[PATCH -next] s390/hmcdrv: Switch to memdup_user_nul() helper

[PATCH -next] s390/hmcdrv: Switch to memdup_user_nul() helper

[Ruan Jinjie]

Use memdup_user_nul() helper instead of open-coding
to simplify the code.

Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
---
 drivers/s390/char/hmcdrv_dev.c | 17 +++++------------
 1 file changed, 5 insertions(+), 12 deletions(-)

diff --git a/drivers/s390/char/hmcdrv_dev.c b/drivers/s390/char/hmcdrv_dev.c
index 8d50c894711f..504cdbf9998d 100644
--- a/drivers/s390/char/hmcdrv_dev.c
+++ b/drivers/s390/char/hmcdrv_dev.c
@@ -252,19 +252,12 @@ static ssize_t hmcdrv_dev_write(struct file *fp, const char __user *ubuf,
 		 fp, (long long) *pos, len);
 
 	if (!fp->private_data) { /* first expect a cmd write */
-		fp->private_data = kmalloc(len + 1, GFP_KERNEL);
-
-		if (!fp->private_data)
-			return -ENOMEM;
-
-		if (!copy_from_user(fp->private_data, ubuf, len)) {
-			((char *)fp->private_data)[len] = '\0';
-			return len;
+		fp->private_data = memdup_user_nul(ubuf, len);
+		if (IS_ERR(fp->private_data)) {
+			fp->private_data = NULL;
+			return PTR_ERR(fp->private_data);
 		}
-
-		kfree(fp->private_data);
-		fp->private_data = NULL;
-		return -EFAULT;
+		return len;
 	}
 
 	retlen = hmcdrv_dev_transfer((char *) fp->private_data,
-- 
2.34.1

Re: [PATCH -next] s390/hmcdrv: Switch to memdup_user_nul() helper

[Heiko Carstens]

On Tue, Aug 15, 2023 at 07:55:46PM +0800, Ruan Jinjie wrote:
> Use memdup_user_nul() helper instead of open-coding
> to simplify the code.
> 
> Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
> ---
>  drivers/s390/char/hmcdrv_dev.c | 17 +++++------------
>  1 file changed, 5 insertions(+), 12 deletions(-)
> 
> diff --git a/drivers/s390/char/hmcdrv_dev.c b/drivers/s390/char/hmcdrv_dev.c
> index 8d50c894711f..504cdbf9998d 100644
> --- a/drivers/s390/char/hmcdrv_dev.c
> +++ b/drivers/s390/char/hmcdrv_dev.c
> @@ -252,19 +252,12 @@ static ssize_t hmcdrv_dev_write(struct file *fp, const char __user *ubuf,
>  		 fp, (long long) *pos, len);
>  
>  	if (!fp->private_data) { /* first expect a cmd write */
> -		fp->private_data = kmalloc(len + 1, GFP_KERNEL);
> -
> -		if (!fp->private_data)
> -			return -ENOMEM;
> -
> -		if (!copy_from_user(fp->private_data, ubuf, len)) {
> -			((char *)fp->private_data)[len] = '\0';
> -			return len;
> +		fp->private_data = memdup_user_nul(ubuf, len);
> +		if (IS_ERR(fp->private_data)) {
> +			fp->private_data = NULL;
> +			return PTR_ERR(fp->private_data);
                        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

This is obviously broken. Sorry, but I'm not willing to review such
"trivial" cleanup patches for subtle bugs anymore - maybe you would
sooner or later succeed and I would apply such a broken patch.

Just don't send such patches; at least I won't take them.