Re: [PATCH] 9p/xen: Fix end of loop tests for list_for_each_entry

From: Stefano Stabellini
Date: Mon Jul 26 2021 - 17:30:47 EST


On Mon, 26 Jul 2021, asmadeus@xxxxxxxxxxxxx wrote:
> Harshvardhan Jha wrote on Sun, Jul 25, 2021 at 11:21:03PM +0530:
> > The list_for_each_entry() iterator, "priv" in this code, can never be
> > NULL so the warning would never be printed.
>
> hm? priv won't be NULL but priv->client won't be client, so it will
> return -EINVAL alright in practice?
>
> This does fix an invalid read after the list head, so there's a real
> bug, but the commit message needs fixing.

Agreed


> > Signed-off-by: Harshvardhan Jha <harshvardhan.jha@xxxxxxxxxx>
> > ---
> > From static analysis. Not tested.
>
> +Stefano in To - I also can't test xen right now :/
> This looks functional to me but if you have a bit of time to spare just
> a mount test can't hurt.

Yes, I did test it successfully. Aside from the commit messaged to be
reworded:

Reviewed-by: Stefano Stabellini <sstabellini@xxxxxxxxxx>
Tested-by: Stefano Stabellini <sstabellini@xxxxxxxxxx>


> > ---
> > net/9p/trans_xen.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/net/9p/trans_xen.c b/net/9p/trans_xen.c
> > index f4fea28e05da..3ec1a51a6944 100644
> > --- a/net/9p/trans_xen.c
> > +++ b/net/9p/trans_xen.c
> > @@ -138,7 +138,7 @@ static bool p9_xen_write_todo(struct xen_9pfs_dataring *ring, RING_IDX size)
> >
> > static int p9_xen_request(struct p9_client *client, struct p9_req_t *p9_req)
> > {
> > - struct xen_9pfs_front_priv *priv = NULL;
> > + struct xen_9pfs_front_priv *priv;
> > RING_IDX cons, prod, masked_cons, masked_prod;
> > unsigned long flags;
> > u32 size = p9_req->tc.size;
> > @@ -151,7 +151,7 @@ static int p9_xen_request(struct p9_client *client, struct p9_req_t *p9_req)
> > break;
> > }
> > read_unlock(&xen_9pfs_lock);
> > - if (!priv || priv->client != client)
> > + if (list_entry_is_head(priv, &xen_9pfs_devs, list))
> > return -EINVAL;
> >
> > num = p9_req->tc.tag % priv->num_rings;