- Linux-stable-mirror - lists.linaro.org

FAILED: patch "[PATCH] intel_idle: Handle older CPUs, which stop the TSC in deeper C" failed to apply to 5.15-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 5.15-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-5.15.y git checkout FETCH_HEAD git cherry-pick -x c157d351460bcf202970e97e611cb6b54a3dd4a4 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025030416-straining-detergent-2e6d@gregkh' --subject-prefix 'PATCH 5.15.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From c157d351460bcf202970e97e611cb6b54a3dd4a4 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner <tglx(a)linutronix.de> Date: Tue, 25 Feb 2025 23:37:08 +0100 Subject: [PATCH] intel_idle: Handle older CPUs, which stop the TSC in deeper C states, correctly The Intel idle driver is preferred over the ACPI processor idle driver, but fails to implement the work around for Core2 generation CPUs, where the TSC stops in C2 and deeper C-states. This causes stalls and boot delays, when the clocksource watchdog does not catch the unstable TSC before the CPU goes deep idle for the first time. The ACPI driver marks the TSC unstable when it detects that the CPU supports C2 or deeper and the CPU does not have a non-stop TSC. Add the equivivalent work around to the Intel idle driver to cure that. Fixes: 18734958e9bf ("intel_idle: Use ACPI _CST for processor models without C-state tables") Reported-by: Fab Stz <fabstz-it(a)yahoo.fr> Signed-off-by: Thomas Gleixner <tglx(a)linutronix.de> Tested-by: Fab Stz <fabstz-it(a)yahoo.fr> Cc: All applicable <stable(a)vger.kernel.org> Closes: https://lore.kernel.org/all/10cf96aa-1276-4bd4-8966-c890377030c3@yahoo.fr Link: https://patch.msgid.link/87bjupfy7f.ffs@tglx Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki(a)intel.com> diff --git a/drivers/idle/intel_idle.c b/drivers/idle/intel_idle.c index 118fe1d37c22..0fdb1d1316c4 100644 --- a/drivers/idle/intel_idle.c +++ b/drivers/idle/intel_idle.c @@ -56,6 +56,7 @@ #include <asm/intel-family.h> #include <asm/mwait.h> #include <asm/spec-ctrl.h> +#include <asm/tsc.h> #include <asm/fpu/api.h> #define INTEL_IDLE_VERSION "0.5.1" @@ -1799,6 +1800,9 @@ static void __init intel_idle_init_cstates_acpi(struct cpuidle_driver *drv) if (intel_idle_state_needs_timer_stop(state)) state->flags |= CPUIDLE_FLAG_TIMER_STOP; + if (cx->type > ACPI_STATE_C1 && !boot_cpu_has(X86_FEATURE_NONSTOP_TSC)) + mark_tsc_unstable("TSC halts in idle"); + state->enter = intel_idle; state->enter_s2idle = intel_idle_s2idle; }

4 months

2
1
0 0

FAILED: patch "[PATCH] intel_idle: Handle older CPUs, which stop the TSC in deeper C" failed to apply to 6.6-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 6.6-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.6.y git checkout FETCH_HEAD git cherry-pick -x c157d351460bcf202970e97e611cb6b54a3dd4a4 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025030415-uncombed-drinking-e339@gregkh' --subject-prefix 'PATCH 6.6.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From c157d351460bcf202970e97e611cb6b54a3dd4a4 Mon Sep 17 00:00:00 2001 From: Thomas Gleixner <tglx(a)linutronix.de> Date: Tue, 25 Feb 2025 23:37:08 +0100 Subject: [PATCH] intel_idle: Handle older CPUs, which stop the TSC in deeper C states, correctly The Intel idle driver is preferred over the ACPI processor idle driver, but fails to implement the work around for Core2 generation CPUs, where the TSC stops in C2 and deeper C-states. This causes stalls and boot delays, when the clocksource watchdog does not catch the unstable TSC before the CPU goes deep idle for the first time. The ACPI driver marks the TSC unstable when it detects that the CPU supports C2 or deeper and the CPU does not have a non-stop TSC. Add the equivivalent work around to the Intel idle driver to cure that. Fixes: 18734958e9bf ("intel_idle: Use ACPI _CST for processor models without C-state tables") Reported-by: Fab Stz <fabstz-it(a)yahoo.fr> Signed-off-by: Thomas Gleixner <tglx(a)linutronix.de> Tested-by: Fab Stz <fabstz-it(a)yahoo.fr> Cc: All applicable <stable(a)vger.kernel.org> Closes: https://lore.kernel.org/all/10cf96aa-1276-4bd4-8966-c890377030c3@yahoo.fr Link: https://patch.msgid.link/87bjupfy7f.ffs@tglx Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki(a)intel.com> diff --git a/drivers/idle/intel_idle.c b/drivers/idle/intel_idle.c index 118fe1d37c22..0fdb1d1316c4 100644 --- a/drivers/idle/intel_idle.c +++ b/drivers/idle/intel_idle.c @@ -56,6 +56,7 @@ #include <asm/intel-family.h> #include <asm/mwait.h> #include <asm/spec-ctrl.h> +#include <asm/tsc.h> #include <asm/fpu/api.h> #define INTEL_IDLE_VERSION "0.5.1" @@ -1799,6 +1800,9 @@ static void __init intel_idle_init_cstates_acpi(struct cpuidle_driver *drv) if (intel_idle_state_needs_timer_stop(state)) state->flags |= CPUIDLE_FLAG_TIMER_STOP; + if (cx->type > ACPI_STATE_C1 && !boot_cpu_has(X86_FEATURE_NONSTOP_TSC)) + mark_tsc_unstable("TSC halts in idle"); + state->enter = intel_idle; state->enter_s2idle = intel_idle_s2idle; }

4 months

2
1
0 0

[PATCH v2] kbuild: userprogs: use correct lld when linking through clang

by Thomas Weißschuh

The userprog infrastructure links objects files through $(CC). Either explicitly by manually calling $(CC) on multiple object files or implicitly by directly compiling a source file to an executable. The documentation at Documentation/kbuild/llvm.rst indicates that ld.lld would be used for linking if LLVM=1 is specified. However clang instead will use either a globally installed cross linker from $PATH called ${target}-ld or fall back to the system linker, which probably does not support crosslinking. For the normal kernel build this is not an issue because the linker is always executed directly, without the compiler being involved. Explicitly pass --ld-path to clang so $(LD) is respected. As clang 13.0.1 is required to build the kernel, this option is available. Fixes: 7f3a59db274c ("kbuild: add infrastructure to build userspace programs") Cc: stable(a)vger.kernel.org # needs wrapping in $(cc-option) for < 6.9 Signed-off-by: Thomas Weißschuh <thomas.weissschuh(a)linutronix.de> --- Reproducer, using nolibc to avoid libc requirements for cross building: $ tail -2 init/Makefile userprogs-always-y += test-llvm test-llvm-userccflags += -nostdlib -nolibc -static -isystem usr/ -include $(srctree)/tools/include/nolibc/nolibc.h $ cat init/test-llvm.c int main(void) { return 0; } $ make ARCH=arm64 LLVM=1 allnoconfig headers_install init/ Validate that init/test-llvm builds and has the correct binary format. --- Changes in v2: - Use --ld-path instead of -fuse-ld - Drop already applied patch - Link to v1: https://lore.kernel.org/r/20250213-kbuild-userprog-fixes-v1-0-f255fb477d98@… --- Makefile | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/Makefile b/Makefile index 96407c1d6be167b04ed5883e455686918c7a75ee..b41c164533d781d010ff8b2522e523164dc375d0 100644 --- a/Makefile +++ b/Makefile @@ -1123,6 +1123,11 @@ endif KBUILD_USERCFLAGS += $(filter -m32 -m64 --target=%, $(KBUILD_CPPFLAGS) $(KBUILD_CFLAGS)) KBUILD_USERLDFLAGS += $(filter -m32 -m64 --target=%, $(KBUILD_CPPFLAGS) $(KBUILD_CFLAGS)) +# userspace programs are linked via the compiler, use the correct linker +ifeq ($(CONFIG_CC_IS_CLANG)$(CONFIG_LD_IS_LLD),yy) +KBUILD_USERLDFLAGS += --ld-path=$(LD) +endif + # make the checker run with the right architecture CHECKFLAGS += --arch=$(ARCH) --- base-commit: 0ad2507d5d93f39619fc42372c347d6006b64319 change-id: 20250213-kbuild-userprog-fixes-4f07b62ae818 Best regards, -- Thomas Weißschuh <thomas.weissschuh(a)linutronix.de>

4 months

3
2
0 0

FAILED: patch "[PATCH] amdgpu/pm/legacy: fix suspend/resume issues" failed to apply to 6.1-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 6.1-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.1.y git checkout FETCH_HEAD git cherry-pick -x 91dcc66b34beb72dde8412421bdc1b4cd40e4fb8 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025030458-properly-playroom-8207@gregkh' --subject-prefix 'PATCH 6.1.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From 91dcc66b34beb72dde8412421bdc1b4cd40e4fb8 Mon Sep 17 00:00:00 2001 From: "chr[]" <chris(a)rudorff.com> Date: Wed, 12 Feb 2025 16:51:38 +0100 Subject: [PATCH] amdgpu/pm/legacy: fix suspend/resume issues resume and irq handler happily races in set_power_state() * amdgpu_legacy_dpm_compute_clocks() needs lock * protect irq work handler * fix dpm_enabled usage v2: fix clang build, integrate Lijo's comments (Alex) Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/2524 Fixes: 3712e7a49459 ("drm/amd/pm: unified lock protections in amdgpu_dpm.c") Reviewed-by: Lijo Lazar <lijo.lazar(a)amd.com> Tested-by: Maciej S. Szmigiero <mail(a)maciej.szmigiero.name> # on Oland PRO Signed-off-by: chr[] <chris(a)rudorff.com> Signed-off-by: Alex Deucher <alexander.deucher(a)amd.com> (cherry picked from commit ee3dc9e204d271c9c7a8d4d38a0bce4745d33e71) Cc: stable(a)vger.kernel.org diff --git a/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c b/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c index 67a8e22b1126..e237ea1185a7 100644 --- a/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c +++ b/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c @@ -3042,6 +3042,7 @@ static int kv_dpm_hw_init(struct amdgpu_ip_block *ip_block) if (!amdgpu_dpm) return 0; + mutex_lock(&adev->pm.mutex); kv_dpm_setup_asic(adev); ret = kv_dpm_enable(adev); if (ret) @@ -3049,6 +3050,8 @@ static int kv_dpm_hw_init(struct amdgpu_ip_block *ip_block) else adev->pm.dpm_enabled = true; amdgpu_legacy_dpm_compute_clocks(adev); + mutex_unlock(&adev->pm.mutex); + return ret; } @@ -3066,32 +3069,42 @@ static int kv_dpm_suspend(struct amdgpu_ip_block *ip_block) { struct amdgpu_device *adev = ip_block->adev; + cancel_work_sync(&adev->pm.dpm.thermal.work); + if (adev->pm.dpm_enabled) { + mutex_lock(&adev->pm.mutex); + adev->pm.dpm_enabled = false; /* disable dpm */ kv_dpm_disable(adev); /* reset the power state */ adev->pm.dpm.current_ps = adev->pm.dpm.requested_ps = adev->pm.dpm.boot_ps; + mutex_unlock(&adev->pm.mutex); } return 0; } static int kv_dpm_resume(struct amdgpu_ip_block *ip_block) { - int ret; + int ret = 0; struct amdgpu_device *adev = ip_block->adev; - if (adev->pm.dpm_enabled) { + if (!amdgpu_dpm) + return 0; + + if (!adev->pm.dpm_enabled) { + mutex_lock(&adev->pm.mutex); /* asic init will reset to the boot state */ kv_dpm_setup_asic(adev); ret = kv_dpm_enable(adev); - if (ret) + if (ret) { adev->pm.dpm_enabled = false; - else + } else { adev->pm.dpm_enabled = true; - if (adev->pm.dpm_enabled) amdgpu_legacy_dpm_compute_clocks(adev); + } + mutex_unlock(&adev->pm.mutex); } - return 0; + return ret; } static bool kv_dpm_is_idle(void *handle) diff --git a/drivers/gpu/drm/amd/pm/legacy-dpm/legacy_dpm.c b/drivers/gpu/drm/amd/pm/legacy-dpm/legacy_dpm.c index e861355ebd75..c7518b13e787 100644 --- a/drivers/gpu/drm/amd/pm/legacy-dpm/legacy_dpm.c +++ b/drivers/gpu/drm/amd/pm/legacy-dpm/legacy_dpm.c @@ -1009,9 +1009,12 @@ void amdgpu_dpm_thermal_work_handler(struct work_struct *work) enum amd_pm_state_type dpm_state = POWER_STATE_TYPE_INTERNAL_THERMAL; int temp, size = sizeof(temp); - if (!adev->pm.dpm_enabled) - return; + mutex_lock(&adev->pm.mutex); + if (!adev->pm.dpm_enabled) { + mutex_unlock(&adev->pm.mutex); + return; + } if (!pp_funcs->read_sensor(adev->powerplay.pp_handle, AMDGPU_PP_SENSOR_GPU_TEMP, (void *)&temp, @@ -1033,4 +1036,5 @@ void amdgpu_dpm_thermal_work_handler(struct work_struct *work) adev->pm.dpm.state = dpm_state; amdgpu_legacy_dpm_compute_clocks(adev->powerplay.pp_handle); + mutex_unlock(&adev->pm.mutex); } diff --git a/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c b/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c index a87dcf0974bc..d6dfe2599ebe 100644 --- a/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c +++ b/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c @@ -7786,6 +7786,7 @@ static int si_dpm_hw_init(struct amdgpu_ip_block *ip_block) if (!amdgpu_dpm) return 0; + mutex_lock(&adev->pm.mutex); si_dpm_setup_asic(adev); ret = si_dpm_enable(adev); if (ret) @@ -7793,6 +7794,7 @@ static int si_dpm_hw_init(struct amdgpu_ip_block *ip_block) else adev->pm.dpm_enabled = true; amdgpu_legacy_dpm_compute_clocks(adev); + mutex_unlock(&adev->pm.mutex); return ret; } @@ -7810,32 +7812,44 @@ static int si_dpm_suspend(struct amdgpu_ip_block *ip_block) { struct amdgpu_device *adev = ip_block->adev; + cancel_work_sync(&adev->pm.dpm.thermal.work); + if (adev->pm.dpm_enabled) { + mutex_lock(&adev->pm.mutex); + adev->pm.dpm_enabled = false; /* disable dpm */ si_dpm_disable(adev); /* reset the power state */ adev->pm.dpm.current_ps = adev->pm.dpm.requested_ps = adev->pm.dpm.boot_ps; + mutex_unlock(&adev->pm.mutex); } + return 0; } static int si_dpm_resume(struct amdgpu_ip_block *ip_block) { - int ret; + int ret = 0; struct amdgpu_device *adev = ip_block->adev; - if (adev->pm.dpm_enabled) { + if (!amdgpu_dpm) + return 0; + + if (!adev->pm.dpm_enabled) { /* asic init will reset to the boot state */ + mutex_lock(&adev->pm.mutex); si_dpm_setup_asic(adev); ret = si_dpm_enable(adev); - if (ret) + if (ret) { adev->pm.dpm_enabled = false; - else + } else { adev->pm.dpm_enabled = true; - if (adev->pm.dpm_enabled) amdgpu_legacy_dpm_compute_clocks(adev); + } + mutex_unlock(&adev->pm.mutex); } - return 0; + + return ret; } static bool si_dpm_is_idle(void *handle)

4 months

2
1
0 0

FAILED: patch "[PATCH] amdgpu/pm/legacy: fix suspend/resume issues" failed to apply to 6.6-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 6.6-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.6.y git checkout FETCH_HEAD git cherry-pick -x 91dcc66b34beb72dde8412421bdc1b4cd40e4fb8 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025030458-lying-nintendo-f777@gregkh' --subject-prefix 'PATCH 6.6.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From 91dcc66b34beb72dde8412421bdc1b4cd40e4fb8 Mon Sep 17 00:00:00 2001 From: "chr[]" <chris(a)rudorff.com> Date: Wed, 12 Feb 2025 16:51:38 +0100 Subject: [PATCH] amdgpu/pm/legacy: fix suspend/resume issues resume and irq handler happily races in set_power_state() * amdgpu_legacy_dpm_compute_clocks() needs lock * protect irq work handler * fix dpm_enabled usage v2: fix clang build, integrate Lijo's comments (Alex) Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/2524 Fixes: 3712e7a49459 ("drm/amd/pm: unified lock protections in amdgpu_dpm.c") Reviewed-by: Lijo Lazar <lijo.lazar(a)amd.com> Tested-by: Maciej S. Szmigiero <mail(a)maciej.szmigiero.name> # on Oland PRO Signed-off-by: chr[] <chris(a)rudorff.com> Signed-off-by: Alex Deucher <alexander.deucher(a)amd.com> (cherry picked from commit ee3dc9e204d271c9c7a8d4d38a0bce4745d33e71) Cc: stable(a)vger.kernel.org diff --git a/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c b/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c index 67a8e22b1126..e237ea1185a7 100644 --- a/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c +++ b/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c @@ -3042,6 +3042,7 @@ static int kv_dpm_hw_init(struct amdgpu_ip_block *ip_block) if (!amdgpu_dpm) return 0; + mutex_lock(&adev->pm.mutex); kv_dpm_setup_asic(adev); ret = kv_dpm_enable(adev); if (ret) @@ -3049,6 +3050,8 @@ static int kv_dpm_hw_init(struct amdgpu_ip_block *ip_block) else adev->pm.dpm_enabled = true; amdgpu_legacy_dpm_compute_clocks(adev); + mutex_unlock(&adev->pm.mutex); + return ret; } @@ -3066,32 +3069,42 @@ static int kv_dpm_suspend(struct amdgpu_ip_block *ip_block) { struct amdgpu_device *adev = ip_block->adev; + cancel_work_sync(&adev->pm.dpm.thermal.work); + if (adev->pm.dpm_enabled) { + mutex_lock(&adev->pm.mutex); + adev->pm.dpm_enabled = false; /* disable dpm */ kv_dpm_disable(adev); /* reset the power state */ adev->pm.dpm.current_ps = adev->pm.dpm.requested_ps = adev->pm.dpm.boot_ps; + mutex_unlock(&adev->pm.mutex); } return 0; } static int kv_dpm_resume(struct amdgpu_ip_block *ip_block) { - int ret; + int ret = 0; struct amdgpu_device *adev = ip_block->adev; - if (adev->pm.dpm_enabled) { + if (!amdgpu_dpm) + return 0; + + if (!adev->pm.dpm_enabled) { + mutex_lock(&adev->pm.mutex); /* asic init will reset to the boot state */ kv_dpm_setup_asic(adev); ret = kv_dpm_enable(adev); - if (ret) + if (ret) { adev->pm.dpm_enabled = false; - else + } else { adev->pm.dpm_enabled = true; - if (adev->pm.dpm_enabled) amdgpu_legacy_dpm_compute_clocks(adev); + } + mutex_unlock(&adev->pm.mutex); } - return 0; + return ret; } static bool kv_dpm_is_idle(void *handle) diff --git a/drivers/gpu/drm/amd/pm/legacy-dpm/legacy_dpm.c b/drivers/gpu/drm/amd/pm/legacy-dpm/legacy_dpm.c index e861355ebd75..c7518b13e787 100644 --- a/drivers/gpu/drm/amd/pm/legacy-dpm/legacy_dpm.c +++ b/drivers/gpu/drm/amd/pm/legacy-dpm/legacy_dpm.c @@ -1009,9 +1009,12 @@ void amdgpu_dpm_thermal_work_handler(struct work_struct *work) enum amd_pm_state_type dpm_state = POWER_STATE_TYPE_INTERNAL_THERMAL; int temp, size = sizeof(temp); - if (!adev->pm.dpm_enabled) - return; + mutex_lock(&adev->pm.mutex); + if (!adev->pm.dpm_enabled) { + mutex_unlock(&adev->pm.mutex); + return; + } if (!pp_funcs->read_sensor(adev->powerplay.pp_handle, AMDGPU_PP_SENSOR_GPU_TEMP, (void *)&temp, @@ -1033,4 +1036,5 @@ void amdgpu_dpm_thermal_work_handler(struct work_struct *work) adev->pm.dpm.state = dpm_state; amdgpu_legacy_dpm_compute_clocks(adev->powerplay.pp_handle); + mutex_unlock(&adev->pm.mutex); } diff --git a/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c b/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c index a87dcf0974bc..d6dfe2599ebe 100644 --- a/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c +++ b/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c @@ -7786,6 +7786,7 @@ static int si_dpm_hw_init(struct amdgpu_ip_block *ip_block) if (!amdgpu_dpm) return 0; + mutex_lock(&adev->pm.mutex); si_dpm_setup_asic(adev); ret = si_dpm_enable(adev); if (ret) @@ -7793,6 +7794,7 @@ static int si_dpm_hw_init(struct amdgpu_ip_block *ip_block) else adev->pm.dpm_enabled = true; amdgpu_legacy_dpm_compute_clocks(adev); + mutex_unlock(&adev->pm.mutex); return ret; } @@ -7810,32 +7812,44 @@ static int si_dpm_suspend(struct amdgpu_ip_block *ip_block) { struct amdgpu_device *adev = ip_block->adev; + cancel_work_sync(&adev->pm.dpm.thermal.work); + if (adev->pm.dpm_enabled) { + mutex_lock(&adev->pm.mutex); + adev->pm.dpm_enabled = false; /* disable dpm */ si_dpm_disable(adev); /* reset the power state */ adev->pm.dpm.current_ps = adev->pm.dpm.requested_ps = adev->pm.dpm.boot_ps; + mutex_unlock(&adev->pm.mutex); } + return 0; } static int si_dpm_resume(struct amdgpu_ip_block *ip_block) { - int ret; + int ret = 0; struct amdgpu_device *adev = ip_block->adev; - if (adev->pm.dpm_enabled) { + if (!amdgpu_dpm) + return 0; + + if (!adev->pm.dpm_enabled) { /* asic init will reset to the boot state */ + mutex_lock(&adev->pm.mutex); si_dpm_setup_asic(adev); ret = si_dpm_enable(adev); - if (ret) + if (ret) { adev->pm.dpm_enabled = false; - else + } else { adev->pm.dpm_enabled = true; - if (adev->pm.dpm_enabled) amdgpu_legacy_dpm_compute_clocks(adev); + } + mutex_unlock(&adev->pm.mutex); } - return 0; + + return ret; } static bool si_dpm_is_idle(void *handle)

4 months

2
1
0 0

FAILED: patch "[PATCH] amdgpu/pm/legacy: fix suspend/resume issues" failed to apply to 6.12-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 6.12-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.12.y git checkout FETCH_HEAD git cherry-pick -x 91dcc66b34beb72dde8412421bdc1b4cd40e4fb8 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025030457-pamphlet-galley-603b@gregkh' --subject-prefix 'PATCH 6.12.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From 91dcc66b34beb72dde8412421bdc1b4cd40e4fb8 Mon Sep 17 00:00:00 2001 From: "chr[]" <chris(a)rudorff.com> Date: Wed, 12 Feb 2025 16:51:38 +0100 Subject: [PATCH] amdgpu/pm/legacy: fix suspend/resume issues resume and irq handler happily races in set_power_state() * amdgpu_legacy_dpm_compute_clocks() needs lock * protect irq work handler * fix dpm_enabled usage v2: fix clang build, integrate Lijo's comments (Alex) Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/2524 Fixes: 3712e7a49459 ("drm/amd/pm: unified lock protections in amdgpu_dpm.c") Reviewed-by: Lijo Lazar <lijo.lazar(a)amd.com> Tested-by: Maciej S. Szmigiero <mail(a)maciej.szmigiero.name> # on Oland PRO Signed-off-by: chr[] <chris(a)rudorff.com> Signed-off-by: Alex Deucher <alexander.deucher(a)amd.com> (cherry picked from commit ee3dc9e204d271c9c7a8d4d38a0bce4745d33e71) Cc: stable(a)vger.kernel.org diff --git a/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c b/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c index 67a8e22b1126..e237ea1185a7 100644 --- a/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c +++ b/drivers/gpu/drm/amd/pm/legacy-dpm/kv_dpm.c @@ -3042,6 +3042,7 @@ static int kv_dpm_hw_init(struct amdgpu_ip_block *ip_block) if (!amdgpu_dpm) return 0; + mutex_lock(&adev->pm.mutex); kv_dpm_setup_asic(adev); ret = kv_dpm_enable(adev); if (ret) @@ -3049,6 +3050,8 @@ static int kv_dpm_hw_init(struct amdgpu_ip_block *ip_block) else adev->pm.dpm_enabled = true; amdgpu_legacy_dpm_compute_clocks(adev); + mutex_unlock(&adev->pm.mutex); + return ret; } @@ -3066,32 +3069,42 @@ static int kv_dpm_suspend(struct amdgpu_ip_block *ip_block) { struct amdgpu_device *adev = ip_block->adev; + cancel_work_sync(&adev->pm.dpm.thermal.work); + if (adev->pm.dpm_enabled) { + mutex_lock(&adev->pm.mutex); + adev->pm.dpm_enabled = false; /* disable dpm */ kv_dpm_disable(adev); /* reset the power state */ adev->pm.dpm.current_ps = adev->pm.dpm.requested_ps = adev->pm.dpm.boot_ps; + mutex_unlock(&adev->pm.mutex); } return 0; } static int kv_dpm_resume(struct amdgpu_ip_block *ip_block) { - int ret; + int ret = 0; struct amdgpu_device *adev = ip_block->adev; - if (adev->pm.dpm_enabled) { + if (!amdgpu_dpm) + return 0; + + if (!adev->pm.dpm_enabled) { + mutex_lock(&adev->pm.mutex); /* asic init will reset to the boot state */ kv_dpm_setup_asic(adev); ret = kv_dpm_enable(adev); - if (ret) + if (ret) { adev->pm.dpm_enabled = false; - else + } else { adev->pm.dpm_enabled = true; - if (adev->pm.dpm_enabled) amdgpu_legacy_dpm_compute_clocks(adev); + } + mutex_unlock(&adev->pm.mutex); } - return 0; + return ret; } static bool kv_dpm_is_idle(void *handle) diff --git a/drivers/gpu/drm/amd/pm/legacy-dpm/legacy_dpm.c b/drivers/gpu/drm/amd/pm/legacy-dpm/legacy_dpm.c index e861355ebd75..c7518b13e787 100644 --- a/drivers/gpu/drm/amd/pm/legacy-dpm/legacy_dpm.c +++ b/drivers/gpu/drm/amd/pm/legacy-dpm/legacy_dpm.c @@ -1009,9 +1009,12 @@ void amdgpu_dpm_thermal_work_handler(struct work_struct *work) enum amd_pm_state_type dpm_state = POWER_STATE_TYPE_INTERNAL_THERMAL; int temp, size = sizeof(temp); - if (!adev->pm.dpm_enabled) - return; + mutex_lock(&adev->pm.mutex); + if (!adev->pm.dpm_enabled) { + mutex_unlock(&adev->pm.mutex); + return; + } if (!pp_funcs->read_sensor(adev->powerplay.pp_handle, AMDGPU_PP_SENSOR_GPU_TEMP, (void *)&temp, @@ -1033,4 +1036,5 @@ void amdgpu_dpm_thermal_work_handler(struct work_struct *work) adev->pm.dpm.state = dpm_state; amdgpu_legacy_dpm_compute_clocks(adev->powerplay.pp_handle); + mutex_unlock(&adev->pm.mutex); } diff --git a/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c b/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c index a87dcf0974bc..d6dfe2599ebe 100644 --- a/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c +++ b/drivers/gpu/drm/amd/pm/legacy-dpm/si_dpm.c @@ -7786,6 +7786,7 @@ static int si_dpm_hw_init(struct amdgpu_ip_block *ip_block) if (!amdgpu_dpm) return 0; + mutex_lock(&adev->pm.mutex); si_dpm_setup_asic(adev); ret = si_dpm_enable(adev); if (ret) @@ -7793,6 +7794,7 @@ static int si_dpm_hw_init(struct amdgpu_ip_block *ip_block) else adev->pm.dpm_enabled = true; amdgpu_legacy_dpm_compute_clocks(adev); + mutex_unlock(&adev->pm.mutex); return ret; } @@ -7810,32 +7812,44 @@ static int si_dpm_suspend(struct amdgpu_ip_block *ip_block) { struct amdgpu_device *adev = ip_block->adev; + cancel_work_sync(&adev->pm.dpm.thermal.work); + if (adev->pm.dpm_enabled) { + mutex_lock(&adev->pm.mutex); + adev->pm.dpm_enabled = false; /* disable dpm */ si_dpm_disable(adev); /* reset the power state */ adev->pm.dpm.current_ps = adev->pm.dpm.requested_ps = adev->pm.dpm.boot_ps; + mutex_unlock(&adev->pm.mutex); } + return 0; } static int si_dpm_resume(struct amdgpu_ip_block *ip_block) { - int ret; + int ret = 0; struct amdgpu_device *adev = ip_block->adev; - if (adev->pm.dpm_enabled) { + if (!amdgpu_dpm) + return 0; + + if (!adev->pm.dpm_enabled) { /* asic init will reset to the boot state */ + mutex_lock(&adev->pm.mutex); si_dpm_setup_asic(adev); ret = si_dpm_enable(adev); - if (ret) + if (ret) { adev->pm.dpm_enabled = false; - else + } else { adev->pm.dpm_enabled = true; - if (adev->pm.dpm_enabled) amdgpu_legacy_dpm_compute_clocks(adev); + } + mutex_unlock(&adev->pm.mutex); } - return 0; + + return ret; } static bool si_dpm_is_idle(void *handle)

4 months

2
1
0 0

[PATCH v2 1/4] drm/xe/vm: Validate userptr during gpu vma prefetching

by Thomas Hellström

If a userptr vma subject to prefetching was already invalidated or invalidated during the prefetch operation, the operation would repeatedly return -EAGAIN which would typically cause an infinite loop. Validate the userptr to ensure this doesn't happen. v2: - Don't fallthrough from UNMAP to PREFETCH (Matthew Brost) Fixes: 5bd24e78829a ("drm/xe/vm: Subclass userptr vmas") Fixes: 617eebb9c480 ("drm/xe: Fix array of binds") Cc: Matthew Brost <matthew.brost(a)intel.com> Cc: <stable(a)vger.kernel.org> # v6.9+ Suggested-by: Matthew Brost <matthew.brost(a)intel.com> Signed-off-by: Thomas Hellström <thomas.hellstrom(a)linux.intel.com> --- drivers/gpu/drm/xe/xe_vm.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c index 996000f2424e..6fdc17be619e 100644 --- a/drivers/gpu/drm/xe/xe_vm.c +++ b/drivers/gpu/drm/xe/xe_vm.c @@ -2306,8 +2306,17 @@ static int vm_bind_ioctl_ops_parse(struct xe_vm *vm, struct drm_gpuva_ops *ops, break; } case DRM_GPUVA_OP_UNMAP: + xe_vma_ops_incr_pt_update_ops(vops, op->tile_mask); + break; case DRM_GPUVA_OP_PREFETCH: - /* FIXME: Need to skip some prefetch ops */ + vma = gpuva_to_vma(op->base.prefetch.va); + + if (xe_vma_is_userptr(vma)) { + err = xe_vma_userptr_pin_pages(to_userptr_vma(vma)); + if (err) + return err; + } + xe_vma_ops_incr_pt_update_ops(vops, op->tile_mask); break; default: -- 2.48.1

4 months

2
1
0 0

FAILED: patch "[PATCH] efi: Don't map the entire mokvar table to determine its size" failed to apply to 6.13-stable tree

by gregkh＠linuxfoundation.org

The patch below does not apply to the 6.13-stable tree. If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <stable(a)vger.kernel.org>. To reproduce the conflict and resubmit, you may use the following commands: git fetch https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/ linux-6.13.y git checkout FETCH_HEAD git cherry-pick -x 2b90e7ace79774a3540ce569e000388f8d22c9e0 # <resolve conflicts, build, test, etc.> git commit -s git send-email --to '<stable(a)vger.kernel.org>' --in-reply-to '2025030421-disgrace-wince-4786@gregkh' --subject-prefix 'PATCH 6.13.y' HEAD^.. Possible dependencies: thanks, greg k-h ------------------ original commit in Linus's tree ------------------ From 2b90e7ace79774a3540ce569e000388f8d22c9e0 Mon Sep 17 00:00:00 2001 From: Peter Jones <pjones(a)redhat.com> Date: Wed, 26 Feb 2025 15:18:39 -0500 Subject: [PATCH] efi: Don't map the entire mokvar table to determine its size Currently, when validating the mokvar table, we (re)map the entire table on each iteration of the loop, adding space as we discover new entries. If the table grows over a certain size, this fails due to limitations of early_memmap(), and we get a failure and traceback: ------------[ cut here ]------------ WARNING: CPU: 0 PID: 0 at mm/early_ioremap.c:139 __early_ioremap+0xef/0x220 ... Call Trace: <TASK> ? __early_ioremap+0xef/0x220 ? __warn.cold+0x93/0xfa ? __early_ioremap+0xef/0x220 ? report_bug+0xff/0x140 ? early_fixup_exception+0x5d/0xb0 ? early_idt_handler_common+0x2f/0x3a ? __early_ioremap+0xef/0x220 ? efi_mokvar_table_init+0xce/0x1d0 ? setup_arch+0x864/0xc10 ? start_kernel+0x6b/0xa10 ? x86_64_start_reservations+0x24/0x30 ? x86_64_start_kernel+0xed/0xf0 ? common_startup_64+0x13e/0x141 </TASK> ---[ end trace 0000000000000000 ]--- mokvar: Failed to map EFI MOKvar config table pa=0x7c4c3000, size=265187. Mapping the entire structure isn't actually necessary, as we don't ever need more than one entry header mapped at once. Changes efi_mokvar_table_init() to only map each entry header, not the entire table, when determining the table size. Since we're not mapping any data past the variable name, it also changes the code to enforce that each variable name is NUL terminated, rather than attempting to verify it in place. Cc: <stable(a)vger.kernel.org> Signed-off-by: Peter Jones <pjones(a)redhat.com> Signed-off-by: Ard Biesheuvel <ardb(a)kernel.org> diff --git a/drivers/firmware/efi/mokvar-table.c b/drivers/firmware/efi/mokvar-table.c index 5ed0602c2f75..d865cb1dbaad 100644 --- a/drivers/firmware/efi/mokvar-table.c +++ b/drivers/firmware/efi/mokvar-table.c @@ -103,7 +103,6 @@ void __init efi_mokvar_table_init(void) void *va = NULL; unsigned long cur_offset = 0; unsigned long offset_limit; - unsigned long map_size = 0; unsigned long map_size_needed = 0; unsigned long size; struct efi_mokvar_table_entry *mokvar_entry; @@ -134,48 +133,34 @@ void __init efi_mokvar_table_init(void) */ err = -EINVAL; while (cur_offset + sizeof(*mokvar_entry) <= offset_limit) { - mokvar_entry = va + cur_offset; - map_size_needed = cur_offset + sizeof(*mokvar_entry); - if (map_size_needed > map_size) { - if (va) - early_memunmap(va, map_size); - /* - * Map a little more than the fixed size entry - * header, anticipating some data. It's safe to - * do so as long as we stay within current memory - * descriptor. - */ - map_size = min(map_size_needed + 2*EFI_PAGE_SIZE, - offset_limit); - va = early_memremap(efi.mokvar_table, map_size); - if (!va) { - pr_err("Failed to map EFI MOKvar config table pa=0x%lx, size=%lu.\n", - efi.mokvar_table, map_size); - return; - } - mokvar_entry = va + cur_offset; + if (va) + early_memunmap(va, sizeof(*mokvar_entry)); + va = early_memremap(efi.mokvar_table + cur_offset, sizeof(*mokvar_entry)); + if (!va) { + pr_err("Failed to map EFI MOKvar config table pa=0x%lx, size=%zu.\n", + efi.mokvar_table + cur_offset, sizeof(*mokvar_entry)); + return; } + mokvar_entry = va; /* Check for last sentinel entry */ if (mokvar_entry->name[0] == '\0') { if (mokvar_entry->data_size != 0) break; err = 0; + map_size_needed = cur_offset + sizeof(*mokvar_entry); break; } - /* Sanity check that the name is null terminated */ - size = strnlen(mokvar_entry->name, - sizeof(mokvar_entry->name)); - if (size >= sizeof(mokvar_entry->name)) - break; + /* Enforce that the name is NUL terminated */ + mokvar_entry->name[sizeof(mokvar_entry->name) - 1] = '\0'; /* Advance to the next entry */ - cur_offset = map_size_needed + mokvar_entry->data_size; + cur_offset += sizeof(*mokvar_entry) + mokvar_entry->data_size; } if (va) - early_memunmap(va, map_size); + early_memunmap(va, sizeof(*mokvar_entry)); if (err) { pr_err("EFI MOKvar config table is not valid\n"); return;

4 months

1
0
0 0

[PATCH v3 3/3] drm/xe/userptr: Unmap userptrs in the mmu notifier

by Thomas Hellström

If userptr pages are freed after a call to the xe mmu notifier, the device will not be blocked out from theoretically accessing these pages unless they are also unmapped from the iommu, and this violates some aspects of the iommu-imposed security. Ensure that userptrs are unmapped in the mmu notifier to mitigate this. A naive attempt would try to free the sg table, but the sg table itself may be accessed by a concurrent bind operation, so settle for only unmapping. v3: - Update lockdep asserts. - Fix a typo (Matthew Auld) Fixes: 81e058a3e7fd ("drm/xe: Introduce helper to populate userptr") Cc: Oak Zeng <oak.zeng(a)intel.com> Cc: Matthew Auld <matthew.auld(a)intel.com> Cc: <stable(a)vger.kernel.org> # v6.10+ Signed-off-by: Thomas Hellström <thomas.hellstrom(a)linux.intel.com> Reviewed-by: Matthew Auld <matthew.auld(a)intel.com> --- drivers/gpu/drm/xe/xe_hmm.c | 51 ++++++++++++++++++++++++++------ drivers/gpu/drm/xe/xe_hmm.h | 2 ++ drivers/gpu/drm/xe/xe_vm.c | 4 +++ drivers/gpu/drm/xe/xe_vm_types.h | 4 +++ 4 files changed, 52 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_hmm.c b/drivers/gpu/drm/xe/xe_hmm.c index d33dd7bc3f41..68a5bc1cf77b 100644 --- a/drivers/gpu/drm/xe/xe_hmm.c +++ b/drivers/gpu/drm/xe/xe_hmm.c @@ -150,6 +150,45 @@ static int xe_build_sg(struct xe_device *xe, struct hmm_range *range, DMA_ATTR_SKIP_CPU_SYNC | DMA_ATTR_NO_KERNEL_MAPPING); } +static void xe_hmm_userptr_set_mapped(struct xe_userptr_vma *uvma) +{ + struct xe_userptr *userptr = &uvma->userptr; + struct xe_vm *vm = xe_vma_vm(&uvma->vma); + + lockdep_assert_held_write(&vm->lock); + lockdep_assert_held(&vm->userptr.notifier_lock); + + mutex_lock(&userptr->unmap_mutex); + xe_assert(vm->xe, !userptr->mapped); + userptr->mapped = true; + mutex_unlock(&userptr->unmap_mutex); +} + +void xe_hmm_userptr_unmap(struct xe_userptr_vma *uvma) +{ + struct xe_userptr *userptr = &uvma->userptr; + struct xe_vma *vma = &uvma->vma; + bool write = !xe_vma_read_only(vma); + struct xe_vm *vm = xe_vma_vm(vma); + struct xe_device *xe = vm->xe; + + if (!lockdep_is_held_type(&vm->userptr.notifier_lock, 0) && + !lockdep_is_held_type(&vm->lock, 0) && + !(vma->gpuva.flags & XE_VMA_DESTROYED)) { + /* Don't unmap in exec critical section. */ + xe_vm_assert_held(vm); + /* Don't unmap while mapping the sg. */ + lockdep_assert_held(&vm->lock); + } + + mutex_lock(&userptr->unmap_mutex); + if (userptr->sg && userptr->mapped) + dma_unmap_sgtable(xe->drm.dev, userptr->sg, + write ? DMA_BIDIRECTIONAL : DMA_TO_DEVICE, 0); + userptr->mapped = false; + mutex_unlock(&userptr->unmap_mutex); +} + /** * xe_hmm_userptr_free_sg() - Free the scatter gather table of userptr * @uvma: the userptr vma which hold the scatter gather table @@ -161,16 +200,9 @@ static int xe_build_sg(struct xe_device *xe, struct hmm_range *range, void xe_hmm_userptr_free_sg(struct xe_userptr_vma *uvma) { struct xe_userptr *userptr = &uvma->userptr; - struct xe_vma *vma = &uvma->vma; - bool write = !xe_vma_read_only(vma); - struct xe_vm *vm = xe_vma_vm(vma); - struct xe_device *xe = vm->xe; - struct device *dev = xe->drm.dev; - - xe_assert(xe, userptr->sg); - dma_unmap_sgtable(dev, userptr->sg, - write ? DMA_BIDIRECTIONAL : DMA_TO_DEVICE, 0); + xe_assert(xe_vma_vm(&uvma->vma)->xe, userptr->sg); + xe_hmm_userptr_unmap(uvma); sg_free_table(userptr->sg); userptr->sg = NULL; } @@ -295,6 +327,7 @@ int xe_hmm_userptr_populate_range(struct xe_userptr_vma *uvma, xe_mark_range_accessed(&hmm_range, write); userptr->sg = &userptr->sgt; + xe_hmm_userptr_set_mapped(uvma); userptr->notifier_seq = hmm_range.notifier_seq; up_read(&vm->userptr.notifier_lock); kvfree(pfns); diff --git a/drivers/gpu/drm/xe/xe_hmm.h b/drivers/gpu/drm/xe/xe_hmm.h index 9602cb7d976d..0ea98d8e7bbc 100644 --- a/drivers/gpu/drm/xe/xe_hmm.h +++ b/drivers/gpu/drm/xe/xe_hmm.h @@ -13,4 +13,6 @@ struct xe_userptr_vma; int xe_hmm_userptr_populate_range(struct xe_userptr_vma *uvma, bool is_mm_mmap_locked); void xe_hmm_userptr_free_sg(struct xe_userptr_vma *uvma); + +void xe_hmm_userptr_unmap(struct xe_userptr_vma *uvma); #endif diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c index dd422ac95dc0..3dbd3d38008a 100644 --- a/drivers/gpu/drm/xe/xe_vm.c +++ b/drivers/gpu/drm/xe/xe_vm.c @@ -621,6 +621,8 @@ static void __vma_userptr_invalidate(struct xe_vm *vm, struct xe_userptr_vma *uv err = xe_vm_invalidate_vma(vma); XE_WARN_ON(err); } + + xe_hmm_userptr_unmap(uvma); } static bool vma_userptr_invalidate(struct mmu_interval_notifier *mni, @@ -1039,6 +1041,7 @@ static struct xe_vma *xe_vma_create(struct xe_vm *vm, INIT_LIST_HEAD(&userptr->invalidate_link); INIT_LIST_HEAD(&userptr->repin_link); vma->gpuva.gem.offset = bo_offset_or_userptr; + mutex_init(&userptr->unmap_mutex); err = mmu_interval_notifier_insert(&userptr->notifier, current->mm, @@ -1080,6 +1083,7 @@ static void xe_vma_destroy_late(struct xe_vma *vma) * them anymore */ mmu_interval_notifier_remove(&userptr->notifier); + mutex_destroy(&userptr->unmap_mutex); xe_vm_put(vm); } else if (xe_vma_is_null(vma)) { xe_vm_put(vm); diff --git a/drivers/gpu/drm/xe/xe_vm_types.h b/drivers/gpu/drm/xe/xe_vm_types.h index 1fe79bf23b6b..eca73c4197d4 100644 --- a/drivers/gpu/drm/xe/xe_vm_types.h +++ b/drivers/gpu/drm/xe/xe_vm_types.h @@ -59,12 +59,16 @@ struct xe_userptr { struct sg_table *sg; /** @notifier_seq: notifier sequence number */ unsigned long notifier_seq; + /** @unmap_mutex: Mutex protecting dma-unmapping */ + struct mutex unmap_mutex; /** * @initial_bind: user pointer has been bound at least once. * write: vm->userptr.notifier_lock in read mode and vm->resv held. * read: vm->userptr.notifier_lock in write mode or vm->resv held. */ bool initial_bind; + /** @mapped: Whether the @sgt sg-table is dma-mapped. Protected by @unmap_mutex. */ + bool mapped; #if IS_ENABLED(CONFIG_DRM_XE_USERPTR_INVAL_INJECT) u32 divisor; #endif -- 2.48.1

4 months

1
0
0 0

[PATCH v3 2/3] drm/xe/hmm: Don't dereference struct page pointers without notifier lock

by Thomas Hellström

The pnfs that we obtain from hmm_range_fault() point to pages that we don't have a reference on, and the guarantee that they are still in the cpu page-tables is that the notifier lock must be held and the notifier seqno is still valid. So while building the sg table and marking the pages accesses / dirty we need to hold this lock with a validated seqno. However, the lock is reclaim tainted which makes sg_alloc_table_from_pages_segment() unusable, since it internally allocates memory. Instead build the sg-table manually. For the non-iommu case this might lead to fewer coalesces, but if that's a problem it can be fixed up later in the resource cursor code. For the iommu case, the whole sg-table may still be coalesced to a single contigous device va region. This avoids marking pages that we don't own dirty and accessed, and it also avoid dereferencing struct pages that we don't own. v2: - Use assert to check whether hmm pfns are valid (Matthew Auld) - Take into account that large pages may cross range boundaries (Matthew Auld) v3: - Don't unnecessarily check for a non-freed sg-table. (Matthew Auld) - Add a missing up_read() in an error path. (Matthew Auld) Fixes: 81e058a3e7fd ("drm/xe: Introduce helper to populate userptr") Cc: Oak Zeng <oak.zeng(a)intel.com> Cc: <stable(a)vger.kernel.org> # v6.10+ Signed-off-by: Thomas Hellström <thomas.hellstrom(a)linux.intel.com> Reviewed-by: Matthew Auld <matthew.auld(a)intel.com> --- drivers/gpu/drm/xe/xe_hmm.c | 112 +++++++++++++++++++++++++++--------- 1 file changed, 86 insertions(+), 26 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_hmm.c b/drivers/gpu/drm/xe/xe_hmm.c index c56738fa713b..d33dd7bc3f41 100644 --- a/drivers/gpu/drm/xe/xe_hmm.c +++ b/drivers/gpu/drm/xe/xe_hmm.c @@ -42,6 +42,42 @@ static void xe_mark_range_accessed(struct hmm_range *range, bool write) } } +static int xe_alloc_sg(struct xe_device *xe, struct sg_table *st, + struct hmm_range *range, struct rw_semaphore *notifier_sem) +{ + unsigned long i, npages, hmm_pfn; + unsigned long num_chunks = 0; + int ret; + + /* HMM docs says this is needed. */ + ret = down_read_interruptible(notifier_sem); + if (ret) + return ret; + + if (mmu_interval_read_retry(range->notifier, range->notifier_seq)) { + up_read(notifier_sem); + return -EAGAIN; + } + + npages = xe_npages_in_range(range->start, range->end); + for (i = 0; i < npages;) { + unsigned long len; + + hmm_pfn = range->hmm_pfns[i]; + xe_assert(xe, hmm_pfn & HMM_PFN_VALID); + + len = 1UL << hmm_pfn_to_map_order(hmm_pfn); + + /* If order > 0 the page may extend beyond range->start */ + len -= (hmm_pfn & ~HMM_PFN_FLAGS) & (len - 1); + i += len; + num_chunks++; + } + up_read(notifier_sem); + + return sg_alloc_table(st, num_chunks, GFP_KERNEL); +} + /** * xe_build_sg() - build a scatter gather table for all the physical pages/pfn * in a hmm_range. dma-map pages if necessary. dma-address is save in sg table @@ -50,6 +86,7 @@ static void xe_mark_range_accessed(struct hmm_range *range, bool write) * @range: the hmm range that we build the sg table from. range->hmm_pfns[] * has the pfn numbers of pages that back up this hmm address range. * @st: pointer to the sg table. + * @notifier_sem: The xe notifier lock. * @write: whether we write to this range. This decides dma map direction * for system pages. If write we map it bi-diretional; otherwise * DMA_TO_DEVICE @@ -76,38 +113,41 @@ static void xe_mark_range_accessed(struct hmm_range *range, bool write) * Returns 0 if successful; -ENOMEM if fails to allocate memory */ static int xe_build_sg(struct xe_device *xe, struct hmm_range *range, - struct sg_table *st, bool write) + struct sg_table *st, + struct rw_semaphore *notifier_sem, + bool write) { + unsigned long npages = xe_npages_in_range(range->start, range->end); struct device *dev = xe->drm.dev; - struct page **pages; - u64 i, npages; - int ret; + struct scatterlist *sgl; + struct page *page; + unsigned long i, j; - npages = xe_npages_in_range(range->start, range->end); - pages = kvmalloc_array(npages, sizeof(*pages), GFP_KERNEL); - if (!pages) - return -ENOMEM; + lockdep_assert_held(notifier_sem); - for (i = 0; i < npages; i++) { - pages[i] = hmm_pfn_to_page(range->hmm_pfns[i]); - xe_assert(xe, !is_device_private_page(pages[i])); - } + i = 0; + for_each_sg(st->sgl, sgl, st->nents, j) { + unsigned long hmm_pfn, size; - ret = sg_alloc_table_from_pages_segment(st, pages, npages, 0, npages << PAGE_SHIFT, - xe_sg_segment_size(dev), GFP_KERNEL); - if (ret) - goto free_pages; + hmm_pfn = range->hmm_pfns[i]; + page = hmm_pfn_to_page(hmm_pfn); + xe_assert(xe, !is_device_private_page(page)); + + size = 1UL << hmm_pfn_to_map_order(hmm_pfn); + size -= page_to_pfn(page) & (size - 1); + i += size; - ret = dma_map_sgtable(dev, st, write ? DMA_BIDIRECTIONAL : DMA_TO_DEVICE, - DMA_ATTR_SKIP_CPU_SYNC | DMA_ATTR_NO_KERNEL_MAPPING); - if (ret) { - sg_free_table(st); - st = NULL; + if (unlikely(j == st->nents - 1)) { + if (i > npages) + size -= (i - npages); + sg_mark_end(sgl); + } + sg_set_page(sgl, page, size << PAGE_SHIFT, 0); } + xe_assert(xe, i == npages); -free_pages: - kvfree(pages); - return ret; + return dma_map_sgtable(dev, st, write ? DMA_BIDIRECTIONAL : DMA_TO_DEVICE, + DMA_ATTR_SKIP_CPU_SYNC | DMA_ATTR_NO_KERNEL_MAPPING); } /** @@ -235,16 +275,36 @@ int xe_hmm_userptr_populate_range(struct xe_userptr_vma *uvma, if (ret) goto free_pfns; - ret = xe_build_sg(vm->xe, &hmm_range, &userptr->sgt, write); + ret = xe_alloc_sg(vm->xe, &userptr->sgt, &hmm_range, &vm->userptr.notifier_lock); if (ret) goto free_pfns; + ret = down_read_interruptible(&vm->userptr.notifier_lock); + if (ret) + goto free_st; + + if (mmu_interval_read_retry(hmm_range.notifier, hmm_range.notifier_seq)) { + ret = -EAGAIN; + goto out_unlock; + } + + ret = xe_build_sg(vm->xe, &hmm_range, &userptr->sgt, + &vm->userptr.notifier_lock, write); + if (ret) + goto out_unlock; + xe_mark_range_accessed(&hmm_range, write); userptr->sg = &userptr->sgt; userptr->notifier_seq = hmm_range.notifier_seq; + up_read(&vm->userptr.notifier_lock); + kvfree(pfns); + return 0; +out_unlock: + up_read(&vm->userptr.notifier_lock); +free_st: + sg_free_table(&userptr->sgt); free_pfns: kvfree(pfns); return ret; } - -- 2.48.1

4 months

1
0
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror