[PATCH V1] accel/amdxdna: Fix out-of-bounds memset in command slot handling

[PATCH V1] accel/amdxdna: Fix out-of-bounds memset in command slot handling

[Lizhi Hou]

The remaining space in a command slot may be smaller than the size of
the command header. Clearing the command header with memset() before
verifying the available slot space can result in an out-of-bounds write
and memory corruption.

Fix this by moving the memset() call after the size validation.

Fixes: 3d32eb7a5ecf ("accel/amdxdna: Fix cu_idx being cleared by memset() during command setup")
Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
---
 drivers/accel/amdxdna/aie2_message.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/accel/amdxdna/aie2_message.c b/drivers/accel/amdxdna/aie2_message.c
index 7d7dcfeaf794..8fbbc3280468 100644
--- a/drivers/accel/amdxdna/aie2_message.c
+++ b/drivers/accel/amdxdna/aie2_message.c
@@ -694,11 +694,11 @@ aie2_cmdlist_fill_npu_cf(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *siz
 	u32 cmd_len;
 	void *cmd;
 
-	memset(npu_slot, 0, sizeof(*npu_slot));
 	cmd = amdxdna_cmd_get_payload(cmd_bo, &cmd_len);
 	if (*size < sizeof(*npu_slot) + cmd_len)
 		return -EINVAL;
 
+	memset(npu_slot, 0, sizeof(*npu_slot));
 	npu_slot->cu_idx = amdxdna_cmd_get_cu_idx(cmd_bo);
 	if (npu_slot->cu_idx == INVALID_CU_IDX)
 		return -EINVAL;
@@ -719,7 +719,6 @@ aie2_cmdlist_fill_npu_dpu(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *si
 	u32 cmd_len;
 	u32 arg_sz;
 
-	memset(npu_slot, 0, sizeof(*npu_slot));
 	sn = amdxdna_cmd_get_payload(cmd_bo, &cmd_len);
 	arg_sz = cmd_len - sizeof(*sn);
 	if (cmd_len < sizeof(*sn) || arg_sz > MAX_NPU_ARGS_SIZE)
@@ -728,6 +727,7 @@ aie2_cmdlist_fill_npu_dpu(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *si
 	if (*size < sizeof(*npu_slot) + arg_sz)
 		return -EINVAL;
 
+	memset(npu_slot, 0, sizeof(*npu_slot));
 	npu_slot->cu_idx = amdxdna_cmd_get_cu_idx(cmd_bo);
 	if (npu_slot->cu_idx == INVALID_CU_IDX)
 		return -EINVAL;
@@ -751,7 +751,6 @@ aie2_cmdlist_fill_npu_preempt(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t
 	u32 cmd_len;
 	u32 arg_sz;
 
-	memset(npu_slot, 0, sizeof(*npu_slot));
 	pd = amdxdna_cmd_get_payload(cmd_bo, &cmd_len);
 	arg_sz = cmd_len - sizeof(*pd);
 	if (cmd_len < sizeof(*pd) || arg_sz > MAX_NPU_ARGS_SIZE)
@@ -760,6 +759,7 @@ aie2_cmdlist_fill_npu_preempt(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t
 	if (*size < sizeof(*npu_slot) + arg_sz)
 		return -EINVAL;
 
+	memset(npu_slot, 0, sizeof(*npu_slot));
 	npu_slot->cu_idx = amdxdna_cmd_get_cu_idx(cmd_bo);
 	if (npu_slot->cu_idx == INVALID_CU_IDX)
 		return -EINVAL;
@@ -787,7 +787,6 @@ aie2_cmdlist_fill_npu_elf(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *si
 	u32 cmd_len;
 	u32 arg_sz;
 
-	memset(npu_slot, 0, sizeof(*npu_slot));
 	pd = amdxdna_cmd_get_payload(cmd_bo, &cmd_len);
 	arg_sz = cmd_len - sizeof(*pd);
 	if (cmd_len < sizeof(*pd) || arg_sz > MAX_NPU_ARGS_SIZE)
@@ -796,6 +795,7 @@ aie2_cmdlist_fill_npu_elf(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *si
 	if (*size < sizeof(*npu_slot) + arg_sz)
 		return -EINVAL;
 
+	memset(npu_slot, 0, sizeof(*npu_slot));
 	npu_slot->type = EXEC_NPU_TYPE_ELF;
 	npu_slot->inst_buf_addr = pd->inst_buf;
 	npu_slot->save_buf_addr = pd->save_buf;
-- 
2.34.1

Re: [PATCH V1] accel/amdxdna: Fix out-of-bounds memset in command slot handling

[Mario Limonciello]

On 2/17/26 12:54 PM, Lizhi Hou wrote:
> The remaining space in a command slot may be smaller than the size of
> the command header. Clearing the command header with memset() before
> verifying the available slot space can result in an out-of-bounds write
> and memory corruption.
> 
> Fix this by moving the memset() call after the size validation.
> 
> Fixes: 3d32eb7a5ecf ("accel/amdxdna: Fix cu_idx being cleared by memset() during command setup")
> Signed-off-by: Lizhi Hou <lizhi.hou@amd.com>
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
> ---
>   drivers/accel/amdxdna/aie2_message.c | 8 ++++----
>   1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/accel/amdxdna/aie2_message.c b/drivers/accel/amdxdna/aie2_message.c
> index 7d7dcfeaf794..8fbbc3280468 100644
> --- a/drivers/accel/amdxdna/aie2_message.c
> +++ b/drivers/accel/amdxdna/aie2_message.c
> @@ -694,11 +694,11 @@ aie2_cmdlist_fill_npu_cf(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *siz
>   	u32 cmd_len;
>   	void *cmd;
>   
> -	memset(npu_slot, 0, sizeof(*npu_slot));
>   	cmd = amdxdna_cmd_get_payload(cmd_bo, &cmd_len);
>   	if (*size < sizeof(*npu_slot) + cmd_len)
>   		return -EINVAL;
>   
> +	memset(npu_slot, 0, sizeof(*npu_slot));
>   	npu_slot->cu_idx = amdxdna_cmd_get_cu_idx(cmd_bo);
>   	if (npu_slot->cu_idx == INVALID_CU_IDX)
>   		return -EINVAL;
> @@ -719,7 +719,6 @@ aie2_cmdlist_fill_npu_dpu(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *si
>   	u32 cmd_len;
>   	u32 arg_sz;
>   
> -	memset(npu_slot, 0, sizeof(*npu_slot));
>   	sn = amdxdna_cmd_get_payload(cmd_bo, &cmd_len);
>   	arg_sz = cmd_len - sizeof(*sn);
>   	if (cmd_len < sizeof(*sn) || arg_sz > MAX_NPU_ARGS_SIZE)
> @@ -728,6 +727,7 @@ aie2_cmdlist_fill_npu_dpu(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *si
>   	if (*size < sizeof(*npu_slot) + arg_sz)
>   		return -EINVAL;
>   
> +	memset(npu_slot, 0, sizeof(*npu_slot));
>   	npu_slot->cu_idx = amdxdna_cmd_get_cu_idx(cmd_bo);
>   	if (npu_slot->cu_idx == INVALID_CU_IDX)
>   		return -EINVAL;
> @@ -751,7 +751,6 @@ aie2_cmdlist_fill_npu_preempt(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t
>   	u32 cmd_len;
>   	u32 arg_sz;
>   
> -	memset(npu_slot, 0, sizeof(*npu_slot));
>   	pd = amdxdna_cmd_get_payload(cmd_bo, &cmd_len);
>   	arg_sz = cmd_len - sizeof(*pd);
>   	if (cmd_len < sizeof(*pd) || arg_sz > MAX_NPU_ARGS_SIZE)
> @@ -760,6 +759,7 @@ aie2_cmdlist_fill_npu_preempt(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t
>   	if (*size < sizeof(*npu_slot) + arg_sz)
>   		return -EINVAL;
>   
> +	memset(npu_slot, 0, sizeof(*npu_slot));
>   	npu_slot->cu_idx = amdxdna_cmd_get_cu_idx(cmd_bo);
>   	if (npu_slot->cu_idx == INVALID_CU_IDX)
>   		return -EINVAL;
> @@ -787,7 +787,6 @@ aie2_cmdlist_fill_npu_elf(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *si
>   	u32 cmd_len;
>   	u32 arg_sz;
>   
> -	memset(npu_slot, 0, sizeof(*npu_slot));
>   	pd = amdxdna_cmd_get_payload(cmd_bo, &cmd_len);
>   	arg_sz = cmd_len - sizeof(*pd);
>   	if (cmd_len < sizeof(*pd) || arg_sz > MAX_NPU_ARGS_SIZE)
> @@ -796,6 +795,7 @@ aie2_cmdlist_fill_npu_elf(struct amdxdna_gem_obj *cmd_bo, void *slot, size_t *si
>   	if (*size < sizeof(*npu_slot) + arg_sz)
>   		return -EINVAL;
>   
> +	memset(npu_slot, 0, sizeof(*npu_slot));
>   	npu_slot->type = EXEC_NPU_TYPE_ELF;
>   	npu_slot->inst_buf_addr = pd->inst_buf;
>   	npu_slot->save_buf_addr = pd->save_buf;

Claude review: accel/amdxdna: Fix out-of-bounds memset in command slot handling

[Claude Code Review Bot]

Patch Review

The fix is applied consistently across all four affected functions: `aie2_cmdlist_fill_npu_cf`, `aie2_cmdlist_fill_npu_dpu`, `aie2_cmdlist_fill_npu_preempt`, and `aie2_cmdlist_fill_npu_elf`.

The pattern in each case is the same. Taking `aie2_cmdlist_fill_npu_cf` as representative:

> -	memset(npu_slot, 0, sizeof(*npu_slot));
>  	cmd = amdxdna_cmd_get_payload(cmd_bo, &cmd_len);
>  	if (*size < sizeof(*npu_slot) + cmd_len)
>  		return -EINVAL;
>
> +	memset(npu_slot, 0, sizeof(*npu_slot));
>  	npu_slot->cu_idx = amdxdna_cmd_get_cu_idx(cmd_bo);

The `slot` pointer comes from `cmdbuf_abo->mem.kva + offset` where the available space is `cmdbuf_abo->mem.size - offset`. When multiple commands are chained, `offset` advances with each slot, and the remaining space can become smaller than `sizeof(struct cmd_chain_slot_npu)` (which is 56+ bytes with the packed struct header fields). Writing `sizeof(*npu_slot)` bytes via memset before checking that there is room would corrupt whatever follows the buffer.

Moving the memset after the size validation is the correct fix. After the check passes, `*size >= sizeof(*npu_slot) + payload_size`, so the memset of `sizeof(*npu_slot)` bytes is guaranteed to be in bounds.

One minor observation on `aie2_cmdlist_fill_npu_elf`: unlike the other three functions, this one does not set `npu_slot->cu_idx` and instead immediately sets `npu_slot->type`. The memset still correctly appears after the size check, so the fix is consistent. This just means `cu_idx` will be left at zero from the memset, which appears intentional since ELF-type slots apparently don't need a CU index.

The commit message is accurate, the Fixes tag correctly identifies the commit that introduced the regression, and the change is minimal and correct. No issues found.

---
Generated by Claude Code Patch Reviewer

Claude review: accel/amdxdna: Fix out-of-bounds memset in command slot handling

[Claude Code Review Bot]

Overall Series Review

Subject: accel/amdxdna: Fix out-of-bounds memset in command slot handling
Author: Lizhi Hou <lizhi.hou@amd.com>
Patches: 2
Reviewed: 2026-02-18T06:45:36.268063

---

This is a single-patch fix for the `accel/amdxdna` driver that addresses an out-of-bounds memset in four command slot filling functions. The bug was introduced by commit 3d32eb7a5ecf which moved the `memset()` call to occur before validation of available slot space. The fix is straightforward: move each `memset()` call to after the size checks that verify the slot buffer is large enough.

The bug is real and reachable. In the chain execution path (`aie2_cmdlist_chain_exec`), the remaining buffer size (`cmdbuf_abo->mem.size - offset`) shrinks with each iteration as slots are packed sequentially. If the remaining space is smaller than `sizeof(struct cmd_chain_slot_npu)`, the old code would `memset()` past the buffer boundary before the `*size < sizeof(*npu_slot) + ...` check could return `-EINVAL`. The fix is correct and minimal.

No cross-patch concerns since this is a single patch.

---
Generated by Claude Code Patch Reviewer