[Crash-utility] [PATCH 0/5] crash: Small bundle of fixes for Xen

[Daniel Kiper]

On Thu, Jul 12, 2012 at 03:47:57PM -0400, Dave Anderson wrote:
>
>
> ----- Original Message -----
> >
> >
> > ----- Original Message -----
> > >
> > >
> > > ----- Original Message -----
> > > > On Thu, Jul 05, 2012 at 05:29:50PM +0200, Daniel Kiper wrote:
> > > > > On Thu, Jul 05, 2012 at 11:15:29AM -0400, Dave Anderson wrote:
> > > > > >
> > > > > >
> > > > > > ----- Original Message -----
> > > > > > > Hi,
> > > > > > >
> > > > > > > It looks that Xen support for crash have not been maintained
> > > > > > > since 2009. I am trying to fix this. Here it is small
> > > > > > > bundle of fixes:
> > > > > > >   - xen: Always calculate max_cpus value,
> > > > > > >   - xen: Read only crash notes for onlined CPUs,
> > > > > > >   - x86/xen: Read variables from dynamically allocated per_cpu data,
> > > > > > >   - xen: Get idle data from alternative source,
> > > > > > >   - xen: Read data correctly from dynamically allocated console ring too.
> > > > > > >
> > > > > > > Daniel
> > > > > >
> > > > > > Daniel,
> > > > > >
> > > > > > Can you absolutely confirm that these changes are all
> > > > > > backwards-compatible?
> > > > >
> > > > > I have not done tests with older versions of Xen. However,
> > > > > I made all patches with backward compatibility in mind. Anyway,
> > > > > If you wish I could do tests with let's say Xen Ver. 3.4
> > > > > and Xen Linux Kernel Ver. 2.6.18 at least.
> > > >
> > > > I think Dave is worried about it breaking with the version that is bundled
> > > > with RHEL5U8?
> > >
> > > Correct...
> > >
> > > Dave
> > >
> >
> > And as it turns out, it does break RHEL5 xen, at least the changes
> > you made for the log command:
> >
> >   crash> log
> >   log: invalid kernel virtual address: 616e69625f64616f  type:
> >   "conring contents"
> >   crash>
> >
>
> Daniel,
>
> Your patch breaks the RHEL5 xen hypervisor log command because "conring" used
> to be declared as a statically-sized array:
>
>   crash> whatis conring
>   char conring[16384];
>   crash>
>
> It appears from your patch that "conring" is now declared as a pointer,
> given that your patch does this:
>
> --- crash-6.0.8.orig/xen_hyper_command.c        2012-06-29 16:59:18.000000000 +0200
> +++ crash-6.0.8/xen_hyper_command.c     2012-07-05 16:22:35.000000000 +0200
> @@ -590,24 +590,31 @@ xen_hyper_dump_log(void)
>         ulong conring;
>         char *buf;
>         char last;
> +       uint32_t conring_size;
>
> -       conring = symbol_value("conring");
> +       get_symbol_data("conring", sizeof(ulong), &conring);
>         get_symbol_data("conringc", sizeof(uint), &conringc);
>         get_symbol_data("conringp", sizeof(uint), &conringp);
> ...
>
> If I change your patch such that the end result looks like this:
>
>         if (get_symbol_type("conring", NULL, NULL) == TYPE_CODE_ARRAY)
>                 conring = symbol_value("conring");
>         else
>                 get_symbol_data("conring", sizeof(ulong), &conring);
>
> it works for RHEL5 xen hypervisor.
>
> Does that work for you?

Thanks. It works. I though that there is an issue with conring symbol too.
Have you done other tests on RHEL5?

I discoverd additional crash issue with Xen core dumps. I am working
on fix for that. I would like to add relevant patch for next release
of this bundle of patches. I am going to do all tests for new release
(probably next week). Is it OK for you?

Daniel