Hi
I have noticed that in ext2 (I haven't read the other FS code) call
mark_buffer_dirty() with spin_locks held. I think that it is better
to use the construction:
__mark_buffer_dirty()
<...>
spinlock_unlock();
balance_dirty();
I attached the patch for ext2 against test8-pre6. Stephen or Tytso,
do you mean to comment on that??
I did the patch because I found lock contention in the superblock lock
when running dbench 48.
Later, Juan.
PD. If you think that the approach is right I can't audit the rest of
the filesystems searching for that problem.
diff -urN --exclude-from=/home/lfcia/quintela/work/kernel/exclude
base/fs/ext2/ialloc.c working/fs/ext2/ialloc.c
--- base/fs/ext2/ialloc.c Wed Sep 6 00:37:31 2000
+++ working/fs/ext2/ialloc.c Fri Sep 8 01:54:42 2000
@@ -230,12 +230,12 @@
gdp->bg_used_dirs_count =
cpu_to_le16(le16_to_cpu(gdp->bg_used_dirs_count) - 1);
}
- mark_buffer_dirty(bh2);
+ __mark_buffer_dirty(bh2);
es->s_free_inodes_count =
cpu_to_le32(le32_to_cpu(es->s_free_inodes_count) + 1);
- mark_buffer_dirty(sb->u.ext2_sb.s_sbh);
+ __mark_buffer_dirty(sb->u.ext2_sb.s_sbh);
}
- mark_buffer_dirty(bh);
+ __mark_buffer_dirty(bh);
if (sb->s_flags & MS_SYNCHRONOUS) {
ll_rw_block (WRITE, 1, &bh);
wait_on_buffer (bh);
@@ -243,6 +243,7 @@
sb->s_dirt = 1;
error_return:
unlock_super (sb);
+ balance_dirty(bh->b_dev);
}
/*
@@ -388,7 +389,7 @@
"bit already set for inode %d", j);
goto repeat;
}
- mark_buffer_dirty(bh);
+ __mark_buffer_dirty(bh);
if (sb->s_flags & MS_SYNCHRONOUS) {
ll_rw_block (WRITE, 1, &bh);
wait_on_buffer (bh);
@@ -404,7 +405,7 @@
return NULL;
}
gdp->bg_free_inodes_count = 0;
- mark_buffer_dirty(bh2);
+ __mark_buffer_dirty(bh2);
}
goto repeat;
}
@@ -423,10 +424,10 @@
if (S_ISDIR(mode))
gdp->bg_used_dirs_count =
cpu_to_le16(le16_to_cpu(gdp->bg_used_dirs_count) + 1);
- mark_buffer_dirty(bh2);
+ __mark_buffer_dirty(bh2);
es->s_free_inodes_count =
cpu_to_le32(le32_to_cpu(es->s_free_inodes_count) - 1);
- mark_buffer_dirty(sb->u.ext2_sb.s_sbh);
+ __mark_buffer_dirty(sb->u.ext2_sb.s_sbh);
sb->s_dirt = 1;
inode->i_mode = mode;
inode->i_sb = sb;
@@ -464,6 +465,7 @@
mark_inode_dirty(inode);
unlock_super (sb);
+ balance_dirty(bh->b_dev);
if(DQUOT_ALLOC_INODE(sb, inode)) {
sb->dq_op->drop(inode);
inode->i_nlink = 0;
--
In theory, practice and theory are the same, but in practice they
are different -- Larry McVoy
-
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to [EMAIL PROTECTED]
