sysctl_net_inet_tcp_ident panic

[John Klos]

Hi,

I have an amd64 system running netbsd-6 from 2-September-2013. Today,
while a few machines were scp'ing, this happened:


uvm_fault(0xfffffe80c01971d0, 0x0, 1) -> e
fatal page fault in supervisor mode
trap type 6 code 0 rip ffffffff804a9fbf cs 8 rflags 10246 cr2 44 cpl 0
rsp ffff8001229c8960
panic: trap
cpu2: Begin traceback...
printf_nolog() at netbsd:printf_nolog
startlwp() at netbsd:startlwp
alltraps() at netbsd:alltraps+0x96
sysctl_net_inet_tcp_ident() at netbsd:sysctl_net_inet_tcp_ident+0x3ee
sysctl_dispatch() at netbsd:sysctl_dispatch+0xb2
sys___sysctl() at netbsd:sys___sysctl+0xeb
syscall() at netbsd:syscall+0xc4
cpu2: End traceback...

dumping to dev 0,1 offset 212671
dump succeeded


rebooting...


Strangely, it took a while for the system to come up afterwards (I wasn't
there, and the filesystem is journaled), but eventually it did.

Any ideas?

Thanks,
John

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de

[Christos Zoulas]

Post by John Klos
Hi,
I have an amd64 system running netbsd-6 from 2-September-2013. Today,
uvm_fault(0xfffffe80c01971d0, 0x0, 1) -> e
fatal page fault in supervisor mode
trap type 6 code 0 rip ffffffff804a9fbf cs 8 rflags 10246 cr2 44 cpl 0
rsp ffff8001229c8960
panic: trap
cpu2: Begin traceback...
printf_nolog() at netbsd:printf_nolog
startlwp() at netbsd:startlwp
alltraps() at netbsd:alltraps+0x96
sysctl_net_inet_tcp_ident() at netbsd:sysctl_net_inet_tcp_ident+0x3ee
sysctl_dispatch() at netbsd:sysctl_dispatch+0xb2
sys___sysctl() at netbsd:sys___sysctl+0xeb
syscall() at netbsd:syscall+0xc4
cpu2: End traceback...
dumping to dev 0,1 offset 212671
dump succeeded

Looks like a NULL pointer dereference. using objdump you can find
the instruction sequence in the file. Then you can compile the source
file with -gstabs -S and get an assembly file that has the line
numbers with the code. At this point you should be able to see the
line number that caused the fault. Also set ddb.onpanic=2 and run
a ps to see what process ran the sysctl. Of course if the crash
worked, you can figure the same by running gdb on the core file.

christos


--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de

[Brian Marcotte]

Post by John Klos
uvm_fault(0xfffffe80c01971d0, 0x0, 1) -> e
fatal page fault in supervisor mode
trap type 6 code 0 rip ffffffff804a9fbf cs 8 rflags 10246 cr2 44 cpl 0
rsp ffff8001229c8960
panic: trap
cpu2: Begin traceback...
printf_nolog() at netbsd:printf_nolog
startlwp() at netbsd:startlwp
alltraps() at netbsd:alltraps+0x96
sysctl_net_inet_tcp_ident() at netbsd:sysctl_net_inet_tcp_ident+0x3ee
sysctl_dispatch() at netbsd:sysctl_dispatch+0xb2
sys___sysctl() at netbsd:sys___sysctl+0xeb
syscall() at netbsd:syscall+0xc4

I've had a panic which looks similar to that:

panic: kernel diagnostic assertion "cred != NULL" failed: file ...
kern/kern_auth.c, line 300
cpu3: Begin traceback...
kern_assert ... at netbsd:kern_assert+0x23
kauth_cred_geteuid ... at netbsd:kauth_cred_geteuid+0x43
sysctl_net_inet_tcp_ident ... at netbsd:sysctl_net_inet_tcp_ident+0x2fd
sysctl_dispatch ... at netbsd:sysctl_dispatch+0xc7
sys___sysctl ... at netbsd:sys___sysctl+0xea
syscall ... at netbsd:syscall+0xaa

More details are in PR #48098.

--
- Brian

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de