[Bug 219153] head, stable/11, release/11.0.1: libkvm (& more?) not updated to handle powerpc/powerpc64 ET_DYN based vmcore.* 's and such

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=219153

--- Comment #11 from Mark Millard <markmi at dsl-only.net> ---
(In reply to Mark Millard from comment #9)

[This note is limited to contexts with gcc
4.2.1 based kernels.]

Summary after avoiding a user error: looks
like there is a default of debug.minidump=1
that means that the process information is
not in the vmcore file. I'll have to generate
new cores with that disabled.

Details:

I really needed to use:

ps -M /var/crash/vmcore.7 -N /usr/lib/debug/boot/kernel/kernel.debug

because when I look at the vmcore.*'s I'm
not using the kernel that fails periodically.
(Not using the kernel build that I wanted to
use kgdb with to look at its crashes.)
(A debug kernel build works but a production
build of the same sources crashes in a
pid 11 thread [idle thread] eventually.

I manually

unload
boot kergcd

instead of using /boot/kernel/kernel when I
do not want the kernel to fail for what I'm
doing.

/boot/kernel/kernel
and
/usr/lib/debug/boot/kernel/kernel.debug

are a matching pair for the failing kernel.

truss shows where ps extracts the
nprocs figure and such. My own
calculations via looking at:

readelf -a /var/crash/vmcore.7 | more
and
readelf -a /usr/lib/debug/boot/kernel/kernel.debug

agrees with what I see in:

cat /var/crash/vmcore.7 | hd | more

The address shown in ddb matches
my calculations as well.

But to do this with matching files I had
to use some more recent vmcore.* files
because other experiments had updated
the kernel (and world).

So here is what I found based on having a
matching -N for the -M :

nprocs=52 (0x36)

(which is good) but when it gets into:

        static int
        kvm_proclist(kvm_t *kd, int what, int arg, struct proc *p,
            struct kinfo_proc *bp, int maxcnt)

and its code:

                for (; cnt < maxcnt && p != NULL; p = LIST_NEXT(&proc, p_list))
{
                        memset(kp, 0, sizeof *kp);
                        if (KREAD(kd, (u_long)p, &proc)) {
                                _kvm_err(kd, kd->program, "can't read proc at
%p", p);
                                return (-1);
                        }

the _kvm_err is being called for the first p value:

(gdb) print p
$4 = (struct proc *) 0x873c370

for which no VirtAddr/MemSiz combination
in the vmcore.9 file spans representing
that address:

# readelf -a /var/crash/vmcore.9 
ELF Header:
  Magic:   7f 45 4c 46 01 02 01 ff 00 00 00 00 00 00 00 00 
  Class:                             ELF32
  Data:                              2's complement, big endian
  Version:                           1 (current)
  OS/ABI:                            StandAlone
  ABI Version:                       0
  Type:                              CORE (Core file)
  Machine:                           PowerPC 32-bit
  Version:                           0
  Entry point address:               0
  Start of program headers:          52 (bytes into file)
  Start of section headers:          0 (bytes into file)
  Flags:                             0
  Size of this header:               52 (bytes)
  Size of program headers:           32 (bytes)
  Number of program headers:         3
  Size of section headers:           40 (bytes)
  Number of section headers:         0 (0)
  Section header string table index: 0

Elf file type is CORE (Core file)
Entry point 0x0
There are 3 program headers, starting at offset 52

Program Headers:
  Type           Offset   VirtAddr   PhysAddr   FileSiz MemSiz  Flg Align
  LOAD           0x001000 0x008fe000 0xffffffff 0x6d7000 0x6d7000 R   0x1000
  LOAD           0x6d8000 0xd0005000 0xffffffff 0x18000 0x18000 R   0x1000
  LOAD           0x6f0000 0xd001d000 0xffffffff 0x01000 0x01000 R   0x1000

There are no sections in this file.

There is no dynamic section in this file.

-- 
You are receiving this mail because:
You are the assignee for the bug.