diff options
| author | Alexander Bluhm <bluhm@cvs.openbsd.org> | 2009-01-22 18:14:22 +0000 |
|---|---|---|
| committer | Alexander Bluhm <bluhm@cvs.openbsd.org> | 2009-01-22 18:14:22 +0000 |
| commit | c4d4977b737b888428e268839d61a44331509067 (patch) | |
| tree | 547a1734893d170a6645bf1cbffb9ff8597446be | |
| parent | 164d8fd6c869aefbaa95a41b3c00356b035747f7 (diff) | |
After the NFS server had sent a reply of size readsize or readdirsize,
the udp receive space got completely filled up. Even if the next
packet from the server was a small reply, it got dropped by
udp_input(). After a second the client resent the NFS request.
Doubling rcvreserve reduces the chance of retransmits by having
enough recv space for multiple NFS replies even if there comes a
big one.
found with pedro@, ok blambert@, thib@, pedro@
| -rw-r--r-- | sys/nfs/nfs_socket.c | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/sys/nfs/nfs_socket.c b/sys/nfs/nfs_socket.c index 016b8a27a8b..1901285d079 100644 --- a/sys/nfs/nfs_socket.c +++ b/sys/nfs/nfs_socket.c @@ -1,4 +1,4 @@ -/* $OpenBSD: nfs_socket.c,v 1.74 2009/01/22 13:51:37 bluhm Exp $ */ +/* $OpenBSD: nfs_socket.c,v 1.75 2009/01/22 18:14:21 bluhm Exp $ */ /* $NetBSD: nfs_socket.c,v 1.27 1996/04/15 20:20:00 thorpej Exp $ */ /* @@ -248,8 +248,8 @@ nfs_connect(nmp, rep) so->so_snd.sb_timeo = 0; if (nmp->nm_sotype == SOCK_DGRAM) { sndreserve = nmp->nm_wsize + NFS_MAXPKTHDR; - rcvreserve = max(nmp->nm_rsize, nmp->nm_readdirsize) + - NFS_MAXPKTHDR; + rcvreserve = (max(nmp->nm_rsize, nmp->nm_readdirsize) + + NFS_MAXPKTHDR) * 2; } else if (nmp->nm_sotype == SOCK_SEQPACKET) { sndreserve = (nmp->nm_wsize + NFS_MAXPKTHDR) * 2; rcvreserve = (max(nmp->nm_rsize, nmp->nm_readdirsize) + |