Seeing the dreaded "ZFS: i/o error - all block copies unavailable" on 9.0-CURRENT

[Matt Reimer]

On Tue, Feb 23, 2010 at 8:20 AM, John Baldwin <jhb at freebsd.org> wrote:

> On Thursday 18 February 2010 10:38:38 pm Chris wrote:
> > > I'm seeing some difference in the bootloader output between using the
> > > F12 workaround or not:
> > >
> > > F12 workaround mode:
> > >
> > > FreeBSD/i386 boot
> > > Default: zroot:/boot/zfsloader
> > > boot: status
> > > config:
> > >
> > >             NAME     STATE
> > >             zroot    ONLINE
> > >        gpt/disk0   ONLINE
> > >
> > >
> > > No F12 key pressed:
> > >
> > > FreeBSD/i386 boot
> > > Default: zroot:/boot/kernel/kernel
> > > config:
> > >
> > >              NAME     STATE
> > >               zroot    ONLINE
> > >          gpt/disk0   ONLINE
> > >
> > >
> > > What's up with the differences between the Default boot setting?
> > >
> >
> > Sorry for the spam but I'm seeing something else very interesting....
> > If I boot to Good Mode (F12) and break to the "boot:" prompt, the
> > system boots if I type /boot/zfsloader. However, if I type
> > /boot/kernel/kernel (as was seen in the Default setting when I didn't
> > hit F12),  I get a crash that looks like the following:
> > -----------------------------------------
> > FreeBSD/i386 boot
> > Default: zroot:/boot/zfsloader
> > boot: /boot/kernel/kernel
> > int=00000006    err=00000000  efl=00010882  eip=00200000
> > eax=0018b070  ebx=ed7706bd ecx=c72e2ca8 edx=00000000
> > esi=00000040   edi=000929d0  ebp=0009279c esp=00009401
> > cs=0008  ds=0010   es=0010  fs=0010  gs=0010  ss=0010
> > cs:eip=ff ff ff ff ff ff ff ff 7f-00 00 00 00 00 00 00 00
> >            00 00 00 00 00 00 00 00-00 00 00 00 00 00 00 00
> > ss:esp=36 80 0d 07 90 00 00 01-e9 c4 fc ff ff aa ac 84
> >              c0 0f 84 86 00 00 00 a8-80 74 f2 88 c5 b0 3d aa
> > BTX halted
> > -----------------------------------------
>
> I'm not sure how well we support booting a kernel from boot2 anymore, much
> less from zfsboot.  Using /boot/loader (or /boot/zfsloader) is probably the
> only tested (and realistically supported) case.


Yeah, the last time I tried this it didn't work, but I didn't hunt down why.

Matt

>