Reiser4 on PowerPC, oops and more

[Carl Henrik Lunde <chlunde+reiser@ping.uio.no>]

Hi, I know it's not supposed to work (yet), but I thought you might be
interrested in these logs (from a 32 bit PPC machine):

First I tried mounting a partition formatted with mkfs.reiserfs4 on the
PowerPC machine (2.6.9-rc3-mm-3):

ef_hash_table: 8192 buckets
z_hash_table: 8192 buckets
z_hash_table: 8192 buckets
j_hash_table: 16384 buckets
loading reiser4 bitmap......done (771 jiffies)
d_cursor_hash_table: 256 buckets
reiser4[mount(823)]: key_warning (fs/reiser4/plugin/object.c:97)[nikita-717]:
code: -2 at fs/reiser4/tree_walk.c:193
WARNING: Error for inode 720575940379279360 (-2)
for key: (910200000000000:0:0:2:a00000000000000:0)[*]
inode: ino: 0, count: 1, link: 1, mode: 0, size: 0
        uid: 0, gid: 0, dev: 3145733, rdev: 0
        atime: [0,0], mtime: [0,0], ctime: [0,0]
        blkbits: 12, blksize: 0, blocks: 0, bytes: 0
        version: 0, generation: 0, state: 72, flags: 0
        is_reiser4_inode: 1
        key: (910200000000000:1:ffffffffffffffff:0:a00000000000000:0)[stat data]
        file: (nil)
        dir: (nil)
        perm: (nil)
        formatting: (nil)
        hash: (nil)
        sd: (nil)
        sd_seal: version: 0, block: 0
seal key: (0:0:0:0:0:0)[]
seal coord: item_pos = 0, unit_pos 0, tween=before unit, iplug=0
        sd_coord: item_pos = 0, unit_pos 0, tween=invalid, iplug=255
        flags: 0x0, extmask: 0x0, pmask: 0, locality: 653057130340810752



Then I tried mounting a file made with mkfs.reiser4/IA32 on the
PowerPC machine:

$ sudo mount -o loop reiser4-from-ia32 /mnt

Message from syslogd@pitr at Sun Oct 10 21:25:03 2004 ...
pitr kernel: reiser4 panicked cowardly: assertion failed: atomic_read(&node->d_count) > 0

loop: loaded (max 8 devices)
ef_hash_table: 8192 buckets
z_hash_table: 8192 buckets
z_hash_table: 8192 buckets
j_hash_table: 16384 buckets
loading reiser4 bitmap...<4>reiser4[mount(1083)]: check_struct_bnode (fs/reiser4/plugin/space/bitmap.c:725)[vpf-1361]:
WARNING: Checksum for the bitmap block 18 is incorrect
reiser4[mount(1083)]: jrelse_tail (fs/reiser4/jnode.c:1023)[nikita-489]:
reiser4 panicked cowardly: assertion failed: atomic_read(&node->d_count) > 0
pid: 1083, comm: mount
        lock stack:
.... nr_signaled 0
.... curpri high
.... current locks:
        atom: no atom
kernel BUG in reiser4_do_panic at fs/reiser4/debug.c:131!
Oops: Exception in kernel mode, sig: 5 [#1]
NIP: C009E194 LR: C009DFE8 SP: D2879910 REGS: d2879860 TRAP: 0700    Not tainted
MSR: 00029032 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 11
TASK = d32103b0[1083] 'mount' THREAD: d2878000Last syscall: 120 
GPR00: C043A69C D2879910 D32103B0 00000000 00000001 FFFFFFFF C0440000 C0390000 
GPR08: C043A69C C043A69C C043A69C C043A69C 82002424 1002AE6C 100C0000 100A0000 
GPR16: 00000000 00000000 10010000 7FFFF718 C0340000 C0360000 C0320000 C0370000 
GPR24: D74A3010 C0910C00 D53C72E0 00000000 00000001 C0450000 4B1B5D0B D2879D00 
NIP [c009e194] reiser4_do_panic+0x28c/0x2a0
LR [c009dfe8] reiser4_do_panic+0xe0/0x2a0
Call trace:
 [c00a2504] jrelse_tail+0xd8/0x1b0
 [c019d80c] release+0x18/0x1c0
 [c019da18] done_bnode+0x64/0x68
 [c01a0f88] destroy_allocator_bitmap+0xdc/0x2c8
 [c01a0dec] init_allocator_bitmap+0x1d4/0x294
 [c01a21a4] try_init_format40+0x380/0x4d0
 [c01a2310] get_ready_format40+0x1c/0x17c
 [c0115958] _init_disk_format+0x2c/0x80
 [c0116044] reiser4_fill_super+0x54/0xdc
 [c0069b78] get_sb_bdev+0x150/0x1cc
 [c01075d4] reiser4_get_sb+0x18/0x28
 [c0069e7c] do_kern_mount+0x64/0x128
 [c008126c] do_new_mount+0x94/0xc4
 [c00819c8] do_mount+0x158/0x190
 [c0081e50] sys_mount+0xa4/0xf4


So there are endian-related bugs both in the user space tools and
kernel driver.

Here are the kernel messages from 2.6.8 + debian patches +
reiserfs4-patches from debian, when mounting an image made on IA32:

reiser4[mount(26854)]: check_struct_bnode (fs/reiser4/plugin/space/bitmap.c:724)[vpf-1361]:
WARNING: Checksum for the bitmap block 18 is incorrect
Oops: kernel access of bad area, sig: 11 [#1]
NIP: C009B720 LR: C00A4F80 SP: CF157CA0 REGS: cf157bf0 TRAP: 0300    Not tainted
MSR: 00009032 EE: 1 PR: 0 FP: 0 ME: 1 IR/DR: 11
DAR: 00000000, DSISR: 40000000
TASK = d4bef320[26854] 'mount' THREAD: cf156000Last syscall: 120 
GPR00: C00A4F80 CF157CA0 D4BEF320 00000000 C000B32C 00E00000 00000000 00049032 
GPR08: 00000000 00000000 00002000 00000001 0724EA46 1002AE6C 100C0000 100A0000 
GPR16: 00000000 100EF3C8 00000000 00000000 7FFFFAE9 00000000 10020000 00000000 
GPR24: 7FFFFAF1 C02728F0 CF157D30 C0935200 C7100818 C7100818 C7100840 C7100818 
NIP [c009b720] znodes_tree_done+0x3c/0x288
LR [c00a4f80] done_tree+0x4c/0x78
Call trace:
 [c00a4f80] done_tree+0x4c/0x78
 [c00c2dd8] _done_formatted_fake+0x28/0x48
 [c00c31f8] done_super+0x48/0x64
 [c00c329c] reiser4_fill_super+0x88/0x90
 [c0066fc4] get_sb_bdev+0x150/0x1c0
 [c00bc4cc] reiser4_get_sb+0x18/0x28
 [c00672ac] do_kern_mount+0x64/0x128
 [c007deec] do_new_mount+0x94/0xc4
 [c007e618] do_mount+0x158/0x190
 [c007eaa0] sys_mount+0xa4/0xf4
 [c0007d30] ret_from_syscall+0x0/0x4c


(I tried reiser4progs 1.0.2 too.  Note that the -mm3 kernel is
probably not stable on PowerPC, programs such as firefox crashed at
startup.)

I also took a quick search for non-byteswapped 32 bit integers, and
found that at least the alloc block checksum is not byteswapped (when
creating an image on PPC).

-- 
Carl Henrik Lunde