* [PATCH 01/10] habanalabs: don't clear previous f/w indications
@ 2021-11-07 20:51 Oded Gabbay
2021-11-07 20:51 ` [PATCH 02/10] habanalabs: handle abort scenario for user interrupt Oded Gabbay
` (8 more replies)
0 siblings, 9 replies; 10+ messages in thread
From: Oded Gabbay @ 2021-11-07 20:51 UTC (permalink / raw)
To: linux-kernel; +Cc: Ohad Sharabi
From: Ohad Sharabi <osharabi@habana.ai>
Once we read indication of whether f/w is doing the reset, we don't
want to clear it, until the next time we read this indication.
Otherwise, we might be in a state of wrong indication.
Signed-off-by: Ohad Sharabi <osharabi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
drivers/misc/habanalabs/common/firmware_if.c | 19 +++++--------------
1 file changed, 5 insertions(+), 14 deletions(-)
diff --git a/drivers/misc/habanalabs/common/firmware_if.c b/drivers/misc/habanalabs/common/firmware_if.c
index c68ad4d7b1bb..9addcfba6a8b 100644
--- a/drivers/misc/habanalabs/common/firmware_if.c
+++ b/drivers/misc/habanalabs/common/firmware_if.c
@@ -1247,8 +1247,7 @@ static void hl_fw_preboot_update_state(struct hl_device *hdev)
* 3. FW application - a. Fetch fw application security status
* b. Check whether hard reset is done by fw app
*/
- prop->hard_reset_done_by_fw =
- !!(cpu_boot_dev_sts0 & CPU_BOOT_DEV_STS0_FW_HARD_RST_EN);
+ prop->hard_reset_done_by_fw = !!(cpu_boot_dev_sts0 & CPU_BOOT_DEV_STS0_FW_HARD_RST_EN);
dev_dbg(hdev->dev, "Firmware preboot boot device status0 %#x\n",
cpu_boot_dev_sts0);
@@ -1915,17 +1914,13 @@ static void hl_fw_boot_fit_update_state(struct hl_device *hdev,
hdev->fw_loader.fw_comp_loaded |= FW_TYPE_BOOT_CPU;
- /* Clear reset status since we need to read it again from boot CPU */
- prop->hard_reset_done_by_fw = false;
-
/* Read boot_cpu status bits */
if (prop->fw_preboot_cpu_boot_dev_sts0 & CPU_BOOT_DEV_STS0_ENABLED) {
prop->fw_bootfit_cpu_boot_dev_sts0 =
RREG32(cpu_boot_dev_sts0_reg);
- if (prop->fw_bootfit_cpu_boot_dev_sts0 &
- CPU_BOOT_DEV_STS0_FW_HARD_RST_EN)
- prop->hard_reset_done_by_fw = true;
+ prop->hard_reset_done_by_fw = !!(prop->fw_bootfit_cpu_boot_dev_sts0 &
+ CPU_BOOT_DEV_STS0_FW_HARD_RST_EN);
dev_dbg(hdev->dev, "Firmware boot CPU status0 %#x\n",
prop->fw_bootfit_cpu_boot_dev_sts0);
@@ -2125,16 +2120,12 @@ static void hl_fw_linux_update_state(struct hl_device *hdev,
hdev->fw_loader.fw_comp_loaded |= FW_TYPE_LINUX;
- /* Clear reset status since we need to read again from app */
- prop->hard_reset_done_by_fw = false;
-
/* Read FW application security bits */
if (prop->fw_cpu_boot_dev_sts0_valid) {
prop->fw_app_cpu_boot_dev_sts0 = RREG32(cpu_boot_dev_sts0_reg);
- if (prop->fw_app_cpu_boot_dev_sts0 &
- CPU_BOOT_DEV_STS0_FW_HARD_RST_EN)
- prop->hard_reset_done_by_fw = true;
+ prop->hard_reset_done_by_fw = !!(prop->fw_app_cpu_boot_dev_sts0 &
+ CPU_BOOT_DEV_STS0_FW_HARD_RST_EN);
if (prop->fw_app_cpu_boot_dev_sts0 &
CPU_BOOT_DEV_STS0_GIC_PRIVILEGED_EN)
--
2.25.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH 02/10] habanalabs: handle abort scenario for user interrupt
2021-11-07 20:51 [PATCH 01/10] habanalabs: don't clear previous f/w indications Oded Gabbay
@ 2021-11-07 20:51 ` Oded Gabbay
2021-11-07 20:51 ` [PATCH 03/10] habanalabs: add dedicated message towards f/w to set power Oded Gabbay
` (7 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Oded Gabbay @ 2021-11-07 20:51 UTC (permalink / raw)
To: linux-kernel; +Cc: Bharat Jauhari
From: Bharat Jauhari <bjauhari@habana.ai>
In case of device reset, the driver does a force trigger on all waiting
users to release them from waiting. However, the driver does not handle
error scenario while waiting.
hl_interrupt_wait_ioctl() now exits the wait in case of an error with
abort status.
Signed-off-by: Bharat Jauhari <bjauhari@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
.../habanalabs/common/command_submission.c | 34 +++++++++----------
1 file changed, 16 insertions(+), 18 deletions(-)
diff --git a/drivers/misc/habanalabs/common/command_submission.c b/drivers/misc/habanalabs/common/command_submission.c
index 4c8000fd246c..41b48929cd59 100644
--- a/drivers/misc/habanalabs/common/command_submission.c
+++ b/drivers/misc/habanalabs/common/command_submission.c
@@ -2768,7 +2768,7 @@ static int hl_cs_wait_ioctl(struct hl_fpriv *hpriv, void *data)
static int _hl_interrupt_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
u32 timeout_us, u64 user_address,
u64 target_value, u16 interrupt_offset,
- enum hl_cs_wait_status *status,
+ u32 *status,
u64 *timestamp)
{
struct hl_user_pending_interrupt *pend;
@@ -2815,13 +2815,14 @@ static int _hl_interrupt_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
}
if (completion_value >= target_value) {
- *status = CS_WAIT_STATUS_COMPLETED;
+ *status = HL_WAIT_CS_STATUS_COMPLETED;
/* There was no interrupt, we assume the completion is now. */
pend->fence.timestamp = ktime_get();
- } else
- *status = CS_WAIT_STATUS_BUSY;
+ } else {
+ *status = HL_WAIT_CS_STATUS_BUSY;
+ }
- if (!timeout_us || (*status == CS_WAIT_STATUS_COMPLETED))
+ if (!timeout_us || (*status == HL_WAIT_CS_STATUS_COMPLETED))
goto remove_pending_user_interrupt;
wait_again:
@@ -2850,7 +2851,13 @@ static int _hl_interrupt_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
}
if (completion_value >= target_value) {
- *status = CS_WAIT_STATUS_COMPLETED;
+ *status = HL_WAIT_CS_STATUS_COMPLETED;
+ } else if (pend->fence.error) {
+ dev_err_ratelimited(hdev->dev,
+ "interrupt based wait ioctl aborted(error:%d) due to a reset cycle initiated\n",
+ pend->fence.error);
+ /* set the command completion status as ABORTED */
+ *status = HL_WAIT_CS_STATUS_ABORTED;
} else {
timeout = completion_rc;
goto wait_again;
@@ -2861,7 +2868,7 @@ static int _hl_interrupt_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
interrupt->interrupt_id);
rc = -EINTR;
} else {
- *status = CS_WAIT_STATUS_BUSY;
+ *status = HL_WAIT_CS_STATUS_BUSY;
}
remove_pending_user_interrupt:
@@ -2883,7 +2890,7 @@ static int hl_interrupt_wait_ioctl(struct hl_fpriv *hpriv, void *data)
struct hl_device *hdev = hpriv->hdev;
struct asic_fixed_properties *prop;
union hl_wait_cs_args *args = data;
- enum hl_cs_wait_status status;
+ u32 status = HL_WAIT_CS_STATUS_BUSY;
u64 timestamp;
int rc;
@@ -2926,22 +2933,13 @@ static int hl_interrupt_wait_ioctl(struct hl_fpriv *hpriv, void *data)
}
memset(args, 0, sizeof(*args));
+ args->out.status = status;
if (timestamp) {
args->out.timestamp_nsec = timestamp;
args->out.flags |= HL_WAIT_CS_STATUS_FLAG_TIMESTAMP_VLD;
}
- switch (status) {
- 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;
}
--
2.25.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH 03/10] habanalabs: add dedicated message towards f/w to set power
2021-11-07 20:51 [PATCH 01/10] habanalabs: don't clear previous f/w indications Oded Gabbay
2021-11-07 20:51 ` [PATCH 02/10] habanalabs: handle abort scenario for user interrupt Oded Gabbay
@ 2021-11-07 20:51 ` Oded Gabbay
2021-11-07 20:51 ` [PATCH 04/10] habanalabs: rename reset flags Oded Gabbay
` (6 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Oded Gabbay @ 2021-11-07 20:51 UTC (permalink / raw)
To: linux-kernel; +Cc: Rajaravi Krishna Katta
From: Rajaravi Krishna Katta <rkatta@habana.ai>
CPUCP_PACKET_POWER_GET packet type was used for both
hl_get_power() and hl_set_power().
To align with other sensor functions hl_set_power()
should use CPUCP_PACKET_POWER_SET.
This packet will only be used with newer ASICs, so need to add
a compatibility flag to the asic properties to indicate whether to use
this packet or the GET packet.
Signed-off-by: Rajaravi Krishna Katta <rkatta@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
drivers/misc/habanalabs/common/habanalabs.h | 3 +++
drivers/misc/habanalabs/common/hwmon.c | 8 +++++++-
drivers/misc/habanalabs/gaudi/gaudi.c | 2 ++
drivers/misc/habanalabs/goya/goya.c | 2 ++
drivers/misc/habanalabs/include/common/cpucp_if.h | 4 ++++
5 files changed, 18 insertions(+), 1 deletion(-)
diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index 5fc9cfd892e8..dc61f7031c38 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -546,6 +546,8 @@ struct hl_hints_range {
* @dynamic_fw_load: is dynamic FW load is supported.
* @gic_interrupts_enable: true if FW is not blocking GIC controller,
* false otherwise.
+ * @use_get_power_for_reset_history: To support backward compatibility for Goya
+ * and Gaudi
*/
struct asic_fixed_properties {
struct hw_queue_properties *hw_queues_props;
@@ -626,6 +628,7 @@ struct asic_fixed_properties {
u8 iatu_done_by_fw;
u8 dynamic_fw_load;
u8 gic_interrupts_enable;
+ u8 use_get_power_for_reset_history;
};
/**
diff --git a/drivers/misc/habanalabs/common/hwmon.c b/drivers/misc/habanalabs/common/hwmon.c
index e33f65be8a00..70182b42940d 100644
--- a/drivers/misc/habanalabs/common/hwmon.c
+++ b/drivers/misc/habanalabs/common/hwmon.c
@@ -677,12 +677,18 @@ int hl_set_power(struct hl_device *hdev,
int sensor_index, u32 attr, long value)
{
struct cpucp_packet pkt;
+ struct asic_fixed_properties *prop = &hdev->asic_prop;
int rc;
memset(&pkt, 0, sizeof(pkt));
- pkt.ctl = cpu_to_le32(CPUCP_PACKET_POWER_GET <<
+ if (prop->use_get_power_for_reset_history)
+ pkt.ctl = cpu_to_le32(CPUCP_PACKET_POWER_GET <<
CPUCP_PKT_CTL_OPCODE_SHIFT);
+ else
+ pkt.ctl = cpu_to_le32(CPUCP_PACKET_POWER_SET <<
+ CPUCP_PKT_CTL_OPCODE_SHIFT);
+
pkt.sensor_index = __cpu_to_le16(sensor_index);
pkt.type = __cpu_to_le16(attr);
pkt.value = __cpu_to_le64(value);
diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index 1dcce1bc976f..738ad2498439 100644
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -665,6 +665,8 @@ static int gaudi_set_fixed_properties(struct hl_device *hdev)
prop->clk_pll_index = HL_GAUDI_MME_PLL;
prop->max_freq_value = GAUDI_MAX_CLK_FREQ;
+ prop->use_get_power_for_reset_history = true;
+
return 0;
}
diff --git a/drivers/misc/habanalabs/goya/goya.c b/drivers/misc/habanalabs/goya/goya.c
index ce06103292a0..959eb21dcc69 100644
--- a/drivers/misc/habanalabs/goya/goya.c
+++ b/drivers/misc/habanalabs/goya/goya.c
@@ -475,6 +475,8 @@ int goya_set_fixed_properties(struct hl_device *hdev)
prop->clk_pll_index = HL_GOYA_MME_PLL;
+ prop->use_get_power_for_reset_history = true;
+
return 0;
}
diff --git a/drivers/misc/habanalabs/include/common/cpucp_if.h b/drivers/misc/habanalabs/include/common/cpucp_if.h
index ae13231fda94..17927968e19a 100644
--- a/drivers/misc/habanalabs/include/common/cpucp_if.h
+++ b/drivers/misc/habanalabs/include/common/cpucp_if.h
@@ -376,6 +376,9 @@ enum pq_init_status {
* and QMANs. The f/w will return a bitmask where each bit represents
* a different engine or QMAN according to enum cpucp_idle_mask.
* The bit will be 1 if the engine is NOT idle.
+ *
+ * CPUCP_PACKET_POWER_SET -
+ * Resets power history of device to 0
*/
enum cpucp_packet_id {
@@ -421,6 +424,7 @@ enum cpucp_packet_id {
CPUCP_PACKET_NIC_STAT_REGS_CLR, /* internal */
CPUCP_PACKET_NIC_STAT_REGS_ALL_GET, /* internal */
CPUCP_PACKET_IS_IDLE_CHECK, /* internal */
+ CPUCP_PACKET_POWER_SET, /* internal */
};
#define CPUCP_PACKET_FENCE_VAL 0xFE8CE7A5
--
2.25.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH 04/10] habanalabs: rename reset flags
2021-11-07 20:51 [PATCH 01/10] habanalabs: don't clear previous f/w indications Oded Gabbay
2021-11-07 20:51 ` [PATCH 02/10] habanalabs: handle abort scenario for user interrupt Oded Gabbay
2021-11-07 20:51 ` [PATCH 03/10] habanalabs: add dedicated message towards f/w to set power Oded Gabbay
@ 2021-11-07 20:51 ` Oded Gabbay
2021-11-07 20:51 ` [PATCH 05/10] habanalabs: change wait for interrupt timeout to 64 bit Oded Gabbay
` (5 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Oded Gabbay @ 2021-11-07 20:51 UTC (permalink / raw)
To: linux-kernel; +Cc: Bharat Jauhari
From: Bharat Jauhari <bjauhari@habana.ai>
Rename reset flags for better readability as compared to
HL_RESET_CAUSE* enum shared with the f/w.
Signed-off-by: Bharat Jauhari <bjauhari@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
.../habanalabs/common/command_submission.c | 2 +-
drivers/misc/habanalabs/common/device.c | 40 +++++++++----------
drivers/misc/habanalabs/common/habanalabs.h | 28 ++++++-------
drivers/misc/habanalabs/common/memory.c | 2 +-
drivers/misc/habanalabs/common/sysfs.c | 2 +-
drivers/misc/habanalabs/gaudi/gaudi.c | 14 ++++---
drivers/misc/habanalabs/goya/goya.c | 10 ++---
7 files changed, 50 insertions(+), 48 deletions(-)
diff --git a/drivers/misc/habanalabs/common/command_submission.c b/drivers/misc/habanalabs/common/command_submission.c
index 41b48929cd59..9ebcd9894d83 100644
--- a/drivers/misc/habanalabs/common/command_submission.c
+++ b/drivers/misc/habanalabs/common/command_submission.c
@@ -767,7 +767,7 @@ static void cs_timedout(struct work_struct *work)
if (likely(!skip_reset_on_timeout)) {
if (hdev->reset_on_lockup)
- hl_device_reset(hdev, HL_RESET_TDR);
+ hl_device_reset(hdev, HL_DRV_RESET_TDR);
else
hdev->needs_reset = true;
}
diff --git a/drivers/misc/habanalabs/common/device.c b/drivers/misc/habanalabs/common/device.c
index 9674e2520532..eb5800b403b6 100644
--- a/drivers/misc/habanalabs/common/device.c
+++ b/drivers/misc/habanalabs/common/device.c
@@ -95,7 +95,7 @@ static void hpriv_release(struct kref *ref)
if ((hdev->reset_if_device_not_idle && !device_is_idle)
|| hdev->reset_upon_device_release)
- hl_device_reset(hdev, HL_RESET_DEVICE_RELEASE);
+ hl_device_reset(hdev, HL_DRV_RESET_DEV_RELEASE);
/* Now we can mark the compute_ctx as empty. Even if a reset is running in a different
* thread, we don't care because the in_reset is marked so if a user will try to open
@@ -330,10 +330,10 @@ static void device_hard_reset_pending(struct work_struct *work)
u32 flags;
int rc;
- flags = HL_RESET_HARD | HL_RESET_FROM_RESET_THREAD;
+ flags = HL_DRV_RESET_HARD | HL_DRV_RESET_FROM_RESET_THR;
if (device_reset_work->fw_reset)
- flags |= HL_RESET_FW;
+ flags |= HL_DRV_RESET_BYPASS_REQ_TO_FW;
rc = hl_device_reset(hdev, flags);
if ((rc == -EBUSY) && !hdev->device_fini_pending) {
@@ -541,7 +541,7 @@ static void hl_device_heartbeat(struct work_struct *work)
goto reschedule;
dev_err(hdev->dev, "Device heartbeat failed!\n");
- hl_device_reset(hdev, HL_RESET_HARD | HL_RESET_HEARTBEAT);
+ hl_device_reset(hdev, HL_DRV_RESET_HARD | HL_DRV_RESET_HEARTBEAT);
return;
@@ -552,7 +552,7 @@ static void hl_device_heartbeat(struct work_struct *work)
* If control reached here, then at least one heartbeat work has been
* scheduled since last reset/init cycle.
* So if the device is not already in reset cycle, reset the flag
- * prev_reset_trigger as no reset occurred with HL_RESET_FW_FATAL_ERR
+ * prev_reset_trigger as no reset occurred with HL_DRV_RESET_FW_FATAL_ERR
* status for at least one heartbeat. From this point driver restarts
* tracking future consecutive fatal errors.
*/
@@ -831,7 +831,7 @@ int hl_device_resume(struct hl_device *hdev)
hdev->disabled = false;
atomic_set(&hdev->in_reset, 0);
- rc = hl_device_reset(hdev, HL_RESET_HARD);
+ rc = hl_device_reset(hdev, HL_DRV_RESET_HARD);
if (rc) {
dev_err(hdev->dev, "Failed to reset device during resume\n");
goto disable_device;
@@ -948,15 +948,15 @@ static void handle_reset_trigger(struct hl_device *hdev, u32 flags)
* ('in_reset' makes sure of it). This makes sure that
* 'reset_cause' will continue holding its 1st recorded reason!
*/
- if (flags & HL_RESET_HEARTBEAT) {
+ if (flags & HL_DRV_RESET_HEARTBEAT) {
hdev->curr_reset_cause = HL_RESET_CAUSE_HEARTBEAT;
- cur_reset_trigger = HL_RESET_HEARTBEAT;
- } else if (flags & HL_RESET_TDR) {
+ cur_reset_trigger = HL_DRV_RESET_HEARTBEAT;
+ } else if (flags & HL_DRV_RESET_TDR) {
hdev->curr_reset_cause = HL_RESET_CAUSE_TDR;
- cur_reset_trigger = HL_RESET_TDR;
- } else if (flags & HL_RESET_FW_FATAL_ERR) {
+ cur_reset_trigger = HL_DRV_RESET_TDR;
+ } else if (flags & HL_DRV_RESET_FW_FATAL_ERR) {
hdev->curr_reset_cause = HL_RESET_CAUSE_UNKNOWN;
- cur_reset_trigger = HL_RESET_FW_FATAL_ERR;
+ cur_reset_trigger = HL_DRV_RESET_FW_FATAL_ERR;
} else {
hdev->curr_reset_cause = HL_RESET_CAUSE_UNKNOWN;
}
@@ -979,8 +979,8 @@ static void handle_reset_trigger(struct hl_device *hdev, u32 flags)
* If F/W is performing the reset, no need to send it a message to disable
* PCI access
*/
- if ((flags & HL_RESET_HARD) &&
- !(flags & (HL_RESET_HEARTBEAT | HL_RESET_FW))) {
+ if ((flags & HL_DRV_RESET_HARD) &&
+ !(flags & (HL_DRV_RESET_HEARTBEAT | HL_DRV_RESET_BYPASS_REQ_TO_FW))) {
/* Disable PCI access from device F/W so he won't send
* us additional interrupts. We disable MSI/MSI-X at
* the halt_engines function and we can't have the F/W
@@ -1025,9 +1025,9 @@ int hl_device_reset(struct hl_device *hdev, u32 flags)
return 0;
}
- hard_reset = !!(flags & HL_RESET_HARD);
- from_hard_reset_thread = !!(flags & HL_RESET_FROM_RESET_THREAD);
- fw_reset = !!(flags & HL_RESET_FW);
+ hard_reset = !!(flags & HL_DRV_RESET_HARD);
+ from_hard_reset_thread = !!(flags & HL_DRV_RESET_FROM_RESET_THR);
+ fw_reset = !!(flags & HL_DRV_RESET_BYPASS_REQ_TO_FW);
if (!hard_reset && !hdev->supports_soft_reset) {
hard_instead_soft = true;
@@ -1035,7 +1035,7 @@ int hl_device_reset(struct hl_device *hdev, u32 flags)
}
if (hdev->reset_upon_device_release &&
- (flags & HL_RESET_DEVICE_RELEASE)) {
+ (flags & HL_DRV_RESET_DEV_RELEASE)) {
dev_dbg(hdev->dev,
"Perform %s-reset upon device release\n",
hard_reset ? "hard" : "soft");
@@ -1075,7 +1075,7 @@ int hl_device_reset(struct hl_device *hdev, u32 flags)
if (hard_reset)
dev_info(hdev->dev, "Going to reset device\n");
- else if (flags & HL_RESET_DEVICE_RELEASE)
+ else if (flags & HL_DRV_RESET_DEV_RELEASE)
dev_info(hdev->dev,
"Going to reset device after it was released by user\n");
else
@@ -1171,7 +1171,7 @@ int hl_device_reset(struct hl_device *hdev, u32 flags)
hdev->hard_reset_pending = false;
if (hdev->reset_trigger_repeated &&
- (hdev->prev_reset_trigger == HL_RESET_FW_FATAL_ERR)) {
+ (hdev->prev_reset_trigger == HL_DRV_RESET_FW_FATAL_ERR)) {
/* if there 2 back to back resets from FW,
* ensure driver puts the driver in a unusable state
*/
diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index dc61f7031c38..92d12c8ba569 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -120,37 +120,37 @@ enum hl_mmu_page_table_location {
/*
* Reset Flags
*
- * - HL_RESET_HARD
+ * - HL_DRV_RESET_HARD
* If set do hard reset to all engines. If not set reset just
* compute/DMA engines.
*
- * - HL_RESET_FROM_RESET_THREAD
+ * - HL_DRV_RESET_FROM_RESET_THR
* Set if the caller is the hard-reset thread
*
- * - HL_RESET_HEARTBEAT
+ * - HL_DRV_RESET_HEARTBEAT
* Set if reset is due to heartbeat
*
- * - HL_RESET_TDR
+ * - HL_DRV_RESET_TDR
* Set if reset is due to TDR
*
- * - HL_RESET_DEVICE_RELEASE
+ * - HL_DRV_RESET_DEV_RELEASE
* Set if reset is due to device release
*
- * - HL_RESET_FW
+ * - HL_DRV_RESET_BYPASS_REQ_TO_FW
* F/W will perform the reset. No need to ask it to reset the device. This is relevant
* only when running with secured f/w
*
- * - HL_RESET_FW_FATAL_ERR
+ * - HL_DRV_RESET_FW_FATAL_ERR
* Set if reset is due to a fatal error from FW
*/
-#define HL_RESET_HARD (1 << 0)
-#define HL_RESET_FROM_RESET_THREAD (1 << 1)
-#define HL_RESET_HEARTBEAT (1 << 2)
-#define HL_RESET_TDR (1 << 3)
-#define HL_RESET_DEVICE_RELEASE (1 << 4)
-#define HL_RESET_FW (1 << 5)
-#define HL_RESET_FW_FATAL_ERR (1 << 6)
+#define HL_DRV_RESET_HARD (1 << 0)
+#define HL_DRV_RESET_FROM_RESET_THR (1 << 1)
+#define HL_DRV_RESET_HEARTBEAT (1 << 2)
+#define HL_DRV_RESET_TDR (1 << 3)
+#define HL_DRV_RESET_DEV_RELEASE (1 << 4)
+#define HL_DRV_RESET_BYPASS_REQ_TO_FW (1 << 5)
+#define HL_DRV_RESET_FW_FATAL_ERR (1 << 6)
#define HL_MAX_SOBS_PER_MONITOR 8
diff --git a/drivers/misc/habanalabs/common/memory.c b/drivers/misc/habanalabs/common/memory.c
index cd3640617d02..530f8b4fadd2 100644
--- a/drivers/misc/habanalabs/common/memory.c
+++ b/drivers/misc/habanalabs/common/memory.c
@@ -316,7 +316,7 @@ static int free_phys_pg_pack(struct hl_device *hdev,
}
if (rc && !hdev->disabled)
- hl_device_reset(hdev, HL_RESET_HARD);
+ hl_device_reset(hdev, HL_DRV_RESET_HARD);
end:
kvfree(phys_pg_pack->pages);
diff --git a/drivers/misc/habanalabs/common/sysfs.c b/drivers/misc/habanalabs/common/sysfs.c
index 42c1769ad25d..aee0cc4d6155 100644
--- a/drivers/misc/habanalabs/common/sysfs.c
+++ b/drivers/misc/habanalabs/common/sysfs.c
@@ -236,7 +236,7 @@ static ssize_t hard_reset_store(struct device *dev,
dev_warn(hdev->dev, "Hard-Reset requested through sysfs\n");
- hl_device_reset(hdev, HL_RESET_HARD);
+ hl_device_reset(hdev, HL_DRV_RESET_HARD);
out:
return count;
diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index 738ad2498439..2724ab3747f2 100644
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -8003,7 +8003,7 @@ static void gaudi_handle_eqe(struct hl_device *hdev,
case GAUDI_EVENT_NIC0_CS_DBG_DERR ... GAUDI_EVENT_NIC4_CS_DBG_DERR:
gaudi_print_irq_info(hdev, event_type, true);
gaudi_handle_ecc_event(hdev, event_type, &eq_entry->ecc_data);
- fw_fatal_err_flag = HL_RESET_FW_FATAL_ERR;
+ fw_fatal_err_flag = HL_DRV_RESET_FW_FATAL_ERR;
goto reset_device;
case GAUDI_EVENT_GIC500:
@@ -8011,7 +8011,7 @@ static void gaudi_handle_eqe(struct hl_device *hdev,
case GAUDI_EVENT_L2_RAM_ECC:
case GAUDI_EVENT_PLL0 ... GAUDI_EVENT_PLL17:
gaudi_print_irq_info(hdev, event_type, false);
- fw_fatal_err_flag = HL_RESET_FW_FATAL_ERR;
+ fw_fatal_err_flag = HL_DRV_RESET_FW_FATAL_ERR;
goto reset_device;
case GAUDI_EVENT_HBM0_SPI_0:
@@ -8022,7 +8022,7 @@ static void gaudi_handle_eqe(struct hl_device *hdev,
gaudi_hbm_read_interrupts(hdev,
gaudi_hbm_event_to_dev(event_type),
&eq_entry->hbm_ecc_data);
- fw_fatal_err_flag = HL_RESET_FW_FATAL_ERR;
+ fw_fatal_err_flag = HL_DRV_RESET_FW_FATAL_ERR;
goto reset_device;
case GAUDI_EVENT_HBM0_SPI_1:
@@ -8205,9 +8205,11 @@ static void gaudi_handle_eqe(struct hl_device *hdev,
reset_device:
if (hdev->asic_prop.fw_security_enabled)
- hl_device_reset(hdev, HL_RESET_HARD | HL_RESET_FW | fw_fatal_err_flag);
+ hl_device_reset(hdev, HL_DRV_RESET_HARD
+ | HL_DRV_RESET_BYPASS_REQ_TO_FW
+ | fw_fatal_err_flag);
else if (hdev->hard_reset_on_fw_events)
- hl_device_reset(hdev, HL_RESET_HARD | fw_fatal_err_flag);
+ hl_device_reset(hdev, HL_DRV_RESET_HARD | fw_fatal_err_flag);
else
hl_fw_unmask_irq(hdev, event_type);
}
@@ -8260,7 +8262,7 @@ static int gaudi_mmu_invalidate_cache(struct hl_device *hdev, bool is_hard,
if (rc) {
dev_err_ratelimited(hdev->dev,
"MMU cache invalidation timeout\n");
- hl_device_reset(hdev, HL_RESET_HARD);
+ hl_device_reset(hdev, HL_DRV_RESET_HARD);
}
return rc;
diff --git a/drivers/misc/habanalabs/goya/goya.c b/drivers/misc/habanalabs/goya/goya.c
index 959eb21dcc69..3bbcab7da25e 100644
--- a/drivers/misc/habanalabs/goya/goya.c
+++ b/drivers/misc/habanalabs/goya/goya.c
@@ -4838,14 +4838,14 @@ void goya_handle_eqe(struct hl_device *hdev, struct hl_eq_entry *eq_entry)
case GOYA_ASYNC_EVENT_ID_L2_RAM_ECC:
goya_print_irq_info(hdev, event_type, false);
if (hdev->hard_reset_on_fw_events)
- hl_device_reset(hdev, (HL_RESET_HARD |
- HL_RESET_FW_FATAL_ERR));
+ hl_device_reset(hdev, (HL_DRV_RESET_HARD |
+ HL_DRV_RESET_FW_FATAL_ERR));
break;
case GOYA_ASYNC_EVENT_ID_PSOC_GPIO_05_SW_RESET:
goya_print_irq_info(hdev, event_type, false);
if (hdev->hard_reset_on_fw_events)
- hl_device_reset(hdev, HL_RESET_HARD);
+ hl_device_reset(hdev, HL_DRV_RESET_HARD);
break;
case GOYA_ASYNC_EVENT_ID_PCIE_DEC:
@@ -4905,7 +4905,7 @@ void goya_handle_eqe(struct hl_device *hdev, struct hl_eq_entry *eq_entry)
goya_print_irq_info(hdev, event_type, false);
goya_print_out_of_sync_info(hdev, &eq_entry->pkt_sync_err);
if (hdev->hard_reset_on_fw_events)
- hl_device_reset(hdev, HL_RESET_HARD);
+ hl_device_reset(hdev, HL_DRV_RESET_HARD);
else
hl_fw_unmask_irq(hdev, event_type);
break;
@@ -5239,7 +5239,7 @@ static int goya_mmu_invalidate_cache(struct hl_device *hdev, bool is_hard,
if (rc) {
dev_err_ratelimited(hdev->dev,
"MMU cache invalidation timeout\n");
- hl_device_reset(hdev, HL_RESET_HARD);
+ hl_device_reset(hdev, HL_DRV_RESET_HARD);
}
return rc;
--
2.25.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH 05/10] habanalabs: change wait for interrupt timeout to 64 bit
2021-11-07 20:51 [PATCH 01/10] habanalabs: don't clear previous f/w indications Oded Gabbay
` (2 preceding siblings ...)
2021-11-07 20:51 ` [PATCH 04/10] habanalabs: rename reset flags Oded Gabbay
@ 2021-11-07 20:51 ` Oded Gabbay
2021-11-07 20:51 ` [PATCH 06/10] habanalabs: expand clock throttling information uAPI Oded Gabbay
` (4 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Oded Gabbay @ 2021-11-07 20:51 UTC (permalink / raw)
To: linux-kernel; +Cc: Dani Liberman
From: Dani Liberman <dliberman@habana.ai>
In order to increase maximum wait-for-interrupt timeout, change it
to 64 bit variable. This wait is used only by newer ASICs, so no
problem in changing this interface at this time.
Signed-off-by: Dani Liberman <dliberman@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
.../habanalabs/common/command_submission.c | 22 ++++++++++++++-----
include/uapi/misc/habanalabs.h | 18 +++++++++------
2 files changed, 28 insertions(+), 12 deletions(-)
diff --git a/drivers/misc/habanalabs/common/command_submission.c b/drivers/misc/habanalabs/common/command_submission.c
index 9ebcd9894d83..54a5425a77a0 100644
--- a/drivers/misc/habanalabs/common/command_submission.c
+++ b/drivers/misc/habanalabs/common/command_submission.c
@@ -2765,8 +2765,23 @@ static int hl_cs_wait_ioctl(struct hl_fpriv *hpriv, void *data)
return 0;
}
+static inline unsigned long hl_usecs64_to_jiffies(const u64 usecs)
+{
+ if (usecs <= U32_MAX)
+ return usecs_to_jiffies(usecs);
+
+ /*
+ * If the value in nanoseconds is larger than 64 bit, use the largest
+ * 64 bit value.
+ */
+ if (usecs >= ((u64)(U64_MAX / NSEC_PER_USEC)))
+ return nsecs_to_jiffies(U64_MAX);
+
+ return nsecs_to_jiffies(usecs * NSEC_PER_USEC);
+}
+
static int _hl_interrupt_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
- u32 timeout_us, u64 user_address,
+ u64 timeout_us, u64 user_address,
u64 target_value, u16 interrupt_offset,
u32 *status,
u64 *timestamp)
@@ -2778,10 +2793,7 @@ static int _hl_interrupt_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
long completion_rc;
int rc = 0;
- if (timeout_us == U32_MAX)
- timeout = timeout_us;
- else
- timeout = usecs_to_jiffies(timeout_us);
+ timeout = hl_usecs64_to_jiffies(timeout_us);
hl_ctx_get(hdev, ctx);
diff --git a/include/uapi/misc/habanalabs.h b/include/uapi/misc/habanalabs.h
index 00b309590499..c5760acebdd1 100644
--- a/include/uapi/misc/habanalabs.h
+++ b/include/uapi/misc/habanalabs.h
@@ -911,14 +911,18 @@ struct hl_wait_cs_in {
*/
__u32 flags;
- /* Multi CS API info- valid entries in multi-CS array */
- __u8 seq_arr_len;
- __u8 pad[3];
+ union {
+ struct {
+ /* Multi CS API info- valid entries in multi-CS array */
+ __u8 seq_arr_len;
+ __u8 pad[7];
+ };
- /* Absolute timeout to wait for an interrupt in microseconds.
- * Relevant only when HL_WAIT_CS_FLAGS_INTERRUPT is set
- */
- __u32 interrupt_timeout_us;
+ /* Absolute timeout to wait for an interrupt in microseconds.
+ * Relevant only when HL_WAIT_CS_FLAGS_INTERRUPT is set
+ */
+ __u64 interrupt_timeout_us;
+ };
};
#define HL_WAIT_CS_STATUS_COMPLETED 0
--
2.25.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH 06/10] habanalabs: expand clock throttling information uAPI
2021-11-07 20:51 [PATCH 01/10] habanalabs: don't clear previous f/w indications Oded Gabbay
` (3 preceding siblings ...)
2021-11-07 20:51 ` [PATCH 05/10] habanalabs: change wait for interrupt timeout to 64 bit Oded Gabbay
@ 2021-11-07 20:51 ` Oded Gabbay
2021-11-07 20:51 ` [PATCH 07/10] habanalabs/gaudi: Fix collective wait bug Oded Gabbay
` (3 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Oded Gabbay @ 2021-11-07 20:51 UTC (permalink / raw)
To: linux-kernel; +Cc: Ofir Bitton
From: Ofir Bitton <obitton@habana.ai>
In addition to the clock throttling reason, user should be able
to obtain also the start time and the duration of the throttling
event.
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/common/device.c | 3 ++
drivers/misc/habanalabs/common/habanalabs.h | 31 +++++++++++++++++--
.../misc/habanalabs/common/habanalabs_ioctl.c | 27 ++++++++++++++--
drivers/misc/habanalabs/gaudi/gaudi.c | 22 ++++++++++---
drivers/misc/habanalabs/goya/goya.c | 25 ++++++++++++---
include/uapi/misc/habanalabs.h | 16 ++++++++--
6 files changed, 110 insertions(+), 14 deletions(-)
diff --git a/drivers/misc/habanalabs/common/device.c b/drivers/misc/habanalabs/common/device.c
index eb5800b403b6..0da5a55490ff 100644
--- a/drivers/misc/habanalabs/common/device.c
+++ b/drivers/misc/habanalabs/common/device.c
@@ -455,6 +455,7 @@ static int device_early_init(struct hl_device *hdev)
INIT_LIST_HEAD(&hdev->fpriv_list);
mutex_init(&hdev->fpriv_list_lock);
atomic_set(&hdev->in_reset, 0);
+ mutex_init(&hdev->clk_throttling.lock);
return 0;
@@ -495,6 +496,8 @@ static void device_early_fini(struct hl_device *hdev)
mutex_destroy(&hdev->fpriv_list_lock);
+ mutex_destroy(&hdev->clk_throttling.lock);
+
hl_cb_mgr_fini(hdev, &hdev->kernel_cb_mgr);
kfree(hdev->hl_chip_info);
diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index 92d12c8ba569..fc201537f7a9 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -2378,6 +2378,32 @@ struct multi_cs_data {
u8 update_ts;
};
+/**
+ * struct hl_clk_throttle_timestamp - current/last clock throttling timestamp
+ * @start: timestamp taken when 'start' event is received in driver
+ * @end: timestamp taken when 'end' event is received in driver
+ */
+struct hl_clk_throttle_timestamp {
+ ktime_t start;
+ ktime_t end;
+};
+
+/**
+ * struct hl_clk_throttle - keeps current/last clock throttling timestamps
+ * @timestamp: timestamp taken by driver and firmware, index 0 refers to POWER
+ * index 1 refers to THERMAL
+ * @lock: protects this structure as it can be accessed from both event queue
+ * context and info_ioctl context
+ * @current_reason: bitmask represents the current clk throttling reasons
+ * @aggregated_reason: bitmask represents aggregated clk throttling reasons since driver load
+ */
+struct hl_clk_throttle {
+ struct hl_clk_throttle_timestamp timestamp[HL_CLK_THROTTLE_TYPE_MAX];
+ struct mutex lock;
+ u32 current_reason;
+ u32 aggregated_reason;
+};
+
/**
* struct hl_device - habanalabs device structure.
* @pdev: pointer to PCI device, can be NULL in case of simulator device.
@@ -2445,6 +2471,7 @@ struct multi_cs_data {
* @pci_mem_region: array of memory regions in the PCI
* @state_dump_specs: constants and dictionaries needed to dump system state.
* @multi_cs_completion: array of multi-CS completion.
+ * @clk_throttling: holds information about current/previous clock throttling events
* @dram_used_mem: current DRAM memory consumption.
* @timeout_jiffies: device CS timeout value.
* @max_power: the max power of the device, as configured by the sysadmin. This
@@ -2474,7 +2501,6 @@ struct multi_cs_data {
* @high_pll: high PLL profile frequency.
* @soft_reset_cnt: number of soft reset since the driver was loaded.
* @hard_reset_cnt: number of hard reset since the driver was loaded.
- * @clk_throttling_reason: bitmask represents the current clk throttling reasons
* @id: device minor.
* @id_control: minor of the control device
* @cpu_pci_msb_addr: 50-bit extension bits for the device CPU's 40-bit
@@ -2604,6 +2630,8 @@ struct hl_device {
struct multi_cs_completion multi_cs_completion[
MULTI_CS_MAX_USER_CTX];
+ struct hl_clk_throttle clk_throttling;
+
u32 *stream_master_qid_arr;
atomic64_t dram_used_mem;
u64 timeout_jiffies;
@@ -2622,7 +2650,6 @@ struct hl_device {
u32 high_pll;
u32 soft_reset_cnt;
u32 hard_reset_cnt;
- u32 clk_throttling_reason;
u16 id;
u16 id_control;
u16 cpu_pci_msb_addr;
diff --git a/drivers/misc/habanalabs/common/habanalabs_ioctl.c b/drivers/misc/habanalabs/common/habanalabs_ioctl.c
index 86c3257d9ae1..19726c6b642a 100644
--- a/drivers/misc/habanalabs/common/habanalabs_ioctl.c
+++ b/drivers/misc/habanalabs/common/habanalabs_ioctl.c
@@ -313,15 +313,38 @@ static int pci_counters_info(struct hl_fpriv *hpriv, struct hl_info_args *args)
static int clk_throttle_info(struct hl_fpriv *hpriv, struct hl_info_args *args)
{
+ void __user *out = (void __user *) (uintptr_t) args->return_pointer;
struct hl_device *hdev = hpriv->hdev;
struct hl_info_clk_throttle clk_throttle = {0};
+ ktime_t end_time, zero_time = ktime_set(0, 0);
u32 max_size = args->return_size;
- void __user *out = (void __user *) (uintptr_t) args->return_pointer;
+ int i;
if ((!max_size) || (!out))
return -EINVAL;
- clk_throttle.clk_throttling_reason = hdev->clk_throttling_reason;
+ mutex_lock(&hdev->clk_throttling.lock);
+
+ clk_throttle.clk_throttling_reason = hdev->clk_throttling.current_reason;
+
+ for (i = 0 ; i < HL_CLK_THROTTLE_TYPE_MAX ; i++) {
+ if (!(hdev->clk_throttling.aggregated_reason & BIT(i)))
+ continue;
+
+ clk_throttle.clk_throttling_timestamp_us[i] =
+ ktime_to_us(hdev->clk_throttling.timestamp[i].start);
+
+ if (ktime_compare(hdev->clk_throttling.timestamp[i].end, zero_time))
+ end_time = ktime_get();
+ else
+ end_time = hdev->clk_throttling.timestamp[i].end;
+
+ clk_throttle.clk_throttling_duration_ns[i] =
+ ktime_to_ns(ktime_sub(end_time,
+ hdev->clk_throttling.timestamp[i].start));
+
+ }
+ mutex_unlock(&hdev->clk_throttling.lock);
return copy_to_user(out, &clk_throttle,
min((size_t) max_size, sizeof(clk_throttle))) ? -EFAULT : 0;
diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index 2724ab3747f2..b4814369062e 100644
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -7925,27 +7925,39 @@ static int tpc_krn_event_to_tpc_id(u16 tpc_dec_event_type)
static void gaudi_print_clk_change_info(struct hl_device *hdev,
u16 event_type)
{
+ ktime_t zero_time = ktime_set(0, 0);
+
+ mutex_lock(&hdev->clk_throttling.lock);
+
switch (event_type) {
case GAUDI_EVENT_FIX_POWER_ENV_S:
- hdev->clk_throttling_reason |= HL_CLK_THROTTLE_POWER;
+ hdev->clk_throttling.current_reason |= HL_CLK_THROTTLE_POWER;
+ hdev->clk_throttling.aggregated_reason |= HL_CLK_THROTTLE_POWER;
+ hdev->clk_throttling.timestamp[HL_CLK_THROTTLE_TYPE_POWER].start = ktime_get();
+ hdev->clk_throttling.timestamp[HL_CLK_THROTTLE_TYPE_POWER].end = zero_time;
dev_info_ratelimited(hdev->dev,
"Clock throttling due to power consumption\n");
break;
case GAUDI_EVENT_FIX_POWER_ENV_E:
- hdev->clk_throttling_reason &= ~HL_CLK_THROTTLE_POWER;
+ hdev->clk_throttling.current_reason &= ~HL_CLK_THROTTLE_POWER;
+ hdev->clk_throttling.timestamp[HL_CLK_THROTTLE_TYPE_POWER].end = ktime_get();
dev_info_ratelimited(hdev->dev,
"Power envelop is safe, back to optimal clock\n");
break;
case GAUDI_EVENT_FIX_THERMAL_ENV_S:
- hdev->clk_throttling_reason |= HL_CLK_THROTTLE_THERMAL;
+ hdev->clk_throttling.current_reason |= HL_CLK_THROTTLE_THERMAL;
+ hdev->clk_throttling.aggregated_reason |= HL_CLK_THROTTLE_THERMAL;
+ hdev->clk_throttling.timestamp[HL_CLK_THROTTLE_TYPE_THERMAL].start = ktime_get();
+ hdev->clk_throttling.timestamp[HL_CLK_THROTTLE_TYPE_THERMAL].end = zero_time;
dev_info_ratelimited(hdev->dev,
"Clock throttling due to overheating\n");
break;
case GAUDI_EVENT_FIX_THERMAL_ENV_E:
- hdev->clk_throttling_reason &= ~HL_CLK_THROTTLE_THERMAL;
+ hdev->clk_throttling.current_reason &= ~HL_CLK_THROTTLE_THERMAL;
+ hdev->clk_throttling.timestamp[HL_CLK_THROTTLE_TYPE_THERMAL].end = ktime_get();
dev_info_ratelimited(hdev->dev,
"Thermal envelop is safe, back to optimal clock\n");
break;
@@ -7955,6 +7967,8 @@ static void gaudi_print_clk_change_info(struct hl_device *hdev,
event_type);
break;
}
+
+ mutex_unlock(&hdev->clk_throttling.lock);
}
static void gaudi_handle_eqe(struct hl_device *hdev,
diff --git a/drivers/misc/habanalabs/goya/goya.c b/drivers/misc/habanalabs/goya/goya.c
index 3bbcab7da25e..7b3683f2a6dc 100644
--- a/drivers/misc/habanalabs/goya/goya.c
+++ b/drivers/misc/habanalabs/goya/goya.c
@@ -4768,24 +4768,39 @@ static int goya_unmask_irq(struct hl_device *hdev, u16 event_type)
static void goya_print_clk_change_info(struct hl_device *hdev, u16 event_type)
{
+ ktime_t zero_time = ktime_set(0, 0);
+
+ mutex_lock(&hdev->clk_throttling.lock);
+
switch (event_type) {
case GOYA_ASYNC_EVENT_ID_FIX_POWER_ENV_S:
- hdev->clk_throttling_reason |= HL_CLK_THROTTLE_POWER;
+ hdev->clk_throttling.current_reason |= HL_CLK_THROTTLE_POWER;
+ hdev->clk_throttling.aggregated_reason |= HL_CLK_THROTTLE_POWER;
+ hdev->clk_throttling.timestamp[HL_CLK_THROTTLE_TYPE_POWER].start = ktime_get();
+ hdev->clk_throttling.timestamp[HL_CLK_THROTTLE_TYPE_POWER].end = zero_time;
dev_info_ratelimited(hdev->dev,
"Clock throttling due to power consumption\n");
break;
+
case GOYA_ASYNC_EVENT_ID_FIX_POWER_ENV_E:
- hdev->clk_throttling_reason &= ~HL_CLK_THROTTLE_POWER;
+ hdev->clk_throttling.current_reason &= ~HL_CLK_THROTTLE_POWER;
+ hdev->clk_throttling.timestamp[HL_CLK_THROTTLE_TYPE_POWER].end = ktime_get();
dev_info_ratelimited(hdev->dev,
"Power envelop is safe, back to optimal clock\n");
break;
+
case GOYA_ASYNC_EVENT_ID_FIX_THERMAL_ENV_S:
- hdev->clk_throttling_reason |= HL_CLK_THROTTLE_THERMAL;
+ hdev->clk_throttling.current_reason |= HL_CLK_THROTTLE_THERMAL;
+ hdev->clk_throttling.aggregated_reason |= HL_CLK_THROTTLE_THERMAL;
+ hdev->clk_throttling.timestamp[HL_CLK_THROTTLE_TYPE_THERMAL].start = ktime_get();
+ hdev->clk_throttling.timestamp[HL_CLK_THROTTLE_TYPE_THERMAL].end = zero_time;
dev_info_ratelimited(hdev->dev,
"Clock throttling due to overheating\n");
break;
+
case GOYA_ASYNC_EVENT_ID_FIX_THERMAL_ENV_E:
- hdev->clk_throttling_reason &= ~HL_CLK_THROTTLE_THERMAL;
+ hdev->clk_throttling.current_reason &= ~HL_CLK_THROTTLE_THERMAL;
+ hdev->clk_throttling.timestamp[HL_CLK_THROTTLE_TYPE_THERMAL].end = ktime_get();
dev_info_ratelimited(hdev->dev,
"Thermal envelop is safe, back to optimal clock\n");
break;
@@ -4795,6 +4810,8 @@ static void goya_print_clk_change_info(struct hl_device *hdev, u16 event_type)
event_type);
break;
}
+
+ mutex_unlock(&hdev->clk_throttling.lock);
}
void goya_handle_eqe(struct hl_device *hdev, struct hl_eq_entry *eq_entry)
diff --git a/include/uapi/misc/habanalabs.h b/include/uapi/misc/habanalabs.h
index c5760acebdd1..257b9630773e 100644
--- a/include/uapi/misc/habanalabs.h
+++ b/include/uapi/misc/habanalabs.h
@@ -473,15 +473,27 @@ struct hl_info_pci_counters {
__u64 replay_cnt;
};
-#define HL_CLK_THROTTLE_POWER 0x1
-#define HL_CLK_THROTTLE_THERMAL 0x2
+enum hl_clk_throttling_type {
+ HL_CLK_THROTTLE_TYPE_POWER,
+ HL_CLK_THROTTLE_TYPE_THERMAL,
+ HL_CLK_THROTTLE_TYPE_MAX
+};
+
+/* clk_throttling_reason masks */
+#define HL_CLK_THROTTLE_POWER (1 << HL_CLK_THROTTLE_TYPE_POWER)
+#define HL_CLK_THROTTLE_THERMAL (1 << HL_CLK_THROTTLE_TYPE_THERMAL)
/**
* struct hl_info_clk_throttle - clock throttling reason
* @clk_throttling_reason: each bit represents a clk throttling reason
+ * @clk_throttling_timestamp_us: represents CPU timestamp in microseconds of the start-event
+ * @clk_throttling_duration_ns: the clock throttle time in nanosec
*/
struct hl_info_clk_throttle {
__u32 clk_throttling_reason;
+ __u32 pad;
+ __u64 clk_throttling_timestamp_us[HL_CLK_THROTTLE_TYPE_MAX];
+ __u64 clk_throttling_duration_ns[HL_CLK_THROTTLE_TYPE_MAX];
};
/**
--
2.25.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH 07/10] habanalabs/gaudi: Fix collective wait bug
2021-11-07 20:51 [PATCH 01/10] habanalabs: don't clear previous f/w indications Oded Gabbay
` (4 preceding siblings ...)
2021-11-07 20:51 ` [PATCH 06/10] habanalabs: expand clock throttling information uAPI Oded Gabbay
@ 2021-11-07 20:51 ` Oded Gabbay
2021-11-07 20:51 ` [PATCH 08/10] habanalabs: refactor wait-for-user-interrupt function Oded Gabbay
` (2 subsequent siblings)
8 siblings, 0 replies; 10+ messages in thread
From: Oded Gabbay @ 2021-11-07 20:51 UTC (permalink / raw)
To: linux-kernel; +Cc: farah kassabri
From: farah kassabri <fkassabri@habana.ai>
In Signaling-From-Graph case, the driver didn't set the hw_sob pointer
at the right place, which is needed for the cs completion
check prior to start sending all the master/slaves jobs to device.
Signed-off-by: farah kassabri <fkassabri@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
drivers/misc/habanalabs/gaudi/gaudi.c | 15 +++++++++++----
1 file changed, 11 insertions(+), 4 deletions(-)
diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index b4814369062e..a9e279bfebae 100644
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -1276,6 +1276,7 @@ static int gaudi_collective_wait_init_cs(struct hl_cs *cs)
container_of(cs->signal_fence, struct hl_cs_compl, base_fence);
struct hl_cs_compl *cs_cmpl =
container_of(cs->fence, struct hl_cs_compl, base_fence);
+ struct hl_cs_encaps_sig_handle *handle = cs->encaps_sig_hdl;
struct gaudi_collective_properties *cprop;
u32 stream, queue_id, sob_group_offset;
struct gaudi_device *gaudi;
@@ -1288,10 +1289,16 @@ static int gaudi_collective_wait_init_cs(struct hl_cs *cs)
gaudi = hdev->asic_specific;
cprop = &gaudi->collective_props;
- /* In encaps signals case the SOB info will be retrieved from
- * the handle in gaudi_collective_slave_init_job.
- */
- if (!cs->encaps_signals) {
+ if (cs->encaps_signals) {
+ cs_cmpl->hw_sob = handle->hw_sob;
+ /* at this checkpoint we only need the hw_sob pointer
+ * for the completion check before start going over the jobs
+ * of the master/slaves, the sob_value will be taken later on
+ * in gaudi_collective_slave_init_job depends on each
+ * job wait offset value.
+ */
+ cs_cmpl->sob_val = 0;
+ } else {
/* copy the SOB id and value of the signal CS */
cs_cmpl->hw_sob = signal_cs_cmpl->hw_sob;
cs_cmpl->sob_val = signal_cs_cmpl->sob_val;
--
2.25.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH 08/10] habanalabs: refactor wait-for-user-interrupt function
2021-11-07 20:51 [PATCH 01/10] habanalabs: don't clear previous f/w indications Oded Gabbay
` (5 preceding siblings ...)
2021-11-07 20:51 ` [PATCH 07/10] habanalabs/gaudi: Fix collective wait bug Oded Gabbay
@ 2021-11-07 20:51 ` Oded Gabbay
2021-11-07 20:51 ` [PATCH 09/10] habanalabs: add new opcodes for INFO IOCTL Oded Gabbay
2021-11-07 20:51 ` [PATCH 10/10] habanalabs: make hdev creation code more readable Oded Gabbay
8 siblings, 0 replies; 10+ messages in thread
From: Oded Gabbay @ 2021-11-07 20:51 UTC (permalink / raw)
To: linux-kernel; +Cc: Bharat Jauhari
From: Bharat Jauhari <bjauhari@habana.ai>
Refactor the wait-for-user-interrupt routine to make it more
generic for re-use for other user exposed h/w interfaces in future
ASICs.
Signed-off-by: Bharat Jauhari <bjauhari@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
.../habanalabs/common/command_submission.c | 22 +++++++------------
1 file changed, 8 insertions(+), 14 deletions(-)
diff --git a/drivers/misc/habanalabs/common/command_submission.c b/drivers/misc/habanalabs/common/command_submission.c
index 54a5425a77a0..e97b21988dea 100644
--- a/drivers/misc/habanalabs/common/command_submission.c
+++ b/drivers/misc/habanalabs/common/command_submission.c
@@ -2782,12 +2782,12 @@ static inline unsigned long hl_usecs64_to_jiffies(const u64 usecs)
static int _hl_interrupt_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
u64 timeout_us, u64 user_address,
- u64 target_value, u16 interrupt_offset,
+ u64 target_value, struct hl_user_interrupt *interrupt,
+
u32 *status,
u64 *timestamp)
{
struct hl_user_pending_interrupt *pend;
- struct hl_user_interrupt *interrupt;
unsigned long timeout, flags;
u64 completion_value;
long completion_rc;
@@ -2805,11 +2805,6 @@ static int _hl_interrupt_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
hl_fence_init(&pend->fence, ULONG_MAX);
- if (interrupt_offset == HL_COMMON_USER_INTERRUPT_ID)
- interrupt = &hdev->common_user_interrupt;
- else
- interrupt = &hdev->user_interrupt[interrupt_offset];
-
/* Add pending user interrupt to relevant list for the interrupt
* handler to monitor
*/
@@ -2898,9 +2893,10 @@ static int _hl_interrupt_wait_ioctl(struct hl_device *hdev, struct hl_ctx *ctx,
static int hl_interrupt_wait_ioctl(struct hl_fpriv *hpriv, void *data)
{
- u16 interrupt_id, interrupt_offset, first_interrupt, last_interrupt;
+ u16 interrupt_id, first_interrupt, last_interrupt;
struct hl_device *hdev = hpriv->hdev;
struct asic_fixed_properties *prop;
+ struct hl_user_interrupt *interrupt;
union hl_wait_cs_args *args = data;
u32 status = HL_WAIT_CS_STATUS_BUSY;
u64 timestamp;
@@ -2913,8 +2909,7 @@ static int hl_interrupt_wait_ioctl(struct hl_fpriv *hpriv, void *data)
return -EPERM;
}
- interrupt_id =
- FIELD_GET(HL_WAIT_CS_FLAGS_INTERRUPT_MASK, args->in.flags);
+ interrupt_id = FIELD_GET(HL_WAIT_CS_FLAGS_INTERRUPT_MASK, args->in.flags);
first_interrupt = prop->first_available_user_msix_interrupt;
last_interrupt = prop->first_available_user_msix_interrupt +
@@ -2927,15 +2922,14 @@ static int hl_interrupt_wait_ioctl(struct hl_fpriv *hpriv, void *data)
}
if (interrupt_id == HL_COMMON_USER_INTERRUPT_ID)
- interrupt_offset = HL_COMMON_USER_INTERRUPT_ID;
+ interrupt = &hdev->common_user_interrupt;
else
- interrupt_offset = interrupt_id - first_interrupt;
+ interrupt = &hdev->user_interrupt[interrupt_id - first_interrupt];
rc = _hl_interrupt_wait_ioctl(hdev, hpriv->ctx,
args->in.interrupt_timeout_us, args->in.addr,
- args->in.target, interrupt_offset, &status,
+ args->in.target, interrupt, &status,
×tamp);
-
if (rc) {
if (rc != -EINTR)
dev_err_ratelimited(hdev->dev,
--
2.25.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH 09/10] habanalabs: add new opcodes for INFO IOCTL
2021-11-07 20:51 [PATCH 01/10] habanalabs: don't clear previous f/w indications Oded Gabbay
` (6 preceding siblings ...)
2021-11-07 20:51 ` [PATCH 08/10] habanalabs: refactor wait-for-user-interrupt function Oded Gabbay
@ 2021-11-07 20:51 ` Oded Gabbay
2021-11-07 20:51 ` [PATCH 10/10] habanalabs: make hdev creation code more readable Oded Gabbay
8 siblings, 0 replies; 10+ messages in thread
From: Oded Gabbay @ 2021-11-07 20:51 UTC (permalink / raw)
To: linux-kernel; +Cc: farah kassabri
From: farah kassabri <fkassabri@habana.ai>
Add implementation for new opcodes in the INFO IOCTL:
1. Retrieve the replaced DRAM rows from f/w.
2. Retrieve the pending DRAM rows from f/w.
Signed-off-by: farah kassabri <fkassabri@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
drivers/misc/habanalabs/common/firmware_if.c | 66 +++++++++++++++++++
drivers/misc/habanalabs/common/habanalabs.h | 3 +
.../misc/habanalabs/common/habanalabs_ioctl.c | 43 ++++++++++++
.../misc/habanalabs/include/common/cpucp_if.h | 33 +++++++++-
include/uapi/misc/habanalabs.h | 4 ++
5 files changed, 148 insertions(+), 1 deletion(-)
diff --git a/drivers/misc/habanalabs/common/firmware_if.c b/drivers/misc/habanalabs/common/firmware_if.c
index 9addcfba6a8b..70e992bdbde7 100644
--- a/drivers/misc/habanalabs/common/firmware_if.c
+++ b/drivers/misc/habanalabs/common/firmware_if.c
@@ -972,6 +972,72 @@ int hl_fw_cpucp_power_get(struct hl_device *hdev, u64 *power)
return rc;
}
+int hl_fw_dram_replaced_row_get(struct hl_device *hdev,
+ struct cpucp_hbm_row_info *info)
+{
+ struct cpucp_hbm_row_info *cpucp_repl_rows_info_cpu_addr;
+ dma_addr_t cpucp_repl_rows_info_dma_addr;
+ struct cpucp_packet pkt = {};
+ u64 result;
+ int rc;
+
+ cpucp_repl_rows_info_cpu_addr =
+ hdev->asic_funcs->cpu_accessible_dma_pool_alloc(hdev,
+ sizeof(struct cpucp_hbm_row_info),
+ &cpucp_repl_rows_info_dma_addr);
+ if (!cpucp_repl_rows_info_cpu_addr) {
+ dev_err(hdev->dev,
+ "Failed to allocate DMA memory for CPU-CP replaced rows info packet\n");
+ return -ENOMEM;
+ }
+
+ memset(cpucp_repl_rows_info_cpu_addr, 0, sizeof(struct cpucp_hbm_row_info));
+
+ pkt.ctl = cpu_to_le32(CPUCP_PACKET_HBM_REPLACED_ROWS_INFO_GET <<
+ CPUCP_PKT_CTL_OPCODE_SHIFT);
+ pkt.addr = cpu_to_le64(cpucp_repl_rows_info_dma_addr);
+ pkt.data_max_size = cpu_to_le32(sizeof(struct cpucp_hbm_row_info));
+
+ rc = hdev->asic_funcs->send_cpu_message(hdev, (u32 *) &pkt, sizeof(pkt),
+ HL_CPUCP_INFO_TIMEOUT_USEC, &result);
+ if (rc) {
+ dev_err(hdev->dev,
+ "Failed to handle CPU-CP replaced rows info pkt, error %d\n", rc);
+ goto out;
+ }
+
+ memcpy(info, cpucp_repl_rows_info_cpu_addr, sizeof(*info));
+
+out:
+ hdev->asic_funcs->cpu_accessible_dma_pool_free(hdev,
+ sizeof(struct cpucp_hbm_row_info),
+ cpucp_repl_rows_info_cpu_addr);
+
+ return rc;
+}
+
+int hl_fw_dram_pending_row_get(struct hl_device *hdev, u32 *pend_rows_num)
+{
+ struct cpucp_packet pkt;
+ u64 result;
+ int rc;
+
+ memset(&pkt, 0, sizeof(pkt));
+
+ pkt.ctl = cpu_to_le32(CPUCP_PACKET_HBM_PENDING_ROWS_STATUS << CPUCP_PKT_CTL_OPCODE_SHIFT);
+
+ rc = hdev->asic_funcs->send_cpu_message(hdev, (u32 *) &pkt, sizeof(pkt), 0, &result);
+ if (rc) {
+ dev_err(hdev->dev,
+ "Failed to handle CPU-CP pending rows info pkt, error %d\n", rc);
+ goto out;
+ }
+
+ *pend_rows_num = (u32) result;
+out:
+ return rc;
+}
+
void hl_fw_ask_hard_reset_without_linux(struct hl_device *hdev)
{
struct static_fw_load_mgr *static_loader =
diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index fc201537f7a9..a19563c416ac 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -3012,6 +3012,9 @@ int hl_fw_dynamic_send_protocol_cmd(struct hl_device *hdev,
struct fw_load_mgr *fw_loader,
enum comms_cmd cmd, unsigned int size,
bool wait_ok, u32 timeout);
+int hl_fw_dram_replaced_row_get(struct hl_device *hdev,
+ struct cpucp_hbm_row_info *info);
+int hl_fw_dram_pending_row_get(struct hl_device *hdev, u32 *pend_rows_num);
int hl_pci_bars_map(struct hl_device *hdev, const char * const name[3],
bool is_wc[3]);
int hl_pci_elbi_read(struct hl_device *hdev, u64 addr, u32 *data);
diff --git a/drivers/misc/habanalabs/common/habanalabs_ioctl.c b/drivers/misc/habanalabs/common/habanalabs_ioctl.c
index 19726c6b642a..68c655acdec8 100644
--- a/drivers/misc/habanalabs/common/habanalabs_ioctl.c
+++ b/drivers/misc/habanalabs/common/habanalabs_ioctl.c
@@ -503,6 +503,43 @@ static int open_stats_info(struct hl_fpriv *hpriv, struct hl_info_args *args)
min((size_t) max_size, sizeof(open_stats_info))) ? -EFAULT : 0;
}
+static int dram_pending_rows_info(struct hl_fpriv *hpriv, struct hl_info_args *args)
+{
+ struct hl_device *hdev = hpriv->hdev;
+ u32 max_size = args->return_size;
+ u32 pend_rows_num = 0;
+ void __user *out = (void __user *) (uintptr_t) args->return_pointer;
+ int rc;
+
+ if ((!max_size) || (!out))
+ return -EINVAL;
+
+ rc = hl_fw_dram_pending_row_get(hdev, &pend_rows_num);
+ if (rc)
+ return rc;
+
+ return copy_to_user(out, &pend_rows_num,
+ min_t(size_t, max_size, sizeof(pend_rows_num))) ? -EFAULT : 0;
+}
+
+static int dram_replaced_rows_info(struct hl_fpriv *hpriv, struct hl_info_args *args)
+{
+ struct hl_device *hdev = hpriv->hdev;
+ u32 max_size = args->return_size;
+ struct cpucp_hbm_row_info info = {0};
+ void __user *out = (void __user *) (uintptr_t) args->return_pointer;
+ int rc;
+
+ if ((!max_size) || (!out))
+ return -EINVAL;
+
+ rc = hl_fw_dram_replaced_row_get(hdev, &info);
+ if (rc)
+ return rc;
+
+ return copy_to_user(out, &info, min_t(size_t, max_size, sizeof(info))) ? -EFAULT : 0;
+}
+
static int _hl_info_ioctl(struct hl_fpriv *hpriv, void *data,
struct device *dev)
{
@@ -589,6 +626,12 @@ static int _hl_info_ioctl(struct hl_fpriv *hpriv, void *data,
case HL_INFO_OPEN_STATS:
return open_stats_info(hpriv, args);
+ case HL_INFO_DRAM_REPLACED_ROWS:
+ return dram_replaced_rows_info(hpriv, args);
+
+ case HL_INFO_DRAM_PENDING_ROWS:
+ return dram_pending_rows_info(hpriv, args);
+
default:
dev_err(dev, "Invalid request %d\n", args->op);
rc = -ENOTTY;
diff --git a/drivers/misc/habanalabs/include/common/cpucp_if.h b/drivers/misc/habanalabs/include/common/cpucp_if.h
index 17927968e19a..5e19c763f3f0 100644
--- a/drivers/misc/habanalabs/include/common/cpucp_if.h
+++ b/drivers/misc/habanalabs/include/common/cpucp_if.h
@@ -1,6 +1,6 @@
/* SPDX-License-Identifier: GPL-2.0
*
- * Copyright 2020 HabanaLabs, Ltd.
+ * Copyright 2021 HabanaLabs, Ltd.
* All Rights Reserved.
*
*/
@@ -377,6 +377,13 @@ enum pq_init_status {
* a different engine or QMAN according to enum cpucp_idle_mask.
* The bit will be 1 if the engine is NOT idle.
*
+ * CPUCP_PACKET_HBM_REPLACED_ROWS_INFO_GET -
+ * Fetch all HBM replaced-rows and prending to be replaced rows data.
+ *
+ * CPUCP_PACKET_HBM_PENDING_ROWS_STATUS -
+ * Fetch status of HBM rows pending replacement and need a reboot to
+ * be replaced.
+ *
* CPUCP_PACKET_POWER_SET -
* Resets power history of device to 0
*/
@@ -424,6 +431,8 @@ enum cpucp_packet_id {
CPUCP_PACKET_NIC_STAT_REGS_CLR, /* internal */
CPUCP_PACKET_NIC_STAT_REGS_ALL_GET, /* internal */
CPUCP_PACKET_IS_IDLE_CHECK, /* internal */
+ CPUCP_PACKET_HBM_REPLACED_ROWS_INFO_GET,/* internal */
+ CPUCP_PACKET_HBM_PENDING_ROWS_STATUS, /* internal */
CPUCP_PACKET_POWER_SET, /* internal */
};
@@ -692,6 +701,7 @@ struct eq_generic_event {
#define CPUCP_MAX_NIC_LANES (CPUCP_MAX_NICS * CPUCP_LANES_PER_NIC)
#define CPUCP_NIC_MASK_ARR_LEN ((CPUCP_MAX_NICS + 63) / 64)
#define CPUCP_NIC_POLARITY_ARR_LEN ((CPUCP_MAX_NIC_LANES + 63) / 64)
+#define CPUCP_HBM_ROW_REPLACE_MAX 32
struct cpucp_sensor {
__le32 type;
@@ -837,4 +847,25 @@ struct cpucp_nic_status {
__le32 high_ber_cnt;
};
+enum cpucp_hbm_row_replace_cause {
+ REPLACE_CAUSE_DOUBLE_ECC_ERR,
+ REPLACE_CAUSE_MULTI_SINGLE_ECC_ERR,
+};
+
+struct cpucp_hbm_row_info {
+ __u8 hbm_idx;
+ __u8 pc;
+ __u8 sid;
+ __u8 bank_idx;
+ __le16 row_addr;
+ __u8 replaced_row_cause; /* enum cpucp_hbm_row_replace_cause */
+ __u8 pad;
+};
+
+struct cpucp_hbm_row_replaced_rows_info {
+ __le16 num_replaced_rows;
+ __u8 pad[6];
+ struct cpucp_hbm_row_info replaced_rows[CPUCP_HBM_ROW_REPLACE_MAX];
+};
+
#endif /* CPUCP_IF_H */
diff --git a/include/uapi/misc/habanalabs.h b/include/uapi/misc/habanalabs.h
index 257b9630773e..9b4d72897061 100644
--- a/include/uapi/misc/habanalabs.h
+++ b/include/uapi/misc/habanalabs.h
@@ -334,6 +334,8 @@ enum hl_server_type {
* HL_INFO_TOTAL_ENERGY - Retrieve total energy consumption
* HL_INFO_PLL_FREQUENCY - Retrieve PLL frequency
* HL_INFO_OPEN_STATS - Retrieve info regarding recent device open calls
+ * HL_INFO_DRAM_REPLACED_ROWS - Retrieve DRAM replaced rows info
+ * HL_INFO_DRAM_PENDING_ROWS - Retrieve DRAM pending rows num
*/
#define HL_INFO_HW_IP_INFO 0
#define HL_INFO_HW_EVENTS 1
@@ -353,6 +355,8 @@ enum hl_server_type {
#define HL_INFO_PLL_FREQUENCY 16
#define HL_INFO_POWER 17
#define HL_INFO_OPEN_STATS 18
+#define HL_INFO_DRAM_REPLACED_ROWS 21
+#define HL_INFO_DRAM_PENDING_ROWS 22
#define HL_INFO_VERSION_MAX_LEN 128
#define HL_INFO_CARD_NAME_MAX_LEN 16
--
2.25.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH 10/10] habanalabs: make hdev creation code more readable
2021-11-07 20:51 [PATCH 01/10] habanalabs: don't clear previous f/w indications Oded Gabbay
` (7 preceding siblings ...)
2021-11-07 20:51 ` [PATCH 09/10] habanalabs: add new opcodes for INFO IOCTL Oded Gabbay
@ 2021-11-07 20:51 ` Oded Gabbay
8 siblings, 0 replies; 10+ messages in thread
From: Oded Gabbay @ 2021-11-07 20:51 UTC (permalink / raw)
To: linux-kernel
Divide the code into 3 different parts:
- Copy kernel parameters
- Setting device behaivor per asic
- Fixup of various device parameters according to the device behaivor.
In addition, remove non-relevant code for upstream (simulator support).
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
---
drivers/misc/habanalabs/common/habanalabs.h | 6 +-
.../misc/habanalabs/common/habanalabs_drv.c | 123 +++++++++---------
2 files changed, 61 insertions(+), 68 deletions(-)
diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index a19563c416ac..6b33fbd72fd8 100644
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -1,6 +1,6 @@
/* SPDX-License-Identifier: GPL-2.0
*
- * Copyright 2016-2019 HabanaLabs, Ltd.
+ * Copyright 2016-2021 HabanaLabs, Ltd.
* All Rights Reserved.
*
*/
@@ -62,7 +62,6 @@
#define HL_CPUCP_EEPROM_TIMEOUT_USEC 10000000 /* 10s */
#define HL_FW_STATUS_POLL_INTERVAL_USEC 10000 /* 10ms */
-#define HL_FW_STATUS_PLDM_POLL_INTERVAL_USEC 300000000 /* 300s */
#define HL_PCI_ELBI_TIMEOUT_MSEC 10 /* 10ms */
@@ -2823,9 +2822,6 @@ bool hl_device_operational(struct hl_device *hdev,
enum hl_device_status *status);
enum hl_device_status hl_device_status(struct hl_device *hdev);
int hl_device_set_debug_mode(struct hl_device *hdev, bool enable);
-int create_hdev(struct hl_device **dev, struct pci_dev *pdev,
- enum hl_asic_type asic_type, int minor);
-void destroy_hdev(struct hl_device *hdev);
int hl_hw_queues_create(struct hl_device *hdev);
void hl_hw_queues_destroy(struct hl_device *hdev);
int hl_hw_queue_send_cb_no_cmpl(struct hl_device *hdev, u32 hw_queue_id,
diff --git a/drivers/misc/habanalabs/common/habanalabs_drv.c b/drivers/misc/habanalabs/common/habanalabs_drv.c
index 5989826701bc..85034f2f2e89 100644
--- a/drivers/misc/habanalabs/common/habanalabs_drv.c
+++ b/drivers/misc/habanalabs/common/habanalabs_drv.c
@@ -1,7 +1,7 @@
// SPDX-License-Identifier: GPL-2.0
/*
- * Copyright 2016-2019 HabanaLabs, Ltd.
+ * Copyright 2016-2021 HabanaLabs, Ltd.
* All Rights Reserved.
*
*/
@@ -263,6 +263,7 @@ int hl_device_open_ctrl(struct inode *inode, struct file *filp)
static void set_driver_behavior_per_device(struct hl_device *hdev)
{
+ hdev->pldm = 0;
hdev->fw_components = FW_TYPE_ALL_TYPES;
hdev->cpu_queues_enable = 1;
hdev->heartbeat = 1;
@@ -279,23 +280,53 @@ static void set_driver_behavior_per_device(struct hl_device *hdev)
hdev->axi_drain = 0;
}
-/*
+static void copy_kernel_module_params_to_device(struct hl_device *hdev)
+{
+ hdev->major = hl_major;
+ hdev->memory_scrub = memory_scrub;
+ hdev->reset_on_lockup = reset_on_lockup;
+ hdev->boot_error_status_mask = boot_error_status_mask;
+
+ if (timeout_locked)
+ hdev->timeout_jiffies = msecs_to_jiffies(timeout_locked * 1000);
+ else
+ hdev->timeout_jiffies = MAX_SCHEDULE_TIMEOUT;
+
+}
+
+static int fixup_device_params(struct hl_device *hdev)
+{
+ hdev->asic_prop.fw_security_enabled = is_asic_secured(hdev->asic_type);
+
+ hdev->fw_poll_interval_usec = HL_FW_STATUS_POLL_INTERVAL_USEC;
+
+ hdev->stop_on_err = true;
+ hdev->curr_reset_cause = HL_RESET_CAUSE_UNKNOWN;
+ hdev->prev_reset_trigger = HL_RESET_TRIGGER_DEFAULT;
+
+ /* Enable only after the initialization of the device */
+ hdev->disabled = true;
+
+ /* Set default DMA mask to 32 bits */
+ hdev->dma_mask = 32;
+
+ return 0;
+}
+
+/**
* create_hdev - create habanalabs device instance
*
* @dev: will hold the pointer to the new habanalabs device structure
* @pdev: pointer to the pci device
- * @asic_type: in case of simulator device, which device is it
- * @minor: in case of simulator device, the minor of the device
*
* Allocate memory for habanalabs device and initialize basic fields
* Identify the ASIC type
* Allocate ID (minor) for the device (only for real devices)
*/
-int create_hdev(struct hl_device **dev, struct pci_dev *pdev,
- enum hl_asic_type asic_type, int minor)
+static int create_hdev(struct hl_device **dev, struct pci_dev *pdev)
{
+ int main_id, ctrl_id = 0, rc = 0;
struct hl_device *hdev;
- int rc, main_id, ctrl_id = 0;
*dev = NULL;
@@ -303,72 +334,39 @@ int create_hdev(struct hl_device **dev, struct pci_dev *pdev,
if (!hdev)
return -ENOMEM;
- /* First, we must find out which ASIC are we handling. This is needed
- * to configure the behavior of the driver (kernel parameters)
- */
- if (pdev) {
- hdev->asic_type = get_asic_type(pdev->device);
- if (hdev->asic_type == ASIC_INVALID) {
- dev_err(&pdev->dev, "Unsupported ASIC\n");
- rc = -ENODEV;
- goto free_hdev;
- }
- } else {
- hdev->asic_type = asic_type;
- }
-
- if (pdev)
- hdev->asic_prop.fw_security_enabled =
- is_asic_secured(hdev->asic_type);
- else
- hdev->asic_prop.fw_security_enabled = false;
+ /* can be NULL in case of simulator device */
+ hdev->pdev = pdev;
/* Assign status description string */
- strncpy(hdev->status[HL_DEVICE_STATUS_OPERATIONAL],
- "operational", HL_STR_MAX);
- strncpy(hdev->status[HL_DEVICE_STATUS_IN_RESET],
- "in reset", HL_STR_MAX);
- strncpy(hdev->status[HL_DEVICE_STATUS_MALFUNCTION],
- "disabled", HL_STR_MAX);
- strncpy(hdev->status[HL_DEVICE_STATUS_NEEDS_RESET],
- "needs reset", HL_STR_MAX);
+ strncpy(hdev->status[HL_DEVICE_STATUS_OPERATIONAL], "operational", HL_STR_MAX);
+ strncpy(hdev->status[HL_DEVICE_STATUS_IN_RESET], "in reset", HL_STR_MAX);
+ strncpy(hdev->status[HL_DEVICE_STATUS_MALFUNCTION], "disabled", HL_STR_MAX);
+ strncpy(hdev->status[HL_DEVICE_STATUS_NEEDS_RESET], "needs reset", HL_STR_MAX);
strncpy(hdev->status[HL_DEVICE_STATUS_IN_DEVICE_CREATION],
"in device creation", HL_STR_MAX);
- hdev->major = hl_major;
- hdev->reset_on_lockup = reset_on_lockup;
- hdev->memory_scrub = memory_scrub;
- hdev->boot_error_status_mask = boot_error_status_mask;
- hdev->stop_on_err = true;
+ /* First, we must find out which ASIC are we handling. This is needed
+ * to configure the behavior of the driver (kernel parameters)
+ */
+ hdev->asic_type = get_asic_type(pdev->device);
+ if (hdev->asic_type == ASIC_INVALID) {
+ dev_err(&pdev->dev, "Unsupported ASIC\n");
+ rc = -ENODEV;
+ goto free_hdev;
+ }
- hdev->pldm = 0;
+ copy_kernel_module_params_to_device(hdev);
set_driver_behavior_per_device(hdev);
- hdev->fw_poll_interval_usec = hdev->pldm ? HL_FW_STATUS_PLDM_POLL_INTERVAL_USEC :
- HL_FW_STATUS_POLL_INTERVAL_USEC;
-
- hdev->curr_reset_cause = HL_RESET_CAUSE_UNKNOWN;
- hdev->prev_reset_trigger = HL_RESET_TRIGGER_DEFAULT;
-
- if (timeout_locked)
- hdev->timeout_jiffies = msecs_to_jiffies(timeout_locked * 1000);
- else
- hdev->timeout_jiffies = MAX_SCHEDULE_TIMEOUT;
-
- hdev->disabled = true;
- hdev->pdev = pdev; /* can be NULL in case of simulator device */
-
- /* Set default DMA mask to 32 bits */
- hdev->dma_mask = 32;
+ fixup_device_params(hdev);
mutex_lock(&hl_devs_idr_lock);
/* Always save 2 numbers, 1 for main device and 1 for control.
* They must be consecutive
*/
- main_id = idr_alloc(&hl_devs_idr, hdev, 0, HL_MAX_MINORS,
- GFP_KERNEL);
+ main_id = idr_alloc(&hl_devs_idr, hdev, 0, HL_MAX_MINORS, GFP_KERNEL);
if (main_id >= 0)
ctrl_id = idr_alloc(&hl_devs_idr, hdev, main_id + 1,
@@ -408,7 +406,7 @@ int create_hdev(struct hl_device **dev, struct pci_dev *pdev,
* @dev: pointer to the habanalabs device structure
*
*/
-void destroy_hdev(struct hl_device *hdev)
+static void destroy_hdev(struct hl_device *hdev)
{
/* Remove device from the device list */
mutex_lock(&hl_devs_idr_lock);
@@ -447,7 +445,7 @@ static int hl_pmops_resume(struct device *dev)
return hl_device_resume(hdev);
}
-/*
+/**
* hl_pci_probe - probe PCI habanalabs devices
*
* @pdev: pointer to pci device
@@ -457,8 +455,7 @@ static int hl_pmops_resume(struct device *dev)
* Create a new habanalabs device and initialize it according to the
* device's type
*/
-static int hl_pci_probe(struct pci_dev *pdev,
- const struct pci_device_id *id)
+static int hl_pci_probe(struct pci_dev *pdev, const struct pci_device_id *id)
{
struct hl_device *hdev;
int rc;
@@ -467,7 +464,7 @@ static int hl_pci_probe(struct pci_dev *pdev,
" device found [%04x:%04x] (rev %x)\n",
(int)pdev->vendor, (int)pdev->device, (int)pdev->revision);
- rc = create_hdev(&hdev, pdev, ASIC_INVALID, -1);
+ rc = create_hdev(&hdev, pdev);
if (rc)
return rc;
--
2.25.1
^ permalink raw reply related [flat|nested] 10+ messages in thread
end of thread, other threads:[~2021-11-07 20:52 UTC | newest]
Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-11-07 20:51 [PATCH 01/10] habanalabs: don't clear previous f/w indications Oded Gabbay
2021-11-07 20:51 ` [PATCH 02/10] habanalabs: handle abort scenario for user interrupt Oded Gabbay
2021-11-07 20:51 ` [PATCH 03/10] habanalabs: add dedicated message towards f/w to set power Oded Gabbay
2021-11-07 20:51 ` [PATCH 04/10] habanalabs: rename reset flags Oded Gabbay
2021-11-07 20:51 ` [PATCH 05/10] habanalabs: change wait for interrupt timeout to 64 bit Oded Gabbay
2021-11-07 20:51 ` [PATCH 06/10] habanalabs: expand clock throttling information uAPI Oded Gabbay
2021-11-07 20:51 ` [PATCH 07/10] habanalabs/gaudi: Fix collective wait bug Oded Gabbay
2021-11-07 20:51 ` [PATCH 08/10] habanalabs: refactor wait-for-user-interrupt function Oded Gabbay
2021-11-07 20:51 ` [PATCH 09/10] habanalabs: add new opcodes for INFO IOCTL Oded Gabbay
2021-11-07 20:51 ` [PATCH 10/10] habanalabs: make hdev creation code more readable Oded Gabbay
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).