From patchwork Fri Nov 15 11:45:46 2019 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: Abdiel Janulgue X-Patchwork-Id: 11246139 Return-Path: Received: from mail.kernel.org (pdx-korg-mail-1.web.codeaurora.org [172.30.200.123]) by pdx-korg-patchwork-2.web.codeaurora.org (Postfix) with ESMTP id 4B6606C1 for ; Fri, 15 Nov 2019 11:45:57 +0000 (UTC) Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id 3491B2072D for ; Fri, 15 Nov 2019 11:45:57 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 3491B2072D Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=linux.intel.com Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=intel-gfx-bounces@lists.freedesktop.org Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id BF8EA6E4CF; Fri, 15 Nov 2019 11:45:55 +0000 (UTC) X-Original-To: intel-gfx@lists.freedesktop.org Delivered-To: intel-gfx@lists.freedesktop.org Received: from mga04.intel.com (mga04.intel.com [192.55.52.120]) by gabe.freedesktop.org (Postfix) with ESMTPS id 6D4F86E4BB for ; Fri, 15 Nov 2019 11:45:54 +0000 (UTC) X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from fmsmga003.fm.intel.com ([10.253.24.29]) by fmsmga104.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 15 Nov 2019 03:45:54 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.68,308,1569308400"; d="scan'208";a="257738209" Received: from thijsmet-mobl2.ger.corp.intel.com (HELO skylake-nuc.ger.corp.intel.com) ([10.249.254.150]) by FMSMGA003.fm.intel.com with ESMTP; 15 Nov 2019 03:45:52 -0800 From: Abdiel Janulgue To: intel-gfx@lists.freedesktop.org Date: Fri, 15 Nov 2019 13:45:46 +0200 Message-Id: <20191115114549.23716-1-abdiel.janulgue@linux.intel.com> X-Mailer: git-send-email 2.23.0 MIME-Version: 1.0 Subject: [Intel-gfx] [PATCH 1/4] drm/i915: Allow i915 to manage the vma offset nodes instead of drm core X-BeenThere: intel-gfx@lists.freedesktop.org X-Mailman-Version: 2.1.23 Precedence: list List-Id: Intel graphics driver community testing & development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: Matthew Auld Errors-To: intel-gfx-bounces@lists.freedesktop.org Sender: "Intel-gfx" Have i915 replace the core drm_gem_mmap implementation to overcome its limitation in having only a single mmap offset node per gem object. This change allows us to have multiple mmap offsets per object and enables a mmapping instance to use unique fault-handlers per user vma. This allows i915 to store extra data within vma->vm_private_data and assign the pagefault ops for each mmap instance allowing objects to use multiple fault handlers depending on its backing storage. v2: - Fix race condition exposed by gem_mmap_gtt@close-race. Simplify lifetime management of the mmap offset objects be ensuring it is owned by the parent gem object instead of refcounting. - Track mmo used by fencing to Avoid locking when revoking mmaps during GPU reset. - Rebase. v3: - Simplify mmo tracking v4: - use vma->mmo in __i915_gem_object_release_mmap_gtt v5: - Revoke CPU mmaps on i915_gem_object_unbind() since unlike GTT mmaps, they don't have bound i915_vmas objects. Rebase. v6: Minor tweaks, header re-org (Chris) Signed-off-by: Abdiel Janulgue Cc: Matthew Auld Cc: Joonas Lahtinen Cc: Chris Wilson --- drivers/gpu/drm/i915/gem/i915_gem_domain.c | 3 +- drivers/gpu/drm/i915/gem/i915_gem_mman.c | 243 +++++++++++++++--- drivers/gpu/drm/i915/gem/i915_gem_mman.h | 28 ++ drivers/gpu/drm/i915/gem/i915_gem_object.c | 14 + drivers/gpu/drm/i915/gem/i915_gem_object.h | 7 +- .../gpu/drm/i915/gem/i915_gem_object_types.h | 19 ++ drivers/gpu/drm/i915/gem/i915_gem_pages.c | 3 + drivers/gpu/drm/i915/gem/i915_gem_tiling.c | 1 + .../drm/i915/gem/selftests/i915_gem_mman.c | 21 +- drivers/gpu/drm/i915/gt/intel_reset.c | 7 +- drivers/gpu/drm/i915/i915_drv.c | 10 +- drivers/gpu/drm/i915/i915_drv.h | 3 - drivers/gpu/drm/i915/i915_gem.c | 3 +- drivers/gpu/drm/i915/i915_getparam.c | 1 + drivers/gpu/drm/i915/i915_vma.c | 3 +- drivers/gpu/drm/i915/i915_vma.h | 3 + 16 files changed, 313 insertions(+), 56 deletions(-) create mode 100644 drivers/gpu/drm/i915/gem/i915_gem_mman.h diff --git a/drivers/gpu/drm/i915/gem/i915_gem_domain.c b/drivers/gpu/drm/i915/gem/i915_gem_domain.c index e2af63af67ad..fd1c6f12b77a 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_domain.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_domain.c @@ -13,6 +13,7 @@ #include "i915_gem_object.h" #include "i915_vma.h" #include "i915_gem_lmem.h" +#include "i915_gem_mman.h" static void __i915_gem_object_flush_for_display(struct drm_i915_gem_object *obj) { @@ -255,7 +256,7 @@ int i915_gem_object_set_cache_level(struct drm_i915_gem_object *obj, } if (obj->userfault_count) - __i915_gem_object_release_mmap(obj); + __i915_gem_object_release_mmap_gtt(obj); /* * As we no longer need a fence for GTT access, diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/i915_gem_mman.c index d60973603cc1..e602dfb44532 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_mman.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.c @@ -14,6 +14,7 @@ #include "i915_gem_gtt.h" #include "i915_gem_ioctls.h" #include "i915_gem_object.h" +#include "i915_gem_mman.h" #include "i915_trace.h" #include "i915_vma.h" @@ -219,7 +220,8 @@ vm_fault_t i915_gem_fault(struct vm_fault *vmf) { #define MIN_CHUNK_PAGES (SZ_1M >> PAGE_SHIFT) struct vm_area_struct *area = vmf->vma; - struct drm_i915_gem_object *obj = to_intel_bo(area->vm_private_data); + struct i915_mmap_offset *priv = area->vm_private_data; + struct drm_i915_gem_object *obj = priv->obj; struct drm_device *dev = obj->base.dev; struct drm_i915_private *i915 = to_i915(dev); struct intel_runtime_pm *rpm = &i915->runtime_pm; @@ -312,6 +314,9 @@ vm_fault_t i915_gem_fault(struct vm_fault *vmf) list_add(&obj->userfault_link, &i915->ggtt.userfault_list); mutex_unlock(&i915->ggtt.vm.mutex); + /* Track the mmo associated with the fenced vma */ + vma->mmo = priv; + if (IS_ACTIVE(CONFIG_DRM_I915_USERFAULT_AUTOSUSPEND)) intel_wakeref_auto(&i915->ggtt.userfault_wakeref, msecs_to_jiffies_timeout(CONFIG_DRM_I915_USERFAULT_AUTOSUSPEND)); @@ -358,28 +363,19 @@ vm_fault_t i915_gem_fault(struct vm_fault *vmf) } } -void __i915_gem_object_release_mmap(struct drm_i915_gem_object *obj) +void __i915_gem_object_release_mmap_gtt(struct drm_i915_gem_object *obj) { struct i915_vma *vma; GEM_BUG_ON(!obj->userfault_count); - obj->userfault_count = 0; - list_del(&obj->userfault_link); - drm_vma_node_unmap(&obj->base.vma_node, - obj->base.dev->anon_inode->i_mapping); - for_each_ggtt_vma(vma, obj) - i915_vma_unset_userfault(vma); + i915_vma_revoke_mmap(vma); + + GEM_BUG_ON(obj->userfault_count); } /** - * i915_gem_object_release_mmap - remove physical page mappings - * @obj: obj in question - * - * Preserve the reservation of the mmapping with the DRM core code, but - * relinquish ownership of the pages back to the system. - * * It is vital that we remove the page mapping if we have mapped a tiled * object through the GTT and then lose the fence register due to * resource pressure. Similarly if the object has been moved out of the @@ -387,7 +383,7 @@ void __i915_gem_object_release_mmap(struct drm_i915_gem_object *obj) * mapping will then trigger a page fault on the next user access, allowing * fixup by i915_gem_fault(). */ -void i915_gem_object_release_mmap(struct drm_i915_gem_object *obj) +static void i915_gem_object_release_mmap_gtt(struct drm_i915_gem_object *obj) { struct drm_i915_private *i915 = to_i915(obj->base.dev); intel_wakeref_t wakeref; @@ -406,7 +402,7 @@ void i915_gem_object_release_mmap(struct drm_i915_gem_object *obj) if (!obj->userfault_count) goto out; - __i915_gem_object_release_mmap(obj); + __i915_gem_object_release_mmap_gtt(obj); /* Ensure that the CPU's PTE are revoked and there are not outstanding * memory transactions from userspace before we return. The TLB @@ -422,15 +418,63 @@ void i915_gem_object_release_mmap(struct drm_i915_gem_object *obj) intel_runtime_pm_put(&i915->runtime_pm, wakeref); } -static int create_mmap_offset(struct drm_i915_gem_object *obj) +void i915_gem_object_release_mmap_offset(struct drm_i915_gem_object *obj) +{ + struct i915_mmap_offset *mmo; + + mutex_lock(&obj->mmo_lock); + list_for_each_entry(mmo, &obj->mmap_offsets, offset) { + /* vma_node_unmap for GTT mmaps handled already in + * __i915_gem_object_release_mmap_gtt + */ + if (mmo->mmap_type != I915_MMAP_TYPE_GTT) + drm_vma_node_unmap(&mmo->vma_node, + obj->base.dev->anon_inode->i_mapping); + } + mutex_unlock(&obj->mmo_lock); +} + +/** + * i915_gem_object_release_mmap - remove physical page mappings + * @obj: obj in question + * + * Preserve the reservation of the mmapping with the DRM core code, but + * relinquish ownership of the pages back to the system. + */ +void i915_gem_object_release_mmap(struct drm_i915_gem_object *obj) +{ + i915_gem_object_release_mmap_gtt(obj); + i915_gem_object_release_mmap_offset(obj); +} + +static void init_mmap_offset(struct drm_i915_gem_object *obj, + struct i915_mmap_offset *mmo) +{ + mutex_lock(&obj->mmo_lock); + list_add(&mmo->offset, &obj->mmap_offsets); + mutex_unlock(&obj->mmo_lock); + + mmo->obj = obj; + mmo->dev = obj->base.dev; +} + +static int create_mmap_offset(struct drm_i915_gem_object *obj, + struct i915_mmap_offset *mmo) { struct drm_i915_private *i915 = to_i915(obj->base.dev); struct intel_gt *gt = &i915->gt; + struct drm_device *dev = obj->base.dev; int err; - err = drm_gem_create_mmap_offset(&obj->base); - if (likely(!err)) + drm_vma_node_reset(&mmo->vma_node); + if (mmo->file) + drm_vma_node_allow(&mmo->vma_node, mmo->file); + err = drm_vma_offset_add(dev->vma_offset_manager, &mmo->vma_node, + obj->base.size / PAGE_SIZE); + if (likely(!err)) { + init_mmap_offset(obj, mmo); return 0; + } /* Attempt to reap some mmap space from dead objects */ err = intel_gt_retire_requests_timeout(gt, MAX_SCHEDULE_TIMEOUT); @@ -438,34 +482,56 @@ static int create_mmap_offset(struct drm_i915_gem_object *obj) return err; i915_gem_drain_freed_objects(i915); - return drm_gem_create_mmap_offset(&obj->base); + err = drm_vma_offset_add(dev->vma_offset_manager, &mmo->vma_node, + obj->base.size / PAGE_SIZE); + if (err) + return err; + + init_mmap_offset(obj, mmo); + return 0; } -int -i915_gem_mmap_gtt(struct drm_file *file, - struct drm_device *dev, - u32 handle, - u64 *offset) +static int +__assign_gem_object_mmap_data(struct drm_file *file, + u32 handle, + enum i915_mmap_type mmap_type, + u64 *offset) { struct drm_i915_gem_object *obj; + struct i915_mmap_offset *mmo; int ret; - if (!i915_ggtt_has_aperture(&to_i915(dev)->ggtt)) - return -ENODEV; - obj = i915_gem_object_lookup(file, handle); if (!obj) return -ENOENT; - if (i915_gem_object_never_bind_ggtt(obj)) { + if (mmap_type == I915_MMAP_TYPE_GTT && + i915_gem_object_never_bind_ggtt(obj)) { ret = -ENODEV; goto out; } - ret = create_mmap_offset(obj); - if (ret == 0) - *offset = drm_vma_node_offset_addr(&obj->base.vma_node); + if (mmap_type != I915_MMAP_TYPE_GTT && + !i915_gem_object_has_struct_page(obj)) { + ret = -ENODEV; + goto out; + } + + mmo = kzalloc(sizeof(*mmo), GFP_KERNEL); + if (!mmo) { + ret = -ENOMEM; + goto out; + } + + mmo->file = file; + ret = create_mmap_offset(obj, mmo); + if (ret) { + kfree(mmo); + goto out; + } + mmo->mmap_type = mmap_type; + *offset = drm_vma_node_offset_addr(&mmo->vma_node); out: i915_gem_object_put(obj); return ret; @@ -492,7 +558,118 @@ i915_gem_mmap_gtt_ioctl(struct drm_device *dev, void *data, { struct drm_i915_gem_mmap_gtt *args = data; - return i915_gem_mmap_gtt(file, dev, args->handle, &args->offset); + return __assign_gem_object_mmap_data(file, args->handle, + I915_MMAP_TYPE_GTT, + &args->offset); +} + +void i915_mmap_offset_destroy(struct i915_mmap_offset *mmo, struct mutex *mutex) +{ + if (mmo->file) + drm_vma_node_revoke(&mmo->vma_node, mmo->file); + drm_vma_offset_remove(mmo->dev->vma_offset_manager, &mmo->vma_node); + + mutex_lock(mutex); + list_del(&mmo->offset); + mutex_unlock(mutex); + + kfree(mmo); +} + +static void i915_gem_vm_open(struct vm_area_struct *vma) +{ + struct i915_mmap_offset *mmo = vma->vm_private_data; + struct drm_i915_gem_object *obj = mmo->obj; + + GEM_BUG_ON(!obj); + i915_gem_object_get(obj); +} + +static void i915_gem_vm_close(struct vm_area_struct *vma) +{ + struct i915_mmap_offset *mmo = vma->vm_private_data; + struct drm_i915_gem_object *obj = mmo->obj; + + GEM_BUG_ON(!obj); + i915_gem_object_put(obj); +} + +static const struct vm_operations_struct i915_gem_gtt_vm_ops = { + .fault = i915_gem_fault, + .open = i915_gem_vm_open, + .close = i915_gem_vm_close, +}; + +/* This overcomes the limitation in drm_gem_mmap's assignment of a + * drm_gem_object as the vma->vm_private_data. Since we need to + * be able to resolve multiple mmap offsets which could be tied + * to a single gem object. + */ +int i915_gem_mmap(struct file *filp, struct vm_area_struct *vma) +{ + struct drm_vma_offset_node *node; + struct drm_file *priv = filp->private_data; + struct drm_device *dev = priv->minor->dev; + struct i915_mmap_offset *mmo = NULL; + struct drm_gem_object *obj = NULL; + + if (drm_dev_is_unplugged(dev)) + return -ENODEV; + + drm_vma_offset_lock_lookup(dev->vma_offset_manager); + node = drm_vma_offset_exact_lookup_locked(dev->vma_offset_manager, + vma->vm_pgoff, + vma_pages(vma)); + if (likely(node)) { + mmo = container_of(node, struct i915_mmap_offset, + vma_node); + /* + * In our dependency chain, the drm_vma_offset_node + * depends on the validity of the mmo, which depends on + * the gem object. However the only reference we have + * at this point is the mmo (as the parent of the node). + * Try to check if the gem object was at least cleared. + */ + if (!mmo || !mmo->obj) { + drm_vma_offset_unlock_lookup(dev->vma_offset_manager); + return -EINVAL; + } + /* + * Skip 0-refcnted objects as it is in the process of being + * destroyed and will be invalid when the vma manager lock + * is released. + */ + obj = &mmo->obj->base; + if (!kref_get_unless_zero(&obj->refcount)) + obj = NULL; + + } + drm_vma_offset_unlock_lookup(dev->vma_offset_manager); + + if (!obj) + return -EINVAL; + + if (!drm_vma_node_is_allowed(node, priv)) { + drm_gem_object_put_unlocked(obj); + return -EACCES; + } + + if (to_intel_bo(obj)->readonly) { + if (vma->vm_flags & VM_WRITE) { + drm_gem_object_put_unlocked(obj); + return -EINVAL; + } + vma->vm_flags &= ~VM_MAYWRITE; + } + + vma->vm_flags |= VM_PFNMAP | VM_DONTEXPAND | VM_DONTDUMP; + vma->vm_page_prot = pgprot_writecombine(vm_get_page_prot(vma->vm_flags)); + vma->vm_page_prot = pgprot_decrypted(vma->vm_page_prot); + vma->vm_private_data = mmo; + + vma->vm_ops = &i915_gem_gtt_vm_ops; + + return 0; } #if IS_ENABLED(CONFIG_DRM_I915_SELFTEST) diff --git a/drivers/gpu/drm/i915/gem/i915_gem_mman.h b/drivers/gpu/drm/i915/gem/i915_gem_mman.h new file mode 100644 index 000000000000..25a3c4d6cd65 --- /dev/null +++ b/drivers/gpu/drm/i915/gem/i915_gem_mman.h @@ -0,0 +1,28 @@ +/* + * SPDX-License-Identifier: MIT + * + * Copyright © 2019 Intel Corporation + */ + +#ifndef __I915_GEM_MMAN_H__ +#define __I915_GEM_MMAN_H__ + +#include +#include + +struct drm_device; +struct drm_file; +struct drm_i915_gem_object; +struct file; +struct i915_mmap_offset; +struct mutex; + +int i915_gem_mmap(struct file *filp, struct vm_area_struct *vma); +vm_fault_t i915_gem_fault(struct vm_fault *vmf); +void i915_mmap_offset_destroy(struct i915_mmap_offset *mmo, struct mutex *mutex); + +void __i915_gem_object_release_mmap_gtt(struct drm_i915_gem_object *obj); +void i915_gem_object_release_mmap(struct drm_i915_gem_object *obj); +void i915_gem_object_release_mmap_offset(struct drm_i915_gem_object *obj); + +#endif diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.c b/drivers/gpu/drm/i915/gem/i915_gem_object.c index db103d3c8760..c4f4f10c852d 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.c @@ -29,6 +29,7 @@ #include "i915_drv.h" #include "i915_gem_clflush.h" #include "i915_gem_context.h" +#include "i915_gem_mman.h" #include "i915_gem_object.h" #include "i915_globals.h" #include "i915_trace.h" @@ -61,6 +62,9 @@ void i915_gem_object_init(struct drm_i915_gem_object *obj, INIT_LIST_HEAD(&obj->lut_list); + mutex_init(&obj->mmo_lock); + INIT_LIST_HEAD(&obj->mmap_offsets); + init_rcu_head(&obj->rcu); obj->ops = ops; @@ -158,6 +162,8 @@ static void __i915_gem_free_objects(struct drm_i915_private *i915, wakeref = intel_runtime_pm_get(&i915->runtime_pm); llist_for_each_entry_safe(obj, on, freed, freed) { + struct i915_mmap_offset *mmo, *mn; + trace_i915_gem_object_destroy(obj); if (!list_empty(&obj->vma.list)) { @@ -183,6 +189,14 @@ static void __i915_gem_free_objects(struct drm_i915_private *i915, spin_unlock(&obj->vma.lock); } + i915_gem_object_release_mmap(obj); + + list_for_each_entry_safe(mmo, mn, &obj->mmap_offsets, offset) { + mmo->obj = NULL; + i915_mmap_offset_destroy(mmo, &obj->mmo_lock); + } + + GEM_BUG_ON(!list_empty(&obj->mmap_offsets)); GEM_BUG_ON(atomic_read(&obj->bind_count)); GEM_BUG_ON(obj->userfault_count); GEM_BUG_ON(!list_empty(&obj->lut_list)); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object.h b/drivers/gpu/drm/i915/gem/i915_gem_object.h index e5750d506cc9..9b61f18606e0 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object.h @@ -132,13 +132,13 @@ void i915_gem_object_unlock_fence(struct drm_i915_gem_object *obj, static inline void i915_gem_object_set_readonly(struct drm_i915_gem_object *obj) { - obj->base.vma_node.readonly = true; + obj->readonly = true; } static inline bool i915_gem_object_is_readonly(const struct drm_i915_gem_object *obj) { - return obj->base.vma_node.readonly; + return obj->readonly; } static inline bool @@ -387,9 +387,6 @@ static inline void i915_gem_object_unpin_map(struct drm_i915_gem_object *obj) i915_gem_object_unpin_pages(obj); } -void __i915_gem_object_release_mmap(struct drm_i915_gem_object *obj); -void i915_gem_object_release_mmap(struct drm_i915_gem_object *obj); - void i915_gem_object_flush_write_domain(struct drm_i915_gem_object *obj, unsigned int flush_domains); diff --git a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h index 15f8297dc34e..632d6b804cfb 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_object_types.h +++ b/drivers/gpu/drm/i915/gem/i915_gem_object_types.h @@ -63,6 +63,20 @@ struct drm_i915_gem_object_ops { void (*release)(struct drm_i915_gem_object *obj); }; +enum i915_mmap_type { + I915_MMAP_TYPE_GTT = 0, +}; + +struct i915_mmap_offset { + struct drm_device *dev; + struct drm_vma_offset_node vma_node; + struct drm_i915_gem_object *obj; + struct drm_file *file; + enum i915_mmap_type mmap_type; + + struct list_head offset; +}; + struct drm_i915_gem_object { struct drm_gem_object base; @@ -118,6 +132,11 @@ struct drm_i915_gem_object { unsigned int userfault_count; struct list_head userfault_link; + /* Protects access to mmap offsets */ + struct mutex mmo_lock; + struct list_head mmap_offsets; + bool readonly:1; + I915_SELFTEST_DECLARE(struct list_head st_link); unsigned long flags; diff --git a/drivers/gpu/drm/i915/gem/i915_gem_pages.c b/drivers/gpu/drm/i915/gem/i915_gem_pages.c index f402c2c415c2..75197ca696a8 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_pages.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_pages.c @@ -8,6 +8,7 @@ #include "i915_gem_object.h" #include "i915_scatterlist.h" #include "i915_gem_lmem.h" +#include "i915_gem_mman.h" void __i915_gem_object_set_pages(struct drm_i915_gem_object *obj, struct sg_table *pages, @@ -207,6 +208,8 @@ int __i915_gem_object_put_pages(struct drm_i915_gem_object *obj) goto unlock; } + i915_gem_object_release_mmap_offset(obj); + /* * ->put_pages might need to allocate memory for the bit17 swizzle * array, hence protect them from being reaped by removing them from gtt diff --git a/drivers/gpu/drm/i915/gem/i915_gem_tiling.c b/drivers/gpu/drm/i915/gem/i915_gem_tiling.c index 1fa592d82af5..6c7825a2dc2a 100644 --- a/drivers/gpu/drm/i915/gem/i915_gem_tiling.c +++ b/drivers/gpu/drm/i915/gem/i915_gem_tiling.c @@ -11,6 +11,7 @@ #include "i915_drv.h" #include "i915_gem.h" #include "i915_gem_ioctls.h" +#include "i915_gem_mman.h" #include "i915_gem_object.h" /** diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c index 9f1a69027a04..63e24f6b0d40 100644 --- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c +++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_mman.c @@ -564,15 +564,20 @@ static bool assert_mmap_offset(struct drm_i915_private *i915, int expected) { struct drm_i915_gem_object *obj; + /* Ownership transferred to parent gem object in create_mmap_offset */ + struct i915_mmap_offset *mmo = kzalloc(sizeof(*mmo), GFP_KERNEL); int err; obj = i915_gem_object_create_internal(i915, size); if (IS_ERR(obj)) return PTR_ERR(obj); - err = create_mmap_offset(obj); + err = create_mmap_offset(obj, mmo); + if (err) + kfree(mmo); i915_gem_object_put(obj); + return err == expected; } @@ -608,6 +613,8 @@ static int igt_mmap_offset_exhaustion(void *arg) struct drm_mm *mm = &i915->drm.vma_offset_manager->vm_addr_space_mm; struct drm_i915_gem_object *obj; struct drm_mm_node *hole, *next; + /* Ownership transferred to parent gem object in create_mmap_offset */ + struct i915_mmap_offset *mmo = kzalloc(sizeof(*mmo), GFP_KERNEL); int loop, err; /* Disable background reaper */ @@ -672,9 +679,10 @@ static int igt_mmap_offset_exhaustion(void *arg) goto out; } - err = create_mmap_offset(obj); + err = create_mmap_offset(obj, mmo); if (err) { pr_err("Unable to insert object into reclaimed hole\n"); + kfree(mmo); goto err_obj; } @@ -919,6 +927,15 @@ static int igt_mmap_gtt_revoke(void *arg) } GEM_BUG_ON(atomic_read(&obj->bind_count)); + if (type != I915_MMAP_TYPE_GTT) { + __i915_gem_object_put_pages(obj); + if (i915_gem_object_has_pages(obj)) { + pr_err("Failed to put-pages object!\n"); + err = -EINVAL; + goto out_unmap; + } + } + err = check_absent(addr, obj->base.size); if (err) goto out_unmap; diff --git a/drivers/gpu/drm/i915/gt/intel_reset.c b/drivers/gpu/drm/i915/gt/intel_reset.c index b7007cd78c6f..777eaa483676 100644 --- a/drivers/gpu/drm/i915/gt/intel_reset.c +++ b/drivers/gpu/drm/i915/gt/intel_reset.c @@ -678,8 +678,13 @@ static void revoke_mmaps(struct intel_gt *gt) continue; GEM_BUG_ON(vma->fence != >->ggtt->fence_regs[i]); - node = &vma->obj->base.vma_node; + + if (!vma->mmo) + continue; + + node = &vma->mmo->vma_node; vma_offset = vma->ggtt_view.partial.offset << PAGE_SHIFT; + unmap_mapping_range(gt->i915->drm.anon_inode->i_mapping, drm_vma_node_offset_addr(node) + vma_offset, vma->size, diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c index 8d7049792d62..5fe071640893 100644 --- a/drivers/gpu/drm/i915/i915_drv.c +++ b/drivers/gpu/drm/i915/i915_drv.c @@ -2662,18 +2662,12 @@ const struct dev_pm_ops i915_pm_ops = { .runtime_resume = intel_runtime_resume, }; -static const struct vm_operations_struct i915_gem_vm_ops = { - .fault = i915_gem_fault, - .open = drm_gem_vm_open, - .close = drm_gem_vm_close, -}; - static const struct file_operations i915_driver_fops = { .owner = THIS_MODULE, .open = drm_open, .release = drm_release, .unlocked_ioctl = drm_ioctl, - .mmap = drm_gem_mmap, + .mmap = i915_gem_mmap, .poll = drm_poll, .read = drm_read, .compat_ioctl = i915_compat_ioctl, @@ -2762,7 +2756,6 @@ static struct drm_driver driver = { .gem_close_object = i915_gem_close_object, .gem_free_object_unlocked = i915_gem_free_object, - .gem_vm_ops = &i915_gem_vm_ops, .prime_handle_to_fd = drm_gem_prime_handle_to_fd, .prime_fd_to_handle = drm_gem_prime_fd_to_handle, @@ -2773,7 +2766,6 @@ static struct drm_driver driver = { .get_scanout_position = i915_get_crtc_scanoutpos, .dumb_create = i915_gem_dumb_create, - .dumb_map_offset = i915_gem_mmap_gtt, .ioctls = i915_ioctls, .num_ioctls = ARRAY_SIZE(i915_ioctls), .fops = &i915_driver_fops, diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h index 1779f600fcfb..a37c77ba5c40 100644 --- a/drivers/gpu/drm/i915/i915_drv.h +++ b/drivers/gpu/drm/i915/i915_drv.h @@ -1861,8 +1861,6 @@ i915_mutex_lock_interruptible(struct drm_device *dev) int i915_gem_dumb_create(struct drm_file *file_priv, struct drm_device *dev, struct drm_mode_create_dumb *args); -int i915_gem_mmap_gtt(struct drm_file *file_priv, struct drm_device *dev, - u32 handle, u64 *offset); int i915_gem_mmap_gtt_version(void); int __must_check i915_gem_set_global_seqno(struct drm_device *dev, u32 seqno); @@ -1886,7 +1884,6 @@ void i915_gem_driver_release(struct drm_i915_private *dev_priv); void i915_gem_suspend(struct drm_i915_private *dev_priv); void i915_gem_suspend_late(struct drm_i915_private *dev_priv); void i915_gem_resume(struct drm_i915_private *dev_priv); -vm_fault_t i915_gem_fault(struct vm_fault *vmf); int i915_gem_open(struct drm_i915_private *i915, struct drm_file *file); void i915_gem_release(struct drm_device *dev, struct drm_file *file); diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c index 43c532756c7c..1b1aa7fc15ba 100644 --- a/drivers/gpu/drm/i915/i915_gem.c +++ b/drivers/gpu/drm/i915/i915_gem.c @@ -44,6 +44,7 @@ #include "gem/i915_gem_clflush.h" #include "gem/i915_gem_context.h" #include "gem/i915_gem_ioctls.h" +#include "gem/i915_gem_mman.h" #include "gem/i915_gem_pm.h" #include "gt/intel_engine_user.h" #include "gt/intel_gt.h" @@ -854,7 +855,7 @@ void i915_gem_runtime_suspend(struct drm_i915_private *i915) list_for_each_entry_safe(obj, on, &i915->ggtt.userfault_list, userfault_link) - __i915_gem_object_release_mmap(obj); + __i915_gem_object_release_mmap_gtt(obj); /* * The fence will be lost when the device powers down. If any were diff --git a/drivers/gpu/drm/i915/i915_getparam.c b/drivers/gpu/drm/i915/i915_getparam.c index cf8a8c3ef047..54fce81d5724 100644 --- a/drivers/gpu/drm/i915/i915_getparam.c +++ b/drivers/gpu/drm/i915/i915_getparam.c @@ -2,6 +2,7 @@ * SPDX-License-Identifier: MIT */ +#include "gem/i915_gem_mman.h" #include "gt/intel_engine_user.h" #include "i915_drv.h" diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c index e5512f26e20a..5108dcb2e5b1 100644 --- a/drivers/gpu/drm/i915/i915_vma.c +++ b/drivers/gpu/drm/i915/i915_vma.c @@ -1054,7 +1054,7 @@ static void __i915_vma_iounmap(struct i915_vma *vma) void i915_vma_revoke_mmap(struct i915_vma *vma) { - struct drm_vma_offset_node *node = &vma->obj->base.vma_node; + struct drm_vma_offset_node *node; u64 vma_offset; lockdep_assert_held(&vma->vm->mutex); @@ -1065,6 +1065,7 @@ void i915_vma_revoke_mmap(struct i915_vma *vma) GEM_BUG_ON(!i915_vma_is_map_and_fenceable(vma)); GEM_BUG_ON(!vma->obj->userfault_count); + node = &vma->mmo->vma_node; vma_offset = vma->ggtt_view.partial.offset << PAGE_SHIFT; unmap_mapping_range(vma->vm->i915->drm.anon_inode->i_mapping, drm_vma_node_offset_addr(node) + vma_offset, diff --git a/drivers/gpu/drm/i915/i915_vma.h b/drivers/gpu/drm/i915/i915_vma.h index 465932813bc5..f09f4f513c41 100644 --- a/drivers/gpu/drm/i915/i915_vma.h +++ b/drivers/gpu/drm/i915/i915_vma.h @@ -63,6 +63,9 @@ struct i915_vma { u64 display_alignment; struct i915_page_sizes page_sizes; + /* mmap-offset associated with fencing for this vma */ + struct i915_mmap_offset *mmo; + u32 fence_size; u32 fence_alignment;