From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Fri, 28 Mar 2025 15:29:04 +0000 (-0400)
Subject: bcachefs: Don't return 0 size holes from bch2_seek_hole()
X-Git-Tag: v6.15-rc1~83^2~21
X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=af3d4c276a9171d142044effb2c43fb6df92a787;p=thirdparty%2Flinux.git

bcachefs: Don't return 0 size holes from bch2_seek_hole()

The hole we find in the btree might be fully dirty in the page cache. If
so, keep searching.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---

diff --git a/fs/bcachefs/fs-io.c b/fs/bcachefs/fs-io.c
index cc366786f0f9a..c80ed3a54e709 100644
--- a/fs/bcachefs/fs-io.c
+++ b/fs/bcachefs/fs-io.c
@@ -1008,10 +1008,19 @@ static loff_t bch2_seek_hole(struct file *file, u64 offset)
 					? MAX_LFS_FILESIZE
 					: k.k->p.offset << 9;
 
+				/*
+				 * Found a hole in the btree, now make sure it's
+				 * a hole in the pagecache. We might have to
+				 * keep searching if this hole is entirely dirty
+				 * in the page cache:
+				 */
 				bch2_trans_unlock(trans);
-				next_hole = bch2_seek_pagecache_hole(&inode->v,
-						start_offset, end_offset, 0, false);
-				break;
+				loff_t pagecache_hole = bch2_seek_pagecache_hole(&inode->v,
+								start_offset, end_offset, 0, false);
+				if (pagecache_hole < end_offset) {
+					next_hole = pagecache_hole;
+					break;
+				}
 			} else {
 				offset = max(offset, bkey_start_offset(k.k) << 9);
 			}