[PATCH] sched: ext3: fix scheduling latencies in ext3

The attached patch fixes long scheduling latencies in the ext3 code, and it
also cleans up the existing lock-break functionality to use the new
primitives.

This patch has been in the -VP patchset for quite some time.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

fs/ext3/balloc.c

@@ -378,6 +378,11 @@ void ext3_free_blocks_sb(handle_t *handle, struct super_block *sb,
 		}
 		jbd_lock_bh_state(bitmap_bh);
 #endif
+		if (need_resched()) {
+			jbd_unlock_bh_state(bitmap_bh);
+			cond_resched();
+			jbd_lock_bh_state(bitmap_bh);
+		}
 		/* @@@ This prevents newly-allocated data from being
 		 * freed and then reallocated within the same
 		 * transaction. 

fs/ext3/ialloc.c

@@ -737,6 +737,7 @@ unsigned long ext3_count_free_inodes (struct super_block * sb)
 		if (!gdp)
 			continue;
 		desc_count += le16_to_cpu(gdp->bg_free_inodes_count);
+		cond_resched();
 	}
 	return desc_count;
 #endif

fs/ext3/namei.c

@@ -675,6 +675,7 @@ static int dx_make_map (struct ext3_dir_entry_2 *de, int size,
 			map_tail->hash = h.hash;
 			map_tail->offs = (u32) ((char *) de - base);
 			count++;
+			cond_resched();
 		}
 		/* XXX: do we need to check rec_len == 0 case? -Chris */
 		de = (struct ext3_dir_entry_2 *) ((char *) de + le16_to_cpu(de->rec_len));

fs/ext3/super.c

@@ -1266,6 +1266,8 @@ static int ext3_fill_super (struct super_block *sb, void *data, int silent)
 	sbi->s_resuid = EXT3_DEF_RESUID;
 	sbi->s_resgid = EXT3_DEF_RESGID;
 
+	unlock_kernel();
+
 	blocksize = sb_min_blocksize(sb, EXT3_MIN_BLOCK_SIZE);
 	if (!blocksize) {
 		printk(KERN_ERR "EXT3-fs: unable to set blocksize\n");
@@ -1622,6 +1624,7 @@ static int ext3_fill_super (struct super_block *sb, void *data, int silent)
 	percpu_counter_mod(&sbi->s_dirs_counter,
 		ext3_count_dirs(sb));
 
+	lock_kernel();
 	return 0;
 
 cantfind_ext3:
@@ -1646,6 +1649,7 @@ static int ext3_fill_super (struct super_block *sb, void *data, int silent)
 out_fail:
 	sb->s_fs_info = NULL;
 	kfree(sbi);
+	lock_kernel();
 	return -EINVAL;
 }
 
@@ -2170,9 +2174,11 @@ int ext3_statfs (struct super_block * sb, struct kstatfs * buf)
 		 * block group descriptors.  If the sparse superblocks
 		 * feature is turned on, then not all groups have this.
 		 */
-		for (i = 0; i < ngroups; i++)
+		for (i = 0; i < ngroups; i++) {
 			overhead += ext3_bg_has_super(sb, i) +
 				ext3_bg_num_gdb(sb, i);
+			cond_resched();
+		}
 
 		/*
 		 * Every block group has an inode bitmap, a block

fs/jbd/checkpoint.c

@@ -333,6 +333,10 @@ int log_do_checkpoint(journal_t *journal)
 				break;
 			}
 			retry = __flush_buffer(journal, jh, bhs, &batch_count, &drop_count);
+			if (cond_resched_lock(&journal->j_list_lock)) {
+				retry = 1;
+				break;
+			}
 		} while (jh != last_jh && !retry);
 
 		if (batch_count)
@@ -487,6 +491,14 @@ int __journal_clean_checkpoint_list(journal_t *journal)
 				/* Use trylock because of the ranknig */
 				if (jbd_trylock_bh_state(jh2bh(jh)))
 					ret += __try_to_free_cp_buf(jh);
+				/*
+				 * This function only frees up some memory
+				 * if possible so we dont have an obligation
+				 * to finish processing. Bail out if preemption
+				 * requested:
+				 */
+				if (need_resched())
+					goto out;
 			} while (jh != last_jh);
 		}
 	} while (transaction != last_transaction);

fs/jbd/commit.c

@@ -262,7 +262,7 @@ void journal_commit_transaction(journal_t *journal)
 			__journal_file_buffer(jh, commit_transaction,
 						BJ_Locked);
 			jbd_unlock_bh_state(bh);
-			if (need_resched()) {
+			if (lock_need_resched(&journal->j_list_lock)) {
 				spin_unlock(&journal->j_list_lock);
 				goto write_out_data;
 			}
@@ -288,7 +288,7 @@ void journal_commit_transaction(journal_t *journal)
 				jbd_unlock_bh_state(bh);
 				journal_remove_journal_head(bh);
 				put_bh(bh);
-				if (need_resched()) {
+				if (lock_need_resched(&journal->j_list_lock)) {
 					spin_unlock(&journal->j_list_lock);
 					goto write_out_data;
 				}
@@ -333,11 +333,7 @@ void journal_commit_transaction(journal_t *journal)
 			jbd_unlock_bh_state(bh);
 		}
 		put_bh(bh);
-		if (need_resched()) {
-			spin_unlock(&journal->j_list_lock);
-			cond_resched();
-			spin_lock(&journal->j_list_lock);
-		}
+		cond_resched_lock(&journal->j_list_lock);
 	}
 	spin_unlock(&journal->j_list_lock);
 
@@ -545,6 +541,8 @@ void journal_commit_transaction(journal_t *journal)
 			wait_on_buffer(bh);
 			goto wait_for_iobuf;
 		}
+		if (cond_resched())
+			goto wait_for_iobuf;
 
 		if (unlikely(!buffer_uptodate(bh)))
 			err = -EIO;
@@ -599,6 +597,8 @@ void journal_commit_transaction(journal_t *journal)
 			wait_on_buffer(bh);
 			goto wait_for_ctlbuf;
 		}
+		if (cond_resched())
+			goto wait_for_ctlbuf;
 
 		if (unlikely(!buffer_uptodate(bh)))
 			err = -EIO;
@@ -695,6 +695,7 @@ void journal_commit_transaction(journal_t *journal)
 	J_ASSERT(commit_transaction->t_shadow_list == NULL);
 	J_ASSERT(commit_transaction->t_log_list == NULL);
 
+restart_loop:
 	while (commit_transaction->t_forget) {
 		transaction_t *cp_transaction;
 		struct buffer_head *bh;
@@ -768,6 +769,8 @@ void journal_commit_transaction(journal_t *journal)
 			release_buffer_page(bh);
 		}
 		spin_unlock(&journal->j_list_lock);
+		if (cond_resched())
+			goto restart_loop;
 	}
 
 	/* Done with this transaction! */

fs/jbd/recovery.c

@@ -354,6 +354,8 @@ static int do_one_pass(journal_t *journal,
 		struct buffer_head *	obh;
 		struct buffer_head *	nbh;
 
+		cond_resched();		/* We're under lock_kernel() */
+
 		/* If we already know where to stop the log traversal,
 		 * check right now that we haven't gone past the end of
 		 * the log. */