Kernel Boot Option Question
Joseph Salisbury
joseph.salisbury at canonical.com
Mon Feb 28 23:56:01 UTC 2011
On 02/25/2011 10:27 PM, Ben Hutchings wrote:
> On Fri, 2011-02-25 at 17:42 -0500, Joseph Salisbury wrote:
>> On 02/25/2011 05:23 PM, Herton Ronaldo Krzesinski wrote:
>>> On Fri, Feb 25, 2011 at 04:25:56PM -0500, Joseph Salisbury wrote:
>>>> Hello,
>>>>
>>>> There seems to be an issue when a motherboards CMOS clock reports a time
>>>> and date older than the one that was recorded during the last successful
>>>> boot. When this happens the system will fail to boot.
>>>>
>>>> Is there a way to get the kernel to ignore a bad date and time from the
>>>> BIOS at boot up, or an option to disable this check entirely?
>>>
>>> Well, just the fact of CMOS clock going backwards causing a subsequent
>>> boot to fail is odd, it shouldn't happen and I think it's not the real
>>> cause, may be a side effect of another issue.
>>>
>>> Do you have more details about the issue and machine where this happens,
>>> environment, etc.?
>>
>> It happens on all the machines I've tried. The one I have next to me is
>> a netbook running Natty, but this also happens on Maverick and Lucid.
>> It appears to be related to fsck. I changed the date in my bios to be a
>> month ahead, it boots once, but then a reboot triggers the fsck and the
>> system then hangs. Attached is a screen shot from a system having this
>> issue.
>>
>> The easy solution is to ensure the bios clock is set properly. I was
>> wondering if there is a boot option to disable the kernel from checking
>> the bios clock, but maybe this needs to be done for specific reasons.
>
> This looks like a regression in e2fsck. It never used to treat this
> sort of time anomaly as requiring manual attention.
>
> (Also, it's crazy that fsck is not called with the -y option by default.
> Only a tiny minority of users know enough to make better choices than
> just saying yes to everything.)
>
> Ben.
>
It looks like there is an upstart job /etc/init/hwclock.conf. It
basically runs the hwclock command after sourcing: /etc/default/rcS.
Looking in /etc/default/rcS, there is an option named "FSCKFIX", which
seems to be defaulted to no. The rcS man page states: "fsck is invoked
with the -a option which means "autorepair" when this option is set to
yes. That might be a way around the hang in fsck.
I found that the following option was available in /etc/default/rcS on
prior releases of Ubuntu: HWCLOCKACCESS. However, I don't see it on
Lucid, Maverick or Natty. I haven't looked at the code to know if this
option is gone, hidden, or not needed now due to changes.
Thanks,
Joe
More information about the kernel-team
mailing list