At Tue, 20 Apr 2021 16:53:58 -0700, "Greg A. Woods" <woods%planix.ca@localhost> wrote: Subject: Re: Xen FreeBSD domU block I/O problem on -current only affects reads > 1024 bytes > > With the gracious help of RVP <rvp%SDF.ORG@localhost> I have been able to identify > better what is actually going wrong with FreeBSD's access to NetBSD dom0 > xbdback(4) storage. > > It seems that in certain circumstances (e.g. in newfs and the test > program) whenever FreeBSD issues a read of more than 1024 bytes only the > first 1024 bytes are correct -- the rest of the bytes returned come from > somewhere else on the disk, which appears to be starting at six(6) > sectors after where they were supposed to have come from. Note that > this corresponds to exactly 4096 bytes offset from the beginning of the > read. Reviving this old thread with some new info.... It seems ZFS either doesn't issue large read requests, and/or it works around the problem in some other way. With the help of a custom FreeBSD kernel with ZFS compiled in, and booting it as a PVH domU kernel, and with the new(ish) FreeBSD (14.0) way of installing with a ZFS root, I have a couple of domUs running just fine now, one even recovered old zpools on the machine where I first experienced this problem! As soon as possible, especially if I can dredge up another test server, I'll test plain UFS again with a NetBSD 10.0_RC2 kernel as dom0. -- Greg A. Woods <gwoods%acm.org@localhost> Kelowna, BC +1 250 762-7675 RoboHack <woods%robohack.ca@localhost> Planix, Inc. <woods%planix.com@localhost> Avoncote Farms <woods%avoncote.ca@localhost>
Attachment:
pgpyliXxE0DTD.pgp
Description: OpenPGP Digital Signature