diff options
Diffstat (limited to 'meta-v1000/recipes-kernel/linux/linux-yocto-4.14.71/1164-drm-amdkfd-Use-ref-count-to-prevent-kfd_process-dest.patch')
| -rw-r--r-- | meta-v1000/recipes-kernel/linux/linux-yocto-4.14.71/1164-drm-amdkfd-Use-ref-count-to-prevent-kfd_process-dest.patch | 309 |
1 files changed, 309 insertions, 0 deletions
diff --git a/meta-v1000/recipes-kernel/linux/linux-yocto-4.14.71/1164-drm-amdkfd-Use-ref-count-to-prevent-kfd_process-dest.patch b/meta-v1000/recipes-kernel/linux/linux-yocto-4.14.71/1164-drm-amdkfd-Use-ref-count-to-prevent-kfd_process-dest.patch new file mode 100644 index 00000000..1a595119 --- /dev/null +++ b/meta-v1000/recipes-kernel/linux/linux-yocto-4.14.71/1164-drm-amdkfd-Use-ref-count-to-prevent-kfd_process-dest.patch @@ -0,0 +1,309 @@ +From a3e6779469a3bb88b61226d87eb7055651a4a687 Mon Sep 17 00:00:00 2001 +From: Felix Kuehling <Felix.Kuehling@amd.com> +Date: Tue, 26 Jul 2016 17:41:11 -0400 +Subject: [PATCH 1164/4131] drm/amdkfd: Use ref count to prevent kfd_process + destruction + +Use a reference counter instead of a lock to prevent process +destruction while functions running out of process context are using +the kfd_process structure. In many cases these functions don't need +the structure to be locked. In the few cases that really do need the +process lock, take it explicitly. + +This helps simplify lock dependencies between the process lock and +other locks. + +Change-Id: Ie8191f5a9f503b576d8f1eed5f2b14eca804d6ac +Signed-off-by: Felix Kuehling <Felix.Kuehling@amd.com> +--- + drivers/gpu/drm/amd/amdkfd/kfd_device.c | 10 +++++---- + .../gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 2 +- + drivers/gpu/drm/amd/amdkfd/kfd_events.c | 19 ++++++++-------- + drivers/gpu/drm/amd/amdkfd/kfd_priv.h | 1 + + drivers/gpu/drm/amd/amdkfd/kfd_process.c | 26 +++++++++++++++------- + drivers/gpu/drm/amd/amdkfd/kfd_rdma.c | 6 +++-- + 6 files changed, 40 insertions(+), 24 deletions(-) + +diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c +index 0111510..6ac3ec6 100644 +--- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c ++++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c +@@ -675,7 +675,8 @@ int kgd2kfd_quiesce_mm(struct kfd_dev *kfd, struct mm_struct *mm) + + /* Because we are called from arbitrary context (workqueue) as opposed + * to process context, kfd_process could attempt to exit while we are +- * running so the lookup function returns a read-locked process. */ ++ * running so the lookup function increments the process ref count. ++ */ + p = kfd_lookup_process_by_mm(mm); + if (!p) + return -ENODEV; +@@ -685,7 +686,7 @@ int kgd2kfd_quiesce_mm(struct kfd_dev *kfd, struct mm_struct *mm) + if (pdd) + r = process_evict_queues(kfd->dqm, &pdd->qpd); + +- up_read(&p->lock); ++ kfd_unref_process(p); + return r; + } + +@@ -701,7 +702,8 @@ int kgd2kfd_resume_mm(struct kfd_dev *kfd, struct mm_struct *mm) + + /* Because we are called from arbitrary context (workqueue) as opposed + * to process context, kfd_process could attempt to exit while we are +- * running so the lookup function returns a read-locked process. */ ++ * running so the lookup function increments the process ref count. ++ */ + p = kfd_lookup_process_by_mm(mm); + if (!p) + return -ENODEV; +@@ -718,7 +720,7 @@ int kgd2kfd_resume_mm(struct kfd_dev *kfd, struct mm_struct *mm) + up_read(&mm->mmap_sem); + } + +- up_read(&p->lock); ++ kfd_unref_process(p); + return r; + } + +diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +index df9b3f3..5f1fbbe 100644 +--- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c ++++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +@@ -1581,7 +1581,7 @@ int kfd_process_vm_fault(struct device_queue_manager *dqm, + pdd = kfd_get_process_device_data(dqm->dev, p); + if (pdd) + ret = process_evict_queues(dqm, &pdd->qpd); +- up_read(&p->lock); ++ kfd_unref_process(p); + + return ret; + } +diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_events.c +index 5f7aa78..6a354f3 100644 +--- a/drivers/gpu/drm/amd/amdkfd/kfd_events.c ++++ b/drivers/gpu/drm/amd/amdkfd/kfd_events.c +@@ -661,7 +661,8 @@ void kfd_signal_event_interrupt(unsigned int pasid, uint32_t partial_id, + + /* Because we are called from arbitrary context (workqueue) as opposed + * to process context, kfd_process could attempt to exit while we are +- * running so the lookup function returns a read-locked process. */ ++ * running so the lookup function increments the process ref count. ++ */ + struct kfd_process *p = kfd_lookup_process_by_pasid(pasid); + if (!p) + return; /* Presumably process exited. */ +@@ -691,7 +692,7 @@ void kfd_signal_event_interrupt(unsigned int pasid, uint32_t partial_id, + } + + mutex_unlock(&p->event_mutex); +- up_read(&p->lock); ++ kfd_unref_process(p); + } + + static struct kfd_event_waiter *alloc_event_waiters(uint32_t num_events) +@@ -1024,7 +1025,7 @@ void kfd_signal_iommu_event(struct kfd_dev *dev, unsigned int pasid, + /* + * Because we are called from arbitrary context (workqueue) as opposed + * to process context, kfd_process could attempt to exit while we are +- * running so the lookup function returns a read-locked process. ++ * running so the lookup function increments the process ref count. + */ + struct kfd_process *p = kfd_lookup_process_by_pasid(pasid); + struct mm_struct *mm; +@@ -1037,7 +1038,7 @@ void kfd_signal_iommu_event(struct kfd_dev *dev, unsigned int pasid, + */ + mm = get_task_mm(p->lead_thread); + if (!mm) { +- up_read(&p->lock); ++ kfd_unref_process(p); + return; /* Process is exiting */ + } + +@@ -1080,7 +1081,8 @@ void kfd_signal_iommu_event(struct kfd_dev *dev, unsigned int pasid, + &memory_exception_data); + + mutex_unlock(&p->event_mutex); +- up_read(&p->lock); ++ ++ kfd_unref_process(p); + } + + void kfd_signal_hw_exception_event(unsigned int pasid) +@@ -1088,7 +1090,7 @@ void kfd_signal_hw_exception_event(unsigned int pasid) + /* + * Because we are called from arbitrary context (workqueue) as opposed + * to process context, kfd_process could attempt to exit while we are +- * running so the lookup function returns a read-locked process. ++ * running so the lookup function increments the process ref count. + */ + struct kfd_process *p = kfd_lookup_process_by_pasid(pasid); + +@@ -1101,7 +1103,7 @@ void kfd_signal_hw_exception_event(unsigned int pasid) + lookup_events_by_type_and_signal(p, KFD_EVENT_TYPE_HW_EXCEPTION, NULL); + + mutex_unlock(&p->event_mutex); +- up_read(&p->lock); ++ kfd_unref_process(p); + } + + void kfd_signal_vm_fault_event(struct kfd_dev *dev, unsigned int pasid, +@@ -1138,7 +1140,6 @@ void kfd_signal_vm_fault_event(struct kfd_dev *dev, unsigned int pasid, + } + + mutex_unlock(&p->event_mutex); +- up_read(&p->lock); +- ++ kfd_unref_process(p); + } + +diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +index cea4efd..420aa64 100644 +--- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h ++++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +@@ -685,6 +685,7 @@ struct kfd_process *kfd_create_process(struct file *filep); + struct kfd_process *kfd_get_process(const struct task_struct *); + struct kfd_process *kfd_lookup_process_by_pasid(unsigned int pasid); + struct kfd_process *kfd_lookup_process_by_mm(const struct mm_struct *mm); ++void kfd_unref_process(struct kfd_process *p); + + struct kfd_process_device *kfd_bind_process_to_device(struct kfd_dev *dev, + struct kfd_process *p); +diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c +index 889d165..03b701a 100644 +--- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c ++++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c +@@ -56,6 +56,7 @@ static struct workqueue_struct *kfd_process_wq; + + static struct kfd_process *find_process(const struct task_struct *thread, + bool lock); ++static void kfd_process_ref_release(struct kref *ref); + static struct kfd_process *create_process(const struct task_struct *thread); + static int kfd_process_init_cwsr(struct kfd_process *p, struct file *filep); + +@@ -260,21 +261,26 @@ static struct kfd_process *find_process_by_mm(const struct mm_struct *mm) + } + + static struct kfd_process *find_process(const struct task_struct *thread, +- bool lock) ++ bool ref) + { + struct kfd_process *p; + int idx; + + idx = srcu_read_lock(&kfd_processes_srcu); + p = find_process_by_mm(thread->mm); +- if (p && lock) +- down_read(&p->lock); ++ if (p && ref) ++ kref_get(&p->ref); + srcu_read_unlock(&kfd_processes_srcu, idx); + + return p; + } + +-/* This returns with process->lock read-locked. */ ++void kfd_unref_process(struct kfd_process *p) ++{ ++ kref_put(&p->ref, kfd_process_ref_release); ++} ++ ++/* This increments the process->ref counter. */ + struct kfd_process *kfd_lookup_process_by_pid(struct pid *pid) + { + struct task_struct *task = NULL; +@@ -853,6 +859,8 @@ void kfd_process_iommu_unbind_callback(struct kfd_dev *dev, unsigned int pasid) + + mutex_unlock(get_dbgmgr_mutex()); + ++ down_write(&p->lock); ++ + pdd = kfd_get_process_device_data(dev, p); + if (pdd->reset_wavefronts) { + dbgdev_wave_reset_wavefronts(pdd->dev, p); +@@ -860,6 +868,8 @@ void kfd_process_iommu_unbind_callback(struct kfd_dev *dev, unsigned int pasid) + } + + up_write(&p->lock); ++ ++ kfd_unref_process(p); + } + + struct kfd_process_device *kfd_get_first_process_device_data(struct kfd_process *p) +@@ -991,7 +1001,7 @@ void kfd_process_device_remove_obj_handle(struct kfd_process_device *pdd, + kfree(buf_obj); + } + +-/* This returns with process->lock read-locked. */ ++/* This increments the process->ref counter. */ + struct kfd_process *kfd_lookup_process_by_pasid(unsigned int pasid) + { + struct kfd_process *p, *ret_p = NULL; +@@ -1001,7 +1011,7 @@ struct kfd_process *kfd_lookup_process_by_pasid(unsigned int pasid) + + hash_for_each_rcu(kfd_processes_table, temp, p, kfd_processes) { + if (p->pasid == pasid) { +- down_read(&p->lock); ++ kref_get(&p->ref); + ret_p = p; + break; + } +@@ -1012,7 +1022,7 @@ struct kfd_process *kfd_lookup_process_by_pasid(unsigned int pasid) + return ret_p; + } + +-/* This returns with process->lock read-locked. */ ++/* This increments the process->ref counter. */ + struct kfd_process *kfd_lookup_process_by_mm(const struct mm_struct *mm) + { + struct kfd_process *p; +@@ -1021,7 +1031,7 @@ struct kfd_process *kfd_lookup_process_by_mm(const struct mm_struct *mm) + + p = find_process_by_mm(mm); + if (p != NULL) +- down_read(&p->lock); ++ kref_get(&p->ref); + + srcu_read_unlock(&kfd_processes_srcu, idx); + +diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_rdma.c b/drivers/gpu/drm/amd/amdkfd/kfd_rdma.c +index f58b57e..56bf9a2 100644 +--- a/drivers/gpu/drm/amd/amdkfd/kfd_rdma.c ++++ b/drivers/gpu/drm/amd/amdkfd/kfd_rdma.c +@@ -80,6 +80,7 @@ static int get_pages(uint64_t address, uint64_t length, struct pid *pid, + __func__); + return -EINVAL; + } ++ down_read(&p->lock); + + buf_obj = kfd_process_find_bo_from_interval(p, address, last); + if (!buf_obj) { +@@ -127,6 +128,7 @@ static int get_pages(uint64_t address, uint64_t length, struct pid *pid, + kfree(rdma_cb_data); + out: + up_read(&p->lock); ++ kfd_unref_process(p); + + return ret; + } +@@ -203,7 +205,7 @@ static int put_pages(struct amd_p2p_info **p_p2p_data) + if (!ret) + *p_p2p_data = NULL; + +- up_read(&p->lock); ++ kfd_unref_process(p); + + return ret; + } +@@ -232,7 +234,7 @@ static int is_gpu_address(uint64_t address, struct pid *pid) + + buf_obj = kfd_process_find_bo_from_interval(p, address, address); + +- up_read(&p->lock); ++ kfd_unref_process(p); + if (!buf_obj) + return 0; + else +-- +2.7.4 + |