> On 03/04/17 14:42, Daniel Kiper wrote:
> > On Fri, Mar 31, 2017 at 12:14:38PM +0200, Juergen Gross wrote:
> >> For kdump to work correctly it needs the physical address of
> >> vmcoreinfo_note. When running as dom0 this means the virtual address
> >> has to be translated to the related machine address.
> >>
> >> paddr_vmcoreinfo_note() is meant to do the translation via
> >> __pa_symbol() only, but being attributed "weak" it can be replaced
> >> easily in Xen case.
> >>
> >> Signed-off-by: Juergen Gross <[email protected]>
> >
> > Have you tested this patch with latest crash tool? Do dom0 and Xen
> > hypervisor analysis work without any issue (at least basic commands
> > like dmesg, bt, ps, etc.)? If yes for both you can add:
> >
> > Reviewed-by: Daniel Kiper <[email protected]>
>
> This patch isn't for dump analysis, but for dump creation. Petr has
I know that. However, it may have impact on crash analysis. So,
I would expect that you or anybody else in your behalf will do
at least minimal crash tool tests.
> verified that the dump is in the expected format. Please ask Petr
> for further details, e.g. user side modifications being necessary.
So, if Petr did relevant tests that is nice. However, then, IMO, this
patch begs Petr Tested-by.
Daniel
On Tue, 4 Apr 2017 12:42:53 -0700 (PDT)
Daniel Kiper <[email protected]> wrote:
> > On 03/04/17 14:42, Daniel Kiper wrote:
> > > On Fri, Mar 31, 2017 at 12:14:38PM +0200, Juergen Gross wrote:
> > >> For kdump to work correctly it needs the physical address of
> > >> vmcoreinfo_note. When running as dom0 this means the virtual address
> > >> has to be translated to the related machine address.
> > >>
> > >> paddr_vmcoreinfo_note() is meant to do the translation via
> > >> __pa_symbol() only, but being attributed "weak" it can be replaced
> > >> easily in Xen case.
> > >>
> > >> Signed-off-by: Juergen Gross <[email protected]>
> > >
> > > Have you tested this patch with latest crash tool? Do dom0 and Xen
> > > hypervisor analysis work without any issue (at least basic commands
> > > like dmesg, bt, ps, etc.)? If yes for both you can add:
> > >
> > > Reviewed-by: Daniel Kiper <[email protected]>
> >
> > This patch isn't for dump analysis, but for dump creation. Petr has
>
> I know that. However, it may have impact on crash analysis. So,
> I would expect that you or anybody else in your behalf will do
> at least minimal crash tool tests.
>
> > verified that the dump is in the expected format. Please ask Petr
> > for further details, e.g. user side modifications being necessary.
>
> So, if Petr did relevant tests that is nice. However, then, IMO, this
> patch begs Petr Tested-by.
Actually, I tested with this patch applied on top of kernel 4.4 (SLES
12 SP2). It matches what traditional Xen had always done, so I am quite
confident it will work with a later kernel, but to give my Tested-by,
let me first re-run the test on master, hopefully until today EOB.
HTH,
Petr T
On Wed, 5 Apr 2017 13:13:00 +0200
Petr Tesarik <[email protected]> wrote:
> On Tue, 4 Apr 2017 12:42:53 -0700 (PDT)
> Daniel Kiper <[email protected]> wrote:
>
>[...]
> > So, if Petr did relevant tests that is nice. However, then, IMO, this
> > patch begs Petr Tested-by.
>
> Actually, I tested with this patch applied on top of kernel 4.4 (SLES
> 12 SP2). It matches what traditional Xen had always done, so I am quite
> confident it will work with a later kernel, but to give my Tested-by,
> let me first re-run the test on master, hopefully until today EOB.
It took me much longer than anticipated (I had some trouble setting up
the host again), but I can confirm that the patch works as expected on
top of 4.11-rc5.
Without the patch, makedumpfile in the crash kernel complains:
/proc/vmcore doesn't contain vmcoreinfo.
With the patch applied, dumping still fails later because of an
unrelated bug in makedumpfile, but I was able to extract the kernel
message buffer with "makedumpfile --dump-dmesg". This already confirms
VMCOREINFO presence and usability.
So finally,
Tested-by: Petr Tesarik <[email protected]>
Petr Tesarik
On Fri, Apr 07, 2017 at 11:16:22AM +0200, Petr Tesarik wrote:
> On Wed, 5 Apr 2017 13:13:00 +0200
> Petr Tesarik <[email protected]> wrote:
>
> > On Tue, 4 Apr 2017 12:42:53 -0700 (PDT)
> > Daniel Kiper <[email protected]> wrote:
> >
> >[...]
> > > So, if Petr did relevant tests that is nice. However, then, IMO, this
> > > patch begs Petr Tested-by.
> >
> > Actually, I tested with this patch applied on top of kernel 4.4 (SLES
> > 12 SP2). It matches what traditional Xen had always done, so I am quite
> > confident it will work with a later kernel, but to give my Tested-by,
> > let me first re-run the test on master, hopefully until today EOB.
>
> It took me much longer than anticipated (I had some trouble setting up
> the host again), but I can confirm that the patch works as expected on
No problem. I know how it works.
> top of 4.11-rc5.
Great!
> Without the patch, makedumpfile in the crash kernel complains:
>
> /proc/vmcore doesn't contain vmcoreinfo.
Though, I would like to ask you to do crash tool tests too.
Could you do that?
> With the patch applied, dumping still fails later because of an
> unrelated bug in makedumpfile, but I was able to extract the kernel
> message buffer with "makedumpfile --dump-dmesg". This already confirms
> VMCOREINFO presence and usability.
Is it Xen specific issue or more generic one?
Daniel
On Mon, 10 Apr 2017 22:49:33 +0200
Daniel Kiper <[email protected]> wrote:
> On Fri, Apr 07, 2017 at 11:16:22AM +0200, Petr Tesarik wrote:
> > On Wed, 5 Apr 2017 13:13:00 +0200
> > Petr Tesarik <[email protected]> wrote:
> >
> > > On Tue, 4 Apr 2017 12:42:53 -0700 (PDT)
> > > Daniel Kiper <[email protected]> wrote:
> > >
> > >[...]
> > > > So, if Petr did relevant tests that is nice. However, then, IMO, this
> > > > patch begs Petr Tested-by.
> > >
> > > Actually, I tested with this patch applied on top of kernel 4.4 (SLES
> > > 12 SP2). It matches what traditional Xen had always done, so I am quite
> > > confident it will work with a later kernel, but to give my Tested-by,
> > > let me first re-run the test on master, hopefully until today EOB.
> >
> > It took me much longer than anticipated (I had some trouble setting up
> > the host again), but I can confirm that the patch works as expected on
>
> No problem. I know how it works.
>
> > top of 4.11-rc5.
>
> Great!
>
> > Without the patch, makedumpfile in the crash kernel complains:
> >
> > /proc/vmcore doesn't contain vmcoreinfo.
>
> Though, I would like to ask you to do crash tool tests too.
> Could you do that?
I'm not sure what kind of tests you want me to perform. I'm not able to
save a dump file without the patch.
Well ... I could reserve enough memory to launch a crash session
directly in the kdump environment. I'll get to you later.
> > With the patch applied, dumping still fails later because of an
> > unrelated bug in makedumpfile, but I was able to extract the kernel
> > message buffer with "makedumpfile --dump-dmesg". This already confirms
> > VMCOREINFO presence and usability.
>
> Is it Xen specific issue or more generic one?
It is Xen-specific, and I even have a patch now, about to be sent to
the kexec ML.
Petr T