diff --git a/drivers/gpu/drm/i915/i915_dma.c b/drivers/gpu/drm/i915/i915_dma.c index 2e7f03ad5ee2e9e84e9de0588ce78d8ad67f86f2..9933c26017edae26a2bab5879e8be0543735c027 100644 --- a/drivers/gpu/drm/i915/i915_dma.c +++ b/drivers/gpu/drm/i915/i915_dma.c @@ -1336,12 +1336,17 @@ static int i915_load_modeset_init(struct drm_device *dev) intel_power_domains_init_hw(dev_priv); + /* + * We enable some interrupt sources in our postinstall hooks, so mark + * interrupts as enabled _before_ actually enabling them to avoid + * special cases in our ordering checks. + */ + dev_priv->pm._irqs_disabled = false; + ret = drm_irq_install(dev, dev->pdev->irq); if (ret) goto cleanup_gem_stolen; - dev_priv->pm._irqs_disabled = false; - /* Important: The output setup functions called by modeset_init need * working irqs for e.g. gmbus and dp aux transfers. */ intel_modeset_init(dev); diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 7a830eac5ba30a7859eb65313d9ee567c5d324eb..3524306d8cfbf7b9d7e93b2bd076774eb69b36bc 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -184,6 +184,7 @@ enum hpd_pin { if ((1 << (domain)) & (mask)) struct drm_i915_private; +struct i915_mm_struct; struct i915_mmu_object; enum intel_dpll_id { @@ -1506,9 +1507,8 @@ struct drm_i915_private { struct i915_gtt gtt; /* VM representing the global address space */ struct i915_gem_mm mm; -#if defined(CONFIG_MMU_NOTIFIER) - DECLARE_HASHTABLE(mmu_notifiers, 7); -#endif + DECLARE_HASHTABLE(mm_structs, 7); + struct mutex mm_lock; /* Kernel Modesetting */ @@ -1814,8 +1814,8 @@ struct drm_i915_gem_object { unsigned workers :4; #define I915_GEM_USERPTR_MAX_WORKERS 15 - struct mm_struct *mm; - struct i915_mmu_object *mn; + struct i915_mm_struct *mm; + struct i915_mmu_object *mmu_object; struct work_struct *work; } userptr; }; diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index ba7f5c6bb50d1f7e5b886ea20e74ed58a9ae5247..ad55b06a3cb1c69754fe60947a183e0af6604e46 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -1590,10 +1590,13 @@ int i915_gem_fault(struct vm_area_struct *vma, struct vm_fault *vmf) out: switch (ret) { case -EIO: - /* If this -EIO is due to a gpu hang, give the reset code a - * chance to clean up the mess. Otherwise return the proper - * SIGBUS. */ - if (i915_terminally_wedged(&dev_priv->gpu_error)) { + /* + * We eat errors when the gpu is terminally wedged to avoid + * userspace unduly crashing (gl has no provisions for mmaps to + * fail). But any other -EIO isn't ours (e.g. swap in failure) + * and so needs to be reported. + */ + if (!i915_terminally_wedged(&dev_priv->gpu_error)) { ret = VM_FAULT_SIGBUS; break; } diff --git a/drivers/gpu/drm/i915/i915_gem_userptr.c b/drivers/gpu/drm/i915/i915_gem_userptr.c index fe69fc837d9ee25f33b3dbd15664c2c33b9ca721..d384139973790e09d2da3583861c7224b1596f44 100644 --- a/drivers/gpu/drm/i915/i915_gem_userptr.c +++ b/drivers/gpu/drm/i915/i915_gem_userptr.c @@ -32,6 +32,15 @@ #include #include +struct i915_mm_struct { + struct mm_struct *mm; + struct drm_device *dev; + struct i915_mmu_notifier *mn; + struct hlist_node node; + struct kref kref; + struct work_struct work; +}; + #if defined(CONFIG_MMU_NOTIFIER) #include @@ -41,16 +50,12 @@ struct i915_mmu_notifier { struct mmu_notifier mn; struct rb_root objects; struct list_head linear; - struct drm_device *dev; - struct mm_struct *mm; - struct work_struct work; - unsigned long count; unsigned long serial; bool has_linear; }; struct i915_mmu_object { - struct i915_mmu_notifier *mmu; + struct i915_mmu_notifier *mn; struct interval_tree_node it; struct list_head link; struct drm_i915_gem_object *obj; @@ -96,18 +101,18 @@ static void *invalidate_range__linear(struct i915_mmu_notifier *mn, unsigned long start, unsigned long end) { - struct i915_mmu_object *mmu; + struct i915_mmu_object *mo; unsigned long serial; restart: serial = mn->serial; - list_for_each_entry(mmu, &mn->linear, link) { + list_for_each_entry(mo, &mn->linear, link) { struct drm_i915_gem_object *obj; - if (mmu->it.last < start || mmu->it.start > end) + if (mo->it.last < start || mo->it.start > end) continue; - obj = mmu->obj; + obj = mo->obj; drm_gem_object_reference(&obj->base); spin_unlock(&mn->lock); @@ -160,130 +165,47 @@ static const struct mmu_notifier_ops i915_gem_userptr_notifier = { }; static struct i915_mmu_notifier * -__i915_mmu_notifier_lookup(struct drm_device *dev, struct mm_struct *mm) -{ - struct drm_i915_private *dev_priv = to_i915(dev); - struct i915_mmu_notifier *mmu; - - /* Protected by dev->struct_mutex */ - hash_for_each_possible(dev_priv->mmu_notifiers, mmu, node, (unsigned long)mm) - if (mmu->mm == mm) - return mmu; - - return NULL; -} - -static struct i915_mmu_notifier * -i915_mmu_notifier_get(struct drm_device *dev, struct mm_struct *mm) +i915_mmu_notifier_create(struct mm_struct *mm) { - struct drm_i915_private *dev_priv = to_i915(dev); - struct i915_mmu_notifier *mmu; + struct i915_mmu_notifier *mn; int ret; - lockdep_assert_held(&dev->struct_mutex); - - mmu = __i915_mmu_notifier_lookup(dev, mm); - if (mmu) - return mmu; - - mmu = kmalloc(sizeof(*mmu), GFP_KERNEL); - if (mmu == NULL) + mn = kmalloc(sizeof(*mn), GFP_KERNEL); + if (mn == NULL) return ERR_PTR(-ENOMEM); - spin_lock_init(&mmu->lock); - mmu->dev = dev; - mmu->mn.ops = &i915_gem_userptr_notifier; - mmu->mm = mm; - mmu->objects = RB_ROOT; - mmu->count = 0; - mmu->serial = 1; - INIT_LIST_HEAD(&mmu->linear); - mmu->has_linear = false; - - /* Protected by mmap_sem (write-lock) */ - ret = __mmu_notifier_register(&mmu->mn, mm); + spin_lock_init(&mn->lock); + mn->mn.ops = &i915_gem_userptr_notifier; + mn->objects = RB_ROOT; + mn->serial = 1; + INIT_LIST_HEAD(&mn->linear); + mn->has_linear = false; + + /* Protected by mmap_sem (write-lock) */ + ret = __mmu_notifier_register(&mn->mn, mm); if (ret) { - kfree(mmu); + kfree(mn); return ERR_PTR(ret); } - /* Protected by dev->struct_mutex */ - hash_add(dev_priv->mmu_notifiers, &mmu->node, (unsigned long)mm); - return mmu; + return mn; } -static void -__i915_mmu_notifier_destroy_worker(struct work_struct *work) +static void __i915_mmu_notifier_update_serial(struct i915_mmu_notifier *mn) { - struct i915_mmu_notifier *mmu = container_of(work, typeof(*mmu), work); - mmu_notifier_unregister(&mmu->mn, mmu->mm); - kfree(mmu); -} - -static void -__i915_mmu_notifier_destroy(struct i915_mmu_notifier *mmu) -{ - lockdep_assert_held(&mmu->dev->struct_mutex); - - /* Protected by dev->struct_mutex */ - hash_del(&mmu->node); - - /* Our lock ordering is: mmap_sem, mmu_notifier_scru, struct_mutex. - * We enter the function holding struct_mutex, therefore we need - * to drop our mutex prior to calling mmu_notifier_unregister in - * order to prevent lock inversion (and system-wide deadlock) - * between the mmap_sem and struct-mutex. Hence we defer the - * unregistration to a workqueue where we hold no locks. - */ - INIT_WORK(&mmu->work, __i915_mmu_notifier_destroy_worker); - schedule_work(&mmu->work); -} - -static void __i915_mmu_notifier_update_serial(struct i915_mmu_notifier *mmu) -{ - if (++mmu->serial == 0) - mmu->serial = 1; -} - -static bool i915_mmu_notifier_has_linear(struct i915_mmu_notifier *mmu) -{ - struct i915_mmu_object *mn; - - list_for_each_entry(mn, &mmu->linear, link) - if (mn->is_linear) - return true; - - return false; -} - -static void -i915_mmu_notifier_del(struct i915_mmu_notifier *mmu, - struct i915_mmu_object *mn) -{ - lockdep_assert_held(&mmu->dev->struct_mutex); - - spin_lock(&mmu->lock); - list_del(&mn->link); - if (mn->is_linear) - mmu->has_linear = i915_mmu_notifier_has_linear(mmu); - else - interval_tree_remove(&mn->it, &mmu->objects); - __i915_mmu_notifier_update_serial(mmu); - spin_unlock(&mmu->lock); - - /* Protected against _add() by dev->struct_mutex */ - if (--mmu->count == 0) - __i915_mmu_notifier_destroy(mmu); + if (++mn->serial == 0) + mn->serial = 1; } static int -i915_mmu_notifier_add(struct i915_mmu_notifier *mmu, - struct i915_mmu_object *mn) +i915_mmu_notifier_add(struct drm_device *dev, + struct i915_mmu_notifier *mn, + struct i915_mmu_object *mo) { struct interval_tree_node *it; int ret; - ret = i915_mutex_lock_interruptible(mmu->dev); + ret = i915_mutex_lock_interruptible(dev); if (ret) return ret; @@ -291,11 +213,11 @@ i915_mmu_notifier_add(struct i915_mmu_notifier *mmu, * remove the objects from the interval tree) before we do * the check for overlapping objects. */ - i915_gem_retire_requests(mmu->dev); + i915_gem_retire_requests(dev); - spin_lock(&mmu->lock); - it = interval_tree_iter_first(&mmu->objects, - mn->it.start, mn->it.last); + spin_lock(&mn->lock); + it = interval_tree_iter_first(&mn->objects, + mo->it.start, mo->it.last); if (it) { struct drm_i915_gem_object *obj; @@ -312,86 +234,122 @@ i915_mmu_notifier_add(struct i915_mmu_notifier *mmu, obj = container_of(it, struct i915_mmu_object, it)->obj; if (!obj->userptr.workers) - mmu->has_linear = mn->is_linear = true; + mn->has_linear = mo->is_linear = true; else ret = -EAGAIN; } else - interval_tree_insert(&mn->it, &mmu->objects); + interval_tree_insert(&mo->it, &mn->objects); if (ret == 0) { - list_add(&mn->link, &mmu->linear); - __i915_mmu_notifier_update_serial(mmu); + list_add(&mo->link, &mn->linear); + __i915_mmu_notifier_update_serial(mn); } - spin_unlock(&mmu->lock); - mutex_unlock(&mmu->dev->struct_mutex); + spin_unlock(&mn->lock); + mutex_unlock(&dev->struct_mutex); return ret; } +static bool i915_mmu_notifier_has_linear(struct i915_mmu_notifier *mn) +{ + struct i915_mmu_object *mo; + + list_for_each_entry(mo, &mn->linear, link) + if (mo->is_linear) + return true; + + return false; +} + +static void +i915_mmu_notifier_del(struct i915_mmu_notifier *mn, + struct i915_mmu_object *mo) +{ + spin_lock(&mn->lock); + list_del(&mo->link); + if (mo->is_linear) + mn->has_linear = i915_mmu_notifier_has_linear(mn); + else + interval_tree_remove(&mo->it, &mn->objects); + __i915_mmu_notifier_update_serial(mn); + spin_unlock(&mn->lock); +} + static void i915_gem_userptr_release__mmu_notifier(struct drm_i915_gem_object *obj) { - struct i915_mmu_object *mn; + struct i915_mmu_object *mo; - mn = obj->userptr.mn; - if (mn == NULL) + mo = obj->userptr.mmu_object; + if (mo == NULL) return; - i915_mmu_notifier_del(mn->mmu, mn); - obj->userptr.mn = NULL; + i915_mmu_notifier_del(mo->mn, mo); + kfree(mo); + + obj->userptr.mmu_object = NULL; +} + +static struct i915_mmu_notifier * +i915_mmu_notifier_find(struct i915_mm_struct *mm) +{ + if (mm->mn == NULL) { + down_write(&mm->mm->mmap_sem); + mutex_lock(&to_i915(mm->dev)->mm_lock); + if (mm->mn == NULL) + mm->mn = i915_mmu_notifier_create(mm->mm); + mutex_unlock(&to_i915(mm->dev)->mm_lock); + up_write(&mm->mm->mmap_sem); + } + return mm->mn; } static int i915_gem_userptr_init__mmu_notifier(struct drm_i915_gem_object *obj, unsigned flags) { - struct i915_mmu_notifier *mmu; - struct i915_mmu_object *mn; + struct i915_mmu_notifier *mn; + struct i915_mmu_object *mo; int ret; if (flags & I915_USERPTR_UNSYNCHRONIZED) return capable(CAP_SYS_ADMIN) ? 0 : -EPERM; - down_write(&obj->userptr.mm->mmap_sem); - ret = i915_mutex_lock_interruptible(obj->base.dev); - if (ret == 0) { - mmu = i915_mmu_notifier_get(obj->base.dev, obj->userptr.mm); - if (!IS_ERR(mmu)) - mmu->count++; /* preemptive add to act as a refcount */ - else - ret = PTR_ERR(mmu); - mutex_unlock(&obj->base.dev->struct_mutex); - } - up_write(&obj->userptr.mm->mmap_sem); - if (ret) - return ret; + if (WARN_ON(obj->userptr.mm == NULL)) + return -EINVAL; - mn = kzalloc(sizeof(*mn), GFP_KERNEL); - if (mn == NULL) { - ret = -ENOMEM; - goto destroy_mmu; - } + mn = i915_mmu_notifier_find(obj->userptr.mm); + if (IS_ERR(mn)) + return PTR_ERR(mn); - mn->mmu = mmu; - mn->it.start = obj->userptr.ptr; - mn->it.last = mn->it.start + obj->base.size - 1; - mn->obj = obj; + mo = kzalloc(sizeof(*mo), GFP_KERNEL); + if (mo == NULL) + return -ENOMEM; - ret = i915_mmu_notifier_add(mmu, mn); - if (ret) - goto free_mn; + mo->mn = mn; + mo->it.start = obj->userptr.ptr; + mo->it.last = mo->it.start + obj->base.size - 1; + mo->obj = obj; - obj->userptr.mn = mn; + ret = i915_mmu_notifier_add(obj->base.dev, mn, mo); + if (ret) { + kfree(mo); + return ret; + } + + obj->userptr.mmu_object = mo; return 0; +} + +static void +i915_mmu_notifier_free(struct i915_mmu_notifier *mn, + struct mm_struct *mm) +{ + if (mn == NULL) + return; -free_mn: + mmu_notifier_unregister(&mn->mn, mm); kfree(mn); -destroy_mmu: - mutex_lock(&obj->base.dev->struct_mutex); - if (--mmu->count == 0) - __i915_mmu_notifier_destroy(mmu); - mutex_unlock(&obj->base.dev->struct_mutex); - return ret; } #else @@ -413,15 +371,114 @@ i915_gem_userptr_init__mmu_notifier(struct drm_i915_gem_object *obj, return 0; } + +static void +i915_mmu_notifier_free(struct i915_mmu_notifier *mn, + struct mm_struct *mm) +{ +} + #endif +static struct i915_mm_struct * +__i915_mm_struct_find(struct drm_i915_private *dev_priv, struct mm_struct *real) +{ + struct i915_mm_struct *mm; + + /* Protected by dev_priv->mm_lock */ + hash_for_each_possible(dev_priv->mm_structs, mm, node, (unsigned long)real) + if (mm->mm == real) + return mm; + + return NULL; +} + +static int +i915_gem_userptr_init__mm_struct(struct drm_i915_gem_object *obj) +{ + struct drm_i915_private *dev_priv = to_i915(obj->base.dev); + struct i915_mm_struct *mm; + int ret = 0; + + /* During release of the GEM object we hold the struct_mutex. This + * precludes us from calling mmput() at that time as that may be + * the last reference and so call exit_mmap(). exit_mmap() will + * attempt to reap the vma, and if we were holding a GTT mmap + * would then call drm_gem_vm_close() and attempt to reacquire + * the struct mutex. So in order to avoid that recursion, we have + * to defer releasing the mm reference until after we drop the + * struct_mutex, i.e. we need to schedule a worker to do the clean + * up. + */ + mutex_lock(&dev_priv->mm_lock); + mm = __i915_mm_struct_find(dev_priv, current->mm); + if (mm == NULL) { + mm = kmalloc(sizeof(*mm), GFP_KERNEL); + if (mm == NULL) { + ret = -ENOMEM; + goto out; + } + + kref_init(&mm->kref); + mm->dev = obj->base.dev; + + mm->mm = current->mm; + atomic_inc(¤t->mm->mm_count); + + mm->mn = NULL; + + /* Protected by dev_priv->mm_lock */ + hash_add(dev_priv->mm_structs, + &mm->node, (unsigned long)mm->mm); + } else + kref_get(&mm->kref); + + obj->userptr.mm = mm; +out: + mutex_unlock(&dev_priv->mm_lock); + return ret; +} + +static void +__i915_mm_struct_free__worker(struct work_struct *work) +{ + struct i915_mm_struct *mm = container_of(work, typeof(*mm), work); + i915_mmu_notifier_free(mm->mn, mm->mm); + mmdrop(mm->mm); + kfree(mm); +} + +static void +__i915_mm_struct_free(struct kref *kref) +{ + struct i915_mm_struct *mm = container_of(kref, typeof(*mm), kref); + + /* Protected by dev_priv->mm_lock */ + hash_del(&mm->node); + mutex_unlock(&to_i915(mm->dev)->mm_lock); + + INIT_WORK(&mm->work, __i915_mm_struct_free__worker); + schedule_work(&mm->work); +} + +static void +i915_gem_userptr_release__mm_struct(struct drm_i915_gem_object *obj) +{ + if (obj->userptr.mm == NULL) + return; + + kref_put_mutex(&obj->userptr.mm->kref, + __i915_mm_struct_free, + &to_i915(obj->base.dev)->mm_lock); + obj->userptr.mm = NULL; +} + struct get_pages_work { struct work_struct work; struct drm_i915_gem_object *obj; struct task_struct *task; }; - #if IS_ENABLED(CONFIG_SWIOTLB) #define swiotlb_active() swiotlb_nr_tbl() #else @@ -479,7 +536,7 @@ __i915_gem_userptr_get_pages_worker(struct work_struct *_work) if (pvec == NULL) pvec = drm_malloc_ab(num_pages, sizeof(struct page *)); if (pvec != NULL) { - struct mm_struct *mm = obj->userptr.mm; + struct mm_struct *mm = obj->userptr.mm->mm; down_read(&mm->mmap_sem); while (pinned < num_pages) { @@ -545,7 +602,7 @@ i915_gem_userptr_get_pages(struct drm_i915_gem_object *obj) pvec = NULL; pinned = 0; - if (obj->userptr.mm == current->mm) { + if (obj->userptr.mm->mm == current->mm) { pvec = kmalloc(num_pages*sizeof(struct page *), GFP_TEMPORARY | __GFP_NOWARN | __GFP_NORETRY); if (pvec == NULL) { @@ -651,17 +708,13 @@ static void i915_gem_userptr_release(struct drm_i915_gem_object *obj) { i915_gem_userptr_release__mmu_notifier(obj); - - if (obj->userptr.mm) { - mmput(obj->userptr.mm); - obj->userptr.mm = NULL; - } + i915_gem_userptr_release__mm_struct(obj); } static int i915_gem_userptr_dmabuf_export(struct drm_i915_gem_object *obj) { - if (obj->userptr.mn) + if (obj->userptr.mmu_object) return 0; return i915_gem_userptr_init__mmu_notifier(obj, 0); @@ -736,7 +789,6 @@ i915_gem_userptr_ioctl(struct drm_device *dev, void *data, struct drm_file *file return -ENODEV; } - /* Allocate the new object */ obj = i915_gem_object_alloc(dev); if (obj == NULL) return -ENOMEM; @@ -754,8 +806,8 @@ i915_gem_userptr_ioctl(struct drm_device *dev, void *data, struct drm_file *file * at binding. This means that we need to hook into the mmu_notifier * in order to detect if the mmu is destroyed. */ - ret = -ENOMEM; - if ((obj->userptr.mm = get_task_mm(current))) + ret = i915_gem_userptr_init__mm_struct(obj); + if (ret == 0) ret = i915_gem_userptr_init__mmu_notifier(obj, args->flags); if (ret == 0) ret = drm_gem_handle_create(file, &obj->base, &handle); @@ -772,9 +824,8 @@ i915_gem_userptr_ioctl(struct drm_device *dev, void *data, struct drm_file *file int i915_gem_init_userptr(struct drm_device *dev) { -#if defined(CONFIG_MMU_NOTIFIER) struct drm_i915_private *dev_priv = to_i915(dev); - hash_init(dev_priv->mmu_notifiers); -#endif + mutex_init(&dev_priv->mm_lock); + hash_init(dev_priv->mm_structs); return 0; } diff --git a/drivers/gpu/drm/i915/i915_reg.h b/drivers/gpu/drm/i915/i915_reg.h index e4d7607da2c482ee6e31576e746de5ca71b6e28a..f29b44c86a2f7354c29dd0ce3ca3a5cfc7ae723c 100644 --- a/drivers/gpu/drm/i915/i915_reg.h +++ b/drivers/gpu/drm/i915/i915_reg.h @@ -334,16 +334,20 @@ #define GFX_OP_DESTBUFFER_INFO ((0x3<<29)|(0x1d<<24)|(0x8e<<16)|1) #define GFX_OP_DRAWRECT_INFO ((0x3<<29)|(0x1d<<24)|(0x80<<16)|(0x3)) #define GFX_OP_DRAWRECT_INFO_I965 ((0x7900<<16)|0x2) -#define SRC_COPY_BLT_CMD ((2<<29)|(0x43<<22)|4) + +#define COLOR_BLT_CMD (2<<29 | 0x40<<22 | (5-2)) +#define SRC_COPY_BLT_CMD ((2<<29)|(0x43<<22)|4) #define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) #define XY_MONO_SRC_COPY_IMM_BLT ((2<<29)|(0x71<<22)|5) -#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) -#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) +#define BLT_WRITE_A (2<<20) +#define BLT_WRITE_RGB (1<<20) +#define BLT_WRITE_RGBA (BLT_WRITE_RGB | BLT_WRITE_A) #define BLT_DEPTH_8 (0<<24) #define BLT_DEPTH_16_565 (1<<24) #define BLT_DEPTH_16_1555 (2<<24) #define BLT_DEPTH_32 (3<<24) -#define BLT_ROP_GXCOPY (0xcc<<16) +#define BLT_ROP_SRC_COPY (0xcc<<16) +#define BLT_ROP_COLOR_COPY (0xf0<<16) #define XY_SRC_COPY_BLT_SRC_TILED (1<<15) /* 965+ only */ #define XY_SRC_COPY_BLT_DST_TILED (1<<11) /* 965+ only */ #define CMD_OP_DISPLAYBUFFER_INFO ((0x0<<29)|(0x14<<23)|2) diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c index 16371a444426d190d73b4baddd38cd8abeae66b2..2d068edd1adc41bc8d71cd919a55361de093b19d 100644 --- a/drivers/gpu/drm/i915/intel_ringbuffer.c +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c @@ -1363,54 +1363,66 @@ i965_dispatch_execbuffer(struct intel_engine_cs *ring, /* Just userspace ABI convention to limit the wa batch bo to a resonable size */ #define I830_BATCH_LIMIT (256*1024) +#define I830_TLB_ENTRIES (2) +#define I830_WA_SIZE max(I830_TLB_ENTRIES*4096, I830_BATCH_LIMIT) static int i830_dispatch_execbuffer(struct intel_engine_cs *ring, u64 offset, u32 len, unsigned flags) { + u32 cs_offset = ring->scratch.gtt_offset; int ret; - if (flags & I915_DISPATCH_PINNED) { - ret = intel_ring_begin(ring, 4); - if (ret) - return ret; + ret = intel_ring_begin(ring, 6); + if (ret) + return ret; - intel_ring_emit(ring, MI_BATCH_BUFFER); - intel_ring_emit(ring, offset | (flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE)); - intel_ring_emit(ring, offset + len - 8); - intel_ring_emit(ring, MI_NOOP); - intel_ring_advance(ring); - } else { - u32 cs_offset = ring->scratch.gtt_offset; + /* Evict the invalid PTE TLBs */ + intel_ring_emit(ring, COLOR_BLT_CMD | BLT_WRITE_RGBA); + intel_ring_emit(ring, BLT_DEPTH_32 | BLT_ROP_COLOR_COPY | 4096); + intel_ring_emit(ring, I830_TLB_ENTRIES << 16 | 4); /* load each page */ + intel_ring_emit(ring, cs_offset); + intel_ring_emit(ring, 0xdeadbeef); + intel_ring_emit(ring, MI_NOOP); + intel_ring_advance(ring); + if ((flags & I915_DISPATCH_PINNED) == 0) { if (len > I830_BATCH_LIMIT) return -ENOSPC; - ret = intel_ring_begin(ring, 9+3); + ret = intel_ring_begin(ring, 6 + 2); if (ret) return ret; - /* Blit the batch (which has now all relocs applied) to the stable batch - * scratch bo area (so that the CS never stumbles over its tlb - * invalidation bug) ... */ - intel_ring_emit(ring, XY_SRC_COPY_BLT_CMD | - XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - intel_ring_emit(ring, BLT_DEPTH_32 | BLT_ROP_GXCOPY | 4096); - intel_ring_emit(ring, 0); - intel_ring_emit(ring, (DIV_ROUND_UP(len, 4096) << 16) | 1024); + + /* Blit the batch (which has now all relocs applied) to the + * stable batch scratch bo area (so that the CS never + * stumbles over its tlb invalidation bug) ... + */ + intel_ring_emit(ring, SRC_COPY_BLT_CMD | BLT_WRITE_RGBA); + intel_ring_emit(ring, BLT_DEPTH_32 | BLT_ROP_SRC_COPY | 4096); + intel_ring_emit(ring, DIV_ROUND_UP(len, 4096) << 16 | 1024); intel_ring_emit(ring, cs_offset); - intel_ring_emit(ring, 0); intel_ring_emit(ring, 4096); intel_ring_emit(ring, offset); + intel_ring_emit(ring, MI_FLUSH); + intel_ring_emit(ring, MI_NOOP); + intel_ring_advance(ring); /* ... and execute it. */ - intel_ring_emit(ring, MI_BATCH_BUFFER); - intel_ring_emit(ring, cs_offset | (flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE)); - intel_ring_emit(ring, cs_offset + len - 8); - intel_ring_advance(ring); + offset = cs_offset; } + ret = intel_ring_begin(ring, 4); + if (ret) + return ret; + + intel_ring_emit(ring, MI_BATCH_BUFFER); + intel_ring_emit(ring, offset | (flags & I915_DISPATCH_SECURE ? 0 : MI_BATCH_NON_SECURE)); + intel_ring_emit(ring, offset + len - 8); + intel_ring_emit(ring, MI_NOOP); + intel_ring_advance(ring); + return 0; } @@ -2200,7 +2212,7 @@ int intel_init_render_ring_buffer(struct drm_device *dev) /* Workaround batchbuffer to combat CS tlb bug. */ if (HAS_BROKEN_CS_TLB(dev)) { - obj = i915_gem_alloc_object(dev, I830_BATCH_LIMIT); + obj = i915_gem_alloc_object(dev, I830_WA_SIZE); if (obj == NULL) { DRM_ERROR("Failed to allocate batch bo\n"); return -ENOMEM; diff --git a/drivers/gpu/drm/i915/intel_tv.c b/drivers/gpu/drm/i915/intel_tv.c index c69d3ce1b3d69a6f7e9c943ed5be3dc68decdde7..c14341ca3ef9386a0be1b432206f00c15ddae09b 100644 --- a/drivers/gpu/drm/i915/intel_tv.c +++ b/drivers/gpu/drm/i915/intel_tv.c @@ -854,6 +854,10 @@ intel_enable_tv(struct intel_encoder *encoder) struct drm_device *dev = encoder->base.dev; struct drm_i915_private *dev_priv = dev->dev_private; + /* Prevents vblank waits from timing out in intel_tv_detect_type() */ + intel_wait_for_vblank(encoder->base.dev, + to_intel_crtc(encoder->base.crtc)->pipe); + I915_WRITE(TV_CTL, I915_READ(TV_CTL) | TV_ENC_ENABLE); }