Boot fails @r264070

[Nikolai Lifanov]

On 04/03/14 12:06, Benjamin Kaduk wrote:
> On Thu, 3 Apr 2014, David Wolfskill wrote:
> 
>> And on my laptop, I had no problems (using AHCI).  The build machine,
>> though, starts OK, then comes to an inglorious end:
>>
>> Trying to mount root from ufs:/dev/aacd0s4a [rw]...
>> mountroot: waiting for device /dev/aacd0s4a ...
>> Mounting from ufs:/dev/aacd0s4a failed with error 19.
>>
> [...]
>>
>> mountroot>
>>
>>
>> So... I have serial console access; is there any debugging I might
>> do to help figure out what's wrong?  Or is what's wrong already
>> known?
>>
>> I have been able to reboot from this point to one of the other
>> slices; I could probably also unload the default kernel & load
>> yesterday's (err... make that "the day before's" -- yesterday's
>> didn't build -- @r263983).
> 
> I'm also having some trouble booting (into single user mode, so as to
> run the installworld), at r264039.  It seems that ciss never attaches,
> so my da0 (which is supposed to be on ciss0) does not appear.  My
> kernel.old is much older, __FreeBSD_version 1100005, but does boot. 
> However, attempting to boot kernel.old in either single-user mode or
> verbse mode also drops me to mountroot> .  That would seem to indicate
> some hardware problem or timing issues more than a problem with the code
> itself, so I had been holding off on posting until I could get more
> information.
> 
> Unfortunately, my serial console is not working at the moment, so
> getting more information is a bit challenging.  Diffing the dmesg from
> the good and bad boots would probably be helpful for you to do.
> 
> -Ben

I reported a similar problem on freebsd-current@ earlier today.
Fortunately, my boot zpool is still fine, but a RocketRAID controller
only shows 2/4 disks, which just happens to be a full leg of another
raid10-like zpool.

Try reverting revisions r264007-264013. This fixes the problem for me.

- Nikolai Lifanov