[PATCH 5/7] accel/habanalabs: sync f/w events interrupt in hard reset
Oded Gabbay
ogabbay at kernel.org
Thu Mar 30 07:22:11 UTC 2023
From: Tal Cohen <talcohen at habana.ai>
Receiving events from FW, while the device is in hard reset, causes
a warning message in Driver log. The message may point to a
problem in the Driver or FW. But It also can appear as a result
of events that have been sent from FW just before the hard reset.
In order to avoid receiving events from FW while the device is in reset
and is already in 'disabled' mode, sync the f/w events interrupt right
before setting the device to 'disabled'.
Signed-off-by: Tal Cohen <talcohen at habana.ai>
Reviewed-by: Oded Gabbay <ogabbay at kernel.org>
Signed-off-by: Oded Gabbay <ogabbay at kernel.org>
---
drivers/accel/habanalabs/common/device.c | 55 +++++++++++---------
drivers/accel/habanalabs/common/habanalabs.h | 2 +
drivers/accel/habanalabs/gaudi/gaudi.c | 3 ++
drivers/accel/habanalabs/gaudi2/gaudi2.c | 1 +
drivers/accel/habanalabs/goya/goya.c | 1 +
5 files changed, 37 insertions(+), 25 deletions(-)
diff --git a/drivers/accel/habanalabs/common/device.c b/drivers/accel/habanalabs/common/device.c
index 3c1af9d43b65..fabfc501ef54 100644
--- a/drivers/accel/habanalabs/common/device.c
+++ b/drivers/accel/habanalabs/common/device.c
@@ -1380,13 +1380,41 @@ static void device_disable_open_processes(struct hl_device *hdev, bool control_d
mutex_unlock(fd_lock);
}
+static void send_disable_pci_access(struct hl_device *hdev, u32 flags)
+{
+ /* If reset is due to heartbeat, device CPU is no responsive in
+ * which case no point sending PCI disable message to it.
+ */
+ 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
+ * sending us interrupts after that. We need to disable
+ * the access here because if the device is marked
+ * disable, the message won't be send. Also, in case
+ * of heartbeat, the device CPU is marked as disable
+ * so this message won't be sent
+ */
+ if (hl_fw_send_pci_access_msg(hdev, CPUCP_PACKET_DISABLE_PCI_ACCESS, 0x0)) {
+ dev_warn(hdev->dev, "Failed to disable FW's PCI access\n");
+ return;
+ }
+
+ /* verify that last EQs are handled before disabled is set */
+ if (hdev->cpu_queues_enable)
+ synchronize_irq(pci_irq_vector(hdev->pdev,
+ hdev->asic_prop.eq_interrupt_id));
+ }
+}
+
static void handle_reset_trigger(struct hl_device *hdev, u32 flags)
{
u32 cur_reset_trigger = HL_RESET_TRIGGER_DEFAULT;
/* No consecutive mechanism when user context exists */
if (hdev->is_compute_ctx_active)
- goto disable_pci;
+ return;
/*
* 'reset cause' is being updated here, because getting here
@@ -1418,30 +1446,6 @@ static void handle_reset_trigger(struct hl_device *hdev, u32 flags)
} else {
hdev->reset_info.reset_trigger_repeated = 1;
}
-
- /* If reset is due to heartbeat, device CPU is no responsive in
- * which case no point sending PCI disable message to it.
- *
- * If F/W is performing the reset, no need to send it a message to disable
- * PCI access
- */
-
-disable_pci:
- 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
- * sending us interrupts after that. We need to disable
- * the access here because if the device is marked
- * disable, the message won't be send. Also, in case
- * of heartbeat, the device CPU is marked as disable
- * so this message won't be sent
- */
- if (hl_fw_send_pci_access_msg(hdev, CPUCP_PACKET_DISABLE_PCI_ACCESS, 0x0))
- dev_warn(hdev->dev,
- "Failed to disable FW's PCI access\n");
- }
}
/*
@@ -1562,6 +1566,7 @@ int hl_device_reset(struct hl_device *hdev, u32 flags)
escalate_reset_flow:
handle_reset_trigger(hdev, flags);
+ send_disable_pci_access(hdev, flags);
/* This also blocks future CS/VM/JOB completion operations */
hdev->disabled = true;
diff --git a/drivers/accel/habanalabs/common/habanalabs.h b/drivers/accel/habanalabs/common/habanalabs.h
index 7b6ad3d7dbaa..8c3bcc50e560 100644
--- a/drivers/accel/habanalabs/common/habanalabs.h
+++ b/drivers/accel/habanalabs/common/habanalabs.h
@@ -662,6 +662,7 @@ struct hl_hints_range {
* @user_interrupt_count: number of user interrupts.
* @user_dec_intr_count: number of decoder interrupts exposed to user.
* @tpc_interrupt_id: interrupt id for TPC to use in order to raise events towards the host.
+ * @eq_interrupt_id: interrupt id for EQ, uses to synchronize EQ interrupts in hard-reset.
* @unexpected_user_error_interrupt_id: interrupt id used to indicate an unexpected user error.
* @cache_line_size: device cache line size.
* @server_type: Server type that the ASIC is currently installed in.
@@ -793,6 +794,7 @@ struct asic_fixed_properties {
u16 user_interrupt_count;
u16 user_dec_intr_count;
u16 tpc_interrupt_id;
+ u16 eq_interrupt_id;
u16 unexpected_user_error_interrupt_id;
u16 cache_line_size;
u16 server_type;
diff --git a/drivers/accel/habanalabs/gaudi/gaudi.c b/drivers/accel/habanalabs/gaudi/gaudi.c
index 08a4b1cf2b42..2ad8e4efce7f 100644
--- a/drivers/accel/habanalabs/gaudi/gaudi.c
+++ b/drivers/accel/habanalabs/gaudi/gaudi.c
@@ -682,6 +682,9 @@ static int gaudi_set_fixed_properties(struct hl_device *hdev)
prop->first_available_user_interrupt = USHRT_MAX;
prop->tpc_interrupt_id = USHRT_MAX;
+ /* single msi */
+ prop->eq_interrupt_id = 0;
+
for (i = 0 ; i < HL_MAX_DCORES ; i++)
prop->first_available_cq[i] = USHRT_MAX;
diff --git a/drivers/accel/habanalabs/gaudi2/gaudi2.c b/drivers/accel/habanalabs/gaudi2/gaudi2.c
index ce85308d03e9..554020026da8 100644
--- a/drivers/accel/habanalabs/gaudi2/gaudi2.c
+++ b/drivers/accel/habanalabs/gaudi2/gaudi2.c
@@ -2439,6 +2439,7 @@ static int gaudi2_set_fixed_properties(struct hl_device *hdev)
prop->first_available_user_interrupt = GAUDI2_IRQ_NUM_USER_FIRST;
prop->tpc_interrupt_id = GAUDI2_IRQ_NUM_TPC_ASSERT;
+ prop->eq_interrupt_id = GAUDI2_IRQ_NUM_EVENT_QUEUE;
prop->unexpected_user_error_interrupt_id = GAUDI2_IRQ_NUM_UNEXPECTED_ERROR;
prop->first_available_cq[0] = GAUDI2_RESERVED_CQ_NUMBER;
diff --git a/drivers/accel/habanalabs/goya/goya.c b/drivers/accel/habanalabs/goya/goya.c
index 07d67878eac5..fb0ac9df841a 100644
--- a/drivers/accel/habanalabs/goya/goya.c
+++ b/drivers/accel/habanalabs/goya/goya.c
@@ -473,6 +473,7 @@ int goya_set_fixed_properties(struct hl_device *hdev)
prop->first_available_user_interrupt = USHRT_MAX;
prop->tpc_interrupt_id = USHRT_MAX;
+ prop->eq_interrupt_id = GOYA_EVENT_QUEUE_MSIX_IDX;
for (i = 0 ; i < HL_MAX_DCORES ; i++)
prop->first_available_cq[i] = USHRT_MAX;
--
2.40.0
More information about the dri-devel
mailing list