Re: [Outreachy kernel] [PATCH] staging: lustre: lent: add missing __user __user annotation

[Greg KH <gregkh@linuxfoundation.org>]

On Fri, Oct 09, 2015 at 12:30:13AM +0300, Ioana Ciornei wrote:
> Add missing __user annotation when using copy_{to/from}_user() to indicate that is
> a pointer from userspace and that it should not be derefenced.
> Also, remove sparse warnings such as:
> 
> drivers/staging/lustre/lnet/selftest/console.c:1474:26: warning: incorrect type in argument 1 (different address spaces)
>   1 drivers/staging/lustre/lnet/selftest/console.c:1474:26:    expected void [noderef] <asn:1>*to
>   2 drivers/staging/lustre/lnet/selftest/console.c:1474:26:    got struct sfw_counters_t [usertype] *[assigned] sfwk_stat
> 
> Signed-off-by: Ioana Ciornei <ciorneiioana@gmail.com>
> ---
>  drivers/staging/lustre/lnet/selftest/console.c | 44 +++++++++++++-------------
>  1 file changed, 22 insertions(+), 22 deletions(-)
> 
> diff --git a/drivers/staging/lustre/lnet/selftest/console.c b/drivers/staging/lustre/lnet/selftest/console.c
> index 024aaee..7ba9022 100644
> --- a/drivers/staging/lustre/lnet/selftest/console.c
> +++ b/drivers/staging/lustre/lnet/selftest/console.c
> @@ -384,9 +384,9 @@ lstcon_sesrpc_readent(int transop, srpc_msg_t *msg,
>  	case LST_TRANS_SESQRY:
>  		rep = &msg->msg_body.dbg_reply;
>  
> -		if (copy_to_user(&ent_up->rpe_priv[0],
> +		if (copy_to_user((void __user *)&ent_up->rpe_priv[0],
>  				     &rep->dbg_timeout, sizeof(int)) ||
> -		    copy_to_user(&ent_up->rpe_payload[0],
> +		    copy_to_user((void __user *)&ent_up->rpe_payload[0],
>  				     &rep->dbg_name, LST_NAME_SIZE))
>  			return -EFAULT;
>  

The __user annotation in the lustre code is a mess, lots of work needs
to be done here, but it's complex and tricky, and in the end is going to
probably require some API changes to do correctly.  I strongly recommend
just leaving it alone and letting the lustre developers fix it up
properly sometime in the future, papering over the warnings with these
casts just gives the impression that the issue is fixed, when it really
isn't.

sorry, but I can't take this change.

greg k-h