diff options
| author |   bluhm <bluhm@openbsd.org> | 2016-09-02 13:28:21 +0000 |
|---|---|---|
| committer | bluhm <bluhm@openbsd.org> | 2016-09-02 13:28:21 +0000 |
| commit | 6b8b9a1c2922c006e7713298cc1cf5fe355511fb (patch) | |
| tree | c16bf9314552a9579f0342c6be161f08ef342f59 | |
| parent | As done in httpd, (re-)initialize ps_what in all processes. This is (diff) | |
| download | wireguard-openbsd-6b8b9a1c2922c006e7713298cc1cf5fe355511fb.tar.xz wireguard-openbsd-6b8b9a1c2922c006e7713298cc1cf5fe355511fb.zip | |
After allocating a single 64 KB mbuf cluster in sosend(), the sending
socket buffer had no space anymore. The default mbuf space limit
was only 32 KB. So no more data from user-land was accepted. As
tcp_output() keeps the mbuf cluster for retransmits, it will be
freed only after all ACKs have been received. That has killed our
TCP send performance totally. To allow cycling through the mbufs
periodically, we need space for at least 3 of them.
Reported by Andreas Bartelt; testing with mikeb@; OK mikeb@ claudio@
| -rw-r--r-- | sys/kern/uipc_socket2.c | 5 |
1 files changed, 3 insertions, 2 deletions
diff --git a/sys/kern/uipc_socket2.c b/sys/kern/uipc_socket2.c index f86e3546b0c..d0f8abb3d7d 100644 --- a/sys/kern/uipc_socket2.c +++ b/sys/kern/uipc_socket2.c @@ -1,4 +1,4 @@ -/* $OpenBSD: uipc_socket2.c,v 1.64 2016/06/28 14:47:00 tedu Exp $ */ +/* $OpenBSD: uipc_socket2.c,v 1.65 2016/09/02 13:28:21 bluhm Exp $ */ /* $NetBSD: uipc_socket2.c,v 1.11 1996/02/04 02:17:55 christos Exp $ */ /* @@ -397,7 +397,8 @@ sbreserve(struct sockbuf *sb, u_long cc) if (cc == 0 || cc > sb_max) return (1); sb->sb_hiwat = cc; - sb->sb_mbmax = min(cc * 2, sb_max + (sb_max / MCLBYTES) * MSIZE); + sb->sb_mbmax = max(3 * MAXMCLBYTES, + min(cc * 2, sb_max + (sb_max / MCLBYTES) * MSIZE)); if (sb->sb_lowat > sb->sb_hiwat) sb->sb_lowat = sb->sb_hiwat; return (0); |