[PATCH] fs/hfsplus: expand s_vhdr_buf size to avoid slab oob

[PATCH] fs/hfsplus: expand s_vhdr_buf size to avoid slab oob

[Edward AD]

The memory allocated to s_vhdr_buf in the function hfsplus-read_wrapper is 
too small, resulting in a slab out of bounds issue when copying data with 
copy_page_from_iter_atomic.

When allocating memory to s_vhdr_buf, take the maximum value between 
hfsplus_min_io_size(sb) and PAGE_SIZE to avoid similar issues.

Reported-and-tested-by: syzbot+4a2376bc62e59406c414@syzkaller.appspotmail.com
Signed-off-by: Edward AD <twuufnxlz@gmail.com>
---
 fs/hfsplus/wrapper.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/fs/hfsplus/wrapper.c b/fs/hfsplus/wrapper.c
index 0b791adf02e5..56bee8dbe532 100644
--- a/fs/hfsplus/wrapper.c
+++ b/fs/hfsplus/wrapper.c
@@ -163,7 +163,7 @@ int hfsplus_read_wrapper(struct super_block *sb)
 	struct hfsplus_sb_info *sbi = HFSPLUS_SB(sb);
 	struct hfsplus_wd wd;
 	sector_t part_start, part_size;
-	u32 blocksize;
+	u32 blocksize, bufsize;
 	int error = 0;
 
 	error = -EINVAL;
@@ -175,10 +175,11 @@ int hfsplus_read_wrapper(struct super_block *sb)
 		goto out;
 
 	error = -ENOMEM;
-	sbi->s_vhdr_buf = kmalloc(hfsplus_min_io_size(sb), GFP_KERNEL);
+	bufsize = max_t(u32, hfsplus_min_io_size(sb), PAGE_SIZE);
+	sbi->s_vhdr_buf = kmalloc(bufsize, GFP_KERNEL);
 	if (!sbi->s_vhdr_buf)
 		goto out;
-	sbi->s_backup_vhdr_buf = kmalloc(hfsplus_min_io_size(sb), GFP_KERNEL);
+	sbi->s_backup_vhdr_buf = kmalloc(bufsize, GFP_KERNEL);
 	if (!sbi->s_backup_vhdr_buf)
 		goto out_free_vhdr;
 
-- 
2.25.1

Re: [PATCH] fs/hfsplus: expand s_vhdr_buf size to avoid slab oob

[Dmitry Vyukov]

Hi Edward,

Was this patch lost? I don't see it merged anywhere.
This bug is still present in most kernels out there.

Re: [syzbot] [hfs?] [mm?] KASAN: slab-out-of-bounds Read in generic_perform_write

[syzbot]

syzbot suspects this issue was fixed by commit:

commit b432163ebd15a0fb74051949cb61456d6c55ccbd
Author: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Date:   Thu Jan 30 14:03:41 2025 +0000

    fs/ntfs3: Update inode->i_mapping->a_ops on compression state

bisection log:  https://syzkaller.appspot.com/x/bisect.txt?x=115c5b68580000
start commit:   fc033cf25e61 Linux 6.13-rc5
git tree:       upstream
kernel config:  https://syzkaller.appspot.com/x/.config?x=ba7cde9482d6bb6
dashboard link: https://syzkaller.appspot.com/bug?extid=4a2376bc62e59406c414
syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=1654aac4580000
C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=13ff26df980000

If the result looks correct, please mark the issue as fixed by replying with:

#syz fix: fs/ntfs3: Update inode->i_mapping->a_ops on compression state

For information about bisection process see: https://goo.gl/tpsmEJ#bisection