Re: Bugs on big endian systems

[Hans Reiser <reiser@namesys.com>]

Thanks Jeff and Pavel.  Nikita, send Pavel the usual agreement, and 
we'll send to Linus unless someone finds an objection.  Edward, read 
this patch, test it on alpha,
 and approve it.  Elena, test it on x86.

Hans

Jeff Mahoney wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
>
> Pavel -
>
> Thanks. The good news is that the filesystem wouldn't have panicked
> because of the SB_ONDISK_RESERVED_FOR_JOURNAL(s) bug. The bad news is
> that it wouldn't have mattered because the filesystem could never have
> been mounted anyway due to the journal header version mismatch.
>
> I'm surprised nobody's ever tried to mount a relocated journal enabled
> filesystem on a big endian machine.
>
> Hans -
>
> This bug appears in both 2.4 and 2.6 kernels and should be fixed in both.
>
> - -Jeff
>
> Pavel Bartusek wrote:
> | Hi list,
> |
> | The attached patch will fix the bugs which appears only on big endian
> | systems in the kernel 2.4.25.
> |
> | regards
> |
> |
> | Pavel Bartusek
> | Software Engineering
> |
> | SYSGO Real-Time Solutions AG | Embedded and Real-Time Software
> | Lise-Meitner-Str.15
> | 89081 Ulm, Germany
> |
> | Voice: +49-731-9533-1295
> | FAX:   +49-731-94683-10
> | www.sysgo.com | www.elinos.com | www.osek.de
> |
> |
> | 
> ------------------------------------------------------------------------
> |
> | --- linux.ori/fs/reiserfs/journal.c    Mon Aug 25 13:44:43 2003
> | +++ linux/fs/reiserfs/journal.c    Wed Mar  3 13:26:23 2004
> | @@ -2094,7 +2094,7 @@
> |
> |      /* make sure that journal matches to the super block */
> |      if (is_reiserfs_jr(rs) &&
> | -        jh->jh_journal.jp_journal_magic != sb_jp_journal_magic(rs)) {
> | +        le32_to_cpu (jh->jh_journal.jp_journal_magic) !=
> sb_jp_journal_magic(rs)) {
> |      char jname[ 32 ];
> |      char fname[ 32 ];
> |     
> | @@ -2102,7 +2102,7 @@
> |      strcpy( fname, kdevname( p_s_sb->s_dev ) );
> |      printk("journal-460: journal header magic %x (device %s) does 
> not "
> |             "match magic found in super block %x (device %s)\n",
> | -        jh->jh_journal.jp_journal_magic, jname,
> | +        le32_to_cpu (jh->jh_journal.jp_journal_magic), jname,
> |          sb_jp_journal_magic(rs), fname);
> |      brelse (bhjh);
> |      goto free_and_return;
> | --- linux.ori/include/linux/reiserfs_fs.h    Mon Aug 25 13:44:44 2003
> | +++ linux/include/linux/reiserfs_fs.h    Wed Mar  3 13:25:45 2004
> | @@ -219,7 +219,7 @@
> |  #define SB_ONDISK_JOURNAL_DEVICE(s) \
> |           le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev))
> |  #define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \
> | -         le32_to_cpu
> ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
> | +         le16_to_cpu
> ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal))
> |
> |  #define is_block_in_log_or_reserved_area(s, block) \
> |           block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \
>
>
> - --
> Jeff Mahoney
> SuSE Labs
> jeffm@suse.com
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.2.2 (GNU/Linux)
> Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
>
> iD8DBQFARk4/LPWxlyuTD7IRAkIZAJ0VR8aYna2n3EK/BfZ2SBwKOe9OPQCfYzDH
> c8D4gSwDzr93HivtH2FMx5M=
> =Jj1U
> -----END PGP SIGNATURE-----
>
>


-- 
Hans