Date: Mon, 21 Jan 2019 09:43:36 +0000 From: bugzilla-noreply@freebsd.org To: net@FreeBSD.org Subject: [Bug 234846] [lagg] race condition when adding port Message-ID: <bug-234846-7501-GOiIdz9IVL@https.bugs.freebsd.org/bugzilla/> In-Reply-To: <bug-234846-7501@https.bugs.freebsd.org/bugzilla/> References: <bug-234846-7501@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=234846 --- Comment #15 from Alexandre martins <alexandre.martins@stormshield.eu> --- Hello Stephen, I'm sorry, I have made a mistake when I give the revision of the bug, It's the 317696 that introduce the bug, when the lock was split in two (SX plus RM). I'm resume all of my story: - We are working on a base of FreeBSD 10.3, and faced crashes with LACP - We updated the if_lagg code to the revision 321477, which solved our initial problem - However, another crash is occurring (the one that I put the backtrace here) - When I searched the cause of the race, I take the wrong lock change (333612 instead of 321477) The root cause of the problem remain the same: ether_input call lagg_input with the lacp port not fully initialized. The function lagg_input locks the (RM / EPOCK) in read mode but that lock is not taken in lagg_port_create/lagg_port_destroy, who is locked by the SX. The first patch with LAGG_PORT_DISABLED works for us. My apologies for the confusion. Best regards, Alexandre -- You are receiving this mail because: You are on the CC list for the bug.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-234846-7501-GOiIdz9IVL>