2021-11-19 19:28:56

by Kees Cook

Subject: [PATCH] hfsplus: Use struct_group_attr() for memcpy() region

In preparation for FORTIFY_SOURCE performing compile-time and run-time
field bounds checking for memset(), avoid intentionally writing across
neighboring fields.

Add struct_group() to mark the "info" region (containing struct DInfo
and struct DXInfo structs) in struct hfsplus_cat_folder and struct
hfsplus_cat_file that are written into directly, so the compiler can
correctly reason about the expected size of the writes.

"pahole" shows no size nor member offset changes to struct
hfsplus_cat_folder nor struct hfsplus_cat_file. "objdump -d" shows no
object code changes.

Signed-off-by: Kees Cook <[email protected]>
---
fs/hfsplus/hfsplus_raw.h | 12 ++++++++----
fs/hfsplus/xattr.c | 4 ++--
2 files changed, 10 insertions(+), 6 deletions(-)

diff --git a/fs/hfsplus/hfsplus_raw.h b/fs/hfsplus/hfsplus_raw.h
index 456e87aec7fd..68b4240c6191 100644
--- a/fs/hfsplus/hfsplus_raw.h
+++ b/fs/hfsplus/hfsplus_raw.h
@@ -260,8 +260,10 @@ struct hfsplus_cat_folder {
__be32 access_date;
__be32 backup_date;
struct hfsplus_perm permissions;
- struct DInfo user_info;
- struct DXInfo finder_info;
+ struct_group_attr(info, __packed,
+ struct DInfo user_info;
+ struct DXInfo finder_info;
+ );
__be32 text_encoding;
__be32 subfolders; /* Subfolder count in HFSX. Reserved in HFS+. */
} __packed;
@@ -294,8 +296,10 @@ struct hfsplus_cat_file {
__be32 access_date;
__be32 backup_date;
struct hfsplus_perm permissions;
- struct FInfo user_info;
- struct FXInfo finder_info;
+ struct_group_attr(info, __packed,
+ struct FInfo user_info;
+ struct FXInfo finder_info;
+ );
__be32 text_encoding;
u32 reserved2;

diff --git a/fs/hfsplus/xattr.c b/fs/hfsplus/xattr.c
index e2855ceefd39..49891b12c415 100644
--- a/fs/hfsplus/xattr.c
+++ b/fs/hfsplus/xattr.c
@@ -296,7 +296,7 @@ int __hfsplus_setxattr(struct inode *inode, const char *name,
sizeof(hfsplus_cat_entry));
if (be16_to_cpu(entry.type) == HFSPLUS_FOLDER) {
if (size == folder_finderinfo_len) {
- memcpy(&entry.folder.user_info, value,
+ memcpy(&entry.folder.info, value,
folder_finderinfo_len);
hfs_bnode_write(cat_fd.bnode, &entry,
cat_fd.entryoffset,
@@ -309,7 +309,7 @@ int __hfsplus_setxattr(struct inode *inode, const char *name,
}
} else if (be16_to_cpu(entry.type) == HFSPLUS_FILE) {
if (size == file_finderinfo_len) {
- memcpy(&entry.file.user_info, value,
+ memcpy(&entry.file.info, value,
file_finderinfo_len);
hfs_bnode_write(cat_fd.bnode, &entry,
cat_fd.entryoffset,
--
2.30.2

2021-11-23 14:28:04

by Christian Brauner

Subject: Re: [PATCH] hfsplus: Use struct_group_attr() for memcpy() region

On Fri, Nov 19, 2021 at 11:28:51AM -0800, Kees Cook wrote:
> In preparation for FORTIFY_SOURCE performing compile-time and run-time
> field bounds checking for memset(), avoid intentionally writing across
> neighboring fields.
>
> Add struct_group() to mark the "info" region (containing struct DInfo
> and struct DXInfo structs) in struct hfsplus_cat_folder and struct
> hfsplus_cat_file that are written into directly, so the compiler can
> correctly reason about the expected size of the writes.
>
> "pahole" shows no size nor member offset changes to struct
> hfsplus_cat_folder nor struct hfsplus_cat_file. "objdump -d" shows no
> object code changes.
>
> Signed-off-by: Kees Cook <[email protected]>
> ---

Looks good.
Acked-by: Christian Brauner <[email protected]>