[Bug tdep/31442] New: [gdb/tdep, arm, SIGILL] FAIL: gdb.threads/sigstep-threads.exp: step 92

[Bug tdep/31442] New: [gdb/tdep, arm, SIGILL] FAIL: gdb.threads/sigstep-threads.exp: step 92

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=31442

            Bug ID: 31442
           Summary: [gdb/tdep, arm, SIGILL] FAIL:
                    gdb.threads/sigstep-threads.exp: step 92
           Product: gdb
           Version: HEAD
            Status: NEW
          Severity: normal
          Priority: P2
         Component: tdep
          Assignee: unassigned at sourceware dot org
          Reporter: vries at gcc dot gnu.org
  Target Milestone: ---

On arm-linux, I ran into:
...
(gdb) step^M
^M
Thread 2 "sigstep-threads" received signal SIGUSR1, User defined signal 1.^M
^M
Thread 2 "sigstep-threads" received signal SIGILL, Illegal instruction.^M
[Switching to Thread 0xf76a5420 (LWP 2340)]^M
0xaaaaa5e4 in handler (signo=10) at
/home/rock/gdb/src/gdb/testsuite/gdb.threads/sigstep-threads.c:32^M
32        var++;                /* step-1 */^M
(gdb) FAIL: gdb.threads/sigstep-threads.exp: step 92
...

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug tdep/31442] [gdb/tdep, arm, SIGILL] FAIL: gdb.threads/sigstep-threads.exp: step 92

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=31442

--- Comment #1 from Tom de Vries <vries at gcc dot gnu.org> ---
This could be related to kernel commits:
...
commit 5f41f9198f296091c6a58bc2e86af1e9f019b2a3
Author: Marek Szyprowski <m.szyprowski@samsung.com>
Date:   Tue May 28 09:38:14 2019 +0100

    ARM: 8864/1: Add workaround for I-Cache line size mismatch between CPU
cores
...
and:
...
commit 116c81f427ff6c5380850963e3fb8798cc821d2b
Author: Suzuki K Poulose <suzuki.poulose@arm.com>
Date:   Fri Sep 9 14:07:16 2016 +0100

    arm64: Work around systems with mismatched cache line sizes
...

I'm using 32-bit arm user land on 64-bit kernel on aarch64, so I'm not sure
which commit would apply.  The arm commit first appears in v5.3, the arm64 one
in v4.9.

The kernel used is 4.4.190.

This is because the pinebook came with
https://github.com/mrfixit2001/debian_desktop , and the update script never
work for me, so I hacked the user land to be plain debian, but I got stuck with
the kernel from this (
https://github.com/mrfixit2001/debian_desktop/releases/tag/190905 ) release.

So, it could be that I don't have the necessary fix for this, I'm not sure how
to check.

I confirmed that this is a big.little issue, I ran with taskset -c 0-3 and
taskset -c 4-5 and the issue no longer reproduces.

I looked into upgrading the kernel to 6.1, but it seems that this is difficult
because /boot is read-only due to being flash, so I'm avoiding that for now.

I'm going to work around this by running tests using taskset -c 4-5,
restricting testing to the 2 big cores.

-- 
You are receiving this mail because:
You are on the CC list for the bug.

[Bug external/31442] [gdb/external, arm, SIGILL] FAIL: gdb.threads/sigstep-threads.exp: step 92

[vries at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=31442

Tom de Vries <vries at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |MOVED
             Status|NEW                         |RESOLVED
            Summary|[gdb/tdep, arm, SIGILL]     |[gdb/external, arm, SIGILL]
                   |FAIL:                       |FAIL:
                   |gdb.threads/sigstep-threads |gdb.threads/sigstep-threads
                   |.exp: step 92               |.exp: step 92
          Component|tdep                        |external

--- Comment #2 from Tom de Vries <vries at gcc dot gnu.org> ---
(In reply to Tom de Vries from comment #1)
> I looked into upgrading the kernel to 6.1, but it seems that this is
> difficult because /boot is read-only due to being flash, so I'm avoiding
> that for now.

I gave it a try, and ... was left with an non-booting system.

I managed to recover by booting from an sd card (which involved opening the
laptop to recover the micro sd card that slid next to rather than into the
slot).

Then I managed to overwrite the broken installation on the hard-disk and chose
the manjaro distro, because it's supposedly the best supported distro for
pinebook pro.  That got me a 6.1.x kernel, which after upgrading got me a 6.6.7
kernel. 

The manjaro install also got me a 64-bit kernel + 64-bit user land.

I managed to build a chroot with 32-bit debian, which is about as close as I
can get to the old 64-bit kernel + 32-bit userland setup. [ I used arch-chroot,
hand-edited to unset pid_unshare and mount_unshare, otherwise we run into
warnings due to not finding /proc/PID/mem. ]

I ran this test-case a 100 times, and the problem doesn't reproduce.

-- 
You are receiving this mail because:
You are on the CC list for the bug.