--- Comment #2 from Christophe Leroy (christophe.leroy@c-s.fr) ---
Could you please provide the content of /sys/kernel/debug/kernel_page_tables
together with the dmesg, and also an 'objdump -h' of the failing modules ?
--
You are receiving this mail because:
You are watching the assignee of the bug.
https://bugzilla.kernel.org/show_bug.cgi?id=204819
Erhard F. (erhard_f@mailbox.org) changed:
What |Removed |Added
----------------------------------------------------------------------------
Attachment #284931|0 |1
is obsolete| |
--- Comment #3 from Erhard F. (erhard_f@mailbox.org) ---
--- Comment #8 from Christophe Leroy (christophe.leroy@c-s.fr) ---
You get:
Sep 13 17:43:49 T600 kernel: BUG: Unable to handle kernel data access at
0xfe205150
However this area is properly mapped RW:
0xfe205000-0xfe205fff 0x297d6000 4K rw present
dirty accessed
There must be something wrong with hash table update.
--
You are receiving this mail because:
You are watching the assignee of the bug.
--- Comment #9 from Erhard F. (erhard_f@mailbox.org) ---
I can apply the patch but how do I apply the manually fixed kernel_page_tables?
Not quite sure how to proceed.
--
You are receiving this mail because:
You are watching the assignee of the bug.
--- Comment #10 from Christophe Leroy (christophe.leroy@c-s.fr) ---
No no, what I mean by 'manually fixed kernel_page_tables' is that I fixed the
file you attached so that it has the same values as what it would have had with
the patch applied. Don't worry.
--
You are receiving this mail because:
You are watching the assignee of the bug.
--- Comment #11 from Christophe Leroy (christophe.leroy@c-s.fr) ---
I have carefully reviewed the flushing calls and have not been able to identify
any issue.
Maybe a SMP issue ? Does this problem also happen without CONFIG_SMP ?
--
You are receiving this mail because:
You are watching the assignee of the bug.
--- Comment #22 from Erhard F. (erhard_f@mailbox.org) ---
(In reply to Christophe Leroy from comment #11)
I have carefully reviewed the flushing calls and have not been able to
identify any issue.
Maybe a SMP issue ? Does this problem also happen without CONFIG_SMP ?
For my 1st (v2) retest I applied the "powerpc/ptdump: Fix addresses display on
PPC32" patch you linked on top of -rc8.
For my 2nd (v3) retest I applied the patch and disabled SMP in the .config.
As you can see the issue still persists without SMP. Also different
'problematic' modules popup even without SMP.
--
You are receiving this mail because:
You are watching the assignee of the bug.
--- Comment #23 from Christophe Leroy (christophe.leroy@c-s.fr) ---
Thanks. The third one is interesting, as it shows :
Sep 16 12:57:39 T600 kernel: BUG: Unable to handle kernel data access at
0xfe295404
0xfe295000-0xfe295fff 0x00d05000 4K user r present
accessed
So unlike the other cases, this one shows that the area is still read-only and
pointing to the zero shadow area.
Wondering whether for the other ones the area is not allocated to a further
module loader after the failing one.
Are you able to dump kernel_page_tables just after each module load failure
before loading any additional module ?
--
You are receiving this mail because:
You are watching the assignee of the bug.
Ahh, now we are getting closer...
No, I did the testing with a vanilla 5.3-rc8 kernel assuming that
powerpc/kasan: Fix shadow area set up for modules
powerpc/kasan: Fix parallel loading of modules
were already in 5.3-rc8. Now I realize I was wrong.
In -rc8 the latest one (2019-07-31) is:
powerpc/kasan: fix early boot failure on PPC32
The other two patches are dated 2019-08-20 and not in 5.3-rc8 or the newly
released 5.3.
Sorry for the noise! I will close the bug for now and retest as soon as the 2
patches are in 5.4-rcx or 5.3.x. If it still seems valid that time I will
re-open.
--
You are receiving this mail because:
You are watching the assignee of the bug.
https://bugzilla.kernel.org/show_bug.cgi?id=204819
Michael Ellerman (michael@ellerman.id.au) changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |michael@ellerman.id.au
--- Comment #26 from Michael Ellerman (michael@ellerman.id.au) ---
If you want to test things that are not quite in mainline you can always try
the "merge" branch of the powerpc tree. That is a merge of master, next and
fixes, so should have any fixes that are in the pipeline.
--
You are receiving this mail because:
You are watching the assignee of the bug.