| 1 |
Hi, |
| 2 |
|
| 3 |
As i look that you want to stay with the latest-fix on reiser-fs. I send you the |
| 4 |
e-mail i got on the linux-kernel mailing-list. So you can incorporate the patch in |
| 5 |
your own kernel or wait for next 2.4.3-preX |
| 6 |
|
| 7 |
As filesystem stuff are quiet critical (especially for me). |
| 8 |
|
| 9 |
Philippe Namias |
| 10 |
|
| 11 |
PS: please tell me if this kind of fix must be send on the mailing-list? |
| 12 |
|
| 13 |
Thanks |
| 14 |
|
| 15 |
--------------------------------------- |
| 16 |
Hello ! |
| 17 |
|
| 18 |
The Reiserfs development team resubmits five revised Reiserfs patches |
| 19 |
for the version after Linux 2.4.3-pre4. Their descriptions are: |
| 20 |
|
| 21 |
1. "prealloc.diff" --a fix for possible preallocated blocks leakage |
| 22 |
after a system crash |
| 23 |
|
| 24 |
Reiserfs blocks preallocation for big files in the same manner as |
| 25 |
ext2. The preallocation code changes both the bitmap and the free |
| 26 |
blocks count in the superblock. It is possible after a crash to |
| 27 |
have unattached disk blocks which do not belong to any file |
| 28 |
(i.e. free space leak). Unlike ext2, reiserfs does not check for |
| 29 |
consistency of a filesystem that was not cleanly unmounted. There |
| 30 |
is no simple way to recover free space. |
| 31 |
|
| 32 |
For a given transaction, the fix is to, at the transaction end, |
| 33 |
discard the preallocations for those inodes which have one. |
| 34 |
|
| 35 |
2. "objectid-sharing.diff" -- object id sharing problem fix |
| 36 |
|
| 37 |
Fix for reiserfs/inode.c:reiserfs_iget () The fix makes sure that |
| 38 |
the reiserfs_iget function really got the inode of the object for |
| 39 |
which it is looking. |
| 40 |
|
| 41 |
3. "namei.diff" -- improvements of reiserfs_add_entry() function |
| 42 |
|
| 43 |
a. Error handling |
| 44 |
|
| 45 |
There is a place in reiserfs_add_entry() code where fs corruption |
| 46 |
(existence of hidden entries--objects temporarily used in |
| 47 |
renaming), I/O errors, or several other occurrences, can produce an |
| 48 |
error message which has no useful information. |
| 49 |
|
| 50 |
We have had several bug reports about this vs-7032 message in the |
| 51 |
older 3.5 version. Each time we were unable to reproduce it. We do |
| 52 |
not understand the reason. This error handling fix will make such |
| 53 |
bug reports much more informative for us if they should occur in |
| 54 |
the newer version. |
| 55 |
|
| 56 |
b. Unnecessary re-searching is avoided |
| 57 |
|
| 58 |
In reiserfs_add_entry(), we don't need to scan a whole bit_string |
| 59 |
for non-zero bits to be able to use a smallest generation counter |
| 60 |
to avoid re-searching for the dir entry insertion point. |
| 61 |
|
| 62 |
4. "messages.diff" |
| 63 |
This adds missed error numbers to reiserfs error |
| 64 |
messages |
| 65 |
|
| 66 |
5. "objectid_map.diff" |
| 67 |
It adds a better debugging message when the oid map is corrupted, |
| 68 |
and (more importantly) fixes a macro in objectid.c to put parens |
| 69 |
around the stuff passed in. |
| 70 |
|
| 71 |
-- |
| 72 |
Thanks, |
| 73 |
Alex. |
| 74 |
|
| 75 |
------------------------------------------------------- |
Archives