You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
After setting hw.ixl.enable_head_writeback=0, I still see an issue with pkt-gen -f txseq where all packets are dropped. It looks like that the hw pointers are showing all slots are available but the user pointers are showing the ring is full. So poll is returning ready over and over again but then there are no available slots and I get stuck in this loop where the user pointers are never updated.
@@ -2117,6 +2118,7 @@ txseq_body(void *data)
break;
D("poll error on queue %d: %s", targ->me,
rv ? strerror(errno) : "timeout");
+ continue
// goto quit;
}
if (pfd.revents & POLLERR) {
@@ -2129,6 +2131,11 @@ txseq_body(void *data)
/* If no room poll() again. */
space = nm_ring_space(ring);
if (!space) {
+ if (ioctl(pfd.fd, NIOCTXSYNC, NULL) < 0) {+ D("ioctl error on queue %d: %s", targ->me,+ strerror(errno));+ goto quit;+ }
continue;
}
similarly for sender_body.
The text was updated successfully, but these errors were encountered:
After setting
hw.ixl.enable_head_writeback=0
, I still see an issue withpkt-gen -f txseq
where all packets are dropped. It looks like that the hw pointers are showing all slots are available but the user pointers are showing the ring is full. So poll is returning ready over and over again but then there are no available slots and I get stuck in this loop where the user pointers are never updated.similarly for
sender_body
.The text was updated successfully, but these errors were encountered: