drm/msm: de-struct_mutex-ification
Subsystems (6)
cover drm/msm: de-struct_mutex-ification
From: Rob Clark <[email protected]>
This doesn't remove *all* the struct_mutex, but it covers the worst of it, ie. shrinker/madvise/free/retire. The submit path still uses struct_mutex, but it still needs *something* serialize a portion of the submit path, and lock_stat mostly just shows the lock contention there being with other submits. And there are a few other bits of struct_mutex usage in less critical paths (debugfs, etc). But this seems like a reasonable step in the right direction.
v2: teach lockdep about shrinker locking patters (danvet) and
convert to obj->resv locking (danvet)
Rob Clark (22):
drm/msm/gem: Add obj->lock wrappers
drm/msm/gem: Rename internal get_iova_locked helper drm/msm/gem: Move prototypes to msm_gem.h drm/msm/gem: Add some _locked() helpers drm/msm/gem: Move locking in shrinker path drm/msm/submit: Move copy_from_user ahead of locking bos drm/msm: Do rpm get sooner in the submit path drm/msm/gem: Switch over to obj->resv for locking drm/msm: Use correct drm_gem_object_put() in fail case drm/msm: Drop chatty trace
drm/msm: Move update_fences()
drm/msm: Add priv->mm_lock to protect active/inactive lists drm/msm: Document and rename preempt_lock drm/msm: Protect ring->submits with it's own lock drm/msm: Refcount submits
drm/msm: Remove obj->gpu
drm/msm: Drop struct_mutex from the retire path drm/msm: Drop struct_mutex in free_object() path drm/msm: remove msm_gem_free_work
drm/msm: drop struct_mutex in madvise path drm/msm: Drop struct_mutex in shrinker path drm/msm: Don't implicit-sync if only a single ring
drivers/gpu/drm/msm/adreno/a5xx_gpu.c | 4 +- drivers/gpu/drm/msm/adreno/a5xx_preempt.c | 12 +- drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 4 +- drivers/gpu/drm/msm/disp/mdp4/mdp4_crtc.c | 1 + drivers/gpu/drm/msm/disp/mdp5/mdp5_crtc.c | 1 + drivers/gpu/drm/msm/dsi/dsi_host.c | 1 + drivers/gpu/drm/msm/msm_debugfs.c | 7 + drivers/gpu/drm/msm/msm_drv.c | 21 +- drivers/gpu/drm/msm/msm_drv.h | 73 ++----- drivers/gpu/drm/msm/msm_fbdev.c | 1 + drivers/gpu/drm/msm/msm_gem.c | 245 ++++++++++------------ drivers/gpu/drm/msm/msm_gem.h | 131 ++++++++++-- drivers/gpu/drm/msm/msm_gem_shrinker.c | 81 +++---- drivers/gpu/drm/msm/msm_gem_submit.c | 154 +++++++++----- drivers/gpu/drm/msm/msm_gpu.c | 98 +++++---- drivers/gpu/drm/msm/msm_gpu.h | 5 +- drivers/gpu/drm/msm/msm_ringbuffer.c | 3 +- drivers/gpu/drm/msm/msm_ringbuffer.h | 13 +- 18 files changed, 459 insertions(+), 396 deletions(-)
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 1/22 drm/msm/gem: Add obj->lock wrappers
From: Rob Clark <[email protected]>
This will make it easier to transition over to obj->resv locking for everything that is per-bo locking.
--- drivers/gpu/drm/msm/msm_gem.c | 99 ++++++++++++++++------------------- drivers/gpu/drm/msm/msm_gem.h | 28 ++++++++++ 2 files changed, 74 insertions(+), 53 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +46 -53
@@ -178,15 +178,15 @@ struct page **msm_gem_get_pages(struct drm_gem_object *obj)struct msm_gem_object *msm_obj =to_msm_bo (obj);struct page **p; -mutex_lock (&msm_obj->lock); +msm_gem_lock (obj);if (WARN_ON (msm_obj->madv !=MSM_MADV_WILLNEED )) { -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);return ERR_PTR (-EBUSY ); } p =get_pages (obj); -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);return p; }
@@ -252,14 +252,14 @@ vm_fault_t msm_gem_fault(struct vm_fault *vmf)* vm_ops.open/drm_gem_mmap_obj and close get and put * a reference on obj. So, we dont need to hold one here. */ - err =mutex_lock_interruptible (&msm_obj->lock); + err =msm_gem_lock_interruptible (obj);if (err) { ret =VM_FAULT_NOPAGE ; goto out; }if (WARN_ON (msm_obj->madv !=MSM_MADV_WILLNEED )) { -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);return VM_FAULT_SIGBUS ; }
@@ -280,7 +280,7 @@ vm_fault_t msm_gem_fault(struct vm_fault *vmf) ret =vmf_insert_mixed (vma, vmf->address,__pfn_to_pfn_t (pfn,PFN_DEV )); out_unlock: -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj); out:return ret; }
@@ -289,10 +289,9 @@ vm_fault_t msm_gem_fault(struct vm_fault *vmf)static uint64_t mmap_offset (struct drm_gem_object *obj) {struct drm_device *dev = obj->dev; -struct msm_gem_object *msm_obj =to_msm_bo (obj);int ret; -WARN_ON (!mutex_is_locked (&msm_obj->lock)); +WARN_ON (!msm_gem_is_locked (obj));/* Make it mmapable */ ret =drm_gem_create_mmap_offset (obj);
@@ -308,11 +307,10 @@ static uint64_t mmap_offset(struct drm_gem_object *obj)uint64_t msm_gem_mmap_offset (struct drm_gem_object *obj) {uint64_t offset; -struct msm_gem_object *msm_obj =to_msm_bo (obj); -mutex_lock (&msm_obj->lock); +msm_gem_lock (obj); offset =mmap_offset (obj); -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);return offset; }
@@ -322,7 +320,7 @@ static struct msm_gem_vma *add_vma(struct drm_gem_object *obj,struct msm_gem_object *msm_obj =to_msm_bo (obj);struct msm_gem_vma *vma; -WARN_ON (!mutex_is_locked (&msm_obj->lock)); +WARN_ON (!msm_gem_is_locked (obj)); vma =kzalloc (sizeof (*vma),GFP_KERNEL );if (!vma)
@@ -341,7 +339,7 @@ static struct msm_gem_vma *lookup_vma(struct drm_gem_object *obj,struct msm_gem_object *msm_obj =to_msm_bo (obj);struct msm_gem_vma *vma; -WARN_ON (!mutex_is_locked (&msm_obj->lock)); +WARN_ON (!msm_gem_is_locked (obj));list_for_each_entry (vma, &msm_obj->vmas, list) {if (vma->aspace == aspace)
@@ -360,14 +358,14 @@ static void del_vma(struct msm_gem_vma *vma)kfree (vma); } -/* Called with msm_obj->lock locked */ +/* Called with msm_obj locked */ static void put_iova (struct drm_gem_object *obj) {struct msm_gem_object *msm_obj =to_msm_bo (obj);struct msm_gem_vma *vma, *tmp; -WARN_ON (!mutex_is_locked (&msm_obj->lock)); +WARN_ON (!msm_gem_is_locked (obj));list_for_each_entry_safe (vma, tmp, &msm_obj->vmas, list) {if (vma->aspace) {
@@ -382,11 +380,10 @@ static int msm_gem_get_iova_locked(struct drm_gem_object *obj, struct msm_gem_address_space *aspace, uint64_t *iova, u64 range_start, u64 range_end) { - struct msm_gem_object *msm_obj = to_msm_bo(obj); struct msm_gem_vma *vma;int ret =0 ; -WARN_ON (!mutex_is_locked (&msm_obj->lock)); +WARN_ON (!msm_gem_is_locked (obj)); vma =lookup_vma (obj, aspace);
@@ -421,7 +418,7 @@ static int msm_gem_pin_iova(struct drm_gem_object *obj,if (msm_obj->flags &MSM_BO_MAP_PRIV ) prot |=IOMMU_PRIV ; -WARN_ON (!mutex_is_locked (&msm_obj->lock)); +WARN_ON (!msm_gem_is_locked (obj));if (WARN_ON (msm_obj->madv !=MSM_MADV_WILLNEED ))return -EBUSY ;
@@ -446,11 +443,10 @@ int msm_gem_get_and_pin_iova_range(struct drm_gem_object *obj, struct msm_gem_address_space *aspace, uint64_t *iova, u64 range_start, u64 range_end) { - struct msm_gem_object *msm_obj = to_msm_bo(obj); u64 local;int ret; -mutex_lock (&msm_obj->lock); +msm_gem_lock (obj); ret =msm_gem_get_iova_locked (obj, aspace, &local, range_start, range_end);
@@ -461,7 +457,7 @@ int msm_gem_get_and_pin_iova_range(struct drm_gem_object *obj,if (!ret) *iova = local; -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);return ret; }
@@ -479,12 +475,11 @@ int msm_gem_get_and_pin_iova(struct drm_gem_object *obj,int msm_gem_get_iova (struct drm_gem_object *obj,struct msm_gem_address_space *aspace,uint64_t *iova) { -struct msm_gem_object *msm_obj =to_msm_bo (obj);int ret; -mutex_lock (&msm_obj->lock); +msm_gem_lock (obj); ret =msm_gem_get_iova_locked (obj, aspace, iova,0 ,U64_MAX ); -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);return ret; }
@@ -495,12 +490,11 @@ int msm_gem_get_iova(struct drm_gem_object *obj,uint64_t msm_gem_iova (struct drm_gem_object *obj,struct msm_gem_address_space *aspace) { -struct msm_gem_object *msm_obj =to_msm_bo (obj);struct msm_gem_vma *vma; -mutex_lock (&msm_obj->lock); +msm_gem_lock (obj); vma =lookup_vma (obj, aspace); -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);WARN_ON (!vma);return vma ? vma->iova :0 ;
@@ -514,16 +508,15 @@ uint64_t msm_gem_iova(struct drm_gem_object *obj,void msm_gem_unpin_iova (struct drm_gem_object *obj,struct msm_gem_address_space *aspace) { -struct msm_gem_object *msm_obj =to_msm_bo (obj);struct msm_gem_vma *vma; -mutex_lock (&msm_obj->lock); +msm_gem_lock (obj); vma =lookup_vma (obj, aspace);if (!WARN_ON (!vma))msm_gem_unmap_vma (aspace, vma); -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj); } int msm_gem_dumb_create(struct drm_file *file, struct drm_device *dev,
@@ -564,20 +557,20 @@ static void *get_vaddr(struct drm_gem_object *obj, unsigned madv)if (obj->import_attach)return ERR_PTR (-ENODEV ); -mutex_lock (&msm_obj->lock); +msm_gem_lock (obj);if (WARN_ON (msm_obj->madv > madv)) {DRM_DEV_ERROR (obj->dev->dev,"Invalid madv state: %u vs %u\n" , msm_obj->madv, madv); -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);return ERR_PTR (-EBUSY ); }/* increment vmap_count *before* vmap() call, so shrinker can -* check vmap_count (is_vunmapable()) outside of msm_obj->lock. +* check vmap_count (is_vunmapable()) outside of msm_obj lock. * This guarantees that we won't try to msm_gem_vunmap() this * same object from within the vmap() call (while we already -* hold msm_obj->lock) +* hold msm_obj lock) */ msm_obj->vmap_count++;
@@ -595,12 +588,12 @@ static void *get_vaddr(struct drm_gem_object *obj, unsigned madv) } } -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);return msm_obj->vaddr; fail: msm_obj->vmap_count--; -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);return ERR_PTR (ret); }
@@ -624,10 +617,10 @@ void msm_gem_put_vaddr(struct drm_gem_object *obj) {struct msm_gem_object *msm_obj =to_msm_bo (obj); -mutex_lock (&msm_obj->lock); +msm_gem_lock (obj);WARN_ON (msm_obj->vmap_count <1 ); msm_obj->vmap_count--; -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj); }/* Update madvise status, returns true if not purged, else
@@ -637,7 +630,7 @@ int msm_gem_madvise(struct drm_gem_object *obj, unsigned madv) {struct msm_gem_object *msm_obj =to_msm_bo (obj); -mutex_lock (&msm_obj->lock); +msm_gem_lock (obj);WARN_ON (!mutex_is_locked (&obj->dev->struct_mutex));
@@ -646,7 +639,7 @@ int msm_gem_madvise(struct drm_gem_object *obj, unsigned madv) madv = msm_obj->madv; -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);return (madv != __MSM_MADV_PURGED); }
@@ -683,14 +676,14 @@ void msm_gem_purge(struct drm_gem_object *obj, enum msm_gem_lock subclass)invalidate_mapping_pages (file_inode (obj->filp)->i_mapping,0 , (loff_t)-1 ); -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj); }static void msm_gem_vunmap_locked (struct drm_gem_object *obj) {struct msm_gem_object *msm_obj =to_msm_bo (obj); -WARN_ON (!mutex_is_locked (&msm_obj->lock)); +WARN_ON (!msm_gem_is_locked (obj));if (!msm_obj->vaddr ||WARN_ON (!is_vunmapable (msm_obj)))return ;
@@ -705,7 +698,7 @@ void msm_gem_vunmap(struct drm_gem_object *obj, enum msm_gem_lock subclass)mutex_lock_nested (&msm_obj->lock, subclass);msm_gem_vunmap_locked (obj); -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj); }/* must be called before _move_to_active().. */
@@ -816,7 +809,7 @@ void msm_gem_describe(struct drm_gem_object *obj, struct seq_file *m)uint64_t off =drm_vma_node_start (&obj->vma_node);const char *madv; -mutex_lock (&msm_obj->lock); +msm_gem_lock (obj);switch (msm_obj->madv) {case __MSM_MADV_PURGED:
@@ -884,7 +877,7 @@ void msm_gem_describe(struct drm_gem_object *obj, struct seq_file *m)describe_fence (fence,"Exclusive" , m);rcu_read_unlock (); -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj); }void msm_gem_describe_objects (struct list_head *list,struct seq_file *m)
@@ -929,7 +922,7 @@ static void free_object(struct msm_gem_object *msm_obj)list_del (&msm_obj->mm_list); -mutex_lock (&msm_obj->lock); +msm_gem_lock (obj);put_iova (obj);
@@ -950,7 +943,7 @@ static void free_object(struct msm_gem_object *msm_obj)drm_gem_object_release (obj); -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);kfree (msm_obj); }
@@ -1070,10 +1063,10 @@ static struct drm_gem_object *_msm_gem_new(struct drm_device *dev,struct msm_gem_vma *vma;struct page **pages; -mutex_lock (&msm_obj->lock); +msm_gem_lock (obj); vma =add_vma (obj,NULL ); -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);if (IS_ERR (vma)) { ret =PTR_ERR (vma); goto fail;
@@ -1157,22 +1150,22 @@ struct drm_gem_object *msm_gem_import(struct drm_device *dev, npages = size /PAGE_SIZE ; msm_obj =to_msm_bo (obj); -mutex_lock (&msm_obj->lock); +msm_gem_lock (obj); msm_obj->sgt = sgt; msm_obj->pages =kvmalloc_array (npages,sizeof (struct page *),GFP_KERNEL );if (!msm_obj->pages) { -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj); ret = -ENOMEM ; goto fail; } ret =drm_prime_sg_to_page_addr_arrays (sgt, msm_obj->pages,NULL , npages);if (ret) { -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj); goto fail; } -mutex_unlock (&msm_obj->lock); +msm_gem_unlock (obj);mutex_lock (&dev->struct_mutex);list_add_tail (&msm_obj->mm_list, &priv->inactive_list);
drivers/gpu/drm/msm/msm_gem.h +28 -0
@@ -93,6 +93,34 @@ struct msm_gem_object { };#define to_msm_bo (x) container_of(x, struct msm_gem_object, base) +static inline void +msm_gem_lock (struct drm_gem_object *obj) +{ +struct msm_gem_object *msm_obj =to_msm_bo (obj); +mutex_lock (&msm_obj->lock); +} + +static inline int +msm_gem_lock_interruptible (struct drm_gem_object *obj) +{ +struct msm_gem_object *msm_obj =to_msm_bo (obj); +return mutex_lock_interruptible (&msm_obj->lock); +} + +static inline void +msm_gem_unlock (struct drm_gem_object *obj) +{ +struct msm_gem_object *msm_obj =to_msm_bo (obj); +mutex_unlock (&msm_obj->lock); +} + +static inline bool +msm_gem_is_locked (struct drm_gem_object *obj) +{ +struct msm_gem_object *msm_obj =to_msm_bo (obj); +return mutex_is_locked (&msm_obj->lock); +} +static inline bool is_active (struct msm_gem_object *msm_obj) {return atomic_read (&msm_obj->active_count);
Signature · 1 line
2.26.2
PATCH v2 2/22 drm/msm/gem: Rename internal get_iova_locked helper
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
We'll need to introduce a _locked() version of msm_gem_get_iova(), so we need to make that name available.
--- drivers/gpu/drm/msm/msm_gem.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +3 -3
@@ -376,7 +376,7 @@ put_iova(struct drm_gem_object *obj) } } -static int msm_gem_get_iova_locked (struct drm_gem_object *obj, +static int get_iova_locked (struct drm_gem_object *obj,struct msm_gem_address_space *aspace,uint64_t *iova,u64 range_start,u64 range_end) {
@@ -448,7 +448,7 @@ int msm_gem_get_and_pin_iova_range(struct drm_gem_object *obj,msm_gem_lock (obj); - ret =msm_gem_get_iova_locked (obj, aspace, &local, + ret =get_iova_locked (obj, aspace, &local, range_start, range_end);if (!ret)
@@ -478,7 +478,7 @@ int msm_gem_get_iova(struct drm_gem_object *obj,int ret;msm_gem_lock (obj); - ret =msm_gem_get_iova_locked (obj, aspace, iova,0 ,U64_MAX ); + ret =get_iova_locked (obj, aspace, iova,0 ,U64_MAX );msm_gem_unlock (obj);return ret;
Signature · 1 line
2.26.2
PATCH v2 3/22 drm/msm/gem: Move prototypes to msm_gem.h
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
--- drivers/gpu/drm/msm/disp/mdp4/mdp4_crtc.c | 1 + drivers/gpu/drm/msm/disp/mdp5/mdp5_crtc.c | 1 + drivers/gpu/drm/msm/dsi/dsi_host.c | 1 + drivers/gpu/drm/msm/msm_drv.h | 54 ---------------------- drivers/gpu/drm/msm/msm_fbdev.c | 1 + drivers/gpu/drm/msm/msm_gem.h | 56 +++++++++++++++++++++++ 6 files changed, 60 insertions(+), 54 deletions(-)
drivers/gpu/drm/msm/disp/mdp4/mdp4_crtc.c +1 -0
@@ -11,6 +11,7 @@#include <drm/drm_vblank.h> #include "mdp4_kms.h" +#include "msm_gem.h" struct mdp4_crtc {struct drm_crtc base;
drivers/gpu/drm/msm/disp/mdp5/mdp5_crtc.c +1 -0
@@ -15,6 +15,7 @@#include <drm/drm_vblank.h> #include "mdp5_kms.h" +#include "msm_gem.h" #define CURSOR_WIDTH 64#define CURSOR_HEIGHT 64
drivers/gpu/drm/msm/dsi/dsi_host.c +1 -0
@@ -26,6 +26,7 @@#include "sfpb.xml.h" #include "dsi_cfg.h" #include "msm_kms.h" +#include "msm_gem.h" #define DSI_RESET_TOGGLE_DELAY_MS 20
drivers/gpu/drm/msm/msm_drv.h +0 -54
74 lines not on this page · raw message
drivers/gpu/drm/msm/msm_fbdev.c +1 -0
@@ -9,6 +9,7 @@#include <drm/drm_fourcc.h> #include "msm_drv.h" +#include "msm_gem.h" #include "msm_kms.h" extern int msm_gem_mmap_obj(struct drm_gem_object *obj,
drivers/gpu/drm/msm/msm_gem.h +56 -0
62 lines not on this page · raw message
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 4/22 drm/msm/gem: Add some _locked() helpers
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
When we cut-over to using dma_resv_lock/etc instead of msm_obj->lock, we'll need these for the submit path (where resv->lock is already held).
--- drivers/gpu/drm/msm/msm_gem.c | 50 +++++++++++++++++++++++++++-------- drivers/gpu/drm/msm/msm_gem.h | 4 +++ 2 files changed, 43 insertions(+), 11 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +39 -11
@@ -435,18 +435,14 @@ static int msm_gem_pin_iova(struct drm_gem_object *obj, msm_obj->sgt, obj->size >> PAGE_SHIFT); } -/* -* get iova and pin it. Should have a matching put -* limits iova to specified range (in pages) -*/ -int msm_gem_get_and_pin_iova_range (struct drm_gem_object *obj, +static int get_and_pin_iova_range_locked (struct drm_gem_object *obj,struct msm_gem_address_space *aspace,uint64_t *iova,u64 range_start,u64 range_end) {u64 local;int ret; -msm_gem_lock (obj); +WARN_ON (!msm_gem_is_locked (obj)); ret =get_iova_locked (obj, aspace, &local, range_start, range_end);
@@ -457,10 +453,32 @@ int msm_gem_get_and_pin_iova_range(struct drm_gem_object *obj,if (!ret) *iova = local; +return ret; +} + +/* +* get iova and pin it. Should have a matching put +* limits iova to specified range (in pages) +*/ +int msm_gem_get_and_pin_iova_range (struct drm_gem_object *obj, +struct msm_gem_address_space *aspace,uint64_t *iova, +u64 range_start,u64 range_end) +{ +int ret; + +msm_gem_lock (obj); + ret =get_and_pin_iova_range_locked (obj, aspace, iova, range_start, range_end);msm_gem_unlock (obj); +return ret; } +int msm_gem_get_and_pin_iova_locked (struct drm_gem_object *obj, +struct msm_gem_address_space *aspace,uint64_t *iova) +{ +return get_and_pin_iova_range_locked (obj, aspace, iova,0 ,U64_MAX ); +} +/* get iova and pin it. Should have a matching put */ int msm_gem_get_and_pin_iova (struct drm_gem_object *obj,struct msm_gem_address_space *aspace,uint64_t *iova)
@@ -501,21 +519,31 @@ uint64_t msm_gem_iova(struct drm_gem_object *obj, }/* -* Unpin a iova by updating the reference counts. The memory isn't actually -* purged until something else (shrinker, mm_notifier, destroy, etc) decides -* to get rid of it +* Locked variant of msm_gem_unpin_iova() */ -void msm_gem_unpin_iova (struct drm_gem_object *obj, +void msm_gem_unpin_iova_locked (struct drm_gem_object *obj,struct msm_gem_address_space *aspace) {struct msm_gem_vma *vma; -msm_gem_lock (obj); +WARN_ON (!msm_gem_is_locked (obj)); + vma =lookup_vma (obj, aspace);if (!WARN_ON (!vma))msm_gem_unmap_vma (aspace, vma); +} +/* +* Unpin a iova by updating the reference counts. The memory isn't actually +* purged until something else (shrinker, mm_notifier, destroy, etc) decides +* to get rid of it +*/ +void msm_gem_unpin_iova (struct drm_gem_object *obj, +struct msm_gem_address_space *aspace) +{ +msm_gem_lock (obj); +msm_gem_unpin_iova_locked (obj, aspace);msm_gem_unlock (obj); }
drivers/gpu/drm/msm/msm_gem.h +4 -0
@@ -103,10 +103,14 @@ int msm_gem_get_iova(struct drm_gem_object *obj,int msm_gem_get_and_pin_iova_range (struct drm_gem_object *obj,struct msm_gem_address_space *aspace,uint64_t *iova,u64 range_start,u64 range_end); +int msm_gem_get_and_pin_iova_locked (struct drm_gem_object *obj, +struct msm_gem_address_space *aspace,uint64_t *iova);int msm_gem_get_and_pin_iova (struct drm_gem_object *obj,struct msm_gem_address_space *aspace,uint64_t *iova);uint64_t msm_gem_iova (struct drm_gem_object *obj,struct msm_gem_address_space *aspace); +void msm_gem_unpin_iova_locked (struct drm_gem_object *obj, +struct msm_gem_address_space *aspace);void msm_gem_unpin_iova (struct drm_gem_object *obj,struct msm_gem_address_space *aspace);struct page **msm_gem_get_pages (struct drm_gem_object *obj);
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 5/22 drm/msm/gem: Move locking in shrinker path
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
Move grabbing the bo lock into shrinker, with a msm_gem_trylock() to skip over bo's that are already locked. This gets rid of the nested lock classes.
--- drivers/gpu/drm/msm/msm_gem.c | 24 +++++---------------- drivers/gpu/drm/msm/msm_gem.h | 29 ++++++++++---------------- drivers/gpu/drm/msm/msm_gem_shrinker.c | 27 +++++++++++++++++------- 3 files changed, 35 insertions(+), 45 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +5 -19
@@ -17,8 +17,6 @@#include "msm_gpu.h" #include "msm_mmu.h" -static void msm_gem_vunmap_locked (struct drm_gem_object *obj); -static dma_addr_t physaddr (struct drm_gem_object *obj) {
@@ -672,20 +670,19 @@ int msm_gem_madvise(struct drm_gem_object *obj, unsigned madv)return (madv != __MSM_MADV_PURGED); } -void msm_gem_purge (struct drm_gem_object *obj,enum msm_gem_lock subclass) +void msm_gem_purge (struct drm_gem_object *obj) {struct drm_device *dev = obj->dev;struct msm_gem_object *msm_obj =to_msm_bo (obj);WARN_ON (!mutex_is_locked (&dev->struct_mutex)); +WARN_ON (!msm_gem_is_locked (obj));WARN_ON (!is_purgeable (msm_obj));WARN_ON (obj->import_attach); -mutex_lock_nested (&msm_obj->lock, subclass); -put_iova (obj); -msm_gem_vunmap_locked (obj); +msm_gem_vunmap (obj);put_pages (obj);
@@ -703,11 +700,9 @@ void msm_gem_purge(struct drm_gem_object *obj, enum msm_gem_lock subclass)invalidate_mapping_pages (file_inode (obj->filp)->i_mapping,0 , (loff_t)-1 ); - -msm_gem_unlock (obj); } -static void msm_gem_vunmap_locked (struct drm_gem_object *obj) +void msm_gem_vunmap (struct drm_gem_object *obj) {struct msm_gem_object *msm_obj =to_msm_bo (obj);
@@ -720,15 +715,6 @@ static void msm_gem_vunmap_locked(struct drm_gem_object *obj) msm_obj->vaddr =NULL ; } -void msm_gem_vunmap (struct drm_gem_object *obj,enum msm_gem_lock subclass) -{ -struct msm_gem_object *msm_obj =to_msm_bo (obj); - -mutex_lock_nested (&msm_obj->lock, subclass); -msm_gem_vunmap_locked (obj); -msm_gem_unlock (obj); -} -/* must be called before _move_to_active().. */ int msm_gem_sync_object (struct drm_gem_object *obj,struct msm_fence_context *fctx,bool exclusive)
@@ -965,7 +951,7 @@ static void free_object(struct msm_gem_object *msm_obj)drm_prime_gem_destroy (obj, msm_obj->sgt); }else { -msm_gem_vunmap_locked (obj); +msm_gem_vunmap (obj);put_pages (obj); }
drivers/gpu/drm/msm/msm_gem.h +11 -18
@@ -160,6 +160,13 @@ msm_gem_lock(struct drm_gem_object *obj)mutex_lock (&msm_obj->lock); } +static inline bool __must_check +msm_gem_trylock (struct drm_gem_object *obj) +{ +struct msm_gem_object *msm_obj =to_msm_bo (obj); +return mutex_trylock_recursive (&msm_obj->lock) ==MUTEX_TRYLOCK_SUCCESS ; +} +static inline int msm_gem_lock_interruptible (struct drm_gem_object *obj) {
@@ -188,6 +195,7 @@ static inline bool is_active(struct msm_gem_object *msm_obj)static inline bool is_purgeable (struct msm_gem_object *msm_obj) { +WARN_ON (!msm_gem_is_locked (&msm_obj->base));WARN_ON (!mutex_is_locked (&msm_obj->base.dev->struct_mutex));return (msm_obj->madv ==MSM_MADV_DONTNEED ) && msm_obj->sgt && !msm_obj->base.dma_buf && !msm_obj->base.import_attach;
@@ -195,27 +203,12 @@ static inline bool is_purgeable(struct msm_gem_object *msm_obj)static inline bool is_vunmapable (struct msm_gem_object *msm_obj) { +WARN_ON (!msm_gem_is_locked (&msm_obj->base));return (msm_obj->vmap_count ==0 ) && msm_obj->vaddr; } -/* The shrinker can be triggered while we hold objA->lock, and need -* to grab objB->lock to purge it. Lockdep just sees these as a single -* class of lock, so we use subclasses to teach it the difference. -* -* OBJ_LOCK_NORMAL is implicit (ie. normal mutex_lock() call), and -* OBJ_LOCK_SHRINKER is used by shrinker. -* -* It is *essential* that we never go down paths that could trigger the -* shrinker for a purgable object. This is ensured by checking that -* msm_obj->madv == MSM_MADV_WILLNEED. -*/ -enum msm_gem_lock { -OBJ_LOCK_NORMAL , -OBJ_LOCK_SHRINKER , -}; - -void msm_gem_purge (struct drm_gem_object *obj,enum msm_gem_lock subclass); -void msm_gem_vunmap (struct drm_gem_object *obj,enum msm_gem_lock subclass); +void msm_gem_purge (struct drm_gem_object *obj); +void msm_gem_vunmap (struct drm_gem_object *obj);void msm_gem_free_work (struct work_struct *work);/* Created per submit-ioctl, to track bo's and cmdstream bufs, etc,
drivers/gpu/drm/msm/msm_gem_shrinker.c +19 -8
@@ -52,8 +52,11 @@ msm_gem_shrinker_count(struct shrinker *shrinker, struct shrink_control *sc)return 0 ;list_for_each_entry (msm_obj, &priv->inactive_list, mm_list) { +if (!msm_gem_trylock (&msm_obj->base)) +continue ;if (is_purgeable (msm_obj)) count += msm_obj->base.size >>PAGE_SHIFT ; +msm_gem_unlock (&msm_obj->base); }if (unlock)
@@ -78,10 +81,13 @@ msm_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc)list_for_each_entry (msm_obj, &priv->inactive_list, mm_list) {if (freed >= sc->nr_to_scan)break ; +if (!msm_gem_trylock (&msm_obj->base)) +continue ;if (is_purgeable (msm_obj)) { -msm_gem_purge (&msm_obj->base,OBJ_LOCK_SHRINKER ); +msm_gem_purge (&msm_obj->base); freed += msm_obj->base.size >>PAGE_SHIFT ; } +msm_gem_unlock (&msm_obj->base); }if (unlock)
@@ -107,15 +113,20 @@ msm_gem_shrinker_vmap(struct notifier_block *nb, unsigned long event, void *ptr)return NOTIFY_DONE ;list_for_each_entry (msm_obj, &priv->inactive_list, mm_list) { +if (!msm_gem_trylock (&msm_obj->base)) +continue ;if (is_vunmapable (msm_obj)) { -msm_gem_vunmap (&msm_obj->base,OBJ_LOCK_SHRINKER ); -/* since we don't know any better, lets bail after a few -* and if necessary the shrinker will be invoked again. -* Seems better than unmapping *everything* -*/ -if (++unmapped >=15 ) -break ; +msm_gem_vunmap (&msm_obj->base); + unmapped++; } +msm_gem_unlock (&msm_obj->base); + +/* since we don't know any better, lets bail after a few +* and if necessary the shrinker will be invoked again. +* Seems better than unmapping *everything* +*/ +if (++unmapped >=15 ) +break ; }if (unlock)
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 6/22 drm/msm/submit: Move copy_from_user ahead of locking bos
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
We cannot switch to using obj->resv for locking without first moving all the copy_from_user() ahead of submit_lock_objects(). Otherwise in the mm fault path we aquire mm->mmap_sem before obj lock, but in the submit path the order is reversed.
--- drivers/gpu/drm/msm/msm_gem.h | 3 + drivers/gpu/drm/msm/msm_gem_submit.c | 121 ++++++++++++++++----------- 2 files changed, 76 insertions(+), 48 deletions(-)
drivers/gpu/drm/msm/msm_gem.h +3 -0
@@ -238,7 +238,10 @@ struct msm_gem_submit {uint32_t type;uint32_t size;/* in dwords */ uint64_t iova; +uint32_t offset;/* in dwords */ uint32_t idx;/* cmdstream buffer idx in bos[] */ +uint32_t nr_relocs; +struct drm_msm_gem_submit_reloc *relocs; } *cmd;/* array of size nr_cmds */ struct {uint32_t flags;
drivers/gpu/drm/msm/msm_gem_submit.c +73 -48
181 lines not on this page · raw message
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 7/22 drm/msm: Do rpm get sooner in the submit path
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
Unfortunately, due to an dev_pm_opp locking interaction with mm->mmap_sem, we need to do pm get before aquiring obj locks, otherwise we can have anger lockdep with the chain:
opp_table_lock --> &mm->mmap_sem --> reservation_ww_class_mutex
For an explicit fencing userspace, the impact should be minimal as we do all the fence waits before this point. It could result in some needless resumes in error cases, etc.
--- drivers/gpu/drm/msm/msm_gem_submit.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-)
drivers/gpu/drm/msm/msm_gem_submit.c +13 -2
@@ -744,11 +744,20 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data, ret =submit_lookup_objects (submit, args, file);if (ret) - goto out; + goto out_pre_pm; ret =submit_lookup_cmds (submit, args, file);if (ret) - goto out; + goto out_pre_pm; + +/* +* Thanks to dev_pm_opp opp_table_lock interactions with mm->mmap_sem +* in the resume path, we need to to rpm get before we lock objs. +* Which unfortunately might involve powering up the GPU sooner than +* is necessary. But at least in the explicit fencing case, we will +* have already done all the fence waiting. +*/ +pm_runtime_get_sync (&gpu->pdev->dev);/* copy_*_user while holding a ww ticket upsets lockdep */ ww_acquire_init (&submit->ticket, &reservation_ww_class);
@@ -825,6 +834,8 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data, out: +pm_runtime_put (&gpu->pdev->dev); +out_pre_pm:submit_cleanup (submit);if (has_ww_ticket)ww_acquire_fini (&submit->ticket);
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 8/22 drm/msm/gem: Switch over to obj->resv for locking
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
--- drivers/gpu/drm/msm/msm_gem.c | 4 +--- drivers/gpu/drm/msm/msm_gem.h | 16 +++++----------- drivers/gpu/drm/msm/msm_gem_submit.c | 4 ++-- drivers/gpu/drm/msm/msm_gpu.c | 2 +- 4 files changed, 9 insertions(+), 17 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +1 -3
@@ -955,9 +955,9 @@ static void free_object(struct msm_gem_object *msm_obj)put_pages (obj); } +msm_gem_unlock (obj);drm_gem_object_release (obj); -msm_gem_unlock (obj);kfree (msm_obj); }
@@ -1029,8 +1029,6 @@ static int msm_gem_new_impl(struct drm_device *dev,if (!msm_obj)return -ENOMEM ; -mutex_init (&msm_obj->lock); - msm_obj->flags = flags; msm_obj->madv =MSM_MADV_WILLNEED ;
drivers/gpu/drm/msm/msm_gem.h +5 -11
@@ -85,7 +85,6 @@ struct msm_gem_object {* an IOMMU. Also used for stolen/splashscreen buffer. */ struct drm_mm_node *vram_node; -struct mutex lock;/* Protects resources associated with bo */ char name[32 ];/* Identifier to print for the debugfs files */
@@ -156,36 +155,31 @@ void msm_gem_describe_objects(struct list_head *list, struct seq_file *m);static inline void msm_gem_lock (struct drm_gem_object *obj) { -struct msm_gem_object *msm_obj =to_msm_bo (obj); -mutex_lock (&msm_obj->lock); +dma_resv_lock (obj->resv,NULL ); }static inline bool __must_check msm_gem_trylock (struct drm_gem_object *obj) { -struct msm_gem_object *msm_obj =to_msm_bo (obj); -return mutex_trylock_recursive (&msm_obj->lock) ==MUTEX_TRYLOCK_SUCCESS ; +return dma_resv_trylock (obj->resv); }static inline int msm_gem_lock_interruptible (struct drm_gem_object *obj) { -struct msm_gem_object *msm_obj =to_msm_bo (obj); -return mutex_lock_interruptible (&msm_obj->lock); +return dma_resv_lock_interruptible (obj->resv,NULL ); }static inline void msm_gem_unlock (struct drm_gem_object *obj) { -struct msm_gem_object *msm_obj =to_msm_bo (obj); -mutex_unlock (&msm_obj->lock); +dma_resv_unlock (obj->resv); }static inline bool msm_gem_is_locked (struct drm_gem_object *obj) { -struct msm_gem_object *msm_obj =to_msm_bo (obj); -return mutex_is_locked (&msm_obj->lock); +return dma_resv_is_locked (obj->resv); }static inline bool is_active (struct msm_gem_object *msm_obj)
drivers/gpu/drm/msm/msm_gem_submit.c +2 -2
@@ -215,7 +215,7 @@ static void submit_unlock_unpin_bo(struct msm_gem_submit *submit,struct msm_gem_object *msm_obj = submit->bos[i].obj;if (submit->bos[i].flags &BO_PINNED ) -msm_gem_unpin_iova (&msm_obj->base, submit->aspace); +msm_gem_unpin_iova_locked (&msm_obj->base, submit->aspace);if (submit->bos[i].flags &BO_LOCKED )dma_resv_unlock (msm_obj->base.resv);
@@ -318,7 +318,7 @@ static int submit_pin_objects(struct msm_gem_submit *submit)uint64_t iova;/* if locking succeeded, pin bo: */ - ret =msm_gem_get_and_pin_iova (&msm_obj->base, + ret =msm_gem_get_and_pin_iova_locked (&msm_obj->base, submit->aspace, &iova);if (ret)
drivers/gpu/drm/msm/msm_gpu.c +1 -1
@@ -784,7 +784,7 @@ void msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit)/* submit takes a reference to the bo and iova until retired: */ drm_gem_object_get (&msm_obj->base); -msm_gem_get_and_pin_iova (&msm_obj->base, submit->aspace, &iova); +msm_gem_get_and_pin_iova_locked (&msm_obj->base, submit->aspace, &iova);if (submit->bos[i].flags &MSM_SUBMIT_BO_WRITE )dma_resv_add_excl_fence (drm_obj->resv, submit->fence);
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 9/22 drm/msm: Use correct drm_gem_object_put() in fail case
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
We only want to use the _unlocked() variant in the unlocked case.
--- drivers/gpu/drm/msm/msm_gem.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-)
drivers/gpu/drm/msm/msm_gem.c +5 -1
@@ -1120,7 +1120,11 @@ static struct drm_gem_object *_msm_gem_new(struct drm_device *dev,return obj; fail: -drm_gem_object_put (obj); +if (struct_mutex_locked) { +drm_gem_object_put_locked (obj); + }else { +drm_gem_object_put (obj); + }return ERR_PTR (ret); }
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 10/22 drm/msm: Drop chatty trace
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
It is somewhat redundant with the gpu tracepoints, and anyways not too useful to justify spamming the log when debug traces are enabled.
--- drivers/gpu/drm/msm/msm_gpu.c | 1 - 1 file changed, 1 deletion(-)
drivers/gpu/drm/msm/msm_gpu.c +0 -1
@@ -535,7 +535,6 @@ static void recover_worker(struct work_struct *work)static void hangcheck_timer_reset (struct msm_gpu *gpu) { -DBG ("%s" , gpu->name);mod_timer (&gpu->hangcheck_timer,round_jiffies_up (jiffies +DRM_MSM_HANGCHECK_JIFFIES )); }
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 11/22 drm/msm: Move update_fences()
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
Small cleanup, update_fences() is used in the hangcheck path, but also in the normal retire path.
--- drivers/gpu/drm/msm/msm_gpu.c | 28 ++++++++++++++-------------- 1 file changed, 14 insertions(+), 14 deletions(-)
drivers/gpu/drm/msm/msm_gpu.c +14 -14
@@ -265,6 +265,20 @@ int msm_gpu_hw_init(struct msm_gpu *gpu)return ret; } +static void update_fences (struct msm_gpu *gpu,struct msm_ringbuffer *ring, +uint32_t fence) +{ +struct msm_gem_submit *submit; + +list_for_each_entry (submit, &ring->submits, node) { +if (submit->seqno > fence) +break ; + +msm_update_fence (submit->ring->fctx, + submit->fence->seqno); + } +} +#ifdef CONFIG_DEV_COREDUMP static ssize_t msm_gpu_devcoredump_read (char *buffer,loff_t offset,size_t count,void *data,size_t datalen)
@@ -411,20 +425,6 @@ static void msm_gpu_crashstate_capture(struct msm_gpu *gpu,* Hangcheck detection for locked gpu: */ -static void update_fences (struct msm_gpu *gpu,struct msm_ringbuffer *ring, -uint32_t fence) -{ -struct msm_gem_submit *submit; - -list_for_each_entry (submit, &ring->submits, node) { -if (submit->seqno > fence) -break ; - -msm_update_fence (submit->ring->fctx, - submit->fence->seqno); - } -} -static struct msm_gem_submit *find_submit (struct msm_ringbuffer *ring,uint32_t fence) {
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 12/22 drm/msm: Add priv->mm_lock to protect active/inactive lists
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
Rather than relying on the big dev->struct_mutex hammer, introduce a more specific lock for protecting the bo lists.
--- drivers/gpu/drm/msm/msm_debugfs.c | 7 +++++++ drivers/gpu/drm/msm/msm_drv.c | 7 +++++++ drivers/gpu/drm/msm/msm_drv.h | 13 +++++++++++- drivers/gpu/drm/msm/msm_gem.c | 28 +++++++++++++++----------- drivers/gpu/drm/msm/msm_gem_shrinker.c | 12 +++++++++++ drivers/gpu/drm/msm/msm_gpu.h | 5 ++++- 6 files changed, 58 insertions(+), 14 deletions(-)
drivers/gpu/drm/msm/msm_debugfs.c +7 -0
@@ -112,6 +112,11 @@ static int msm_gem_show(struct drm_device *dev, struct seq_file *m) {struct msm_drm_private *priv = dev->dev_private;struct msm_gpu *gpu = priv->gpu; +int ret; + + ret =mutex_lock_interruptible (&priv->mm_lock); +if (ret) +return ret;if (gpu) {seq_printf (m,"Active Objects (%s):\n" , gpu->name);
@@ -121,6 +126,8 @@ static int msm_gem_show(struct drm_device *dev, struct seq_file *m)seq_printf (m,"Inactive Objects:\n" );msm_gem_describe_objects (&priv->inactive_list, m); +mutex_unlock (&priv->mm_lock); +return 0 ; }
drivers/gpu/drm/msm/msm_drv.c +7 -0
@@ -7,6 +7,7 @@#include <linux/dma-mapping.h> #include <linux/kthread.h> +#include <linux/sched/mm.h> #include <linux/uaccess.h> #include <uapi/linux/sched/types.h>
@@ -441,6 +442,12 @@ static int msm_drm_init(struct device *dev, struct drm_driver *drv)init_llist_head (&priv->free_list);INIT_LIST_HEAD (&priv->inactive_list); +mutex_init (&priv->mm_lock); + +/* Teach lockdep about lock ordering wrt. shrinker: */ +fs_reclaim_acquire (GFP_KERNEL ); +might_lock (&priv->mm_lock); +fs_reclaim_release (GFP_KERNEL );drm_mode_config_init (ddev);
drivers/gpu/drm/msm/msm_drv.h +12 -1
@@ -174,8 +174,19 @@ struct msm_drm_private {struct msm_rd_state *hangrd;/* debugfs to dump hanging submits */ struct msm_perf_state *perf; -/* list of GEM objects: */ +/* +* List of inactive GEM objects. Every bo is either in the inactive_list +* or gpu->active_list (for the gpu it is active on[1]) +* +* These lists are protected by mm_lock. If struct_mutex is involved, it +* should be aquired prior to mm_lock. One should *not* hold mm_lock in +* get_pages()/vmap()/etc paths, as they can trigger the shrinker. +* +* [1] if someone ever added support for the old 2d cores, there could be +* more than one gpu object +*/ struct list_head inactive_list; +struct mutex mm_lock;/* worker for delayed free of objects: */ struct work_struct free_work;
drivers/gpu/drm/msm/msm_gem.c +16 -12
@@ -753,13 +753,17 @@ int msm_gem_sync_object(struct drm_gem_object *obj,void msm_gem_active_get (struct drm_gem_object *obj,struct msm_gpu *gpu) {struct msm_gem_object *msm_obj =to_msm_bo (obj); -WARN_ON (!mutex_is_locked (&obj->dev->struct_mutex)); +struct msm_drm_private *priv = obj->dev->dev_private; + +might_sleep ();WARN_ON (msm_obj->madv !=MSM_MADV_WILLNEED );if (!atomic_fetch_inc (&msm_obj->active_count)) { +mutex_lock (&priv->mm_lock); msm_obj->gpu = gpu;list_del_init (&msm_obj->mm_list);list_add_tail (&msm_obj->mm_list, &gpu->active_list); +mutex_unlock (&priv->mm_lock); } }
@@ -768,12 +772,14 @@ void msm_gem_active_put(struct drm_gem_object *obj)struct msm_gem_object *msm_obj =to_msm_bo (obj);struct msm_drm_private *priv = obj->dev->dev_private; -WARN_ON (!mutex_is_locked (&obj->dev->struct_mutex)); +might_sleep ();if (!atomic_dec_return (&msm_obj->active_count)) { +mutex_lock (&priv->mm_lock); msm_obj->gpu =NULL ;list_del_init (&msm_obj->mm_list);list_add_tail (&msm_obj->mm_list, &priv->inactive_list); +mutex_unlock (&priv->mm_lock); } }
@@ -928,13 +934,16 @@ static void free_object(struct msm_gem_object *msm_obj) {struct drm_gem_object *obj = &msm_obj->base;struct drm_device *dev = obj->dev; +struct msm_drm_private *priv = dev->dev_private;WARN_ON (!mutex_is_locked (&dev->struct_mutex));/* object should not be on active list: */ WARN_ON (is_active (msm_obj)); +mutex_lock (&priv->mm_lock);list_del (&msm_obj->mm_list); +mutex_unlock (&priv->mm_lock);msm_gem_lock (obj);
@@ -1108,14 +1117,9 @@ static struct drm_gem_object *_msm_gem_new(struct drm_device *dev,mapping_set_gfp_mask (obj->filp->f_mapping,GFP_HIGHUSER ); } -if (struct_mutex_locked) { -WARN_ON (!mutex_is_locked (&dev->struct_mutex)); -list_add_tail (&msm_obj->mm_list, &priv->inactive_list); - }else { -mutex_lock (&dev->struct_mutex); -list_add_tail (&msm_obj->mm_list, &priv->inactive_list); -mutex_unlock (&dev->struct_mutex); - } +mutex_lock (&priv->mm_lock); +list_add_tail (&msm_obj->mm_list, &priv->inactive_list); +mutex_unlock (&priv->mm_lock);return obj;
@@ -1183,9 +1187,9 @@ struct drm_gem_object *msm_gem_import(struct drm_device *dev,msm_gem_unlock (obj); -mutex_lock (&dev->struct_mutex); +mutex_lock (&priv->mm_lock);list_add_tail (&msm_obj->mm_list, &priv->inactive_list); -mutex_unlock (&dev->struct_mutex); +mutex_unlock (&priv->mm_lock);return obj;
drivers/gpu/drm/msm/msm_gem_shrinker.c +12 -0
@@ -51,6 +51,8 @@ msm_gem_shrinker_count(struct shrinker *shrinker, struct shrink_control *sc)if (!msm_gem_shrinker_lock (dev, &unlock))return 0 ; +mutex_lock (&priv->mm_lock); +list_for_each_entry (msm_obj, &priv->inactive_list, mm_list) {if (!msm_gem_trylock (&msm_obj->base))continue ;
@@ -59,6 +61,8 @@ msm_gem_shrinker_count(struct shrinker *shrinker, struct shrink_control *sc)msm_gem_unlock (&msm_obj->base); } +mutex_unlock (&priv->mm_lock); +if (unlock)mutex_unlock (&dev->struct_mutex);
@@ -78,6 +82,8 @@ msm_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc)if (!msm_gem_shrinker_lock (dev, &unlock))return SHRINK_STOP ; +mutex_lock (&priv->mm_lock); +list_for_each_entry (msm_obj, &priv->inactive_list, mm_list) {if (freed >= sc->nr_to_scan)break ;
@@ -90,6 +96,8 @@ msm_gem_shrinker_scan(struct shrinker *shrinker, struct shrink_control *sc)msm_gem_unlock (&msm_obj->base); } +mutex_unlock (&priv->mm_lock); +if (unlock)mutex_unlock (&dev->struct_mutex);
@@ -112,6 +120,8 @@ msm_gem_shrinker_vmap(struct notifier_block *nb, unsigned long event, void *ptr)if (!msm_gem_shrinker_lock (dev, &unlock))return NOTIFY_DONE ; +mutex_lock (&priv->mm_lock); +list_for_each_entry (msm_obj, &priv->inactive_list, mm_list) {if (!msm_gem_trylock (&msm_obj->base))continue ;
@@ -129,6 +139,8 @@ msm_gem_shrinker_vmap(struct notifier_block *nb, unsigned long event, void *ptr)break ; } +mutex_unlock (&priv->mm_lock); +if (unlock)mutex_unlock (&dev->struct_mutex);
drivers/gpu/drm/msm/msm_gpu.h +4 -1
@@ -94,7 +94,10 @@ struct msm_gpu {struct msm_ringbuffer *rb[MSM_GPU_MAX_RINGS ];int nr_rings; -/* list of GEM active objects: */ +/* +* List of GEM active objects on this gpu. Protected by +* msm_drm_private::mm_lock +*/ struct list_head active_list;/* does gpu need hw_init? */
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 13/22 drm/msm: Document and rename preempt_lock
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
Before adding another lock, give ring->lock a more descriptive name.
--- drivers/gpu/drm/msm/adreno/a5xx_gpu.c | 4 ++-- drivers/gpu/drm/msm/adreno/a5xx_preempt.c | 12 ++++++------ drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 4 ++-- drivers/gpu/drm/msm/msm_ringbuffer.c | 2 +- drivers/gpu/drm/msm/msm_ringbuffer.h | 7 ++++++- 5 files changed, 17 insertions(+), 12 deletions(-)
drivers/gpu/drm/msm/adreno/a5xx_gpu.c +2 -2
@@ -36,7 +36,7 @@ void a5xx_flush(struct msm_gpu *gpu, struct msm_ringbuffer *ring,OUT_RING (ring,upper_32_bits (shadowptr (a5xx_gpu, ring))); } -spin_lock_irqsave (&ring->lock, flags); +spin_lock_irqsave (&ring->preempt_lock, flags);/* Copy the shadow to the actual register */ ring->cur = ring->next;
@@ -44,7 +44,7 @@ void a5xx_flush(struct msm_gpu *gpu, struct msm_ringbuffer *ring,/* Make sure to wrap wptr if we need to */ wptr =get_wptr (ring); -spin_unlock_irqrestore (&ring->lock, flags); +spin_unlock_irqrestore (&ring->preempt_lock, flags);/* Make sure everything is posted before making a decision */ mb ();
drivers/gpu/drm/msm/adreno/a5xx_preempt.c +6 -6
@@ -45,9 +45,9 @@ static inline void update_wptr(struct msm_gpu *gpu, struct msm_ringbuffer *ring)if (!ring)return ; -spin_lock_irqsave (&ring->lock, flags); +spin_lock_irqsave (&ring->preempt_lock, flags); wptr =get_wptr (ring); -spin_unlock_irqrestore (&ring->lock, flags); +spin_unlock_irqrestore (&ring->preempt_lock, flags);gpu_write (gpu,REG_A5XX_CP_RB_WPTR , wptr); }
@@ -62,9 +62,9 @@ static struct msm_ringbuffer *get_next_ring(struct msm_gpu *gpu)bool empty;struct msm_ringbuffer *ring = gpu->rb[i]; -spin_lock_irqsave (&ring->lock, flags); +spin_lock_irqsave (&ring->preempt_lock, flags); empty = (get_wptr (ring) == ring->memptrs->rptr); -spin_unlock_irqrestore (&ring->lock, flags); +spin_unlock_irqrestore (&ring->preempt_lock, flags);if (!empty)return ring;
@@ -132,9 +132,9 @@ void a5xx_preempt_trigger(struct msm_gpu *gpu) }/* Make sure the wptr doesn't update while we're in motion */ -spin_lock_irqsave (&ring->lock, flags); +spin_lock_irqsave (&ring->preempt_lock, flags); a5xx_gpu->preempt[ring->id]->wptr =get_wptr (ring); -spin_unlock_irqrestore (&ring->lock, flags); +spin_unlock_irqrestore (&ring->preempt_lock, flags);/* Set the address of the incoming preemption record */ gpu_write64(gpu, REG_A5XX_CP_CONTEXT_SWITCH_RESTORE_ADDR_LO,
drivers/gpu/drm/msm/adreno/a6xx_gpu.c +2 -2
@@ -65,7 +65,7 @@ static void a6xx_flush(struct msm_gpu *gpu, struct msm_ringbuffer *ring)OUT_RING (ring,upper_32_bits (shadowptr (a6xx_gpu, ring))); } -spin_lock_irqsave (&ring->lock, flags); +spin_lock_irqsave (&ring->preempt_lock, flags);/* Copy the shadow to the actual register */ ring->cur = ring->next;
@@ -73,7 +73,7 @@ static void a6xx_flush(struct msm_gpu *gpu, struct msm_ringbuffer *ring)/* Make sure to wrap wptr if we need to */ wptr =get_wptr (ring); -spin_unlock_irqrestore (&ring->lock, flags); +spin_unlock_irqrestore (&ring->preempt_lock, flags);/* Make sure everything is posted before making a decision */ mb ();
drivers/gpu/drm/msm/msm_ringbuffer.c +1 -1
@@ -46,7 +46,7 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int id, ring->memptrs_iova = memptrs_iova;INIT_LIST_HEAD (&ring->submits); -spin_lock_init (&ring->lock); +spin_lock_init (&ring->preempt_lock);snprintf (name,sizeof (name),"gpu-ring-%d" , ring->id);
drivers/gpu/drm/msm/msm_ringbuffer.h +6 -1
@@ -46,7 +46,12 @@ struct msm_ringbuffer {struct msm_rbmemptrs *memptrs;uint64_t memptrs_iova;struct msm_fence_context *fctx; -spinlock_t lock; + +/* +* preempt_lock protects preemption and serializes wptr updates against +* preemption. Can be aquired from irq context. +*/ +spinlock_t preempt_lock; }; struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int id,
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 14/22 drm/msm: Protect ring->submits with it's own lock
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
One less place to rely on dev->struct_mutex.
--- drivers/gpu/drm/msm/msm_gem_submit.c | 2 ++ drivers/gpu/drm/msm/msm_gpu.c | 37 ++++++++++++++++++++++------ drivers/gpu/drm/msm/msm_ringbuffer.c | 1 + drivers/gpu/drm/msm/msm_ringbuffer.h | 6 +++++ 4 files changed, 39 insertions(+), 7 deletions(-)
drivers/gpu/drm/msm/msm_gem_submit.c +2 -0
@@ -65,7 +65,9 @@ void msm_gem_submit_free(struct msm_gem_submit *submit)unsigned i;dma_fence_put (submit->fence); +spin_lock (&submit->ring->submit_lock);list_del (&submit->node); +spin_unlock (&submit->ring->submit_lock);put_pid (submit->pid);msm_submitqueue_put (submit->queue);
drivers/gpu/drm/msm/msm_gpu.c +30 -7
@@ -270,6 +270,7 @@ static void update_fences(struct msm_gpu *gpu, struct msm_ringbuffer *ring, {struct msm_gem_submit *submit; +spin_lock (&ring->submit_lock);list_for_each_entry (submit, &ring->submits, node) {if (submit->seqno > fence)break ;
@@ -277,6 +278,7 @@ static void update_fences(struct msm_gpu *gpu, struct msm_ringbuffer *ring,msm_update_fence (submit->ring->fctx, submit->fence->seqno); } +spin_unlock (&ring->submit_lock); }#ifdef CONFIG_DEV_COREDUMP
@@ -430,11 +432,14 @@ find_submit(struct msm_ringbuffer *ring, uint32_t fence) {struct msm_gem_submit *submit; -WARN_ON (!mutex_is_locked (&ring->gpu->dev->struct_mutex)); - -list_for_each_entry (submit, &ring->submits, node) -if (submit->seqno == fence) +spin_lock (&ring->submit_lock); +list_for_each_entry (submit, &ring->submits, node) { +if (submit->seqno == fence) { +spin_unlock (&ring->submit_lock);return submit; + } + } +spin_unlock (&ring->submit_lock);return NULL ; }
@@ -523,8 +528,10 @@ static void recover_worker(struct work_struct *work)for (i =0 ; i < gpu->nr_rings; i++) {struct msm_ringbuffer *ring = gpu->rb[i]; +spin_lock (&ring->submit_lock);list_for_each_entry (submit, &ring->submits, node) gpu->funcs->submit (gpu, submit); +spin_unlock (&ring->submit_lock); } }
@@ -711,7 +718,6 @@ static void retire_submit(struct msm_gpu *gpu, struct msm_ringbuffer *ring,static void retire_submits (struct msm_gpu *gpu) {struct drm_device *dev = gpu->dev; -struct msm_gem_submit *submit, *tmp;int i;WARN_ON (!mutex_is_locked (&dev->struct_mutex));
@@ -720,9 +726,24 @@ static void retire_submits(struct msm_gpu *gpu)for (i =0 ; i < gpu->nr_rings; i++) {struct msm_ringbuffer *ring = gpu->rb[i]; -list_for_each_entry_safe (submit, tmp, &ring->submits, node) { -if (dma_fence_is_signaled (submit->fence)) +while (true) { +struct msm_gem_submit *submit =NULL ; + +spin_lock (&ring->submit_lock); + submit =list_first_entry_or_null (&ring->submits, + struct msm_gem_submit, node); +spin_unlock (&ring->submit_lock); + +/* +* If no submit, we are done. If submit->fence hasn't +* been signalled, then later submits are not signalled +* either, so we are also done. +*/ +if (submit &&dma_fence_is_signaled (submit->fence)) {retire_submit (gpu, ring, submit); + }else { +break ; + } } } }
@@ -765,7 +786,9 @@ void msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit) submit->seqno = ++ring->seqno; +spin_lock (&ring->submit_lock);list_add_tail (&submit->node, &ring->submits); +spin_unlock (&ring->submit_lock);msm_rd_dump_submit (priv->rd, submit,NULL );
drivers/gpu/drm/msm/msm_ringbuffer.c +1 -0
@@ -46,6 +46,7 @@ struct msm_ringbuffer *msm_ringbuffer_new(struct msm_gpu *gpu, int id, ring->memptrs_iova = memptrs_iova;INIT_LIST_HEAD (&ring->submits); +spin_lock_init (&ring->submit_lock);spin_lock_init (&ring->preempt_lock);snprintf (name,sizeof (name),"gpu-ring-%d" , ring->id);
drivers/gpu/drm/msm/msm_ringbuffer.h +6 -0
@@ -39,7 +39,13 @@ struct msm_ringbuffer {int id;struct drm_gem_object *bo;uint32_t *start, *end, *cur, *next; + +/* +* List of in-flight submits on this ring. Protected by submit_lock. +*/ struct list_head submits; +spinlock_t submit_lock; +uint64_t iova;uint32_t seqno;uint32_t hangcheck_fence;
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 15/22 drm/msm: Refcount submits
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
Before we remove dev->struct_mutex from the retire path, we have to deal with the situation of a submit retiring before the submit ioctl returns.
To deal with this, ring->submits will hold a reference to the submit, which is dropped when the submit is retired. And the submit ioctl path holds it's own ref, which it drops when it is done with the submit.
Also, add to submit list *after* getting/pinning bo's, to prevent badness in case the completed fence is corrupted, and retire_worker mistakenly believes the submit is done too early.
--- drivers/gpu/drm/msm/msm_drv.h | 1 - drivers/gpu/drm/msm/msm_gem.h | 13 +++++++++++++ drivers/gpu/drm/msm/msm_gem_submit.c | 11 +++++------ drivers/gpu/drm/msm/msm_gpu.c | 21 ++++++++++++++++----- 4 files changed, 34 insertions(+), 12 deletions(-)
drivers/gpu/drm/msm/msm_drv.h +0 -1
@@ -277,7 +277,6 @@ void msm_unregister_mmu(struct drm_device *dev, struct msm_mmu *mmu);bool msm_use_mmu (struct drm_device *dev); -void msm_gem_submit_free (struct msm_gem_submit *submit);int msm_ioctl_gem_submit (struct drm_device *dev,void *data,struct drm_file *file);
drivers/gpu/drm/msm/msm_gem.h +13 -0
@@ -211,6 +211,7 @@ void msm_gem_free_work(struct work_struct *work);* lasts for the duration of the submit-ioctl. */ struct msm_gem_submit { +struct kref ref;struct drm_device *dev;struct msm_gpu *gpu;struct msm_gem_address_space *aspace;
@@ -247,6 +248,18 @@ struct msm_gem_submit { } bos[]; }; +void __msm_gem_submit_destroy (struct kref *kref); + +static inline void msm_gem_submit_get (struct msm_gem_submit *submit) +{ +kref_get (&submit->ref); +} + +static inline void msm_gem_submit_put (struct msm_gem_submit *submit) +{ +kref_put (&submit->ref, __msm_gem_submit_destroy); +} +/* helper to determine of a buffer in submit should be dumped, used for both * devcoredump and debugfs cmdstream dumping: */
drivers/gpu/drm/msm/msm_gem_submit.c +5 -6
@@ -42,6 +42,7 @@ static struct msm_gem_submit *submit_create(struct drm_device *dev,if (!submit)return NULL ; +kref_init (&submit->ref); submit->dev = dev; submit->aspace = queue->ctx->aspace; submit->gpu = gpu;
@@ -60,14 +61,13 @@ static struct msm_gem_submit *submit_create(struct drm_device *dev,return submit; } -void msm_gem_submit_free (struct msm_gem_submit *submit) +void __msm_gem_submit_destroy (struct kref *kref) { +struct msm_gem_submit *submit = + container_of(kref, struct msm_gem_submit, ref);unsigned i;dma_fence_put (submit->fence); -spin_lock (&submit->ring->submit_lock); -list_del (&submit->node); -spin_unlock (&submit->ring->submit_lock);put_pid (submit->pid);msm_submitqueue_put (submit->queue);
@@ -841,8 +841,7 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,submit_cleanup (submit);if (has_ww_ticket)ww_acquire_fini (&submit->ticket); -if (ret) -msm_gem_submit_free (submit); +msm_gem_submit_put (submit); out_unlock:if (ret && (out_fence_fd >=0 ))put_unused_fd (out_fence_fd);
drivers/gpu/drm/msm/msm_gpu.c +16 -5
@@ -712,7 +712,12 @@ static void retire_submit(struct msm_gpu *gpu, struct msm_ringbuffer *ring,pm_runtime_mark_last_busy (&gpu->pdev->dev);pm_runtime_put_autosuspend (&gpu->pdev->dev); -msm_gem_submit_free (submit); + +spin_lock (&ring->submit_lock); +list_del (&submit->node); +spin_unlock (&ring->submit_lock); + +msm_gem_submit_put (submit); }static void retire_submits (struct msm_gpu *gpu)
@@ -786,10 +791,6 @@ void msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit) submit->seqno = ++ring->seqno; -spin_lock (&ring->submit_lock); -list_add_tail (&submit->node, &ring->submits); -spin_unlock (&ring->submit_lock); -msm_rd_dump_submit (priv->rd, submit,NULL );update_sw_cntrs (gpu);
@@ -816,6 +817,16 @@ void msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit)msm_gem_active_get (drm_obj, gpu); } +/* +* ring->submits holds a ref to the submit, to deal with the case +* that a submit completes before msm_ioctl_gem_submit() returns. +*/ +msm_gem_submit_get (submit); + +spin_lock (&ring->submit_lock); +list_add_tail (&submit->node, &ring->submits); +spin_unlock (&ring->submit_lock); + gpu->funcs->submit (gpu, submit); priv->lastctx = submit->queue->ctx;
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 16/22 drm/msm: Remove obj->gpu
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
It cannot be atomically updated with obj->active_count, and the only purpose is a useless WARN_ON() (which becomes a buggy WARN_ON() once retire_submits() is not serialized with incoming submits via struct_mutex)
--- drivers/gpu/drm/msm/msm_gem.c | 2 -- drivers/gpu/drm/msm/msm_gem.h | 1 - drivers/gpu/drm/msm/msm_gpu.c | 5 ----- 3 files changed, 8 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +0 -2
@@ -760,7 +760,6 @@ void msm_gem_active_get(struct drm_gem_object *obj, struct msm_gpu *gpu)if (!atomic_fetch_inc (&msm_obj->active_count)) {mutex_lock (&priv->mm_lock); - msm_obj->gpu = gpu;list_del_init (&msm_obj->mm_list);list_add_tail (&msm_obj->mm_list, &gpu->active_list);mutex_unlock (&priv->mm_lock);
@@ -776,7 +775,6 @@ void msm_gem_active_put(struct drm_gem_object *obj)if (!atomic_dec_return (&msm_obj->active_count)) {mutex_lock (&priv->mm_lock); - msm_obj->gpu =NULL ;list_del_init (&msm_obj->mm_list);list_add_tail (&msm_obj->mm_list, &priv->inactive_list);mutex_unlock (&priv->mm_lock);
drivers/gpu/drm/msm/msm_gem.h +0 -1
@@ -64,7 +64,6 @@ struct msm_gem_object {* */ struct list_head mm_list; -struct msm_gpu *gpu;/* non-null if active */ /* Transiently in the process of submit ioctl, objects associated * with the submit are on submit->bo_list.. this only lasts for
drivers/gpu/drm/msm/msm_gpu.c +0 -5
@@ -800,11 +800,6 @@ void msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit)struct drm_gem_object *drm_obj = &msm_obj->base;uint64_t iova; -/* can't happen yet.. but when we add 2d support we'll have -* to deal w/ cross-ring synchronization: -*/ -WARN_ON (is_active (msm_obj) && (msm_obj->gpu != gpu)); -/* submit takes a reference to the bo and iova until retired: */ drm_gem_object_get (&msm_obj->base);msm_gem_get_and_pin_iova_locked (&msm_obj->base, submit->aspace, &iova);
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 17/22 drm/msm: Drop struct_mutex from the retire path
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
Now that we are not relying on dev->struct_mutex to protect the ring->submits lists, drop the struct_mutex lock.
--- drivers/gpu/drm/msm/msm_gpu.c | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-)
drivers/gpu/drm/msm/msm_gpu.c +1 -7
@@ -707,7 +707,7 @@ static void retire_submit(struct msm_gpu *gpu, struct msm_ringbuffer *ring,msm_gem_active_put (&msm_obj->base);msm_gem_unpin_iova (&msm_obj->base, submit->aspace); -drm_gem_object_put_locked (&msm_obj->base); +drm_gem_object_put (&msm_obj->base); }pm_runtime_mark_last_busy (&gpu->pdev->dev);
@@ -722,11 +722,8 @@ static void retire_submit(struct msm_gpu *gpu, struct msm_ringbuffer *ring,static void retire_submits (struct msm_gpu *gpu) { -struct drm_device *dev = gpu->dev;int i; -WARN_ON (!mutex_is_locked (&dev->struct_mutex)); -/* Retire the commits starting with highest priority */ for (i =0 ; i < gpu->nr_rings; i++) {struct msm_ringbuffer *ring = gpu->rb[i];
@@ -756,15 +753,12 @@ static void retire_submits(struct msm_gpu *gpu)static void retire_worker (struct work_struct *work) { struct msm_gpu *gpu = container_of(work, struct msm_gpu, retire_work); -struct drm_device *dev = gpu->dev;int i;for (i =0 ; i < gpu->nr_rings; i++)update_fences (gpu, gpu->rb[i], gpu->rb[i]->memptrs->fence); -mutex_lock (&dev->struct_mutex);retire_submits (gpu); -mutex_unlock (&dev->struct_mutex); }/* call from irq handler to schedule work to retire bo's */
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 18/22 drm/msm: Drop struct_mutex in free_object() path
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
Now that active_list/inactive_list is protected by mm_lock, we no longer need dev->struct_mutex in the free_object() path.
--- drivers/gpu/drm/msm/msm_gem.c | 8 -------- 1 file changed, 8 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +0 -8
@@ -934,8 +934,6 @@ static void free_object(struct msm_gem_object *msm_obj)struct drm_device *dev = obj->dev;struct msm_drm_private *priv = dev->dev_private; -WARN_ON (!mutex_is_locked (&dev->struct_mutex)); -/* object should not be on active list: */ WARN_ON (is_active (msm_obj));
@@ -972,20 +970,14 @@ void msm_gem_free_work(struct work_struct *work) {struct msm_drm_private *priv = container_of(work, struct msm_drm_private, free_work); -struct drm_device *dev = priv->dev;struct llist_node *freed;struct msm_gem_object *msm_obj, *next;while ((freed =llist_del_all (&priv->free_list))) { - -mutex_lock (&dev->struct_mutex); -llist_for_each_entry_safe (msm_obj, next, freed, freed)free_object (msm_obj); -mutex_unlock (&dev->struct_mutex); -if (need_resched ())break ; }
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 19/22 drm/msm: remove msm_gem_free_work
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
Now that we don't need struct_mutex in the free path, we can get rid of the asynchronous free altogether.
--- drivers/gpu/drm/msm/msm_drv.c | 3 --- drivers/gpu/drm/msm/msm_drv.h | 5 ----- drivers/gpu/drm/msm/msm_gem.c | 27 --------------------------- drivers/gpu/drm/msm/msm_gem.h | 1 - 4 files changed, 36 deletions(-)
drivers/gpu/drm/msm/msm_drv.c +0 -3
@@ -438,9 +438,6 @@ static int msm_drm_init(struct device *dev, struct drm_driver *drv) priv->wq =alloc_ordered_workqueue ("msm" ,0 ); -INIT_WORK (&priv->free_work, msm_gem_free_work); -init_llist_head (&priv->free_list); -INIT_LIST_HEAD (&priv->inactive_list);mutex_init (&priv->mm_lock);
drivers/gpu/drm/msm/msm_drv.h +0 -5
@@ -188,10 +188,6 @@ struct msm_drm_private {struct list_head inactive_list;struct mutex mm_lock; -/* worker for delayed free of objects: */ -struct work_struct free_work; -struct llist_head free_list; -struct workqueue_struct *wq;unsigned int num_planes;
@@ -291,7 +287,6 @@ struct drm_gem_object *msm_gem_prime_import_sg_table(struct drm_device *dev, struct dma_buf_attachment *attach, struct sg_table *sg);int msm_gem_prime_pin (struct drm_gem_object *obj);void msm_gem_prime_unpin (struct drm_gem_object *obj); -void msm_gem_free_work (struct work_struct *work);int msm_framebuffer_prepare (struct drm_framebuffer *fb,struct msm_gem_address_space *aspace);
drivers/gpu/drm/msm/msm_gem.c +0 -27
@@ -924,16 +924,6 @@ void msm_gem_free_object(struct drm_gem_object *obj)struct drm_device *dev = obj->dev;struct msm_drm_private *priv = dev->dev_private; -if (llist_add (&msm_obj->freed, &priv->free_list)) -queue_work (priv->wq, &priv->free_work); -} - -static void free_object (struct msm_gem_object *msm_obj) -{ -struct drm_gem_object *obj = &msm_obj->base; -struct drm_device *dev = obj->dev; -struct msm_drm_private *priv = dev->dev_private; -/* object should not be on active list: */ WARN_ON (is_active (msm_obj));
@@ -966,23 +956,6 @@ static void free_object(struct msm_gem_object *msm_obj)kfree (msm_obj); } -void msm_gem_free_work (struct work_struct *work) -{ -struct msm_drm_private *priv = - container_of(work, struct msm_drm_private, free_work); -struct llist_node *freed; -struct msm_gem_object *msm_obj, *next; - -while ((freed =llist_del_all (&priv->free_list))) { -llist_for_each_entry_safe (msm_obj, next, - freed, freed) -free_object (msm_obj); - -if (need_resched ()) -break ; - } -} -/* convenience method to construct a GEM buffer object, and userspace handle */ int msm_gem_new_handle(struct drm_device *dev, struct drm_file *file, uint32_t size, uint32_t flags, uint32_t *handle,
drivers/gpu/drm/msm/msm_gem.h +0 -1
@@ -202,7 +202,6 @@ static inline bool is_vunmapable(struct msm_gem_object *msm_obj)void msm_gem_purge (struct drm_gem_object *obj);void msm_gem_vunmap (struct drm_gem_object *obj); -void msm_gem_free_work (struct work_struct *work);/* Created per submit-ioctl, to track bo's and cmdstream bufs, etc, * associated with the cmdstream submission for synchronization (and
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 20/22 drm/msm: drop struct_mutex in madvise path
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
The obj->lock is sufficient for what we need.
This *does* have the implication that userspace can try to shoot themselves in the foot by racing madvise(DONTNEED) with submit. But the result will be about the same if they did madvise(DONTNEED) before the submit ioctl, ie. they might not get want they want if they race with shrinker. But iova fault handling is robust enough, and userspace is only shooting it's own foot.
--- drivers/gpu/drm/msm/msm_drv.c | 11 ++--------- drivers/gpu/drm/msm/msm_gem.c | 4 +--- drivers/gpu/drm/msm/msm_gem.h | 2 -- 3 files changed, 3 insertions(+), 14 deletions(-)
drivers/gpu/drm/msm/msm_drv.c +2 -9
@@ -912,14 +912,9 @@ static int msm_ioctl_gem_madvise(struct drm_device *dev, void *data,return -EINVAL ; } - ret =mutex_lock_interruptible (&dev->struct_mutex); -if (ret) -return ret; - obj =drm_gem_object_lookup (file, args->handle);if (!obj) { - ret = -ENOENT ; - goto unlock; +return -ENOENT ; } ret =msm_gem_madvise (obj, args->madv);
@@ -928,10 +923,8 @@ static int msm_ioctl_gem_madvise(struct drm_device *dev, void *data, ret =0 ; } -drm_gem_object_put_locked (obj); +drm_gem_object_put (obj); -unlock: -mutex_unlock (&dev->struct_mutex);return ret; }
drivers/gpu/drm/msm/msm_gem.c +1 -3
@@ -658,8 +658,6 @@ int msm_gem_madvise(struct drm_gem_object *obj, unsigned madv)msm_gem_lock (obj); -WARN_ON (!mutex_is_locked (&obj->dev->struct_mutex)); -if (msm_obj->madv != __MSM_MADV_PURGED) msm_obj->madv = madv;
@@ -676,7 +674,6 @@ void msm_gem_purge(struct drm_gem_object *obj)struct msm_gem_object *msm_obj =to_msm_bo (obj);WARN_ON (!mutex_is_locked (&dev->struct_mutex)); -WARN_ON (!msm_gem_is_locked (obj));WARN_ON (!is_purgeable (msm_obj));WARN_ON (obj->import_attach);
@@ -756,6 +753,7 @@ void msm_gem_active_get(struct drm_gem_object *obj, struct msm_gpu *gpu)struct msm_drm_private *priv = obj->dev->dev_private;might_sleep (); +WARN_ON (!msm_gem_is_locked (obj));WARN_ON (msm_obj->madv !=MSM_MADV_WILLNEED );if (!atomic_fetch_inc (&msm_obj->active_count)) {
drivers/gpu/drm/msm/msm_gem.h +0 -2
@@ -188,8 +188,6 @@ static inline bool is_active(struct msm_gem_object *msm_obj)static inline bool is_purgeable (struct msm_gem_object *msm_obj) { -WARN_ON (!msm_gem_is_locked (&msm_obj->base)); -WARN_ON (!mutex_is_locked (&msm_obj->base.dev->struct_mutex));return (msm_obj->madv ==MSM_MADV_DONTNEED ) && msm_obj->sgt && !msm_obj->base.dma_buf && !msm_obj->base.import_attach; }
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 21/22 drm/msm: Drop struct_mutex in shrinker path
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
Now that the inactive_list is protected by mm_lock, and everything else on per-obj basis is protected by obj->lock, we no longer depend on struct_mutex.
--- drivers/gpu/drm/msm/msm_gem.c | 1 - drivers/gpu/drm/msm/msm_gem_shrinker.c | 54 -------------------------- 2 files changed, 55 deletions(-)
drivers/gpu/drm/msm/msm_gem.c +0 -1
@@ -673,7 +673,6 @@ void msm_gem_purge(struct drm_gem_object *obj)struct drm_device *dev = obj->dev;struct msm_gem_object *msm_obj =to_msm_bo (obj); -WARN_ON (!mutex_is_locked (&dev->struct_mutex));WARN_ON (!is_purgeable (msm_obj));WARN_ON (obj->import_attach);
drivers/gpu/drm/msm/msm_gem_shrinker.c +0 -54
101 lines not on this page · raw message
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
PATCH v2 22/22 drm/msm: Don't implicit-sync if only a single ring
In reply to Rob ClarkFrom: Rob Clark <[email protected]>
Any cross-device sync use-cases *must* use explicit sync. And if there is only a single ring (no-preemption), everything is FIFO order and there is no need to implicit-sync.
Mesa should probably just always use MSM_SUBMIT_NO_IMPLICIT, as behavior is undefined when fences are not used to synchronize buffer usage across contexts (which is the only case where multiple different priority rings could come into play).
--- drivers/gpu/drm/msm/msm_gem_submit.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)
drivers/gpu/drm/msm/msm_gem_submit.c +4 -3
@@ -277,7 +277,7 @@ static int submit_lock_objects(struct msm_gem_submit *submit)return ret; } -static int submit_fence_sync (struct msm_gem_submit *submit,bool no_implicit) +static int submit_fence_sync (struct msm_gem_submit *submit,bool implicit_sync) {int i, ret =0 ;
@@ -297,7 +297,7 @@ static int submit_fence_sync(struct msm_gem_submit *submit, bool no_implicit)return ret; } -if (no_implicit) +if (!implicit_sync)continue ; ret = msm_gem_sync_object(&msm_obj->base, submit->ring->fctx,
@@ -768,7 +768,8 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,if (ret) goto out; - ret =submit_fence_sync (submit, !!(args->flags &MSM_SUBMIT_NO_IMPLICIT )); + ret =submit_fence_sync (submit, (gpu->nr_rings >1 ) && + !(args->flags &MSM_SUBMIT_NO_IMPLICIT ));if (ret) goto out;
Signature · 6 lines
2.26.2 _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
In reply to Rob ClarkQuoted 5 lines · Rob Clark
From: Rob Clark <[email protected]>
Unfortunately, due to an
dev_pm_opplocking interaction withmm->mmap_sem, we need to do pm get before aquiring obj locks, otherwise we can have anger lockdep with the chain:
tbh this sounds like a bug in that subsystem, since it means we cannot use said subsystem in mmap handlers either.
So if you have some remapping unit or need to wake up your gpu to blt the buffer into system memory first, we're toast. That doesn't sound right. So maybe Cc: pm folks and figure out how to fix this long term properly? Imo not a good reason to hold up this patch set, since unwrangling mmap_sem tends to be work ...
-Daniel
Quoted 51 lines
opp_table_lock-->&mm->mmap_sem-->reservation_ww_class_mutexFor an explicit fencing userspace, the impact should be minimal as we do all the fence waits before this point. It could result in some needless resumes in error cases, etc.
Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/msm_gem_submit.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-)drivers/gpu/drm/msm/msm_gem_submit.c +13
-2@@ -744,11 +744,20 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data, ret =submit_lookup_objects (submit, args, file);if (ret) - goto out; + goto out_pre_pm; ret =submit_lookup_cmds (submit, args, file);if (ret) - goto out; + goto out_pre_pm; + +/* +* Thanks to dev_pm_opp opp_table_lock interactions with mm->mmap_sem +* in the resume path, we need to to rpm get before we lock objs. +* Which unfortunately might involve powering up the GPU sooner than +* is necessary. But at least in the explicit fencing case, we will +* have already done all the fence waiting. +*/ +pm_runtime_get_sync (&gpu->pdev->dev);/* copy_*_user while holding a ww ticket upsets lockdep */ ww_acquire_init (&submit->ticket, &reservation_ww_class);@@ -825,6 +834,8 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data, out: +pm_runtime_put (&gpu->pdev->dev); +out_pre_pm:submit_cleanup (submit);if (has_ww_ticket)ww_acquire_fini (&submit->ticket);Signature · 1 line
2.26.2
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH v2 22/22] drm/msm: Don't implicit-sync if only a single ring
In reply to Rob ClarkQuoted 10 lines · Rob Clark
From: Rob Clark <[email protected]>
Any cross-device sync use-cases *must* use explicit sync. And if there is only a single ring (no-preemption), everything is FIFO order and there is no need to implicit-sync.
Mesa should probably just always use
MSM_SUBMIT_NO_IMPLICIT, as behavior is undefined when fences are not used to synchronize buffer usage across contexts (which is the only case where multiple different priority rings could come into play).
Uh does this mean msm is broken on dri2/3 and wayland? Or I'm I just confused by your commit message?
Since for these protocols we do expect implicit sync accross processes to work. Even across devices (and nvidia have actually provided quite a bunch of patches to make this work in i915 - ttm based drivers get this right, plus dumb scanout drivers using the right helpers also get this all right).
-Daniel
Quoted 41 lines
Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/msm_gem_submit.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)drivers/gpu/drm/msm/msm_gem_submit.c +4
-3@@ -277,7 +277,7 @@ static int submit_lock_objects(struct msm_gem_submit *submit)return ret; } -static int submit_fence_sync (struct msm_gem_submit *submit,bool no_implicit) +static int submit_fence_sync (struct msm_gem_submit *submit,bool implicit_sync) {int i, ret =0 ;@@ -297,7 +297,7 @@ static int submit_fence_sync(struct msm_gem_submit *submit, bool no_implicit)return ret; } -if (no_implicit) +if (!implicit_sync)continue ; ret = msm_gem_sync_object(&msm_obj->base, submit->ring->fctx,@@ -768,7 +768,8 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,if (ret) goto out; - ret =submit_fence_sync (submit, !!(args->flags &MSM_SUBMIT_NO_IMPLICIT )); + ret =submit_fence_sync (submit, (gpu->nr_rings >1 ) && + !(args->flags &MSM_SUBMIT_NO_IMPLICIT ));if (ret) goto out;Signature · 1 line
2.26.2
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH v2 22/22] drm/msm: Don't implicit-sync if only a single ring
Quoted 8 lines · Daniel Vetter
Quoted 6 lines · Rob Clark
From: Rob Clark <[email protected]>
Any cross-device sync use-cases *must* use explicit sync. And if there is only a single ring (no-preemption), everything is FIFO order and there is no need to implicit-sync.
Quoted 7 lines
Quoted 4 lines
Mesa should probably just always use
MSM_SUBMIT_NO_IMPLICIT, as behavior is undefined when fences are not used to synchronize buffer usage across contexts (which is the only case where multiple different priority rings could come into play).Uh does this mean msm is broken on dri2/3 and wayland? Or I'm I just confused by your commit message?
No, I don't think so. If there is only a single priority level ringbuffer (ie. no preemption to higher priority ring) then everything is inherently FIFO order.
For cases where we are sharing buffers with something external to drm, explicit sync will be used. And we don't implicit sync with display, otherwise x11 (frontbuffer rendering) would not work
BR,
-R
Quoted 53 lines
Since for these protocols we do expect implicit sync accross processes to work. Even across devices (and nvidia have actually provided quite a bunch of patches to make this work in i915 - ttm based drivers get this right, plus dumb scanout drivers using the right helpers also get this all right).
-DanielQuoted 39 lines
Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/msm_gem_submit.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)drivers/gpu/drm/msm/msm_gem_submit.c +4
-4@@ -277,7 +277,7 @@ static int submit_lock_objects(struct msm_gem_submit *submit)return ret; } -static int submit_fence_sync (struct msm_gem_submit *submit,bool no_implicit) +static int submit_fence_sync (struct msm_gem_submit *submit,bool implicit_sync) {int i, ret =0 ;@@ -297,7 +297,7 @@ static int submit_fence_sync(struct msm_gem_submit *submit, bool no_implicit)return ret; } -if (no_implicit) +if (!implicit_sync)continue ; ret = msm_gem_sync_object(&msm_obj->base, submit->ring->fctx,@@ -768,7 +768,8 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,if (ret) goto out; - ret =submit_fence_sync (submit, !!(args->flags &MSM_SUBMIT_NO_IMPLICIT )); + ret =submit_fence_sync (submit, (gpu->nr_rings >1 ) && + !(args->flags &MSM_SUBMIT_NO_IMPLICIT ));if (ret) goto out; --Quoted 2 lines
2.26.2
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
In reply to Daniel VetterQuoted 8 lines · Daniel Vetter
Quoted 5 lines · Rob Clark
From: Rob Clark <[email protected]>
Unfortunately, due to an
dev_pm_opplocking interaction withmm->mmap_sem, we need to do pm get before aquiring obj locks, otherwise we can have anger lockdep with the chain:
Quoted 8 lines
tbh this sounds like a bug in that subsystem, since it means we cannot use said subsystem in mmap handlers either.
So if you have some remapping unit or need to wake up your gpu to blt the buffer into system memory first, we're toast. That doesn't sound right. So maybe Cc: pm folks and figure out how to fix this long term properly? Imo not a good reason to hold up this patch set, since unwrangling
mmap_semtends to be work ...
+ a couple of PM folks
Looks like it has been this way for quite some time, so I guess the overlap between things using dev_pm_opp and mmap is low..
fwiw, example splat so folks can see the locking interaction I am talking about.. I suspect the pm_opp interaction with mm->mmap_sem is from the debugfs calls while opp_table_lock is held?
[ 15.627855] ======================================================
[ 15.634202] WARNING: possible circular locking dependency detected
[ 15.640550] 5.4.70 #41 Not tainted
[ 15.644050] ------------------------------------------------------
[ 15.650397] chrome/1805 is trying to acquire lock:
[ 15.655314] ffffffed90720738 (opp_table_lock){+.+.}, at:
_find_opp_table+0x34/0x74
[ 15.663092]
[ 15.663092] but task is already holding lock:
[ 15.669082] ffffff80ff3911a8 (reservation_ww_class_mutex){+.+.},
at: submit_lock_objects+0x70/0x1ec
[ 15.678369]
[ 15.678369] which lock already depends on the new lock.
[ 15.678369]
[ 15.686764]
[ 15.686764] the existing dependency chain (in reverse order) is:
[ 15.694438]
[ 15.694438] -> #3 (reservation_ww_class_mutex){+.+.}:
[ 15.701146] __mutex_lock_common+0xec/0xc0c
[ 15.705978] ww_mutex_lock_interruptible+0x5c/0xc4
[ 15.711432] msm_gem_fault+0x2c/0x124
[ 15.715731] __do_fault+0x40/0x16c
[ 15.719766] handle_mm_fault+0x7cc/0xd98
[ 15.724337] do_page_fault+0x230/0x3b4
[ 15.728721] do_translation_fault+0x5c/0x78
[ 15.733558] do_mem_abort+0x4c/0xb4
[ 15.737680] el0_da+0x1c/0x20
[ 15.741266]
[ 15.741266] -> #2 (&mm->mmap_sem){++++}:
[ 15.746809] __might_fault+0x70/0x98
[ 15.751022] compat_filldir+0xf8/0x48c
[ 15.755412] dcache_readdir+0x70/0x1dc
[ 15.759808] iterate_dir+0xd4/0x180
[ 15.763931] __arm64_compat_sys_getdents+0xa0/0x19c
[ 15.769476] el0_svc_common+0xa8/0x178
[ 15.773861] el0_svc_compat_handler+0x2c/0x40
[ 15.778868] el0_svc_compat+0x8/0x10
[ 15.783075]
[ 15.783075] -> #1 (&sb->s_type->i_mutex_key#3){++++}:
[ 15.789788] down_write+0x54/0x16c
[ 15.793826] debugfs_remove_recursive+0x50/0x158
[ 15.799108] opp_debug_unregister+0x34/0x114
[ 15.804028] dev_pm_opp_put_opp_table+0xd0/0x14c
[ 15.809308] dev_pm_opp_put_clkname+0x3c/0x50
[ 15.814318] msm_dsi_host_destroy+0xb0/0xcc
[ 15.819149] dsi_destroy+0x40/0x58
[ 15.823184] dsi_bind+0x90/0x170
[ 15.827041] component_bind_all+0xf0/0x208
[ 15.831787] msm_drm_init+0x188/0x60c
[ 15.836084] msm_drm_bind+0x24/0x30
[ 15.840205] try_to_bring_up_master+0x15c/0x1a4
[ 15.845396] __component_add+0x98/0x14c
[ 15.849878] component_add+0x28/0x34
[ 15.854086] dp_display_probe+0x324/0x370
[ 15.858744] platform_drv_probe+0x90/0xb0
[ 15.863400] really_probe+0x134/0x2ec
[ 15.867699] driver_probe_device+0x64/0xfc
[ 15.872443] __device_attach_driver+0x8c/0xa4
[ 15.877459] bus_for_each_drv+0x90/0xd8
[ 15.881939] __device_attach+0xc0/0x148
[ 15.886420] device_initial_probe+0x20/0x2c
[ 15.891254] bus_probe_device+0x34/0x94
[ 15.895726] deferred_probe_work_func+0x78/0xb4
[ 15.900914] process_one_work+0x30c/0x5d0
[ 15.905573] worker_thread+0x240/0x3f0
[ 15.909959] kthread+0x144/0x154
[ 15.913809] ret_from_fork+0x10/0x18
[ 15.918016]
[ 15.918016] -> #0 (opp_table_lock){+.+.}:
[ 15.923660] __lock_acquire+0xee4/0x2450
[ 15.928230] lock_acquire+0x1cc/0x210
[ 15.932527] __mutex_lock_common+0xec/0xc0c
[ 15.937359] mutex_lock_nested+0x40/0x50
[ 15.941928] _find_opp_table+0x34/0x74
[ 15.946312] dev_pm_opp_find_freq_exact+0x2c/0xdc
[ 15.951680] a6xx_gmu_resume+0xc8/0xecc
[ 15.952812] fscrypt: AES-256-CTS-CBC using implementation "cts-cbc-aes-ce"
[ 15.956161] a6xx_pm_resume+0x148/0x200
[ 15.956166] adreno_resume+0x28/0x34
[ 15.956171] pm_generic_runtime_resume+0x34/0x48
[ 15.956174] __rpm_callback+0x70/0x10c
[ 15.956176] rpm_callback+0x34/0x8c
[ 15.956179] rpm_resume+0x414/0x550
[ 15.956182] __pm_runtime_resume+0x7c/0xa0
[ 15.956185] msm_gpu_submit+0x60/0x1c0
[ 15.956190] msm_ioctl_gem_submit+0xadc/0xb60
[ 16.003961] drm_ioctl_kernel+0x9c/0x118
[ 16.008532] drm_ioctl+0x27c/0x408
[ 16.012562] drm_compat_ioctl+0xcc/0xdc
[ 16.017038] __se_compat_sys_ioctl+0x100/0x206c
[ 16.022224] __arm64_compat_sys_ioctl+0x20/0x2c
[ 16.027412] el0_svc_common+0xa8/0x178
[ 16.031800] el0_svc_compat_handler+0x2c/0x40
[ 16.036810] el0_svc_compat+0x8/0x10
[ 16.041021]
[ 16.041021] other info that might help us debug this:
[ 16.041021]
[ 16.049235] Chain exists of:
[ 16.049235] opp_table_lock --> &mm->mmap_sem --> reservation_ww_class_mutex
[ 16.049235]
[ 16.061014] Possible unsafe locking scenario:
[ 16.061014]
[ 16.067091] CPU0 CPU1
[ 16.071750] ---- ----
[ 16.076399] lock(reservation_ww_class_mutex);
[ 16.081059] lock(&mm->mmap_sem);
[ 16.087134] lock(reservation_ww_class_mutex);
[ 16.094369] lock(opp_table_lock);
[ 16.097961]
[ 16.097961] *** DEADLOCK ***
[ 16.097961]
[ 16.104038] 3 locks held by chrome/1805:
[ 16.108068] #0: ffffff80fb20c0d8 (&dev->struct_mutex){+.+.}, at:
msm_ioctl_gem_submit+0x264/0xb60
[ 16.117264] #1: ffffff80dd712c70
(reservation_ww_class_acquire){+.+.}, at:
msm_ioctl_gem_submit+0x8e8/0xb60
[ 16.127357] #2: ffffff80ff3911a8
(reservation_ww_class_mutex){+.+.}, at: submit_lock_objects+0x70/0x1ec
[ 16.137089]
[ 16.137089] stack backtrace:
[ 16.141567] CPU: 4 PID: 1805 Comm: chrome Not tainted 5.4.70 #41
[ 16.147733] Hardware name: Google Lazor (rev1+) with LTE (DT)
[ 16.153632] Call trace:
[ 16.156154] dump_backtrace+0x0/0x158
[ 16.159924] show_stack+0x20/0x2c
[ 16.163340] dump_stack+0xc8/0x160
[ 16.166840] print_circular_bug+0x2c4/0x2c8
[ 16.171144] check_noncircular+0x1a8/0x1b0
[ 16.175351] __lock_acquire+0xee4/0x2450
[ 16.179382] lock_acquire+0x1cc/0x210
[ 16.183146] __mutex_lock_common+0xec/0xc0c
[ 16.187450] mutex_lock_nested+0x40/0x50
[ 16.191481] _find_opp_table+0x34/0x74
[ 16.195344] dev_pm_opp_find_freq_exact+0x2c/0xdc
[ 16.200178] a6xx_gmu_resume+0xc8/0xecc
[ 16.204120] a6xx_pm_resume+0x148/0x200
[ 16.208064] adreno_resume+0x28/0x34
[ 16.211743] pm_generic_runtime_resume+0x34/0x48
[ 16.216488] __rpm_callback+0x70/0x10c
[ 16.220342] rpm_callback+0x34/0x8c
[ 16.223933] rpm_resume+0x414/0x550
[ 16.227524] __pm_runtime_resume+0x7c/0xa0
[ 16.231731] msm_gpu_submit+0x60/0x1c0
[ 16.235586] msm_ioctl_gem_submit+0xadc/0xb60
[ 16.240066] drm_ioctl_kernel+0x9c/0x118
[ 16.244097] drm_ioctl+0x27c/0x408
[ 16.247602] drm_compat_ioctl+0xcc/0xdc
[ 16.251546] __se_compat_sys_ioctl+0x100/0x206c
[ 16.256204] __arm64_compat_sys_ioctl+0x20/0x2c
[ 16.260861] el0_svc_common+0xa8/0x178
[ 16.264716] el0_svc_compat_handler+0x2c/0x40
[ 16.269196] el0_svc_compat+0x8/0x10BR,
-R
Quoted 58 lines
-Daniel
Quoted 49 lines
opp_table_lock-->&mm->mmap_sem-->reservation_ww_class_mutexFor an explicit fencing userspace, the impact should be minimal as we do all the fence waits before this point. It could result in some needless resumes in error cases, etc.
Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/msm_gem_submit.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-)drivers/gpu/drm/msm/msm_gem_submit.c +13
-3@@ -744,11 +744,20 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data, ret =submit_lookup_objects (submit, args, file);if (ret) - goto out; + goto out_pre_pm; ret =submit_lookup_cmds (submit, args, file);if (ret) - goto out; + goto out_pre_pm; + +/* +* Thanks to dev_pm_opp opp_table_lock interactions with mm->mmap_sem +* in the resume path, we need to to rpm get before we lock objs. +* Which unfortunately might involve powering up the GPU sooner than +* is necessary. But at least in the explicit fencing case, we will +* have already done all the fence waiting. +*/ +pm_runtime_get_sync (&gpu->pdev->dev);/* copy_*_user while holding a ww ticket upsets lockdep */ ww_acquire_init (&submit->ticket, &reservation_ww_class);@@ -825,6 +834,8 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data, out: +pm_runtime_put (&gpu->pdev->dev); +out_pre_pm:submit_cleanup (submit);if (has_ww_ticket)ww_acquire_fini (&submit->ticket); --Quoted 2 lines
2.26.2
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
Re: [PATCH v2 22/22] drm/msm: Don't implicit-sync if only a single ring
In reply to Rob ClarkQuoted 14 lines · Rob Clark
Quoted 13 lines · Daniel Vetter
Quoted 10 lines · Rob Clark
From: Rob Clark <[email protected]>
Any cross-device sync use-cases *must* use explicit sync. And if there is only a single ring (no-preemption), everything is FIFO order and there is no need to implicit-sync.
Mesa should probably just always use
MSM_SUBMIT_NO_IMPLICIT, as behavior is undefined when fences are not used to synchronize buffer usage across contexts (which is the only case where multiple different priority rings could come into play).
Quoted 6 lines
Quoted 2 lines
Uh does this mean msm is broken on dri2/3 and wayland? Or I'm I just confused by your commit message?
No, I don't think so. If there is only a single priority level ringbuffer (ie. no preemption to higher priority ring) then everything is inherently FIFO order.
Well eventually you get a scheduler I guess/hope :-)
Quoted 3 lines
For cases where we are sharing buffers with something external to drm, explicit sync will be used. And we don't implicit sync with display, otherwise x11 (frontbuffer rendering) would not work
Uh now I'm even more confused. The implicit sync fences in dma_resv are kinda for everyone. That's also why dma_resv with the common locking approach is a useful idea.
So display should definitely support implicit sync, and iirc msm does have the helper hooked up.
Wrt other subsystems, I guess passing dma_fence around somehow doesn't fit into v4l (the patches never landed), so v4l doesn't do any kind of sync right now. But this could be fixed. Not sure what else is going on.
So I guess I still have no idea why you put that into the commit message.
btw for what you're trying to do yourself, the way to do this is to allocate a fence timeline for your engine, compare fences, and no-op them all out if their own the same timeline.
-Daniel
Quoted 57 lines
BR,
-RQuoted 53 lines
Since for these protocols we do expect implicit sync accross processes to work. Even across devices (and nvidia have actually provided quite a bunch of patches to make this work in i915 - ttm based drivers get this right, plus dumb scanout drivers using the right helpers also get this all right).
-DanielQuoted 39 lines
Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/msm_gem_submit.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)drivers/gpu/drm/msm/msm_gem_submit.c +4
-4@@ -277,7 +277,7 @@ static int submit_lock_objects(struct msm_gem_submit *submit)return ret; } -static int submit_fence_sync (struct msm_gem_submit *submit,bool no_implicit) +static int submit_fence_sync (struct msm_gem_submit *submit,bool implicit_sync) {int i, ret =0 ;@@ -297,7 +297,7 @@ static int submit_fence_sync(struct msm_gem_submit *submit, bool no_implicit)return ret; } -if (no_implicit) +if (!implicit_sync)continue ; ret = msm_gem_sync_object(&msm_obj->base, submit->ring->fctx,@@ -768,7 +768,8 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,if (ret) goto out; - ret =submit_fence_sync (submit, !!(args->flags &MSM_SUBMIT_NO_IMPLICIT )); + ret =submit_fence_sync (submit, (gpu->nr_rings >1 ) && + !(args->flags &MSM_SUBMIT_NO_IMPLICIT ));if (ret) goto out; --Quoted 2 lines
2.26.2
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch
Re: [Freedreno] [PATCH v2 22/22] drm/msm: Don't implicit-sync if only a single ring
Quoted 16 lines · Daniel Vetter
Quoted 14 lines · Rob Clark
Quoted 13 lines · Daniel Vetter
Quoted 10 lines · Rob Clark
From: Rob Clark <[email protected]>
Any cross-device sync use-cases *must* use explicit sync. And if there is only a single ring (no-preemption), everything is FIFO order and there is no need to implicit-sync.
Mesa should probably just always use
MSM_SUBMIT_NO_IMPLICIT, as behavior is undefined when fences are not used to synchronize buffer usage across contexts (which is the only case where multiple different priority rings could come into play).
Quoted 8 lines
Quoted 6 lines
Quoted 2 lines
Uh does this mean msm is broken on dri2/3 and wayland? Or I'm I just confused by your commit message?
No, I don't think so. If there is only a single priority level ringbuffer (ie. no preemption to higher priority ring) then everything is inherently FIFO order.
Well eventually you get a scheduler I guess/hope :-)
we do have one currently for some gens, but not others.. hence the check for # of rings. (Ie. there is a ring per priority level, if only one ring, that means no preemption/scheduler)
Quoted 10 lines
Quoted 3 lines
For cases where we are sharing buffers with something external to drm, explicit sync will be used. And we don't implicit sync with display, otherwise x11 (frontbuffer rendering) would not work
Uh now I'm even more confused. The implicit sync fences in
dma_resvare kinda for everyone. That's also whydma_resvwith the common locking approach is a useful idea.So display should definitely support implicit sync, and iirc msm does have the helper hooked up.
yup
Quoted 9 lines
Wrt other subsystems, I guess passing
dma_fencearound somehow doesn't fit into v4l (the patches never landed), so v4l doesn't do any kind of sync right now. But this could be fixed. Not sure what else is going on.So I guess I still have no idea why you put that into the commit message.
btw for what you're trying to do yourself, the way to do this is to allocate a fence timeline for your engine, compare fences, and no-op them all out if their own the same timeline.
we do that already (with a fence timeline per-ring, in the case of gens which support multiple rings / preemption).. this patch just short-circuits that in the case where we already knows the fences will of the same timeline
BR,
-R
Quoted 68 lines
-Daniel
Quoted 57 lines
BR,
-RQuoted 53 lines
Since for these protocols we do expect implicit sync accross processes to work. Even across devices (and nvidia have actually provided quite a bunch of patches to make this work in i915 - ttm based drivers get this right, plus dumb scanout drivers using the right helpers also get this all right).
-DanielQuoted 39 lines
Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/msm_gem_submit.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)drivers/gpu/drm/msm/msm_gem_submit.c +4
-4@@ -277,7 +277,7 @@ static int submit_lock_objects(struct msm_gem_submit *submit)return ret; } -static int submit_fence_sync (struct msm_gem_submit *submit,bool no_implicit) +static int submit_fence_sync (struct msm_gem_submit *submit,bool implicit_sync) {int i, ret =0 ;@@ -297,7 +297,7 @@ static int submit_fence_sync(struct msm_gem_submit *submit, bool no_implicit)return ret; } -if (no_implicit) +if (!implicit_sync)continue ; ret = msm_gem_sync_object(&msm_obj->base, submit->ring->fctx,@@ -768,7 +768,8 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,if (ret) goto out; - ret =submit_fence_sync (submit, !!(args->flags &MSM_SUBMIT_NO_IMPLICIT )); + ret =submit_fence_sync (submit, (gpu->nr_rings >1 ) && + !(args->flags &MSM_SUBMIT_NO_IMPLICIT ));if (ret) goto out; --Quoted 2 lines
2.26.2
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno
Re: [Freedreno] [PATCH v2 22/22] drm/msm: Don't implicit-sync if only a single ring
Quoted 25 lines · Rob Clark
Quoted 23 lines · Daniel Vetter
Quoted 20 lines · Rob Clark
Quoted 8 lines · Daniel Vetter
Quoted 6 lines · Rob Clark
From: Rob Clark <[email protected]>
Any cross-device sync use-cases *must* use explicit sync. And if there is only a single ring (no-preemption), everything is FIFO order and there is no need to implicit-sync.
Quoted 7 lines
Quoted 4 lines
Mesa should probably just always use
MSM_SUBMIT_NO_IMPLICIT, as behavior is undefined when fences are not used to synchronize buffer usage across contexts (which is the only case where multiple different priority rings could come into play).Uh does this mean msm is broken on dri2/3 and wayland? Or I'm I just confused by your commit message?
No, I don't think so. If there is only a single priority level ringbuffer (ie. no preemption to higher priority ring) then everything is inherently FIFO order.
Quoted 5 lines
Quoted 1 line
Well eventually you get a scheduler I guess/hope :-)
we do have one currently for some gens, but not others.. hence the check for # of rings. (Ie. there is a ring per priority level, if only one ring, that means no preemption/scheduler)
Even without preempt a scheduler is somewhat useful, if you have a very spammy client. Of course it assumes that everyone submits reasonably short workloads, otherwise nothing you can do.
Quoted 19 lines
Quoted 10 lines
Quoted 3 lines
For cases where we are sharing buffers with something external to drm, explicit sync will be used. And we don't implicit sync with display, otherwise x11 (frontbuffer rendering) would not work
Uh now I'm even more confused. The implicit sync fences in
dma_resvare kinda for everyone. That's also whydma_resvwith the common locking approach is a useful idea.So display should definitely support implicit sync, and iirc msm does have the helper hooked up.
yup
Quoted 6 lines
Wrt other subsystems, I guess passing
dma_fencearound somehow doesn't fit into v4l (the patches never landed), so v4l doesn't do any kind of sync right now. But this could be fixed. Not sure what else is going on.So I guess I still have no idea why you put that into the commit message.
Quoted 8 lines
Quoted 3 lines
btw for what you're trying to do yourself, the way to do this is to allocate a fence timeline for your engine, compare fences, and no-op them all out if their own the same timeline.
we do that already (with a fence timeline per-ring, in the case of gens which support multiple rings / preemption).. this patch just short-circuits that in the case where we already knows the fences will of the same timeline
Ok so I think it's all good, no misunderstanding, but the commit message. I think if you delete the first sentence that cross-device sync must use explicit fences then it all makes sense and is consistent. Or clarify it that this is cross-engine sync with explicit internal synchronization, to differentiate it against cross-device sync (as seen by userspace, like different drm_device instances) and explicit dma_fence synchronization controlled by userspace. -Daniel
Quoted 71 lines
BR,
-RQuoted 68 lines
-Daniel
Quoted 57 lines
BR,
-RQuoted 53 lines
Since for these protocols we do expect implicit sync accross processes to work. Even across devices (and nvidia have actually provided quite a bunch of patches to make this work in i915 - ttm based drivers get this right, plus dumb scanout drivers using the right helpers also get this all right).
-DanielQuoted 39 lines
Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/msm_gem_submit.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)drivers/gpu/drm/msm/msm_gem_submit.c +4
-4@@ -277,7 +277,7 @@ static int submit_lock_objects(struct msm_gem_submit *submit)return ret; } -static int submit_fence_sync (struct msm_gem_submit *submit,bool no_implicit) +static int submit_fence_sync (struct msm_gem_submit *submit,bool implicit_sync) {int i, ret =0 ;@@ -297,7 +297,7 @@ static int submit_fence_sync(struct msm_gem_submit *submit, bool no_implicit)return ret; } -if (no_implicit) +if (!implicit_sync)continue ; ret = msm_gem_sync_object(&msm_obj->base, submit->ring->fctx,@@ -768,7 +768,8 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,if (ret) goto out; - ret =submit_fence_sync (submit, !!(args->flags &MSM_SUBMIT_NO_IMPLICIT )); + ret =submit_fence_sync (submit, (gpu->nr_rings >1 ) && + !(args->flags &MSM_SUBMIT_NO_IMPLICIT ));if (ret) goto out; --Quoted 2 lines
2.26.2
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
_______________________________________________
Freedreno mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/freedreno
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
In reply to Rob ClarkQuoted 18 lines · Rob Clark
Quoted 8 lines · Daniel Vetter
Quoted 5 lines · Rob Clark
From: Rob Clark <[email protected]>
Unfortunately, due to an
dev_pm_opplocking interaction withmm->mmap_sem, we need to do pm get before aquiring obj locks, otherwise we can have anger lockdep with the chain:Quoted 8 lines
tbh this sounds like a bug in that subsystem, since it means we cannot use said subsystem in mmap handlers either.
So if you have some remapping unit or need to wake up your gpu to blt the buffer into system memory first, we're toast. That doesn't sound right. So maybe Cc: pm folks and figure out how to fix this long term properly? Imo not a good reason to hold up this patch set, since unwrangling
mmap_semtends to be work ...
Quoted 8 lines
+ a couple of PM folks
Looks like it has been this way for quite some time, so I guess the overlap between things using
dev_pm_oppand mmap is low..fwiw, example splat so folks can see the locking interaction I am talking about.. I suspect the
pm_oppinteraction withmm->mmap_semis from the debugfs calls whileopp_table_lockis held?
I am not very sure about why this circular locking dependency is happening here and how exactly can we fix it. The OPP core works under the opp_table_lock, from within which it creates/remove the debugfs stuff as well.
Quoted 207 lines
[ 15.627855] ====================================================== [ 15.634202] WARNING: possible circular locking dependency detected [ 15.640550] 5.4.70 #41 Not tainted [ 15.644050] ------------------------------------------------------ [ 15.650397] chrome/1805 is trying to acquire lock: [ 15.655314] ffffffed90720738 (opp_table_lock){+.+.}, at: _find_opp_table+0x34/0x74 [ 15.663092] [ 15.663092] but task is already holding lock: [ 15.669082] ffffff80ff3911a8 (reservation_ww_class_mutex){+.+.}, at: submit_lock_objects+0x70/0x1ec [ 15.678369] [ 15.678369] which lock already depends on the new lock. [ 15.678369] [ 15.686764] [ 15.686764] the existing dependency chain (in reverse order) is: [ 15.694438] [ 15.694438] -> #3 (reservation_ww_class_mutex){+.+.}: [ 15.701146] __mutex_lock_common+0xec/0xc0c [ 15.705978] ww_mutex_lock_interruptible+0x5c/0xc4 [ 15.711432] msm_gem_fault+0x2c/0x124 [ 15.715731] __do_fault+0x40/0x16c [ 15.719766] handle_mm_fault+0x7cc/0xd98 [ 15.724337] do_page_fault+0x230/0x3b4 [ 15.728721] do_translation_fault+0x5c/0x78 [ 15.733558] do_mem_abort+0x4c/0xb4 [ 15.737680] el0_da+0x1c/0x20 [ 15.741266] [ 15.741266] -> #2 (&mm->mmap_sem){++++}: [ 15.746809] __might_fault+0x70/0x98 [ 15.751022] compat_filldir+0xf8/0x48c [ 15.755412] dcache_readdir+0x70/0x1dc [ 15.759808] iterate_dir+0xd4/0x180 [ 15.763931] __arm64_compat_sys_getdents+0xa0/0x19c [ 15.769476] el0_svc_common+0xa8/0x178 [ 15.773861] el0_svc_compat_handler+0x2c/0x40 [ 15.778868] el0_svc_compat+0x8/0x10 [ 15.783075] [ 15.783075] -> #1 (&sb->s_type->i_mutex_key#3){++++}: [ 15.789788] down_write+0x54/0x16c [ 15.793826] debugfs_remove_recursive+0x50/0x158 [ 15.799108] opp_debug_unregister+0x34/0x114 [ 15.804028] dev_pm_opp_put_opp_table+0xd0/0x14c [ 15.809308] dev_pm_opp_put_clkname+0x3c/0x50 [ 15.814318] msm_dsi_host_destroy+0xb0/0xcc [ 15.819149] dsi_destroy+0x40/0x58 [ 15.823184] dsi_bind+0x90/0x170 [ 15.827041] component_bind_all+0xf0/0x208 [ 15.831787] msm_drm_init+0x188/0x60c [ 15.836084] msm_drm_bind+0x24/0x30 [ 15.840205] try_to_bring_up_master+0x15c/0x1a4 [ 15.845396] __component_add+0x98/0x14c [ 15.849878] component_add+0x28/0x34 [ 15.854086] dp_display_probe+0x324/0x370 [ 15.858744] platform_drv_probe+0x90/0xb0 [ 15.863400] really_probe+0x134/0x2ec [ 15.867699] driver_probe_device+0x64/0xfc [ 15.872443] __device_attach_driver+0x8c/0xa4 [ 15.877459] bus_for_each_drv+0x90/0xd8 [ 15.881939] __device_attach+0xc0/0x148 [ 15.886420] device_initial_probe+0x20/0x2c [ 15.891254] bus_probe_device+0x34/0x94 [ 15.895726] deferred_probe_work_func+0x78/0xb4 [ 15.900914] process_one_work+0x30c/0x5d0 [ 15.905573] worker_thread+0x240/0x3f0 [ 15.909959] kthread+0x144/0x154 [ 15.913809] ret_from_fork+0x10/0x18 [ 15.918016] [ 15.918016] -> #0 (opp_table_lock){+.+.}: [ 15.923660] __lock_acquire+0xee4/0x2450 [ 15.928230] lock_acquire+0x1cc/0x210 [ 15.932527] __mutex_lock_common+0xec/0xc0c [ 15.937359] mutex_lock_nested+0x40/0x50 [ 15.941928] _find_opp_table+0x34/0x74 [ 15.946312] dev_pm_opp_find_freq_exact+0x2c/0xdc [ 15.951680] a6xx_gmu_resume+0xc8/0xecc [ 15.952812] fscrypt: AES-256-CTS-CBC using implementation "cts-cbc-aes-ce" [ 15.956161] a6xx_pm_resume+0x148/0x200 [ 15.956166] adreno_resume+0x28/0x34 [ 15.956171] pm_generic_runtime_resume+0x34/0x48 [ 15.956174] __rpm_callback+0x70/0x10c [ 15.956176] rpm_callback+0x34/0x8c [ 15.956179] rpm_resume+0x414/0x550 [ 15.956182] __pm_runtime_resume+0x7c/0xa0 [ 15.956185] msm_gpu_submit+0x60/0x1c0 [ 15.956190] msm_ioctl_gem_submit+0xadc/0xb60 [ 16.003961] drm_ioctl_kernel+0x9c/0x118 [ 16.008532] drm_ioctl+0x27c/0x408 [ 16.012562] drm_compat_ioctl+0xcc/0xdc [ 16.017038] __se_compat_sys_ioctl+0x100/0x206c [ 16.022224] __arm64_compat_sys_ioctl+0x20/0x2c [ 16.027412] el0_svc_common+0xa8/0x178 [ 16.031800] el0_svc_compat_handler+0x2c/0x40 [ 16.036810] el0_svc_compat+0x8/0x10 [ 16.041021] [ 16.041021] other info that might help us debug this: [ 16.041021] [ 16.049235] Chain exists of: [ 16.049235] opp_table_lock --> &mm->mmap_sem --> reservation_ww_class_mutex [ 16.049235] [ 16.061014] Possible unsafe locking scenario: [ 16.061014] [ 16.067091] CPU0 CPU1 [ 16.071750] ---- ---- [ 16.076399] lock(reservation_ww_class_mutex); [ 16.081059] lock(&mm->mmap_sem); [ 16.087134] lock(reservation_ww_class_mutex); [ 16.094369] lock(opp_table_lock); [ 16.097961] [ 16.097961] *** DEADLOCK *** [ 16.097961] [ 16.104038] 3 locks held by chrome/1805: [ 16.108068] #0: ffffff80fb20c0d8 (&dev->struct_mutex){+.+.}, at: msm_ioctl_gem_submit+0x264/0xb60 [ 16.117264] #1: ffffff80dd712c70 (reservation_ww_class_acquire){+.+.}, at: msm_ioctl_gem_submit+0x8e8/0xb60 [ 16.127357] #2: ffffff80ff3911a8 (reservation_ww_class_mutex){+.+.}, at: submit_lock_objects+0x70/0x1ec [ 16.137089] [ 16.137089] stack backtrace: [ 16.141567] CPU: 4 PID: 1805 Comm: chrome Not tainted 5.4.70 #41 [ 16.147733] Hardware name: Google Lazor (rev1+) with LTE (DT) [ 16.153632] Call trace: [ 16.156154] dump_backtrace+0x0/0x158 [ 16.159924] show_stack+0x20/0x2c [ 16.163340] dump_stack+0xc8/0x160 [ 16.166840] print_circular_bug+0x2c4/0x2c8 [ 16.171144] check_noncircular+0x1a8/0x1b0 [ 16.175351] __lock_acquire+0xee4/0x2450 [ 16.179382] lock_acquire+0x1cc/0x210 [ 16.183146] __mutex_lock_common+0xec/0xc0c [ 16.187450] mutex_lock_nested+0x40/0x50 [ 16.191481] _find_opp_table+0x34/0x74 [ 16.195344] dev_pm_opp_find_freq_exact+0x2c/0xdc [ 16.200178] a6xx_gmu_resume+0xc8/0xecc [ 16.204120] a6xx_pm_resume+0x148/0x200 [ 16.208064] adreno_resume+0x28/0x34 [ 16.211743] pm_generic_runtime_resume+0x34/0x48 [ 16.216488] __rpm_callback+0x70/0x10c [ 16.220342] rpm_callback+0x34/0x8c [ 16.223933] rpm_resume+0x414/0x550 [ 16.227524] __pm_runtime_resume+0x7c/0xa0 [ 16.231731] msm_gpu_submit+0x60/0x1c0 [ 16.235586] msm_ioctl_gem_submit+0xadc/0xb60 [ 16.240066] drm_ioctl_kernel+0x9c/0x118 [ 16.244097] drm_ioctl+0x27c/0x408 [ 16.247602] drm_compat_ioctl+0xcc/0xdc [ 16.251546] __se_compat_sys_ioctl+0x100/0x206c [ 16.256204] __arm64_compat_sys_ioctl+0x20/0x2c [ 16.260861] el0_svc_common+0xa8/0x178 [ 16.264716] el0_svc_compat_handler+0x2c/0x40 [ 16.269196] el0_svc_compat+0x8/0x10BR,
-RQuoted 50 lines
-Daniel
Quoted 48 lines
opp_table_lock-->&mm->mmap_sem-->reservation_ww_class_mutexFor an explicit fencing userspace, the impact should be minimal as we do all the fence waits before this point. It could result in some needless resumes in error cases, etc.
Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/msm_gem_submit.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-)drivers/gpu/drm/msm/msm_gem_submit.c +13
-2@@ -744,11 +744,20 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data, ret =submit_lookup_objects (submit, args, file);if (ret) - goto out; + goto out_pre_pm; ret =submit_lookup_cmds (submit, args, file);if (ret) - goto out; + goto out_pre_pm; + +/* +* Thanks to dev_pm_opp opp_table_lock interactions with mm->mmap_sem +* in the resume path, we need to to rpm get before we lock objs. +* Which unfortunately might involve powering up the GPU sooner than +* is necessary. But at least in the explicit fencing case, we will +* have already done all the fence waiting. +*/ +pm_runtime_get_sync (&gpu->pdev->dev);/* copy_*_user while holding a ww ticket upsets lockdep */ ww_acquire_init (&submit->ticket, &reservation_ww_class);@@ -825,6 +834,8 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data, out: +pm_runtime_put (&gpu->pdev->dev); +out_pre_pm:submit_cleanup (submit);if (has_ww_ticket)ww_acquire_fini (&submit->ticket);
Signature · 1 line
viresh
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 25 lines · Viresh Kumar
Quoted 23 lines · Rob Clark
Quoted 8 lines · Daniel Vetter
Quoted 5 lines · Rob Clark
From: Rob Clark <[email protected]>
Unfortunately, due to an
dev_pm_opplocking interaction withmm->mmap_sem, we need to do pm get before aquiring obj locks, otherwise we can have anger lockdep with the chain:Quoted 8 lines
tbh this sounds like a bug in that subsystem, since it means we cannot use said subsystem in mmap handlers either.
So if you have some remapping unit or need to wake up your gpu to blt the buffer into system memory first, we're toast. That doesn't sound right. So maybe Cc: pm folks and figure out how to fix this long term properly? Imo not a good reason to hold up this patch set, since unwrangling
mmap_semtends to be work ...+ a couple of PM folks
Looks like it has been this way for quite some time, so I guess the overlap between things using
dev_pm_oppand mmap is low..
Quoted 8 lines
Quoted 3 lines
fwiw, example splat so folks can see the locking interaction I am talking about.. I suspect the
pm_oppinteraction withmm->mmap_semis from the debugfs calls whileopp_table_lockis held?I am not very sure about why this circular locking dependency is happening here and how exactly can we fix it. The OPP core works under the
opp_table_lock, from within which it creates/remove the debugfs stuff as well.
Yeah that's bad practice. Generally you shouldn't need to hold locks in setup/teardown code, since there's no other thread which can possible hold a reference to anything your touching anymore. Ofc excluding quickly grabbing/dropping a lock to insert/remove objects into lists and stuff.
The other reason is that especially with anything related to sysfs or debugfs, the locking dependencies you're pulling in are enormous: vfs locks pull in mm locks (due to mmap) and at that point there's pretty much nothing left you're allowed to hold while acquiring such a lock. For simple drivers this is no issue, but for fancy drivers (like gpu drivers) which need to interact with core mm) this means your subsystem is a major pain to use.
Usually the correct fix is to only hold your subsystem locks in setup/teardown when absolutely required, and fix any data inconsistency issues by reordering your setup/teardown code: When you register as the last step and unregister as the first step, there's no need for any additional locking. And hence no need to call debugfs functions while holding your subsystem locks.
The catch phrase I use for this is "don't solve object lifetime issues with locking". Instead use refcounting and careful ordering in setup/teardown code.
I think Rob has found some duct-tape, so this isn't an immediate issue, but would be really nice to get fixed.
Cheers, Daniel
Quoted 210 lines
Quoted 201 lines
[ 15.627855] ====================================================== [ 15.634202] WARNING: possible circular locking dependency detected [ 15.640550] 5.4.70 #41 Not tainted [ 15.644050] ------------------------------------------------------ [ 15.650397] chrome/1805 is trying to acquire lock: [ 15.655314] ffffffed90720738 (opp_table_lock){+.+.}, at: _find_opp_table+0x34/0x74 [ 15.663092] [ 15.663092] but task is already holding lock: [ 15.669082] ffffff80ff3911a8 (reservation_ww_class_mutex){+.+.}, at: submit_lock_objects+0x70/0x1ec [ 15.678369] [ 15.678369] which lock already depends on the new lock. [ 15.678369] [ 15.686764] [ 15.686764] the existing dependency chain (in reverse order) is: [ 15.694438] [ 15.694438] -> #3 (reservation_ww_class_mutex){+.+.}: [ 15.701146] __mutex_lock_common+0xec/0xc0c [ 15.705978] ww_mutex_lock_interruptible+0x5c/0xc4 [ 15.711432] msm_gem_fault+0x2c/0x124 [ 15.715731] __do_fault+0x40/0x16c [ 15.719766] handle_mm_fault+0x7cc/0xd98 [ 15.724337] do_page_fault+0x230/0x3b4 [ 15.728721] do_translation_fault+0x5c/0x78 [ 15.733558] do_mem_abort+0x4c/0xb4 [ 15.737680] el0_da+0x1c/0x20 [ 15.741266] [ 15.741266] -> #2 (&mm->mmap_sem){++++}: [ 15.746809] __might_fault+0x70/0x98 [ 15.751022] compat_filldir+0xf8/0x48c [ 15.755412] dcache_readdir+0x70/0x1dc [ 15.759808] iterate_dir+0xd4/0x180 [ 15.763931] __arm64_compat_sys_getdents+0xa0/0x19c [ 15.769476] el0_svc_common+0xa8/0x178 [ 15.773861] el0_svc_compat_handler+0x2c/0x40 [ 15.778868] el0_svc_compat+0x8/0x10 [ 15.783075] [ 15.783075] -> #1 (&sb->s_type->i_mutex_key#3){++++}: [ 15.789788] down_write+0x54/0x16c [ 15.793826] debugfs_remove_recursive+0x50/0x158 [ 15.799108] opp_debug_unregister+0x34/0x114 [ 15.804028] dev_pm_opp_put_opp_table+0xd0/0x14c [ 15.809308] dev_pm_opp_put_clkname+0x3c/0x50 [ 15.814318] msm_dsi_host_destroy+0xb0/0xcc [ 15.819149] dsi_destroy+0x40/0x58 [ 15.823184] dsi_bind+0x90/0x170 [ 15.827041] component_bind_all+0xf0/0x208 [ 15.831787] msm_drm_init+0x188/0x60c [ 15.836084] msm_drm_bind+0x24/0x30 [ 15.840205] try_to_bring_up_master+0x15c/0x1a4 [ 15.845396] __component_add+0x98/0x14c [ 15.849878] component_add+0x28/0x34 [ 15.854086] dp_display_probe+0x324/0x370 [ 15.858744] platform_drv_probe+0x90/0xb0 [ 15.863400] really_probe+0x134/0x2ec [ 15.867699] driver_probe_device+0x64/0xfc [ 15.872443] __device_attach_driver+0x8c/0xa4 [ 15.877459] bus_for_each_drv+0x90/0xd8 [ 15.881939] __device_attach+0xc0/0x148 [ 15.886420] device_initial_probe+0x20/0x2c [ 15.891254] bus_probe_device+0x34/0x94 [ 15.895726] deferred_probe_work_func+0x78/0xb4 [ 15.900914] process_one_work+0x30c/0x5d0 [ 15.905573] worker_thread+0x240/0x3f0 [ 15.909959] kthread+0x144/0x154 [ 15.913809] ret_from_fork+0x10/0x18 [ 15.918016] [ 15.918016] -> #0 (opp_table_lock){+.+.}: [ 15.923660] __lock_acquire+0xee4/0x2450 [ 15.928230] lock_acquire+0x1cc/0x210 [ 15.932527] __mutex_lock_common+0xec/0xc0c [ 15.937359] mutex_lock_nested+0x40/0x50 [ 15.941928] _find_opp_table+0x34/0x74 [ 15.946312] dev_pm_opp_find_freq_exact+0x2c/0xdc [ 15.951680] a6xx_gmu_resume+0xc8/0xecc [ 15.952812] fscrypt: AES-256-CTS-CBC using implementation "cts-cbc-aes-ce" [ 15.956161] a6xx_pm_resume+0x148/0x200 [ 15.956166] adreno_resume+0x28/0x34 [ 15.956171] pm_generic_runtime_resume+0x34/0x48 [ 15.956174] __rpm_callback+0x70/0x10c [ 15.956176] rpm_callback+0x34/0x8c [ 15.956179] rpm_resume+0x414/0x550 [ 15.956182] __pm_runtime_resume+0x7c/0xa0 [ 15.956185] msm_gpu_submit+0x60/0x1c0 [ 15.956190] msm_ioctl_gem_submit+0xadc/0xb60 [ 16.003961] drm_ioctl_kernel+0x9c/0x118 [ 16.008532] drm_ioctl+0x27c/0x408 [ 16.012562] drm_compat_ioctl+0xcc/0xdc [ 16.017038] __se_compat_sys_ioctl+0x100/0x206c [ 16.022224] __arm64_compat_sys_ioctl+0x20/0x2c [ 16.027412] el0_svc_common+0xa8/0x178 [ 16.031800] el0_svc_compat_handler+0x2c/0x40 [ 16.036810] el0_svc_compat+0x8/0x10 [ 16.041021] [ 16.041021] other info that might help us debug this: [ 16.041021] [ 16.049235] Chain exists of: [ 16.049235] opp_table_lock --> &mm->mmap_sem --> reservation_ww_class_mutex [ 16.049235] [ 16.061014] Possible unsafe locking scenario: [ 16.061014] [ 16.067091] CPU0 CPU1 [ 16.071750] ---- ---- [ 16.076399] lock(reservation_ww_class_mutex); [ 16.081059] lock(&mm->mmap_sem); [ 16.087134] lock(reservation_ww_class_mutex); [ 16.094369] lock(opp_table_lock); [ 16.097961] [ 16.097961] *** DEADLOCK *** [ 16.097961] [ 16.104038] 3 locks held by chrome/1805: [ 16.108068] #0: ffffff80fb20c0d8 (&dev->struct_mutex){+.+.}, at: msm_ioctl_gem_submit+0x264/0xb60 [ 16.117264] #1: ffffff80dd712c70 (reservation_ww_class_acquire){+.+.}, at: msm_ioctl_gem_submit+0x8e8/0xb60 [ 16.127357] #2: ffffff80ff3911a8 (reservation_ww_class_mutex){+.+.}, at: submit_lock_objects+0x70/0x1ec [ 16.137089] [ 16.137089] stack backtrace: [ 16.141567] CPU: 4 PID: 1805 Comm: chrome Not tainted 5.4.70 #41 [ 16.147733] Hardware name: Google Lazor (rev1+) with LTE (DT) [ 16.153632] Call trace: [ 16.156154] dump_backtrace+0x0/0x158 [ 16.159924] show_stack+0x20/0x2c [ 16.163340] dump_stack+0xc8/0x160 [ 16.166840] print_circular_bug+0x2c4/0x2c8 [ 16.171144] check_noncircular+0x1a8/0x1b0 [ 16.175351] __lock_acquire+0xee4/0x2450 [ 16.179382] lock_acquire+0x1cc/0x210 [ 16.183146] __mutex_lock_common+0xec/0xc0c [ 16.187450] mutex_lock_nested+0x40/0x50 [ 16.191481] _find_opp_table+0x34/0x74 [ 16.195344] dev_pm_opp_find_freq_exact+0x2c/0xdc [ 16.200178] a6xx_gmu_resume+0xc8/0xecc [ 16.204120] a6xx_pm_resume+0x148/0x200 [ 16.208064] adreno_resume+0x28/0x34 [ 16.211743] pm_generic_runtime_resume+0x34/0x48 [ 16.216488] __rpm_callback+0x70/0x10c [ 16.220342] rpm_callback+0x34/0x8c [ 16.223933] rpm_resume+0x414/0x550 [ 16.227524] __pm_runtime_resume+0x7c/0xa0 [ 16.231731] msm_gpu_submit+0x60/0x1c0 [ 16.235586] msm_ioctl_gem_submit+0xadc/0xb60 [ 16.240066] drm_ioctl_kernel+0x9c/0x118 [ 16.244097] drm_ioctl+0x27c/0x408 [ 16.247602] drm_compat_ioctl+0xcc/0xdc [ 16.251546] __se_compat_sys_ioctl+0x100/0x206c [ 16.256204] __arm64_compat_sys_ioctl+0x20/0x2c [ 16.260861] el0_svc_common+0xa8/0x178 [ 16.264716] el0_svc_compat_handler+0x2c/0x40 [ 16.269196] el0_svc_compat+0x8/0x10BR,
-RQuoted 44 lines
-Daniel
Quoted 42 lines
opp_table_lock-->&mm->mmap_sem-->reservation_ww_class_mutexFor an explicit fencing userspace, the impact should be minimal as we do all the fence waits before this point. It could result in some needless resumes in error cases, etc.
Signed-off-by Rob Clark <[email protected]>--- drivers/gpu/drm/msm/msm_gem_submit.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-)drivers/gpu/drm/msm/msm_gem_submit.c +11
-2@@ -744,11 +744,20 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data, ret =submit_lookup_objects (submit, args, file);if (ret) - goto out; + goto out_pre_pm; ret =submit_lookup_cmds (submit, args, file);if (ret) - goto out; + goto out_pre_pm; + +/* +* Thanks to dev_pm_opp opp_table_lock interactions with mm->mmap_sem +* in the resume path, we need to to rpm get before we lock objs. +* Which unfortunately might involve powering up the GPU sooner than +* is necessary. But at least in the explicit fencing case, we will +* have already done all the fence waiting. +*/ +pm_runtime_get_sync (&gpu->pdev->dev);/* copy_*_user while holding a ww ticket upsets lockdep */ ww_acquire_init (&submit->ticket, &reservation_ww_class);@@ -825,6 +834,8 @@ int msm_ioctl_gem_submit(struct drm_device *dev, void *data,Quoted 6 lines
Quoted 6 lines
drivers/gpu/drm/msm/msm_gem_submit.c int msm_ioctl_gem_submit(struct drm_device *dev, void *data, · line 841
Quoted 6 lines
drivers/gpu/drm/msm/msm_gem_submit.c +2
-0out: +pm_runtime_put (&gpu->pdev->dev); +out_pre_pm:submit_cleanup (submit);if (has_ww_ticket)ww_acquire_fini (&submit->ticket);--
viresh
Signature · 3 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 21 lines · Daniel Vetter
Yeah that's bad practice. Generally you shouldn't need to hold locks in setup/teardown code, since there's no other thread which can possible hold a reference to anything your touching anymore. Ofc excluding quickly grabbing/dropping a lock to insert/remove objects into lists and stuff.
The other reason is that especially with anything related to sysfs or debugfs, the locking dependencies you're pulling in are enormous: vfs locks pull in mm locks (due to mmap) and at that point there's pretty much nothing left you're allowed to hold while acquiring such a lock. For simple drivers this is no issue, but for fancy drivers (like gpu drivers) which need to interact with core mm) this means your subsystem is a major pain to use.
Usually the correct fix is to only hold your subsystem locks in setup/teardown when absolutely required, and fix any data inconsistency issues by reordering your setup/teardown code: When you register as the last step and unregister as the first step, there's no need for any additional locking. And hence no need to call debugfs functions while holding your subsystem locks.
Quoted 3 lines
The catch phrase I use for this is "don't solve object lifetime issues with locking". Instead use refcounting and careful ordering in setup/teardown code.
This is exactly what I have done in the OPP core, the locks were taken only when really necessary, though as we have seen now I have missed that at a single place and that should be fixed as well. Will do that, thanks.
Signature · 1 line
viresh
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 23 lines · Viresh Kumar
Quoted 21 lines · Daniel Vetter
Yeah that's bad practice. Generally you shouldn't need to hold locks in setup/teardown code, since there's no other thread which can possible hold a reference to anything your touching anymore. Ofc excluding quickly grabbing/dropping a lock to insert/remove objects into lists and stuff.
The other reason is that especially with anything related to sysfs or debugfs, the locking dependencies you're pulling in are enormous: vfs locks pull in mm locks (due to mmap) and at that point there's pretty much nothing left you're allowed to hold while acquiring such a lock. For simple drivers this is no issue, but for fancy drivers (like gpu drivers) which need to interact with core mm) this means your subsystem is a major pain to use.
Usually the correct fix is to only hold your subsystem locks in setup/teardown when absolutely required, and fix any data inconsistency issues by reordering your setup/teardown code: When you register as the last step and unregister as the first step, there's no need for any additional locking. And hence no need to call debugfs functions while holding your subsystem locks.
Quoted 8 lines
Quoted 3 lines
The catch phrase I use for this is "don't solve object lifetime issues with locking". Instead use refcounting and careful ordering in setup/teardown code.
This is exactly what I have done in the OPP core, the locks were taken only when really necessary, though as we have seen now I have missed that at a single place and that should be fixed as well. Will do that, thanks.
Excellent. If the fix is small enough can you push it into 5.10? That way drm/msm doesn't have to carry the temporary solution for 5.11 (the issue only pops up with the locking rework, which teaches lockdep a few more things about what's going on as a side effect). -Daniel
Signature · 7 lines
Daniel Vetter Software Engineer, Intel Corporation http://blog.ffwll.ch _______________________________________________ dri-devel mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
In reply to Viresh KumarQuoted 23 lines · Viresh Kumar
Quoted 21 lines · Daniel Vetter
Yeah that's bad practice. Generally you shouldn't need to hold locks in setup/teardown code, since there's no other thread which can possible hold a reference to anything your touching anymore. Ofc excluding quickly grabbing/dropping a lock to insert/remove objects into lists and stuff.
The other reason is that especially with anything related to sysfs or debugfs, the locking dependencies you're pulling in are enormous: vfs locks pull in mm locks (due to mmap) and at that point there's pretty much nothing left you're allowed to hold while acquiring such a lock. For simple drivers this is no issue, but for fancy drivers (like gpu drivers) which need to interact with core mm) this means your subsystem is a major pain to use.
Usually the correct fix is to only hold your subsystem locks in setup/teardown when absolutely required, and fix any data inconsistency issues by reordering your setup/teardown code: When you register as the last step and unregister as the first step, there's no need for any additional locking. And hence no need to call debugfs functions while holding your subsystem locks.
Quoted 8 lines
Quoted 3 lines
The catch phrase I use for this is "don't solve object lifetime issues with locking". Instead use refcounting and careful ordering in setup/teardown code.
This is exactly what I have done in the OPP core, the locks were taken only when really necessary, though as we have seen now I have missed that at a single place and that should be fixed as well. Will do that, thanks.
I do have an easy enough way to repro the issue, so if you have a patch I can certainly test it.
BR,
-R
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 28 lines · Rob Clark
Quoted 27 lines · Viresh Kumar
Quoted 24 lines · Daniel Vetter
Yeah that's bad practice. Generally you shouldn't need to hold locks in setup/teardown code, since there's no other thread which can possible hold a reference to anything your touching anymore. Ofc excluding quickly grabbing/dropping a lock to insert/remove objects into lists and stuff.
The other reason is that especially with anything related to sysfs or debugfs, the locking dependencies you're pulling in are enormous: vfs locks pull in mm locks (due to mmap) and at that point there's pretty much nothing left you're allowed to hold while acquiring such a lock. For simple drivers this is no issue, but for fancy drivers (like gpu drivers) which need to interact with core mm) this means your subsystem is a major pain to use.
Usually the correct fix is to only hold your subsystem locks in setup/teardown when absolutely required, and fix any data inconsistency issues by reordering your setup/teardown code: When you register as the last step and unregister as the first step, there's no need for any additional locking. And hence no need to call debugfs functions while holding your subsystem locks.
The catch phrase I use for this is "don't solve object lifetime issues with locking". Instead use refcounting and careful ordering in setup/teardown code.
Quoted 7 lines
Quoted 4 lines
This is exactly what I have done in the OPP core, the locks were taken only when really necessary, though as we have seen now I have missed that at a single place and that should be fixed as well. Will do that, thanks.
I do have an easy enough way to repro the issue, so if you have a patch I can certainly test it.
Does this fix it for you ? There is one more place still left where we are taking the opp_table_lock while adding stuff to debugfs and that's not that straight forward to fix. But I didn't see that path in your circular dependency trace, so who knows :)
drivers/opp/core.c +4 -3
@@ -1181,6 +1181,10 @@ static void _opp_table_kref_release(struct kref *kref)struct opp_device *opp_dev, *temp;int i; +/* Drop the lock as soon as we can */ +list_del (&opp_table->node); +mutex_unlock (&opp_table_lock); +_of_clear_opp_table (opp_table);/* Release clk */
@@ -1208,10 +1212,7 @@ static void _opp_table_kref_release(struct kref *kref)mutex_destroy (&opp_table->genpd_virt_dev_lock);mutex_destroy (&opp_table->lock); -list_del (&opp_table->node);kfree (opp_table); - -mutex_unlock (&opp_table_lock); }void dev_pm_opp_put_opp_table (struct opp_table *opp_table)
Signature · 1 line
viresh
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 35 lines · Viresh Kumar
Quoted 33 lines · Rob Clark
Quoted 31 lines · Viresh Kumar
Quoted 21 lines · Daniel Vetter
Yeah that's bad practice. Generally you shouldn't need to hold locks in setup/teardown code, since there's no other thread which can possible hold a reference to anything your touching anymore. Ofc excluding quickly grabbing/dropping a lock to insert/remove objects into lists and stuff.
The other reason is that especially with anything related to sysfs or debugfs, the locking dependencies you're pulling in are enormous: vfs locks pull in mm locks (due to mmap) and at that point there's pretty much nothing left you're allowed to hold while acquiring such a lock. For simple drivers this is no issue, but for fancy drivers (like gpu drivers) which need to interact with core mm) this means your subsystem is a major pain to use.
Usually the correct fix is to only hold your subsystem locks in setup/teardown when absolutely required, and fix any data inconsistency issues by reordering your setup/teardown code: When you register as the last step and unregister as the first step, there's no need for any additional locking. And hence no need to call debugfs functions while holding your subsystem locks.
Quoted 3 lines
The catch phrase I use for this is "don't solve object lifetime issues with locking". Instead use refcounting and careful ordering in setup/teardown code.
This is exactly what I have done in the OPP core, the locks were taken only when really necessary, though as we have seen now I have missed that at a single place and that should be fixed as well. Will do that, thanks.
Quoted 7 lines
Quoted 2 lines
I do have an easy enough way to repro the issue, so if you have a patch I can certainly test it.
Does this fix it for you ? There is one more place still left where we are taking the
opp_table_lockwhile adding stuff to debugfs and that's not that straight forward to fix. But I didn't see that path in your circular dependency trace, so who knows :)
Nope, I suspect any creation of debugfs files will be problematic.
(btw, _add_opp_dev_unlocked() looks like it should be called _add_opp_dev_locked()?)
It does look like 'struct opp_table' is already refcnt'd, so I suspect you could replace holding opp_table_lock while calling into debugfs with holding a reference to the opp_table instead?
BR,
-R
[ +0.074543] ======================================================
[ +0.006347] WARNING: possible circular locking dependency detected
[ +0.006349] 5.4.72 #14 Not tainted
[ +0.003501] ------------------------------------------------------
[ +0.006350] chrome/1865 is trying to acquire lock:
[ +0.004922] ffffffdd34921750 (opp_table_lock){+.+.}, at:_find_opp_table+0x34/0x74
[ +0.007779]
but task is already holding lock:
[ +0.005989] ffffff81f0fc71a8 (reservation_ww_class_mutex){+.+.}, at:submit_lock_objects+0x70/0x1ec
[ +0.001132] fscrypt: AES-256-CTS-CBC using implementation "cts-cbc-aes-ce" [ +0.008156]
which lock already depends on the new lock.
[ +0.000002]
the existing dependency chain (in reverse order) is:
[ +0.000002]
-> #4 (reservation_ww_class_mutex){+.+.}:
[ +0.000009] __mutex_lock_common+0xec/0xc0c
[ +0.000004] ww_mutex_lock_interruptible+0x5c/0xc4
[ +0.000003] msm_gem_fault+0x2c/0x124
[ +0.000005] __do_fault+0x40/0x16c
[ +0.000003] handle_mm_fault+0x7cc/0xd98
[ +0.000005] do_page_fault+0x230/0x3b4
[ +0.000003] do_translation_fault+0x5c/0x78
[ +0.000004] do_mem_abort+0x4c/0xb4
[ +0.000006] el0_da+0x1c/0x20
[ +0.069917]
-> #3 (&mm->mmap_sem){++++}:
[ +0.005548] __might_fault+0x70/0x98
[ +0.004209] compat_filldir+0xf8/0x48c
[ +0.004394] dcache_readdir+0x70/0x1dc
[ +0.004394] iterate_dir+0xd4/0x180
[ +0.004119] __arm64_compat_sys_getdents+0xa0/0x19c
[ +0.005549] el0_svc_common+0xa8/0x178
[ +0.004394] el0_svc_compat_handler+0x2c/0x40
[ +0.005007] el0_svc_compat+0x8/0x10
[ +0.004205]
-> #2 (&sb->s_type->i_mutex_key#3){++++}:
[ +0.006708] down_write+0x54/0x16c
[ +0.004034] start_creating+0x68/0x128
[ +0.004392] debugfs_create_dir+0x28/0x114
[ +0.004747] opp_debug_register+0x8c/0xc0
[ +0.004657] _add_opp_dev_unlocked+0x5c/0x70
[ +0.004920] _add_opp_dev+0x38/0x58
[ +0.004118] _opp_get_opp_table+0xdc/0x1ac
[ +0.004745] dev_pm_opp_get_opp_table_indexed+0x24/0x30
[ +0.005899] dev_pm_opp_of_add_table_indexed+0x48/0x84
[ +0.005813] of_genpd_add_provider_onecell+0xc0/0x1b8
[ +0.005724] rpmhpd_probe+0x240/0x268
[ +0.004307] platform_drv_probe+0x90/0xb0
[ +0.004654] really_probe+0x134/0x2ec
[ +0.004304] driver_probe_device+0x64/0xfc
[ +0.004746] __device_attach_driver+0x8c/0xa4
[ +0.005008] bus_for_each_drv+0x90/0xd8
[ +0.004481] __device_attach+0xc0/0x148
[ +0.004480] device_initial_probe+0x20/0x2c
[ +0.004832] bus_probe_device+0x34/0x94
[ +0.004482] device_add+0x1fc/0x3b0
[ +0.004121] of_device_add+0x3c/0x4c
[ +0.004206] of_platform_device_create_pdata+0xb8/0xfc
[ +0.005811] of_platform_bus_create+0x1e4/0x368
[ +0.005185] of_platform_populate+0x70/0xbc
[ +0.004833] devm_of_platform_populate+0x58/0xa0
[ +0.005283] rpmh_rsc_probe+0x36c/0x3cc
[ +0.004481] platform_drv_probe+0x90/0xb0
[ +0.004657] really_probe+0x134/0x2ec
[ +0.004305] driver_probe_device+0x64/0xfc
[ +0.004745] __device_attach_driver+0x8c/0xa4
[ +0.005007] bus_for_each_drv+0x90/0xd8
[ +0.004480] __device_attach+0xc0/0x148
[ +0.004481] device_initial_probe+0x20/0x2c
[ +0.004833] bus_probe_device+0x34/0x94
[ +0.004481] device_add+0x1fc/0x3b0
[ +0.004119] of_device_add+0x3c/0x4c
[ +0.004206] of_platform_device_create_pdata+0xb8/0xfc
[ +0.005811] of_platform_bus_create+0x1e4/0x368
[ +0.005185] of_platform_bus_create+0x230/0x368
[ +0.005185] of_platform_populate+0x70/0xbc
[ +0.004836] of_platform_default_populate_init+0xa8/0xc0
[ +0.005986] do_one_initcall+0x1c8/0x3fc
[ +0.004572] do_initcall_level+0xb4/0x10c
[ +0.004657] do_basic_setup+0x30/0x48
[ +0.004304] kernel_init_freeable+0x124/0x1a4
[ +0.005009] kernel_init+0x14/0x104
[ +0.004119] ret_from_fork+0x10/0x18
[ +0.004205]
-> #1 (&opp_table->lock){+.+.}:
[ +0.005815] __mutex_lock_common+0xec/0xc0c
[ +0.004832] mutex_lock_nested+0x40/0x50
[ +0.004570] _add_opp_dev+0x2c/0x58
[ +0.004119] _opp_get_opp_table+0xdc/0x1ac
[ +0.004745] dev_pm_opp_get_opp_table_indexed+0x24/0x30
[ +0.005899] dev_pm_opp_of_add_table_indexed+0x48/0x84
[ +0.005814] of_genpd_add_provider_onecell+0xc0/0x1b8
[ +0.005721] rpmhpd_probe+0x240/0x268
[ +0.004306] platform_drv_probe+0x90/0xb0
[ +0.004656] really_probe+0x134/0x2ec
[ +0.004305] driver_probe_device+0x64/0xfc
[ +0.004745] __device_attach_driver+0x8c/0xa4
[ +0.005007] bus_for_each_drv+0x90/0xd8
[ +0.004481] __device_attach+0xc0/0x148
[ +0.004481] device_initial_probe+0x20/0x2c
[ +0.004832] bus_probe_device+0x34/0x94
[ +0.004481] device_add+0x1fc/0x3b0
[ +0.004119] of_device_add+0x3c/0x4c
[ +0.004206] of_platform_device_create_pdata+0xb8/0xfc
[ +0.005810] of_platform_bus_create+0x1e4/0x368
[ +0.005197] of_platform_populate+0x70/0xbc
[ +0.004832] devm_of_platform_populate+0x58/0xa0
[ +0.005284] rpmh_rsc_probe+0x36c/0x3cc
[ +0.004480] platform_drv_probe+0x90/0xb0
[ +0.004658] really_probe+0x134/0x2ec
[ +0.004301] driver_probe_device+0x64/0xfc
[ +0.004745] __device_attach_driver+0x8c/0xa4
[ +0.005007] bus_for_each_drv+0x90/0xd8
[ +0.004480] __device_attach+0xc0/0x148
[ +0.004481] device_initial_probe+0x20/0x2c
[ +0.004831] bus_probe_device+0x34/0x94
[ +0.004482] device_add+0x1fc/0x3b0
[ +0.004118] of_device_add+0x3c/0x4c
[ +0.004214] of_platform_device_create_pdata+0xb8/0xfc
[ +0.005817] of_platform_bus_create+0x1e4/0x368
[ +0.005186] of_platform_bus_create+0x230/0x368
[ +0.005188] of_platform_populate+0x70/0xbc
[ +0.004832] of_platform_default_populate_init+0xa8/0xc0
[ +0.005987] do_one_initcall+0x1c8/0x3fc
[ +0.004569] do_initcall_level+0xb4/0x10c
[ +0.004657] do_basic_setup+0x30/0x48
[ +0.004305] kernel_init_freeable+0x124/0x1a4
[ +0.005008] kernel_init+0x14/0x104
[ +0.004119] ret_from_fork+0x10/0x18
[ +0.004206]
-> #0 (opp_table_lock){+.+.}:
[ +0.005640] __lock_acquire+0xee4/0x2450
[ +0.004570] lock_acquire+0x1cc/0x210
[ +0.004305] __mutex_lock_common+0xec/0xc0c
[ +0.004833] mutex_lock_nested+0x40/0x50
[ +0.004570] _find_opp_table+0x34/0x74
[ +0.004393] dev_pm_opp_find_freq_exact+0x2c/0xdc
[ +0.005372] a6xx_gmu_resume+0xc8/0xecc
[ +0.004480] a6xx_pm_resume+0x148/0x200
[ +0.004482] adreno_resume+0x28/0x34
[ +0.004209] pm_generic_runtime_resume+0x34/0x48
[ +0.005283] __rpm_callback+0x70/0x10c
[ +0.004393] rpm_callback+0x34/0x8c
[ +0.004119] rpm_resume+0x414/0x550
[ +0.004119] __pm_runtime_resume+0x7c/0xa0
[ +0.004746] msm_gpu_submit+0x60/0x1c0
[ +0.004394] msm_ioctl_gem_submit+0xadc/0xb60
[ +0.005010] drm_ioctl_kernel+0x9c/0x118
[ +0.004569] drm_ioctl+0x27c/0x408
[ +0.004034] drm_compat_ioctl+0xcc/0xdc
[ +0.004483] __se_compat_sys_ioctl+0x100/0x206c
[ +0.005186] __arm64_compat_sys_ioctl+0x20/0x2c
[ +0.005187] el0_svc_common+0xa8/0x178
[ +0.004393] el0_svc_compat_handler+0x2c/0x40
[ +0.005009] el0_svc_compat+0x8/0x10
[ +0.004205]other info that might help us debug this:
[ +0.008213] Chain exists of:
opp_table_lock --> &mm->mmap_sem --> reservation_ww_class_mutex
[ +0.011780] Possible unsafe locking scenario:
[ +0.006082] CPU0 CPU1
[ +0.004660] ---- ----
[ +0.004656] lock(reservation_ww_class_mutex);
[ +0.004657] lock(&mm->mmap_sem);
[ +0.006079] lock(reservation_ww_class_mutex);
[ +0.007237] lock(opp_table_lock);
[ +0.003592]
*** DEADLOCK ***[ +0.006084] 3 locks held by chrome/1865:
[ +0.004031] #0: ffffff81edecc0d8 (&dev->struct_mutex){+.+.}, at:msm_ioctl_gem_submit+0x264/0xb60
[ +0.009198] #1: ffffff81d0000870 (reservation_ww_class_acquire){+.+.}, at:msm_ioctl_gem_submit+0x8e8/0xb60
[ +0.010086] #2: ffffff81f0fc71a8 (reservation_ww_class_mutex){+.+.}, at: submit_lock_objects+0x70/0x1ec [ +0.009735]
stack backtrace: [ +0.004482] CPU: 0 PID: 1865 Comm: chrome Not tainted 5.4.72 #14 [ +0.006173] Hardware name: Google Lazor (rev1+) with LTE (DT) [ +0.005899] Call trace: [ +0.002515] dump_backtrace+0x0/0x158 [ +0.003768] show_stack+0x20/0x2c [ +0.003407] dump_stack+0xc8/0x160 [ +0.003506] print_circular_bug+0x2c4/0x2c8 [ +0.004305] check_noncircular+0x1a8/0x1b0 [ +0.004206] __lock_acquire+0xee4/0x2450 [ +0.004032] lock_acquire+0x1cc/0x210 [ +0.003768] __mutex_lock_common+0xec/0xc0c [ +0.004305] mutex_lock_nested+0x40/0x50 [ +0.004033] _find_opp_table+0x34/0x74 [ +0.003855] dev_pm_opp_find_freq_exact+0x2c/0xdc [ +0.004833] a6xx_gmu_resume+0xc8/0xecc [ +0.003943] a6xx_pm_resume+0x148/0x200 [ +0.003944] adreno_resume+0x28/0x34 [ +0.003681] pm_generic_runtime_resume+0x34/0x48 [ +0.004745] __rpm_callback+0x70/0x10c [ +0.003854] rpm_callback+0x34/0x8c [ +0.003592] rpm_resume+0x414/0x550 [ +0.003592] __pm_runtime_resume+0x7c/0xa0 [ +0.004207] msm_gpu_submit+0x60/0x1c0 [ +0.003855] msm_ioctl_gem_submit+0xadc/0xb60 [ +0.004481] drm_ioctl_kernel+0x9c/0x118 [ +0.004031] drm_ioctl+0x27c/0x408 [ +0.003504] drm_compat_ioctl+0xcc/0xdc [ +0.003945] __se_compat_sys_ioctl+0x100/0x206c [ +0.004658] __arm64_compat_sys_ioctl+0x20/0x2c [ +0.004659] el0_svc_common+0xa8/0x178 [ +0.003855] el0_svc_compat_handler+0x2c/0x40 [ +0.004480] el0_svc_compat+0x8/0x10
Quoted 29 lines
drivers/opp/core.c +4
-4@@ -1181,6 +1181,10 @@ static void _opp_table_kref_release(struct kref *kref)struct opp_device *opp_dev, *temp;int i; +/* Drop the lock as soon as we can */ +list_del (&opp_table->node); +mutex_unlock (&opp_table_lock); +_of_clear_opp_table (opp_table);/* Release clk */ @@ -1208,10 +1212,7 @@ static void _opp_table_kref_release(struct kref *kref)mutex_destroy (&opp_table->genpd_virt_dev_lock);mutex_destroy (&opp_table->lock); -list_del (&opp_table->node);kfree (opp_table); - -mutex_unlock (&opp_table_lock); }void dev_pm_opp_put_opp_table (struct opp_table *opp_table) --viresh
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 1 line · Rob Clark
Nope, I suspect any creation of debugfs files will be problematic.
Yeah, so it only fixed part of the problem.
Quoted 6 lines
(btw,
_add_opp_dev_unlocked()looks like it should be called_add_opp_dev_locked()?)It does look like 'struct
opp_table' is already refcnt'd, so I suspect you could replace holdingopp_table_lockwhile calling into debugfs with holding a reference to theopp_tableinstead?
It isn't that straight forward unfortunately, we need to make sure the table doesn't get allocated for the same device twice, so find+allocate needs to happen within a locked region.
I have taken, not so straight forward, approach to fixing this issue, lets see if this fixes it or not.
-------------------------8<-------------------------
drivers/opp/core.c +51 -23
@@ -29,6 +29,8 @@LIST_HEAD (opp_tables);/* Lock to allow exclusive modification to the device and opp lists */ DEFINE_MUTEX (opp_table_lock); +/* Flag indicating that opp_tables list is being updated at the moment */ +static bool opp_tables_busy;static struct opp_device *_find_opp_dev (const struct device *dev,struct opp_table *opp_table)
@@ -1036,8 +1038,8 @@ static void _remove_opp_dev(struct opp_device *opp_dev,kfree (opp_dev); } -static struct opp_device *_add_opp_dev_unlocked (const struct device *dev, -struct opp_table *opp_table) +struct opp_device *_add_opp_dev (const struct device *dev, +struct opp_table *opp_table) {struct opp_device *opp_dev;
@@ -1048,7 +1050,9 @@ static struct opp_device *_add_opp_dev_unlocked(const struct device *dev,/* Initialize opp-dev */ opp_dev->dev = dev; +mutex_lock (&opp_table->lock);list_add (&opp_dev->node, &opp_table->dev_list); +mutex_unlock (&opp_table->lock);/* Create debugfs entries for the opp_table */ opp_debug_register (opp_dev, opp_table);
@@ -1056,18 +1060,6 @@ static struct opp_device *_add_opp_dev_unlocked(const struct device *dev,return opp_dev; } -struct opp_device *_add_opp_dev (const struct device *dev, -struct opp_table *opp_table) -{ -struct opp_device *opp_dev; - -mutex_lock (&opp_table->lock); - opp_dev =_add_opp_dev_unlocked (dev, opp_table); -mutex_unlock (&opp_table->lock); - -return opp_dev; -} -static struct opp_table *_allocate_opp_table (struct device *dev,int index) {struct opp_table *opp_table;
@@ -1121,8 +1113,6 @@ static struct opp_table *_allocate_opp_table(struct device *dev, int index)INIT_LIST_HEAD (&opp_table->opp_list);kref_init (&opp_table->kref); -/* Secure the device table modification */ -list_add (&opp_table->node, &opp_tables);return opp_table; err:
@@ -1135,27 +1125,64 @@ void _get_opp_table_kref(struct opp_table *opp_table)kref_get (&opp_table->kref); } +/* +* We need to make sure that the OPP table for a device doesn't get added twice, +* if this routine gets called in parallel with the same device pointer. +* +* The simplest way to enforce that is to perform everything (find existing +* table and if not found, create a new one) under the opp_table_lock, so only +* one creator gets access to the same. But that expands the critical section +* under the lock and may end up causing circular dependencies with frameworks +* like debugfs, interconnect or clock framework as they may be direct or +* indirect users of OPP core. +* +* And for that reason we have to go for a bit tricky implementation here, which +* uses the opp_tables_busy flag to indicate if another creator is in the middle +* of adding an OPP table and others should wait for it to finish. +*/ static struct opp_table *_opp_get_opp_table (struct device *dev,int index) {struct opp_table *opp_table; -/* Hold our table modification lock here */ +again:mutex_lock (&opp_table_lock); opp_table =_find_opp_table_unlocked (dev);if (!IS_ERR (opp_table)) goto unlock; +/* +* The opp_tables list or an OPP table's dev_list is getting updated by +* another user, wait for it to finish. +*/ +if (unlikely (opp_tables_busy)) { +mutex_unlock (&opp_table_lock); +cpu_relax (); + goto again; + } + + opp_tables_busy = true; opp_table =_managed_opp (dev, index); + +/* Drop the lock to reduce the size of critical section */ +mutex_unlock (&opp_table_lock); +if (opp_table) { -if (!_add_opp_dev_unlocked (dev, opp_table)) { +if (!_add_opp_dev (dev, opp_table)) {dev_pm_opp_put_opp_table (opp_table); opp_table =ERR_PTR (-ENOMEM ); } - goto unlock; + +mutex_lock (&opp_table_lock); + }else { + opp_table =_allocate_opp_table (dev, index); + +mutex_lock (&opp_table_lock); +if (!IS_ERR (opp_table)) +list_add (&opp_table->node, &opp_tables); } - opp_table =_allocate_opp_table (dev, index); + opp_tables_busy = false; unlock:mutex_unlock (&opp_table_lock);
@@ -1181,6 +1208,10 @@ static void _opp_table_kref_release(struct kref *kref)struct opp_device *opp_dev, *temp;int i; +/* Drop the lock as soon as we can */ +list_del (&opp_table->node); +mutex_unlock (&opp_table_lock); +_of_clear_opp_table (opp_table);/* Release clk */
@@ -1208,10 +1239,7 @@ static void _opp_table_kref_release(struct kref *kref)mutex_destroy (&opp_table->genpd_virt_dev_lock);mutex_destroy (&opp_table->lock); -list_del (&opp_table->node);kfree (opp_table); - -mutex_unlock (&opp_table_lock); }void dev_pm_opp_put_opp_table (struct opp_table *opp_table)
Signature · 1 line
viresh
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 152 lines · Viresh Kumar
It isn't that straight forward unfortunately, we need to make sure the table doesn't get allocated for the same device twice, so find+allocate needs to happen within a locked region.
I have taken, not so straight forward, approach to fixing this issue, lets see if this fixes it or not.
-------------------------8<-------------------------
drivers/opp/core.c +51
-20@@ -29,6 +29,8 @@LIST_HEAD (opp_tables);/* Lock to allow exclusive modification to the device and opp lists */ DEFINE_MUTEX (opp_table_lock); +/* Flag indicating that opp_tables list is being updated at the moment */ +static bool opp_tables_busy;static struct opp_device *_find_opp_dev (const struct device *dev,struct opp_table *opp_table)@@ -1036,8 +1038,8 @@ static void _remove_opp_dev(struct opp_device *opp_dev,kfree (opp_dev); } -static struct opp_device *_add_opp_dev_unlocked (const struct device *dev, -struct opp_table *opp_table) +struct opp_device *_add_opp_dev (const struct device *dev, +struct opp_table *opp_table) {struct opp_device *opp_dev;@@ -1048,7 +1050,9 @@ static struct opp_device *_add_opp_dev_unlocked(const struct device *dev,/* Initialize opp-dev */ opp_dev->dev = dev; +mutex_lock (&opp_table->lock);list_add (&opp_dev->node, &opp_table->dev_list); +mutex_unlock (&opp_table->lock);/* Create debugfs entries for the opp_table */ opp_debug_register (opp_dev, opp_table);@@ -1056,18 +1060,6 @@ static struct opp_device *_add_opp_dev_unlocked(const struct device *dev,return opp_dev; } -struct opp_device *_add_opp_dev (const struct device *dev, -struct opp_table *opp_table) -{ -struct opp_device *opp_dev; - -mutex_lock (&opp_table->lock); - opp_dev =_add_opp_dev_unlocked (dev, opp_table); -mutex_unlock (&opp_table->lock); - -return opp_dev; -} -static struct opp_table *_allocate_opp_table (struct device *dev,int index) {struct opp_table *opp_table;@@ -1121,8 +1113,6 @@ static struct opp_table *_allocate_opp_table(struct device *dev, int index)INIT_LIST_HEAD (&opp_table->opp_list);kref_init (&opp_table->kref); -/* Secure the device table modification */ -list_add (&opp_table->node, &opp_tables);return opp_table; err:@@ -1135,27 +1125,64 @@ void _get_opp_table_kref(struct opp_table *opp_table)kref_get (&opp_table->kref); } +/* +* We need to make sure that the OPP table for a device doesn't get added twice, +* if this routine gets called in parallel with the same device pointer. +* +* The simplest way to enforce that is to perform everything (find existing +* table and if not found, create a new one) under the opp_table_lock, so only +* one creator gets access to the same. But that expands the critical section +* under the lock and may end up causing circular dependencies with frameworks +* like debugfs, interconnect or clock framework as they may be direct or +* indirect users of OPP core. +* +* And for that reason we have to go for a bit tricky implementation here, which +* uses the opp_tables_busy flag to indicate if another creator is in the middle +* of adding an OPP table and others should wait for it to finish. +*/ static struct opp_table *_opp_get_opp_table (struct device *dev,int index) {struct opp_table *opp_table; -/* Hold our table modification lock here */ +again:mutex_lock (&opp_table_lock); opp_table =_find_opp_table_unlocked (dev);if (!IS_ERR (opp_table)) goto unlock; +/* +* The opp_tables list or an OPP table's dev_list is getting updated by +* another user, wait for it to finish. +*/ +if (unlikely (opp_tables_busy)) { +mutex_unlock (&opp_table_lock); +cpu_relax (); + goto again; + } + + opp_tables_busy = true; opp_table =_managed_opp (dev, index); + +/* Drop the lock to reduce the size of critical section */ +mutex_unlock (&opp_table_lock); +if (opp_table) { -if (!_add_opp_dev_unlocked (dev, opp_table)) { +if (!_add_opp_dev (dev, opp_table)) {dev_pm_opp_put_opp_table (opp_table); opp_table =ERR_PTR (-ENOMEM ); } - goto unlock; + +mutex_lock (&opp_table_lock); + }else { + opp_table =_allocate_opp_table (dev, index); + +mutex_lock (&opp_table_lock); +if (!IS_ERR (opp_table)) +list_add (&opp_table->node, &opp_tables); } - opp_table =_allocate_opp_table (dev, index); + opp_tables_busy = false; unlock:mutex_unlock (&opp_table_lock);@@ -1181,6 +1208,10 @@ static void _opp_table_kref_release(struct kref *kref)struct opp_device *opp_dev, *temp;int i; +/* Drop the lock as soon as we can */ +list_del (&opp_table->node); +mutex_unlock (&opp_table_lock); +_of_clear_opp_table (opp_table);/* Release clk */
drivers/opp/core.c static void _opp_table_kref_release(struct kref *kref) · line 1245
Quoted 10 lines
drivers/opp/core.c +0
-3@@ -1208,10 +1239,7 @@ static void _opp_table_kref_release(struct kref *kref)mutex_destroy (&opp_table->genpd_virt_dev_lock);mutex_destroy (&opp_table->lock); -list_del (&opp_table->node);kfree (opp_table); - -mutex_unlock (&opp_table_lock); }void dev_pm_opp_put_opp_table (struct opp_table *opp_table)
Rob, Ping.
Signature · 1 line
viresh
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 163 lines · Viresh Kumar
Quoted 161 lines · Viresh Kumar
It isn't that straight forward unfortunately, we need to make sure the table doesn't get allocated for the same device twice, so find+allocate needs to happen within a locked region.
I have taken, not so straight forward, approach to fixing this issue, lets see if this fixes it or not.
-------------------------8<-------------------------
drivers/opp/core.c +51
-23@@ -29,6 +29,8 @@LIST_HEAD (opp_tables);/* Lock to allow exclusive modification to the device and opp lists */ DEFINE_MUTEX (opp_table_lock); +/* Flag indicating that opp_tables list is being updated at the moment */ +static bool opp_tables_busy;static struct opp_device *_find_opp_dev (const struct device *dev,struct opp_table *opp_table)@@ -1036,8 +1038,8 @@ static void _remove_opp_dev(struct opp_device *opp_dev,kfree (opp_dev); } -static struct opp_device *_add_opp_dev_unlocked (const struct device *dev, -struct opp_table *opp_table) +struct opp_device *_add_opp_dev (const struct device *dev, +struct opp_table *opp_table) {struct opp_device *opp_dev;@@ -1048,7 +1050,9 @@ static struct opp_device *_add_opp_dev_unlocked(const struct device *dev,/* Initialize opp-dev */ opp_dev->dev = dev; +mutex_lock (&opp_table->lock);list_add (&opp_dev->node, &opp_table->dev_list); +mutex_unlock (&opp_table->lock);/* Create debugfs entries for the opp_table */ opp_debug_register (opp_dev, opp_table);@@ -1056,18 +1060,6 @@ static struct opp_device *_add_opp_dev_unlocked(const struct device *dev,return opp_dev; } -struct opp_device *_add_opp_dev (const struct device *dev, -struct opp_table *opp_table) -{ -struct opp_device *opp_dev; - -mutex_lock (&opp_table->lock); - opp_dev =_add_opp_dev_unlocked (dev, opp_table); -mutex_unlock (&opp_table->lock); - -return opp_dev; -} -static struct opp_table *_allocate_opp_table (struct device *dev,int index) {struct opp_table *opp_table;@@ -1121,8 +1113,6 @@ static struct opp_table *_allocate_opp_table(struct device *dev, int index)INIT_LIST_HEAD (&opp_table->opp_list);kref_init (&opp_table->kref); -/* Secure the device table modification */ -list_add (&opp_table->node, &opp_tables);return opp_table; err:@@ -1135,27 +1125,64 @@ void _get_opp_table_kref(struct opp_table *opp_table)kref_get (&opp_table->kref); } +/* +* We need to make sure that the OPP table for a device doesn't get added twice, +* if this routine gets called in parallel with the same device pointer. +* +* The simplest way to enforce that is to perform everything (find existing +* table and if not found, create a new one) under the opp_table_lock, so only +* one creator gets access to the same. But that expands the critical section +* under the lock and may end up causing circular dependencies with frameworks +* like debugfs, interconnect or clock framework as they may be direct or +* indirect users of OPP core. +* +* And for that reason we have to go for a bit tricky implementation here, which +* uses the opp_tables_busy flag to indicate if another creator is in the middle +* of adding an OPP table and others should wait for it to finish. +*/ static struct opp_table *_opp_get_opp_table (struct device *dev,int index) {struct opp_table *opp_table; -/* Hold our table modification lock here */ +again:mutex_lock (&opp_table_lock); opp_table =_find_opp_table_unlocked (dev);if (!IS_ERR (opp_table)) goto unlock; +/* +* The opp_tables list or an OPP table's dev_list is getting updated by +* another user, wait for it to finish. +*/ +if (unlikely (opp_tables_busy)) { +mutex_unlock (&opp_table_lock); +cpu_relax (); + goto again; + } + + opp_tables_busy = true; opp_table =_managed_opp (dev, index); + +/* Drop the lock to reduce the size of critical section */ +mutex_unlock (&opp_table_lock); +if (opp_table) { -if (!_add_opp_dev_unlocked (dev, opp_table)) { +if (!_add_opp_dev (dev, opp_table)) {dev_pm_opp_put_opp_table (opp_table); opp_table =ERR_PTR (-ENOMEM ); } - goto unlock; + +mutex_lock (&opp_table_lock); + }else { + opp_table =_allocate_opp_table (dev, index); + +mutex_lock (&opp_table_lock); +if (!IS_ERR (opp_table)) +list_add (&opp_table->node, &opp_tables); } - opp_table =_allocate_opp_table (dev, index); + opp_tables_busy = false; unlock:mutex_unlock (&opp_table_lock);@@ -1181,6 +1208,10 @@ static void _opp_table_kref_release(struct kref *kref)struct opp_device *opp_dev, *temp;int i; +/* Drop the lock as soon as we can */ +list_del (&opp_table->node); +mutex_unlock (&opp_table_lock); +_of_clear_opp_table (opp_table);/* Release clk */ @@ -1208,10 +1239,7 @@ static void _opp_table_kref_release(struct kref *kref)mutex_destroy (&opp_table->genpd_virt_dev_lock);mutex_destroy (&opp_table->lock); -list_del (&opp_table->node);kfree (opp_table); - -mutex_unlock (&opp_table_lock); }
Quoted 4 lines
drivers/opp/core.c static void _opp_table_kref_release(struct kref *kref) · line 1245
Quoted 1 line
drivers/opp/core.c +0
-0void dev_pm_opp_put_opp_table (struct opp_table *opp_table)Rob, Ping.
sorry, it didn't apply cleanly (which I guess is due to some other dependencies that need to be picked back to v5.4 product kernel), and due to some other things I'm in middle of debugging I didn't have time yet to switch to v5.10-rc or look at what else needs to cherry-picked..
If you could, pushing a branch with this patch somewhere would be a bit easier to work with (ie. fetch && cherry-pick is easier to deal with than picking things from list)
BR,
-R
_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 9 lines · Rob Clark
sorry, it didn't apply cleanly (which I guess is due to some other dependencies that need to be picked back to v5.4 product kernel), and due to some other things I'm in middle of debugging I didn't have time yet to switch to v5.10-rc or look at what else needs to cherry-picked..
If you could, pushing a branch with this patch somewhere would be a bit easier to work with (ie. fetch && cherry-pick is easier to deal with than picking things from list)
It has been in linux-next for a few days. Here is the HEAD to pick from. There are few patches there since rc1.
commit 203e29749cc0 ("opp: Allocate the OPP table outside of opp_table_lock")
Signature · 1 line
viresh
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 8 lines · Viresh Kumar
Quoted 6 lines · Rob Clark
sorry, it didn't apply cleanly (which I guess is due to some other dependencies that need to be picked back to v5.4 product kernel), and due to some other things I'm in middle of debugging I didn't have time yet to switch to v5.10-rc or look at what else needs to cherry-picked..
Quoted 9 lines
Quoted 3 lines
If you could, pushing a branch with this patch somewhere would be a bit easier to work with (ie. fetch && cherry-pick is easier to deal with than picking things from list)
It has been in linux-next for a few days. Here is the HEAD to pick from. There are few patches there since rc1.
commit
203e29749cc0("opp: Allocate the OPP table outside ofopp_table_lock")
sorry for the delay, with that cherry-picked, I'm getting a whole lot of:
[ 10.191497] WARNING: CPU: 7 PID: 52 at drivers/opp/of.c:115
_find_table_of_opp_np+0x8c/0x94
[ 10.191502] Modules linked in:
[ 10.191517] CPU: 7 PID: 52 Comm: kworker/7:1 Tainted: G W
5.10.0-rc2+ #2
[ 10.191522] Hardware name: Google Lazor (rev1+) with LTE (DT)
[ 10.191537] Workqueue: events deferred_probe_work_func
[ 10.191551] pstate: 60c00009 (nZCv daif +PAN +UAO -TCO BTYPE=--)
[ 10.202819] mmc0: CQHCI version 5.10
[ 10.206038] pc : _find_table_of_opp_np+0x8c/0x94
[ 10.206045] lr : _find_table_of_opp_np+0x34/0x94
[ 10.206050] sp : ffffffc010373810
[ 10.206054] x29: ffffffc010373810 x28: ffffff94c5a3d170
[ 10.206070] x27: ffffff94c5a3d168
[ 10.249098] mmc0: SDHCI controller on 7c4000.sdhci [7c4000.sdhci]
using ADMA 64-bit
[ 10.251366] x26: ffffff94c580c000
[ 10.251374] x25: 0000000000000001 x24: ffffff963f02c750
[ 10.251385] x23: 0000000000000000 x22: ffffff94c5aabc80
[ 10.251397] x21: ffffff963f021c78 x20: ffffff94c5a75800
[ 10.256963] sdhci_msm 7c4000.sdhci: mmc0: CQE init: success
[ 10.260376]
[ 10.260380] x19: ffffff963f02c750 x18: 0000000000000004
[ 10.260392] x17: 000000000000002c x16: ffffffe2468e1e78
[ 10.260404] x15: ffffffe246df3eb8 x14: ffffffff52f45308
[ 10.311816] x13: 0000000000000000 x12: ffffffe24541aef0
[ 10.317298] x11: ffffffe246df3eb8 x10: fffffffefe60e678
[ 10.322776] x9 : 0000000000000000 x8 : ffffffb3f89a7000
[ 10.328258] x7 : ffffffe245c5d9d0 x6 : 0000000000000000
[ 10.333730] x5 : 0000000000000080 x4 : 0000000000000001
[ 10.339206] x3 : 0000000000000000 x2 : 0000000000000006
[ 10.344684] x1 : ffffffe24684aa88 x0 : 0000000000000000
[ 10.350158] Call trace:
[ 10.352695] _find_table_of_opp_np+0x8c/0x94
[ 10.353507] mmc0: Command Queue Engine enabled
[ 10.357095] _of_init_opp_table+0x15c/0x1e4
[ 10.357103] _opp_get_opp_table+0x168/0x280
[ 10.357110] dev_pm_opp_set_clkname+0x28/0xcc
[ 10.357119] dpu_bind+0x50/0x1a4
[ 10.357128] component_bind_all+0xf4/0x20c
[ 10.357138] msm_drm_init+0x180/0x588
[ 10.361815] mmc0: new HS400 Enhanced strobe MMC card at address 0001
[ 10.366050] msm_drm_bind+0x1c/0x24
[ 10.366057] try_to_bring_up_master+0x160/0x1a8
[ 10.366065] component_master_add_with_match+0xc4/0x108
[ 10.366072] msm_pdev_probe+0x214/0x2a4
[ 10.366081] platform_drv_probe+0x94/0xb4
[ 10.374415] mmcblk0: mmc0:0001 DA4064 58.2 GiB
[ 10.374871] really_probe+0x138/0x348
[ 10.374881] driver_probe_device+0x80/0xb8
[ 10.379483] mmcblk0boot0: mmc0:0001 DA4064 partition 1 4.00 MiB
[ 10.382446] __device_attach_driver+0x90/0xa8
[ 10.382453] bus_for_each_drv+0x84/0xcc
[ 10.382459] __device_attach+0xc0/0x148
[ 10.382466] device_initial_probe+0x18/0x20
[ 10.382473] bus_probe_device+0x38/0x98
[ 10.382483] deferred_probe_work_func+0x7c/0xb8
[ 10.387402] mmcblk0boot1: mmc0:0001 DA4064 partition 2 4.00 MiB
[ 10.392780] process_one_work+0x314/0x60c
[ 10.392786] worker_thread+0x238/0x3e8
[ 10.392793] kthread+0x148/0x158
[ 10.392800] ret_from_fork+0x10/0x18
[ 10.392809] CPU: 7 PID: 52 Comm: kworker/7:1 Tainted: G W
5.10.0-rc2+ #2
[ 10.397683] mmcblk0rpmb: mmc0:0001 DA4064 partition 3 16.0 MiB,
chardev (241:0)
[ 10.401051] Hardware name: Google Lazor (rev1+) with LTE (DT)
[ 10.401062] Workqueue: events deferred_probe_work_func
[ 10.401069] Call trace:
[ 10.401077] dump_backtrace+0x0/0x1b4
[ 10.401087] show_stack+0x1c/0x24
[ 10.427111] mmcblk0: p1 p2 p3 p4 p5 p6 p7 p8 p9 p10 p11 p12
[ 10.427156] dump_stack+0xdc/0x158
[ 10.427165] __warn+0xd8/0x16c
[ 10.427173] report_bug+0x88/0xe0
[ 10.427179] bug_handler+0x24/0x6c
[ 10.535574] brk_handler+0x78/0xb4
[ 10.539090] do_debug_exception+0x1a4/0x208
[ 10.543395] el1_sync_handler+0x8c/0x110
[ 10.547434] el1_sync+0x7c/0x100
[ 10.550762] _find_table_of_opp_np+0x8c/0x94
[ 10.555166] _of_init_opp_table+0x15c/0x1e4
[ 10.559472] _opp_get_opp_table+0x168/0x280
[ 10.563779] dev_pm_opp_set_clkname+0x28/0xcc
[ 10.568270] dpu_bind+0x50/0x1a4
[ 10.571607] component_bind_all+0xf4/0x20c
[ 10.575826] msm_drm_init+0x180/0x588
[ 10.579603] msm_drm_bind+0x1c/0x24
[ 10.583205] try_to_bring_up_master+0x160/0x1a8
[ 10.587877] component_master_add_with_match+0xc4/0x108
[ 10.593251] msm_pdev_probe+0x214/0x2a4
[ 10.597203] platform_drv_probe+0x94/0xb4
[ 10.601334] really_probe+0x138/0x348
[ 10.605110] driver_probe_device+0x80/0xb8
[ 10.609329] __device_attach_driver+0x90/0xa8
[ 10.613821] bus_for_each_drv+0x84/0xcc
[ 10.617774] __device_attach+0xc0/0x148
[ 10.621729] device_initial_probe+0x18/0x20
[ 10.626044] bus_probe_device+0x38/0x98
[ 10.629998] deferred_probe_work_func+0x7c/0xb8
[ 10.634668] process_one_work+0x314/0x60c
[ 10.638797] worker_thread+0x238/0x3e8
[ 10.642661] kthread+0x148/0x158
[ 10.645997] ret_from_fork+0x10/0x18
[ 10.649683] irq event stamp: 117274
[ 10.653290] hardirqs last enabled at (117273):
[<ffffffe245ed8430>] _raw_spin_unlock_irqrestore+0x60/0x94
[ 10.663213] hardirqs last disabled at (117274):
[<ffffffe245420ea0>] do_debug_exception+0x60/0x208
[ 10.672420] softirqs last enabled at (116976):
[<ffffffe245400eec>] __do_softirq+0x4bc/0x540
[ 10.681184] softirqs last disabled at (116971):
[<ffffffe24547dd10>] __irq_exit_rcu+0x118/0x138
[ 10.690123] ---[ end trace 00b127c206a99072 ]---Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 13 lines · Rob Clark
Quoted 12 lines · Viresh Kumar
Quoted 9 lines · Rob Clark
sorry, it didn't apply cleanly (which I guess is due to some other dependencies that need to be picked back to v5.4 product kernel), and due to some other things I'm in middle of debugging I didn't have time yet to switch to v5.10-rc or look at what else needs to cherry-picked..
If you could, pushing a branch with this patch somewhere would be a bit easier to work with (ie. fetch && cherry-pick is easier to deal with than picking things from list)
Quoted 7 lines
Quoted 5 lines
It has been in linux-next for a few days. Here is the HEAD to pick from. There are few patches there since rc1.
commit
203e29749cc0("opp: Allocate the OPP table outside ofopp_table_lock")sorry for the delay, with that cherry-picked, I'm getting a whole lot of:
Ahh, sorry about that and thanks for reporting it. Here is the fix:
drivers/opp/of.c +3 -2
@@ -112,8 +112,6 @@ static struct opp_table *_find_table_of_opp_np(struct device_node *opp_np)struct opp_table *opp_table;struct device_node *opp_table_np; -lockdep_assert_held (&opp_table_lock); - opp_table_np =of_get_parent (opp_np);if (!opp_table_np) goto err;
@@ -121,12 +119,15 @@ static struct opp_table *_find_table_of_opp_np(struct device_node *opp_np)/* It is safe to put the node now as all we need now is its address */ of_node_put (opp_table_np); +mutex_lock (&opp_table_lock);list_for_each_entry (opp_table, &opp_tables, node) {if (opp_table_np == opp_table->np) {_get_opp_table_kref (opp_table); +mutex_unlock (&opp_table_lock);return opp_table; } } +mutex_unlock (&opp_table_lock); err:return ERR_PTR (-ENODEV );
Signature · 1 line
viresh
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 44 lines · Viresh Kumar
Quoted 13 lines · Rob Clark
Quoted 12 lines · Viresh Kumar
Quoted 9 lines · Rob Clark
sorry, it didn't apply cleanly (which I guess is due to some other dependencies that need to be picked back to v5.4 product kernel), and due to some other things I'm in middle of debugging I didn't have time yet to switch to v5.10-rc or look at what else needs to cherry-picked..
If you could, pushing a branch with this patch somewhere would be a bit easier to work with (ie. fetch && cherry-pick is easier to deal with than picking things from list)
Quoted 7 lines
Quoted 5 lines
It has been in linux-next for a few days. Here is the HEAD to pick from. There are few patches there since rc1.
commit
203e29749cc0("opp: Allocate the OPP table outside ofopp_table_lock")sorry for the delay, with that cherry-picked, I'm getting a whole lot of:
Ahh, sorry about that and thanks for reporting it. Here is the fix:
drivers/opp/of.c +1
-2@@ -112,8 +112,6 @@ static struct opp_table *_find_table_of_opp_np(struct device_node *opp_np)struct opp_table *opp_table;struct device_node *opp_table_np; -lockdep_assert_held (&opp_table_lock); - opp_table_np =of_get_parent (opp_np);if (!opp_table_np) goto err;@@ -121,12 +119,15 @@ static struct opp_table *_find_table_of_opp_np(struct device_node *opp_np)/* It is safe to put the node now as all we need now is its address */ of_node_put (opp_table_np); +mutex_lock (&opp_table_lock);list_for_each_entry (opp_table, &opp_tables, node) {
drivers/opp/of.c static struct opp_table *_find_table_of_opp_np(struct device_node *opp_np) · line 133
Quoted 10 lines
drivers/opp/of.c +2
-0if (opp_table_np == opp_table->np) {_get_opp_table_kref (opp_table); +mutex_unlock (&opp_table_lock);return opp_table; } } +mutex_unlock (&opp_table_lock); err:return ERR_PTR (-ENODEV );
Ping.
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
In reply to Viresh KumarQuoted 45 lines · Viresh Kumar
Quoted 13 lines · Rob Clark
Quoted 12 lines · Viresh Kumar
Quoted 9 lines · Rob Clark
sorry, it didn't apply cleanly (which I guess is due to some other dependencies that need to be picked back to v5.4 product kernel), and due to some other things I'm in middle of debugging I didn't have time yet to switch to v5.10-rc or look at what else needs to cherry-picked..
If you could, pushing a branch with this patch somewhere would be a bit easier to work with (ie. fetch && cherry-pick is easier to deal with than picking things from list)
Quoted 7 lines
Quoted 5 lines
It has been in linux-next for a few days. Here is the HEAD to pick from. There are few patches there since rc1.
commit
203e29749cc0("opp: Allocate the OPP table outside ofopp_table_lock")sorry for the delay, with that cherry-picked, I'm getting a whole lot of:
Ahh, sorry about that and thanks for reporting it. Here is the fix:
drivers/opp/of.c +1
-2@@ -112,8 +112,6 @@ static struct opp_table *_find_table_of_opp_np(struct device_node *opp_np)struct opp_table *opp_table;struct device_node *opp_table_np; -lockdep_assert_held (&opp_table_lock); - opp_table_np =of_get_parent (opp_np);if (!opp_table_np) goto err;@@ -121,12 +119,15 @@ static struct opp_table *_find_table_of_opp_np(struct device_node *opp_np)/* It is safe to put the node now as all we need now is its address */ of_node_put (opp_table_np); +mutex_lock (&opp_table_lock);list_for_each_entry (opp_table, &opp_tables, node) {
drivers/opp/of.c static struct opp_table *_find_table_of_opp_np(struct device_node *opp_np) · line 133
Quoted 10 lines
drivers/opp/of.c +2
-0if (opp_table_np == opp_table->np) {_get_opp_table_kref (opp_table); +mutex_unlock (&opp_table_lock);return opp_table; } } +mutex_unlock (&opp_table_lock); err:return ERR_PTR (-ENODEV );
With that on top of the previous patch,
[ 26.378245] ======================================================
[ 26.384595] WARNING: possible circular locking dependency detected
[ 26.390947] 5.10.0-rc2+ #6 Not tainted
[ 26.394804] ------------------------------------------------------
[ 26.401155] chrome/1886 is trying to acquire lock:
[ 26.406087] ffffffe5e264aa88 (opp_table_lock){+.+.}-{3:3}, at:
_find_opp_table+0x38/0x78
[ 26.414436]
[ 26.414436] but task is already holding lock:
[ 26.420423] ffffffb0283935b0
(reservation_ww_class_mutex){+.+.}-{3:3}, at:
submit_lock_objects+0x70/0x1ec
[ 26.430268]
[ 26.430268] which lock already depends on the new lock.
[ 26.430268]
[ 26.438661]
[ 26.438661] the existing dependency chain (in reverse order) is:
[ 26.446343]
[ 26.446343] -> #3 (reservation_ww_class_mutex){+.+.}-{3:3}:
[ 26.453603] lock_acquire+0x23c/0x30c
[ 26.457910] __mutex_lock_common+0xdc/0xbc4
[ 26.462743] ww_mutex_lock_interruptible+0x84/0xec
[ 26.468203] msm_gem_fault+0x30/0x138
[ 26.472507] __do_fault+0x44/0x184
[ 26.476541] handle_mm_fault+0x754/0xc50
[ 26.481117] do_page_fault+0x230/0x354
[ 26.485507] do_translation_fault+0x40/0x54
[ 26.490338] do_mem_abort+0x44/0xac
[ 26.494469] el0_sync_compat_handler+0x15c/0x190
[ 26.499756] el0_sync_compat+0x144/0x180
[ 26.504328]
[ 26.504328] -> #2 (&mm->mmap_lock){++++}-{3:3}:
[ 26.510511] lock_acquire+0x23c/0x30c
[ 26.514813] __might_fault+0x60/0x80
[ 26.519034] compat_filldir+0x118/0x4d0
[ 26.523519] dcache_readdir+0x74/0x1e0
[ 26.527907] iterate_dir+0xd4/0x198
[ 26.532037] __arm64_compat_sys_getdents+0x6c/0x168
[ 26.537583] el0_svc_common+0xa4/0x174
[ 26.541970] do_el0_svc_compat+0x20/0x30
[ 26.546543] el0_sync_compat_handler+0x124/0x190
[ 26.551828] el0_sync_compat+0x144/0x180
[ 26.556399]
[ 26.556399] -> #1 (&sb->s_type->i_mutex_key#2){++++}-{3:3}:
[ 26.563660] lock_acquire+0x23c/0x30c
[ 26.567963] down_write+0x80/0x1dc
[ 26.571997] simple_recursive_removal+0x48/0x238
[ 26.577288] debugfs_remove+0x5c/0x78
[ 26.581592] opp_debug_unregister+0x34/0x118
[ 26.586521] dev_pm_opp_put_opp_table+0xd0/0x14c
[ 26.591802] dev_pm_opp_put_clkname+0x40/0x54
[ 26.596820] msm_dsi_host_destroy+0xe0/0x108
[ 26.601748] dsi_destroy+0x40/0x58
[ 26.605789] dsi_bind+0x8c/0x16c
[ 26.609648] component_bind_all+0xf4/0x20c
[ 26.614399] msm_drm_init+0x180/0x588
[ 26.618696] msm_drm_bind+0x1c/0x24
[ 26.622822] try_to_bring_up_master+0x160/0x1a8
[ 26.628014] component_master_add_with_match+0xc4/0x108
[ 26.633918] msm_pdev_probe+0x214/0x2a4
[ 26.638405] platform_drv_probe+0x94/0xb4
[ 26.643066] really_probe+0x138/0x348
[ 26.647365] driver_probe_device+0x80/0xb8
[ 26.652111] device_driver_attach+0x50/0x70
[ 26.656951] __driver_attach+0xb4/0xc8
[ 26.661340] bus_for_each_dev+0x80/0xc8
[ 26.665825] driver_attach+0x28/0x30
[ 26.670038] bus_add_driver+0x100/0x1d4
[ 26.674524] driver_register+0x68/0xfc
[ 26.678910] __platform_driver_register+0x48/0x50
[ 26.684284] msm_drm_register+0x64/0x68
[ 26.688766] do_one_initcall+0x1ac/0x3e4
[ 26.693341] do_initcall_level+0xa0/0xb8
[ 26.697910] do_initcalls+0x58/0x94
[ 26.702039] do_basic_setup+0x28/0x30
[ 26.706338] kernel_init_freeable+0x190/0x1d0
[ 26.711352] kernel_init+0x18/0x10c
[ 26.715481] ret_from_fork+0x10/0x18
[ 26.719692]
[ 26.719692] -> #0 (opp_table_lock){+.+.}-{3:3}:
[ 26.725883] check_noncircular+0x12c/0x134
[ 26.730627] __lock_acquire+0x2288/0x2b2c
[ 26.735284] lock_acquire+0x23c/0x30c
[ 26.739584] __mutex_lock_common+0xdc/0xbc4
[ 26.744426] mutex_lock_nested+0x50/0x58
[ 26.748998] _find_opp_table+0x38/0x78
[ 26.753395] dev_pm_opp_find_freq_exact+0x2c/0xdc
[ 26.758771] a6xx_gmu_resume+0xcc/0xed0
[ 26.763255] a6xx_pm_resume+0x140/0x174
[ 26.767741] adreno_resume+0x24/0x2c
[ 26.771956] pm_generic_runtime_resume+0x2c/0x3c
[ 26.777244] __rpm_callback+0x74/0x114
[ 26.781633] rpm_callback+0x30/0x84
[ 26.785759] rpm_resume+0x3c8/0x4f0
[ 26.789884] __pm_runtime_resume+0x80/0xa4
[ 26.794631] msm_gpu_submit+0x60/0x228
[ 26.799019] msm_ioctl_gem_submit+0xba0/0xc1c
[ 26.804038] drm_ioctl_kernel+0xa0/0x11c
[ 26.808608] drm_ioctl+0x240/0x3dc
[ 26.812653] drm_compat_ioctl+0xd4/0xe4
[ 26.817141] __arm64_compat_sys_ioctl+0xc4/0xf8
[ 26.822331] el0_svc_common+0xa4/0x174
[ 26.826718] do_el0_svc_compat+0x20/0x30
[ 26.831291] el0_sync_compat_handler+0x124/0x190
[ 26.836577] el0_sync_compat+0x144/0x180
[ 26.841148]
[ 26.841148] other info that might help us debug this:
[ 26.841148]
[ 26.849361] Chain exists of:
[ 26.849361] opp_table_lock --> &mm->mmap_lock -->
reservation_ww_class_mutex
[ 26.849361]
[ 26.861249] Possible unsafe locking scenario:
[ 26.861249]
[ 26.867334] CPU0 CPU1
[ 26.871990] ---- ----
[ 26.876647] lock(reservation_ww_class_mutex);
[ 26.881309] lock(&mm->mmap_lock);
[ 26.887487] lock(reservation_ww_class_mutex);
[ 26.894730] lock(opp_table_lock);
[ 26.898327]
[ 26.898327] *** DEADLOCK ***
[ 26.898327]
[ 26.904410] 3 locks held by chrome/1886:
[ 26.908447] #0: ffffffb005bd9138 (&dev->struct_mutex){+.+.}-{3:3},
at: msm_ioctl_gem_submit+0x238/0xc1c
[ 26.918199] #1: ffffffb02251fa70
(reservation_ww_class_acquire){+.+.}-{0:0}, at:
msm_ioctl_gem_submit+0x978/0xc1c
[ 26.928843] #2: ffffffb0283935b0
(reservation_ww_class_mutex){+.+.}-{3:3}, at:
submit_lock_objects+0x70/0x1ec
[ 26.939126]
[ 26.939126] stack backtrace:
[ 26.943612] CPU: 5 PID: 1886 Comm: chrome Not tainted 5.10.0-rc2+ #6
[ 26.950137] Hardware name: Google Lazor (rev1+) with LTE (DT)
[ 26.956039] Call trace:
[ 26.958566] dump_backtrace+0x0/0x1b4
[ 26.962333] show_stack+0x1c/0x24
[ 26.965754] dump_stack+0xdc/0x158
[ 26.969251] print_circular_bug+0x308/0x338
[ 26.973550] check_noncircular+0x12c/0x134
[ 26.977762] __lock_acquire+0x2288/0x2b2c
[ 26.981889] lock_acquire+0x23c/0x30c
[ 26.985658] __mutex_lock_common+0xdc/0xbc4
[ 26.989961] mutex_lock_nested+0x50/0x58
[ 26.993994] _find_opp_table+0x38/0x78
[ 26.997852] dev_pm_opp_find_freq_exact+0x2c/0xdc
[ 27.002690] a6xx_gmu_resume+0xcc/0xed0
[ 27.006635] a6xx_pm_resume+0x140/0x174
[ 27.010580] adreno_resume+0x24/0x2c
[ 27.014259] pm_generic_runtime_resume+0x2c/0x3c
[ 27.019008] __rpm_callback+0x74/0x114
[ 27.022868] rpm_callback+0x30/0x84
[ 27.026455] rpm_resume+0x3c8/0x4f0
[ 27.030042] __pm_runtime_resume+0x80/0xa4
[ 27.034259] msm_gpu_submit+0x60/0x228
[ 27.038117] msm_ioctl_gem_submit+0xba0/0xc1c
[ 27.042604] drm_ioctl_kernel+0xa0/0x11c
[ 27.046637] drm_ioctl+0x240/0x3dc
[ 27.050139] drm_compat_ioctl+0xd4/0xe4
[ 27.054088] __arm64_compat_sys_ioctl+0xc4/0xf8
[ 27.058748] el0_svc_common+0xa4/0x174
[ 27.062606] do_el0_svc_compat+0x20/0x30
[ 27.066647] el0_sync_compat_handler+0x124/0x190
[ 27.071393] el0_sync_compat+0x144/0x180BR,
-R
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 1 line · Rob Clark
With that on top of the previous patch,
Don't you still have this ? Which fixed the lockdep in the remove path.
https://lore.kernel.org/lkml/20201022080644.2ck4okrxygmkuatn@vireshk-i7/
To make it clear you need these patches to fix the OPP stuff:
//From 5.10-rc3 (the one from the above link).
commit e0df59de670b ("opp: Reduce the size of critical section in _opp_table_kref_release()")
//Below two from linux-next
commit ef43f01ac069 ("opp: Always add entries in dev_list with opp_table->lock held") commit 27c09484dd3d ("opp: Allocate the OPP table outside of opp_table_lock")
This matches the diff I gave you earlier.
Signature · 1 line
viresh
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 10 lines · Viresh Kumar
Quoted 1 line · Rob Clark
With that on top of the previous patch,
Don't you still have this ? Which fixed the lockdep in the remove path.
https://lore.kernel.org/lkml/20201022080644.2ck4okrxygmkuatn@vireshk-i7/
To make it clear you need these patches to fix the OPP stuff:
Quoted 9 lines
//From 5.10-rc3 (the one from the above link).
commite0df59de670b("opp: Reduce the size of critical section in_opp_table_kref_release()")//Below two from linux-next
commitef43f01ac069("opp: Always add entries indev_listwithopp_table->lockheld") commit27c09484dd3d("opp: Allocate the OPP table outside ofopp_table_lock")This matches the diff I gave you earlier.
no, I did not have all three, only "opp: Allocate the OPP table outside of opp_table_lock" plus the fixup. But with all three:
[ 27.072188] ======================================================
[ 27.078542] WARNING: possible circular locking dependency detected
[ 27.084897] 5.10.0-rc2+ #1 Not tainted
[ 27.088750] ------------------------------------------------------
[ 27.095103] chrome/1897 is trying to acquire lock:
[ 27.100031] ffffffdb14e4aa88 (opp_table_lock){+.+.}-{3:3}, at:
_find_opp_table+0x38/0x78
[ 27.108379]
[ 27.108379] but task is already holding lock:
[ 27.114373] ffffff8e2c8f91b0
(reservation_ww_class_mutex){+.+.}-{3:3}, at:
submit_lock_objects+0x70/0x1ec
[ 27.124212]
[ 27.124212] which lock already depends on the new lock.
[ 27.124212]
[ 27.132604]
[ 27.132604] the existing dependency chain (in reverse order) is:
[ 27.140290]
[ 27.140290] -> #4 (reservation_ww_class_mutex){+.+.}-{3:3}:
[ 27.147544] lock_acquire+0x23c/0x30c
[ 27.151848] __mutex_lock_common+0xdc/0xbc4
[ 27.156685] ww_mutex_lock_interruptible+0x84/0xec
[ 27.162142] msm_gem_fault+0x30/0x138
[ 27.166443] __do_fault+0x44/0x184
[ 27.170479] handle_mm_fault+0x754/0xc50
[ 27.175053] do_page_fault+0x230/0x354
[ 27.179444] do_translation_fault+0x40/0x54
[ 27.184277] do_mem_abort+0x44/0xac
[ 27.188402] el0_sync_compat_handler+0x15c/0x190
[ 27.193680] el0_sync_compat+0x144/0x180
[ 27.198244]
[ 27.198244] -> #3 (&mm->mmap_lock){++++}-{3:3}:
[ 27.204435] lock_acquire+0x23c/0x30c
[ 27.208738] __might_fault+0x60/0x80
[ 27.212951] compat_filldir+0x118/0x4d0
[ 27.217434] dcache_readdir+0x74/0x1e0
[ 27.221825] iterate_dir+0xd4/0x198
[ 27.225947] __arm64_compat_sys_getdents+0x6c/0x168
[ 27.231495] el0_svc_common+0xa4/0x174
[ 27.235886] do_el0_svc_compat+0x20/0x30
[ 27.240461] el0_sync_compat_handler+0x124/0x190
[ 27.245746] el0_sync_compat+0x144/0x180
[ 27.250310]
[ 27.250310] -> #2 (&sb->s_type->i_mutex_key#2){++++}-{3:3}:
[ 27.257569] lock_acquire+0x23c/0x30c
[ 27.261877] down_write+0x80/0x1dc
[ 27.265912] simple_recursive_removal+0x48/0x238
[ 27.271193] debugfs_remove+0x5c/0x78
[ 27.275502] opp_debug_remove_one+0x18/0x20
[ 27.280343] _opp_kref_release+0x40/0x74
[ 27.284917] dev_pm_opp_put_unlocked+0x44/0x64
[ 27.290015] _opp_remove_all_static+0x5c/0x90
[ 27.295029] dev_pm_opp_remove_table+0x70/0x90
[ 27.300129] dev_pm_opp_of_remove_table+0x14/0x1c
[ 27.305504] msm_dsi_host_destroy+0xd8/0x108
[ 27.310434] dsi_destroy+0x40/0x58
[ 27.314469] dsi_bind+0x8c/0x16c
[ 27.318329] component_bind_all+0xf4/0x20c
[ 27.323081] msm_drm_init+0x180/0x588
[ 27.327382] msm_drm_bind+0x1c/0x24
[ 27.331503] try_to_bring_up_master+0x160/0x1a8
[ 27.336696] component_master_add_with_match+0xc4/0x108
[ 27.342597] msm_pdev_probe+0x214/0x2a4
[ 27.347076] platform_drv_probe+0x94/0xb4
[ 27.351739] really_probe+0x138/0x348
[ 27.356041] driver_probe_device+0x80/0xb8
[ 27.360788] device_driver_attach+0x50/0x70
[ 27.365621] __driver_attach+0xb4/0xc8
[ 27.370012] bus_for_each_dev+0x80/0xc8
[ 27.374495] driver_attach+0x28/0x30
[ 27.378712] bus_add_driver+0x100/0x1d4
[ 27.383188] driver_register+0x68/0xfc
[ 27.387579] __platform_driver_register+0x48/0x50
[ 27.392957] msm_drm_register+0x64/0x68
[ 27.397434] do_one_initcall+0x1ac/0x3e4
[ 27.402011] do_initcall_level+0xa0/0xb8
[ 27.406583] do_initcalls+0x58/0x94
[ 27.410704] do_basic_setup+0x28/0x30
[ 27.415008] kernel_init_freeable+0x190/0x1d0
[ 27.420024] kernel_init+0x18/0x10c
[ 27.424146] ret_from_fork+0x10/0x18
[ 27.428362]
[ 27.428362] -> #1 (&opp_table->lock){+.+.}-{3:3}:
[ 27.434725] lock_acquire+0x23c/0x30c
[ 27.439028] __mutex_lock_common+0xdc/0xbc4
[ 27.443862] mutex_lock_nested+0x50/0x58
[ 27.448436] _find_opp_table_unlocked+0x44/0xb4
[ 27.453626] _opp_get_opp_table+0x3c/0x280
[ 27.458375] dev_pm_opp_get_opp_table_indexed+0x14/0x1c
[ 27.464281] of_genpd_add_provider_onecell+0xd8/0x1c0
[ 27.470019] rpmhpd_probe+0x244/0x26c
[ 27.474323] platform_drv_probe+0x94/0xb4
[ 27.478985] really_probe+0x138/0x348
[ 27.483287] driver_probe_device+0x80/0xb8
[ 27.488033] __device_attach_driver+0x90/0xa8
[ 27.493047] bus_for_each_drv+0x84/0xcc
[ 27.497524] __device_attach+0xc0/0x148
[ 27.502007] device_initial_probe+0x18/0x20
[ 27.506840] bus_probe_device+0x38/0x98
[ 27.511317] device_add+0x214/0x3c8
[ 27.515443] of_device_add+0x3c/0x48
[ 27.519654] of_platform_device_create_pdata+0xac/0xec
[ 27.525473] of_platform_bus_create+0x1cc/0x348
[ 27.530664] of_platform_populate+0x78/0xc8
[ 27.535496] devm_of_platform_populate+0x5c/0xa4
[ 27.540779] rpmh_rsc_probe+0x370/0x3d0
[ 27.545253] platform_drv_probe+0x94/0xb4
[ 27.549916] really_probe+0x138/0x348
[ 27.554223] driver_probe_device+0x80/0xb8
[ 27.558971] __device_attach_driver+0x90/0xa8
[ 27.563988] bus_for_each_drv+0x84/0xcc
[ 27.568465] __device_attach+0xc0/0x148
[ 27.572942] device_initial_probe+0x18/0x20
[ 27.577778] bus_probe_device+0x38/0x98
[ 27.582263] fw_devlink_resume+0xdc/0x110
[ 27.586930] of_platform_default_populate_init+0xb8/0xd0
[ 27.592923] do_one_initcall+0x1ac/0x3e4
[ 27.597489] do_initcall_level+0xa0/0xb8
[ 27.602051] do_initcalls+0x58/0x94
[ 27.606175] do_basic_setup+0x28/0x30
[ 27.610472] kernel_init_freeable+0x190/0x1d0
[ 27.615493] kernel_init+0x18/0x10c
[ 27.619616] ret_from_fork+0x10/0x18
[ 27.623823]
[ 27.623823] -> #0 (opp_table_lock){+.+.}-{3:3}:
[ 27.630006] check_noncircular+0x12c/0x134
[ 27.634757] __lock_acquire+0x2288/0x2b2c
[ 27.639419] lock_acquire+0x23c/0x30c
[ 27.643727] __mutex_lock_common+0xdc/0xbc4
[ 27.648566] mutex_lock_nested+0x50/0x58
[ 27.653133] _find_opp_table+0x38/0x78
[ 27.657520] dev_pm_opp_find_freq_exact+0x2c/0xdc
[ 27.662890] a6xx_gmu_resume+0xcc/0xed0
[ 27.667372] a6xx_pm_resume+0x140/0x174
[ 27.671849] adreno_resume+0x24/0x2c
[ 27.676070] pm_generic_runtime_resume+0x2c/0x3c
[ 27.681351] __rpm_callback+0x74/0x114
[ 27.685741] rpm_callback+0x30/0x84
[ 27.689865] rpm_resume+0x3c8/0x4f0
[ 27.693989] __pm_runtime_resume+0x80/0xa4
[ 27.698742] msm_gpu_submit+0x60/0x228
[ 27.703136] msm_ioctl_gem_submit+0xba0/0xc1c
[ 27.708158] drm_ioctl_kernel+0xa0/0x11c
[ 27.712724] drm_ioctl+0x240/0x3dc
[ 27.716762] drm_compat_ioctl+0xd4/0xe4
[ 27.721244] __arm64_compat_sys_ioctl+0xc4/0xf8
[ 27.726435] el0_svc_common+0xa4/0x174
[ 27.730827] do_el0_svc_compat+0x20/0x30
[ 27.735395] el0_sync_compat_handler+0x124/0x190
[ 27.740675] el0_sync_compat+0x144/0x180
[ 27.745240]
[ 27.745240] other info that might help us debug this:
[ 27.745240]
[ 27.753459] Chain exists of:
[ 27.753459] opp_table_lock --> &mm->mmap_lock -->
reservation_ww_class_mutex
[ 27.753459]
[ 27.765342] Possible unsafe locking scenario:
[ 27.765342]
[ 27.771422] CPU0 CPU1
[ 27.776085] ---- ----
[ 27.780747] lock(reservation_ww_class_mutex);
[ 27.785413] lock(&mm->mmap_lock);
[ 27.791591] lock(reservation_ww_class_mutex);
[ 27.798833] lock(opp_table_lock);
[ 27.802428]
[ 27.802428] *** DEADLOCK ***
[ 27.802428]
[ 27.808506] 3 locks held by chrome/1897:
[ 27.812540] #0: ffffff8e05f91138 (&dev->struct_mutex){+.+.}-{3:3},
at: msm_ioctl_gem_submit+0x238/0xc1c
[ 27.822295] #1: ffffff8e1ebd2670
(reservation_ww_class_acquire){+.+.}-{0:0}, at:
msm_ioctl_gem_submit+0x978/0xc1c
[ 27.832930] #2: ffffff8e2c8f91b0
(reservation_ww_class_mutex){+.+.}-{3:3}, at:
submit_lock_objects+0x70/0x1ec
[ 27.843216]
[ 27.843216] stack backtrace:
[ 27.847702] CPU: 5 PID: 1897 Comm: chrome Not tainted 5.10.0-rc2+ #1
[ 27.854235] Hardware name: Google Lazor (rev1+) with LTE (DT)
[ 27.860142] Call trace:
[ 27.862662] dump_backtrace+0x0/0x1b4
[ 27.866426] show_stack+0x1c/0x24
[ 27.869847] dump_stack+0xdc/0x158
[ 27.873349] print_circular_bug+0x308/0x338
[ 27.877647] check_noncircular+0x12c/0x134
[ 27.881858] __lock_acquire+0x2288/0x2b2c
[ 27.885984] lock_acquire+0x23c/0x30c
[ 27.889753] __mutex_lock_common+0xdc/0xbc4
[ 27.894054] mutex_lock_nested+0x50/0x58
[ 27.898086] _find_opp_table+0x38/0x78
[ 27.901946] dev_pm_opp_find_freq_exact+0x2c/0xdc
[ 27.906784] a6xx_gmu_resume+0xcc/0xed0
[ 27.910734] a6xx_pm_resume+0x140/0x174
[ 27.914684] adreno_resume+0x24/0x2c
[ 27.918363] pm_generic_runtime_resume+0x2c/0x3c
[ 27.923113] __rpm_callback+0x74/0x114
[ 27.926975] rpm_callback+0x30/0x84
[ 27.930565] rpm_resume+0x3c8/0x4f0
[ 27.934154] __pm_runtime_resume+0x80/0xa4
[ 27.938373] msm_gpu_submit+0x60/0x228
[ 27.942233] msm_ioctl_gem_submit+0xba0/0xc1c
[ 27.946713] drm_ioctl_kernel+0xa0/0x11c
[ 27.950749] drm_ioctl+0x240/0x3dc
[ 27.954256] drm_compat_ioctl+0xd4/0xe4
[ 27.958207] __arm64_compat_sys_ioctl+0xc4/0xf8
[ 27.962871] el0_svc_common+0xa4/0x174
[ 27.966731] do_el0_svc_compat+0x20/0x30
[ 27.970766] el0_sync_compat_handler+0x124/0x190
[ 27.975516] el0_sync_compat+0x144/0x180_______________________________________________
dri-devel mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/dri-devel
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 5 lines · Rob Clark
Quoted 4 lines · Viresh Kumar
Quoted 1 line · Rob Clark
With that on top of the previous patch,
Quoted 8 lines
Quoted 8 lines
Don't you still have this ? Which fixed the lockdep in the remove path.
https://lore.kernel.org/lkml/20201022080644.2ck4okrxygmkuatn@vireshk-i7/
To make it clear you need these patches to fix the OPP stuff:
//From 5.10-rc3 (the one from the above link).
commite0df59de670b("opp: Reduce the size of critical section in_opp_table_kref_release()")
This fixes debugfs stuff while the OPP table is removed.
Quoted 3 lines
Quoted 3 lines
//Below two from linux-next
commitef43f01ac069("opp: Always add entries indev_listwithopp_table->lockheld") commit27c09484dd3d("opp: Allocate the OPP table outside ofopp_table_lock")
This fixes debugfs stuff while the OPP table is added.
Quoted 5 lines
Quoted 2 lines
This matches the diff I gave you earlier.
no, I did not have all three, only "opp: Allocate the OPP table outside of
opp_table_lock" plus the fixup. But with all three:
And looking at the lockdep you gave now, it looks like we have a problem with OPP table's internal lock (opp_table->lock) as well apart from the global opp_table_lock.
I wish there was a way for me to reproduce the lockdep :(
I know this is exhausting for both of us and I really want to be over with it as soon as possible, this really should be the last patch here, please try this along with other two. This fixes the debugfs thing while the OPPs in the OPP table are removed (they are already added without a lock around debugfs stuff).
AFAIU, there is no further debugfs stuff that happens from within the locks and so this really should be the last patch unless I missed something.
Signature · 17 lines
viresh -------------------------8<------------------------- From: Viresh Kumar <[email protected]> Date: Thu, 19 Nov 2020 11:24:32 +0530 Subject: [PATCH] opp: Reduce the size of critical section in _opp_kref_release() There is a lot of stuff here which can be done outside of the opp_table->lock, do that. This helps avoiding a circular dependency lockdeps around debugfs. Reported-by: Rob Clark <[email protected]> Signed-off-by: Viresh Kumar <[email protected]> --- drivers/opp/core.c | 94 +++++++++++++++++++++++----------------------- 1 file changed, 47 insertions(+), 47 deletions(-)
drivers/opp/core.c +47 -47
@@ -1251,9 +1251,14 @@ void _opp_free(struct dev_pm_opp *opp)kfree (opp); } -static void _opp_kref_release (struct dev_pm_opp *opp, -struct opp_table *opp_table) +static void _opp_kref_release (struct kref *kref) { + struct dev_pm_opp *opp = container_of(kref, struct dev_pm_opp, kref); +struct opp_table *opp_table = opp->opp_table; + +list_del (&opp->node); +mutex_unlock (&opp_table->lock); +/* * Notify the changes in the availability of the operable * frequency/voltage list.
@@ -1261,27 +1266,9 @@ static void _opp_kref_release(struct dev_pm_opp *opp,blocking_notifier_call_chain (&opp_table->head,OPP_EVENT_REMOVE , opp);_of_opp_free_required_opps (opp_table, opp);opp_debug_remove_one (opp); -list_del (&opp->node);kfree (opp); } -static void _opp_kref_release_unlocked (struct kref *kref) -{ - struct dev_pm_opp *opp = container_of(kref, struct dev_pm_opp, kref); -struct opp_table *opp_table = opp->opp_table; - -_opp_kref_release (opp, opp_table); -} - -static void _opp_kref_release_locked (struct kref *kref) -{ - struct dev_pm_opp *opp = container_of(kref, struct dev_pm_opp, kref); -struct opp_table *opp_table = opp->opp_table; - -_opp_kref_release (opp, opp_table); -mutex_unlock (&opp_table->lock); -} -void dev_pm_opp_get (struct dev_pm_opp *opp) {kref_get (&opp->kref);
@@ -1289,16 +1276,10 @@ void dev_pm_opp_get(struct dev_pm_opp *opp)void dev_pm_opp_put (struct dev_pm_opp *opp) { -kref_put_mutex (&opp->kref, _opp_kref_release_locked, - &opp->opp_table->lock); +kref_put_mutex (&opp->kref, _opp_kref_release, &opp->opp_table->lock); }EXPORT_SYMBOL_GPL (dev_pm_opp_put); -static void dev_pm_opp_put_unlocked (struct dev_pm_opp *opp) -{ -kref_put (&opp->kref, _opp_kref_release_unlocked); -} -/** * dev_pm_opp_remove() - Remove an OPP from OPP table * @dev: device for which we do this operation
@@ -1342,30 +1323,49 @@ void dev_pm_opp_remove(struct device *dev, unsigned long freq) }EXPORT_SYMBOL_GPL (dev_pm_opp_remove); +static struct dev_pm_opp *_opp_get_next (struct opp_table *opp_table, +bool dynamic) +{ +struct dev_pm_opp *opp =NULL , *temp; + +mutex_lock (&opp_table->lock); +list_for_each_entry (temp, &opp_table->opp_list, node) { +if (dynamic == temp->dynamic) { + opp = temp; +break ; + } + } + +mutex_unlock (&opp_table->lock); +return opp; +} +bool _opp_remove_all_static (struct opp_table *opp_table) { -struct dev_pm_opp *opp, *tmp; -bool ret = true; +struct dev_pm_opp *opp;mutex_lock (&opp_table->lock);if (!opp_table->parsed_static_opps) { - ret = false; - goto unlock; +mutex_unlock (&opp_table->lock); +return false; } -if (--opp_table->parsed_static_opps) - goto unlock; - -list_for_each_entry_safe (opp, tmp, &opp_table->opp_list, node) { -if (!opp->dynamic) -dev_pm_opp_put_unlocked (opp); +if (--opp_table->parsed_static_opps) { +mutex_unlock (&opp_table->lock); +return true; } -unlock:mutex_unlock (&opp_table->lock); -return ret; +/* +* Can't remove the OPP from under the lock, debugfs removal needs to +* happen lock less to avoid circular dependency issues. +*/ +while ((opp =_opp_get_next (opp_table, false))) +dev_pm_opp_put (opp); + +return true; }/**
@@ -1377,21 +1377,21 @@ bool _opp_remove_all_static(struct opp_table *opp_table)void dev_pm_opp_remove_all_dynamic (struct device *dev) {struct opp_table *opp_table; -struct dev_pm_opp *opp, *temp; +struct dev_pm_opp *opp;int count =0 ; opp_table =_find_opp_table (dev);if (IS_ERR (opp_table))return ; -mutex_lock (&opp_table->lock); -list_for_each_entry_safe (opp, temp, &opp_table->opp_list, node) { -if (opp->dynamic) { -dev_pm_opp_put_unlocked (opp); - count++; - } +/* +* Can't remove the OPP from under the lock, debugfs removal needs to +* happen lock less to avoid circular dependency issues. +*/ +while ((opp =_opp_get_next (opp_table, true))) { +dev_pm_opp_put (opp); + count++; } -mutex_unlock (&opp_table->lock);/* Drop the references taken by dev_pm_opp_add() */ while (count--)
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 41 lines · Viresh Kumar
Quoted 5 lines · Rob Clark
Quoted 4 lines · Viresh Kumar
Quoted 1 line · Rob Clark
With that on top of the previous patch,
Quoted 8 lines
Quoted 8 lines
Don't you still have this ? Which fixed the lockdep in the remove path.
https://lore.kernel.org/lkml/20201022080644.2ck4okrxygmkuatn@vireshk-i7/
To make it clear you need these patches to fix the OPP stuff:
//From 5.10-rc3 (the one from the above link).
commite0df59de670b("opp: Reduce the size of critical section in_opp_table_kref_release()")This fixes debugfs stuff while the OPP table is removed.
Quoted 3 lines
Quoted 3 lines
//Below two from linux-next
commitef43f01ac069("opp: Always add entries indev_listwithopp_table->lockheld") commit27c09484dd3d("opp: Allocate the OPP table outside ofopp_table_lock")This fixes debugfs stuff while the OPP table is added.
Quoted 5 lines
Quoted 2 lines
This matches the diff I gave you earlier.
no, I did not have all three, only "opp: Allocate the OPP table outside of
opp_table_lock" plus the fixup. But with all three:And looking at the lockdep you gave now, it looks like we have a problem with OPP table's internal lock (
opp_table->lock) as well apart from the globalopp_table_lock.I wish there was a way for me to reproduce the lockdep :(
I know this is exhausting for both of us and I really want to be over with it as soon as possible, this really should be the last patch here, please try this along with other two. This fixes the debugfs thing while the OPPs in the OPP table are removed (they are already added without a lock around debugfs stuff).
Quoted 3 lines
AFAIU, there is no further debugfs stuff that happens from within the locks and so this really should be the last patch unless I missed something.
Rob, were you able to test this patch ?
Signature · 1 line
viresh
Re: [PATCH v2 07/22] drm/msm: Do rpm get sooner in the submit path
Quoted 42 lines · Viresh Kumar
Quoted 41 lines · Viresh Kumar
Quoted 5 lines · Rob Clark
Quoted 4 lines · Viresh Kumar
Quoted 1 line · Rob Clark
With that on top of the previous patch,
Quoted 8 lines
Quoted 8 lines
Don't you still have this ? Which fixed the lockdep in the remove path.
https://lore.kernel.org/lkml/20201022080644.2ck4okrxygmkuatn@vireshk-i7/
To make it clear you need these patches to fix the OPP stuff:
//From 5.10-rc3 (the one from the above link).
commite0df59de670b("opp: Reduce the size of critical section in_opp_table_kref_release()")This fixes debugfs stuff while the OPP table is removed.
Quoted 3 lines
Quoted 3 lines
//Below two from linux-next
commitef43f01ac069("opp: Always add entries indev_listwithopp_table->lockheld") commit27c09484dd3d("opp: Allocate the OPP table outside ofopp_table_lock")This fixes debugfs stuff while the OPP table is added.
Quoted 5 lines
Quoted 2 lines
This matches the diff I gave you earlier.
no, I did not have all three, only "opp: Allocate the OPP table outside of
opp_table_lock" plus the fixup. But with all three:And looking at the lockdep you gave now, it looks like we have a problem with OPP table's internal lock (
opp_table->lock) as well apart from the globalopp_table_lock.I wish there was a way for me to reproduce the lockdep :(
I know this is exhausting for both of us and I really want to be over with it as soon as possible, this really should be the last patch here, please try this along with other two. This fixes the debugfs thing while the OPPs in the OPP table are removed (they are already added without a lock around debugfs stuff).
Quoted 5 lines
Quoted 3 lines
AFAIU, there is no further debugfs stuff that happens from within the locks and so this really should be the last patch unless I missed something.
Rob, were you able to test this patch ?
FWIW, this patch and everything else I had is merged into Linus's master. You can test 5.11-rc1 to see if you still see a lockdep or not.
Signature · 1 line
viresh