[PATCH] habanalabs: add missing counter update

[PATCH] habanalabs: add missing counter update

[Oded Gabbay]

The global CS drop-on-reset counter wasn't updated together with
the context counter.

Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
 drivers/misc/habanalabs/common/hw_queue.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/misc/habanalabs/common/hw_queue.c b/drivers/misc/habanalabs/common/hw_queue.c
index d1d30fb36410..7caf868d1585 100644
--- a/drivers/misc/habanalabs/common/hw_queue.c
+++ b/drivers/misc/habanalabs/common/hw_queue.c
@@ -529,6 +529,7 @@ int hl_hw_queue_schedule_cs(struct hl_cs *cs)
 	hdev->asic_funcs->hw_queues_lock(hdev);
 
 	if (!hl_device_operational(hdev, &status)) {
+		atomic64_inc(&cntr->device_in_reset_drop_cnt);
 		atomic64_inc(&ctx->cs_counters.device_in_reset_drop_cnt);
 		dev_err(hdev->dev,
 			"device is %s, CS rejected!\n", hdev->status[status]);
-- 
2.17.1

[PATCH] habanalabs: add support for cs with timestamp

[Oded Gabbay]

From: Ofir Bitton <obitton@habana.ai>

add support for user to request a timestamp upon
cs completion.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
 .../habanalabs/common/command_submission.c    | 44 ++++++++++++++-----
 drivers/misc/habanalabs/common/habanalabs.h   |  4 ++
 include/uapi/misc/habanalabs.h                |  6 ++-
 3 files changed, 41 insertions(+), 13 deletions(-)

diff --git a/drivers/misc/habanalabs/common/command_submission.c b/drivers/misc/habanalabs/common/command_submission.c
index f91b17480588..bd2f54399020 100644
--- a/drivers/misc/habanalabs/common/command_submission.c
+++ b/drivers/misc/habanalabs/common/command_submission.c
@@ -26,7 +26,7 @@ enum hl_cs_wait_status {
 static void job_wq_completion(struct work_struct *work);
 static int _hl_cs_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
 				u64 timeout_us, u64 seq,
-				enum hl_cs_wait_status *status);
+				enum hl_cs_wait_status *status, s64 *timestamp);
 static void cs_do_release(struct kref *ref);
 
 static void hl_sob_reset(struct kref *ref)
@@ -150,6 +150,7 @@ static void hl_fence_init(struct hl_fence *fence)
 {
 	kref_init(&fence->refcount);
 	fence->error = 0;
+	fence->timestamp = ktime_set(0, 0);
 	init_completion(&fence->completion);
 }
 
@@ -404,6 +405,8 @@ static void cs_do_release(struct kref *ref)
 	else if (!cs->submitted)
 		cs->fence->error = -EBUSY;
 
+	if (cs->timestamp)
+		cs->fence->timestamp = ktime_get();
 	complete_all(&cs->fence->completion);
 	hl_fence_put(cs->fence);
 
@@ -734,7 +737,8 @@ static int hl_cs_sanity_checks(struct hl_fpriv *hpriv, union hl_cs_args *args)
 		return -EBUSY;
 	}
 
-	cs_type_flags = args->in.cs_flags & ~HL_CS_FLAGS_FORCE_RESTORE;
+	cs_type_flags = args->in.cs_flags &
+			~(HL_CS_FLAGS_FORCE_RESTORE | HL_CS_FLAGS_TIMESTAMP);
 
 	if (unlikely(cs_type_flags && !is_power_of_2(cs_type_flags))) {
 		dev_err(hdev->dev,
@@ -798,7 +802,7 @@ static int hl_cs_copy_chunk_array(struct hl_device *hdev,
 }
 
 static int cs_ioctl_default(struct hl_fpriv *hpriv, void __user *chunks,
-				u32 num_chunks, u64 *cs_seq)
+				u32 num_chunks, u64 *cs_seq, bool timestamp)
 {
 	bool int_queues_only = true;
 	struct hl_device *hdev = hpriv->hdev;
@@ -825,6 +829,7 @@ static int cs_ioctl_default(struct hl_fpriv *hpriv, void __user *chunks,
 		goto free_cs_chunk_array;
 	}
 
+	cs->timestamp = !!timestamp;
 	*cs_seq = cs->sequence;
 
 	hl_debugfs_add_cs(cs);
@@ -995,7 +1000,7 @@ static int hl_cs_ctx_switch(struct hl_fpriv *hpriv, union hl_cs_args *args,
 			rc = 0;
 		} else {
 			rc = cs_ioctl_default(hpriv, chunks, num_chunks,
-						cs_seq);
+						cs_seq, false);
 		}
 
 		mutex_unlock(&hpriv->restore_phase_mutex);
@@ -1013,7 +1018,7 @@ static int hl_cs_ctx_switch(struct hl_fpriv *hpriv, union hl_cs_args *args,
 wait_again:
 			ret = _hl_cs_wait_ioctl(hdev, ctx,
 					jiffies_to_usecs(hdev->timeout_jiffies),
-					*cs_seq, &status);
+					*cs_seq, &status, NULL);
 			if (ret) {
 				if (ret == -ERESTARTSYS) {
 					usleep_range(100, 200);
@@ -1154,7 +1159,7 @@ static int cs_ioctl_signal_wait_create_jobs(struct hl_device *hdev,
 
 static int cs_ioctl_signal_wait(struct hl_fpriv *hpriv, enum hl_cs_type cs_type,
 				void __user *chunks, u32 num_chunks,
-				u64 *cs_seq)
+				u64 *cs_seq, bool timestamp)
 {
 	struct hl_cs_chunk *cs_chunk_array, *chunk;
 	struct hw_queue_properties *hw_queue_prop;
@@ -1259,6 +1264,8 @@ static int cs_ioctl_signal_wait(struct hl_fpriv *hpriv, enum hl_cs_type cs_type,
 		goto free_cs_chunk_array;
 	}
 
+	cs->timestamp = !!timestamp;
+
 	/*
 	 * Save the signal CS fence for later initialization right before
 	 * hanging the wait CS on the queue.
@@ -1334,10 +1341,11 @@ int hl_cs_ioctl(struct hl_fpriv *hpriv, void *data)
 	case CS_TYPE_WAIT:
 	case CS_TYPE_COLLECTIVE_WAIT:
 		rc = cs_ioctl_signal_wait(hpriv, cs_type, chunks, num_chunks,
-						&cs_seq);
+			&cs_seq, args->in.cs_flags & HL_CS_FLAGS_TIMESTAMP);
 		break;
 	default:
-		rc = cs_ioctl_default(hpriv, chunks, num_chunks, &cs_seq);
+		rc = cs_ioctl_default(hpriv, chunks, num_chunks, &cs_seq,
+				args->in.cs_flags & HL_CS_FLAGS_TIMESTAMP);
 		break;
 	}
 
@@ -1353,13 +1361,16 @@ int hl_cs_ioctl(struct hl_fpriv *hpriv, void *data)
 
 static int _hl_cs_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
 				u64 timeout_us, u64 seq,
-				enum hl_cs_wait_status *status)
+				enum hl_cs_wait_status *status, s64 *timestamp)
 {
 	struct hl_fence *fence;
 	unsigned long timeout;
 	int rc = 0;
 	long completion_rc;
 
+	if (timestamp)
+		*timestamp = 0;
+
 	if (timeout_us == MAX_SCHEDULE_TIMEOUT)
 		timeout = timeout_us;
 	else
@@ -1382,10 +1393,13 @@ static int _hl_cs_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
 				wait_for_completion_interruptible_timeout(
 					&fence->completion, timeout);
 
-		if (completion_rc > 0)
+		if (completion_rc > 0) {
 			*status = CS_WAIT_STATUS_COMPLETED;
-		else
+			if (timestamp)
+				*timestamp = ktime_to_ns(fence->timestamp);
+		} else {
 			*status = CS_WAIT_STATUS_BUSY;
+		}
 
 		if (fence->error == -ETIMEDOUT)
 			rc = -ETIMEDOUT;
@@ -1411,10 +1425,11 @@ int hl_cs_wait_ioctl(struct hl_fpriv *hpriv, void *data)
 	union hl_wait_cs_args *args = data;
 	enum hl_cs_wait_status status;
 	u64 seq = args->in.seq;
+	s64 timestamp;
 	int rc;
 
 	rc = _hl_cs_wait_ioctl(hdev, hpriv->ctx, args->in.timeout_us, seq,
-				&status);
+				&status, &timestamp);
 
 	memset(args, 0, sizeof(*args));
 
@@ -1439,6 +1454,11 @@ int hl_cs_wait_ioctl(struct hl_fpriv *hpriv, void *data)
 		return rc;
 	}
 
+	if (timestamp) {
+		args->out.flags |= HL_WAIT_CS_STATUS_FLAG_TIMESTAMP_VLD;
+		args->out.timestamp_nsec = timestamp;
+	}
+
 	switch (status) {
 	case CS_WAIT_STATUS_GONE:
 		args->out.flags |= HL_WAIT_CS_STATUS_FLAG_GONE;
diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index 9c9c8b24c47a..8e2d164d97e8 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -481,12 +481,14 @@ struct asic_fixed_properties {
  * @completion: fence is implemented using completion
  * @refcount: refcount for this fence
  * @error: mark this fence with error
+ * @timestamp: timestamp upon completion
  *
  */
 struct hl_fence {
 	struct completion	completion;
 	struct kref		refcount;
 	int			error;
+	ktime_t			timestamp;
 };
 
 /**
@@ -1127,6 +1129,7 @@ struct hl_userptr {
  * @tdr_active: true if TDR was activated for this CS (to prevent
  *		double TDR activation).
  * @aborted: true if CS was aborted due to some device error.
+ * @timestamp: true if a timestmap must be captured upon completion
  */
 struct hl_cs {
 	u16			*jobs_in_queue_cnt;
@@ -1147,6 +1150,7 @@ struct hl_cs {
 	u8			timedout;
 	u8			tdr_active;
 	u8			aborted;
+	u8			timestamp;
 };
 
 /**
diff --git a/include/uapi/misc/habanalabs.h b/include/uapi/misc/habanalabs.h
index 808d20da024a..6eff4e05eccb 100644
--- a/include/uapi/misc/habanalabs.h
+++ b/include/uapi/misc/habanalabs.h
@@ -592,6 +592,7 @@ struct hl_cs_chunk {
 #define HL_CS_FLAGS_SIGNAL		0x2
 #define HL_CS_FLAGS_WAIT		0x4
 #define HL_CS_FLAGS_COLLECTIVE_WAIT	0x8
+#define HL_CS_FLAGS_TIMESTAMP		0x20
 
 #define HL_CS_STATUS_SUCCESS		0
 
@@ -662,13 +663,16 @@ struct hl_wait_cs_in {
 #define HL_WAIT_CS_STATUS_ABORTED	3
 #define HL_WAIT_CS_STATUS_INTERRUPTED	4
 
-#define HL_WAIT_CS_STATUS_FLAG_GONE	0x1
+#define HL_WAIT_CS_STATUS_FLAG_GONE		0x1
+#define HL_WAIT_CS_STATUS_FLAG_TIMESTAMP_VLD	0x2
 
 struct hl_wait_cs_out {
 	/* HL_WAIT_CS_STATUS_* */
 	__u32 status;
 	/* HL_WAIT_CS_STATUS_FLAG* */
 	__u32 flags;
+	/* valid only if HL_WAIT_CS_STATUS_FLAG_TIMESTAMP_VLD is set */
+	__s64 timestamp_nsec;
 };
 
 union hl_wait_cs_args {
-- 
2.17.1

[PATCH] habanalabs: add ull to PLL masks

[Oded Gabbay]

From: Alon Mizrahi <amizrahi@habana.ai>

These defines are 64-bit defines so they need ull suffix.

Signed-off-by: Alon Mizrahi <amizrahi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
 drivers/misc/habanalabs/include/common/cpucp_if.h | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/misc/habanalabs/include/common/cpucp_if.h b/drivers/misc/habanalabs/include/common/cpucp_if.h
index 554f82271d5f..00bd9b392f93 100644
--- a/drivers/misc/habanalabs/include/common/cpucp_if.h
+++ b/drivers/misc/habanalabs/include/common/cpucp_if.h
@@ -317,13 +317,13 @@ enum cpucp_packet_id {
 #define CPUCP_PKT_CTL_OPCODE_MASK	0x1FFF0000
 
 #define CPUCP_PKT_RES_PLL_OUT0_SHIFT	0
-#define CPUCP_PKT_RES_PLL_OUT0_MASK	0x000000000000FFFF
+#define CPUCP_PKT_RES_PLL_OUT0_MASK	0x000000000000FFFFull
 #define CPUCP_PKT_RES_PLL_OUT1_SHIFT	16
-#define CPUCP_PKT_RES_PLL_OUT1_MASK	0x00000000FFFF0000
+#define CPUCP_PKT_RES_PLL_OUT1_MASK	0x00000000FFFF0000ull
 #define CPUCP_PKT_RES_PLL_OUT2_SHIFT	32
-#define CPUCP_PKT_RES_PLL_OUT2_MASK	0x0000FFFF00000000
+#define CPUCP_PKT_RES_PLL_OUT2_MASK	0x0000FFFF00000000ull
 #define CPUCP_PKT_RES_PLL_OUT3_SHIFT	48
-#define CPUCP_PKT_RES_PLL_OUT3_MASK	0xFFFF000000000000
+#define CPUCP_PKT_RES_PLL_OUT3_MASK	0xFFFF000000000000ull
 
 struct cpucp_packet {
 	union {
-- 
2.17.1

[PATCH] habanalabs: indicate to user that a cs is gone

[Oded Gabbay]

From: Ofir Bitton <obitton@habana.ai>

We want to indicate to the user that a certain command submission
is finished long time ago and it is no longer in database.
This means no further information regarding this cs can be obtained.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
 .../habanalabs/common/command_submission.c    | 68 ++++++++++++++-----
 include/uapi/misc/habanalabs.h                |  5 +-
 2 files changed, 54 insertions(+), 19 deletions(-)

diff --git a/drivers/misc/habanalabs/common/command_submission.c b/drivers/misc/habanalabs/common/command_submission.c
index 7309dd2b88a9..f91b17480588 100644
--- a/drivers/misc/habanalabs/common/command_submission.c
+++ b/drivers/misc/habanalabs/common/command_submission.c
@@ -11,9 +11,22 @@
 #include <linux/uaccess.h>
 #include <linux/slab.h>
 
+/**
+ * enum hl_cs_wait_status - cs wait status
+ * @CS_WAIT_STATUS_BUSY: cs was not completed yet
+ * @CS_WAIT_STATUS_COMPLETED: cs completed
+ * @CS_WAIT_STATUS_GONE: cs completed but fence is already gone
+ */
+enum hl_cs_wait_status {
+	CS_WAIT_STATUS_BUSY,
+	CS_WAIT_STATUS_COMPLETED,
+	CS_WAIT_STATUS_GONE
+};
+
 static void job_wq_completion(struct work_struct *work);
-static long _hl_cs_wait_ioctl(struct hl_device *hdev,
-		struct hl_ctx *ctx, u64 timeout_us, u64 seq);
+static int _hl_cs_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
+				u64 timeout_us, u64 seq,
+				enum hl_cs_wait_status *status);
 static void cs_do_release(struct kref *ref);
 
 static void hl_sob_reset(struct kref *ref)
@@ -942,7 +955,7 @@ static int hl_cs_ctx_switch(struct hl_fpriv *hpriv, union hl_cs_args *args,
 	int rc = 0, do_ctx_switch;
 	void __user *chunks;
 	u32 num_chunks, tmp;
-	long ret;
+	int ret;
 
 	do_ctx_switch = atomic_cmpxchg(&ctx->thread_ctx_switch_token, 1, 0);
 
@@ -996,18 +1009,19 @@ static int hl_cs_ctx_switch(struct hl_fpriv *hpriv, union hl_cs_args *args,
 
 		/* Need to wait for restore completion before execution phase */
 		if (num_chunks) {
+			enum hl_cs_wait_status status;
 wait_again:
 			ret = _hl_cs_wait_ioctl(hdev, ctx,
 					jiffies_to_usecs(hdev->timeout_jiffies),
-					*cs_seq);
-			if (ret <= 0) {
+					*cs_seq, &status);
+			if (ret) {
 				if (ret == -ERESTARTSYS) {
 					usleep_range(100, 200);
 					goto wait_again;
 				}
 
 				dev_err(hdev->dev,
-					"Restore CS for context %d failed to complete %ld\n",
+					"Restore CS for context %d failed to complete %d\n",
 					ctx->asid, ret);
 				rc = -ENOEXEC;
 				goto out;
@@ -1337,12 +1351,14 @@ int hl_cs_ioctl(struct hl_fpriv *hpriv, void *data)
 	return rc;
 }
 
-static long _hl_cs_wait_ioctl(struct hl_device *hdev,
-		struct hl_ctx *ctx, u64 timeout_us, u64 seq)
+static int _hl_cs_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
+				u64 timeout_us, u64 seq,
+				enum hl_cs_wait_status *status)
 {
 	struct hl_fence *fence;
 	unsigned long timeout;
-	long rc;
+	int rc = 0;
+	long completion_rc;
 
 	if (timeout_us == MAX_SCHEDULE_TIMEOUT)
 		timeout = timeout_us;
@@ -1360,11 +1376,17 @@ static long _hl_cs_wait_ioctl(struct hl_device *hdev,
 				seq, ctx->cs_sequence);
 	} else if (fence) {
 		if (!timeout_us)
-			rc = completion_done(&fence->completion);
+			completion_rc = completion_done(&fence->completion);
 		else
-			rc = wait_for_completion_interruptible_timeout(
+			completion_rc =
+				wait_for_completion_interruptible_timeout(
 					&fence->completion, timeout);
 
+		if (completion_rc > 0)
+			*status = CS_WAIT_STATUS_COMPLETED;
+		else
+			*status = CS_WAIT_STATUS_BUSY;
+
 		if (fence->error == -ETIMEDOUT)
 			rc = -ETIMEDOUT;
 		else if (fence->error == -EIO)
@@ -1375,7 +1397,7 @@ static long _hl_cs_wait_ioctl(struct hl_device *hdev,
 		dev_dbg(hdev->dev,
 			"Can't wait on seq %llu because current CS is at seq %llu (Fence is gone)\n",
 			seq, ctx->cs_sequence);
-		rc = 1;
+		*status = CS_WAIT_STATUS_GONE;
 	}
 
 	hl_ctx_put(ctx);
@@ -1387,14 +1409,16 @@ int hl_cs_wait_ioctl(struct hl_fpriv *hpriv, void *data)
 {
 	struct hl_device *hdev = hpriv->hdev;
 	union hl_wait_cs_args *args = data;
+	enum hl_cs_wait_status status;
 	u64 seq = args->in.seq;
-	long rc;
+	int rc;
 
-	rc = _hl_cs_wait_ioctl(hdev, hpriv->ctx, args->in.timeout_us, seq);
+	rc = _hl_cs_wait_ioctl(hdev, hpriv->ctx, args->in.timeout_us, seq,
+				&status);
 
 	memset(args, 0, sizeof(*args));
 
-	if (rc < 0) {
+	if (rc) {
 		if (rc == -ERESTARTSYS) {
 			dev_err_ratelimited(hdev->dev,
 				"user process got signal while waiting for CS handle %llu\n",
@@ -1415,10 +1439,18 @@ int hl_cs_wait_ioctl(struct hl_fpriv *hpriv, void *data)
 		return rc;
 	}
 
-	if (rc == 0)
-		args->out.status = HL_WAIT_CS_STATUS_BUSY;
-	else
+	switch (status) {
+	case CS_WAIT_STATUS_GONE:
+		args->out.flags |= HL_WAIT_CS_STATUS_FLAG_GONE;
+		fallthrough;
+	case CS_WAIT_STATUS_COMPLETED:
 		args->out.status = HL_WAIT_CS_STATUS_COMPLETED;
+		break;
+	case CS_WAIT_STATUS_BUSY:
+	default:
+		args->out.status = HL_WAIT_CS_STATUS_BUSY;
+		break;
+	}
 
 	return 0;
 }
diff --git a/include/uapi/misc/habanalabs.h b/include/uapi/misc/habanalabs.h
index 96eea49f48bc..808d20da024a 100644
--- a/include/uapi/misc/habanalabs.h
+++ b/include/uapi/misc/habanalabs.h
@@ -662,10 +662,13 @@ struct hl_wait_cs_in {
 #define HL_WAIT_CS_STATUS_ABORTED	3
 #define HL_WAIT_CS_STATUS_INTERRUPTED	4
 
+#define HL_WAIT_CS_STATUS_FLAG_GONE	0x1
+
 struct hl_wait_cs_out {
 	/* HL_WAIT_CS_STATUS_* */
 	__u32 status;
-	__u32 pad;
+	/* HL_WAIT_CS_STATUS_FLAG* */
+	__u32 flags;
 };
 
 union hl_wait_cs_args {
-- 
2.17.1