On 8/13/23 10:38, Helge Deller wrote:
On 8/13/23 10:09, Sam James wrote:
Upgrading from 6.1.37 or so to 6.4.10, I get:
```
[...]
devtmpfs: mounted
Freeing unused kernel image (initmem) memory: 656K
Write protected read-only-after-init data: 41k
Run /sbin/init as init process
process 'usr/lib/systemd/systemd' started with executable stack
_______________________________
< Your System ate a SPARC! Gah! >
-------------------------------
\ ^__^
(__)\ )\/\
U ||----w |
|| ||
init (pid 1): Spinlock was trashed (code 1)
CPU: 1 PID: 1 Comm: init Not tainted 6.4.10 #1
Hardware name: 9000/800/rp3440
YZrvWESTHLNXBCVMcbcbcbcbOGFRQPDI
PSW: 00001000000001001111111100001111 Not tainted
r00-03 000000ff0804ff0f 000000004b4e8de0 00000000404021d8 000000004b4e8e80
r04-07 0000000040f4bda0 000000004b64c000 000000004b699800 000000004d1ad000
r08-11 000000004b699864 000000000c574000 0000000000000000 0000000040001e1c
r12-15 0000000000001000 0000000040ed5900 0000000000000001 0000000000000000
r16-19 000003fff0fff000 000000004101d5a0 0000000040f8d5a0 0000000000001a46
r20-23 0000000000000000 0000000000000000 0000000000000000 0000000000000080
r24-27 0000000000000000 0000000000000000 000000004b699864 0000000040f4bda0
r28-31 00000000000ce800 000000004b4e8e50 000000004b4e8ec0 00000000000ce800
sr00-03 0000000000000000 0000000000000000 0000000000000000 00000000000ce800
sr04-07 0000000000000000 0000000000000000 0000000000000000 0000000000000000
IASQ: 0000000000000000 0000000000000000 IAOQ: 0000000040e8466c 0000000040e84670
IIR: 0000c006 ISR: 0000000010240000 IOR: 0000002da6699864
CPU: 1 CR30: 000000004b46a010 CR31: ffffffffffffffff
ORIG_R28: 0000000000000000
IAOQ[0]: _raw_spin_lock+0x1c/0x58
IAOQ[1]: _raw_spin_lock+0x20/0x58
RP(r2): __pmd_alloc+0xb8/0x1b8
Backtrace:
[<00000000404021d8>] __pmd_alloc+0xb8/0x1b8
[<00000000404165c8>] move_page_tables.part.0+0x788/0x7a0
[<0000000040417630>] move_page_tables+0x38/0x50
[<000000004047f82c>] shift_arg_pages+0x12c/0x2f0
[<000000004047fc28>] setup_arg_pages+0x238/0x390
[<000000004051b538>] load_elf_binary+0x660/0x1b38
[<0000000040480f54>] bprm_execve+0x41c/0x8c8
[<000000004048323c>] kernel_execve+0x20c/0x2b8
[<0000000040e63924>] run_init_process+0x164/0x198
[<0000000040e63990>] try_to_run_init_process+0x38/0xa0
[<0000000040e81dd0>] kernel_init+0x290/0x340
[<00000000401a6020>] ret_from_kernel_thread+0x20/0x28
CPU: 1 PID: 1 Comm: init Not tainted 6.4.10 #1
Hardware name: 9000/800/rp3440
Backtrace:
[<00000000401ae4a8>] show_stack+0x70/0x90
[<0000000040e80d88>] dump_stack_lvl+0xd8/0x128
[<0000000040e80e0c>] dump_stack+0x34/0x48
[<00000000401ae6b8>] die_if_kernel+0x1d0/0x388
[<00000000401af6c8>] handle_interruption+0xbe0/0xcd8
[<00000000401a707c>] intr_check_sig+0x0/0x3c
Kernel panic - not syncing: Fatal exception
```
Note that I did apply
1. https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/tree/queue-6.4/io_uring-parisc-adjust-pgoff-in-io_uring-mmap-for-parisc.patch
2. https://git.kernel.org/pub/scm/linux/kernel/git/stable/stable-queue.git/tree/queue-6.4/parisc-fix-lightweight-spinlock-checks-to-not-break-futexes.patch
on top of 6.4.10, which I thought would solve any spinlock problems.
Hmm.. I thought patch 2 should have fixed it.
Do you have TLB_PTLOCK enabled? If so, maybe I missed fixing that in patch (2)? I'll check..
I was right.
TLB_PTLOCK and LIGHTWEIGHT_SPINLOCK_CHECK currently can't be enabled at the same
time, as it will then crash as shown above.
Please disable TLB_PTLOCK for now until I can come up with a patch.
Helge
"Spinlock was trashed" is coming from CONFIG_LIGHTWEIGHT_SPINLOCK_CHECK.
Disabling that config option should at least bring your kernel up.
You might want to add this too:
https://patchwork.kernel.org/project/linux-parisc/patch/ZNep5EcYskP9HtGD@p100/
(lockdep: Fix static memory detection even more)
But I don't think it's related.
Helge
