Re: New builds won't boot (fwd)

On Mon, Jun 29, 2009 at 4:56 PM, Chris
Hedley<freebsd-current_at_chrishedley.com> wrote:
> On Fri, 26 Jun 2009, Chris Hedley wrote:
>>
>> On Fri, 26 Jun 2009, Chris Hedley wrote:
>>>
>>> On Sun, 7 Jun 2009, Tim Kientzle wrote:
>>>>
>>>> * Latest checkout date of a kernel that does boot.
>>>> * Earliest checkout date of a kernel that doesn't boot.
>>
>>        ...
>>>
>>> As for the versions of the kernel, I've narrowed it down to a half-day
>>> window, which is hopefully useful: the cvsup-specified dates I have are
>>> 2009.02.18.12.00.00 (working) and 2009.02.19.00.00.00 (not working).
>>
>> Just to pontificate for a moment, I notice that there's quite a few
>> changes to the ATA subsystem that afternoon.  This might be significant as
>> the Supermicro SATA controllers I use (a pair of AOC-SAT2-MV8 cards, 8 port
>> PCI-X things) aren't entirely trouble-free: when they're up and running
>> they're fine, but they don't reset on reboot, even with the current (? 1.0c,
>> anyway) firmware which causes problems rescanning the drives.  This could be
>> a complete red herring, but I wondered if it may be related.
>
> I've been fiddling about with this a bit more and I'm not sure it's just a
> zfs problem as it seems gmirror also can't find its entries.  Attempting to
> boot of a non-zfs partition also leads to it hanging near the end of the
> kernel's boot messages.
>
> Anybody got any ideas of how I can get my system or AOC-SAT2-MV8 to work
> with the post 18th Feb kernels?  Or of any ddb wizardry to further probe it
> for what it might be doing while it's apparently thinking about things?
>



I went through the commits from that time period and identified
possible candidates:

svn commit: r188755 - head/sys/dev/ata

 Remove unused variable.

svn commit: r188761 - in stable/7: lib/libc lib/libc/string
lib/libc/sys sys sys/contrib/pf sys/dev/ath/ath_hal sys/d\
ev/cxgb sys/kern sys/net sys/netinet sys/netinet6 sys/sys

r188144:
    Standardize the various prison_foo_ip[46] functions and prison_if to
    return zero on success and an error code otherwise.  The possible errors
    are EADDRNOTAVAIL if an address being checked for doesn't match the
    prison, and EAFNOSUPPORT if the prison doesn't have any addresses in
    that address family.  For most callers of these functions, use the
    returned error code instead of e.g. a hard-coded EADDRNOTAVAIL or
    EINVAL.


svn commit: r188763 - head/sys/dev/ata
 Make ch->dma.free() called symmetrically to ch->dma.alloc().

svn commit: r188765 - in head/sys/dev/ata: . chipsets

Log:
 As soon as they called in only same one place (ata_pcichannel_attach()),
 join allocate() and dmainit() atapci subdriver's channel initialization
 methods into single ch_attach() method.



 As opposite to ch_attach() add new ch_detach() method to deallocate/disable
 channel.

svn commit: r188743 - head/sys/dev/aac
Log:
 Use outbound message register 0 instead of mailbox 7 in
 aac_{rx,rkt}_get_fwstatus, as done in Adaptec's vendor driver as well as
 the Linux drivers.

 Submitted by:   jkim, from Adaptec's driver