Re: Kernel crash before login prompt

Hello Andrew,

Can we have a short discussions on the below issue.
We are not able to debug further.

Please let me know your time zone and convenient time.

Regards
Faisal




Juniper Business Use Only

From: Sha Faisal <[email protected]>
Date: Friday, 30 June 2023 at 16:45
To: [email protected] <[email protected]>, Mahesh Vardhamanaiah <[email protected]>
Cc: [email protected] <[email protected]>, Steve Kiernan <[email protected]>, Amit Verma <[email protected]>
Subject: Re: Kernel crash before login prompt
Hello Andrew,
Below is the disas of sched_pickcpu()


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

1325    {

   0xffff000000223724 <+0>:     sub     sp, sp, #0x90

   0xffff000000223728 <+4>:     stp     x29, x30, [sp, #48]

   0xffff00000022372c <+8>:     stp     x28, x27, [sp, #64]

   0xffff000000223730 <+12>:    stp     x26, x25, [sp, #80]

   0xffff000000223734 <+16>:    stp     x24, x23, [sp, #96]

   0xffff000000223738 <+20>:    stp     x22, x21, [sp, #112]

   0xffff00000022373c <+24>:    stp     x20, x19, [sp, #128]

   0xffff000000223740 <+28>:    add     x29, sp, #0x30



1326            struct cpu_group *cg, *ccg;

1327            struct td_sched *ts;

1328            struct tdq *tdq;

1329            cpuset_t *mask;

1330            int cpu, pri, r, self, intr;

1331

1332            self = PCPU_GET(cpuid);

   0xffff000000223744 <+32>:    ldr     w19, [x18, #60]

   0xffff000000223748 <+36>:    mov     w21, w1

   0xffff00000022374c <+40>:    mov     x20, x0



Also, below is the show allpcpu log


db> show allpcpu
Current CPU: 0

cpuid        = 0
dynamic pcpu = 0x2886100
curthread    = 0xffff0000008166a0: pid 0 tid 100000 critnest 1 "swapper"
curpcb       = 0xffff00000068bb40
fpcurthread  = none
idlethread   = 0xffff000041b82000: tid 100003 "idle: cpu0"
spin locks held:

cpuid        = 1
dynamic pcpu = 0x41393300
curthread    = none
curpcb       = 0
fpcurthread  = none
idlethread   = 0xffff000041b81880: tid 100004 "idle: cpu1"
spin locks held:

cpuid        = 2
dynamic pcpu = 0x4139e300
curthread    = none
curpcb       = 0
fpcurthread  = none
idlethread   = 0xffff000041b81260: tid 100005 "idle: cpu2"
spin locks held:

cpuid        = 3
dynamic pcpu = 0x413a9300
curthread    = none
curpcb       = 0
fpcurthread  = none
idlethread   = 0xffff000041b80c40: tid 100006 "idle: cpu3"
spin locks held:



Regards

Faisal


From: [email protected] <[email protected]>
Date: Friday, 30 June 2023 at 16:26
To: Mahesh Vardhamanaiah <[email protected]>
Cc: Sha Faisal <[email protected]>, [email protected] <[email protected]>, Steve Kiernan <[email protected]>, Amit Verma <[email protected]>
Subject: Re: Kernel crash before login prompt
[External Email. Be cautious of content]


On 2023-06-30 10:48, Mahesh Vardhamanaiah wrote:
> Hi Andrew,
>
> We are observing that self is getting -65536 at line 1329 in
> sched_pickcpu() before panic.

What is the disassembly around reading the cpuid? It looks like it is
getting 0xffff0000 when it should be zero.

>
> From the code flow
>
> sched_add [1](struct thread [2] *td [3], int flags [4])->
> sched_pickcpu()->PCPU_GET
>
> as below
>
> static int
>
>  1321 [5] sched_pickcpu [6](struct thread [2] *td [3], int flags [4])
>
>  1322 [7] {
>
>  1323 [8]         struct cpu_group [9] *cg [10], *ccg;
>
>  1324 [11]         struct td_sched [12] *ts [13];
>
>  1325 [14]         struct tdq [15] *tdq [15];
>
>  1326 [16]         cpuset_t [17] *mask [18];
>
>  1327 [19]         int cpu [20], pri, r [21], self [22], intr;
>
>  1328 [23]
>
>  1329 [24]         self [22] = PCPU_GET [25](cpuid [26]);
>
> PCPU_GET is defined differently in Freebsd14 compared to Freebsd12 but
> I am  unable to dump the pcpup in db> prompt is it proper ?
>
> Is it expected to be a global ?

You can print all pcpu structs with "show allpcpu".

Andrew