Inline function (difficult debug)

Thu Jun 4 13:00:53 UTC 2009

John Baldwin wrote:
> On Thursday 04 June 2009 5:24:07 am Menshikov Konstantin wrote:
>   
>> Hi.
>> My system FreeBSD 7.1 RELEASE periodically freeze.
>>
>> I`m compiling kernel with WITNESS and get backtrace.
>>
>> #0 doadump () at pcpu.h:195
>> #1 0xffffffff801a899c in db_fncall (dummy1=Variable "dummy1" is not 
>> available.
>> ) at /usr/src/sys/ddb/db_command.c:516
>> #2 0xffffffff801a8ecf in db_command (last_cmdp=0xffffffff80825688, 
>> cmd_table=0x0, dopager=1)
>> at /usr/src/sys/ddb/db_command.c:413
>> #3 0xffffffff801a90e0 in db_command_loop () at 
>> /usr/src/sys/ddb/db_command.c:466
>> #4 0xffffffff801aaa19 in db_trap (type=Variable "type" is not available.
>> ) at /usr/src/sys/ddb/db_main.c:228
>> #5 0xffffffff803d5e55 in kdb_trap (type=3, code=0, 
>> tf=0xffffffffabf284c0) at /usr/src/sys/kern/subr_kdb.c:524
>> #6 0xffffffff8059a0e5 in trap (frame=0xffffffffabf284c0) at 
>> /usr/src/sys/amd64/amd64/trap.c:538
>> #7 0xffffffff8057f60e in calltrap () at 
>> /usr/src/sys/amd64/amd64/exception.S:209
>> #8 0xffffffff803d602d in kdb_enter_why (why=0xffffffff80628137 
>> "witness", msg=0xa <Address 0xa out of bounds>)
>> at cpufunc.h:63
>> #9 0xffffffff803e70c3 in witness_checkorder (lock=Variable "lock" is not 
>> available.
>> ) at /usr/src/sys/kern/subr_witness.c:1126
>> #10 0xffffffff8039e341 in _mtx_lock_spin_flags (m=0xffffffff8084fad0, 
>> opts=0, file=dwarf2_read_address: Corrupted DWARF expression.
>> ) at /usr/src/sys/kern/kern_mutex.c:227
>> #11 0xffffffff80319838 in sc_puts (scp=0xffffffff8084f980, 
>> buf=0xffffffffabf286a7 "a@\az\200ЪЪЪЪ", len=1)
>> at /usr/src/sys/dev/syscons/syscons.c:2519
>> #12 0xffffffff8031b3da in sc_cnputc (cd=Variable "cd" is not available.
>> ) at /usr/src/sys/dev/syscons/syscons.c:1561
>> #13 0xffffffff803f9d89 in cnputc (c=97) at /usr/src/sys/kern/tty_cons.c:632
>> #14 0xffffffff803d993b in putchar (c=97, arg=Variable "arg" is not 
>> available.
>> ) at /usr/src/sys/kern/subr_prf.c:421
>> #15 0xffffffff803d7f22 in kvprintf (fmt=0xffffffff80628b89 "cquiring 
>> duplicate lock of same type: \"%s\"\n",
>> func=0xffffffff803d98d0 <putchar>, arg=0xffffffffabf28890, radix=10, 
>> ap=Variable "ap" is not available.
>> ) at /usr/src/sys/kern/subr_prf.c:674
>> #16 0xffffffff803d9384 in printf (fmt=Variable "fmt" is not available.
>> ) at /usr/src/sys/kern/subr_prf.c:314
>> #17 0xffffffff803e705d in witness_checkorder (lock=0xffffffff808668f8, 
>> flags=0,
>> file=0xffffffff80626cf8 "/usr/src/sys/kern/subr_sleepqueue.c", line=232) 
>> at /usr/src/sys/kern/subr_witness.c:948
>> #18 0xffffffff8039e341 in _mtx_lock_spin_flags (m=0xffffffff808668f8, 
>> opts=0, file=dwarf2_read_address: Corrupted DWARF expression.
>> ) at /usr/src/sys/kern/kern_mutex.c:227
>> #19 0xffffffff803b2271 in wakeup (ident=0xffffffff80850ac0) at 
>> /usr/src/sys/kern/kern_synch.c:341
>> #20 0xffffffff803aed95 in tdsignal (p=0xffffff00017978f0, 
>> td=0xffffff000146a6e0, sig=14, ksi=Variable "ksi" is not available.
>> )
>> at /usr/src/sys/kern/kern_sig.c:2292
>> #21 0xffffffff803ba2de in realitexpire (arg=Variable "arg" is not available.
>> ) at /usr/src/sys/kern/kern_time.c:669
>> #22 0xffffffff803bbe1a in softclock (dummy=Variable "dummy" is not 
>> available.
>> ) at /usr/src/sys/kern/kern_timeout.c:274
>> #23 0xffffffff8038c120 in ithread_loop (arg=0xffffff00010ebc00) at 
>> /usr/src/sys/kern/kern_intr.c:1088
>> #24 0xffffffff8038978a in fork_exit (callout=0xffffffff8038c040 
>> <ithread_loop>, arg=0xffffff00010ebc00,
>> frame=0xffffffffabf28c80) at /usr/src/sys/kern/kern_fork.c:804
>> #25 0xffffffff8057f9de in fork_trampoline () at 
>> /usr/src/sys/amd64/amd64/exception.S:455
>>
>> But file /usr/src/sys/kern/kern_sig.c:2292
>> hasn`t run function wake up on string 2292
>> I`m think, what compiler use inline function, therefore code and trace 
>> differ.
>> Tell me please, how compile kernel without any inline function.
>>     
>
> makeoptions DEBUG="-g -fno-inline"
>
> However. I think you already have enough information to debug this.  Can you 
> go to frame 17 and do an 'l'?  I think the two things you want to print out 
> are "i", "*lock", "*lock1", and "*lock2".
>
>   
Unfortunately, the kernel is not compiled with an option-fno-inline.
cc -c -O2 -frename-registers -pipe -fno-strict-aliasing  -std=c99 -g 
-fno-inline -Wall -Wredundant-decls -Wnested-externs 
-Wstrict-prototypes  -Wmissing-prototypes -Wpointer-arith -Winline 
-Wcast-qual  -Wundef -Wno-pointer-sign -fformat-extensions -nostdinc  
-I. -I/usr/src/sys -I/usr/src/sys/contrib/altq -D_KERNEL 
-DHAVE_KERNEL_OPTION_HEADERS -include opt_global.h -fno-common 
-finline-limit=8000 --param inline-unit-growth=100 --param 
large-function-growth=1000  -fno-omit-frame-pointer -mcmodel=kernel 
-mno-red-zone  -mfpmath=387 -mno-sse -mno-sse2 -mno-mmx -mno-3dnow  
-msoft-float -fno-asynchronous-unwind-tables -ffreestanding -Werror  
/usr/src/sys/cam/cam_periph.c
cc1: warnings being treated as errors
/usr/src/sys/cam/cam_periph.c: In function 'scsi_extract_sense':
/usr/src/sys/cam/scsi/scsi_all.h:1215: warning: function 
'scsi_extract_sense' can never be inlined because it is suppressed using 
-fno-inline

I`m try build with -O0 options.

lock in frame 17
(kgdb) print {struct mtx} 0xffffffff808668f8
$4 = {lock_object = {lo_name = 0xffffffff806270eb "sleepq chain", 
lo_type = 0xffffffff806270eb "sleepq chain",
    lo_flags = 720896, lo_witness_data = {lod_list = {stqe_next = 
0xffffffff808793e0}, lod_witness = 0xffffffff808793e0}},
  mtx_lock = 4, mtx_recurse = 0}
next in list
(kgdb) print {struct mtx} 0xffffffff808793e0
$5 = {lock_object = {lo_name = 0xffffffff806270eb "sleepq chain", 
lo_type = 0xffffffff807e8860 "\031\006b\200ЪЪЪЪ\n",
    lo_flags = 2156368928, lo_witness_data = {lod_list = {stqe_next = 
0xffffffff80879420},
      lod_witness = 0xffffffff80879420}}, mtx_lock = 
18446744071571216096, mtx_recurse = 2153913763}
(kgdb) print {struct mtx} 0xffffffff80879420
next
$6 = {lock_object = {lo_name = 0xffffffff806205df "process slock", 
lo_type = 0xffffffff807e8860 "\031\006b\200ЪЪЪЪ\n",
    lo_flags = 2156368992, lo_witness_data = {lod_list = {stqe_next = 
0xffffffff80879460},
      lod_witness = 0xffffffff80879460}}, mtx_lock = 
18446744071571216160, mtx_recurse = 2153889520}

As far as I understand, 2 times are requested blocking of one class.
Besides, too great values mtx_recurse it is wrong.