Date: Tue, 23 Mar 2021 22:14:20 +0000 From: Rick Macklem <rmacklem@uoguelph.ca> To: "freebsd-current@freebsd.org" <freebsd-current@freebsd.org> Cc: Alexander Motin <mav@FreeBSD.org>, Peter Eriksson <pen@lysator.liu.se>, Garrett Wollman <wollman@hergotha.csail.mit.edu> Subject: testers for NFS server patch needed Message-ID: <YQXPR0101MB0968F25AD458BC91F163025FDD649@YQXPR0101MB0968.CANPRD01.PROD.OUTLOOK.COM>
index | next in thread | raw e-mail
[-- Attachment #1 --] Hi, A thread over on freebsd-net@ discusses a situation where Linux clients seem to get stuck when the TCP connection is partially torn down. (FIN_WAIT_2 on the client and CLOSE_WAIT on the server.) The thread is here: http://docs.FreeBSD.org/cgi/mid.cgi?YQXPR0101MB0968FB1FF0FC481CE37E9A81DD649 Although the underlying cause is not known, I think that the attached patch, which adds soshutdown(...SHUT_WR) calls where the socket is no longer useful, might fix the problem/ If you have a FreeBSD NFS server (especially if you have Linux clients) where you can patch the server, please try the attached patch and let me know how it goes. rick [-- Attachment #2 --] --- sys/rpc/svc.c.xprtdied 2021-03-20 13:36:50.798335000 -0700 +++ sys/rpc/svc.c 2021-03-20 17:30:41.448270000 -0700 @@ -203,6 +203,8 @@ svcpool_cleanup(SVCPOOL *pool) mtx_unlock(&grp->sg_lock); } TAILQ_FOREACH_SAFE(xprt, &cleanup, xp_link, nxprt) { + if (xprt->xp_socket != NULL) + soshutdown(xprt->xp_socket, SHUT_WR); SVC_RELEASE(xprt); } @@ -388,6 +390,8 @@ xprt_unregister(SVCXPRT *xprt) xprt_unregister_locked(xprt); mtx_unlock(&grp->sg_lock); + if (xprt->xp_socket != NULL) + soshutdown(xprt->xp_socket, SHUT_WR); SVC_RELEASE(xprt); } @@ -1078,6 +1082,7 @@ svc_checkidle(SVCGROUP *grp) mtx_unlock(&grp->sg_lock); TAILQ_FOREACH_SAFE(xprt, &cleanup, xp_link, nxprt) { + soshutdown(xprt->xp_socket, SHUT_WR); SVC_RELEASE(xprt); } mtx_lock(&grp->sg_lock);help
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?YQXPR0101MB0968F25AD458BC91F163025FDD649>