Date: Mon, 2 Dec 2019 19:57:39 +0000 (UTC) From: "Kenneth D. Merry" <ken@FreeBSD.org> To: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: svn commit: r355299 - head/sys/cam/scsi Message-ID: <201912021957.xB2JvdUe044607@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: ken Date: Mon Dec 2 19:57:39 2019 New Revision: 355299 URL: https://svnweb.freebsd.org/changeset/base/355299 Log: Fix a hang introduced in r351599. My changes in 351599 (kindly committed by avg) made the cd(4) media check asynchronous to avoid a sleep while holding a mutex. There was a difficult to reproduce bug with those changes that caused a hang on boot on some single processor machines/VMs. Leandro Lupori managed to reproduce the bug, diagnose it, and supplied a patch! Here is his analysis, from the PR: ====== I was able to reproduce the problem described in comment#14. Actually, I wasn't trying to reproduce it, I just started seeing it a few weeks ago, in CURRENT. I can reproduce it consistently, by using QEMU to run a PowerPC64 VM with a single core/thread (-smp 1). It happens only when there is no media in the emulated CD-ROM, a device that QEMU adds by default, unless -nodefaults is specified in command line. I've debugged it and this is what I've found: 1- After the CD probe is successful, GEOM will try to open the device, which will end up calling cdcheckmedia(), that sets CD state to CD_STATE_MEDIA_PREVENT. 2- Next, scsi_prevent() is executed and succeeds, the CD_FLAG_DISC_LOCKED flag is set and CD state moves to CD_STATE_MEDIA_SIZE. 3- Next, scsi_read_capacity() is executed and fails, state is set to CD_STATE_MEDIA_ALLOW, cdmediaprobedone() is called and wakes up cdcheckmedia(). 4- Then, when cdstart() is invoked to process CD_STATE_MEDIA_ALLOW, it first checks if CD_FLAG_DISC_LOCKED is set, and if so skips directly to CD_STATE_MEDIA_SIZE state. This will repeat the steps of bullet 3, entering an infinite MEDIA_SIZE command loop. When there is a least another core/thread, the GEOM thread that performed the initial cdopen() will get scheduled again, closing the CD device, that will call cdprevent(PR_ALLOW) that clears the CD_FLAG_DISC_LOCKED flag and breaks the loop. So, apparently, the problem is CD_STATE_MEDIA_ALLOW being skipped when CD_FLAG_DISC_LOCKED is set. If I understand correctly, in this case, the state should be advanced to CD_STATE_MEDIA size only when the current state is CD_STATE_MEDIA_PREVENT. ===== PR: kern/219857 Submitted by: Leandro Lupori <leandro.lupori@gmail.com> MFC after: 1 week Modified: head/sys/cam/scsi/scsi_cd.c Modified: head/sys/cam/scsi/scsi_cd.c ============================================================================== --- head/sys/cam/scsi/scsi_cd.c Mon Dec 2 19:57:20 2019 (r355298) +++ head/sys/cam/scsi/scsi_cd.c Mon Dec 2 19:57:39 2019 (r355299) @@ -1032,7 +1032,8 @@ cdstart(struct cam_periph *periph, union ccb *start_cc * If the CD is already locked, we don't need to do this. * Move on to the capacity check. */ - if ((softc->flags & CD_FLAG_DISC_LOCKED) != 0) { + if (softc->state == CD_STATE_MEDIA_PREVENT + && (softc->flags & CD_FLAG_DISC_LOCKED) != 0) { softc->state = CD_STATE_MEDIA_SIZE; xpt_release_ccb(start_ccb); xpt_schedule(periph, CAM_PRIORITY_NORMAL);
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201912021957.xB2JvdUe044607>