nosmp causes 'ata1: COMRESET failed' on vmlinuz-2.6.23.1-42.fc8

[Todd Denniston]

Todd Denniston wrote, On 11/08/2007 10:26 AM:
> Chuck Ebbert wrote, On 11/07/2007 03:32 PM:
>> On 11/07/2007 10:13 AM, Todd Denniston wrote:
>>> Folks,
>>> While attempting to get xorg working[1] in vesa and fbdev mode with the
>>> CF52 I have, I tried using the nosmp boot option and received an
>>> unexpected result on boot.  Fortunately I don't require nosmp to get the
>>> system working, but I definitely was not expecting the nosmp option to
>>> cause a failure of the ata subsystem.
>>>
>>
>> Try maxcpus=1
>>
> normal boot (no nosmp or maxcpus=1)
> grub outputs:
> kernel /vmlinuz-2.6.23.1-42.fc8 ro root=LABEL=/ rhgb quiet
> 
> grub outputs:
> kernel /vmlinuz-2.6.23.1-42.fc8 ro root=LABEL=/ maxcpus=1 rhgb quiet
> maxcpus=1 bios    result
> --------- ----    ------
> !use      smp     Boots and works normally, modulo X problems.
> !use      !smp    Boots and works normally, modulo X problems.
> use       !smp    Boots and works normally, modulo X problems.
> use       smp     Boots and works normally, modulo X problems.
> 
> grub outputs:
> kernel /vmlinuz-2.6.23.1-42.fc8 ro root=LABEL=/ nosmp rhgb quiet
> nosmp    bios    result
> -----    ----    ------
> !use     smp     Boots and works normally, modulo X problems.
> !use     !smp    Boots and works normally, modulo X problems.
> use      smp     Fails with ata1: COMRESET failed
> use      !smp    Fails with ata1: COMRESET failed
> 
> Interesting...
> `nosmp` !(exactly)= `maxcpus=1` ???
> 

Still trying to understand this from a personal perspective, I did just a 
little more research...
`nosmp` = `maxcpus=0`[1]
and I get the same COMRESET failed for maxcpus=0.

also from [1]
`nosmp` ~= `maxcpus=1 noapic`
but this gets through the boot process.

looking at [1] and [2]
`maxcpus=1 nolapic`
also gets through the boot process, just slower, and a lot slower (several 
minutes) at HAL start.

`maxcpus=1  noapic nolapic` & `noapic nolapic maxcpus=1`
also gets through the boot process, just slower, and a lot slower (several 
minutes) at udev start.

`maxcpus=0 lapic` & `lapic maxcpus=0`
gives COMRESET failed.

And just to see if defining them turns something off correctly:
`maxcpus=0 noapic nolapic` & `noapic nolapic maxcpus=0`
gives COMRESET failed.


So it seems that unlike what is said in [1] [2] and [3]
`maxcpus=0` != `maxcpus=1  noapic`

And to me it still seems that nosmp/maxcpus=0 should not have caused a fault 
with the ata bus.

[1] http://www.linuxhq.com/kernel/v2.6/23/Documentation/kernel-parameters.txt
[2] http://www.ussg.iu.edu/hypermail/linux/kernel/0708.2/0321.html
[3] http://bugzilla.kernel.org/show_bug.cgi?id=1641


-- 
Todd Denniston
Crane Division, Naval Surface Warfare Center (NSWC Crane)
Harnessing the Power of Technology for the Warfighter