linux-kernel - [PATCH 1/9] fs: make block_read_full

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20241125114419.903270-2-yi.zhang@huaweicloud.com>
Date: Mon, 25 Nov 2024 19:44:11 +0800
From: Zhang Yi <yi.zhang@...weicloud.com>
To: linux-ext4@...r.kernel.org
Cc: linux-fsdevel@...r.kernel.org,
	linux-kernel@...r.kernel.org,
	tytso@....edu,
	willy@...radead.org,
	adilger.kernel@...ger.ca,
	jack@...e.cz,
	brauner@...nel.org,
	yi.zhang@...wei.com,
	yi.zhang@...weicloud.com,
	chengzhihao1@...wei.com,
	yukuai3@...wei.com,
	yangerkun@...wei.com
Subject: [PATCH 1/9] fs: make block_read_full_folio() support large folio

From: Zhang Yi <yi.zhang@...wei.com>

block_read_full_folio() uses an on-stack array to save the un-uptodated
buffers. This approach is acceptable for small folios, as there are not
too many buffers in a single folio. However, if the folio is a large
folio, the number of buffers could reach up to 2,000 for a 2MB folio
size with a 1KB block size. Therefore, we cannot use the 'arr' array for
large folios, instead, we should iterate through the buffers using 'bh =
bh->b_this_page'.

Signed-off-by: Zhang Yi <yi.zhang@...wei.com>
---
 fs/buffer.c | 34 +++++++++++++++++-----------------
 1 file changed, 17 insertions(+), 17 deletions(-)

diff --git a/fs/buffer.c b/fs/buffer.c
index 1fc9a50def0b..54e608c8912e 100644
--- a/fs/buffer.c
+++ b/fs/buffer.c
@@ -2366,7 +2366,7 @@ int block_read_full_folio(struct folio *folio, get_block_t *get_block)
 {
 	struct inode *inode = folio->mapping->host;
 	sector_t iblock, lblock;
-	struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
+	struct buffer_head *bh, *head;
 	size_t blocksize;
 	int nr, i;
 	int fully_mapped = 1;
@@ -2377,8 +2377,6 @@ int block_read_full_folio(struct folio *folio, get_block_t *get_block)
 	if (IS_ENABLED(CONFIG_FS_VERITY) && IS_VERITY(inode))
 		limit = inode->i_sb->s_maxbytes;
 
-	VM_BUG_ON_FOLIO(folio_test_large(folio), folio);
-
 	head = folio_create_buffers(folio, inode, 0);
 	blocksize = head->b_size;
 
@@ -2416,7 +2414,10 @@ int block_read_full_folio(struct folio *folio, get_block_t *get_block)
 			if (buffer_uptodate(bh))
 				continue;
 		}
-		arr[nr++] = bh;
+		/* lock the buffer */
+		lock_buffer(bh);
+		mark_buffer_async_read(bh);
+		nr++;
 	} while (i++, iblock++, (bh = bh->b_this_page) != head);
 
 	if (fully_mapped)
@@ -2431,25 +2432,24 @@ int block_read_full_folio(struct folio *folio, get_block_t *get_block)
 		return 0;
 	}
 
-	/* Stage two: lock the buffers */
-	for (i = 0; i < nr; i++) {
-		bh = arr[i];
-		lock_buffer(bh);
-		mark_buffer_async_read(bh);
-	}
-
 	/*
-	 * Stage 3: start the IO.  Check for uptodateness
-	 * inside the buffer lock in case another process reading
-	 * the underlying blockdev brought it uptodate (the sct fix).
+	 * Start the IO.  Check for uptodateness inside the buffer lock
+	 * in case another process reading the underlying blockdev brought
+	 * it uptodate (the sct fix).
 	 */
-	for (i = 0; i < nr; i++) {
-		bh = arr[i];
+	bh = head;
+	do {
+		if (!buffer_async_read(bh))
+			continue;
+
 		if (buffer_uptodate(bh))
 			end_buffer_async_read(bh, 1);
 		else
 			submit_bh(REQ_OP_READ, bh);
-	}
+
+		nr--;
+	} while (nr && (bh = bh->b_this_page) != head);
+
 	return 0;
 }
 EXPORT_SYMBOL(block_read_full_folio);
-- 
2.46.1