All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH 1/6] habanalabs: hide memory manager page shift
@ 2022-05-02  8:05 Oded Gabbay
  2022-05-02  8:05 ` [PATCH 2/6] habanalabs: add put by handle method to memory manager Oded Gabbay
                   ` (4 more replies)
  0 siblings, 5 replies; 6+ messages in thread
From: Oded Gabbay @ 2022-05-02  8:05 UTC (permalink / raw)
  To: linux-kernel; +Cc: Yuri Nudelman

From: Yuri Nudelman <ynudelman@habana.ai>

The new unified memory manager uses page offset to pass buffer handle
during the mmap operation. One problem with this approach is that it
requires the handle to always be divisible by the page size, else, the
user would not be able to pass it correctly as an argument to the mmap
system call.

Previously, this was achieved by shifting the handle left after alloc
operation, and shifting it right before get operation. This was done in
the user code. This creates code duplication, and, what's worse,
requires some knowledge from the user regarding the handle internal
structure, hurting the encapsulation.

This patch encloses all the page shifts inside memory manager functions.
This way, the user can take the handle as a black box, and simply use
it, without any concert about how it actually works.

Signed-off-by: Yuri Nudelman <ynudelman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
 .../habanalabs/common/command_submission.c    | 12 ++-------
 drivers/misc/habanalabs/common/device.c       |  3 ++-
 drivers/misc/habanalabs/common/habanalabs.h   | 20 ++++++++------
 drivers/misc/habanalabs/common/memory.c       | 10 +++----
 drivers/misc/habanalabs/common/memory_mgr.c   | 27 +++++++++----------
 5 files changed, 32 insertions(+), 40 deletions(-)

diff --git a/drivers/misc/habanalabs/common/command_submission.c b/drivers/misc/habanalabs/common/command_submission.c
index 6c13ae3ffa7c..a189157be104 100644
--- a/drivers/misc/habanalabs/common/command_submission.c
+++ b/drivers/misc/habanalabs/common/command_submission.c
@@ -2952,7 +2952,7 @@ static int _hl_interrupt_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
 				bool register_ts_record, u64 ts_handle, u64 ts_offset,
 				u32 *status, u64 *timestamp)
 {
-	u32 cq_patched_handle, ts_patched_handle;
+	u32 cq_patched_handle;
 	struct hl_user_pending_interrupt *pend;
 	struct hl_mmap_mem_buf *buf;
 	struct hl_cb *cq_cb;
@@ -2974,15 +2974,7 @@ static int _hl_interrupt_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
 	if (register_ts_record) {
 		dev_dbg(hdev->dev, "Timestamp registration: interrupt id: %u, ts offset: %llu, cq_offset: %llu\n",
 					interrupt->interrupt_id, ts_offset, cq_counters_offset);
-
-		/* TODO:
-		 * See if this can be removed.
-		 * Embedding type in handle will no longer be needed as soon as we
-		 * switch to using a single memory manager for all memory types.
-		 * We may still need the page shift, though.
-		 */
-		ts_patched_handle = lower_32_bits(ts_handle >> PAGE_SHIFT);
-		buf = hl_mmap_mem_buf_get(mmg, ts_patched_handle);
+		buf = hl_mmap_mem_buf_get(mmg, ts_handle);
 		if (!buf) {
 			rc = -EINVAL;
 			goto put_cq_cb;
diff --git a/drivers/misc/habanalabs/common/device.c b/drivers/misc/habanalabs/common/device.c
index 350cd61e06c1..bd74e0359473 100644
--- a/drivers/misc/habanalabs/common/device.c
+++ b/drivers/misc/habanalabs/common/device.c
@@ -384,13 +384,14 @@ static int hl_mmap(struct file *filp, struct vm_area_struct *vma)
 	}
 
 	vm_pgoff = vma->vm_pgoff;
-	vma->vm_pgoff = HL_MMAP_OFFSET_VALUE_GET(vm_pgoff);
 
 	switch (vm_pgoff & HL_MMAP_TYPE_MASK) {
 	case HL_MMAP_TYPE_CB:
+		vma->vm_pgoff = HL_MMAP_OFFSET_VALUE_GET(vm_pgoff);
 		return hl_cb_mmap(hpriv, vma);
 
 	case HL_MMAP_TYPE_BLOCK:
+		vma->vm_pgoff = HL_MMAP_OFFSET_VALUE_GET(vm_pgoff);
 		return hl_hw_block_mmap(hpriv, vma);
 
 	case HL_MMAP_TYPE_TS_BUFF:
diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index 19f6af59cfc9..ea5cfea21a2c 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -750,21 +750,25 @@ struct hl_mem_mgr {
 };
 
 /**
- * struct hl_mmap_mem_buf_ops - describes unified memory manager buffer behavior
+ * struct hl_mmap_mem_buf_behavior - describes unified memory manager buffer behavior
+ * @mem_id: memory type identifier, embedded in the handle and used to identify
+ *          the memory type by handle.
  * @alloc: callback executed on buffer allocation, shall allocate the memory,
  *         set it under buffer private, and set mappable size.
  * @mmap: callback executed on mmap, must map the buffer to vma
  * @release: callback executed on release, must free the resources used by the buffer
  */
-struct hl_mmap_mem_buf_ops {
+struct hl_mmap_mem_buf_behavior {
+	u64 mem_id;
+
 	int (*alloc)(struct hl_mmap_mem_buf *buf, gfp_t gfp, void *args);
 	int (*mmap)(struct hl_mmap_mem_buf *buf, struct vm_area_struct *vma, void *args);
 	void (*release)(struct hl_mmap_mem_buf *buf);
 };
 
 /**
- * struct hl_mmap_mem_buf_ops - describes a single unified memory buffer
- * @ops: buffer behavior
+ * struct hl_mmap_mem_buf - describes a single unified memory buffer
+ * @behavior: buffer behavior
  * @mmg: back pointer to the unified memory manager
  * @refcount: reference counter for buffer users
  * @private: pointer to buffer behavior private data
@@ -776,14 +780,14 @@ struct hl_mmap_mem_buf_ops {
  * @handle: the buffer id in mmg handles store
  */
 struct hl_mmap_mem_buf {
-	struct hl_mmap_mem_buf_ops *ops;
+	struct hl_mmap_mem_buf_behavior *behavior;
 	struct hl_mem_mgr *mmg;
 	struct kref refcount;
 	void *private;
 	atomic_t mmap;
 	u64 real_mapped_size;
 	u64 mappable_size;
-	u32 handle;
+	u64 handle;
 };
 
 /**
@@ -3288,11 +3292,11 @@ void hl_mem_mgr_fini(struct hl_mem_mgr *mmg);
 int hl_mem_mgr_mmap(struct hl_mem_mgr *mmg, struct vm_area_struct *vma,
 		    void *args);
 struct hl_mmap_mem_buf *hl_mmap_mem_buf_get(struct hl_mem_mgr *mmg,
-						   u32 handle);
+						   u64 handle);
 int hl_mmap_mem_buf_put(struct hl_mmap_mem_buf *buf);
 struct hl_mmap_mem_buf *
 hl_mmap_mem_buf_alloc(struct hl_mem_mgr *mmg,
-		      struct hl_mmap_mem_buf_ops *behavior, gfp_t gfp,
+		      struct hl_mmap_mem_buf_behavior *behavior, gfp_t gfp,
 		      void *args);
 
 #ifdef CONFIG_DEBUG_FS
diff --git a/drivers/misc/habanalabs/common/memory.c b/drivers/misc/habanalabs/common/memory.c
index 6face45c57e3..e7a0c44c487d 100644
--- a/drivers/misc/habanalabs/common/memory.c
+++ b/drivers/misc/habanalabs/common/memory.c
@@ -2140,7 +2140,8 @@ static int hl_ts_alloc_buf(struct hl_mmap_mem_buf *buf, gfp_t gfp, void *args)
 	return -ENOMEM;
 }
 
-static struct hl_mmap_mem_buf_ops hl_ts_behavior = {
+static struct hl_mmap_mem_buf_behavior hl_ts_behavior = {
+	.mem_id = HL_MMAP_TYPE_TS_BUFF,
 	.mmap = hl_ts_mmap,
 	.alloc = hl_ts_alloc_buf,
 	.release = ts_buff_release,
@@ -2175,12 +2176,7 @@ static int allocate_timestamps_buffers(struct hl_fpriv *hpriv, struct hl_mem_in
 	if (!buf)
 		return -ENOMEM;
 
-	/* TODO:
-	 * Remove HL_MMAP_TYPE_TS_BUFF.
-	 * Embedding type in handle will no longer be needed as soon as we
-	 * switch to using a single memory manager for all memory types.
-	 */
-	*handle = ((u64)buf->handle | HL_MMAP_TYPE_TS_BUFF) << PAGE_SHIFT;
+	*handle = buf->handle;
 
 	return 0;
 }
diff --git a/drivers/misc/habanalabs/common/memory_mgr.c b/drivers/misc/habanalabs/common/memory_mgr.c
index 1bc23362e61d..1cc2f2ebe3b6 100644
--- a/drivers/misc/habanalabs/common/memory_mgr.c
+++ b/drivers/misc/habanalabs/common/memory_mgr.c
@@ -17,16 +17,16 @@
  * @return Find the buffer in the store and return a pointer to its descriptor.
  *         Increase buffer refcount. If not found - return NULL.
  */
-struct hl_mmap_mem_buf *hl_mmap_mem_buf_get(struct hl_mem_mgr *mmg, u32 handle)
+struct hl_mmap_mem_buf *hl_mmap_mem_buf_get(struct hl_mem_mgr *mmg, u64 handle)
 {
 	struct hl_mmap_mem_buf *buf;
 
 	spin_lock(&mmg->lock);
-	buf = idr_find(&mmg->handles, handle);
+	buf = idr_find(&mmg->handles, lower_32_bits(handle >> PAGE_SHIFT));
 	if (!buf) {
 		spin_unlock(&mmg->lock);
 		dev_warn(mmg->dev,
-			 "Buff get failed, no match to handle %u\n", handle);
+			 "Buff get failed, no match to handle %llu\n", handle);
 		return NULL;
 	}
 	kref_get(&buf->refcount);
@@ -51,8 +51,8 @@ static void hl_mmap_mem_buf_release(struct kref *kref)
 	idr_remove(&buf->mmg->handles, lower_32_bits(buf->handle >> PAGE_SHIFT));
 	spin_unlock(&buf->mmg->lock);
 
-	if (buf->ops->release)
-		buf->ops->release(buf);
+	if (buf->behavior->release)
+		buf->behavior->release(buf);
 
 	kfree(buf);
 }
@@ -83,7 +83,7 @@ int hl_mmap_mem_buf_put(struct hl_mmap_mem_buf *buf)
  */
 struct hl_mmap_mem_buf *
 hl_mmap_mem_buf_alloc(struct hl_mem_mgr *mmg,
-		      struct hl_mmap_mem_buf_ops *behavior, gfp_t gfp,
+		      struct hl_mmap_mem_buf_behavior *behavior, gfp_t gfp,
 		      void *args)
 {
 	struct hl_mmap_mem_buf *buf;
@@ -102,19 +102,18 @@ hl_mmap_mem_buf_alloc(struct hl_mem_mgr *mmg,
 		goto free_buf;
 	}
 
-	buf->handle = rc;
 	buf->mmg = mmg;
-	buf->ops = behavior;
+	buf->behavior = behavior;
+	buf->handle = (((u64)rc | buf->behavior->mem_id) << PAGE_SHIFT);
 	kref_init(&buf->refcount);
 
-	rc = buf->ops->alloc(buf, gfp, args);
+	rc = buf->behavior->alloc(buf, gfp, args);
 	if (rc) {
 		dev_err(mmg->dev, "Failure in buffer alloc callback %d\n",
 			rc);
 		goto remove_idr;
 	}
 
-	dev_dbg(mmg->dev, "Created buff object handle %u\n", buf->handle);
 	return buf;
 
 remove_idr:
@@ -169,20 +168,20 @@ int hl_mem_mgr_mmap(struct hl_mem_mgr *mmg, struct vm_area_struct *vma,
 {
 	struct hl_mmap_mem_buf *buf;
 	u64 user_mem_size;
-	u32 handle;
+	u64 handle;
 	int rc;
 
 	/* We use the page offset to hold the idr and thus we need to clear
 	 * it before doing the mmap itself
 	 */
-	handle = vma->vm_pgoff;
+	handle = vma->vm_pgoff << PAGE_SHIFT;
 	vma->vm_pgoff = 0;
 
 	/* Reference was taken here */
 	buf = hl_mmap_mem_buf_get(mmg, handle);
 	if (!buf) {
 		dev_err(mmg->dev,
-			"Memory mmap failed, no match to handle %u\n", handle);
+			"Memory mmap failed, no match to handle %llu\n", handle);
 		return -EINVAL;
 	}
 
@@ -223,7 +222,7 @@ int hl_mem_mgr_mmap(struct hl_mem_mgr *mmg, struct vm_area_struct *vma,
 
 	vma->vm_private_data = buf;
 
-	rc = buf->ops->mmap(buf, vma, args);
+	rc = buf->behavior->mmap(buf, vma, args);
 	if (rc) {
 		atomic_set(&buf->mmap, 0);
 		goto put_mem;
-- 
2.25.1


^ permalink raw reply related	[flat|nested] 6+ messages in thread
* [PATCH 2/6] habanalabs: add put by handle method to memory manager
  2022-05-02  8:05 [PATCH 1/6] habanalabs: hide memory manager page shift Oded Gabbay
@ 2022-05-02  8:05 ` Oded Gabbay
  2022-05-02  8:05 ` [PATCH 3/6] habanalabs/gaudi: set arbitration timeout to a high value Oded Gabbay
                   ` (3 subsequent siblings)
  4 siblings, 0 replies; 6+ messages in thread
From: Oded Gabbay @ 2022-05-02  8:05 UTC (permalink / raw)
  To: linux-kernel; +Cc: Yuri Nudelman

From: Yuri Nudelman <ynudelman@habana.ai>

Putting object by its handle and not by object pointer is useful in
some finalization flows that do not have object pointer available.
It eliminates the need to first get the object and then perform
put twice.

Signed-off-by: Yuri Nudelman <ynudelman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
 drivers/misc/habanalabs/common/habanalabs.h |  1 +
 drivers/misc/habanalabs/common/memory_mgr.c | 76 +++++++++++++++++++--
 2 files changed, 71 insertions(+), 6 deletions(-)

diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index ea5cfea21a2c..cfe349ebd0bb 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -3293,6 +3293,7 @@ int hl_mem_mgr_mmap(struct hl_mem_mgr *mmg, struct vm_area_struct *vma,
 		    void *args);
 struct hl_mmap_mem_buf *hl_mmap_mem_buf_get(struct hl_mem_mgr *mmg,
 						   u64 handle);
+int hl_mmap_mem_buf_put_handle(struct hl_mem_mgr *mmg, u64 handle);
 int hl_mmap_mem_buf_put(struct hl_mmap_mem_buf *buf);
 struct hl_mmap_mem_buf *
 hl_mmap_mem_buf_alloc(struct hl_mem_mgr *mmg,
diff --git a/drivers/misc/habanalabs/common/memory_mgr.c b/drivers/misc/habanalabs/common/memory_mgr.c
index 1cc2f2ebe3b6..9f3ab6cf25d3 100644
--- a/drivers/misc/habanalabs/common/memory_mgr.c
+++ b/drivers/misc/habanalabs/common/memory_mgr.c
@@ -14,8 +14,8 @@
  * @mmg: parent unifed memory manager
  * @handle: requested buffer handle
  *
- * @return Find the buffer in the store and return a pointer to its descriptor.
- *         Increase buffer refcount. If not found - return NULL.
+ * Find the buffer in the store and return a pointer to its descriptor.
+ * Increase buffer refcount. If not found - return NULL.
  */
 struct hl_mmap_mem_buf *hl_mmap_mem_buf_get(struct hl_mem_mgr *mmg, u64 handle)
 {
@@ -34,6 +34,23 @@ struct hl_mmap_mem_buf *hl_mmap_mem_buf_get(struct hl_mem_mgr *mmg, u64 handle)
 	return buf;
 }
 
+/**
+ * hl_mmap_mem_buf_destroy - destroy the unused buffer
+ *
+ * @buf: memory manager buffer descriptor
+ *
+ * Internal function, used as a final step of buffer release. Shall be invoked
+ * only when the buffer is no longer in use (removed from idr). Will call the
+ * release callback (if applicable), and free the memory.
+ */
+static void hl_mmap_mem_buf_destroy(struct hl_mmap_mem_buf *buf)
+{
+	if (buf->behavior->release)
+		buf->behavior->release(buf);
+
+	kfree(buf);
+}
+
 /**
  * hl_mmap_mem_buf_release - release buffer
  *
@@ -51,10 +68,23 @@ static void hl_mmap_mem_buf_release(struct kref *kref)
 	idr_remove(&buf->mmg->handles, lower_32_bits(buf->handle >> PAGE_SHIFT));
 	spin_unlock(&buf->mmg->lock);
 
-	if (buf->behavior->release)
-		buf->behavior->release(buf);
+	hl_mmap_mem_buf_destroy(buf);
+}
 
-	kfree(buf);
+/**
+ * hl_mmap_mem_buf_remove_idr_locked - remove handle from idr
+ *
+ * @kref: kref that reached 0.
+ *
+ * Internal function, used for kref put by handle. Assumes mmg lock is taken.
+ * Will remove the buffer from idr, without destroying it.
+ */
+static void hl_mmap_mem_buf_remove_idr_locked(struct kref *kref)
+{
+	struct hl_mmap_mem_buf *buf =
+		container_of(kref, struct hl_mmap_mem_buf, refcount);
+
+	idr_remove(&buf->mmg->handles, lower_32_bits(buf->handle >> PAGE_SHIFT));
 }
 
 /**
@@ -71,7 +101,41 @@ int hl_mmap_mem_buf_put(struct hl_mmap_mem_buf *buf)
 }
 
 /**
- * hl_mmap_mem_buf_alloc - allocate a new mappable buffer
+ * hl_mmap_mem_buf_put_handle - decrease the reference to the buffer with the
+ *                              given handle.
+ *
+ * @mmg: parent unifed memory manager
+ * @handle: requested buffer handle
+ *
+ * Decrease the reference to the buffer, and release it if it was the last one.
+ * Shall not be called from an interrupt context. Return -EINVAL if handle was
+ * not found, else return the put outcome (0 or 1).
+ */
+int hl_mmap_mem_buf_put_handle(struct hl_mem_mgr *mmg, u64 handle)
+{
+	struct hl_mmap_mem_buf *buf;
+
+	spin_lock(&mmg->lock);
+	buf = idr_find(&mmg->handles, lower_32_bits(handle >> PAGE_SHIFT));
+	if (!buf) {
+		spin_unlock(&mmg->lock);
+		dev_warn(mmg->dev,
+			 "Buff put failed, no match to handle %llu\n", handle);
+		return -EINVAL;
+	}
+
+	if (kref_put(&buf->refcount, hl_mmap_mem_buf_remove_idr_locked)) {
+		spin_unlock(&mmg->lock);
+		hl_mmap_mem_buf_destroy(buf);
+		return 1;
+	}
+
+	spin_unlock(&mmg->lock);
+	return 0;
+}
+
+/**
+ * @hl_mmap_mem_buf_alloc - allocate a new mappable buffer
  *
  * @mmg: parent unifed memory manager
  * @behavior: behavior object describing this buffer polymorphic behavior
-- 
2.25.1


^ permalink raw reply related	[flat|nested] 6+ messages in thread
* [PATCH 3/6] habanalabs/gaudi: set arbitration timeout to a high value
  2022-05-02  8:05 [PATCH 1/6] habanalabs: hide memory manager page shift Oded Gabbay
  2022-05-02  8:05 ` [PATCH 2/6] habanalabs: add put by handle method to memory manager Oded Gabbay
@ 2022-05-02  8:05 ` Oded Gabbay
  2022-05-02  8:05 ` [PATCH 4/6] habanalabs: unified memory manager new code for CB flow Oded Gabbay
                   ` (2 subsequent siblings)
  4 siblings, 0 replies; 6+ messages in thread
From: Oded Gabbay @ 2022-05-02  8:05 UTC (permalink / raw)
  To: linux-kernel; +Cc: Ofir Bitton

From: Ofir Bitton <obitton@habana.ai>

In certain workloads, arbitration timeout might expire although
no actual issue present. Hence, we set timeout to a very high value.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
 drivers/misc/habanalabs/gaudi/gaudi.c | 27 +++++++++++----------------
 1 file changed, 11 insertions(+), 16 deletions(-)

diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index 7828337eedce..e74ea4c2d008 100644
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -95,7 +95,7 @@
 
 #define GAUDI_NUM_OF_QM_ARB_ERR_CAUSE	3
 
-#define GAUDI_ARB_WDT_TIMEOUT		0x1000000
+#define GAUDI_ARB_WDT_TIMEOUT		0xEE6b27FF /* 8 seconds */
 
 #define GAUDI_CLK_GATE_DEBUGFS_MASK	(\
 		BIT(GAUDI_ENGINE_ID_MME_0) |\
@@ -2811,9 +2811,8 @@ static void gaudi_init_pci_dma_qman(struct hl_device *hdev, int dma_id,
 		WREG32(mmDMA0_QM_ARB_ERR_MSG_EN + dma_qm_offset,
 				QM_ARB_ERR_MSG_EN_MASK);
 
-		/* Increase ARB WDT to support streams architecture */
-		WREG32(mmDMA0_QM_ARB_SLV_CHOISE_WDT + dma_qm_offset,
-				GAUDI_ARB_WDT_TIMEOUT);
+		/* Set timeout to maximum */
+		WREG32(mmDMA0_QM_ARB_SLV_CHOISE_WDT + dma_qm_offset, GAUDI_ARB_WDT_TIMEOUT);
 
 		WREG32(mmDMA0_QM_GLBL_PROT + dma_qm_offset,
 				QMAN_EXTERNAL_MAKE_TRUSTED);
@@ -2990,9 +2989,8 @@ static void gaudi_init_hbm_dma_qman(struct hl_device *hdev, int dma_id,
 		WREG32(mmDMA0_QM_ARB_ERR_MSG_EN + dma_qm_offset,
 				QM_ARB_ERR_MSG_EN_MASK);
 
-		/* Increase ARB WDT to support streams architecture */
-		WREG32(mmDMA0_QM_ARB_SLV_CHOISE_WDT + dma_qm_offset,
-				GAUDI_ARB_WDT_TIMEOUT);
+		/* Set timeout to maximum */
+		WREG32(mmDMA0_QM_ARB_SLV_CHOISE_WDT + dma_qm_offset, GAUDI_ARB_WDT_TIMEOUT);
 
 		WREG32(mmDMA0_QM_GLBL_CFG1 + dma_qm_offset, 0);
 		WREG32(mmDMA0_QM_GLBL_PROT + dma_qm_offset,
@@ -3127,9 +3125,8 @@ static void gaudi_init_mme_qman(struct hl_device *hdev, u32 mme_offset,
 		WREG32(mmMME0_QM_ARB_ERR_MSG_EN + mme_offset,
 				QM_ARB_ERR_MSG_EN_MASK);
 
-		/* Increase ARB WDT to support streams architecture */
-		WREG32(mmMME0_QM_ARB_SLV_CHOISE_WDT + mme_offset,
-				GAUDI_ARB_WDT_TIMEOUT);
+		/* Set timeout to maximum */
+		WREG32(mmMME0_QM_ARB_SLV_CHOISE_WDT + mme_offset, GAUDI_ARB_WDT_TIMEOUT);
 
 		WREG32(mmMME0_QM_GLBL_CFG1 + mme_offset, 0);
 		WREG32(mmMME0_QM_GLBL_PROT + mme_offset,
@@ -3261,9 +3258,8 @@ static void gaudi_init_tpc_qman(struct hl_device *hdev, u32 tpc_offset,
 		WREG32(mmTPC0_QM_ARB_ERR_MSG_EN + tpc_offset,
 				QM_ARB_ERR_MSG_EN_MASK);
 
-		/* Increase ARB WDT to support streams architecture */
-		WREG32(mmTPC0_QM_ARB_SLV_CHOISE_WDT + tpc_offset,
-				GAUDI_ARB_WDT_TIMEOUT);
+		/* Set timeout to maximum */
+		WREG32(mmTPC0_QM_ARB_SLV_CHOISE_WDT + tpc_offset, GAUDI_ARB_WDT_TIMEOUT);
 
 		WREG32(mmTPC0_QM_GLBL_CFG1 + tpc_offset, 0);
 		WREG32(mmTPC0_QM_GLBL_PROT + tpc_offset,
@@ -3412,9 +3408,8 @@ static void gaudi_init_nic_qman(struct hl_device *hdev, u32 nic_offset,
 		WREG32(mmNIC0_QM0_ARB_ERR_MSG_EN + nic_offset,
 				QM_ARB_ERR_MSG_EN_MASK);
 
-		/* Increase ARB WDT to support streams architecture */
-		WREG32(mmNIC0_QM0_ARB_SLV_CHOISE_WDT + nic_offset,
-				GAUDI_ARB_WDT_TIMEOUT);
+		/* Set timeout to maximum */
+		WREG32(mmNIC0_QM0_ARB_SLV_CHOISE_WDT + nic_offset, GAUDI_ARB_WDT_TIMEOUT);
 
 		WREG32(mmNIC0_QM0_GLBL_CFG1 + nic_offset, 0);
 		WREG32(mmNIC0_QM0_GLBL_PROT + nic_offset,
-- 
2.25.1


^ permalink raw reply related	[flat|nested] 6+ messages in thread
* [PATCH 4/6] habanalabs: unified memory manager new code for CB flow
  2022-05-02  8:05 [PATCH 1/6] habanalabs: hide memory manager page shift Oded Gabbay
  2022-05-02  8:05 ` [PATCH 2/6] habanalabs: add put by handle method to memory manager Oded Gabbay
  2022-05-02  8:05 ` [PATCH 3/6] habanalabs/gaudi: set arbitration timeout to a high value Oded Gabbay
@ 2022-05-02  8:05 ` Oded Gabbay
  2022-05-02  8:05 ` [PATCH 5/6] habanalabs: use unified memory manager " Oded Gabbay
  2022-05-02  8:05 ` [PATCH 6/6] habanalabs: add device memory scrub ability through debugfs Oded Gabbay
  4 siblings, 0 replies; 6+ messages in thread
From: Oded Gabbay @ 2022-05-02  8:05 UTC (permalink / raw)
  To: linux-kernel; +Cc: Yuri Nudelman

From: Yuri Nudelman <ynudelman@habana.ai>

This commit adds the new code needed for command buffer flow using the
new unified memory manager, without changing the actual functionality.

Signed-off-by: Yuri Nudelman <ynudelman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
 .../misc/habanalabs/common/command_buffer.c   | 156 ++++++++++++++++++
 drivers/misc/habanalabs/common/habanalabs.h   |   8 +
 2 files changed, 164 insertions(+)

diff --git a/drivers/misc/habanalabs/common/command_buffer.c b/drivers/misc/habanalabs/common/command_buffer.c
index a507110f6443..a2a97f4876bd 100644
--- a/drivers/misc/habanalabs/common/command_buffer.c
+++ b/drivers/misc/habanalabs/common/command_buffer.c
@@ -378,6 +378,162 @@ int hl_cb_destroy(struct hl_device *hdev, struct hl_cb_mgr *mgr, u64 cb_handle)
 	return rc;
 }
 
+struct hl_cb_mmap_mem_alloc_args {
+	struct hl_device *hdev;
+	struct hl_ctx *ctx;
+	u32 cb_size;
+	bool internal_cb;
+	bool map_cb;
+};
+
+static void hl_cb_mmap_mem_release(struct hl_mmap_mem_buf *buf)
+{
+	struct hl_cb *cb = buf->private;
+
+	hl_debugfs_remove_cb(cb);
+
+	if (cb->is_mmu_mapped)
+		cb_unmap_mem(cb->ctx, cb);
+
+	hl_ctx_put(cb->ctx);
+
+	cb_do_release(cb->hdev, cb);
+}
+
+static int hl_cb_mmap_mem_alloc(struct hl_mmap_mem_buf *buf, gfp_t gfp, void *args)
+{
+	struct hl_cb_mmap_mem_alloc_args *cb_args = args;
+	struct hl_cb *cb;
+	int rc, ctx_id = cb_args->ctx->asid;
+	bool alloc_new_cb = true;
+
+	if (!cb_args->internal_cb) {
+		/* Minimum allocation must be PAGE SIZE */
+		if (cb_args->cb_size < PAGE_SIZE)
+			cb_args->cb_size = PAGE_SIZE;
+
+		if (ctx_id == HL_KERNEL_ASID_ID &&
+				cb_args->cb_size <= cb_args->hdev->asic_prop.cb_pool_cb_size) {
+
+			spin_lock(&cb_args->hdev->cb_pool_lock);
+			if (!list_empty(&cb_args->hdev->cb_pool)) {
+				cb = list_first_entry(&cb_args->hdev->cb_pool,
+						typeof(*cb), pool_list);
+				list_del(&cb->pool_list);
+				spin_unlock(&cb_args->hdev->cb_pool_lock);
+				alloc_new_cb = false;
+			} else {
+				spin_unlock(&cb_args->hdev->cb_pool_lock);
+				dev_dbg(cb_args->hdev->dev, "CB pool is empty\n");
+			}
+		}
+	}
+
+	if (alloc_new_cb) {
+		cb = hl_cb_alloc(cb_args->hdev, cb_args->cb_size, ctx_id, cb_args->internal_cb);
+		if (!cb)
+			return -ENOMEM;
+	}
+
+	cb->hdev = cb_args->hdev;
+	cb->ctx = cb_args->ctx;
+	cb->buf = buf;
+	cb->buf->mappable_size = cb->size;
+	cb->buf->private = cb;
+
+	hl_ctx_get(cb_args->hdev, cb->ctx);
+
+	if (cb_args->map_cb) {
+		if (ctx_id == HL_KERNEL_ASID_ID) {
+			dev_err(cb_args->hdev->dev,
+				"CB mapping is not supported for kernel context\n");
+			rc = -EINVAL;
+			goto release_cb;
+		}
+
+		rc = cb_map_mem(cb_args->ctx, cb);
+		if (rc)
+			goto release_cb;
+	}
+
+	hl_debugfs_add_cb(cb);
+
+	return 0;
+
+release_cb:
+	hl_ctx_put(cb->ctx);
+	cb_do_release(cb_args->hdev, cb);
+
+	return rc;
+}
+
+static int hl_cb_mmap_unified_mem_mgr(struct hl_mmap_mem_buf *buf,
+				      struct vm_area_struct *vma, void *args)
+{
+	struct hl_cb *cb = buf->private;
+
+	return cb->hdev->asic_funcs->mmap(cb->hdev, vma, cb->kernel_address,
+					cb->bus_address, cb->size);
+}
+
+static struct hl_mmap_mem_buf_behavior cb_behavior = {
+	.mem_id = HL_MMAP_TYPE_CB,
+	.alloc = hl_cb_mmap_mem_alloc,
+	.release = hl_cb_mmap_mem_release,
+	.mmap = hl_cb_mmap_unified_mem_mgr,
+};
+
+int hl_cb_create_unified_mem_mgr(struct hl_device *hdev, struct hl_mem_mgr *mmg,
+			struct hl_ctx *ctx, u32 cb_size, bool internal_cb,
+			bool map_cb, u64 *handle)
+{
+	struct hl_cb_mmap_mem_alloc_args args = {
+		.hdev = hdev,
+		.ctx = ctx,
+		.cb_size = cb_size,
+		.internal_cb = internal_cb,
+		.map_cb = map_cb,
+	};
+	struct hl_mmap_mem_buf *buf;
+	int ctx_id = ctx->asid;
+
+	if ((hdev->disabled) || (hdev->reset_info.in_reset && (ctx_id != HL_KERNEL_ASID_ID))) {
+		dev_warn_ratelimited(hdev->dev,
+			"Device is disabled or in reset. Can't create new CBs\n");
+		return -EBUSY;
+	}
+
+	if (cb_size > SZ_2M) {
+		dev_err(hdev->dev, "CB size %d must be less than %d\n",
+			cb_size, SZ_2M);
+		return -EINVAL;
+	}
+
+	buf = hl_mmap_mem_buf_alloc(
+		mmg, &cb_behavior,
+		ctx_id == HL_KERNEL_ASID_ID ? GFP_ATOMIC : GFP_KERNEL, &args);
+	if (!buf)
+		return -ENOMEM;
+
+	*handle = buf->handle;
+
+	return 0;
+}
+
+int hl_cb_destroy_unified_mem_mgr(struct hl_mem_mgr *mmg, u64 cb_handle)
+{
+	int rc;
+
+	rc = hl_mmap_mem_buf_put_handle(mmg, cb_handle);
+	if (rc < 0)
+		return rc; /* Invalid handle */
+
+	if (rc == 0)
+		dev_warn(mmg->dev, "CB 0x%llx is destroyed while still in use\n", cb_handle);
+
+	return 0;
+}
+
 static int hl_cb_info(struct hl_device *hdev, struct hl_cb_mgr *mgr,
 			u64 cb_handle, u32 flags, u32 *usage_cnt, u64 *device_va)
 {
diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index cfe349ebd0bb..47a5baf652b7 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -796,6 +796,7 @@ struct hl_mmap_mem_buf {
  * @hdev: pointer to device this CB belongs to.
  * @ctx: pointer to the CB owner's context.
  * @lock: spinlock to protect mmap flows.
+ * @buf: back pointer to the parent mappable memory buffer
  * @debugfs_list: node in debugfs list of command buffers.
  * @pool_list: node in pool list of command buffers.
  * @va_block_list: list of virtual addresses blocks of the CB if it is mapped to
@@ -816,6 +817,7 @@ struct hl_cb {
 	struct hl_device	*hdev;
 	struct hl_ctx		*ctx;
 	spinlock_t		lock;
+	struct hl_mmap_mem_buf	*buf;
 	struct list_head	debugfs_list;
 	struct list_head	pool_list;
 	struct list_head	va_block_list;
@@ -2662,6 +2664,7 @@ struct hl_reset_info {
  * @cs_mirror_list: CS mirror list for TDR.
  * @cs_mirror_lock: protects cs_mirror_list.
  * @kernel_cb_mgr: command buffer manager for creating/destroying/handling CBs.
+ * @kernel_mem_mgr: memory manager for memory buffers with lifespan of driver.
  * @event_queue: event queue for IRQ from CPU-CP.
  * @dma_pool: DMA pool for small allocations.
  * @cpu_accessible_dma_mem: Host <-> CPU-CP shared memory CPU address.
@@ -2794,6 +2797,7 @@ struct hl_device {
 	struct list_head		cs_mirror_list;
 	spinlock_t			cs_mirror_lock;
 	struct hl_cb_mgr		kernel_cb_mgr;
+	struct hl_mem_mgr		kernel_mem_mgr;
 	struct hl_eq			event_queue;
 	struct dma_pool			*dma_pool;
 	void				*cpu_accessible_dma_mem;
@@ -3098,6 +3102,10 @@ void hl_sysfs_fini(struct hl_device *hdev);
 int hl_hwmon_init(struct hl_device *hdev);
 void hl_hwmon_fini(struct hl_device *hdev);
 
+int hl_cb_create_unified_mem_mgr(struct hl_device *hdev, struct hl_mem_mgr *mmg,
+			struct hl_ctx *ctx, u32 cb_size, bool internal_cb,
+			bool map_cb, u64 *handle);
+int hl_cb_destroy_unified_mem_mgr(struct hl_mem_mgr *mmg, u64 cb_handle);
 int hl_cb_create(struct hl_device *hdev, struct hl_cb_mgr *mgr,
 			struct hl_ctx *ctx, u32 cb_size, bool internal_cb,
 			bool map_cb, u64 *handle);
-- 
2.25.1


^ permalink raw reply related	[flat|nested] 6+ messages in thread
* [PATCH 5/6] habanalabs: use unified memory manager for CB flow
  2022-05-02  8:05 [PATCH 1/6] habanalabs: hide memory manager page shift Oded Gabbay
                   ` (2 preceding siblings ...)
  2022-05-02  8:05 ` [PATCH 4/6] habanalabs: unified memory manager new code for CB flow Oded Gabbay
@ 2022-05-02  8:05 ` Oded Gabbay
  2022-05-02  8:05 ` [PATCH 6/6] habanalabs: add device memory scrub ability through debugfs Oded Gabbay
  4 siblings, 0 replies; 6+ messages in thread
From: Oded Gabbay @ 2022-05-02  8:05 UTC (permalink / raw)
  To: linux-kernel; +Cc: Yuri Nudelman

From: Yuri Nudelman <ynudelman@habana.ai>

With the new code required for the flow added, we can now switch
to using the new memory manager infrastructure, removing the old code.

Signed-off-by: Yuri Nudelman <ynudelman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
 .../misc/habanalabs/common/command_buffer.c   | 362 ++----------------
 .../habanalabs/common/command_submission.c    |  21 +-
 drivers/misc/habanalabs/common/debugfs.c      |   6 +-
 drivers/misc/habanalabs/common/device.c       |  12 +-
 drivers/misc/habanalabs/common/habanalabs.h   |  47 +--
 .../misc/habanalabs/common/habanalabs_drv.c   |   2 -
 drivers/misc/habanalabs/gaudi/gaudi.c         |  42 +-
 drivers/misc/habanalabs/goya/goya.c           |  34 +-
 8 files changed, 72 insertions(+), 454 deletions(-)

diff --git a/drivers/misc/habanalabs/common/command_buffer.c b/drivers/misc/habanalabs/common/command_buffer.c
index a2a97f4876bd..fd9ef32ea6a0 100644
--- a/drivers/misc/habanalabs/common/command_buffer.c
+++ b/drivers/misc/habanalabs/common/command_buffer.c
@@ -160,24 +160,6 @@ static void cb_do_release(struct hl_device *hdev, struct hl_cb *cb)
 	}
 }
 
-static void cb_release(struct kref *ref)
-{
-	struct hl_device *hdev;
-	struct hl_cb *cb;
-
-	cb = container_of(ref, struct hl_cb, refcount);
-	hdev = cb->hdev;
-
-	hl_debugfs_remove_cb(cb);
-
-	if (cb->is_mmu_mapped)
-		cb_unmap_mem(cb->ctx, cb);
-
-	hl_ctx_put(cb->ctx);
-
-	cb_do_release(hdev, cb);
-}
-
 static struct hl_cb *hl_cb_alloc(struct hl_device *hdev, u32 cb_size,
 					int ctx_id, bool internal_cb)
 {
@@ -238,146 +220,6 @@ static struct hl_cb *hl_cb_alloc(struct hl_device *hdev, u32 cb_size,
 	return cb;
 }
 
-int hl_cb_create(struct hl_device *hdev, struct hl_cb_mgr *mgr,
-			struct hl_ctx *ctx, u32 cb_size, bool internal_cb,
-			bool map_cb, u64 *handle)
-{
-	struct hl_cb *cb;
-	bool alloc_new_cb = true;
-	int rc, ctx_id = ctx->asid;
-
-	/*
-	 * Can't use generic function to check this because of special case
-	 * where we create a CB as part of the reset process
-	 */
-	if ((hdev->disabled) || (hdev->reset_info.in_reset && (ctx_id != HL_KERNEL_ASID_ID))) {
-		dev_warn_ratelimited(hdev->dev,
-			"Device is disabled or in reset. Can't create new CBs\n");
-		rc = -EBUSY;
-		goto out_err;
-	}
-
-	if (cb_size > SZ_2M) {
-		dev_err(hdev->dev, "CB size %d must be less than %d\n",
-			cb_size, SZ_2M);
-		rc = -EINVAL;
-		goto out_err;
-	}
-
-	if (!internal_cb) {
-		/* Minimum allocation must be PAGE SIZE */
-		if (cb_size < PAGE_SIZE)
-			cb_size = PAGE_SIZE;
-
-		if (ctx_id == HL_KERNEL_ASID_ID &&
-				cb_size <= hdev->asic_prop.cb_pool_cb_size) {
-
-			spin_lock(&hdev->cb_pool_lock);
-			if (!list_empty(&hdev->cb_pool)) {
-				cb = list_first_entry(&hdev->cb_pool,
-						typeof(*cb), pool_list);
-				list_del(&cb->pool_list);
-				spin_unlock(&hdev->cb_pool_lock);
-				alloc_new_cb = false;
-			} else {
-				spin_unlock(&hdev->cb_pool_lock);
-				dev_dbg(hdev->dev, "CB pool is empty\n");
-			}
-		}
-	}
-
-	if (alloc_new_cb) {
-		cb = hl_cb_alloc(hdev, cb_size, ctx_id, internal_cb);
-		if (!cb) {
-			rc = -ENOMEM;
-			goto out_err;
-		}
-	}
-
-	cb->hdev = hdev;
-	cb->ctx = ctx;
-	hl_ctx_get(hdev, cb->ctx);
-
-	if (map_cb) {
-		if (ctx_id == HL_KERNEL_ASID_ID) {
-			dev_err(hdev->dev,
-				"CB mapping is not supported for kernel context\n");
-			rc = -EINVAL;
-			goto release_cb;
-		}
-
-		rc = cb_map_mem(ctx, cb);
-		if (rc)
-			goto release_cb;
-	}
-
-	spin_lock(&mgr->cb_lock);
-	rc = idr_alloc(&mgr->cb_handles, cb, 1, 0, GFP_ATOMIC);
-	spin_unlock(&mgr->cb_lock);
-
-	if (rc < 0) {
-		dev_err(hdev->dev, "Failed to allocate IDR for a new CB\n");
-		goto unmap_mem;
-	}
-
-	cb->id = (u64) rc;
-
-	kref_init(&cb->refcount);
-	spin_lock_init(&cb->lock);
-
-	/*
-	 * idr is 32-bit so we can safely OR it with a mask that is above
-	 * 32 bit
-	 */
-	*handle = cb->id | HL_MMAP_TYPE_CB;
-	*handle <<= PAGE_SHIFT;
-
-	hl_debugfs_add_cb(cb);
-
-	return 0;
-
-unmap_mem:
-	if (cb->is_mmu_mapped)
-		cb_unmap_mem(cb->ctx, cb);
-release_cb:
-	hl_ctx_put(cb->ctx);
-	cb_do_release(hdev, cb);
-out_err:
-	*handle = 0;
-
-	return rc;
-}
-
-int hl_cb_destroy(struct hl_device *hdev, struct hl_cb_mgr *mgr, u64 cb_handle)
-{
-	struct hl_cb *cb;
-	u32 handle;
-	int rc = 0;
-
-	/*
-	 * handle was given to user to do mmap, I need to shift it back to
-	 * how the idr module gave it to me
-	 */
-	cb_handle >>= PAGE_SHIFT;
-	handle = (u32) cb_handle;
-
-	spin_lock(&mgr->cb_lock);
-
-	cb = idr_find(&mgr->cb_handles, handle);
-	if (cb) {
-		idr_remove(&mgr->cb_handles, handle);
-		spin_unlock(&mgr->cb_lock);
-		kref_put(&cb->refcount, cb_release);
-	} else {
-		spin_unlock(&mgr->cb_lock);
-		dev_err(hdev->dev,
-			"CB destroy failed, no match to handle 0x%x\n", handle);
-		rc = -EINVAL;
-	}
-
-	return rc;
-}
-
 struct hl_cb_mmap_mem_alloc_args {
 	struct hl_device *hdev;
 	struct hl_ctx *ctx;
@@ -467,7 +309,7 @@ static int hl_cb_mmap_mem_alloc(struct hl_mmap_mem_buf *buf, gfp_t gfp, void *ar
 	return rc;
 }
 
-static int hl_cb_mmap_unified_mem_mgr(struct hl_mmap_mem_buf *buf,
+static int hl_cb_mmap(struct hl_mmap_mem_buf *buf,
 				      struct vm_area_struct *vma, void *args)
 {
 	struct hl_cb *cb = buf->private;
@@ -480,10 +322,10 @@ static struct hl_mmap_mem_buf_behavior cb_behavior = {
 	.mem_id = HL_MMAP_TYPE_CB,
 	.alloc = hl_cb_mmap_mem_alloc,
 	.release = hl_cb_mmap_mem_release,
-	.mmap = hl_cb_mmap_unified_mem_mgr,
+	.mmap = hl_cb_mmap,
 };
 
-int hl_cb_create_unified_mem_mgr(struct hl_device *hdev, struct hl_mem_mgr *mmg,
+int hl_cb_create(struct hl_device *hdev, struct hl_mem_mgr *mmg,
 			struct hl_ctx *ctx, u32 cb_size, bool internal_cb,
 			bool map_cb, u64 *handle)
 {
@@ -520,7 +362,7 @@ int hl_cb_create_unified_mem_mgr(struct hl_device *hdev, struct hl_mem_mgr *mmg,
 	return 0;
 }
 
-int hl_cb_destroy_unified_mem_mgr(struct hl_mem_mgr *mmg, u64 cb_handle)
+int hl_cb_destroy(struct hl_mem_mgr *mmg, u64 cb_handle)
 {
 	int rc;
 
@@ -529,33 +371,23 @@ int hl_cb_destroy_unified_mem_mgr(struct hl_mem_mgr *mmg, u64 cb_handle)
 		return rc; /* Invalid handle */
 
 	if (rc == 0)
-		dev_warn(mmg->dev, "CB 0x%llx is destroyed while still in use\n", cb_handle);
+		dev_dbg(mmg->dev, "CB 0x%llx is destroyed while still in use\n", cb_handle);
 
 	return 0;
 }
 
-static int hl_cb_info(struct hl_device *hdev, struct hl_cb_mgr *mgr,
-			u64 cb_handle, u32 flags, u32 *usage_cnt, u64 *device_va)
+static int hl_cb_info(struct hl_mem_mgr *mmg,
+			u64 handle, u32 flags, u32 *usage_cnt, u64 *device_va)
 {
 	struct hl_vm_va_block *va_block;
 	struct hl_cb *cb;
-	u32 handle;
 	int rc = 0;
 
-	/* The CB handle was given to user to do mmap, so need to shift it back
-	 * to the value which was allocated by the IDR module.
-	 */
-	cb_handle >>= PAGE_SHIFT;
-	handle = (u32) cb_handle;
-
-	spin_lock(&mgr->cb_lock);
-
-	cb = idr_find(&mgr->cb_handles, handle);
+	cb = hl_cb_get(mmg, handle);
 	if (!cb) {
-		dev_err(hdev->dev,
-			"CB info failed, no match to handle 0x%x\n", handle);
-		rc = -EINVAL;
-		goto out;
+		dev_err(mmg->dev,
+			"CB info failed, no match to handle 0x%llx\n", handle);
+		return -EINVAL;
 	}
 
 	if (flags & HL_CB_FLAGS_GET_DEVICE_VA) {
@@ -563,7 +395,7 @@ static int hl_cb_info(struct hl_device *hdev, struct hl_cb_mgr *mgr,
 		if (va_block) {
 			*device_va = va_block->start;
 		} else {
-			dev_err(hdev->dev, "CB is not mapped to the device's MMU\n");
+			dev_err(mmg->dev, "CB is not mapped to the device's MMU\n");
 			rc = -EINVAL;
 			goto out;
 		}
@@ -572,7 +404,7 @@ static int hl_cb_info(struct hl_device *hdev, struct hl_cb_mgr *mgr,
 	}
 
 out:
-	spin_unlock(&mgr->cb_lock);
+	hl_cb_put(cb);
 	return rc;
 }
 
@@ -600,7 +432,7 @@ int hl_cb_ioctl(struct hl_fpriv *hpriv, void *data)
 				args->in.cb_size, HL_MAX_CB_SIZE);
 			rc = -EINVAL;
 		} else {
-			rc = hl_cb_create(hdev, &hpriv->cb_mgr, hpriv->ctx,
+			rc = hl_cb_create(hdev, &hpriv->mem_mgr, hpriv->ctx,
 					args->in.cb_size, false,
 					!!(args->in.flags & HL_CB_FLAGS_MAP),
 					&handle);
@@ -611,12 +443,12 @@ int hl_cb_ioctl(struct hl_fpriv *hpriv, void *data)
 		break;
 
 	case HL_CB_OP_DESTROY:
-		rc = hl_cb_destroy(hdev, &hpriv->cb_mgr,
+		rc = hl_cb_destroy(&hpriv->mem_mgr,
 					args->in.cb_handle);
 		break;
 
 	case HL_CB_OP_INFO:
-		rc = hl_cb_info(hdev, &hpriv->cb_mgr, args->in.cb_handle,
+		rc = hl_cb_info(&hpriv->mem_mgr, args->in.cb_handle,
 				args->in.flags,
 				&usage_cnt,
 				&device_va);
@@ -639,163 +471,20 @@ int hl_cb_ioctl(struct hl_fpriv *hpriv, void *data)
 	return rc;
 }
 
-static void cb_vm_close(struct vm_area_struct *vma)
-{
-	struct hl_cb *cb = (struct hl_cb *) vma->vm_private_data;
-	long new_mmap_size;
-
-	new_mmap_size = cb->mmap_size - (vma->vm_end - vma->vm_start);
-
-	if (new_mmap_size > 0) {
-		cb->mmap_size = new_mmap_size;
-		return;
-	}
-
-	spin_lock(&cb->lock);
-	cb->mmap = false;
-	spin_unlock(&cb->lock);
-
-	hl_cb_put(cb);
-	vma->vm_private_data = NULL;
-}
-
-static const struct vm_operations_struct cb_vm_ops = {
-	.close = cb_vm_close
-};
-
-int hl_cb_mmap(struct hl_fpriv *hpriv, struct vm_area_struct *vma)
+struct hl_cb *hl_cb_get(struct hl_mem_mgr *mmg, u64 handle)
 {
-	struct hl_device *hdev = hpriv->hdev;
-	struct hl_cb *cb;
-	u32 handle, user_cb_size;
-	int rc;
-
-	/* We use the page offset to hold the idr and thus we need to clear
-	 * it before doing the mmap itself
-	 */
-	handle = vma->vm_pgoff;
-	vma->vm_pgoff = 0;
-
-	/* reference was taken here */
-	cb = hl_cb_get(hdev, &hpriv->cb_mgr, handle);
-	if (!cb) {
-		dev_err(hdev->dev,
-			"CB mmap failed, no match to handle 0x%x\n", handle);
-		return -EINVAL;
-	}
-
-	/* Validation check */
-	user_cb_size = vma->vm_end - vma->vm_start;
-	if (user_cb_size != ALIGN(cb->size, PAGE_SIZE)) {
-		dev_err(hdev->dev,
-			"CB mmap failed, mmap size 0x%lx != 0x%x cb size\n",
-			vma->vm_end - vma->vm_start, cb->size);
-		rc = -EINVAL;
-		goto put_cb;
-	}
-
-	if (!access_ok((void __user *) (uintptr_t) vma->vm_start,
-							user_cb_size)) {
-		dev_err(hdev->dev,
-			"user pointer is invalid - 0x%lx\n",
-			vma->vm_start);
-
-		rc = -EINVAL;
-		goto put_cb;
-	}
-
-	spin_lock(&cb->lock);
-
-	if (cb->mmap) {
-		dev_err(hdev->dev,
-			"CB mmap failed, CB already mmaped to user\n");
-		rc = -EINVAL;
-		goto release_lock;
-	}
-
-	cb->mmap = true;
-
-	spin_unlock(&cb->lock);
-
-	vma->vm_ops = &cb_vm_ops;
-
-	/*
-	 * Note: We're transferring the cb reference to
-	 * vma->vm_private_data here.
-	 */
-
-	vma->vm_private_data = cb;
-
-	rc = hdev->asic_funcs->mmap(hdev, vma, cb->kernel_address,
-					cb->bus_address, cb->size);
-	if (rc) {
-		spin_lock(&cb->lock);
-		cb->mmap = false;
-		goto release_lock;
-	}
-
-	cb->mmap_size = cb->size;
-	vma->vm_pgoff = handle;
-
-	return 0;
-
-release_lock:
-	spin_unlock(&cb->lock);
-put_cb:
-	hl_cb_put(cb);
-	return rc;
-}
-
-struct hl_cb *hl_cb_get(struct hl_device *hdev, struct hl_cb_mgr *mgr,
-			u32 handle)
-{
-	struct hl_cb *cb;
-
-	spin_lock(&mgr->cb_lock);
-	cb = idr_find(&mgr->cb_handles, handle);
+	struct hl_mmap_mem_buf *buf;
 
-	if (!cb) {
-		spin_unlock(&mgr->cb_lock);
-		dev_warn(hdev->dev,
-			"CB get failed, no match to handle 0x%x\n", handle);
+	buf = hl_mmap_mem_buf_get(mmg, handle);
+	if (!buf)
 		return NULL;
-	}
-
-	kref_get(&cb->refcount);
-
-	spin_unlock(&mgr->cb_lock);
-
-	return cb;
+	return buf->private;
 
 }
 
 void hl_cb_put(struct hl_cb *cb)
 {
-	kref_put(&cb->refcount, cb_release);
-}
-
-void hl_cb_mgr_init(struct hl_cb_mgr *mgr)
-{
-	spin_lock_init(&mgr->cb_lock);
-	idr_init(&mgr->cb_handles);
-}
-
-void hl_cb_mgr_fini(struct hl_device *hdev, struct hl_cb_mgr *mgr)
-{
-	struct hl_cb *cb;
-	struct idr *idp;
-	u32 id;
-
-	idp = &mgr->cb_handles;
-
-	idr_for_each_entry(idp, cb, id) {
-		if (kref_put(&cb->refcount, cb_release) != 1)
-			dev_err(hdev->dev,
-				"CB %d for CTX ID %d is still alive\n",
-				id, cb->ctx->asid);
-	}
-
-	idr_destroy(&mgr->cb_handles);
+	hl_mmap_mem_buf_put(cb->buf);
 }
 
 struct hl_cb *hl_cb_kernel_create(struct hl_device *hdev, u32 cb_size,
@@ -805,7 +494,7 @@ struct hl_cb *hl_cb_kernel_create(struct hl_device *hdev, u32 cb_size,
 	struct hl_cb *cb;
 	int rc;
 
-	rc = hl_cb_create(hdev, &hdev->kernel_cb_mgr, hdev->kernel_ctx, cb_size,
+	rc = hl_cb_create(hdev, &hdev->kernel_mem_mgr, hdev->kernel_ctx, cb_size,
 				internal_cb, false, &cb_handle);
 	if (rc) {
 		dev_err(hdev->dev,
@@ -813,8 +502,7 @@ struct hl_cb *hl_cb_kernel_create(struct hl_device *hdev, u32 cb_size,
 		return NULL;
 	}
 
-	cb_handle >>= PAGE_SHIFT;
-	cb = hl_cb_get(hdev, &hdev->kernel_cb_mgr, (u32) cb_handle);
+	cb = hl_cb_get(&hdev->kernel_mem_mgr, cb_handle);
 	/* hl_cb_get should never fail here */
 	if (!cb) {
 		dev_crit(hdev->dev, "Kernel CB handle invalid 0x%x\n",
@@ -825,7 +513,7 @@ struct hl_cb *hl_cb_kernel_create(struct hl_device *hdev, u32 cb_size,
 	return cb;
 
 destroy_cb:
-	hl_cb_destroy(hdev, &hdev->kernel_cb_mgr, cb_handle << PAGE_SHIFT);
+	hl_cb_destroy(&hdev->kernel_mem_mgr, cb_handle);
 
 	return NULL;
 }
diff --git a/drivers/misc/habanalabs/common/command_submission.c b/drivers/misc/habanalabs/common/command_submission.c
index a189157be104..41206c5d6e7b 100644
--- a/drivers/misc/habanalabs/common/command_submission.c
+++ b/drivers/misc/habanalabs/common/command_submission.c
@@ -1071,17 +1071,14 @@ static int validate_queue_index(struct hl_device *hdev,
 }
 
 static struct hl_cb *get_cb_from_cs_chunk(struct hl_device *hdev,
-					struct hl_cb_mgr *cb_mgr,
+					struct hl_mem_mgr *mmg,
 					struct hl_cs_chunk *chunk)
 {
 	struct hl_cb *cb;
-	u32 cb_handle;
 
-	cb_handle = (u32) (chunk->cb_handle >> PAGE_SHIFT);
-
-	cb = hl_cb_get(hdev, cb_mgr, cb_handle);
+	cb = hl_cb_get(mmg, chunk->cb_handle);
 	if (!cb) {
-		dev_err(hdev->dev, "CB handle 0x%x invalid\n", cb_handle);
+		dev_err(hdev->dev, "CB handle 0x%llx invalid\n", chunk->cb_handle);
 		return NULL;
 	}
 
@@ -1343,7 +1340,7 @@ static int cs_ioctl_default(struct hl_fpriv *hpriv, void __user *chunks,
 		}
 
 		if (is_kernel_allocated_cb) {
-			cb = get_cb_from_cs_chunk(hdev, &hpriv->cb_mgr, chunk);
+			cb = get_cb_from_cs_chunk(hdev, &hpriv->mem_mgr, chunk);
 			if (!cb) {
 				atomic64_inc(
 					&ctx->cs_counters.validation_drop_cnt);
@@ -1771,7 +1768,7 @@ static int cs_ioctl_signal_wait_create_jobs(struct hl_device *hdev,
 	 */
 	job->patched_cb = job->user_cb;
 	job->job_cb_size = job->user_cb_size;
-	hl_cb_destroy(hdev, &hdev->kernel_cb_mgr, cb->id << PAGE_SHIFT);
+	hl_cb_destroy(&hdev->kernel_mem_mgr, cb->buf->handle);
 
 	/* increment refcount as for external queues we get completion */
 	cs_get(cs);
@@ -2946,13 +2943,12 @@ static int ts_buff_get_kernel_ts_record(struct hl_mmap_mem_buf *buf,
 }
 
 static int _hl_interrupt_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
-				struct hl_cb_mgr *cb_mgr, struct hl_mem_mgr *mmg,
+				struct hl_mem_mgr *cb_mmg, struct hl_mem_mgr *mmg,
 				u64 timeout_us, u64 cq_counters_handle,	u64 cq_counters_offset,
 				u64 target_value, struct hl_user_interrupt *interrupt,
 				bool register_ts_record, u64 ts_handle, u64 ts_offset,
 				u32 *status, u64 *timestamp)
 {
-	u32 cq_patched_handle;
 	struct hl_user_pending_interrupt *pend;
 	struct hl_mmap_mem_buf *buf;
 	struct hl_cb *cq_cb;
@@ -2964,8 +2960,7 @@ static int _hl_interrupt_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
 
 	hl_ctx_get(hdev, ctx);
 
-	cq_patched_handle = lower_32_bits(cq_counters_handle >> PAGE_SHIFT);
-	cq_cb = hl_cb_get(hdev, cb_mgr, cq_patched_handle);
+	cq_cb = hl_cb_get(cb_mmg, cq_counters_handle);
 	if (!cq_cb) {
 		rc = -EINVAL;
 		goto put_ctx;
@@ -3250,7 +3245,7 @@ static int hl_interrupt_wait_ioctl(struct hl_fpriv *hpriv, void *data)
 		interrupt = &hdev->user_interrupt[interrupt_id - first_interrupt];
 
 	if (args->in.flags & HL_WAIT_CS_FLAGS_INTERRUPT_KERNEL_CQ)
-		rc = _hl_interrupt_wait_ioctl(hdev, hpriv->ctx, &hpriv->cb_mgr, &hpriv->mem_mgr,
+		rc = _hl_interrupt_wait_ioctl(hdev, hpriv->ctx, &hpriv->mem_mgr, &hpriv->mem_mgr,
 				args->in.interrupt_timeout_us, args->in.cq_counters_handle,
 				args->in.cq_counters_offset,
 				args->in.target, interrupt,
diff --git a/drivers/misc/habanalabs/common/debugfs.c b/drivers/misc/habanalabs/common/debugfs.c
index a94f01713efd..a9c4f2d4139d 100644
--- a/drivers/misc/habanalabs/common/debugfs.c
+++ b/drivers/misc/habanalabs/common/debugfs.c
@@ -126,9 +126,9 @@ static int command_buffers_show(struct seq_file *s, void *data)
 		}
 		seq_printf(s,
 			"   %03llu        %d    0x%08x      %d          %d          %d\n",
-			cb->id, cb->ctx->asid, cb->size,
-			kref_read(&cb->refcount),
-			cb->mmap, atomic_read(&cb->cs_cnt));
+			cb->buf->handle, cb->ctx->asid, cb->size,
+			kref_read(&cb->buf->refcount),
+			atomic_read(&cb->buf->mmap), atomic_read(&cb->cs_cnt));
 	}
 
 	spin_unlock(&dev_entry->cb_spinlock);
diff --git a/drivers/misc/habanalabs/common/device.c b/drivers/misc/habanalabs/common/device.c
index bd74e0359473..dbec98736a31 100644
--- a/drivers/misc/habanalabs/common/device.c
+++ b/drivers/misc/habanalabs/common/device.c
@@ -324,7 +324,6 @@ static int hl_device_release(struct inode *inode, struct file *filp)
 	 */
 	hl_release_pending_user_interrupts(hpriv->hdev);
 
-	hl_cb_mgr_fini(hdev, &hpriv->cb_mgr);
 	hl_mem_mgr_fini(&hpriv->mem_mgr);
 	hl_ctx_mgr_fini(hdev, &hpriv->ctx_mgr);
 
@@ -386,14 +385,11 @@ static int hl_mmap(struct file *filp, struct vm_area_struct *vma)
 	vm_pgoff = vma->vm_pgoff;
 
 	switch (vm_pgoff & HL_MMAP_TYPE_MASK) {
-	case HL_MMAP_TYPE_CB:
-		vma->vm_pgoff = HL_MMAP_OFFSET_VALUE_GET(vm_pgoff);
-		return hl_cb_mmap(hpriv, vma);
-
 	case HL_MMAP_TYPE_BLOCK:
 		vma->vm_pgoff = HL_MMAP_OFFSET_VALUE_GET(vm_pgoff);
 		return hl_hw_block_mmap(hpriv, vma);
 
+	case HL_MMAP_TYPE_CB:
 	case HL_MMAP_TYPE_TS_BUFF:
 		return hl_mem_mgr_mmap(&hpriv->mem_mgr, vma, NULL);
 	}
@@ -616,7 +612,7 @@ static int device_early_init(struct hl_device *hdev)
 	if (rc)
 		goto free_chip_info;
 
-	hl_cb_mgr_init(&hdev->kernel_cb_mgr);
+	hl_mem_mgr_init(hdev->dev, &hdev->kernel_mem_mgr);
 
 	hdev->device_reset_work.wq =
 			create_singlethread_workqueue("hl_device_reset");
@@ -645,7 +641,7 @@ static int device_early_init(struct hl_device *hdev)
 	return 0;
 
 free_cb_mgr:
-	hl_cb_mgr_fini(hdev, &hdev->kernel_cb_mgr);
+	hl_mem_mgr_fini(&hdev->kernel_mem_mgr);
 free_chip_info:
 	kfree(hdev->hl_chip_info);
 free_ts_free_wq:
@@ -684,7 +680,7 @@ static void device_early_fini(struct hl_device *hdev)
 
 	mutex_destroy(&hdev->clk_throttling.lock);
 
-	hl_cb_mgr_fini(hdev, &hdev->kernel_cb_mgr);
+	hl_mem_mgr_fini(&hdev->kernel_mem_mgr);
 
 	kfree(hdev->hl_chip_info);
 
diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index 47a5baf652b7..496d61ee07c5 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -703,26 +703,6 @@ struct hl_cs_compl {
  * Command Buffers
  */
 
-/**
- * struct hl_cb_mgr - describes a Command Buffer Manager.
- * @cb_lock: protects cb_handles.
- * @cb_handles: an idr to hold all command buffer handles.
- */
-struct hl_cb_mgr {
-	spinlock_t		cb_lock;
-	struct idr		cb_handles; /* protected by cb_lock */
-};
-
-/**
- * struct hl_ts_mgr - describes the timestamp registration memory manager.
- * @ts_lock: protects ts_handles.
- * @ts_handles: an idr to hold all ts bufferes handles.
- */
-struct hl_ts_mgr {
-	spinlock_t		ts_lock;
-	struct idr		ts_handles;
-};
-
 /**
  * struct hl_ts_buff - describes a timestamp buffer.
  * @kernel_buff_address: Holds the internal buffer's kernel virtual address.
@@ -792,42 +772,32 @@ struct hl_mmap_mem_buf {
 
 /**
  * struct hl_cb - describes a Command Buffer.
- * @refcount: reference counter for usage of the CB.
  * @hdev: pointer to device this CB belongs to.
  * @ctx: pointer to the CB owner's context.
- * @lock: spinlock to protect mmap flows.
  * @buf: back pointer to the parent mappable memory buffer
  * @debugfs_list: node in debugfs list of command buffers.
  * @pool_list: node in pool list of command buffers.
  * @va_block_list: list of virtual addresses blocks of the CB if it is mapped to
  *                 the device's MMU.
- * @id: the CB's ID.
  * @kernel_address: Holds the CB's kernel virtual address.
  * @bus_address: Holds the CB's DMA address.
- * @mmap_size: Holds the CB's size that was mmaped.
  * @size: holds the CB's size.
  * @cs_cnt: holds number of CS that this CB participates in.
- * @mmap: true if the CB is currently mmaped to user.
  * @is_pool: true if CB was acquired from the pool, false otherwise.
  * @is_internal: internaly allocated
  * @is_mmu_mapped: true if the CB is mapped to the device's MMU.
  */
 struct hl_cb {
-	struct kref		refcount;
 	struct hl_device	*hdev;
 	struct hl_ctx		*ctx;
-	spinlock_t		lock;
 	struct hl_mmap_mem_buf	*buf;
 	struct list_head	debugfs_list;
 	struct list_head	pool_list;
 	struct list_head	va_block_list;
-	u64			id;
 	void			*kernel_address;
 	dma_addr_t		bus_address;
-	u32			mmap_size;
 	u32			size;
 	atomic_t		cs_cnt;
-	u8			mmap;
 	u8			is_pool;
 	u8			is_internal;
 	u8			is_mmu_mapped;
@@ -1982,7 +1952,6 @@ struct hl_fpriv {
 	struct pid		*taskpid;
 	struct hl_ctx		*ctx;
 	struct hl_ctx_mgr	ctx_mgr;
-	struct hl_cb_mgr	cb_mgr;
 	struct hl_mem_mgr	mem_mgr;
 	struct list_head	debugfs_list;
 	struct list_head	dev_node;
@@ -2663,7 +2632,6 @@ struct hl_reset_info {
  * @kernel_queues: array of hl_hw_queue.
  * @cs_mirror_list: CS mirror list for TDR.
  * @cs_mirror_lock: protects cs_mirror_list.
- * @kernel_cb_mgr: command buffer manager for creating/destroying/handling CBs.
  * @kernel_mem_mgr: memory manager for memory buffers with lifespan of driver.
  * @event_queue: event queue for IRQ from CPU-CP.
  * @dma_pool: DMA pool for small allocations.
@@ -2796,7 +2764,6 @@ struct hl_device {
 	struct hl_hw_queue		*kernel_queues;
 	struct list_head		cs_mirror_list;
 	spinlock_t			cs_mirror_lock;
-	struct hl_cb_mgr		kernel_cb_mgr;
 	struct hl_mem_mgr		kernel_mem_mgr;
 	struct hl_eq			event_queue;
 	struct dma_pool			*dma_pool;
@@ -3102,21 +3069,13 @@ void hl_sysfs_fini(struct hl_device *hdev);
 int hl_hwmon_init(struct hl_device *hdev);
 void hl_hwmon_fini(struct hl_device *hdev);
 
-int hl_cb_create_unified_mem_mgr(struct hl_device *hdev, struct hl_mem_mgr *mmg,
-			struct hl_ctx *ctx, u32 cb_size, bool internal_cb,
-			bool map_cb, u64 *handle);
-int hl_cb_destroy_unified_mem_mgr(struct hl_mem_mgr *mmg, u64 cb_handle);
-int hl_cb_create(struct hl_device *hdev, struct hl_cb_mgr *mgr,
+int hl_cb_create(struct hl_device *hdev, struct hl_mem_mgr *mmg,
 			struct hl_ctx *ctx, u32 cb_size, bool internal_cb,
 			bool map_cb, u64 *handle);
-int hl_cb_destroy(struct hl_device *hdev, struct hl_cb_mgr *mgr, u64 cb_handle);
-int hl_cb_mmap(struct hl_fpriv *hpriv, struct vm_area_struct *vma);
+int hl_cb_destroy(struct hl_mem_mgr *mmg, u64 cb_handle);
 int hl_hw_block_mmap(struct hl_fpriv *hpriv, struct vm_area_struct *vma);
-struct hl_cb *hl_cb_get(struct hl_device *hdev,	struct hl_cb_mgr *mgr,
-			u32 handle);
+struct hl_cb *hl_cb_get(struct hl_mem_mgr *mmg, u64 handle);
 void hl_cb_put(struct hl_cb *cb);
-void hl_cb_mgr_init(struct hl_cb_mgr *mgr);
-void hl_cb_mgr_fini(struct hl_device *hdev, struct hl_cb_mgr *mgr);
 struct hl_cb *hl_cb_kernel_create(struct hl_device *hdev, u32 cb_size,
 					bool internal_cb);
 int hl_cb_pool_init(struct hl_device *hdev);
diff --git a/drivers/misc/habanalabs/common/habanalabs_drv.c b/drivers/misc/habanalabs/common/habanalabs_drv.c
index 57218e03c222..70203433e6cd 100644
--- a/drivers/misc/habanalabs/common/habanalabs_drv.c
+++ b/drivers/misc/habanalabs/common/habanalabs_drv.c
@@ -138,7 +138,6 @@ int hl_device_open(struct inode *inode, struct file *filp)
 	kref_init(&hpriv->refcount);
 	nonseekable_open(inode, filp);
 
-	hl_cb_mgr_init(&hpriv->cb_mgr);
 	hl_ctx_mgr_init(&hpriv->ctx_mgr);
 	hl_mem_mgr_init(hpriv->hdev->dev, &hpriv->mem_mgr);
 
@@ -197,7 +196,6 @@ int hl_device_open(struct inode *inode, struct file *filp)
 
 out_err:
 	mutex_unlock(&hdev->fpriv_list_lock);
-	hl_cb_mgr_fini(hpriv->hdev, &hpriv->cb_mgr);
 	hl_mem_mgr_fini(&hpriv->mem_mgr);
 	hl_ctx_mgr_fini(hpriv->hdev, &hpriv->ctx_mgr);
 	filp->private_data = NULL;
diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index e74ea4c2d008..8e9bdbac512e 100644
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -1011,7 +1011,7 @@ static int _gaudi_init_tpc_mem(struct hl_device *hdev,
 
 release_cb:
 	hl_cb_put(cb);
-	hl_cb_destroy(hdev, &hdev->kernel_cb_mgr, cb->id << PAGE_SHIFT);
+	hl_cb_destroy(&hdev->kernel_mem_mgr, cb->buf->handle);
 
 	return rc;
 }
@@ -1473,7 +1473,7 @@ static int gaudi_collective_wait_create_job(struct hl_device *hdev,
 		job->patched_cb = NULL;
 
 	job->job_cb_size = job->user_cb_size;
-	hl_cb_destroy(hdev, &hdev->kernel_cb_mgr, cb->id << PAGE_SHIFT);
+	hl_cb_destroy(&hdev->kernel_mem_mgr, cb->buf->handle);
 
 	/* increment refcount as for external queues we get completion */
 	if (hw_queue_prop->type == QUEUE_TYPE_EXT)
@@ -5525,7 +5525,7 @@ static int gaudi_patch_cb(struct hl_device *hdev,
 static int gaudi_parse_cb_mmu(struct hl_device *hdev,
 		struct hl_cs_parser *parser)
 {
-	u64 patched_cb_handle;
+	u64 handle;
 	u32 patched_cb_size;
 	struct hl_cb *user_cb;
 	int rc;
@@ -5541,9 +5541,9 @@ static int gaudi_parse_cb_mmu(struct hl_device *hdev,
 	else
 		parser->patched_cb_size = parser->user_cb_size;
 
-	rc = hl_cb_create(hdev, &hdev->kernel_cb_mgr, hdev->kernel_ctx,
+	rc = hl_cb_create(hdev, &hdev->kernel_mem_mgr, hdev->kernel_ctx,
 				parser->patched_cb_size, false, false,
-				&patched_cb_handle);
+				&handle);
 
 	if (rc) {
 		dev_err(hdev->dev,
@@ -5552,13 +5552,10 @@ static int gaudi_parse_cb_mmu(struct hl_device *hdev,
 		return rc;
 	}
 
-	patched_cb_handle >>= PAGE_SHIFT;
-	parser->patched_cb = hl_cb_get(hdev, &hdev->kernel_cb_mgr,
-				(u32) patched_cb_handle);
+	parser->patched_cb = hl_cb_get(&hdev->kernel_mem_mgr, handle);
 	/* hl_cb_get should never fail */
 	if (!parser->patched_cb) {
-		dev_crit(hdev->dev, "DMA CB handle invalid 0x%x\n",
-			(u32) patched_cb_handle);
+		dev_crit(hdev->dev, "DMA CB handle invalid 0x%llx\n", handle);
 		rc = -EFAULT;
 		goto out;
 	}
@@ -5598,8 +5595,7 @@ static int gaudi_parse_cb_mmu(struct hl_device *hdev,
 	 * cb_put will release it, but here we want to remove it from the
 	 * idr
 	 */
-	hl_cb_destroy(hdev, &hdev->kernel_cb_mgr,
-					patched_cb_handle << PAGE_SHIFT);
+	hl_cb_destroy(&hdev->kernel_mem_mgr, handle);
 
 	return rc;
 }
@@ -5607,7 +5603,7 @@ static int gaudi_parse_cb_mmu(struct hl_device *hdev,
 static int gaudi_parse_cb_no_mmu(struct hl_device *hdev,
 		struct hl_cs_parser *parser)
 {
-	u64 patched_cb_handle;
+	u64 handle;
 	int rc;
 
 	rc = gaudi_validate_cb(hdev, parser, false);
@@ -5615,22 +5611,19 @@ static int gaudi_parse_cb_no_mmu(struct hl_device *hdev,
 	if (rc)
 		goto free_userptr;
 
-	rc = hl_cb_create(hdev, &hdev->kernel_cb_mgr, hdev->kernel_ctx,
+	rc = hl_cb_create(hdev, &hdev->kernel_mem_mgr, hdev->kernel_ctx,
 				parser->patched_cb_size, false, false,
-				&patched_cb_handle);
+				&handle);
 	if (rc) {
 		dev_err(hdev->dev,
 			"Failed to allocate patched CB for DMA CS %d\n", rc);
 		goto free_userptr;
 	}
 
-	patched_cb_handle >>= PAGE_SHIFT;
-	parser->patched_cb = hl_cb_get(hdev, &hdev->kernel_cb_mgr,
-				(u32) patched_cb_handle);
+	parser->patched_cb = hl_cb_get(&hdev->kernel_mem_mgr, handle);
 	/* hl_cb_get should never fail here */
 	if (!parser->patched_cb) {
-		dev_crit(hdev->dev, "DMA CB handle invalid 0x%x\n",
-				(u32) patched_cb_handle);
+		dev_crit(hdev->dev, "DMA CB handle invalid 0x%llx\n", handle);
 		rc = -EFAULT;
 		goto out;
 	}
@@ -5647,8 +5640,7 @@ static int gaudi_parse_cb_no_mmu(struct hl_device *hdev,
 	 * cb_put will release it, but here we want to remove it from the
 	 * idr
 	 */
-	hl_cb_destroy(hdev, &hdev->kernel_cb_mgr,
-				patched_cb_handle << PAGE_SHIFT);
+	hl_cb_destroy(&hdev->kernel_mem_mgr, handle);
 
 free_userptr:
 	if (rc)
@@ -5761,7 +5753,6 @@ static int gaudi_memset_device_memory(struct hl_device *hdev, u64 addr,
 	struct hl_cs_job *job;
 	u32 cb_size, ctl, err_cause;
 	struct hl_cb *cb;
-	u64 id;
 	int rc;
 
 	cb = hl_cb_kernel_create(hdev, PAGE_SIZE, false);
@@ -5828,9 +5819,8 @@ static int gaudi_memset_device_memory(struct hl_device *hdev, u64 addr,
 	}
 
 release_cb:
-	id = cb->id;
 	hl_cb_put(cb);
-	hl_cb_destroy(hdev, &hdev->kernel_cb_mgr, id << PAGE_SHIFT);
+	hl_cb_destroy(&hdev->kernel_mem_mgr, cb->buf->handle);
 
 	return rc;
 }
@@ -5893,7 +5883,7 @@ static int gaudi_memset_registers(struct hl_device *hdev, u64 reg_base,
 
 release_cb:
 	hl_cb_put(cb);
-	hl_cb_destroy(hdev, &hdev->kernel_cb_mgr, cb->id << PAGE_SHIFT);
+	hl_cb_destroy(&hdev->kernel_mem_mgr, cb->buf->handle);
 
 	return rc;
 }
diff --git a/drivers/misc/habanalabs/goya/goya.c b/drivers/misc/habanalabs/goya/goya.c
index 75736ccdfe4b..f8fb6dc04269 100644
--- a/drivers/misc/habanalabs/goya/goya.c
+++ b/drivers/misc/habanalabs/goya/goya.c
@@ -4001,7 +4001,7 @@ static int goya_patch_cb(struct hl_device *hdev,
 static int goya_parse_cb_mmu(struct hl_device *hdev,
 		struct hl_cs_parser *parser)
 {
-	u64 patched_cb_handle;
+	u64 handle;
 	u32 patched_cb_size;
 	struct hl_cb *user_cb;
 	int rc;
@@ -4014,9 +4014,9 @@ static int goya_parse_cb_mmu(struct hl_device *hdev,
 	parser->patched_cb_size = parser->user_cb_size +
 			sizeof(struct packet_msg_prot) * 2;
 
-	rc = hl_cb_create(hdev, &hdev->kernel_cb_mgr, hdev->kernel_ctx,
+	rc = hl_cb_create(hdev, &hdev->kernel_mem_mgr, hdev->kernel_ctx,
 				parser->patched_cb_size, false, false,
-				&patched_cb_handle);
+				&handle);
 
 	if (rc) {
 		dev_err(hdev->dev,
@@ -4025,13 +4025,10 @@ static int goya_parse_cb_mmu(struct hl_device *hdev,
 		return rc;
 	}
 
-	patched_cb_handle >>= PAGE_SHIFT;
-	parser->patched_cb = hl_cb_get(hdev, &hdev->kernel_cb_mgr,
-				(u32) patched_cb_handle);
+	parser->patched_cb = hl_cb_get(&hdev->kernel_mem_mgr, handle);
 	/* hl_cb_get should never fail here */
 	if (!parser->patched_cb) {
-		dev_crit(hdev->dev, "DMA CB handle invalid 0x%x\n",
-			(u32) patched_cb_handle);
+		dev_crit(hdev->dev, "DMA CB handle invalid 0x%llx\n", handle);
 		rc = -EFAULT;
 		goto out;
 	}
@@ -4071,8 +4068,7 @@ static int goya_parse_cb_mmu(struct hl_device *hdev,
 	 * cb_put will release it, but here we want to remove it from the
 	 * idr
 	 */
-	hl_cb_destroy(hdev, &hdev->kernel_cb_mgr,
-					patched_cb_handle << PAGE_SHIFT);
+	hl_cb_destroy(&hdev->kernel_mem_mgr, handle);
 
 	return rc;
 }
@@ -4080,7 +4076,7 @@ static int goya_parse_cb_mmu(struct hl_device *hdev,
 static int goya_parse_cb_no_mmu(struct hl_device *hdev,
 				struct hl_cs_parser *parser)
 {
-	u64 patched_cb_handle;
+	u64 handle;
 	int rc;
 
 	rc = goya_validate_cb(hdev, parser, false);
@@ -4088,22 +4084,19 @@ static int goya_parse_cb_no_mmu(struct hl_device *hdev,
 	if (rc)
 		goto free_userptr;
 
-	rc = hl_cb_create(hdev, &hdev->kernel_cb_mgr, hdev->kernel_ctx,
+	rc = hl_cb_create(hdev, &hdev->kernel_mem_mgr, hdev->kernel_ctx,
 				parser->patched_cb_size, false, false,
-				&patched_cb_handle);
+				&handle);
 	if (rc) {
 		dev_err(hdev->dev,
 			"Failed to allocate patched CB for DMA CS %d\n", rc);
 		goto free_userptr;
 	}
 
-	patched_cb_handle >>= PAGE_SHIFT;
-	parser->patched_cb = hl_cb_get(hdev, &hdev->kernel_cb_mgr,
-				(u32) patched_cb_handle);
+	parser->patched_cb = hl_cb_get(&hdev->kernel_mem_mgr, handle);
 	/* hl_cb_get should never fail here */
 	if (!parser->patched_cb) {
-		dev_crit(hdev->dev, "DMA CB handle invalid 0x%x\n",
-			(u32) patched_cb_handle);
+		dev_crit(hdev->dev, "DMA CB handle invalid 0x%llx\n", handle);
 		rc = -EFAULT;
 		goto out;
 	}
@@ -4120,8 +4113,7 @@ static int goya_parse_cb_no_mmu(struct hl_device *hdev,
 	 * cb_put will release it, but here we want to remove it from the
 	 * idr
 	 */
-	hl_cb_destroy(hdev, &hdev->kernel_cb_mgr,
-				patched_cb_handle << PAGE_SHIFT);
+	hl_cb_destroy(&hdev->kernel_mem_mgr, handle);
 
 free_userptr:
 	if (rc)
@@ -4853,7 +4845,7 @@ static int goya_memset_device_memory(struct hl_device *hdev, u64 addr, u64 size,
 
 release_cb:
 	hl_cb_put(cb);
-	hl_cb_destroy(hdev, &hdev->kernel_cb_mgr, cb->id << PAGE_SHIFT);
+	hl_cb_destroy(&hdev->kernel_mem_mgr, cb->buf->handle);
 
 	return rc;
 }
-- 
2.25.1


^ permalink raw reply related	[flat|nested] 6+ messages in thread
* [PATCH 6/6] habanalabs: add device memory scrub ability through debugfs
  2022-05-02  8:05 [PATCH 1/6] habanalabs: hide memory manager page shift Oded Gabbay
                   ` (3 preceding siblings ...)
  2022-05-02  8:05 ` [PATCH 5/6] habanalabs: use unified memory manager " Oded Gabbay
@ 2022-05-02  8:05 ` Oded Gabbay
  4 siblings, 0 replies; 6+ messages in thread
From: Oded Gabbay @ 2022-05-02  8:05 UTC (permalink / raw)
  To: linux-kernel; +Cc: Dafna Hirschfeld

From: Dafna Hirschfeld <dhirschfeld@habana.ai>

Add the ability to scrub the device memory with a given value.
Add file 'dram_mem_scrub_val' to set the value
and a file 'dram_mem_scrub' to scrub the dram.

This is very important to help during automated tests, when you want
the CI system to randomize the memory before training certain
DL topologies.

Signed-off-by: Dafna Hirschfeld <dhirschfeld@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
 .../ABI/testing/debugfs-driver-habanalabs     | 14 ++++++
 drivers/misc/habanalabs/common/debugfs.c      | 49 +++++++++++++++++++
 drivers/misc/habanalabs/common/habanalabs.h   |  6 +++
 .../misc/habanalabs/common/habanalabs_drv.c   |  8 +++
 drivers/misc/habanalabs/gaudi/gaudi.c         | 18 ++++---
 drivers/misc/habanalabs/goya/goya.c           |  6 +++
 drivers/misc/habanalabs/goya/goyaP.h          |  1 +
 7 files changed, 94 insertions(+), 8 deletions(-)

diff --git a/Documentation/ABI/testing/debugfs-driver-habanalabs b/Documentation/ABI/testing/debugfs-driver-habanalabs
index 84bf3da2bb27..0f8d20fe343f 100644
--- a/Documentation/ABI/testing/debugfs-driver-habanalabs
+++ b/Documentation/ABI/testing/debugfs-driver-habanalabs
@@ -170,6 +170,20 @@ KernelVersion:  5.1
 Contact:        ogabbay@kernel.org
 Description:    Sets the state of the third S/W led on the device
 
+What:           /sys/kernel/debug/habanalabs/hl<n>/memory_scrub
+Date:           May 2022
+KernelVersion:  5.19
+Contact:        dhirschfeld@habana.ai
+Description:    Allows the root user to scrub the dram memory. The scrubbing
+                value can be set using the debugfs file memory_scrub_val.
+
+What:           /sys/kernel/debug/habanalabs/hl<n>/memory_scrub_val
+Date:           May 2022
+KernelVersion:  5.19
+Contact:        dhirschfeld@habana.ai
+Description:    The value to which the dram will be set to when the user
+                scrubs the dram using 'memory_scrub' debugfs file
+
 What:           /sys/kernel/debug/habanalabs/hl<n>/mmu
 Date:           Jan 2019
 KernelVersion:  5.1
diff --git a/drivers/misc/habanalabs/common/debugfs.c b/drivers/misc/habanalabs/common/debugfs.c
index a9c4f2d4139d..c6744bfc6da4 100644
--- a/drivers/misc/habanalabs/common/debugfs.c
+++ b/drivers/misc/habanalabs/common/debugfs.c
@@ -538,6 +538,39 @@ static int engines_show(struct seq_file *s, void *data)
 	return 0;
 }
 
+static ssize_t hl_memory_scrub(struct file *f, const char __user *buf,
+					size_t count, loff_t *ppos)
+{
+	struct hl_dbg_device_entry *entry = file_inode(f)->i_private;
+	struct hl_device *hdev = entry->hdev;
+	u64 val = entry->memory_scrub_val;
+	int rc;
+
+	if (!hl_device_operational(hdev, NULL)) {
+		dev_warn_ratelimited(hdev->dev, "Can't scrub memory, device is not operational\n");
+		return -EIO;
+	}
+
+	mutex_lock(&hdev->fpriv_list_lock);
+	if (hdev->is_compute_ctx_active) {
+		mutex_unlock(&hdev->fpriv_list_lock);
+		dev_err(hdev->dev, "can't scrub dram, context exist\n");
+		return -EBUSY;
+	}
+	hdev->is_in_dram_scrub = true;
+	mutex_unlock(&hdev->fpriv_list_lock);
+
+	rc = hdev->asic_funcs->scrub_device_dram(hdev, val);
+
+	mutex_lock(&hdev->fpriv_list_lock);
+	hdev->is_in_dram_scrub = false;
+	mutex_unlock(&hdev->fpriv_list_lock);
+
+	if (rc)
+		return rc;
+	return count;
+}
+
 static bool hl_is_device_va(struct hl_device *hdev, u64 addr)
 {
 	struct asic_fixed_properties *prop = &hdev->asic_prop;
@@ -1316,6 +1349,11 @@ static ssize_t hl_timeout_locked_write(struct file *f, const char __user *buf,
 	return count;
 }
 
+static const struct file_operations hl_mem_scrub_fops = {
+	.owner = THIS_MODULE,
+	.write = hl_memory_scrub,
+};
+
 static const struct file_operations hl_data32b_fops = {
 	.owner = THIS_MODULE,
 	.read = hl_data_read32,
@@ -1475,6 +1513,17 @@ void hl_debugfs_add_device(struct hl_device *hdev)
 	dev_entry->root = debugfs_create_dir(dev_name(hdev->dev),
 						hl_debug_root);
 
+	debugfs_create_x64("memory_scrub_val",
+				0644,
+				dev_entry->root,
+				&dev_entry->memory_scrub_val);
+
+	debugfs_create_file("memory_scrub",
+				0200,
+				dev_entry->root,
+				dev_entry,
+				&hl_mem_scrub_fops);
+
 	debugfs_create_x64("addr",
 				0644,
 				dev_entry->root,
diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index 496d61ee07c5..59150caa98a2 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -1246,6 +1246,7 @@ struct fw_load_mgr {
  *                           its implementation is not trivial when the driver
  *                           is loaded in simulation mode (not upstreamed).
  * @scrub_device_mem: Scrub device memory given an address and size
+ * @scrub_device_dram: Scrub the dram memory of the device.
  * @get_int_queue_base: get the internal queue base address.
  * @test_queues: run simple test on all queues for sanity check.
  * @asic_dma_pool_zalloc: small DMA allocation of coherent memory from DMA pool.
@@ -1357,6 +1358,7 @@ struct hl_asic_funcs {
 	void (*asic_dma_free_coherent)(struct hl_device *hdev, size_t size,
 					void *cpu_addr, dma_addr_t dma_handle);
 	int (*scrub_device_mem)(struct hl_device *hdev, u64 addr, u64 size);
+	int (*scrub_device_dram)(struct hl_device *hdev, u64 val);
 	void* (*get_int_queue_base)(struct hl_device *hdev, u32 queue_id,
 				dma_addr_t *dma_handle, u16 *queue_len);
 	int (*test_queues)(struct hl_device *hdev);
@@ -2011,6 +2013,7 @@ struct hl_debugfs_entry {
  * @addr: next address to read/write from/to in read/write32.
  * @mmu_addr: next virtual address to translate to physical address in mmu_show.
  * @userptr_lookup: the target user ptr to look up for on demand.
+ * @memory_scrub_val: the value to which the dram will be scrubbed to using cb scrub_device_dram
  * @mmu_asid: ASID to use while translating in mmu_show.
  * @state_dump_head: index of the latest state dump
  * @i2c_bus: generic u8 debugfs file for bus value to use in i2c_data_read.
@@ -2041,6 +2044,7 @@ struct hl_dbg_device_entry {
 	u64				addr;
 	u64				mmu_addr;
 	u64				userptr_lookup;
+	u64				memory_scrub_val;
 	u32				mmu_asid;
 	u32				state_dump_head;
 	u8				i2c_bus;
@@ -2704,6 +2708,7 @@ struct hl_reset_info {
  * @id_control: minor of the control device
  * @cpu_pci_msb_addr: 50-bit extension bits for the device CPU's 40-bit
  *                    addresses.
+ * @is_in_dram_scrub: true if dram scrub operation is on going.
  * @disabled: is device disabled.
  * @late_init_done: is late init stage was done during initialization.
  * @hwmon_initialized: is H/W monitor sensors was initialized.
@@ -2834,6 +2839,7 @@ struct hl_device {
 	u16				id;
 	u16				id_control;
 	u16				cpu_pci_msb_addr;
+	u8				is_in_dram_scrub;
 	u8				disabled;
 	u8				late_init_done;
 	u8				hwmon_initialized;
diff --git a/drivers/misc/habanalabs/common/habanalabs_drv.c b/drivers/misc/habanalabs/common/habanalabs_drv.c
index 70203433e6cd..1210de39d661 100644
--- a/drivers/misc/habanalabs/common/habanalabs_drv.c
+++ b/drivers/misc/habanalabs/common/habanalabs_drv.c
@@ -158,6 +158,14 @@ int hl_device_open(struct inode *inode, struct file *filp)
 		goto out_err;
 	}
 
+	if (hdev->is_in_dram_scrub) {
+		dev_dbg_ratelimited(hdev->dev,
+			"Can't open %s during dram scrub\n",
+			dev_name(hdev->dev));
+		rc = -EAGAIN;
+		goto out_err;
+	}
+
 	if (hdev->compute_ctx_in_release) {
 		dev_dbg_ratelimited(hdev->dev,
 			"Can't open %s because another user is still releasing it\n",
diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index 8e9bdbac512e..179d9d32f9a3 100644
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -4740,12 +4740,11 @@ static void gaudi_dma_free_coherent(struct hl_device *hdev, size_t size,
 	dma_free_coherent(&hdev->pdev->dev, size, cpu_addr, fixed_dma_handle);
 }
 
-static int gaudi_hbm_scrubbing(struct hl_device *hdev)
+int gaudi_scrub_device_dram(struct hl_device *hdev, u64 val)
 {
 	struct asic_fixed_properties *prop = &hdev->asic_prop;
 	u64  cur_addr = DRAM_BASE_ADDR_USER;
-	u32 val;
-	u32 chunk_size;
+	u32 chunk_size, busy;
 	int rc, dma_id;
 
 	while (cur_addr < prop->dram_end_address) {
@@ -4759,8 +4758,10 @@ static int gaudi_hbm_scrubbing(struct hl_device *hdev)
 				"Doing HBM scrubbing for 0x%09llx - 0x%09llx\n",
 				cur_addr, cur_addr + chunk_size);
 
-			WREG32(mmDMA0_CORE_SRC_BASE_LO + dma_offset, 0xdeadbeaf);
-			WREG32(mmDMA0_CORE_SRC_BASE_HI + dma_offset, 0xdeadbeaf);
+			WREG32(mmDMA0_CORE_SRC_BASE_LO + dma_offset,
+					lower_32_bits(val));
+			WREG32(mmDMA0_CORE_SRC_BASE_HI + dma_offset,
+					upper_32_bits(val));
 			WREG32(mmDMA0_CORE_DST_BASE_LO + dma_offset,
 						lower_32_bits(cur_addr));
 			WREG32(mmDMA0_CORE_DST_BASE_HI + dma_offset,
@@ -4783,8 +4784,8 @@ static int gaudi_hbm_scrubbing(struct hl_device *hdev)
 			rc = hl_poll_timeout(
 				hdev,
 				mmDMA0_CORE_STS0 + dma_offset,
-				val,
-				((val & DMA0_CORE_STS0_BUSY_MASK) == 0),
+				busy,
+				((busy & DMA0_CORE_STS0_BUSY_MASK) == 0),
 				1000,
 				HBM_SCRUBBING_TIMEOUT_US);
 
@@ -4838,7 +4839,7 @@ static int gaudi_scrub_device_mem(struct hl_device *hdev, u64 addr, u64 size)
 		}
 
 		/* Scrub HBM using all DMA channels in parallel */
-		rc = gaudi_hbm_scrubbing(hdev);
+		rc = gaudi_scrub_device_dram(hdev, 0xdeadbeaf);
 		if (rc)
 			dev_err(hdev->dev,
 				"Failed to clear HBM in mem scrub all\n");
@@ -9208,6 +9209,7 @@ static const struct hl_asic_funcs gaudi_funcs = {
 	.asic_dma_alloc_coherent = gaudi_dma_alloc_coherent,
 	.asic_dma_free_coherent = gaudi_dma_free_coherent,
 	.scrub_device_mem = gaudi_scrub_device_mem,
+	.scrub_device_dram = gaudi_scrub_device_dram,
 	.get_int_queue_base = gaudi_get_int_queue_base,
 	.test_queues = gaudi_test_queues,
 	.asic_dma_pool_zalloc = gaudi_dma_pool_zalloc,
diff --git a/drivers/misc/habanalabs/goya/goya.c b/drivers/misc/habanalabs/goya/goya.c
index f8fb6dc04269..7983c362da67 100644
--- a/drivers/misc/habanalabs/goya/goya.c
+++ b/drivers/misc/habanalabs/goya/goya.c
@@ -5434,6 +5434,11 @@ static int goya_mmu_prefetch_cache_range(struct hl_device *hdev, u32 flags, u32
 	return 0;
 }
 
+int goya_scrub_device_dram(struct hl_device *hdev, u64 val)
+{
+	return -EOPNOTSUPP;
+}
+
 static const struct hl_asic_funcs goya_funcs = {
 	.early_init = goya_early_init,
 	.early_fini = goya_early_fini,
@@ -5452,6 +5457,7 @@ static const struct hl_asic_funcs goya_funcs = {
 	.asic_dma_alloc_coherent = goya_dma_alloc_coherent,
 	.asic_dma_free_coherent = goya_dma_free_coherent,
 	.scrub_device_mem = goya_scrub_device_mem,
+	.scrub_device_dram = goya_scrub_device_dram,
 	.get_int_queue_base = goya_get_int_queue_base,
 	.test_queues = goya_test_queues,
 	.asic_dma_pool_zalloc = goya_dma_pool_zalloc,
diff --git a/drivers/misc/habanalabs/goya/goyaP.h b/drivers/misc/habanalabs/goya/goyaP.h
index 647f57402616..10118d4bc87d 100644
--- a/drivers/misc/habanalabs/goya/goyaP.h
+++ b/drivers/misc/habanalabs/goya/goyaP.h
@@ -177,6 +177,7 @@ struct goya_device {
 	enum hl_pm_mng_profile		pm_mng_profile;
 };
 
+int goya_scrub_device_dram(struct hl_device *hdev, u64 val);
 int goya_set_fixed_properties(struct hl_device *hdev);
 int goya_mmu_init(struct hl_device *hdev);
 void goya_init_dma_qmans(struct hl_device *hdev);
-- 
2.25.1


^ permalink raw reply related	[flat|nested] 6+ messages in thread
end of thread, other threads:[~2022-05-02  8:08 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-05-02  8:05 [PATCH 1/6] habanalabs: hide memory manager page shift Oded Gabbay
2022-05-02  8:05 ` [PATCH 2/6] habanalabs: add put by handle method to memory manager Oded Gabbay
2022-05-02  8:05 ` [PATCH 3/6] habanalabs/gaudi: set arbitration timeout to a high value Oded Gabbay
2022-05-02  8:05 ` [PATCH 4/6] habanalabs: unified memory manager new code for CB flow Oded Gabbay
2022-05-02  8:05 ` [PATCH 5/6] habanalabs: use unified memory manager " Oded Gabbay
2022-05-02  8:05 ` [PATCH 6/6] habanalabs: add device memory scrub ability through debugfs Oded Gabbay

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.