Unfortunately, an erratum with engineering sample that I was not aware
of was exposed by adding pinctrl configuration to the icicle kit.
When routed to MSS IOs, i2c signals are never anything other than tied
low. Being an FPGA, a Libero workaround for this problem was created,
that involves routing i2c signals to the FPGA fabric when the MSS IO
option is selected in the configurator and then back to the intended pin
using the debug "fabric test" capability. This is invisible to user
facing information in the tooling and not mentioned in reference designs
documentation. It manifests solely in the .xml output from the MSS
configuration that the HSS firmware uses to configure the device, which
Linux now overwrites using the pinctrl information. As a result, I never
noticed this.
My original submission had the engineering sample configuration, but I
modified it on application after I was told it didn't work, not
realising that the report came from a colleague with a production
device, where the erratum was fixed and the workaround not automatically
implemented by Libero when creating a design.
Move this part of the pinctrl configuration out of the shared portion of
the icicle device trees, into the portions that are specific to
engineering sample and production devices so that the different settings
for i2c pins can be dealt with.
Although the reference design only has this workaround in place for
i2c1, as i2c0 is genuinely fabric routed, move it too since the
erratum affects both controllers.
x86/bug: Put HAVE_ARCH_BUG_FORMAT_ARGS WARN definitions inside __ASSEMBLER__
Extend the !assembly #ifdef guarding x86's custom WARN helpers to cover the
WARN macros themselves, as they aren't assembly friendly. This helps make
it clear that things like __WARN_validate_printf() don't need a dummy
definition for assembly code.
x86/bug: Add printf() validation to HAVE_ARCH_BUG_FORMAT_ARGS WARNs
Add explicit printf() validation for x86-64's newfangled WARN
implementation, as most (all?) compilers fail to detect basic formatting
issues without the annotation. E.g. even goofs like printing a u64 as a
string aren't detected:
32-bit x86 doesn't support HAVE_ARCH_BUG_FORMAT_ARGS and uses generic
implementations that provide printf() validation. This means there's
now a big blind spot is code that is strictly x86-64. Inconveniently,
new features are also frequently x86-64-only.
Fix the blind 64-bit blind spot.
[ dhansen: changelog tweaks to flesh out the 64-bit-only details ]
DOORBELL_BLOCK_n[0-63] is an array of GPU control register pages.
Each block is memory-mappable and contains a single DOORBELL register
used to trigger actions in the GPU.
Add definitions for the DOORBELL_BLOCK registers using the register! macro
so they can be used by future Tyr interfaces.
Define the MMU_CONTROL register block with the kernel's register! macro
and replace the existing hand-written MMU register definitions with typed
register and field accessors.
This adds typed definitions for the MMU IRQ registers and the per-address
space MMU_AS_CONTROL registers, including TRANSTAB, MEMATTR, LOCKADDR,
COMMAND, FAULTSTATUS, STATUS, and TRANSCFG. It also introduces typed
enums for MMU commands, fault types, access types, address space modes,
memory attributes, and related MMU configuration fields.
For logical 64-bit MMU registers that are exposed as split 32-bit MMIO
registers, define both the typed 64-bit view and explicit low/high 32-bit
registers so the register layout remains documented without relying on
native 64-bit MMIO accesses.
This reduces open-coded bit manipulation, keeps MMU register layout
knowledge in one place, and makes the definitions easier to read and
maintain.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com> Co-developed-by: Daniel Almeida <daniel.almeida@collabora.com> Signed-off-by: Daniel Almeida <daniel.almeida@collabora.com> Reviewed-by: Daniel Almeida <daniel.almeida@collabora.com> Signed-off-by: Deborah Brouwer <deborah.brouwer@collabora.com> Link: https://patch.msgid.link/20260409-b4-tyr-use-register-macro-v5-v5-4-8abfff8a0204@collabora.com
[aliceryhl: reformat long comment] Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Define the JOB_CONTROL register block with the kernel's register! macro
and replace the existing hand-written JOB IRQ register definitions with
typed register and field accessors.
This adds typed definitions for the raw status, clear, mask, and status
registers, including the per-CSG interrupt bits and the global interface
interrupt bit.
This reduces open-coded bit manipulation, keeps the JOB_CONTROL register
layout in one place, and makes the definitions easier to read and
maintain.
Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com> Co-developed-by: Daniel Almeida <daniel.almeida@collabora.com> Signed-off-by: Daniel Almeida <daniel.almeida@collabora.com> Reviewed-by: Daniel Almeida <daniel.almeida@collabora.com> Signed-off-by: Deborah Brouwer <deborah.brouwer@collabora.com> Link: https://patch.msgid.link/20260409-b4-tyr-use-register-macro-v5-v5-3-8abfff8a0204@collabora.com Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Currently, Tyr prints just the upper 16 bits of the GPU_ID in the hex id
field, namely ARCH_MAJOR, ARCH_MINOR, ARCH_REV, and PRODUCT_MAJOR. The
VERSION_* fields are already printed separately as "major", "minor", and
"status".
Print the full 32-bit GPU_ID register instead of shifting it, so the hex
id reflects the complete register contents.
Before this change:
mali-g610 id 0xa867 major 0x0 minor 0x0 status 0x5
After this change:
mali-g610 GPU_ID 0xa8670005 major 0x0 minor 0x0 status 0x5
Daniel Almeida [Thu, 9 Apr 2026 17:51:24 +0000 (10:51 -0700)]
drm/tyr: Use register! macro for GPU_CONTROL
Define the GPU_CONTROL register block with the kernel's register! macro
and switch the current GPU control paths over to the new typed register
definitions.
This replaces manual register constants, bit masks, shifts, and the
hand-written GpuId parsing code with typed register and field accessors.
It also adds helpers for combining split 64-bit registers and uses the new
definitions in reset, L2 power-on, and GPU info readout/logging paths.
This reduces open-coded bit manipulation making the code easier to read
and maintain.
Acked-by: Boris Brezillon <boris.brezillon@collabora.com> Signed-off-by: Daniel Almeida <daniel.almeida@collabora.com> Co-developed-by: Deborah Brouwer <deborah.brouwer@collabora.com> Signed-off-by: Deborah Brouwer <deborah.brouwer@collabora.com> Link: https://patch.msgid.link/20260409-b4-tyr-use-register-macro-v5-v5-1-8abfff8a0204@collabora.com
[aliceryhl: reformat long comment] Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Eliot Courtney [Thu, 23 Apr 2026 07:11:44 +0000 (16:11 +0900)]
gpu: nova-core: simplify and_then with condition to filter
This code is more simply expressed using Option::filter instead of the
and_then with conditional.
This fixes the following warning with latest nightly Rust clippy build:
warning: manual implementation of `Option::filter`
--> drivers/gpu/nova-core/firmware.rs:391:14
|
391 | .and_then(|hdr| {
| ______________^
392 | | if hdr.bin_magic == BIN_MAGIC {
393 | | Some(hdr)
394 | | } else {
... |
397 | | })
| |______________^ help: try: `filter(|hdr| hdr.bin_magic == BIN_MAGIC)`
|
= help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#manual_filter
= note: `-D clippy::manual-filter` implied by `-D warnings`
= help: to override `-D warnings` add `#[allow(clippy::manual_filter)]`
Cc: stable@vger.kernel.org Fixes: d6cb7319e64e ("gpu: nova-core: firmware: add support for common firmware header") Signed-off-by: Eliot Courtney <ecourtney@nvidia.com> Acked-by: Danilo Krummrich <dakr@kernel.org> Reviewed-by: Alice Ryhl <aliceryhl@google.com> Reviewed-by: Gary Guo <gary@garyguo.net> Link: https://patch.msgid.link/20260423-fix-filter-v1-1-b3b197c65daf@nvidia.com
[aliceryhl: add Fixes: tag and quote the warning it fixes] Signed-off-by: Alice Ryhl <aliceryhl@google.com>
Commit 5f4f954bba12 ("dt-bindings: bluetooth: bring the HW description
closer to reality for wcn6855") changed the vddrfa1p7-supply to 1p8
for whatever reason.
The schematics footprint for this chip definitely says 7 on the input
leg and the driver still expects 1p7. Bring it back.
Fixes: 5f4f954bba12 ("dt-bindings: bluetooth: bring the HW description closer to reality for wcn6855") Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@oss.qualcomm.com> Reviewed-by: Abel Vesa <abel.vesa@oss.qualcomm.com> Reviewed-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com> Acked-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com> Signed-off-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com> Link: https://lore.kernel.org/r/20260225-topic-wcn6855_pmu_dtbdings-v3-1-576ec5c4e631@oss.qualcomm.com Signed-off-by: Bjorn Andersson <andersson@kernel.org>
Thomas Weißschuh [Sat, 18 Apr 2026 10:20:00 +0000 (12:20 +0200)]
tools/nolibc: handle 64-bit system call arguments on MIPS N32
The N32 system call ABI expects 64-bit values directly in registers.
This does not work on nolibc currently, as a 'long' is only 32 bits
wide. Switch the system call wrappers to use 'long long' instead which
can handle 64-bit values on N32. As on N64 'long' and 'long long' are
the same, this does not change the behavior there.
Thomas Weißschuh [Sat, 18 Apr 2026 10:19:59 +0000 (12:19 +0200)]
tools/nolibc: handle 64-bit system call arguments on x32
The x32 system call ABI expects 64-bit values directly in registers.
This does not work on nolibc currently, as a 'long' is only 32 bits
wide. Switch the system call wrappers to use 'long long' instead which
can handle 64-bit values on x32. As on x86_64 'long' and 'long long' are
the same, this does not change the behavior there.
Thomas Weißschuh [Sat, 18 Apr 2026 10:19:58 +0000 (12:19 +0200)]
tools/nolibc: cast pointers returned from system calls through integers
Currently all system call wrappers return 'long' integers which can be
directly cast to 'void *' if the returned value is actually a pointer.
An upcoming change will change the system call wrappers to sometimes
return 'long long' which can not be cast to a pointer directly.
Add explicit cast through 'long' to prepare for this.
Thomas Weißschuh [Sat, 18 Apr 2026 10:19:57 +0000 (12:19 +0200)]
tools/nolibc: add __nolibc_arg_to_reg()
In the architecture specific system call glue, all arguments are
currently casted to 'long' to fit into registers. This works for
pointers as 'long' has the same size as pointers.
However the system call registers for X32 and MIPS N32 need to be
'long long' to work correctly for 64-bit values expected by the system
call ABI. Casting a pointer to a 'long long' will produce a compiler
warning while casting 64-bit integers to 'long' will truncate those.
Add a helper which can be used to correctly cast both pointers and
integers into 'long long' registers. Cast the pointers through
'unsigned' to avoid any sign extensions.
Both builtins have been available since at least GCC 3 and clang 3.
Thomas Weißschuh [Sat, 18 Apr 2026 10:19:56 +0000 (12:19 +0200)]
tools/nolibc: also handle _llseek system call
On some architectures the llseek system call contains a leading
underscore. Treat it the same way as llseek and prefer it over the
plain lseek system call as is necessary for 64-bit offset handling.
remoteproc: Dead code cleanup in Kconfig for STM32_RPROC
There is already an 'if REMOTEPROC' condition wrapping this config option,
making the 'depends on REMOTEPROC' statement a duplicate dependency
(dead code).
I propose leaving the outer 'if REMOTEPROC...endif' and removing the
individual 'depends on REMOTEPROC' statement.
This dead code was found by kconfirm, a static analysis tool for Kconfig.
Jiri Olsa [Fri, 24 Apr 2026 15:39:05 +0000 (17:39 +0200)]
bpf: Remove obsolete WARN_ON call
The WARN_ON call in bpf_trampoline_update could never hit, because we
direct the code path with (total == 0) to out label, which effectively
skips the WARN_ON call.
The WARN_ON made sense back then when it checked tr->selector, but now
with total being set just inside the function it's useless.
Daniel Palmer [Sat, 25 Apr 2026 11:13:15 +0000 (20:13 +0900)]
tools/nolibc: Don't use stack protector before setting it up
The stack protector is configured in _start_c() so we shouldn't
use it before then.
Add __nolibc_no_stack_protector to _start_c() to avoid the compiler
generating stack protector code for _start_c() and thus using it
before its configured.
====================
bpf: range_within() must check cnum ranges instead of min/max pairs
This is a follow-up for series [1].
is_state_visited() should check cnum's subset relations using
cnum-based operations, not min/max projections. See patch #1 for
detailed explanation and patch #2 for an example of buggy program
accepted by verifier w/o this fix.
Updated veristat performance numbers compared to master before [1]
merge follow. Measurements done for the same set of selftests/
sched_ext/meta/cilium programs as in [1].
Net increase: 98K insn 88 programs
Net decrease: -282K insn 52 programs
Raw stats filtered as -f insns_pct>1 -f !insns<10000:
========= selftests: master vs experiment =========
Eduard Zingerman [Sat, 25 Apr 2026 22:48:24 +0000 (15:48 -0700)]
selftests/bpf: a test for proper cnums compare in is_state_visited()
Test case demonstrating a bug in cnum comparison logic fixed by
previous commit. A pruning point is reached with r6 in two states:
1. 32-bit range of [0x7FFFFFF0, U32_MAX] ∪ [0, 0x10]
2. 32-bit range of [0x100, 0x200]
At pruning point the buggy is_state_visited() logic would assume that
would assume range (2) to be a subset of (1) and fail to explore the
path performing division by zero.
Eduard Zingerman [Sat, 25 Apr 2026 22:48:23 +0000 (15:48 -0700)]
bpf: range_within() must check cnum ranges instead of min/max pairs
states.c:range_within() must be updated to properly check if
cnum-based range in an old state is a superset of a range in the cur
state. Currently it makes the decision using min/max accessors:
reg_umin(old) <= reg_umin(cur) <= reg_umax(old)
This is wrong for cnums that cross both UT_MAX/0 and ST_MAX/ST_MIN
boundaries. Consider cnum32{base=0x7FFFFFF0, size=0x80000020},
which represents values [0x7FFFFFF0, ..., U32_MAX, 0, ..., 0x10].
Its projections are u32_min/max=0/U32_MAX, s32_min/max=S32_MIN/MAX.
A register with range [0x100, 0x200] (which lies entirely in the gap
of the wrapping range) would pass the min/max check despite having no
overlap with the actual cnum arc.
This commit replaces min/max comparison with cnum{32,64}_is_subset()
operation. The operation implementation is verified using cbmc model
checker in [1].
Thomas Weißschuh [Wed, 22 Apr 2026 12:32:33 +0000 (14:32 +0200)]
selftests: harness: Restore order of test functions
The recent addition of explicit constructor orders for fixture tests
broke the ordering of those relative to non-fixture tests and the
reverse-constructor-order detection.
Restore the ordering of the test functions relative to each other by
using the same explicit test order for all test registrations and
__constructor_order_first().
Rename the constant, as it is not specific to TEST_F() anymore.
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.
None of the remaining boards put useful data into the platform_data
structures, so effectively this only works with DT based probing.
Remove all code that references this data, to stop using the legacy
gpiolib interfaces. The pxa27x version already supports gpio
descriptors, while the pxa25x version now does it the same way.
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.
Konrad Dybcio [Mon, 20 Apr 2026 11:40:28 +0000 (13:40 +0200)]
usb: typec: mux: ps883x: Power the retimer off when not in use
When there's nothing going through the retimer, there's no reason to
keep it online. Put it in reset when possible to save power.
Also, remove the register cache-compare optimization as it makes little
sense now that the chip resets during almost all transitions and
tracking the validity of that cache becomes a headache.
Suggested-by: Jack Pham <jack.pham@oss.qualcomm.com> Signed-off-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com> Acked-by: Heikki Krogerus <heikki.krogerus@linux.intel.com> Tested-by: Anthony Ruhier <aruhier@mailbox.org> Link: https://patch.msgid.link/20260420-topic-ps883x_unused_reset-v1-1-7aabf7004d2a@oss.qualcomm.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
accel/amdxdna: Set default DPM level based on QoS for temporal-only mode
The QoS request provided when creating a hardware context is currently
ignored when operating in temporal-only mode. Change this to use resource
allocation through xrs_allocate_resource(), which sets the default DPM
level according to the QoS request.
When multiple hardware contexts are active, track their required DPM
levels and set the default DPM level to the highest among them.
usb: usblp: fix uninitialized heap leak via LPGETSTATUS ioctl
Just like in a previous problem in this driver, usblp_ctrl_msg() will
collapse the usb_control_msg() return value to 0/-errno, discarding the
actual number of bytes transferred.
Ideally that short command should be detected and error out, but many
printers are known to send "incorrect" responses back so we can't just
do that.
statusbuf is kmalloc(8) at probe time and never filled before the first
LPGETSTATUS ioctl.
usblp_read_status() requests 1 byte. If a malicious printer responds
with zero bytes, *statusbuf is one byte of stale kmalloc heap,
sign-extended into the local int status, which the LPGETSTATUS path then
copy_to_user()s directly to the ioctl caller.
Fix this all by just zapping out the memory buffer when allocated at
probe time. If a later call does a short read, the data will be
identical to what the device sent it the last time, so there is no
"leak" of information happening.
usb: usblp: fix heap leak in IEEE 1284 device ID via short response
usblp_ctrl_msg() collapses the usb_control_msg() return value to
0/-errno, discarding the actual number of bytes transferred. A broken
printer can complete the GET_DEVICE_ID control transfer short and the
driver has no way to know.
usblp_cache_device_id_string() reads the 2-byte big-endian length prefix
from the response and trusts it (clamped only to the buffer bounds).
The buffer is kmalloc(1024) at probe time. A device that sends exactly
two bytes (e.g. 0x03 0xFF, claiming a 1023-byte ID) leaves
device_id_string[2..1022] holding stale kmalloc heap.
That stale data is then exposed:
- via the ieee1284_id sysfs attribute (sprintf("%s", buf+2), truncated
at the first NUL in the stale heap), and
- via the IOCNR_GET_DEVICE_ID ioctl, which copy_to_user()s the full
claimed length regardless of NULs, up to 1021 bytes of uninitialized
heap, with the leak size chosen by the device.
Fix this up by just zapping the buffer with zeros before each request
sent to the device.
usb: dwc3: Move GUID programming after PHY initialization
The Linux Version Code is currently written to the GUID register before
PHY initialization. Certain PHY implementations (such as Synopsys eUSB
PHY performing link_sw_reset) clear the GUID register to its default
value during initialization, causing the kernel version information to
be lost.
Move the GUID register programming to occur after PHY initialization
completes to ensure the Linux version information persists.
The port in debug accessory mode can be either a source or sink. The
previous tcpm_port_is_debug() function only checked for source port.
Commit 8db73e6a42b6 ("usb: typec: tcpm: allow sink (ufp) to toggle into
accessory mode debug") changed the detection logic to support both roles,
but left some logic in _tcpm_cc_change() unchanged, This causes the state
machine to transition to an incorrect state when operating as a sink in
debug accessory mode. Log as below:
It should go to SNK_ATTACH_WAIT instead of SRC_ATTACH_WAIT state.
To fix this, add tcpm_port_is_debug_source() and tcpm_port_is_debug_sink()
helper to explicitly identify the power mode in debug accessory mode.
Update the state transition logic in _tcpm_cc_change() to ensure the state
machine transitions comply with Type-C specification. Also update the logic
in run_state_machine() to keep consistency.
Fixes: 8db73e6a42b6 ("usb: typec: tcpm: allow sink (ufp) to toggle into accessory mode debug") Cc: stable <stable@kernel.org> Signed-off-by: Xu Yang <xu.yang_2@nxp.com> Acked-by: Heikki Krogerus <heikki.krogerus@linux.intel.com> Reviewed-by: Amit Sunil Dhamne <amitsd@google.com> Link: https://patch.msgid.link/20260424074009.2979266-1-xu.yang_2@nxp.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
usb: typec: tcpm: reset internal port states on soft reset AMS
Reset internal port states (such as vdm_sm_running and
explicit_contract) on soft reset AMS as the port needs to negotiate a
new contract. The consequence of leaving the states in as-is cond are as
follows:
* port is in SRC power role and an explicit contract is negotiated
with the port partner (in sink role)
* port partner sends a Soft Reset AMS while VDM State Machine is
running
* port accepts the Soft Reset request and the port advertises src caps
* port partner sends a Request message but since the explicit_contract
and vdm_sm_running are true from previous negotiation, the port ends
up sending Soft Reset instead of Accept msg.
Felix Gu [Tue, 7 Apr 2026 13:21:22 +0000 (21:21 +0800)]
usb: ulpi: fix memory leak on ulpi_register() error paths
Commit 01af542392b5 ("usb: ulpi: fix double free in
ulpi_register_interface() error path") removed kfree(ulpi) from
ulpi_register_interface() to fix a double-free when device_register()
fails.
But when ulpi_of_register() or ulpi_read_id() fail before
device_register() is called, the ulpi allocation is leaked.
Add kfree(ulpi) on both error paths to properly clean up the allocation.
Fixes: 01af542392b5 ("usb: ulpi: fix double free in ulpi_register_interface() error path") Cc: stable <stable@kernel.org> Signed-off-by: Felix Gu <ustc.gu@gmail.com> Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com> Link: https://patch.msgid.link/20260407-ulpi-v1-1-f3fafe53f7b2@gmail.com Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Aaro Koskinen [Mon, 13 Apr 2026 18:49:12 +0000 (21:49 +0300)]
USB: omap_udc: DMA: Don't enable burst 4 mode
Commit 65111084c63d7 ("USB: more omap_udc updates (dma and omap1710)")
added setting for DMA burst 4 mode. But I think this should be undone for
two reasons:
- It breaks DMA on 15xx boards - transfers just silently stall.
- On newer OMAP1 boards, like Nokia 770 (omap1710), there is no measurable
performance impact when testing TCP throughput with g_ether with large
15000 byte MTU size.
It's also worth noting that when the original change was made, the
OMAP_DMA_DATA_BURST_4 handling in arch/arm/plat-omap/dma.c was broken, and
actually resulted in the same as the OMAP_DMA_DATA_BURST_DIS i.e. burst
disabled. This was fixed not until a couple kernel releases later in an
unrelated commit 1a8bfa1eb998a ("[ARM] 3142/1: OMAP 2/5: Update files
common to omap1 and omap2").
So based on this it seems there was never really a very good reason to
enable this burst mode in omap_udc, so remove it now to allow 15xx DMA
to work again (it provides 2x throughput compared to PIO mode).
Fixes: 65111084c63d ("[PATCH] USB: more omap_udc updates (dma and omap1710)") Cc: stable <stable@kernel.org> Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi> Link: https://patch.msgid.link/ad06qHLclWHeSGnV@darkstar.musicnaut.iki.fi Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Chen-Yu Tsai [Tue, 24 Mar 2026 16:43:54 +0000 (00:43 +0800)]
soc: sunxi: sram: Add H616 SRAM regions
The Allwinner H616 has two switchable peripheral SRAM regions:
- The VE SRAM is a 2 MB dedicated SRAM for the Video Engine. CPU access
to this region is enabled by default. CPU access can be disabled,
after which reads will show the same stale value for all addresses,
while writes are ignored.
The mux value for this region is different from previous generations.
- The SRAM C region is an alias of the first 128 KB of VE SRAM, plus 64
KB of DE SRAM. The latter is otherwise unaccessible from the CPU. When
CPU access is disabled, the whole region reads as zero, while writes
are ignored.
The mux value for this region is the same as on the A64 and H6.
Add data for the VE SRAM. The register values were taken from the BSP
vendor kernel.
Chen-Yu Tsai [Tue, 24 Mar 2026 16:43:52 +0000 (00:43 +0800)]
soc: sunxi: sram: Allow SRAM to be claimed multiple times
On the H616, the SRAM C region is an alias mapping to part of the VE
SRAM (accessible in whole at a different address) and part of the DE
SRAM (otherwise unaccessible). As such both the VE and DE need to claim
this SRAM region to prevent access from the CPU.
The SRAM claim API is designed so that a "claim" routes the SRAM to the
peripheral device, disabling access from the CPU. So long as the written
register value is the same for all the claimants involved, allowing
multiple or repeated claims is trivial. This is indeed the case for all
supported SRAM regions. The only known SRAM region to have multiple
different settings is the SRAM C2 region; this can be claimed by the AE,
CE, or ACE (assumed to be AE + CE). This region is not supported, and
likely will never be needed nor supported, as there is no documentation
for the peripherals involved.
Change the SRAM region "claimed" field from a boolean to a reference
count. A claim will increment the count, while a release decreases it.
The first claim will trigger the register value write. The driver
otherwise behaves as before.
Chen-Yu Tsai [Tue, 24 Mar 2026 16:43:50 +0000 (00:43 +0800)]
dt-bindings: sram: sunxi-sram: Add H616 SRAM regions
The Allwinner H616 has two switchable peripheral SRAM regions:
- The VE SRAM is a 2 MB dedicated SRAM for the Video Engine. CPU access
to this region is enabled by default. CPU access can be disabled,
after which reads will show the same stale value for all addresses,
while writes are ignored.
The mux value for this region is different from previous generations.
- The SRAM C region is an alias of the first 128 KB of VE SRAM, plus 64
KB of DE SRAM. The latter is otherwise unaccessible from the CPU. When
CPU access is disabled, the whole region reads as zero, while writes
are ignored.
The mux value for this region is the same as on the A64 and H6.
Chen-Yu Tsai [Tue, 24 Mar 2026 16:43:49 +0000 (00:43 +0800)]
dt-bindings: sram: Document Allwinner H616 VE SRAM
The Allwinner H616 has two switchable peripheral SRAM regions:
- The VE SRAM is a 2 MB dedicated SRAM for the Video Engine. CPU access
to this region is enabled by default. CPU access can be disabled,
after which reads will show the same stale value for all addresses,
while writes are ignored.
The mux value for this region is different from previous generations,
and thus needs a completely new compatible.
- The SRAM C region is an alias of the first 128 KB of VE SRAM, plus 64
KB of DE SRAM. The latter is otherwise unaccessible from the CPU. When
CPU access is disabled, the whole region reads as zero, while writes
are ignored.
The mux value for this region is the same as on the A64 and H6. The
existing compatible for the A64 already covers this.
Add the compatible for the VE SRAM to the list of covered compatibles in
the generic SRAM region binding.
Revert b0e85c3c8554 ("PCI: Add Kconfig options for MPS/MRRS strategy"),
which allowed build-time selection of the "off", "default", "safe",
"performance", or "peer2peer" strategies for MPS and MRRS configuration.
These strategies can be selected at boot-time using the
"pci=pcie_bus_tune_*" kernel parameters.
Per the discussion mentioned below, these Kconfig options were added to
work around a hardware defect in a WiFi device used in a cable modem. The
defect occurred only when the device was configured with MPS=128, and
Kconfig was a way to avoid that setting. It was easier for the modem
vendor to use Kconfig and update the kernel image than to change the kernel
parameters.
Neither Kconfig nor kernel parameters are a complete solution because the
broken WiFi device may be used in other systems where it may be configured
with MPS=128 and be susceptible to the defect.
Remove the Kconfig settings to simplify the MPS code. If we can identify
the WiFi device in question, we may be able to make a generic quirk to
avoid the problem on all system.
This is not a fix and should not be backported to previous kernels.
Matt Evans [Thu, 23 Apr 2026 17:30:51 +0000 (10:30 -0700)]
PCI/P2PDMA: Avoid returning a provider for non_mappable_bars
Extend the checks in pcim_p2pdma_init() and pcim_p2pdma_provider() to
exclude functions that have pdev->non_mappable_bars set.
Consumers such as VFIO were previously able to map these for access by the
CPU or P2P. Update the comment on non_mappable_bars to show it refers to
any access, not just userspace CPU access.
Fixes: 372d6d1b8ae3c ("PCI/P2PDMA: Refactor to separate core P2P functionality from memory allocation") Suggested-by: Alex Williamson <alex@shazbot.org> Signed-off-by: Matt Evans <mattev@meta.com> Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Niklas Schnelle <schnelle@linux.ibm.com> Reviewed-by: Alex Williamson <alex@shazbot.org> Link: https://patch.msgid.link/20260423173051.1999679-1-mattev@meta.com
PCI/pwrctrl: Do not try to power on/off devices that don't need pwrctrl
pci_pwrctrl_is_required() detects whether a device needs PCI pwrctrl
support. It is currently used in pci_pwrctrl_create_device(), but not in
pci_pwrctrl_power_{on/off}_device() APIs. This leads to pwrctrl core trying
to power on/off incompatible devices like USB hub downstream ports defined
in DT.
Add this check to prevent pwrctrl core from poking at wrong devices.
Fixes: b35cf3b6aa1e ("PCI/pwrctrl: Add APIs to power on/off pwrctrl devices") Reported-by: Krishna Chaitanya Chundru <krishna.chundru@oss.qualcomm.com> Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@oss.qualcomm.com>
[bhelgaas: split pci_pwrctrl_is_required() move to separate patch] Signed-off-by: Bjorn Helgaas <bhelgaas@google.com> Reviewed-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com> Link: https://patch.msgid.link/20260421104102.12322-1-manivannan.sadhasivam@oss.qualcomm.com
Marco Nenciarini [Fri, 17 Apr 2026 13:24:37 +0000 (15:24 +0200)]
PCI/IOV: Skip VF Resizable BAR restore on read error
sriov_restore_vf_rebar_state() uses the VF Resizable BAR Control register
to decide how many VF BARs to restore (nbars) and which VF BAR each
iteration addresses (bar_idx). bar_idx indexes into dev->sriov->barsz[],
which has only PCI_SRIOV_NUM_BARS (6) entries.
When a device does not respond, config reads typically return
PCI_ERROR_RESPONSE (~0). Both fields are 3 bits wide, so nbars and bar_idx
both evaluate to 7. The barsz[] access then goes out of bounds. UBSAN
reports this as:
UBSAN: array-index-out-of-bounds in drivers/pci/iov.c:948:51 index 7 is out of range for type 'resource_size_t [6]'
Observed on an NVIDIA RTX PRO 1000 GPU (GB207GLM) that stopped responding
during a failed GC6 power state exit. The subsequent pci_restore_state()
invoked sriov_restore_vf_rebar_state() while config reads returned
0xffffffff, triggering the splat.
Bail out if any VF Resizable BAR Control read returns PCI_ERROR_RESPONSE.
No further VF BARs are touched, which is safe because a config read that
returns PCI_ERROR_RESPONSE indicates the device is unreachable and
restoration is pointless. This mirrors the guard in
pci_restore_rebar_state().
Marco Nenciarini [Fri, 17 Apr 2026 13:24:36 +0000 (15:24 +0200)]
PCI: Skip Resizable BAR restore on read error
pci_restore_rebar_state() uses the Resizable BAR Control register to decide
how many BARs to restore (nbars) and which BAR each iteration addresses
(bar_idx).
When a device does not respond, config reads typically return
PCI_ERROR_RESPONSE (~0). Both fields are 3 bits wide, so nbars and bar_idx
both evaluate to 7, past the spec's valid ranges for both fields.
pci_resource_n() then returns an unrelated resource slot, whose size is
used to derive a nonsensical value written back to the Resizable BAR
Control register.
Bail out if any Resizable BAR Control read returns PCI_ERROR_RESPONSE. No
further BARs are touched, which is safe because a config read that returns
PCI_ERROR_RESPONSE indicates the device is unreachable and restoration is
pointless.
PCI: Stop setting cached power state to 'unknown' on unbind
When a PCI device is unbound from its driver, pci_device_remove() sets the
cached power state in pci_dev->current_state to PCI_UNKNOWN. This was
introduced by commit 2449e06a5696 ("PCI: reset pci device state to unknown
state for resume") to invalidate the cached power state in case the system
is subsequently put to sleep.
For bound devices, the cached power state is set to PCI_UNKNOWN in
pci_pm_suspend_noirq(), immediately before entering system sleep.
Extend to unbound devices for consistency.
This obviates the need to change the cached power state on unbind, so stop
doing so.
ntfs: fix error handling in ntfs_write_iomap_end_resident()
When ntfs_attr_get_search_ctx() fails and returns NULL, the function
returned early without calling put_page(ipage).
Fix this by jumping to err_out label on error. The err_out path now
properly releases the page and the mutex, with a NULL check for
the search context.
Reported-by: DaeMyung Kang <charsyam@gmail.com> Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Alexey Klimov [Wed, 25 Mar 2026 00:26:32 +0000 (00:26 +0000)]
arm64: dts: exynos850: Add syscon-poweroff node
Without poweroff node Exynos850-based board continue to draw current
(around ~60 mA with my test setup) after poweroff. Kernel also reports
different lockup problems and RCU stalls warnings continuosly after
last kernel messages about hardware being switched off.
Turns out we missed a write to PMU's PS_HOLD_CONTROL (PMU + 0x30c)
register that actually switches the SoC off.
Add poweroff node that implements this.
With this change the current draw after power off is in range of few
milliampers and lockup messages are no more.
ntfs: fix VCN overflow in ntfs_mapping_pairs_decompress()
In ntfs_mapping_pairs_decompress(), lowest_vcn is read from
on-disk metadata and used as the initial vcn without validation.
A malformed value can introduce an invalid (e.g. negative) vcn,
corrupting the runlist from the start.
Additionally, the accumulation
vcn += deltaxcn
does not check for s64 overflow. A crafted mapping pairs array
can wrap vcn to a negative value, breaking the monotonically-
increasing invariant relied upon by ntfs_rl_vcn_to_lcn() and
related helpers.
Fix this by validating lowest_vcn and using check_add_overflow()
for vcn accumulation.
DaeMyung Kang [Sun, 26 Apr 2026 04:02:32 +0000 (13:02 +0900)]
ntfs: fix WSL symlink target leak on reparse failure
ntfs_reparse_set_wsl_symlink() converts the symlink target into an
allocated NLS string and transfers ownership to ni->target only after
ntfs_set_ntfs_reparse_data() succeeds. If setting the reparse data fails,
the converted target is left unreferenced and leaks.
Free the converted target on the reparse update failure path. Use kfree()
for the other local failure path as well, matching the ntfs_ucstonls()
allocation contract.
Fixes: fc053f05ca28 ("ntfs: add reparse and ea operations") Signed-off-by: DaeMyung Kang <charsyam@gmail.com> Reviewed-by: Hyunchul Lee <hyc.lee@gmail.com> Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
DaeMyung Kang [Sun, 26 Apr 2026 04:02:31 +0000 (13:02 +0900)]
ntfs: fix NULL dereference in ntfs_index_walk_down()
ntfs_index_walk_down() allocates ictx->ib when descending from the root
into an index allocation block. If that allocation fails, the old code
still passes the NULL buffer to ntfs_ib_read(), which can write through
it via ntfs_inode_attr_pread().
Allocate the index block into a temporary pointer and return -ENOMEM
before changing the index context on allocation failure. Also propagate
ERR_PTR() through ntfs_index_next() and ntfs_readdir() so walk-down
allocation or index block read failures are not mistaken for normal
index iteration inside the filesystem.
ntfs_readdir() keeps the existing userspace-visible behavior of
suppressing readdir errors after marking end_in_iterate; this change only
prevents the walk-down failure path from dereferencing NULL internally.
The failure was reproduced with failslab fail-nth injection on getdents64;
the original module hits a NULL pointer dereference in memcpy_orig through
ntfs_ib_read(), while the patched module reaches the same
ntfs_index_walk_down() allocation failure without crashing.
Fixes: 0a8ac0c1fa0b ("ntfs: update directory operations") Signed-off-by: DaeMyung Kang <charsyam@gmail.com> Reviewed-by: Hyunchul Lee <hyc.lee@gmail.com> Signed-off-by: Namjae Jeon <linkinjeon@kernel.org>
Haoyu Lu [Fri, 10 Apr 2026 13:37:35 +0000 (21:37 +0800)]
mtd: mtdsuper: replace simple_strtoul with kstrtouint
Modernize string-to-number conversion in mtdsuper.c by replacing
simple_strtoul with kstrtouint. This change provides proper type safety
for MTD device numbers which are non-negative integers.
Using kstrtouint avoids unsigned long to int conversion and is more
appropriate for device indices. The debug output format specifier is
updated to %u for unsigned int.
Signed-off-by: Haoyu Lu <hechushiguitu666@gmail.com> Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Haoyu Lu [Fri, 10 Apr 2026 13:37:34 +0000 (21:37 +0800)]
mtd: mtdoops: replace simple_strtoul with kstrtouint
Replace deprecated simple_strtoul with kstrtouint for better error
handling and type safety. The kstrtouint function provides stricter
validation, automatically rejecting inputs like "123abc" that
simple_strtoul would partially accept.
Using kstrtouint avoids unsigned long to int conversion and is more
appropriate for MTD device indices which are non-negative integers.
Signed-off-by: Haoyu Lu <hechushiguitu666@gmail.com> Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
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: Miquel Raynal <miquel.raynal@bootlin.com>
mtd: spinand: winbond: Fix ODTR write VCR on W35NxxJW
In most scenarios this variant is actually unused (VCR is written in
SSDR mode), but we need to provide an octal variant. The address is 24
bits but is sent over 4 bytes MSB first. This means we need to shift the
register address by one extra byte for the address to be correct.
I didn't catch this initially because the volatile register region is
256 bytes wide, so the write-then-read procedure did work with the small
register addresses I was using at that time: 0 and 1.
mtd: spinand: winbond: Set the packed page read flag to W35N02/04JW
Both W35N02JW and W35N04JW diverge from W35N01JW when it comes to the
"data read" operation in ODTR mode. In order to stuff more address
bits (up to 18), the second command byte is replaced by the most
significant address bits, keeping the number of address bytes to 2.
mtd: spinand: Add support for packed read data ODTR commands
Some devices stuff address bits in the double byte opcode (in place of
the repeated byte) in order to be able to increase the size of the
devices, without adding extra address bytes.
Create a flag to identify those devices. When the flag is set, use the
"packed" variant for the read data operation.
mtd: spi-nor: debugfs: fix out-of-bounds read in spi_nor_params_show()
Sashiko noticed an out-of-bounds read [1].
In spi_nor_params_show(), the snor_f_names array is passed to
spi_nor_print_flags() using sizeof(snor_f_names).
Since snor_f_names is an array of pointers, sizeof() returns the total
number of bytes occupied by the pointers
(element_count * sizeof(void *))
rather than the element count itself. On 64-bit systems, this makes the
passed length 8x larger than intended.
Inside spi_nor_print_flags(), the 'names_len' argument is used to
bounds-check the 'names' array access. An out-of-bounds read occurs
if a flag bit is set that exceeds the array's actual element count
but is within the inflated byte-size count.
Correct this by using ARRAY_SIZE() to pass the actual number of
string pointers in the array.
David Carlier [Fri, 24 Apr 2026 18:36:38 +0000 (19:36 +0100)]
mm/userfaultfd: detect VMA type change after copy retry in mfill_copy_folio_retry()
mfill_copy_folio_retry() drops mmap_lock for the copy_from_user() call.
During this window, the VMA can be replaced with a different type (e.g.
hugetlb), making the caller's ops pointer stale. Subsequent use of the
stale ops would dispatch into the wrong per-vma handlers.
Capture the VMA's ops via vma_uffd_ops() before dropping the lock and
compare against the current vma_uffd_ops() after re-acquiring it.
Return -EAGAIN if they differ so the operation can be retried. This
avoids comparing against the caller's ops which may have been
overridden to anon_uffd_ops for MAP_PRIVATE file-backed mappings.
Link: https://lore.kernel.org/20260424183638.196227-1-devnexen@gmail.com Fixes: 6ab703034f14 ("userfaultfd: mfill_atomic(): remove retry logic") Reported-by: Usama Arif <usama.arif@linux.dev> Closes: https://lore.kernel.org/all/20260410114809.3592720-1-usama.arif@linux.dev/ Signed-off-by: David Carlier <devnexen@gmail.com> Acked-by: Mike Rapoport (Microsoft) <rppt@kernel.org> Cc: Jann Horn <jannh@google.com> Cc: Liam Howlett <liam.howlett@oracle.com> Cc: Lorenzo Stoakes <ljs@kernel.org> Cc: Peter Xu <peterx@redhat.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
SeongJae Park [Sun, 19 Apr 2026 16:10:02 +0000 (09:10 -0700)]
mm/damon/stat: detect and use fresh enabled value
DAMON_STAT updates 'enabled' parameter value, which represents the running
status of its kdamond, when the user explicitly requests start/stop of the
kdamond. The kdamond can, however, be stopped even if the user explicitly
requested the stop, if ctx->regions_score_histogram allocation failure at
beginning of the execution of the kdamond. Hence, if the kdamond is
stopped by the allocation failure, the value of the parameter can be
stale.
Users could show the stale value and be confused. The problem will only
rarely happen in real and common setups because the allocation is arguably
too small to fail. Also, unlike the similar bugs that are now fixed in
DAMON_RECLAIM and DAMON_LRU_SORT, kdamond can be restarted in this case,
because DAMON_STAT force-updates the enabled parameter value for user
inputs. The bug is a bug, though.
The issue stems from the fact that there are multiple events that can
change the status, and following all the events is challenging.
Dynamically detect and use the fresh status for the parameters when those
are requested.
SeongJae Park [Sun, 19 Apr 2026 16:10:01 +0000 (09:10 -0700)]
mm/damon/lru_sort: detect and use fresh enabled and kdamond_pid values
DAMON_LRU_SORT updates 'enabled' and 'kdamond_pid' parameter values, which
represents the running status of its kdamond, when the user explicitly
requests start/stop of the kdamond. The kdamond can, however, be stopped
in events other than the explicit user request in the following three
events.
1. ctx->regions_score_histogram allocation failure at beginning of the
execution,
2. damon_commit_ctx() failure due to invalid user input, and
3. damon_commit_ctx() failure due to its internal allocation failures.
Hence, if the kdamond is stopped by the above three events, the values of
the status parameters can be stale. Users could show the stale values and
be confused. This is already bad, but the real consequence is worse.
DAMON_LRU_SORT avoids unnecessary damon_start() and damon_stop() calls
based on the 'enabled' parameter value. And the update of 'enabled'
parameter value depends on the damon_start() and damon_stop() call
results. Hence, once the kdamond has stopped by the unintentional events,
the user cannot restart the kdamond before the system reboot. For
example, the issue can be reproduced via below steps.
The problem will only rarely happen in real and common setups for the
following reasons. The allocation failures are unlikely in such setups
since those allocations are arguably too small to fail. Also sane users
on real production environments may not commit wrong input parameters.
But once it happens, the consequence is quite bad. And the bug is a bug.
The issue stems from the fact that there are multiple events that can
change the status, and following all the events is challenging.
Dynamically detect and use the fresh status for the parameters when those
are requested.
Link: https://lore.kernel.org/20260419161003.79176-3-sj@kernel.org Fixes: 40e983cca927 ("mm/damon: introduce DAMON-based LRU-lists Sorting") Co-developed-by: Liew Rui Yan <aethernet65535@gmail.com> Signed-off-by: Liew Rui Yan <aethernet65535@gmail.com> Signed-off-by: SeongJae Park <sj@kernel.org> Cc: <stable@vger.kernel.org> # 6.0.x Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
SeongJae Park [Sun, 19 Apr 2026 16:10:00 +0000 (09:10 -0700)]
mm/damon/reclaim: detect and use fresh enabled and kdamond_pid values
Patch series "mm/damon/modules: detect and use fresh status", v3.
DAMON modules including DAMON_RECLAIM, DAMON_LRU_SORT and DAMON_STAT
commonly expose the kdamond running status via their parameters. Under
certain scenarios including wrong user inputs and memory allocation
failures, those parameter values can be stale. It can confuse users. For
DAMON_RECLAIM and DAMON_LRU_SORT, it even makes the kdamond unable to be
restarted before the system reboot.
The problem comes from the fact that there are multiple events for the
status changes and it is difficult to follow up all the scenarios. Fix
the issue by detecting and using the status on demand, instead of using a
cached status that is difficult to be updated.
Patches 1-3 fix the bugs in DAMON_RECLAIM, DAMON_LRU_SORT and DAMON_STAT
in the order.
This patch (of 3):
DAMON_RECLAIM updates 'enabled' and 'kdamond_pid' parameter values, which
represents the running status of its kdamond, when the user explicitly
requests start/stop of the kdamond. The kdamond can, however, be stopped
in events other than the explicit user request in the following three
events.
1. ctx->regions_score_histogram allocation failure at beginning of the
execution,
2. damon_commit_ctx() failure due to invalid user input, and
3. damon_commit_ctx() failure due to its internal allocation failures.
Hence, if the kdamond is stopped by the above three events, the values of
the status parameters can be stale. Users could show the stale values and
be confused. This is already bad, but the real consequence is worse.
DAMON_RECLAIM avoids unnecessary damon_start() and damon_stop() calls
based on the 'enabled' parameter value. And the update of 'enabled'
parameter value depends on the damon_start() and damon_stop() call
results. Hence, once the kdamond has stopped by the unintentional events,
the user cannot restart the kdamond before the system reboot. For
example, the issue can be reproduced via below steps.
The problem will only rarely happen in real and common setups for the
following reasons. The allocation failures are unlikely in such setups
since those allocations are arguably too small to fail. Also sane users
on real production environments may not commit wrong input parameters.
But once it happens, the consequence is quite bad. And the bug is a bug.
The issue stems from the fact that there are multiple events that can
change the status, and following all the events is challenging.
Dynamically detect and use the fresh status for the parameters when those
are requested.
Mark Brown [Thu, 16 Apr 2026 18:40:56 +0000 (19:40 +0100)]
selftests/mm: specify requirement for PROC_MEM_ALWAYS_FORCE=y
Several of the mm selftests made use of /proc/pid/mem as part of their
operation but we do not specify this in the config fragment for them, at
least mkdirty and ksm_functional_tests have this requirement.
This has been working fine in practice since PROC_MEM_ALWAYS_FORCE was the
default setting but commit 599bbba5a36f ("proc: make PROC_MEM_FORCE_PTRACE
the Kconfig default") that is no longer the case, meaning that tests run
on kernels built based on defconfigs have started having the new more
restrictive default and failing. Add PROC_MEM_ALWAYS_FORCE to the config
fragment for the mm selftests.
Thanks to Aishwarya TCV for spotting the issue and identifying the commit
that introduced it.
Link: https://lore.kernel.org/20260416-selftests-mm-proc-mem-always-force-v1-1-3f5865153c67@kernel.org Fixes: 599bbba5a36f ("proc: make PROC_MEM_FORCE_PTRACE the Kconfig default") Signed-off-by: Mark Brown <broonie@kernel.org> Reported-by: Aishwarya TCV <aishwarya.tcv@arm.com> Reviewed-by: Mike Rapoport (Microsoft) <rppt@kernel.org> Acked-by: David Hildenbrand (Arm) <david@kernel.org> Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com> Reviewed-by: Dev Jain <dev.jain@arm.com> Cc: Lorenzo Stoakes <ljs@kernel.org> Cc: Michal Hocko <mhocko@suse.com> Cc: Mike Rapoport <rppt@kernel.org> Cc: Shuah Khan <shuah@kernel.org> Cc: Suren Baghdasaryan <surenb@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
SeongJae Park [Thu, 23 Apr 2026 15:02:52 +0000 (08:02 -0700)]
mm/damon/sysfs-schemes: protect path kfree() with damon_sysfs_lock
damon_sysfs_quot_goal->path can be read and written by users, via DAMON
sysfs 'path' file. It can also be indirectly read, for the parameters
{on,off}line committing to DAMON. The reads for parameters committing are
protected by damon_sysfs_lock to avoid the sysfs files being destroyed
while any of the parameters are being read. But the user-driven direct
reads and writes are not protected by any lock, while the write is
deallocating the path-pointing buffer. As a result, the readers could
read the already freed buffer (user-after-free). Note that the user-reads
don't race when the same open file is used by the writer, due to kernfs's
open file locking. Nonetheless, doing the reads and writes with separate
open files would be common. Fix it by protecting both the user-direct
reads and writes with damon_sysfs_lock.
SeongJae Park [Thu, 23 Apr 2026 15:02:51 +0000 (08:02 -0700)]
mm/damon/sysfs-schemes: protect memcg_path kfree() with damon_sysfs_lock
Patch series "mm/damon/sysfs-schemes: fix use-after-free for [memcg_]path".
Reads of 'memcg_path' and 'path' files in DAMON sysfs interface could race
with their writes, results in use-after-free. Fix those.
This patch (of 2):
damon_sysfs_scheme_filter->mmecg_path can be read and written by users,
via DAMON sysfs memcg_path file. It can also be indirectly read, for the
parameters {on,off}line committing to DAMON. The reads for parameters
committing are protected by damon_sysfs_lock to avoid the sysfs files
being destroyed while any of the parameters are being read. But the
user-driven direct reads and writes are not protected by any lock, while
the write is deallocating the memcg_path-pointing buffer. As a result,
the readers could read the already freed buffer (user-after-free). Note
that the user-reads don't race when the same open file is used by the
writer, due to kernfs's open file locking. Nonetheless, doing the reads
and writes with separate open files would be common. Fix it by protecting
both the user-direct reads and writes with damon_sysfs_lock.
Li Wang [Thu, 23 Apr 2026 13:26:49 +0000 (21:26 +0800)]
MAINTAINERS: update Li Wang's email address
Update my email address as my work email account is no longer in use.
Also update .mailmap.
Link: https://lore.kernel.org/20260423132649.31126-1-li.wang@linux.dev Signed-off-by: Li Wang <li.wang@linux.dev> Reviewed-by: Cyril Hrubis <chrubis@suse.cz> Reviewed-by: Petr Vorel <pvorel@suse.cz> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Jakub Kacinski <kuba@kernel.org> Cc: Li Wang <liwang@redhat.com> Cc: Martin Kepplinger <martink@posteo.de> Cc: Shannon Nelson <sln@onemain.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
mm/hugetlb_cma: round up per_node before logging it
When the user requests a total hugetlb CMA size without per-node
specification, hugetlb_cma_reserve() computes per_node from
hugetlb_cma_size and the number of nodes that have memory
So the actually reserved per_node size is multiple of (PAGE_SIZE <<
order), but the logged per_node is not rounded up, so it may be smaller
than the actual reserved size.
For example, as the existing comment describes, if a 3 GB area is
requested on a machine with 4 NUMA nodes that have memory, 1 GB is
allocated on the first three nodes, but the printed log is
hugetlb_cma: reserve 3072 MiB, up to 768 MiB per node
Round per_node up to (PAGE_SIZE << order) before logging so that the
printed log always matches the actual reserved size. No functional change
to the actual reservation size, as the following case analysis shows
Pedro Falcato [Wed, 22 Apr 2026 12:37:26 +0000 (13:37 +0100)]
MAINTAINERS: fix regex pattern in CORE MM category
The pattern "include/linux/page[-_]*" matches every file that starts with
"page", because it's a regex and not a glob (so it has the meaning of
include/linux/page + match [-_] 0+ times).
Fix it up into a more regex-correct expression. Doing so reduces CC's
drastically in patches that touch pagemap.h (which is maintained as part
of PAGE CACHE).
As a side-effect, move linux/pageblock-flags.h explicitly under PAGE
ALLOCATOR.
Link: https://lore.kernel.org/linux-mm/20260422005608.342028-1-fmayle@google.com/ Link: https://lore.kernel.org/20260422123726.517220-1-pfalcato@suse.de Signed-off-by: Pedro Falcato <pfalcato@suse.de> Reviewed-by: Lorenzo Stoakes <ljs@kernel.org> Acked-by: David Hildenbrand (Arm) <david@kernel.org> Acked-by: Vlastimil Babka (SUSE) <vbabka@kernel.org> Reviewed-by: SeongJae Park <sj@kernel.org> Cc: Liam Howlett <liam.howlett@oracle.com> Cc: Michal Hocko <mhocko@suse.com> Cc: Mike Rapoport <rppt@kernel.org> Cc: Suren Baghdasaryan <surenb@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Lorenzo Stoakes [Tue, 21 Apr 2026 10:21:50 +0000 (11:21 +0100)]
mm/vma: do not try to unmap a VMA if mmap_prepare() invoked from mmap()
The mmap_prepare hook functionality includes the ability to invoke
mmap_prepare() from the mmap() hook of existing 'stacked' drivers, that is
ones which are capable of calling the mmap hooks of other drivers/file
systems (e.g. overlayfs, shm).
As part of the mmap_prepare action functionality, we deal with errors by
unmapping the VMA should one arise. This works in the usual mmap_prepare
case, as we invoke this action at the last moment, when the VMA is
established in the maple tree.
However, the mmap() hook passes a not-fully-established VMA pointer to the
caller (which is the motivation behind the mmap_prepare() work), which is
detached.
So attempting to unmap a VMA in this state will be problematic, with the
most obvious symptom being a warning in vma_mark_detached(), because the
VMA is already detached.
It's also unncessary - the mmap() handler will clean up the VMA on error.
So to fix this issue, this patch propagates whether or not an mmap action
is being completed via the compatibility layer or directly.
If the former, then we do not attempt VMA cleanup, if the latter, then we
do.
This patch also updates the userland VMA tests to reflect the change.
Link: https://lore.kernel.org/20260421102150.189982-1-ljs@kernel.org Fixes: ac0a3fc9c07d ("mm: add ability to take further action in vm_area_desc") Signed-off-by: Lorenzo Stoakes <ljs@kernel.org> Reported-by: syzbot+db390288d141a1dccf96@syzkaller.appspotmail.com Closes: https://lore.kernel.org/all/69e69734.050a0220.24bfd3.0027.GAE@google.com/ Cc: David Hildenbrand <david@kernel.org> Cc: Jann Horn <jannh@google.com> Cc: Liam Howlett <liam.howlett@oracle.com> Cc: Michal Hocko <mhocko@suse.com> Cc: Mike Rapoport <rppt@kernel.org> Cc: Pedro Falcato <pfalcato@suse.de> Cc: Suren Baghdasaryan <surenb@google.com> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Joanne Koong [Thu, 26 Mar 2026 23:46:29 +0000 (16:46 -0700)]
mm: start background writeback based on per-wb threshold for strictlimit BDIs
The proactive nr_dirty > gdtc->bg_thresh check in balance_dirty_pages()
only checks the global dirty threshold to start background writeback while
the writer is still free-running, but for strictlimit BDIs (eg fuse), the
per-wb dirty count can exceed the per-wb background threshold while the
global threshold is not yet exceeded, so background writeback for this
case never gets proactively started.
Add a per-wb threshold check for strictlimit BDIs so that background
writeback is started when wb_dirty exceeds wb_bg_thresh, which drains
dirty pages before the writer hits the throttle wall, matching the
proactive behavior that the global check provides for non-strictlimit
BDIs.
fio runs on fuse show about a 3-4% improvement in perf for buffered
writes:
fio --name=writeback_test --ioengine=psync --rw=write --bs=128k \
--size=2G --numjobs=4 --ramp_time=10 --runtime=20 \
--time_based --group_reporting=1 --direct=0
Link: https://lore.kernel.org/20260326234629.840938-2-joannelkoong@gmail.com Signed-off-by: Joanne Koong <joannelkoong@gmail.com> Reviewed-by: Jan Kara <jack@suse.cz> Cc: Matthew Wilcox (Oracle) <willy@infradead.org> Cc: Miklos Szeredi <miklos@szeredi.hu> Cc: Christoph Hellwig <hch@infradead.org> Cc: Johannes Weiner <hannes@cmpxchg.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Fix two error handling issues in kho_add_subtree(), where it doesn't
handle the error path correctly.
1. If fdt_setprop() fails after the subnode has been created, the
subnode is not removed. This leaves an incomplete node in the FDT
(missing "preserved-data" or "blob-size" properties).
2. The fdt_setprop() return value (an FDT error code) is stored
directly in err and returned to the caller, which expects -errno.
Fix both by storing fdt_setprop() results in fdt_err, jumping to a new
out_del_node label that removes the subnode on failure, and only setting
err = 0 on the success path, otherwise returning -ENOMEM (instead of
FDT_ERR_ errors that would come from fdt_setprop).
No user-visible changes. This patch fixes error handling in the KHO
(Kexec HandOver) subsystem, which is used to preserve data across kexec
reboots. The fix only affects a rare failure path during kexec
preparation — specifically when the kernel runs out of space in the
Flattened Device Tree buffer while registering preserved memory regions.
In the unlikely event that this error path was triggered, the old code
would leave a malformed node in the device tree and return an incorrect
error code to the calling subsystem, which could lead to confusing log
messages or incorrect recovery decisions. With this fix, the incomplete
node is properly cleaned up and the appropriate errno value is propagated,
this error code is not returned to the user.
liveupdate: fix return value on session allocation failure
When session allocation fails during deserialization, the global 'err'
variable was not updated before returning. This caused subsequent calls
to luo_session_deserialize() to incorrectly report success.
Ensure 'err' is set to the error code from PTR_ERR(session). This ensures
that an error is correctly returned to userspace when it attempts to open
/dev/liveupdate in the new kernel if deserialization failed.
Link: https://lore.kernel.org/20260415193738.515491-1-pasha.tatashin@soleen.com Signed-off-by: Pasha Tatashin <pasha.tatashin@soleen.com> Reviewed-by: Pratyush Yadav (Google) <pratyush@kernel.org> Cc: David Matlack <dmatlack@google.com> Cc: Mike Rapoport <rppt@kernel.org> Cc: Samiullah Khawaja <skhawaja@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Marco Elver [Mon, 20 Apr 2026 11:47:26 +0000 (13:47 +0200)]
vmalloc: fix buffer overflow in vrealloc_node_align()
Commit 4c5d3365882d ("mm/vmalloc: allow to set node and align in
vrealloc") added the ability to force a new allocation if the current
pointer is on the wrong NUMA node, or if an alignment constraint is not
met, even if the user is shrinking the allocation.
On this path (need_realloc), the code allocates a new object of 'size'
bytes and then memcpy()s 'old_size' bytes into it. If the request is to
shrink the object (size < old_size), this results in an out-of-bounds
write on the new buffer.
Fix this by bounding the copy length by the new allocation size.
Link: https://lore.kernel.org/20260420114805.3572606-2-elver@google.com Fixes: 4c5d3365882d ("mm/vmalloc: allow to set node and align in vrealloc") Signed-off-by: Marco Elver <elver@google.com> Reported-by: Harry Yoo (Oracle) <harry@kernel.org> Reviewed-by: Uladzislau Rezki (Sony) <urezki@gmail.com> Acked-by: Vlastimil Babka (SUSE) <vbabka@kernel.org> Reviewed-by: Harry Yoo (Oracle) <harry@kernel.org> Cc: <stable@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
pmdomain: mediatek: fix use-after-free in scpsys_get_bus_protection_legacy()
In scpsys_get_bus_protection_legacy(), of_find_node_with_property()
returns a device node with its reference count incremented. The function
then calls of_node_put(node) before checking whether
syscon_regmap_lookup_by_phandle() returns an error. If an error occurs,
dev_err_probe() dereferences the node pointer to print diagnostic
information, but the node memory may have already been freed due to the
earlier of_node_put(), leading to a use-after-free vulnerability.
Fix this by moving the of_node_put() call after the error check, ensuring
the node is still valid when accessed in the error path.
ACPI: arm64: cpuidle: Tolerate platforms with no deep PSCI idle states
Commit cac173bea57d ("ACPI: processor: idle: Rework the handling of
acpi_processor_ffh_lpi_probe()") moved the acpi_processor_ffh_lpi_probe()
call from acpi_processor_setup_cpuidle_dev(), where its return value was
ignored, to acpi_processor_get_power_info(), where it is now treated as
a hard failure. As a result, platforms where psci_acpi_cpu_init_idle()
returned -ENODEV stopped registering any cpuidle states, forcing CPUs to
busy-poll when idle.
On NVIDIA Grace (aarch64) systems with PSCIv1.1, pr->power.count is 1
(only WFI, no deep PSCI states beyond it), so the previous
"count = pr->power.count - 1; if (count <= 0) return -ENODEV;" check
returned -ENODEV for all 72 CPUs and disabled cpuidle entirely.
The lpi_states count is already validated in acpi_processor_get_lpi_info(),
so the check here is redundant. Simplify the loop to iterate over
lpi_states[1..power.count). When only WFI is present, the loop body
simply does not execute and the function returns 0, which is the correct
outcome: there is nothing to validate for FFH and no error to report.
Suggested-by: Huisong Li <lihuisong@huawei.com> Cc: stable@vger.kernel.org Fixes: cac173bea57d ("ACPI: processor: idle: Rework the handling of acpi_processor_ffh_lpi_probe()") Signed-off-by: Breno Leitao <leitao@debian.org> Reviewed-by: Sudeep Holla <sudeep.holla@kernel.org> Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
ALSA: caiaq: fix usb_dev refcount leak on probe failure
create_card() takes a reference on the USB device with usb_get_dev()
and stores the matching usb_put_dev() in card_free(), which is
installed as the snd_card's ->private_free destructor.
However, ->private_free is only assigned near the end of init_card(),
after several failure points (usb_set_interface(), EP type checks,
usb_submit_urb(), the EP1_CMD_GET_DEVICE_INFO exchange, and its
timeout). When any of those fail, init_card() returns an error to
snd_probe(), which calls snd_card_free(card). Because ->private_free
is still NULL, card_free() never runs, the usb_get_dev() reference
is not dropped, and the struct usb_device leaks along with its
descriptor allocations and device_private.
syzbot reproduces this with a malformed UAC3 device whose only valid
altsetting is 0; init_card()'s usb_set_interface(usb_dev, 0, 1) call
fails with -EIO and triggers the leak.
Move the ->private_free assignment into create_card(), immediately
after usb_get_dev(), so that every error path reaching snd_card_free()
balances the reference. card_free()'s callees (snd_usb_caiaq_input_free,
free_urbs, kfree) already tolerate the partially-initialized state
because the chip private area is zero-initialized by snd_card_new().
Fixes: 80bb50e2d459 ("ALSA: caiaq: take a reference on the USB device in create_card()") Reported-by: syzbot+2afd7e71155c7e241560@syzkaller.appspotmail.com Closes: https://syzkaller.appspot.com/bug?extid=2afd7e71155c7e241560 Tested-by: syzbot+2afd7e71155c7e241560@syzkaller.appspotmail.com Cc: stable@vger.kernel.org Signed-off-by: Deepanshu Kartikey <kartikey406@gmail.com> Link: https://patch.msgid.link/20260426001934.70813-1-kartikey406@gmail.com Signed-off-by: Takashi Iwai <tiwai@suse.de>
projects / thirdparty / kernel / linux.git / log
