Re: Kernel crash before login prompt

Hello Andrew,

The addr2line for the address points to,


[sfaisal@svl-bsdx-02 /b/sfaisal/stable_main/src]$ addr2line -e ../obj/arm64/juniper/kernels/JNPR-ARM64-WIT/kernel/kernel.full -f 0xffff0000002244dc

sched_add

/b/sfaisal/stable_main/src/sys/kern/sched_ule.c:1309


Sched_setcpu is the exact function which calls TDQ_LOCK(__mtx_lock_spin_flags) , and sched_setcpu is an inline called from sched_add()

Regards
Faisal



Juniper Business Use Only

From: [email protected] <[email protected]>
Date: Thursday, 29 June 2023 at 20:17
To: Sha Faisal <[email protected]>
Cc: [email protected] <[email protected]>, Steve Kiernan <[email protected]>, Mahesh Vardhamanaiah <[email protected]>
Subject: Re: Kernel crash before login prompt
[External Email. Be cautious of content]


Can you find out which function 0xffff0000002244dc is from (i.e. the
caller of __mtx_lock_spin_flags)? It looks like it is in sched_pickcpu,
but it also looks like sched_pickcpu should return before calling
__mtx_lock_spin_flags based on the boot log.

Andrew

On 2023-06-29 14:26, Sha Faisal wrote:
> Below are the values you asked for
>
> Just after dpcpu_init() in initarm()
>
> --------In initarm and pcpup->pc_dynamic = 0x2886480
>
> --------In initarm and dpcpu_off[0] = 0x2886480
>
> --------In initarm and DPCPU_START = 0xffff000000681b80
>
> --------In initarm and DPCPU_STOP = 0xffff000000687300
>
> --------In initarm and DPCPU_BYTES = 0x5780
>
> --------In initarm and DPCPU_SIZE = 0x6000
>
> --------In initarm and PAGE_SIZE = 0x1000
>
> --------In initarm and pcpu = 0xffff00000095bd80
>
> --------In initarm and pcpu_entry_timerstate is 0xffff00000068ac80----
>
>
> --------In initarm and __start_set_pcpu is 0----
>
> --------In initarm and pcpu_entry_tdq is 0xffff00000068ac80----
>
> ---<<BOOT>>---
>
> Below are after it crashes on db prompt:
>
> db> p __start_set_pcpu
> ffff000000681b80
> db> p __stop_set_pcpu
> ffff000000687300
> db>  show dpcpu_off
> dpcpu_off[ 0] = 0x2886480 (+ DPCPU_START = 0xffff000002f08000)
>
> db> p pcpu_entry_tdq
>
> ffff000000681c80
>
> Regards
>
> Faisal
>
>  Juniper Business Use Only
>  From: [email protected] <[email protected]>
> on behalf of Andrew Turner <[email protected]>
> Date: Thursday, 29 June 2023 at 15:27
> To: Sha Faisal <[email protected]>
> Cc: [email protected] <[email protected]>, Steve Kiernan
> <[email protected]>, Mahesh Vardhamanaiah <[email protected]>
> Subject: Re: Kernel crash before login prompt
>
> [External Email. Be cautious of content]
>
> It looks like that exception is due to DPCPU_ID_PTR calculating an
> invalid pointer.
>
> Can you print the contents of the dpcpu_off array, and the addresses
> of __start_set_pcpu and pcpu_entry_tdq.
>
> Andrew
>
>> On 28 Jun 2023, at 19:45, Sha Faisal <[email protected]> wrote:
>>
>> Hello Andrew,
>>
>> Attached is the boot log,  also at the end I have pasted the remote
>> gdb backtrace
>>
>> Regards
>>
>> faisal
>>
>> Juniper Business Use Only
>>
>> From: Andrew Turner <[email protected]>
>> Date: Wednesday, 28 June 2023 at 20:47
>> To: Sha Faisal <[email protected]>
>> Cc: [email protected] <[email protected]>, Steve Kiernan
>> <[email protected]>, Mahesh Vardhamanaiah <[email protected]>
>> Subject: Re: Kernel crash before login prompt
>>
>> [External Email. Be cautious of content]
>>
>>> On 28 Jun 2023, at 13:18, Sha Faisal <[email protected]> wrote:
>>>
>>> Hello FreeBSD team,
>>>
>>> We have upgraded our FreeBSD kernel from 12 to main branch and
>>> with latest main branch one of our boxes with
>>>
>>> Broadcom iProc Northstar2 BCM5871X SoC (Quad-core ARM Cortex-A57
>>> 64-bit 2GHz ARMv8-A CPU architecture) is crashing at
>>> mtx_spin_lock_flags(), with data_abort exception.
>>>
>>> Sometimes the exception is raised at DPCPCU_PTR() functions.
>>>
>>> Has anyone tried to boot freeBSD14 on the above mentioned BCM SoC?
>>>
>>>
>>> Please share your inputs and comment
>>
>> I don’t know of any issues with either of these, and don’t know
>> of anyone running FreeBSD on this SoC.
>>
>> Can you provide more details on what is failing, e.g. the boot log?
>>
>> Andrew