[PATCH 2/4] habanalabs: add support for multiple SOBs per monitor

From: Oded Gabbay
Date: Mon Nov 02 2020 - 14:58:16 EST

Next message: Oded Gabbay: "[PATCH 4/4] habanalabs: remove duplicate check"
Previous message: Oded Gabbay: "[PATCH] habanalabs: don't init vm module if no MMU"
In reply to: Oded Gabbay: "[PATCH] habanalabs: don't init vm module if no MMU"
Next in thread: Oded Gabbay: "[PATCH 4/4] habanalabs: remove duplicate check"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

From: Ofir Bitton <obitton@xxxxxxxxx>

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@xxxxxxxxx>
Reviewed-by: Oded Gabbay <ogabbay@xxxxxxxxxx>
Signed-off-by: Oded Gabbay <ogabbay@xxxxxxxxxx>
---
.../habanalabs/common/command_submission.c | 32 ++++
drivers/misc/habanalabs/common/habanalabs.h | 16 +-
drivers/misc/habanalabs/common/hw_queue.c | 6 +-
drivers/misc/habanalabs/gaudi/gaudi.c | 137 ++++++++++++------
drivers/misc/habanalabs/goya/goya.c | 9 +-
5 files changed, 143 insertions(+), 57 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 58b4097235d9..7307e0b88b44 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 613681c2cdcc..ca625789d78d 100644
--- a/drivers/misc/habanalabs/common/hw_queue.c
+++ b/drivers/misc/habanalabs/common/hw_queue.c
@@ -408,7 +408,7 @@ 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);

@@ -455,10 +455,12 @@ 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);
diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index 46dced9d1eec..930b26b1f445 100644
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -6380,14 +6380,16 @@ static u32 gaudi_get_wait_cb_size(struct hl_device *hdev)
sizeof(struct packet_msg_prot) * 2;
}

-static void gaudi_gen_signal_cb(struct hl_device *hdev, void *data, u16 sob_id)
+static u32 gaudi_gen_signal_cb(struct hl_device *hdev, void *data, u16 sob_id,
+ u32 size)
{
struct hl_cb *cb = (struct hl_cb *) data;
struct packet_msg_short *pkt;
- u32 value, ctl;
+ u32 value, ctl, pkt_size = sizeof(*pkt);

- pkt = (struct packet_msg_short *) (uintptr_t) cb->kernel_address;
- memset(pkt, 0, sizeof(*pkt));
+ pkt = (struct packet_msg_short *) (uintptr_t) (cb->kernel_address +
+ size);
+ memset(pkt, 0, pkt_size);

/* Inc by 1, Mode ADD */
value = FIELD_PREP(GAUDI_PKT_SHORT_VAL_SOB_SYNC_VAL_MASK, 1);
@@ -6403,6 +6405,8 @@ static void gaudi_gen_signal_cb(struct hl_device *hdev, void *data, u16 sob_id)

pkt->value = cpu_to_le32(value);
pkt->ctl = cpu_to_le32(ctl);
+
+ return size + pkt_size;
}

static u32 gaudi_add_mon_msg_short(struct packet_msg_short *pkt, u32 value,
@@ -6425,21 +6429,42 @@ static u32 gaudi_add_mon_msg_short(struct packet_msg_short *pkt, u32 value,
return pkt_size;
}

-static u32 gaudi_add_arm_monitor_pkt(struct packet_msg_short *pkt, u16 sob_id,
- u16 sob_val, u16 addr)
+static u32 gaudi_add_arm_monitor_pkt(struct hl_device *hdev,
+ struct packet_msg_short *pkt, u16 sob_base, u8 sob_mask,
+ u16 sob_val, u16 mon_id)
{
+ u64 monitor_base;
u32 ctl, value, pkt_size = sizeof(*pkt);
- u8 mask = ~(1 << (sob_id & 0x7));
+ u16 msg_addr_offset;
+ u8 mask;
+
+ if (hl_gen_sob_mask(sob_base, sob_mask, &mask)) {
+ dev_err(hdev->dev,
+ "sob_base %u (mask %#x) is not valid\n",
+ sob_base, sob_mask);
+ return 0;
+ }
+
+ /*
+ * monitor_base should be the content of the base0 address registers,
+ * so it will be added to the msg short offsets
+ */
+ monitor_base = mmSYNC_MNGR_W_S_SYNC_MNGR_OBJS_MON_PAY_ADDRL_0;
+
+ msg_addr_offset =
+ (mmSYNC_MNGR_W_S_SYNC_MNGR_OBJS_MON_ARM_0 + mon_id * 4) -
+ monitor_base;

memset(pkt, 0, pkt_size);

- value = FIELD_PREP(GAUDI_PKT_SHORT_VAL_MON_SYNC_GID_MASK, sob_id / 8);
+ /* Monitor config packet: bind the monitor to a sync object */
+ value = FIELD_PREP(GAUDI_PKT_SHORT_VAL_MON_SYNC_GID_MASK, sob_base / 8);
value |= FIELD_PREP(GAUDI_PKT_SHORT_VAL_MON_SYNC_VAL_MASK, sob_val);
value |= FIELD_PREP(GAUDI_PKT_SHORT_VAL_MON_MODE_MASK,
0); /* GREATER OR EQUAL*/
value |= FIELD_PREP(GAUDI_PKT_SHORT_VAL_MON_MASK_MASK, mask);

- ctl = FIELD_PREP(GAUDI_PKT_SHORT_CTL_ADDR_MASK, addr);
+ ctl = FIELD_PREP(GAUDI_PKT_SHORT_CTL_ADDR_MASK, msg_addr_offset);
ctl |= FIELD_PREP(GAUDI_PKT_SHORT_CTL_OP_MASK, 0); /* write the value */
ctl |= FIELD_PREP(GAUDI_PKT_SHORT_CTL_BASE_MASK, 2); /* W_S MON base */
ctl |= FIELD_PREP(GAUDI_PKT_SHORT_CTL_OPCODE_MASK, PACKET_MSG_SHORT);
@@ -6474,60 +6499,61 @@ static u32 gaudi_add_fence_pkt(struct packet_fence *pkt)
return pkt_size;
}

-static void gaudi_gen_wait_cb(struct hl_device *hdev,
- struct hl_gen_wait_properties *prop)
+static int gaudi_get_fence_addr(struct hl_device *hdev, u32 queue_id, u64 *addr)
{
- struct hl_cb *cb = (struct hl_cb *) prop->data;
- void *buf = (void *) (uintptr_t) cb->kernel_address;
- u64 monitor_base, fence_addr = 0;
- u32 size = 0;
- u16 msg_addr_offset;
+ u32 offset;

- switch (prop->q_idx) {
+ switch (queue_id) {
case GAUDI_QUEUE_ID_DMA_0_0:
- fence_addr = mmDMA0_QM_CP_FENCE2_RDATA_0;
+ offset = mmDMA0_QM_CP_FENCE2_RDATA_0;
break;
case GAUDI_QUEUE_ID_DMA_0_1:
- fence_addr = mmDMA0_QM_CP_FENCE2_RDATA_1;
+ offset = mmDMA0_QM_CP_FENCE2_RDATA_1;
break;
case GAUDI_QUEUE_ID_DMA_0_2:
- fence_addr = mmDMA0_QM_CP_FENCE2_RDATA_2;
+ offset = mmDMA0_QM_CP_FENCE2_RDATA_2;
break;
case GAUDI_QUEUE_ID_DMA_0_3:
- fence_addr = mmDMA0_QM_CP_FENCE2_RDATA_3;
+ offset = mmDMA0_QM_CP_FENCE2_RDATA_3;
break;
case GAUDI_QUEUE_ID_DMA_1_0:
- fence_addr = mmDMA1_QM_CP_FENCE2_RDATA_0;
+ offset = mmDMA1_QM_CP_FENCE2_RDATA_0;
break;
case GAUDI_QUEUE_ID_DMA_1_1:
- fence_addr = mmDMA1_QM_CP_FENCE2_RDATA_1;
+ offset = mmDMA1_QM_CP_FENCE2_RDATA_1;
break;
case GAUDI_QUEUE_ID_DMA_1_2:
- fence_addr = mmDMA1_QM_CP_FENCE2_RDATA_2;
+ offset = mmDMA1_QM_CP_FENCE2_RDATA_2;
break;
case GAUDI_QUEUE_ID_DMA_1_3:
- fence_addr = mmDMA1_QM_CP_FENCE2_RDATA_3;
+ offset = mmDMA1_QM_CP_FENCE2_RDATA_3;
break;
case GAUDI_QUEUE_ID_DMA_5_0:
- fence_addr = mmDMA5_QM_CP_FENCE2_RDATA_0;
+ offset = mmDMA5_QM_CP_FENCE2_RDATA_0;
break;
case GAUDI_QUEUE_ID_DMA_5_1:
- fence_addr = mmDMA5_QM_CP_FENCE2_RDATA_1;
+ offset = mmDMA5_QM_CP_FENCE2_RDATA_1;
break;
case GAUDI_QUEUE_ID_DMA_5_2:
- fence_addr = mmDMA5_QM_CP_FENCE2_RDATA_2;
+ offset = mmDMA5_QM_CP_FENCE2_RDATA_2;
break;
case GAUDI_QUEUE_ID_DMA_5_3:
- fence_addr = mmDMA5_QM_CP_FENCE2_RDATA_3;
+ offset = mmDMA5_QM_CP_FENCE2_RDATA_3;
break;
default:
- /* queue index should be valid here */
- dev_crit(hdev->dev, "wrong queue id %d for wait packet\n",
- prop->q_idx);
- return;
+ return -EINVAL;
}

- fence_addr += CFG_BASE;
+ *addr = CFG_BASE + offset;
+
+ return 0;
+}
+
+static u32 gaudi_add_mon_pkts(void *buf, u16 mon_id, u64 fence_addr)
+{
+ u64 monitor_base;
+ u32 size = 0;
+ u16 msg_addr_offset;

/*
* monitor_base should be the content of the base0 address registers,
@@ -6536,15 +6562,17 @@ static void gaudi_gen_wait_cb(struct hl_device *hdev,
monitor_base = mmSYNC_MNGR_W_S_SYNC_MNGR_OBJS_MON_PAY_ADDRL_0;

/* First monitor config packet: low address of the sync */
- msg_addr_offset = (mmSYNC_MNGR_W_S_SYNC_MNGR_OBJS_MON_PAY_ADDRL_0 +
- prop->mon_id * 4) - monitor_base;
+ msg_addr_offset =
+ (mmSYNC_MNGR_W_S_SYNC_MNGR_OBJS_MON_PAY_ADDRL_0 + mon_id * 4) -
+ monitor_base;

size += gaudi_add_mon_msg_short(buf + size, (u32) fence_addr,
msg_addr_offset);

/* Second monitor config packet: high address of the sync */
- msg_addr_offset = (mmSYNC_MNGR_W_S_SYNC_MNGR_OBJS_MON_PAY_ADDRH_0 +
- prop->mon_id * 4) - monitor_base;
+ msg_addr_offset =
+ (mmSYNC_MNGR_W_S_SYNC_MNGR_OBJS_MON_PAY_ADDRH_0 + mon_id * 4) -
+ monitor_base;

size += gaudi_add_mon_msg_short(buf + size, (u32) (fence_addr >> 32),
msg_addr_offset);
@@ -6553,20 +6581,35 @@ static void gaudi_gen_wait_cb(struct hl_device *hdev,
* Third monitor config packet: the payload, i.e. what to write when the
* sync triggers
*/
- msg_addr_offset = (mmSYNC_MNGR_W_S_SYNC_MNGR_OBJS_MON_PAY_DATA_0 +
- prop->mon_id * 4) - monitor_base;
+ msg_addr_offset =
+ (mmSYNC_MNGR_W_S_SYNC_MNGR_OBJS_MON_PAY_DATA_0 + mon_id * 4) -
+ monitor_base;

size += gaudi_add_mon_msg_short(buf + size, 1, msg_addr_offset);

- /* Fourth monitor config packet: bind the monitor to a sync object */
- msg_addr_offset =
- (mmSYNC_MNGR_W_S_SYNC_MNGR_OBJS_MON_ARM_0 + prop->mon_id * 4) -
- monitor_base;
- size += gaudi_add_arm_monitor_pkt(buf + size, prop->sob_id,
- prop->sob_val, msg_addr_offset);
+ return size;
+}
+
+u32 gaudi_gen_wait_cb(struct hl_device *hdev,
+ struct hl_gen_wait_properties *prop)
+{
+ struct hl_cb *cb = (struct hl_cb *) prop->data;
+ void *buf = (void *) (uintptr_t) cb->kernel_address;
+ u64 fence_addr = 0;
+ u32 size = prop->size;

- /* Fence packet */
+ if (gaudi_get_fence_addr(hdev, prop->q_idx, &fence_addr)) {
+ dev_crit(hdev->dev, "wrong queue id %d for wait packet\n",
+ prop->q_idx);
+ return 0;
+ }
+
+ size += gaudi_add_mon_pkts(buf + size, prop->mon_id, fence_addr);
+ size += gaudi_add_arm_monitor_pkt(hdev, buf + size, prop->sob_base,
+ prop->sob_mask, prop->sob_val, prop->mon_id);
size += gaudi_add_fence_pkt(buf + size);
+
+ return size;
}

static void gaudi_reset_sob(struct hl_device *hdev, void *data)
diff --git a/drivers/misc/habanalabs/goya/goya.c b/drivers/misc/habanalabs/goya/goya.c
index cd1366f10fbe..e8bf0b79cd67 100644
--- a/drivers/misc/habanalabs/goya/goya.c
+++ b/drivers/misc/habanalabs/goya/goya.c
@@ -5292,15 +5292,16 @@ static u32 goya_get_wait_cb_size(struct hl_device *hdev)
return 0;
}

-static void goya_gen_signal_cb(struct hl_device *hdev, void *data, u16 sob_id)
+static u32 goya_gen_signal_cb(struct hl_device *hdev, void *data, u16 sob_id,
+ u32 size)
{
-
+ return 0;
}

-static void goya_gen_wait_cb(struct hl_device *hdev,
+static u32 goya_gen_wait_cb(struct hl_device *hdev,
struct hl_gen_wait_properties *prop)
{
-
+ return 0;
}

static void goya_reset_sob(struct hl_device *hdev, void *data)
--
2.17.1

Next message: Oded Gabbay: "[PATCH 4/4] habanalabs: remove duplicate check"
Previous message: Oded Gabbay: "[PATCH] habanalabs: don't init vm module if no MMU"
In reply to: Oded Gabbay: "[PATCH] habanalabs: don't init vm module if no MMU"
Next in thread: Oded Gabbay: "[PATCH 4/4] habanalabs: remove duplicate check"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]