Re: [Xen-devel] Problems booting with 3ware Controllers

["Sven Kretzschmar" <sven.kretzschmar@xxxxxx>]

Attached is a complete very long dump, which I gzipped
before posting now (first post failed, because it was too big --
I didn't look at the size before posting, mea culpa :-/ )
This is a boot dump from a 2.4. Xen kernel and only
_one_ 3ware card, in order to simplify things.
I did let it run some time to be sure enough that it's really an infinite 
loop...
I think the line:
"scsi: device set offline - not ready or command retry failed after host reset: 
host 1 channel 0 id X lun 0"
is interesting.
Every time it is printed, the 'id X' gets incremented by one (range 1-5 in this 
dump).
I am not an SCSI expert, but shouldn't this always be the same id in this case 
?!
Did you test other SCSI controlers at the lab ? Is this only a 3ware specific 
problem ?

HTH,
  Sven


*********** REPLY SEPARATOR  ***********

On 11.08.2004 at 09:51 Christian Limpach wrote:

>On Wed, Aug 11, 2004 at 01:47:11AM +0200, Sven Kretzschmar wrote:
>> Unresolved symbols in ...../3w-xxxx.o
>> depmod:   force_evtchn_callback
>>
>> looks like a missing EXPORT_SYMBOL statement, but I am
>> not sure if this has anything to do with my problems when
>> booting with the 3ware drivers compiled into the kernel; I think
>> it hasn't ...
>
>Keir has fixed this.
>
>> OK, I attached now the output while booting with TW_DEBUG=1.
>> Caveat: It's from the 2.4. Xenolinux, because that compiled quicker :-/
>> Don't pay too much attention at that part:
>> "3w-xxxx: tw_check_bits(): No expected bits (0x13025000).
>> 3w-xxxx: tw_poll_status(): Unexpected bits."
>> ...because it also shows up when booting a vanilla 2.4.26 kernel
>> (successfully !), but without the endless loop.
>
>Are you sure it gets into an endless loop?  Because you indicated
>earlier that it does find both cards, so it should find the 2nd
>card after a long stream of tw_check_bits/tw_poll_status messages.
>Because the interesting part will be after it found both cards and
>tries to use them...  At least PCI setup und interrupt routing
>looks allright.
>
>> BTW: How can I catch the output of a normal 2.4.26 kernel via
>> serial line ? What kernel options do I have to set exactly ?
>> I tried "com1=115200,8n1 console=ttyS0" like with xenolinux,
>> but it did not work ?
>
>You need to enable the serial drivers (on by default) and there's
>a console on serial line just next to it.  The syntax for the
>command line is: console=ttyS0,115200n8
>I don't know if a com1= option is supported...
>
>> Tomorrow I will also try to boot only with one 3ware controller;
>> currently, there are 2 controllers connected to the machine.
>> And I will send you the output of a 2.6. kernel, if still necessary
>
>Please try to get the output after the endless loop or confirm that
>it doesn't find the 2nd card without TW_DEBUG set.
>
>    christian

Attachment: bootdump.txt.gz
Description: Binary data