Re: [PATCH] scsi:sg: add sg_remove_request in sg_write

[Douglas Gilbert]

On 2020-04-13 10:13 p.m., Wu Bo wrote:
> From: Wu Bo <wubo40@xxxxxxxxxx>
>
> If the __copy_from_user function return failed,
> it should call sg_remove_request in sg_write.

This is a fix.

Acked-by: Douglas Gilbert <dgilbert@xxxxxxxxxxxx>
> Signed-off-by: Wu Bo <wubo40@xxxxxxxxxx>
> ---
>  Âdrivers/scsi/sg.c | 4 +++-
>  Â1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/scsi/sg.c b/drivers/scsi/sg.c
> index 4e6af59..ff3f532 100644
> --- a/drivers/scsi/sg.c
> +++ b/drivers/scsi/sg.c
> @@ -685,8 +685,10 @@ static int get_sg_io_pack_id(int *pack_id, void __user 
> *buf, size_t count)
>  ÂÂÂÂÂÂÂ hp->flags = input_size; /* structure abuse ... */
>  ÂÂÂÂÂÂÂ hp->pack_id = old_hdr.pack_id;
>  ÂÂÂÂÂÂÂ hp->usr_ptr = NULL;
> -ÂÂÂÂÂÂ if (copy_from_user(cmnd, buf, cmd_size))
> +ÂÂÂÂÂÂ if (copy_from_user(cmnd, buf, cmd_size)) {
> +ÂÂÂÂÂÂÂÂÂÂÂÂÂÂ sg_remove_request(sfp, srp);
>  ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ return -EFAULT;
> +ÂÂÂÂÂÂ }
>  ÂÂÂÂÂÂÂ /*
>  ÂÂÂÂÂÂÂÂ * SG_DXFER_TO_FROM_DEV is functionally equivalent to SG_DXFER_FROM_DEV,
>  ÂÂÂÂÂÂÂÂ * but is is possible that the app intended SG_DXFER_TO_DEV, because there
> --
> 1.8.3.1