.
Suggested-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Pengjie Zhang <zhangpengjie2@huawei.com>
Signed-off-by: Will Deacon <will@kernel.org>
Replace sprintf() with sysfs_emit() in sysfs show functions, which is
preferred for formatting sysfs output because it provides safer bounds
checking.
While the current code only emits fixed strings that fit easily within
PAGE_SIZE, use sysfs_emit() to follow secure coding best practices.
Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev>
Signed-off-by: Will Deacon <will@kernel.org>
In mana_hwc_rx_event_handler(), resp->response.hwc_msg_id is read from
DMA-coherent memory and bounds-checked, then mana_hwc_handle_resp()
re-reads the same field from the same DMA buffer for test_bit() and
pointer arithmetic.
DMA-coherent memory is mapped uncacheable on x86 and is shared,
unencrypted, in Confidential VMs (SEV-SNP/TDX), so each load goes
directly to host-visible memory. A H/W can modify the value
between the check and the use, bypassing the bounds validation.
Fix this by reading hwc_msg_id exactly once using READ_ONCE() into a
stack-local variable in mana_hwc_rx_event_handler(), and passing the
validated value as a parameter to mana_hwc_handle_resp().
Fixes: ca9c54d2d6 ("net: mana: Add a driver for Microsoft Azure Network Adapter (MANA)")
Signed-off-by: Erni Sri Satya Vennela <ernis@linux.microsoft.com>
Link: https://patch.msgid.link/20260514194156.466823-1-ernis@linux.microsoft.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Add qcom,glymur-mdss to the qcom_smmu_actlr_client_of_match[]
table to configure the SMMU-500 context bank for the display
subsystem (MDSS) on the Glymur platform.
The settings disable the next-page prefetcher while keeping
macro TLB caching enabled. Without this entry,
qcom_smmu_set_actlr_dev() finds no match for the glymur MDSS
device and leaves the context bank ACTLR at its reset value.
Signed-off-by: Lokanadha M R <lokanadha.r@oss.qualcomm.com>
Reviewed-by: Pankaj Patil <pankaj.patil@oss.qualcomm.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Signed-off-by: Will Deacon <will@kernel.org>
We have some inconsistency where multiple errata for the same component
share the same Kconfig workaround; some are one ID per line, some are
smooshed together, and some are entirely separate entries. Standardise
on the single entry, one ID per line format so that things render nice
and consistently in the HTML docs, and it's simple and clear to add new
IDs to existing workarounds without churning the table too much.
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Robin Murphy <robin.murphy@arm.com>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Will Deacon <will@kernel.org>
POR_EL0 is expected to be:
- Saved in the poe_context record
- Reset to POR_EL0_INIT when invoking the signal handler
- Restored from poe_context when returning from the signal handler
Add a new test, poe_restore, to check that the save/reset/restore
mechanism is working as intended. See commit 2e8a1acea8 ("arm64:
signal: Improve POR_EL0 handling to avoid uaccess failures") for
more details.
This commit did not handle the case where poe_context is missing
correctly. This was recently fixed; add a new test,
poe_missing_poe_context, to check this case.
Note: td->pass is only set to true at the very end, as an unexpected
signal may occur in case of failure (especially in
poe_missing_poe_context if POR_EL0 is restored to an invalid value).
Failures are tracked with a global, failed_check.
Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
In preparation to adding further POE signal tests, move
get_por_el0() to test_signals_utils.h and add set_por_el0().
Reviewed-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
Add the POE feature to the signal tests framework, to allow tests to
require it.
Reviewed-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
get_header() wants the size of the reserved area in struct
sigcontext, but instead we pass it the size of the entire struct.
This could in theory result in an out-of-bounds read (if the signal
frame is malformed).
Fix this using one of the existing macros from
tools/testing/selftests/arm64/signal/testcases/testcases.h.
This issue was reported by Sashiko on a patch that copied this
portion of the code.
Link: https://sashiko.dev/#/patchset/20260421144252.1440365-1-kevin.brodsky%40arm.com
Fixes: f5b5ea51f7 ("selftests: mm: make protection_keys test work on arm64")
Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
Reviewed-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Will Deacon <will@kernel.org>
I volunteered to maintain the phy-zynqmp driver as part of my work on
the ZynqMP DPSUB driver. Now that Tomi has taken over the DPSUB, it
makes sense for him to handle the phy-zynqmp driver as well.
Signed-off-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Acked-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
Acked-by: Radhey Shyam Pandey <radhey.shyam.pandey@amd.com>
Link: https://patch.msgid.link/20260519082253.40142-1-laurent.pinchart@ideasonboard.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Avoid BUG_ON() while checking for inconsistent page table state conditions
and instead replace them with VM_WARN_ON_ONCE().
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Ryan Roberts <ryan.roberts@arm.com>
Cc: David Hildenbrand <david@kernel.org>
Cc: linux-kernel@vger.kernel.org
Cc: linux-arm-kernel@lists.infradead.org
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Suggested-by: David Hildenbrand (Arm) <david@kernel.org>
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
In i2c_smbus_xfer_emulated(), there are two buffers: msgbuf0 and
msgbuf1, which are used to save a series of messages, as mentioned in
the comment. According to the value of the variable 'size', msgbuf0 is
initialized to various values. In contrast, msgbuf1 is left
uninitialized until the function i2c_transfer() is invoked. However,
msgbuf1 is not always initialized on all possible execution paths
(implementation) of i2c_transfer(). Thus, it is possible that msgbuf1
may still be uninitialized even after the invocation of the function
i2c_transfer(), especially when the return value of i2c_transfer() is
not checked properly. In the following execution, the uninitialized
msgbuf1 will be used, such as for security checks. Since uninitialized
values can be random and arbitrary, this will cause undefined behaviors
or even check bypass. For example, it is expected that if the value of
'size' is I2C_SMBUS_BLOCK_PROC_CALL, the value of data->block[0] should
not be larger than I2C_SMBUS_BLOCK_MAX. This patch initializes the first
byte of msgbuf1 with 0 to avoid such undefined behaviors or security
issues.
Signed-off-by: Wenwen Wang <wang6495@umn.edu>
[wsa: reworded commit message a little]
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Add changes to support new ranking table version. Version 2 of the
heterogeneous ranking table provides static CPU rankings. Version 3 adds
dynamic ranking table support on newer AMD platforms. These changes ensure
that platforms still reporting version 2 continue to function with the
existing static ranking path, avoiding regressions on older hardware
that does not supply a dynamic ranking table.
Signed-off-by: Krishnamoorthi M <krishnamoorthi.m@amd.com>
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
Link: https://patch.msgid.link/20260507190926.1211726-1-krishnamoorthi.m@amd.com
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Commit 5e8323c3d5 ("dt-bindings: arm-smmu: qcom: Add compatible for
Hawi SoC") was intended for the APSS SMMU but was mistakenly placed
under the Adreno GPU SMMU section. Since that compatible is also valid
for the Hawi GPU SMMU, keep that commit as-is and add proper
documentation for the Hawi APSS SMMU here.
Signed-off-by: Mukesh Ojha <mukesh.ojha@oss.qualcomm.com>
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@oss.qualcomm.com>
Signed-off-by: Will Deacon <will@kernel.org>
Make sure those helpers are always inlined and instrumentation safe.
Suggested-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Leonardo Bras <leo.bras@arm.com>
Signed-off-by: Will Deacon <will@kernel.org>
Daniel Golle says:
====================
net: dsa: mt7530: assorted fixes
A batch of small, independent fixes for the MediaTek MT7530 family DSA
driver, addressing long-standing correctness issues that surface on
hardware with bridge VLAN filtering enabled, on link-local frame
reception, and during bridge join/leave transitions.
====================
Link: https://patch.msgid.link/cover.1778766629.git.daniel@makrotopia.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
With bridge VLAN filtering enabled on a port configured as untagged
member of the bridge PVID, ingress untagged frames do not reach the
corresponding bridge VLAN upper interface (br-lan.<vid>). ARP and
similar traffic is visible on the physical port but not delivered
to the VLAN sub-interface.
The MT7530/MT7531 forwards frames to the CPU port with the user
port's PVID tag applied even when the frame ingressed untagged on
the wire, because the CPU port is set to MT7530_VLAN_EG_CONSISTENT
and is a tagged member of the VLAN entry created for the bridge
VLAN. The DSA core then sees a hwaccel-tagged frame whose VID
matches the port's PVID, which the bridge does not treat as the
untagged-on-the-wire frame that the user expects.
Set ds->untag_vlan_aware_bridge_pvid in the mt7530 and mt7531
setup paths so the DSA core strips that hwaccel tag in software
when the parsed VID matches the bridge port's PVID, restoring the
on-the-wire frame as the bridge expects to see it.
Link: https://github.com/openwrt/openwrt/issues/18576
Fixes: 83163f7dca ("net: dsa: mediatek: add VLAN support for MT7530")
Signed-off-by: Edward Parker <edward@topnotchit.com>
[daniel@makrotopia.org: improve commit message]
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Link: https://patch.msgid.link/85d25ea1b26d3c907f815649f2e0bde6560282a3.1778766629.git.daniel@makrotopia.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
After a VLAN-aware bridge is destroyed, creating any VLAN-unaware
bridge loses all connectivity. The VID 0 VLAN table entry used by
VLAN-unaware ports in FALLBACK mode gets corrupted during VLAN-aware
operation: mt7530_hw_vlan_add() overwrites its EG_CON flag with
VTAG_EN and bridge teardown removes ports from its PORT_MEM.
The cleanup code that should restore it never runs because the current
port's dp->vlan_filtering flag is still true when checked (DSA updates
it only after the driver callback returns). Even when restored, the
deferred VLAN deletion events from the switchdev workqueue can corrupt
VID 0 again after the restoration.
Skip the current port in the all_user_ports_removed check, call
mt7530_setup_vlan0() to restore the VID 0 entry, and protect VID 0
from being modified by bridge VLAN operations in port_vlan_add and
port_vlan_del since it is managed exclusively by mt7530_setup_vlan0().
Remove the CPU port PCR and PVC register writes which were clobbering
PORT_VLAN mode and VLAN_ATTR with wrong values.
Fixes: 83163f7dca ("net: dsa: mediatek: add VLAN support for MT7530")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Link: https://patch.msgid.link/da8bdaf08b2427a9057e6cb33e26d41f8a8d5000.1778766629.git.daniel@makrotopia.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The BPC, RGAC1 and RGAC2 registers control the handling of link-local
frames with reserved MAC DAs (01:80:C2:00:00:0x). These frames are
correctly trapped to the CPU port, but the egress VLAN tag attribute was
set to MT7530_VLAN_EG_UNTAGGED which causes the switch to strip any
VLAN tags from trapped frames before they reach the CPU.
This causes VLAN-tagged link-local frames (STP BPDUs, LLDP, PTP Peer
Delay Requests) to arrive at the CPU without their VLAN tag, so they
are delivered to the base network interface instead of the VLAN
sub-interface. The DSA local_termination selftest confirms this: all
link-local protocol tests on VLAN upper interfaces fail.
Set the EG_TAG attribute to MT7530_VLAN_EG_DISABLED (system default)
so that the switch does not modify VLAN tags in trapped frames. This
way VLAN-tagged frames retain their original tag and are delivered to
the correct VLAN sub-interface, matching the behavior of non-trapped
frames which pass through without VLAN tag modification.
Fixes: 69ddba9d17 ("net: dsa: mt7530: fix handling of all link-local frames")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Acked-by: Chester A. Unal <chester.a.unal@arinc9.com>
Link: https://patch.msgid.link/891e0cd34db2a5fe20ceb73283a81fb5f71427ca.1778766629.git.daniel@makrotopia.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
The DSA forwarding selftests bridge_vlan_aware.sh and
bridge_vlan_unaware.sh configure the bridge with ageing_time set to
LOW_AGEING_TIME (1000 centiseconds, i.e. 10 seconds) and then run
learning_test() in lib.sh, which expects a learned FDB entry to be
removed after ageing_time + 10 seconds. On MT7530/MT7531 the entry
persisted past the deadline and the "Found FDB record when should
not" assertion failed.
With msecs=10000, the algorithm in mt7530_set_ageing_time() finds
AGE_CNT=0 and AGE_UNIT=9 as the first exact match (starting the
search from tmp_age_count=0). The per-entry aging counter is
initialized to AGE_CNT when a MAC address is learned, so with
AGE_CNT=0 new entries start with a counter value of 0, which the
hardware treats as "already aged" and never removes, effectively
disabling aging.
Fix this by starting the search from tmp_age_count=1 to ensure
entries always have a non-zero initial aging counter. For a
10-second ageing time this yields AGE_CNT=1 and AGE_UNIT=4 instead:
the timer ticks every 5 seconds and entries are removed after 2
ticks.
Starting the search at AGE_CNT=1 raises the minimum representable
ageing time from 1 to 2 seconds. Without bounds, a stale ageing_time
of 1 second would now make the loop fall through without setting
age_count and age_unit, leaving them uninitialized when written to
the MT7530_AAC hardware register. Set ds->ageing_time_min and
ds->ageing_time_max so the DSA core validates the range before the
callback is invoked, and drop the now-redundant range check from
mt7530_set_ageing_time().
Fixes: ea6d5c924e ("net: dsa: mt7530: support setting ageing time")
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Link: https://patch.msgid.link/7788ded12dc07b1bce329ec35fa70f4b45f3f9b7.1778766629.git.daniel@makrotopia.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Add a generic PHY driver for the TI DS125DF111 Multi-Protocol
Dual-Channel Retimer. The driver currently supports only 10G and 1G link
speeds but it can easily extended to also cover other usecases.
Since the available datasheet (https://www.ti.com/lit/gpn/DS125DF111)
does not name the registers, the name for the macros were determined by
their usage pattern.
A PHY device is created for each of the two channels present on the
retimer. This allows for independent configuration of the two channels.
This capability is especially important on retimers which have more than
2 channels that can be, depending on the board design, connected in
multiple different ways to the SerDes lanes.
Signed-off-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://patch.msgid.link/20260518142026.3098496-4-ioana.ciornei@nxp.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Various names for Qualcomm as a company are used in user-visible config
options: QCOM, Qualcomm and Qualcomm Technologies. Switch to unified
"Qualcomm" so it will be easier for users to identify the options when
for example running menuconfig.
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@oss.qualcomm.com>
Signed-off-by: Will Deacon <will@kernel.org>
The PCIe SSC configuration for the RK3588 and RK3576 SoCs required
additional tuning which is missing. When adding these same SSC
configurations for both of these two SoCs, as well as upcoming
platforms, it's obvious the SSC setup code was largely duplicated
across the platform-specific configuration functions. This becomes
harder to maintain as more platforms are added.
So extract the common SSC logic into a shared helper function,
rk_combphy_common_cfg_ssc(). This cleans up the per-platform drivers
and centralizes the standard configuration as possible.
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Reviewed-by: Heiko Stuebner <heiko@sntech.de>
Signed-off-by: Shawn Lin <shawn.lin@rock-chips.com>
Link: https://patch.msgid.link/1777251433-110466-1-git-send-email-shawn.lin@rock-chips.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The ARM architecture added the HDBSS feature and descriptions of
related registers (HDBSSBR/HDBSSPROD) in the DDI0601(ID121123) version,
add them to Linux.
Signed-off-by: eillon <yezhenyu2@huawei.com>
Signed-off-by: Tian Zheng <zhengtian10@huawei.com>
Signed-off-by: Will Deacon <will@kernel.org>
The package versioning scheme does not enable smooth upgrades from "rc"
releases to the corresponding stable releases (e.g. 7.0.0-rc7 -> 7.0.0)
because pacman considers that a downgrade due to the underscore in
pkgver (e.g. 7.0.0_rc7), see e.g. vercmp(8) for an explanation of the
package version comparison used by pacman. Package versions which are
derived from said releases (e.g. built from git revisions) are
similarly affected. Fix this by modifying pkgver in order to remove the
hyphen from kernel versions containing "-rcN", where N is a
non-negative integer.
Acked-by: Thomas Weißschuh <linux@weissschuh.net>
Signed-off-by: Viktor Jägersküpper <viktor_jaegerskuepper@freenet.de>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Link: https://patch.msgid.link/20260515215913.92481-1-viktor_jaegerskuepper@freenet.de
Fixes: c8578539de ("kbuild: add script and target to generate pacman package")
Signed-off-by: Nicolas Schier <nsc@kernel.org>
Several functions in scripts/mod/file2alias.c build the module alias
string by repeatedly appending into a fixed-size on-stack buffer:
char alias[256] = {};
...
sprintf(alias + strlen(alias), "%X,*", i);
This pattern is unbounded and silently corrupts the stack when the
formatted output exceeds the destination size. Two functions in this
file are realistically reachable with input that overflows their
buffer:
1. do_input_entry() appends across nine bitmap classes
(evbit/keybit/relbit/absbit/mscbit/ledbit/sndbit/ffbit/swbit). The
keybit case alone scans bits from INPUT_DEVICE_ID_KEY_MIN_INTERESTING
(0x71) to INPUT_DEVICE_ID_KEY_MAX (0x2ff), 655 iterations; if a
MODULE_DEVICE_TABLE(input, ...) populates keybit[] densely, the
emission reaches ~3132 bytes — overflowing the 256-byte buffer by
about 12x. include/linux/mod_devicetable.h declares storage for the
full bit range ("keybit[INPUT_DEVICE_ID_KEY_MAX / BITS_PER_LONG + 1]"),
so the worst case is reachable per the ABI.
2. do_dmi_entry() emits one ":<prefix>*<filtered_substr>*" segment per
matched DMI field, up to 4 matches per dmi_system_id. Each substr
is sized as char[79] in struct dmi_strmatch (mod_devicetable.h:584),
and dmi_ascii_filter() copies it verbatim into the alias buffer
without bounds. Worst case: 4 × (1 + 3 + 1 + 79 + 1) = 336 bytes
into alias[256], an 80-byte overflow.
No driver in the current tree triggers either case — every in-tree
INPUT_DEVICE_ID_MATCH_KEYBIT user populates keybit[] very sparsely
(1-3 bits), and no in-tree dmi_system_id has four maximally-long
matches. The concern is defense-in-depth: both unbounded sprintf
chains are silent stack-corruption primitives in a host build tool,
and the buffer sizes have not been revisited since the corresponding
code was first introduced.
The other do_*_entry() handlers in this file (do_usb_entry,
do_cpu_entry, do_typec_entry, ...) were audited and are bounded by
their input field sizes (uint16 IDs, fixed-length keys); their alias
buffers do not need this treatment.
Reproduced under AddressSanitizer with a stand-alone harness mirroring
do_input on a fully-populated keybit:
==18319==ERROR: AddressSanitizer: stack-buffer-overflow
WRITE of size 2 at offset 288 in frame [32, 288) 'alias'
#6 do_input poc.c:44
Stack-canary build:
Abort trap: 6 (strlen(alias)=3134, cap was 256-1)
Add a small alias_append() helper around vsnprintf with a remaining-
space check and call fatal() on overflow, matching the modpost style
for unrecoverable build conditions. do_input() takes the buffer size
as a new parameter; do_input_entry() and do_dmi_entry() pass
sizeof(alias) at every call site. dmi_ascii_filter() takes the
remaining buffer size as well and aborts on truncation. This bounds
every write into the on-stack buffers and turns the latent overflow
into a clean build error if it is ever reached.
Fixes: 1d8f430c15 ("[PATCH] Input: add modalias support")
Reviewed-by: Randy Dunlap <rdunlap@infradead.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Hasan Basbunar <basbunarhasan@gmail.com>
Link: https://patch.msgid.link/20260505161102.44087-1-basbunarhasan@gmail.com
Signed-off-by: Nicolas Schier <nsc@kernel.org>
The connector number in a UCSI CCI notification is a 7-bit field
supplied by the PPM. ucsi_connector_change() uses it to index the
ucsi->connector[] array without checking it against the number of
connectors the PPM reported at init time, so a buggy or malicious PPM
(EC firmware, or an I2C-attached UCSI controller on the ccg / stm32g0 /
glink transports) can drive schedule_work() on memory past the end of
the array.
Reject connector numbers that are zero or exceed cap.num_connectors
before dereferencing the array.
Assisted-by: gkh_clanker_t1000
Cc: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Cc: Benson Leung <bleung@chromium.org>
Cc: Jameson Thies <jthies@google.com>
Cc: Nathan Rebello <nathan.c.rebello@gmail.com>
Cc: Johan Hovold <johan@kernel.org>
Cc: Pooja Katiyar <pooja.katiyar@intel.com>
Cc: Hsin-Te Yuan <yuanhsinte@chromium.org>
Cc: Abel Vesa <abelvesa@kernel.org>
Cc: stable <stable@kernel.org>
Reviewed-by: Abel Vesa <abel.vesa@oss.qualcomm.com>
Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Reviewed-by: Benson Leung <bleung@chromium.org>
Link: https://patch.msgid.link/2026051351-truck-steadfast-df48@gregkh
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
ucsi_displayport_vdm() handles a DP_CMD_CONFIGURE by copying the first
payload VDO from data[], but unlike the equivalent handler in
altmodes/displayport.c it does not check that count covers a VDO beyond
the header. A header-only Configure VDM (count == 1) would read one u32
past the caller's array.
In the normal UCSI path the caller controls count, so this is hardening
for non-standard delivery paths. NAK and bail when no configuration VDO
is present, matching the generic DP altmode driver's existing guard.
Assisted-by: gkh_clanker_t1000
Cc: Pooja Katiyar <pooja.katiyar@intel.com>
Cc: Johan Hovold <johan@kernel.org>
Cc: stable <stable@kernel.org>
Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Link: https://patch.msgid.link/2026051351-vividly-flattered-eb3d@gregkh
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
svdm_consume_modes() checks pmdata->altmodes against the array size once
before the loop over the count, but forgot to check the bound at every
point in the loop.
In the well-behaved SVDM discovery flow this is harmless because each of
at most SVID_DISCOVERY_MAX SVIDs contributes at most MODE_DISCOVERY_MAX
modes, exactly filling altmode_desc[ALTMODE_DISCOVERY_MAX]. But the
CMDT_RSP_ACK handler in tcpm_pd_svdm() does not correlate an incoming
ACK with any request the port actually sent. Once port->partner is set,
an unsolicited Discover Modes ACK is consumed unconditionally. A broken
or malicious port partner can therefore drive altmodes to
ALTMODE_DISCOVERY_MAX - 1 via the normal flow, and then send one extra
Discover Modes ACK with seven VDOs. Because the pre-loop check passes,
the loop could then writes up to five entries past altmode_desc[]. For
mode_data_prime the next field in struct tcpm_port is the
partner_altmode[] pointer array, which then receives partner-chosen
SVID/VDO bytes.
Move the bound check inside the loop so the array can never be indexed
past ALTMODE_DISCOVERY_MAX regardless of how many VDOs the partner
supplies or how the function was reached.
Assisted-by: gkh_clanker_t1000
Cc: Badhri Jagan Sridharan <badhri@google.com>
Cc: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Cc: stable <stable@kernel.org>
Link: https://patch.msgid.link/2026051351-reshuffle-skillful-90af@gregkh
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Properly validate the count passed from a device when calling
svdm_consume_identity() or svdm_consume_identity_sop_prime() as the
device-controlled value could index off of the static arrays, which
could leak data.
Assisted-by: gkh_clanker_t1000
Cc: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Cc: stable <stable@kernel.org>
Reviewed-by: Badhri Jagan Sridharan <badhri@google.com>
Link: https://patch.msgid.link/2026051350-plated-salute-0efe@gregkh
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A broken/malicious port can transmit a CRC-valid frame whose header
advertises up to seven data objects but whose body carries fewer than
that. Check for this, and rightfully reject the message, instead of
reading from uninitialized stack memory.
Assisted-by: gkh_clanker_t1000
Cc: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Cc: "André Draszik" <andre.draszik@linaro.org>
Cc: Badhri Jagan Sridharan <badhri@google.com>
Cc: Amit Sunil Dhamne <amitsd@google.com>
Cc: stable <stable@kernel.org>
Link: https://patch.msgid.link/2026051350-sitter-canopener-9045@gregkh
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A broken/malicious device can send the incorrect count for a status
update VDO, which will cause the kernel to read uninitialized stack data
and send it off elsewhere.
Fix this up by correctly verifying the count for the update object.
Assisted-by: gkh_clanker_t1000
Cc: stable <stable@kernel.org>
Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Link: https://patch.msgid.link/2026051350-reacquire-sculpture-4244@gregkh
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
wcove_read_rx_buffer() copies the PD RX FIFO into the caller's
struct pd_message with
for (i = 0; i < USBC_RXINFO_RXBYTES(info); i++)
regmap_read(wcove->regmap, USBC_RX_DATA + i, msg + i);
which has two problems:
USBC_RXINFO_RXBYTES() is a 5-bit field (max 31) while struct pd_message
is 30 bytes (__le16 header + __le32 payload[PD_MAX_PAYLOAD], packed).
The byte count latched in RXINFO is the number of bytes the port partner
put on the wire, so a malicious partner that transmits a 31-byte frame
can drive the loop one byte past the destination if the WCOVE BMC
receiver does not enforce the PD object-count limit in hardware. The
existing FIXME flagged this as unverified.
Independently, regmap_read() takes an unsigned int * and stores a full
unsigned int at the destination. Passing the byte pointer msg + i means
each iteration writes four bytes; the high three are zero (val_bits is
8) and are normally overwritten by the next iteration, but the final
iteration's high bytes are not. With RXBYTES == 30 the i == 29 iteration
already writes three zero bytes past msg, which sits on the IRQ thread's
stack in wcove_typec_irq().
Clamp the loop to sizeof(struct pd_message) and read each register into
a local before storing only its low byte, so the copy can never exceed
the destination regardless of what RXINFO reports.
Assisted-by: gkh_clanker_t1000
Cc: stable <stable@kernel.org>
Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Link: https://patch.msgid.link/2026051347-clustered-deflected-9543@gregkh
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This includes three patches that harden the XDomain property parser
against possible malicious hosts.
All these have been in linux-next with no reported issues.
-----BEGIN PGP SIGNATURE-----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=fKQQ
-----END PGP SIGNATURE-----
Merge tag 'thunderbolt-for-v7.1-rc5' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/westeri/thunderbolt into usb-linus
Mika writes:
thunderbolt: Fixes for v7.1-rc5
This includes three patches that harden the XDomain property parser
against possible malicious hosts.
All these have been in linux-next with no reported issues.
* tag 'thunderbolt-for-v7.1-rc5' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/westeri/thunderbolt:
thunderbolt: property: Cap recursion depth in __tb_property_parse_dir()
thunderbolt: property: Reject dir_len < 4 to prevent size_t underflow
thunderbolt: property: Reject u32 wrap in tb_property_entry_valid()
RT migration is done aggressively. When a CPU schedules out a high
priority RT task for a lower priority task, it will look to see if there's
any RT tasks that are waiting to run on another CPU that is of higher
priority than the task this CPU is about to run. If it finds one, it will
pull that task over to the CPU and allow it to run there instead.
Normally, this pulling is done by looking at the RT overloaded mask (rto)
which contains all the CPUs in the scheduler domain with RT tasks that are
waiting to run due to a higher priority RT task currently running on their
CPU. The CPU that is about to schedule a lower priority task will grab the
rq lock of the overloaded CPU and move the RT task from that CPU's runqueue
to the local one and schedule the higher priority RT task.
This caused issues when a lot of CPUs would schedule a lower priority task
at the same time. They would all try to grab the same runqueue lock of
the CPU with the overloaded RT tasks. Only the first CPU that got in will
get that task. All the others would wait until they got the runqueue lock
and see there's nothing to pull and do nothing. On systems with lots of
CPUs, this caused a large latency (up to 500us) which is beyond what
PREEMPT_RT is to allow.
The solution to that was to create an RT_PUSH_IPI logic. When any CPU
wanted to pull a task, instead of grabbing the runqueue lock of the
overloaded CPU, it would start by sending an IPI to the overloaded CPU,
and that IPI handler would have the CPU with the waiting RT task do a push
instead. Then that handler would send an IPI to the next CPU with
overloaded RT tasks, and so on. Note, after the first CPU starts this
process, if another CPU wanted to do a pull, it would see that the process
has already begun and would only increment a counter to have the IPIs
continue again.
The RT_PUSH_IPI solved the latency problem with PREEMPT_RT but could cause
a new issue with non PREEMPT_RT. Namely, softirqs run in a threaded
context on PREEMPT_RT but they can run in an interrupt context in non-RT.
If an IPI lands on a CPU that has just woken up multiple RT tasks and the
current CPU is running a non RT or a low priority RT task, instead of
doing a push, it would simply do a schedule on that CPU. But if a softirq
was also executing on this CPU, the schedule would need to wait until the
softirq finished. Until then, the CPU would still be considered overloaded
as there are RT tasks still waiting to run on it.
A live lock occurred on a workload that was doing heavy networking traffic
on a large machine where the softirqs would run 500us out of 750us. And it
would also be waking up RT tasks, causing the RT pull logic to be
constantly executed.
When a softirq triggered on a CPU with RT tasks queued but not running
yet, and the other CPUs would see this CPU as being overloaded, they would
send an IPI over to it. The CPU would notice that the waiting RT tasks are
of higher priority than the currently running task and simply schedule
that CPU instead. But because the softirq was executing, before it could
schedule, it would receive another IPI to do the same. The amount of IPIs
would slow down the currently running softirq so much that before it could
return back to task context, it would execute another softirq never
allowing the CPU to schedule. This live locked that CPU.
As RT_PUSH_IPI was created to help PREEMPT_RT, make it default off if
PREEMPT_RT is not enabled.
Fixes: b6366f048e ("sched/rt: Use IPI to trigger RT task push migration instead of pulling")
Closes: https://lore.kernel.org/all/20260506235716.2530720-1-tj@kernel.org/
Reported-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: stable@vger.kernel.org
Link: https://patch.msgid.link/20260515103740.25ccbed8@gandalf.local.home
