habanalabs: add support for multiple SOBs per monitor

Support advanced monitor functionality to monitor more than a
single SOB. In addition expand all CB generation functions
with buffer offset in order to put in them multiple packets that are
generated by different functions.

Signed-off-by: Ofir Bitton <obitton@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

3 files changed, 48 insertions, 6 deletions

diff --git a/drivers/misc/habanalabs/common/command_submission.c b/drivers/misc/habanalabs/common/command_submission.c
index 9d49dd1558af..0d82c7dd93d0 100644
--- a/drivers/misc/habanalabs/common/command_submission.c
+++ b/drivers/misc/habanalabs/common/command_submission.c
@@ -38,6 +38,38 @@ void hl_sob_reset_error(struct kref *ref)
 			hw_sob->q_idx, hw_sob->sob_id);
 }
 
+/**
+ * hl_gen_sob_mask() - Generates a sob mask to be used in a monitor arm packet
+ * @sob_base: sob base id
+ * @sob_mask: sob user mask, each bit represents a sob offset from sob base
+ * @mask: generated mask
+ *
+ * Return: 0 if given parameters are valid
+ */
+int hl_gen_sob_mask(u16 sob_base, u8 sob_mask, u8 *mask)
+{
+	int i;
+
+	if (sob_mask == 0)
+		return -EINVAL;
+
+	if (sob_mask == 0x1) {
+		*mask = ~(1 << (sob_base & 0x7));
+	} else {
+		/* find msb in order to verify sob range is valid */
+		for (i = BITS_PER_BYTE - 1 ; i >= 0 ; i--)
+			if (BIT(i) & sob_mask)
+				break;
+
+		if (i > (HL_MAX_SOBS_PER_MONITOR - (sob_base & 0x7) - 1))
+			return -EINVAL;
+
+		*mask = ~sob_mask;
+	}
+
+	return 0;
+}
+
 static void hl_fence_release(struct kref *kref)
 {
 	struct hl_fence *fence =
diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index 0f128154bfa2..c61967213f89 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -77,20 +77,26 @@
 
 #define HL_MAX_DCORES			4
 
+#define HL_MAX_SOBS_PER_MONITOR	8
+
 /**
  * struct hl_gen_wait_properties - properties for generating a wait CB
  * @data: command buffer
  * @q_idx: queue id is used to extract fence register address
- * @sob_id: SOB id to use in this wait CB
+ * @size: offset in command buffer
+ * @sob_base: SOB base to use in this wait CB
  * @sob_val: SOB value to wait for
  * @mon_id: monitor to use in this wait CB
+ * @sob_mask: each bit represents a SOB offset from sob_base to be used
  */
 struct hl_gen_wait_properties {
 	void	*data;
 	u32	q_idx;
-	u16	sob_id;
+	u32	size;
+	u16	sob_base;
 	u16	sob_val;
 	u16	mon_id;
+	u8	sob_mask;
 };
 
 /**
@@ -844,8 +850,9 @@ struct hl_asic_funcs {
 	int (*load_boot_fit_to_device)(struct hl_device *hdev);
 	u32 (*get_signal_cb_size)(struct hl_device *hdev);
 	u32 (*get_wait_cb_size)(struct hl_device *hdev);
-	void (*gen_signal_cb)(struct hl_device *hdev, void *data, u16 sob_id);
-	void (*gen_wait_cb)(struct hl_device *hdev,
+	u32 (*gen_signal_cb)(struct hl_device *hdev, void *data, u16 sob_id,
+			u32 size);
+	u32 (*gen_wait_cb)(struct hl_device *hdev,
 			struct hl_gen_wait_properties *prop);
 	void (*reset_sob)(struct hl_device *hdev, void *data);
 	void (*set_dma_mask_from_fw)(struct hl_device *hdev);
@@ -1927,6 +1934,7 @@ void hl_cs_rollback_all(struct hl_device *hdev);
 struct hl_cs_job *hl_cs_allocate_job(struct hl_device *hdev,
 		enum hl_queue_type queue_type, bool is_kernel_allocated_cb);
 void hl_sob_reset_error(struct kref *ref);
+int hl_gen_sob_mask(u16 sob_base, u8 sob_mask, u8 *mask);
 void hl_fence_put(struct hl_fence *fence);
 void hl_fence_get(struct hl_fence *fence);
 
diff --git a/drivers/misc/habanalabs/common/hw_queue.c b/drivers/misc/habanalabs/common/hw_queue.c
index f2d094063159..be1d0e2c99d8 100644
--- a/drivers/misc/habanalabs/common/hw_queue.c
+++ b/drivers/misc/habanalabs/common/hw_queue.c
@@ -407,7 +407,7 @@ static void init_signal_cs(struct hl_device *hdev,
 		cs_cmpl->hw_sob->sob_id, cs_cmpl->sob_val, q_idx);
 
 	hdev->asic_funcs->gen_signal_cb(hdev, job->patched_cb,
-				cs_cmpl->hw_sob->sob_id);
+				cs_cmpl->hw_sob->sob_id, 0);
 
 	kref_get(&hw_sob->kref);
 
@@ -454,10 +454,12 @@ static void init_wait_cs(struct hl_device *hdev, struct hl_cs *cs,
 		prop->base_mon_id, q_idx);
 
 	wait_prop.data = (void *) job->patched_cb;
-	wait_prop.sob_id = cs_cmpl->hw_sob->sob_id;
+	wait_prop.sob_base = cs_cmpl->hw_sob->sob_id;
+	wait_prop.sob_mask = 0x1;
 	wait_prop.sob_val = cs_cmpl->sob_val;
 	wait_prop.mon_id = prop->base_mon_id;
 	wait_prop.q_idx = q_idx;
+	wait_prop.size = 0;
 	hdev->asic_funcs->gen_wait_cb(hdev, &wait_prop);
 
 	kref_get(&cs_cmpl->hw_sob->kref);