Date: Sun, 26 Aug 2018 03:44:43 +0700 From: Eugene Grosbein <eugen@grosbein.net> To: David Cross <dcrosstech@gmail.com>, FreeBSD Hackers <freebsd-hackers@freebsd.org> Subject: Re: Weird USB DA behavior Message-ID: <83388164-39c8-14e7-d0e7-304cbfdcb796@grosbein.net> In-Reply-To: <CAM9edeP4wTeEpPfpTYBsn1YjOVh2a%2BwYR-OL1Tv0Vv4LXPNUgw@mail.gmail.com> References: <CAM9edePfxANDxXAjgQsZPXzPc3Ezw4Pn%2BdaVcnkaHx1oY%2BUoDA@mail.gmail.com> <20180825010023.GD45503@funkthat.com> <CD43A15C-74B2-4F29-ADB5-B831A0CD5BF6@gmail.com> <20180825052330.GE45503@funkthat.com> <CAM9edeP4wTeEpPfpTYBsn1YjOVh2a%2BwYR-OL1Tv0Vv4LXPNUgw@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
26.08.2018 3:22, David Cross wrote: > I turned debugging ALL the way up on geli (3) and noticed the hangs always > happened when geli handed off a 20480 length read to the layer below (in > this case mirror).. I used the rebugging output to create a dummy program > called 'pread' to try to simulate these failures, and I was quite > successful. Attached is 'pread.c' which takes on stdin a offset and a > length to read. In that version I overwrote it to always ask for 32k., and > that works every time. If I eliminate that line and let the input data > control it, on some of the 20480 (and a different set each time) it hangs, > after the hang the pread returns _0_ (and obviously no errno), no kernel > logs. "pread" is run directly against /dev/da0, so no mirror or geli > layers at this point. The lists strips non-textual attachments, so requestdata.txt get passed but pread.c was stripped from the message. You need to re-send it changing name to something like pread.c.txt
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?83388164-39c8-14e7-d0e7-304cbfdcb796>