- Dec 26, 2024
-
-
Rob Clark authored
On mmu-500, stall-on-fault seems to stall all context banks, causing the GMU to misbehave. So limit this feature to smmu-v2 for now. This fixes an issue with an older mesa bug taking outo the system because of GMU going off into the year. https://lore.kernel.org/linux-iommu/CAF6AEGsFeibyH5_a7m=9PMKReD4Xb+c=dEGsbCN0_UHiJK1vbA@mail.gmail.com/T/#ma88e83db2e824a3834bda205391d36fe4de2d068 Signed-off-by:
Rob Clark <robdclark@chromium.org>
-
Vignesh Raman authored
Force db410c to host mode to fix network issue which results in failure to mount root fs via NFS. See gfx-ci/linux@cb72a629 Use apq8016-sbc-usb-host.dtb which allows the USB controllers to work in host mode. Signed-off-by:
Vignesh Raman <vignesh.raman@collabora.com>
-
Signed-off-by:Daniel Stone <daniels@collabora.com>
-
Vignesh Raman authored
Gen10 Arm Mali firmware is added to rootfs, so we can build panthor driver as module. Signed-off-by: -
Enabling the GPU power domain requires that the GPU regulator is enabled. The regulator is enabled at boot time, but automatically gets disabled when there are no users. If the GPU driver is not probed at boot time or rebound while the system is running the system will try to enable the power domain before the regulator is enabled resulting in a failure hanging the whole system. Avoid this by adding an explicit dependency. Reported-by:
Adrián Martínez Larumbe <adrian.larumbe@collabora.com> Tested-by:
Heiko Stuebner <heiko@sntech.de> Tested-by: Adrian Larumbe <adrian.larumbe@collabora.com> # On Rock 5B Signed-off-by:
Sebastian Reichel <sebastian.reichel@collabora.com>
-
Some power domains require extra voltages to be applied. For example trying to enable the GPU power domain on RK3588 fails when the SoC does not have VDD GPU enabled. The same is expected to happen for the NPU, which also has a dedicated supply line. We get the regulator using devm_of_regulator_get(), so a missing dependency in the devicetree is handled gracefully by printing a warning and creating a dummy regulator. This is necessary, since existing DTs do not have the regulator described. They might still work if the regulator is marked as always-on. It is also working if the regulator is enabled at boot time and the GPU driver is probed before the kernel disables unused regulators. The regulator itself is not acquired at driver probe time, since that creates an unsolvable circular dependency. The power domain driver must be probed early, since SoC peripherals need it. Regulators on the other hand depend on SoC peripherals like SPI, I2C or GPIO. MediaTek does not run into this, since they have two power domain drivers. Tested-by:
-
Add optional support for a voltage supply required to enable a power domain. The binding follows the way it is handled by the Mediatek binding to keep things consistent. This will initially be used by the RK3588 GPU power domain, which fails to be enabled when the GPU regulator is not enabled. It is not limited to that platform, since older generations have similar requirements. They worked around this by marking the regulators as always-on instead of describing the dependency. Reviewed-by:
Rob Herring (Arm) <robh@kernel.org> Signed-off-by:
-
Rework the logic, so that the function exits early when the power domain state is already correct to reduce code indentation. No functional change intended. Reviewed-by:
-
Currently rockchip_do_pmu_set_power_domain prints a warning if there have been errors turning on the power domain, but it does not return any errors and rockchip_pd_power() tries to continue setting up the QOS registers. This usually results in accessing unpowered registers, which triggers an SError and a full system hang. This improves the error handling by forwarding the error to avoid kernel panics. Reviewed-by:
-
Use the cleanup infrastructure to handle the mutex, which slightly improve code readability for this function. Reviewed-by:
-
The Rockchip power-domain controller also plans to make use of per-domain regulators similar to the MediaTek power-domain controller. Since existing DTs are missing the regulator information, the kernel should fallback to the automatically created dummy regulator if necessary. Thus the version without the _optional suffix is needed. The Rockchip driver plans to use the managed version, but to be consistent with existing code the unmanaged version is added at the same time. Tested-by:
-
Vignesh Raman authored
Enable DRM support for ARM Mali CSF-based GPUs and other configs to boot RK3588. Signed-off-by: -
Vignesh Raman authored
Signed-off-by: -
iid_tokens will replace $CI_JOB_TOKEN predefined variable in GitLab 18.0 See: https://docs.gitlab.com/ee/update/deprecations.html#default-cicd-job-token-ci_job_token-scope-changed This is a port of mesa/mesa!28916 in Mesa Signed-off-by:
Sergi Blanch Torne <sergi.blanch.torne@collabora.com>
-
The bucket in use to upload artifacts to S3, mesa-lava, has an expiration period of a month. If there isn't a kernel uprev in a month, the build job should be retriggered to regenerate them. Meanwhile, some jobs can fail simply because the expired artifacts. A year is a time window that ensures a kernel uprev. Signed-off-by: -
Vignesh Raman authored
Signed-off-by: -
These limits were always defined as 0, but that didn't cause any issue since the driver had hardcoded limits. In commit b4e13e1a ("scsi: ufs: qcom: Add multiple frequency support for MAX_CORE_CLK_1US_CYCLES") the hardcoded limits were removed and the driver started reading them from DT, causing UFS to stop working on MSM8996. Add real UniPro clock limits to fix UFS. Signed-off-by:
Yassine Oudjana <y.oudjana@protonmail.com> Fixes: 57fc67ef ("arm64: dts: qcom: msm8996: Add ufs related nodes") Reviewed-by:
Konrad Dybcio <konrad.dybcio@linaro.org>
-
This reverts commit 3d07a411.
-
This reverts commit 6ab502bc.
-
Signed-off-by:
David Heidelberg <david@ixit.cz> -
These ones will be needed to make use fo the NN and TP units in the NPUs based on Vivante IP. Signed-off-by:
Tomeu Vizoso <tomeu@tomeuvizoso.net>
Acked-by: Christian Gmeiner <cgmeiner@igalia.com>
-
Signed-off-by:David Heidelberg <david@ixit.cz> -
Signed-off-by:David Heidelberg <david@ixit.cz> -
We need `amdgpu.gttsize=3072` for the testing on machines with =< 4G RAM. This reverts commit f1f6f48a.
-
This is found popular brands such as StarTech.com or Delock, and has been a source of frustration to quite a few people, if I can trust Amazon comments complaining about Linux support via the official out-of-the-tree driver. Signed-off-by:Martin Peres <martin.peres@mupuf.org>
-
Change for v6.4: - rename sc7180-trogdor-kingoftown-r1 to sc7180-trogdor-kingoftown - enable BCM2835 on ARM64 (Raspberry Pi) thx @ Eric Change for v6.6: - source symlink in modules directory no longer exists - enable QRTR Signed-off-by:David Heidelberg <david@ixit.cz> -
These configs are optimized to test graphics hardware. All machines boot directly from NFS and essencial modules are present directly in the kernel. Can be applied together with default architecture defconfigs by running: ./scripts/kconfig/merge_config.sh ${DEFCONFIG} kernel/configs/mesa3d-ci_"${KERNEL_ARCH}".config Signed-off-by:David Heidelberg <david@ixit.cz> -
Needed to keep the GPU alive in our farm setup. Signed-off-by:David Heidelberg <david@ixit.cz> -
When the special handling of qcom,adreno-smmu was moved into qcom_smmu_create(), it was overlooked that we didn't have all the required entries in qcom_smmu_impl_of_match. So we stopped getting adreno_smmu_priv on sc7180, breaking per-process pgtables. Fixes: 30b912a0 ("iommu/arm-smmu-qcom: Move the qcom,adreno-smmu check into qcom_smmu_create") Signed-off-by:
-
This reverts commit 5b17993f1d2e988ec7f67f2d796754396dc413c3. Signed-off-by:Eric Anholt <eric@anholt.net>
-
Currently this is just set from the special qcom,adreno-smmu iommu. But we are not *yet* using that for a5xx. Meanwhile we still need the HUPCF bit set so iommu faults in CI do not take out the CP and cause a follow- on cascade of fail. On qcom systems, it is safe to set HUPCF everywhere so do that. Signed-off-by: -
This is a workaround for old cheza firmware v2: - add ifdef for ARM64 since the change isn't relevant outside of it Signed-off-by: -
This reverts commit 21ea4b62. Otherwise, cheza fails to probe drm-msm.
-
- Dec 15, 2024
-
-
Linus Torvalds authored
-
git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arcLinus Torvalds authored
Pull ARC fixes from Vineet Gupta: - Sundry build and misc fixes * tag 'arc-6.13-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc: ARC: build: Try to guess GCC variant of cross compiler ARC: bpf: Correct conditional check in 'check_jmp_32' ARC: dts: Replace deprecated snps,nr-gpios property for snps,dw-apb-gpio-port devices ARC: build: Use __force to suppress per-CPU cmpxchg warnings ARC: fix reference of dependency for PAE40 config ARC: build: disallow invalid PAE40 + 4K page config arc: rename aux.h to arc_aux.h
-
git://git.kernel.org/pub/scm/linux/kernel/git/efi/efiLinus Torvalds authored
Pull EFI fixes from Ard Biesheuvel: - Limit EFI zboot to GZIP and ZSTD before it comes in wider use - Fix inconsistent error when looking up a non-existent file in efivarfs with a name that does not adhere to the NAME-GUID format - Drop some unused code * tag 'efi-fixes-for-v6.13-1' of git://git.kernel.org/pub/scm/linux/kernel/git/efi/efi: efi/esrt: remove esre_attribute::store() efivarfs: Fix error on non-existent file efi/zboot: Limit compression options to GZIP and ZSTD
-
git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linuxLinus Torvalds authored
Pull i2c fixes from Wolfram Sang: "i2c host fixes: PNX used the wrong unit for timeouts, Nomadik was missing a sentinel, and RIIC was missing rounding up" * tag 'i2c-for-6.13-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux: i2c: riic: Always round-up when calculating bus period i2c: nomadik: Add missing sentinel to match table i2c: pnx: Fix timeout in wait functions
-
git://git.kernel.org/pub/scm/linux/kernel/git/ras/rasLinus Torvalds authored
Pull EDAC fix from Borislav Petkov: - Make sure amd64_edac loads successfully on certain Zen4 memory configurations * tag 'edac_urgent_for_v6.13_rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras: EDAC/amd64: Simplify ECC check on unified memory controllers
-
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tipLinus Torvalds authored
Pull irq fixes from Borislav Petkov: - Disable the secure programming interface of the GIC500 chip in the RK3399 SoC to fix interrupt priority assignment and even make a dead machine boot again when the gic-v3 driver enables pseudo NMIs - Correct the declaration of a percpu variable to fix several sparse warnings * tag 'irq_urgent_for_v6.13_rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: irqchip/gic-v3: Work around insecure GIC integrations irqchip/gic: Correct declaration of *percpu_base pointer in union gic_base
-
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tipLinus Torvalds authored
Pull scheduler fixes from Borislav Petkov: - Prevent incorrect dequeueing of the deadline dlserver helper task and fix its time accounting - Properly track the CFS runqueue runnable stats - Check the total number of all queued tasks in a sched fair's runqueue hierarchy before deciding to stop the tick - Fix the scheduling of the task that got woken last (NEXT_BUDDY) by preventing those from being delayed * tag 'sched_urgent_for_v6.13_rc3-p2' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: sched/dlserver: Fix dlserver time accounting sched/dlserver: Fix dlserver double enqueue sched/eevdf: More PELT vs DELAYED_DEQUEUE sched/fair: Fix sched_can_stop_tick() for fair tasks sched/fair: Fix NEXT_BUDDY
-
