From da79de97d254145dcb7c08c978b1093eac15ec9c Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Mon, 22 Nov 2010 11:12:46 +0000 Subject: [PATCH 1/2] drm/i915/sdvo: Only enable HDMI encodings only if the commandset is supported As we conflated intel_sdvo->is_hdmi with both having HDMI support on the ADD along with having HDMI support on the monitor, we would attempt to use HDMI encodings even if the interface did not support those commands. Reported-by: Simon Farnsworth Signed-off-by: Chris Wilson Tested-by: Simon Farnsworth Reviewed-by: Simon Farnsworth --- drivers/gpu/drm/i915/intel_sdvo.c | 25 ++++++++++++++----------- 1 file changed, 14 insertions(+), 11 deletions(-) diff --git a/drivers/gpu/drm/i915/intel_sdvo.c b/drivers/gpu/drm/i915/intel_sdvo.c index de158b76bcd5..843182528b79 100644 --- a/drivers/gpu/drm/i915/intel_sdvo.c +++ b/drivers/gpu/drm/i915/intel_sdvo.c @@ -107,7 +107,8 @@ struct intel_sdvo { * This is set if we treat the device as HDMI, instead of DVI. */ bool is_hdmi; - bool has_audio; + bool has_hdmi_monitor; + bool has_hdmi_audio; /** * This is set if we detect output of sdvo device as LVDS and @@ -1023,7 +1024,7 @@ static void intel_sdvo_mode_set(struct drm_encoder *encoder, if (!intel_sdvo_set_target_input(intel_sdvo)) return; - if (intel_sdvo->is_hdmi && + if (intel_sdvo->has_hdmi_monitor && !intel_sdvo_set_avi_infoframe(intel_sdvo)) return; @@ -1063,7 +1064,7 @@ static void intel_sdvo_mode_set(struct drm_encoder *encoder, } if (intel_crtc->pipe == 1) sdvox |= SDVO_PIPE_B_SELECT; - if (intel_sdvo->has_audio) + if (intel_sdvo->has_hdmi_audio) sdvox |= SDVO_AUDIO_ENABLE; if (INTEL_INFO(dev)->gen >= 4) { @@ -1388,8 +1389,10 @@ intel_sdvo_hdmi_sink_detect(struct drm_connector *connector) /* DDC bus is shared, match EDID to connector type */ if (edid->input & DRM_EDID_INPUT_DIGITAL) { status = connector_status_connected; - intel_sdvo->is_hdmi = drm_detect_hdmi_monitor(edid); - intel_sdvo->has_audio = drm_detect_monitor_audio(edid); + if (intel_sdvo->is_hdmi) { + intel_sdvo->has_hdmi_monitor = drm_detect_hdmi_monitor(edid); + intel_sdvo->has_hdmi_audio = drm_detect_monitor_audio(edid); + } } connector->display_info.raw_edid = NULL; kfree(edid); @@ -1398,7 +1401,7 @@ intel_sdvo_hdmi_sink_detect(struct drm_connector *connector) if (status == connector_status_connected) { struct intel_sdvo_connector *intel_sdvo_connector = to_intel_sdvo_connector(connector); if (intel_sdvo_connector->force_audio) - intel_sdvo->has_audio = intel_sdvo_connector->force_audio > 0; + intel_sdvo->has_hdmi_audio = intel_sdvo_connector->force_audio > 0; } return status; @@ -1713,12 +1716,12 @@ intel_sdvo_set_property(struct drm_connector *connector, intel_sdvo_connector->force_audio = val; - if (val > 0 && intel_sdvo->has_audio) + if (val > 0 && intel_sdvo->has_hdmi_audio) return 0; - if (val < 0 && !intel_sdvo->has_audio) + if (val < 0 && !intel_sdvo->has_hdmi_audio) return 0; - intel_sdvo->has_audio = val > 0; + intel_sdvo->has_hdmi_audio = val > 0; goto done; } @@ -2070,6 +2073,8 @@ intel_sdvo_dvi_init(struct intel_sdvo *intel_sdvo, int device) intel_sdvo_set_colorimetry(intel_sdvo, SDVO_COLORIMETRY_RGB256); connector->connector_type = DRM_MODE_CONNECTOR_HDMIA; + + intel_sdvo_add_hdmi_properties(intel_sdvo_connector); intel_sdvo->is_hdmi = true; } intel_sdvo->base.clone_mask = ((1 << INTEL_SDVO_NON_TV_CLONE_BIT) | @@ -2077,8 +2082,6 @@ intel_sdvo_dvi_init(struct intel_sdvo *intel_sdvo, int device) intel_sdvo_connector_init(intel_sdvo_connector, intel_sdvo); - intel_sdvo_add_hdmi_properties(intel_sdvo_connector); - return true; } From bcf50e2775bbc3101932d8e4ab8c7902aa4163b4 Mon Sep 17 00:00:00 2001 From: Chris Wilson Date: Sun, 21 Nov 2010 22:07:12 +0000 Subject: [PATCH 2/2] drm/i915: Handle pagefaults in execbuffer user relocations Currently if we hit a pagefault when applying a user relocation for the execbuffer, we bail and return EFAULT to the application. Instead, we need to unwind, drop the dev->struct_mutex, copy all the relocation entries to a vmalloc array (to avoid any potential circular deadlocks when resolving the pagefault), retake the mutex and then apply the relocations. Afterwards, we need to again drop the lock and copy the vmalloc array back to userspace. v2: Incorporate feedback from Daniel Vetter. Reported-by: Daniel Vetter Signed-off-by: Chris Wilson Reviewed-by: Daniel Vetter --- drivers/gpu/drm/i915/i915_gem.c | 481 ++++++++++++++++++++------------ 1 file changed, 303 insertions(+), 178 deletions(-) diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index bc4164590054..92b097dbe4ff 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -3254,192 +3254,230 @@ i915_gem_object_set_cpu_read_domain_range(struct drm_gem_object *obj, return 0; } -/** - * Pin an object to the GTT and evaluate the relocations landing in it. - */ static int -i915_gem_execbuffer_relocate(struct drm_i915_gem_object *obj, - struct drm_file *file_priv, - struct drm_i915_gem_exec_object2 *entry) +i915_gem_execbuffer_relocate_entry(struct drm_i915_gem_object *obj, + struct drm_file *file_priv, + struct drm_i915_gem_exec_object2 *entry, + struct drm_i915_gem_relocation_entry *reloc) { struct drm_device *dev = obj->base.dev; - drm_i915_private_t *dev_priv = dev->dev_private; - struct drm_i915_gem_relocation_entry __user *user_relocs; - struct drm_gem_object *target_obj = NULL; - uint32_t target_handle = 0; - int i, ret = 0; + struct drm_gem_object *target_obj; + uint32_t target_offset; + int ret = -EINVAL; - user_relocs = (void __user *)(uintptr_t)entry->relocs_ptr; - for (i = 0; i < entry->relocation_count; i++) { - struct drm_i915_gem_relocation_entry reloc; - uint32_t target_offset; + target_obj = drm_gem_object_lookup(dev, file_priv, + reloc->target_handle); + if (target_obj == NULL) + return -ENOENT; - if (__copy_from_user_inatomic(&reloc, - user_relocs+i, - sizeof(reloc))) { - ret = -EFAULT; - break; - } - - if (reloc.target_handle != target_handle) { - drm_gem_object_unreference(target_obj); - - target_obj = drm_gem_object_lookup(dev, file_priv, - reloc.target_handle); - if (target_obj == NULL) { - ret = -ENOENT; - break; - } - - target_handle = reloc.target_handle; - } - target_offset = to_intel_bo(target_obj)->gtt_offset; + target_offset = to_intel_bo(target_obj)->gtt_offset; #if WATCH_RELOC - DRM_INFO("%s: obj %p offset %08x target %d " - "read %08x write %08x gtt %08x " - "presumed %08x delta %08x\n", - __func__, - obj, - (int) reloc.offset, - (int) reloc.target_handle, - (int) reloc.read_domains, - (int) reloc.write_domain, - (int) target_offset, - (int) reloc.presumed_offset, - reloc.delta); + DRM_INFO("%s: obj %p offset %08x target %d " + "read %08x write %08x gtt %08x " + "presumed %08x delta %08x\n", + __func__, + obj, + (int) reloc->offset, + (int) reloc->target_handle, + (int) reloc->read_domains, + (int) reloc->write_domain, + (int) target_offset, + (int) reloc->presumed_offset, + reloc->delta); #endif - /* The target buffer should have appeared before us in the - * exec_object list, so it should have a GTT space bound by now. - */ - if (target_offset == 0) { - DRM_ERROR("No GTT space found for object %d\n", - reloc.target_handle); - ret = -EINVAL; - break; - } - - /* Validate that the target is in a valid r/w GPU domain */ - if (reloc.write_domain & (reloc.write_domain - 1)) { - DRM_ERROR("reloc with multiple write domains: " - "obj %p target %d offset %d " - "read %08x write %08x", - obj, reloc.target_handle, - (int) reloc.offset, - reloc.read_domains, - reloc.write_domain); - ret = -EINVAL; - break; - } - if (reloc.write_domain & I915_GEM_DOMAIN_CPU || - reloc.read_domains & I915_GEM_DOMAIN_CPU) { - DRM_ERROR("reloc with read/write CPU domains: " - "obj %p target %d offset %d " - "read %08x write %08x", - obj, reloc.target_handle, - (int) reloc.offset, - reloc.read_domains, - reloc.write_domain); - ret = -EINVAL; - break; - } - if (reloc.write_domain && target_obj->pending_write_domain && - reloc.write_domain != target_obj->pending_write_domain) { - DRM_ERROR("Write domain conflict: " - "obj %p target %d offset %d " - "new %08x old %08x\n", - obj, reloc.target_handle, - (int) reloc.offset, - reloc.write_domain, - target_obj->pending_write_domain); - ret = -EINVAL; - break; - } - - target_obj->pending_read_domains |= reloc.read_domains; - target_obj->pending_write_domain |= reloc.write_domain; - - /* If the relocation already has the right value in it, no - * more work needs to be done. - */ - if (target_offset == reloc.presumed_offset) - continue; - - /* Check that the relocation address is valid... */ - if (reloc.offset > obj->base.size - 4) { - DRM_ERROR("Relocation beyond object bounds: " - "obj %p target %d offset %d size %d.\n", - obj, reloc.target_handle, - (int) reloc.offset, (int) obj->base.size); - ret = -EINVAL; - break; - } - if (reloc.offset & 3) { - DRM_ERROR("Relocation not 4-byte aligned: " - "obj %p target %d offset %d.\n", - obj, reloc.target_handle, - (int) reloc.offset); - ret = -EINVAL; - break; - } - - /* and points to somewhere within the target object. */ - if (reloc.delta >= target_obj->size) { - DRM_ERROR("Relocation beyond target object bounds: " - "obj %p target %d delta %d size %d.\n", - obj, reloc.target_handle, - (int) reloc.delta, (int) target_obj->size); - ret = -EINVAL; - break; - } - - reloc.delta += target_offset; - if (obj->base.write_domain == I915_GEM_DOMAIN_CPU) { - uint32_t page_offset = reloc.offset & ~PAGE_MASK; - char *vaddr; - - vaddr = kmap_atomic(obj->pages[reloc.offset >> PAGE_SHIFT]); - *(uint32_t *)(vaddr + page_offset) = reloc.delta; - kunmap_atomic(vaddr); - } else { - uint32_t __iomem *reloc_entry; - void __iomem *reloc_page; - - ret = i915_gem_object_set_to_gtt_domain(&obj->base, 1); - if (ret) - break; - - /* Map the page containing the relocation we're going to perform. */ - reloc.offset += obj->gtt_offset; - reloc_page = io_mapping_map_atomic_wc(dev_priv->mm.gtt_mapping, - reloc.offset & PAGE_MASK); - reloc_entry = (uint32_t __iomem *) - (reloc_page + (reloc.offset & ~PAGE_MASK)); - iowrite32(reloc.delta, reloc_entry); - io_mapping_unmap_atomic(reloc_page); - } - - /* and update the user's relocation entry */ - reloc.presumed_offset = target_offset; - if (__copy_to_user_inatomic(&user_relocs[i].presumed_offset, - &reloc.presumed_offset, - sizeof(reloc.presumed_offset))) { - ret = -EFAULT; - break; - } + /* The target buffer should have appeared before us in the + * exec_object list, so it should have a GTT space bound by now. + */ + if (target_offset == 0) { + DRM_ERROR("No GTT space found for object %d\n", + reloc->target_handle); + goto err; } + /* Validate that the target is in a valid r/w GPU domain */ + if (reloc->write_domain & (reloc->write_domain - 1)) { + DRM_ERROR("reloc with multiple write domains: " + "obj %p target %d offset %d " + "read %08x write %08x", + obj, reloc->target_handle, + (int) reloc->offset, + reloc->read_domains, + reloc->write_domain); + goto err; + } + if (reloc->write_domain & I915_GEM_DOMAIN_CPU || + reloc->read_domains & I915_GEM_DOMAIN_CPU) { + DRM_ERROR("reloc with read/write CPU domains: " + "obj %p target %d offset %d " + "read %08x write %08x", + obj, reloc->target_handle, + (int) reloc->offset, + reloc->read_domains, + reloc->write_domain); + goto err; + } + if (reloc->write_domain && target_obj->pending_write_domain && + reloc->write_domain != target_obj->pending_write_domain) { + DRM_ERROR("Write domain conflict: " + "obj %p target %d offset %d " + "new %08x old %08x\n", + obj, reloc->target_handle, + (int) reloc->offset, + reloc->write_domain, + target_obj->pending_write_domain); + goto err; + } + + target_obj->pending_read_domains |= reloc->read_domains; + target_obj->pending_write_domain |= reloc->write_domain; + + /* If the relocation already has the right value in it, no + * more work needs to be done. + */ + if (target_offset == reloc->presumed_offset) + goto out; + + /* Check that the relocation address is valid... */ + if (reloc->offset > obj->base.size - 4) { + DRM_ERROR("Relocation beyond object bounds: " + "obj %p target %d offset %d size %d.\n", + obj, reloc->target_handle, + (int) reloc->offset, + (int) obj->base.size); + goto err; + } + if (reloc->offset & 3) { + DRM_ERROR("Relocation not 4-byte aligned: " + "obj %p target %d offset %d.\n", + obj, reloc->target_handle, + (int) reloc->offset); + goto err; + } + + /* and points to somewhere within the target object. */ + if (reloc->delta >= target_obj->size) { + DRM_ERROR("Relocation beyond target object bounds: " + "obj %p target %d delta %d size %d.\n", + obj, reloc->target_handle, + (int) reloc->delta, + (int) target_obj->size); + goto err; + } + + reloc->delta += target_offset; + if (obj->base.write_domain == I915_GEM_DOMAIN_CPU) { + uint32_t page_offset = reloc->offset & ~PAGE_MASK; + char *vaddr; + + vaddr = kmap_atomic(obj->pages[reloc->offset >> PAGE_SHIFT]); + *(uint32_t *)(vaddr + page_offset) = reloc->delta; + kunmap_atomic(vaddr); + } else { + struct drm_i915_private *dev_priv = dev->dev_private; + uint32_t __iomem *reloc_entry; + void __iomem *reloc_page; + + ret = i915_gem_object_set_to_gtt_domain(&obj->base, 1); + if (ret) + goto err; + + /* Map the page containing the relocation we're going to perform. */ + reloc->offset += obj->gtt_offset; + reloc_page = io_mapping_map_atomic_wc(dev_priv->mm.gtt_mapping, + reloc->offset & PAGE_MASK); + reloc_entry = (uint32_t __iomem *) + (reloc_page + (reloc->offset & ~PAGE_MASK)); + iowrite32(reloc->delta, reloc_entry); + io_mapping_unmap_atomic(reloc_page); + } + + /* and update the user's relocation entry */ + reloc->presumed_offset = target_offset; + +out: + ret = 0; +err: drm_gem_object_unreference(target_obj); return ret; } static int -i915_gem_execbuffer_pin(struct drm_device *dev, - struct drm_file *file, - struct drm_gem_object **object_list, - struct drm_i915_gem_exec_object2 *exec_list, - int count) +i915_gem_execbuffer_relocate_object(struct drm_i915_gem_object *obj, + struct drm_file *file_priv, + struct drm_i915_gem_exec_object2 *entry) +{ + struct drm_i915_gem_relocation_entry __user *user_relocs; + int i, ret; + + user_relocs = (void __user *)(uintptr_t)entry->relocs_ptr; + for (i = 0; i < entry->relocation_count; i++) { + struct drm_i915_gem_relocation_entry reloc; + + if (__copy_from_user_inatomic(&reloc, + user_relocs+i, + sizeof(reloc))) + return -EFAULT; + + ret = i915_gem_execbuffer_relocate_entry(obj, file_priv, entry, &reloc); + if (ret) + return ret; + + if (__copy_to_user_inatomic(&user_relocs[i].presumed_offset, + &reloc.presumed_offset, + sizeof(reloc.presumed_offset))) + return -EFAULT; + } + + return 0; +} + +static int +i915_gem_execbuffer_relocate_object_slow(struct drm_i915_gem_object *obj, + struct drm_file *file_priv, + struct drm_i915_gem_exec_object2 *entry, + struct drm_i915_gem_relocation_entry *relocs) +{ + int i, ret; + + for (i = 0; i < entry->relocation_count; i++) { + ret = i915_gem_execbuffer_relocate_entry(obj, file_priv, entry, &relocs[i]); + if (ret) + return ret; + } + + return 0; +} + +static int +i915_gem_execbuffer_relocate(struct drm_device *dev, + struct drm_file *file, + struct drm_gem_object **object_list, + struct drm_i915_gem_exec_object2 *exec_list, + int count) +{ + int i, ret; + + for (i = 0; i < count; i++) { + struct drm_i915_gem_object *obj = to_intel_bo(object_list[i]); + obj->base.pending_read_domains = 0; + obj->base.pending_write_domain = 0; + ret = i915_gem_execbuffer_relocate_object(obj, file, + &exec_list[i]); + if (ret) + return ret; + } + + return 0; +} + +static int +i915_gem_execbuffer_reserve(struct drm_device *dev, + struct drm_file *file, + struct drm_gem_object **object_list, + struct drm_i915_gem_exec_object2 *exec_list, + int count) { struct drm_i915_private *dev_priv = dev->dev_private; int ret, i, retry; @@ -3501,6 +3539,87 @@ i915_gem_execbuffer_pin(struct drm_device *dev, return 0; } +static int +i915_gem_execbuffer_relocate_slow(struct drm_device *dev, + struct drm_file *file, + struct drm_gem_object **object_list, + struct drm_i915_gem_exec_object2 *exec_list, + int count) +{ + struct drm_i915_gem_relocation_entry *reloc; + int i, total, ret; + + for (i = 0; i < count; i++) { + struct drm_i915_gem_object *obj = to_intel_bo(object_list[i]); + obj->in_execbuffer = false; + } + + mutex_unlock(&dev->struct_mutex); + + total = 0; + for (i = 0; i < count; i++) + total += exec_list[i].relocation_count; + + reloc = drm_malloc_ab(total, sizeof(*reloc)); + if (reloc == NULL) { + mutex_lock(&dev->struct_mutex); + return -ENOMEM; + } + + total = 0; + for (i = 0; i < count; i++) { + struct drm_i915_gem_relocation_entry __user *user_relocs; + + user_relocs = (void __user *)(uintptr_t)exec_list[i].relocs_ptr; + + if (copy_from_user(reloc+total, user_relocs, + exec_list[i].relocation_count * + sizeof(*reloc))) { + ret = -EFAULT; + mutex_lock(&dev->struct_mutex); + goto err; + } + + total += exec_list[i].relocation_count; + } + + ret = i915_mutex_lock_interruptible(dev); + if (ret) { + mutex_lock(&dev->struct_mutex); + goto err; + } + + ret = i915_gem_execbuffer_reserve(dev, file, + object_list, exec_list, + count); + if (ret) + goto err; + + total = 0; + for (i = 0; i < count; i++) { + struct drm_i915_gem_object *obj = to_intel_bo(object_list[i]); + obj->base.pending_read_domains = 0; + obj->base.pending_write_domain = 0; + ret = i915_gem_execbuffer_relocate_object_slow(obj, file, + &exec_list[i], + reloc + total); + if (ret) + goto err; + + total += exec_list[i].relocation_count; + } + + /* Leave the user relocations as are, this is the painfully slow path, + * and we want to avoid the complication of dropping the lock whilst + * having buffers reserved in the aperture and so causing spurious + * ENOSPC for random operations. + */ + +err: + drm_free_large(reloc); + return ret; +} + static int i915_gem_execbuffer_move_to_gpu(struct drm_device *dev, struct drm_file *file, @@ -3781,18 +3900,24 @@ i915_gem_do_execbuffer(struct drm_device *dev, void *data, } /* Move the objects en-masse into the GTT, evicting if necessary. */ - ret = i915_gem_execbuffer_pin(dev, file, - object_list, exec_list, - args->buffer_count); + ret = i915_gem_execbuffer_reserve(dev, file, + object_list, exec_list, + args->buffer_count); if (ret) goto err; /* The objects are in their final locations, apply the relocations. */ - for (i = 0; i < args->buffer_count; i++) { - struct drm_i915_gem_object *obj = to_intel_bo(object_list[i]); - obj->base.pending_read_domains = 0; - obj->base.pending_write_domain = 0; - ret = i915_gem_execbuffer_relocate(obj, file, &exec_list[i]); + ret = i915_gem_execbuffer_relocate(dev, file, + object_list, exec_list, + args->buffer_count); + if (ret) { + if (ret == -EFAULT) { + ret = i915_gem_execbuffer_relocate_slow(dev, file, + object_list, + exec_list, + args->buffer_count); + BUG_ON(!mutex_is_locked(&dev->struct_mutex)); + } if (ret) goto err; }