Merge tag 'misc-habanalabs-fixes-2021-01-21' of https://git.kernel.org/pub/scm/linux/kernel/git/ogabbay/linux into char-misc-linus

Oded writes:

This tag contains the following bug fixes for 5.11-rc5/6:

- Clear the fence field in the PCI counters packet before sending
  the packet to the F/W. Not clearing it might cause the driver
  and F/W to get out-of-sync

- Fix backward compatibility in the uapi of IDLE check that is
  part of the INFO IOCTL.

- Tell the F/W to not access the Host (device outbound) while
  the driver removes the device. If that happens, the server
  might crash.

* tag 'misc-habanalabs-fixes-2021-01-21' of https://git.kernel.org/pub/scm/linux/kernel/git/ogabbay/linux:
  habanalabs: disable FW events on device removal
  habanalabs: fix backward compatibility of idle check
  habanalabs: zero pci counters packet before submit to FW

3 files changed, 16 insertions, 0 deletions

diff --git a/drivers/misc/habanalabs/common/device.c b/drivers/misc/habanalabs/common/device.c
index 1ea57d86caa3..69d04eca767f 100644
--- a/drivers/misc/habanalabs/common/device.c
+++ b/drivers/misc/habanalabs/common/device.c
@@ -1487,6 +1487,15 @@ void hl_device_fini(struct hl_device *hdev)
 		}
 	}
 
+	/* Disable PCI access from device F/W so it 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
+	 */
+	hl_fw_send_pci_access_msg(hdev,	CPUCP_PACKET_DISABLE_PCI_ACCESS);
+
 	/* Mark device as disabled */
 	hdev->disabled = true;
 
diff --git a/drivers/misc/habanalabs/common/firmware_if.c b/drivers/misc/habanalabs/common/firmware_if.c
index 20f77f58edef..c9a12980218a 100644
--- a/drivers/misc/habanalabs/common/firmware_if.c
+++ b/drivers/misc/habanalabs/common/firmware_if.c
@@ -402,6 +402,10 @@ int hl_fw_cpucp_pci_counters_get(struct hl_device *hdev,
 	}
 	counters->rx_throughput = result;
 
+	memset(&pkt, 0, sizeof(pkt));
+	pkt.ctl = cpu_to_le32(CPUCP_PACKET_PCIE_THROUGHPUT_GET <<
+			CPUCP_PKT_CTL_OPCODE_SHIFT);
+
 	/* Fetch PCI tx counter */
 	pkt.index = cpu_to_le32(cpucp_pcie_throughput_tx);
 	rc = hdev->asic_funcs->send_cpu_message(hdev, (u32 *) &pkt, sizeof(pkt),
@@ -414,6 +418,7 @@ int hl_fw_cpucp_pci_counters_get(struct hl_device *hdev,
 	counters->tx_throughput = result;
 
 	/* Fetch PCI replay counter */
+	memset(&pkt, 0, sizeof(pkt));
 	pkt.ctl = cpu_to_le32(CPUCP_PACKET_PCIE_REPLAY_CNT_GET <<
 			CPUCP_PKT_CTL_OPCODE_SHIFT);
 
diff --git a/drivers/misc/habanalabs/common/habanalabs_ioctl.c b/drivers/misc/habanalabs/common/habanalabs_ioctl.c
index 12efbd9d2e3a..d25892d61ec9 100644
--- a/drivers/misc/habanalabs/common/habanalabs_ioctl.c
+++ b/drivers/misc/habanalabs/common/habanalabs_ioctl.c
@@ -133,6 +133,8 @@ static int hw_idle(struct hl_device *hdev, struct hl_info_args *args)
 
 	hw_idle.is_idle = hdev->asic_funcs->is_device_idle(hdev,
 					&hw_idle.busy_engines_mask_ext, NULL);
+	hw_idle.busy_engines_mask =
+			lower_32_bits(hw_idle.busy_engines_mask_ext);
 
 	return copy_to_user(out, &hw_idle,
 		min((size_t) max_size, sizeof(hw_idle))) ? -EFAULT : 0;