All of lore.kernel.org
 help / color / mirror / Atom feed
From: Kyle Zeng <kylebot@openai.com>
To: Viacheslav Dubeyko <slava@dubeyko.com>
Cc: linux-fsdevel@vger.kernel.org, outbounddisclosures@openai.com
Subject: Re: [PATCH v2] hfsplus: validate thread record before delete key rebuild
Date: Tue, 7 Jul 2026 11:48:26 -0700	[thread overview]
Message-ID: <ak1J-ouL-DjO6pa3@com-75606> (raw)
In-Reply-To: <deaf9198e4e937310f92001628e48bb9d8b88844.camel@dubeyko.com>

[-- Attachment #1: Type: text/plain, Size: 156 bytes --]

Hi Slava,

A new patch following your suggestion is attached. Please let me know if
you prefer to receive the new patch in a separate thread.

Thanks,
Kyle

[-- Attachment #2: 0001-hfsplus-validate-thread-record-before-delete-key-rebuild.patch --]
[-- Type: text/plain, Size: 4659 bytes --]

From 25375348bf897058e2c47806bc9cf1d3dd6b2647 Mon Sep 17 00:00:00 2001
From: Kyle Zeng <kylebot@openai.com>
Date: Tue, 7 Jul 2026 18:37:32 +0000
Subject: [PATCH v3] hfsplus: validate thread record before delete key rebuild

hfsplus_delete_cat() is called with str == NULL when the last open
reference to an unlinked HFS+ hardlink backing inode is closed. In that
case, the function finds the catalog thread by CNID and rebuilds the
catalog key from thread.nodeName.

That reconstruction path reads thread.nodeName.length directly from the
catalog B-tree into fd.search_key and then copies length * 2 bytes into
fd.search_key->cat.name.unicode. It does not first check that the found
record is a thread record, that the record size matches the thread name
length, or that nodeName.length fits in the fixed HFS+ catalog key
buffer.

A corrupted image can therefore provide an oversized thread name length
and make hfs_bnode_read() write past the catalog search-key allocation.

Read the CNID record through hfsplus_brec_read_cat(), which performs the
hfs_brec_find() lookup via hfs_brec_read() before validating the catalog
record. Reject non-thread records before building the delete key from
the validated thread name.

Move the HFS+ catalog thread name bound into hfsplus_brec_read_cat() so
it is checked together with the record-size consistency validation, and
factor the thread-record-type check into a shared HFS/HFS+ helper.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Assisted-by: Codex:gpt-5.5
Signed-off-by: Kyle Zeng <kylebot@openai.com>
---
 fs/hfsplus/bfind.c         |  5 +++++
 fs/hfsplus/catalog.c       | 30 ++++++++++++------------------
 include/linux/hfs_common.h |  5 +++++
 3 files changed, 22 insertions(+), 18 deletions(-)

diff --git a/fs/hfsplus/bfind.c b/fs/hfsplus/bfind.c
index 9a55fa6d5294..417c3014f4c7 100644
--- a/fs/hfsplus/bfind.c
+++ b/fs/hfsplus/bfind.c
@@ -322,6 +322,11 @@ int hfsplus_brec_read_cat(struct hfs_find_data *fd, hfsplus_cat_entry *entry)
 			pr_err("thread record too short (got %u)\n", fd->entrylength);
 			return -EIO;
 		}
+		if (be16_to_cpu(entry->thread.nodeName.length) >
+		    HFSPLUS_MAX_STRLEN) {
+			pr_err("catalog name length corrupted\n");
+			return -EIO;
+		}
 		expected_size = hfsplus_cat_thread_size(&entry->thread);
 		break;
 	default:
diff --git a/fs/hfsplus/catalog.c b/fs/hfsplus/catalog.c
index 776ce36cf076..1a4b0beae135 100644
--- a/fs/hfsplus/catalog.c
+++ b/fs/hfsplus/catalog.c
@@ -204,16 +204,11 @@ int hfsplus_find_cat(struct super_block *sb, u32 cnid,
 		return err;
 
 	type = be16_to_cpu(tmp.type);
-	if (type != HFSPLUS_FOLDER_THREAD && type != HFSPLUS_FILE_THREAD) {
+	if (!is_hfs_thread_record_type(type)) {
 		pr_err("found bad thread record in catalog\n");
 		return -EIO;
 	}
 
-	if (be16_to_cpu(tmp.thread.nodeName.length) > 255) {
-		pr_err("catalog name length corrupted\n");
-		return -EIO;
-	}
-
 	hfsplus_cat_build_key_uni(fd->search_key,
 		be32_to_cpu(tmp.thread.parentID),
 		&tmp.thread.nodeName);
@@ -350,23 +345,22 @@ int hfsplus_delete_cat(u32 cnid, struct inode *dir, const struct qstr *str)
 		goto out;
 
 	if (!str) {
-		int len;
+		hfsplus_cat_entry entry = {0};
 
 		hfsplus_cat_build_key_with_cnid(sb, fd.search_key, cnid);
-		err = hfs_brec_find(&fd, hfs_find_rec_by_key);
+		err = hfsplus_brec_read_cat(&fd, &entry);
 		if (err)
 			goto out;
 
-		off = fd.entryoffset +
-			offsetof(struct hfsplus_cat_thread, nodeName);
-		fd.search_key->cat.parent = cpu_to_be32(dir->i_ino);
-		hfs_bnode_read(fd.bnode,
-			&fd.search_key->cat.name.length, off, 2);
-		len = be16_to_cpu(fd.search_key->cat.name.length) * 2;
-		hfs_bnode_read(fd.bnode,
-			&fd.search_key->cat.name.unicode,
-			off + 2, len);
-		fd.search_key->key_len = cpu_to_be16(6 + len);
+		type = be16_to_cpu(entry.type);
+		if (!is_hfs_thread_record_type(type)) {
+			pr_err("found bad thread record in catalog\n");
+			err = -EIO;
+			goto out;
+		}
+
+		hfsplus_cat_build_key_uni(fd.search_key, dir->i_ino,
+					  &entry.thread.nodeName);
 	} else {
 		err = hfsplus_cat_build_key(sb, fd.search_key, dir->i_ino, str);
 		if (unlikely(err))
diff --git a/include/linux/hfs_common.h b/include/linux/hfs_common.h
index 45fb4c9ff9f5..c4862166efc9 100644
--- a/include/linux/hfs_common.h
+++ b/include/linux/hfs_common.h
@@ -92,6 +92,11 @@
 #define HFS_CDR_THD		0x03	/* folder (directory) thread */
 #define HFS_CDR_FTH		0x04	/* file thread */
 
+static inline bool is_hfs_thread_record_type(u16 type)
+{
+	return type == HFS_CDR_THD || type == HFS_CDR_FTH;
+}
+
 /* legal values for hfs_ext_key.FkType and hfs_file.fork */
 #define HFS_FK_DATA		0x00
 #define HFS_FK_RSRC		0xFF
-- 
2.54.0


  parent reply	other threads:[~2026-07-07 18:48 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-07-01  6:35 [PATCH v2] hfsplus: validate thread record before delete key rebuild Kyle Zeng
2026-07-03  0:13 ` Viacheslav Dubeyko
2026-07-03 19:08   ` Viacheslav Dubeyko
2026-07-07 18:48   ` Kyle Zeng [this message]
2026-07-07 18:53     ` Viacheslav Dubeyko

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=ak1J-ouL-DjO6pa3@com-75606 \
    --to=kylebot@openai.com \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=outbounddisclosures@openai.com \
    --cc=slava@dubeyko.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.