RE: [bug report] RDMA/iw_cxgb4: Handle return value of c4iw_ofld_send() in abort_arp_failure()

["Steve Wise" <swise-7bPotxP6k4+P2YhJcF5u+vpXobYPEAuW@public.gmane.org>]

> Any update on this?  It really looks buggy to me.
>

Hey Dan, I think you're correct.  Seems like abort_arp_failure() needs to take a
ref before calling cxgb4_ofld_send() because cxgb4_ofld_send() will free the skb
if there is a fatal error.  So abort_arp_failure() will deref the skb if
cxgb4_ofld_send() suceeds.  Otherwise, the deref will happen in _put_ep_safe().
The same issue exists with FAKE_CPL_PASS_PUT_EP_SAFE too, and maybe all the arp
handlers.


Steve.


 
> regards,
> dan carpenter
> 
> On Thu, Feb 23, 2017 at 06:58:09PM +0300, Dan Carpenter wrote:
> > Hello Hariprasad S,
> >
> > The patch 761e19a504af: "RDMA/iw_cxgb4: Handle return value of
> > c4iw_ofld_send() in abort_arp_failure()" from May 6, 2016, leads to
> > the following static checker warning:
> >
> > 	drivers/infiniband/hw/cxgb4/cm.c:575 abort_arp_failure()
> > 	warn: passing freed memory 'skb'
> >
> > drivers/infiniband/hw/cxgb4/cm.c
> >    559  /*
> >    560   * Handle an ARP failure for a CPL_ABORT_REQ.  Change it into a no
RST
> variant
> >    561   * and send it along.
> >    562   */
> >    563  static void abort_arp_failure(void *handle, struct sk_buff *skb)
> >    564  {
> >    565          int ret;
> >    566          struct c4iw_ep *ep = handle;
> >    567          struct c4iw_rdev *rdev = &ep->com.dev->rdev;
> >    568          struct cpl_abort_req *req = cplhdr(skb);
> >    569
> >    570          PDBG("%s rdev %p\n", __func__, rdev);
> >    571          req->cmd = CPL_ABORT_NO_RST;
> >    572          ret = c4iw_ofld_send(rdev, skb);
> >    573          if (ret) {
> >
> > If c4iw_ofld_send() fails then it frees skb.
> >
> >    574                  __state_set(&ep->com, DEAD);
> >    575                  queue_arp_failure_cpl(ep, skb,
FAKE_CPL_PUT_EP_SAFE);
> >                                                   ^^^
> > Used after free here.  These are ref counted data so it might actually
> > be that I have misread it and miscounted the references so perhaps it's
> > OK, but it looks might suspect at a glance.
> >
> >    576          }
> >    577  }
> >
> > regards,
> > dan carpenter
> --
> To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html