accel/amdxdna: Support firmware debug buffer

To collect firmware debug information, the userspace application allocates
a AMDXDNA_BO_DEV buffer object through DRM_IOCTL_AMDXDNA_CREATE_BO.
Then it associates the buffer with the hardware context through
DRM_IOCTL_AMDXDNA_CONFIG_HWCTX which requests firmware to bind the buffer
through a mailbox command. The firmware then writes the debug data into
this buffer. The buffer can be mapped into userspace so that
applications can retrieve and analyze the firmware debug information.

Reviewed-by: Mario Limonciello (AMD) <[e-mail]>
Signed-off-by: Lizhi Hou <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]

drivers/accel/amdxdna/TODO

···
 - Add debugfs support
-- Add debug BO support

drivers/accel/amdxdna/aie2_ctx.c

···
 }
 
 static int
-aie2_sched_nocmd_resp_handler(void *handle, void __iomem *data, size_t size)
+aie2_sched_drvcmd_resp_handler(void *handle, void __iomem *data, size_t size)
 {
 	struct amdxdna_sched_job *job = handle;
 	int ret = 0;
-	u32 status;
 
 	if (unlikely(!data))
 		goto out;
···
 		goto out;
 	}
 
-	status = readl(data);
-	XDNA_DBG(job->hwctx->client->xdna, "Resp status 0x%x", status);
+	job->drv_cmd->result = readl(data);
 
 out:
 	aie2_sched_notify(job);
···
 	kref_get(&job->refcnt);
 	fence = dma_fence_get(job->fence);
 
-	if (unlikely(!cmd_abo)) {
-		ret = aie2_sync_bo(hwctx, job, aie2_sched_nocmd_resp_handler);
+	if (job->drv_cmd) {
+		switch (job->drv_cmd->opcode) {
+		case SYNC_DEBUG_BO:
+			ret = aie2_sync_bo(hwctx, job, aie2_sched_drvcmd_resp_handler);
+			break;
+		case ATTACH_DEBUG_BO:
+			ret = aie2_config_debug_bo(hwctx, job, aie2_sched_drvcmd_resp_handler);
+			break;
+		default:
+			ret = -EINVAL;
+			break;
+		}
 		goto out;
 	}
 
···
 	return ret;
 }
 
+static void aie2_cmd_wait(struct amdxdna_hwctx *hwctx, u64 seq)
+{
+	struct dma_fence *out_fence = aie2_cmd_get_out_fence(hwctx, seq);
+
+	if (!out_fence) {
+		XDNA_ERR(hwctx->client->xdna, "Failed to get fence");
+		return;
+	}
+
+	dma_fence_wait_timeout(out_fence, false, MAX_SCHEDULE_TIMEOUT);
+	dma_fence_put(out_fence);
+}
+
+static int aie2_hwctx_cfg_debug_bo(struct amdxdna_hwctx *hwctx, u32 bo_hdl,
+				   bool attach)
+{
+	struct amdxdna_client *client = hwctx->client;
+	struct amdxdna_dev *xdna = client->xdna;
+	struct amdxdna_drv_cmd cmd = { 0 };
+	struct amdxdna_gem_obj *abo;
+	u64 seq;
+	int ret;
+
+	abo = amdxdna_gem_get_obj(client, bo_hdl, AMDXDNA_BO_DEV);
+	if (!abo) {
+		XDNA_ERR(xdna, "Get bo %d failed", bo_hdl);
+		return -EINVAL;
+	}
+
+	if (attach) {
+		if (abo->assigned_hwctx != AMDXDNA_INVALID_CTX_HANDLE) {
+			ret = -EBUSY;
+			goto put_obj;
+		}
+		cmd.opcode = ATTACH_DEBUG_BO;
+	} else {
+		if (abo->assigned_hwctx != hwctx->id) {
+			ret = -EINVAL;
+			goto put_obj;
+		}
+		cmd.opcode = DETACH_DEBUG_BO;
+	}
+
+	ret = amdxdna_cmd_submit(client, &cmd, AMDXDNA_INVALID_BO_HANDLE,
+				 &bo_hdl, 1, hwctx->id, &seq);
+	if (ret) {
+		XDNA_ERR(xdna, "Submit command failed");
+		goto put_obj;
+	}
+
+	aie2_cmd_wait(hwctx, seq);
+	if (cmd.result) {
+		XDNA_ERR(xdna, "Response failure 0x%x", cmd.result);
+		goto put_obj;
+	}
+
+	if (attach)
+		abo->assigned_hwctx = hwctx->id;
+	else
+		abo->assigned_hwctx = AMDXDNA_INVALID_CTX_HANDLE;
+
+	XDNA_DBG(xdna, "Config debug BO %d to %s", bo_hdl, hwctx->name);
+
+put_obj:
+	amdxdna_gem_put_obj(abo);
+	return ret;
+}
+
 int aie2_hwctx_config(struct amdxdna_hwctx *hwctx, u32 type, u64 value, void *buf, u32 size)
 {
 	struct amdxdna_dev *xdna = hwctx->client->xdna;
···
 	case DRM_AMDXDNA_HWCTX_CONFIG_CU:
 		return aie2_hwctx_cu_config(hwctx, buf, size);
 	case DRM_AMDXDNA_HWCTX_ASSIGN_DBG_BUF:
+		return aie2_hwctx_cfg_debug_bo(hwctx, (u32)value, true);
 	case DRM_AMDXDNA_HWCTX_REMOVE_DBG_BUF:
-		return -EOPNOTSUPP;
+		return aie2_hwctx_cfg_debug_bo(hwctx, (u32)value, false);
 	default:
 		XDNA_DBG(xdna, "Not supported type %d", type);
 		return -EOPNOTSUPP;
 	}
+}
+
+int aie2_hwctx_sync_debug_bo(struct amdxdna_hwctx *hwctx, u32 debug_bo_hdl)
+{
+	struct amdxdna_client *client = hwctx->client;
+	struct amdxdna_dev *xdna = client->xdna;
+	struct amdxdna_drv_cmd cmd = { 0 };
+	u64 seq;
+	int ret;
+
+	cmd.opcode = SYNC_DEBUG_BO;
+	ret = amdxdna_cmd_submit(client, &cmd, AMDXDNA_INVALID_BO_HANDLE,
+				 &debug_bo_hdl, 1, hwctx->id, &seq);
+	if (ret) {
+		XDNA_ERR(xdna, "Submit command failed");
+		return ret;
+	}
+
+	aie2_cmd_wait(hwctx, seq);
+	if (cmd.result) {
+		XDNA_ERR(xdna, "Response failure 0x%x", cmd.result);
+		return ret;
+	}
+
+	return 0;
 }
 
 static int aie2_populate_range(struct amdxdna_gem_obj *abo)

drivers/accel/amdxdna/aie2_message.c

···
 	int ret = 0;
 
 	req.src_addr = 0;
-	req.dst_addr = abo->mem.dev_addr - hwctx->client->dev_heap->mem.dev_addr;
+	req.dst_addr = amdxdna_dev_bo_offset(abo);
 	req.size = abo->mem.size;
 
 	/* Device to Host */
···
 	}
 
 	return 0;
+}
+
+int aie2_config_debug_bo(struct amdxdna_hwctx *hwctx, struct amdxdna_sched_job *job,
+			 int (*notify_cb)(void *, void __iomem *, size_t))
+{
+	struct mailbox_channel *chann = hwctx->priv->mbox_chann;
+	struct amdxdna_gem_obj *abo = to_xdna_obj(job->bos[0]);
+	struct amdxdna_dev *xdna = hwctx->client->xdna;
+	struct config_debug_bo_req req;
+	struct xdna_mailbox_msg msg;
+
+	if (job->drv_cmd->opcode == ATTACH_DEBUG_BO)
+		req.config = DEBUG_BO_REGISTER;
+	else
+		req.config = DEBUG_BO_UNREGISTER;
+
+	req.offset = amdxdna_dev_bo_offset(abo);
+	req.size = abo->mem.size;
+
+	XDNA_DBG(xdna, "offset 0x%llx size 0x%llx config %d",
+		 req.offset, req.size, req.config);
+
+	msg.handle = job;
+	msg.notify_cb = notify_cb;
+	msg.send_data = (u8 *)&req;
+	msg.send_size = sizeof(req);
+	msg.opcode = MSG_OP_CONFIG_DEBUG_BO;
+
+	return xdna_mailbox_send_msg(chann, &msg, TX_TIMEOUT);
 }

drivers/accel/amdxdna/aie2_msg_priv.h

···
 	MSG_OP_CONFIG_CU                   = 0x11,
 	MSG_OP_CHAIN_EXEC_BUFFER_CF        = 0x12,
 	MSG_OP_CHAIN_EXEC_DPU              = 0x13,
+	MSG_OP_CONFIG_DEBUG_BO             = 0x14,
 	MSG_OP_MAX_XRT_OPCODE,
 	MSG_OP_SUSPEND                     = 0x101,
 	MSG_OP_RESUME                      = 0x102,
···
 } __packed;
 
 struct sync_bo_resp {
+	enum aie2_msg_status	status;
+} __packed;
+
+#define DEBUG_BO_UNREGISTER 0
+#define DEBUG_BO_REGISTER   1
+struct config_debug_bo_req {
+	__u64	offset;
+	__u64	size;
+	/*
+	 * config operations.
+	 *   DEBUG_BO_REGISTER: Register debug buffer
+	 *   DEBUG_BO_UNREGISTER: Unregister debug buffer
+	 */
+	__u32	config;
+} __packed;
+
+struct config_debug_bo_resp {
 	enum aie2_msg_status	status;
 } __packed;
 #endif /* _AIE2_MSG_PRIV_H_ */

drivers/accel/amdxdna/aie2_pci.c

···
 	.hwctx_init = aie2_hwctx_init,
 	.hwctx_fini = aie2_hwctx_fini,
 	.hwctx_config = aie2_hwctx_config,
+	.hwctx_sync_debug_bo = aie2_hwctx_sync_debug_bo,
 	.cmd_submit = aie2_cmd_submit,
 	.hmm_invalidate = aie2_hmm_invalidate,
 	.get_array = aie2_get_array,

drivers/accel/amdxdna/aie2_pci.h

···
 			       int (*notify_cb)(void *, void __iomem *, size_t));
 int aie2_sync_bo(struct amdxdna_hwctx *hwctx, struct amdxdna_sched_job *job,
 		 int (*notify_cb)(void *, void __iomem *, size_t));
+int aie2_config_debug_bo(struct amdxdna_hwctx *hwctx, struct amdxdna_sched_job *job,
+			 int (*notify_cb)(void *, void __iomem *, size_t));
 
 /* aie2_hwctx.c */
 int aie2_hwctx_init(struct amdxdna_hwctx *hwctx);
 void aie2_hwctx_fini(struct amdxdna_hwctx *hwctx);
 int aie2_hwctx_config(struct amdxdna_hwctx *hwctx, u32 type, u64 value, void *buf, u32 size);
+int aie2_hwctx_sync_debug_bo(struct amdxdna_hwctx *hwctx, u32 debug_bo_hdl);
 void aie2_hwctx_suspend(struct amdxdna_client *client);
 int aie2_hwctx_resume(struct amdxdna_client *client);
 int aie2_cmd_submit(struct amdxdna_hwctx *hwctx, struct amdxdna_sched_job *job, u64 *seq);

drivers/accel/amdxdna/amdxdna_ctx.c

···
 	return ret;
 }
 
+int amdxdna_hwctx_sync_debug_bo(struct amdxdna_client *client, u32 debug_bo_hdl)
+{
+	struct amdxdna_dev *xdna = client->xdna;
+	struct amdxdna_hwctx *hwctx;
+	struct amdxdna_gem_obj *abo;
+	struct drm_gem_object *gobj;
+	int ret, idx;
+
+	if (!xdna->dev_info->ops->hwctx_sync_debug_bo)
+		return -EOPNOTSUPP;
+
+	gobj = drm_gem_object_lookup(client->filp, debug_bo_hdl);
+	if (!gobj)
+		return -EINVAL;
+
+	abo = to_xdna_obj(gobj);
+	guard(mutex)(&xdna->dev_lock);
+	idx = srcu_read_lock(&client->hwctx_srcu);
+	hwctx = xa_load(&client->hwctx_xa, abo->assigned_hwctx);
+	if (!hwctx) {
+		ret = -EINVAL;
+		goto unlock_srcu;
+	}
+
+	ret = xdna->dev_info->ops->hwctx_sync_debug_bo(hwctx, debug_bo_hdl);
+
+unlock_srcu:
+	srcu_read_unlock(&client->hwctx_srcu, idx);
+	drm_gem_object_put(gobj);
+	return ret;
+}
+
 static void
 amdxdna_arg_bos_put(struct amdxdna_sched_job *job)
 {
···
 }
 
 int amdxdna_cmd_submit(struct amdxdna_client *client,
+		       struct amdxdna_drv_cmd *drv_cmd,
 		       u32 cmd_bo_hdl, u32 *arg_bo_hdls, u32 arg_bo_cnt,
 		       u32 hwctx_hdl, u64 *seq)
 {
···
 	if (!job)
 		return -ENOMEM;
 
+	job->drv_cmd = drv_cmd;
+
 	if (cmd_bo_hdl != AMDXDNA_INVALID_BO_HANDLE) {
 		job->cmd_bo = amdxdna_gem_get_obj(client, cmd_bo_hdl, AMDXDNA_BO_CMD);
 		if (!job->cmd_bo) {
···
 			ret = -EINVAL;
 			goto free_job;
 		}
-	} else {
-		job->cmd_bo = NULL;
 	}
 
 	ret = amdxdna_arg_bos_lookup(client, job, arg_bo_hdls, arg_bo_cnt);
···
 		}
 	}
 
-	ret = amdxdna_cmd_submit(client, cmd_bo_hdl, arg_bo_hdls,
+	ret = amdxdna_cmd_submit(client, NULL, cmd_bo_hdl, arg_bo_hdls,
 				 args->arg_count, args->hwctx, &args->seq);
 	if (ret)
 		XDNA_DBG(xdna, "Submit cmds failed, ret %d", ret);

drivers/accel/amdxdna/amdxdna_ctx.h

···
 #define drm_job_to_xdna_job(j) \
 	container_of(j, struct amdxdna_sched_job, base)
 
+enum amdxdna_job_opcode {
+	SYNC_DEBUG_BO,
+	ATTACH_DEBUG_BO,
+	DETACH_DEBUG_BO,
+};
+
+struct amdxdna_drv_cmd {
+	enum amdxdna_job_opcode	opcode;
+	u32			result;
+};
+
 struct amdxdna_sched_job {
 	struct drm_sched_job	base;
 	struct kref		refcnt;
···
 	struct dma_fence	*out_fence;
 	bool			job_done;
 	u64			seq;
+	struct amdxdna_drv_cmd	*drv_cmd;
 	struct amdxdna_gem_obj	*cmd_bo;
 	size_t			bo_cnt;
 	struct drm_gem_object	*bos[] __counted_by(bo_cnt);
···
 void amdxdna_hwctx_remove_all(struct amdxdna_client *client);
 int amdxdna_hwctx_walk(struct amdxdna_client *client, void *arg,
 		       int (*walk)(struct amdxdna_hwctx *hwctx, void *arg));
+int amdxdna_hwctx_sync_debug_bo(struct amdxdna_client *client, u32 debug_bo_hdl);
 
 int amdxdna_cmd_submit(struct amdxdna_client *client,
-		       u32 cmd_bo_hdls, u32 *arg_bo_hdls, u32 arg_bo_cnt,
+		       struct amdxdna_drv_cmd *drv_cmd, u32 cmd_bo_hdls,
+		       u32 *arg_bo_hdls, u32 arg_bo_cnt,
 		       u32 hwctx_hdl, u64 *seq);
 
 int amdxdna_cmd_wait(struct amdxdna_client *client, u32 hwctx_hdl,

drivers/accel/amdxdna/amdxdna_gem.c

···
 	XDNA_DBG(xdna, "Sync bo %d offset 0x%llx, size 0x%llx\n",
 		 args->handle, args->offset, args->size);
 
+	if (args->direction == SYNC_DIRECT_FROM_DEVICE)
+		ret = amdxdna_hwctx_sync_debug_bo(abo->client, args->handle);
+
 put_obj:
 	drm_gem_object_put(gobj);
 	return ret;

drivers/accel/amdxdna/amdxdna_gem.h

···
 #define _AMDXDNA_GEM_H_
 
 #include <linux/hmm.h>
+#include "amdxdna_pci_drv.h"
 
 struct amdxdna_umap {
 	struct vm_area_struct		*vma;
···
 static inline void amdxdna_gem_put_obj(struct amdxdna_gem_obj *abo)
 {
 	drm_gem_object_put(to_gobj(abo));
+}
+
+static inline u64 amdxdna_dev_bo_offset(struct amdxdna_gem_obj *abo)
+{
+	return abo->mem.dev_addr - abo->client->dev_heap->mem.dev_addr;
 }
 
 void amdxdna_umap_put(struct amdxdna_umap *mapp);

drivers/accel/amdxdna/amdxdna_pci_drv.c

···
  * 0.0: Initial version
  * 0.1: Support getting all hardware contexts by DRM_IOCTL_AMDXDNA_GET_ARRAY
  * 0.2: Support getting last error hardware error
+ * 0.3: Support firmware debug buffer
  */
 #define AMDXDNA_DRIVER_MAJOR		0
-#define AMDXDNA_DRIVER_MINOR		2
+#define AMDXDNA_DRIVER_MINOR		3
 
 /*
  * Bind the driver base on (vendor_id, device_id) pair and later use the

drivers/accel/amdxdna/amdxdna_pci_drv.h

···
 	int (*hwctx_init)(struct amdxdna_hwctx *hwctx);
 	void (*hwctx_fini)(struct amdxdna_hwctx *hwctx);
 	int (*hwctx_config)(struct amdxdna_hwctx *hwctx, u32 type, u64 value, void *buf, u32 size);
+	int (*hwctx_sync_debug_bo)(struct amdxdna_hwctx *hwctx, u32 debug_bo_hdl);
 	void (*hmm_invalidate)(struct amdxdna_gem_obj *abo, unsigned long cur_seq);
 	int (*cmd_submit)(struct amdxdna_hwctx *hwctx, struct amdxdna_sched_job *job, u64 *seq);
 	int (*get_aie_info)(struct amdxdna_client *client, struct amdxdna_drm_get_info *args);

drivers/accel/amdxdna/npu1_regs.c

···
 
 const struct rt_config npu1_default_rt_cfg[] = {
 	{ 2, 1, AIE2_RT_CFG_INIT }, /* PDI APP LOAD MODE */
+	{ 4, 1, AIE2_RT_CFG_INIT }, /* Debug BO */
 	{ 1, 1, AIE2_RT_CFG_CLK_GATING }, /* Clock gating on */
 	{ 0 },
 };

drivers/accel/amdxdna/npu4_regs.c

···
 
 const struct rt_config npu4_default_rt_cfg[] = {
 	{ 5, 1, AIE2_RT_CFG_INIT }, /* PDI APP LOAD MODE */
+	{ 10, 1, AIE2_RT_CFG_INIT }, /* DEBUG BUF */
 	{ 1, 1, AIE2_RT_CFG_CLK_GATING }, /* Clock gating on */
 	{ 2, 1, AIE2_RT_CFG_CLK_GATING }, /* Clock gating on */
 	{ 3, 1, AIE2_RT_CFG_CLK_GATING }, /* Clock gating on */