> >Are you running a release or a debug snv_33 kernel?
> >
> svn_33, /var/adm/messages says
...
> Mar 8 20:52:29 ferrari unix: [ID 592667 kern.notice] r10:
> d000d000d000d r11: fffffffffff2fff2 r12: 0
...
> Mar 8 20:52:29 ferrari genunix: [ID 655072 kern.notice] fffffe80003d88a0
> unix:die+b1 ()
> Mar 8 20:52:29 ferrari genunix: [ID 655072 kern.notice] fffffe80003d89a0
> unix:trap+392 ()
> Mar 8 20:52:29 ferrari genunix: [ID 655072 kern.notice] fffffe80003d89b0
> unix:cmntrap+140 ()
> Mar 8 20:52:29 ferrari genunix: [ID 655072 kern.notice] fffffe80003d8b00
> udfs:ud_xlate_to_daddr+69 ()
The udfs module that I compiled from on-20060228 sources accesses
register %r10 as a pointer at ud_xlate_to_daddr+69, and the contents
of register %r10 (d000d000d000d) that is listed in your /var/adm/messages
looks quite strange.
% dis -F ud_xlate_to_daddr /kernel/fs/amd64/udfs
**** DISASSEMBLER ****
disassembly for /kernel/fs/amd64/udfs
section .text
ud_xlate_to_daddr()
ud_xlate_to_daddr: 55 pushq %rbp
ud_xlate_to_daddr+0x1: 48 8b ec movq %rsp,%rbp
ud_xlate_to_daddr+0x4: 48 83 ec 38 subq $0x38,%rsp
....
ud_xlate_to_daddr+0x54: 4d 0f b7 c7 movzwq %r15w,%r8
ud_xlate_to_daddr+0x58: 4d 69 c0 88 00 00 00 imulq $0x88,%r8,%r8
ud_xlate_to_daddr+0x5f: 4d 8b 95 a8 00 00 00 movq 0xa8(%r13),%r10
ud_xlate_to_daddr+0x66: 4d 03 d0 addq %r8,%r10
ud_xlate_to_daddr+0x69: 41 8b 02 movl (%r10),%eax
<<<<<<<<<<<<<<<<<<
ud_xlate_to_daddr+0x6c: 83 f8 01 cmpl $0x1,%eax
ud_xlate_to_daddr+0x6f: 75 49 jne +0x4b
<ud_xlate_to_daddr+0xba>
....
> I already filed a bug but do not have the bug id yet.
Bug ID: 6395818
Synopsis: Solaris crahes while writing to UDFS formatted DVD-RAM media
http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6395818
Are you using the Panasonic DVD-RAM media, that is
preformatted with an UDFS filesystem? Or did you use Solaris'
"mkfs -F udfs" command to create a new UDFS filesystem?
> > using a release kernel, you may want to repeat your DVD-RAM udfs experiment
> > with the
> > kernel running with kernel heap memory checking enabled: add "set
> > kmem_flags=0xf"
> > to /etc/system, and reboot.
> >
> Is this setting also useful for svn_33?
Yes. Kernel runs a bit slower when heap checking is enabled;
sometimes it helps catching errors in the kernel that you wouldn't
notice otherwise or that lead to crashes much later in some other
part of the kernel.
This message posted from opensolaris.org
