The clean up in idxd_setup_wqs() has had a couple bugs because the error
handling is a bit subtle. It's simpler to just re-write it in a cleaner
way. The issues here are:
1) If "idxd->max_wqs" is <= 0 then we call put_device(conf_dev) when
"conf_dev" hasn't been initialized.
2) If kzalloc_node() fails then again "conf_dev" is invalid. It's
either uninitialized or it points to the "conf_dev" from the
previous iteration so it leads to a double free.
It's better to free partial loop iterations within the loop and then
the unwinding at the end can handle whole loop iterations. I also
renamed the labels to describe what the goto does and not where the goto
was located.
Fixes: 3fd2f4bc01 ("dmaengine: idxd: fix memory leak in error handling path of idxd_setup_wqs")
Reported-by: Colin Ian King <colin.i.king@gmail.com>
Closes: https://lore.kernel.org/all/20250811095836.1642093-1-colin.i.king@gmail.com/
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/aJnJW3iYTDDCj9sk@stanley.mountain
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Replace memset(0) followed by strscpy() with strscpy_pad() to improve
idxd_load_iaa_device_defaults(). This avoids zeroing the memory before
copying the strings and ensures the destination buffers are only written
to once, simplifying the code and improving efficiency.
Signed-off-by: Thorsten Blum <thorsten.blum@linux.dev>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Link: https://lore.kernel.org/r/20250810225858.2953-2-thorsten.blum@linux.dev
Signed-off-by: Vinod Koul <vkoul@kernel.org>
A recent refactor introduced a misplaced put_device() call, resulting in a
reference count underflow during module unload.
There is no need to add additional put_device() calls for idxd groups,
engines, or workqueues. Although the commit claims: "Note, this also
fixes the missing put_device() for idxd groups, engines, and wqs."
It appears no such omission actually existed. The required cleanup is
already handled by the call chain:
idxd_unregister_devices() -> device_unregister() -> put_device()
Extend idxd_cleanup() to handle the remaining necessary cleanup and
remove idxd_cleanup_internals(), which duplicates deallocation logic
for idxd, engines, groups, and workqueues. Memory management is also
properly handled through the Linux device model.
Fixes: a409e919ca ("dmaengine: idxd: Refactor remove call with idxd_cleanup() helper")
Signed-off-by: Yi Sun <yi.sun@intel.com>
Tested-by: Shuai Xue <xueshuai@linux.alibaba.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Link: https://lore.kernel.org/r/20250729150313.1934101-3-yi.sun@intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The call to idxd_free() introduces a duplicate put_device() leading to a
reference count underflow:
refcount_t: underflow; use-after-free.
WARNING: CPU: 15 PID: 4428 at lib/refcount.c:28 refcount_warn_saturate+0xbe/0x110
...
Call Trace:
<TASK>
idxd_remove+0xe4/0x120 [idxd]
pci_device_remove+0x3f/0xb0
device_release_driver_internal+0x197/0x200
driver_detach+0x48/0x90
bus_remove_driver+0x74/0xf0
pci_unregister_driver+0x2e/0xb0
idxd_exit_module+0x34/0x7a0 [idxd]
__do_sys_delete_module.constprop.0+0x183/0x280
do_syscall_64+0x54/0xd70
entry_SYSCALL_64_after_hwframe+0x76/0x7e
The idxd_unregister_devices() which is invoked at the very beginning of
idxd_remove(), already takes care of the necessary put_device() through the
following call path:
idxd_unregister_devices() -> device_unregister() -> put_device()
In addition, when CONFIG_DEBUG_KOBJECT_RELEASE is enabled, put_device() may
trigger asynchronous cleanup via schedule_delayed_work(). If idxd_free() is
called immediately after, it can result in a use-after-free.
Remove the improper idxd_free() to avoid both the refcount underflow and
potential memory corruption during module unload.
Fixes: d5449ff1b0 ("dmaengine: idxd: Add missing idxd cleanup to fix memory leak in remove call")
Signed-off-by: Yi Sun <yi.sun@intel.com>
Tested-by: Shuai Xue <xueshuai@linux.alibaba.com>
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Link: https://lore.kernel.org/r/20250729150313.1934101-2-yi.sun@intel.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add the undocumented NVIDIA APBDMA compatibles and "clock-names" which
are already in use. There doesn't appear to be any per compatible
differences.
Signed-off-by: Rob Herring (Arm) <robh@kernel.org>
Link: https://lore.kernel.org/r/20250812203308.727731-1-robh@kernel.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add shutdown callback to ensure that DMA operations are properly stopped
and resources are released during system shutdown or kexec operations.
Fix incorrect PM state handling in the remove function that was causing
clock disable warnings during the shutdown operations, which was not
implemented earlier. The original logic used pm_runtime_enabled() check
after calling the pm_runtime_disable(), would always evaluate to true
after the disable call, which leads to unconditionally calling the
runtime_suspend regardless of the device's actual power state.
During shutdown, the device may already be suspended with clock disabled
from the autosuspend timer, causing the clock framework to warn about
the double-disable attempt. The pm_runtime_active() function checks the
actual device power state rather than the PM subsystem's enabled/disabled
status. ensuring the runtime_suspend is only called when the device is in
active power state. This prevents clock warnings during shutdown while
maintaining proper cleanup during normal remove operations.
Signed-off-by: Abin Joseph <abin.joseph@amd.com>
Reviewed-by: Radhey Shyam Pandey <radhey.shyam.pandey@amd.com>
Link: https://lore.kernel.org/r/20250722070255.28944-1-abin.joseph@amd.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The Radxa Cubie A5E has a 3-color LED. The green and blue LEDs are wired
to GPIO pins on the SoC, and the green one is lit by default to serve as
a power indicator. The red LED is wired to the M.2 slot.
Add device nodes for the green and blue LEDs.
A default "heartbeat" trigger is set for the green power LED, though in
practice it might be better if it were inverted, i.e. lit most of the
time.
Acked-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Link: https://patch.msgid.link/20250812175927.2199219-1-wens@kernel.org
Signed-off-by: Chen-Yu Tsai <wens@csie.org>
Use core driver model helper dev_err_probe() defined at
drivers/base/core.c in driver probe path to propagate errors.
standardize and improve the code of deferred probe error handling
using this helper for ingenic usb phy driver.
Inspired by,
commit a787e5400a ("driver core: add device probe log helper")
Signed-off-by: Akhilesh Patil <akhilesh@ee.iitb.ac.in>
Cc: Andrzej Hajda <a.hajda@samsung.com>
Link: https://lore.kernel.org/r/aIIMW971BYsIk4As@bhairav-test.ee.iitb.ac.in
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add compatible string for the supported eDP PHY on QCS8300 platform.
QCS8300 have the same eDP PHY with SA8775P.
Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Yongxing Mou <quic_yongmou@quicinc.com>
Link: https://lore.kernel.org/r/20250730072725.1433360-1-quic_yongmou@quicinc.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Relocate the debug print in rcar_gen3_enable_vbus_ctrl() to appear after
the `val` variable is assigned and updated based on the VBUS state. This
ensures that the debug log reflects the actual register value being
written, improving debugging accuracy.
Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://lore.kernel.org/r/20250808215209.3692744-6-prabhakar.mahadev-lad.rj@bp.renesas.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Add initial support for the Renesas RZ/T2H SoC to the R-Car Gen3 USB2 PHY
driver. The RZ/T2H SoC requires configuration of additional
hardware-specific bits for proper VBUS level control and OTG operation.
Introduce the `vblvl_ctrl` flag in the SoC-specific driver data to enable
handling of VBUS level selection logic using `VBCTRL.VBLVL` bits. This is
required for managing the VBUS status detection and drive logic based on
SoC-specific needs.
Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://lore.kernel.org/r/20250808215209.3692744-5-prabhakar.mahadev-lad.rj@bp.renesas.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Update the PHY driver to support SoC-specific OBINT enable bits by
introducing the `obint_enable_bits` field in the `rcar_gen3_phy_drv_data`
structure. This allows each SoC to specify bits required.
Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://lore.kernel.org/r/20250808215209.3692744-4-prabhakar.mahadev-lad.rj@bp.renesas.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Store the SoC-specific driver data pointer (struct rcar_gen3_phy_drv_data)
directly in struct rcar_gen3_chan instead of copying individual flags into
separate channel members.
Obtain the drvdata with of_device_get_match_data() in probe and assign it
to channel->phy_data. Update all call sites to reference
`channel->phy_data->*` for SoC-specific behaviour (for example no_adp_ctrl
and utmi_ctrl). Remove the redundant soc_no_adp_ctrl and utmi_ctrl fields
from struct rcar_gen3_chan.
This simplifies the probe path, reduces duplication, and makes it easier
to extend the driver with additional platform-specific fields in the
future.
Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Link: https://lore.kernel.org/r/20250808215209.3692744-3-prabhakar.mahadev-lad.rj@bp.renesas.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Document the USB2 PHY controller for the Renesas RZ/T2H (r9a09g077) and
RZ/N2H (r9a09g087) SoCs. These SoCs share the same PHY block, which is
similar to the one on RZ/G2L but differs in clocks, resets, and register
bits. To account for these differences, a new compatible string
`renesas,usb2-phy-r9a09g077` is introduced.
The RZ/N2H SoC uses the same PHY as RZ/T2H, so it reuses the RZ/T2H
compatible string as a fallback.
Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Acked-by: Conor Dooley <conor.dooley@microchip.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Link: https://lore.kernel.org/r/20250808215209.3692744-2-prabhakar.mahadev-lad.rj@bp.renesas.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Register a typec mux in order to change the PHY mode on the Type-C
mux events depending on the mode and the svid when in Altmode setup.
The DisplayPort phy should be left enabled if is still powered on
by the DRM DisplayPort controller, so bail out until the DisplayPort
PHY is not powered off.
The Type-C Mode/SVID only changes on plug/unplug, and USB SAFE states
will be set in between of USB-Only, Combo and DisplayPort Only so
this will leave enough time to the DRM DisplayPort controller to
turn of the DisplayPort PHY.
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
[konrad: renaming, rewording, bug fixes]
Signed-off-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on Lenovo Thinkpad T14S
Link: https://lore.kernel.org/r/20250807-topic-4ln_dp_respin-v4-5-43272d6eca92@oss.qualcomm.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Introduce an enum for the QMP Combo PHY modes, use it in the
QMP commmon phy init function and default to COMBO mode.
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
[konrad: some renaming and rewording]
Signed-off-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on Lenovo Thinkpad T14S
Link: https://lore.kernel.org/r/20250807-topic-4ln_dp_respin-v4-4-43272d6eca92@oss.qualcomm.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Switching the PHY Mode requires the DisplayPort PHY to be powered off,
keep track of the DisplayPort phy power state.
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Signed-off-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on Lenovo Thinkpad T14S
Link: https://lore.kernel.org/r/20250807-topic-4ln_dp_respin-v4-3-43272d6eca92@oss.qualcomm.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
There are a numbers of ""modes"" involved: USB mode, Type-C mode (with
its altmodes), phy_mode and QMP_PHY mode (DP/combo/USB/off).
Rename the generic sounding 'mode' to 'phy_mode' to hopefully make
the code easier to follow.
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Signed-off-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on Lenovo Thinkpad T14S
Link: https://lore.kernel.org/r/20250807-topic-4ln_dp_respin-v4-2-43272d6eca92@oss.qualcomm.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
The QMP USB3/DP Combo PHY can work in 3 modes:
- DisplayPort Only
- USB3 Only
- USB3 + DisplayPort Combo mode
In order to switch between those modes, the PHY needs to receive
Type-C events, allow marking to the phy with the mode-switch
property in order to allow the PHY to Type-C events.
Reference usb-switch.yaml as a simpler way to allow the mode-switch
property instead of duplicating the property definition.
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Reviewed-by: Krzysztof Kozlowski <krzk@kernel.org>
Signed-off-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on Lenovo Thinkpad T14S
Link: https://lore.kernel.org/r/20250807-topic-4ln_dp_respin-v4-1-43272d6eca92@oss.qualcomm.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Replace comma between expressions with semicolons.
Using a ',' in place of a ';' can have unintended side effects.
Although that is not the case here, it is seems best to use ';'
unless ',' is intended.
Found by inspection.
No functional change intended.
Compile tested only.
Signed-off-by: Chen Ni <nichen@iscas.ac.cn>
Link: https://lore.kernel.org/r/20250814013943.2905307-1-nichen@iscas.ac.cn
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Replace the magic number '0xff' with CLK_CTL_DIV_MASK macro for finding
actual clock in renesas_sdhi_set_clock().
Signed-off-by: Biju Das <biju.das.jz@bp.renesas.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Link: https://lore.kernel.org/r/20250820104808.94562-1-biju.das.jz@bp.renesas.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Clang warns (or errors with CONFIG_WERROR=y):
drivers/mmc/host/sdhci-cadence.c:297:9: error: variable 'hrs37_mode' is uninitialized when used here [-Werror,-Wuninitialized]
297 | writel(hrs37_mode, hrs37_reg);
| ^~~~~~~~~~
drivers/mmc/host/sdhci-cadence.c:291:16: note: initialize the variable 'hrs37_mode' to silence this warning
291 | u32 hrs37_mode;
| ^
| = 0
A previous revision assigned SDHCI_CDNS_HRS37_MODE_MMC_HS200 to
hrs37_mode in a switch statement but the final revision moved to a
simple if statement. Pass that as the value to writel() and
remove hrs37_mode, clearing up the warning.
Fixes: 60613a8b9b ("mmc: sdhci-cadence: implement multi-block read gap tuning")
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Tested-by: Benoît Monin <benoit.monin@bootlin.com>
Link: https://lore.kernel.org/r/20250819-mmc-sdhci-cadence-fix-uninit-hrs37_mode-v1-1-94aa2d0c438a@kernel.org
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
As part of
38a4968b31 ("x86/percpu/64: Remove INIT_PER_CPU macros"),
the only use of the config option X86_64_SMP in the kernel tree is removed. As
this config option X86_64_SMP is just equivalent to X86_64 && SMP, the source
code in the tree just uses that expression in the few places where needed. Note
further that this option cannot be explicitly enabled or disabled when
configuring the kernel build configuration.
Drop this needless and unused config option. No functional change.
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@redhat.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/20250723071211.622802-1-lukas.bulwahn@redhat.com
irq_domain_create_simple() takes fwnode as the first argument. It can be
extracted from the struct device using dev_fwnode() helper instead of using
of_node with of_fwnode_handle().
So use the dev_fwnode() helper.
Signed-off-by: Jiri Slaby (SUSE) <jirislaby@kernel.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Yazen Ghannam <yazen.ghannam@amd.com>
Acked-by: Dinh Nguyen <dinguyen@kernel.org>
Link: https://lore.kernel.org/20250723062631.1830757-1-jirislaby@kernel.org
So far we set a number of expected Allwinner USB PHY instances for any
given compatible string, and would fail if we do not find every PHY
described properly in the DT (missing reset/PMU/clocks).
This is somewhat redundant, as the DT only describes the resources for
the implemented PHYs, but goes in line with being strict about firmware
descriptions, and rather fixing things in the driver code, based on the
compatible string.
However this causes issues when we make a mistake, like we did recently
on the A523: there are actually three USB PHYs, not two, as we assumed.
Changing the number in the driver and the compatible string would cause
all kinds of compatibility issues, both with older and newer DTs.
To avoid problems with newer kernels and older or newer DTs, we can change
the driver code to deduce the number of PHY instances from what's
described in the DT. This has the added advantage of not requiring new
compatible strings for new SoCs when just the number of PHYs change, which
already happened and might occur again in the future.
Drop the num_phys member from the config struct, and remove the fixed
number of PHYs from each SoC's config description. Then enumerate the
usb<x>_reset properties for all of the maximum four PHY instances, and
just stop once we cannot find such a property anymore.
The binding describes the reset property as mandatory for each PHY
instance, and each DT in the kernel tree matches exactly the num_phys
value in the current driver code, so we can rely on that.
Apart from being more future proof, this will solve the A523 mishap:
Older DTs would just describe two PHYs, whereas newer ones would feature
all three of them. In any case we would get a valid number, matching the
other nodes in the DT. Older kernels would always enumerate two PHYs,
which at least does not cause any regressions.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Link: https://lore.kernel.org/r/20250819001522.13011-1-andre.przywara@arm.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Extend the "Disable NULL query for Anyhit Shader" tuning to Xe2
(graphics version 2000+) platforms, in addition to Xe3.
This sets the DIS_NULL_QUERY bit in RT_CTRL to disable null query
for Anyhit shaders on both Xe2 and Xe3.
This is a change in behavior that can regress a userspace not
prepared for it. However it's not feasible to change dynamically
the option per client or per exec queue via an opt-in flag. Mesa
is already prepared for that and it got propagated to their
stable versions. Even if it was possible, at this point adding a
flag would mean mesa would also need to propagate such a flag to
their stable versions, otherwise the previous fix would not be
used.
Signed-off-by: Nitin Gote <nitin.r.gote@intel.com>
Acked-by: Tapani Pälli <tapani.palli@intel.com>
Link: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35044
Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com>
Link: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/35044
Link: https://lore.kernel.org/r/20250819061151.1272622-1-nitin.r.gote@intel.com
Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
Set the `support_hdcp` bit to enable the connector to register content
protection during initialization.
Signed-off-by: Hsin-Yi Wang <hsinyi@chromium.org>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Fei Shao <fshao@chromium.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://lore.kernel.org/r/20250812082135.3351172-3-fshao@chromium.org
Some bridges can update HDCP status based on userspace requests if they
support HDCP.
The HDCP property is created after connector initialization and before
registration, just like other connector properties.
Add the content protection property to the connector if a bridge
supports HDCP.
Signed-off-by: Hsin-Yi Wang <hsinyi@chromium.org>
Reviewed-by: Sean Paul <seanpaul@chromium.org>
Signed-off-by: Fei Shao <fshao@chromium.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://lore.kernel.org/r/20250812082135.3351172-2-fshao@chromium.org
Rename is_signed_type() to is_signed_var() to avoid colliding with a macro
of the same name defined by tools' linux/overflow.h. This fixes warnings
(and presumably potential test failures) in tests that utilize the
selftests harness and happen to (indirectly) include overflow.h.
In file included from tools/include/linux/bits.h:34,
from tools/include/linux/bitops.h:14,
from tools/include/linux/hashtable.h:13,
from include/kvm_util.h:11,
from x86/userspace_msr_exit_test.c:11:
tools/include/linux/overflow.h:31:9: error: "is_signed_type" redefined [-Werror]
31 | #define is_signed_type(type) (((type)(-1)) < (type)1)
| ^~~~~~~~~~~~~~
In file included from include/kvm_test_harness.h:11,
from x86/userspace_msr_exit_test.c:9:
../kselftest_harness.h:754:9: note: this is the location of the previous definition
754 | #define is_signed_type(var) (!!(((__typeof__(var))(-1)) < (__typeof__(var))1))
| ^~~~~~~~~~~~~~
Use a separate definition, at least for now, as many selftests build
without tools/include in their include path.
Fixes: fc92099902 ("tools headers: Synchronize linux/bits.h with the kernel sources")
Cc: Vincent Mailhol <mailhol.vincent@wanadoo.fr>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Mark Brown <broonie@kernel.org>
Link: https://lore.kernel.org/r/20250624231930.583689-1-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
With pch_pic device, its register is based on MMIO address space,
different access size 1/2/4/8 is supported. And base address should
be naturally aligned with its access size, here add alignment check
in its register access emulation function.
Cc: stable@vger.kernel.org
Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
Since using kvm_get_vcpu() may fail to retrieve the vCPU context,
kvm_get_vcpu_by_id() should be used instead.
Fixes: 8e3054261b ("LoongArch: KVM: Add IPI user mode read and write function")
Fixes: 3956a52bc0 ("LoongArch: KVM: Add EIOINTC read and write functions")
Reviewed-by: Yanteng Si <siyanteng@cqsoftware.com.cm>
Signed-off-by: Song Gao <gaosong@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
Add the flag KVM_LARCH_LBT checking in function kvm_own_lbt(), so that
it can be called safely rather than duplicated enabling again.
Cc: stable@vger.kernel.org
Signed-off-by: Bibo Mao <maobibo@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
Fprobe event accepts wildcards for the target functions, but unless user
specifies its event name, it makes an event with the wildcards.
/sys/kernel/tracing # echo 'f mutex*' >> dynamic_events
/sys/kernel/tracing # cat dynamic_events
f:fprobes/mutex*__entry mutex*
/sys/kernel/tracing # ls events/fprobes/
enable filter mutex*__entry
To fix this, replace the wildcard ('*') with an underscore.
Link: https://lore.kernel.org/all/175535345114.282990.12294108192847938710.stgit@devnote2/
Fixes: 334e5519c3 ("tracing/probes: Add fprobe events for tracing function entry and exit.")
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Cc: stable@vger.kernel.org
GCC inlining heuristics prevent code growth due to inlining into cold
paths. This causes GCC to emit a partially specialized version of
__flogr for non-constant input for all occurrences on cold paths.
This happens since the overhead seen during inlining includes setting
up a union register_pair, calling flogr, and extracting and casting
the result. This overhead is not removed until the function is
lowered into RTL. But this happens after inlining.
For -ftrivial-var-auto-init=zero builds, an additional initialization
of the union register_pair adds another statement to be inlinined.
This is unneeded since the even register is initialized anyway and the
odd register is not an input register. It is only marked as such
since the whole pair has to be marked as a read/write output register.
Mark the union register_pair as uninitialized to get rid of this
statement. This, however, does not change the code since the
initialization happens when part of the register pair is written.
Nevertheless, GCC function size approximation during inlining is
reduced by one statement.
Force inlining of flogr and also flatten some other functions that
should be leaf functions but are called in cold context, like, e.g.,
__init functions.
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Juergen Christ <jchrist@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
Use a simpler algorithm to calculate the result of ffs() and fls64().
This generates slightly better code and increases readability.
Kernel image size is reduced by ~3kb (gcc 15.1.0 + defconfig).
Suggested-by: Nina Schoetterl-Glausch <nsg@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
To improve readability and prepare for future extensions, move the
memory hotplug code to a separate file. In addtion to it, add required
headers in sclp_mem.c and remove unnecessary headers in sclp_cmd.c
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Sumanth Korikkar <sumanthk@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
To improve readability, refactor sclp_cmd.c:
* Move defines and structures to the beginning.
* Reverse x-mas tree usage.
* Remove spaces after casting to eliminate checkpatch warnings.
* Remove unnecessary comments.
* Reframe certain comments.
* Convert all unsigned long long to unsigned long since unsigned long
long is a leftover from the 31/32-bit era.
* Use correct format specifiers.
* Add braces to for loops with bodies containing more than one line.
* Sort header files and remove unnecessary ones.
* Use __packed and __aligned instead of __attribute((packed,
aligned(8))) in structures.
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Sumanth Korikkar <sumanthk@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
Since the identity mapping is pinned to address zero the lowcore is always
also mapped to address zero, this happens regardless of the relocate_lowcore
command line option. If the option is specified the lowcore is mapped
twice, instead of only once.
This means that NULL pointer accesses will succeed instead of causing an
exception (low address protection still applies, but covers only parts).
To fix this never map the first two pages of physical memory with the
identity mapping.
Fixes: 32db401965 ("s390/mm: Pin identity mapping base to zero")
Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
Tracing code called by the SCLP interrupt handler contains early exits
if the SCCB address associated with an interrupt is NULL. This check is
performed after physical to virtual address translation.
If the kernel identity mapping does not start at address zero, the
resulting virtual address is never zero, so that the NULL checks won't
work. Subsequently this may result in incorrect accesses to the first
page of the identity mapping.
Fix this by introducing a function that handles the NULL case before
address translation.
Fixes: ada1da31ce ("s390/sclp: sort out physical vs virtual pointers usage")
Cc: stable@vger.kernel.org
Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Peter Oberparleiter <oberpar@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
Similar to powerpc set HZ to 1000. See also commit a206d23340
("powerpc/defconfigs: Set HZ=1000 on ppc64 and powernv defconfigs").
Besides other this will reduce the latency seen with synchronize_rcu().
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
Production silicon for CS36L63 has some small differences compared to
pre-production silicon. Remove soundwire clock workaround as no
longer necessary. We don't want to do tricks with low-level clocking
controls if we don't need to.
Fixes: 978858791c ("ASoC: cs35l56: Add initial support for CS35L63 for I2C and SoundWire")
Signed-off-by: Stefan Binding <sbinding@opensource.cirrus.com>
Link: https://patch.msgid.link/20250820142209.127575-4-sbinding@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
CS35L63 uses different algorithm IDs from CS35L56.
Add a new mechanism to handle different alg IDs between parts in the
CS35L56 driver.
Fixes: 978858791c ("ASoC: cs35l56: Add initial support for CS35L63 for I2C and SoundWire")
Signed-off-by: Richard Fitzgerald <rf@opensource.cirrus.com>
Signed-off-by: Stefan Binding <sbinding@opensource.cirrus.com>
Link: https://patch.msgid.link/20250820142209.127575-3-sbinding@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Production silicon for CS36L63 has some small differences compared to
pre-production silicon. Update firmware addresses, which are different.
No product was ever released with pre-production silicon so there is no
need for the driver to include support for it.
Fixes: 978858791c ("ASoC: cs35l56: Add initial support for CS35L63 for I2C and SoundWire")
Signed-off-by: Stefan Binding <sbinding@opensource.cirrus.com>
Link: https://patch.msgid.link/20250820142209.127575-2-sbinding@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Commit 57fbad15c2 ("stackleak: Rename STACKLEAK to KSTACK_ERASE")
misses the stackframe.h part for LoongArch, so fix it.
Fixes: 57fbad15c2 ("stackleak: Rename STACKLEAK to KSTACK_ERASE")
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
Now if preemption happens between protected_save_fpu_context() and
protected_save_lbt_context(), FTOP context is lost. Because FTOP is
saved by protected_save_lbt_context() but protected_save_fpu_context()
disables TM before that. So save LBT before FPU in setup_sigcontext()
to avoid this potential risk.
Signed-off-by: Hanlu Li <lihanlu@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
When enabling CONFIG_KASAN, CONFIG_PREEMPT_VOLUNTARY_BUILD and
CONFIG_PREEMPT_VOLUNTARY at the same time, there will be soft deadlock,
the relevant logs are as follows:
rcu: INFO: rcu_sched self-detected stall on CPU
...
Call Trace:
[<900000000024f9e4>] show_stack+0x5c/0x180
[<90000000002482f4>] dump_stack_lvl+0x94/0xbc
[<9000000000224544>] rcu_dump_cpu_stacks+0x1fc/0x280
[<900000000037ac80>] rcu_sched_clock_irq+0x720/0xf88
[<9000000000396c34>] update_process_times+0xb4/0x150
[<90000000003b2474>] tick_nohz_handler+0xf4/0x250
[<9000000000397e28>] __hrtimer_run_queues+0x1d0/0x428
[<9000000000399b2c>] hrtimer_interrupt+0x214/0x538
[<9000000000253634>] constant_timer_interrupt+0x64/0x80
[<9000000000349938>] __handle_irq_event_percpu+0x78/0x1a0
[<9000000000349a78>] handle_irq_event_percpu+0x18/0x88
[<9000000000354c00>] handle_percpu_irq+0x90/0xf0
[<9000000000348c74>] handle_irq_desc+0x94/0xb8
[<9000000001012b28>] handle_cpu_irq+0x68/0xa0
[<9000000001def8c0>] handle_loongarch_irq+0x30/0x48
[<9000000001def958>] do_vint+0x80/0xd0
[<9000000000268a0c>] kasan_mem_to_shadow.part.0+0x2c/0x2a0
[<90000000006344f4>] __asan_load8+0x4c/0x120
[<900000000025c0d0>] module_frob_arch_sections+0x5c8/0x6b8
[<90000000003895f0>] load_module+0x9e0/0x2958
[<900000000038b770>] __do_sys_init_module+0x208/0x2d0
[<9000000001df0c34>] do_syscall+0x94/0x190
[<900000000024d6fc>] handle_syscall+0xbc/0x158
After analysis, this is because the slow speed of loading the amdgpu
module leads to the long time occupation of the cpu and then the soft
deadlock.
When loading a module, module_frob_arch_sections() tries to figure out
the number of PLTs/GOTs that will be needed to handle all the RELAs. It
will call the count_max_entries() to find in an out-of-order date which
counting algorithm has O(n^2) complexity.
To make it faster, we sort the relocation list by info and addend. That
way, to check for a duplicate relocation, it just needs to compare with
the previous entry. This reduces the complexity of the algorithm to O(n
log n), as done in commit d4e0340919 ("arm64/module: Optimize module
load time by optimizing PLT counting"). This gives sinificant reduction
in module load time for modules with large number of relocations.
After applying this patch, the soft deadlock problem has been solved,
and the kernel starts normally without "Call Trace".
Using the default configuration to test some modules, the results are as
follows:
Module Size
ip_tables 36K
fat 143K
radeon 2.5MB
amdgpu 16MB
Without this patch:
Module Module load time (ms) Count(PLTs/GOTs)
ip_tables 18 59/6
fat 0 162/14
radeon 54 1221/84
amdgpu 1411 4525/1098
With this patch:
Module Module load time (ms) Count(PLTs/GOTs)
ip_tables 18 59/6
fat 0 162/14
radeon 22 1221/84
amdgpu 45 4525/1098
Fixes: fcdfe9d22b ("LoongArch: Add ELF and module support")
Signed-off-by: Kanglong Wang <wangkanglong@loongson.cn>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
