Re: [GIT PULL 0/9] EFI fixes for v5.7-rc

Re: [GIT PULL 0/9] EFI fixes for v5.7-rc

[Dave Young]

Cc kexec list.
On 04/10/20 at 09:08am, Ard Biesheuvel wrote:
> On Fri, 10 Apr 2020 at 01:57, Theodore Y. Ts'o <tytso@mit.edu> wrote:
> >
> > On Thu, Apr 09, 2020 at 11:29:06PM +0200, Ard Biesheuvel wrote:
> > > > What happens is that the kexec'ed kernel immediately crashes, at which
> > > > point we drop back into the BIOS, and then it boots the Debain 4.19.0
> > > > distro kernel instead of the kernel to be tested boot.  Since we lose
> > > > the boot command line that was used from the kexec, the gce-xfstests
> > > > image retries the kexec, which fails, and the failing kexec repeats
> > > > until I manually kill the VM.
> > >
> > > Does this help at all?
> > >
> > > diff --git a/arch/x86/include/asm/efi.h b/arch/x86/include/asm/efi.h
> > > index 781170d36f50..52f8138243df 100644
> > > --- a/arch/x86/include/asm/efi.h
> > > +++ b/arch/x86/include/asm/efi.h
> > > @@ -180,6 +180,7 @@ extern void __init
> > > efi_uv1_memmap_phys_epilog(pgd_t *save_pgd);
> > >
> > >  struct efi_setup_data {
> > >         u64 fw_vendor;
> > > +       u64 __unused;
> > >         u64 tables;
> > >         u64 smbios;
> > >         u64 reserved[8];
> >
> >
> > Tested-by: Theodore Ts'o <tytso@mit.edu>
> >
> 
> OK, I'll spin a proper patch
> 
> > Yep, that fixed it.  Thanks!!
> >
> > I wonder if this structure definition should be moved something like
> > arch/x86/include/uapi/asm/efi.h so it's more obvious that the
> > structure layout is used externally to the kernel?
> >
> 
> Well, 95% of the data structures used by EFI are based on the UEFI
> spec, so the base assumption is really that we cannot make changes
> like these to begin with. But I'll add a DON'T TOUCH comment here in
> any case.
> 

The runtime cleanup looks a very good one, but I also missed that,
userspace kexec-tools will break with the efi setup_data changes. But
kexec_file_load will just work with the cleanup applied.

Ard, could you add kexec list in cc when you send the fix out?

Thanks
Dave


_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: [GIT PULL 0/9] EFI fixes for v5.7-rc

[Theodore Y. Ts'o]

On Fri, Apr 10, 2020 at 09:54:42PM +0800, Dave Young wrote:
> 
> The runtime cleanup looks a very good one, but I also missed that,
> userspace kexec-tools will break with the efi setup_data changes. But
> kexec_file_load will just work with the cleanup applied.

Hmmm, I wonder if there could be some kselftest or kunit tests that
would make it easier to pick up these sorts of regressions earlier?

      	      	     	     	      	    - Ted

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: [GIT PULL 0/9] EFI fixes for v5.7-rc

[Dave Young]

On 04/11/20 at 03:43pm, Theodore Y. Ts'o wrote:
> On Fri, Apr 10, 2020 at 09:54:42PM +0800, Dave Young wrote:
> > 
> > The runtime cleanup looks a very good one, but I also missed that,
> > userspace kexec-tools will break with the efi setup_data changes. But
> > kexec_file_load will just work with the cleanup applied.
> 
> Hmmm, I wonder if there could be some kselftest or kunit tests that
> would make it easier to pick up these sorts of regressions earlier?

I thought about that before, but did not go with any actual actions.
kexec test needs a system reboot, Kdump is even harder to test, that is
the reason I hesitated about.

But since the breakage happens here and there frequently, it is time to
try it.  I think I will play with it, but I might be slow because of
other things,  welcome to post patches if anyone is interested :)

Thanks
Dave


_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec