On Wed, Mar 28, 2018 at 09:14:25PM +0530, Varun Prakash wrote:
> On Wed, Mar 21, 2018 at 09:12:00PM -0400, Martin K. Petersen wrote:
> >
> > Varun: Please look at this. Thanks!
> >
> > > What happened to this one?
> > >
> > > regards,
> > > dan carpenter
> > >
> > >
> > > On Wed, Nov 29, 2017 at 02:42:20PM +0300, Dan Carpenter wrote:
> > >> The story is that Smatch marks skb->data as untrusted and so it
> > >> complains about this code:
> > >>
> > >> drivers/scsi/cxgbi/cxgb4i/cxgb4i.c:2111 t4_uld_rx_handler()
> > >> error: buffer overflow 'cxgb4i_cplhandlers' 239 <= 255.
> > >>
> > >> I don't know the code very well, but it looks like a reasonable warning
> > >> message. Let's address it by adding a sanity check to make sure "opc"
> > >> is within bounds.
> > >>
> > >> Fixes: bbc02c7e9d34 ("cxgb4: Add register, message, and FW definitions")
> > >> Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> > >>
> > >> diff --git a/drivers/scsi/cxgbi/cxgb4i/cxgb4i.c b/drivers/scsi/cxgbi/cxgb4i/cxgb4i.c
> > >> index 266eddf17a99..94b2d5660a07 100644
> > >> --- a/drivers/scsi/cxgbi/cxgb4i/cxgb4i.c
> > >> +++ b/drivers/scsi/cxgbi/cxgb4i/cxgb4i.c
> > >> @@ -2108,12 +2108,12 @@ static int t4_uld_rx_handler(void *handle, const __be64 *rsp,
> > >> log_debug(1 << CXGBI_DBG_TOE,
> > >> "cdev %p, opcode 0x%x(0x%x,0x%x), skb %p.\n",
> > >> cdev, opc, rpl->ot.opcode_tid, ntohl(rpl->ot.opcode_tid), skb);
> > >> - if (cxgb4i_cplhandlers[opc])
> > >> - cxgb4i_cplhandlers[opc](cdev, skb);
> > >> - else {
> > >> + if (opc >= ARRAY_SIZE(cxgb4i_cplhandlers) || !cxgb4i_cplhandlers[opc]) {
> > >> pr_err("No handler for opcode 0x%x.\n", opc);
> > >> __kfree_skb(skb);
> > >> + return 0;
> > >> }
> > >> + cxgb4i_cplhandlers[opc](cdev, skb);
> > >> return 0;
> > >> nomem:
> > >> log_debug(1 << CXGBI_DBG_TOE, "OOM bailing out.\n");
> > >
> > >
>
> This check is not necessary but we can add it to avoid warning.
Is the reason it's not necessary, because the skb->data comes from the
firmware and we trust it? The v5 declares the array as having 256
elements which also solves this warning. And cxgbit_uld_lro_rx_handler()
has a bounds check. So it seems pretty normal to prevent the array
overflow by force as well as by trust.
> The commit mentioned in "Fixes" is not correct, this code was added in commit
> "7b36b6e [SCSI] cxgb4i v5: iscsi driver"
Yeah. You're right. I can resend with an updated commit message.
regards,
dan carpenter
