[Intel-gfx] [PATCH] drm/i915: Move context management under GEM
Chris Wilson
chris at chris-wilson.co.uk
Thu Oct 3 13:35:49 UTC 2019
Keep track of the GEM contexts underneath i915->gem.contexts and assign
them their own lock for the purposes of list management.
v2: Focus on lock tracking; ctx->vm is protected by ctx->mutex
Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
---
Another missed ctx->mutex around __assign_ppgtt.
---
drivers/gpu/drm/i915/gem/i915_gem_context.c | 183 +++++++++---------
drivers/gpu/drm/i915/gem/i915_gem_context.h | 27 ++-
.../gpu/drm/i915/gem/i915_gem_context_types.h | 2 +-
.../gpu/drm/i915/gem/i915_gem_execbuffer.c | 2 +-
drivers/gpu/drm/i915/gem/i915_gem_userptr.c | 3 +-
.../gpu/drm/i915/gem/selftests/huge_pages.c | 36 ++--
.../drm/i915/gem/selftests/i915_gem_context.c | 168 ++++++++--------
.../gpu/drm/i915/gem/selftests/mock_context.c | 4 +-
drivers/gpu/drm/i915/gt/intel_context.c | 10 +-
drivers/gpu/drm/i915/gt/intel_lrc.c | 5 +
drivers/gpu/drm/i915/gt/selftest_context.c | 24 +--
drivers/gpu/drm/i915/gt/selftest_hangcheck.c | 39 ++--
drivers/gpu/drm/i915/gt/selftest_lrc.c | 6 +-
.../gpu/drm/i915/gt/selftest_workarounds.c | 20 +-
drivers/gpu/drm/i915/gvt/scheduler.c | 24 +--
drivers/gpu/drm/i915/i915_debugfs.c | 50 +++--
drivers/gpu/drm/i915/i915_drv.c | 2 -
drivers/gpu/drm/i915/i915_drv.h | 27 +--
drivers/gpu/drm/i915/i915_gem.c | 10 +-
drivers/gpu/drm/i915/i915_gem_gtt.c | 4 +-
drivers/gpu/drm/i915/i915_perf.c | 24 ++-
drivers/gpu/drm/i915/i915_sysfs.c | 77 ++++----
drivers/gpu/drm/i915/i915_trace.h | 2 +-
drivers/gpu/drm/i915/selftests/i915_gem.c | 8 -
.../gpu/drm/i915/selftests/i915_gem_evict.c | 3 -
drivers/gpu/drm/i915/selftests/i915_gem_gtt.c | 15 +-
drivers/gpu/drm/i915/selftests/i915_request.c | 12 +-
drivers/gpu/drm/i915/selftests/i915_vma.c | 7 +-
.../gpu/drm/i915/selftests/mock_gem_device.c | 6 +-
29 files changed, 411 insertions(+), 389 deletions(-)
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.c b/drivers/gpu/drm/i915/gem/i915_gem_context.c
index ae588544bf3a..dadb820b0502 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.c
@@ -218,9 +218,12 @@ static struct i915_gem_engines *default_engines(struct i915_gem_context *ctx)
static void i915_gem_context_free(struct i915_gem_context *ctx)
{
- lockdep_assert_held(&ctx->i915->drm.struct_mutex);
GEM_BUG_ON(!i915_gem_context_is_closed(ctx));
+ spin_lock(&ctx->i915->gem.contexts.lock);
+ list_del(&ctx->link);
+ spin_unlock(&ctx->i915->gem.contexts.lock);
+
free_engines(rcu_access_pointer(ctx->engines));
mutex_destroy(&ctx->engines_mutex);
@@ -230,67 +233,54 @@ static void i915_gem_context_free(struct i915_gem_context *ctx)
kfree(ctx->name);
put_pid(ctx->pid);
- list_del(&ctx->link);
mutex_destroy(&ctx->mutex);
kfree_rcu(ctx, rcu);
}
-static void contexts_free(struct drm_i915_private *i915)
+static void contexts_free_all(struct llist_node *list)
{
- struct llist_node *freed = llist_del_all(&i915->contexts.free_list);
struct i915_gem_context *ctx, *cn;
- lockdep_assert_held(&i915->drm.struct_mutex);
-
- llist_for_each_entry_safe(ctx, cn, freed, free_link)
+ llist_for_each_entry_safe(ctx, cn, list, free_link)
i915_gem_context_free(ctx);
}
-static void contexts_free_first(struct drm_i915_private *i915)
+static void contexts_flush_free(struct i915_gem_contexts *gc)
{
- struct i915_gem_context *ctx;
- struct llist_node *freed;
-
- lockdep_assert_held(&i915->drm.struct_mutex);
-
- freed = llist_del_first(&i915->contexts.free_list);
- if (!freed)
- return;
-
- ctx = container_of(freed, typeof(*ctx), free_link);
- i915_gem_context_free(ctx);
+ contexts_free_all(llist_del_all(&gc->free_list));
}
static void contexts_free_worker(struct work_struct *work)
{
- struct drm_i915_private *i915 =
- container_of(work, typeof(*i915), contexts.free_work);
+ struct i915_gem_contexts *gc =
+ container_of(work, typeof(*gc), free_work);
- mutex_lock(&i915->drm.struct_mutex);
- contexts_free(i915);
- mutex_unlock(&i915->drm.struct_mutex);
+ contexts_flush_free(gc);
}
void i915_gem_context_release(struct kref *ref)
{
struct i915_gem_context *ctx = container_of(ref, typeof(*ctx), ref);
- struct drm_i915_private *i915 = ctx->i915;
+ struct i915_gem_contexts *gc = &ctx->i915->gem.contexts;
trace_i915_context_free(ctx);
- if (llist_add(&ctx->free_link, &i915->contexts.free_list))
- queue_work(i915->wq, &i915->contexts.free_work);
+ if (llist_add(&ctx->free_link, &gc->free_list))
+ queue_work(ctx->i915->wq, &gc->free_work);
}
static void context_close(struct i915_gem_context *ctx)
{
- i915_gem_context_set_closed(ctx);
+ struct i915_address_space *vm;
- if (ctx->vm)
- i915_vm_close(ctx->vm);
+ i915_gem_context_set_closed(ctx);
mutex_lock(&ctx->mutex);
+ vm = i915_gem_context_vm(ctx);
+ if (vm)
+ i915_vm_close(vm);
+
ctx->file_priv = ERR_PTR(-EBADF);
/*
@@ -317,7 +307,6 @@ __create_context(struct drm_i915_private *i915)
return ERR_PTR(-ENOMEM);
kref_init(&ctx->ref);
- list_add_tail(&ctx->link, &i915->contexts.list);
ctx->i915 = i915;
ctx->sched.priority = I915_USER_PRIORITY(I915_PRIORITY_NORMAL);
mutex_init(&ctx->mutex);
@@ -343,6 +332,10 @@ __create_context(struct drm_i915_private *i915)
for (i = 0; i < ARRAY_SIZE(ctx->hang_timestamp); i++)
ctx->hang_timestamp[i] = jiffies - CONTEXT_FAST_HANG_JIFFIES;
+ spin_lock(&i915->gem.contexts.lock);
+ list_add_tail(&ctx->link, &i915->gem.contexts.list);
+ spin_unlock(&i915->gem.contexts.lock);
+
return ctx;
err_free:
@@ -372,11 +365,11 @@ static void __apply_ppgtt(struct intel_context *ce, void *vm)
static struct i915_address_space *
__set_ppgtt(struct i915_gem_context *ctx, struct i915_address_space *vm)
{
- struct i915_address_space *old = ctx->vm;
+ struct i915_address_space *old = i915_gem_context_vm(ctx);
GEM_BUG_ON(old && i915_vm_is_4lvl(vm) != i915_vm_is_4lvl(old));
- ctx->vm = i915_vm_open(vm);
+ rcu_assign_pointer(ctx->vm, i915_vm_open(vm));
context_apply_all(ctx, __apply_ppgtt, vm);
return old;
@@ -385,7 +378,7 @@ __set_ppgtt(struct i915_gem_context *ctx, struct i915_address_space *vm)
static void __assign_ppgtt(struct i915_gem_context *ctx,
struct i915_address_space *vm)
{
- if (vm == ctx->vm)
+ if (vm == rcu_access_pointer(ctx->vm))
return;
vm = __set_ppgtt(ctx, vm);
@@ -417,27 +410,25 @@ static void __assign_timeline(struct i915_gem_context *ctx,
}
static struct i915_gem_context *
-i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags)
+i915_gem_create_context(struct drm_i915_private *i915, unsigned int flags)
{
struct i915_gem_context *ctx;
- lockdep_assert_held(&dev_priv->drm.struct_mutex);
-
if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE &&
- !HAS_EXECLISTS(dev_priv))
+ !HAS_EXECLISTS(i915))
return ERR_PTR(-EINVAL);
- /* Reap the most stale context */
- contexts_free_first(dev_priv);
+ /* Reap the stale contexts */
+ contexts_flush_free(&i915->gem.contexts);
- ctx = __create_context(dev_priv);
+ ctx = __create_context(i915);
if (IS_ERR(ctx))
return ctx;
- if (HAS_FULL_PPGTT(dev_priv)) {
+ if (HAS_FULL_PPGTT(i915)) {
struct i915_ppgtt *ppgtt;
- ppgtt = i915_ppgtt_create(dev_priv);
+ ppgtt = i915_ppgtt_create(i915);
if (IS_ERR(ppgtt)) {
DRM_DEBUG_DRIVER("PPGTT setup failed (%ld)\n",
PTR_ERR(ppgtt));
@@ -445,14 +436,17 @@ i915_gem_create_context(struct drm_i915_private *dev_priv, unsigned int flags)
return ERR_CAST(ppgtt);
}
+ mutex_lock(&ctx->mutex);
__assign_ppgtt(ctx, &ppgtt->vm);
+ mutex_unlock(&ctx->mutex);
+
i915_vm_put(&ppgtt->vm);
}
if (flags & I915_CONTEXT_CREATE_FLAGS_SINGLE_TIMELINE) {
struct intel_timeline *timeline;
- timeline = intel_timeline_create(&dev_priv->gt, NULL);
+ timeline = intel_timeline_create(&i915->gt, NULL);
if (IS_ERR(timeline)) {
context_close(ctx);
return ERR_CAST(timeline);
@@ -497,48 +491,40 @@ i915_gem_context_create_kernel(struct drm_i915_private *i915, int prio)
return ctx;
}
-static void init_contexts(struct drm_i915_private *i915)
+static void init_contexts(struct i915_gem_contexts *gc)
{
- mutex_init(&i915->contexts.mutex);
- INIT_LIST_HEAD(&i915->contexts.list);
-
- /* Using the simple ida interface, the max is limited by sizeof(int) */
- BUILD_BUG_ON(MAX_CONTEXT_HW_ID > INT_MAX);
- BUILD_BUG_ON(GEN11_MAX_CONTEXT_HW_ID > INT_MAX);
- ida_init(&i915->contexts.hw_ida);
- INIT_LIST_HEAD(&i915->contexts.hw_id_list);
+ spin_lock_init(&gc->lock);
+ INIT_LIST_HEAD(&gc->list);
- INIT_WORK(&i915->contexts.free_work, contexts_free_worker);
- init_llist_head(&i915->contexts.free_list);
+ INIT_WORK(&gc->free_work, contexts_free_worker);
+ init_llist_head(&gc->free_list);
}
-int i915_gem_contexts_init(struct drm_i915_private *dev_priv)
+int i915_gem_init_contexts(struct drm_i915_private *i915)
{
struct i915_gem_context *ctx;
/* Reassure ourselves we are only called once */
- GEM_BUG_ON(dev_priv->kernel_context);
+ GEM_BUG_ON(i915->kernel_context);
- init_contexts(dev_priv);
+ init_contexts(&i915->gem.contexts);
/* lowest priority; idle task */
- ctx = i915_gem_context_create_kernel(dev_priv, I915_PRIORITY_MIN);
+ ctx = i915_gem_context_create_kernel(i915, I915_PRIORITY_MIN);
if (IS_ERR(ctx)) {
DRM_ERROR("Failed to create default global context\n");
return PTR_ERR(ctx);
}
- dev_priv->kernel_context = ctx;
+ i915->kernel_context = ctx;
DRM_DEBUG_DRIVER("%s context support initialized\n",
- DRIVER_CAPS(dev_priv)->has_logical_contexts ?
+ DRIVER_CAPS(i915)->has_logical_contexts ?
"logical" : "fake");
return 0;
}
-void i915_gem_contexts_fini(struct drm_i915_private *i915)
+void i915_gem_driver_release__contexts(struct drm_i915_private *i915)
{
- lockdep_assert_held(&i915->drm.struct_mutex);
-
destroy_kernel_context(&i915->kernel_context);
}
@@ -557,11 +543,16 @@ static int vm_idr_cleanup(int id, void *p, void *data)
static int gem_context_register(struct i915_gem_context *ctx,
struct drm_i915_file_private *fpriv)
{
+ struct i915_address_space *vm;
int ret;
ctx->file_priv = fpriv;
- if (ctx->vm)
- ctx->vm->file = fpriv;
+
+ mutex_lock(&ctx->mutex);
+ vm = i915_gem_context_vm(ctx);
+ if (vm)
+ WRITE_ONCE(vm->file, fpriv); /* XXX */
+ mutex_unlock(&ctx->mutex);
ctx->pid = get_task_pid(current, PIDTYPE_PID);
ctx->name = kasprintf(GFP_KERNEL, "%s[%d]",
@@ -598,9 +589,7 @@ int i915_gem_context_open(struct drm_i915_private *i915,
idr_init(&file_priv->context_idr);
idr_init_base(&file_priv->vm_idr, 1);
- mutex_lock(&i915->drm.struct_mutex);
ctx = i915_gem_create_context(i915, 0);
- mutex_unlock(&i915->drm.struct_mutex);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
goto err;
@@ -628,6 +617,7 @@ int i915_gem_context_open(struct drm_i915_private *i915,
void i915_gem_context_close(struct drm_file *file)
{
struct drm_i915_file_private *file_priv = file->driver_priv;
+ struct drm_i915_private *i915 = file_priv->dev_priv;
idr_for_each(&file_priv->context_idr, context_idr_cleanup, NULL);
idr_destroy(&file_priv->context_idr);
@@ -636,6 +626,8 @@ void i915_gem_context_close(struct drm_file *file)
idr_for_each(&file_priv->vm_idr, vm_idr_cleanup, NULL);
idr_destroy(&file_priv->vm_idr);
mutex_destroy(&file_priv->vm_idr_lock);
+
+ contexts_flush_free(&i915->gem.contexts);
}
int i915_gem_vm_create_ioctl(struct drm_device *dev, void *data,
@@ -814,16 +806,12 @@ static int get_ppgtt(struct drm_i915_file_private *file_priv,
struct i915_address_space *vm;
int ret;
- if (!ctx->vm)
+ if (!rcu_access_pointer(ctx->vm))
return -ENODEV;
- /* XXX rcu acquire? */
- ret = mutex_lock_interruptible(&ctx->i915->drm.struct_mutex);
- if (ret)
- return ret;
-
+ rcu_read_lock();
vm = i915_vm_get(ctx->vm);
- mutex_unlock(&ctx->i915->drm.struct_mutex);
+ rcu_read_unlock();
ret = mutex_lock_interruptible(&file_priv->vm_idr_lock);
if (ret)
@@ -932,7 +920,7 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
if (args->size)
return -EINVAL;
- if (!ctx->vm)
+ if (!rcu_access_pointer(ctx->vm))
return -ENODEV;
if (upper_32_bits(args->value))
@@ -946,17 +934,20 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
if (!vm)
return -ENOENT;
- err = mutex_lock_interruptible(&ctx->i915->drm.struct_mutex);
+ err = mutex_lock_interruptible(&ctx->mutex);
if (err)
goto out;
- if (vm == ctx->vm)
+ if (i915_gem_context_is_closed(ctx)) {
+ err = -ENOENT;
+ goto out;
+ }
+
+ if (vm == rcu_access_pointer(ctx->vm))
goto unlock;
/* Teardown the existing obj:vma cache, it will have to be rebuilt. */
- mutex_lock(&ctx->mutex);
lut_close(ctx);
- mutex_unlock(&ctx->mutex);
old = __set_ppgtt(ctx, vm);
@@ -976,8 +967,7 @@ static int set_ppgtt(struct drm_i915_file_private *file_priv,
}
unlock:
- mutex_unlock(&ctx->i915->drm.struct_mutex);
-
+ mutex_unlock(&ctx->mutex);
out:
i915_vm_put(vm);
return err;
@@ -1833,10 +1823,11 @@ static int clone_vm(struct i915_gem_context *dst,
struct i915_gem_context *src)
{
struct i915_address_space *vm;
+ int err = 0;
rcu_read_lock();
do {
- vm = READ_ONCE(src->vm);
+ vm = rcu_dereference(src->vm);
if (!vm)
break;
@@ -1858,7 +1849,7 @@ static int clone_vm(struct i915_gem_context *dst,
* it cannot be reallocated elsewhere.
*/
- if (vm == READ_ONCE(src->vm))
+ if (vm == rcu_access_pointer(src->vm))
break;
i915_vm_put(vm);
@@ -1866,11 +1857,16 @@ static int clone_vm(struct i915_gem_context *dst,
rcu_read_unlock();
if (vm) {
- __assign_ppgtt(dst, vm);
+ if (!mutex_lock_interruptible(&dst->mutex)) {
+ __assign_ppgtt(dst, vm);
+ mutex_unlock(&dst->mutex);
+ } else {
+ err = -EINTR;
+ }
i915_vm_put(vm);
}
- return 0;
+ return err;
}
static int create_clone(struct i915_user_extension __user *ext, void *data)
@@ -1960,12 +1956,7 @@ int i915_gem_context_create_ioctl(struct drm_device *dev, void *data,
return -EIO;
}
- ret = i915_mutex_lock_interruptible(dev);
- if (ret)
- return ret;
-
ext_data.ctx = i915_gem_create_context(i915, args->flags);
- mutex_unlock(&dev->struct_mutex);
if (IS_ERR(ext_data.ctx))
return PTR_ERR(ext_data.ctx);
@@ -2092,10 +2083,12 @@ int i915_gem_context_getparam_ioctl(struct drm_device *dev, void *data,
case I915_CONTEXT_PARAM_GTT_SIZE:
args->size = 0;
- if (ctx->vm)
- args->value = ctx->vm->total;
+ rcu_read_lock();
+ if (rcu_access_pointer(ctx->vm))
+ args->value = rcu_dereference(ctx->vm)->total;
else
args->value = to_i915(dev)->ggtt.vm.total;
+ rcu_read_unlock();
break;
case I915_CONTEXT_PARAM_NO_ERROR_CAPTURE:
@@ -2161,7 +2154,7 @@ int i915_gem_context_setparam_ioctl(struct drm_device *dev, void *data,
int i915_gem_context_reset_stats_ioctl(struct drm_device *dev,
void *data, struct drm_file *file)
{
- struct drm_i915_private *dev_priv = to_i915(dev);
+ struct drm_i915_private *i915 = to_i915(dev);
struct drm_i915_reset_stats *args = data;
struct i915_gem_context *ctx;
int ret;
@@ -2183,7 +2176,7 @@ int i915_gem_context_reset_stats_ioctl(struct drm_device *dev,
*/
if (capable(CAP_SYS_ADMIN))
- args->reset_count = i915_reset_count(&dev_priv->gpu_error);
+ args->reset_count = i915_reset_count(&i915->gpu_error);
else
args->reset_count = 0;
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context.h b/drivers/gpu/drm/i915/gem/i915_gem_context.h
index 50bc27d30c03..9234586830d1 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context.h
@@ -11,7 +11,9 @@
#include "gt/intel_context.h"
+#include "i915_drv.h"
#include "i915_gem.h"
+#include "i915_gem_gtt.h"
#include "i915_scheduler.h"
#include "intel_device_info.h"
@@ -118,8 +120,8 @@ static inline bool i915_gem_context_is_kernel(struct i915_gem_context *ctx)
}
/* i915_gem_context.c */
-int __must_check i915_gem_contexts_init(struct drm_i915_private *dev_priv);
-void i915_gem_contexts_fini(struct drm_i915_private *dev_priv);
+int __must_check i915_gem_init_contexts(struct drm_i915_private *i915);
+void i915_gem_driver_release__contexts(struct drm_i915_private *i915);
int i915_gem_context_open(struct drm_i915_private *i915,
struct drm_file *file);
@@ -158,6 +160,27 @@ static inline void i915_gem_context_put(struct i915_gem_context *ctx)
kref_put(&ctx->ref, i915_gem_context_release);
}
+static inline struct i915_address_space *
+i915_gem_context_vm(struct i915_gem_context *ctx)
+{
+ return rcu_dereference_protected(ctx->vm, lockdep_is_held(&ctx->mutex));
+}
+
+static inline struct i915_address_space *
+i915_gem_context_get_vm_rcu(struct i915_gem_context *ctx)
+{
+ struct i915_address_space *vm;
+
+ rcu_read_lock();
+ vm = rcu_dereference(ctx->vm);
+ if (!vm)
+ vm = &ctx->i915->ggtt.vm;
+ vm = i915_vm_get(vm);
+ rcu_read_unlock();
+
+ return vm;
+}
+
static inline struct i915_gem_engines *
i915_gem_context_engines(struct i915_gem_context *ctx)
{
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
index 87be27877e22..ab8e1367dfc8 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
+++ b/drivers/gpu/drm/i915/gem/i915_gem_context_types.h
@@ -88,7 +88,7 @@ struct i915_gem_context {
* In other modes, this is a NULL pointer with the expectation that
* the caller uses the shared global GTT.
*/
- struct i915_address_space *vm;
+ struct i915_address_space __rcu *vm;
/**
* @pid: process id of creator
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
index 88a881be12ec..98816c35ffc3 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_execbuffer.c
@@ -728,7 +728,7 @@ static int eb_select_context(struct i915_execbuffer *eb)
return -ENOENT;
eb->gem_context = ctx;
- if (ctx->vm)
+ if (rcu_access_pointer(ctx->vm))
eb->invalid_flags |= EXEC_OBJECT_NEEDS_GTT;
eb->context_flags = 0;
diff --git a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
index 1738a15eb911..4f970474013f 100644
--- a/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
+++ b/drivers/gpu/drm/i915/gem/i915_gem_userptr.c
@@ -758,7 +758,8 @@ i915_gem_userptr_ioctl(struct drm_device *dev,
* On almost all of the older hw, we cannot tell the GPU that
* a page is readonly.
*/
- vm = dev_priv->kernel_context->vm;
+ vm = rcu_dereference_protected(dev_priv->kernel_context->vm,
+ true); /* static vm */
if (!vm || !vm->has_read_only)
return -ENODEV;
}
diff --git a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
index 98b2a6ccfcc1..3314858f3046 100644
--- a/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
+++ b/drivers/gpu/drm/i915/gem/selftests/huge_pages.c
@@ -1322,15 +1322,15 @@ static int igt_ppgtt_pin_update(void *arg)
struct i915_gem_context *ctx = arg;
struct drm_i915_private *dev_priv = ctx->i915;
unsigned long supported = INTEL_INFO(dev_priv)->page_sizes;
- struct i915_address_space *vm = ctx->vm;
struct drm_i915_gem_object *obj;
struct i915_gem_engines_iter it;
+ struct i915_address_space *vm;
struct intel_context *ce;
struct i915_vma *vma;
unsigned int flags = PIN_USER | PIN_OFFSET_FIXED;
unsigned int n;
int first, last;
- int err;
+ int err = 0;
/*
* Make sure there's no funny business when doing a PIN_UPDATE -- in the
@@ -1340,9 +1340,10 @@ static int igt_ppgtt_pin_update(void *arg)
* huge-gtt-pages.
*/
- if (!vm || !i915_vm_is_4lvl(vm)) {
+ vm = i915_gem_context_get_vm_rcu(ctx);
+ if (!i915_vm_is_4lvl(vm)) {
pr_info("48b PPGTT not supported, skipping\n");
- return 0;
+ goto out_vm;
}
first = ilog2(I915_GTT_PAGE_SIZE_64K);
@@ -1451,6 +1452,8 @@ static int igt_ppgtt_pin_update(void *arg)
i915_vma_close(vma);
out_put:
i915_gem_object_put(obj);
+out_vm:
+ i915_vm_put(vm);
return err;
}
@@ -1460,7 +1463,7 @@ static int igt_tmpfs_fallback(void *arg)
struct i915_gem_context *ctx = arg;
struct drm_i915_private *i915 = ctx->i915;
struct vfsmount *gemfs = i915->mm.gemfs;
- struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm;
+ struct i915_address_space *vm = i915_gem_context_get_vm_rcu(ctx);
struct drm_i915_gem_object *obj;
struct i915_vma *vma;
u32 *vaddr;
@@ -1510,6 +1513,7 @@ static int igt_tmpfs_fallback(void *arg)
out_restore:
i915->mm.gemfs = gemfs;
+ i915_vm_put(vm);
return err;
}
@@ -1517,14 +1521,14 @@ static int igt_shrink_thp(void *arg)
{
struct i915_gem_context *ctx = arg;
struct drm_i915_private *i915 = ctx->i915;
- struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm;
+ struct i915_address_space *vm = i915_gem_context_get_vm_rcu(ctx);
struct drm_i915_gem_object *obj;
struct i915_gem_engines_iter it;
struct intel_context *ce;
struct i915_vma *vma;
unsigned int flags = PIN_USER;
unsigned int n;
- int err;
+ int err = 0;
/*
* Sanity check shrinking huge-paged object -- make sure nothing blows
@@ -1533,12 +1537,14 @@ static int igt_shrink_thp(void *arg)
if (!igt_can_allocate_thp(i915)) {
pr_info("missing THP support, skipping\n");
- return 0;
+ goto out_vm;
}
obj = i915_gem_object_create_shmem(i915, SZ_2M);
- if (IS_ERR(obj))
- return PTR_ERR(obj);
+ if (IS_ERR(obj)) {
+ err = PTR_ERR(obj);
+ goto out_vm;
+ }
vma = i915_vma_instance(obj, vm, NULL);
if (IS_ERR(vma)) {
@@ -1607,6 +1613,8 @@ static int igt_shrink_thp(void *arg)
i915_vma_close(vma);
out_put:
i915_gem_object_put(obj);
+out_vm:
+ i915_vm_put(vm);
return err;
}
@@ -1675,6 +1683,7 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *i915)
};
struct drm_file *file;
struct i915_gem_context *ctx;
+ struct i915_address_space *vm;
intel_wakeref_t wakeref;
int err;
@@ -1699,8 +1708,11 @@ int i915_gem_huge_page_live_selftests(struct drm_i915_private *i915)
goto out_unlock;
}
- if (ctx->vm)
- ctx->vm->scrub_64K = true;
+ mutex_lock(&ctx->mutex);
+ vm = i915_gem_context_vm(ctx);
+ if (vm)
+ WRITE_ONCE(vm->scrub_64K, true);
+ mutex_unlock(&ctx->mutex);
err = i915_subtests(tests, ctx);
diff --git a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
index 2fb31ada2fa7..d44fa9d356f1 100644
--- a/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/i915_gem_context.c
@@ -53,19 +53,17 @@ static int live_nop_switch(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(&i915->drm.struct_mutex);
-
ctx = kcalloc(nctx, sizeof(*ctx), GFP_KERNEL);
if (!ctx) {
err = -ENOMEM;
- goto out_unlock;
+ goto out_file;
}
for (n = 0; n < nctx; n++) {
ctx[n] = live_context(i915, file);
if (IS_ERR(ctx[n])) {
err = PTR_ERR(ctx[n]);
- goto out_unlock;
+ goto out_file;
}
}
@@ -79,7 +77,7 @@ static int live_nop_switch(void *arg)
rq = igt_request_alloc(ctx[n], engine);
if (IS_ERR(rq)) {
err = PTR_ERR(rq);
- goto out_unlock;
+ goto out_file;
}
i915_request_add(rq);
}
@@ -87,7 +85,7 @@ static int live_nop_switch(void *arg)
pr_err("Failed to populated %d contexts\n", nctx);
intel_gt_set_wedged(&i915->gt);
err = -EIO;
- goto out_unlock;
+ goto out_file;
}
times[1] = ktime_get_raw();
@@ -97,7 +95,7 @@ static int live_nop_switch(void *arg)
err = igt_live_test_begin(&t, i915, __func__, engine->name);
if (err)
- goto out_unlock;
+ goto out_file;
end_time = jiffies + i915_selftest.timeout_jiffies;
for_each_prime_number_from(prime, 2, 8192) {
@@ -107,7 +105,7 @@ static int live_nop_switch(void *arg)
rq = igt_request_alloc(ctx[n % nctx], engine);
if (IS_ERR(rq)) {
err = PTR_ERR(rq);
- goto out_unlock;
+ goto out_file;
}
/*
@@ -143,7 +141,7 @@ static int live_nop_switch(void *arg)
err = igt_live_test_end(&t);
if (err)
- goto out_unlock;
+ goto out_file;
pr_info("Switch latencies on %s: 1 = %lluns, %lu = %lluns\n",
engine->name,
@@ -151,8 +149,7 @@ static int live_nop_switch(void *arg)
prime - 1, div64_u64(ktime_to_ns(times[1]), prime - 1));
}
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
+out_file:
mock_file_free(i915, file);
return err;
}
@@ -253,12 +250,10 @@ static int live_parallel_switch(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(&i915->drm.struct_mutex);
-
ctx = live_context(i915, file);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
- goto out_locked;
+ goto out_file;
}
engines = i915_gem_context_lock_engines(ctx);
@@ -268,7 +263,7 @@ static int live_parallel_switch(void *arg)
if (!data) {
i915_gem_context_unlock_engines(ctx);
err = -ENOMEM;
- goto out_locked;
+ goto out;
}
m = 0; /* Use the first context as our template for the engines */
@@ -276,7 +271,7 @@ static int live_parallel_switch(void *arg)
err = intel_context_pin(ce);
if (err) {
i915_gem_context_unlock_engines(ctx);
- goto out_locked;
+ goto out;
}
data[m++].ce[0] = intel_context_get(ce);
}
@@ -287,7 +282,7 @@ static int live_parallel_switch(void *arg)
ctx = live_context(i915, file);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
- goto out_locked;
+ goto out;
}
for (m = 0; m < count; m++) {
@@ -296,20 +291,18 @@ static int live_parallel_switch(void *arg)
ce = intel_context_create(ctx, data[m].ce[0]->engine);
if (IS_ERR(ce))
- goto out_locked;
+ goto out;
err = intel_context_pin(ce);
if (err) {
intel_context_put(ce);
- goto out_locked;
+ goto out;
}
data[m].ce[n] = ce;
}
}
- mutex_unlock(&i915->drm.struct_mutex);
-
for (fn = func; !err && *fn; fn++) {
struct igt_live_test t;
int n;
@@ -354,8 +347,7 @@ static int live_parallel_switch(void *arg)
mutex_unlock(&i915->drm.struct_mutex);
}
- mutex_lock(&i915->drm.struct_mutex);
-out_locked:
+out:
for (n = 0; n < count; n++) {
for (m = 0; m < ARRAY_SIZE(data->ce); m++) {
if (!data[n].ce[m])
@@ -365,8 +357,8 @@ static int live_parallel_switch(void *arg)
intel_context_put(data[n].ce[m]);
}
}
- mutex_unlock(&i915->drm.struct_mutex);
kfree(data);
+out_file:
mock_file_free(i915, file);
return err;
}
@@ -626,11 +618,9 @@ static int igt_ctx_exec(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(&i915->drm.struct_mutex);
-
err = igt_live_test_begin(&t, i915, __func__, engine->name);
if (err)
- goto out_unlock;
+ goto out_file;
ncontexts = 0;
ndwords = 0;
@@ -642,7 +632,7 @@ static int igt_ctx_exec(void *arg)
ctx = kernel_context(i915);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
- goto out_unlock;
+ goto out_file;
}
ce = i915_gem_context_get_engine(ctx, engine->legacy_idx);
@@ -654,7 +644,7 @@ static int igt_ctx_exec(void *arg)
err = PTR_ERR(obj);
intel_context_put(ce);
kernel_context_close(ctx);
- goto out_unlock;
+ goto out_file;
}
}
@@ -663,17 +653,18 @@ static int igt_ctx_exec(void *arg)
pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
ndwords, dw, max_dwords(obj),
engine->name,
- yesno(!!ctx->vm), err);
+ yesno(!!rcu_access_pointer(ctx->vm)),
+ err);
intel_context_put(ce);
kernel_context_close(ctx);
- goto out_unlock;
+ goto out_file;
}
err = throttle(ce, tq, ARRAY_SIZE(tq));
if (err) {
intel_context_put(ce);
kernel_context_close(ctx);
- goto out_unlock;
+ goto out_file;
}
if (++dw == max_dwords(obj)) {
@@ -703,11 +694,10 @@ static int igt_ctx_exec(void *arg)
dw += rem;
}
-out_unlock:
+out_file:
throttle_release(tq, ARRAY_SIZE(tq));
if (igt_live_test_end(&t))
err = -EIO;
- mutex_unlock(&i915->drm.struct_mutex);
mock_file_free(i915, file);
if (err)
@@ -742,22 +732,20 @@ static int igt_shared_ctx_exec(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(&i915->drm.struct_mutex);
-
parent = live_context(i915, file);
if (IS_ERR(parent)) {
err = PTR_ERR(parent);
- goto out_unlock;
+ goto out_file;
}
if (!parent->vm) { /* not full-ppgtt; nothing to share */
err = 0;
- goto out_unlock;
+ goto out_file;
}
err = igt_live_test_begin(&t, i915, __func__, "");
if (err)
- goto out_unlock;
+ goto out_file;
for_each_engine(engine, i915, id) {
unsigned long ncontexts, ndwords, dw;
@@ -781,7 +769,9 @@ static int igt_shared_ctx_exec(void *arg)
goto out_test;
}
+ mutex_lock(&ctx->mutex);
__assign_ppgtt(ctx, parent->vm);
+ mutex_unlock(&ctx->mutex);
ce = i915_gem_context_get_engine(ctx, engine->legacy_idx);
GEM_BUG_ON(IS_ERR(ce));
@@ -801,7 +791,8 @@ static int igt_shared_ctx_exec(void *arg)
pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
ndwords, dw, max_dwords(obj),
engine->name,
- yesno(!!ctx->vm), err);
+ yesno(!!rcu_access_pointer(ctx->vm)),
+ err);
intel_context_put(ce);
kernel_context_close(ctx);
goto out_test;
@@ -840,17 +831,13 @@ static int igt_shared_ctx_exec(void *arg)
dw += rem;
}
- mutex_unlock(&i915->drm.struct_mutex);
i915_gem_drain_freed_objects(i915);
- mutex_lock(&i915->drm.struct_mutex);
}
out_test:
throttle_release(tq, ARRAY_SIZE(tq));
if (igt_live_test_end(&t))
err = -EIO;
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
-
+out_file:
mock_file_free(i915, file);
return err;
}
@@ -1222,8 +1209,6 @@ __igt_ctx_sseu(struct drm_i915_private *i915,
if (flags & TEST_RESET)
igt_global_reset_lock(&i915->gt);
- mutex_lock(&i915->drm.struct_mutex);
-
ctx = live_context(i915, file);
if (IS_ERR(ctx)) {
ret = PTR_ERR(ctx);
@@ -1278,8 +1263,6 @@ __igt_ctx_sseu(struct drm_i915_private *i915,
i915_gem_object_put(obj);
out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
-
if (flags & TEST_RESET)
igt_global_reset_unlock(&i915->gt);
@@ -1339,23 +1322,24 @@ static int igt_ctx_readonly(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(&i915->drm.struct_mutex);
-
err = igt_live_test_begin(&t, i915, __func__, "");
if (err)
- goto out_unlock;
+ goto out_file;
ctx = live_context(i915, file);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
- goto out_unlock;
+ goto out_file;
}
- vm = ctx->vm ?: &i915->ggtt.alias->vm;
+ rcu_read_lock();
+ vm = rcu_dereference(ctx->vm) ?: &i915->ggtt.alias->vm;
if (!vm || !vm->has_read_only) {
+ rcu_read_unlock();
err = 0;
- goto out_unlock;
+ goto out_file;
}
+ rcu_read_unlock();
ndwords = 0;
dw = 0;
@@ -1373,7 +1357,7 @@ static int igt_ctx_readonly(void *arg)
if (IS_ERR(obj)) {
err = PTR_ERR(obj);
i915_gem_context_unlock_engines(ctx);
- goto out_unlock;
+ goto out_file;
}
if (prandom_u32_state(&prng) & 1)
@@ -1384,15 +1368,17 @@ static int igt_ctx_readonly(void *arg)
if (err) {
pr_err("Failed to fill dword %lu [%lu/%lu] with gpu (%s) [full-ppgtt? %s], err=%d\n",
ndwords, dw, max_dwords(obj),
- ce->engine->name, yesno(!!ctx->vm), err);
+ ce->engine->name,
+ yesno(!!rcu_access_pointer(ctx->vm)),
+ err);
i915_gem_context_unlock_engines(ctx);
- goto out_unlock;
+ goto out_file;
}
err = throttle(ce, tq, ARRAY_SIZE(tq));
if (err) {
i915_gem_context_unlock_engines(ctx);
- goto out_unlock;
+ goto out_file;
}
if (++dw == max_dwords(obj)) {
@@ -1424,20 +1410,19 @@ static int igt_ctx_readonly(void *arg)
dw += rem;
}
-out_unlock:
+out_file:
throttle_release(tq, ARRAY_SIZE(tq));
if (igt_live_test_end(&t))
err = -EIO;
- mutex_unlock(&i915->drm.struct_mutex);
mock_file_free(i915, file);
return err;
}
-static int check_scratch(struct i915_gem_context *ctx, u64 offset)
+static int check_scratch(struct i915_address_space *vm, u64 offset)
{
struct drm_mm_node *node =
- __drm_mm_interval_first(&ctx->vm->mm,
+ __drm_mm_interval_first(&vm->mm,
offset, offset + sizeof(u32) - 1);
if (!node || node->start > offset)
return 0;
@@ -1455,6 +1440,7 @@ static int write_to_scratch(struct i915_gem_context *ctx,
{
struct drm_i915_private *i915 = ctx->i915;
struct drm_i915_gem_object *obj;
+ struct i915_address_space *vm;
struct i915_request *rq;
struct i915_vma *vma;
u32 *cmd;
@@ -1487,17 +1473,18 @@ static int write_to_scratch(struct i915_gem_context *ctx,
intel_gt_chipset_flush(engine->gt);
- vma = i915_vma_instance(obj, ctx->vm, NULL);
+ vm = i915_gem_context_get_vm_rcu(ctx);
+ vma = i915_vma_instance(obj, vm, NULL);
if (IS_ERR(vma)) {
err = PTR_ERR(vma);
- goto err;
+ goto err_vm;
}
err = i915_vma_pin(vma, 0, 0, PIN_USER | PIN_OFFSET_FIXED);
if (err)
- goto err;
+ goto err_vm;
- err = check_scratch(ctx, offset);
+ err = check_scratch(vm, offset);
if (err)
goto err_unpin;
@@ -1523,6 +1510,7 @@ static int write_to_scratch(struct i915_gem_context *ctx,
i915_request_add(rq);
+ i915_vm_put(vm);
return 0;
skip_request:
@@ -1531,6 +1519,8 @@ static int write_to_scratch(struct i915_gem_context *ctx,
i915_request_add(rq);
err_unpin:
i915_vma_unpin(vma);
+err_vm:
+ i915_vm_put(vm);
err:
i915_gem_object_put(obj);
return err;
@@ -1542,6 +1532,7 @@ static int read_from_scratch(struct i915_gem_context *ctx,
{
struct drm_i915_private *i915 = ctx->i915;
struct drm_i915_gem_object *obj;
+ struct i915_address_space *vm;
const u32 RCS_GPR0 = 0x2600; /* not all engines have their own GPR! */
const u32 result = 0x100;
struct i915_request *rq;
@@ -1586,17 +1577,18 @@ static int read_from_scratch(struct i915_gem_context *ctx,
intel_gt_chipset_flush(engine->gt);
- vma = i915_vma_instance(obj, ctx->vm, NULL);
+ vm = i915_gem_context_get_vm_rcu(ctx);
+ vma = i915_vma_instance(obj, vm, NULL);
if (IS_ERR(vma)) {
err = PTR_ERR(vma);
- goto err;
+ goto err_vm;
}
err = i915_vma_pin(vma, 0, 0, PIN_USER | PIN_OFFSET_FIXED);
if (err)
- goto err;
+ goto err_vm;
- err = check_scratch(ctx, offset);
+ err = check_scratch(vm, offset);
if (err)
goto err_unpin;
@@ -1627,12 +1619,12 @@ static int read_from_scratch(struct i915_gem_context *ctx,
err = i915_gem_object_set_to_cpu_domain(obj, false);
i915_gem_object_unlock(obj);
if (err)
- goto err;
+ goto err_vm;
cmd = i915_gem_object_pin_map(obj, I915_MAP_WB);
if (IS_ERR(cmd)) {
err = PTR_ERR(cmd);
- goto err;
+ goto err_vm;
}
*value = cmd[result / sizeof(*cmd)];
@@ -1647,6 +1639,8 @@ static int read_from_scratch(struct i915_gem_context *ctx,
i915_request_add(rq);
err_unpin:
i915_vma_unpin(vma);
+err_vm:
+ i915_vm_put(vm);
err:
i915_gem_object_put(obj);
return err;
@@ -1677,27 +1671,25 @@ static int igt_vm_isolation(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(&i915->drm.struct_mutex);
-
err = igt_live_test_begin(&t, i915, __func__, "");
if (err)
- goto out_unlock;
+ goto out_file;
ctx_a = live_context(i915, file);
if (IS_ERR(ctx_a)) {
err = PTR_ERR(ctx_a);
- goto out_unlock;
+ goto out_file;
}
ctx_b = live_context(i915, file);
if (IS_ERR(ctx_b)) {
err = PTR_ERR(ctx_b);
- goto out_unlock;
+ goto out_file;
}
/* We can only test vm isolation, if the vm are distinct */
if (ctx_a->vm == ctx_b->vm)
- goto out_unlock;
+ goto out_file;
vm_total = ctx_a->vm->total;
GEM_BUG_ON(ctx_b->vm->total != vm_total);
@@ -1726,7 +1718,7 @@ static int igt_vm_isolation(void *arg)
err = read_from_scratch(ctx_b, engine,
offset, &value);
if (err)
- goto out_unlock;
+ goto out_file;
if (value) {
pr_err("%s: Read %08x from scratch (offset 0x%08x_%08x), after %lu reads!\n",
@@ -1735,7 +1727,7 @@ static int igt_vm_isolation(void *arg)
lower_32_bits(offset),
this);
err = -EINVAL;
- goto out_unlock;
+ goto out_file;
}
this++;
@@ -1745,11 +1737,9 @@ static int igt_vm_isolation(void *arg)
pr_info("Checked %lu scratch offsets across %d engines\n",
count, RUNTIME_INFO(i915)->num_engines);
-out_unlock:
+out_file:
if (igt_live_test_end(&t))
err = -EIO;
- mutex_unlock(&i915->drm.struct_mutex);
-
mock_file_free(i915, file);
return err;
}
@@ -1781,13 +1771,9 @@ static int mock_context_barrier(void *arg)
* a request; useful for retiring old state after loading new.
*/
- mutex_lock(&i915->drm.struct_mutex);
-
ctx = mock_context(i915, "mock");
- if (!ctx) {
- err = -ENOMEM;
- goto unlock;
- }
+ if (!ctx)
+ return -ENOMEM;
counter = 0;
err = context_barrier_task(ctx, 0,
@@ -1860,8 +1846,6 @@ static int mock_context_barrier(void *arg)
out:
mock_context_close(ctx);
-unlock:
- mutex_unlock(&i915->drm.struct_mutex);
return err;
#undef pr_fmt
#define pr_fmt(x) x
diff --git a/drivers/gpu/drm/i915/gem/selftests/mock_context.c b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
index 0104f16b1327..fd799defd8be 100644
--- a/drivers/gpu/drm/i915/gem/selftests/mock_context.c
+++ b/drivers/gpu/drm/i915/gem/selftests/mock_context.c
@@ -65,7 +65,7 @@ void mock_context_close(struct i915_gem_context *ctx)
void mock_init_contexts(struct drm_i915_private *i915)
{
- init_contexts(i915);
+ init_contexts(&i915->gem.contexts);
}
struct i915_gem_context *
@@ -74,8 +74,6 @@ live_context(struct drm_i915_private *i915, struct drm_file *file)
struct i915_gem_context *ctx;
int err;
- lockdep_assert_held(&i915->drm.struct_mutex);
-
ctx = i915_gem_create_context(i915, 0);
if (IS_ERR(ctx))
return ctx;
diff --git a/drivers/gpu/drm/i915/gt/intel_context.c b/drivers/gpu/drm/i915/gt/intel_context.c
index 35a40c2820a2..be34d97ac18f 100644
--- a/drivers/gpu/drm/i915/gt/intel_context.c
+++ b/drivers/gpu/drm/i915/gt/intel_context.c
@@ -221,12 +221,20 @@ intel_context_init(struct intel_context *ce,
struct i915_gem_context *ctx,
struct intel_engine_cs *engine)
{
+ struct i915_address_space *vm;
+
GEM_BUG_ON(!engine->cops);
kref_init(&ce->ref);
ce->gem_context = ctx;
- ce->vm = i915_vm_get(ctx->vm ?: &engine->gt->ggtt->vm);
+ rcu_read_lock();
+ vm = rcu_dereference(ctx->vm);
+ if (vm)
+ ce->vm = i915_vm_get(vm);
+ else
+ ce->vm = i915_vm_get(&engine->gt->ggtt->vm);
+ rcu_read_unlock();
if (ctx->timeline)
ce->timeline = intel_timeline_get(ctx->timeline);
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 468438fb47af..471c23d0cffb 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -981,6 +981,11 @@ __execlists_schedule_in(struct i915_request *rq)
ce->lrc_desc |=
(u64)(engine->context_tag++ % NUM_CONTEXT_TAG) <<
GEN11_SW_CTX_ID_SHIFT;
+
+#define GEN11_MAX_CONTEXT_HW_ID (1<<11) /* exclusive */
+/* in Gen12 ID 0x7FF is reserved to indicate idle */
+#define GEN12_MAX_CONTEXT_HW_ID (GEN11_MAX_CONTEXT_HW_ID - 1)
+
BUILD_BUG_ON(NUM_CONTEXT_TAG > GEN12_MAX_CONTEXT_HW_ID);
}
diff --git a/drivers/gpu/drm/i915/gt/selftest_context.c b/drivers/gpu/drm/i915/gt/selftest_context.c
index 86cffbb0a9cb..7c838a57e174 100644
--- a/drivers/gpu/drm/i915/gt/selftest_context.c
+++ b/drivers/gpu/drm/i915/gt/selftest_context.c
@@ -155,13 +155,9 @@ static int live_context_size(void *arg)
* HW tries to write past the end of one.
*/
- mutex_lock(>->i915->drm.struct_mutex);
-
fixme = kernel_context(gt->i915);
- if (IS_ERR(fixme)) {
- err = PTR_ERR(fixme);
- goto unlock;
- }
+ if (IS_ERR(fixme))
+ return PTR_ERR(fixme);
for_each_engine(engine, gt->i915, id) {
struct {
@@ -201,8 +197,6 @@ static int live_context_size(void *arg)
}
kernel_context_close(fixme);
-unlock:
- mutex_unlock(>->i915->drm.struct_mutex);
return err;
}
@@ -305,12 +299,10 @@ static int live_active_context(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(>->i915->drm.struct_mutex);
-
fixme = live_context(gt->i915, file);
if (IS_ERR(fixme)) {
err = PTR_ERR(fixme);
- goto unlock;
+ goto out_file;
}
for_each_engine(engine, gt->i915, id) {
@@ -323,8 +315,7 @@ static int live_active_context(void *arg)
break;
}
-unlock:
- mutex_unlock(>->i915->drm.struct_mutex);
+out_file:
mock_file_free(gt->i915, file);
return err;
}
@@ -418,12 +409,10 @@ static int live_remote_context(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(>->i915->drm.struct_mutex);
-
fixme = live_context(gt->i915, file);
if (IS_ERR(fixme)) {
err = PTR_ERR(fixme);
- goto unlock;
+ goto out_file;
}
for_each_engine(engine, gt->i915, id) {
@@ -436,8 +425,7 @@ static int live_remote_context(void *arg)
break;
}
-unlock:
- mutex_unlock(>->i915->drm.struct_mutex);
+out_file:
mock_file_free(gt->i915, file);
return err;
}
diff --git a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
index ffbb3d23b887..e8a40df79bd0 100644
--- a/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
+++ b/drivers/gpu/drm/i915/gt/selftest_hangcheck.c
@@ -58,9 +58,7 @@ static int hang_init(struct hang *h, struct intel_gt *gt)
memset(h, 0, sizeof(*h));
h->gt = gt;
- mutex_lock(>->i915->drm.struct_mutex);
h->ctx = kernel_context(gt->i915);
- mutex_unlock(>->i915->drm.struct_mutex);
if (IS_ERR(h->ctx))
return PTR_ERR(h->ctx);
@@ -133,7 +131,7 @@ static struct i915_request *
hang_create_request(struct hang *h, struct intel_engine_cs *engine)
{
struct intel_gt *gt = h->gt;
- struct i915_address_space *vm = h->ctx->vm ?: &engine->gt->ggtt->vm;
+ struct i915_address_space *vm = i915_gem_context_get_vm_rcu(h->ctx);
struct drm_i915_gem_object *obj;
struct i915_request *rq = NULL;
struct i915_vma *hws, *vma;
@@ -143,12 +141,15 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine)
int err;
obj = i915_gem_object_create_internal(gt->i915, PAGE_SIZE);
- if (IS_ERR(obj))
+ if (IS_ERR(obj)) {
+ i915_vm_put(vm);
return ERR_CAST(obj);
+ }
vaddr = i915_gem_object_pin_map(obj, i915_coherent_map_type(gt->i915));
if (IS_ERR(vaddr)) {
i915_gem_object_put(obj);
+ i915_vm_put(vm);
return ERR_CAST(vaddr);
}
@@ -159,16 +160,22 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine)
h->batch = vaddr;
vma = i915_vma_instance(h->obj, vm, NULL);
- if (IS_ERR(vma))
+ if (IS_ERR(vma)) {
+ i915_vm_put(vm);
return ERR_CAST(vma);
+ }
hws = i915_vma_instance(h->hws, vm, NULL);
- if (IS_ERR(hws))
+ if (IS_ERR(hws)) {
+ i915_vm_put(vm);
return ERR_CAST(hws);
+ }
err = i915_vma_pin(vma, 0, 0, PIN_USER);
- if (err)
+ if (err) {
+ i915_vm_put(vm);
return ERR_PTR(err);
+ }
err = i915_vma_pin(hws, 0, 0, PIN_USER);
if (err)
@@ -266,6 +273,7 @@ hang_create_request(struct hang *h, struct intel_engine_cs *engine)
i915_vma_unpin(hws);
unpin_vma:
i915_vma_unpin(vma);
+ i915_vm_put(vm);
return err ? ERR_PTR(err) : rq;
}
@@ -382,9 +390,7 @@ static int igt_reset_nop(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(>->i915->drm.struct_mutex);
ctx = live_context(gt->i915, file);
- mutex_unlock(>->i915->drm.struct_mutex);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
goto out;
@@ -458,9 +464,7 @@ static int igt_reset_nop_engine(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(>->i915->drm.struct_mutex);
ctx = live_context(gt->i915, file);
- mutex_unlock(>->i915->drm.struct_mutex);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
goto out;
@@ -705,9 +709,7 @@ static int active_engine(void *data)
return PTR_ERR(file);
for (count = 0; count < ARRAY_SIZE(ctx); count++) {
- mutex_lock(&engine->i915->drm.struct_mutex);
ctx[count] = live_context(engine->i915, file);
- mutex_unlock(&engine->i915->drm.struct_mutex);
if (IS_ERR(ctx[count])) {
err = PTR_ERR(ctx[count]);
while (--count)
@@ -1291,6 +1293,7 @@ static int igt_reset_evict_ppgtt(void *arg)
{
struct intel_gt *gt = arg;
struct i915_gem_context *ctx;
+ struct i915_address_space *vm;
struct drm_file *file;
int err;
@@ -1298,18 +1301,20 @@ static int igt_reset_evict_ppgtt(void *arg)
if (IS_ERR(file))
return PTR_ERR(file);
- mutex_lock(>->i915->drm.struct_mutex);
ctx = live_context(gt->i915, file);
- mutex_unlock(>->i915->drm.struct_mutex);
if (IS_ERR(ctx)) {
err = PTR_ERR(ctx);
goto out;
}
err = 0;
- if (ctx->vm) /* aliasing == global gtt locking, covered above */
- err = __igt_reset_evict_vma(gt, ctx->vm,
+ vm = i915_gem_context_get_vm_rcu(ctx);
+ if (!i915_is_ggtt(vm)) {
+ /* aliasing == global gtt locking, covered above */
+ err = __igt_reset_evict_vma(gt, vm,
evict_vma, EXEC_OBJECT_WRITE);
+ }
+ i915_vm_put(vm);
out:
mock_file_free(gt->i915, file);
diff --git a/drivers/gpu/drm/i915/gt/selftest_lrc.c b/drivers/gpu/drm/i915/gt/selftest_lrc.c
index 04c1cf573642..8dc42c5c7569 100644
--- a/drivers/gpu/drm/i915/gt/selftest_lrc.c
+++ b/drivers/gpu/drm/i915/gt/selftest_lrc.c
@@ -1631,7 +1631,11 @@ static int smoke_submit(struct preempt_smoke *smoke,
int err = 0;
if (batch) {
- vma = i915_vma_instance(batch, ctx->vm, NULL);
+ struct i915_address_space *vm;
+
+ vm = i915_gem_context_get_vm_rcu(ctx);
+ vma = i915_vma_instance(batch, vm, NULL);
+ i915_vm_put(vm);
if (IS_ERR(vma))
return PTR_ERR(vma);
diff --git a/drivers/gpu/drm/i915/gt/selftest_workarounds.c b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
index 4ee2e2babd0d..943b003beba6 100644
--- a/drivers/gpu/drm/i915/gt/selftest_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/selftest_workarounds.c
@@ -355,6 +355,7 @@ static int check_whitelist_across_reset(struct intel_engine_cs *engine,
static struct i915_vma *create_batch(struct i915_gem_context *ctx)
{
struct drm_i915_gem_object *obj;
+ struct i915_address_space *vm;
struct i915_vma *vma;
int err;
@@ -362,7 +363,9 @@ static struct i915_vma *create_batch(struct i915_gem_context *ctx)
if (IS_ERR(obj))
return ERR_CAST(obj);
- vma = i915_vma_instance(obj, ctx->vm, NULL);
+ vm = i915_gem_context_get_vm_rcu(ctx);
+ vma = i915_vma_instance(obj, vm, NULL);
+ i915_vm_put(vm);
if (IS_ERR(vma)) {
err = PTR_ERR(vma);
goto err_obj;
@@ -463,12 +466,15 @@ static int check_dirty_whitelist(struct i915_gem_context *ctx,
0xffff00ff,
0xffffffff,
};
+ struct i915_address_space *vm;
struct i915_vma *scratch;
struct i915_vma *batch;
int err = 0, i, v;
u32 *cs, *results;
- scratch = create_scratch(ctx->vm, 2 * ARRAY_SIZE(values) + 1);
+ vm = i915_gem_context_get_vm_rcu(ctx);
+ scratch = create_scratch(vm, 2 * ARRAY_SIZE(values) + 1);
+ i915_vm_put(vm);
if (IS_ERR(scratch))
return PTR_ERR(scratch);
@@ -1010,6 +1016,7 @@ static int live_isolated_whitelist(void *arg)
return 0;
for (i = 0; i < ARRAY_SIZE(client); i++) {
+ struct i915_address_space *vm;
struct i915_gem_context *c;
c = kernel_context(i915);
@@ -1018,22 +1025,27 @@ static int live_isolated_whitelist(void *arg)
goto err;
}
- client[i].scratch[0] = create_scratch(c->vm, 1024);
+ vm = i915_gem_context_get_vm_rcu(c);
+
+ client[i].scratch[0] = create_scratch(vm, 1024);
if (IS_ERR(client[i].scratch[0])) {
err = PTR_ERR(client[i].scratch[0]);
+ i915_vm_put(vm);
kernel_context_close(c);
goto err;
}
- client[i].scratch[1] = create_scratch(c->vm, 1024);
+ client[i].scratch[1] = create_scratch(vm, 1024);
if (IS_ERR(client[i].scratch[1])) {
err = PTR_ERR(client[i].scratch[1]);
i915_vma_unpin_and_release(&client[i].scratch[0], 0);
+ i915_vm_put(vm);
kernel_context_close(c);
goto err;
}
client[i].ctx = c;
+ i915_vm_put(vm);
}
for_each_engine(engine, i915, id) {
diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
index 03f567084548..6850f1f40241 100644
--- a/drivers/gpu/drm/i915/gvt/scheduler.c
+++ b/drivers/gpu/drm/i915/gvt/scheduler.c
@@ -365,7 +365,8 @@ static void set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload,
struct i915_gem_context *ctx)
{
struct intel_vgpu_mm *mm = workload->shadow_mm;
- struct i915_ppgtt *ppgtt = i915_vm_to_ppgtt(ctx->vm);
+ struct i915_ppgtt *ppgtt =
+ i915_vm_to_ppgtt(i915_gem_context_get_vm_rcu(ctx));
int i = 0;
if (mm->ppgtt_mm.root_entry_type == GTT_TYPE_PPGTT_ROOT_L4_ENTRY) {
@@ -378,6 +379,8 @@ static void set_context_ppgtt_from_shadow(struct intel_vgpu_workload *workload,
px_dma(pd) = mm->ppgtt_mm.shadow_pdps[i];
}
}
+
+ i915_vm_put(&ppgtt->vm);
}
static int
@@ -1230,20 +1233,18 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
struct intel_vgpu_submission *s = &vgpu->submission;
struct intel_engine_cs *engine;
struct i915_gem_context *ctx;
+ struct i915_ppgtt *ppgtt;
enum intel_engine_id i;
int ret;
- mutex_lock(&i915->drm.struct_mutex);
-
ctx = i915_gem_context_create_kernel(i915, I915_PRIORITY_MAX);
- if (IS_ERR(ctx)) {
- ret = PTR_ERR(ctx);
- goto out_unlock;
- }
+ if (IS_ERR(ctx))
+ return PTR_ERR(ctx);
i915_gem_context_set_force_single_submission(ctx);
- i915_context_ppgtt_root_save(s, i915_vm_to_ppgtt(ctx->vm));
+ ppgtt = i915_vm_to_ppgtt(i915_gem_context_get_vm_rcu(ctx));
+ i915_context_ppgtt_root_save(s, ppgtt);
for_each_engine(engine, i915, i) {
struct intel_context *ce;
@@ -1288,12 +1289,12 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
atomic_set(&s->running_workload_num, 0);
bitmap_zero(s->tlb_handle_pending, I915_NUM_ENGINES);
+ i915_vm_put(&ppgtt->vm);
i915_gem_context_put(ctx);
- mutex_unlock(&i915->drm.struct_mutex);
return 0;
out_shadow_ctx:
- i915_context_ppgtt_root_restore(s, i915_vm_to_ppgtt(ctx->vm));
+ i915_context_ppgtt_root_restore(s, ppgtt);
for_each_engine(engine, i915, i) {
if (IS_ERR(s->shadow[i]))
break;
@@ -1301,9 +1302,8 @@ int intel_vgpu_setup_submission(struct intel_vgpu *vgpu)
intel_context_unpin(s->shadow[i]);
intel_context_put(s->shadow[i]);
}
+ i915_vm_put(&ppgtt->vm);
i915_gem_context_put(ctx);
-out_unlock:
- mutex_unlock(&i915->drm.struct_mutex);
return ret;
}
diff --git a/drivers/gpu/drm/i915/i915_debugfs.c b/drivers/gpu/drm/i915/i915_debugfs.c
index 0e90ac608e07..b04cebc26eca 100644
--- a/drivers/gpu/drm/i915/i915_debugfs.c
+++ b/drivers/gpu/drm/i915/i915_debugfs.c
@@ -316,12 +316,18 @@ static void print_context_stats(struct seq_file *m,
struct drm_i915_private *i915)
{
struct file_stats kstats = {};
- struct i915_gem_context *ctx;
+ struct i915_gem_context *ctx, *cn;
- list_for_each_entry(ctx, &i915->contexts.list, link) {
+ spin_lock(&i915->gem.contexts.lock);
+ list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) {
struct i915_gem_engines_iter it;
struct intel_context *ce;
+ if (!kref_get_unless_zero(&ctx->ref))
+ continue;
+
+ spin_unlock(&i915->gem.contexts.lock);
+
for_each_gem_engine(ce,
i915_gem_context_lock_engines(ctx), it) {
intel_context_lock_pinned(ce);
@@ -338,7 +344,9 @@ static void print_context_stats(struct seq_file *m,
i915_gem_context_unlock_engines(ctx);
if (!IS_ERR_OR_NULL(ctx->file_priv)) {
- struct file_stats stats = { .vm = ctx->vm, };
+ struct file_stats stats = {
+ .vm = rcu_access_pointer(ctx->vm),
+ };
struct drm_file *file = ctx->file_priv->file;
struct task_struct *task;
char name[80];
@@ -355,7 +363,12 @@ static void print_context_stats(struct seq_file *m,
print_file_stats(m, name, stats);
}
+
+ spin_lock(&i915->gem.contexts.lock);
+ list_safe_reset_next(ctx, cn, link);
+ i915_gem_context_put(ctx);
}
+ spin_unlock(&i915->gem.contexts.lock);
print_file_stats(m, "[k]contexts", kstats);
}
@@ -363,7 +376,6 @@ static void print_context_stats(struct seq_file *m,
static int i915_gem_object_info(struct seq_file *m, void *data)
{
struct drm_i915_private *i915 = node_to_i915(m->private);
- int ret;
seq_printf(m, "%u shrinkable [%u free] objects, %llu bytes\n",
i915->mm.shrink_count,
@@ -372,12 +384,7 @@ static int i915_gem_object_info(struct seq_file *m, void *data)
seq_putc(m, '\n');
- ret = mutex_lock_interruptible(&i915->drm.struct_mutex);
- if (ret)
- return ret;
-
print_context_stats(m, i915);
- mutex_unlock(&i915->drm.struct_mutex);
return 0;
}
@@ -1579,19 +1586,19 @@ static void describe_ctx_ring(struct seq_file *m, struct intel_ring *ring)
static int i915_context_status(struct seq_file *m, void *unused)
{
- struct drm_i915_private *dev_priv = node_to_i915(m->private);
- struct drm_device *dev = &dev_priv->drm;
- struct i915_gem_context *ctx;
- int ret;
-
- ret = mutex_lock_interruptible(&dev->struct_mutex);
- if (ret)
- return ret;
+ struct drm_i915_private *i915 = node_to_i915(m->private);
+ struct i915_gem_context *ctx, *cn;
- list_for_each_entry(ctx, &dev_priv->contexts.list, link) {
+ spin_lock(&i915->gem.contexts.lock);
+ list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) {
struct i915_gem_engines_iter it;
struct intel_context *ce;
+ if (!kref_get_unless_zero(&ctx->ref))
+ continue;
+
+ spin_unlock(&i915->gem.contexts.lock);
+
seq_puts(m, "HW context ");
if (ctx->pid) {
struct task_struct *task;
@@ -1626,9 +1633,12 @@ static int i915_context_status(struct seq_file *m, void *unused)
i915_gem_context_unlock_engines(ctx);
seq_putc(m, '\n');
- }
- mutex_unlock(&dev->struct_mutex);
+ spin_lock(&i915->gem.contexts.lock);
+ list_safe_reset_next(ctx, cn, link);
+ i915_gem_context_put(ctx);
+ }
+ spin_unlock(&i915->gem.contexts.lock);
return 0;
}
diff --git a/drivers/gpu/drm/i915/i915_drv.c b/drivers/gpu/drm/i915/i915_drv.c
index 5323e4fa55d9..024da582ba0f 100644
--- a/drivers/gpu/drm/i915/i915_drv.c
+++ b/drivers/gpu/drm/i915/i915_drv.c
@@ -1665,10 +1665,8 @@ static void i915_driver_postclose(struct drm_device *dev, struct drm_file *file)
{
struct drm_i915_file_private *file_priv = file->driver_priv;
- mutex_lock(&dev->struct_mutex);
i915_gem_context_close(file);
i915_gem_release(dev, file);
- mutex_unlock(&dev->struct_mutex);
kfree_rcu(file_priv, rcu);
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index cb63b2bd0ce8..35b610d52379 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -1536,25 +1536,6 @@ struct drm_i915_private {
int audio_power_refcount;
u32 audio_freq_cntrl;
- struct {
- struct mutex mutex;
- struct list_head list;
- struct llist_head free_list;
- struct work_struct free_work;
-
- /* The hw wants to have a stable context identifier for the
- * lifetime of the context (for OA, PASID, faults, etc).
- * This is limited in execlists to 21 bits.
- */
- struct ida hw_ida;
-#define MAX_CONTEXT_HW_ID (1<<21) /* exclusive */
-#define MAX_GUC_CONTEXT_HW_ID (1 << 20) /* exclusive */
-#define GEN11_MAX_CONTEXT_HW_ID (1<<11) /* exclusive */
-/* in Gen12 ID 0x7FF is reserved to indicate idle */
-#define GEN12_MAX_CONTEXT_HW_ID (GEN11_MAX_CONTEXT_HW_ID - 1)
- struct list_head hw_id_list;
- } contexts;
-
u32 fdi_rx_config;
/* Shadow for DISPLAY_PHY_CONTROL which can't be safely read */
@@ -1710,6 +1691,14 @@ struct drm_i915_private {
struct {
struct notifier_block pm_notifier;
+
+ struct i915_gem_contexts {
+ spinlock_t lock; /* locks list */
+ struct list_head list;
+
+ struct llist_head free_list;
+ struct work_struct free_work;
+ } contexts;
} gem;
/* For i945gm vblank irq vs. C3 workaround */
diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 5a664bdead8c..f6db415985d5 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -1266,7 +1266,7 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
goto err_unlock;
}
- ret = i915_gem_contexts_init(dev_priv);
+ ret = i915_gem_init_contexts(dev_priv);
if (ret) {
GEM_BUG_ON(ret == -EIO);
goto err_scratch;
@@ -1348,7 +1348,7 @@ int i915_gem_init(struct drm_i915_private *dev_priv)
}
err_context:
if (ret != -EIO)
- i915_gem_contexts_fini(dev_priv);
+ i915_gem_driver_release__contexts(dev_priv);
err_scratch:
intel_gt_driver_release(&dev_priv->gt);
err_unlock:
@@ -1416,11 +1416,9 @@ void i915_gem_driver_remove(struct drm_i915_private *dev_priv)
void i915_gem_driver_release(struct drm_i915_private *dev_priv)
{
- mutex_lock(&dev_priv->drm.struct_mutex);
intel_engines_cleanup(dev_priv);
- i915_gem_contexts_fini(dev_priv);
+ i915_gem_driver_release__contexts(dev_priv);
intel_gt_driver_release(&dev_priv->gt);
- mutex_unlock(&dev_priv->drm.struct_mutex);
intel_wa_list_free(&dev_priv->gt_wa_list);
@@ -1430,7 +1428,7 @@ void i915_gem_driver_release(struct drm_i915_private *dev_priv)
i915_gem_drain_freed_objects(dev_priv);
- WARN_ON(!list_empty(&dev_priv->contexts.list));
+ WARN_ON(!list_empty(&dev_priv->gem.contexts.list));
}
void i915_gem_init_mmio(struct drm_i915_private *i915)
diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.c b/drivers/gpu/drm/i915/i915_gem_gtt.c
index 1d26634ca597..7b15bb891970 100644
--- a/drivers/gpu/drm/i915/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/i915_gem_gtt.c
@@ -1366,7 +1366,9 @@ static int gen8_init_scratch(struct i915_address_space *vm)
if (vm->has_read_only &&
vm->i915->kernel_context &&
vm->i915->kernel_context->vm) {
- struct i915_address_space *clone = vm->i915->kernel_context->vm;
+ struct i915_address_space *clone =
+ rcu_dereference_protected(vm->i915->kernel_context->vm,
+ true); /* static */
GEM_BUG_ON(!clone->has_read_only);
diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index ecfbc37b738b..231388d06c82 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -1853,8 +1853,8 @@ static int gen8_configure_all_contexts(struct i915_perf_stream *stream,
};
#undef ctx_flexeuN
struct intel_engine_cs *engine;
- struct i915_gem_context *ctx;
- int i;
+ struct i915_gem_context *ctx, *cn;
+ int i, err;
for (i = 2; i < ARRAY_SIZE(regs); i++)
regs[i].value = oa_config_flex_reg(oa_config, regs[i].reg);
@@ -1877,16 +1877,27 @@ static int gen8_configure_all_contexts(struct i915_perf_stream *stream,
* context. Contexts idle at the time of reconfiguration are not
* trapped behind the barrier.
*/
- list_for_each_entry(ctx, &i915->contexts.list, link) {
- int err;
-
+ spin_lock(&i915->gem.contexts.lock);
+ list_for_each_entry_safe(ctx, cn, &i915->gem.contexts.list, link) {
if (ctx == i915->kernel_context)
continue;
+ if (!kref_get_unless_zero(&ctx->ref))
+ continue;
+
+ spin_unlock(&i915->gem.contexts.lock);
+
err = gen8_configure_context(ctx, regs, ARRAY_SIZE(regs));
- if (err)
+ if (err) {
+ i915_gem_context_put(ctx);
return err;
+ }
+
+ spin_lock(&i915->gem.contexts.lock);
+ list_safe_reset_next(ctx, cn, link);
+ i915_gem_context_put(ctx);
}
+ spin_unlock(&i915->gem.contexts.lock);
/*
* After updating all other contexts, we need to modify ourselves.
@@ -1895,7 +1906,6 @@ static int gen8_configure_all_contexts(struct i915_perf_stream *stream,
*/
for_each_uabi_engine(engine, i915) {
struct intel_context *ce = engine->kernel_context;
- int err;
if (engine->class != RENDER_CLASS)
continue;
diff --git a/drivers/gpu/drm/i915/i915_sysfs.c b/drivers/gpu/drm/i915/i915_sysfs.c
index 034b8abc5062..2a72407a6380 100644
--- a/drivers/gpu/drm/i915/i915_sysfs.c
+++ b/drivers/gpu/drm/i915/i915_sysfs.c
@@ -144,9 +144,9 @@ static const struct attribute_group media_rc6_attr_group = {
};
#endif
-static int l3_access_valid(struct drm_i915_private *dev_priv, loff_t offset)
+static int l3_access_valid(struct drm_i915_private *i915, loff_t offset)
{
- if (!HAS_L3_DPF(dev_priv))
+ if (!HAS_L3_DPF(i915))
return -EPERM;
if (offset % 4 != 0)
@@ -164,31 +164,24 @@ i915_l3_read(struct file *filp, struct kobject *kobj,
loff_t offset, size_t count)
{
struct device *kdev = kobj_to_dev(kobj);
- struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev);
- struct drm_device *dev = &dev_priv->drm;
+ struct drm_i915_private *i915 = kdev_minor_to_i915(kdev);
int slice = (int)(uintptr_t)attr->private;
int ret;
- count = round_down(count, 4);
-
- ret = l3_access_valid(dev_priv, offset);
+ ret = l3_access_valid(i915, offset);
if (ret)
return ret;
+ count = round_down(count, 4);
count = min_t(size_t, GEN7_L3LOG_SIZE - offset, count);
+ memset(buf, 0, count);
- ret = i915_mutex_lock_interruptible(dev);
- if (ret)
- return ret;
-
- if (dev_priv->l3_parity.remap_info[slice])
+ spin_lock(&i915->gem.contexts.lock);
+ if (i915->l3_parity.remap_info[slice])
memcpy(buf,
- dev_priv->l3_parity.remap_info[slice] + (offset/4),
+ i915->l3_parity.remap_info[slice] + offset / 4,
count);
- else
- memset(buf, 0, count);
-
- mutex_unlock(&dev->struct_mutex);
+ spin_unlock(&i915->gem.contexts.lock);
return count;
}
@@ -199,46 +192,46 @@ i915_l3_write(struct file *filp, struct kobject *kobj,
loff_t offset, size_t count)
{
struct device *kdev = kobj_to_dev(kobj);
- struct drm_i915_private *dev_priv = kdev_minor_to_i915(kdev);
- struct drm_device *dev = &dev_priv->drm;
+ struct drm_i915_private *i915 = kdev_minor_to_i915(kdev);
struct i915_gem_context *ctx;
int slice = (int)(uintptr_t)attr->private;
- u32 **remap_info;
+ u32 *remap_info, *freeme = NULL;
int ret;
- ret = l3_access_valid(dev_priv, offset);
+ ret = l3_access_valid(i915, offset);
if (ret)
return ret;
- ret = i915_mutex_lock_interruptible(dev);
- if (ret)
- return ret;
+ remap_info = kzalloc(GEN7_L3LOG_SIZE, GFP_KERNEL);
+ if (!remap_info)
+ return -ENOMEM;
- remap_info = &dev_priv->l3_parity.remap_info[slice];
- if (!*remap_info) {
- *remap_info = kzalloc(GEN7_L3LOG_SIZE, GFP_KERNEL);
- if (!*remap_info) {
- ret = -ENOMEM;
- goto out;
- }
+ spin_lock(&i915->gem.contexts.lock);
+
+ if (i915->l3_parity.remap_info[slice]) {
+ freeme = remap_info;
+ remap_info = i915->l3_parity.remap_info[slice];
+ } else {
+ i915->l3_parity.remap_info[slice] = remap_info;
}
- /* TODO: Ideally we really want a GPU reset here to make sure errors
- * aren't propagated. Since I cannot find a stable way to reset the GPU
- * at this point it is left as a TODO.
- */
- memcpy(*remap_info + (offset/4), buf, count);
+ count = round_down(count, 4);
+ memcpy(remap_info + offset / 4, buf, count);
/* NB: We defer the remapping until we switch to the context */
- list_for_each_entry(ctx, &dev_priv->contexts.list, link)
- ctx->remap_slice |= (1<<slice);
+ list_for_each_entry(ctx, &i915->gem.contexts.list, link)
+ ctx->remap_slice |= BIT(slice);
- ret = count;
+ spin_unlock(&i915->gem.contexts.lock);
+ kfree(freeme);
-out:
- mutex_unlock(&dev->struct_mutex);
+ /*
+ * TODO: Ideally we really want a GPU reset here to make sure errors
+ * aren't propagated. Since I cannot find a stable way to reset the GPU
+ * at this point it is left as a TODO.
+ */
- return ret;
+ return count;
}
static const struct bin_attribute dpf_attrs = {
diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h
index 1f2cf6cfafb5..7ef7a1e1664c 100644
--- a/drivers/gpu/drm/i915/i915_trace.h
+++ b/drivers/gpu/drm/i915/i915_trace.h
@@ -952,7 +952,7 @@ DECLARE_EVENT_CLASS(i915_context,
TP_fast_assign(
__entry->dev = ctx->i915->drm.primary->index;
__entry->ctx = ctx;
- __entry->vm = ctx->vm;
+ __entry->vm = rcu_access_pointer(ctx->vm);
),
TP_printk("dev=%u, ctx=%p, ctx_vm=%p",
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem.c b/drivers/gpu/drm/i915/selftests/i915_gem.c
index 0346c3e5b6b6..bfa40a5b6d98 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem.c
@@ -138,11 +138,9 @@ static int igt_gem_suspend(void *arg)
return PTR_ERR(file);
err = -ENOMEM;
- mutex_lock(&i915->drm.struct_mutex);
ctx = live_context(i915, file);
if (!IS_ERR(ctx))
err = switch_to_context(i915, ctx);
- mutex_unlock(&i915->drm.struct_mutex);
if (err)
goto out;
@@ -157,9 +155,7 @@ static int igt_gem_suspend(void *arg)
pm_resume(i915);
- mutex_lock(&i915->drm.struct_mutex);
err = switch_to_context(i915, ctx);
- mutex_unlock(&i915->drm.struct_mutex);
out:
mock_file_free(i915, file);
return err;
@@ -177,11 +173,9 @@ static int igt_gem_hibernate(void *arg)
return PTR_ERR(file);
err = -ENOMEM;
- mutex_lock(&i915->drm.struct_mutex);
ctx = live_context(i915, file);
if (!IS_ERR(ctx))
err = switch_to_context(i915, ctx);
- mutex_unlock(&i915->drm.struct_mutex);
if (err)
goto out;
@@ -196,9 +190,7 @@ static int igt_gem_hibernate(void *arg)
pm_resume(i915);
- mutex_lock(&i915->drm.struct_mutex);
err = switch_to_context(i915, ctx);
- mutex_unlock(&i915->drm.struct_mutex);
out:
mock_file_free(i915, file);
return err;
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
index f39f0282e78c..0af9a58d011d 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_evict.c
@@ -473,7 +473,6 @@ static int igt_evict_contexts(void *arg)
}
count = 0;
- mutex_lock(&i915->drm.struct_mutex);
onstack_fence_init(&fence);
do {
struct i915_request *rq;
@@ -510,8 +509,6 @@ static int igt_evict_contexts(void *arg)
count++;
err = 0;
} while(1);
- mutex_unlock(&i915->drm.struct_mutex);
-
onstack_fence_fini(&fence);
pr_info("Submitted %lu contexts/requests on %s\n",
count, engine->name);
diff --git a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
index e40e6cfa51f1..8d8121c02161 100644
--- a/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
+++ b/drivers/gpu/drm/i915/selftests/i915_gem_gtt.c
@@ -1246,6 +1246,7 @@ static int exercise_mock(struct drm_i915_private *i915,
unsigned long end_time))
{
const u64 limit = totalram_pages() << PAGE_SHIFT;
+ struct i915_address_space *vm;
struct i915_gem_context *ctx;
IGT_TIMEOUT(end_time);
int err;
@@ -1254,7 +1255,9 @@ static int exercise_mock(struct drm_i915_private *i915,
if (!ctx)
return -ENOMEM;
- err = func(i915, ctx->vm, 0, min(ctx->vm->total, limit), end_time);
+ vm = i915_gem_context_get_vm_rcu(ctx);
+ err = func(i915, vm, 0, min(vm->total, limit), end_time);
+ i915_vm_put(vm);
mock_context_close(ctx);
return err;
@@ -1801,15 +1804,15 @@ static int igt_cs_tlb(void *arg)
goto out_unlock;
}
- vm = ctx->vm;
- if (!vm)
- goto out_unlock;
+ vm = i915_gem_context_get_vm_rcu(ctx);
+ if (i915_is_ggtt(vm))
+ goto out_vm;
/* Create two pages; dummy we prefill the TLB, and intended */
bbe = i915_gem_object_create_internal(i915, PAGE_SIZE);
if (IS_ERR(bbe)) {
err = PTR_ERR(bbe);
- goto out_unlock;
+ goto out_vm;
}
batch = i915_gem_object_pin_map(bbe, I915_MAP_WC);
@@ -2014,6 +2017,8 @@ static int igt_cs_tlb(void *arg)
i915_gem_object_put(act);
out_put_bbe:
i915_gem_object_put(bbe);
+out_vm:
+ i915_vm_put(vm);
out_unlock:
mutex_unlock(&i915->drm.struct_mutex);
mock_file_free(i915, file);
diff --git a/drivers/gpu/drm/i915/selftests/i915_request.c b/drivers/gpu/drm/i915/selftests/i915_request.c
index d7d68c6a6bd5..0897a7b04944 100644
--- a/drivers/gpu/drm/i915/selftests/i915_request.c
+++ b/drivers/gpu/drm/i915/selftests/i915_request.c
@@ -181,9 +181,7 @@ static int igt_request_rewind(void *arg)
struct intel_context *ce;
int err = -EINVAL;
- mutex_lock(&i915->drm.struct_mutex);
ctx[0] = mock_context(i915, "A");
- mutex_unlock(&i915->drm.struct_mutex);
ce = i915_gem_context_get_engine(ctx[0], RCS0);
GEM_BUG_ON(IS_ERR(ce));
@@ -197,9 +195,7 @@ static int igt_request_rewind(void *arg)
i915_request_get(request);
i915_request_add(request);
- mutex_lock(&i915->drm.struct_mutex);
ctx[1] = mock_context(i915, "B");
- mutex_unlock(&i915->drm.struct_mutex);
ce = i915_gem_context_get_engine(ctx[1], RCS0);
GEM_BUG_ON(IS_ERR(ce));
@@ -438,9 +434,7 @@ static int mock_breadcrumbs_smoketest(void *arg)
}
for (n = 0; n < t.ncontexts; n++) {
- mutex_lock(&t.engine->i915->drm.struct_mutex);
t.contexts[n] = mock_context(t.engine->i915, "mock");
- mutex_unlock(&t.engine->i915->drm.struct_mutex);
if (!t.contexts[n]) {
ret = -ENOMEM;
goto out_contexts;
@@ -734,9 +728,9 @@ static int live_empty_request(void *arg)
static struct i915_vma *recursive_batch(struct drm_i915_private *i915)
{
struct i915_gem_context *ctx = i915->kernel_context;
- struct i915_address_space *vm = ctx->vm ?: &i915->ggtt.vm;
struct drm_i915_gem_object *obj;
const int gen = INTEL_GEN(i915);
+ struct i915_address_space *vm;
struct i915_vma *vma;
u32 *cmd;
int err;
@@ -745,7 +739,9 @@ static struct i915_vma *recursive_batch(struct drm_i915_private *i915)
if (IS_ERR(obj))
return ERR_CAST(obj);
+ vm = i915_gem_context_get_vm_rcu(ctx);
vma = i915_vma_instance(obj, vm, NULL);
+ i915_vm_put(vm);
if (IS_ERR(vma)) {
err = PTR_ERR(vma);
goto err;
@@ -1220,9 +1216,7 @@ static int live_breadcrumbs_smoketest(void *arg)
}
for (n = 0; n < t[0].ncontexts; n++) {
- mutex_lock(&i915->drm.struct_mutex);
t[0].contexts[n] = live_context(i915, file);
- mutex_unlock(&i915->drm.struct_mutex);
if (!t[0].contexts[n]) {
ret = -ENOMEM;
goto out_contexts;
diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c b/drivers/gpu/drm/i915/selftests/i915_vma.c
index ac1ff558eb90..58b5f40a07dd 100644
--- a/drivers/gpu/drm/i915/selftests/i915_vma.c
+++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
@@ -24,6 +24,7 @@
#include <linux/prime_numbers.h>
+#include "gem/i915_gem_context.h"
#include "gem/selftests/mock_context.h"
#include "i915_scatterlist.h"
@@ -38,7 +39,7 @@ static bool assert_vma(struct i915_vma *vma,
{
bool ok = true;
- if (vma->vm != ctx->vm) {
+ if (vma->vm != rcu_access_pointer(ctx->vm)) {
pr_err("VMA created with wrong VM\n");
ok = false;
}
@@ -113,11 +114,13 @@ static int create_vmas(struct drm_i915_private *i915,
list_for_each_entry(obj, objects, st_link) {
for (pinned = 0; pinned <= 1; pinned++) {
list_for_each_entry(ctx, contexts, link) {
- struct i915_address_space *vm = ctx->vm;
+ struct i915_address_space *vm;
struct i915_vma *vma;
int err;
+ vm = i915_gem_context_get_vm_rcu(ctx);
vma = checked_vma_instance(obj, vm, NULL);
+ i915_vm_put(vm);
if (IS_ERR(vma))
return PTR_ERR(vma);
diff --git a/drivers/gpu/drm/i915/selftests/mock_gem_device.c b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
index 4e6cde0d4859..335f37ba98de 100644
--- a/drivers/gpu/drm/i915/selftests/mock_gem_device.c
+++ b/drivers/gpu/drm/i915/selftests/mock_gem_device.c
@@ -59,11 +59,9 @@ static void mock_device_release(struct drm_device *dev)
i915_gem_drain_workqueue(i915);
- mutex_lock(&i915->drm.struct_mutex);
for_each_engine(engine, i915, id)
mock_engine_free(engine);
- i915_gem_contexts_fini(i915);
- mutex_unlock(&i915->drm.struct_mutex);
+ i915_gem_driver_release__contexts(i915);
intel_timelines_fini(i915);
@@ -206,7 +204,7 @@ struct drm_i915_private *mock_gem_device(void)
return i915;
err_context:
- i915_gem_contexts_fini(i915);
+ i915_gem_driver_release__contexts(i915);
err_engine:
mock_engine_free(i915->engine[RCS0]);
err_unlock:
--
2.23.0
More information about the Intel-gfx
mailing list