Date: Sat, 14 Mar 2020 19:51:55 +0000 (UTC) From: Patrick Kelsey <pkelsey@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r358996 - head/sys/net Message-ID: <202003141951.02EJptLc099988@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: pkelsey Date: Sat Mar 14 19:51:55 2020 New Revision: 358996 URL: https://svnweb.freebsd.org/changeset/base/358996 Log: Fix iflib zero-length fragment handling The dmamap for zero-length fragments should not be unloaded, as doing so breaks the the cluster-reuse logic in _iflib_fl_refill(). All zero-length fragments are now handled by the assemble_segments() path so that the cluster-reuse logic there does not have to be replicated in the small-single-fragment-packet path of iflib_rxd_pkt_get(). Packets consisting entirely of zero-length fragments (which result in a NULL mbuf pointer) are now properly tolerated. This allows drivers (such as the vmx driver) to pass such packets to iflib when a descriptor error occurs during packet reception, the advantage being that the refill of descriptors associated with the error packet are handled via the existing iflib machinery without having to duplicate parts of that machinery in the driver to handle that error case. Reviewed by: avg, erj, gallatin MFC after: 1 week Differential Revision: https://reviews.freebsd.org/D23945 Modified: head/sys/net/iflib.c Modified: head/sys/net/iflib.c ============================================================================== --- head/sys/net/iflib.c Sat Mar 14 19:43:44 2020 (r358995) +++ head/sys/net/iflib.c Sat Mar 14 19:51:55 2020 (r358996) @@ -2550,7 +2550,8 @@ rxd_frag_to_sd(iflib_rxq_t rxq, if_rxd_frag_t irf, boo MPASS(fl->ifl_cidx == cidx); bus_dmamap_sync(fl->ifl_buf_tag, map, BUS_DMASYNC_POSTREAD); - if (rxq->pfil != NULL && PFIL_HOOKED_IN(rxq->pfil) && pf_rv != NULL) { + if (rxq->pfil != NULL && PFIL_HOOKED_IN(rxq->pfil) && pf_rv != NULL && + irf->irf_len != 0) { payload = *sd->ifsd_cl; payload += ri->iri_pad; len = ri->iri_len - ri->iri_pad; @@ -2587,7 +2588,7 @@ rxd_frag_to_sd(iflib_rxq_t rxq, if_rxd_frag_t irf, boo *pf_rv = PFIL_PASS; } - if (unload) + if (unload && irf->irf_len != 0) bus_dmamap_unload(fl->ifl_buf_tag, map); fl->ifl_cidx = (fl->ifl_cidx + 1) & (fl->ifl_size-1); if (__predict_false(fl->ifl_cidx == 0)) @@ -2672,6 +2673,7 @@ iflib_rxd_pkt_get(iflib_rxq_t rxq, if_rxd_info_t ri) /* should I merge this back in now that the two paths are basically duplicated? */ if (ri->iri_nfrags == 1 && + ri->iri_frags[0].irf_len != 0 && ri->iri_frags[0].irf_len <= MIN(IFLIB_RX_COPY_THRESH, MHLEN)) { m = rxd_frag_to_sd(rxq, &ri->iri_frags[0], false, &sd, &pf_rv, ri); @@ -2688,6 +2690,8 @@ iflib_rxd_pkt_get(iflib_rxq_t rxq, if_rxd_info_t ri) } } else { m = assemble_segments(rxq, ri, &sd, &pf_rv); + if (m == NULL) + return (NULL); if (pf_rv != PFIL_PASS && pf_rv != PFIL_REALLOCED) return (m); }
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?202003141951.02EJptLc099988>