Quick test on real hardware requested to test qemu

Quick test on real hardware requested to test qemu

[John Paul Adrian Glaubitz]

Hi!

Could someone install the "debfoster" command please on real hardware and test whether
it crashes like it does on qemu?

root@pacman:~# debfoster
debfoster: /usr/include/atomic_ops/sysdeps/loadstore/atomic_load.h:31: AO_load: Assertion `((size_t)addr & (sizeof(*addr) - 1)) == 0' failed.
Aborted
root@pacman:~#

Thanks,
Adrian

-- 
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - glaubitz@debian.org
`. `'   Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913

Re: Quick test on real hardware requested to test qemu

[Andreas Schwab]

On Sep 22 2020, John Paul Adrian Glaubitz wrote:

> root@pacman:~# debfoster
> debfoster: /usr/include/atomic_ops/sysdeps/loadstore/atomic_load.h:31: AO_load: Assertion `((size_t)addr & (sizeof(*addr) - 1)) == 0' failed.

That a typical bad alignment check.  It should use __alignof__, not
sizeof.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 7578 EB47 D4E5 4D69 2510  2552 DF73 E780 A9DA AEC1
"And now for something completely different."

Re: Quick test on real hardware requested to test qemu

[John Paul Adrian Glaubitz]

On 9/22/20 3:22 PM, Andreas Schwab wrote:
> On Sep 22 2020, John Paul Adrian Glaubitz wrote:
> 
>> root@pacman:~# debfoster
>> debfoster: /usr/include/atomic_ops/sysdeps/loadstore/atomic_load.h:31: AO_load: Assertion `((size_t)addr & (sizeof(*addr) - 1)) == 0' failed.
> 
> That a typical bad alignment check.  It should use __alignof__, not
> sizeof.

Aha, I was assuming something like that already but I wasn't sure since libatomic-ops wasn't
updated for a while in Debian [1].

Adrian

> [1] https://packages.qa.debian.org/liba/libatomic-ops.html

-- 
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - glaubitz@debian.org
`. `'   Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913

Re: Quick test on real hardware requested to test qemu

[John Paul Adrian Glaubitz]

On 9/22/20 3:24 PM, John Paul Adrian Glaubitz wrote:
> On 9/22/20 3:22 PM, Andreas Schwab wrote:
>> On Sep 22 2020, John Paul Adrian Glaubitz wrote:
>>
>>> root@pacman:~# debfoster
>>> debfoster: /usr/include/atomic_ops/sysdeps/loadstore/atomic_load.h:31: AO_load: Assertion `((size_t)addr & (sizeof(*addr) - 1)) == 0' failed.
>>
>> That a typical bad alignment check.  It should use __alignof__, not
>> sizeof.
> 
> Aha, I was assuming something like that already but I wasn't sure since libatomic-ops wasn't
> updated for a while in Debian [1].

What would be the correct check then in [1]?

Adrian

> [1] https://github.com/ivmai/libatomic_ops/blob/master/src/atomic_ops/sysdeps/loadstore/atomic_load.h#L31

-- 
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - glaubitz@debian.org
`. `'   Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913

Re: Quick test on real hardware requested to test qemu

[John Paul Adrian Glaubitz]

On 9/22/20 3:27 PM, John Paul Adrian Glaubitz wrote:
>>> That a typical bad alignment check.  It should use __alignof__, not
>>> sizeof.
>>
>> Aha, I was assuming something like that already but I wasn't sure since libatomic-ops wasn't
>> updated for a while in Debian [1].
> 
> What would be the correct check then in [1]?
I was just re-reading the code and realized I'm dumb. Never mind.

s/sizeof/__alignof__/ should help.

Adrian

-- 
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - glaubitz@debian.org
`. `'   Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913

Re: Quick test on real hardware requested to test qemu

[John Paul Adrian Glaubitz]

On 9/22/20 5:23 PM, John Paul Adrian Glaubitz wrote:> On 9/22/20 3:27 PM, John Paul Adrian Glaubitz wrote:
>>>> That a typical bad alignment check.  It should use __alignof__, not
>>>> sizeof.
>>>
>>> Aha, I was assuming something like that already but I wasn't sure since libatomic-ops wasn't
>>> updated for a while in Debian [1].
>>
>> What would be the correct check then in [1]?
>
> I was just re-reading the code and realized I'm dumb. Never mind.
> 
> s/sizeof/__alignof__/ should help.

Hmm, now it fails with __alignof__():

(sid-m68k-sbuild)root@epyc:/# debfoster -f
debfoster: /usr/include/atomic_ops/sysdeps/loadstore/atomic_load.h:31: AO_load: Assertion `((size_t)addr & (__alignof__(*addr) - 1)) == 0' failed.
qemu: uncaught target signal 6 (Aborted) - core dumped
Aborted
(sid-m68k-sbuild)root@epyc:/#

Adrian

-- 
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - glaubitz@debian.org
`. `'   Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913

Re: Quick test on real hardware requested to test qemu

[John Paul Adrian Glaubitz]

On 9/22/20 6:06 PM, John Paul Adrian Glaubitz wrote:
> Hmm, now it fails with __alignof__():
> 
> (sid-m68k-sbuild)root@epyc:/# debfoster -f
> debfoster: /usr/include/atomic_ops/sysdeps/loadstore/atomic_load.h:31: AO_load: Assertion `((size_t)addr & (__alignof__(*addr) - 1)) == 0' failed.
> qemu: uncaught target signal 6 (Aborted) - core dumped
> Aborted
> (sid-m68k-sbuild)root@epyc:/#

It's actually a bug in libgc and was fixed here [1].

Adrian

> [1] https://github.com/ivmai/bdwgc/commit/b55459ad47d9aaa75c9a4b0f99a3b867592ab8e0

-- 
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer - glaubitz@debian.org
`. `'   Freie Universitaet Berlin - glaubitz@physik.fu-berlin.de
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913

Re: Quick test on real hardware requested to test qemu

[Geert Uytterhoeven]

Hi Adrian,

On Tue, Sep 22, 2020 at 5:23 PM John Paul Adrian Glaubitz
<glaubitz@physik.fu-berlin.de> wrote:
> On 9/22/20 3:27 PM, John Paul Adrian Glaubitz wrote:
> >>> That a typical bad alignment check.  It should use __alignof__, not
> >>> sizeof.
> >>
> >> Aha, I was assuming something like that already but I wasn't sure since libatomic-ops wasn't
> >> updated for a while in Debian [1].
> >
> > What would be the correct check then in [1]?
> I was just re-reading the code and realized I'm dumb. Never mind.
>
> s/sizeof/__alignof__/ should help.

Are 32-bit accesses to an address of the form 4 * n + 2 atomic?

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds