Date: Sat, 14 Apr 2018 19:59:37 +0000 From: bugzilla-noreply@freebsd.org To: bugs@FreeBSD.org Subject: [Bug 227285] File descriptor passing does not work reliably on SMP system Message-ID: <bug-227285-227-1QFS4DFUWm@https.bugs.freebsd.org/bugzilla/> In-Reply-To: <bug-227285-227@https.bugs.freebsd.org/bugzilla/> References: <bug-227285-227@https.bugs.freebsd.org/bugzilla/>
next in thread | previous in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D227285 --- Comment #7 from Jan Kokem=C3=BCller <jan.kokemueller@gmail.com> --- Created attachment 192515 --> https://bugs.freebsd.org/bugzilla/attachment.cgi?id=3D192515&action= =3Dedit patch for a GC that detects cycles I've tried to modify the GC so that it detects cycles in all "in flight" sockets and only destroys sockets that are in such a cycle. With this patch= the test program doesn't exit anymore. I wonder if there can be false negatives with a GC that just detects cycles. DragonFly has some interesting comments[1]. DragonFly still suffers from th= is race, though. It looks like Linux had the same race when they still used a mark & sweep GC instead of a cycle detecting one. Search for "rewrite AF_UNIX garbage collector". The BUGS section in the man page for send(2) looks related to this problem. This seems to be a long standing issue. [1]: https://github.com/DragonFlyBSD/DragonFlyBSD/blob/b9f81f6ad2f9ae32fb3ec268a= 1ea2edac10f373e/sys/kern/uipc_usrreq.c#L72 --=20 You are receiving this mail because: You are the assignee for the bug.=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-227285-227-1QFS4DFUWm>