[uml-devel] SEGV in hostfs open_dir() in latest CVS version

[Henrik N. <hn...@ma...>]

hostfs crashes reliably on ls -l.

ip is in malloc.c:chunk_malloc(), which usually indicates trashed
memory.

#0  panic (fmt=0x1010a100 "Kernel mode fault at addr 0x%lx, ip 0x%lx")
at panic.c:54
#1  0x1009e1b7 in segv (address=270025508, ip=269161786, is_write=2,
is_user=0) at trap_kern.c:60
#2  0x1009ea1e in segv_handler (sig=11, sc=0x50f2fad0, usermode=0) at
trap_user.c:301
#3  0x1009ea8e in sig_handler (sig=11) at trap_user.c:327
#4  0x100a6cf8 in __restore () at
../sysdeps/unix/sysv/linux/i386/sigaction.c:127
#5  0x100b22d2 in __libc_calloc (n=1, elem_size=4144) at malloc.c:3699
#6  0x100b4f5d in __opendir (name=0x500576a4 "/data/uml/lib/modules") at
../sysdeps/unix/opendir.c:135
#7  0x10099a1c in open_dir (path=0x500576a4 "/data/uml/lib/modules",
err_out=0x50f2feb8) at hostfs_user.c:82
#8  0x10098e5a in hostfs_readdir (file=0x50f9812c, ent=0x50f2ff08,
filldir=0x10036874 <filldir>) at hostfs_kern.c:176
#9  0x1003656d in vfs_readdir (file=0x50f9812c, filler=0x10036874
<filldir>, buf=0x50f2ff08) at readdir.c:27
#10 0x10036a74 in sys_getdents (fd=3, dirent=0xbf7ff044, count=3391) at
readdir.c:228
#11 0x1009d040 in execute_syscall (regs={regs = {3, 3212832836, 3391,
134571016, 3, 3212836296, 4294967258, 43, 43, 0, 0, 141, 1074434317, 35,
646, 3212832832, 43}}) at syscall_kern.c:348
#12 0x1009d14a in syscall_handler (unused=0x0) at syscall_user.c:73

I can reproduce it with close to 100% probability, but sometimes it does
not crash on the first attempt..

I boot singleuser with root on hostfs, and then issue ls -l.. sometimes
it crashes on the first directory, sometimes on the second attempt.

ls -l in /dev (devfs) works fine, but as soon as I "ls -l" a hostfs
directory one or two times it crashes.

Would it be possible to run "uml" under a malloc debugger? I'll try
anyway...

--
Henrik Nordstrom
MARA Systems