diff options
Diffstat (limited to 'drivers/gpu/drm')
45 files changed, 680 insertions, 426 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index e2ae9ba147ba..5cb33ac99f70 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -73,10 +73,10 @@ amdgpu_ctx_to_drm_sched_prio(int32_t ctx_prio) return DRM_SCHED_PRIORITY_NORMAL; case AMDGPU_CTX_PRIORITY_VERY_LOW: - return DRM_SCHED_PRIORITY_MIN; + return DRM_SCHED_PRIORITY_LOW; case AMDGPU_CTX_PRIORITY_LOW: - return DRM_SCHED_PRIORITY_MIN; + return DRM_SCHED_PRIORITY_LOW; case AMDGPU_CTX_PRIORITY_NORMAL: return DRM_SCHED_PRIORITY_NORMAL; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index 62bb7fc7448a..71a5cf37b472 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -325,7 +325,7 @@ void amdgpu_job_stop_all_jobs_on_sched(struct drm_gpu_scheduler *sched) int i; /* Signal all jobs not yet scheduled */ - for (i = sched->num_rqs - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) { + for (i = DRM_SCHED_PRIORITY_KERNEL; i < sched->num_rqs; i++) { struct drm_sched_rq *rq = sched->sched_rq[i]; spin_lock(&rq->lock); list_for_each_entry(s_entity, &rq->entities, list) { diff --git a/drivers/gpu/drm/armada/armada_crtc.c b/drivers/gpu/drm/armada/armada_crtc.c index 52d2c942d3d2..c78687c755a8 100644 --- a/drivers/gpu/drm/armada/armada_crtc.c +++ b/drivers/gpu/drm/armada/armada_crtc.c @@ -7,8 +7,9 @@ #include <linux/clk.h> #include <linux/component.h> #include <linux/module.h> -#include <linux/of_device.h> +#include <linux/of.h> #include <linux/platform_device.h> +#include <linux/property.h> #include <drm/drm_atomic.h> #include <drm/drm_atomic_helper.h> @@ -1012,26 +1013,17 @@ armada_lcd_bind(struct device *dev, struct device *master, void *data) int irq = platform_get_irq(pdev, 0); const struct armada_variant *variant; struct device_node *port = NULL; + struct device_node *np, *parent = dev->of_node; if (irq < 0) return irq; - if (!dev->of_node) { - const struct platform_device_id *id; - id = platform_get_device_id(pdev); - if (!id) - return -ENXIO; - - variant = (const struct armada_variant *)id->driver_data; - } else { - const struct of_device_id *match; - struct device_node *np, *parent = dev->of_node; - - match = of_match_device(dev->driver->of_match_table, dev); - if (!match) - return -ENXIO; + variant = device_get_match_data(dev); + if (!variant) + return -ENXIO; + if (parent) { np = of_get_child_by_name(parent, "ports"); if (np) parent = np; @@ -1041,8 +1033,6 @@ armada_lcd_bind(struct device *dev, struct device *master, void *data) dev_err(dev, "no port node found in %pOF\n", parent); return -ENXIO; } - - variant = match->data; } return armada_drm_crtc_create(drm, dev, res, irq, variant, port); diff --git a/drivers/gpu/drm/aspeed/aspeed_gfx_drv.c b/drivers/gpu/drm/aspeed/aspeed_gfx_drv.c index 78122b35a0cb..a7a6b70220eb 100644 --- a/drivers/gpu/drm/aspeed/aspeed_gfx_drv.c +++ b/drivers/gpu/drm/aspeed/aspeed_gfx_drv.c @@ -6,10 +6,10 @@ #include <linux/irq.h> #include <linux/mfd/syscon.h> #include <linux/module.h> -#include <linux/of.h> -#include <linux/of_device.h> +#include <linux/mod_devicetable.h> #include <linux/of_reserved_mem.h> #include <linux/platform_device.h> +#include <linux/property.h> #include <linux/regmap.h> #include <linux/reset.h> @@ -143,7 +143,6 @@ static int aspeed_gfx_load(struct drm_device *drm) struct aspeed_gfx *priv = to_aspeed_gfx(drm); struct device_node *np = pdev->dev.of_node; const struct aspeed_gfx_config *config; - const struct of_device_id *match; struct resource *res; int ret; @@ -152,10 +151,9 @@ static int aspeed_gfx_load(struct drm_device *drm) if (IS_ERR(priv->base)) return PTR_ERR(priv->base); - match = of_match_device(aspeed_gfx_match, &pdev->dev); - if (!match) + config = device_get_match_data(&pdev->dev); + if (!config) return -EINVAL; - config = match->data; priv->dac_reg = config->dac_reg; priv->int_clr_reg = config->int_clear_reg; diff --git a/drivers/gpu/drm/drm_atomic_state_helper.c b/drivers/gpu/drm/drm_atomic_state_helper.c index 784e63d70a42..54975de44a0e 100644 --- a/drivers/gpu/drm/drm_atomic_state_helper.c +++ b/drivers/gpu/drm/drm_atomic_state_helper.c @@ -275,6 +275,20 @@ void __drm_atomic_helper_plane_state_reset(struct drm_plane_state *plane_state, plane_state->normalized_zpos = val; } } + + if (plane->hotspot_x_property) { + if (!drm_object_property_get_default_value(&plane->base, + plane->hotspot_x_property, + &val)) + plane_state->hotspot_x = val; + } + + if (plane->hotspot_y_property) { + if (!drm_object_property_get_default_value(&plane->base, + plane->hotspot_y_property, + &val)) + plane_state->hotspot_y = val; + } } EXPORT_SYMBOL(__drm_atomic_helper_plane_state_reset); diff --git a/drivers/gpu/drm/drm_atomic_uapi.c b/drivers/gpu/drm/drm_atomic_uapi.c index 98d3b10c08ae..aee4a65d4959 100644 --- a/drivers/gpu/drm/drm_atomic_uapi.c +++ b/drivers/gpu/drm/drm_atomic_uapi.c @@ -593,6 +593,22 @@ static int drm_atomic_plane_set_property(struct drm_plane *plane, } else if (plane->funcs->atomic_set_property) { return plane->funcs->atomic_set_property(plane, state, property, val); + } else if (property == plane->hotspot_x_property) { + if (plane->type != DRM_PLANE_TYPE_CURSOR) { + drm_dbg_atomic(plane->dev, + "[PLANE:%d:%s] is not a cursor plane: 0x%llx\n", + plane->base.id, plane->name, val); + return -EINVAL; + } + state->hotspot_x = val; + } else if (property == plane->hotspot_y_property) { + if (plane->type != DRM_PLANE_TYPE_CURSOR) { + drm_dbg_atomic(plane->dev, + "[PLANE:%d:%s] is not a cursor plane: 0x%llx\n", + plane->base.id, plane->name, val); + return -EINVAL; + } + state->hotspot_y = val; } else { drm_dbg_atomic(plane->dev, "[PLANE:%d:%s] unknown property [PROP:%d:%s]\n", @@ -653,6 +669,10 @@ drm_atomic_plane_get_property(struct drm_plane *plane, *val = state->scaling_filter; } else if (plane->funcs->atomic_get_property) { return plane->funcs->atomic_get_property(plane, state, property, val); + } else if (property == plane->hotspot_x_property) { + *val = state->hotspot_x; + } else if (property == plane->hotspot_y_property) { + *val = state->hotspot_y; } else { drm_dbg_atomic(dev, "[PLANE:%d:%s] unknown property [PROP:%d:%s]\n", @@ -1006,13 +1026,28 @@ out: return ret; } +static int drm_atomic_check_prop_changes(int ret, uint64_t old_val, uint64_t prop_value, + struct drm_property *prop) +{ + if (ret != 0 || old_val != prop_value) { + drm_dbg_atomic(prop->dev, + "[PROP:%d:%s] No prop can be changed during async flip\n", + prop->base.id, prop->name); + return -EINVAL; + } + + return 0; +} + int drm_atomic_set_property(struct drm_atomic_state *state, struct drm_file *file_priv, struct drm_mode_object *obj, struct drm_property *prop, - uint64_t prop_value) + u64 prop_value, + bool async_flip) { struct drm_mode_object *ref; + u64 old_val; int ret; if (!drm_property_change_valid_get(prop, prop_value, &ref)) @@ -1029,6 +1064,13 @@ int drm_atomic_set_property(struct drm_atomic_state *state, break; } + if (async_flip) { + ret = drm_atomic_connector_get_property(connector, connector_state, + prop, &old_val); + ret = drm_atomic_check_prop_changes(ret, old_val, prop_value, prop); + break; + } + ret = drm_atomic_connector_set_property(connector, connector_state, file_priv, prop, prop_value); @@ -1044,6 +1086,13 @@ int drm_atomic_set_property(struct drm_atomic_state *state, break; } + if (async_flip) { + ret = drm_atomic_crtc_get_property(crtc, crtc_state, + prop, &old_val); + ret = drm_atomic_check_prop_changes(ret, old_val, prop_value, prop); + break; + } + ret = drm_atomic_crtc_set_property(crtc, crtc_state, prop, prop_value); break; @@ -1051,6 +1100,7 @@ int drm_atomic_set_property(struct drm_atomic_state *state, case DRM_MODE_OBJECT_PLANE: { struct drm_plane *plane = obj_to_plane(obj); struct drm_plane_state *plane_state; + struct drm_mode_config *config = &plane->dev->mode_config; plane_state = drm_atomic_get_plane_state(state, plane); if (IS_ERR(plane_state)) { @@ -1058,6 +1108,21 @@ int drm_atomic_set_property(struct drm_atomic_state *state, break; } + if (async_flip && prop != config->prop_fb_id) { + ret = drm_atomic_plane_get_property(plane, plane_state, + prop, &old_val); + ret = drm_atomic_check_prop_changes(ret, old_val, prop_value, prop); + break; + } + + if (async_flip && plane_state->plane->type != DRM_PLANE_TYPE_PRIMARY) { + drm_dbg_atomic(prop->dev, + "[OBJECT:%d] Only primary planes can be changed during async flip\n", + obj->id); + ret = -EINVAL; + break; + } + ret = drm_atomic_plane_set_property(plane, plane_state, file_priv, prop, prop_value); @@ -1323,6 +1388,18 @@ static void complete_signaling(struct drm_device *dev, kfree(fence_state); } +static void +set_async_flip(struct drm_atomic_state *state) +{ + struct drm_crtc *crtc; + struct drm_crtc_state *crtc_state; + int i; + + for_each_new_crtc_in_state(state, crtc, crtc_state, i) { + crtc_state->async_flip = true; + } +} + int drm_mode_atomic_ioctl(struct drm_device *dev, void *data, struct drm_file *file_priv) { @@ -1337,6 +1414,7 @@ int drm_mode_atomic_ioctl(struct drm_device *dev, struct drm_out_fence_state *fence_state; int ret = 0; unsigned int i, j, num_fences; + bool async_flip = false; /* disallow for drivers not supporting atomic: */ if (!drm_core_check_feature(dev, DRIVER_ATOMIC)) @@ -1363,9 +1441,13 @@ int drm_mode_atomic_ioctl(struct drm_device *dev, } if (arg->flags & DRM_MODE_PAGE_FLIP_ASYNC) { - drm_dbg_atomic(dev, - "commit failed: invalid flag DRM_MODE_PAGE_FLIP_ASYNC\n"); - return -EINVAL; + if (!dev->mode_config.async_page_flip) { + drm_dbg_atomic(dev, + "commit failed: DRM_MODE_PAGE_FLIP_ASYNC not supported\n"); + return -EINVAL; + } + + async_flip = true; } /* can't test and expect an event at the same time. */ @@ -1450,8 +1532,8 @@ retry: goto out; } - ret = drm_atomic_set_property(state, file_priv, - obj, prop, prop_value); + ret = drm_atomic_set_property(state, file_priv, obj, + prop, prop_value, async_flip); if (ret) { drm_mode_object_put(obj); goto out; @@ -1468,6 +1550,9 @@ retry: if (ret) goto out; + if (arg->flags & DRM_MODE_PAGE_FLIP_ASYNC) + set_async_flip(state); + if (arg->flags & DRM_MODE_ATOMIC_TEST_ONLY) { ret = drm_atomic_check_only(state); } else if (arg->flags & DRM_MODE_ATOMIC_NONBLOCK) { diff --git a/drivers/gpu/drm/drm_crtc_internal.h b/drivers/gpu/drm/drm_crtc_internal.h index 6b646e0783be..a514d5207e41 100644 --- a/drivers/gpu/drm/drm_crtc_internal.h +++ b/drivers/gpu/drm/drm_crtc_internal.h @@ -253,7 +253,7 @@ int drm_atomic_set_property(struct drm_atomic_state *state, struct drm_file *file_priv, struct drm_mode_object *obj, struct drm_property *prop, - uint64_t prop_value); + u64 prop_value, bool async_flip); int drm_atomic_get_property(struct drm_mode_object *obj, struct drm_property *property, uint64_t *val); diff --git a/drivers/gpu/drm/drm_damage_helper.c b/drivers/gpu/drm/drm_damage_helper.c index d8b2955e88fd..afb02aae707b 100644 --- a/drivers/gpu/drm/drm_damage_helper.c +++ b/drivers/gpu/drm/drm_damage_helper.c @@ -241,7 +241,8 @@ drm_atomic_helper_damage_iter_init(struct drm_atomic_helper_damage_iter *iter, iter->plane_src.x2 = (src.x2 >> 16) + !!(src.x2 & 0xFFFF); iter->plane_src.y2 = (src.y2 >> 16) + !!(src.y2 & 0xFFFF); - if (!iter->clips || !drm_rect_equals(&state->src, &old_state->src)) { + if (!iter->clips || state->ignore_damage_clips || + !drm_rect_equals(&state->src, &old_state->src)) { iter->clips = NULL; iter->num_clips = 0; iter->full_update = true; diff --git a/drivers/gpu/drm/drm_ioctl.c b/drivers/gpu/drm/drm_ioctl.c index 44fda68c28ae..ea303386e688 100644 --- a/drivers/gpu/drm/drm_ioctl.c +++ b/drivers/gpu/drm/drm_ioctl.c @@ -301,6 +301,10 @@ static int drm_getcap(struct drm_device *dev, void *data, struct drm_file *file_ case DRM_CAP_CRTC_IN_VBLANK_EVENT: req->value = 1; break; + case DRM_CAP_ATOMIC_ASYNC_PAGE_FLIP: + req->value = drm_core_check_feature(dev, DRIVER_ATOMIC) && + dev->mode_config.async_page_flip; + break; default: return -EINVAL; } @@ -361,6 +365,15 @@ drm_setclientcap(struct drm_device *dev, void *data, struct drm_file *file_priv) return -EINVAL; file_priv->writeback_connectors = req->value; break; + case DRM_CLIENT_CAP_CURSOR_PLANE_HOTSPOT: + if (!drm_core_check_feature(dev, DRIVER_CURSOR_HOTSPOT)) + return -EOPNOTSUPP; + if (!file_priv->atomic) + return -EINVAL; + if (req->value > 1) + return -EINVAL; + file_priv->supports_virtualized_cursor_plane = req->value; + break; default: return -EINVAL; } diff --git a/drivers/gpu/drm/drm_mode_object.c b/drivers/gpu/drm/drm_mode_object.c index ac0d2ce3f870..0e8355063eee 100644 --- a/drivers/gpu/drm/drm_mode_object.c +++ b/drivers/gpu/drm/drm_mode_object.c @@ -538,7 +538,7 @@ retry: obj_to_connector(obj), prop_value); } else { - ret = drm_atomic_set_property(state, file_priv, obj, prop, prop_value); + ret = drm_atomic_set_property(state, file_priv, obj, prop, prop_value, false); if (ret) goto out; ret = drm_atomic_commit(state); diff --git a/drivers/gpu/drm/drm_plane.c b/drivers/gpu/drm/drm_plane.c index 24e7998d1731..9e8e4c60983d 100644 --- a/drivers/gpu/drm/drm_plane.c +++ b/drivers/gpu/drm/drm_plane.c @@ -230,6 +230,103 @@ static int create_in_format_blob(struct drm_device *dev, struct drm_plane *plane return 0; } +/** + * DOC: hotspot properties + * + * HOTSPOT_X: property to set mouse hotspot x offset. + * HOTSPOT_Y: property to set mouse hotspot y offset. + * + * When the plane is being used as a cursor image to display a mouse pointer, + * the "hotspot" is the offset within the cursor image where mouse events + * are expected to go. + * + * Positive values move the hotspot from the top-left corner of the cursor + * plane towards the right and bottom. + * + * Most display drivers do not need this information because the + * hotspot is not actually connected to anything visible on screen. + * However, this is necessary for display drivers like the para-virtualized + * drivers (eg qxl, vbox, virtio, vmwgfx), that are attached to a user console + * with a mouse pointer. Since these consoles are often being remoted over a + * network, they would otherwise have to wait to display the pointer movement to + * the user until a full network round-trip has occurred. New mouse events have + * to be sent from the user's console, over the network to the virtual input + * devices, forwarded to the desktop for processing, and then the cursor plane's + * position can be updated and sent back to the user's console over the network. + * Instead, with the hotspot information, the console can anticipate the new + * location, and draw the mouse cursor there before the confirmation comes in. + * To do that correctly, the user's console must be able predict how the + * desktop will process mouse events, which normally requires the desktop's + * mouse topology information, ie where each CRTC sits in the mouse coordinate + * space. This is typically sent to the para-virtualized drivers using some + * driver-specific method, and the driver then forwards it to the console by + * way of the virtual display device or hypervisor. + * + * The assumption is generally made that there is only one cursor plane being + * used this way at a time, and that the desktop is feeding all mouse devices + * into the same global pointer. Para-virtualized drivers that require this + * should only be exposing a single cursor plane, or find some other way + * to coordinate with a userspace desktop that supports multiple pointers. + * If the hotspot properties are set, the cursor plane is therefore assumed to be + * used only for displaying a mouse cursor image, and the position of the combined + * cursor plane + offset can therefore be used for coordinating with input from a + * mouse device. + * + * The cursor will then be drawn either at the location of the plane in the CRTC + * console, or as a free-floating cursor plane on the user's console + * corresponding to their desktop mouse position. + * + * DRM clients which would like to work correctly on drivers which expose + * hotspot properties should advertise DRM_CLIENT_CAP_CURSOR_PLANE_HOTSPOT. + * Setting this property on drivers which do not special case + * cursor planes will return EOPNOTSUPP, which can be used by userspace to + * gauge requirements of the hardware/drivers they're running on. Advertising + * DRM_CLIENT_CAP_CURSOR_PLANE_HOTSPOT implies that the userspace client will be + * correctly setting the hotspot properties. + */ + +/** + * drm_plane_create_hotspot_properties - creates the mouse hotspot + * properties and attaches them to the given cursor plane + * + * @plane: drm cursor plane + * + * This function enables the mouse hotspot property on a given + * cursor plane. Look at the documentation for hotspot properties + * to get a better understanding for what they're used for. + * + * RETURNS: + * Zero for success or -errno + */ +static int drm_plane_create_hotspot_properties(struct drm_plane *plane) +{ + struct drm_property *prop_x; + struct drm_property *prop_y; + + drm_WARN_ON(plane->dev, + !drm_core_check_feature(plane->dev, + DRIVER_CURSOR_HOTSPOT)); + + prop_x = drm_property_create_signed_range(plane->dev, 0, "HOTSPOT_X", + INT_MIN, INT_MAX); + if (IS_ERR(prop_x)) + return PTR_ERR(prop_x); + + prop_y = drm_property_create_signed_range(plane->dev, 0, "HOTSPOT_Y", + INT_MIN, INT_MAX); + if (IS_ERR(prop_y)) { + drm_property_destroy(plane->dev, prop_x); + return PTR_ERR(prop_y); + } + + drm_object_attach_property(&plane->base, prop_x, 0); + drm_object_attach_property(&plane->base, prop_y, 0); + plane->hotspot_x_property = prop_x; + plane->hotspot_y_property = prop_y; + + return 0; +} + __printf(9, 0) static int __drm_universal_plane_init(struct drm_device *dev, struct drm_plane *plane, @@ -348,6 +445,10 @@ static int __drm_universal_plane_init(struct drm_device *dev, drm_object_attach_property(&plane->base, config->prop_src_w, 0); drm_object_attach_property(&plane->base, config->prop_src_h, 0); } + if (drm_core_check_feature(dev, DRIVER_CURSOR_HOTSPOT) && + type == DRM_PLANE_TYPE_CURSOR) { + drm_plane_create_hotspot_properties(plane); + } if (format_modifier_count) create_in_format_blob(dev, plane); @@ -678,6 +779,19 @@ int drm_mode_getplane_res(struct drm_device *dev, void *data, !file_priv->universal_planes) continue; + /* + * If we're running on a virtualized driver then, + * unless userspace advertizes support for the + * virtualized cursor plane, disable cursor planes + * because they'll be broken due to missing cursor + * hotspot info. + */ + if (plane->type == DRM_PLANE_TYPE_CURSOR && + drm_core_check_feature(dev, DRIVER_CURSOR_HOTSPOT) && + file_priv->atomic && + !file_priv->supports_virtualized_cursor_plane) + continue; + if (drm_lease_held(file_priv, plane->base.id)) { if (count < plane_resp->count_planes && put_user(plane->base.id, plane_ptr + count)) @@ -1052,8 +1166,10 @@ static int drm_mode_cursor_universal(struct drm_crtc *crtc, return PTR_ERR(fb); } - fb->hot_x = req->hot_x; - fb->hot_y = req->hot_y; + if (plane->hotspot_x_property && plane->state) + plane->state->hotspot_x = req->hot_x; + if (plane->hotspot_y_property && plane->state) + plane->state->hotspot_y = req->hot_y; } else { fb = NULL; } @@ -1442,6 +1558,36 @@ out: * Drivers implementing damage can use drm_atomic_helper_damage_iter_init() and * drm_atomic_helper_damage_iter_next() helper iterator function to get damage * rectangles clipped to &drm_plane_state.src. + * + * Note that there are two types of damage handling: frame damage and buffer + * damage, the type of damage handling implemented depends on a driver's upload + * target. Drivers implementing a per-plane or per-CRTC upload target need to + * handle frame damage, while drivers implementing a per-buffer upload target + * need to handle buffer damage. + * + * The existing damage helpers only support the frame damage type, there is no + * buffer age support or similar damage accumulation algorithm implemented yet. + * + * Only drivers handling frame damage can use the mentioned damage helpers to + * iterate over the damaged regions. Drivers that handle buffer damage, must set + * &drm_plane_state.ignore_damage_clips for drm_atomic_helper_damage_iter_init() + * to know that damage clips should be ignored and return &drm_plane_state.src + * as the damage rectangle, to force a full plane update. + * + * Drivers with a per-buffer upload target could compare the &drm_plane_state.fb + * of the old and new plane states to determine if the framebuffer attached to a + * plane has changed or not since the last plane update. If &drm_plane_state.fb + * has changed, then &drm_plane_state.ignore_damage_clips must be set to true. + * + * That is because drivers with a per-plane upload target, expect the backing + * storage buffer to not change for a given plane. If the upload buffer changes + * between page flips, the new upload buffer has to be updated as a whole. This + * can be improved in the future if support for frame damage is added to the DRM + * damage helpers, similarly to how user-space already handle this case as it is + * explained in the following documents: + * + * https://registry.khronos.org/EGL/extensions/KHR/EGL_KHR_swap_buffers_with_damage.txt + * https://emersion.fr/blog/2019/intro-to-damage-tracking/ */ /** diff --git a/drivers/gpu/drm/exynos/exynos_drm_gsc.c b/drivers/gpu/drm/exynos/exynos_drm_gsc.c index 34cdabc30b4f..35771fb4e85d 100644 --- a/drivers/gpu/drm/exynos/exynos_drm_gsc.c +++ b/drivers/gpu/drm/exynos/exynos_drm_gsc.c @@ -11,9 +11,10 @@ #include <linux/component.h> #include <linux/kernel.h> #include <linux/mfd/syscon.h> -#include <linux/of_device.h> +#include <linux/mod_devicetable.h> #include <linux/platform_device.h> #include <linux/pm_runtime.h> +#include <linux/property.h> #include <linux/regmap.h> #include <drm/drm_fourcc.h> @@ -103,7 +104,7 @@ struct gsc_context { unsigned int num_formats; void __iomem *regs; - const char **clk_names; + const char *const *clk_names; struct clk *clocks[GSC_MAX_CLOCKS]; int num_clocks; struct gsc_scaler sc; @@ -1217,7 +1218,7 @@ static const unsigned int gsc_tiled_formats[] = { static int gsc_probe(struct platform_device *pdev) { struct device *dev = &pdev->dev; - struct gsc_driverdata *driver_data; + const struct gsc_driverdata *driver_data; struct exynos_drm_ipp_formats *formats; struct gsc_context *ctx; int num_formats, ret, i, j; @@ -1226,7 +1227,7 @@ static int gsc_probe(struct platform_device *pdev) if (!ctx) return -ENOMEM; - driver_data = (struct gsc_driverdata *)of_device_get_match_data(dev); + driver_data = device_get_match_data(dev); ctx->dev = dev; ctx->num_clocks = driver_data->num_clocks; ctx->clk_names = driver_data->clk_names; diff --git a/drivers/gpu/drm/imagination/pvr_free_list.c b/drivers/gpu/drm/imagination/pvr_free_list.c index c61fd417edcb..5e51bc980751 100644 --- a/drivers/gpu/drm/imagination/pvr_free_list.c +++ b/drivers/gpu/drm/imagination/pvr_free_list.c @@ -255,7 +255,7 @@ pvr_free_list_insert_pages_locked(struct pvr_free_list *free_list, if (!num_pages) break; - }; + } /* clang-format on */ /* Make sure our free_list update is flushed. */ diff --git a/drivers/gpu/drm/imagination/pvr_queue.c b/drivers/gpu/drm/imagination/pvr_queue.c index d65c3fbedf5a..5ed9c98fb599 100644 --- a/drivers/gpu/drm/imagination/pvr_queue.c +++ b/drivers/gpu/drm/imagination/pvr_queue.c @@ -1292,7 +1292,7 @@ struct pvr_queue *pvr_queue_create(struct pvr_context *ctx, goto err_release_ufo; err = drm_sched_entity_init(&queue->entity, - DRM_SCHED_PRIORITY_MIN, + DRM_SCHED_PRIORITY_KERNEL, &sched, 1, &ctx->faulty); if (err) goto err_sched_fini; diff --git a/drivers/gpu/drm/imagination/pvr_rogue_fwif_sf.h b/drivers/gpu/drm/imagination/pvr_rogue_fwif_sf.h index 571954182f33..56e11009e123 100644 --- a/drivers/gpu/drm/imagination/pvr_rogue_fwif_sf.h +++ b/drivers/gpu/drm/imagination/pvr_rogue_fwif_sf.h @@ -497,7 +497,7 @@ static const struct rogue_km_stid_fmt stid_fmts[] = { { ROGUE_FW_LOG_CREATESFID(213, ROGUE_FW_GROUP_MAIN, 1), "Safety Watchdog threshold period set to 0x%x clock cycles" }, { ROGUE_FW_LOG_CREATESFID(214, ROGUE_FW_GROUP_MAIN, 0), - "MTS Safety Event trigged by the safety watchdog." }, + "MTS Safety Event triggered by the safety watchdog." }, { ROGUE_FW_LOG_CREATESFID(215, ROGUE_FW_GROUP_MAIN, 3), "DM%d USC tasks range limit 0 - %d, stride %d" }, { ROGUE_FW_LOG_CREATESFID(216, ROGUE_FW_GROUP_MAIN, 1), @@ -1114,7 +1114,7 @@ static const struct rogue_km_stid_fmt stid_fmts[] = { { ROGUE_FW_LOG_CREATESFID(39, ROGUE_FW_GROUP_SPM, 2), "3DMemFree matches freelist 0x%08x (FL type = %u)" }, { ROGUE_FW_LOG_CREATESFID(40, ROGUE_FW_GROUP_SPM, 0), - "Raise the 3DMemFreeDedected flag" }, + "Raise the 3DMemFreeDetected flag" }, { ROGUE_FW_LOG_CREATESFID(41, ROGUE_FW_GROUP_SPM, 1), "Wait for pending grow on Freelist 0x%08x" }, { ROGUE_FW_LOG_CREATESFID(42, ROGUE_FW_GROUP_SPM, 1), diff --git a/drivers/gpu/drm/imagination/pvr_vm.c b/drivers/gpu/drm/imagination/pvr_vm.c index 3ad1366294b9..e0d74d9a6190 100644 --- a/drivers/gpu/drm/imagination/pvr_vm.c +++ b/drivers/gpu/drm/imagination/pvr_vm.c @@ -64,6 +64,12 @@ struct pvr_vm_context { struct drm_gem_object dummy_gem; }; +static inline +struct pvr_vm_context *to_pvr_vm_context(struct drm_gpuvm *gpuvm) +{ + return container_of(gpuvm, struct pvr_vm_context, gpuvm_mgr); +} + struct pvr_vm_context *pvr_vm_context_get(struct pvr_vm_context *vm_ctx) { if (vm_ctx) @@ -224,6 +230,7 @@ pvr_vm_bind_op_map_init(struct pvr_vm_bind_op *bind_op, struct pvr_gem_object *pvr_obj, u64 offset, u64 device_addr, u64 size) { + struct drm_gem_object *obj = gem_from_pvr_gem(pvr_obj); const bool is_user = vm_ctx == vm_ctx->pvr_dev->kernel_vm_ctx; const u64 pvr_obj_size = pvr_gem_object_size(pvr_obj); struct sg_table *sgt; @@ -238,17 +245,18 @@ pvr_vm_bind_op_map_init(struct pvr_vm_bind_op *bind_op, return -EINVAL; } - if (!pvr_device_addr_and_size_are_valid(device_addr, size) || + if (!pvr_device_addr_and_size_are_valid(vm_ctx, device_addr, size) || offset & ~PAGE_MASK || size & ~PAGE_MASK || offset >= pvr_obj_size || offset_plus_size > pvr_obj_size) return -EINVAL; bind_op->type = PVR_VM_BIND_TYPE_MAP; - bind_op->gpuvm_bo = drm_gpuvm_bo_create(&vm_ctx->gpuvm_mgr, - gem_from_pvr_gem(pvr_obj)); - if (!bind_op->gpuvm_bo) - return -ENOMEM; + dma_resv_lock(obj->resv, NULL); + bind_op->gpuvm_bo = drm_gpuvm_bo_obtain(&vm_ctx->gpuvm_mgr, obj); + dma_resv_unlock(obj->resv); + if (IS_ERR(bind_op->gpuvm_bo)) + return PTR_ERR(bind_op->gpuvm_bo); bind_op->new_va = kzalloc(sizeof(*bind_op->new_va), GFP_KERNEL); bind_op->prev_va = kzalloc(sizeof(*bind_op->prev_va), GFP_KERNEL); @@ -293,7 +301,7 @@ pvr_vm_bind_op_unmap_init(struct pvr_vm_bind_op *bind_op, { int err; - if (!pvr_device_addr_and_size_are_valid(device_addr, size)) + if (!pvr_device_addr_and_size_are_valid(vm_ctx, device_addr, size)) return -EINVAL; bind_op->type = PVR_VM_BIND_TYPE_UNMAP; @@ -503,6 +511,7 @@ pvr_device_addr_is_valid(u64 device_addr) /** * pvr_device_addr_and_size_are_valid() - Tests whether a device-virtual * address and associated size are both valid. + * @vm_ctx: Target VM context. * @device_addr: Virtual device address to test. * @size: Size of the range based at @device_addr to test. * @@ -521,16 +530,18 @@ pvr_device_addr_is_valid(u64 device_addr) * * %false otherwise. */ bool -pvr_device_addr_and_size_are_valid(u64 device_addr, u64 size) +pvr_device_addr_and_size_are_valid(struct pvr_vm_context *vm_ctx, + u64 device_addr, u64 size) { return pvr_device_addr_is_valid(device_addr) && + drm_gpuvm_range_valid(&vm_ctx->gpuvm_mgr, device_addr, size) && size != 0 && (size & ~PVR_DEVICE_PAGE_MASK) == 0 && (device_addr + size <= PVR_PAGE_TABLE_ADDR_SPACE_SIZE); } void pvr_gpuvm_free(struct drm_gpuvm *gpuvm) { - + kfree(to_pvr_vm_context(gpuvm)); } static const struct drm_gpuvm_ops pvr_vm_gpuva_ops = { @@ -650,12 +661,11 @@ pvr_vm_context_release(struct kref *ref_count) WARN_ON(pvr_vm_unmap(vm_ctx, vm_ctx->gpuvm_mgr.mm_start, vm_ctx->gpuvm_mgr.mm_range)); - drm_gpuvm_put(&vm_ctx->gpuvm_mgr); pvr_mmu_context_destroy(vm_ctx->mmu_ctx); drm_gem_private_object_fini(&vm_ctx->dummy_gem); mutex_destroy(&vm_ctx->lock); - kfree(vm_ctx); + drm_gpuvm_put(&vm_ctx->gpuvm_mgr); } /** diff --git a/drivers/gpu/drm/imagination/pvr_vm.h b/drivers/gpu/drm/imagination/pvr_vm.h index cf8b97553dc8..f2a6463f2b05 100644 --- a/drivers/gpu/drm/imagination/pvr_vm.h +++ b/drivers/gpu/drm/imagination/pvr_vm.h @@ -29,7 +29,8 @@ struct drm_exec; /* Functions defined in pvr_vm.c */ bool pvr_device_addr_is_valid(u64 device_addr); -bool pvr_device_addr_and_size_are_valid(u64 device_addr, u64 size); +bool pvr_device_addr_and_size_are_valid(struct pvr_vm_context *vm_ctx, + u64 device_addr, u64 size); struct pvr_vm_context *pvr_vm_create_context(struct pvr_device *pvr_dev, bool is_userspace_context); diff --git a/drivers/gpu/drm/imx/ipuv3/imx-ldb.c b/drivers/gpu/drm/imx/ipuv3/imx-ldb.c index 989eca32d325..53840ab054c7 100644 --- a/drivers/gpu/drm/imx/ipuv3/imx-ldb.c +++ b/drivers/gpu/drm/imx/ipuv3/imx-ldb.c @@ -12,8 +12,10 @@ #include <linux/mfd/syscon.h> #include <linux/mfd/syscon/imx6q-iomuxc-gpr.h> #include <linux/module.h> -#include <linux/of_device.h> +#include <linux/of.h> #include <linux/of_graph.h> +#include <linux/platform_device.h> +#include <linux/property.h> #include <linux/regmap.h> #include <linux/videodev2.h> @@ -617,7 +619,6 @@ static int imx_ldb_probe(struct platform_device *pdev) { struct device *dev = &pdev->dev; struct device_node *np = dev->of_node; - const struct of_device_id *of_id = of_match_device(imx_ldb_dt_ids, dev); struct device_node *child; struct imx_ldb *imx_ldb; int dual; @@ -638,9 +639,7 @@ static int imx_ldb_probe(struct platform_device *pdev) regmap_write(imx_ldb->regmap, IOMUXC_GPR2, 0); imx_ldb->dev = dev; - - if (of_id) - imx_ldb->lvds_mux = of_id->data; + imx_ldb->lvds_mux = device_get_match_data(dev); dual = of_property_read_bool(np, "fsl,dual-channel"); if (dual) diff --git a/drivers/gpu/drm/msm/msm_gpu.h b/drivers/gpu/drm/msm/msm_gpu.h index 4252e3839fbc..2bfcb222e353 100644 --- a/drivers/gpu/drm/msm/msm_gpu.h +++ b/drivers/gpu/drm/msm/msm_gpu.h @@ -347,7 +347,7 @@ struct msm_gpu_perfcntr { * DRM_SCHED_PRIORITY_KERNEL priority level is treated specially in some * cases, so we don't use it (no need for kernel generated jobs). */ -#define NR_SCHED_PRIORITIES (1 + DRM_SCHED_PRIORITY_HIGH - DRM_SCHED_PRIORITY_MIN) +#define NR_SCHED_PRIORITIES (1 + DRM_SCHED_PRIORITY_LOW - DRM_SCHED_PRIORITY_HIGH) /** * struct msm_file_private - per-drm_file context diff --git a/drivers/gpu/drm/mxsfb/mxsfb_drv.c b/drivers/gpu/drm/mxsfb/mxsfb_drv.c index 625c1bfc4173..b483ef48216a 100644 --- a/drivers/gpu/drm/mxsfb/mxsfb_drv.c +++ b/drivers/gpu/drm/mxsfb/mxsfb_drv.c @@ -11,9 +11,10 @@ #include <linux/clk.h> #include <linux/dma-mapping.h> #include <linux/io.h> +#include <linux/mod_devicetable.h> #include <linux/module.h> -#include <linux/of_device.h> #include <linux/platform_device.h> +#include <linux/property.h> #include <linux/pm_runtime.h> #include <drm/drm_atomic_helper.h> @@ -346,18 +347,13 @@ MODULE_DEVICE_TABLE(of, mxsfb_dt_ids); static int mxsfb_probe(struct platform_device *pdev) { struct drm_device *drm; - const struct of_device_id *of_id = - of_match_device(mxsfb_dt_ids, &pdev->dev); int ret; - if (!pdev->dev.of_node) - return -ENODEV; - drm = drm_dev_alloc(&mxsfb_driver, &pdev->dev); if (IS_ERR(drm)) return PTR_ERR(drm); - ret = mxsfb_load(drm, of_id->data); + ret = mxsfb_load(drm, device_get_match_data(&pdev->dev)); if (ret) goto err_free; diff --git a/drivers/gpu/drm/nouveau/nouveau_abi16.c b/drivers/gpu/drm/nouveau/nouveau_abi16.c index 2edd7bb13fae..a04156ca8390 100644 --- a/drivers/gpu/drm/nouveau/nouveau_abi16.c +++ b/drivers/gpu/drm/nouveau/nouveau_abi16.c @@ -127,21 +127,14 @@ nouveau_abi16_chan_fini(struct nouveau_abi16 *abi16, { struct nouveau_abi16_ntfy *ntfy, *temp; - /* When a client exits without waiting for it's queued up jobs to - * finish it might happen that we fault the channel. This is due to - * drm_file_free() calling drm_gem_release() before the postclose() - * callback. Hence, we can't tear down this scheduler entity before - * uvmm mappings are unmapped. Currently, we can't detect this case. - * - * However, this should be rare and harmless, since the channel isn't - * needed anymore. - */ - nouveau_sched_entity_fini(&chan->sched_entity); + /* Cancel all jobs from the entity's queue. */ + drm_sched_entity_fini(&chan->sched.entity); - /* wait for all activity to stop before cleaning up */ if (chan->chan) nouveau_channel_idle(chan->chan); + nouveau_sched_fini(&chan->sched); + /* cleanup notifier state */ list_for_each_entry_safe(ntfy, temp, &chan->notifiers, head) { nouveau_abi16_ntfy_fini(chan, ntfy); @@ -344,8 +337,8 @@ nouveau_abi16_ioctl_channel_alloc(ABI16_IOCTL_ARGS) if (ret) goto done; - ret = nouveau_sched_entity_init(&chan->sched_entity, &drm->sched, - drm->sched_wq); + ret = nouveau_sched_init(&chan->sched, drm, drm->sched_wq, + chan->chan->dma.ib_max); if (ret) goto done; diff --git a/drivers/gpu/drm/nouveau/nouveau_abi16.h b/drivers/gpu/drm/nouveau/nouveau_abi16.h index 9f538486c10e..1f5e243c0c75 100644 --- a/drivers/gpu/drm/nouveau/nouveau_abi16.h +++ b/drivers/gpu/drm/nouveau/nouveau_abi16.h @@ -26,7 +26,7 @@ struct nouveau_abi16_chan { struct nouveau_bo *ntfy; struct nouveau_vma *ntfy_vma; struct nvkm_mm heap; - struct nouveau_sched_entity sched_entity; + struct nouveau_sched sched; }; struct nouveau_abi16 { diff --git a/drivers/gpu/drm/nouveau/nouveau_bo.c b/drivers/gpu/drm/nouveau/nouveau_bo.c index 7afad86da64b..b7dda486a7ea 100644 --- a/drivers/gpu/drm/nouveau/nouveau_bo.c +++ b/drivers/gpu/drm/nouveau/nouveau_bo.c @@ -1061,17 +1061,18 @@ nouveau_bo_move(struct ttm_buffer_object *bo, bool evict, { struct nouveau_drm *drm = nouveau_bdev(bo->bdev); struct nouveau_bo *nvbo = nouveau_bo(bo); + struct drm_gem_object *obj = &bo->base; struct ttm_resource *old_reg = bo->resource; struct nouveau_drm_tile *new_tile = NULL; int ret = 0; - if (new_reg->mem_type == TTM_PL_TT) { ret = nouveau_ttm_tt_bind(bo->bdev, bo->ttm, new_reg); if (ret) return ret; } + drm_gpuvm_bo_gem_evict(obj, evict); nouveau_bo_move_ntfy(bo, new_reg); ret = ttm_bo_wait_ctx(bo, ctx); if (ret) @@ -1136,6 +1137,7 @@ out: out_ntfy: if (ret) { nouveau_bo_move_ntfy(bo, bo->resource); + drm_gpuvm_bo_gem_evict(obj, !evict); } return ret; } diff --git a/drivers/gpu/drm/nouveau/nouveau_drm.c b/drivers/gpu/drm/nouveau/nouveau_drm.c index f603eaef1560..6f6c31a9937b 100644 --- a/drivers/gpu/drm/nouveau/nouveau_drm.c +++ b/drivers/gpu/drm/nouveau/nouveau_drm.c @@ -201,9 +201,9 @@ nouveau_cli_fini(struct nouveau_cli *cli) WARN_ON(!list_empty(&cli->worker)); usif_client_fini(cli); + nouveau_sched_fini(&cli->sched); if (uvmm) nouveau_uvmm_fini(uvmm); - nouveau_sched_entity_fini(&cli->sched_entity); nouveau_vmm_fini(&cli->svm); nouveau_vmm_fini(&cli->vmm); nvif_mmu_dtor(&cli->mmu); @@ -310,8 +310,17 @@ nouveau_cli_init(struct nouveau_drm *drm, const char *sname, cli->mem = &mems[ret]; - ret = nouveau_sched_entity_init(&cli->sched_entity, &drm->sched, - drm->sched_wq); + /* Don't pass in the (shared) sched_wq in order to let + * nouveau_sched_init() create a dedicated one for VM_BIND jobs. + * + * This is required to ensure that for VM_BIND jobs free_job() work and + * run_job() work can always run concurrently and hence, free_job() work + * can never stall run_job() work. For EXEC jobs we don't have this + * requirement, since EXEC job's free_job() does not require to take any + * locks which indirectly or directly are held for allocations + * elsewhere. + */ + ret = nouveau_sched_init(&cli->sched, drm, NULL, 1); if (ret) goto done; @@ -582,13 +591,16 @@ nouveau_drm_device_init(struct drm_device *dev) nvif_parent_ctor(&nouveau_parent, &drm->parent); drm->master.base.object.parent = &drm->parent; - ret = nouveau_sched_init(drm); - if (ret) + drm->sched_wq = alloc_workqueue("nouveau_sched_wq_shared", 0, + WQ_MAX_ACTIVE); + if (!drm->sched_wq) { + ret = -ENOMEM; goto fail_alloc; + } ret = nouveau_cli_init(drm, "DRM-master", &drm->master); if (ret) - goto fail_sched; + goto fail_wq; ret = nouveau_cli_init(drm, "DRM", &drm->client); if (ret) @@ -658,8 +670,8 @@ fail_ttm: nouveau_cli_fini(&drm->client); fail_master: nouveau_cli_fini(&drm->master); -fail_sched: - nouveau_sched_fini(drm); +fail_wq: + destroy_workqueue(drm->sched_wq); fail_alloc: nvif_parent_dtor(&drm->parent); kfree(drm); @@ -711,10 +723,9 @@ nouveau_drm_device_fini(struct drm_device *dev) } mutex_unlock(&drm->clients_lock); - nouveau_sched_fini(drm); - nouveau_cli_fini(&drm->client); nouveau_cli_fini(&drm->master); + destroy_workqueue(drm->sched_wq); nvif_parent_dtor(&drm->parent); mutex_destroy(&drm->clients_lock); kfree(drm); diff --git a/drivers/gpu/drm/nouveau/nouveau_drv.h b/drivers/gpu/drm/nouveau/nouveau_drv.h index 7f7051df84a6..8a6d94c8b163 100644 --- a/drivers/gpu/drm/nouveau/nouveau_drv.h +++ b/drivers/gpu/drm/nouveau/nouveau_drv.h @@ -98,7 +98,7 @@ struct nouveau_cli { bool disabled; } uvmm; - struct nouveau_sched_entity sched_entity; + struct nouveau_sched sched; const struct nvif_mclass *mem; @@ -258,6 +258,9 @@ struct nouveau_drm { u64 context_base; } *runl; + /* Workqueue used for channel schedulers. */ + struct workqueue_struct *sched_wq; + /* context for accelerated drm-internal operations */ struct nouveau_channel *cechan; struct nouveau_channel *channel; @@ -298,10 +301,6 @@ struct nouveau_drm { struct mutex lock; bool component_registered; } audio; - - struct drm_gpu_scheduler sched; - struct workqueue_struct *sched_wq; - }; static inline struct nouveau_drm * diff --git a/drivers/gpu/drm/nouveau/nouveau_exec.c b/drivers/gpu/drm/nouveau/nouveau_exec.c index 9a5ef574744b..bc5d71b79ab2 100644 --- a/drivers/gpu/drm/nouveau/nouveau_exec.c +++ b/drivers/gpu/drm/nouveau/nouveau_exec.c @@ -1,7 +1,5 @@ // SPDX-License-Identifier: MIT -#include <drm/drm_exec.h> - #include "nouveau_drv.h" #include "nouveau_gem.h" #include "nouveau_mem.h" @@ -86,14 +84,12 @@ */ static int -nouveau_exec_job_submit(struct nouveau_job *job) +nouveau_exec_job_submit(struct nouveau_job *job, + struct drm_gpuvm_exec *vme) { struct nouveau_exec_job *exec_job = to_nouveau_exec_job(job); struct nouveau_cli *cli = job->cli; struct nouveau_uvmm *uvmm = nouveau_cli_uvmm(cli); - struct drm_exec *exec = &job->exec; - struct drm_gem_object *obj; - unsigned long index; int ret; /* Create a new fence, but do not emit yet. */ @@ -102,52 +98,29 @@ nouveau_exec_job_submit(struct nouveau_job *job) return ret; nouveau_uvmm_lock(uvmm); - drm_exec_init(exec, DRM_EXEC_INTERRUPTIBLE_WAIT | - DRM_EXEC_IGNORE_DUPLICATES); - drm_exec_until_all_locked(exec) { - struct drm_gpuva *va; - - drm_gpuvm_for_each_va(va, &uvmm->base) { - if (unlikely(va == &uvmm->base.kernel_alloc_node)) - continue; - - ret = drm_exec_prepare_obj(exec, va->gem.obj, 1); - drm_exec_retry_on_contention(exec); - if (ret) - goto err_uvmm_unlock; - } + ret = drm_gpuvm_exec_lock(vme); + if (ret) { + nouveau_uvmm_unlock(uvmm); + return ret; } nouveau_uvmm_unlock(uvmm); - drm_exec_for_each_locked_object(exec, index, obj) { - struct nouveau_bo *nvbo = nouveau_gem_object(obj); - - ret = nouveau_bo_validate(nvbo, true, false); - if (ret) - goto err_exec_fini; + ret = drm_gpuvm_exec_validate(vme); + if (ret) { + drm_gpuvm_exec_unlock(vme); + return ret; } return 0; - -err_uvmm_unlock: - nouveau_uvmm_unlock(uvmm); -err_exec_fini: - drm_exec_fini(exec); - return ret; - } static void -nouveau_exec_job_armed_submit(struct nouveau_job *job) +nouveau_exec_job_armed_submit(struct nouveau_job *job, + struct drm_gpuvm_exec *vme) { - struct drm_exec *exec = &job->exec; - struct drm_gem_object *obj; - unsigned long index; - - drm_exec_for_each_locked_object(exec, index, obj) - dma_resv_add_fence(obj->resv, job->done_fence, job->resv_usage); - - drm_exec_fini(exec); + drm_gpuvm_exec_resv_add_fence(vme, job->done_fence, + job->resv_usage, job->resv_usage); + drm_gpuvm_exec_unlock(vme); } static struct dma_fence * @@ -192,6 +165,7 @@ nouveau_exec_job_free(struct nouveau_job *job) { struct nouveau_exec_job *exec_job = to_nouveau_exec_job(job); + nouveau_job_done(job); nouveau_job_free(job); kfree(exec_job->fence); @@ -211,8 +185,6 @@ nouveau_exec_job_timeout(struct nouveau_job *job) NV_PRINTK(warn, job->cli, "job timeout, channel %d killed!\n", chan->chid); - nouveau_sched_entity_fini(job->entity); - return DRM_GPU_SCHED_STAT_NOMINAL; } @@ -259,10 +231,12 @@ nouveau_exec_job_init(struct nouveau_exec_job **pjob, } } + args.file_priv = __args->file_priv; job->chan = __args->chan; - args.sched_entity = __args->sched_entity; - args.file_priv = __args->file_priv; + args.sched = __args->sched; + /* Plus one to account for the HW fence. */ + args.credits = job->push.count + 1; args.in_sync.count = __args->in_sync.count; args.in_sync.s = __args->in_sync.s; @@ -415,7 +389,7 @@ nouveau_exec_ioctl_exec(struct drm_device *dev, if (ret) goto out; - args.sched_entity = &chan16->sched_entity; + args.sched = &chan16->sched; args.file_priv = file_priv; args.chan = chan; diff --git a/drivers/gpu/drm/nouveau/nouveau_exec.h b/drivers/gpu/drm/nouveau/nouveau_exec.h index 5488d337bcc0..9b3b151facfd 100644 --- a/drivers/gpu/drm/nouveau/nouveau_exec.h +++ b/drivers/gpu/drm/nouveau/nouveau_exec.h @@ -3,16 +3,12 @@ #ifndef __NOUVEAU_EXEC_H__ #define __NOUVEAU_EXEC_H__ -#include <drm/drm_exec.h> - #include "nouveau_drv.h" #include "nouveau_sched.h" struct nouveau_exec_job_args { struct drm_file *file_priv; - struct nouveau_sched_entity *sched_entity; - - struct drm_exec exec; + struct nouveau_sched *sched; struct nouveau_channel *chan; struct { diff --git a/drivers/gpu/drm/nouveau/nouveau_sched.c b/drivers/gpu/drm/nouveau/nouveau_sched.c index e03fd2bc8a11..dd98f6910f9c 100644 --- a/drivers/gpu/drm/nouveau/nouveau_sched.c +++ b/drivers/gpu/drm/nouveau/nouveau_sched.c @@ -12,30 +12,28 @@ #include "nouveau_abi16.h" #include "nouveau_sched.h" -/* FIXME - * - * We want to make sure that jobs currently executing can't be deferred by - * other jobs competing for the hardware. Otherwise we might end up with job - * timeouts just because of too many clients submitting too many jobs. We don't - * want jobs to time out because of system load, but because of the job being - * too bulky. - * - * For now allow for up to 16 concurrent jobs in flight until we know how many - * rings the hardware can process in parallel. - */ -#define NOUVEAU_SCHED_HW_SUBMISSIONS 16 #define NOUVEAU_SCHED_JOB_TIMEOUT_MS 10000 +/* Starts at 0, since the DRM scheduler interprets those parameters as (initial) + * index to the run-queue array. + */ +enum nouveau_sched_priority { + NOUVEAU_SCHED_PRIORITY_SINGLE = DRM_SCHED_PRIORITY_KERNEL, + NOUVEAU_SCHED_PRIORITY_COUNT, +}; + int nouveau_job_init(struct nouveau_job *job, struct nouveau_job_args *args) { - struct nouveau_sched_entity *entity = args->sched_entity; + struct nouveau_sched *sched = args->sched; int ret; + INIT_LIST_HEAD(&job->entry); + job->file_priv = args->file_priv; job->cli = nouveau_cli(args->file_priv); - job->entity = entity; + job->sched = sched; job->sync = args->sync; job->resv_usage = args->resv_usage; @@ -86,10 +84,10 @@ nouveau_job_init(struct nouveau_job *job, ret = -ENOMEM; goto err_free_objs; } - } - ret = drm_sched_job_init(&job->base, &entity->base, 1, NULL); + ret = drm_sched_job_init(&job->base, &sched->entity, + args->credits, NULL); if (ret) goto err_free_chains; @@ -109,6 +107,27 @@ return ret; } void +nouveau_job_fini(struct nouveau_job *job) +{ + dma_fence_put(job->done_fence); + drm_sched_job_cleanup(&job->base); + + job->ops->free(job); +} + +void +nouveau_job_done(struct nouveau_job *job) +{ + struct nouveau_sched *sched = job->sched; + + spin_lock(&sched->job.list.lock); + list_del(&job->entry); + spin_unlock(&sched->job.list.lock); + + wake_up(&sched->job.wq); +} + +void nouveau_job_free(struct nouveau_job *job) { kfree(job->in_sync.data); @@ -117,13 +136,6 @@ nouveau_job_free(struct nouveau_job *job) kfree(job->out_sync.chains); } -void nouveau_job_fini(struct nouveau_job *job) -{ - dma_fence_put(job->done_fence); - drm_sched_job_cleanup(&job->base); - job->ops->free(job); -} - static int sync_find_fence(struct nouveau_job *job, struct drm_nouveau_sync *sync, @@ -261,8 +273,13 @@ nouveau_job_fence_attach(struct nouveau_job *job) int nouveau_job_submit(struct nouveau_job *job) { - struct nouveau_sched_entity *entity = to_nouveau_sched_entity(job->base.entity); + struct nouveau_sched *sched = job->sched; struct dma_fence *done_fence = NULL; + struct drm_gpuvm_exec vm_exec = { + .vm = &nouveau_cli_uvmm(job->cli)->base, + .flags = DRM_EXEC_IGNORE_DUPLICATES, + .num_fences = 1, + }; int ret; ret = nouveau_job_add_deps(job); @@ -276,46 +293,29 @@ nouveau_job_submit(struct nouveau_job *job) /* Make sure the job appears on the sched_entity's queue in the same * order as it was submitted. */ - mutex_lock(&entity->mutex); + mutex_lock(&sched->mutex); /* Guarantee we won't fail after the submit() callback returned * successfully. */ if (job->ops->submit) { - ret = job->ops->submit(job); + ret = job->ops->submit(job, &vm_exec); if (ret) goto err_cleanup; } + /* Submit was successful; add the job to the schedulers job list. */ + spin_lock(&sched->job.list.lock); + list_add(&job->entry, &sched->job.list.head); + spin_unlock(&sched->job.list.lock); + drm_sched_job_arm(&job->base); job->done_fence = dma_fence_get(&job->base.s_fence->finished); if (job->sync) done_fence = dma_fence_get(job->done_fence); - /* If a sched job depends on a dma-fence from a job from the same GPU - * scheduler instance, but a different scheduler entity, the GPU - * scheduler does only wait for the particular job to be scheduled, - * rather than for the job to fully complete. This is due to the GPU - * scheduler assuming that there is a scheduler instance per ring. - * However, the current implementation, in order to avoid arbitrary - * amounts of kthreads, has a single scheduler instance while scheduler - * entities represent rings. - * - * As a workaround, set the DRM_SCHED_FENCE_DONT_PIPELINE for all - * out-fences in order to force the scheduler to wait for full job - * completion for dependent jobs from different entities and same - * scheduler instance. - * - * There is some work in progress [1] to address the issues of firmware - * schedulers; once it is in-tree the scheduler topology in Nouveau - * should be re-worked accordingly. - * - * [1] https://lore.kernel.org/dri-devel/20230801205103.627779-1-matthew.brost@intel.com/ - */ - set_bit(DRM_SCHED_FENCE_DONT_PIPELINE, &job->done_fence->flags); - if (job->ops->armed_submit) - job->ops->armed_submit(job); + job->ops->armed_submit(job, &vm_exec); nouveau_job_fence_attach(job); @@ -326,7 +326,7 @@ nouveau_job_submit(struct nouveau_job *job) drm_sched_entity_push_job(&job->base); - mutex_unlock(&entity->mutex); + mutex_unlock(&sched->mutex); if (done_fence) { dma_fence_wait(done_fence, true); @@ -336,20 +336,13 @@ nouveau_job_submit(struct nouveau_job *job) return 0; err_cleanup: - mutex_unlock(&entity->mutex); + mutex_unlock(&sched->mutex); nouveau_job_fence_attach_cleanup(job); err: job->state = NOUVEAU_JOB_SUBMIT_FAILED; return ret; } -bool -nouveau_sched_entity_qwork(struct nouveau_sched_entity *entity, - struct work_struct *work) -{ - return queue_work(entity->sched_wq, work); -} - static struct dma_fence * nouveau_job_run(struct nouveau_job *job) { @@ -399,50 +392,82 @@ nouveau_sched_free_job(struct drm_sched_job *sched_job) nouveau_job_fini(job); } -int nouveau_sched_entity_init(struct nouveau_sched_entity *entity, - struct drm_gpu_scheduler *sched, - struct workqueue_struct *sched_wq) -{ - mutex_init(&entity->mutex); - spin_lock_init(&entity->job.list.lock); - INIT_LIST_HEAD(&entity->job.list.head); - init_waitqueue_head(&entity->job.wq); - - entity->sched_wq = sched_wq; - return drm_sched_entity_init(&entity->base, - DRM_SCHED_PRIORITY_NORMAL, - &sched, 1, NULL); -} - -void -nouveau_sched_entity_fini(struct nouveau_sched_entity *entity) -{ - drm_sched_entity_destroy(&entity->base); -} - static const struct drm_sched_backend_ops nouveau_sched_ops = { .run_job = nouveau_sched_run_job, .timedout_job = nouveau_sched_timedout_job, .free_job = nouveau_sched_free_job, }; -int nouveau_sched_init(struct nouveau_drm *drm) +int +nouveau_sched_init(struct nouveau_sched *sched, struct nouveau_drm *drm, + struct workqueue_struct *wq, u32 credit_limit) { - struct drm_gpu_scheduler *sched = &drm->sched; + struct drm_gpu_scheduler *drm_sched = &sched->base; + struct drm_sched_entity *entity = &sched->entity; long job_hang_limit = msecs_to_jiffies(NOUVEAU_SCHED_JOB_TIMEOUT_MS); + int ret; - drm->sched_wq = create_singlethread_workqueue("nouveau_sched_wq"); - if (!drm->sched_wq) - return -ENOMEM; + if (!wq) { + wq = alloc_workqueue("nouveau_sched_wq_%d", 0, WQ_MAX_ACTIVE, + current->pid); + if (!wq) + return -ENOMEM; + + sched->wq = wq; + } - return drm_sched_init(sched, &nouveau_sched_ops, NULL, - DRM_SCHED_PRIORITY_COUNT, - NOUVEAU_SCHED_HW_SUBMISSIONS, 0, job_hang_limit, - NULL, NULL, "nouveau_sched", drm->dev->dev); + ret = drm_sched_init(drm_sched, &nouveau_sched_ops, wq, + NOUVEAU_SCHED_PRIORITY_COUNT, + credit_limit, 0, job_hang_limit, + NULL, NULL, "nouveau_sched", drm->dev->dev); + if (ret) + goto fail_wq; + + /* Using DRM_SCHED_PRIORITY_KERNEL, since that's what we're required to use + * when we want to have a single run-queue only. + * + * It's not documented, but one will find out when trying to use any + * other priority running into faults, because the scheduler uses the + * priority as array index. + * + * Can't use NOUVEAU_SCHED_PRIORITY_SINGLE either, because it's not + * matching the enum type used in drm_sched_entity_init(). + */ + ret = drm_sched_entity_init(entity, DRM_SCHED_PRIORITY_KERNEL, + &drm_sched, 1, NULL); + if (ret) + goto fail_sched; + + mutex_init(&sched->mutex); + spin_lock_init(&sched->job.list.lock); + INIT_LIST_HEAD(&sched->job.list.head); + init_waitqueue_head(&sched->job.wq); + + return 0; + +fail_sched: + drm_sched_fini(drm_sched); +fail_wq: + if (sched->wq) + destroy_workqueue(sched->wq); + return ret; } -void nouveau_sched_fini(struct nouveau_drm *drm) +void +nouveau_sched_fini(struct nouveau_sched *sched) { - destroy_workqueue(drm->sched_wq); - drm_sched_fini(&drm->sched); + struct drm_gpu_scheduler *drm_sched = &sched->base; + struct drm_sched_entity *entity = &sched->entity; + + rmb(); /* for list_empty to work without lock */ + wait_event(sched->job.wq, list_empty(&sched->job.list.head)); + + drm_sched_entity_fini(entity); + drm_sched_fini(drm_sched); + + /* Destroy workqueue after scheduler tear down, otherwise it might still + * be in use. + */ + if (sched->wq) + destroy_workqueue(sched->wq); } diff --git a/drivers/gpu/drm/nouveau/nouveau_sched.h b/drivers/gpu/drm/nouveau/nouveau_sched.h index 27ac19792597..a6528f5981e6 100644 --- a/drivers/gpu/drm/nouveau/nouveau_sched.h +++ b/drivers/gpu/drm/nouveau/nouveau_sched.h @@ -5,7 +5,7 @@ #include <linux/types.h> -#include <drm/drm_exec.h> +#include <drm/drm_gpuvm.h> #include <drm/gpu_scheduler.h> #include "nouveau_drv.h" @@ -26,7 +26,8 @@ enum nouveau_job_state { struct nouveau_job_args { struct drm_file *file_priv; - struct nouveau_sched_entity *sched_entity; + struct nouveau_sched *sched; + u32 credits; enum dma_resv_usage resv_usage; bool sync; @@ -49,12 +50,12 @@ struct nouveau_job { enum nouveau_job_state state; - struct nouveau_sched_entity *entity; + struct nouveau_sched *sched; + struct list_head entry; struct drm_file *file_priv; struct nouveau_cli *cli; - struct drm_exec exec; enum dma_resv_usage resv_usage; struct dma_fence *done_fence; @@ -76,8 +77,8 @@ struct nouveau_job { /* If .submit() returns without any error, it is guaranteed that * armed_submit() is called. */ - int (*submit)(struct nouveau_job *); - void (*armed_submit)(struct nouveau_job *); + int (*submit)(struct nouveau_job *, struct drm_gpuvm_exec *); + void (*armed_submit)(struct nouveau_job *, struct drm_gpuvm_exec *); struct dma_fence *(*run)(struct nouveau_job *); void (*free)(struct nouveau_job *); enum drm_gpu_sched_stat (*timeout)(struct nouveau_job *); @@ -90,20 +91,17 @@ int nouveau_job_ucopy_syncs(struct nouveau_job_args *args, int nouveau_job_init(struct nouveau_job *job, struct nouveau_job_args *args); -void nouveau_job_free(struct nouveau_job *job); - -int nouveau_job_submit(struct nouveau_job *job); void nouveau_job_fini(struct nouveau_job *job); +int nouveau_job_submit(struct nouveau_job *job); +void nouveau_job_done(struct nouveau_job *job); +void nouveau_job_free(struct nouveau_job *job); -#define to_nouveau_sched_entity(entity) \ - container_of((entity), struct nouveau_sched_entity, base) - -struct nouveau_sched_entity { - struct drm_sched_entity base; +struct nouveau_sched { + struct drm_gpu_scheduler base; + struct drm_sched_entity entity; + struct workqueue_struct *wq; struct mutex mutex; - struct workqueue_struct *sched_wq; - struct { struct { struct list_head head; @@ -113,15 +111,8 @@ struct nouveau_sched_entity { } job; }; -int nouveau_sched_entity_init(struct nouveau_sched_entity *entity, - struct drm_gpu_scheduler *sched, - struct workqueue_struct *sched_wq); -void nouveau_sched_entity_fini(struct nouveau_sched_entity *entity); - -bool nouveau_sched_entity_qwork(struct nouveau_sched_entity *entity, - struct work_struct *work); - -int nouveau_sched_init(struct nouveau_drm *drm); -void nouveau_sched_fini(struct nouveau_drm *drm); +int nouveau_sched_init(struct nouveau_sched *sched, struct nouveau_drm *drm, + struct workqueue_struct *wq, u32 credit_limit); +void nouveau_sched_fini(struct nouveau_sched *sched); #endif diff --git a/drivers/gpu/drm/nouveau/nouveau_uvmm.c b/drivers/gpu/drm/nouveau/nouveau_uvmm.c index eda7bb8624f1..4b236da79509 100644 --- a/drivers/gpu/drm/nouveau/nouveau_uvmm.c +++ b/drivers/gpu/drm/nouveau/nouveau_uvmm.c @@ -438,8 +438,9 @@ nouveau_uvma_region_complete(struct nouveau_uvma_region *reg) static void op_map_prepare_unwind(struct nouveau_uvma *uvma) { + struct drm_gpuva *va = &uvma->va; nouveau_uvma_gem_put(uvma); - drm_gpuva_remove(&uvma->va); + drm_gpuva_remove(va); nouveau_uvma_free(uvma); } @@ -468,6 +469,7 @@ nouveau_uvmm_sm_prepare_unwind(struct nouveau_uvmm *uvmm, break; case DRM_GPUVA_OP_REMAP: { struct drm_gpuva_op_remap *r = &op->remap; + struct drm_gpuva *va = r->unmap->va; if (r->next) op_map_prepare_unwind(new->next); @@ -475,7 +477,7 @@ nouveau_uvmm_sm_prepare_unwind(struct nouveau_uvmm *uvmm, if (r->prev) op_map_prepare_unwind(new->prev); - op_unmap_prepare_unwind(r->unmap->va); + op_unmap_prepare_unwind(va); break; } case DRM_GPUVA_OP_UNMAP: @@ -634,6 +636,7 @@ nouveau_uvmm_sm_prepare(struct nouveau_uvmm *uvmm, goto unwind; } } + break; } case DRM_GPUVA_OP_REMAP: { @@ -960,6 +963,12 @@ nouveau_uvmm_bind_job_free(struct kref *kref) { struct nouveau_uvmm_bind_job *job = container_of(kref, struct nouveau_uvmm_bind_job, kref); + struct bind_job_op *op, *next; + + list_for_each_op_safe(op, next, &job->ops) { + list_del(&op->entry); + kfree(op); + } nouveau_job_free(&job->base); kfree(job); @@ -1001,14 +1010,16 @@ bind_validate_op(struct nouveau_job *job, static void bind_validate_map_sparse(struct nouveau_job *job, u64 addr, u64 range) { - struct nouveau_uvmm_bind_job *bind_job; - struct nouveau_sched_entity *entity = job->entity; + struct nouveau_sched *sched = job->sched; + struct nouveau_job *__job; struct bind_job_op *op; u64 end = addr + range; again: - spin_lock(&entity->job.list.lock); - list_for_each_entry(bind_job, &entity->job.list.head, entry) { + spin_lock(&sched->job.list.lock); + list_for_each_entry(__job, &sched->job.list.head, entry) { + struct nouveau_uvmm_bind_job *bind_job = to_uvmm_bind_job(__job); + list_for_each_op(op, &bind_job->ops) { if (op->op == OP_UNMAP) { u64 op_addr = op->va.addr; @@ -1016,7 +1027,7 @@ again: if (!(end <= op_addr || addr >= op_end)) { nouveau_uvmm_bind_job_get(bind_job); - spin_unlock(&entity->job.list.lock); + spin_unlock(&sched->job.list.lock); wait_for_completion(&bind_job->complete); nouveau_uvmm_bind_job_put(bind_job); goto again; @@ -1024,7 +1035,7 @@ again: } } } - spin_unlock(&entity->job.list.lock); + spin_unlock(&sched->job.list.lock); } static int @@ -1135,12 +1146,52 @@ bind_link_gpuvas(struct bind_job_op *bop) } static int -nouveau_uvmm_bind_job_submit(struct nouveau_job *job) +bind_lock_validate(struct nouveau_job *job, struct drm_exec *exec, + unsigned int num_fences) +{ + struct nouveau_uvmm_bind_job *bind_job = to_uvmm_bind_job(job); + struct bind_job_op *op; + int ret; + + list_for_each_op(op, &bind_job->ops) { + struct drm_gpuva_op *va_op; + + if (!op->ops) + continue; + + drm_gpuva_for_each_op(va_op, op->ops) { + struct drm_gem_object *obj = op_gem_obj(va_op); + + if (unlikely(!obj)) + continue; + + ret = drm_exec_prepare_obj(exec, obj, num_fences); + if (ret) + return ret; + + /* Don't validate GEMs backing mappings we're about to + * unmap, it's not worth the effort. + */ + if (va_op->op == DRM_GPUVA_OP_UNMAP) + continue; + + ret = nouveau_bo_validate(nouveau_gem_object(obj), + true, false); + if (ret) + return ret; + } + } + + return 0; +} + +static int +nouveau_uvmm_bind_job_submit(struct nouveau_job *job, + struct drm_gpuvm_exec *vme) { struct nouveau_uvmm *uvmm = nouveau_cli_uvmm(job->cli); struct nouveau_uvmm_bind_job *bind_job = to_uvmm_bind_job(job); - struct nouveau_sched_entity *entity = job->entity; - struct drm_exec *exec = &job->exec; + struct drm_exec *exec = &vme->exec; struct bind_job_op *op; int ret; @@ -1157,6 +1208,8 @@ nouveau_uvmm_bind_job_submit(struct nouveau_job *job) dma_resv_unlock(obj->resv); if (IS_ERR(op->vm_bo)) return PTR_ERR(op->vm_bo); + + drm_gpuvm_bo_extobj_add(op->vm_bo); } ret = bind_validate_op(job, op); @@ -1179,6 +1232,7 @@ nouveau_uvmm_bind_job_submit(struct nouveau_job *job) * unwind all GPU VA space changes on failure. */ nouveau_uvmm_lock(uvmm); + list_for_each_op(op, &bind_job->ops) { switch (op->op) { case OP_MAP_SPARSE: @@ -1290,55 +1344,13 @@ nouveau_uvmm_bind_job_submit(struct nouveau_job *job) } } - drm_exec_init(exec, DRM_EXEC_INTERRUPTIBLE_WAIT | - DRM_EXEC_IGNORE_DUPLICATES); + drm_exec_init(exec, vme->flags); drm_exec_until_all_locked(exec) { - list_for_each_op(op, &bind_job->ops) { - struct drm_gpuva_op *va_op; - - if (IS_ERR_OR_NULL(op->ops)) - continue; - - drm_gpuva_for_each_op(va_op, op->ops) { - struct drm_gem_object *obj = op_gem_obj(va_op); - - if (unlikely(!obj)) - continue; - - ret = drm_exec_prepare_obj(exec, obj, 1); - drm_exec_retry_on_contention(exec); - if (ret) { - op = list_last_op(&bind_job->ops); - goto unwind; - } - } - } - } - - list_for_each_op(op, &bind_job->ops) { - struct drm_gpuva_op *va_op; - - if (IS_ERR_OR_NULL(op->ops)) - continue; - - drm_gpuva_for_each_op(va_op, op->ops) { - struct drm_gem_object *obj = op_gem_obj(va_op); - - if (unlikely(!obj)) - continue; - - /* Don't validate GEMs backing mappings we're about to - * unmap, it's not worth the effort. - */ - if (unlikely(va_op->op == DRM_GPUVA_OP_UNMAP)) - continue; - - ret = nouveau_bo_validate(nouveau_gem_object(obj), - true, false); - if (ret) { - op = list_last_op(&bind_job->ops); - goto unwind; - } + ret = bind_lock_validate(job, exec, vme->num_fences); + drm_exec_retry_on_contention(exec); + if (ret) { + op = list_last_op(&bind_job->ops); + goto unwind; } } @@ -1375,10 +1387,6 @@ nouveau_uvmm_bind_job_submit(struct nouveau_job *job) } nouveau_uvmm_unlock(uvmm); - spin_lock(&entity->job.list.lock); - list_add(&bind_job->entry, &entity->job.list.head); - spin_unlock(&entity->job.list.lock); - return 0; unwind_continue: @@ -1413,21 +1421,17 @@ unwind: } nouveau_uvmm_unlock(uvmm); - drm_exec_fini(exec); + drm_gpuvm_exec_unlock(vme); return ret; } static void -nouveau_uvmm_bind_job_armed_submit(struct nouveau_job *job) +nouveau_uvmm_bind_job_armed_submit(struct nouveau_job *job, + struct drm_gpuvm_exec *vme) { - struct drm_exec *exec = &job->exec; - struct drm_gem_object *obj; - unsigned long index; - - drm_exec_for_each_locked_object(exec, index, obj) - dma_resv_add_fence(obj->resv, job->done_fence, job->resv_usage); - - drm_exec_fini(exec); + drm_gpuvm_exec_resv_add_fence(vme, job->done_fence, + job->resv_usage, job->resv_usage); + drm_gpuvm_exec_unlock(vme); } static struct dma_fence * @@ -1465,14 +1469,11 @@ out: } static void -nouveau_uvmm_bind_job_free_work_fn(struct work_struct *work) +nouveau_uvmm_bind_job_cleanup(struct nouveau_job *job) { - struct nouveau_uvmm_bind_job *bind_job = - container_of(work, struct nouveau_uvmm_bind_job, work); - struct nouveau_job *job = &bind_job->base; + struct nouveau_uvmm_bind_job *bind_job = to_uvmm_bind_job(job); struct nouveau_uvmm *uvmm = nouveau_cli_uvmm(job->cli); - struct nouveau_sched_entity *entity = job->entity; - struct bind_job_op *op, *next; + struct bind_job_op *op; list_for_each_op(op, &bind_job->ops) { struct drm_gem_object *obj = op->gem.obj; @@ -1524,38 +1525,17 @@ nouveau_uvmm_bind_job_free_work_fn(struct work_struct *work) drm_gem_object_put(obj); } - spin_lock(&entity->job.list.lock); - list_del(&bind_job->entry); - spin_unlock(&entity->job.list.lock); - + nouveau_job_done(job); complete_all(&bind_job->complete); - wake_up(&entity->job.wq); - - /* Remove and free ops after removing the bind job from the job list to - * avoid races against bind_validate_map_sparse(). - */ - list_for_each_op_safe(op, next, &bind_job->ops) { - list_del(&op->entry); - kfree(op); - } nouveau_uvmm_bind_job_put(bind_job); } -static void -nouveau_uvmm_bind_job_free_qwork(struct nouveau_job *job) -{ - struct nouveau_uvmm_bind_job *bind_job = to_uvmm_bind_job(job); - struct nouveau_sched_entity *entity = job->entity; - - nouveau_sched_entity_qwork(entity, &bind_job->work); -} - static struct nouveau_job_ops nouveau_bind_job_ops = { .submit = nouveau_uvmm_bind_job_submit, .armed_submit = nouveau_uvmm_bind_job_armed_submit, .run = nouveau_uvmm_bind_job_run, - .free = nouveau_uvmm_bind_job_free_qwork, + .free = nouveau_uvmm_bind_job_cleanup, }; static int @@ -1616,7 +1596,6 @@ nouveau_uvmm_bind_job_init(struct nouveau_uvmm_bind_job **pjob, return ret; INIT_LIST_HEAD(&job->ops); - INIT_LIST_HEAD(&job->entry); for (i = 0; i < __args->op.count; i++) { ret = bind_job_op_from_uop(&op, &__args->op.s[i]); @@ -1627,11 +1606,12 @@ nouveau_uvmm_bind_job_init(struct nouveau_uvmm_bind_job **pjob, } init_completion(&job->complete); - INIT_WORK(&job->work, nouveau_uvmm_bind_job_free_work_fn); - args.sched_entity = __args->sched_entity; args.file_priv = __args->file_priv; + args.sched = __args->sched; + args.credits = 1; + args.in_sync.count = __args->in_sync.count; args.in_sync.s = __args->in_sync.s; @@ -1757,7 +1737,7 @@ nouveau_uvmm_ioctl_vm_bind(struct drm_device *dev, if (ret) return ret; - args.sched_entity = &cli->sched_entity; + args.sched = &cli->sched; args.file_priv = file_priv; ret = nouveau_uvmm_vm_bind(&args); @@ -1815,8 +1795,17 @@ nouveau_uvmm_free(struct drm_gpuvm *gpuvm) kfree(uvmm); } +static int +nouveau_uvmm_bo_validate(struct drm_gpuvm_bo *vm_bo, struct drm_exec *exec) +{ + struct nouveau_bo *nvbo = nouveau_gem_object(vm_bo->obj); + + return nouveau_bo_validate(nvbo, true, false); +} + static const struct drm_gpuvm_ops gpuvm_ops = { .vm_free = nouveau_uvmm_free, + .vm_bo_validate = nouveau_uvmm_bo_validate, }; int @@ -1900,12 +1889,8 @@ nouveau_uvmm_fini(struct nouveau_uvmm *uvmm) MA_STATE(mas, &uvmm->region_mt, 0, 0); struct nouveau_uvma_region *reg; struct nouveau_cli *cli = uvmm->vmm.cli; - struct nouveau_sched_entity *entity = &cli->sched_entity; struct drm_gpuva *va, *next; - rmb(); /* for list_empty to work without lock */ - wait_event(entity->job.wq, list_empty(&entity->job.list.head)); - nouveau_uvmm_lock(uvmm); drm_gpuvm_for_each_va_safe(va, next, &uvmm->base) { struct nouveau_uvma *uvma = uvma_from_va(va); diff --git a/drivers/gpu/drm/nouveau/nouveau_uvmm.h b/drivers/gpu/drm/nouveau/nouveau_uvmm.h index f0a6d98ace4f..9d3c348581eb 100644 --- a/drivers/gpu/drm/nouveau/nouveau_uvmm.h +++ b/drivers/gpu/drm/nouveau/nouveau_uvmm.h @@ -44,8 +44,6 @@ struct nouveau_uvmm_bind_job { struct nouveau_job base; struct kref kref; - struct list_head entry; - struct work_struct work; struct completion complete; /* struct bind_job_op */ @@ -54,7 +52,7 @@ struct nouveau_uvmm_bind_job { struct nouveau_uvmm_bind_job_args { struct drm_file *file_priv; - struct nouveau_sched_entity *sched_entity; + struct nouveau_sched *sched; unsigned int flags; diff --git a/drivers/gpu/drm/omapdrm/dss/dispc.c b/drivers/gpu/drm/omapdrm/dss/dispc.c index c26aab4939fa..993691b3cc7e 100644 --- a/drivers/gpu/drm/omapdrm/dss/dispc.c +++ b/drivers/gpu/drm/omapdrm/dss/dispc.c @@ -22,11 +22,11 @@ #include <linux/hardirq.h> #include <linux/platform_device.h> #include <linux/pm_runtime.h> +#include <linux/property.h> #include <linux/sizes.h> #include <linux/mfd/syscon.h> #include <linux/regmap.h> #include <linux/of.h> -#include <linux/of_device.h> #include <linux/component.h> #include <linux/sys_soc.h> #include <drm/drm_fourcc.h> @@ -4765,7 +4765,7 @@ static int dispc_bind(struct device *dev, struct device *master, void *data) if (soc) dispc->feat = soc->data; else - dispc->feat = of_match_device(dispc_of_match, &pdev->dev)->data; + dispc->feat = device_get_match_data(&pdev->dev); r = dispc_errata_i734_wa_init(dispc); if (r) diff --git a/drivers/gpu/drm/omapdrm/dss/dss.c b/drivers/gpu/drm/omapdrm/dss/dss.c index 02955f976845..988888e164d7 100644 --- a/drivers/gpu/drm/omapdrm/dss/dss.c +++ b/drivers/gpu/drm/omapdrm/dss/dss.c @@ -22,12 +22,13 @@ #include <linux/pinctrl/consumer.h> #include <linux/platform_device.h> #include <linux/pm_runtime.h> +#include <linux/property.h> #include <linux/gfp.h> #include <linux/sizes.h> #include <linux/mfd/syscon.h> #include <linux/regmap.h> #include <linux/of.h> -#include <linux/of_device.h> +#include <linux/of_platform.h> #include <linux/of_graph.h> #include <linux/regulator/consumer.h> #include <linux/suspend.h> @@ -1445,7 +1446,7 @@ static int dss_probe(struct platform_device *pdev) if (soc) dss->feat = soc->data; else - dss->feat = of_match_device(dss_of_match, &pdev->dev)->data; + dss->feat = device_get_match_data(&pdev->dev); /* Map I/O registers, get and setup clocks. */ dss->base = devm_platform_ioremap_resource(pdev, 0); diff --git a/drivers/gpu/drm/qxl/qxl_display.c b/drivers/gpu/drm/qxl/qxl_display.c index 404b0483bb7c..c6d35c33d5d6 100644 --- a/drivers/gpu/drm/qxl/qxl_display.c +++ b/drivers/gpu/drm/qxl/qxl_display.c @@ -485,7 +485,6 @@ static int qxl_primary_atomic_check(struct drm_plane *plane, static int qxl_primary_apply_cursor(struct qxl_device *qdev, struct drm_plane_state *plane_state) { - struct drm_framebuffer *fb = plane_state->fb; struct qxl_crtc *qcrtc = to_qxl_crtc(plane_state->crtc); struct qxl_cursor_cmd *cmd; struct qxl_release *release; @@ -510,8 +509,8 @@ static int qxl_primary_apply_cursor(struct qxl_device *qdev, cmd = (struct qxl_cursor_cmd *)qxl_release_map(qdev, release); cmd->type = QXL_CURSOR_SET; - cmd->u.set.position.x = plane_state->crtc_x + fb->hot_x; - cmd->u.set.position.y = plane_state->crtc_y + fb->hot_y; + cmd->u.set.position.x = plane_state->crtc_x + plane_state->hotspot_x; + cmd->u.set.position.y = plane_state->crtc_y + plane_state->hotspot_y; cmd->u.set.shape = qxl_bo_physical_address(qdev, qcrtc->cursor_bo, 0); @@ -531,7 +530,6 @@ out_free_release: static int qxl_primary_move_cursor(struct qxl_device *qdev, struct drm_plane_state *plane_state) { - struct drm_framebuffer *fb = plane_state->fb; struct qxl_crtc *qcrtc = to_qxl_crtc(plane_state->crtc); struct qxl_cursor_cmd *cmd; struct qxl_release *release; @@ -554,8 +552,8 @@ static int qxl_primary_move_cursor(struct qxl_device *qdev, cmd = (struct qxl_cursor_cmd *)qxl_release_map(qdev, release); cmd->type = QXL_CURSOR_MOVE; - cmd->u.position.x = plane_state->crtc_x + fb->hot_x; - cmd->u.position.y = plane_state->crtc_y + fb->hot_y; + cmd->u.position.x = plane_state->crtc_x + plane_state->hotspot_x; + cmd->u.position.y = plane_state->crtc_y + plane_state->hotspot_y; qxl_release_unmap(qdev, release, &cmd->release_info); qxl_release_fence_buffer_objects(release); @@ -851,8 +849,8 @@ static int qxl_plane_prepare_fb(struct drm_plane *plane, struct qxl_bo *old_cursor_bo = qcrtc->cursor_bo; qcrtc->cursor_bo = qxl_create_cursor(qdev, user_bo, - new_state->fb->hot_x, - new_state->fb->hot_y); + new_state->hotspot_x, + new_state->hotspot_y); qxl_free_cursor(old_cursor_bo); } diff --git a/drivers/gpu/drm/qxl/qxl_drv.c b/drivers/gpu/drm/qxl/qxl_drv.c index 46de4f171970..beee5563031a 100644 --- a/drivers/gpu/drm/qxl/qxl_drv.c +++ b/drivers/gpu/drm/qxl/qxl_drv.c @@ -285,7 +285,7 @@ static const struct drm_ioctl_desc qxl_ioctls[] = { }; static struct drm_driver qxl_driver = { - .driver_features = DRIVER_GEM | DRIVER_MODESET | DRIVER_ATOMIC, + .driver_features = DRIVER_GEM | DRIVER_MODESET | DRIVER_ATOMIC | DRIVER_CURSOR_HOTSPOT, .dumb_create = qxl_mode_dumb_create, .dumb_map_offset = drm_gem_ttm_dumb_map_offset, diff --git a/drivers/gpu/drm/scheduler/sched_entity.c b/drivers/gpu/drm/scheduler/sched_entity.c index 4d42b1e4daa6..3c4f5a392b06 100644 --- a/drivers/gpu/drm/scheduler/sched_entity.c +++ b/drivers/gpu/drm/scheduler/sched_entity.c @@ -81,12 +81,16 @@ int drm_sched_entity_init(struct drm_sched_entity *entity, */ pr_warn("%s: called with uninitialized scheduler\n", __func__); } else if (num_sched_list) { - /* The "priority" of an entity cannot exceed the number - * of run-queues of a scheduler. + /* The "priority" of an entity cannot exceed the number of run-queues of a + * scheduler. Protect against num_rqs being 0, by converting to signed. Choose + * the lowest priority available. */ - if (entity->priority >= sched_list[0]->num_rqs) - entity->priority = max_t(u32, sched_list[0]->num_rqs, - DRM_SCHED_PRIORITY_MIN); + if (entity->priority >= sched_list[0]->num_rqs) { + drm_err(sched_list[0], "entity with out-of-bounds priority:%u num_rqs:%u\n", + entity->priority, sched_list[0]->num_rqs); + entity->priority = max_t(s32, (s32) sched_list[0]->num_rqs - 1, + (s32) DRM_SCHED_PRIORITY_KERNEL); + } entity->rq = sched_list[0]->sched_rq[entity->priority]; } diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c index 044a8c4875ba..682aebe96db7 100644 --- a/drivers/gpu/drm/scheduler/sched_main.c +++ b/drivers/gpu/drm/scheduler/sched_main.c @@ -1051,8 +1051,9 @@ drm_sched_select_entity(struct drm_gpu_scheduler *sched) struct drm_sched_entity *entity; int i; - /* Kernel run queue has higher priority than normal run queue*/ - for (i = sched->num_rqs - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) { + /* Start with the highest priority. + */ + for (i = DRM_SCHED_PRIORITY_KERNEL; i < sched->num_rqs; i++) { entity = drm_sched_policy == DRM_SCHED_POLICY_FIFO ? drm_sched_rq_select_entity_fifo(sched, sched->sched_rq[i]) : drm_sched_rq_select_entity_rr(sched, sched->sched_rq[i]); @@ -1291,7 +1292,7 @@ int drm_sched_init(struct drm_gpu_scheduler *sched, if (!sched->sched_rq) goto Out_free; sched->num_rqs = num_rqs; - for (i = DRM_SCHED_PRIORITY_MIN; i < sched->num_rqs; i++) { + for (i = DRM_SCHED_PRIORITY_KERNEL; i < sched->num_rqs; i++) { sched->sched_rq[i] = kzalloc(sizeof(*sched->sched_rq[i]), GFP_KERNEL); if (!sched->sched_rq[i]) goto Out_unroll; @@ -1312,7 +1313,7 @@ int drm_sched_init(struct drm_gpu_scheduler *sched, sched->ready = true; return 0; Out_unroll: - for (--i ; i >= DRM_SCHED_PRIORITY_MIN; i--) + for (--i ; i >= DRM_SCHED_PRIORITY_KERNEL; i--) kfree(sched->sched_rq[i]); Out_free: kfree(sched->sched_rq); @@ -1338,7 +1339,7 @@ void drm_sched_fini(struct drm_gpu_scheduler *sched) drm_sched_wqueue_stop(sched); - for (i = sched->num_rqs - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) { + for (i = DRM_SCHED_PRIORITY_KERNEL; i < sched->num_rqs; i++) { struct drm_sched_rq *rq = sched->sched_rq[i]; spin_lock(&rq->lock); @@ -1390,9 +1391,7 @@ void drm_sched_increase_karma(struct drm_sched_job *bad) if (bad->s_priority != DRM_SCHED_PRIORITY_KERNEL) { atomic_inc(&bad->karma); - for (i = DRM_SCHED_PRIORITY_MIN; - i < min_t(typeof(sched->num_rqs), sched->num_rqs, DRM_SCHED_PRIORITY_KERNEL); - i++) { + for (i = DRM_SCHED_PRIORITY_HIGH; i < sched->num_rqs; i++) { struct drm_sched_rq *rq = sched->sched_rq[i]; spin_lock(&rq->lock); diff --git a/drivers/gpu/drm/ttm/ttm_bo.c b/drivers/gpu/drm/ttm/ttm_bo.c index e58b7e249816..edf10618fe2b 100644 --- a/drivers/gpu/drm/ttm/ttm_bo.c +++ b/drivers/gpu/drm/ttm/ttm_bo.c @@ -370,7 +370,13 @@ static void ttm_bo_release(struct kref *kref) spin_unlock(&bo->bdev->lru_lock); INIT_WORK(&bo->delayed_delete, ttm_bo_delayed_delete); - queue_work(bdev->wq, &bo->delayed_delete); + + /* Schedule the worker on the closest NUMA node. This + * improves performance since system memory might be + * cleared on free and that is best done on a CPU core + * close to it. + */ + queue_work_node(bdev->pool.nid, bdev->wq, &bo->delayed_delete); return; } diff --git a/drivers/gpu/drm/ttm/ttm_device.c b/drivers/gpu/drm/ttm/ttm_device.c index d48b39132b32..f5187b384ae9 100644 --- a/drivers/gpu/drm/ttm/ttm_device.c +++ b/drivers/gpu/drm/ttm/ttm_device.c @@ -204,7 +204,8 @@ int ttm_device_init(struct ttm_device *bdev, const struct ttm_device_funcs *func if (ret) return ret; - bdev->wq = alloc_workqueue("ttm", WQ_MEM_RECLAIM | WQ_HIGHPRI, 16); + bdev->wq = alloc_workqueue("ttm", + WQ_MEM_RECLAIM | WQ_HIGHPRI | WQ_UNBOUND, 16); if (!bdev->wq) { ttm_global_release(); return -ENOMEM; @@ -213,7 +214,8 @@ int ttm_device_init(struct ttm_device *bdev, const struct ttm_device_funcs *func bdev->funcs = funcs; ttm_sys_man_init(bdev); - ttm_pool_init(&bdev->pool, dev, NUMA_NO_NODE, use_dma_alloc, use_dma32); + + ttm_pool_init(&bdev->pool, dev, dev_to_node(dev), use_dma_alloc, use_dma32); bdev->vma_manager = vma_manager; spin_lock_init(&bdev->lru_lock); diff --git a/drivers/gpu/drm/vboxvideo/vbox_drv.c b/drivers/gpu/drm/vboxvideo/vbox_drv.c index 047b95812334..cd9e66a06596 100644 --- a/drivers/gpu/drm/vboxvideo/vbox_drv.c +++ b/drivers/gpu/drm/vboxvideo/vbox_drv.c @@ -182,7 +182,7 @@ DEFINE_DRM_GEM_FOPS(vbox_fops); static const struct drm_driver driver = { .driver_features = - DRIVER_MODESET | DRIVER_GEM | DRIVER_ATOMIC, + DRIVER_MODESET | DRIVER_GEM | DRIVER_ATOMIC | DRIVER_CURSOR_HOTSPOT, .fops = &vbox_fops, .name = DRIVER_NAME, diff --git a/drivers/gpu/drm/vboxvideo/vbox_mode.c b/drivers/gpu/drm/vboxvideo/vbox_mode.c index 341edd982cb3..9ff3bade9795 100644 --- a/drivers/gpu/drm/vboxvideo/vbox_mode.c +++ b/drivers/gpu/drm/vboxvideo/vbox_mode.c @@ -429,8 +429,8 @@ static void vbox_cursor_atomic_update(struct drm_plane *plane, flags = VBOX_MOUSE_POINTER_VISIBLE | VBOX_MOUSE_POINTER_SHAPE | VBOX_MOUSE_POINTER_ALPHA; hgsmi_update_pointer_shape(vbox->guest_pool, flags, - min_t(u32, max(fb->hot_x, 0), width), - min_t(u32, max(fb->hot_y, 0), height), + min_t(u32, max(new_state->hotspot_x, 0), width), + min_t(u32, max(new_state->hotspot_y, 0), height), width, height, vbox->cursor_data, data_size); mutex_unlock(&vbox->hw_mutex); diff --git a/drivers/gpu/drm/virtio/virtgpu_drv.c b/drivers/gpu/drm/virtio/virtgpu_drv.c index 4334c7608408..f8e9abe647b9 100644 --- a/drivers/gpu/drm/virtio/virtgpu_drv.c +++ b/drivers/gpu/drm/virtio/virtgpu_drv.c @@ -177,7 +177,7 @@ static const struct drm_driver driver = { * out via drm_device::driver_features: */ .driver_features = DRIVER_MODESET | DRIVER_GEM | DRIVER_RENDER | DRIVER_ATOMIC | - DRIVER_SYNCOBJ | DRIVER_SYNCOBJ_TIMELINE, + DRIVER_SYNCOBJ | DRIVER_SYNCOBJ_TIMELINE | DRIVER_CURSOR_HOTSPOT, .open = virtio_gpu_driver_open, .postclose = virtio_gpu_driver_postclose, diff --git a/drivers/gpu/drm/virtio/virtgpu_plane.c b/drivers/gpu/drm/virtio/virtgpu_plane.c index a2e045f3a000..a72a2dbda031 100644 --- a/drivers/gpu/drm/virtio/virtgpu_plane.c +++ b/drivers/gpu/drm/virtio/virtgpu_plane.c @@ -79,6 +79,8 @@ static int virtio_gpu_plane_atomic_check(struct drm_plane *plane, { struct drm_plane_state *new_plane_state = drm_atomic_get_new_plane_state(state, plane); + struct drm_plane_state *old_plane_state = drm_atomic_get_old_plane_state(state, + plane); bool is_cursor = plane->type == DRM_PLANE_TYPE_CURSOR; struct drm_crtc_state *crtc_state; int ret; @@ -86,6 +88,14 @@ static int virtio_gpu_plane_atomic_check(struct drm_plane *plane, if (!new_plane_state->fb || WARN_ON(!new_plane_state->crtc)) return 0; + /* + * Ignore damage clips if the framebuffer attached to the plane's state + * has changed since the last plane update (page-flip). In this case, a + * full plane update should happen because uploads are done per-buffer. + */ + if (old_plane_state->fb != new_plane_state->fb) + new_plane_state->ignore_damage_clips = true; + crtc_state = drm_atomic_get_crtc_state(state, new_plane_state->crtc); if (IS_ERR(crtc_state)) @@ -323,16 +333,16 @@ static void virtio_gpu_cursor_plane_update(struct drm_plane *plane, DRM_DEBUG("update, handle %d, pos +%d+%d, hot %d,%d\n", handle, plane->state->crtc_x, plane->state->crtc_y, - plane->state->fb ? plane->state->fb->hot_x : 0, - plane->state->fb ? plane->state->fb->hot_y : 0); + plane->state->hotspot_x, + plane->state->hotspot_y); output->cursor.hdr.type = cpu_to_le32(VIRTIO_GPU_CMD_UPDATE_CURSOR); output->cursor.resource_id = cpu_to_le32(handle); if (plane->state->fb) { output->cursor.hot_x = - cpu_to_le32(plane->state->fb->hot_x); + cpu_to_le32(plane->state->hotspot_x); output->cursor.hot_y = - cpu_to_le32(plane->state->fb->hot_y); + cpu_to_le32(plane->state->hotspot_y); } else { output->cursor.hot_x = cpu_to_le32(0); output->cursor.hot_y = cpu_to_le32(0); diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c index 8b24ecf60e3e..d3e308fdfd5b 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_drv.c @@ -1611,7 +1611,7 @@ static const struct file_operations vmwgfx_driver_fops = { static const struct drm_driver driver = { .driver_features = - DRIVER_MODESET | DRIVER_RENDER | DRIVER_ATOMIC | DRIVER_GEM, + DRIVER_MODESET | DRIVER_RENDER | DRIVER_ATOMIC | DRIVER_GEM | DRIVER_CURSOR_HOTSPOT, .ioctls = vmw_ioctls, .num_ioctls = ARRAY_SIZE(vmw_ioctls), .master_set = vmw_master_set, diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c index 818b7f109f53..5fd0ccaa0b41 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_kms.c @@ -768,13 +768,8 @@ vmw_du_cursor_plane_atomic_update(struct drm_plane *plane, struct vmw_plane_state *old_vps = vmw_plane_state_to_vps(old_state); s32 hotspot_x, hotspot_y; - hotspot_x = du->hotspot_x; - hotspot_y = du->hotspot_y; - - if (new_state->fb) { - hotspot_x += new_state->fb->hot_x; - hotspot_y += new_state->fb->hot_y; - } + hotspot_x = du->hotspot_x + new_state->hotspot_x; + hotspot_y = du->hotspot_y + new_state->hotspot_y; du->cursor_surface = vps->surf; du->cursor_bo = vps->bo; @@ -837,10 +832,21 @@ int vmw_du_primary_plane_atomic_check(struct drm_plane *plane, { struct drm_plane_state *new_state = drm_atomic_get_new_plane_state(state, plane); + struct drm_plane_state *old_state = drm_atomic_get_old_plane_state(state, + plane); struct drm_crtc_state *crtc_state = NULL; struct drm_framebuffer *new_fb = new_state->fb; + struct drm_framebuffer *old_fb = old_state->fb; int ret; + /* + * Ignore damage clips if the framebuffer attached to the plane's state + * has changed since the last plane update (page-flip). In this case, a + * full plane update should happen because uploads are done per-buffer. + */ + if (old_fb != new_fb) + new_state->ignore_damage_clips = true; + if (new_state->crtc) crtc_state = drm_atomic_get_new_crtc_state(state, new_state->crtc); |