Date: Fri, 14 Apr 2017 14:10:56 -0700 From: Maxim Sobolev <sobomax@sippysoft.com> To: "Ngie Cooper (yaneurabeya)" <yaneurabeya@gmail.com> Cc: Peter Wemm <peter@wemm.org>, svn-src-head@freebsd.org, svn-src-all@freebsd.org, src-committers <src-committers@freebsd.org>, Hiren Panchasara <hiren@freebsd.org> Subject: Re: svn commit: r316874 - head/sys/kern Message-ID: <CAH7qZfvPQpmTkypWdv1K=9yKF6was1sXuXY7McAeTrC8mvYZbg@mail.gmail.com> In-Reply-To: <513B1846-02E3-49E0-BAA2-CA923EDC5DFF@gmail.com> References: <201704141723.v3EHNS3B043902@repo.freebsd.org> <12133324.zHp9vo75vZ@overcee.wemm.org> <CAH7qZfuxM2pWvPh-LGP8_vrSpm==hTCWjnoZoXSvaFSgAaS8pg@mail.gmail.com> <6942092.OsECkgAp13@overcee.wemm.org> <513B1846-02E3-49E0-BAA2-CA923EDC5DFF@gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
[-- Attachment #1 --] Peter, Ngie, Looks like out of that refactoring came a logical bug that is present in the head, which causes syslod to first to shutdown the socket for reading and then try to select/recv on it (which is somewhat stupid). And that issue has been masked by shutdown() on datagram socket becoming effectively a NOP in 11 & head 20 months ago. It only affects head though, 11-stable still has the old code which does not include that half-closed socket into the select list. Attached patch is expected to fix head, Peter, it would be nice if you can give it a try (restoring latest changes into uipc_sockets.c) and let me know if it helps. Thanks! On Fri, Apr 14, 2017 at 12:48 PM, Ngie Cooper (yaneurabeya) < yaneurabeya@gmail.com> wrote: > > > On Apr 14, 2017, at 12:46, Peter Wemm <peter@wemm.org> wrote: > > > > On Friday, April 14, 2017 12:41:52 PM Maxim Sobolev wrote: > >> Thanks, Peter. I will try to look into this asap. > > > > I don't understand what is going on yet. Presumably there must be other > > changes in play that affect udp/select sometime between the original 2015 > > change and this. The syslogd -s code is Old(TM). I'm also wondering > whether > > the -s code even works at all since the 2015 / r285910 change... > > syslogd has been refactored a lot on ^/head. I don’t think it’s safe to > say that the ^/head and ^/stable/11 and earlier copies will function the > same. > Thanks, > -Ngie > -- Maksym Sobolyev Sippy Software, Inc. Internet Telephony (VoIP) Experts Tel (Canada): +1-778-783-0474 Tel (Toll-Free): +1-855-747-7779 Fax: +1-866-857-6942 Web: http://www.sippysoft.com MSN: sales@sippysoft.com Skype: SippySoft [-- Attachment #2 --] Index: syslogd.c =================================================================== --- syslogd.c (revision 316854) +++ syslogd.c (working copy) @@ -702,7 +702,7 @@ sizeof(fd_mask)); STAILQ_FOREACH(sl, &shead, next) { - if (sl->sl_socket != -1) + if (sl->sl_socket != -1 && sl->sl_recv != NULL) FD_SET(sl->sl_socket, fdsr); } i = select(fdsrmax + 1, fdsr, NULL, NULL, @@ -2877,6 +2877,7 @@ struct addrinfo hints, *res, *res0; int error; char *cp; + int (*sl_recv)(struct socklist *); /* * We have to handle this case for backwards compatibility: * If there are two (or more) colons but no '[' and ']', @@ -3003,6 +3004,7 @@ } dprintf("new socket fd is %d\n", s); listen(s, 5); + sl_recv = socklist_recv_sock; dprintf("shutdown\n"); if (SecureMode || res->ai_family == AF_LOCAL) { /* Forbid communication in secure mode. */ @@ -3013,6 +3015,7 @@ die(0); } dprintf("listening on socket\n"); + sl_recv = NULL; } else dprintf("sending on socket\n"); addsock(res->ai_addr, res->ai_addrlen, @@ -3019,7 +3022,7 @@ &(struct socklist){ .sl_socket = s, .sl_peer = pe, - .sl_recv = socklist_recv_sock + .sl_recv = sl_recv }); } freeaddrinfo(res0);
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAH7qZfvPQpmTkypWdv1K=9yKF6was1sXuXY7McAeTrC8mvYZbg>