March 2024 - Linux-stable-mirror

+ selftests-mm-sigbus-wp-test-requires-uffd_feature_wp_hugetlbfs_shmem.patch added to mm-hotfixes-unstable branch

by Andrew Morton

The patch titled Subject: selftests/mm: sigbus-wp test requires UFFD_FEATURE_WP_HUGETLBFS_SHMEM has been added to the -mm mm-hotfixes-unstable branch. Its filename is selftests-mm-sigbus-wp-test-requires-uffd_feature_wp_hugetlbfs_shmem.patch This patch will shortly appear at https://git.kernel.org/pub/scm/linux/kernel/git/akpm/25-new.git/tree/patche… This patch will later appear in the mm-hotfixes-unstable branch at git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm Before you just go and hit "reply", please: a) Consider who else should be cc'ed b) Prefer to cc a suitable mailing list as well c) Ideally: find the original patch on the mailing list and do a reply-to-all to that, adding suitable additional cc's *** Remember to use Documentation/process/submit-checklist.rst when testing your code *** The -mm tree is included into linux-next via the mm-everything branch at git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm and is updated there every 2-3 working days ------------------------------------------------------ From: Edward Liaw <edliaw(a)google.com> Subject: selftests/mm: sigbus-wp test requires UFFD_FEATURE_WP_HUGETLBFS_SHMEM Date: Thu, 21 Mar 2024 23:20:21 +0000 The sigbus-wp test requires the UFFD_FEATURE_WP_HUGETLBFS_SHMEM flag for shmem and hugetlb targets. Otherwise it is not backwards compatible with kernels <5.19 and fails with EINVAL. Link: https://lkml.kernel.org/r/20240321232023.2064975-1-edliaw@google.com Fixes: 73c1ea939b65 ("selftests/mm: move uffd sig/events tests into uffd unit tests") Signed-off-by: Edward Liaw <edliaw(a)google.com> Cc: Shuah Khan <shuah(a)kernel.org> Cc: Peter Xu <peterx(a)redhat.com Cc: <stable(a)vger.kernel.org> Signed-off-by: Andrew Morton <akpm(a)linux-foundation.org> --- tools/testing/selftests/mm/uffd-unit-tests.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) --- a/tools/testing/selftests/mm/uffd-unit-tests.c~selftests-mm-sigbus-wp-test-requires-uffd_feature_wp_hugetlbfs_shmem +++ a/tools/testing/selftests/mm/uffd-unit-tests.c @@ -1427,7 +1427,8 @@ uffd_test_case_t uffd_tests[] = { .uffd_fn = uffd_sigbus_wp_test, .mem_targets = MEM_ALL, .uffd_feature_required = UFFD_FEATURE_SIGBUS | - UFFD_FEATURE_EVENT_FORK | UFFD_FEATURE_PAGEFAULT_FLAG_WP, + UFFD_FEATURE_EVENT_FORK | UFFD_FEATURE_PAGEFAULT_FLAG_WP | + UFFD_FEATURE_WP_HUGETLBFS_SHMEM, }, { .name = "events", _ Patches currently in -mm which might be from edliaw(a)google.com are selftests-mm-sigbus-wp-test-requires-uffd_feature_wp_hugetlbfs_shmem.patch

11 months, 3 weeks

1
0
0 0

[PATCH v2] drm/i915/gt: Report full vm address range

by Andi Shyti

Commit 9bb66c179f50 ("drm/i915: Reserve some kernel space per vm") has reserved an object for kernel space usage. Userspace, though, needs to know the full address range. In the former patch the reserved space was substructed from the total amount of the VM space. Add it back when the user requests the GTT size through ioctl (I915_CONTEXT_PARAM_GTT_SIZE). Fixes: 9bb66c179f50 ("drm/i915: Reserve some kernel space per vm") Signed-off-by: Andi Shyti <andi.shyti(a)linux.intel.com> Cc: Andrzej Hajda <andrzej.hajda(a)intel.com> Cc: Chris Wilson <chris.p.wilson(a)linux.intel.com> Cc: Lionel Landwerlin <lionel.g.landwerlin(a)intel.com> Cc: Michal Mrozek <michal.mrozek(a)intel.com> Cc: Nirmoy Das <nirmoy.das(a)intel.com> Cc: <stable(a)vger.kernel.org> # v6.2+ Acked-by: Michal Mrozek <michal.mrozek(a)intel.com> Acked-by: Lionel Landwerlin <lionel.g.landwerlin(a)intel.com> --- Hi, Just proposing a different implementation that doesn't affect i915 internally but provides the same result. Instead of not substracting the space during the reservation, I add it back during the ioctl call. All the "vm->rsvd.vma->node.size" looks a bit ugly, but that's how it is. Maybe a comment can help to understand better why there is this addition. I kept the Ack from Michal and Lionel, because the outcome from userspace perspactive doesn't really change. Andi drivers/gpu/drm/i915/gem/i915_gem_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c index 81f65cab1330..60d9e7fe33b3 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_context.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c @@ -2454,7 +2454,7 @@ int i915_gem_context_getparam_ioctl(struct drm_device *dev, void *data, case I915_CONTEXT_PARAM_GTT_SIZE: args->size = 0; vm = i915_gem_context_get_eb_vm(ctx); - args->value = vm->total; + args->value = vm->total + vm->rsvd.vma->node.size; i915_vm_put(vm); break; -- 2.43.0

11 months, 3 weeks

2
1
0 0

[PATCH 2/7] drm/xe: Rework rebinding

by Thomas Hellström

Instead of handling the vm's rebind fence separately, which is error prone if they are not strictly ordered, attach rebind fences as kernel fences to the vm's resv. Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Cc: Rodrigo Vivi <rodrigo.vivi(a)intel.com> Cc: Matthew Brost <matthew.brost(a)intel.com> Cc: <stable(a)vger.kernel.org> # v6.8+ Signed-off-by: Thomas Hellström <thomas.hellstrom(a)linux.intel.com> --- drivers/gpu/drm/xe/xe_exec.c | 31 +++---------------------------- drivers/gpu/drm/xe/xe_pt.c | 2 +- drivers/gpu/drm/xe/xe_vm.c | 27 +++++++++------------------ drivers/gpu/drm/xe/xe_vm.h | 2 +- drivers/gpu/drm/xe/xe_vm_types.h | 3 --- 5 files changed, 14 insertions(+), 51 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_exec.c b/drivers/gpu/drm/xe/xe_exec.c index 7692ebfe7d47..759497d4a102 100644 --- a/drivers/gpu/drm/xe/xe_exec.c +++ b/drivers/gpu/drm/xe/xe_exec.c @@ -152,7 +152,6 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file) struct drm_exec *exec = &vm_exec.exec; u32 i, num_syncs = 0, num_ufence = 0; struct xe_sched_job *job; - struct dma_fence *rebind_fence; struct xe_vm *vm; bool write_locked, skip_retry = false; ktime_t end = 0; @@ -294,35 +293,11 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file) * Rebind any invalidated userptr or evicted BOs in the VM, non-compute * VM mode only. */ - rebind_fence = xe_vm_rebind(vm, false); - if (IS_ERR(rebind_fence)) { - err = PTR_ERR(rebind_fence); + err = xe_vm_rebind(vm, false); + if (err) goto err_put_job; - } - - /* - * We store the rebind_fence in the VM so subsequent execs don't get - * scheduled before the rebinds of userptrs / evicted BOs is complete. - */ - if (rebind_fence) { - dma_fence_put(vm->rebind_fence); - vm->rebind_fence = rebind_fence; - } - if (vm->rebind_fence) { - if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, - &vm->rebind_fence->flags)) { - dma_fence_put(vm->rebind_fence); - vm->rebind_fence = NULL; - } else { - dma_fence_get(vm->rebind_fence); - err = drm_sched_job_add_dependency(&job->drm, - vm->rebind_fence); - if (err) - goto err_put_job; - } - } - /* Wait behind munmap style rebinds */ + /* Wait behind rebinds */ if (!xe_vm_in_lr_mode(vm)) { err = drm_sched_job_add_resv_dependencies(&job->drm, xe_vm_resv(vm), diff --git a/drivers/gpu/drm/xe/xe_pt.c b/drivers/gpu/drm/xe/xe_pt.c index 21bc0d13fccf..0484ed5b495f 100644 --- a/drivers/gpu/drm/xe/xe_pt.c +++ b/drivers/gpu/drm/xe/xe_pt.c @@ -1298,7 +1298,7 @@ __xe_pt_bind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_exec_queue } /* add shared fence now for pagetable delayed destroy */ - dma_resv_add_fence(xe_vm_resv(vm), fence, !rebind && + dma_resv_add_fence(xe_vm_resv(vm), fence, rebind || last_munmap_rebind ? DMA_RESV_USAGE_KERNEL : DMA_RESV_USAGE_BOOKKEEP); diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c index 80d43d75b1da..35fba6e3f889 100644 --- a/drivers/gpu/drm/xe/xe_vm.c +++ b/drivers/gpu/drm/xe/xe_vm.c @@ -522,7 +522,6 @@ static void preempt_rebind_work_func(struct work_struct *w) { struct xe_vm *vm = container_of(w, struct xe_vm, preempt.rebind_work); struct drm_exec exec; - struct dma_fence *rebind_fence; unsigned int fence_count = 0; LIST_HEAD(preempt_fences); ktime_t end = 0; @@ -568,18 +567,11 @@ static void preempt_rebind_work_func(struct work_struct *w) if (err) goto out_unlock; - rebind_fence = xe_vm_rebind(vm, true); - if (IS_ERR(rebind_fence)) { - err = PTR_ERR(rebind_fence); + err = xe_vm_rebind(vm, true); + if (err) goto out_unlock; - } - - if (rebind_fence) { - dma_fence_wait(rebind_fence, false); - dma_fence_put(rebind_fence); - } - /* Wait on munmap style VM unbinds */ + /* Wait on rebinds and munmap style VM unbinds */ wait = dma_resv_wait_timeout(xe_vm_resv(vm), DMA_RESV_USAGE_KERNEL, false, MAX_SCHEDULE_TIMEOUT); @@ -773,14 +765,14 @@ xe_vm_bind_vma(struct xe_vma *vma, struct xe_exec_queue *q, struct xe_sync_entry *syncs, u32 num_syncs, bool first_op, bool last_op); -struct dma_fence *xe_vm_rebind(struct xe_vm *vm, bool rebind_worker) +int xe_vm_rebind(struct xe_vm *vm, bool rebind_worker) { - struct dma_fence *fence = NULL; + struct dma_fence *fence; struct xe_vma *vma, *next; lockdep_assert_held(&vm->lock); if (xe_vm_in_lr_mode(vm) && !rebind_worker) - return NULL; + return 0; xe_vm_assert_held(vm); list_for_each_entry_safe(vma, next, &vm->rebind_list, @@ -788,17 +780,17 @@ struct dma_fence *xe_vm_rebind(struct xe_vm *vm, bool rebind_worker) xe_assert(vm->xe, vma->tile_present); list_del_init(&vma->combined_links.rebind); - dma_fence_put(fence); if (rebind_worker) trace_xe_vma_rebind_worker(vma); else trace_xe_vma_rebind_exec(vma); fence = xe_vm_bind_vma(vma, NULL, NULL, 0, false, false); if (IS_ERR(fence)) - return fence; + return PTR_ERR(fence); + dma_fence_put(fence); } - return fence; + return 0; } static void xe_vma_free(struct xe_vma *vma) @@ -1588,7 +1580,6 @@ static void vm_destroy_work_func(struct work_struct *w) XE_WARN_ON(vm->pt_root[id]); trace_xe_vm_free(vm); - dma_fence_put(vm->rebind_fence); kfree(vm); } diff --git a/drivers/gpu/drm/xe/xe_vm.h b/drivers/gpu/drm/xe/xe_vm.h index 6df1f1c7f85d..4853354336f2 100644 --- a/drivers/gpu/drm/xe/xe_vm.h +++ b/drivers/gpu/drm/xe/xe_vm.h @@ -207,7 +207,7 @@ int __xe_vm_userptr_needs_repin(struct xe_vm *vm); int xe_vm_userptr_check_repin(struct xe_vm *vm); -struct dma_fence *xe_vm_rebind(struct xe_vm *vm, bool rebind_worker); +int xe_vm_rebind(struct xe_vm *vm, bool rebind_worker); int xe_vm_invalidate_vma(struct xe_vma *vma); diff --git a/drivers/gpu/drm/xe/xe_vm_types.h b/drivers/gpu/drm/xe/xe_vm_types.h index 5747f136d24d..badf3945083d 100644 --- a/drivers/gpu/drm/xe/xe_vm_types.h +++ b/drivers/gpu/drm/xe/xe_vm_types.h @@ -177,9 +177,6 @@ struct xe_vm { */ struct list_head rebind_list; - /** @rebind_fence: rebind fence from execbuf */ - struct dma_fence *rebind_fence; - /** * @destroy_work: worker to destroy VM, needed as a dma_fence signaling * from an irq context can be last put and the destroy needs to be able -- 2.44.0

11 months, 3 weeks

2
2
0 0

cherry-pick request into stable for: HID: lenovo: Add middleclick_workaround sysfs knob for cptkbd

by Mikhail Khvoinitsky

Hello, Please, add the following commit: 2814646f76f8518326964f12ff20aaee70ba154d HID: lenovo: Add middleclick_workaround sysfs knob for cptkbd Into stable versions: 5.10, 5.15, 6.1, 6.6, 6.7, 6.8 This versions received automatic backport of my previous attempts to detect properly-behaving patched firmware which suffer from false-positives causing regression for users of factory lenovo firmware. The commit adds an explicit sysfs control hence avoids any room for false positives and fixes the regression. Thanks.

11 months, 3 weeks

1
0
0 0

Re: efivarfs fixes without the commit being fixed in 6.1 and 6.6 (resending without html)

by Ted Brandston

Hi, this is my first time posting to a kernel list (third try, finally figured out the html-free -- sorry for the noise). I noticed that in the 6.6 kernel there's a fix commit from Ard [1] but not the commit it's fixing ("efivarfs: Add uid/gid mount options"). Same thing in 6.1 [2]. The commit being fixed doesn't appear until 6.7 [3]. I'm not familiar with this code so it's unclear to me if this might cause problems, but I figured I should point it out. [1]: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/fs/… [2]: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/log/fs/efi… [3]: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/log/fs/efi… Thanks!

11 months, 3 weeks

3
3
0 0

[PATCH] HID: i2c-hid: remove I2C_HID_READ_PENDING flag to prevent lock-up

by Nam Cao

The flag I2C_HID_READ_PENDING is used to serialize I2C operations. However, this is not necessary, because I2C core already has its own locking for that. More importantly, this flag can cause a lock-up: if the flag is set in i2c_hid_xfer() and an interrupt happens, the interrupt handler (i2c_hid_irq) will check this flag and return immediately without doing anything, then the interrupt handler will be invoked again in an infinite loop. Since interrupt handler is an RT task, it takes over the CPU and the flag-clearing task never gets scheduled, thus we have a lock-up. Delete this unnecessary flag. Reported-and-tested-by: Eva Kurchatova <nyandarknessgirl(a)gmail.com> Closes: https://lore.kernel.org/r/CA+eeCSPUDpUg76ZO8dszSbAGn+UHjcyv8F1J-CUPVARAzEtW… Fixes: 4a200c3b9a40 ("HID: i2c-hid: introduce HID over i2c specification implementation") Cc: <stable(a)vger.kernel.org> Signed-off-by: Nam Cao <namcao(a)linutronix.de> --- drivers/hid/i2c-hid/i2c-hid-core.c | 9 --------- 1 file changed, 9 deletions(-) diff --git a/drivers/hid/i2c-hid/i2c-hid-core.c b/drivers/hid/i2c-hid/i2c-hid-core.c index 2df1ab3c31cc..1c86c97688e9 100644 --- a/drivers/hid/i2c-hid/i2c-hid-core.c +++ b/drivers/hid/i2c-hid/i2c-hid-core.c @@ -64,7 +64,6 @@ /* flags */ #define I2C_HID_STARTED 0 #define I2C_HID_RESET_PENDING 1 -#define I2C_HID_READ_PENDING 2 #define I2C_HID_PWR_ON 0x00 #define I2C_HID_PWR_SLEEP 0x01 @@ -190,15 +189,10 @@ static int i2c_hid_xfer(struct i2c_hid *ihid, msgs[n].len = recv_len; msgs[n].buf = recv_buf; n++; - - set_bit(I2C_HID_READ_PENDING, &ihid->flags); } ret = i2c_transfer(client->adapter, msgs, n); - if (recv_len) - clear_bit(I2C_HID_READ_PENDING, &ihid->flags); - if (ret != n) return ret < 0 ? ret : -EIO; @@ -556,9 +550,6 @@ static irqreturn_t i2c_hid_irq(int irq, void *dev_id) { struct i2c_hid *ihid = dev_id; - if (test_bit(I2C_HID_READ_PENDING, &ihid->flags)) - return IRQ_HANDLED; - i2c_hid_get_input(ihid); return IRQ_HANDLED; -- 2.39.2

11 months, 3 weeks

2
1
0 0

[PATCH 4/7] drm/xe: Rework rebinding

by Thomas Hellström

Instead of handling the vm's rebind fence separately, which is error prone if they are not strictly ordered, attach rebind fences as kernel fences to the vm's resv. Fixes: dd08ebf6c352 ("drm/xe: Introduce a new DRM driver for Intel GPUs") Cc: Rodrigo Vivi <rodrigo.vivi(a)intel.com> Cc: Matthew Brost <matthew.brost(a)intel.com> Cc: <stable(a)vger.kernel.org> # v6.8+ Signed-off-by: Thomas Hellström <thomas.hellstrom(a)linux.intel.com> --- drivers/gpu/drm/xe/xe_exec.c | 31 +++---------------------------- drivers/gpu/drm/xe/xe_pt.c | 2 +- drivers/gpu/drm/xe/xe_vm.c | 27 +++++++++------------------ drivers/gpu/drm/xe/xe_vm.h | 2 +- drivers/gpu/drm/xe/xe_vm_types.h | 3 --- 5 files changed, 14 insertions(+), 51 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_exec.c b/drivers/gpu/drm/xe/xe_exec.c index 511d23426a5e..397a49b731f1 100644 --- a/drivers/gpu/drm/xe/xe_exec.c +++ b/drivers/gpu/drm/xe/xe_exec.c @@ -127,7 +127,6 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file) struct drm_exec *exec = &vm_exec.exec; u32 i, num_syncs = 0, num_ufence = 0; struct xe_sched_job *job; - struct dma_fence *rebind_fence; struct xe_vm *vm; bool write_locked, skip_retry = false; ktime_t end = 0; @@ -269,35 +268,11 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file) * Rebind any invalidated userptr or evicted BOs in the VM, non-compute * VM mode only. */ - rebind_fence = xe_vm_rebind(vm, false); - if (IS_ERR(rebind_fence)) { - err = PTR_ERR(rebind_fence); + err = xe_vm_rebind(vm, false); + if (err) goto err_put_job; - } - - /* - * We store the rebind_fence in the VM so subsequent execs don't get - * scheduled before the rebinds of userptrs / evicted BOs is complete. - */ - if (rebind_fence) { - dma_fence_put(vm->rebind_fence); - vm->rebind_fence = rebind_fence; - } - if (vm->rebind_fence) { - if (test_bit(DMA_FENCE_FLAG_SIGNALED_BIT, - &vm->rebind_fence->flags)) { - dma_fence_put(vm->rebind_fence); - vm->rebind_fence = NULL; - } else { - dma_fence_get(vm->rebind_fence); - err = drm_sched_job_add_dependency(&job->drm, - vm->rebind_fence); - if (err) - goto err_put_job; - } - } - /* Wait behind munmap style rebinds */ + /* Wait behind rebinds */ if (!xe_vm_in_lr_mode(vm)) { err = drm_sched_job_add_resv_dependencies(&job->drm, xe_vm_resv(vm), diff --git a/drivers/gpu/drm/xe/xe_pt.c b/drivers/gpu/drm/xe/xe_pt.c index 7add08b2954e..92cd660fbcef 100644 --- a/drivers/gpu/drm/xe/xe_pt.c +++ b/drivers/gpu/drm/xe/xe_pt.c @@ -1304,7 +1304,7 @@ __xe_pt_bind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_exec_queue } /* add shared fence now for pagetable delayed destroy */ - dma_resv_add_fence(xe_vm_resv(vm), fence, !rebind && + dma_resv_add_fence(xe_vm_resv(vm), fence, rebind || last_munmap_rebind ? DMA_RESV_USAGE_KERNEL : DMA_RESV_USAGE_BOOKKEEP); diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c index cff8db605743..33a85e702e2c 100644 --- a/drivers/gpu/drm/xe/xe_vm.c +++ b/drivers/gpu/drm/xe/xe_vm.c @@ -529,7 +529,6 @@ static void preempt_rebind_work_func(struct work_struct *w) { struct xe_vm *vm = container_of(w, struct xe_vm, preempt.rebind_work); struct drm_exec exec; - struct dma_fence *rebind_fence; unsigned int fence_count = 0; LIST_HEAD(preempt_fences); ktime_t end = 0; @@ -575,18 +574,11 @@ static void preempt_rebind_work_func(struct work_struct *w) if (err) goto out_unlock; - rebind_fence = xe_vm_rebind(vm, true); - if (IS_ERR(rebind_fence)) { - err = PTR_ERR(rebind_fence); + err = xe_vm_rebind(vm, true); + if (err) goto out_unlock; - } - - if (rebind_fence) { - dma_fence_wait(rebind_fence, false); - dma_fence_put(rebind_fence); - } - /* Wait on munmap style VM unbinds */ + /* Wait on rebinds and munmap style VM unbinds */ wait = dma_resv_wait_timeout(xe_vm_resv(vm), DMA_RESV_USAGE_KERNEL, false, MAX_SCHEDULE_TIMEOUT); @@ -780,14 +772,14 @@ xe_vm_bind_vma(struct xe_vma *vma, struct xe_exec_queue *q, struct xe_sync_entry *syncs, u32 num_syncs, bool first_op, bool last_op); -struct dma_fence *xe_vm_rebind(struct xe_vm *vm, bool rebind_worker) +int xe_vm_rebind(struct xe_vm *vm, bool rebind_worker) { - struct dma_fence *fence = NULL; + struct dma_fence *fence; struct xe_vma *vma, *next; lockdep_assert_held(&vm->lock); if (xe_vm_in_lr_mode(vm) && !rebind_worker) - return NULL; + return 0; xe_vm_assert_held(vm); list_for_each_entry_safe(vma, next, &vm->rebind_list, @@ -795,17 +787,17 @@ struct dma_fence *xe_vm_rebind(struct xe_vm *vm, bool rebind_worker) xe_assert(vm->xe, vma->tile_present); list_del_init(&vma->combined_links.rebind); - dma_fence_put(fence); if (rebind_worker) trace_xe_vma_rebind_worker(vma); else trace_xe_vma_rebind_exec(vma); fence = xe_vm_bind_vma(vma, NULL, NULL, 0, false, false); if (IS_ERR(fence)) - return fence; + return PTR_ERR(fence); + dma_fence_put(fence); } - return fence; + return 0; } static void xe_vma_free(struct xe_vma *vma) @@ -1586,7 +1578,6 @@ static void vm_destroy_work_func(struct work_struct *w) XE_WARN_ON(vm->pt_root[id]); trace_xe_vm_free(vm); - dma_fence_put(vm->rebind_fence); kfree(vm); } diff --git a/drivers/gpu/drm/xe/xe_vm.h b/drivers/gpu/drm/xe/xe_vm.h index 47f3960120a0..20009d8b4702 100644 --- a/drivers/gpu/drm/xe/xe_vm.h +++ b/drivers/gpu/drm/xe/xe_vm.h @@ -207,7 +207,7 @@ int __xe_vm_userptr_needs_repin(struct xe_vm *vm); int xe_vm_userptr_check_repin(struct xe_vm *vm); -struct dma_fence *xe_vm_rebind(struct xe_vm *vm, bool rebind_worker); +int xe_vm_rebind(struct xe_vm *vm, bool rebind_worker); int xe_vm_invalidate_vma(struct xe_vma *vma); diff --git a/drivers/gpu/drm/xe/xe_vm_types.h b/drivers/gpu/drm/xe/xe_vm_types.h index 5747f136d24d..badf3945083d 100644 --- a/drivers/gpu/drm/xe/xe_vm_types.h +++ b/drivers/gpu/drm/xe/xe_vm_types.h @@ -177,9 +177,6 @@ struct xe_vm { */ struct list_head rebind_list; - /** @rebind_fence: rebind fence from execbuf */ - struct dma_fence *rebind_fence; - /** * @destroy_work: worker to destroy VM, needed as a dma_fence signaling * from an irq context can be last put and the destroy needs to be able -- 2.44.0

11 months, 3 weeks

1
0
0 0

[PATCH 2/7] drm/xe: Use ring ops TLB invalidation for rebinds

by Thomas Hellström

For each rebind we insert a GuC TLB invalidation and add a corresponding unordered TLB invalidation fence. This might add a huge number of TLB invalidation fences to wait for so rather than doing that, defer the TLB invalidation to the next ring ops for each affected exec queue. Since the TLB is invalidated on exec_queue switch, we need to invalidate once for each affected exec_queue. Fixes: 5387e865d90e ("drm/xe: Add TLB invalidation fence after rebinds issued from execs") Cc: Matthew Brost <matthew.brost(a)intel.com> Cc: <stable(a)vger.kernel.org> # v6.8+ Signed-off-by: Thomas Hellström <thomas.hellstrom(a)linux.intel.com> --- drivers/gpu/drm/xe/xe_exec_queue_types.h | 2 ++ drivers/gpu/drm/xe/xe_pt.c | 5 +++-- drivers/gpu/drm/xe/xe_ring_ops.c | 11 ++++------- drivers/gpu/drm/xe/xe_sched_job.c | 11 +++++++++++ drivers/gpu/drm/xe/xe_sched_job_types.h | 2 ++ drivers/gpu/drm/xe/xe_vm_types.h | 5 +++++ 6 files changed, 27 insertions(+), 9 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_exec_queue_types.h b/drivers/gpu/drm/xe/xe_exec_queue_types.h index 62b3d9d1d7cd..891ad30e906f 100644 --- a/drivers/gpu/drm/xe/xe_exec_queue_types.h +++ b/drivers/gpu/drm/xe/xe_exec_queue_types.h @@ -148,6 +148,8 @@ struct xe_exec_queue { const struct xe_ring_ops *ring_ops; /** @entity: DRM sched entity for this exec queue (1 to 1 relationship) */ struct drm_sched_entity *entity; + /** @tlb_flush_seqno: The seqno of the last rebind tlb flush performed */ + u64 tlb_flush_seqno; /** @lrc: logical ring context for this exec queue */ struct xe_lrc lrc[]; }; diff --git a/drivers/gpu/drm/xe/xe_pt.c b/drivers/gpu/drm/xe/xe_pt.c index a5bb8d20f815..109c26e5689e 100644 --- a/drivers/gpu/drm/xe/xe_pt.c +++ b/drivers/gpu/drm/xe/xe_pt.c @@ -1253,11 +1253,12 @@ __xe_pt_bind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_exec_queue * non-faulting LR, in particular on user-space batch buffer chaining, * it needs to be done here. */ - if ((rebind && !xe_vm_in_lr_mode(vm) && !vm->batch_invalidate_tlb) || - (!rebind && xe_vm_has_scratch(vm) && xe_vm_in_preempt_fence_mode(vm))) { + if ((!rebind && xe_vm_has_scratch(vm) && xe_vm_in_preempt_fence_mode(vm))) { ifence = kzalloc(sizeof(*ifence), GFP_KERNEL); if (!ifence) return ERR_PTR(-ENOMEM); + } else if (rebind && !xe_vm_in_lr_mode(vm) && !vm->batch_invalidate_tlb) { + vm->tlb_flush_seqno++; } rfence = kzalloc(sizeof(*rfence), GFP_KERNEL); diff --git a/drivers/gpu/drm/xe/xe_ring_ops.c b/drivers/gpu/drm/xe/xe_ring_ops.c index c4edffcd4a32..5b2b37b59813 100644 --- a/drivers/gpu/drm/xe/xe_ring_ops.c +++ b/drivers/gpu/drm/xe/xe_ring_ops.c @@ -219,10 +219,9 @@ static void __emit_job_gen12_simple(struct xe_sched_job *job, struct xe_lrc *lrc { u32 dw[MAX_JOB_SIZE_DW], i = 0; u32 ppgtt_flag = get_ppgtt_flag(job); - struct xe_vm *vm = job->q->vm; struct xe_gt *gt = job->q->gt; - if (vm && vm->batch_invalidate_tlb) { + if (job->ring_ops_flush_tlb) { dw[i++] = preparser_disable(true); i = emit_flush_imm_ggtt(xe_lrc_start_seqno_ggtt_addr(lrc), seqno, true, dw, i); @@ -270,7 +269,6 @@ static void __emit_job_gen12_video(struct xe_sched_job *job, struct xe_lrc *lrc, struct xe_gt *gt = job->q->gt; struct xe_device *xe = gt_to_xe(gt); bool decode = job->q->class == XE_ENGINE_CLASS_VIDEO_DECODE; - struct xe_vm *vm = job->q->vm; dw[i++] = preparser_disable(true); @@ -282,13 +280,13 @@ static void __emit_job_gen12_video(struct xe_sched_job *job, struct xe_lrc *lrc, i = emit_aux_table_inv(gt, VE0_AUX_INV, dw, i); } - if (vm && vm->batch_invalidate_tlb) + if (job->ring_ops_flush_tlb) i = emit_flush_imm_ggtt(xe_lrc_start_seqno_ggtt_addr(lrc), seqno, true, dw, i); dw[i++] = preparser_disable(false); - if (!vm || !vm->batch_invalidate_tlb) + if (!job->ring_ops_flush_tlb) i = emit_store_imm_ggtt(xe_lrc_start_seqno_ggtt_addr(lrc), seqno, dw, i); @@ -317,7 +315,6 @@ static void __emit_job_gen12_render_compute(struct xe_sched_job *job, struct xe_gt *gt = job->q->gt; struct xe_device *xe = gt_to_xe(gt); bool lacks_render = !(gt->info.engine_mask & XE_HW_ENGINE_RCS_MASK); - struct xe_vm *vm = job->q->vm; u32 mask_flags = 0; dw[i++] = preparser_disable(true); @@ -327,7 +324,7 @@ static void __emit_job_gen12_render_compute(struct xe_sched_job *job, mask_flags = PIPE_CONTROL_3D_ENGINE_FLAGS; /* See __xe_pt_bind_vma() for a discussion on TLB invalidations. */ - i = emit_pipe_invalidate(mask_flags, vm && vm->batch_invalidate_tlb, dw, i); + i = emit_pipe_invalidate(mask_flags, job->ring_ops_flush_tlb, dw, i); /* hsdes: 1809175790 */ if (has_aux_ccs(xe)) diff --git a/drivers/gpu/drm/xe/xe_sched_job.c b/drivers/gpu/drm/xe/xe_sched_job.c index 8151ddafb940..d55458d915a9 100644 --- a/drivers/gpu/drm/xe/xe_sched_job.c +++ b/drivers/gpu/drm/xe/xe_sched_job.c @@ -250,6 +250,17 @@ bool xe_sched_job_completed(struct xe_sched_job *job) void xe_sched_job_arm(struct xe_sched_job *job) { + struct xe_exec_queue *q = job->q; + struct xe_vm *vm = q->vm; + + if (vm && !xe_sched_job_is_migration(q) && !xe_vm_in_lr_mode(vm) && + vm->tlb_flush_seqno != q->tlb_flush_seqno) { + q->tlb_flush_seqno = vm->tlb_flush_seqno; + job->ring_ops_flush_tlb = true; + } else if (vm && vm->batch_invalidate_tlb) { + job->ring_ops_flush_tlb = true; + } + drm_sched_job_arm(&job->drm); } diff --git a/drivers/gpu/drm/xe/xe_sched_job_types.h b/drivers/gpu/drm/xe/xe_sched_job_types.h index b1d83da50a53..5e12724219fd 100644 --- a/drivers/gpu/drm/xe/xe_sched_job_types.h +++ b/drivers/gpu/drm/xe/xe_sched_job_types.h @@ -39,6 +39,8 @@ struct xe_sched_job { } user_fence; /** @migrate_flush_flags: Additional flush flags for migration jobs */ u32 migrate_flush_flags; + /** @ring_ops_flush_tlb: The ring ops need to flush TLB before payload. */ + bool ring_ops_flush_tlb; /** @batch_addr: batch buffer address of job */ u64 batch_addr[]; }; diff --git a/drivers/gpu/drm/xe/xe_vm_types.h b/drivers/gpu/drm/xe/xe_vm_types.h index ae5fb565f6bf..5747f136d24d 100644 --- a/drivers/gpu/drm/xe/xe_vm_types.h +++ b/drivers/gpu/drm/xe/xe_vm_types.h @@ -264,6 +264,11 @@ struct xe_vm { bool capture_once; } error_capture; + /** + * @tlb_flush_seqno: Required TLB flush seqno for the next exec. + * protected by the vm resv. + */ + u64 tlb_flush_seqno; /** @batch_invalidate_tlb: Always invalidate TLB before batch start */ bool batch_invalidate_tlb; /** @xef: XE file handle for tracking this VM's drm client */ -- 2.44.0

11 months, 3 weeks

1
0
0 0

[PATCH 1/7] drm/xe: Make TLB invalidation fences unordered

by Thomas Hellström

They can actually complete out-of-order, so allocate a unique fence context for each fence. Fixes: 5387e865d90e ("drm/xe: Add TLB invalidation fence after rebinds issued from execs") Cc: Matthew Brost <matthew.brost(a)intel.com> Cc: <stable(a)vger.kernel.org> # v6.8+ Signed-off-by: Thomas Hellström <thomas.hellstrom(a)linux.intel.com> --- drivers/gpu/drm/xe/xe_gt_tlb_invalidation.c | 1 - drivers/gpu/drm/xe/xe_gt_types.h | 7 ------- drivers/gpu/drm/xe/xe_pt.c | 3 +-- 3 files changed, 1 insertion(+), 10 deletions(-) diff --git a/drivers/gpu/drm/xe/xe_gt_tlb_invalidation.c b/drivers/gpu/drm/xe/xe_gt_tlb_invalidation.c index a3c4ffba679d..787cba5e49a1 100644 --- a/drivers/gpu/drm/xe/xe_gt_tlb_invalidation.c +++ b/drivers/gpu/drm/xe/xe_gt_tlb_invalidation.c @@ -63,7 +63,6 @@ int xe_gt_tlb_invalidation_init(struct xe_gt *gt) INIT_LIST_HEAD(&gt->tlb_invalidation.pending_fences); spin_lock_init(&gt->tlb_invalidation.pending_lock); spin_lock_init(&gt->tlb_invalidation.lock); - gt->tlb_invalidation.fence_context = dma_fence_context_alloc(1); INIT_DELAYED_WORK(&gt->tlb_invalidation.fence_tdr, xe_gt_tlb_fence_timeout); diff --git a/drivers/gpu/drm/xe/xe_gt_types.h b/drivers/gpu/drm/xe/xe_gt_types.h index f6da2ad9719f..2143dffcaf11 100644 --- a/drivers/gpu/drm/xe/xe_gt_types.h +++ b/drivers/gpu/drm/xe/xe_gt_types.h @@ -179,13 +179,6 @@ struct xe_gt { * xe_gt_tlb_fence_timeout after the timeut interval is over. */ struct delayed_work fence_tdr; - /** @tlb_invalidation.fence_context: context for TLB invalidation fences */ - u64 fence_context; - /** - * @tlb_invalidation.fence_seqno: seqno to TLB invalidation fences, protected by - * tlb_invalidation.lock - */ - u32 fence_seqno; /** @tlb_invalidation.lock: protects TLB invalidation fences */ spinlock_t lock; } tlb_invalidation; diff --git a/drivers/gpu/drm/xe/xe_pt.c b/drivers/gpu/drm/xe/xe_pt.c index 8d3922d2206e..a5bb8d20f815 100644 --- a/drivers/gpu/drm/xe/xe_pt.c +++ b/drivers/gpu/drm/xe/xe_pt.c @@ -1135,8 +1135,7 @@ static int invalidation_fence_init(struct xe_gt *gt, spin_lock_irq(&gt->tlb_invalidation.lock); dma_fence_init(&ifence->base.base, &invalidation_fence_ops, &gt->tlb_invalidation.lock, - gt->tlb_invalidation.fence_context, - ++gt->tlb_invalidation.fence_seqno); + dma_fence_context_alloc(1), 1); spin_unlock_irq(&gt->tlb_invalidation.lock); INIT_LIST_HEAD(&ifence->base.link); -- 2.44.0

11 months, 3 weeks

1
0
0 0

[PATCH] drm/xe/query: fix gt_id bounds check

by Matthew Auld

The user provided gt_id should always be less than the XE_MAX_GT_PER_TILE. Fixes: 7793d00d1bf5 ("drm/xe: Correlate engine and cpu timestamps with better accuracy") Signed-off-by: Matthew Auld <matthew.auld(a)intel.com> Cc: Nirmoy Das <nirmoy.das(a)intel.com> Cc: <stable(a)vger.kernel.org> # v6.8+ --- drivers/gpu/drm/xe/xe_query.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/gpu/drm/xe/xe_query.c b/drivers/gpu/drm/xe/xe_query.c index fcd8680d2ccc..df407d73e5f5 100644 --- a/drivers/gpu/drm/xe/xe_query.c +++ b/drivers/gpu/drm/xe/xe_query.c @@ -133,7 +133,7 @@ query_engine_cycles(struct xe_device *xe, return -EINVAL; eci = &resp.eci; - if (eci->gt_id > XE_MAX_GT_PER_TILE) + if (eci->gt_id >= XE_MAX_GT_PER_TILE) return -EINVAL; gt = xe_device_get_gt(xe, eci->gt_id); -- 2.44.0

11 months, 3 weeks

2
1
0 0

2025

2024

2023

2022

2021

2020

2019

2018

2017

Linux-stable-mirror March 2024