RE: Re: [lkcd-devel] Re: Re: lcrash and debian

> -----Original Message-----
> From: [email protected]
> [mailto:[email protected]]On Behalf Of Corey
> Minyard
> Sent: Thursday, October 07, 2004 6:00 AM
> To: Rainer Bawidamann
> Cc: [email protected];=20
> [email protected]
> Subject: [lkcd-general] Re: [lkcd-devel] Re: Re: lcrash and debian
>=20
>=20
> Rainer Bawidamann wrote:
>=20
> >Hi Corey!
> >
> >In article <[email protected]>,
> >	Corey Minyard <[email protected]> writes:
> > =20
> >
> >>Plus, I have cross support for crash, which is also an absolute=20
> >>necessity for us since we support host development on different=20
> >>platforms than the target environment.
> >>
> >>If I was to support this, I would maintain the one that does=20
> >>cross-development support, BTW.  I think there are enough people=20
> >>interested in this to make it worthwhile.
> >>   =20
> >>
> >
> >I'm very interested in this. But I haven't found this in=20
> Dave Anderson's
> >crash. Is it included there? If not, is you cross-development support
> >available somewhere?
> > =20
> >
> No, it is not included there.  Since the patch is so huge, he has=20
> decided not to take it (for now).  I suspect he will change his mind=20
> when Redhat's Wind River friends want this function :).
>=20
> I can post the current version on my web site, if you like.  It is=20
> against a somewhat older version of crash.

Here at Brocade we have a cross version of crash that executes on linux =
x86 platforms (e.g. PCs) and which analyzes dumps obtained from PowerPC =
platforms (e.g our FibreChannel switches). The code came to us from =
MontaVista, who was then where we got our kernels from, but I had to =
make several fixes to it in order to get it to deal properly with =
loadable modules. This code deals with the big-endian/little-endian =
issue, PC's being little and PowerPCs being big.

One of the other complexities is dealing with the varying ways that page =
tables are organized on different PowerPC processors. On some PowerPC =
variants, PTEs are 32 bits. On others, they are 64 bits due to the 64 =
bit physical addresses that those processors implement (my understanding =
is that the same issue exists among x86 variants, those with and without =
PAE). crash has to deal with these differences if it is to make sense of =
the page tables contained in the dump files. Changing the code to deal =
with these differences at run time was something that I couldn't figure =
out how to do (and I really didn't have the time to do), so I opted to =
create two versions, one for each PTE size.

Having spent a non-trivial amount of time looking at this code, I think =
creating a cross version of crash that will properly analyze dumps from =
any other processor and any kernel configuration would be a very =
difficult problem. There are just too many variables to deal with at =
run-time.

Jeff Haran
Brocade Communications Systems

> >Is you version able to work cross-endian? I.e. debug a dump from a
> >big-endian system on a little-endian system?
> > =20
> >
> Yes, you can debug cross-endian and between targets with=20
> different type=20
> sizes.
>=20
> -Corey
>=20
>=20
> -------------------------------------------------------
> This SF.net email is sponsored by: IT Product Guide on=20
> ITManagersJournal
> Use IT products in your business? Tell us what you think of=20
> them. Give us
> Your Opinions, Get Free ThinkGeek Gift Certificates! Click to=20
> find out more
> http://productguide.itmanagersjournal.com/guidepromo.tmpl
> _______________________________________________
> Lkcd-general mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/lkcd-general
>=20
>=20


-------------------------------------------------------
This SF.net email is sponsored by: IT Product Guide on ITManagersJournal
Use IT products in your business? Tell us what you think of them. Give us
Your Opinions, Get Free ThinkGeek Gift Certificates! Click to find out more
http://productguide.itmanagersjournal.com/guidepromo.tmpl