Ross Vandegrift wrote:
>
> > > Are there any HOWTOs that anybody knows about on how to do a RAID-5 with
> > > LVM and a Reiser-FS filesystem? Are there are reasons not to do this? If
> > > I can dynamically resize (hopefully using the raidreconf with no
> > > problems) the array does LVM get me anything?
>
> There's currently an issue with 2.4 kernels that causes read starvation while
> heavy writes are going on. As a consequence, if your array is rebuilding while
> you have a journal replay going on, your boot will stall at replaying the
> journal for a very long time.
Yeah also bitmap blocks reading, especially if you formated a large volume...
This is a fix against 2.4.17. (still not in kernel). Let us know, if you have any
problem (or if you want a patch against another kernel).
Edward.
diff -u -r linux/fs/reiserfs/journal.c linux-patched/fs/reiserfs/journal.c
--- linux/fs/reiserfs/journal.c Fri Dec 21 17:42:03 2001
+++ linux-patched/fs/reiserfs/journal.c Fri Feb 1 14:27:51 2002
@@ -71,7 +71,9 @@
static DECLARE_WAIT_QUEUE_HEAD(reiserfs_commit_thread_done) ;
DECLARE_TASK_QUEUE(reiserfs_commit_thread_tq) ;
-#define JOURNAL_TRANS_HALF 1018 /* must be correct to keep the desc and commit structs at 4k */
+#define JOURNAL_TRANS_HALF 1018 /* must be correct to keep the desc and commit
+ structs at 4k */
+#define BUFNR 64 /*read ahead */
/* cnode stat bits. Move these into reiserfs_fs.h */
@@ -1593,6 +1595,41 @@
**
** On exit, it sets things up so the first transaction will work correctly.
*/
+struct buffer_head * reiserfs_breada (kdev_t dev, int block, int bufsize,
+ unsigned int max_block)
+{
+ struct buffer_head * bhlist[BUFNR];
+ unsigned int blocks = BUFNR;
+ struct buffer_head * bh;
+ int i, j;
+
+ bh = getblk (dev, block, bufsize);
+ if (buffer_uptodate (bh))
+ return (bh);
+
+ if (block + BUFNR > max_block) {
+ blocks = max_block - block;
+ }
+ bhlist[0] = bh;
+ j = 1;
+ for (i = 1; i < blocks; i++) {
+ bh = getblk (dev, block + i, bufsize);
+ if (buffer_uptodate (bh)) {
+ brelse (bh);
+ break;
+ }
+ else bhlist[j++] = bh;
+ }
+ ll_rw_block (READ, j, bhlist);
+ for(i = 1; i < j; i++)
+ brelse (bhlist[i]);
+ bh = bhlist[0];
+ wait_on_buffer (bh);
+ if (buffer_uptodate (bh))
+ return bh;
+ brelse (bh);
+ return NULL;
+}
static int journal_read(struct super_block *p_s_sb) {
struct reiserfs_journal_desc *desc ;
unsigned long last_flush_trans_id = 0 ;
@@ -1663,7 +1700,8 @@
** all the valid transactions, and pick out the oldest.
*/
while(continue_replay && cur_dblock < (reiserfs_get_journal_block(p_s_sb) + JOURNAL_BLOCK_COUNT))
{
- d_bh = bread(p_s_sb->s_dev, cur_dblock, p_s_sb->s_blocksize) ;
+ d_bh = reiserfs_breada(p_s_sb->s_dev, cur_dblock, p_s_sb->s_blocksize,
+ reiserfs_get_journal_block(p_s_sb) + JOURNAL_BLOCK_COUNT) ;
ret = journal_transaction_is_valid(p_s_sb, d_bh, &oldest_invalid_trans_id, &newest_mount_id) ;
if (ret == 1) {
desc = (struct reiserfs_journal_desc *)d_bh->b_data ;
diff -u -r linux/fs/reiserfs/super.c linux-patched/fs/reiserfs/super.c
--- linux/fs/reiserfs/super.c Fri Dec 21 17:42:03 2001
+++ linux-patched/fs/reiserfs/super.c Fri Feb 1 14:42:26 2002
@@ -558,28 +558,30 @@
static int read_bitmaps (struct super_block * s)
{
- int i, bmp, dl ;
- struct reiserfs_super_block * rs = SB_DISK_SUPER_BLOCK(s);
+ int i, bmp;
- SB_AP_BITMAP (s) = reiserfs_kmalloc (sizeof (struct buffer_head *) * sb_bmap_nr(rs), GFP_NOFS,
s);
+ SB_AP_BITMAP (s) = reiserfs_kmalloc (sizeof (struct buffer_head *) * SB_BMAP_NR(s), GFP_NOFS,
s);
if (SB_AP_BITMAP (s) == 0)
- return 1;
- memset (SB_AP_BITMAP (s), 0, sizeof (struct buffer_head *) * sb_bmap_nr(rs));
-
- /* reiserfs leaves the first 64k unused so that any partition
- labeling scheme currently used will have enough space. Then we
- need one block for the super. -Hans */
- bmp = (REISERFS_DISK_OFFSET_IN_BYTES / s->s_blocksize) + 1; /* first of bitmap blocks */
- SB_AP_BITMAP (s)[0] = reiserfs_bread (s, bmp, s->s_blocksize);
- if(!SB_AP_BITMAP(s)[0])
- return 1;
- for (i = 1, bmp = dl = s->s_blocksize * 8; i < sb_bmap_nr(rs); i ++) {
- SB_AP_BITMAP (s)[i] = reiserfs_bread (s, bmp, s->s_blocksize);
- if (!SB_AP_BITMAP (s)[i])
- return 1;
- bmp += dl;
+ return 1;
+ for (i = 0, bmp = REISERFS_DISK_OFFSET_IN_BYTES / s->s_blocksize + 1;
+ i < SB_BMAP_NR(s); i++, bmp = s->s_blocksize * 8 * i) {
+ SB_AP_BITMAP (s)[i] = getblk (s->s_dev, bmp, s->s_blocksize);
+ if (!buffer_uptodate(SB_AP_BITMAP(s)[i]))
+ ll_rw_block(READ, 1, SB_AP_BITMAP(s) + i);
}
-
+ for (i = 0; i < SB_BMAP_NR(s); i++) {
+ wait_on_buffer(SB_AP_BITMAP (s)[i]);
+ if (!buffer_uptodate(SB_AP_BITMAP(s)[i])) {
+ reiserfs_warning("sh-2029: reiserfs read_bitmaps: "
+ "bitmap block (#%lu) reading failed\n",
+ SB_AP_BITMAP(s)[i]->b_blocknr);
+ for (i = 0; i < SB_BMAP_NR(s); i++)
+ brelse(SB_AP_BITMAP(s)[i]);
+ reiserfs_kfree(SB_AP_BITMAP(s), sizeof(struct buffer_head *) * SB_BMAP_NR(s), s);
+ SB_AP_BITMAP(s) = NULL;
+ return 1;
+ }
+ }
return 0;
}
> This isn't just a problem with RAID/ReiserFS,
> it's more generalized than that. Also, I do believe there is a path somewhere
> that fixes the issue.
>
> Ross Vandegrift
> ross@willow.seitz.com
> -
> To unsubscribe from this list: send the line "unsubscribe linux-raid" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at http://vger.kernel.org/majordomo-info.html
-
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
