Kernel hang on 6.x

[Hans Petter Selasky]

On Thursday 14 December 2006 20:05, Brian Dean wrote:
> Hi,
>
> We're experiencing a kernel hang on a 6.x quad processor Sun amd64
> based system.  We are able to reproduce it fairly reliably, but the
> environment to do so is not easily replicatable so I cannot provide a
> simple test case.  However, I have been able to build a debug kernel
> and when the system "hangs", I can break to the debugger prompt.  But
> once there, I'm not sure what to do to isolate where the system is
> hung up.  I have confirmed that the hang occurs in both SMP and
> uniprocessor mode.  Here are some system details:
>
> uname -a:
> FreeBSD bb02f54 6.2-BETA2 FreeBSD 6.2-BETA2 #4: Wed Dec 13 11:43:38 EST
> 2006     root at bb02f54:/usr/src/sys/amd64/compile/BBKERN  amd64
>
> FreeBSD 6.2-BETA2 #4: Wed Dec 13 11:43:38 EST 2006
>     root at bb02f54:/usr/src/sys/amd64/compile/BBKERN
> WARNING: WITNESS option enabled, expect reduced performance.
> Timecounter "i8254" frequency 1193182 Hz quality 0
> CPU: Dual Core AMD Opteron(tm) Processor 275 (2193.76-MHz K8-class CPU)
>   Origin = "AuthenticAMD"  Id = 0x20f12  Stepping = 2
>  
> Features=0x178bfbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MC
>A,CMOV,PAT,PSE36,CLFLUSH,MMX,FXSR,SSE,SSE2,HTT> Features2=0x1<SSE3>
>   AMD Features=0xe2500800<SYSCALL,NX,MMX+,FFXSR,LM,3DNow+,3DNow>
>   AMD Features2=0x3<LAHF,CMP>
>   Cores per package: 2
> real memory  = 17179869184 (16384 MB)
> avail memory = 16518569984 (15753 MB)
> ACPI APIC Table: <SUN    X4200   >
>
> The hang appears to occur under heavy memory usage and usually seems
> to happen when the process size approaches the size of swap.
>
> If anyone can offer a suggestion as to what information from the db>
> prompt might help home in on this problem, please let me know.  A
> simple backtrace wasn't terribly englightening, at least to me:
>

Does this happen when you work on the console? Or when you switch from X11 to 
the console. I know that the keyboard driver is called from several places 
without Giant locked. In my new USB keyboard driver I have added several "if 
(!mtx_owned(&Giant)) return XXX;". Maybe you can try adding such to the SUN 
keyboard driver?

Please see:
http://www.turbocat.net/~hselasky/isdn4bsd/sources/src/sys/dev/usb/ukbd.c

> db> bt
> Tracing pid 18 tid 
1000http://www.turbocat.net/~hselasky/isdn4bsd/sources/src/sys/dev/usb/ukbd.c11 
td 0xffffff03e1563980
> kdb_enter() at kdb_enter+0x2f
> scgetc() at scgetc+0x43e
> sckbdevent() at sckbdevent+0x83
> kbdmux_intr() at kbdmux_intr+0x4d
> kbdmux_kbd_intr() at kbdmux_kbd_intr+0x20
> taskqueue_run() at taskqueue_run+0x135
> ithread_loop() at ithread_loop+0x132
> fork_exit() at fork_exit+0x87
> fork_trampoline() at fork_trampoline+0xe
> --- trap 0, rip = 0, rsp = 0xffffffffbf50ad00, rbp = 0 ---
> db>
>
> db> show reg
> cs                 0x8
> ss                0x10
> rax               0x26
> rcx           0x10457a
> rdx                0x1
> rbx                  0
> rsp         0xffffffffbf50aac0
> rbp         0xffffffffbf50aad0
> rsi         0xffffffff80c11000
> rdi                  0
> r8             0xffe00
> r9                 0xa
> r10         0xffffffffbf50a9e0
> r11                0xa
> r12         0xffffffff80957c20  main_softc
> r13         0xffffffff809579c0  main_console
> r14                0x2
> r15                  0
> rip         0xffffffff803fd57f  kdb_enter+0x2f
> rflags           0x286
> dr0                  0
> dr1                  0
> dr2                  0
> dr3                  0
> dr4         0xffff0ff0
> dr5              0x400
> dr6         0xffff0ff0
> dr7              0x400
> kdb_enter+0x2f: nop
> db>
>

--HPS