aboutsummaryrefslogtreecommitdiff
path: root/drivers/gpu
diff options
context:
space:
mode:
authorDave Airlie2024-02-02 13:52:03 +1000
committerDave Airlie2024-02-02 13:52:28 +1000
commit111a3f0afb88e31a6a7b5768d23288e982f12496 (patch)
tree59de887488d5d8df46a090106f4c511711f10144 /drivers/gpu
parent419d8a93757f1fb4a0bd10e9c462a2f6da077ca7 (diff)
parent5f16ee27cd5abd5166e28b2311ac693c204063ff (diff)
Merge tag 'drm-xe-fixes-2024-02-01' of https://gitlab.freedesktop.org/drm/xe/kernel into drm-fixes
UAPI Changes: - Only allow a single user-fence per exec / bind. The reason for this clarification fix is a limitation in the implementation which can be lifted moving forward, if needed. Driver Changes: - A crash fix - A fix for an assert due to missing mem_acces ref - Only allow a single user-fence per exec / bind. - Some sparse warning fixes - Two fixes for compilation failures on various odd combinations of gcc / arch pointed out on LKML. - Fix a fragile partial allocation pointed out on LKML. Cross-driver Change: - A sysfs ABI documentation warning fix This also touches i915 and is acked by i915 maintainers. Signed-off-by: Dave Airlie <airlied@redhat.com> From: Thomas Hellstrom <thomas.hellstrom@linux.intel.com> Link: https://patchwork.freedesktop.org/patch/msgid/ZbuCYdMDVK-kAWC5@fedora
Diffstat (limited to 'drivers/gpu')
-rw-r--r--drivers/gpu/drm/xe/abi/guc_actions_abi.h4
-rw-r--r--drivers/gpu/drm/xe/abi/guc_actions_slpc_abi.h4
-rw-r--r--drivers/gpu/drm/xe/abi/guc_communication_ctb_abi.h8
-rw-r--r--drivers/gpu/drm/xe/abi/guc_klvs_abi.h6
-rw-r--r--drivers/gpu/drm/xe/abi/guc_messages_abi.h20
-rw-r--r--drivers/gpu/drm/xe/xe_exec.c10
-rw-r--r--drivers/gpu/drm/xe/xe_gt_mcr.c4
-rw-r--r--drivers/gpu/drm/xe/xe_gt_pagefault.c11
-rw-r--r--drivers/gpu/drm/xe/xe_guc_pc.c2
-rw-r--r--drivers/gpu/drm/xe/xe_hw_fence.c6
-rw-r--r--drivers/gpu/drm/xe/xe_lrc.c14
-rw-r--r--drivers/gpu/drm/xe/xe_pt.c32
-rw-r--r--drivers/gpu/drm/xe/xe_query.c50
-rw-r--r--drivers/gpu/drm/xe/xe_sync.h5
-rw-r--r--drivers/gpu/drm/xe/xe_vm.c165
-rw-r--r--drivers/gpu/drm/xe/xe_vm.h16
-rw-r--r--drivers/gpu/drm/xe/xe_vm_types.h16
17 files changed, 220 insertions, 153 deletions
diff --git a/drivers/gpu/drm/xe/abi/guc_actions_abi.h b/drivers/gpu/drm/xe/abi/guc_actions_abi.h
index 3062e0e0d467..79ba98a169f9 100644
--- a/drivers/gpu/drm/xe/abi/guc_actions_abi.h
+++ b/drivers/gpu/drm/xe/abi/guc_actions_abi.h
@@ -50,8 +50,8 @@
#define HOST2GUC_SELF_CFG_REQUEST_MSG_LEN (GUC_HXG_REQUEST_MSG_MIN_LEN + 3u)
#define HOST2GUC_SELF_CFG_REQUEST_MSG_0_MBZ GUC_HXG_REQUEST_MSG_0_DATA0
-#define HOST2GUC_SELF_CFG_REQUEST_MSG_1_KLV_KEY (0xffff << 16)
-#define HOST2GUC_SELF_CFG_REQUEST_MSG_1_KLV_LEN (0xffff << 0)
+#define HOST2GUC_SELF_CFG_REQUEST_MSG_1_KLV_KEY (0xffffu << 16)
+#define HOST2GUC_SELF_CFG_REQUEST_MSG_1_KLV_LEN (0xffffu << 0)
#define HOST2GUC_SELF_CFG_REQUEST_MSG_2_VALUE32 GUC_HXG_REQUEST_MSG_n_DATAn
#define HOST2GUC_SELF_CFG_REQUEST_MSG_3_VALUE64 GUC_HXG_REQUEST_MSG_n_DATAn
diff --git a/drivers/gpu/drm/xe/abi/guc_actions_slpc_abi.h b/drivers/gpu/drm/xe/abi/guc_actions_slpc_abi.h
index 811add10c30d..c165e26c0976 100644
--- a/drivers/gpu/drm/xe/abi/guc_actions_slpc_abi.h
+++ b/drivers/gpu/drm/xe/abi/guc_actions_slpc_abi.h
@@ -242,8 +242,8 @@ struct slpc_shared_data {
(HOST2GUC_PC_SLPC_REQUEST_REQUEST_MSG_MIN_LEN + \
HOST2GUC_PC_SLPC_EVENT_MAX_INPUT_ARGS)
#define HOST2GUC_PC_SLPC_REQUEST_MSG_0_MBZ GUC_HXG_REQUEST_MSG_0_DATA0
-#define HOST2GUC_PC_SLPC_REQUEST_MSG_1_EVENT_ID (0xff << 8)
-#define HOST2GUC_PC_SLPC_REQUEST_MSG_1_EVENT_ARGC (0xff << 0)
+#define HOST2GUC_PC_SLPC_REQUEST_MSG_1_EVENT_ID (0xffu << 8)
+#define HOST2GUC_PC_SLPC_REQUEST_MSG_1_EVENT_ARGC (0xffu << 0)
#define HOST2GUC_PC_SLPC_REQUEST_MSG_N_EVENT_DATA_N GUC_HXG_REQUEST_MSG_n_DATAn
#endif
diff --git a/drivers/gpu/drm/xe/abi/guc_communication_ctb_abi.h b/drivers/gpu/drm/xe/abi/guc_communication_ctb_abi.h
index 3b83f907ece4..0b1146d0c997 100644
--- a/drivers/gpu/drm/xe/abi/guc_communication_ctb_abi.h
+++ b/drivers/gpu/drm/xe/abi/guc_communication_ctb_abi.h
@@ -82,11 +82,11 @@ static_assert(sizeof(struct guc_ct_buffer_desc) == 64);
#define GUC_CTB_HDR_LEN 1u
#define GUC_CTB_MSG_MIN_LEN GUC_CTB_HDR_LEN
#define GUC_CTB_MSG_MAX_LEN 256u
-#define GUC_CTB_MSG_0_FENCE (0xffff << 16)
-#define GUC_CTB_MSG_0_FORMAT (0xf << 12)
+#define GUC_CTB_MSG_0_FENCE (0xffffu << 16)
+#define GUC_CTB_MSG_0_FORMAT (0xfu << 12)
#define GUC_CTB_FORMAT_HXG 0u
-#define GUC_CTB_MSG_0_RESERVED (0xf << 8)
-#define GUC_CTB_MSG_0_NUM_DWORDS (0xff << 0)
+#define GUC_CTB_MSG_0_RESERVED (0xfu << 8)
+#define GUC_CTB_MSG_0_NUM_DWORDS (0xffu << 0)
/**
* DOC: CTB HXG Message
diff --git a/drivers/gpu/drm/xe/abi/guc_klvs_abi.h b/drivers/gpu/drm/xe/abi/guc_klvs_abi.h
index 47094b9b044c..0400bc0fccdc 100644
--- a/drivers/gpu/drm/xe/abi/guc_klvs_abi.h
+++ b/drivers/gpu/drm/xe/abi/guc_klvs_abi.h
@@ -31,9 +31,9 @@
*/
#define GUC_KLV_LEN_MIN 1u
-#define GUC_KLV_0_KEY (0xffff << 16)
-#define GUC_KLV_0_LEN (0xffff << 0)
-#define GUC_KLV_n_VALUE (0xffffffff << 0)
+#define GUC_KLV_0_KEY (0xffffu << 16)
+#define GUC_KLV_0_LEN (0xffffu << 0)
+#define GUC_KLV_n_VALUE (0xffffffffu << 0)
/**
* DOC: GuC Self Config KLVs
diff --git a/drivers/gpu/drm/xe/abi/guc_messages_abi.h b/drivers/gpu/drm/xe/abi/guc_messages_abi.h
index 3d199016cf88..29e414c82d56 100644
--- a/drivers/gpu/drm/xe/abi/guc_messages_abi.h
+++ b/drivers/gpu/drm/xe/abi/guc_messages_abi.h
@@ -40,18 +40,18 @@
*/
#define GUC_HXG_MSG_MIN_LEN 1u
-#define GUC_HXG_MSG_0_ORIGIN (0x1 << 31)
+#define GUC_HXG_MSG_0_ORIGIN (0x1u << 31)
#define GUC_HXG_ORIGIN_HOST 0u
#define GUC_HXG_ORIGIN_GUC 1u
-#define GUC_HXG_MSG_0_TYPE (0x7 << 28)
+#define GUC_HXG_MSG_0_TYPE (0x7u << 28)
#define GUC_HXG_TYPE_REQUEST 0u
#define GUC_HXG_TYPE_EVENT 1u
#define GUC_HXG_TYPE_NO_RESPONSE_BUSY 3u
#define GUC_HXG_TYPE_NO_RESPONSE_RETRY 5u
#define GUC_HXG_TYPE_RESPONSE_FAILURE 6u
#define GUC_HXG_TYPE_RESPONSE_SUCCESS 7u
-#define GUC_HXG_MSG_0_AUX (0xfffffff << 0)
-#define GUC_HXG_MSG_n_PAYLOAD (0xffffffff << 0)
+#define GUC_HXG_MSG_0_AUX (0xfffffffu << 0)
+#define GUC_HXG_MSG_n_PAYLOAD (0xffffffffu << 0)
/**
* DOC: HXG Request
@@ -85,8 +85,8 @@
*/
#define GUC_HXG_REQUEST_MSG_MIN_LEN GUC_HXG_MSG_MIN_LEN
-#define GUC_HXG_REQUEST_MSG_0_DATA0 (0xfff << 16)
-#define GUC_HXG_REQUEST_MSG_0_ACTION (0xffff << 0)
+#define GUC_HXG_REQUEST_MSG_0_DATA0 (0xfffu << 16)
+#define GUC_HXG_REQUEST_MSG_0_ACTION (0xffffu << 0)
#define GUC_HXG_REQUEST_MSG_n_DATAn GUC_HXG_MSG_n_PAYLOAD
/**
@@ -117,8 +117,8 @@
*/
#define GUC_HXG_EVENT_MSG_MIN_LEN GUC_HXG_MSG_MIN_LEN
-#define GUC_HXG_EVENT_MSG_0_DATA0 (0xfff << 16)
-#define GUC_HXG_EVENT_MSG_0_ACTION (0xffff << 0)
+#define GUC_HXG_EVENT_MSG_0_DATA0 (0xfffu << 16)
+#define GUC_HXG_EVENT_MSG_0_ACTION (0xffffu << 0)
#define GUC_HXG_EVENT_MSG_n_DATAn GUC_HXG_MSG_n_PAYLOAD
/**
@@ -188,8 +188,8 @@
*/
#define GUC_HXG_FAILURE_MSG_LEN GUC_HXG_MSG_MIN_LEN
-#define GUC_HXG_FAILURE_MSG_0_HINT (0xfff << 16)
-#define GUC_HXG_FAILURE_MSG_0_ERROR (0xffff << 0)
+#define GUC_HXG_FAILURE_MSG_0_HINT (0xfffu << 16)
+#define GUC_HXG_FAILURE_MSG_0_ERROR (0xffffu << 0)
/**
* DOC: HXG Response
diff --git a/drivers/gpu/drm/xe/xe_exec.c b/drivers/gpu/drm/xe/xe_exec.c
index b853feed9ccc..17f26952e665 100644
--- a/drivers/gpu/drm/xe/xe_exec.c
+++ b/drivers/gpu/drm/xe/xe_exec.c
@@ -111,7 +111,7 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
u64 addresses[XE_HW_ENGINE_MAX_INSTANCE];
struct drm_gpuvm_exec vm_exec = {.extra.fn = xe_exec_fn};
struct drm_exec *exec = &vm_exec.exec;
- u32 i, num_syncs = 0;
+ u32 i, num_syncs = 0, num_ufence = 0;
struct xe_sched_job *job;
struct dma_fence *rebind_fence;
struct xe_vm *vm;
@@ -157,6 +157,14 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
SYNC_PARSE_FLAG_LR_MODE : 0));
if (err)
goto err_syncs;
+
+ if (xe_sync_is_ufence(&syncs[i]))
+ num_ufence++;
+ }
+
+ if (XE_IOCTL_DBG(xe, num_ufence > 1)) {
+ err = -EINVAL;
+ goto err_syncs;
}
if (xe_exec_queue_is_parallel(q)) {
diff --git a/drivers/gpu/drm/xe/xe_gt_mcr.c b/drivers/gpu/drm/xe/xe_gt_mcr.c
index 77925b35cf8d..8546cd3cc50d 100644
--- a/drivers/gpu/drm/xe/xe_gt_mcr.c
+++ b/drivers/gpu/drm/xe/xe_gt_mcr.c
@@ -480,7 +480,7 @@ static bool xe_gt_mcr_get_nonterminated_steering(struct xe_gt *gt,
* to synchronize with external clients (e.g., firmware), so a semaphore
* register will also need to be taken.
*/
-static void mcr_lock(struct xe_gt *gt)
+static void mcr_lock(struct xe_gt *gt) __acquires(&gt->mcr_lock)
{
struct xe_device *xe = gt_to_xe(gt);
int ret = 0;
@@ -500,7 +500,7 @@ static void mcr_lock(struct xe_gt *gt)
drm_WARN_ON_ONCE(&xe->drm, ret == -ETIMEDOUT);
}
-static void mcr_unlock(struct xe_gt *gt)
+static void mcr_unlock(struct xe_gt *gt) __releases(&gt->mcr_lock)
{
/* Release hardware semaphore - this is done by writing 1 to the register */
if (GRAPHICS_VERx100(gt_to_xe(gt)) >= 1270)
diff --git a/drivers/gpu/drm/xe/xe_gt_pagefault.c b/drivers/gpu/drm/xe/xe_gt_pagefault.c
index 59a70d2e0a7a..9c2fe1697d6e 100644
--- a/drivers/gpu/drm/xe/xe_gt_pagefault.c
+++ b/drivers/gpu/drm/xe/xe_gt_pagefault.c
@@ -165,7 +165,8 @@ retry_userptr:
goto unlock_vm;
}
- if (!xe_vma_is_userptr(vma) || !xe_vma_userptr_check_repin(vma)) {
+ if (!xe_vma_is_userptr(vma) ||
+ !xe_vma_userptr_check_repin(to_userptr_vma(vma))) {
downgrade_write(&vm->lock);
write_locked = false;
}
@@ -181,11 +182,13 @@ retry_userptr:
/* TODO: Validate fault */
if (xe_vma_is_userptr(vma) && write_locked) {
+ struct xe_userptr_vma *uvma = to_userptr_vma(vma);
+
spin_lock(&vm->userptr.invalidated_lock);
- list_del_init(&vma->userptr.invalidate_link);
+ list_del_init(&uvma->userptr.invalidate_link);
spin_unlock(&vm->userptr.invalidated_lock);
- ret = xe_vma_userptr_pin_pages(vma);
+ ret = xe_vma_userptr_pin_pages(uvma);
if (ret)
goto unlock_vm;
@@ -220,7 +223,7 @@ retry_userptr:
dma_fence_put(fence);
if (xe_vma_is_userptr(vma))
- ret = xe_vma_userptr_check_repin(vma);
+ ret = xe_vma_userptr_check_repin(to_userptr_vma(vma));
vma->usm.tile_invalidated &= ~BIT(tile->id);
unlock_dma_resv:
diff --git a/drivers/gpu/drm/xe/xe_guc_pc.c b/drivers/gpu/drm/xe/xe_guc_pc.c
index f71085228cb3..d91702592520 100644
--- a/drivers/gpu/drm/xe/xe_guc_pc.c
+++ b/drivers/gpu/drm/xe/xe_guc_pc.c
@@ -963,7 +963,9 @@ void xe_guc_pc_fini(struct xe_guc_pc *pc)
struct xe_device *xe = pc_to_xe(pc);
if (xe->info.skip_guc_pc) {
+ xe_device_mem_access_get(xe);
xe_gt_idle_disable_c6(pc_to_gt(pc));
+ xe_device_mem_access_put(xe);
return;
}
diff --git a/drivers/gpu/drm/xe/xe_hw_fence.c b/drivers/gpu/drm/xe/xe_hw_fence.c
index a6094c81f2ad..a5de3e7b0bd6 100644
--- a/drivers/gpu/drm/xe/xe_hw_fence.c
+++ b/drivers/gpu/drm/xe/xe_hw_fence.c
@@ -217,13 +217,13 @@ struct xe_hw_fence *xe_hw_fence_create(struct xe_hw_fence_ctx *ctx,
if (!fence)
return ERR_PTR(-ENOMEM);
- dma_fence_init(&fence->dma, &xe_hw_fence_ops, &ctx->irq->lock,
- ctx->dma_fence_ctx, ctx->next_seqno++);
-
fence->ctx = ctx;
fence->seqno_map = seqno_map;
INIT_LIST_HEAD(&fence->irq_link);
+ dma_fence_init(&fence->dma, &xe_hw_fence_ops, &ctx->irq->lock,
+ ctx->dma_fence_ctx, ctx->next_seqno++);
+
trace_xe_hw_fence_create(fence);
return fence;
diff --git a/drivers/gpu/drm/xe/xe_lrc.c b/drivers/gpu/drm/xe/xe_lrc.c
index b7fa3831b684..0ec5ad2539f1 100644
--- a/drivers/gpu/drm/xe/xe_lrc.c
+++ b/drivers/gpu/drm/xe/xe_lrc.c
@@ -21,10 +21,10 @@
#include "xe_map.h"
#include "xe_vm.h"
-#define CTX_VALID (1 << 0)
-#define CTX_PRIVILEGE (1 << 8)
-#define CTX_ADDRESSING_MODE_SHIFT 3
-#define LEGACY_64B_CONTEXT 3
+#define LRC_VALID (1 << 0)
+#define LRC_PRIVILEGE (1 << 8)
+#define LRC_ADDRESSING_MODE_SHIFT 3
+#define LRC_LEGACY_64B_CONTEXT 3
#define ENGINE_CLASS_SHIFT 61
#define ENGINE_INSTANCE_SHIFT 48
@@ -762,15 +762,15 @@ int xe_lrc_init(struct xe_lrc *lrc, struct xe_hw_engine *hwe,
(q->usm.acc_notify << ACC_NOTIFY_S) |
q->usm.acc_trigger);
- lrc->desc = CTX_VALID;
- lrc->desc |= LEGACY_64B_CONTEXT << CTX_ADDRESSING_MODE_SHIFT;
+ lrc->desc = LRC_VALID;
+ lrc->desc |= LRC_LEGACY_64B_CONTEXT << LRC_ADDRESSING_MODE_SHIFT;
/* TODO: Priority */
/* While this appears to have something about privileged batches or
* some such, it really just means PPGTT mode.
*/
if (vm)
- lrc->desc |= CTX_PRIVILEGE;
+ lrc->desc |= LRC_PRIVILEGE;
if (GRAPHICS_VERx100(xe) < 1250) {
lrc->desc |= (u64)hwe->instance << ENGINE_INSTANCE_SHIFT;
diff --git a/drivers/gpu/drm/xe/xe_pt.c b/drivers/gpu/drm/xe/xe_pt.c
index de1030a47588..e45b37c3f0c2 100644
--- a/drivers/gpu/drm/xe/xe_pt.c
+++ b/drivers/gpu/drm/xe/xe_pt.c
@@ -618,8 +618,8 @@ xe_pt_stage_bind(struct xe_tile *tile, struct xe_vma *vma,
if (!xe_vma_is_null(vma)) {
if (xe_vma_is_userptr(vma))
- xe_res_first_sg(vma->userptr.sg, 0, xe_vma_size(vma),
- &curs);
+ xe_res_first_sg(to_userptr_vma(vma)->userptr.sg, 0,
+ xe_vma_size(vma), &curs);
else if (xe_bo_is_vram(bo) || xe_bo_is_stolen(bo))
xe_res_first(bo->ttm.resource, xe_vma_bo_offset(vma),
xe_vma_size(vma), &curs);
@@ -906,17 +906,17 @@ static void xe_vm_dbg_print_entries(struct xe_device *xe,
#ifdef CONFIG_DRM_XE_USERPTR_INVAL_INJECT
-static int xe_pt_userptr_inject_eagain(struct xe_vma *vma)
+static int xe_pt_userptr_inject_eagain(struct xe_userptr_vma *uvma)
{
- u32 divisor = vma->userptr.divisor ? vma->userptr.divisor : 2;
+ u32 divisor = uvma->userptr.divisor ? uvma->userptr.divisor : 2;
static u32 count;
if (count++ % divisor == divisor - 1) {
- struct xe_vm *vm = xe_vma_vm(vma);
+ struct xe_vm *vm = xe_vma_vm(&uvma->vma);
- vma->userptr.divisor = divisor << 1;
+ uvma->userptr.divisor = divisor << 1;
spin_lock(&vm->userptr.invalidated_lock);
- list_move_tail(&vma->userptr.invalidate_link,
+ list_move_tail(&uvma->userptr.invalidate_link,
&vm->userptr.invalidated);
spin_unlock(&vm->userptr.invalidated_lock);
return true;
@@ -927,7 +927,7 @@ static int xe_pt_userptr_inject_eagain(struct xe_vma *vma)
#else
-static bool xe_pt_userptr_inject_eagain(struct xe_vma *vma)
+static bool xe_pt_userptr_inject_eagain(struct xe_userptr_vma *uvma)
{
return false;
}
@@ -1000,9 +1000,9 @@ static int xe_pt_userptr_pre_commit(struct xe_migrate_pt_update *pt_update)
{
struct xe_pt_migrate_pt_update *userptr_update =
container_of(pt_update, typeof(*userptr_update), base);
- struct xe_vma *vma = pt_update->vma;
- unsigned long notifier_seq = vma->userptr.notifier_seq;
- struct xe_vm *vm = xe_vma_vm(vma);
+ struct xe_userptr_vma *uvma = to_userptr_vma(pt_update->vma);
+ unsigned long notifier_seq = uvma->userptr.notifier_seq;
+ struct xe_vm *vm = xe_vma_vm(&uvma->vma);
int err = xe_pt_vm_dependencies(pt_update->job,
&vm->rftree[pt_update->tile_id],
pt_update->start,
@@ -1023,7 +1023,7 @@ static int xe_pt_userptr_pre_commit(struct xe_migrate_pt_update *pt_update)
*/
do {
down_read(&vm->userptr.notifier_lock);
- if (!mmu_interval_read_retry(&vma->userptr.notifier,
+ if (!mmu_interval_read_retry(&uvma->userptr.notifier,
notifier_seq))
break;
@@ -1032,11 +1032,11 @@ static int xe_pt_userptr_pre_commit(struct xe_migrate_pt_update *pt_update)
if (userptr_update->bind)
return -EAGAIN;
- notifier_seq = mmu_interval_read_begin(&vma->userptr.notifier);
+ notifier_seq = mmu_interval_read_begin(&uvma->userptr.notifier);
} while (true);
/* Inject errors to test_whether they are handled correctly */
- if (userptr_update->bind && xe_pt_userptr_inject_eagain(vma)) {
+ if (userptr_update->bind && xe_pt_userptr_inject_eagain(uvma)) {
up_read(&vm->userptr.notifier_lock);
return -EAGAIN;
}
@@ -1297,7 +1297,7 @@ __xe_pt_bind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_exec_queue
vma->tile_present |= BIT(tile->id);
if (bind_pt_update.locked) {
- vma->userptr.initial_bind = true;
+ to_userptr_vma(vma)->userptr.initial_bind = true;
up_read(&vm->userptr.notifier_lock);
xe_bo_put_commit(&deferred);
}
@@ -1642,7 +1642,7 @@ __xe_pt_unbind_vma(struct xe_tile *tile, struct xe_vma *vma, struct xe_exec_queu
if (!vma->tile_present) {
spin_lock(&vm->userptr.invalidated_lock);
- list_del_init(&vma->userptr.invalidate_link);
+ list_del_init(&to_userptr_vma(vma)->userptr.invalidate_link);
spin_unlock(&vm->userptr.invalidated_lock);
}
up_read(&vm->userptr.notifier_lock);
diff --git a/drivers/gpu/drm/xe/xe_query.c b/drivers/gpu/drm/xe/xe_query.c
index 9b35673b286c..7e924faeeea0 100644
--- a/drivers/gpu/drm/xe/xe_query.c
+++ b/drivers/gpu/drm/xe/xe_query.c
@@ -459,21 +459,21 @@ static size_t calc_topo_query_size(struct xe_device *xe)
sizeof_field(struct xe_gt, fuse_topo.eu_mask_per_dss));
}
-static void __user *copy_mask(void __user *ptr,
- struct drm_xe_query_topology_mask *topo,
- void *mask, size_t mask_size)
+static int copy_mask(void __user **ptr,
+ struct drm_xe_query_topology_mask *topo,
+ void *mask, size_t mask_size)
{
topo->num_bytes = mask_size;
- if (copy_to_user(ptr, topo, sizeof(*topo)))
- return ERR_PTR(-EFAULT);
- ptr += sizeof(topo);
+ if (copy_to_user(*ptr, topo, sizeof(*topo)))
+ return -EFAULT;
+ *ptr += sizeof(topo);
- if (copy_to_user(ptr, mask, mask_size))
- return ERR_PTR(-EFAULT);
- ptr += mask_size;
+ if (copy_to_user(*ptr, mask, mask_size))
+ return -EFAULT;
+ *ptr += mask_size;
- return ptr;
+ return 0;
}
static int query_gt_topology(struct xe_device *xe,
@@ -493,28 +493,28 @@ static int query_gt_topology(struct xe_device *xe,
}
for_each_gt(gt, xe, id) {
+ int err;
+
topo.gt_id = id;
topo.type = DRM_XE_TOPO_DSS_GEOMETRY;
- query_ptr = copy_mask(query_ptr, &topo,
- gt->fuse_topo.g_dss_mask,
- sizeof(gt->fuse_topo.g_dss_mask));
- if (IS_ERR(query_ptr))
- return PTR_ERR(query_ptr);
+ err = copy_mask(&query_ptr, &topo, gt->fuse_topo.g_dss_mask,
+ sizeof(gt->fuse_topo.g_dss_mask));
+ if (err)
+ return err;
topo.type = DRM_XE_TOPO_DSS_COMPUTE;
- query_ptr = copy_mask(query_ptr, &topo,
- gt->fuse_topo.c_dss_mask,
- sizeof(gt->fuse_topo.c_dss_mask));
- if (IS_ERR(query_ptr))
- return PTR_ERR(query_ptr);
+ err = copy_mask(&query_ptr, &topo, gt->fuse_topo.c_dss_mask,
+ sizeof(gt->fuse_topo.c_dss_mask));
+ if (err)
+ return err;
topo.type = DRM_XE_TOPO_EU_PER_DSS;
- query_ptr = copy_mask(query_ptr, &topo,
- gt->fuse_topo.eu_mask_per_dss,
- sizeof(gt->fuse_topo.eu_mask_per_dss));
- if (IS_ERR(query_ptr))
- return PTR_ERR(query_ptr);
+ err = copy_mask(&query_ptr, &topo,
+ gt->fuse_topo.eu_mask_per_dss,
+ sizeof(gt->fuse_topo.eu_mask_per_dss));
+ if (err)
+ return err;
}
return 0;
diff --git a/drivers/gpu/drm/xe/xe_sync.h b/drivers/gpu/drm/xe/xe_sync.h
index d284afbe917c..f43cdcaca6c5 100644
--- a/drivers/gpu/drm/xe/xe_sync.h
+++ b/drivers/gpu/drm/xe/xe_sync.h
@@ -33,4 +33,9 @@ struct dma_fence *
xe_sync_in_fence_get(struct xe_sync_entry *sync, int num_sync,
struct xe_exec_queue *q, struct xe_vm *vm);
+static inline bool xe_sync_is_ufence(struct xe_sync_entry *sync)
+{
+ return !!sync->ufence;
+}
+
#endif
diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
index 53833ab81424..30db264d34a3 100644
--- a/drivers/gpu/drm/xe/xe_vm.c
+++ b/drivers/gpu/drm/xe/xe_vm.c
@@ -46,7 +46,7 @@ static struct drm_gem_object *xe_vm_obj(struct xe_vm *vm)
/**
* xe_vma_userptr_check_repin() - Advisory check for repin needed
- * @vma: The userptr vma
+ * @uvma: The userptr vma
*
* Check if the userptr vma has been invalidated since last successful
* repin. The check is advisory only and can the function can be called
@@ -56,15 +56,17 @@ static struct drm_gem_object *xe_vm_obj(struct xe_vm *vm)
*
* Return: 0 if userptr vma is valid, -EAGAIN otherwise; repin recommended.
*/
-int xe_vma_userptr_check_repin(struct xe_vma *vma)
+int xe_vma_userptr_check_repin(struct xe_userptr_vma *uvma)
{
- return mmu_interval_check_retry(&vma->userptr.notifier,
- vma->userptr.notifier_seq) ?
+ return mmu_interval_check_retry(&uvma->userptr.notifier,
+ uvma->userptr.notifier_seq) ?
-EAGAIN : 0;
}
-int xe_vma_userptr_pin_pages(struct xe_vma *vma)
+int xe_vma_userptr_pin_pages(struct xe_userptr_vma *uvma)
{
+ struct xe_userptr *userptr = &uvma->userptr;
+ struct xe_vma *vma = &uvma->vma;
struct xe_vm *vm = xe_vma_vm(vma);
struct xe_device *xe = vm->xe;
const unsigned long num_pages = xe_vma_size(vma) >> PAGE_SHIFT;
@@ -80,30 +82,30 @@ retry:
if (vma->gpuva.flags & XE_VMA_DESTROYED)
return 0;
- notifier_seq = mmu_interval_read_begin(&vma->userptr.notifier);
- if (notifier_seq == vma->userptr.notifier_seq)
+ notifier_seq = mmu_interval_read_begin(&userptr->notifier);
+ if (notifier_seq == userptr->notifier_seq)
return 0;
pages = kvmalloc_array(num_pages, sizeof(*pages), GFP_KERNEL);
if (!pages)
return -ENOMEM;
- if (vma->userptr.sg) {
+ if (userptr->sg) {
dma_unmap_sgtable(xe->drm.dev,
- vma->userptr.sg,
+ userptr->sg,
read_only ? DMA_TO_DEVICE :
DMA_BIDIRECTIONAL, 0);
- sg_free_table(vma->userptr.sg);
- vma->userptr.sg = NULL;
+ sg_free_table(userptr->sg);
+ userptr->sg = NULL;
}
pinned = ret = 0;
if (in_kthread) {
- if (!mmget_not_zero(vma->userptr.notifier.mm)) {
+ if (!mmget_not_zero(userptr->notifier.mm)) {
ret = -EFAULT;
goto mm_closed;
}
- kthread_use_mm(vma->userptr.notifier.mm);
+ kthread_use_mm(userptr->notifier.mm);
}
while (pinned < num_pages) {
@@ -123,32 +125,32 @@ retry:
}
if (in_kthread) {
- kthread_unuse_mm(vma->userptr.notifier.mm);
- mmput(vma->userptr.notifier.mm);
+ kthread_unuse_mm(userptr->notifier.mm);
+ mmput(userptr->notifier.mm);
}
mm_closed:
if (ret)
goto out;
- ret = sg_alloc_table_from_pages_segment(&vma->userptr.sgt, pages,
+ ret = sg_alloc_table_from_pages_segment(&userptr->sgt, pages,
pinned, 0,
(u64)pinned << PAGE_SHIFT,
xe_sg_segment_size(xe->drm.dev),
GFP_KERNEL);
if (ret) {
- vma->userptr.sg = NULL;
+ userptr->sg = NULL;
goto out;
}
- vma->userptr.sg = &vma->userptr.sgt;
+ userptr->sg = &userptr->sgt;
- ret = dma_map_sgtable(xe->drm.dev, vma->userptr.sg,
+ ret = dma_map_sgtable(xe->drm.dev, userptr->sg,
read_only ? DMA_TO_DEVICE :
DMA_BIDIRECTIONAL,
DMA_ATTR_SKIP_CPU_SYNC |
DMA_ATTR_NO_KERNEL_MAPPING);
if (ret) {
- sg_free_table(vma->userptr.sg);
- vma->userptr.sg = NULL;
+ sg_free_table(userptr->sg);
+ userptr->sg = NULL;
goto out;
}
@@ -167,8 +169,8 @@ out:
kvfree(pages);
if (!(ret < 0)) {
- vma->userptr.notifier_seq = notifier_seq;
- if (xe_vma_userptr_check_repin(vma) == -EAGAIN)
+ userptr->notifier_seq = notifier_seq;
+ if (xe_vma_userptr_check_repin(uvma) == -EAGAIN)
goto retry;
}
@@ -635,7 +637,9 @@ static bool vma_userptr_invalidate(struct mmu_interval_notifier *mni,
const struct mmu_notifier_range *range,
unsigned long cur_seq)
{
- struct xe_vma *vma = container_of(mni, struct xe_vma, userptr.notifier);
+ struct xe_userptr *userptr = container_of(mni, typeof(*userptr), notifier);
+ struct xe_userptr_vma *uvma = container_of(userptr, typeof(*uvma), userptr);
+ struct xe_vma *vma = &uvma->vma;
struct xe_vm *vm = xe_vma_vm(vma);
struct dma_resv_iter cursor;
struct dma_fence *fence;
@@ -651,7 +655,7 @@ static bool vma_userptr_invalidate(struct mmu_interval_notifier *mni,
mmu_interval_set_seq(mni, cur_seq);
/* No need to stop gpu access if the userptr is not yet bound. */
- if (!vma->userptr.initial_bind) {
+ if (!userptr->initial_bind) {
up_write(&vm->userptr.notifier_lock);
return true;
}
@@ -663,7 +667,7 @@ static bool vma_userptr_invalidate(struct mmu_interval_notifier *mni,
if (!xe_vm_in_fault_mode(vm) &&
!(vma->gpuva.flags & XE_VMA_DESTROYED) && vma->tile_present) {
spin_lock(&vm->userptr.invalidated_lock);
- list_move_tail(&vma->userptr.invalidate_link,
+ list_move_tail(&userptr->invalidate_link,
&vm->userptr.invalidated);
spin_unlock(&vm->userptr.invalidated_lock);
}
@@ -703,7 +707,7 @@ static const struct mmu_interval_notifier_ops vma_userptr_notifier_ops = {
int xe_vm_userptr_pin(struct xe_vm *vm)
{
- struct xe_vma *vma, *next;
+ struct xe_userptr_vma *uvma, *next;
int err = 0;
LIST_HEAD(tmp_evict);
@@ -711,22 +715,23 @@ int xe_vm_userptr_pin(struct xe_vm *vm)
/* Collect invalidated userptrs */
spin_lock(&vm->userptr.invalidated_lock);
- list_for_each_entry_safe(vma, next, &vm->userptr.invalidated,
+ list_for_each_entry_safe(uvma, next, &vm->userptr.invalidated,
userptr.invalidate_link) {
- list_del_init(&vma->userptr.invalidate_link);
- list_move_tail(&vma->combined_links.userptr,
+ list_del_init(&uvma->userptr.invalidate_link);
+ list_move_tail(&uvma->userptr.repin_link,
&vm->userptr.repin_list);
}
spin_unlock(&vm->userptr.invalidated_lock);
/* Pin and move to temporary list */
- list_for_each_entry_safe(vma, next, &vm->userptr.repin_list,
- combined_links.userptr) {
- err = xe_vma_userptr_pin_pages(vma);
+ list_for_each_entry_safe(uvma, next, &vm->userptr.repin_list,
+ userptr.repin_link) {
+ err = xe_vma_userptr_pin_pages(uvma);
if (err < 0)
return err;
- list_move_tail(&vma->combined_links.userptr, &vm->rebind_list);
+ list_del_init(&uvma->userptr.repin_link);
+ list_move_tail(&uvma->vma.combined_links.rebind, &vm->rebind_list);
}
return 0;
@@ -782,6 +787,14 @@ struct dma_fence *xe_vm_rebind(struct xe_vm *vm, bool rebind_worker)
return fence;
}
+static void xe_vma_free(struct xe_vma *vma)
+{
+ if (xe_vma_is_userptr(vma))
+ kfree(to_userptr_vma(vma));
+ else
+ kfree(vma);
+}
+
#define VMA_CREATE_FLAG_READ_ONLY BIT(0)
#define VMA_CREATE_FLAG_IS_NULL BIT(1)
@@ -800,14 +813,26 @@ static struct xe_vma *xe_vma_create(struct xe_vm *vm,
xe_assert(vm->xe, start < end);
xe_assert(vm->xe, end < vm->size);
- if (!bo && !is_null) /* userptr */
+ /*
+ * Allocate and ensure that the xe_vma_is_userptr() return
+ * matches what was allocated.
+ */
+ if (!bo && !is_null) {
+ struct xe_userptr_vma *uvma = kzalloc(sizeof(*uvma), GFP_KERNEL);
+
+ if (!uvma)
+ return ERR_PTR(-ENOMEM);
+
+ vma = &uvma->vma;
+ } else {
vma = kzalloc(sizeof(*vma), GFP_KERNEL);
- else
- vma = kzalloc(sizeof(*vma) - sizeof(struct xe_userptr),
- GFP_KERNEL);
- if (!vma) {
- vma = ERR_PTR(-ENOMEM);
- return vma;
+ if (!vma)
+ return ERR_PTR(-ENOMEM);
+
+ if (is_null)
+ vma->gpuva.flags |= DRM_GPUVA_SPARSE;
+ if (bo)
+ vma->gpuva.gem.obj = &bo->ttm.base;
}
INIT_LIST_HEAD(&vma->combined_links.rebind);
@@ -818,8 +843,6 @@ static struct xe_vma *xe_vma_create(struct xe_vm *vm,
vma->gpuva.va.range = end - start + 1;
if (read_only)
vma->gpuva.flags |= XE_VMA_READ_ONLY;
- if (is_null)
- vma->gpuva.flags |= DRM_GPUVA_SPARSE;
for_each_tile(tile, vm->xe, id)
vma->tile_mask |= 0x1 << id;
@@ -836,35 +859,35 @@ static struct xe_vma *xe_vma_create(struct xe_vm *vm,
vm_bo = drm_gpuvm_bo_obtain(vma->gpuva.vm, &bo->ttm.base);
if (IS_ERR(vm_bo)) {
- kfree(vma);
+ xe_vma_free(vma);
return ERR_CAST(vm_bo);
}
drm_gpuvm_bo_extobj_add(vm_bo);
drm_gem_object_get(&bo->ttm.base);
- vma->gpuva.gem.obj = &bo->ttm.base;
vma->gpuva.gem.offset = bo_offset_or_userptr;
drm_gpuva_link(&vma->gpuva, vm_bo);
drm_gpuvm_bo_put(vm_bo);
} else /* userptr or null */ {
if (!is_null) {
+ struct xe_userptr *userptr = &to_userptr_vma(vma)->userptr;
u64 size = end - start + 1;
int err;
- INIT_LIST_HEAD(&vma->userptr.invalidate_link);
+ INIT_LIST_HEAD(&userptr->invalidate_link);
+ INIT_LIST_HEAD(&userptr->repin_link);
vma->gpuva.gem.offset = bo_offset_or_userptr;
- err = mmu_interval_notifier_insert(&vma->userptr.notifier,
+ err = mmu_interval_notifier_insert(&userptr->notifier,
current->mm,
xe_vma_userptr(vma), size,
&vma_userptr_notifier_ops);
if (err) {
- kfree(vma);
- vma = ERR_PTR(err);
- return vma;
+ xe_vma_free(vma);
+ return ERR_PTR(err);
}
- vma->userptr.notifier_seq = LONG_MAX;
+ userptr->notifier_seq = LONG_MAX;
}
xe_vm_get(vm);
@@ -880,13 +903,15 @@ static void xe_vma_destroy_late(struct xe_vma *vma)
bool read_only = xe_vma_read_only(vma);
if (xe_vma_is_userptr(vma)) {
- if (vma->userptr.sg) {
+ struct xe_userptr *userptr = &to_userptr_vma(vma)->userptr;
+
+ if (userptr->sg) {
dma_unmap_sgtable(xe->drm.dev,
- vma->userptr.sg,
+ userptr->sg,
read_only ? DMA_TO_DEVICE :
DMA_BIDIRECTIONAL, 0);
- sg_free_table(vma->userptr.sg);
- vma->userptr.sg = NULL;
+ sg_free_table(userptr->sg);
+ userptr->sg = NULL;
}
/*
@@ -894,7 +919,7 @@ static void xe_vma_destroy_late(struct xe_vma *vma)
* the notifer until we're sure the GPU is not accessing
* them anymore
*/
- mmu_interval_notifier_remove(&vma->userptr.notifier);
+ mmu_interval_notifier_remove(&userptr->notifier);
xe_vm_put(vm);
} else if (xe_vma_is_null(vma)) {
xe_vm_put(vm);
@@ -902,7 +927,7 @@ static void xe_vma_destroy_late(struct xe_vma *vma)
xe_bo_put(xe_vma_bo(vma));
}
- kfree(vma);
+ xe_vma_free(vma);
}
static void vma_destroy_work_func(struct work_struct *w)
@@ -933,7 +958,7 @@ static void xe_vma_destroy(struct xe_vma *vma, struct dma_fence *fence)
xe_assert(vm->xe, vma->gpuva.flags & XE_VMA_DESTROYED);
spin_lock(&vm->userptr.invalidated_lock);
- list_del(&vma->userptr.invalidate_link);
+ list_del(&to_userptr_vma(vma)->userptr.invalidate_link);
spin_unlock(&vm->userptr.invalidated_lock);
} else if (!xe_vma_is_null(vma)) {
xe_bo_assert_held(xe_vma_bo(vma));
@@ -2150,7 +2175,7 @@ static struct xe_vma *new_vma(struct xe_vm *vm, struct drm_gpuva_op_map *op,
drm_exec_fini(&exec);
if (xe_vma_is_userptr(vma)) {
- err = xe_vma_userptr_pin_pages(vma);
+ err = xe_vma_userptr_pin_pages(to_userptr_vma(vma));
if (err) {
prep_vma_destroy(vm, vma, false);
xe_vma_destroy_unlocked(vma);
@@ -2507,7 +2532,7 @@ retry_userptr:
if (err == -EAGAIN && xe_vma_is_userptr(vma)) {
lockdep_assert_held_write(&vm->lock);
- err = xe_vma_userptr_pin_pages(vma);
+ err = xe_vma_userptr_pin_pages(to_userptr_vma(vma));
if (!err)
goto retry_userptr;
@@ -2851,7 +2876,7 @@ int xe_vm_bind_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
struct drm_gpuva_ops **ops = NULL;
struct xe_vm *vm;
struct xe_exec_queue *q = NULL;
- u32 num_syncs;
+ u32 num_syncs, num_ufence = 0;
struct xe_sync_entry *syncs = NULL;
struct drm_xe_vm_bind_op *bind_ops;
LIST_HEAD(ops_list);
@@ -2988,6 +3013,14 @@ int xe_vm_bind_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
SYNC_PARSE_FLAG_DISALLOW_USER_FENCE : 0));
if (err)
goto free_syncs;
+
+ if (xe_sync_is_ufence(&syncs[num_syncs]))
+ num_ufence++;
+ }
+
+ if (XE_IOCTL_DBG(xe, num_ufence > 1)) {
+ err = -EINVAL;
+ goto free_syncs;
}
if (!args->num_binds) {
@@ -3130,8 +3163,8 @@ int xe_vm_invalidate_vma(struct xe_vma *vma)
if (IS_ENABLED(CONFIG_PROVE_LOCKING)) {
if (xe_vma_is_userptr(vma)) {
WARN_ON_ONCE(!mmu_interval_check_retry
- (&vma->userptr.notifier,
- vma->userptr.notifier_seq));
+ (&to_userptr_vma(vma)->userptr.notifier,
+ to_userptr_vma(vma)->userptr.notifier_seq));
WARN_ON_ONCE(!dma_resv_test_signaled(xe_vm_resv(xe_vma_vm(vma)),
DMA_RESV_USAGE_BOOKKEEP));
@@ -3192,11 +3225,11 @@ int xe_analyze_vm(struct drm_printer *p, struct xe_vm *vm, int gt_id)
if (is_null) {
addr = 0;
} else if (is_userptr) {
+ struct sg_table *sg = to_userptr_vma(vma)->userptr.sg;
struct xe_res_cursor cur;
- if (vma->userptr.sg) {
- xe_res_first_sg(vma->userptr.sg, 0, XE_PAGE_SIZE,
- &cur);
+ if (sg) {
+ xe_res_first_sg(sg, 0, XE_PAGE_SIZE, &cur);
addr = xe_res_dma(&cur);
} else {
addr = 0;
diff --git a/drivers/gpu/drm/xe/xe_vm.h b/drivers/gpu/drm/xe/xe_vm.h
index cf2f96e8c1ab..9654a0612fc2 100644
--- a/drivers/gpu/drm/xe/xe_vm.h
+++ b/drivers/gpu/drm/xe/xe_vm.h
@@ -160,6 +160,18 @@ static inline bool xe_vma_is_userptr(struct xe_vma *vma)
return xe_vma_has_no_bo(vma) && !xe_vma_is_null(vma);
}
+/**
+ * to_userptr_vma() - Return a pointer to an embedding userptr vma
+ * @vma: Pointer to the embedded struct xe_vma
+ *
+ * Return: Pointer to the embedding userptr vma
+ */
+static inline struct xe_userptr_vma *to_userptr_vma(struct xe_vma *vma)
+{
+ xe_assert(xe_vma_vm(vma)->xe, xe_vma_is_userptr(vma));
+ return container_of(vma, struct xe_userptr_vma, vma);
+}
+
u64 xe_vm_pdp4_descriptor(struct xe_vm *vm, struct xe_tile *tile);
int xe_vm_create_ioctl(struct drm_device *dev, void *data,
@@ -224,9 +236,9 @@ static inline void xe_vm_reactivate_rebind(struct xe_vm *vm)
}
}
-int xe_vma_userptr_pin_pages(struct xe_vma *vma);
+int xe_vma_userptr_pin_pages(struct xe_userptr_vma *uvma);
-int xe_vma_userptr_check_repin(struct xe_vma *vma);
+int xe_vma_userptr_check_repin(struct xe_userptr_vma *uvma);
bool xe_vm_validate_should_retry(struct drm_exec *exec, int err, ktime_t *end);
diff --git a/drivers/gpu/drm/xe/xe_vm_types.h b/drivers/gpu/drm/xe/xe_vm_types.h
index 63e8a50b88e9..1fec66ae2eb2 100644
--- a/drivers/gpu/drm/xe/xe_vm_types.h
+++ b/drivers/gpu/drm/xe/xe_vm_types.h
@@ -37,6 +37,8 @@ struct xe_vm;
struct xe_userptr {
/** @invalidate_link: Link for the vm::userptr.invalidated list */
struct list_head invalidate_link;
+ /** @userptr: link into VM repin list if userptr. */
+ struct list_head repin_link;
/**
* @notifier: MMU notifier for user pointer (invalidation call back)
*/
@@ -68,8 +70,6 @@ struct xe_vma {
* resv.
*/
union {
- /** @userptr: link into VM repin list if userptr. */
- struct list_head userptr;
/** @rebind: link into VM if this VMA needs rebinding. */
struct list_head rebind;
/** @destroy: link to contested list when VM is being closed. */
@@ -105,11 +105,15 @@ struct xe_vma {
* @pat_index: The pat index to use when encoding the PTEs for this vma.
*/
u16 pat_index;
+};
- /**
- * @userptr: user pointer state, only allocated for VMAs that are
- * user pointers
- */
+/**
+ * struct xe_userptr_vma - A userptr vma subclass
+ * @vma: The vma.
+ * @userptr: Additional userptr information.
+ */
+struct xe_userptr_vma {
+ struct xe_vma vma;
struct xe_userptr userptr;
};