[Unionfs] RHEL kernel oops

[Theresa Noble]

My original intent was to get unionfs working on CentOS 5.1 (using
kernel 2.6.18-53.el5)

I tried using the RHEL5 patch here:
http://download.filesystems.org/unionfs/stable/unionfs-2.4_for_2.6.18-RHEL5.diff.gz

I had to make some modifications to the patch and I got it compiling
with the 53 version of the kernel but I kept getting a kernel oops.

I then tried to get the configuration as close to the one that Dave
had in this posting:
http://www.mail-archive.com/unionfs@mail.fsl.cs.sunysb.edu/msg02959.html

I didn’t have the source rpm he had, I pulled the Red Hat 5.1
kernel-2.6.18-92.1.6.el5 and I had to make a couple kabi check
modifications to get the spec file to work. I followed Dave’s advice
(in the thread above) and added  the following to my spec file:

for i in *.config; do
echo "CONFIG_UNION_FS=m" >> $i
echo "CONFIG_UNION_FS_XATTR=y" >> $i
echo "CONFIG_UNION_FS_DEBUG=y" >> $i
done

Except in other threads Erez suggested to turn CONFIG_UNION_FS_XATTR
off because it isn’t compatible with RHEL. So I set it to “n” instead.

I was able to get it to compile with the unionfs-2.4_for_2.6.18-RHEL5
patch with no modifications at all.

But I get another kernel oops:
root (hd0,0)
Filesystem type is ext2fs, partition type 0x83
kernel /boot/vmlinuz ro root=LABEL=/ console=ttyS0,115200 single quiet
  [Linux-bzImage, setup=0x1e00, size=0x1b80e4]
initrd /boot/initrd.img
  [Linux-initrd @ 0xfc88000, 0x257c44 bytes]
Memory for crash kernel (0x0 to 0x0) notwithin permissible range
------------[ cut here ]------------
kernel BUG at include/linux/dcache.h:323!
invalid opcode: 0000 [#1]
SMP
last sysfs file:
Modules linked in:
CPU:    0
EIP:    0060:[<c047ec91>]    Not tainted VLI
EFLAGS: 00010246   (2.6.18-92.1.6.el5 #1)
EIP is at do_path_lookup+0x120/0x2a5
eax: 00000000   ebx: 00000310   ecx: 00000110   edx: c127992c
esi: c1279910   edi: ffffffe9   ebp: c1279910   esp: c1279884
ds: 007b   es: 007b   ss: 0068
Process swapper (pid: 1, ti=c1279000 task=c1278aa0 task.ti=c1279000)
Stack: cf129000 c0472361 00000310 c1279910 ffffffe9 cf129000 c047f655c1279910

      ffffff9c ffffff9c c1279910 ffffff9c c1279910 c047f7f7 c1279910 00000242

      000081c0 c1279910 ffffff9c 00000242 cf129000 00000002 cf129000 c1279910

Call Trace:

 [<c0472361>] get_empty_filp+0x99/0x164
 [<c047f655>] __path_lookup_intent_open+0x42/0x72
 [<c047f7f7>] open_namei+0x9a/0x602
 [<c046fe9f>] do_filp_open+0x1c/0x31
 [<c0488bca>] mntput_no_expire+0x13/0x6c
 [<c046feef>] do_sys_open+0x3b/0xaf
 [<c046ff90>] sys_open+0x16/0x18
 [<c06ee63f>] do_name+0xc2/0x1bb
 [<c06ed8c7>] write_buffer+0x1a/0x28
 [<c06ed936>] flush_window+0x61/0xaf
 [<c06edad0>] inflate_codes+0x14c/0x3d9
 [<c06ef0f1>] inflate_dynamic+0x520/0x556
 [<c06ef5f5>] unpack_to_rootfs+0x4ce/0x8e0
 [<c06ef9fa>] unpack_to_rootfs+0x8d3/0x8e0
 [<c04269f2>] printk+0x19/0xa0
 [<c06efa7a>] populate_rootfs+0x73/0xc1
 [<c06ea55b>] init+0x12b/0x24c
 [<c0404e0a>] ret_from_fork+0x6/0x1c
 [<c06ea430>] init+0x0/0x24c
 [<c06ea430>] init+0x0/0x24c
 [<c0405c57>] kernel_thread_helper+0x7/0x10
 =======================
Code: 80 40 04 00 00 8b 50 0c eb 7c 80 e5 01 74 2c 8b 45 04 85 c0 74
04 f0 ff 40 28 8b 55 00 85 d2 0f 84 08 01 00 00 8b 02 85 c0 75 08 <0f>
0b 43 01 1b a4 62 c0 f0 ff 02 e9 f2 00 00 00 83 f8 9c 75 73
EIP: [<c047ec91>] do_path_lookup+0x120/0x2a5 SS:ESP 0068:c1279884
 <0>Kernel panic - not syncing: Fatal exception

I see the main possible difference is my kernel version may be
different because I was just using the stock kernel.. I am not sure
what Dave was using. I wasn’t getting a kernel oops when I took out
unionfs.

Does anyone have any ideas?

Thanks!



Theresa