Add results of changing the CSR

It seems to arrive to systemd with all-in-order, but hangs there.

JOURNAL.md

@@ -886,7 +886,83 @@ Continues to hang just after those perl messages:
 May be a long shot, but if we are experiencing the same page fault loop as in
 cincoranch we may as well try.
 
+## 2024-07-09
+
 ### QUESTION: Maybe we can try without out-of-order?
 
 I made a small tool in C to view and change the CSR register that controls the
 in-order/out-of-order. Maybe we can try with the "in-order" setting.
+
+We arrive to execute `systemd`:
+
+
+    Starting interactive shell...
+    + setsid /nix/store/xm3mpj9aldz5r4s5yb7p08jdjv98hj4w-extra-utils/bin/ash -c 'exec /nix/store/xm3mpj9aldz5r4s5yb7p08jdjv98hj4w-extra-utils/bin/ash < /dev/hvc0 >/dev/hvc0 2>/dev/hvc0'
+    [   90.077300] stage-1-init: [Thu Jan  1 00:01:27 UTC 1970] + '[' -n 1 -a i '=' f ]
+    [   90.639760] stage-1-init: [Thu Jan  1 00:01:28 UTC 1970] + '[' -n 1 -a i '=' i ]
+    ~ # [   90.967260] stage-1-init: [Thu Jan  1 00:01:28 UTC 1970] + echo 'Starting interactive shell...'
+    [   91.234980] stage-1-init: [Thu Jan  1 00:01:28 UTC 1970] Starting interactive shell...
+    [   91.569580] stage-1-init: [Thu Jan  1 00:01:29 UTC 1970] + setsid /nix/store/xm3mpj9aldz5r4s5yb7p08jdjv98hj4w-extra-utils/bin/ash -c 'exec /nix/store/xm3mpj9aldz5r4s5yb7p08jdjv98hj4w-extr
+    a-utils/bin/ash < /dev/hvc0 >/dev/hvc0 2>/dev/hvc0'
+    which csrtool
+    /nix/store/xm3mpj9aldz5r4s5yb7p08jdjv98hj4w-extra-utils/bin/csrtool
+    ~ # csrtool
+    CSR 0x801 = 0u
+    ~ # csrtool o
+    unknown 'o', use: mem-in-order, all-in-order or all-out-of-order
+    ~ # csrtool all-in-order
+    CSR 0x801 = 7u
+    ~ # csrtool
+    CSR 0x801 = 7u
+    ~ #
+    + IFS='='
+    + echo init /nix/store/xmagm60y90pfh3yvqanvmaswa0m3cb0a-nixos-system-nixos-riscv-24.11pre-git/init
+    + set -- init /nix/store/xmagm60y90pfh3yvqanvmaswa0m3cb0a-nixos-system-nixos-riscv-24.11pre-git/init
+    + stage2Init=/nix/store/xmagm60y90pfh3yvqanvmaswa0m3cb0a-nixos-system-nixos-riscv-24.11pre-git/init
+    + echo /nix/store/xm3mpj9aldz5r4s5yb7p08jdjv98hj4w-extra-utils/bin/modprobe
+    + basename dm_mod
+    [...]
+    + echo /sbin/modprobe
+    + '[' '!' -e /mnt-root//nix/store/xmagm60y90pfh3yvqanvmaswa0m3cb0a-nixos-system-nixos-riscv-24.11pre-git/init ]
+    + mkdir -m 0755 -p /mnt-root/proc /mnt-root/sys /mnt-root/dev /mnt-root/run
+    + mount --move /proc /mnt-root/proc
+    + mount --move /sys /mnt-root/sys
+    + mount --move /dev /mnt-root/dev
+    + mount --move /run /mnt-root/run
+    + type -P switch_root
+    + exec env -i /nix/store/xm3mpj9aldz5r4s5yb7p08jdjv98hj4w-extra-utils/bin/switch_root /mnt-root /nix/store/xmagm60y90pfh3yvqanvmaswa0m3cb0a-nixos-system-nixos-riscv-24.11pre-git/init
+    
+    <<< NixOS Stage 2 >>>
+    
+    [  967.703320] EXT4-fs (pmem0p2): re-mounted 44444444-4444-4444-8888-888888888888 r/w. Quota mode: none.
+    [  967.928020] booting system configuration /nix/store/xmagm60y90pfh3yvqanvmaswa0m3cb0a-nixos-system-nixos-riscv-24.11pre-git
+    running activation script...
+    [  977.608980] stage-2-init: running activation script...
+    bbbbbbbbbbbbbbbbbbbbsetting up /etc...
+    [ 1084.376420] stage-2-init: setting up /etc...
+    starting systemd...
+
+Not sure if it is a good reproducer, as it is taking around 15 minutes to hang
+in a very large piece of software, while when we have the out-of-order enabled,
+we can hang in half of the time in some script.
+
+Either way, we need to see a backtrace of where it is hanging to understand why
+it does. We may also enable a stage-2 heartbeat to be sure that it is hanging
+the kernel and not only the console.
+
+Another idea is to arrive at a proper bash shell, where we can have debugging
+tools, which may allow us to go slowly until we catch the bug.
+
+### OBSERVATION: Setting memory in-order only causes a hang
+
+Tested with:
+
+    $ csrtool mem-in-order
+
+And it hangs just after exiting the tool.
+
+### QUESTION: Can we see the printk buffer from the host?
+
+If the problem that we are observing is somehow related to the recursive
+segfault of the kernel in Cincoranch, we may be able to see the printk ring
+buffer by directly poking at the memory from the host.