Thanks for looking into this Vladimir!
> reiserfs panic labeled PAP-5680 which you encountered indicates that file
> truncate did not complete. Actually, reiserfs does not have to panic, it
> has to return EIO or something suitable.
Indeed, that would be much better behaviour. Maybe the attached patch is okay.
It compiles fine with 2.6.19.1. However, I don't see how to test it.
> Could it be hardware fault?
In case you saw the smartd warnings concerning /dev/hda, the reiserfs panic
occurred with a reiserfs on an S-ATA software raid 5 (/dev/sd[a-g]1) I got a
few months ago. The P-ATA boot disk is the only old component. I don't
remember any worrying S-ATA or raid messages in my syslog. Unfortunately, I
cannot have a closer look at the moment as I'm away on holiday.
Regards,
Joachim
--- stree.orig 2006-12-11 20:32:53.000000000 +0100
+++ stree.c 2006-12-26 14:53:48.000000000 +0100
@@ -1898,9 +1898,30 @@
goto out;
reiserfs_update_inode_transaction(p_s_inode);
}
- } while (n_file_size > ROUND_UP(n_new_file_size) &&
- search_for_position_by_key(p_s_inode->i_sb, &s_item_key,
- &s_search_path) == POSITION_FOUND);
+ if (n_file_size > ROUND_UP(n_new_file_size)) {
+ retval = search_for_position_by_key(p_s_inode->i_sb,
+ &s_item_key, &s_search_path);
+ if (retval == IO_ERROR) {
+ reiserfs_warning(p_s_inode->i_sb,
+ "jw-5670: reiserfs_do_truncate: "
+ "i/o failure occurred searching for %K",
+ &s_item_key);
+ err = -EIO;
+ goto out;
+ }
+ if (retval == FILE_NOT_FOUND) {
+ reiserfs_warning(p_s_inode->i_sb,
+ "jw-5672: reiserfs_do_truncate: "
+ "file not found searching for %K",
+ &s_item_key);
+ err = -EIO;
+ goto out;
+ }
+ }
+ else
+ retval = POSITION_NOT_FOUND;
+
+ } while (retval == POSITION_FOUND);
RFALSE(n_file_size > ROUND_UP(n_new_file_size),
"PAP-5680: truncate did not finish: new_file_size %Ld, current %Ld, oid %d",
