From 355eac79d2d0ed63e452e8c2c9d8482bf7025abf Mon Sep 17 00:00:00 2001
From: Andy Grover <agrover@versity.com>
Date: Sat, 16 Jan 2021 12:18:16 -0800
Subject: [PATCH] Retry if transaction cannot alloc for fallocate or write

Add a new distinguishable return value (ENOBUFS) from allocator for if
the transaction cannot alloc space. This doesn't mean the filesystem is
full -- opening a new transaction may result in forward progress.

Alter fallocate and get_blocks code to check for this err val and retry
with a new transaction. Handling actual ENOSPC can still happen, of
course.

Add counter called "alloc_trans_retry" and increment it from both spots.

Signed-off-by: Andy Grover <agrover@versity.com>
[zab@versity.com: fixed up write_begin error paths]
---
 kmod/src/alloc.c    |  7 ++++++-
 kmod/src/counters.h |  2 ++
 kmod/src/data.c     | 28 ++++++++++++++++++++--------
 3 files changed, 28 insertions(+), 9 deletions(-)

diff --git a/kmod/src/alloc.c b/kmod/src/alloc.c
index 1d370dda..530e261f 100644
--- a/kmod/src/alloc.c
+++ b/kmod/src/alloc.c
@@ -770,8 +770,13 @@ int scoutfs_alloc_data(struct super_block *sb, struct scoutfs_alloc *alloc,
 	ret = 0;
 out:
 	if (ret < 0) {
+		/*
+		 * Special retval meaning there wasn't space to alloc from
+		 * this txn. Doesn't mean filesystem is completely full.
+		 * Maybe upper layers want to try again.
+		 */
 		if (ret == -ENOENT)
-			ret = -ENOSPC;
+			ret = -ENOBUFS;
 		*blkno_ret = 0;
 		*count_ret = 0;
 	} else {
diff --git a/kmod/src/counters.h b/kmod/src/counters.h
index 6c470e58..819cf67c 100644
--- a/kmod/src/counters.h
+++ b/kmod/src/counters.h
@@ -58,6 +58,8 @@
 	EXPAND_COUNTER(corrupt_symlink_inode_size)		\
 	EXPAND_COUNTER(corrupt_symlink_missing_item)		\
 	EXPAND_COUNTER(corrupt_symlink_not_null_term)		\
+	EXPAND_COUNTER(data_fallocate_enobufs_retry)		\
+	EXPAND_COUNTER(data_write_begin_enobufs_retry)		\
 	EXPAND_COUNTER(dentry_revalidate_error)			\
 	EXPAND_COUNTER(dentry_revalidate_invalid)		\
 	EXPAND_COUNTER(dentry_revalidate_locked)		\
diff --git a/kmod/src/data.c b/kmod/src/data.c
index 73b29e5a..33061d8d 100644
--- a/kmod/src/data.c
+++ b/kmod/src/data.c
@@ -751,6 +751,7 @@ static int scoutfs_write_begin(struct file *file,
 		goto out;
 	}
 
+retry:
 	do {
 		ret = scoutfs_inode_index_start(sb, &ind_seq) ?:
 		      scoutfs_inode_index_prepare(sb, &wbd->ind_locks, inode,
@@ -765,17 +766,22 @@ static int scoutfs_write_begin(struct file *file,
 	flags |= AOP_FLAG_NOFS;
 
 	/* generic write_end updates i_size and calls dirty_inode */
-	ret = scoutfs_dirty_inode_item(inode, wbd->lock);
-	if (ret == 0)
-		ret = block_write_begin(mapping, pos, len, flags, pagep,
-					scoutfs_get_block_write);
-	if (ret)
+	ret = scoutfs_dirty_inode_item(inode, wbd->lock) ?:
+	      block_write_begin(mapping, pos, len, flags, pagep,
+				scoutfs_get_block_write);
+	if (ret < 0) {
 		scoutfs_release_trans(sb);
-out:
-	if (ret) {
 		scoutfs_inode_index_unlock(sb, &wbd->ind_locks);
-		kfree(wbd);
+		if (ret == -ENOBUFS) {
+			/* Retry with a new transaction. */
+			scoutfs_inc_counter(sb, data_write_begin_enobufs_retry);
+			goto retry;
+		}
 	}
+
+out:
+	if (ret < 0)
+		kfree(wbd);
         return ret;
 }
 
@@ -1022,6 +1028,12 @@ long scoutfs_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
 		scoutfs_release_trans(sb);
 		scoutfs_inode_index_unlock(sb, &ind_locks);
 
+		/* txn couldn't meet the request. Let's try with a new txn */
+		if (ret == -ENOBUFS) {
+			scoutfs_inc_counter(sb, data_fallocate_enobufs_retry);
+			continue;
+		}
+
 		if (ret <= 0)
 			goto out;