ah! util-linux umount gets segfault always.


On Sun, Jun 2, 2013 at 8:14 AM, plan9assembler <plan9assembler@gmail.com> wrote:
Hi,

latest musl libc seems to fixes mount segfault error partially.
(mount: /mnt: filesystem mounted, but mount(8) failedOperation timed out)
rebuild and test vanilla util-linux-2.23 result same.
tested util-linux-2.23.1 same.
tested gcc-4.8.1 same.

sscanf "%ms" -> "%s" with malloc doesn't fixed the issue here.
with patch or without it result same.

i compile strace-4.7 to trace the bug, but get some build error:
gcc -Wall -Wwrite-string -g -O2 -o strace strace.o syscall.o count.o util.o desc.o file.o ipc.o io.o ioctl.o mem.o net.o process.o bjm.o quota.o resource.o signal.o sock.o system.o term.o time.o scsi.o stream.o block.o pathtrace.o mtd.o vsprintf.o loop.o
syscall.o:(.rodata+0x28080): undefined reference to `sys_getdents64'
syscall.o:(.rodata+0x2b100): undefined reference to `sys_truncate64'
syscall.o:(.rodata+0x2b118): undefined reference to `sys_ftruncate64'
syscall.o:(.rodata+0x2b388): undefined reference to `sys_getdents64'
syscall.o:(.rodata+0x2dd00): undefined reference to `sys_getdents64'

BTW, i was quite surprised by base packages are so much "glibc-centric"..



On Thu, May 30, 2013 at 5:37 PM, Szabolcs Nagy <nsz@port70.net> wrote:
* plan9assembler <plan9assembler@gmail.com> [2013-05-30 16:26:13 +0900]:
>
> it still gets same segfaults, same bt gdb result.
>

if you get the exact same segfault then you do something wrong
..or malloc(200) is not enough

> and it is clear to me that latest musl libc[2013/05/29] contain new bug,
> because,
> below abnormal operation never happened before. (musl version git pulled at
> 2013/05/03)
>
> # ./mount /dev/sda1 /mnt
> EXT4-fs (sda1): mounted filesystem with ordered data mode. Opts: (null)
>
> < 30 - 40  seconds waiting without return to shell>
>
> mount: /mnt: filesystem mounted, but mount(8) failedOperation timed out
> // <-- this is weird.
> #

works here fine
since you have local modifications i'd check those first

i doubt that latest musl has any related bug

but you could easily prove me wrong with a strace
that shows bad flags passed to some syscall