git: bdaddb76ef1b - main - iflib: Pass up the frames received before a descriptor error

The branch main has been updated by rcm:

URL: https://cgit.FreeBSD.org/src/commit/?id=bdaddb76ef1b9d5537f4a76cff01cdb4795804a4

commit bdaddb76ef1b9d5537f4a76cff01cdb4795804a4
Author:     R. Christian McDonald <rcm@FreeBSD.org>
AuthorDate: 2026-10-19 01:06:41 +0000
Commit:     R. Christian McDonald <rcm@FreeBSD.org>
CommitDate: 2026-10-19 01:06:41 +0000

    iflib: Pass up the frames received before a descriptor error
    
    When the driver's isc_rxd_pkt_get() fails part way through a receive
    pass, iflib_rxeof() jumps straight to the reset request.  The frames it
    already took off the ring and chained for if_input() are neither passed
    up nor freed, so up to budget - 1 frames are lost and their mbufs and
    clusters leak on every such error, and the frames queued to LRO in the
    pass wait for the next one.  Leave the loop instead, pass the chain up
    and flush LRO as on the normal exit, then request the reset as before,
    without refilling.
    
    The path is reachable from several drivers: igc, em, igb and ixgbe
    return EBADMSG for frames received with errors, which the hardware
    passes up only with store-bad-packets enabled; ixl, iavf and ice for
    every descriptor flagged RXE; aq, enetc, mgb and enic for descriptors
    they cannot use.
    
    Reviewed by:            gallatin
    Sponsored by:           Rubicon Communications, LLC ("Netgate")
    Differential Revision:  https://reviews.freebsd.org/D60805
---
 sys/net/iflib.c | 27 ++++++++++++++++-----------
 1 file changed, 16 insertions(+), 11 deletions(-)

diff --git a/sys/net/iflib.c b/sys/net/iflib.c
index 9ca08eaee6e8..d47c108292ae 100644
--- a/sys/net/iflib.c
+++ b/sys/net/iflib.c
@@ -3274,6 +3274,7 @@ iflib_rxeof(iflib_rxq_t rxq, qidx_t budget)
 	struct if_rxd_info ri;
 	int err, budget_left, rx_bytes, rx_pkts;
 	iflib_fl_t fl;
+	bool rx_err;
 #if defined(INET6) || defined(INET)
 	int lro_enabled;
 #endif
@@ -3289,6 +3290,7 @@ iflib_rxeof(iflib_rxq_t rxq, qidx_t budget)
 
 	ifp = ctx->ifc_ifp;
 	mh = mt = NULL;
+	rx_err = false;
 	MPASS(budget > 0);
 	rx_pkts	= rx_bytes = 0;
 	if (sctx->isc_flags & IFLIB_HAS_RXCQ)
@@ -3323,9 +3325,9 @@ iflib_rxeof(iflib_rxq_t rxq, qidx_t budget)
 		ri.iri_frags = rxq->ifr_frags;
 		err = ctx->isc_rxd_pkt_get(ctx->ifc_softc, &ri);
 
-		if (err) {
-			CURVNET_RESTORE();
-			goto err;
+		if (__predict_false(err != 0)) {
+			rx_err = true;
+			break;
 		}
 		rx_pkts += 1;
 		rx_bytes += ri.iri_len;
@@ -3372,8 +3374,10 @@ iflib_rxeof(iflib_rxq_t rxq, qidx_t budget)
 	}
 	CURVNET_RESTORE();
 	/* make sure that we can refill faster than drain */
-	for (i = 0, fl = &rxq->ifr_fl[0]; i < sctx->isc_nfl; i++, fl++)
-		retval |= iflib_fl_refill_all(ctx, fl);
+	if (__predict_true(!rx_err)) {
+		for (i = 0, fl = &rxq->ifr_fl[0]; i < sctx->isc_nfl; i++, fl++)
+			retval |= iflib_fl_refill_all(ctx, fl);
+	}
 
 	if (mh != NULL) {
 		if_input(ifp, mh);
@@ -3389,15 +3393,16 @@ iflib_rxeof(iflib_rxq_t rxq, qidx_t budget)
 #if defined(INET6) || defined(INET)
 	tcp_lro_flush_all(&rxq->ifr_lc);
 #endif
+	if (__predict_false(rx_err)) {
+		STATE_LOCK(ctx);
+		ctx->ifc_flags |= IFC_DO_RESET;
+		iflib_admin_intr_deferred(ctx);
+		STATE_UNLOCK(ctx);
+		return (0);
+	}
 	if (avail != 0 || iflib_rxd_avail(ctx, rxq, *cidxp, 1) != 0)
 		retval |= IFLIB_RXEOF_MORE;
 	return (retval);
-err:
-	STATE_LOCK(ctx);
-	ctx->ifc_flags |= IFC_DO_RESET;
-	iflib_admin_intr_deferred(ctx);
-	STATE_UNLOCK(ctx);
-	return (0);
 }
 
 #define TXD_NOTIFY_COUNT(txq) (((txq)->ift_size / (txq)->ift_update_freq) - 1)