Re: [PATCH V1 3/5] accel/amd_vpci: Add Remote Management(RM) queue infrastructure
From: Christophe JAILLET
Date: Tue Nov 11 2025 - 04:49:11 EST
Le 11/11/2025 à 02:15, David Zhang a écrit :
This patch introduces a Remote Management (RM) queue service, which
provides a way to communicate between the management PCIe function (PF0)
and the embedded firmware running on AMD Versal SoCs.
The RM service implements a hardware-based ring buffer for bidirectional
command and response exchange between the host driver and the firmware.
This patch adds the core infrastructure for:
- Initializing and managing the RM queue
- Submitting commands to the embedded firmware
- Polling for command completion
Subsequent patches will integrate the infrastructure with the firmware
management logic to enable firmware download, status query, and other
control operations.
Co-developed-by: Nishad Saraf <nishads@xxxxxxx>
Signed-off-by: Nishad Saraf <nishads@xxxxxxx>
Signed-off-by: David Zhang <yidong.zhang@xxxxxxx>
...
diff --git a/drivers/accel/amd_vpci/versal-pci-rm-service.h b/drivers/accel/amd_vpci/versal-pci-rm-service.h
new file mode 100644
index 000000000000..d2397a1a672c
--- /dev/null
+++ b/drivers/accel/amd_vpci/versal-pci-rm-service.h
...
+#define RM_CMD_ID_MIN 1
+#define RM_CMD_ID_MAX (BIT(17) - 1)
+#define RM_CMD_SQ_HDR_OPS_MSK GENMASK(15, 0)
+#define RM_CMD_SQ_HDR_SIZE_MSK GENMASK(14, 0)
+#define RM_CMD_SQ_SLOT_SIZE SZ_512
+#define RM_CMD_CQ_SLOT_SIZE SZ_16
+#define RM_CMD_CQ_BUFFER_SIZE (1024 * 1024)
SZ_1M to be consistent with other xx_SIZE macro?
+#define RM_CMD_CQ_BUFFER_OFFSET 0x0
+#define RM_CMD_LOG_PAGE_TYPE_MASK GENMASK(15, 0)
+#define RM_CMD_VMR_CONTROL_MSK GENMASK(10, 8)
+#define RM_CMD_VMR_CONTROL_PS_MASK BIT(9)
+
+#define RM_CMD_WAIT_CONFIG_TIMEOUT msecs_to_jiffies(10 * 1000)
+#define RM_CMD_WAIT_DOWNLOAD_TIMEOUT msecs_to_jiffies(300 * 1000)
secs_to_jiffies() maybe to be less verbose?
+
+#define RM_COMPLETION_TIMER (HZ / 10)
+#define RM_HEALTH_CHECK_TIMER (HZ)
...
CJ