Re: CVS commit: src/sys [freeze on boot]

[Ryo ONODERA <ryo%tetera.org@localhost>]

Ryo ONODERA <ryo%tetera.org@localhost> writes:

> Ryo ONODERA <ryo%tetera.org@localhost> writes:
>
>> Hi,
>>
>> Andrew Doran <ad%netbsd.org@localhost> writes:
>>
>>> Hi,
>>>
>>> This also happened the last time I touched rw_downgrade(), and I backed out
>>> the change then, but both times I don't see the bug.  I have some questions:
>>>
>>> - Are you running DIAGNOSTIC and/or LOCKDEBUG?  I would be very interested
>>>   to see what happens with a LOCKDEBUG kernel here.
>>
>> I will enable LOCKDEBUG and DIAGNOSTIC soon.
>
> Sadly DIAGNOSTIC and LOCKDEBUG with i915drmkms(4) makes my LCD black
> and I cannot see any messages.
> When i915drmkms is disabled, the kernel boots without freeze in
> DIAGNOSTIC and LOCKDEBUG case.
>
> However ioctl(2) to ims(4) causes kernel panic.
> I feel that this panic is not related to the boot freeze.
> See:
> $ crash -M netbsd.10.core -N netbsd.10
> Crash version 9.99.39, image version 9.99.39.
> System panicked: kernel diagnostic assertion "ci->ci_mtx_count == -1" failed: file "/usr/src/sys/kern/kern_synch.c", line 676 mi_switch: cpu0: ci_mtx_count (-2) != -1 (block with spin-mutex held)
> Backtrace from time of crash is available.
> crash> bt
> _KERNEL_OPT_NARCNET() at 0
> _KERNEL_OPT_ACPI_SCANPCI() at _KERNEL_OPT_ACPI_SCANPCI
> sys_reboot() at sys_reboot
> vpanic() at vpanic+0x181
> kern_assert() at kern_assert+0x48
> mi_switch() at mi_switch+0x9b8
> sleepq_block() at sleepq_block+0x1cb
> turnstile_block() at turnstile_block+0x5bd
> mutex_enter() at mutex_enter+0x31d
> iic_acquire_bus() at iic_acquire_bus+0x2a
> ihidev_softintr() at ihidev_softintr+0x27
> softint_dispatch() at softint_dispatch+0xdb
> DDB lost frame for Xsoftintr+0x4f, trying 0xffffda8138ef00f0
> Xsoftintr() at Xsoftintr+0x4f
> --- interrupt ---
> 1c6f7f77b9463525:
> crash>

The black screen and ims(4) panic are not related to your change.
Older src tree with LOCKDEBUG reproduces these problem.

Thank you.

>>> - Do you have an ATI Radeon graphics chip?
>>> - Are you using ZFS?
>>
>> My GPU is in Intel CPU (KabyLake Refresh).
>> And I do not use ZFS at all. All partitions are FFSv2 with WAPBL.
>>
>>> Thanks,
>>> Andrew
>>>
>>>
>>> On Mon, Jan 20, 2020 at 12:41:37PM +0900, Ryo ONODERA wrote:
>>>> Hi,
>>>> 
>>>> After this commit, the kernel stalls just before root file system
>>>> will be found on my NetBSD/amd64 laptop.
>>>> 
>>>> Reverting
>>>> src/sys/kern/kern_rwlock.c to r1.60
>>>> and
>>>> src/sys/sys/rwlock.h to r1.12
>>>> in latest -current tree and I can get the kernel that works like
>>>> before.
>>>> 
>>>> And on another laptop, the problematic kernel stalls before root file
>>>> system detection like my laptop.
>>>> 
>>>> It may be universal problem.
>>>> 
>>>> Could you take a look at this problem?
>>>> 
>>>> Thank you.
>>>> 
-- 
Ryo ONODERA // ryo%tetera.org@localhost
PGP fingerprint = 82A2 DC91 76E0 A10A 8ABB  FD1B F404 27FA C7D1 15F3