Wednesday, 07 August, 2013 11:02 Written by Brian B
The Solaris Crash Analysis Tool is a fantastic solution that is available in “My Oracle Support” (MOS) that can assist those that don’t have a strong background in Solaris internals in looking at potential issues with a system that is in a panic condition.
The built-in modular debugger (mdb) can also augment or at times work faster than SCAT
Here is a very basic walkthrough that I provide to our Collier IT engineers to assist them in initial diagnostics.
There’s much more, and I’ll add some additional walk-throughs later.
1. Useful information can be found in the stack backtrace to search keywords against MOS. Sometimes you get lucky here.
> $c vpanic(127def0, 2a100ed40c0, 0, 0, 3effffff8000000, 1869c00) cpu_deferred_error+0x568(ecc1ecc100000000, 2, 1000060000003a, 600000000, 0, 30001622360) ktl0+0x48(29fff982000, 2a100ed4d78, 30000, 16, 60, 30) pp_load_tlb+0x1e4(29fff980000, 29fff9822c0, 1d00, 29fff980300, 1822f00, 2) ppcopy_common+0x12c(70001d32500, 700030b2500, 1, 1, 29fff982000, 29fff980000) ppcopy+0xc(70001d32500, 700030b2500, 0, 0, 1822348, 70001d32500) do_page_relocate+0x228(2a100ed5120, 2a100ed5128, 700030b2500, 2a100ed53e0, 0, 2a100ed4fb0) page_relocate+0x14(2a100ed5120, 2a100ed5128, 1, 1, 2a100ed53e0, 0) page_lookup_create+0x244(60017811400, 6007c570000, 70001d32500, 0, 2a100ed53e0, 0) swap_getconpage+0xb4(60017811400, 6007c570000, 2000, 0, 2a100ed53c8, 2000) anon_map_getpages+0x474(60010c02008, 0, 200, 109a420, 2a100ed53e0, 1) segvn_fault_anonpages+0x32c(0, 800000, 0, 1, 6001753c2a8, 3) segvn_fault+0x530(300034bc3c0, 300012abc20, 1, 1, 892000, ffffffffff76e000) as_fault+0x4c8(300012abc20, 6001766b9d0, 890000, 60016881390, 186c0b0, 0) pagefault+0xac(890000, 0, 1, 0, 60016881318, 1) trap+0xd50(2a100ed5b90, 8903bb, 0, 1, fea0ad6c, 0) utl0+0x4c(1e, fe8f8104, 9e58, fe8fee34, 7aebd8, fe8fa524) >
2. Status can also give you things like the hostname and the kernel revision they’re running:
> ::status debugging crash dump vmcore.0 (64-bit) from sunbkpsrv5 operating system: 5.10 Generic_142900-13 (sun4u) panic message: UE CE Error(s) dump content: kernel pages only >
3. cpuinfo also shows some good info on what was running when the system panicked
> ::cpuinfo -v
ID ADDR FLG NRUN BSPL PRI RNRN KRNRN SWITCH THREAD PROC
0 0000183a620 1b 7 0 60 no no t-0 3000371fb20 java
| |
RUNNING <--+ +--> PRI THREAD PROC
READY 60 2a1000c7ca0 sched
EXISTS 59 30001e121e0 java
ENABLE 59 30001d293e0 in.mpathd
59 3000371d480 java
59 3000371ce00 java
59 3000371c440 java
59 3000371f4a0 java
ID ADDR FLG NRUN BSPL PRI RNRN KRNRN SWITCH THREAD PROC
1 0000180c000 1d 6 0 59 yes no t-0 30001dc01c0 syslogd
| |
RUNNING <--+ +--> PRI THREAD PROC
QUIESCED 99 2a100237ca0 sched
EXISTS 60 2a100a83ca0 sched
ENABLE 53 3000371c100 java
53 3000371c780 java
51 3000371aaa0 java
50 300032a9940 savecore
>
4. ::ps gives good info on everything running at the time of the crash
> ::ps S PID PPID PGID SID UID FLAGS ADDR NAME R 0 0 0 0 0 0x00000001 0000000001838150 sched R 3 0 0 0 0 0x00020001 0000060012dab848 fsflush R 2 0 0 0 0 0x00020001 0000060012dac468 pageout R 1 0 0 0 0 0x4a004000 0000060012dad088 init R 808 1 807 807 0 0x42000000 0000060016acf890 nbevtmgr R 805 1 7 7 60002 0x4a304102 0000060016746038 java R 764 1 764 764 0 0x42000000 0000060016acec70 dbsrv11 R 712 1 711 711 0 0x42000000 0000060016ad04b0 bpcd R 709 1 708 708 0 0x42000000 00000600167fa040 vnetd R 386 1 385 385 0 0x42000000 0000060016ad10d0 snmpd R 382 1 382 382 25 0x52010000 00000600169a2048 sendmail R 381 1 381 381 0 0x52010000 00000600169a2c68 sendmail R 334 1 334 334 0 0x42000000 0000060016747878 syslogd R 327 1 327 327 0 0x42000000 00000600161c0490 sshd R 324 1 323 323 0 0x42010000 00000600167fb880 smcboot R 326 324 323 323 0 0x42010000 0000060013fba018 smcboot R 325 324 323 323 0 0x42010000 00000600167fac60 smcboot R 275 1 275 275 0 0x42000000 0000060016748498 utmpd R 267 1 266 266 0 0x42000000 00000600159bb860 pbx_exchange R 263 1 263 263 0 0x42000000 00000600159bac40 inetd R 257 1 257 257 0 0x42000000 0000060013e26c30 automountd R 259 257 257 257 0 0x42000000 0000060015d02488 automountd R 251 1 251 251 1 0x42000000 0000060013fbc478 rpcbind R 234 1 234 234 0 0x42010000 00000600161c10b0 cron R 208 1 208 208 0 0x42000000 0000060015d00c48 xntpd R 185 1 7 7 0 0x42000000 0000060013fbd098 iscsid R 155 1 154 154 0 0x42000000 0000060013e28470 in.mpathd R 144 1 144 144 0 0x42000000 00000600159ba020 picld R 139 1 139 139 1 0x42000000 00000600159bd0a0 kcfd R 136 1 136 136 0 0x42000000 0000060012daac28 nscd R 120 1 120 120 0 0x42000000 0000060015d030a8 syseventd R 80 1 79 79 0 0x42020000 0000060013e26010 dhcpagent R 61 1 61 61 0 0x42000000 0000060013fbb858 devfsadm R 9 1 9 9 0 0x42000000 0000060013e29090 svc.configd R 7 1 7 7 0 0x42000000 0000060012daa008 svc.startd R 357 7 7 7 0 0x4a004000 0000060016746c58 rc2 R 702 357 7 7 0 0x4a004000 00000600167490b8 lsvcrun R 703 702 7 7 0 0x4a004000 0000060013e27850 sh R 809 703 7 7 0 0x4a004000 00000600169a3888 pdde R 812 809 7 7 0 0x4a004000 0000060016ace050 pdde R 813 812 7 7 0 0x4a004000 00000600169a44a8 sleep R 342 7 7 7 0 0x4a004000 0000060015d00028 svc-webconsole R 717 342 7 7 0 0x4a004000 00000600169a50c8 sjwcx R 720 717 7 7 0 0x4a004000 00000600167fc4a0 java R 304 7 304 304 0 0x4a004000 0000060013fbac38 ttymon R 290 7 7 7 0 0x4a004000 00000600167fd0c0 svc-dumpadm R 293 290 7 7 0 0x4a004000 00000600161bf870 savecore R 269 7 269 269 0 0x4a014000 00000600161be030 sac R 278 269 269 269 0 0x4a014000 0000060015d01868 ttymon
5. ::panicinfo shows more info on the panic itself
> ::panicinfo
cpu 0
thread 3000371fb20
message UE CE Error(s)
tstate 80001606
g1 1270ce4
g2 127dc00
g3 3effffff8000000
g4 fbfffffe
g5 1
g6 0
g7 3000371fb20
o0 127def0
o1 2a100ed4098
o2 0
o3 0
o4 fc30ffffffffffff
o5 3cf000000000000
o6 2a100ed3761
o7 11020dc
pc 104982c
npc 1049830
y 0
>
6. Find the address of the thread that was executing when the system panicked.
> panic_thread/K panic_thread: panic_thread: 3003acf7020 gt;
7. Run the thread macro against the pointer value from above. Search for the t_procp structure.
> 3003acf7020$<$thread
t_link = 0
t_stk = 0x2a108333ae0
t_startpc = 0
t_bound_cpu = 0x30004b42000
t_affinitycnt = 0
t_bind_cpu = 0xffff
t_flag = 0x1800
t_proc_flag = 0x104
...
t_procp = 0x3005a6713e0 <== use the value here ...
>
8. run the proc2u macro against the pointer from the t_procp structure. Look for the value stored in p_user.u_psargs. This is the full path to the command that was running on the CPU at the time of the system panic.
> 0x3005a6713e0$<proc2u
p_user.u_execsw = execsw+0x28
p_user.u_auxv = [
{
a_type = 0x7d8
a_un = {
a_val = 0xffffffff7fffff90
a_ptr = 0xffffffff7fffff90
a_fcn = 0xffffffff7fffff90
}
...
p_user.u_start = {
tv_sec = 2007 Jun 11 00:00:00
tv_nsec = 0xcf77e0
}
p_user.u_ticks = 0x191b148
p_user.u_comm = [ "bgscollect" ]
p_user.u_psargs = [ "bgscollect -I noInstance -B /usr/adm/best1_7.3.00" ] <== use the value here
p_user.u_argc = 0x5
p_user.u_argv = 0xffffffff7ffffc98 ...
>