Commit 6c161732 authored by Lizhi Hou's avatar Lizhi Hou
Browse files

accel/amdxdna: Fix incorrect size of ERT_START_NPU commands 



When multiple ERT_START_NPU commands are combined in one buffer, the
buffer size calculation is incorrect. Also, the condition to make sure
the buffer size is not beyond 4K is also fixed.

Fixes: aac24309 ("accel/amdxdna: Add command execution")
Reviewed-by: default avatarJacek Lawrynowicz <jacek.lawrynowicz@linux.intel.com>
Reviewed-by: default avatarMaciej Falkowski <maciej.falkowski@linux.intel.com>
Signed-off-by: default avatarLizhi Hou <lizhi.hou@amd.com>
Link: https://lore.kernel.org/r/20250409210013.10854-1-lizhi.hou@amd.com
parent c180b003

drivers/accel/amdxdna/aie2_message.c

+3 −3
Original line number Diff line number Diff line
@@ -525,7 +525,7 @@ aie2_cmdlist_fill_one_slot_cf(void *cmd_buf, u32 offset, 
	if (!payload)

		return -EINVAL;



	if (!slot_cf_has_space(offset, payload_len))

	if (!slot_has_space(*buf, offset, payload_len))

		return -ENOSPC;



	buf->cu_idx = cu_idx;
@@ -558,7 +558,7 @@ aie2_cmdlist_fill_one_slot_dpu(void *cmd_buf, u32 offset, 
	if (payload_len < sizeof(*sn) || arg_sz > MAX_DPU_ARGS_SIZE)

		return -EINVAL;



	if (!slot_dpu_has_space(offset, arg_sz))

	if (!slot_has_space(*buf, offset, arg_sz))

		return -ENOSPC;



	buf->inst_buf_addr = sn->buffer;
@@ -569,7 +569,7 @@ aie2_cmdlist_fill_one_slot_dpu(void *cmd_buf, u32 offset, 
	memcpy(buf->args, sn->prop_args, arg_sz);



	/* Accurate buf size to hint firmware to do necessary copy */

	*size += sizeof(*buf) + arg_sz;

	*size = sizeof(*buf) + arg_sz;

	return 0;

}

drivers/accel/amdxdna/aie2_msg_priv.h

+4 −6
Original line number Diff line number Diff line
@@ -319,18 +319,16 @@ struct async_event_msg_resp { 
} __packed;



#define MAX_CHAIN_CMDBUF_SIZE SZ_4K

#define slot_cf_has_space(offset, payload_size) \

	(MAX_CHAIN_CMDBUF_SIZE - ((offset) + (payload_size)) > \

	 offsetof(struct cmd_chain_slot_execbuf_cf, args[0]))

#define slot_has_space(slot, offset, payload_size)		\

	(MAX_CHAIN_CMDBUF_SIZE >= (offset) + (payload_size) +	\

	 sizeof(typeof(slot)))



struct cmd_chain_slot_execbuf_cf {

	__u32 cu_idx;

	__u32 arg_cnt;

	__u32 args[] __counted_by(arg_cnt);

};



#define slot_dpu_has_space(offset, payload_size) \

	(MAX_CHAIN_CMDBUF_SIZE - ((offset) + (payload_size)) > \

	 offsetof(struct cmd_chain_slot_dpu, args[0]))

struct cmd_chain_slot_dpu {

	__u64 inst_buf_addr;

	__u32 inst_size;