diff options
Diffstat (limited to 'drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c')
| -rw-r--r-- | drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c | 503 |
1 files changed, 278 insertions, 225 deletions
diff --git a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c index a2b8464b3f56..3057f8baa7d2 100644 --- a/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c +++ b/drivers/gpu/drm/vmwgfx/vmwgfx_execbuf.c @@ -1,39 +1,23 @@ // SPDX-License-Identifier: GPL-2.0 OR MIT /************************************************************************** * - * Copyright 2009 - 2015 VMware, Inc., Palo Alto, CA., USA - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. + * Copyright (c) 2009-2025 Broadcom. All Rights Reserved. The term + * “Broadcom” refers to Broadcom Inc. and/or its subsidiaries. * **************************************************************************/ -#include <linux/sync_file.h> +#include "vmwgfx_binding.h" +#include "vmwgfx_bo.h" #include "vmwgfx_drv.h" -#include "vmwgfx_reg.h" -#include <drm/ttm/ttm_bo_api.h> -#include <drm/ttm/ttm_placement.h> +#include "vmwgfx_mksstat.h" #include "vmwgfx_so.h" -#include "vmwgfx_binding.h" -#define VMW_RES_HT_ORDER 12 +#include <drm/ttm/ttm_bo.h> +#include <drm/ttm/ttm_placement.h> + +#include <linux/sync_file.h> +#include <linux/hashtable.h> +#include <linux/vmalloc.h> /* * Helper macro to get dx_ctx_node if available otherwise print an error @@ -64,7 +48,7 @@ */ struct vmw_relocation { struct list_head head; - struct vmw_buffer_object *vbo; + struct vmw_bo *vbo; union { SVGAMobId *mob_loc; SVGAGuestPtr *location; @@ -148,7 +132,7 @@ static int vmw_resource_context_res_add(struct vmw_private *dev_priv, static int vmw_translate_mob_ptr(struct vmw_private *dev_priv, struct vmw_sw_context *sw_context, SVGAMobId *id, - struct vmw_buffer_object **vmw_bo_p); + struct vmw_bo **vmw_bo_p); /** * vmw_ptr_diff - Compute the offset from a to b in bytes * @@ -289,20 +273,26 @@ static void vmw_execbuf_rcache_update(struct vmw_res_cache_entry *rcache, rcache->valid_handle = 0; } +enum vmw_val_add_flags { + vmw_val_add_flag_none = 0, + vmw_val_add_flag_noctx = 1 << 0, +}; + /** - * vmw_execbuf_res_noref_val_add - Add a resource described by an unreferenced - * rcu-protected pointer to the validation list. + * vmw_execbuf_res_val_add - Add a resource to the validation list. * * @sw_context: Pointer to the software context. * @res: Unreferenced rcu-protected pointer to the resource. * @dirty: Whether to change dirty status. + * @flags: specifies whether to use the context or not * * Returns: 0 on success. Negative error code on failure. Typical error codes * are %-EINVAL on inconsistency and %-ESRCH if the resource was doomed. */ -static int vmw_execbuf_res_noref_val_add(struct vmw_sw_context *sw_context, - struct vmw_resource *res, - u32 dirty) +static int vmw_execbuf_res_val_add(struct vmw_sw_context *sw_context, + struct vmw_resource *res, + u32 dirty, + u32 flags) { struct vmw_private *dev_priv = res->dev_priv; int ret; @@ -317,24 +307,30 @@ static int vmw_execbuf_res_noref_val_add(struct vmw_sw_context *sw_context, if (dirty) vmw_validation_res_set_dirty(sw_context->ctx, rcache->private, dirty); - vmw_user_resource_noref_release(); return 0; } - priv_size = vmw_execbuf_res_size(dev_priv, res_type); - ret = vmw_validation_add_resource(sw_context->ctx, res, priv_size, - dirty, (void **)&ctx_info, - &first_usage); - vmw_user_resource_noref_release(); - if (ret) - return ret; + if ((flags & vmw_val_add_flag_noctx) != 0) { + ret = vmw_validation_add_resource(sw_context->ctx, res, 0, dirty, + (void **)&ctx_info, NULL); + if (ret) + return ret; - if (priv_size && first_usage) { - ret = vmw_cmd_ctx_first_setup(dev_priv, sw_context, res, - ctx_info); - if (ret) { - VMW_DEBUG_USER("Failed first usage context setup.\n"); + } else { + priv_size = vmw_execbuf_res_size(dev_priv, res_type); + ret = vmw_validation_add_resource(sw_context->ctx, res, priv_size, + dirty, (void **)&ctx_info, + &first_usage); + if (ret) return ret; + + if (priv_size && first_usage) { + ret = vmw_cmd_ctx_first_setup(dev_priv, sw_context, res, + ctx_info); + if (ret) { + VMW_DEBUG_USER("Failed first usage context setup.\n"); + return ret; + } } } @@ -343,43 +339,6 @@ static int vmw_execbuf_res_noref_val_add(struct vmw_sw_context *sw_context, } /** - * vmw_execbuf_res_noctx_val_add - Add a non-context resource to the resource - * validation list if it's not already on it - * - * @sw_context: Pointer to the software context. - * @res: Pointer to the resource. - * @dirty: Whether to change dirty status. - * - * Returns: Zero on success. Negative error code on failure. - */ -static int vmw_execbuf_res_noctx_val_add(struct vmw_sw_context *sw_context, - struct vmw_resource *res, - u32 dirty) -{ - struct vmw_res_cache_entry *rcache; - enum vmw_res_type res_type = vmw_res_type(res); - void *ptr; - int ret; - - rcache = &sw_context->res_cache[res_type]; - if (likely(rcache->valid && rcache->res == res)) { - if (dirty) - vmw_validation_res_set_dirty(sw_context->ctx, - rcache->private, dirty); - return 0; - } - - ret = vmw_validation_add_resource(sw_context->ctx, res, 0, dirty, - &ptr, NULL); - if (ret) - return ret; - - vmw_execbuf_rcache_update(rcache, res, ptr); - - return 0; -} - -/** * vmw_view_res_val_add - Add a view and the surface it's pointing to to the * validation list * @@ -397,13 +356,13 @@ static int vmw_view_res_val_add(struct vmw_sw_context *sw_context, * First add the resource the view is pointing to, otherwise it may be * swapped out when the view is validated. */ - ret = vmw_execbuf_res_noctx_val_add(sw_context, vmw_view_srf(view), - vmw_view_dirtying(view)); + ret = vmw_execbuf_res_val_add(sw_context, vmw_view_srf(view), + vmw_view_dirtying(view), vmw_val_add_flag_noctx); if (ret) return ret; - return vmw_execbuf_res_noctx_val_add(sw_context, view, - VMW_RES_DIRTY_NONE); + return vmw_execbuf_res_val_add(sw_context, view, VMW_RES_DIRTY_NONE, + vmw_val_add_flag_noctx); } /** @@ -471,11 +430,12 @@ static int vmw_resource_context_res_add(struct vmw_private *dev_priv, vmw_res_type(ctx) == vmw_res_dx_context) { for (i = 0; i < cotable_max; ++i) { res = vmw_context_cotable(ctx, i); - if (IS_ERR(res)) + if (IS_ERR_OR_NULL(res)) continue; - ret = vmw_execbuf_res_noctx_val_add(sw_context, res, - VMW_RES_DIRTY_SET); + ret = vmw_execbuf_res_val_add(sw_context, res, + VMW_RES_DIRTY_SET, + vmw_val_add_flag_noctx); if (unlikely(ret != 0)) return ret; } @@ -489,21 +449,25 @@ static int vmw_resource_context_res_add(struct vmw_private *dev_priv, if (vmw_res_type(entry->res) == vmw_res_view) ret = vmw_view_res_val_add(sw_context, entry->res); else - ret = vmw_execbuf_res_noctx_val_add - (sw_context, entry->res, - vmw_binding_dirtying(entry->bt)); + ret = vmw_execbuf_res_val_add(sw_context, entry->res, + vmw_binding_dirtying(entry->bt), + vmw_val_add_flag_noctx); if (unlikely(ret != 0)) break; } if (has_sm4_context(dev_priv) && vmw_res_type(ctx) == vmw_res_dx_context) { - struct vmw_buffer_object *dx_query_mob; + struct vmw_bo *dx_query_mob; dx_query_mob = vmw_context_get_dx_query_mob(ctx); - if (dx_query_mob) + if (dx_query_mob) { + vmw_bo_placement_set(dx_query_mob, + VMW_BO_DOMAIN_MOB, + VMW_BO_DOMAIN_MOB); ret = vmw_validation_add_bo(sw_context->ctx, - dx_query_mob, true, false); + dx_query_mob); + } } mutex_unlock(&dev_priv->binding_mutex); @@ -619,7 +583,7 @@ static int vmw_resources_reserve(struct vmw_sw_context *sw_context) return ret; if (sw_context->dx_query_mob) { - struct vmw_buffer_object *expected_dx_query_mob; + struct vmw_bo *expected_dx_query_mob; expected_dx_query_mob = vmw_context_get_dx_query_mob(sw_context->dx_query_ctx); @@ -640,10 +604,10 @@ static int vmw_resources_reserve(struct vmw_sw_context *sw_context) * @sw_context: Pointer to the software context. * @res_type: Resource type. * @dirty: Whether to change dirty status. - * @converter: User-space visisble type specific information. + * @converter: User-space visible type specific information. * @id_loc: Pointer to the location in the command buffer currently being parsed * from where the user-space resource id handle is located. - * @p_res: Pointer to pointer to resource validalidation node. Populated on + * @p_res: Pointer to pointer to resource validation node. Populated on * exit. */ static int @@ -657,7 +621,8 @@ vmw_cmd_res_check(struct vmw_private *dev_priv, { struct vmw_res_cache_entry *rcache = &sw_context->res_cache[res_type]; struct vmw_resource *res; - int ret; + int ret = 0; + bool needs_unref = false; if (p_res) *p_res = NULL; @@ -682,17 +647,18 @@ vmw_cmd_res_check(struct vmw_private *dev_priv, if (ret) return ret; - res = vmw_user_resource_noref_lookup_handle - (dev_priv, sw_context->fp->tfile, *id_loc, converter); - if (IS_ERR(res)) { + ret = vmw_user_resource_lookup_handle + (dev_priv, sw_context->fp->tfile, *id_loc, converter, &res); + if (ret != 0) { VMW_DEBUG_USER("Could not find/use resource 0x%08x.\n", (unsigned int) *id_loc); - return PTR_ERR(res); + return ret; } + needs_unref = true; - ret = vmw_execbuf_res_noref_val_add(sw_context, res, dirty); + ret = vmw_execbuf_res_val_add(sw_context, res, dirty, vmw_val_add_flag_none); if (unlikely(ret != 0)) - return ret; + goto res_check_done; if (rcache->valid && rcache->res == res) { rcache->valid_handle = true; @@ -707,7 +673,11 @@ vmw_cmd_res_check(struct vmw_private *dev_priv, if (p_res) *p_res = res; - return 0; +res_check_done: + if (needs_unref) + vmw_resource_unreference(&res); + + return ret; } /** @@ -720,7 +690,7 @@ vmw_cmd_res_check(struct vmw_private *dev_priv, static int vmw_rebind_all_dx_query(struct vmw_resource *ctx_res) { struct vmw_private *dev_priv = ctx_res->dev_priv; - struct vmw_buffer_object *dx_query_mob; + struct vmw_bo *dx_query_mob; VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdDXBindAllQuery); dx_query_mob = vmw_context_get_dx_query_mob(ctx_res); @@ -735,7 +705,7 @@ static int vmw_rebind_all_dx_query(struct vmw_resource *ctx_res) cmd->header.id = SVGA_3D_CMD_DX_BIND_ALL_QUERY; cmd->header.size = sizeof(cmd->body); cmd->body.cid = ctx_res->id; - cmd->body.mobid = dx_query_mob->base.resource->start; + cmd->body.mobid = dx_query_mob->tbo.resource->start; vmw_cmd_commit(dev_priv, sizeof(*cmd)); vmw_context_bind_dx_query(ctx_res, dx_query_mob); @@ -1034,7 +1004,7 @@ static int vmw_cmd_present_check(struct vmw_private *dev_priv, * after successful submission of the current command batch. */ static int vmw_query_bo_switch_prepare(struct vmw_private *dev_priv, - struct vmw_buffer_object *new_query_bo, + struct vmw_bo *new_query_bo, struct vmw_sw_context *sw_context) { struct vmw_res_cache_entry *ctx_entry = @@ -1046,24 +1016,24 @@ static int vmw_query_bo_switch_prepare(struct vmw_private *dev_priv, if (unlikely(new_query_bo != sw_context->cur_query_bo)) { - if (unlikely(new_query_bo->base.resource->num_pages > 4)) { + if (unlikely(PFN_UP(new_query_bo->tbo.resource->size) > 4)) { VMW_DEBUG_USER("Query buffer too large.\n"); return -EINVAL; } if (unlikely(sw_context->cur_query_bo != NULL)) { sw_context->needs_post_query_barrier = true; + vmw_bo_placement_set_default_accelerated(sw_context->cur_query_bo); ret = vmw_validation_add_bo(sw_context->ctx, - sw_context->cur_query_bo, - dev_priv->has_mob, false); + sw_context->cur_query_bo); if (unlikely(ret != 0)) return ret; } sw_context->cur_query_bo = new_query_bo; + vmw_bo_placement_set_default_accelerated(dev_priv->dummy_query_bo); ret = vmw_validation_add_bo(sw_context->ctx, - dev_priv->dummy_query_bo, - dev_priv->has_mob, false); + dev_priv->dummy_query_bo); if (unlikely(ret != 0)) return ret; } @@ -1162,22 +1132,23 @@ static void vmw_query_bo_switch_commit(struct vmw_private *dev_priv, static int vmw_translate_mob_ptr(struct vmw_private *dev_priv, struct vmw_sw_context *sw_context, SVGAMobId *id, - struct vmw_buffer_object **vmw_bo_p) + struct vmw_bo **vmw_bo_p) { - struct vmw_buffer_object *vmw_bo; + struct vmw_bo *vmw_bo, *tmp_bo; uint32_t handle = *id; struct vmw_relocation *reloc; int ret; vmw_validation_preload_bo(sw_context->ctx); - vmw_bo = vmw_user_bo_noref_lookup(sw_context->fp->tfile, handle); - if (IS_ERR(vmw_bo)) { - VMW_DEBUG_USER("Could not find or use MOB buffer.\n"); + ret = vmw_user_bo_lookup(sw_context->filp, handle, &vmw_bo); + if (ret != 0) { + drm_dbg(&dev_priv->drm, "Could not find or use MOB buffer.\n"); return PTR_ERR(vmw_bo); } - - ret = vmw_validation_add_bo(sw_context->ctx, vmw_bo, true, false); - vmw_user_bo_noref_release(); + vmw_bo_placement_set(vmw_bo, VMW_BO_DOMAIN_MOB, VMW_BO_DOMAIN_MOB); + ret = vmw_validation_add_bo(sw_context->ctx, vmw_bo); + tmp_bo = vmw_bo; + vmw_user_bo_unref(&tmp_bo); if (unlikely(ret != 0)) return ret; @@ -1217,22 +1188,24 @@ static int vmw_translate_mob_ptr(struct vmw_private *dev_priv, static int vmw_translate_guest_ptr(struct vmw_private *dev_priv, struct vmw_sw_context *sw_context, SVGAGuestPtr *ptr, - struct vmw_buffer_object **vmw_bo_p) + struct vmw_bo **vmw_bo_p) { - struct vmw_buffer_object *vmw_bo; + struct vmw_bo *vmw_bo, *tmp_bo; uint32_t handle = ptr->gmrId; struct vmw_relocation *reloc; int ret; vmw_validation_preload_bo(sw_context->ctx); - vmw_bo = vmw_user_bo_noref_lookup(sw_context->fp->tfile, handle); - if (IS_ERR(vmw_bo)) { - VMW_DEBUG_USER("Could not find or use GMR region.\n"); + ret = vmw_user_bo_lookup(sw_context->filp, handle, &vmw_bo); + if (ret != 0) { + drm_dbg(&dev_priv->drm, "Could not find or use GMR region.\n"); return PTR_ERR(vmw_bo); } - - ret = vmw_validation_add_bo(sw_context->ctx, vmw_bo, false, false); - vmw_user_bo_noref_release(); + vmw_bo_placement_set(vmw_bo, VMW_BO_DOMAIN_GMR | VMW_BO_DOMAIN_VRAM, + VMW_BO_DOMAIN_GMR | VMW_BO_DOMAIN_VRAM); + ret = vmw_validation_add_bo(sw_context->ctx, vmw_bo); + tmp_bo = vmw_bo; + vmw_user_bo_unref(&tmp_bo); if (unlikely(ret != 0)) return ret; @@ -1276,6 +1249,8 @@ static int vmw_cmd_dx_define_query(struct vmw_private *dev_priv, return -EINVAL; cotable_res = vmw_context_cotable(ctx_node->ctx, SVGA_COTABLE_DXQUERY); + if (IS_ERR_OR_NULL(cotable_res)) + return cotable_res ? PTR_ERR(cotable_res) : -EINVAL; ret = vmw_cotable_notify(cotable_res, cmd->body.queryId); return ret; @@ -1297,7 +1272,7 @@ static int vmw_cmd_dx_bind_query(struct vmw_private *dev_priv, SVGA3dCmdHeader *header) { VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdDXBindQuery); - struct vmw_buffer_object *vmw_bo; + struct vmw_bo *vmw_bo; int ret; cmd = container_of(header, typeof(*cmd), header); @@ -1380,7 +1355,7 @@ static int vmw_cmd_end_gb_query(struct vmw_private *dev_priv, struct vmw_sw_context *sw_context, SVGA3dCmdHeader *header) { - struct vmw_buffer_object *vmw_bo; + struct vmw_bo *vmw_bo; VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdEndGBQuery); int ret; @@ -1410,7 +1385,7 @@ static int vmw_cmd_end_query(struct vmw_private *dev_priv, struct vmw_sw_context *sw_context, SVGA3dCmdHeader *header) { - struct vmw_buffer_object *vmw_bo; + struct vmw_bo *vmw_bo; VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdEndQuery); int ret; @@ -1456,7 +1431,7 @@ static int vmw_cmd_wait_gb_query(struct vmw_private *dev_priv, struct vmw_sw_context *sw_context, SVGA3dCmdHeader *header) { - struct vmw_buffer_object *vmw_bo; + struct vmw_bo *vmw_bo; VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdWaitForGBQuery); int ret; @@ -1484,7 +1459,7 @@ static int vmw_cmd_wait_query(struct vmw_private *dev_priv, struct vmw_sw_context *sw_context, SVGA3dCmdHeader *header) { - struct vmw_buffer_object *vmw_bo; + struct vmw_bo *vmw_bo; VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdWaitForQuery); int ret; @@ -1521,7 +1496,8 @@ static int vmw_cmd_dma(struct vmw_private *dev_priv, struct vmw_sw_context *sw_context, SVGA3dCmdHeader *header) { - struct vmw_buffer_object *vmw_bo = NULL; + struct vmw_bo *vmw_bo = NULL; + struct vmw_resource *res; struct vmw_surface *srf = NULL; VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdSurfaceDMA); int ret; @@ -1545,7 +1521,7 @@ static int vmw_cmd_dma(struct vmw_private *dev_priv, return ret; /* Make sure DMA doesn't cross BO boundaries. */ - bo_size = vmw_bo->base.base.size; + bo_size = vmw_bo->tbo.base.size; if (unlikely(cmd->body.guest.ptr.offset > bo_size)) { VMW_DEBUG_USER("Invalid DMA offset.\n"); return -EINVAL; @@ -1557,18 +1533,24 @@ static int vmw_cmd_dma(struct vmw_private *dev_priv, dirty = (cmd->body.transfer == SVGA3D_WRITE_HOST_VRAM) ? VMW_RES_DIRTY_SET : 0; - ret = vmw_cmd_res_check(dev_priv, sw_context, vmw_res_surface, - dirty, user_surface_converter, - &cmd->body.host.sid, NULL); + ret = vmw_cmd_res_check(dev_priv, sw_context, vmw_res_surface, dirty, + user_surface_converter, &cmd->body.host.sid, + NULL); if (unlikely(ret != 0)) { if (unlikely(ret != -ERESTARTSYS)) VMW_DEBUG_USER("could not find surface for DMA.\n"); return ret; } - srf = vmw_res_to_srf(sw_context->res_cache[vmw_res_surface].res); + res = sw_context->res_cache[vmw_res_surface].res; + if (!res) { + VMW_DEBUG_USER("Invalid DMA surface.\n"); + return -EINVAL; + } - vmw_kms_cursor_snoop(srf, sw_context->fp->tfile, &vmw_bo->base, header); + srf = vmw_res_to_srf(res); + vmw_kms_cursor_snoop(srf, sw_context->fp->tfile, &vmw_bo->tbo, + header); return 0; } @@ -1631,7 +1613,7 @@ static int vmw_cmd_tex_state(struct vmw_private *dev_priv, { VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdSetTextureState); SVGA3dTextureState *last_state = (SVGA3dTextureState *) - ((unsigned long) header + header->size + sizeof(header)); + ((unsigned long) header + header->size + sizeof(*header)); SVGA3dTextureState *cur_state = (SVGA3dTextureState *) ((unsigned long) header + sizeof(*cmd)); struct vmw_resource *ctx; @@ -1687,7 +1669,7 @@ static int vmw_cmd_check_define_gmrfb(struct vmw_private *dev_priv, struct vmw_sw_context *sw_context, void *buf) { - struct vmw_buffer_object *vmw_bo; + struct vmw_bo *vmw_bo; struct { uint32_t header; @@ -1718,7 +1700,7 @@ static int vmw_cmd_res_switch_backup(struct vmw_private *dev_priv, struct vmw_resource *res, uint32_t *buf_id, unsigned long backup_offset) { - struct vmw_buffer_object *vbo; + struct vmw_bo *vbo; void *info; int ret; @@ -2002,7 +1984,7 @@ static int vmw_cmd_set_shader(struct vmw_private *dev_priv, cmd = container_of(header, typeof(*cmd), header); - if (cmd->body.type >= SVGA3D_SHADERTYPE_PREDX_MAX) { + if (!vmw_shadertype_is_valid(VMW_SM_LEGACY, cmd->body.type)) { VMW_DEBUG_USER("Illegal shader type %u.\n", (unsigned int) cmd->body.type); return -EINVAL; @@ -2026,8 +2008,9 @@ static int vmw_cmd_set_shader(struct vmw_private *dev_priv, res = vmw_shader_lookup(vmw_context_res_man(ctx), cmd->body.shid, cmd->body.type); if (!IS_ERR(res)) { - ret = vmw_execbuf_res_noctx_val_add(sw_context, res, - VMW_RES_DIRTY_NONE); + ret = vmw_execbuf_res_val_add(sw_context, res, + VMW_RES_DIRTY_NONE, + vmw_val_add_flag_noctx); if (unlikely(ret != 0)) return ret; @@ -2124,8 +2107,6 @@ vmw_cmd_dx_set_single_constant_buffer(struct vmw_private *dev_priv, SVGA3dCmdHeader *header) { VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdDXSetSingleConstantBuffer); - SVGA3dShaderType max_shader_num = has_sm5_context(dev_priv) ? - SVGA3D_NUM_SHADERTYPE : SVGA3D_NUM_SHADERTYPE_DX10; struct vmw_resource *res = NULL; struct vmw_ctx_validation_info *ctx_node = VMW_GET_CTX_NODE(sw_context); @@ -2142,6 +2123,14 @@ vmw_cmd_dx_set_single_constant_buffer(struct vmw_private *dev_priv, if (unlikely(ret != 0)) return ret; + if (!vmw_shadertype_is_valid(dev_priv->sm_type, cmd->body.type) || + cmd->body.slot >= SVGA3D_DX_MAX_CONSTBUFFERS) { + VMW_DEBUG_USER("Illegal const buffer shader %u slot %u.\n", + (unsigned int) cmd->body.type, + (unsigned int) cmd->body.slot); + return -EINVAL; + } + binding.bi.ctx = ctx_node->ctx; binding.bi.res = res; binding.bi.bt = vmw_ctx_binding_cb; @@ -2150,16 +2139,46 @@ vmw_cmd_dx_set_single_constant_buffer(struct vmw_private *dev_priv, binding.size = cmd->body.sizeInBytes; binding.slot = cmd->body.slot; - if (binding.shader_slot >= max_shader_num || - binding.slot >= SVGA3D_DX_MAX_CONSTBUFFERS) { - VMW_DEBUG_USER("Illegal const buffer shader %u slot %u.\n", - (unsigned int) cmd->body.type, - (unsigned int) binding.slot); + vmw_binding_add(ctx_node->staged, &binding.bi, binding.shader_slot, + binding.slot); + + return 0; +} + +/** + * vmw_cmd_dx_set_constant_buffer_offset - Validate + * SVGA_3D_CMD_DX_SET_VS/PS/GS/HS/DS/CS_CONSTANT_BUFFER_OFFSET command. + * + * @dev_priv: Pointer to a device private struct. + * @sw_context: The software context being used for this batch. + * @header: Pointer to the command header in the command stream. + */ +static int +vmw_cmd_dx_set_constant_buffer_offset(struct vmw_private *dev_priv, + struct vmw_sw_context *sw_context, + SVGA3dCmdHeader *header) +{ + VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdDXSetConstantBufferOffset); + + struct vmw_ctx_validation_info *ctx_node = VMW_GET_CTX_NODE(sw_context); + u32 shader_slot; + + if (!has_sm5_context(dev_priv)) + return -EINVAL; + + if (!ctx_node) + return -EINVAL; + + cmd = container_of(header, typeof(*cmd), header); + if (cmd->body.slot >= SVGA3D_DX_MAX_CONSTBUFFERS) { + VMW_DEBUG_USER("Illegal const buffer slot %u.\n", + (unsigned int) cmd->body.slot); return -EINVAL; } - vmw_binding_add(ctx_node->staged, &binding.bi, binding.shader_slot, - binding.slot); + shader_slot = cmd->header.id - SVGA_3D_CMD_DX_SET_VS_CONSTANT_BUFFER_OFFSET; + vmw_binding_cb_offset_update(ctx_node->staged, shader_slot, + cmd->body.slot, cmd->body.offsetInBytes); return 0; } @@ -2178,15 +2197,13 @@ static int vmw_cmd_dx_set_shader_res(struct vmw_private *dev_priv, { VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdDXSetShaderResources) = container_of(header, typeof(*cmd), header); - SVGA3dShaderType max_allowed = has_sm5_context(dev_priv) ? - SVGA3D_SHADERTYPE_MAX : SVGA3D_SHADERTYPE_DX10_MAX; u32 num_sr_view = (cmd->header.size - sizeof(cmd->body)) / sizeof(SVGA3dShaderResourceViewId); if ((u64) cmd->body.startView + (u64) num_sr_view > (u64) SVGA3D_DX_MAX_SRVIEWS || - cmd->body.type >= max_allowed) { + !vmw_shadertype_is_valid(dev_priv->sm_type, cmd->body.type)) { VMW_DEBUG_USER("Invalid shader binding.\n"); return -EINVAL; } @@ -2210,8 +2227,6 @@ static int vmw_cmd_dx_set_shader(struct vmw_private *dev_priv, SVGA3dCmdHeader *header) { VMW_DECLARE_CMD_VAR(*cmd, SVGA3dCmdDXSetShader); - SVGA3dShaderType max_allowed = has_sm5_context(dev_priv) ? - SVGA3D_SHADERTYPE_MAX : SVGA3D_SHADERTYPE_DX10_MAX; struct vmw_resource *res = NULL; struct vmw_ctx_validation_info *ctx_node = VMW_GET_CTX_NODE(sw_context); struct vmw_ctx_bindinfo_shader binding; @@ -2222,8 +2237,7 @@ static int vmw_cmd_dx_set_shader(struct vmw_private *dev_priv, cmd = container_of(header, typeof(*cmd), header); - if (cmd->body.type >= max_allowed || - cmd->body.type < SVGA3D_SHADERTYPE_MIN) { + if (!vmw_shadertype_is_valid(dev_priv->sm_type, cmd->body.type)) { VMW_DEBUG_USER("Illegal shader type %u.\n", (unsigned int) cmd->body.type); return -EINVAL; @@ -2236,8 +2250,9 @@ static int vmw_cmd_dx_set_shader(struct vmw_private *dev_priv, return PTR_ERR(res); } - ret = vmw_execbuf_res_noctx_val_add(sw_context, res, - VMW_RES_DIRTY_NONE); + ret = vmw_execbuf_res_val_add(sw_context, res, + VMW_RES_DIRTY_NONE, + vmw_val_add_flag_noctx); if (ret) return ret; } @@ -2364,7 +2379,7 @@ static int vmw_cmd_dx_set_rendertargets(struct vmw_private *dev_priv, sizeof(SVGA3dRenderTargetViewId); int ret; - if (num_rt_view > SVGA3D_MAX_SIMULTANEOUS_RENDER_TARGETS) { + if (num_rt_view > SVGA3D_DX_MAX_RENDER_TARGETS) { VMW_DEBUG_USER("Invalid DX Rendertarget binding.\n"); return -EINVAL; } @@ -2461,6 +2476,8 @@ static int vmw_cmd_dx_view_define(struct vmw_private *dev_priv, return ret; res = vmw_context_cotable(ctx_node->ctx, vmw_view_cotables[view_type]); + if (IS_ERR_OR_NULL(res)) + return res ? PTR_ERR(res) : -EINVAL; ret = vmw_cotable_notify(res, cmd->defined_id); if (unlikely(ret != 0)) return ret; @@ -2546,6 +2563,8 @@ static int vmw_cmd_dx_so_define(struct vmw_private *dev_priv, so_type = vmw_so_cmd_to_type(header->id); res = vmw_context_cotable(ctx_node->ctx, vmw_so_cotables[so_type]); + if (IS_ERR_OR_NULL(res)) + return res ? PTR_ERR(res) : -EINVAL; cmd = container_of(header, typeof(*cmd), header); ret = vmw_cotable_notify(res, cmd->defined_id); @@ -2664,6 +2683,8 @@ static int vmw_cmd_dx_define_shader(struct vmw_private *dev_priv, return -EINVAL; res = vmw_context_cotable(ctx_node->ctx, SVGA_COTABLE_DXSHADER); + if (IS_ERR_OR_NULL(res)) + return res ? PTR_ERR(res) : -EINVAL; ret = vmw_cotable_notify(res, cmd->body.shaderId); if (ret) return ret; @@ -2738,8 +2759,8 @@ static int vmw_cmd_dx_bind_shader(struct vmw_private *dev_priv, return PTR_ERR(res); } - ret = vmw_execbuf_res_noctx_val_add(sw_context, res, - VMW_RES_DIRTY_NONE); + ret = vmw_execbuf_res_val_add(sw_context, res, VMW_RES_DIRTY_NONE, + vmw_val_add_flag_noctx); if (ret) { VMW_DEBUG_USER("Error creating resource validation node.\n"); return ret; @@ -2915,7 +2936,7 @@ static int vmw_cmd_set_uav(struct vmw_private *dev_priv, if (!has_sm5_context(dev_priv)) return -EINVAL; - if (num_uav > SVGA3D_MAX_UAVIEWS) { + if (num_uav > vmw_max_num_uavs(dev_priv)) { VMW_DEBUG_USER("Invalid UAV binding.\n"); return -EINVAL; } @@ -2947,7 +2968,7 @@ static int vmw_cmd_set_cs_uav(struct vmw_private *dev_priv, if (!has_sm5_context(dev_priv)) return -EINVAL; - if (num_uav > SVGA3D_MAX_UAVIEWS) { + if (num_uav > vmw_max_num_uavs(dev_priv)) { VMW_DEBUG_USER("Invalid UAV binding.\n"); return -EINVAL; } @@ -2985,6 +3006,8 @@ static int vmw_cmd_dx_define_streamoutput(struct vmw_private *dev_priv, } res = vmw_context_cotable(ctx_node->ctx, SVGA_COTABLE_STREAMOUTPUT); + if (IS_ERR_OR_NULL(res)) + return res ? PTR_ERR(res) : -EINVAL; ret = vmw_cotable_notify(res, cmd->body.soid); if (ret) return ret; @@ -3059,8 +3082,8 @@ static int vmw_cmd_dx_bind_streamoutput(struct vmw_private *dev_priv, vmw_dx_streamoutput_set_size(res, cmd->body.sizeInBytes); - ret = vmw_execbuf_res_noctx_val_add(sw_context, res, - VMW_RES_DIRTY_NONE); + ret = vmw_execbuf_res_val_add(sw_context, res, VMW_RES_DIRTY_NONE, + vmw_val_add_flag_noctx); if (ret) { DRM_ERROR("Error creating resource validation node.\n"); return ret; @@ -3109,8 +3132,8 @@ static int vmw_cmd_dx_set_streamoutput(struct vmw_private *dev_priv, return 0; } - ret = vmw_execbuf_res_noctx_val_add(sw_context, res, - VMW_RES_DIRTY_NONE); + ret = vmw_execbuf_res_val_add(sw_context, res, VMW_RES_DIRTY_NONE, + vmw_val_add_flag_noctx); if (ret) { DRM_ERROR("Error creating resource validation node.\n"); return ret; @@ -3525,6 +3548,24 @@ static const struct vmw_cmd_entry vmw_cmd_entries[SVGA_3D_CMD_MAX] = { VMW_CMD_DEF(SVGA_3D_CMD_DX_TRANSFER_FROM_BUFFER, &vmw_cmd_dx_transfer_from_buffer, true, false, true), + VMW_CMD_DEF(SVGA_3D_CMD_DX_SET_VS_CONSTANT_BUFFER_OFFSET, + &vmw_cmd_dx_set_constant_buffer_offset, + true, false, true), + VMW_CMD_DEF(SVGA_3D_CMD_DX_SET_PS_CONSTANT_BUFFER_OFFSET, + &vmw_cmd_dx_set_constant_buffer_offset, + true, false, true), + VMW_CMD_DEF(SVGA_3D_CMD_DX_SET_GS_CONSTANT_BUFFER_OFFSET, + &vmw_cmd_dx_set_constant_buffer_offset, + true, false, true), + VMW_CMD_DEF(SVGA_3D_CMD_DX_SET_HS_CONSTANT_BUFFER_OFFSET, + &vmw_cmd_dx_set_constant_buffer_offset, + true, false, true), + VMW_CMD_DEF(SVGA_3D_CMD_DX_SET_DS_CONSTANT_BUFFER_OFFSET, + &vmw_cmd_dx_set_constant_buffer_offset, + true, false, true), + VMW_CMD_DEF(SVGA_3D_CMD_DX_SET_CS_CONSTANT_BUFFER_OFFSET, + &vmw_cmd_dx_set_constant_buffer_offset, + true, false, true), VMW_CMD_DEF(SVGA_3D_CMD_INTRA_SURFACE_COPY, &vmw_cmd_intra_surface_copy, true, false, true), @@ -3558,6 +3599,10 @@ static const struct vmw_cmd_entry vmw_cmd_entries[SVGA_3D_CMD_MAX] = { &vmw_cmd_dx_define_streamoutput, true, false, true), VMW_CMD_DEF(SVGA_3D_CMD_DX_BIND_STREAMOUTPUT, &vmw_cmd_dx_bind_streamoutput, true, false, true), + VMW_CMD_DEF(SVGA_3D_CMD_DX_DEFINE_RASTERIZER_STATE_V2, + &vmw_cmd_dx_so_define, true, false, true), + VMW_CMD_DEF(SVGA_3D_CMD_DEFINE_GB_SURFACE_V4, + &vmw_cmd_invalid, false, false, true), }; bool vmw_cmd_describe(const void *buf, u32 *size, char const **cmd) @@ -3623,6 +3668,11 @@ static int vmw_cmd_check(struct vmw_private *dev_priv, cmd_id = header->id; + if (header->size > SVGA_CMD_MAX_DATASIZE) { + VMW_DEBUG_USER("SVGA3D command: %d is too big.\n", + cmd_id + SVGA_3D_CMD_BASE); + return -E2BIG; + } *size = header->size + sizeof(SVGA3dCmdHeader); cmd_id -= SVGA_3D_CMD_BASE; @@ -3709,7 +3759,7 @@ static void vmw_apply_relocations(struct vmw_sw_context *sw_context) struct ttm_buffer_object *bo; list_for_each_entry(reloc, &sw_context->bo_relocations, head) { - bo = &reloc->vbo->base; + bo = &reloc->vbo->tbo; switch (bo->resource->mem_type) { case TTM_PL_VRAM: reloc->location->offset += bo->resource->start << PAGE_SHIFT; @@ -3810,7 +3860,6 @@ int vmw_execbuf_fence_commands(struct drm_file *file_priv, * @fence: Pointer to the fenc object. * @fence_handle: User-space fence handle. * @out_fence_fd: exported file descriptor for the fence. -1 if not used - * @sync_file: Only used to clean up in case of an error in this function. * * This function copies fence information to user-space. If copying fails, the * user-space struct drm_vmw_fence_rep::error member is hopefully left @@ -3820,17 +3869,17 @@ int vmw_execbuf_fence_commands(struct drm_file *file_priv, * Also if copying fails, user-space will be unable to signal the fence object * so we wait for it immediately, and then unreference the user-space reference. */ -void +int vmw_execbuf_copy_fence_user(struct vmw_private *dev_priv, struct vmw_fpriv *vmw_fp, int ret, struct drm_vmw_fence_rep __user *user_fence_rep, struct vmw_fence_obj *fence, uint32_t fence_handle, - int32_t out_fence_fd, struct sync_file *sync_file) + int32_t out_fence_fd) { struct drm_vmw_fence_rep fence_rep; if (user_fence_rep == NULL) - return; + return 0; memset(&fence_rep, 0, sizeof(fence_rep)); @@ -3841,8 +3890,7 @@ vmw_execbuf_copy_fence_user(struct vmw_private *dev_priv, fence_rep.handle = fence_handle; fence_rep.seqno = fence->base.seqno; - vmw_update_seqno(dev_priv); - fence_rep.passed_seqno = dev_priv->last_read_seqno; + fence_rep.passed_seqno = vmw_fences_update(dev_priv->fman); } /* @@ -3858,20 +3906,13 @@ vmw_execbuf_copy_fence_user(struct vmw_private *dev_priv, * handle. */ if (unlikely(ret != 0) && (fence_rep.error == 0)) { - if (sync_file) - fput(sync_file->file); - - if (fence_rep.fd != -1) { - put_unused_fd(fence_rep.fd); - fence_rep.fd = -1; - } - - ttm_ref_object_base_unref(vmw_fp->tfile, fence_handle, - TTM_REF_USAGE); + ttm_ref_object_base_unref(vmw_fp->tfile, fence_handle); VMW_DEBUG_USER("Fence copy error. Syncing.\n"); (void) vmw_fence_obj_wait(fence, false, false, VMW_FENCE_WAIT_TIMEOUT); } + + return ret ? -EFAULT : 0; } /** @@ -4015,22 +4056,26 @@ static int vmw_execbuf_tie_context(struct vmw_private *dev_priv, if (ret) return ret; - res = vmw_user_resource_noref_lookup_handle + ret = vmw_user_resource_lookup_handle (dev_priv, sw_context->fp->tfile, handle, - user_context_converter); - if (IS_ERR(res)) { + user_context_converter, &res); + if (ret != 0) { VMW_DEBUG_USER("Could not find or user DX context 0x%08x.\n", (unsigned int) handle); - return PTR_ERR(res); + return ret; } - ret = vmw_execbuf_res_noref_val_add(sw_context, res, VMW_RES_DIRTY_SET); - if (unlikely(ret != 0)) + ret = vmw_execbuf_res_val_add(sw_context, res, VMW_RES_DIRTY_SET, + vmw_val_add_flag_none); + if (unlikely(ret != 0)) { + vmw_resource_unreference(&res); return ret; + } sw_context->dx_ctx_node = vmw_execbuf_info_from_res(sw_context, res); sw_context->man = vmw_context_res_man(res); + vmw_resource_unreference(&res); return 0; } @@ -4049,9 +4094,7 @@ int vmw_execbuf_process(struct drm_file *file_priv, int ret; int32_t out_fence_fd = -1; struct sync_file *sync_file = NULL; - DECLARE_VAL_CONTEXT(val_ctx, &sw_context->res_ht, 1); - - vmw_validation_set_val_mem(&val_ctx, &dev_priv->vvm); + DECLARE_VAL_CONTEXT(val_ctx, sw_context, 1); if (flags & DRM_VMW_EXECBUF_FLAG_EXPORT_FENCE_FD) { out_fence_fd = get_unused_fd_flags(O_CLOEXEC); @@ -4098,6 +4141,7 @@ int vmw_execbuf_process(struct drm_file *file_priv, sw_context->kernel = true; } + sw_context->filp = file_priv; sw_context->fp = vmw_fpriv(file_priv); INIT_LIST_HEAD(&sw_context->ctx_list); sw_context->cur_query_bo = dev_priv->pinned_bo; @@ -4113,14 +4157,6 @@ int vmw_execbuf_process(struct drm_file *file_priv, if (sw_context->staged_bindings) vmw_binding_state_reset(sw_context->staged_bindings); - if (!sw_context->res_ht_initialized) { - ret = drm_ht_create(&sw_context->res_ht, VMW_RES_HT_ORDER); - if (unlikely(ret != 0)) - goto out_unlock; - - sw_context->res_ht_initialized = true; - } - INIT_LIST_HEAD(&sw_context->staged_cmd_res); sw_context->ctx = &val_ctx; ret = vmw_execbuf_tie_context(dev_priv, sw_context, dx_context_handle); @@ -4209,16 +4245,23 @@ int vmw_execbuf_process(struct drm_file *file_priv, (void) vmw_fence_obj_wait(fence, false, false, VMW_FENCE_WAIT_TIMEOUT); + } + } + + ret = vmw_execbuf_copy_fence_user(dev_priv, vmw_fpriv(file_priv), ret, + user_fence_rep, fence, handle, out_fence_fd); + + if (sync_file) { + if (ret) { + /* usercopy of fence failed, put the file object */ + fput(sync_file->file); + put_unused_fd(out_fence_fd); } else { /* Link the fence with the FD created earlier */ fd_install(out_fence_fd, sync_file->file); } } - vmw_execbuf_copy_fence_user(dev_priv, vmw_fpriv(file_priv), ret, - user_fence_rep, fence, handle, out_fence_fd, - sync_file); - /* Don't unreference when handing fence out */ if (unlikely(out_fence != NULL)) { *out_fence = fence; @@ -4236,7 +4279,7 @@ int vmw_execbuf_process(struct drm_file *file_priv, */ vmw_validation_unref_lists(&val_ctx); - return 0; + return ret; out_unlock_binding: mutex_unlock(&dev_priv->binding_mutex); @@ -4325,13 +4368,17 @@ void __vmw_execbuf_release_pinned_bo(struct vmw_private *dev_priv, if (dev_priv->pinned_bo == NULL) goto out_unlock; - ret = vmw_validation_add_bo(&val_ctx, dev_priv->pinned_bo, false, - false); + vmw_bo_placement_set(dev_priv->pinned_bo, + VMW_BO_DOMAIN_GMR | VMW_BO_DOMAIN_VRAM, + VMW_BO_DOMAIN_GMR | VMW_BO_DOMAIN_VRAM); + ret = vmw_validation_add_bo(&val_ctx, dev_priv->pinned_bo); if (ret) goto out_no_reserve; - ret = vmw_validation_add_bo(&val_ctx, dev_priv->dummy_query_bo, false, - false); + vmw_bo_placement_set(dev_priv->dummy_query_bo, + VMW_BO_DOMAIN_GMR | VMW_BO_DOMAIN_VRAM, + VMW_BO_DOMAIN_GMR | VMW_BO_DOMAIN_VRAM); + ret = vmw_validation_add_bo(&val_ctx, dev_priv->dummy_query_bo); if (ret) goto out_no_reserve; @@ -4406,6 +4453,9 @@ int vmw_execbuf_ioctl(struct drm_device *dev, void *data, int ret; struct dma_fence *in_fence = NULL; + MKS_STAT_TIME_DECL(MKSSTAT_KERN_EXECBUF); + MKS_STAT_TIME_PUSH(MKSSTAT_KERN_EXECBUF); + /* * Extend the ioctl argument while maintaining backwards compatibility: * We take different code paths depending on the value of arg->version. @@ -4415,7 +4465,8 @@ int vmw_execbuf_ioctl(struct drm_device *dev, void *data, if (unlikely(arg->version > DRM_VMW_EXECBUF_VERSION || arg->version == 0)) { VMW_DEBUG_USER("Incorrect execbuf version.\n"); - return -EINVAL; + ret = -EINVAL; + goto mksstats_out; } switch (arg->version) { @@ -4435,10 +4486,11 @@ int vmw_execbuf_ioctl(struct drm_device *dev, void *data, if (!in_fence) { VMW_DEBUG_USER("Cannot get imported fence\n"); - return -EINVAL; + ret = -EINVAL; + goto mksstats_out; } - ret = vmw_wait_dma_fence(dev_priv->fman, in_fence); + ret = dma_fence_wait(in_fence, true); if (ret) goto out; } @@ -4453,10 +4505,11 @@ int vmw_execbuf_ioctl(struct drm_device *dev, void *data, if (unlikely(ret != 0)) goto out; - vmw_kms_cursor_post_execbuf(dev_priv); - out: if (in_fence) dma_fence_put(in_fence); + +mksstats_out: + MKS_STAT_TIME_POP(MKSSTAT_KERN_EXECBUF); return ret; } |