Date: Sun, 15 Sep 2013 18:13:11 +0300 From: Andriy Gapon <avg@FreeBSD.org> To: J David <j.david.lists@gmail.com> Cc: "freebsd-fs@freebsd.org" <freebsd-fs@FreeBSD.org>, freebsd-stable <freebsd-stable@FreeBSD.org> Subject: Re: zfs_enable vs zfs_load in loader.conf (but neither works) Message-ID: <5235CE87.2090607@FreeBSD.org> In-Reply-To: <CABXB=RR%2B5toQjRJRuf2yuj1updnTn1FqWnckn6fX4N%2BSr_6T4g@mail.gmail.com> References: <CABXB=RTz6jM=B895Bo6Kp-ZAf2pvTZkm-HfS=PrfX=aMKqjMbw@mail.gmail.com> <523310E2.4050702@FreeBSD.org> <52331179.4030201@FreeBSD.org> <CABXB=RR%2B5toQjRJRuf2yuj1updnTn1FqWnckn6fX4N%2BSr_6T4g@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
on 15/09/2013 17:16 J David said the following: > Thanks very much for the info Andriy. > > On Fri, Sep 13, 2013 at 9:22 AM, Andriy Gapon <avg@freebsd.org> wrote: >> Another piece of information is that neither mountpoint nor canmount property >> affects ZFS root mounting. > > It is mountpoint=legacy that boots on this machine and mountpoint=/ > that can't find init, with no other changes. So clearly under some > obscure edge case, this is not strictly correct. I am sure that the same can be said about almost anything documented about any software... > Did your test include zpool root != filesystem root? Because as you > describe one possible cause of the problem is mounting the wrong > filesystem as root, one wonders if somehow with the mountpoint=/ > setting the zpool root (which has no files at all) is incorrectly > being chosen as fsroot with mountpoint=/ on data/root. I.e. perhaps > somewhere in the code is looking for "legacy" (or skipping anything > with a mountpoint set) and defaulting back to the zpool root if it is > not found? I tried to reproduce with exactly the same configuration as you described. To be specific, yes I used data/root as a root filesystem and I set its mountpoint to "/". > Unfortunately there is no way I know of to check and see what the root > filesystem turned out to be after the failure to find init. That > would reduce speculation about what is happening quite a bit. Can the > kernel debugger extract this info? Unfortunately, I am not sure if it is possible to obtain anu useful information from ddb and saving a crash dump is not possible in pre-init environment. I could write a patch that would print some useful debugging info. Will you able to use it? -- Andriy Gapon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?5235CE87.2090607>