[PATCH v4] staging: rtl8723bs: remove redundant rsp_allocated_buf

[PATCH v4] staging: rtl8723bs: remove redundant rsp_allocated_buf

[Devansh Soni]

The original code allocated extra memory and manually aligned the
rsp_buf pointer to a 4-byte boundary, however kzalloc() guarantees a
minimum of 8-byte alignment so this was unnecessary.

Also, because the pointer was shifted, the original pointer had to be
stored in rsp_allocated_buf just so it could be passed to kfree() later.

Remove the redundant alignment math, remove the extra 4 bytes of padding
from kzalloc() call, and assign memory directly to rsp_buf.

This allows us to remove the rsp_allocated_buf variable from cmd_priv
struct.

Suggested-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Devansh Soni <devanshsoni874@gmail.com>
---
Changes in v4:
- Dropped PTR_ALIGN() entirely as kzalloc() 8-byte alignment is sufficient.
- Removed the +4 byte padding from kzalloc().
- Completely removed the rsp_allocated_buf tracking variable and updated kfree().

Changes in v3:
- Replaced manual bitwise math with PTR_ALIGN() macro based on feedback from v1.
- Updated commit message to detail kzalloc() 8-byte alignment guarantees.

Changes in v2:
- Wrapped commit log text to resolve line length issue noted by Greg.

 drivers/staging/rtl8723bs/core/rtw_cmd.c    | 9 +++------
 drivers/staging/rtl8723bs/include/rtw_cmd.h | 1 -
 2 files changed, 3 insertions(+), 7 deletions(-)

diff --git a/drivers/staging/rtl8723bs/core/rtw_cmd.c b/drivers/staging/rtl8723bs/core/rtw_cmd.c
index b932670f5..fd1ba1b4e 100644
--- a/drivers/staging/rtl8723bs/core/rtw_cmd.c
+++ b/drivers/staging/rtl8723bs/core/rtw_cmd.c
@@ -178,15 +178,12 @@ int rtw_init_cmd_priv(struct	cmd_priv *pcmdpriv)
 
 	pcmdpriv->cmd_buf = PTR_ALIGN(pcmdpriv->cmd_allocated_buf, CMDBUFF_ALIGN_SZ);
 
-	pcmdpriv->rsp_allocated_buf = kzalloc(MAX_RSPSZ + 4, GFP_ATOMIC);
-	if (!pcmdpriv->rsp_allocated_buf) {
+	pcmdpriv->rsp_buf = kzalloc(MAX_RSPSZ, GFP_ATOMIC);
+	if (!pcmdpriv->rsp_buf) {
 		kfree(pcmdpriv->cmd_allocated_buf);
 		return -ENOMEM;
 	}
 
-	pcmdpriv->rsp_buf = pcmdpriv->rsp_allocated_buf + 4 -
-		((SIZE_PTR)(pcmdpriv->rsp_allocated_buf) & 3);
-
 	pcmdpriv->cmd_issued_cnt = 0;
 	pcmdpriv->cmd_done_cnt = 0;
 	pcmdpriv->rsp_cnt = 0;
@@ -232,7 +229,7 @@ void _rtw_free_cmd_priv(struct	cmd_priv *pcmdpriv)
 	if (pcmdpriv) {
 		kfree(pcmdpriv->cmd_allocated_buf);
 
-		kfree(pcmdpriv->rsp_allocated_buf);
+		kfree(pcmdpriv->rsp_buf);
 
 		mutex_destroy(&pcmdpriv->sctx_mutex);
 	}
diff --git a/drivers/staging/rtl8723bs/include/rtw_cmd.h b/drivers/staging/rtl8723bs/include/rtw_cmd.h
index f78aa0d7a..66b970955 100644
--- a/drivers/staging/rtl8723bs/include/rtw_cmd.h
+++ b/drivers/staging/rtl8723bs/include/rtw_cmd.h
@@ -45,7 +45,6 @@
 		u8 *cmd_buf;	/* shall be non-paged, and 4 bytes aligned */
 		u8 *cmd_allocated_buf;
 		u8 *rsp_buf;	/* shall be non-paged, and 4 bytes aligned */
-		u8 *rsp_allocated_buf;
 		u32 cmd_issued_cnt;
 		u32 cmd_done_cnt;
 		u32 rsp_cnt;
-- 
2.54.0

Re: [PATCH v4] staging: rtl8723bs: remove redundant rsp_allocated_buf

[Dan Carpenter]

On Wed, Jun 17, 2026 at 06:08:53PM +0530, Devansh Soni wrote:
> The original code allocated extra memory and manually aligned the
> rsp_buf pointer to a 4-byte boundary, however kzalloc() guarantees a
> minimum of 8-byte alignment so this was unnecessary.
> 
> Also, because the pointer was shifted, the original pointer had to be
> stored in rsp_allocated_buf just so it could be passed to kfree() later.
> 
> Remove the redundant alignment math, remove the extra 4 bytes of padding
> from kzalloc() call, and assign memory directly to rsp_buf.
> 
> This allows us to remove the rsp_allocated_buf variable from cmd_priv
> struct.
> 
> Suggested-by: Dan Carpenter <error27@gmail.com>
> Signed-off-by: Devansh Soni <devanshsoni874@gmail.com>
> ---
> Changes in v4:
> - Dropped PTR_ALIGN() entirely as kzalloc() 8-byte alignment is sufficient.
> - Removed the +4 byte padding from kzalloc().
> - Completely removed the rsp_allocated_buf tracking variable and updated kfree().

Thanks!

Reviewed-by: Dan Carpenter <error27@gmail.com>

regards,
dan carpenter