Re: Analysis of s390 on x86_64

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, 20 May 2019 11:28:30 +0300
Nikolay Borisov <nborisov@xxxxxxxx> wrote:

> On 20.05.19 г. 11:24 ч., Honglei Wang wrote:
> > Hi Nikolay,
> > 
> > I don't think we can do the s390 analysis just use the x86_64 crash
> > utility directly. I'm not familiar with cross build, but seems this
> > linkage might help:
> > 
> > https://developer.ibm.com/linuxonpower/2017/07/06/analyzing-openpower-kernel-dump-information-laptop/  
> 
> This is exactly which got me started into trying to understand what it
> takes to analyse s390 on x86_64. . Hence I've referenced the function
> which is supposed to verify that a particular combination of host and
> target is supported. I'm now interested in understanding why s390 on
> x86_64 is not supported.

Heh, been there already... ;-)

That's for many reasons. Bernhard Walle tried to make a crosscrash
package many years ago. The necessary changes were quite intrusive, and
Dave wrote clearly that this was unlikely to get merged any time soon.

I still thought it was a good idea, so as a Hackweek project I started
writing a tool that can automatically convert the crash utility to a
cross-platform variant. The (abandoned) sources can be found at
GitHub[1]. IIRC this work was based on crash-6.0.5, it most likely fails
miserably with any other version. ;-)

Anyway, the main issue is that crash does not read individual data
fields using gdb primitives. Instead, it often reads a whole structure
and then accesses the fields directly. There are also too many global
variables, sometimes their type depends on the target, etc. etc.

But if you're willing to open this can of worms and clean it up, I will
do my best to support you. ;-)

Petr T

> > Honglei
> > 
> > 
> > On 2019/5/20 15:18, Nikolay Borisov wrote:  
> >> Hello Dave,
> >>
> >> I'd like to better understand why s390-on-x86_64 analysis is not
> >> supported e.g. respective pair in get_current_configuration is
> >> missing? Is it due to the endianness mismatch ? gdb does support
> >> setting the endianness of a target? Or are there other problems?
> >>
> >> -- 
> >> Crash-utility mailing list
> >> Crash-utility@xxxxxxxxxx
> >> https://www.redhat.com/mailman/listinfo/crash-utility  
> >   
> 
> --
> Crash-utility mailing list
> Crash-utility@xxxxxxxxxx
> https://www.redhat.com/mailman/listinfo/crash-utility

Attachment: pgpXq4odbk9SX.pgp
Description: Digitální podpis OpenPGP

--
Crash-utility mailing list
Crash-utility@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/crash-utility

[Index of Archives]     [Fedora Development]     [Fedora Desktop]     [Fedora SELinux]     [Yosemite News]     [KDE Users]     [Fedora Tools]

 

Powered by Linux