Date: Mon, 21 Nov 2016 21:33:04 -0500 From: David Cross <dcrosstech@gmail.com> To: freebsd-hackers@freebsd.org Subject: FreeBSD 11 i386 disk deadlock (I think) Message-ID: <CAM9edeMRGN7rXTYjGERFuHUc4QkNuwFrHv1wQ%2BWaYTdoCzUPqA@mail.gmail.com>
next in thread | raw e-mail | index | archive | help
I noticed this situation on a Soekris machine I have (geode processor), It has happened twice, once in installworld ,once in buildworld. I am not sure how to trigger it more easily, but the symptoms are quite straightforward. Both times this happened in objcopy. And objcopy was stuck in the "vm_wait" wait-channel. The first time I was able to control-C it (after a few Control-Ts that showed it was not progressing) and everything came back, I was then able to re-run the installworld (that had been the source of the objcopy) without incident. That I was able to Control-C it, and re-run it without a disk-wait process, or blocking in the same spot is telling to me. The second time I had wandered away and by the time I got back I was no longer able to interact with the machine. It still passed traffic, but any attempt to open a new ssh connection was greeted with an opened TCP socket and IMMEDIATE reset/termination. I was forced to reset it. While this was happening, it was successfully routing 6 megabytes/second through its interfaces. I am going to try to run this again to see if I can get any more information, it is apparently fairly easy for me to trip.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAM9edeMRGN7rXTYjGERFuHUc4QkNuwFrHv1wQ%2BWaYTdoCzUPqA>