summaryrefslogtreecommitdiff
path: root/drivers/accel
diff options
context:
space:
mode:
Diffstat (limited to 'drivers/accel')
-rw-r--r--drivers/accel/amdxdna/TODO1
-rw-r--r--drivers/accel/amdxdna/aie2_ctx.c72
-rw-r--r--drivers/accel/amdxdna/aie2_message.c6
-rw-r--r--drivers/accel/amdxdna/aie2_msg_priv.h10
-rw-r--r--drivers/accel/amdxdna/aie2_pci.c13
-rw-r--r--drivers/accel/amdxdna/aie2_psp.c4
-rw-r--r--drivers/accel/amdxdna/amdxdna_ctx.c22
-rw-r--r--drivers/accel/amdxdna/amdxdna_gem.c554
-rw-r--r--drivers/accel/amdxdna/amdxdna_gem.h27
-rw-r--r--drivers/accel/amdxdna/amdxdna_pci_drv.c11
-rw-r--r--drivers/accel/amdxdna/amdxdna_pci_drv.h2
-rw-r--r--drivers/accel/drm_accel.c16
-rw-r--r--drivers/accel/habanalabs/Kconfig2
-rw-r--r--drivers/accel/habanalabs/common/device.c25
-rw-r--r--drivers/accel/habanalabs/common/habanalabs_ioctl.c2
-rw-r--r--drivers/accel/habanalabs/common/sysfs.c4
-rw-r--r--drivers/accel/ivpu/ivpu_debugfs.c6
-rw-r--r--drivers/accel/ivpu/ivpu_drv.c15
-rw-r--r--drivers/accel/ivpu/ivpu_drv.h16
-rw-r--r--drivers/accel/ivpu/ivpu_fw.c33
-rw-r--r--drivers/accel/ivpu/ivpu_fw.h1
-rw-r--r--drivers/accel/ivpu/ivpu_gem.c101
-rw-r--r--drivers/accel/ivpu/ivpu_gem.h2
-rw-r--r--drivers/accel/ivpu/ivpu_hw.c6
-rw-r--r--drivers/accel/ivpu/ivpu_hw.h14
-rw-r--r--drivers/accel/ivpu/ivpu_hw_btrs.c134
-rw-r--r--drivers/accel/ivpu/ivpu_hw_btrs.h9
-rw-r--r--drivers/accel/ivpu/ivpu_hw_ip.c1
-rw-r--r--drivers/accel/ivpu/ivpu_ipc.c3
-rw-r--r--drivers/accel/ivpu/ivpu_job.c145
-rw-r--r--drivers/accel/ivpu/ivpu_jsm_msg.c9
-rw-r--r--drivers/accel/ivpu/ivpu_ms.c24
-rw-r--r--drivers/accel/ivpu/ivpu_pm.c47
-rw-r--r--drivers/accel/ivpu/ivpu_sysfs.c49
-rw-r--r--drivers/accel/ivpu/vpu_boot_api.h13
-rw-r--r--drivers/accel/ivpu/vpu_jsm_api.h53
-rw-r--r--drivers/accel/qaic/Kconfig1
-rw-r--r--drivers/accel/qaic/Makefile1
-rw-r--r--drivers/accel/qaic/qaic.h10
-rw-r--r--drivers/accel/qaic/qaic_data.c9
-rw-r--r--drivers/accel/qaic/qaic_debugfs.c2
-rw-r--r--drivers/accel/qaic/qaic_drv.c6
-rw-r--r--drivers/accel/qaic/qaic_ras.c642
-rw-r--r--drivers/accel/qaic/qaic_ras.h10
-rw-r--r--drivers/accel/qaic/qaic_timesync.c4
45 files changed, 1637 insertions, 500 deletions
diff --git a/drivers/accel/amdxdna/TODO b/drivers/accel/amdxdna/TODO
index 5119bccd1917..ad8ac6e315b6 100644
--- a/drivers/accel/amdxdna/TODO
+++ b/drivers/accel/amdxdna/TODO
@@ -1,3 +1,2 @@
-- Add import and export BO support
- Add debugfs support
- Add debug BO support
diff --git a/drivers/accel/amdxdna/aie2_ctx.c b/drivers/accel/amdxdna/aie2_ctx.c
index 00d215ac866e..2cff5419bd2f 100644
--- a/drivers/accel/amdxdna/aie2_ctx.c
+++ b/drivers/accel/amdxdna/aie2_ctx.c
@@ -361,7 +361,7 @@ aie2_sched_job_timedout(struct drm_sched_job *sched_job)
aie2_hwctx_restart(xdna, hwctx);
mutex_unlock(&xdna->dev_lock);
- return DRM_GPU_SCHED_STAT_NOMINAL;
+ return DRM_GPU_SCHED_STAT_RESET;
}
static const struct drm_sched_backend_ops sched_ops = {
@@ -566,7 +566,7 @@ int aie2_hwctx_init(struct amdxdna_hwctx *hwctx)
.size = MAX_CHAIN_CMDBUF_SIZE,
};
- abo = amdxdna_drm_alloc_dev_bo(&xdna->ddev, &args, client->filp, true);
+ abo = amdxdna_drm_alloc_dev_bo(&xdna->ddev, &args, client->filp);
if (IS_ERR(abo)) {
ret = PTR_ERR(abo);
goto free_cmd_bufs;
@@ -758,27 +758,42 @@ int aie2_hwctx_config(struct amdxdna_hwctx *hwctx, u32 type, u64 value, void *bu
static int aie2_populate_range(struct amdxdna_gem_obj *abo)
{
struct amdxdna_dev *xdna = to_xdna_dev(to_gobj(abo)->dev);
- struct mm_struct *mm = abo->mem.notifier.mm;
- struct hmm_range range = { 0 };
+ struct amdxdna_umap *mapp;
unsigned long timeout;
+ struct mm_struct *mm;
+ bool found;
int ret;
- XDNA_INFO_ONCE(xdna, "populate memory range %llx size %lx",
- abo->mem.userptr, abo->mem.size);
- range.notifier = &abo->mem.notifier;
- range.start = abo->mem.userptr;
- range.end = abo->mem.userptr + abo->mem.size;
- range.hmm_pfns = abo->mem.pfns;
- range.default_flags = HMM_PFN_REQ_FAULT;
+ timeout = jiffies + msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT);
+again:
+ found = false;
+ down_write(&xdna->notifier_lock);
+ list_for_each_entry(mapp, &abo->mem.umap_list, node) {
+ if (mapp->invalid) {
+ found = true;
+ break;
+ }
+ }
- if (!mmget_not_zero(mm))
+ if (!found) {
+ abo->mem.map_invalid = false;
+ up_write(&xdna->notifier_lock);
+ return 0;
+ }
+ kref_get(&mapp->refcnt);
+ up_write(&xdna->notifier_lock);
+
+ XDNA_DBG(xdna, "populate memory range %lx %lx",
+ mapp->vma->vm_start, mapp->vma->vm_end);
+ mm = mapp->notifier.mm;
+ if (!mmget_not_zero(mm)) {
+ amdxdna_umap_put(mapp);
return -EFAULT;
+ }
- timeout = jiffies + msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT);
-again:
- range.notifier_seq = mmu_interval_read_begin(&abo->mem.notifier);
+ mapp->range.notifier_seq = mmu_interval_read_begin(&mapp->notifier);
mmap_read_lock(mm);
- ret = hmm_range_fault(&range);
+ ret = hmm_range_fault(&mapp->range);
mmap_read_unlock(mm);
if (ret) {
if (time_after(jiffies, timeout)) {
@@ -786,21 +801,27 @@ again:
goto put_mm;
}
- if (ret == -EBUSY)
+ if (ret == -EBUSY) {
+ amdxdna_umap_put(mapp);
goto again;
+ }
goto put_mm;
}
- down_read(&xdna->notifier_lock);
- if (mmu_interval_read_retry(&abo->mem.notifier, range.notifier_seq)) {
- up_read(&xdna->notifier_lock);
+ down_write(&xdna->notifier_lock);
+ if (mmu_interval_read_retry(&mapp->notifier, mapp->range.notifier_seq)) {
+ up_write(&xdna->notifier_lock);
+ amdxdna_umap_put(mapp);
goto again;
}
- abo->mem.map_invalid = false;
- up_read(&xdna->notifier_lock);
+ mapp->invalid = false;
+ up_write(&xdna->notifier_lock);
+ amdxdna_umap_put(mapp);
+ goto again;
put_mm:
+ amdxdna_umap_put(mapp);
mmput(mm);
return ret;
}
@@ -827,7 +848,8 @@ int aie2_cmd_submit(struct amdxdna_hwctx *hwctx, struct amdxdna_sched_job *job,
goto up_sem;
}
- ret = drm_sched_job_init(&job->base, &hwctx->priv->entity, 1, hwctx);
+ ret = drm_sched_job_init(&job->base, &hwctx->priv->entity, 1, hwctx,
+ hwctx->client->filp->client_id);
if (ret) {
XDNA_ERR(xdna, "DRM job init failed, ret %d", ret);
goto free_chain;
@@ -908,10 +930,6 @@ void aie2_hmm_invalidate(struct amdxdna_gem_obj *abo,
struct drm_gem_object *gobj = to_gobj(abo);
long ret;
- down_write(&xdna->notifier_lock);
- abo->mem.map_invalid = true;
- mmu_interval_set_seq(&abo->mem.notifier, cur_seq);
- up_write(&xdna->notifier_lock);
ret = dma_resv_wait_timeout(gobj->resv, DMA_RESV_USAGE_BOOKKEEP,
true, MAX_SCHEDULE_TIMEOUT);
if (!ret || ret == -ERESTARTSYS)
diff --git a/drivers/accel/amdxdna/aie2_message.c b/drivers/accel/amdxdna/aie2_message.c
index bf4219e32cc1..82412eec9a4b 100644
--- a/drivers/accel/amdxdna/aie2_message.c
+++ b/drivers/accel/amdxdna/aie2_message.c
@@ -525,7 +525,7 @@ aie2_cmdlist_fill_one_slot_cf(void *cmd_buf, u32 offset,
if (!payload)
return -EINVAL;
- if (!slot_cf_has_space(offset, payload_len))
+ if (!slot_has_space(*buf, offset, payload_len))
return -ENOSPC;
buf->cu_idx = cu_idx;
@@ -558,7 +558,7 @@ aie2_cmdlist_fill_one_slot_dpu(void *cmd_buf, u32 offset,
if (payload_len < sizeof(*sn) || arg_sz > MAX_DPU_ARGS_SIZE)
return -EINVAL;
- if (!slot_dpu_has_space(offset, arg_sz))
+ if (!slot_has_space(*buf, offset, arg_sz))
return -ENOSPC;
buf->inst_buf_addr = sn->buffer;
@@ -569,7 +569,7 @@ aie2_cmdlist_fill_one_slot_dpu(void *cmd_buf, u32 offset,
memcpy(buf->args, sn->prop_args, arg_sz);
/* Accurate buf size to hint firmware to do necessary copy */
- *size += sizeof(*buf) + arg_sz;
+ *size = sizeof(*buf) + arg_sz;
return 0;
}
diff --git a/drivers/accel/amdxdna/aie2_msg_priv.h b/drivers/accel/amdxdna/aie2_msg_priv.h
index 4e02e744b470..6df9065b13f6 100644
--- a/drivers/accel/amdxdna/aie2_msg_priv.h
+++ b/drivers/accel/amdxdna/aie2_msg_priv.h
@@ -319,18 +319,16 @@ struct async_event_msg_resp {
} __packed;
#define MAX_CHAIN_CMDBUF_SIZE SZ_4K
-#define slot_cf_has_space(offset, payload_size) \
- (MAX_CHAIN_CMDBUF_SIZE - ((offset) + (payload_size)) > \
- offsetof(struct cmd_chain_slot_execbuf_cf, args[0]))
+#define slot_has_space(slot, offset, payload_size) \
+ (MAX_CHAIN_CMDBUF_SIZE >= (offset) + (payload_size) + \
+ sizeof(typeof(slot)))
+
struct cmd_chain_slot_execbuf_cf {
__u32 cu_idx;
__u32 arg_cnt;
__u32 args[] __counted_by(arg_cnt);
};
-#define slot_dpu_has_space(offset, payload_size) \
- (MAX_CHAIN_CMDBUF_SIZE - ((offset) + (payload_size)) > \
- offsetof(struct cmd_chain_slot_dpu, args[0]))
struct cmd_chain_slot_dpu {
__u64 inst_buf_addr;
__u32 inst_size;
diff --git a/drivers/accel/amdxdna/aie2_pci.c b/drivers/accel/amdxdna/aie2_pci.c
index 5a058e565b01..c6cf7068d23c 100644
--- a/drivers/accel/amdxdna/aie2_pci.c
+++ b/drivers/accel/amdxdna/aie2_pci.c
@@ -512,12 +512,6 @@ static int aie2_init(struct amdxdna_dev *xdna)
goto release_fw;
}
- ret = iommu_dev_enable_feature(&pdev->dev, IOMMU_DEV_FEAT_SVA);
- if (ret) {
- XDNA_ERR(xdna, "Enable PASID failed, ret %d", ret);
- goto free_irq;
- }
-
psp_conf.fw_size = fw->size;
psp_conf.fw_buf = fw->data;
for (i = 0; i < PSP_MAX_REGS; i++)
@@ -526,14 +520,14 @@ static int aie2_init(struct amdxdna_dev *xdna)
if (!ndev->psp_hdl) {
XDNA_ERR(xdna, "failed to create psp");
ret = -ENOMEM;
- goto disable_sva;
+ goto free_irq;
}
xdna->dev_handle = ndev;
ret = aie2_hw_start(xdna);
if (ret) {
XDNA_ERR(xdna, "start npu failed, ret %d", ret);
- goto disable_sva;
+ goto free_irq;
}
ret = aie2_mgmt_fw_query(ndev);
@@ -584,8 +578,6 @@ async_event_free:
aie2_error_async_events_free(ndev);
stop_hw:
aie2_hw_stop(xdna);
-disable_sva:
- iommu_dev_disable_feature(&pdev->dev, IOMMU_DEV_FEAT_SVA);
free_irq:
pci_free_irq_vectors(pdev);
release_fw:
@@ -601,7 +593,6 @@ static void aie2_fini(struct amdxdna_dev *xdna)
aie2_hw_stop(xdna);
aie2_error_async_events_free(ndev);
- iommu_dev_disable_feature(&pdev->dev, IOMMU_DEV_FEAT_SVA);
pci_free_irq_vectors(pdev);
}
diff --git a/drivers/accel/amdxdna/aie2_psp.c b/drivers/accel/amdxdna/aie2_psp.c
index dc3a072ce3b6..f28a060a8810 100644
--- a/drivers/accel/amdxdna/aie2_psp.c
+++ b/drivers/accel/amdxdna/aie2_psp.c
@@ -126,8 +126,8 @@ struct psp_device *aie2m_psp_create(struct drm_device *ddev, struct psp_config *
psp->ddev = ddev;
memcpy(psp->psp_regs, conf->psp_regs, sizeof(psp->psp_regs));
- psp->fw_buf_sz = ALIGN(conf->fw_size, PSP_FW_ALIGN) + PSP_FW_ALIGN;
- psp->fw_buffer = drmm_kmalloc(ddev, psp->fw_buf_sz, GFP_KERNEL);
+ psp->fw_buf_sz = ALIGN(conf->fw_size, PSP_FW_ALIGN);
+ psp->fw_buffer = drmm_kmalloc(ddev, psp->fw_buf_sz + PSP_FW_ALIGN, GFP_KERNEL);
if (!psp->fw_buffer) {
drm_err(ddev, "no memory for fw buffer");
return NULL;
diff --git a/drivers/accel/amdxdna/amdxdna_ctx.c b/drivers/accel/amdxdna/amdxdna_ctx.c
index 43442b9e273b..be073224bd69 100644
--- a/drivers/accel/amdxdna/amdxdna_ctx.c
+++ b/drivers/accel/amdxdna/amdxdna_ctx.c
@@ -496,11 +496,11 @@ static int amdxdna_drm_submit_execbuf(struct amdxdna_client *client,
struct amdxdna_drm_exec_cmd *args)
{
struct amdxdna_dev *xdna = client->xdna;
- u32 *arg_bo_hdls;
+ u32 *arg_bo_hdls = NULL;
u32 cmd_bo_hdl;
int ret;
- if (!args->arg_count || args->arg_count > MAX_ARG_COUNT) {
+ if (args->arg_count > MAX_ARG_COUNT) {
XDNA_ERR(xdna, "Invalid arg bo count %d", args->arg_count);
return -EINVAL;
}
@@ -512,14 +512,16 @@ static int amdxdna_drm_submit_execbuf(struct amdxdna_client *client,
}
cmd_bo_hdl = (u32)args->cmd_handles;
- arg_bo_hdls = kcalloc(args->arg_count, sizeof(u32), GFP_KERNEL);
- if (!arg_bo_hdls)
- return -ENOMEM;
- ret = copy_from_user(arg_bo_hdls, u64_to_user_ptr(args->args),
- args->arg_count * sizeof(u32));
- if (ret) {
- ret = -EFAULT;
- goto free_cmd_bo_hdls;
+ if (args->arg_count) {
+ arg_bo_hdls = kcalloc(args->arg_count, sizeof(u32), GFP_KERNEL);
+ if (!arg_bo_hdls)
+ return -ENOMEM;
+ ret = copy_from_user(arg_bo_hdls, u64_to_user_ptr(args->args),
+ args->arg_count * sizeof(u32));
+ if (ret) {
+ ret = -EFAULT;
+ goto free_cmd_bo_hdls;
+ }
}
ret = amdxdna_cmd_submit(client, cmd_bo_hdl, arg_bo_hdls,
diff --git a/drivers/accel/amdxdna/amdxdna_gem.c b/drivers/accel/amdxdna/amdxdna_gem.c
index 606433d73236..0f85a0105178 100644
--- a/drivers/accel/amdxdna/amdxdna_gem.c
+++ b/drivers/accel/amdxdna/amdxdna_gem.c
@@ -9,7 +9,10 @@
#include <drm/drm_gem.h>
#include <drm/drm_gem_shmem_helper.h>
#include <drm/gpu_scheduler.h>
+#include <linux/dma-buf.h>
+#include <linux/dma-direct.h>
#include <linux/iosys-map.h>
+#include <linux/pagemap.h>
#include <linux/vmalloc.h>
#include "amdxdna_ctx.h"
@@ -18,94 +21,116 @@
#define XDNA_MAX_CMD_BO_SIZE SZ_32K
+MODULE_IMPORT_NS("DMA_BUF");
+
static int
-amdxdna_gem_insert_node_locked(struct amdxdna_gem_obj *abo, bool use_vmap)
+amdxdna_gem_heap_alloc(struct amdxdna_gem_obj *abo)
{
struct amdxdna_client *client = abo->client;
struct amdxdna_dev *xdna = client->xdna;
struct amdxdna_mem *mem = &abo->mem;
+ struct amdxdna_gem_obj *heap;
u64 offset;
u32 align;
int ret;
+ mutex_lock(&client->mm_lock);
+
+ heap = client->dev_heap;
+ if (!heap) {
+ ret = -EINVAL;
+ goto unlock_out;
+ }
+
+ if (heap->mem.userptr == AMDXDNA_INVALID_ADDR) {
+ XDNA_ERR(xdna, "Invalid dev heap userptr");
+ ret = -EINVAL;
+ goto unlock_out;
+ }
+
+ if (mem->size == 0 || mem->size > heap->mem.size) {
+ XDNA_ERR(xdna, "Invalid dev bo size 0x%lx, limit 0x%lx",
+ mem->size, heap->mem.size);
+ ret = -EINVAL;
+ goto unlock_out;
+ }
+
align = 1 << max(PAGE_SHIFT, xdna->dev_info->dev_mem_buf_shift);
- ret = drm_mm_insert_node_generic(&abo->dev_heap->mm, &abo->mm_node,
+ ret = drm_mm_insert_node_generic(&heap->mm, &abo->mm_node,
mem->size, align,
0, DRM_MM_INSERT_BEST);
if (ret) {
XDNA_ERR(xdna, "Failed to alloc dev bo memory, ret %d", ret);
- return ret;
+ goto unlock_out;
}
mem->dev_addr = abo->mm_node.start;
- offset = mem->dev_addr - abo->dev_heap->mem.dev_addr;
- mem->userptr = abo->dev_heap->mem.userptr + offset;
- mem->pages = &abo->dev_heap->base.pages[offset >> PAGE_SHIFT];
- mem->nr_pages = mem->size >> PAGE_SHIFT;
-
- if (use_vmap) {
- mem->kva = vmap(mem->pages, mem->nr_pages, VM_MAP, PAGE_KERNEL);
- if (!mem->kva) {
- XDNA_ERR(xdna, "Failed to vmap");
- drm_mm_remove_node(&abo->mm_node);
- return -EFAULT;
- }
- }
+ offset = mem->dev_addr - heap->mem.dev_addr;
+ mem->userptr = heap->mem.userptr + offset;
+ mem->kva = heap->mem.kva + offset;
- return 0;
+ drm_gem_object_get(to_gobj(heap));
+
+unlock_out:
+ mutex_unlock(&client->mm_lock);
+
+ return ret;
}
-static void amdxdna_gem_obj_free(struct drm_gem_object *gobj)
+static void
+amdxdna_gem_destroy_obj(struct amdxdna_gem_obj *abo)
{
- struct amdxdna_dev *xdna = to_xdna_dev(gobj->dev);
- struct amdxdna_gem_obj *abo = to_xdna_obj(gobj);
- struct iosys_map map = IOSYS_MAP_INIT_VADDR(abo->mem.kva);
+ mutex_destroy(&abo->lock);
+ kfree(abo);
+}
- XDNA_DBG(xdna, "BO type %d xdna_addr 0x%llx", abo->type, abo->mem.dev_addr);
- if (abo->pinned)
- amdxdna_gem_unpin(abo);
+static void
+amdxdna_gem_heap_free(struct amdxdna_gem_obj *abo)
+{
+ struct amdxdna_gem_obj *heap;
- if (abo->type == AMDXDNA_BO_DEV) {
- mutex_lock(&abo->client->mm_lock);
- drm_mm_remove_node(&abo->mm_node);
- mutex_unlock(&abo->client->mm_lock);
+ mutex_lock(&abo->client->mm_lock);
- vunmap(abo->mem.kva);
- drm_gem_object_put(to_gobj(abo->dev_heap));
- drm_gem_object_release(gobj);
- mutex_destroy(&abo->lock);
- kfree(abo);
- return;
- }
+ drm_mm_remove_node(&abo->mm_node);
- if (abo->type == AMDXDNA_BO_DEV_HEAP)
- drm_mm_takedown(&abo->mm);
+ heap = abo->client->dev_heap;
+ drm_gem_object_put(to_gobj(heap));
- drm_gem_vunmap_unlocked(gobj, &map);
- mutex_destroy(&abo->lock);
- drm_gem_shmem_free(&abo->base);
+ mutex_unlock(&abo->client->mm_lock);
}
-static const struct drm_gem_object_funcs amdxdna_gem_dev_obj_funcs = {
- .free = amdxdna_gem_obj_free,
-};
-
static bool amdxdna_hmm_invalidate(struct mmu_interval_notifier *mni,
const struct mmu_notifier_range *range,
unsigned long cur_seq)
{
- struct amdxdna_gem_obj *abo = container_of(mni, struct amdxdna_gem_obj,
- mem.notifier);
- struct amdxdna_dev *xdna = to_xdna_dev(to_gobj(abo)->dev);
+ struct amdxdna_umap *mapp = container_of(mni, struct amdxdna_umap, notifier);
+ struct amdxdna_gem_obj *abo = mapp->abo;
+ struct amdxdna_dev *xdna;
- XDNA_DBG(xdna, "Invalid range 0x%llx, 0x%lx, type %d",
- abo->mem.userptr, abo->mem.size, abo->type);
+ xdna = to_xdna_dev(to_gobj(abo)->dev);
+ XDNA_DBG(xdna, "Invalidating range 0x%lx, 0x%lx, type %d",
+ mapp->vma->vm_start, mapp->vma->vm_end, abo->type);
if (!mmu_notifier_range_blockable(range))
return false;
+ down_write(&xdna->notifier_lock);
+ abo->mem.map_invalid = true;
+ mapp->invalid = true;
+ mmu_interval_set_seq(&mapp->notifier, cur_seq);
+ up_write(&xdna->notifier_lock);
+
xdna->dev_info->ops->hmm_invalidate(abo, cur_seq);
+ if (range->event == MMU_NOTIFY_UNMAP) {
+ down_write(&xdna->notifier_lock);
+ if (!mapp->unmapped) {
+ queue_work(xdna->notifier_wq, &mapp->hmm_unreg_work);
+ mapp->unmapped = true;
+ }
+ up_write(&xdna->notifier_lock);
+ }
+
return true;
}
@@ -113,102 +138,311 @@ static const struct mmu_interval_notifier_ops amdxdna_hmm_ops = {
.invalidate = amdxdna_hmm_invalidate,
};
-static void amdxdna_hmm_unregister(struct amdxdna_gem_obj *abo)
+static void amdxdna_hmm_unregister(struct amdxdna_gem_obj *abo,
+ struct vm_area_struct *vma)
{
struct amdxdna_dev *xdna = to_xdna_dev(to_gobj(abo)->dev);
+ struct amdxdna_umap *mapp;
+
+ down_read(&xdna->notifier_lock);
+ list_for_each_entry(mapp, &abo->mem.umap_list, node) {
+ if (!vma || mapp->vma == vma) {
+ if (!mapp->unmapped) {
+ queue_work(xdna->notifier_wq, &mapp->hmm_unreg_work);
+ mapp->unmapped = true;
+ }
+ if (vma)
+ break;
+ }
+ }
+ up_read(&xdna->notifier_lock);
+}
- if (!xdna->dev_info->ops->hmm_invalidate)
- return;
+static void amdxdna_umap_release(struct kref *ref)
+{
+ struct amdxdna_umap *mapp = container_of(ref, struct amdxdna_umap, refcnt);
+ struct vm_area_struct *vma = mapp->vma;
+ struct amdxdna_dev *xdna;
+
+ mmu_interval_notifier_remove(&mapp->notifier);
+ if (is_import_bo(mapp->abo) && vma->vm_file && vma->vm_file->f_mapping)
+ mapping_clear_unevictable(vma->vm_file->f_mapping);
+
+ xdna = to_xdna_dev(to_gobj(mapp->abo)->dev);
+ down_write(&xdna->notifier_lock);
+ list_del(&mapp->node);
+ up_write(&xdna->notifier_lock);
- mmu_interval_notifier_remove(&abo->mem.notifier);
- kvfree(abo->mem.pfns);
- abo->mem.pfns = NULL;
+ kvfree(mapp->range.hmm_pfns);
+ kfree(mapp);
}
-static int amdxdna_hmm_register(struct amdxdna_gem_obj *abo, unsigned long addr,
- size_t len)
+void amdxdna_umap_put(struct amdxdna_umap *mapp)
+{
+ kref_put(&mapp->refcnt, amdxdna_umap_release);
+}
+
+static void amdxdna_hmm_unreg_work(struct work_struct *work)
+{
+ struct amdxdna_umap *mapp = container_of(work, struct amdxdna_umap,
+ hmm_unreg_work);
+
+ amdxdna_umap_put(mapp);
+}
+
+static int amdxdna_hmm_register(struct amdxdna_gem_obj *abo,
+ struct vm_area_struct *vma)
{
struct amdxdna_dev *xdna = to_xdna_dev(to_gobj(abo)->dev);
+ unsigned long len = vma->vm_end - vma->vm_start;
+ unsigned long addr = vma->vm_start;
+ struct amdxdna_umap *mapp;
u32 nr_pages;
int ret;
if (!xdna->dev_info->ops->hmm_invalidate)
return 0;
- if (abo->mem.pfns)
- return -EEXIST;
+ mapp = kzalloc(sizeof(*mapp), GFP_KERNEL);
+ if (!mapp)
+ return -ENOMEM;
nr_pages = (PAGE_ALIGN(addr + len) - (addr & PAGE_MASK)) >> PAGE_SHIFT;
- abo->mem.pfns = kvcalloc(nr_pages, sizeof(*abo->mem.pfns),
- GFP_KERNEL);
- if (!abo->mem.pfns)
- return -ENOMEM;
+ mapp->range.hmm_pfns = kvcalloc(nr_pages, sizeof(*mapp->range.hmm_pfns),
+ GFP_KERNEL);
+ if (!mapp->range.hmm_pfns) {
+ ret = -ENOMEM;
+ goto free_map;
+ }
- ret = mmu_interval_notifier_insert_locked(&abo->mem.notifier,
+ ret = mmu_interval_notifier_insert_locked(&mapp->notifier,
current->mm,
addr,
len,
&amdxdna_hmm_ops);
if (ret) {
XDNA_ERR(xdna, "Insert mmu notifier failed, ret %d", ret);
- kvfree(abo->mem.pfns);
+ goto free_pfns;
}
- abo->mem.userptr = addr;
+ mapp->range.notifier = &mapp->notifier;
+ mapp->range.start = vma->vm_start;
+ mapp->range.end = vma->vm_end;
+ mapp->range.default_flags = HMM_PFN_REQ_FAULT;
+ mapp->vma = vma;
+ mapp->abo = abo;
+ kref_init(&mapp->refcnt);
+
+ if (abo->mem.userptr == AMDXDNA_INVALID_ADDR)
+ abo->mem.userptr = addr;
+ INIT_WORK(&mapp->hmm_unreg_work, amdxdna_hmm_unreg_work);
+ if (is_import_bo(abo) && vma->vm_file && vma->vm_file->f_mapping)
+ mapping_set_unevictable(vma->vm_file->f_mapping);
+
+ down_write(&xdna->notifier_lock);
+ list_add_tail(&mapp->node, &abo->mem.umap_list);
+ up_write(&xdna->notifier_lock);
+
+ return 0;
+
+free_pfns:
+ kvfree(mapp->range.hmm_pfns);
+free_map:
+ kfree(mapp);
return ret;
}
+static void amdxdna_gem_dev_obj_free(struct drm_gem_object *gobj)
+{
+ struct amdxdna_dev *xdna = to_xdna_dev(gobj->dev);
+ struct amdxdna_gem_obj *abo = to_xdna_obj(gobj);
+
+ XDNA_DBG(xdna, "BO type %d xdna_addr 0x%llx", abo->type, abo->mem.dev_addr);
+ if (abo->pinned)
+ amdxdna_gem_unpin(abo);
+
+ amdxdna_gem_heap_free(abo);
+ drm_gem_object_release(gobj);
+ amdxdna_gem_destroy_obj(abo);
+}
+
+static int amdxdna_insert_pages(struct amdxdna_gem_obj *abo,
+ struct vm_area_struct *vma)
+{
+ struct amdxdna_dev *xdna = to_xdna_dev(to_gobj(abo)->dev);
+ unsigned long num_pages = vma_pages(vma);
+ unsigned long offset = 0;
+ int ret;
+
+ if (!is_import_bo(abo)) {
+ ret = drm_gem_shmem_mmap(&abo->base, vma);
+ if (ret) {
+ XDNA_ERR(xdna, "Failed shmem mmap %d", ret);
+ return ret;
+ }
+
+ /* The buffer is based on memory pages. Fix the flag. */
+ vm_flags_mod(vma, VM_MIXEDMAP, VM_PFNMAP);
+ ret = vm_insert_pages(vma, vma->vm_start, abo->base.pages,
+ &num_pages);
+ if (ret) {
+ XDNA_ERR(xdna, "Failed insert pages %d", ret);
+ vma->vm_ops->close(vma);
+ return ret;
+ }
+
+ return 0;
+ }
+
+ vma->vm_private_data = NULL;
+ vma->vm_ops = NULL;
+ ret = dma_buf_mmap(to_gobj(abo)->dma_buf, vma, 0);
+ if (ret) {
+ XDNA_ERR(xdna, "Failed to mmap dma buf %d", ret);
+ return ret;
+ }
+
+ do {
+ vm_fault_t fault_ret;
+
+ fault_ret = handle_mm_fault(vma, vma->vm_start + offset,
+ FAULT_FLAG_WRITE, NULL);
+ if (fault_ret & VM_FAULT_ERROR) {
+ vma->vm_ops->close(vma);
+ XDNA_ERR(xdna, "Fault in page failed");
+ return -EFAULT;
+ }
+
+ offset += PAGE_SIZE;
+ } while (--num_pages);
+
+ /* Drop the reference drm_gem_mmap_obj() acquired.*/
+ drm_gem_object_put(to_gobj(abo));
+
+ return 0;
+}
+
static int amdxdna_gem_obj_mmap(struct drm_gem_object *gobj,
struct vm_area_struct *vma)
{
+ struct amdxdna_dev *xdna = to_xdna_dev(gobj->dev);
struct amdxdna_gem_obj *abo = to_xdna_obj(gobj);
- unsigned long num_pages;
int ret;
- ret = amdxdna_hmm_register(abo, vma->vm_start, gobj->size);
+ ret = amdxdna_hmm_register(abo, vma);
if (ret)
return ret;
+ ret = amdxdna_insert_pages(abo, vma);
+ if (ret) {
+ XDNA_ERR(xdna, "Failed insert pages, ret %d", ret);
+ goto hmm_unreg;
+ }
+
+ XDNA_DBG(xdna, "BO map_offset 0x%llx type %d userptr 0x%lx size 0x%lx",
+ drm_vma_node_offset_addr(&gobj->vma_node), abo->type,
+ vma->vm_start, gobj->size);
+ return 0;
+
+hmm_unreg:
+ amdxdna_hmm_unregister(abo, vma);
+ return ret;
+}
+
+static int amdxdna_gem_dmabuf_mmap(struct dma_buf *dma_buf, struct vm_area_struct *vma)
+{
+ struct drm_gem_object *gobj = dma_buf->priv;
+ struct amdxdna_gem_obj *abo = to_xdna_obj(gobj);
+ unsigned long num_pages = vma_pages(vma);
+ int ret;
+
+ vma->vm_ops = &drm_gem_shmem_vm_ops;
+ vma->vm_private_data = gobj;
+
+ drm_gem_object_get(gobj);
ret = drm_gem_shmem_mmap(&abo->base, vma);
if (ret)
- goto hmm_unreg;
+ goto put_obj;
- num_pages = gobj->size >> PAGE_SHIFT;
- /* Try to insert the pages */
+ /* The buffer is based on memory pages. Fix the flag. */
vm_flags_mod(vma, VM_MIXEDMAP, VM_PFNMAP);
- ret = vm_insert_pages(vma, vma->vm_start, abo->base.pages, &num_pages);
+ ret = vm_insert_pages(vma, vma->vm_start, abo->base.pages,
+ &num_pages);
if (ret)
- XDNA_ERR(abo->client->xdna, "Failed insert pages, ret %d", ret);
+ goto close_vma;
return 0;
-hmm_unreg:
- amdxdna_hmm_unregister(abo);
+close_vma:
+ vma->vm_ops->close(vma);
+put_obj:
+ drm_gem_object_put(gobj);
return ret;
}
-static vm_fault_t amdxdna_gem_vm_fault(struct vm_fault *vmf)
+static const struct dma_buf_ops amdxdna_dmabuf_ops = {
+ .attach = drm_gem_map_attach,
+ .detach = drm_gem_map_detach,
+ .map_dma_buf = drm_gem_map_dma_buf,
+ .unmap_dma_buf = drm_gem_unmap_dma_buf,
+ .release = drm_gem_dmabuf_release,
+ .mmap = amdxdna_gem_dmabuf_mmap,
+ .vmap = drm_gem_dmabuf_vmap,
+ .vunmap = drm_gem_dmabuf_vunmap,
+};
+
+static struct dma_buf *amdxdna_gem_prime_export(struct drm_gem_object *gobj, int flags)
{
- return drm_gem_shmem_vm_ops.fault(vmf);
+ DEFINE_DMA_BUF_EXPORT_INFO(exp_info);
+
+ exp_info.ops = &amdxdna_dmabuf_ops;
+ exp_info.size = gobj->size;
+ exp_info.flags = flags;
+ exp_info.priv = gobj;
+ exp_info.resv = gobj->resv;
+
+ return drm_gem_dmabuf_export(gobj->dev, &exp_info);
}
-static void amdxdna_gem_vm_open(struct vm_area_struct *vma)
+static void amdxdna_imported_obj_free(struct amdxdna_gem_obj *abo)
{
- drm_gem_shmem_vm_ops.open(vma);
+ dma_buf_unmap_attachment_unlocked(abo->attach, abo->base.sgt, DMA_BIDIRECTIONAL);
+ dma_buf_detach(abo->dma_buf, abo->attach);
+ dma_buf_put(abo->dma_buf);
+ drm_gem_object_release(to_gobj(abo));
+ kfree(abo);
}
-static void amdxdna_gem_vm_close(struct vm_area_struct *vma)
+static void amdxdna_gem_obj_free(struct drm_gem_object *gobj)
{
- struct drm_gem_object *gobj = vma->vm_private_data;
+ struct amdxdna_dev *xdna = to_xdna_dev(gobj->dev);
+ struct amdxdna_gem_obj *abo = to_xdna_obj(gobj);
+ struct iosys_map map = IOSYS_MAP_INIT_VADDR(abo->mem.kva);
+
+ XDNA_DBG(xdna, "BO type %d xdna_addr 0x%llx", abo->type, abo->mem.dev_addr);
+
+ amdxdna_hmm_unregister(abo, NULL);
+ flush_workqueue(xdna->notifier_wq);
+
+ if (abo->pinned)
+ amdxdna_gem_unpin(abo);
- amdxdna_hmm_unregister(to_xdna_obj(gobj));
- drm_gem_shmem_vm_ops.close(vma);
+ if (abo->type == AMDXDNA_BO_DEV_HEAP)
+ drm_mm_takedown(&abo->mm);
+
+ drm_gem_vunmap(gobj, &map);
+ mutex_destroy(&abo->lock);
+
+ if (is_import_bo(abo)) {
+ amdxdna_imported_obj_free(abo);
+ return;
+ }
+
+ drm_gem_shmem_free(&abo->base);
}
-static const struct vm_operations_struct amdxdna_gem_vm_ops = {
- .fault = amdxdna_gem_vm_fault,
- .open = amdxdna_gem_vm_open,
- .close = amdxdna_gem_vm_close,
+static const struct drm_gem_object_funcs amdxdna_gem_dev_obj_funcs = {
+ .free = amdxdna_gem_dev_obj_free,
};
static const struct drm_gem_object_funcs amdxdna_gem_shmem_funcs = {
@@ -220,7 +454,8 @@ static const struct drm_gem_object_funcs amdxdna_gem_shmem_funcs = {
.vmap = drm_gem_shmem_object_vmap,
.vunmap = drm_gem_shmem_object_vunmap,
.mmap = amdxdna_gem_obj_mmap,
- .vm_ops = &amdxdna_gem_vm_ops,
+ .vm_ops = &drm_gem_shmem_vm_ops,
+ .export = amdxdna_gem_prime_export,
};
static struct amdxdna_gem_obj *
@@ -239,6 +474,7 @@ amdxdna_gem_create_obj(struct drm_device *dev, size_t size)
abo->mem.userptr = AMDXDNA_INVALID_ADDR;
abo->mem.dev_addr = AMDXDNA_INVALID_ADDR;
abo->mem.size = size;
+ INIT_LIST_HEAD(&abo->mem.umap_list);
return abo;
}
@@ -258,6 +494,51 @@ amdxdna_gem_create_object_cb(struct drm_device *dev, size_t size)
return to_gobj(abo);
}
+struct drm_gem_object *
+amdxdna_gem_prime_import(struct drm_device *dev, struct dma_buf *dma_buf)
+{
+ struct dma_buf_attachment *attach;
+ struct amdxdna_gem_obj *abo;
+ struct drm_gem_object *gobj;
+ struct sg_table *sgt;
+ int ret;
+
+ get_dma_buf(dma_buf);
+
+ attach = dma_buf_attach(dma_buf, dev->dev);
+ if (IS_ERR(attach)) {
+ ret = PTR_ERR(attach);
+ goto put_buf;
+ }
+
+ sgt = dma_buf_map_attachment_unlocked(attach, DMA_BIDIRECTIONAL);
+ if (IS_ERR(sgt)) {
+ ret = PTR_ERR(sgt);
+ goto fail_detach;
+ }
+
+ gobj = drm_gem_shmem_prime_import_sg_table(dev, attach, sgt);
+ if (IS_ERR(gobj)) {
+ ret = PTR_ERR(gobj);
+ goto fail_unmap;
+ }
+
+ abo = to_xdna_obj(gobj);
+ abo->attach = attach;
+ abo->dma_buf = dma_buf;
+
+ return gobj;
+
+fail_unmap:
+ dma_buf_unmap_attachment_unlocked(attach, sgt, DMA_BIDIRECTIONAL);
+fail_detach:
+ dma_buf_detach(dma_buf, attach);
+put_buf:
+ dma_buf_put(dma_buf);
+
+ return ERR_PTR(ret);
+}
+
static struct amdxdna_gem_obj *
amdxdna_drm_alloc_shmem(struct drm_device *dev,
struct amdxdna_drm_create_bo *args,
@@ -286,6 +567,7 @@ amdxdna_drm_create_dev_heap(struct drm_device *dev,
struct drm_file *filp)
{
struct amdxdna_client *client = filp->driver_priv;
+ struct iosys_map map = IOSYS_MAP_INIT_VADDR(NULL);
struct amdxdna_dev *xdna = to_xdna_dev(dev);
struct drm_gem_shmem_object *shmem;
struct amdxdna_gem_obj *abo;
@@ -312,18 +594,26 @@ amdxdna_drm_create_dev_heap(struct drm_device *dev,
shmem->map_wc = false;
abo = to_xdna_obj(&shmem->base);
-
abo->type = AMDXDNA_BO_DEV_HEAP;
abo->client = client;
abo->mem.dev_addr = client->xdna->dev_info->dev_mem_base;
drm_mm_init(&abo->mm, abo->mem.dev_addr, abo->mem.size);
+ ret = drm_gem_vmap(to_gobj(abo), &map);
+ if (ret) {
+ XDNA_ERR(xdna, "Vmap heap bo failed, ret %d", ret);
+ goto release_obj;
+ }
+ abo->mem.kva = map.vaddr;
+
client->dev_heap = abo;
drm_gem_object_get(to_gobj(abo));
mutex_unlock(&client->mm_lock);
return abo;
+release_obj:
+ drm_gem_object_put(to_gobj(abo));
mm_unlock:
mutex_unlock(&client->mm_lock);
return ERR_PTR(ret);
@@ -332,58 +622,32 @@ mm_unlock:
struct amdxdna_gem_obj *
amdxdna_drm_alloc_dev_bo(struct drm_device *dev,
struct amdxdna_drm_create_bo *args,
- struct drm_file *filp, bool use_vmap)
+ struct drm_file *filp)
{
struct amdxdna_client *client = filp->driver_priv;
struct amdxdna_dev *xdna = to_xdna_dev(dev);
size_t aligned_sz = PAGE_ALIGN(args->size);
- struct amdxdna_gem_obj *abo, *heap;
+ struct amdxdna_gem_obj *abo;
int ret;
- mutex_lock(&client->mm_lock);
- heap = client->dev_heap;
- if (!heap) {
- ret = -EINVAL;
- goto mm_unlock;
- }
-
- if (heap->mem.userptr == AMDXDNA_INVALID_ADDR) {
- XDNA_ERR(xdna, "Invalid dev heap userptr");
- ret = -EINVAL;
- goto mm_unlock;
- }
-
- if (args->size > heap->mem.size) {
- XDNA_ERR(xdna, "Invalid dev bo size 0x%llx, limit 0x%lx",
- args->size, heap->mem.size);
- ret = -EINVAL;
- goto mm_unlock;
- }
-
abo = amdxdna_gem_create_obj(&xdna->ddev, aligned_sz);
- if (IS_ERR(abo)) {
- ret = PTR_ERR(abo);
- goto mm_unlock;
- }
+ if (IS_ERR(abo))
+ return abo;
+
to_gobj(abo)->funcs = &amdxdna_gem_dev_obj_funcs;
abo->type = AMDXDNA_BO_DEV;
abo->client = client;
- abo->dev_heap = heap;
- ret = amdxdna_gem_insert_node_locked(abo, use_vmap);
+
+ ret = amdxdna_gem_heap_alloc(abo);
if (ret) {
XDNA_ERR(xdna, "Failed to alloc dev bo memory, ret %d", ret);
- goto mm_unlock;
+ amdxdna_gem_destroy_obj(abo);
+ return ERR_PTR(ret);
}
- drm_gem_object_get(to_gobj(heap));
drm_gem_private_object_init(&xdna->ddev, to_gobj(abo), aligned_sz);
- mutex_unlock(&client->mm_lock);
return abo;
-
-mm_unlock:
- mutex_unlock(&client->mm_lock);
- return ERR_PTR(ret);
}
static struct amdxdna_gem_obj *
@@ -391,10 +655,10 @@ amdxdna_drm_create_cmd_bo(struct drm_device *dev,
struct amdxdna_drm_create_bo *args,
struct drm_file *filp)
{
+ struct iosys_map map = IOSYS_MAP_INIT_VADDR(NULL);
struct amdxdna_dev *xdna = to_xdna_dev(dev);
struct drm_gem_shmem_object *shmem;
struct amdxdna_gem_obj *abo;
- struct iosys_map map;
int ret;
if (args->size > XDNA_MAX_CMD_BO_SIZE) {
@@ -417,7 +681,7 @@ amdxdna_drm_create_cmd_bo(struct drm_device *dev,
abo->type = AMDXDNA_BO_CMD;
abo->client = filp->driver_priv;
- ret = drm_gem_vmap_unlocked(to_gobj(abo), &map);
+ ret = drm_gem_vmap(to_gobj(abo), &map);
if (ret) {
XDNA_ERR(xdna, "Vmap cmd bo failed, ret %d", ret);
goto release_obj;
@@ -451,7 +715,7 @@ int amdxdna_drm_create_bo_ioctl(struct drm_device *dev, void *data, struct drm_f
abo = amdxdna_drm_create_dev_heap(dev, args, filp);
break;
case AMDXDNA_BO_DEV:
- abo = amdxdna_drm_alloc_dev_bo(dev, args, filp, false);
+ abo = amdxdna_drm_alloc_dev_bo(dev, args, filp);
break;
case AMDXDNA_BO_CMD:
abo = amdxdna_drm_create_cmd_bo(dev, args, filp);
@@ -483,17 +747,13 @@ int amdxdna_gem_pin_nolock(struct amdxdna_gem_obj *abo)
struct amdxdna_dev *xdna = to_xdna_dev(to_gobj(abo)->dev);
int ret;
- switch (abo->type) {
- case AMDXDNA_BO_SHMEM:
- case AMDXDNA_BO_DEV_HEAP:
- ret = drm_gem_shmem_pin(&abo->base);
- break;
- case AMDXDNA_BO_DEV:
- ret = drm_gem_shmem_pin(&abo->dev_heap->base);
- break;
- default:
- ret = -EOPNOTSUPP;
- }
+ if (abo->type == AMDXDNA_BO_DEV)
+ abo = abo->client->dev_heap;
+
+ if (is_import_bo(abo))
+ return 0;
+
+ ret = drm_gem_shmem_pin(&abo->base);
XDNA_DBG(xdna, "BO type %d ret %d", abo->type, ret);
return ret;
@@ -503,9 +763,6 @@ int amdxdna_gem_pin(struct amdxdna_gem_obj *abo)
{
int ret;
- if (abo->type == AMDXDNA_BO_DEV)
- abo = abo->dev_heap;
-
mutex_lock(&abo->lock);
ret = amdxdna_gem_pin_nolock(abo);
mutex_unlock(&abo->lock);
@@ -516,7 +773,10 @@ int amdxdna_gem_pin(struct amdxdna_gem_obj *abo)
void amdxdna_gem_unpin(struct amdxdna_gem_obj *abo)
{
if (abo->type == AMDXDNA_BO_DEV)
- abo = abo->dev_heap;
+ abo = abo->client->dev_heap;
+
+ if (is_import_bo(abo))
+ return;
mutex_lock(&abo->lock);
drm_gem_shmem_unpin(&abo->base);
@@ -606,10 +866,14 @@ int amdxdna_drm_sync_bo_ioctl(struct drm_device *dev,
goto put_obj;
}
- if (abo->type == AMDXDNA_BO_DEV)
- drm_clflush_pages(abo->mem.pages, abo->mem.nr_pages);
- else
+ if (is_import_bo(abo))
+ drm_clflush_sg(abo->base.sgt);
+ else if (abo->mem.kva)
+ drm_clflush_virt_range(abo->mem.kva + args->offset, args->size);
+ else if (abo->base.pages)
drm_clflush_pages(abo->base.pages, gobj->size >> PAGE_SHIFT);
+ else
+ drm_WARN(&xdna->ddev, 1, "Can not get flush memory");
amdxdna_gem_unpin(abo);
diff --git a/drivers/accel/amdxdna/amdxdna_gem.h b/drivers/accel/amdxdna/amdxdna_gem.h
index 8ccc0375dd9d..ae29db94a9d3 100644
--- a/drivers/accel/amdxdna/amdxdna_gem.h
+++ b/drivers/accel/amdxdna/amdxdna_gem.h
@@ -6,6 +6,20 @@
#ifndef _AMDXDNA_GEM_H_
#define _AMDXDNA_GEM_H_
+#include <linux/hmm.h>
+
+struct amdxdna_umap {
+ struct vm_area_struct *vma;
+ struct mmu_interval_notifier notifier;
+ struct hmm_range range;
+ struct work_struct hmm_unreg_work;
+ struct amdxdna_gem_obj *abo;
+ struct list_head node;
+ struct kref refcnt;
+ bool invalid;
+ bool unmapped;
+};
+
struct amdxdna_mem {
u64 userptr;
void *kva;
@@ -13,8 +27,7 @@ struct amdxdna_mem {
size_t size;
struct page **pages;
u32 nr_pages;
- struct mmu_interval_notifier notifier;
- unsigned long *pfns;
+ struct list_head umap_list;
bool map_invalid;
};
@@ -28,12 +41,14 @@ struct amdxdna_gem_obj {
/* Below members is uninitialized when needed */
struct drm_mm mm; /* For AMDXDNA_BO_DEV_HEAP */
- struct amdxdna_gem_obj *dev_heap; /* For AMDXDNA_BO_DEV */
struct drm_mm_node mm_node; /* For AMDXDNA_BO_DEV */
u32 assigned_hwctx;
+ struct dma_buf *dma_buf;
+ struct dma_buf_attachment *attach;
};
#define to_gobj(obj) (&(obj)->base.base)
+#define is_import_bo(obj) ((obj)->attach)
static inline struct amdxdna_gem_obj *to_xdna_obj(struct drm_gem_object *gobj)
{
@@ -47,12 +62,16 @@ static inline void amdxdna_gem_put_obj(struct amdxdna_gem_obj *abo)
drm_gem_object_put(to_gobj(abo));
}
+void amdxdna_umap_put(struct amdxdna_umap *mapp);
+
struct drm_gem_object *
amdxdna_gem_create_object_cb(struct drm_device *dev, size_t size);
+struct drm_gem_object *
+amdxdna_gem_prime_import(struct drm_device *dev, struct dma_buf *dma_buf);
struct amdxdna_gem_obj *
amdxdna_drm_alloc_dev_bo(struct drm_device *dev,
struct amdxdna_drm_create_bo *args,
- struct drm_file *filp, bool use_vmap);
+ struct drm_file *filp);
int amdxdna_gem_pin_nolock(struct amdxdna_gem_obj *abo);
int amdxdna_gem_pin(struct amdxdna_gem_obj *abo);
diff --git a/drivers/accel/amdxdna/amdxdna_pci_drv.c b/drivers/accel/amdxdna/amdxdna_pci_drv.c
index f5b8497cf5ad..f2bf1d374cc7 100644
--- a/drivers/accel/amdxdna/amdxdna_pci_drv.c
+++ b/drivers/accel/amdxdna/amdxdna_pci_drv.c
@@ -226,6 +226,7 @@ const struct drm_driver amdxdna_drm_drv = {
.num_ioctls = ARRAY_SIZE(amdxdna_drm_ioctls),
.gem_create_object = amdxdna_gem_create_object_cb,
+ .gem_prime_import = amdxdna_gem_prime_import,
};
static const struct amdxdna_dev_info *
@@ -266,12 +267,16 @@ static int amdxdna_probe(struct pci_dev *pdev, const struct pci_device_id *id)
fs_reclaim_release(GFP_KERNEL);
}
+ xdna->notifier_wq = alloc_ordered_workqueue("notifier_wq", 0);
+ if (!xdna->notifier_wq)
+ return -ENOMEM;
+
mutex_lock(&xdna->dev_lock);
ret = xdna->dev_info->ops->init(xdna);
mutex_unlock(&xdna->dev_lock);
if (ret) {
XDNA_ERR(xdna, "Hardware init failed, ret %d", ret);
- return ret;
+ goto destroy_notifier_wq;
}
ret = amdxdna_sysfs_init(xdna);
@@ -301,6 +306,8 @@ failed_dev_fini:
mutex_lock(&xdna->dev_lock);
xdna->dev_info->ops->fini(xdna);
mutex_unlock(&xdna->dev_lock);
+destroy_notifier_wq:
+ destroy_workqueue(xdna->notifier_wq);
return ret;
}
@@ -310,6 +317,8 @@ static void amdxdna_remove(struct pci_dev *pdev)
struct device *dev = &pdev->dev;
struct amdxdna_client *client;
+ destroy_workqueue(xdna->notifier_wq);
+
pm_runtime_get_noresume(dev);
pm_runtime_forbid(dev);
diff --git a/drivers/accel/amdxdna/amdxdna_pci_drv.h b/drivers/accel/amdxdna/amdxdna_pci_drv.h
index 37848a8d8031..ab79600911aa 100644
--- a/drivers/accel/amdxdna/amdxdna_pci_drv.h
+++ b/drivers/accel/amdxdna/amdxdna_pci_drv.h
@@ -6,6 +6,7 @@
#ifndef _AMDXDNA_PCI_DRV_H_
#define _AMDXDNA_PCI_DRV_H_
+#include <linux/workqueue.h>
#include <linux/xarray.h>
#define XDNA_INFO(xdna, fmt, args...) drm_info(&(xdna)->ddev, fmt, ##args)
@@ -98,6 +99,7 @@ struct amdxdna_dev {
struct list_head client_list;
struct amdxdna_fw_ver fw_ver;
struct rw_semaphore notifier_lock; /* for mmu notifier*/
+ struct workqueue_struct *notifier_wq;
};
/*
diff --git a/drivers/accel/drm_accel.c b/drivers/accel/drm_accel.c
index aa826033b0ce..ca3357acd127 100644
--- a/drivers/accel/drm_accel.c
+++ b/drivers/accel/drm_accel.c
@@ -20,8 +20,6 @@
DEFINE_XARRAY_ALLOC(accel_minors_xa);
-static struct dentry *accel_debugfs_root;
-
static const struct device_type accel_sysfs_device_minor = {
.name = "accel_minor"
};
@@ -74,17 +72,6 @@ static const struct drm_info_list accel_debugfs_list[] = {
#define ACCEL_DEBUGFS_ENTRIES ARRAY_SIZE(accel_debugfs_list)
/**
- * accel_debugfs_init() - Initialize debugfs for device
- * @dev: Pointer to the device instance.
- *
- * This function creates a root directory for the device in debugfs.
- */
-void accel_debugfs_init(struct drm_device *dev)
-{
- drm_debugfs_dev_init(dev, accel_debugfs_root);
-}
-
-/**
* accel_debugfs_register() - Register debugfs for device
* @dev: Pointer to the device instance.
*
@@ -194,7 +181,6 @@ static const struct file_operations accel_stub_fops = {
void accel_core_exit(void)
{
unregister_chrdev(ACCEL_MAJOR, "accel");
- debugfs_remove(accel_debugfs_root);
accel_sysfs_destroy();
WARN_ON(!xa_empty(&accel_minors_xa));
}
@@ -209,8 +195,6 @@ int __init accel_core_init(void)
goto error;
}
- accel_debugfs_root = debugfs_create_dir("accel", NULL);
-
ret = register_chrdev(ACCEL_MAJOR, "accel", &accel_stub_fops);
if (ret < 0)
DRM_ERROR("Cannot register ACCEL major: %d\n", ret);
diff --git a/drivers/accel/habanalabs/Kconfig b/drivers/accel/habanalabs/Kconfig
index be85336107f9..1919fbb169c7 100644
--- a/drivers/accel/habanalabs/Kconfig
+++ b/drivers/accel/habanalabs/Kconfig
@@ -6,7 +6,7 @@
config DRM_ACCEL_HABANALABS
tristate "HabanaLabs AI accelerators"
depends on DRM_ACCEL
- depends on X86_64
+ depends on X86 && X86_64
depends on PCI && HAS_IOMEM
select GENERIC_ALLOCATOR
select HWMON
diff --git a/drivers/accel/habanalabs/common/device.c b/drivers/accel/habanalabs/common/device.c
index 68eebed3b050..80fa08bf57bd 100644
--- a/drivers/accel/habanalabs/common/device.c
+++ b/drivers/accel/habanalabs/common/device.c
@@ -1066,28 +1066,11 @@ static bool is_pci_link_healthy(struct hl_device *hdev)
return (device_id == hdev->pdev->device);
}
-static void stringify_time_of_last_heartbeat(struct hl_device *hdev, char *time_str, size_t size,
- bool is_pq_hb)
-{
- time64_t seconds = is_pq_hb ? hdev->heartbeat_debug_info.last_pq_heartbeat_ts
- : hdev->heartbeat_debug_info.last_eq_heartbeat_ts;
- struct tm tm;
-
- if (!seconds)
- return;
-
- time64_to_tm(seconds, 0, &tm);
-
- snprintf(time_str, size, "%ld-%02d-%02d %02d:%02d:%02d (UTC)",
- tm.tm_year + 1900, tm.tm_mon, tm.tm_mday, tm.tm_hour, tm.tm_min, tm.tm_sec);
-}
-
static bool hl_device_eq_heartbeat_received(struct hl_device *hdev)
{
struct eq_heartbeat_debug_info *heartbeat_debug_info = &hdev->heartbeat_debug_info;
u32 cpu_q_id = heartbeat_debug_info->cpu_queue_id, pq_pi_mask = (HL_QUEUE_LENGTH << 1) - 1;
struct asic_fixed_properties *prop = &hdev->asic_prop;
- char pq_time_str[64] = "N/A", eq_time_str[64] = "N/A";
if (!prop->cpucp_info.eq_health_check_supported)
return true;
@@ -1095,17 +1078,15 @@ static bool hl_device_eq_heartbeat_received(struct hl_device *hdev)
if (!hdev->eq_heartbeat_received) {
dev_err(hdev->dev, "EQ heartbeat event was not received!\n");
- stringify_time_of_last_heartbeat(hdev, pq_time_str, sizeof(pq_time_str), true);
- stringify_time_of_last_heartbeat(hdev, eq_time_str, sizeof(eq_time_str), false);
dev_err(hdev->dev,
- "EQ: {CI %u, HB counter %u, last HB time: %s}, PQ: {PI: %u, CI: %u (%u), last HB time: %s}\n",
+ "EQ: {CI %u, HB counter %u, last HB time: %ptTs}, PQ: {PI: %u, CI: %u (%u), last HB time: %ptTs}\n",
hdev->event_queue.ci,
heartbeat_debug_info->heartbeat_event_counter,
- eq_time_str,
+ &hdev->heartbeat_debug_info.last_eq_heartbeat_ts,
hdev->kernel_queues[cpu_q_id].pi,
atomic_read(&hdev->kernel_queues[cpu_q_id].ci),
atomic_read(&hdev->kernel_queues[cpu_q_id].ci) & pq_pi_mask,
- pq_time_str);
+ &hdev->heartbeat_debug_info.last_pq_heartbeat_ts);
hl_eq_dump(hdev, &hdev->event_queue);
diff --git a/drivers/accel/habanalabs/common/habanalabs_ioctl.c b/drivers/accel/habanalabs/common/habanalabs_ioctl.c
index 8729a0c57d78..dc80ca921d90 100644
--- a/drivers/accel/habanalabs/common/habanalabs_ioctl.c
+++ b/drivers/accel/habanalabs/common/habanalabs_ioctl.c
@@ -17,8 +17,6 @@
#include <linux/uaccess.h>
#include <linux/vmalloc.h>
-#include <asm/msr.h>
-
/* make sure there is space for all the signed info */
static_assert(sizeof(struct cpucp_info) <= SEC_DEV_INFO_BUF_SZ);
diff --git a/drivers/accel/habanalabs/common/sysfs.c b/drivers/accel/habanalabs/common/sysfs.c
index 9d58efa2ff38..82f66520ec18 100644
--- a/drivers/accel/habanalabs/common/sysfs.c
+++ b/drivers/accel/habanalabs/common/sysfs.c
@@ -446,7 +446,7 @@ static DEVICE_ATTR_RO(parent_device);
static const struct bin_attribute bin_attr_eeprom = {
.attr = {.name = "eeprom", .mode = (0444)},
.size = PAGE_SIZE,
- .read_new = eeprom_read_handler
+ .read = eeprom_read_handler
};
static struct attribute *hl_dev_attrs[] = {
@@ -479,7 +479,7 @@ static const struct bin_attribute *const hl_dev_bin_attrs[] = {
static struct attribute_group hl_dev_attr_group = {
.attrs = hl_dev_attrs,
- .bin_attrs_new = hl_dev_bin_attrs,
+ .bin_attrs = hl_dev_bin_attrs,
};
static struct attribute_group hl_dev_clks_attr_group;
diff --git a/drivers/accel/ivpu/ivpu_debugfs.c b/drivers/accel/ivpu/ivpu_debugfs.c
index 0825851656a2..cd24ccd20ba6 100644
--- a/drivers/accel/ivpu/ivpu_debugfs.c
+++ b/drivers/accel/ivpu/ivpu_debugfs.c
@@ -332,7 +332,7 @@ ivpu_force_recovery_fn(struct file *file, const char __user *user_buf, size_t si
return -EINVAL;
ret = ivpu_rpm_get(vdev);
- if (ret)
+ if (ret < 0)
return ret;
ivpu_pm_trigger_recovery(vdev, "debugfs");
@@ -383,7 +383,7 @@ static int dct_active_set(void *data, u64 active_percent)
return -EINVAL;
ret = ivpu_rpm_get(vdev);
- if (ret)
+ if (ret < 0)
return ret;
if (active_percent)
@@ -455,7 +455,7 @@ priority_bands_fops_write(struct file *file, const char __user *user_buf, size_t
if (ret < 0)
return ret;
- buf[size] = '\0';
+ buf[ret] = '\0';
ret = sscanf(buf, "%u %u %u %u", &band, &grace_period, &process_grace_period,
&process_quantum);
if (ret != 4)
diff --git a/drivers/accel/ivpu/ivpu_drv.c b/drivers/accel/ivpu/ivpu_drv.c
index 4fa73189502e..3d6d52492536 100644
--- a/drivers/accel/ivpu/ivpu_drv.c
+++ b/drivers/accel/ivpu/ivpu_drv.c
@@ -1,6 +1,6 @@
// SPDX-License-Identifier: GPL-2.0-only
/*
- * Copyright (C) 2020-2024 Intel Corporation
+ * Copyright (C) 2020-2025 Intel Corporation
*/
#include <linux/firmware.h>
@@ -164,7 +164,7 @@ static int ivpu_get_param_ioctl(struct drm_device *dev, void *data, struct drm_f
args->value = vdev->platform;
break;
case DRM_IVPU_PARAM_CORE_CLOCK_RATE:
- args->value = ivpu_hw_ratio_to_freq(vdev, vdev->hw->pll.max_ratio);
+ args->value = ivpu_hw_dpu_max_freq_get(vdev);
break;
case DRM_IVPU_PARAM_NUM_CONTEXTS:
args->value = ivpu_get_context_count(vdev);
@@ -374,6 +374,9 @@ int ivpu_boot(struct ivpu_device *vdev)
{
int ret;
+ drm_WARN_ON(&vdev->drm, atomic_read(&vdev->job_timeout_counter));
+ drm_WARN_ON(&vdev->drm, !xa_empty(&vdev->submitted_jobs_xa));
+
/* Update boot params located at first 4KB of FW memory */
ivpu_fw_boot_params_setup(vdev, ivpu_bo_vaddr(vdev->fw->mem));
@@ -421,9 +424,9 @@ void ivpu_prepare_for_reset(struct ivpu_device *vdev)
{
ivpu_hw_irq_disable(vdev);
disable_irq(vdev->irq);
- cancel_work_sync(&vdev->irq_ipc_work);
- cancel_work_sync(&vdev->irq_dct_work);
- cancel_work_sync(&vdev->context_abort_work);
+ flush_work(&vdev->irq_ipc_work);
+ flush_work(&vdev->irq_dct_work);
+ flush_work(&vdev->context_abort_work);
ivpu_ipc_disable(vdev);
ivpu_mmu_disable(vdev);
}
@@ -573,6 +576,7 @@ static int ivpu_dev_init(struct ivpu_device *vdev)
vdev->context_xa_limit.min = IVPU_USER_CONTEXT_MIN_SSID;
vdev->context_xa_limit.max = IVPU_USER_CONTEXT_MAX_SSID;
atomic64_set(&vdev->unique_id_counter, 0);
+ atomic_set(&vdev->job_timeout_counter, 0);
xa_init_flags(&vdev->context_xa, XA_FLAGS_ALLOC | XA_FLAGS_LOCK_IRQ);
xa_init_flags(&vdev->submitted_jobs_xa, XA_FLAGS_ALLOC1);
xa_init_flags(&vdev->db_xa, XA_FLAGS_ALLOC1);
@@ -700,6 +704,7 @@ static struct pci_device_id ivpu_pci_ids[] = {
{ PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_ARL) },
{ PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_LNL) },
{ PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_PTL_P) },
+ { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_WCL) },
{ }
};
MODULE_DEVICE_TABLE(pci, ivpu_pci_ids);
diff --git a/drivers/accel/ivpu/ivpu_drv.h b/drivers/accel/ivpu/ivpu_drv.h
index 92753effb1c9..62ab1c654e63 100644
--- a/drivers/accel/ivpu/ivpu_drv.h
+++ b/drivers/accel/ivpu/ivpu_drv.h
@@ -1,6 +1,6 @@
/* SPDX-License-Identifier: GPL-2.0-only */
/*
- * Copyright (C) 2020-2024 Intel Corporation
+ * Copyright (C) 2020-2025 Intel Corporation
*/
#ifndef __IVPU_DRV_H__
@@ -26,6 +26,7 @@
#define PCI_DEVICE_ID_ARL 0xad1d
#define PCI_DEVICE_ID_LNL 0x643e
#define PCI_DEVICE_ID_PTL_P 0xb03e
+#define PCI_DEVICE_ID_WCL 0xfd3e
#define IVPU_HW_IP_37XX 37
#define IVPU_HW_IP_40XX 40
@@ -154,6 +155,7 @@ struct ivpu_device {
struct mutex submitted_jobs_lock; /* Protects submitted_jobs */
struct xarray submitted_jobs_xa;
struct ivpu_ipc_consumer job_done_consumer;
+ atomic_t job_timeout_counter;
atomic64_t unique_id_counter;
@@ -164,6 +166,7 @@ struct ivpu_device {
int boot;
int jsm;
int tdr;
+ int inference;
int autosuspend;
int d0i3_entry_msg;
int state_dump_msg;
@@ -206,10 +209,11 @@ extern bool ivpu_force_snoop;
#define IVPU_TEST_MODE_D0I3_MSG_ENABLE BIT(5)
#define IVPU_TEST_MODE_MIP_DISABLE BIT(6)
#define IVPU_TEST_MODE_DISABLE_TIMEOUTS BIT(8)
-#define IVPU_TEST_MODE_TURBO BIT(9)
-#define IVPU_TEST_MODE_CLK_RELINQ_DISABLE BIT(10)
-#define IVPU_TEST_MODE_CLK_RELINQ_ENABLE BIT(11)
-#define IVPU_TEST_MODE_D0I2_DISABLE BIT(12)
+#define IVPU_TEST_MODE_TURBO_ENABLE BIT(9)
+#define IVPU_TEST_MODE_TURBO_DISABLE BIT(10)
+#define IVPU_TEST_MODE_CLK_RELINQ_DISABLE BIT(11)
+#define IVPU_TEST_MODE_CLK_RELINQ_ENABLE BIT(12)
+#define IVPU_TEST_MODE_D0I2_DISABLE BIT(13)
extern int ivpu_test_mode;
struct ivpu_file_priv *ivpu_file_priv_get(struct ivpu_file_priv *file_priv);
@@ -239,6 +243,7 @@ static inline int ivpu_hw_ip_gen(struct ivpu_device *vdev)
case PCI_DEVICE_ID_LNL:
return IVPU_HW_IP_40XX;
case PCI_DEVICE_ID_PTL_P:
+ case PCI_DEVICE_ID_WCL:
return IVPU_HW_IP_50XX;
default:
dump_stack();
@@ -255,6 +260,7 @@ static inline int ivpu_hw_btrs_gen(struct ivpu_device *vdev)
return IVPU_HW_BTRS_MTL;
case PCI_DEVICE_ID_LNL:
case PCI_DEVICE_ID_PTL_P:
+ case PCI_DEVICE_ID_WCL:
return IVPU_HW_BTRS_LNL;
default:
dump_stack();
diff --git a/drivers/accel/ivpu/ivpu_fw.c b/drivers/accel/ivpu/ivpu_fw.c
index 7a1bb92d8c81..9db741695401 100644
--- a/drivers/accel/ivpu/ivpu_fw.c
+++ b/drivers/accel/ivpu/ivpu_fw.c
@@ -1,6 +1,6 @@
// SPDX-License-Identifier: GPL-2.0-only
/*
- * Copyright (C) 2020-2024 Intel Corporation
+ * Copyright (C) 2020-2025 Intel Corporation
*/
#include <linux/firmware.h>
@@ -55,18 +55,18 @@ static struct {
int gen;
const char *name;
} fw_names[] = {
- { IVPU_HW_IP_37XX, "vpu_37xx.bin" },
+ { IVPU_HW_IP_37XX, "intel/vpu/vpu_37xx_v1.bin" },
{ IVPU_HW_IP_37XX, "intel/vpu/vpu_37xx_v0.0.bin" },
- { IVPU_HW_IP_40XX, "vpu_40xx.bin" },
+ { IVPU_HW_IP_40XX, "intel/vpu/vpu_40xx_v1.bin" },
{ IVPU_HW_IP_40XX, "intel/vpu/vpu_40xx_v0.0.bin" },
- { IVPU_HW_IP_50XX, "vpu_50xx.bin" },
+ { IVPU_HW_IP_50XX, "intel/vpu/vpu_50xx_v1.bin" },
{ IVPU_HW_IP_50XX, "intel/vpu/vpu_50xx_v0.0.bin" },
};
/* Production fw_names from the table above */
-MODULE_FIRMWARE("intel/vpu/vpu_37xx_v0.0.bin");
-MODULE_FIRMWARE("intel/vpu/vpu_40xx_v0.0.bin");
-MODULE_FIRMWARE("intel/vpu/vpu_50xx_v0.0.bin");
+MODULE_FIRMWARE("intel/vpu/vpu_37xx_v1.bin");
+MODULE_FIRMWARE("intel/vpu/vpu_40xx_v1.bin");
+MODULE_FIRMWARE("intel/vpu/vpu_50xx_v1.bin");
static int ivpu_fw_request(struct ivpu_device *vdev)
{
@@ -233,10 +233,20 @@ static int ivpu_fw_parse(struct ivpu_device *vdev)
fw->dvfs_mode = 0;
fw->sched_mode = ivpu_fw_sched_mode_select(vdev, fw_hdr);
- fw->primary_preempt_buf_size = fw_hdr->preemption_buffer_1_size;
- fw->secondary_preempt_buf_size = fw_hdr->preemption_buffer_2_size;
ivpu_info(vdev, "Scheduler mode: %s\n", fw->sched_mode ? "HW" : "OS");
+ if (fw_hdr->preemption_buffer_1_max_size)
+ fw->primary_preempt_buf_size = fw_hdr->preemption_buffer_1_max_size;
+ else
+ fw->primary_preempt_buf_size = fw_hdr->preemption_buffer_1_size;
+
+ if (fw_hdr->preemption_buffer_2_max_size)
+ fw->secondary_preempt_buf_size = fw_hdr->preemption_buffer_2_max_size;
+ else
+ fw->secondary_preempt_buf_size = fw_hdr->preemption_buffer_2_size;
+ ivpu_dbg(vdev, FW_BOOT, "Preemption buffer sizes: primary %u, secondary %u\n",
+ fw->primary_preempt_buf_size, fw->secondary_preempt_buf_size);
+
if (fw_hdr->ro_section_start_address && !is_within_range(fw_hdr->ro_section_start_address,
fw_hdr->ro_section_size,
fw_hdr->image_load_address,
@@ -534,7 +544,7 @@ static void ivpu_fw_boot_params_print(struct ivpu_device *vdev, struct vpu_boot_
boot_params->d0i3_entry_vpu_ts);
ivpu_dbg(vdev, FW_BOOT, "boot_params.system_time_us = %llu\n",
boot_params->system_time_us);
- ivpu_dbg(vdev, FW_BOOT, "boot_params.power_profile = %u\n",
+ ivpu_dbg(vdev, FW_BOOT, "boot_params.power_profile = 0x%x\n",
boot_params->power_profile);
}
@@ -566,7 +576,6 @@ void ivpu_fw_boot_params_setup(struct ivpu_device *vdev, struct vpu_boot_params
boot_params->magic = VPU_BOOT_PARAMS_MAGIC;
boot_params->vpu_id = to_pci_dev(vdev->drm.dev)->bus->number;
- boot_params->frequency = ivpu_hw_pll_freq_get(vdev);
/*
* This param is a debug firmware feature. It switches default clock
@@ -637,7 +646,7 @@ void ivpu_fw_boot_params_setup(struct ivpu_device *vdev, struct vpu_boot_params
boot_params->d0i3_residency_time_us = 0;
boot_params->d0i3_entry_vpu_ts = 0;
if (IVPU_WA(disable_d0i2))
- boot_params->power_profile = 1;
+ boot_params->power_profile |= BIT(1);
boot_params->system_time_us = ktime_to_us(ktime_get_real());
wmb(); /* Flush WC buffers after writing bootparams */
diff --git a/drivers/accel/ivpu/ivpu_fw.h b/drivers/accel/ivpu/ivpu_fw.h
index 1d0b2bd9d65c..9a3935be1c05 100644
--- a/drivers/accel/ivpu/ivpu_fw.h
+++ b/drivers/accel/ivpu/ivpu_fw.h
@@ -39,6 +39,7 @@ struct ivpu_fw_info {
u64 read_only_addr;
u32 read_only_size;
u32 sched_mode;
+ u64 last_heartbeat;
};
int ivpu_fw_init(struct ivpu_device *vdev);
diff --git a/drivers/accel/ivpu/ivpu_gem.c b/drivers/accel/ivpu/ivpu_gem.c
index 8741c73b92ce..59cfcf3eaded 100644
--- a/drivers/accel/ivpu/ivpu_gem.c
+++ b/drivers/accel/ivpu/ivpu_gem.c
@@ -28,9 +28,19 @@ static inline void ivpu_dbg_bo(struct ivpu_device *vdev, struct ivpu_bo *bo, con
{
ivpu_dbg(vdev, BO,
"%6s: bo %8p vpu_addr %9llx size %8zu ctx %d has_pages %d dma_mapped %d mmu_mapped %d wc %d imported %d\n",
- action, bo, bo->vpu_addr, ivpu_bo_size(bo), bo->ctx ? bo->ctx->id : 0,
+ action, bo, bo->vpu_addr, ivpu_bo_size(bo), bo->ctx_id,
(bool)bo->base.pages, (bool)bo->base.sgt, bo->mmu_mapped, bo->base.map_wc,
- (bool)bo->base.base.import_attach);
+ (bool)drm_gem_is_imported(&bo->base.base));
+}
+
+static inline int ivpu_bo_lock(struct ivpu_bo *bo)
+{
+ return dma_resv_lock(bo->base.base.resv, NULL);
+}
+
+static inline void ivpu_bo_unlock(struct ivpu_bo *bo)
+{
+ dma_resv_unlock(bo->base.base.resv);
}
/*
@@ -43,22 +53,22 @@ static inline void ivpu_dbg_bo(struct ivpu_device *vdev, struct ivpu_bo *bo, con
int __must_check ivpu_bo_pin(struct ivpu_bo *bo)
{
struct ivpu_device *vdev = ivpu_bo_to_vdev(bo);
+ struct sg_table *sgt;
int ret = 0;
- mutex_lock(&bo->lock);
-
ivpu_dbg_bo(vdev, bo, "pin");
- drm_WARN_ON(&vdev->drm, !bo->ctx);
- if (!bo->mmu_mapped) {
- struct sg_table *sgt = drm_gem_shmem_get_pages_sgt(&bo->base);
+ sgt = drm_gem_shmem_get_pages_sgt(&bo->base);
+ if (IS_ERR(sgt)) {
+ ret = PTR_ERR(sgt);
+ ivpu_err(vdev, "Failed to map BO in IOMMU: %d\n", ret);
+ return ret;
+ }
- if (IS_ERR(sgt)) {
- ret = PTR_ERR(sgt);
- ivpu_err(vdev, "Failed to map BO in IOMMU: %d\n", ret);
- goto unlock;
- }
+ ivpu_bo_lock(bo);
+ if (!bo->mmu_mapped) {
+ drm_WARN_ON(&vdev->drm, !bo->ctx);
ret = ivpu_mmu_context_map_sgt(vdev, bo->ctx, bo->vpu_addr, sgt,
ivpu_bo_is_snooped(bo));
if (ret) {
@@ -69,7 +79,7 @@ int __must_check ivpu_bo_pin(struct ivpu_bo *bo)
}
unlock:
- mutex_unlock(&bo->lock);
+ ivpu_bo_unlock(bo);
return ret;
}
@@ -84,7 +94,7 @@ ivpu_bo_alloc_vpu_addr(struct ivpu_bo *bo, struct ivpu_mmu_context *ctx,
if (!drm_dev_enter(&vdev->drm, &idx))
return -ENODEV;
- mutex_lock(&bo->lock);
+ ivpu_bo_lock(bo);
ret = ivpu_mmu_context_insert_node(ctx, range, ivpu_bo_size(bo), &bo->mm_node);
if (!ret) {
@@ -94,9 +104,7 @@ ivpu_bo_alloc_vpu_addr(struct ivpu_bo *bo, struct ivpu_mmu_context *ctx,
ivpu_err(vdev, "Failed to add BO to context %u: %d\n", ctx->id, ret);
}
- ivpu_dbg_bo(vdev, bo, "alloc");
-
- mutex_unlock(&bo->lock);
+ ivpu_bo_unlock(bo);
drm_dev_exit(idx);
@@ -107,7 +115,7 @@ static void ivpu_bo_unbind_locked(struct ivpu_bo *bo)
{
struct ivpu_device *vdev = ivpu_bo_to_vdev(bo);
- lockdep_assert(lockdep_is_held(&bo->lock) || !kref_read(&bo->base.base.refcount));
+ lockdep_assert(dma_resv_held(bo->base.base.resv) || !kref_read(&bo->base.base.refcount));
if (bo->mmu_mapped) {
drm_WARN_ON(&vdev->drm, !bo->ctx);
@@ -122,17 +130,15 @@ static void ivpu_bo_unbind_locked(struct ivpu_bo *bo)
bo->ctx = NULL;
}
- if (bo->base.base.import_attach)
+ if (drm_gem_is_imported(&bo->base.base))
return;
- dma_resv_lock(bo->base.base.resv, NULL);
if (bo->base.sgt) {
dma_unmap_sgtable(vdev->drm.dev, bo->base.sgt, DMA_BIDIRECTIONAL, 0);
sg_free_table(bo->base.sgt);
kfree(bo->base.sgt);
bo->base.sgt = NULL;
}
- dma_resv_unlock(bo->base.base.resv);
}
void ivpu_bo_unbind_all_bos_from_context(struct ivpu_device *vdev, struct ivpu_mmu_context *ctx)
@@ -144,12 +150,12 @@ void ivpu_bo_unbind_all_bos_from_context(struct ivpu_device *vdev, struct ivpu_m
mutex_lock(&vdev->bo_list_lock);
list_for_each_entry(bo, &vdev->bo_list, bo_list_node) {
- mutex_lock(&bo->lock);
+ ivpu_bo_lock(bo);
if (bo->ctx == ctx) {
ivpu_dbg_bo(vdev, bo, "unbind");
ivpu_bo_unbind_locked(bo);
}
- mutex_unlock(&bo->lock);
+ ivpu_bo_unlock(bo);
}
mutex_unlock(&vdev->bo_list_lock);
}
@@ -169,7 +175,6 @@ struct drm_gem_object *ivpu_gem_create_object(struct drm_device *dev, size_t siz
bo->base.pages_mark_dirty_on_put = true; /* VPU can dirty a BO anytime */
INIT_LIST_HEAD(&bo->bo_list_node);
- mutex_init(&bo->lock);
return &bo->base.base;
}
@@ -215,7 +220,7 @@ fail_detach:
return ERR_PTR(ret);
}
-static struct ivpu_bo *ivpu_bo_alloc(struct ivpu_device *vdev, u64 size, u32 flags)
+static struct ivpu_bo *ivpu_bo_alloc(struct ivpu_device *vdev, u64 size, u32 flags, u32 ctx_id)
{
struct drm_gem_shmem_object *shmem;
struct ivpu_bo *bo;
@@ -233,6 +238,7 @@ static struct ivpu_bo *ivpu_bo_alloc(struct ivpu_device *vdev, u64 size, u32 fla
return ERR_CAST(shmem);
bo = to_ivpu_bo(&shmem->base);
+ bo->ctx_id = ctx_id;
bo->base.map_wc = flags & DRM_IVPU_BO_WC;
bo->flags = flags;
@@ -240,6 +246,8 @@ static struct ivpu_bo *ivpu_bo_alloc(struct ivpu_device *vdev, u64 size, u32 fla
list_add_tail(&bo->bo_list_node, &vdev->bo_list);
mutex_unlock(&vdev->bo_list_lock);
+ ivpu_dbg_bo(vdev, bo, "alloc");
+
return bo;
}
@@ -277,12 +285,16 @@ static void ivpu_gem_bo_free(struct drm_gem_object *obj)
list_del(&bo->bo_list_node);
mutex_unlock(&vdev->bo_list_lock);
- drm_WARN_ON(&vdev->drm, !dma_resv_test_signaled(obj->resv, DMA_RESV_USAGE_READ));
+ drm_WARN_ON(&vdev->drm, !drm_gem_is_imported(&bo->base.base) &&
+ !dma_resv_test_signaled(obj->resv, DMA_RESV_USAGE_READ));
+ drm_WARN_ON(&vdev->drm, ivpu_bo_size(bo) == 0);
+ drm_WARN_ON(&vdev->drm, bo->base.vaddr);
ivpu_bo_unbind_locked(bo);
- mutex_destroy(&bo->lock);
+ drm_WARN_ON(&vdev->drm, bo->mmu_mapped);
+ drm_WARN_ON(&vdev->drm, bo->ctx);
- drm_WARN_ON(obj->dev, bo->base.pages_use_count > 1);
+ drm_WARN_ON(obj->dev, refcount_read(&bo->base.pages_use_count) > 1);
drm_gem_shmem_free(&bo->base);
}
@@ -314,7 +326,7 @@ int ivpu_bo_create_ioctl(struct drm_device *dev, void *data, struct drm_file *fi
if (size == 0)
return -EINVAL;
- bo = ivpu_bo_alloc(vdev, size, args->flags);
+ bo = ivpu_bo_alloc(vdev, size, args->flags, file_priv->ctx.id);
if (IS_ERR(bo)) {
ivpu_err(vdev, "Failed to allocate BO: %pe (ctx %u size %llu flags 0x%x)",
bo, file_priv->ctx.id, args->size, args->flags);
@@ -322,7 +334,10 @@ int ivpu_bo_create_ioctl(struct drm_device *dev, void *data, struct drm_file *fi
}
ret = drm_gem_handle_create(file, &bo->base.base, &args->handle);
- if (!ret)
+ if (ret)
+ ivpu_err(vdev, "Failed to create handle for BO: %pe (ctx %u size %llu flags 0x%x)",
+ bo, file_priv->ctx.id, args->size, args->flags);
+ else
args->vpu_addr = bo->vpu_addr;
drm_gem_object_put(&bo->base.base);
@@ -345,7 +360,7 @@ ivpu_bo_create(struct ivpu_device *vdev, struct ivpu_mmu_context *ctx,
drm_WARN_ON(&vdev->drm, !PAGE_ALIGNED(range->end));
drm_WARN_ON(&vdev->drm, !PAGE_ALIGNED(size));
- bo = ivpu_bo_alloc(vdev, size, flags);
+ bo = ivpu_bo_alloc(vdev, size, flags, IVPU_GLOBAL_CONTEXT_MMU_SSID);
if (IS_ERR(bo)) {
ivpu_err(vdev, "Failed to allocate BO: %pe (vpu_addr 0x%llx size %llu flags 0x%x)",
bo, range->start, size, flags);
@@ -361,9 +376,9 @@ ivpu_bo_create(struct ivpu_device *vdev, struct ivpu_mmu_context *ctx,
goto err_put;
if (flags & DRM_IVPU_BO_MAPPABLE) {
- dma_resv_lock(bo->base.base.resv, NULL);
- ret = drm_gem_shmem_vmap(&bo->base, &map);
- dma_resv_unlock(bo->base.base.resv);
+ ivpu_bo_lock(bo);
+ ret = drm_gem_shmem_vmap_locked(&bo->base, &map);
+ ivpu_bo_unlock(bo);
if (ret)
goto err_put;
@@ -386,9 +401,9 @@ void ivpu_bo_free(struct ivpu_bo *bo)
struct iosys_map map = IOSYS_MAP_INIT_VADDR(bo->base.vaddr);
if (bo->flags & DRM_IVPU_BO_MAPPABLE) {
- dma_resv_lock(bo->base.base.resv, NULL);
- drm_gem_shmem_vunmap(&bo->base, &map);
- dma_resv_unlock(bo->base.base.resv);
+ ivpu_bo_lock(bo);
+ drm_gem_shmem_vunmap_locked(&bo->base, &map);
+ ivpu_bo_unlock(bo);
}
drm_gem_object_put(&bo->base.base);
@@ -407,12 +422,12 @@ int ivpu_bo_info_ioctl(struct drm_device *dev, void *data, struct drm_file *file
bo = to_ivpu_bo(obj);
- mutex_lock(&bo->lock);
+ ivpu_bo_lock(bo);
args->flags = bo->flags;
args->mmap_offset = drm_vma_node_offset_addr(&obj->vma_node);
args->vpu_addr = bo->vpu_addr;
args->size = obj->size;
- mutex_unlock(&bo->lock);
+ ivpu_bo_unlock(bo);
drm_gem_object_put(obj);
return ret;
@@ -449,10 +464,10 @@ int ivpu_bo_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *file
static void ivpu_bo_print_info(struct ivpu_bo *bo, struct drm_printer *p)
{
- mutex_lock(&bo->lock);
+ ivpu_bo_lock(bo);
drm_printf(p, "%-9p %-3u 0x%-12llx %-10lu 0x%-8x %-4u",
- bo, bo->ctx ? bo->ctx->id : 0, bo->vpu_addr, bo->base.base.size,
+ bo, bo->ctx_id, bo->vpu_addr, bo->base.base.size,
bo->flags, kref_read(&bo->base.base.refcount));
if (bo->base.pages)
@@ -461,12 +476,12 @@ static void ivpu_bo_print_info(struct ivpu_bo *bo, struct drm_printer *p)
if (bo->mmu_mapped)
drm_printf(p, " mmu_mapped");
- if (bo->base.base.import_attach)
+ if (drm_gem_is_imported(&bo->base.base))
drm_printf(p, " imported");
drm_printf(p, "\n");
- mutex_unlock(&bo->lock);
+ ivpu_bo_unlock(bo);
}
void ivpu_bo_list(struct drm_device *dev, struct drm_printer *p)
diff --git a/drivers/accel/ivpu/ivpu_gem.h b/drivers/accel/ivpu/ivpu_gem.h
index a222a9ec9d61..aa8ff14f7aae 100644
--- a/drivers/accel/ivpu/ivpu_gem.h
+++ b/drivers/accel/ivpu/ivpu_gem.h
@@ -17,10 +17,10 @@ struct ivpu_bo {
struct list_head bo_list_node;
struct drm_mm_node mm_node;
- struct mutex lock; /* Protects: ctx, mmu_mapped, vpu_addr */
u64 vpu_addr;
u32 flags;
u32 job_status; /* Valid only for command buffer */
+ u32 ctx_id;
bool mmu_mapped;
};
diff --git a/drivers/accel/ivpu/ivpu_hw.c b/drivers/accel/ivpu/ivpu_hw.c
index ec9a3629da3a..08dcc31b56f4 100644
--- a/drivers/accel/ivpu/ivpu_hw.c
+++ b/drivers/accel/ivpu/ivpu_hw.c
@@ -94,12 +94,14 @@ static void timeouts_init(struct ivpu_device *vdev)
vdev->timeout.boot = -1;
vdev->timeout.jsm = -1;
vdev->timeout.tdr = -1;
+ vdev->timeout.inference = -1;
vdev->timeout.autosuspend = -1;
vdev->timeout.d0i3_entry_msg = -1;
} else if (ivpu_is_fpga(vdev)) {
vdev->timeout.boot = 50;
vdev->timeout.jsm = 15000;
vdev->timeout.tdr = 30000;
+ vdev->timeout.inference = 900000;
vdev->timeout.autosuspend = -1;
vdev->timeout.d0i3_entry_msg = 500;
vdev->timeout.state_dump_msg = 10000;
@@ -107,6 +109,7 @@ static void timeouts_init(struct ivpu_device *vdev)
vdev->timeout.boot = 50;
vdev->timeout.jsm = 500;
vdev->timeout.tdr = 10000;
+ vdev->timeout.inference = 300000;
vdev->timeout.autosuspend = 100;
vdev->timeout.d0i3_entry_msg = 100;
vdev->timeout.state_dump_msg = 10;
@@ -114,12 +117,13 @@ static void timeouts_init(struct ivpu_device *vdev)
vdev->timeout.boot = 1000;
vdev->timeout.jsm = 500;
vdev->timeout.tdr = 2000;
+ vdev->timeout.inference = 60000;
if (ivpu_hw_ip_gen(vdev) == IVPU_HW_IP_37XX)
vdev->timeout.autosuspend = 10;
else
vdev->timeout.autosuspend = 100;
vdev->timeout.d0i3_entry_msg = 5;
- vdev->timeout.state_dump_msg = 10;
+ vdev->timeout.state_dump_msg = 100;
}
}
diff --git a/drivers/accel/ivpu/ivpu_hw.h b/drivers/accel/ivpu/ivpu_hw.h
index 16435f2756d0..d79668fe1609 100644
--- a/drivers/accel/ivpu/ivpu_hw.h
+++ b/drivers/accel/ivpu/ivpu_hw.h
@@ -1,6 +1,6 @@
/* SPDX-License-Identifier: GPL-2.0-only */
/*
- * Copyright (C) 2020-2024 Intel Corporation
+ * Copyright (C) 2020-2025 Intel Corporation
*/
#ifndef __IVPU_HW_H__
@@ -82,19 +82,19 @@ static inline u64 ivpu_hw_range_size(const struct ivpu_addr_range *range)
return range->end - range->start;
}
-static inline u32 ivpu_hw_ratio_to_freq(struct ivpu_device *vdev, u32 ratio)
+static inline u32 ivpu_hw_dpu_max_freq_get(struct ivpu_device *vdev)
{
- return ivpu_hw_btrs_ratio_to_freq(vdev, ratio);
+ return ivpu_hw_btrs_dpu_max_freq_get(vdev);
}
-static inline void ivpu_hw_irq_clear(struct ivpu_device *vdev)
+static inline u32 ivpu_hw_dpu_freq_get(struct ivpu_device *vdev)
{
- ivpu_hw_ip_irq_clear(vdev);
+ return ivpu_hw_btrs_dpu_freq_get(vdev);
}
-static inline u32 ivpu_hw_pll_freq_get(struct ivpu_device *vdev)
+static inline void ivpu_hw_irq_clear(struct ivpu_device *vdev)
{
- return ivpu_hw_btrs_pll_freq_get(vdev);
+ ivpu_hw_ip_irq_clear(vdev);
}
static inline u32 ivpu_hw_profiling_freq_get(struct ivpu_device *vdev)
diff --git a/drivers/accel/ivpu/ivpu_hw_btrs.c b/drivers/accel/ivpu/ivpu_hw_btrs.c
index 56c56012b980..b236c7234daa 100644
--- a/drivers/accel/ivpu/ivpu_hw_btrs.c
+++ b/drivers/accel/ivpu/ivpu_hw_btrs.c
@@ -1,8 +1,10 @@
// SPDX-License-Identifier: GPL-2.0-only
/*
- * Copyright (C) 2020-2024 Intel Corporation
+ * Copyright (C) 2020-2025 Intel Corporation
*/
+#include <linux/units.h>
+
#include "ivpu_drv.h"
#include "ivpu_hw.h"
#include "ivpu_hw_btrs.h"
@@ -28,17 +30,13 @@
#define BTRS_LNL_ALL_IRQ_MASK ((u32)-1)
-#define BTRS_MTL_WP_CONFIG_1_TILE_5_3_RATIO WP_CONFIG(MTL_CONFIG_1_TILE, MTL_PLL_RATIO_5_3)
-#define BTRS_MTL_WP_CONFIG_1_TILE_4_3_RATIO WP_CONFIG(MTL_CONFIG_1_TILE, MTL_PLL_RATIO_4_3)
-#define BTRS_MTL_WP_CONFIG_2_TILE_5_3_RATIO WP_CONFIG(MTL_CONFIG_2_TILE, MTL_PLL_RATIO_5_3)
-#define BTRS_MTL_WP_CONFIG_2_TILE_4_3_RATIO WP_CONFIG(MTL_CONFIG_2_TILE, MTL_PLL_RATIO_4_3)
-#define BTRS_MTL_WP_CONFIG_0_TILE_PLL_OFF WP_CONFIG(0, 0)
#define PLL_CDYN_DEFAULT 0x80
#define PLL_EPP_DEFAULT 0x80
#define PLL_CONFIG_DEFAULT 0x0
-#define PLL_SIMULATION_FREQ 10000000
-#define PLL_REF_CLK_FREQ 50000000
+#define PLL_REF_CLK_FREQ 50000000ull
+#define PLL_RATIO_TO_FREQ(x) ((x) * PLL_REF_CLK_FREQ)
+
#define PLL_TIMEOUT_US (1500 * USEC_PER_MSEC)
#define IDLE_TIMEOUT_US (5 * USEC_PER_MSEC)
#define TIMEOUT_US (150 * USEC_PER_MSEC)
@@ -62,6 +60,8 @@
#define DCT_ENABLE 0x1
#define DCT_DISABLE 0x0
+static u32 pll_ratio_to_dpu_freq(struct ivpu_device *vdev, u32 ratio);
+
int ivpu_hw_btrs_irqs_clear_with_0_mtl(struct ivpu_device *vdev)
{
REGB_WR32(VPU_HW_BTRS_MTL_INTERRUPT_STAT, BTRS_MTL_ALL_IRQ_MASK);
@@ -156,7 +156,7 @@ static int info_init_mtl(struct ivpu_device *vdev)
hw->tile_fuse = BTRS_MTL_TILE_FUSE_ENABLE_BOTH;
hw->sku = BTRS_MTL_TILE_SKU_BOTH;
- hw->config = BTRS_MTL_WP_CONFIG_2_TILE_4_3_RATIO;
+ hw->config = WP_CONFIG(MTL_CONFIG_2_TILE, MTL_PLL_RATIO_4_3);
return 0;
}
@@ -334,8 +334,8 @@ int ivpu_hw_btrs_wp_drive(struct ivpu_device *vdev, bool enable)
prepare_wp_request(vdev, &wp, enable);
- ivpu_dbg(vdev, PM, "PLL workpoint request: %u Hz, config: 0x%x, epp: 0x%x, cdyn: 0x%x\n",
- PLL_RATIO_TO_FREQ(wp.target), wp.cfg, wp.epp, wp.cdyn);
+ ivpu_dbg(vdev, PM, "PLL workpoint request: %lu MHz, config: 0x%x, epp: 0x%x, cdyn: 0x%x\n",
+ pll_ratio_to_dpu_freq(vdev, wp.target) / HZ_PER_MHZ, wp.cfg, wp.epp, wp.cdyn);
ret = wp_request_send(vdev, &wp);
if (ret) {
@@ -573,6 +573,47 @@ int ivpu_hw_btrs_wait_for_idle(struct ivpu_device *vdev)
return REGB_POLL_FLD(VPU_HW_BTRS_LNL_VPU_STATUS, IDLE, 0x1, IDLE_TIMEOUT_US);
}
+static u32 pll_config_get_mtl(struct ivpu_device *vdev)
+{
+ return REGB_RD32(VPU_HW_BTRS_MTL_CURRENT_PLL);
+}
+
+static u32 pll_config_get_lnl(struct ivpu_device *vdev)
+{
+ return REGB_RD32(VPU_HW_BTRS_LNL_PLL_FREQ);
+}
+
+static u32 pll_ratio_to_dpu_freq_mtl(u16 ratio)
+{
+ return (PLL_RATIO_TO_FREQ(ratio) * 2) / 3;
+}
+
+static u32 pll_ratio_to_dpu_freq_lnl(u16 ratio)
+{
+ return PLL_RATIO_TO_FREQ(ratio) / 2;
+}
+
+static u32 pll_ratio_to_dpu_freq(struct ivpu_device *vdev, u32 ratio)
+{
+ if (ivpu_hw_btrs_gen(vdev) == IVPU_HW_BTRS_MTL)
+ return pll_ratio_to_dpu_freq_mtl(ratio);
+ else
+ return pll_ratio_to_dpu_freq_lnl(ratio);
+}
+
+u32 ivpu_hw_btrs_dpu_max_freq_get(struct ivpu_device *vdev)
+{
+ return pll_ratio_to_dpu_freq(vdev, vdev->hw->pll.max_ratio);
+}
+
+u32 ivpu_hw_btrs_dpu_freq_get(struct ivpu_device *vdev)
+{
+ if (ivpu_hw_btrs_gen(vdev) == IVPU_HW_BTRS_MTL)
+ return pll_ratio_to_dpu_freq_mtl(pll_config_get_mtl(vdev));
+ else
+ return pll_ratio_to_dpu_freq_lnl(pll_config_get_lnl(vdev));
+}
+
/* Handler for IRQs from Buttress core (irqB) */
bool ivpu_hw_btrs_irq_handler_mtl(struct ivpu_device *vdev, int irq)
{
@@ -582,9 +623,12 @@ bool ivpu_hw_btrs_irq_handler_mtl(struct ivpu_device *vdev, int irq)
if (!status)
return false;
- if (REG_TEST_FLD(VPU_HW_BTRS_MTL_INTERRUPT_STAT, FREQ_CHANGE, status))
- ivpu_dbg(vdev, IRQ, "FREQ_CHANGE irq: %08x",
- REGB_RD32(VPU_HW_BTRS_MTL_CURRENT_PLL));
+ if (REG_TEST_FLD(VPU_HW_BTRS_MTL_INTERRUPT_STAT, FREQ_CHANGE, status)) {
+ u32 pll = pll_config_get_mtl(vdev);
+
+ ivpu_dbg(vdev, IRQ, "FREQ_CHANGE irq, wp %08x, %lu MHz",
+ pll, pll_ratio_to_dpu_freq_mtl(pll) / HZ_PER_MHZ);
+ }
if (REG_TEST_FLD(VPU_HW_BTRS_MTL_INTERRUPT_STAT, ATS_ERR, status)) {
ivpu_err(vdev, "ATS_ERR irq 0x%016llx", REGB_RD64(VPU_HW_BTRS_MTL_ATS_ERR_LOG_0));
@@ -633,8 +677,12 @@ bool ivpu_hw_btrs_irq_handler_lnl(struct ivpu_device *vdev, int irq)
queue_work(system_wq, &vdev->irq_dct_work);
}
- if (REG_TEST_FLD(VPU_HW_BTRS_LNL_INTERRUPT_STAT, FREQ_CHANGE, status))
- ivpu_dbg(vdev, IRQ, "FREQ_CHANGE irq: %08x", REGB_RD32(VPU_HW_BTRS_LNL_PLL_FREQ));
+ if (REG_TEST_FLD(VPU_HW_BTRS_LNL_INTERRUPT_STAT, FREQ_CHANGE, status)) {
+ u32 pll = pll_config_get_lnl(vdev);
+
+ ivpu_dbg(vdev, IRQ, "FREQ_CHANGE irq, wp %08x, %lu MHz",
+ pll, pll_ratio_to_dpu_freq_lnl(pll) / HZ_PER_MHZ);
+ }
if (REG_TEST_FLD(VPU_HW_BTRS_LNL_INTERRUPT_STAT, ATS_ERR, status)) {
ivpu_err(vdev, "ATS_ERR LOG1 0x%08x ATS_ERR_LOG2 0x%08x\n",
@@ -717,60 +765,6 @@ void ivpu_hw_btrs_dct_set_status(struct ivpu_device *vdev, bool enable, u32 acti
REGB_WR32(VPU_HW_BTRS_LNL_PCODE_MAILBOX_STATUS, val);
}
-static u32 pll_ratio_to_freq_mtl(u32 ratio, u32 config)
-{
- u32 pll_clock = PLL_REF_CLK_FREQ * ratio;
- u32 cpu_clock;
-
- if ((config & 0xff) == MTL_PLL_RATIO_4_3)
- cpu_clock = pll_clock * 2 / 4;
- else
- cpu_clock = pll_clock * 2 / 5;
-
- return cpu_clock;
-}
-
-u32 ivpu_hw_btrs_ratio_to_freq(struct ivpu_device *vdev, u32 ratio)
-{
- struct ivpu_hw_info *hw = vdev->hw;
-
- if (ivpu_hw_btrs_gen(vdev) == IVPU_HW_BTRS_MTL)
- return pll_ratio_to_freq_mtl(ratio, hw->config);
- else
- return PLL_RATIO_TO_FREQ(ratio);
-}
-
-static u32 pll_freq_get_mtl(struct ivpu_device *vdev)
-{
- u32 pll_curr_ratio;
-
- pll_curr_ratio = REGB_RD32(VPU_HW_BTRS_MTL_CURRENT_PLL);
- pll_curr_ratio &= VPU_HW_BTRS_MTL_CURRENT_PLL_RATIO_MASK;
-
- if (!ivpu_is_silicon(vdev))
- return PLL_SIMULATION_FREQ;
-
- return pll_ratio_to_freq_mtl(pll_curr_ratio, vdev->hw->config);
-}
-
-static u32 pll_freq_get_lnl(struct ivpu_device *vdev)
-{
- u32 pll_curr_ratio;
-
- pll_curr_ratio = REGB_RD32(VPU_HW_BTRS_LNL_PLL_FREQ);
- pll_curr_ratio &= VPU_HW_BTRS_LNL_PLL_FREQ_RATIO_MASK;
-
- return PLL_RATIO_TO_FREQ(pll_curr_ratio);
-}
-
-u32 ivpu_hw_btrs_pll_freq_get(struct ivpu_device *vdev)
-{
- if (ivpu_hw_btrs_gen(vdev) == IVPU_HW_BTRS_MTL)
- return pll_freq_get_mtl(vdev);
- else
- return pll_freq_get_lnl(vdev);
-}
-
u32 ivpu_hw_btrs_telemetry_offset_get(struct ivpu_device *vdev)
{
if (ivpu_hw_btrs_gen(vdev) == IVPU_HW_BTRS_MTL)
diff --git a/drivers/accel/ivpu/ivpu_hw_btrs.h b/drivers/accel/ivpu/ivpu_hw_btrs.h
index 1fd71b4d4ab0..d2d82651976d 100644
--- a/drivers/accel/ivpu/ivpu_hw_btrs.h
+++ b/drivers/accel/ivpu/ivpu_hw_btrs.h
@@ -1,6 +1,6 @@
/* SPDX-License-Identifier: GPL-2.0-only */
/*
- * Copyright (C) 2020-2024 Intel Corporation
+ * Copyright (C) 2020-2025 Intel Corporation
*/
#ifndef __IVPU_HW_BTRS_H__
@@ -13,9 +13,8 @@
#define PLL_PROFILING_FREQ_DEFAULT 38400000
#define PLL_PROFILING_FREQ_HIGH 400000000
-#define PLL_RATIO_TO_FREQ(x) ((x) * PLL_REF_CLK_FREQ)
-#define DCT_DEFAULT_ACTIVE_PERCENT 15u
+#define DCT_DEFAULT_ACTIVE_PERCENT 30u
#define DCT_PERIOD_US 35300u
int ivpu_hw_btrs_info_init(struct ivpu_device *vdev);
@@ -32,12 +31,12 @@ int ivpu_hw_btrs_ip_reset(struct ivpu_device *vdev);
void ivpu_hw_btrs_profiling_freq_reg_set_lnl(struct ivpu_device *vdev);
void ivpu_hw_btrs_ats_print_lnl(struct ivpu_device *vdev);
void ivpu_hw_btrs_clock_relinquish_disable_lnl(struct ivpu_device *vdev);
+u32 ivpu_hw_btrs_dpu_max_freq_get(struct ivpu_device *vdev);
+u32 ivpu_hw_btrs_dpu_freq_get(struct ivpu_device *vdev);
bool ivpu_hw_btrs_irq_handler_mtl(struct ivpu_device *vdev, int irq);
bool ivpu_hw_btrs_irq_handler_lnl(struct ivpu_device *vdev, int irq);
int ivpu_hw_btrs_dct_get_request(struct ivpu_device *vdev, bool *enable);
void ivpu_hw_btrs_dct_set_status(struct ivpu_device *vdev, bool enable, u32 dct_percent);
-u32 ivpu_hw_btrs_pll_freq_get(struct ivpu_device *vdev);
-u32 ivpu_hw_btrs_ratio_to_freq(struct ivpu_device *vdev, u32 ratio);
u32 ivpu_hw_btrs_telemetry_offset_get(struct ivpu_device *vdev);
u32 ivpu_hw_btrs_telemetry_size_get(struct ivpu_device *vdev);
u32 ivpu_hw_btrs_telemetry_enable_get(struct ivpu_device *vdev);
diff --git a/drivers/accel/ivpu/ivpu_hw_ip.c b/drivers/accel/ivpu/ivpu_hw_ip.c
index 823f6a57dc54..2bf9882ab52e 100644
--- a/drivers/accel/ivpu/ivpu_hw_ip.c
+++ b/drivers/accel/ivpu/ivpu_hw_ip.c
@@ -683,6 +683,7 @@ static void pwr_island_delay_set(struct ivpu_device *vdev)
return;
switch (ivpu_device_id(vdev)) {
+ case PCI_DEVICE_ID_WCL:
case PCI_DEVICE_ID_PTL_P:
post = high ? 18 : 0;
post1 = 0;
diff --git a/drivers/accel/ivpu/ivpu_ipc.c b/drivers/accel/ivpu/ivpu_ipc.c
index 0e096fd9b95d..39f83225c181 100644
--- a/drivers/accel/ivpu/ivpu_ipc.c
+++ b/drivers/accel/ivpu/ivpu_ipc.c
@@ -302,7 +302,8 @@ ivpu_ipc_send_receive_internal(struct ivpu_device *vdev, struct vpu_jsm_msg *req
struct ivpu_ipc_consumer cons;
int ret;
- drm_WARN_ON(&vdev->drm, pm_runtime_status_suspended(vdev->drm.dev));
+ drm_WARN_ON(&vdev->drm, pm_runtime_status_suspended(vdev->drm.dev) &&
+ pm_runtime_enabled(vdev->drm.dev));
ivpu_ipc_consumer_add(vdev, &cons, channel, NULL);
diff --git a/drivers/accel/ivpu/ivpu_job.c b/drivers/accel/ivpu/ivpu_job.c
index 004059e4f1e8..060f1fc031d3 100644
--- a/drivers/accel/ivpu/ivpu_job.c
+++ b/drivers/accel/ivpu/ivpu_job.c
@@ -1,6 +1,6 @@
// SPDX-License-Identifier: GPL-2.0-only
/*
- * Copyright (C) 2020-2024 Intel Corporation
+ * Copyright (C) 2020-2025 Intel Corporation
*/
#include <drm/drm_file.h>
@@ -100,6 +100,43 @@ err_free_cmdq:
return NULL;
}
+/**
+ * ivpu_cmdq_get_entry_count - Calculate the number of entries in the command queue.
+ * @cmdq: Pointer to the command queue structure.
+ *
+ * Returns the number of entries that can fit in the command queue memory.
+ */
+static inline u32 ivpu_cmdq_get_entry_count(struct ivpu_cmdq *cmdq)
+{
+ size_t size = ivpu_bo_size(cmdq->mem) - sizeof(struct vpu_job_queue_header);
+
+ return size / sizeof(struct vpu_job_queue_entry);
+}
+
+/**
+ * ivpu_cmdq_get_flags - Get command queue flags based on input flags and test mode.
+ * @vdev: Pointer to the ivpu device structure.
+ * @flags: Input flags to determine the command queue flags.
+ *
+ * Returns the calculated command queue flags, considering both the input flags
+ * and the current test mode settings.
+ */
+static u32 ivpu_cmdq_get_flags(struct ivpu_device *vdev, u32 flags)
+{
+ u32 cmdq_flags = 0;
+
+ if ((flags & DRM_IVPU_CMDQ_FLAG_TURBO) && (ivpu_hw_ip_gen(vdev) >= IVPU_HW_IP_40XX))
+ cmdq_flags |= VPU_JOB_QUEUE_FLAGS_TURBO_MODE;
+
+ /* Test mode can override the TURBO flag coming from the application */
+ if (ivpu_test_mode & IVPU_TEST_MODE_TURBO_ENABLE)
+ cmdq_flags |= VPU_JOB_QUEUE_FLAGS_TURBO_MODE;
+ if (ivpu_test_mode & IVPU_TEST_MODE_TURBO_DISABLE)
+ cmdq_flags &= ~VPU_JOB_QUEUE_FLAGS_TURBO_MODE;
+
+ return cmdq_flags;
+}
+
static void ivpu_cmdq_free(struct ivpu_file_priv *file_priv, struct ivpu_cmdq *cmdq)
{
ivpu_preemption_buffers_free(file_priv->vdev, file_priv, cmdq);
@@ -107,8 +144,7 @@ static void ivpu_cmdq_free(struct ivpu_file_priv *file_priv, struct ivpu_cmdq *c
kfree(cmdq);
}
-static struct ivpu_cmdq *ivpu_cmdq_create(struct ivpu_file_priv *file_priv, u8 priority,
- bool is_legacy)
+static struct ivpu_cmdq *ivpu_cmdq_create(struct ivpu_file_priv *file_priv, u8 priority, u32 flags)
{
struct ivpu_device *vdev = file_priv->vdev;
struct ivpu_cmdq *cmdq = NULL;
@@ -121,10 +157,6 @@ static struct ivpu_cmdq *ivpu_cmdq_create(struct ivpu_file_priv *file_priv, u8 p
ivpu_err(vdev, "Failed to allocate command queue\n");
return NULL;
}
-
- cmdq->priority = priority;
- cmdq->is_legacy = is_legacy;
-
ret = xa_alloc_cyclic(&file_priv->cmdq_xa, &cmdq->id, cmdq, file_priv->cmdq_limit,
&file_priv->cmdq_id_next, GFP_KERNEL);
if (ret < 0) {
@@ -132,7 +164,15 @@ static struct ivpu_cmdq *ivpu_cmdq_create(struct ivpu_file_priv *file_priv, u8 p
goto err_free_cmdq;
}
- ivpu_dbg(vdev, JOB, "Command queue %d created, ctx %d\n", cmdq->id, file_priv->ctx.id);
+ cmdq->entry_count = ivpu_cmdq_get_entry_count(cmdq);
+ cmdq->priority = priority;
+
+ cmdq->jobq = (struct vpu_job_queue *)ivpu_bo_vaddr(cmdq->mem);
+ cmdq->jobq->header.engine_idx = VPU_ENGINE_COMPUTE;
+ cmdq->jobq->header.flags = ivpu_cmdq_get_flags(vdev, flags);
+
+ ivpu_dbg(vdev, JOB, "Command queue %d created, ctx %d, flags 0x%08x\n",
+ cmdq->id, file_priv->ctx.id, cmdq->jobq->header.flags);
return cmdq;
err_free_cmdq:
@@ -188,27 +228,14 @@ static int ivpu_register_db(struct ivpu_file_priv *file_priv, struct ivpu_cmdq *
return ret;
}
-static void ivpu_cmdq_jobq_init(struct ivpu_device *vdev, struct vpu_job_queue *jobq)
+static void ivpu_cmdq_jobq_reset(struct ivpu_device *vdev, struct vpu_job_queue *jobq)
{
- jobq->header.engine_idx = VPU_ENGINE_COMPUTE;
jobq->header.head = 0;
jobq->header.tail = 0;
- if (ivpu_test_mode & IVPU_TEST_MODE_TURBO) {
- ivpu_dbg(vdev, JOB, "Turbo mode enabled");
- jobq->header.flags = VPU_JOB_QUEUE_FLAGS_TURBO_MODE;
- }
-
wmb(); /* Flush WC buffer for jobq->header */
}
-static inline u32 ivpu_cmdq_get_entry_count(struct ivpu_cmdq *cmdq)
-{
- size_t size = ivpu_bo_size(cmdq->mem) - sizeof(struct vpu_job_queue_header);
-
- return size / sizeof(struct vpu_job_queue_entry);
-}
-
static int ivpu_cmdq_register(struct ivpu_file_priv *file_priv, struct ivpu_cmdq *cmdq)
{
struct ivpu_device *vdev = file_priv->vdev;
@@ -219,10 +246,7 @@ static int ivpu_cmdq_register(struct ivpu_file_priv *file_priv, struct ivpu_cmdq
if (cmdq->db_id)
return 0;
- cmdq->entry_count = ivpu_cmdq_get_entry_count(cmdq);
- cmdq->jobq = (struct vpu_job_queue *)ivpu_bo_vaddr(cmdq->mem);
-
- ivpu_cmdq_jobq_init(vdev, cmdq->jobq);
+ ivpu_cmdq_jobq_reset(vdev, cmdq->jobq);
if (vdev->fw->sched_mode == VPU_SCHEDULING_MODE_HW) {
ret = ivpu_hws_cmdq_init(file_priv, cmdq, VPU_ENGINE_COMPUTE, cmdq->priority);
@@ -247,6 +271,10 @@ static int ivpu_cmdq_unregister(struct ivpu_file_priv *file_priv, struct ivpu_cm
if (!cmdq->db_id)
return 0;
+ ret = ivpu_jsm_unregister_db(vdev, cmdq->db_id);
+ if (!ret)
+ ivpu_dbg(vdev, JOB, "DB %d unregistered\n", cmdq->db_id);
+
if (vdev->fw->sched_mode == VPU_SCHEDULING_MODE_HW) {
ret = ivpu_jsm_hws_destroy_cmdq(vdev, file_priv->ctx.id, cmdq->id);
if (!ret)
@@ -254,10 +282,6 @@ static int ivpu_cmdq_unregister(struct ivpu_file_priv *file_priv, struct ivpu_cm
cmdq->id, file_priv->ctx.id);
}
- ret = ivpu_jsm_unregister_db(vdev, cmdq->db_id);
- if (!ret)
- ivpu_dbg(vdev, JOB, "DB %d unregistered\n", cmdq->db_id);
-
xa_erase(&file_priv->vdev->db_xa, cmdq->db_id);
cmdq->db_id = 0;
@@ -291,9 +315,10 @@ static struct ivpu_cmdq *ivpu_cmdq_acquire_legacy(struct ivpu_file_priv *file_pr
break;
if (!cmdq) {
- cmdq = ivpu_cmdq_create(file_priv, priority, true);
+ cmdq = ivpu_cmdq_create(file_priv, priority, 0);
if (!cmdq)
return NULL;
+ cmdq->is_legacy = true;
}
return cmdq;
@@ -470,8 +495,8 @@ static void ivpu_job_destroy(struct ivpu_job *job)
struct ivpu_device *vdev = job->vdev;
u32 i;
- ivpu_dbg(vdev, JOB, "Job destroyed: id %3u ctx %2d engine %d",
- job->job_id, job->file_priv->ctx.id, job->engine_idx);
+ ivpu_dbg(vdev, JOB, "Job destroyed: id %3u ctx %2d cmdq_id %u engine %d",
+ job->job_id, job->file_priv->ctx.id, job->cmdq_id, job->engine_idx);
for (i = 0; i < job->bo_count; i++)
if (job->bos[i])
@@ -564,8 +589,8 @@ static int ivpu_job_signal_and_destroy(struct ivpu_device *vdev, u32 job_id, u32
dma_fence_signal(job->done_fence);
trace_job("done", job);
- ivpu_dbg(vdev, JOB, "Job complete: id %3u ctx %2d engine %d status 0x%x\n",
- job->job_id, job->file_priv->ctx.id, job->engine_idx, job_status);
+ ivpu_dbg(vdev, JOB, "Job complete: id %3u ctx %2d cmdq_id %u engine %d status 0x%x\n",
+ job->job_id, job->file_priv->ctx.id, job->cmdq_id, job->engine_idx, job_status);
ivpu_job_destroy(job);
ivpu_stop_job_timeout_detection(vdev);
@@ -664,8 +689,8 @@ static int ivpu_job_submit(struct ivpu_job *job, u8 priority, u32 cmdq_id)
}
trace_job("submit", job);
- ivpu_dbg(vdev, JOB, "Job submitted: id %3u ctx %2d engine %d prio %d addr 0x%llx next %d\n",
- job->job_id, file_priv->ctx.id, job->engine_idx, cmdq->priority,
+ ivpu_dbg(vdev, JOB, "Job submitted: id %3u ctx %2d cmdq_id %u engine %d prio %d addr 0x%llx next %d\n",
+ job->job_id, file_priv->ctx.id, cmdq->id, job->engine_idx, cmdq->priority,
job->cmd_buf_vpu_addr, cmdq->jobq->header.tail);
mutex_unlock(&file_priv->lock);
@@ -681,8 +706,8 @@ static int ivpu_job_submit(struct ivpu_job *job, u8 priority, u32 cmdq_id)
err_erase_xa:
xa_erase(&vdev->submitted_jobs_xa, job->job_id);
err_unlock:
- mutex_unlock(&vdev->submitted_jobs_lock);
mutex_unlock(&file_priv->lock);
+ mutex_unlock(&vdev->submitted_jobs_lock);
ivpu_rpm_put(vdev);
return ret;
}
@@ -777,7 +802,8 @@ static int ivpu_submit(struct drm_file *file, struct ivpu_file_priv *file_priv,
goto err_free_handles;
}
- ivpu_dbg(vdev, JOB, "Submit ioctl: ctx %u buf_count %u\n", file_priv->ctx.id, buffer_count);
+ ivpu_dbg(vdev, JOB, "Submit ioctl: ctx %u cmdq_id %u buf_count %u\n",
+ file_priv->ctx.id, cmdq_id, buffer_count);
job = ivpu_job_create(file_priv, engine, buffer_count);
if (!job) {
@@ -873,23 +899,31 @@ int ivpu_cmdq_submit_ioctl(struct drm_device *dev, void *data, struct drm_file *
int ivpu_cmdq_create_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
{
struct ivpu_file_priv *file_priv = file->driver_priv;
+ struct ivpu_device *vdev = file_priv->vdev;
struct drm_ivpu_cmdq_create *args = data;
struct ivpu_cmdq *cmdq;
+ int ret;
- if (!ivpu_is_capable(file_priv->vdev, DRM_IVPU_CAP_MANAGE_CMDQ))
+ if (!ivpu_is_capable(vdev, DRM_IVPU_CAP_MANAGE_CMDQ))
return -ENODEV;
if (args->priority > DRM_IVPU_JOB_PRIORITY_REALTIME)
return -EINVAL;
+ ret = ivpu_rpm_get(vdev);
+ if (ret < 0)
+ return ret;
+
mutex_lock(&file_priv->lock);
- cmdq = ivpu_cmdq_create(file_priv, ivpu_job_to_jsm_priority(args->priority), false);
+ cmdq = ivpu_cmdq_create(file_priv, ivpu_job_to_jsm_priority(args->priority), args->flags);
if (cmdq)
args->cmdq_id = cmdq->id;
mutex_unlock(&file_priv->lock);
+ ivpu_rpm_put(vdev);
+
return cmdq ? 0 : -ENOMEM;
}
@@ -899,28 +933,35 @@ int ivpu_cmdq_destroy_ioctl(struct drm_device *dev, void *data, struct drm_file
struct ivpu_device *vdev = file_priv->vdev;
struct drm_ivpu_cmdq_destroy *args = data;
struct ivpu_cmdq *cmdq;
- u32 cmdq_id;
+ u32 cmdq_id = 0;
int ret;
if (!ivpu_is_capable(vdev, DRM_IVPU_CAP_MANAGE_CMDQ))
return -ENODEV;
+ ret = ivpu_rpm_get(vdev);
+ if (ret < 0)
+ return ret;
+
mutex_lock(&file_priv->lock);
cmdq = xa_load(&file_priv->cmdq_xa, args->cmdq_id);
if (!cmdq || cmdq->is_legacy) {
ret = -ENOENT;
- goto err_unlock;
+ } else {
+ cmdq_id = cmdq->id;
+ ivpu_cmdq_destroy(file_priv, cmdq);
+ ret = 0;
}
- cmdq_id = cmdq->id;
- ivpu_cmdq_destroy(file_priv, cmdq);
mutex_unlock(&file_priv->lock);
- ivpu_cmdq_abort_all_jobs(vdev, file_priv->ctx.id, cmdq_id);
- return 0;
-err_unlock:
- mutex_unlock(&file_priv->lock);
+ /* Abort any pending jobs only if cmdq was destroyed */
+ if (!ret)
+ ivpu_cmdq_abort_all_jobs(vdev, file_priv->ctx.id, cmdq_id);
+
+ ivpu_rpm_put(vdev);
+
return ret;
}
@@ -970,7 +1011,8 @@ void ivpu_context_abort_work_fn(struct work_struct *work)
return;
if (vdev->fw->sched_mode == VPU_SCHEDULING_MODE_HW)
- ivpu_jsm_reset_engine(vdev, 0);
+ if (ivpu_jsm_reset_engine(vdev, 0))
+ return;
mutex_lock(&vdev->context_list_lock);
xa_for_each(&vdev->context_xa, ctx_id, file_priv) {
@@ -993,7 +1035,8 @@ void ivpu_context_abort_work_fn(struct work_struct *work)
if (vdev->fw->sched_mode != VPU_SCHEDULING_MODE_HW)
goto runtime_put;
- ivpu_jsm_hws_resume_engine(vdev, 0);
+ if (ivpu_jsm_hws_resume_engine(vdev, 0))
+ return;
/*
* In hardware scheduling mode NPU already has stopped processing jobs
* and won't send us any further notifications, thus we have to free job related resources
diff --git a/drivers/accel/ivpu/ivpu_jsm_msg.c b/drivers/accel/ivpu/ivpu_jsm_msg.c
index 219ab8afefab..0256b2dfefc1 100644
--- a/drivers/accel/ivpu/ivpu_jsm_msg.c
+++ b/drivers/accel/ivpu/ivpu_jsm_msg.c
@@ -7,6 +7,7 @@
#include "ivpu_hw.h"
#include "ivpu_ipc.h"
#include "ivpu_jsm_msg.h"
+#include "ivpu_pm.h"
#include "vpu_jsm_api.h"
const char *ivpu_jsm_msg_type_to_str(enum vpu_ipc_msg_type type)
@@ -163,8 +164,10 @@ int ivpu_jsm_reset_engine(struct ivpu_device *vdev, u32 engine)
ret = ivpu_ipc_send_receive(vdev, &req, VPU_JSM_MSG_ENGINE_RESET_DONE, &resp,
VPU_IPC_CHAN_ASYNC_CMD, vdev->timeout.jsm);
- if (ret)
+ if (ret) {
ivpu_err_ratelimited(vdev, "Failed to reset engine %d: %d\n", engine, ret);
+ ivpu_pm_trigger_recovery(vdev, "Engine reset failed");
+ }
return ret;
}
@@ -354,8 +357,10 @@ int ivpu_jsm_hws_resume_engine(struct ivpu_device *vdev, u32 engine)
ret = ivpu_ipc_send_receive(vdev, &req, VPU_JSM_MSG_HWS_RESUME_ENGINE_DONE, &resp,
VPU_IPC_CHAN_ASYNC_CMD, vdev->timeout.jsm);
- if (ret)
+ if (ret) {
ivpu_err_ratelimited(vdev, "Failed to resume engine %d: %d\n", engine, ret);
+ ivpu_pm_trigger_recovery(vdev, "Engine resume failed");
+ }
return ret;
}
diff --git a/drivers/accel/ivpu/ivpu_ms.c b/drivers/accel/ivpu/ivpu_ms.c
index ffe7b10f8a76..2a043baf10ca 100644
--- a/drivers/accel/ivpu/ivpu_ms.c
+++ b/drivers/accel/ivpu/ivpu_ms.c
@@ -4,6 +4,7 @@
*/
#include <drm/drm_file.h>
+#include <linux/pm_runtime.h>
#include "ivpu_drv.h"
#include "ivpu_gem.h"
@@ -44,6 +45,10 @@ int ivpu_ms_start_ioctl(struct drm_device *dev, void *data, struct drm_file *fil
args->sampling_period_ns < MS_MIN_SAMPLE_PERIOD_NS)
return -EINVAL;
+ ret = ivpu_rpm_get(vdev);
+ if (ret < 0)
+ return ret;
+
mutex_lock(&file_priv->ms_lock);
if (get_instance_by_mask(file_priv, args->metric_group_mask)) {
@@ -96,6 +101,8 @@ err_free_ms:
kfree(ms);
unlock:
mutex_unlock(&file_priv->ms_lock);
+
+ ivpu_rpm_put(vdev);
return ret;
}
@@ -160,6 +167,10 @@ int ivpu_ms_get_data_ioctl(struct drm_device *dev, void *data, struct drm_file *
if (!args->metric_group_mask)
return -EINVAL;
+ ret = ivpu_rpm_get(vdev);
+ if (ret < 0)
+ return ret;
+
mutex_lock(&file_priv->ms_lock);
ms = get_instance_by_mask(file_priv, args->metric_group_mask);
@@ -187,6 +198,7 @@ int ivpu_ms_get_data_ioctl(struct drm_device *dev, void *data, struct drm_file *
unlock:
mutex_unlock(&file_priv->ms_lock);
+ ivpu_rpm_put(vdev);
return ret;
}
@@ -204,11 +216,17 @@ int ivpu_ms_stop_ioctl(struct drm_device *dev, void *data, struct drm_file *file
{
struct ivpu_file_priv *file_priv = file->driver_priv;
struct drm_ivpu_metric_streamer_stop *args = data;
+ struct ivpu_device *vdev = file_priv->vdev;
struct ivpu_ms_instance *ms;
+ int ret;
if (!args->metric_group_mask)
return -EINVAL;
+ ret = ivpu_rpm_get(vdev);
+ if (ret < 0)
+ return ret;
+
mutex_lock(&file_priv->ms_lock);
ms = get_instance_by_mask(file_priv, args->metric_group_mask);
@@ -217,6 +235,7 @@ int ivpu_ms_stop_ioctl(struct drm_device *dev, void *data, struct drm_file *file
mutex_unlock(&file_priv->ms_lock);
+ ivpu_rpm_put(vdev);
return ms ? 0 : -EINVAL;
}
@@ -281,6 +300,9 @@ unlock:
void ivpu_ms_cleanup(struct ivpu_file_priv *file_priv)
{
struct ivpu_ms_instance *ms, *tmp;
+ struct ivpu_device *vdev = file_priv->vdev;
+
+ pm_runtime_get_sync(vdev->drm.dev);
mutex_lock(&file_priv->ms_lock);
@@ -293,6 +315,8 @@ void ivpu_ms_cleanup(struct ivpu_file_priv *file_priv)
free_instance(file_priv, ms);
mutex_unlock(&file_priv->ms_lock);
+
+ pm_runtime_put_autosuspend(vdev->drm.dev);
}
void ivpu_ms_cleanup_all(struct ivpu_device *vdev)
diff --git a/drivers/accel/ivpu/ivpu_pm.c b/drivers/accel/ivpu/ivpu_pm.c
index b5891e91f7ab..eacda1dbe840 100644
--- a/drivers/accel/ivpu/ivpu_pm.c
+++ b/drivers/accel/ivpu/ivpu_pm.c
@@ -33,6 +33,10 @@ static unsigned long ivpu_tdr_timeout_ms;
module_param_named(tdr_timeout_ms, ivpu_tdr_timeout_ms, ulong, 0644);
MODULE_PARM_DESC(tdr_timeout_ms, "Timeout for device hang detection, in milliseconds, 0 - default");
+static unsigned long ivpu_inference_timeout_ms;
+module_param_named(inference_timeout_ms, ivpu_inference_timeout_ms, ulong, 0644);
+MODULE_PARM_DESC(inference_timeout_ms, "Inference maximum duration, in milliseconds, 0 - default");
+
#define PM_RESCHEDULE_LIMIT 5
static void ivpu_pm_prepare_cold_boot(struct ivpu_device *vdev)
@@ -44,6 +48,7 @@ static void ivpu_pm_prepare_cold_boot(struct ivpu_device *vdev)
ivpu_fw_log_reset(vdev);
ivpu_fw_load(vdev);
fw->entry_point = fw->cold_boot_entry_point;
+ fw->last_heartbeat = 0;
}
static void ivpu_pm_prepare_warm_boot(struct ivpu_device *vdev)
@@ -189,7 +194,30 @@ static void ivpu_job_timeout_work(struct work_struct *work)
{
struct ivpu_pm_info *pm = container_of(work, struct ivpu_pm_info, job_timeout_work.work);
struct ivpu_device *vdev = pm->vdev;
+ unsigned long timeout_ms = ivpu_tdr_timeout_ms ? ivpu_tdr_timeout_ms : vdev->timeout.tdr;
+ unsigned long inference_timeout_ms = ivpu_inference_timeout_ms ? ivpu_inference_timeout_ms :
+ vdev->timeout.inference;
+ u64 inference_max_retries;
+ u64 heartbeat;
+
+ if (ivpu_jsm_get_heartbeat(vdev, 0, &heartbeat) || heartbeat <= vdev->fw->last_heartbeat) {
+ ivpu_err(vdev, "Job timeout detected, heartbeat not progressed\n");
+ goto recovery;
+ }
+
+ inference_max_retries = DIV_ROUND_UP(inference_timeout_ms, timeout_ms);
+ if (atomic_fetch_inc(&vdev->job_timeout_counter) >= inference_max_retries) {
+ ivpu_err(vdev, "Job timeout detected, heartbeat limit (%lld) exceeded\n",
+ inference_max_retries);
+ goto recovery;
+ }
+ vdev->fw->last_heartbeat = heartbeat;
+ ivpu_start_job_timeout_detection(vdev);
+ return;
+
+recovery:
+ atomic_set(&vdev->job_timeout_counter, 0);
ivpu_pm_trigger_recovery(vdev, "TDR");
}
@@ -204,6 +232,7 @@ void ivpu_start_job_timeout_detection(struct ivpu_device *vdev)
void ivpu_stop_job_timeout_detection(struct ivpu_device *vdev)
{
cancel_delayed_work_sync(&vdev->pm->job_timeout_work);
+ atomic_set(&vdev->job_timeout_counter, 0);
}
int ivpu_pm_suspend_cb(struct device *dev)
@@ -428,16 +457,17 @@ int ivpu_pm_dct_enable(struct ivpu_device *vdev, u8 active_percent)
active_us = (DCT_PERIOD_US * active_percent) / 100;
inactive_us = DCT_PERIOD_US - active_us;
+ vdev->pm->dct_active_percent = active_percent;
+
+ ivpu_dbg(vdev, PM, "DCT requested %u%% (D0: %uus, D0i2: %uus)\n",
+ active_percent, active_us, inactive_us);
+
ret = ivpu_jsm_dct_enable(vdev, active_us, inactive_us);
if (ret) {
ivpu_err_ratelimited(vdev, "Failed to enable DCT: %d\n", ret);
return ret;
}
- vdev->pm->dct_active_percent = active_percent;
-
- ivpu_dbg(vdev, PM, "DCT set to %u%% (D0: %uus, D0i2: %uus)\n",
- active_percent, active_us, inactive_us);
return 0;
}
@@ -445,15 +475,16 @@ int ivpu_pm_dct_disable(struct ivpu_device *vdev)
{
int ret;
+ vdev->pm->dct_active_percent = 0;
+
+ ivpu_dbg(vdev, PM, "DCT requested to be disabled\n");
+
ret = ivpu_jsm_dct_disable(vdev);
if (ret) {
ivpu_err_ratelimited(vdev, "Failed to disable DCT: %d\n", ret);
return ret;
}
- vdev->pm->dct_active_percent = 0;
-
- ivpu_dbg(vdev, PM, "DCT disabled\n");
return 0;
}
@@ -466,7 +497,7 @@ void ivpu_pm_irq_dct_work_fn(struct work_struct *work)
if (ivpu_hw_btrs_dct_get_request(vdev, &enable))
return;
- if (vdev->pm->dct_active_percent)
+ if (enable)
ret = ivpu_pm_dct_enable(vdev, DCT_DEFAULT_ACTIVE_PERCENT);
else
ret = ivpu_pm_dct_disable(vdev);
diff --git a/drivers/accel/ivpu/ivpu_sysfs.c b/drivers/accel/ivpu/ivpu_sysfs.c
index 97102feaf8dd..268ab7744a8b 100644
--- a/drivers/accel/ivpu/ivpu_sysfs.c
+++ b/drivers/accel/ivpu/ivpu_sysfs.c
@@ -1,10 +1,12 @@
// SPDX-License-Identifier: GPL-2.0-only
/*
- * Copyright (C) 2024 Intel Corporation
+ * Copyright (C) 2024-2025 Intel Corporation
*/
#include <linux/device.h>
#include <linux/err.h>
+#include <linux/pm_runtime.h>
+#include <linux/units.h>
#include "ivpu_drv.h"
#include "ivpu_gem.h"
@@ -90,10 +92,55 @@ sched_mode_show(struct device *dev, struct device_attribute *attr, char *buf)
static DEVICE_ATTR_RO(sched_mode);
+/**
+ * DOC: npu_max_frequency
+ *
+ * The npu_max_frequency shows maximum frequency in MHz of the NPU's data
+ * processing unit
+ */
+static ssize_t
+npu_max_frequency_mhz_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+ struct drm_device *drm = dev_get_drvdata(dev);
+ struct ivpu_device *vdev = to_ivpu_device(drm);
+ u32 freq = ivpu_hw_dpu_max_freq_get(vdev);
+
+ return sysfs_emit(buf, "%lu\n", freq / HZ_PER_MHZ);
+}
+
+static DEVICE_ATTR_RO(npu_max_frequency_mhz);
+
+/**
+ * DOC: npu_current_frequency_mhz
+ *
+ * The npu_current_frequency_mhz shows current frequency in MHz of the NPU's
+ * data processing unit
+ */
+static ssize_t
+npu_current_frequency_mhz_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+ struct drm_device *drm = dev_get_drvdata(dev);
+ struct ivpu_device *vdev = to_ivpu_device(drm);
+ u32 freq = 0;
+
+ /* Read frequency only if device is active, otherwise frequency is 0 */
+ if (pm_runtime_get_if_active(vdev->drm.dev) > 0) {
+ freq = ivpu_hw_dpu_freq_get(vdev);
+
+ pm_runtime_put_autosuspend(vdev->drm.dev);
+ }
+
+ return sysfs_emit(buf, "%lu\n", freq / HZ_PER_MHZ);
+}
+
+static DEVICE_ATTR_RO(npu_current_frequency_mhz);
+
static struct attribute *ivpu_dev_attrs[] = {
&dev_attr_npu_busy_time_us.attr,
&dev_attr_npu_memory_utilization.attr,
&dev_attr_sched_mode.attr,
+ &dev_attr_npu_max_frequency_mhz.attr,
+ &dev_attr_npu_current_frequency_mhz.attr,
NULL,
};
diff --git a/drivers/accel/ivpu/vpu_boot_api.h b/drivers/accel/ivpu/vpu_boot_api.h
index 908e68ea1c39..218468bbbcad 100644
--- a/drivers/accel/ivpu/vpu_boot_api.h
+++ b/drivers/accel/ivpu/vpu_boot_api.h
@@ -26,7 +26,7 @@
* Minor version changes when API backward compatibility is preserved.
* Resets to 0 if Major version is incremented.
*/
-#define VPU_BOOT_API_VER_MINOR 26
+#define VPU_BOOT_API_VER_MINOR 28
/*
* API header changed (field names, documentation, formatting) but API itself has not been changed
@@ -76,8 +76,15 @@ struct vpu_firmware_header {
* submission queue size and device capabilities.
*/
u32 preemption_buffer_2_size;
+ /*
+ * Maximum preemption buffer size that the FW can use: no need for the host
+ * driver to allocate more space than that specified by these fields.
+ * A value of 0 means no declared limit.
+ */
+ u32 preemption_buffer_1_max_size;
+ u32 preemption_buffer_2_max_size;
/* Space reserved for future preemption-related fields. */
- u32 preemption_reserved[6];
+ u32 preemption_reserved[4];
/* FW image read only section start address, 4KB aligned */
u64 ro_section_start_address;
/* FW image read only section size, 4KB aligned */
@@ -134,7 +141,7 @@ enum vpu_trace_destination {
/*
* Processor bit shifts (for loggable HW components).
*/
-#define VPU_TRACE_PROC_BIT_ARM 0
+#define VPU_TRACE_PROC_BIT_RESERVED 0
#define VPU_TRACE_PROC_BIT_LRT 1
#define VPU_TRACE_PROC_BIT_LNN 2
#define VPU_TRACE_PROC_BIT_SHV_0 3
diff --git a/drivers/accel/ivpu/vpu_jsm_api.h b/drivers/accel/ivpu/vpu_jsm_api.h
index 7215c144158c..4b6b2b3d2583 100644
--- a/drivers/accel/ivpu/vpu_jsm_api.h
+++ b/drivers/accel/ivpu/vpu_jsm_api.h
@@ -22,7 +22,7 @@
/*
* Minor version changes when API backward compatibility is preserved.
*/
-#define VPU_JSM_API_VER_MINOR 25
+#define VPU_JSM_API_VER_MINOR 29
/*
* API header changed (field names, documentation, formatting) but API itself has not been changed
@@ -53,8 +53,7 @@
* Engine indexes.
*/
#define VPU_ENGINE_COMPUTE 0
-#define VPU_ENGINE_COPY 1
-#define VPU_ENGINE_NB 2
+#define VPU_ENGINE_NB 1
/*
* VPU status values.
@@ -126,11 +125,13 @@ enum {
* When set, indicates that job queue uses native fences (as inline commands
* in job queue). Such queues may also use legacy fences (as commands in batch buffers).
* When cleared, indicates the job queue only uses legacy fences.
- * NOTE: For queues using native fences, VPU expects that all jobs in the queue
- * are immediately followed by an inline command object. This object is expected
- * to be a fence signal command in most cases, but can also be a NOP in case the host
- * does not need per-job fence signalling. Other inline commands objects can be
- * inserted between "job and inline command" pairs.
+ * NOTES:
+ * 1. For queues using native fences, VPU expects that all jobs in the queue
+ * are immediately followed by an inline command object. This object is expected
+ * to be a fence signal command in most cases, but can also be a NOP in case the host
+ * does not need per-job fence signalling. Other inline commands objects can be
+ * inserted between "job and inline command" pairs.
+ * 2. Native fence queues are only supported on VPU 40xx onwards.
*/
VPU_JOB_QUEUE_FLAGS_USE_NATIVE_FENCE_MASK = (1 << 1U),
@@ -275,6 +276,8 @@ struct vpu_inline_cmd {
u64 value;
/* User VA of the log buffer in which to add log entry on completion. */
u64 log_buffer_va;
+ /* NPU private data. */
+ u64 npu_private_data;
} fence;
/* Other commands do not have a payload. */
/* Payload definition for future inline commands can be inserted here. */
@@ -791,12 +794,22 @@ struct vpu_jsm_metric_streamer_update {
/** Metric group mask that identifies metric streamer instance. */
u64 metric_group_mask;
/**
- * Address and size of the buffer where the VPU will write metric data. If
- * the buffer address is 0 or same as the currently used buffer the VPU will
- * continue writing metric data to the current buffer. In this case the
- * buffer size is ignored and the size of the current buffer is unchanged.
- * If the address is non-zero and differs from the current buffer address the
- * VPU will immediately switch data collection to the new buffer.
+ * Address and size of the buffer where the VPU will write metric data.
+ * This member dictates how the update operation should perform:
+ * 1. client needs information about the number of collected samples and the
+ * amount of data written to the current buffer
+ * 2. client wants to switch to a new buffer
+ *
+ * Case 1. is identified by the buffer address being 0 or the same as the
+ * currently used buffer address. In this case the buffer size is ignored and
+ * the size of the current buffer is unchanged. The VPU will return an update
+ * in the vpu_jsm_metric_streamer_done structure. The internal writing position
+ * into the buffer is not changed.
+ *
+ * Case 2. is identified by the address being non-zero and differs from the
+ * current buffer address. The VPU will immediately switch data collection to
+ * the new buffer. Then the VPU will return an update in the
+ * vpu_jsm_metric_streamer_done structure.
*/
u64 buffer_addr;
u64 buffer_size;
@@ -934,6 +947,7 @@ struct vpu_ipc_msg_payload_hws_priority_band_setup {
/*
* Default quantum in 100ns units for scheduling across processes
* within a priority band
+ * Minimum value supported by NPU is 1ms (10000 in 100ns units).
*/
u32 process_quantum[VPU_HWS_NUM_PRIORITY_BANDS];
/*
@@ -946,8 +960,10 @@ struct vpu_ipc_msg_payload_hws_priority_band_setup {
* in situations when it's starved by the focus band.
*/
u32 normal_band_percentage;
- /* Reserved */
- u32 reserved_0;
+ /*
+ * TDR timeout value in milliseconds. Default value of 0 meaning no timeout.
+ */
+ u32 tdr_timeout;
};
/*
@@ -1024,7 +1040,10 @@ struct vpu_ipc_msg_payload_hws_set_context_sched_properties {
s32 in_process_priority;
/* Zero padding / Reserved */
u32 reserved_1;
- /* Context quantum relative to other contexts of same priority in the same process */
+ /*
+ * Context quantum relative to other contexts of same priority in the same process
+ * Minimum value supported by NPU is 1ms (10000 in 100ns units).
+ */
u64 context_quantum;
/* Grace period when preempting context of the same priority within the same process */
u64 grace_period_same_priority;
diff --git a/drivers/accel/qaic/Kconfig b/drivers/accel/qaic/Kconfig
index a9f866230058..5e405a19c157 100644
--- a/drivers/accel/qaic/Kconfig
+++ b/drivers/accel/qaic/Kconfig
@@ -8,7 +8,6 @@ config DRM_ACCEL_QAIC
depends on DRM_ACCEL
depends on PCI && HAS_IOMEM
depends on MHI_BUS
- depends on MMU
select CRC32
help
Enables driver for Qualcomm's Cloud AI accelerator PCIe cards that are
diff --git a/drivers/accel/qaic/Makefile b/drivers/accel/qaic/Makefile
index 35e883515629..1106b876f737 100644
--- a/drivers/accel/qaic/Makefile
+++ b/drivers/accel/qaic/Makefile
@@ -10,6 +10,7 @@ qaic-y := \
qaic_control.o \
qaic_data.o \
qaic_drv.o \
+ qaic_ras.o \
qaic_timesync.o \
sahara.o
diff --git a/drivers/accel/qaic/qaic.h b/drivers/accel/qaic/qaic.h
index 0dbb8e32e4b9..c31081e42cee 100644
--- a/drivers/accel/qaic/qaic.h
+++ b/drivers/accel/qaic/qaic.h
@@ -167,6 +167,14 @@ struct qaic_device {
struct workqueue_struct *bootlog_wq;
/* Synchronizes access of pages in MHI bootlog device */
struct mutex bootlog_mutex;
+ /* MHI RAS channel device */
+ struct mhi_device *ras_ch;
+ /* Correctable error count */
+ unsigned int ce_count;
+ /* Un-correctable error count */
+ unsigned int ue_count;
+ /* Un-correctable non-fatal error count */
+ unsigned int ue_nf_count;
};
struct qaic_drm_device {
@@ -213,8 +221,6 @@ struct qaic_bo {
bool sliced;
/* Request ID of this BO if it is queued for execution */
u16 req_id;
- /* Handle assigned to this BO */
- u32 handle;
/* Wait on this for completion of DMA transfer of this BO */
struct completion xfer_done;
/*
diff --git a/drivers/accel/qaic/qaic_data.c b/drivers/accel/qaic/qaic_data.c
index 43aba57b48f0..797289e9d780 100644
--- a/drivers/accel/qaic/qaic_data.c
+++ b/drivers/accel/qaic/qaic_data.c
@@ -609,7 +609,7 @@ static int qaic_gem_object_mmap(struct drm_gem_object *obj, struct vm_area_struc
struct scatterlist *sg;
int ret = 0;
- if (obj->import_attach)
+ if (drm_gem_is_imported(obj))
return -EINVAL;
for (sg = bo->sgt->sgl; sg; sg = sg_next(sg)) {
@@ -630,7 +630,7 @@ static void qaic_free_object(struct drm_gem_object *obj)
{
struct qaic_bo *bo = to_qaic_bo(obj);
- if (obj->import_attach) {
+ if (drm_gem_is_imported(obj)) {
/* DMABUF/PRIME Path */
drm_prime_gem_destroy(obj, NULL);
} else {
@@ -731,7 +731,6 @@ int qaic_create_bo_ioctl(struct drm_device *dev, void *data, struct drm_file *fi
if (ret)
goto free_bo;
- bo->handle = args->handle;
drm_gem_object_put(obj);
srcu_read_unlock(&qdev->dev_lock, qdev_rcu_id);
srcu_read_unlock(&usr->qddev_lock, usr_rcu_id);
@@ -870,7 +869,7 @@ static int qaic_prepare_bo(struct qaic_device *qdev, struct qaic_bo *bo,
{
int ret;
- if (bo->base.import_attach)
+ if (drm_gem_is_imported(&bo->base))
ret = qaic_prepare_import_bo(bo, hdr);
else
ret = qaic_prepare_export_bo(qdev, bo, hdr);
@@ -894,7 +893,7 @@ static void qaic_unprepare_export_bo(struct qaic_device *qdev, struct qaic_bo *b
static void qaic_unprepare_bo(struct qaic_device *qdev, struct qaic_bo *bo)
{
- if (bo->base.import_attach)
+ if (drm_gem_is_imported(&bo->base))
qaic_unprepare_import_bo(bo);
else
qaic_unprepare_export_bo(qdev, bo);
diff --git a/drivers/accel/qaic/qaic_debugfs.c b/drivers/accel/qaic/qaic_debugfs.c
index ba0cf2f94732..a991b8198dc4 100644
--- a/drivers/accel/qaic/qaic_debugfs.c
+++ b/drivers/accel/qaic/qaic_debugfs.c
@@ -240,7 +240,6 @@ static int qaic_bootlog_mhi_probe(struct mhi_device *mhi_dev, const struct mhi_d
mhi_unprepare:
mhi_unprepare_from_transfer(mhi_dev);
destroy_workqueue:
- flush_workqueue(qdev->bootlog_wq);
destroy_workqueue(qdev->bootlog_wq);
out:
return ret;
@@ -253,7 +252,6 @@ static void qaic_bootlog_mhi_remove(struct mhi_device *mhi_dev)
qdev = dev_get_drvdata(&mhi_dev->dev);
mhi_unprepare_from_transfer(qdev->bootlog_ch);
- flush_workqueue(qdev->bootlog_wq);
destroy_workqueue(qdev->bootlog_wq);
qdev->bootlog_ch = NULL;
}
diff --git a/drivers/accel/qaic/qaic_drv.c b/drivers/accel/qaic/qaic_drv.c
index 3b415e2c9431..e31bcb0ecfc9 100644
--- a/drivers/accel/qaic/qaic_drv.c
+++ b/drivers/accel/qaic/qaic_drv.c
@@ -29,6 +29,7 @@
#include "mhi_controller.h"
#include "qaic.h"
#include "qaic_debugfs.h"
+#include "qaic_ras.h"
#include "qaic_timesync.h"
#include "sahara.h"
@@ -695,6 +696,10 @@ static int __init qaic_init(void)
if (ret)
pr_debug("qaic: qaic_bootlog_register failed %d\n", ret);
+ ret = qaic_ras_register();
+ if (ret)
+ pr_debug("qaic: qaic_ras_register failed %d\n", ret);
+
return 0;
free_mhi:
@@ -722,6 +727,7 @@ static void __exit qaic_exit(void)
* reinitializing the link_up state after the cleanup is done.
*/
link_up = true;
+ qaic_ras_unregister();
qaic_bootlog_unregister();
qaic_timesync_deinit();
sahara_unregister();
diff --git a/drivers/accel/qaic/qaic_ras.c b/drivers/accel/qaic/qaic_ras.c
new file mode 100644
index 000000000000..914ffc4a9970
--- /dev/null
+++ b/drivers/accel/qaic/qaic_ras.c
@@ -0,0 +1,642 @@
+// SPDX-License-Identifier: GPL-2.0-only
+
+/* Copyright (c) 2020-2021, The Linux Foundation. All rights reserved. */
+/* Copyright (c) 2022-2024 Qualcomm Innovation Center, Inc. All rights reserved. */
+/* Copyright (c) Qualcomm Technologies, Inc. and/or its subsidiaries. */
+
+#include <asm/byteorder.h>
+#include <linux/device.h>
+#include <linux/kernel.h>
+#include <linux/mhi.h>
+
+#include "qaic.h"
+#include "qaic_ras.h"
+
+#define MAGIC 0x55AA
+#define VERSION 0x2
+#define HDR_SZ 12
+#define NUM_TEMP_LVL 3
+#define POWER_BREAK BIT(0)
+
+enum msg_type {
+ MSG_PUSH, /* async push from device */
+ MSG_REQ, /* sync request to device */
+ MSG_RESP, /* sync response from device */
+};
+
+enum err_type {
+ CE, /* correctable error */
+ UE, /* uncorrectable error */
+ UE_NF, /* uncorrectable error that is non-fatal, expect a disruption */
+ ERR_TYPE_MAX,
+};
+
+static const char * const err_type_str[] = {
+ [CE] = "Correctable",
+ [UE] = "Uncorrectable",
+ [UE_NF] = "Uncorrectable Non-Fatal",
+};
+
+static const char * const err_class_str[] = {
+ [CE] = "Warning",
+ [UE] = "Fatal",
+ [UE_NF] = "Warning",
+};
+
+enum err_source {
+ SOC_MEM,
+ PCIE,
+ DDR,
+ SYS_BUS1,
+ SYS_BUS2,
+ NSP_MEM,
+ TSENS,
+};
+
+static const char * const err_src_str[TSENS + 1] = {
+ [SOC_MEM] = "SoC Memory",
+ [PCIE] = "PCIE",
+ [DDR] = "DDR",
+ [SYS_BUS1] = "System Bus source 1",
+ [SYS_BUS2] = "System Bus source 2",
+ [NSP_MEM] = "NSP Memory",
+ [TSENS] = "Temperature Sensors",
+};
+
+struct ras_data {
+ /* header start */
+ /* Magic number to validate the message */
+ u16 magic;
+ /* RAS version number */
+ u16 ver;
+ u32 seq_num;
+ /* RAS message type */
+ u8 type;
+ u8 id;
+ /* Size of RAS message without the header in byte */
+ u16 len;
+ /* header end */
+ s32 result;
+ /*
+ * Error source
+ * 0 : SoC Memory
+ * 1 : PCIE
+ * 2 : DDR
+ * 3 : System Bus source 1
+ * 4 : System Bus source 2
+ * 5 : NSP Memory
+ * 6 : Temperature Sensors
+ */
+ u32 source;
+ /*
+ * Stores the error type, there are three types of error in RAS
+ * 0 : correctable error (CE)
+ * 1 : uncorrectable error (UE)
+ * 2 : uncorrectable error that is non-fatal (UE_NF)
+ */
+ u32 err_type;
+ u32 err_threshold;
+ u32 ce_count;
+ u32 ue_count;
+ u32 intr_num;
+ /* Data specific to error source */
+ u8 syndrome[64];
+} __packed;
+
+struct soc_mem_syndrome {
+ u64 error_address[8];
+} __packed;
+
+struct nsp_mem_syndrome {
+ u32 error_address[8];
+ u8 nsp_id;
+} __packed;
+
+struct ddr_syndrome {
+ u32 count;
+ u32 irq_status;
+ u32 data_31_0[2];
+ u32 data_63_32[2];
+ u32 data_95_64[2];
+ u32 data_127_96[2];
+ u32 addr_lsb;
+ u16 addr_msb;
+ u16 parity_bits;
+ u16 instance;
+ u16 err_type;
+} __packed;
+
+struct tsens_syndrome {
+ u32 threshold_type;
+ s32 temp;
+} __packed;
+
+struct sysbus1_syndrome {
+ u32 slave;
+ u32 err_type;
+ u16 addr[8];
+ u8 instance;
+} __packed;
+
+struct sysbus2_syndrome {
+ u32 lsb3;
+ u32 msb3;
+ u32 lsb2;
+ u32 msb2;
+ u32 ext_id;
+ u16 path;
+ u16 op_type;
+ u16 len;
+ u16 redirect;
+ u8 valid;
+ u8 word_error;
+ u8 non_secure;
+ u8 opc;
+ u8 error_code;
+ u8 trans_type;
+ u8 addr_space;
+ u8 instance;
+} __packed;
+
+struct pcie_syndrome {
+ /* CE info */
+ u32 bad_tlp;
+ u32 bad_dllp;
+ u32 replay_rollover;
+ u32 replay_timeout;
+ u32 rx_err;
+ u32 internal_ce_count;
+ /* UE_NF info */
+ u32 fc_timeout;
+ u32 poison_tlp;
+ u32 ecrc_err;
+ u32 unsupported_req;
+ u32 completer_abort;
+ u32 completion_timeout;
+ /* UE info */
+ u32 addr;
+ u8 index;
+ /*
+ * Flag to indicate specific event of PCIe
+ * BIT(0): Power break (low power)
+ * BIT(1) to BIT(7): Reserved
+ */
+ u8 flag;
+} __packed;
+
+static const char * const threshold_type_str[NUM_TEMP_LVL] = {
+ [0] = "lower",
+ [1] = "upper",
+ [2] = "critical",
+};
+
+static void ras_msg_to_cpu(struct ras_data *msg)
+{
+ struct sysbus1_syndrome *sysbus1_syndrome = (struct sysbus1_syndrome *)&msg->syndrome[0];
+ struct sysbus2_syndrome *sysbus2_syndrome = (struct sysbus2_syndrome *)&msg->syndrome[0];
+ struct soc_mem_syndrome *soc_syndrome = (struct soc_mem_syndrome *)&msg->syndrome[0];
+ struct nsp_mem_syndrome *nsp_syndrome = (struct nsp_mem_syndrome *)&msg->syndrome[0];
+ struct tsens_syndrome *tsens_syndrome = (struct tsens_syndrome *)&msg->syndrome[0];
+ struct pcie_syndrome *pcie_syndrome = (struct pcie_syndrome *)&msg->syndrome[0];
+ struct ddr_syndrome *ddr_syndrome = (struct ddr_syndrome *)&msg->syndrome[0];
+ int i;
+
+ le16_to_cpus(&msg->magic);
+ le16_to_cpus(&msg->ver);
+ le32_to_cpus(&msg->seq_num);
+ le16_to_cpus(&msg->len);
+ le32_to_cpus(&msg->result);
+ le32_to_cpus(&msg->source);
+ le32_to_cpus(&msg->err_type);
+ le32_to_cpus(&msg->err_threshold);
+ le32_to_cpus(&msg->ce_count);
+ le32_to_cpus(&msg->ue_count);
+ le32_to_cpus(&msg->intr_num);
+
+ switch (msg->source) {
+ case SOC_MEM:
+ for (i = 0; i < 8; i++)
+ le64_to_cpus(&soc_syndrome->error_address[i]);
+ break;
+ case PCIE:
+ le32_to_cpus(&pcie_syndrome->bad_tlp);
+ le32_to_cpus(&pcie_syndrome->bad_dllp);
+ le32_to_cpus(&pcie_syndrome->replay_rollover);
+ le32_to_cpus(&pcie_syndrome->replay_timeout);
+ le32_to_cpus(&pcie_syndrome->rx_err);
+ le32_to_cpus(&pcie_syndrome->internal_ce_count);
+ le32_to_cpus(&pcie_syndrome->fc_timeout);
+ le32_to_cpus(&pcie_syndrome->poison_tlp);
+ le32_to_cpus(&pcie_syndrome->ecrc_err);
+ le32_to_cpus(&pcie_syndrome->unsupported_req);
+ le32_to_cpus(&pcie_syndrome->completer_abort);
+ le32_to_cpus(&pcie_syndrome->completion_timeout);
+ le32_to_cpus(&pcie_syndrome->addr);
+ break;
+ case DDR:
+ le16_to_cpus(&ddr_syndrome->instance);
+ le16_to_cpus(&ddr_syndrome->err_type);
+ le32_to_cpus(&ddr_syndrome->count);
+ le32_to_cpus(&ddr_syndrome->irq_status);
+ le32_to_cpus(&ddr_syndrome->data_31_0[0]);
+ le32_to_cpus(&ddr_syndrome->data_31_0[1]);
+ le32_to_cpus(&ddr_syndrome->data_63_32[0]);
+ le32_to_cpus(&ddr_syndrome->data_63_32[1]);
+ le32_to_cpus(&ddr_syndrome->data_95_64[0]);
+ le32_to_cpus(&ddr_syndrome->data_95_64[1]);
+ le32_to_cpus(&ddr_syndrome->data_127_96[0]);
+ le32_to_cpus(&ddr_syndrome->data_127_96[1]);
+ le16_to_cpus(&ddr_syndrome->parity_bits);
+ le16_to_cpus(&ddr_syndrome->addr_msb);
+ le32_to_cpus(&ddr_syndrome->addr_lsb);
+ break;
+ case SYS_BUS1:
+ le32_to_cpus(&sysbus1_syndrome->slave);
+ le32_to_cpus(&sysbus1_syndrome->err_type);
+ for (i = 0; i < 8; i++)
+ le16_to_cpus(&sysbus1_syndrome->addr[i]);
+ break;
+ case SYS_BUS2:
+ le16_to_cpus(&sysbus2_syndrome->op_type);
+ le16_to_cpus(&sysbus2_syndrome->len);
+ le16_to_cpus(&sysbus2_syndrome->redirect);
+ le16_to_cpus(&sysbus2_syndrome->path);
+ le32_to_cpus(&sysbus2_syndrome->ext_id);
+ le32_to_cpus(&sysbus2_syndrome->lsb2);
+ le32_to_cpus(&sysbus2_syndrome->msb2);
+ le32_to_cpus(&sysbus2_syndrome->lsb3);
+ le32_to_cpus(&sysbus2_syndrome->msb3);
+ break;
+ case NSP_MEM:
+ for (i = 0; i < 8; i++)
+ le32_to_cpus(&nsp_syndrome->error_address[i]);
+ break;
+ case TSENS:
+ le32_to_cpus(&tsens_syndrome->threshold_type);
+ le32_to_cpus(&tsens_syndrome->temp);
+ break;
+ }
+}
+
+static void decode_ras_msg(struct qaic_device *qdev, struct ras_data *msg)
+{
+ struct sysbus1_syndrome *sysbus1_syndrome = (struct sysbus1_syndrome *)&msg->syndrome[0];
+ struct sysbus2_syndrome *sysbus2_syndrome = (struct sysbus2_syndrome *)&msg->syndrome[0];
+ struct soc_mem_syndrome *soc_syndrome = (struct soc_mem_syndrome *)&msg->syndrome[0];
+ struct nsp_mem_syndrome *nsp_syndrome = (struct nsp_mem_syndrome *)&msg->syndrome[0];
+ struct tsens_syndrome *tsens_syndrome = (struct tsens_syndrome *)&msg->syndrome[0];
+ struct pcie_syndrome *pcie_syndrome = (struct pcie_syndrome *)&msg->syndrome[0];
+ struct ddr_syndrome *ddr_syndrome = (struct ddr_syndrome *)&msg->syndrome[0];
+ char *class;
+ char *level;
+
+ if (msg->magic != MAGIC) {
+ pci_warn(qdev->pdev, "Dropping RAS message with invalid magic %x\n", msg->magic);
+ return;
+ }
+
+ if (!msg->ver || msg->ver > VERSION) {
+ pci_warn(qdev->pdev, "Dropping RAS message with invalid version %d\n", msg->ver);
+ return;
+ }
+
+ if (msg->type != MSG_PUSH) {
+ pci_warn(qdev->pdev, "Dropping non-PUSH RAS message\n");
+ return;
+ }
+
+ if (msg->len != sizeof(*msg) - HDR_SZ) {
+ pci_warn(qdev->pdev, "Dropping RAS message with invalid len %d\n", msg->len);
+ return;
+ }
+
+ if (msg->err_type >= ERR_TYPE_MAX) {
+ pci_warn(qdev->pdev, "Dropping RAS message with err type %d\n", msg->err_type);
+ return;
+ }
+
+ if (msg->err_type == UE)
+ level = KERN_ERR;
+ else
+ level = KERN_WARNING;
+
+ switch (msg->source) {
+ case SOC_MEM:
+ dev_printk(level, &qdev->pdev->dev, "RAS event.\nClass:%s\nDescription:%s %s %s\nError Threshold for this report %d\nSyndrome:\n 0x%llx\n 0x%llx\n 0x%llx\n 0x%llx\n 0x%llx\n 0x%llx\n 0x%llx\n 0x%llx\n",
+ err_class_str[msg->err_type],
+ err_type_str[msg->err_type],
+ "error from",
+ err_src_str[msg->source],
+ msg->err_threshold,
+ soc_syndrome->error_address[0],
+ soc_syndrome->error_address[1],
+ soc_syndrome->error_address[2],
+ soc_syndrome->error_address[3],
+ soc_syndrome->error_address[4],
+ soc_syndrome->error_address[5],
+ soc_syndrome->error_address[6],
+ soc_syndrome->error_address[7]);
+ break;
+ case PCIE:
+ dev_printk(level, &qdev->pdev->dev, "RAS event.\nClass:%s\nDescription:%s %s %s\nError Threshold for this report %d\n",
+ err_class_str[msg->err_type],
+ err_type_str[msg->err_type],
+ "error from",
+ err_src_str[msg->source],
+ msg->err_threshold);
+
+ switch (msg->err_type) {
+ case CE:
+ /*
+ * Modeled after AER prints. This continues the dev_printk() from a few
+ * lines up. We reduce duplication of code, but also avoid re-printing the
+ * PCI device info so that the end result looks uniform to the log user.
+ */
+ printk(KERN_WARNING pr_fmt("Syndrome:\n Bad TLP count %d\n Bad DLLP count %d\n Replay Rollover count %d\n Replay Timeout count %d\n Recv Error count %d\n Internal CE count %d\n"),
+ pcie_syndrome->bad_tlp,
+ pcie_syndrome->bad_dllp,
+ pcie_syndrome->replay_rollover,
+ pcie_syndrome->replay_timeout,
+ pcie_syndrome->rx_err,
+ pcie_syndrome->internal_ce_count);
+ if (msg->ver > 0x1)
+ pr_warn(" Power break %s\n",
+ pcie_syndrome->flag & POWER_BREAK ? "ON" : "OFF");
+ break;
+ case UE:
+ printk(KERN_ERR pr_fmt("Syndrome:\n Index %d\n Address 0x%x\n"),
+ pcie_syndrome->index, pcie_syndrome->addr);
+ break;
+ case UE_NF:
+ printk(KERN_WARNING pr_fmt("Syndrome:\n FC timeout count %d\n Poisoned TLP count %d\n ECRC error count %d\n Unsupported request count %d\n Completer abort count %d\n Completion timeout count %d\n"),
+ pcie_syndrome->fc_timeout,
+ pcie_syndrome->poison_tlp,
+ pcie_syndrome->ecrc_err,
+ pcie_syndrome->unsupported_req,
+ pcie_syndrome->completer_abort,
+ pcie_syndrome->completion_timeout);
+ break;
+ default:
+ break;
+ }
+ break;
+ case DDR:
+ dev_printk(level, &qdev->pdev->dev, "RAS event.\nClass:%s\nDescription:%s %s %s\nError Threshold for this report %d\nSyndrome:\n Instance %d\n Count %d\n Data 31_0 0x%x 0x%x\n Data 63_32 0x%x 0x%x\n Data 95_64 0x%x 0x%x\n Data 127_96 0x%x 0x%x\n Parity bits 0x%x\n Address msb 0x%x\n Address lsb 0x%x\n",
+ err_class_str[msg->err_type],
+ err_type_str[msg->err_type],
+ "error from",
+ err_src_str[msg->source],
+ msg->err_threshold,
+ ddr_syndrome->instance,
+ ddr_syndrome->count,
+ ddr_syndrome->data_31_0[1],
+ ddr_syndrome->data_31_0[0],
+ ddr_syndrome->data_63_32[1],
+ ddr_syndrome->data_63_32[0],
+ ddr_syndrome->data_95_64[1],
+ ddr_syndrome->data_95_64[0],
+ ddr_syndrome->data_127_96[1],
+ ddr_syndrome->data_127_96[0],
+ ddr_syndrome->parity_bits,
+ ddr_syndrome->addr_msb,
+ ddr_syndrome->addr_lsb);
+ break;
+ case SYS_BUS1:
+ dev_printk(level, &qdev->pdev->dev, "RAS event.\nClass:%s\nDescription:%s %s %s\nError Threshold for this report %d\nSyndrome:\n instance %d\n %s\n err_type %d\n address0 0x%x\n address1 0x%x\n address2 0x%x\n address3 0x%x\n address4 0x%x\n address5 0x%x\n address6 0x%x\n address7 0x%x\n",
+ err_class_str[msg->err_type],
+ err_type_str[msg->err_type],
+ "error from",
+ err_src_str[msg->source],
+ msg->err_threshold,
+ sysbus1_syndrome->instance,
+ sysbus1_syndrome->slave ? "Slave" : "Master",
+ sysbus1_syndrome->err_type,
+ sysbus1_syndrome->addr[0],
+ sysbus1_syndrome->addr[1],
+ sysbus1_syndrome->addr[2],
+ sysbus1_syndrome->addr[3],
+ sysbus1_syndrome->addr[4],
+ sysbus1_syndrome->addr[5],
+ sysbus1_syndrome->addr[6],
+ sysbus1_syndrome->addr[7]);
+ break;
+ case SYS_BUS2:
+ dev_printk(level, &qdev->pdev->dev, "RAS event.\nClass:%s\nDescription:%s %s %s\nError Threshold for this report %d\nSyndrome:\n instance %d\n valid %d\n word error %d\n non-secure %d\n opc %d\n error code %d\n transaction type %d\n address space %d\n operation type %d\n len %d\n redirect %d\n path %d\n ext_id %d\n lsb2 %d\n msb2 %d\n lsb3 %d\n msb3 %d\n",
+ err_class_str[msg->err_type],
+ err_type_str[msg->err_type],
+ "error from",
+ err_src_str[msg->source],
+ msg->err_threshold,
+ sysbus2_syndrome->instance,
+ sysbus2_syndrome->valid,
+ sysbus2_syndrome->word_error,
+ sysbus2_syndrome->non_secure,
+ sysbus2_syndrome->opc,
+ sysbus2_syndrome->error_code,
+ sysbus2_syndrome->trans_type,
+ sysbus2_syndrome->addr_space,
+ sysbus2_syndrome->op_type,
+ sysbus2_syndrome->len,
+ sysbus2_syndrome->redirect,
+ sysbus2_syndrome->path,
+ sysbus2_syndrome->ext_id,
+ sysbus2_syndrome->lsb2,
+ sysbus2_syndrome->msb2,
+ sysbus2_syndrome->lsb3,
+ sysbus2_syndrome->msb3);
+ break;
+ case NSP_MEM:
+ dev_printk(level, &qdev->pdev->dev, "RAS event.\nClass:%s\nDescription:%s %s %s\nError Threshold for this report %d\nSyndrome:\n NSP ID %d\n 0x%x\n 0x%x\n 0x%x\n 0x%x\n 0x%x\n 0x%x\n 0x%x\n 0x%x\n",
+ err_class_str[msg->err_type],
+ err_type_str[msg->err_type],
+ "error from",
+ err_src_str[msg->source],
+ msg->err_threshold,
+ nsp_syndrome->nsp_id,
+ nsp_syndrome->error_address[0],
+ nsp_syndrome->error_address[1],
+ nsp_syndrome->error_address[2],
+ nsp_syndrome->error_address[3],
+ nsp_syndrome->error_address[4],
+ nsp_syndrome->error_address[5],
+ nsp_syndrome->error_address[6],
+ nsp_syndrome->error_address[7]);
+ break;
+ case TSENS:
+ if (tsens_syndrome->threshold_type >= NUM_TEMP_LVL) {
+ pci_warn(qdev->pdev, "Dropping RAS message with invalid temp threshold %d\n",
+ tsens_syndrome->threshold_type);
+ break;
+ }
+
+ if (msg->err_type)
+ class = "Fatal";
+ else if (tsens_syndrome->threshold_type)
+ class = "Critical";
+ else
+ class = "Warning";
+
+ dev_printk(level, &qdev->pdev->dev, "RAS event.\nClass:%s\nDescription:%s %s %s\nError Threshold for this report %d\nSyndrome:\n %s threshold\n %d deg C\n",
+ class,
+ err_type_str[msg->err_type],
+ "error from",
+ err_src_str[msg->source],
+ msg->err_threshold,
+ threshold_type_str[tsens_syndrome->threshold_type],
+ tsens_syndrome->temp);
+ break;
+ }
+
+ /* Uncorrectable errors are fatal */
+ if (msg->err_type == UE)
+ mhi_soc_reset(qdev->mhi_cntrl);
+
+ switch (msg->err_type) {
+ case CE:
+ if (qdev->ce_count != UINT_MAX)
+ qdev->ce_count++;
+ break;
+ case UE:
+ if (qdev->ce_count != UINT_MAX)
+ qdev->ue_count++;
+ break;
+ case UE_NF:
+ if (qdev->ce_count != UINT_MAX)
+ qdev->ue_nf_count++;
+ break;
+ default:
+ /* not possible */
+ break;
+ }
+}
+
+static ssize_t ce_count_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+ struct qaic_device *qdev = pci_get_drvdata(to_pci_dev(dev));
+
+ return snprintf(buf, PAGE_SIZE, "%d\n", qdev->ce_count);
+}
+
+static ssize_t ue_count_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+ struct qaic_device *qdev = pci_get_drvdata(to_pci_dev(dev));
+
+ return snprintf(buf, PAGE_SIZE, "%d\n", qdev->ue_count);
+}
+
+static ssize_t ue_nonfatal_count_show(struct device *dev, struct device_attribute *attr, char *buf)
+{
+ struct qaic_device *qdev = pci_get_drvdata(to_pci_dev(dev));
+
+ return snprintf(buf, PAGE_SIZE, "%d\n", qdev->ue_nf_count);
+}
+
+static DEVICE_ATTR_RO(ce_count);
+static DEVICE_ATTR_RO(ue_count);
+static DEVICE_ATTR_RO(ue_nonfatal_count);
+
+static struct attribute *ras_attrs[] = {
+ &dev_attr_ce_count.attr,
+ &dev_attr_ue_count.attr,
+ &dev_attr_ue_nonfatal_count.attr,
+ NULL,
+};
+
+static struct attribute_group ras_group = {
+ .attrs = ras_attrs,
+};
+
+static int qaic_ras_mhi_probe(struct mhi_device *mhi_dev, const struct mhi_device_id *id)
+{
+ struct qaic_device *qdev = pci_get_drvdata(to_pci_dev(mhi_dev->mhi_cntrl->cntrl_dev));
+ struct ras_data *resp;
+ int ret;
+
+ ret = mhi_prepare_for_transfer(mhi_dev);
+ if (ret)
+ return ret;
+
+ resp = kzalloc(sizeof(*resp), GFP_KERNEL);
+ if (!resp) {
+ mhi_unprepare_from_transfer(mhi_dev);
+ return -ENOMEM;
+ }
+
+ ret = mhi_queue_buf(mhi_dev, DMA_FROM_DEVICE, resp, sizeof(*resp), MHI_EOT);
+ if (ret) {
+ kfree(resp);
+ mhi_unprepare_from_transfer(mhi_dev);
+ return ret;
+ }
+
+ ret = device_add_group(&qdev->pdev->dev, &ras_group);
+ if (ret) {
+ mhi_unprepare_from_transfer(mhi_dev);
+ pci_dbg(qdev->pdev, "ras add sysfs failed %d\n", ret);
+ return ret;
+ }
+
+ dev_set_drvdata(&mhi_dev->dev, qdev);
+ qdev->ras_ch = mhi_dev;
+
+ return ret;
+}
+
+static void qaic_ras_mhi_remove(struct mhi_device *mhi_dev)
+{
+ struct qaic_device *qdev;
+
+ qdev = dev_get_drvdata(&mhi_dev->dev);
+ qdev->ras_ch = NULL;
+ device_remove_group(&qdev->pdev->dev, &ras_group);
+ mhi_unprepare_from_transfer(mhi_dev);
+}
+
+static void qaic_ras_mhi_ul_xfer_cb(struct mhi_device *mhi_dev, struct mhi_result *mhi_result) {}
+
+static void qaic_ras_mhi_dl_xfer_cb(struct mhi_device *mhi_dev, struct mhi_result *mhi_result)
+{
+ struct qaic_device *qdev = dev_get_drvdata(&mhi_dev->dev);
+ struct ras_data *msg = mhi_result->buf_addr;
+ int ret;
+
+ if (mhi_result->transaction_status) {
+ kfree(msg);
+ return;
+ }
+
+ ras_msg_to_cpu(msg);
+ decode_ras_msg(qdev, msg);
+
+ ret = mhi_queue_buf(qdev->ras_ch, DMA_FROM_DEVICE, msg, sizeof(*msg), MHI_EOT);
+ if (ret) {
+ dev_err(&mhi_dev->dev, "Cannot requeue RAS recv buf %d\n", ret);
+ kfree(msg);
+ }
+}
+
+static const struct mhi_device_id qaic_ras_mhi_match_table[] = {
+ { .chan = "QAIC_STATUS", },
+ {},
+};
+
+static struct mhi_driver qaic_ras_mhi_driver = {
+ .id_table = qaic_ras_mhi_match_table,
+ .remove = qaic_ras_mhi_remove,
+ .probe = qaic_ras_mhi_probe,
+ .ul_xfer_cb = qaic_ras_mhi_ul_xfer_cb,
+ .dl_xfer_cb = qaic_ras_mhi_dl_xfer_cb,
+ .driver = {
+ .name = "qaic_ras",
+ },
+};
+
+int qaic_ras_register(void)
+{
+ return mhi_driver_register(&qaic_ras_mhi_driver);
+}
+
+void qaic_ras_unregister(void)
+{
+ mhi_driver_unregister(&qaic_ras_mhi_driver);
+}
diff --git a/drivers/accel/qaic/qaic_ras.h b/drivers/accel/qaic/qaic_ras.h
new file mode 100644
index 000000000000..d44a4eeeb060
--- /dev/null
+++ b/drivers/accel/qaic/qaic_ras.h
@@ -0,0 +1,10 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/* Copyright (c) 2020, The Linux Foundation. All rights reserved. */
+
+#ifndef __QAIC_RAS_H__
+#define __QAIC_RAS_H__
+
+int qaic_ras_register(void);
+void qaic_ras_unregister(void);
+
+#endif /* __QAIC_RAS_H__ */
diff --git a/drivers/accel/qaic/qaic_timesync.c b/drivers/accel/qaic/qaic_timesync.c
index 2473c66309d4..3fac540f8e03 100644
--- a/drivers/accel/qaic/qaic_timesync.c
+++ b/drivers/accel/qaic/qaic_timesync.c
@@ -129,7 +129,7 @@ static void qaic_timesync_dl_xfer_cb(struct mhi_device *mhi_dev, struct mhi_resu
static void qaic_timesync_timer(struct timer_list *t)
{
- struct mqts_dev *mqtsdev = from_timer(mqtsdev, t, timer);
+ struct mqts_dev *mqtsdev = timer_container_of(mqtsdev, t, timer);
struct qts_host_time_sync_msg_data *sync_msg;
u64 device_qtimer_us;
u64 device_qtimer;
@@ -221,7 +221,7 @@ static void qaic_timesync_remove(struct mhi_device *mhi_dev)
{
struct mqts_dev *mqtsdev = dev_get_drvdata(&mhi_dev->dev);
- del_timer_sync(&mqtsdev->timer);
+ timer_delete_sync(&mqtsdev->timer);
mhi_unprepare_from_transfer(mqtsdev->mhi_dev);
kfree(mqtsdev->sync_msg);
kfree(mqtsdev);