Re: [PATCH] can: etas_es58x: Replace 0-element raw_msg array

On Wed. 18 Aug 2021 at 15:48, Kees Cook [off-list ref] wrote:

On Wed, Aug 18, 2021 at 02:13:51PM +0900, Vincent MAILHOL wrote:

quoted

On Wed. 18 Aug 2021 at 12:40, Kees Cook [off-list ref] wrote:

quoted

While raw_msg isn't a fixed size, it does have a maximum size. Adjust the
struct to represent this and avoid the following warning when building
with -Wzero-length-bounds:

drivers/net/can/usb/etas_es58x/es58x_fd.c: In function 'es58x_fd_tx_can_msg':
drivers/net/can/usb/etas_es58x/es58x_fd.c:360:35: warning: array subscript 65535 is outside the bounds of an interior zero-length array 'u8[0]' {aka 'unsigned char[]'} [-Wzero-length-bounds]
  360 |  tx_can_msg = (typeof(tx_can_msg))&es58x_fd_urb_cmd->raw_msg[msg_len];
      |                                   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from drivers/net/can/usb/etas_es58x/es58x_core.h:22,
                 from drivers/net/can/usb/etas_es58x/es58x_fd.c:17:
drivers/net/can/usb/etas_es58x/es58x_fd.h:231:6: note: while referencing 'raw_msg'
  231 |   u8 raw_msg[0];
      |      ^~~~~~~

Cc: Wolfgang Grandegger <redacted>
Cc: Marc Kleine-Budde <mkl@pengutronix.de>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Arunachalam Santhanam <redacted>
Cc: Vincent Mailhol <redacted>
Cc: linux-can@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: Kees Cook <redacted>
---
 drivers/net/can/usb/etas_es58x/es581_4.h  | 2 +-
 drivers/net/can/usb/etas_es58x/es58x_fd.h | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/can/usb/etas_es58x/es581_4.h b/drivers/net/can/usb/etas_es58x/es581_4.h
index 4bc60a6df697..af38c4938859 100644
--- a/drivers/net/can/usb/etas_es58x/es581_4.h
+++ b/drivers/net/can/usb/etas_es58x/es581_4.h

@@ -192,7 +192,7 @@ struct es581_4_urb_cmd {
                struct es581_4_rx_cmd_ret rx_cmd_ret;
                __le64 timestamp;
                u8 rx_cmd_ret_u8;
-               u8 raw_msg[0];
+               u8 raw_msg[USHRT_MAX];
        } __packed;

        __le16 reserved_for_crc16_do_not_use;

diff --git a/drivers/net/can/usb/etas_es58x/es58x_fd.h b/drivers/net/can/usb/etas_es58x/es58x_fd.h
index ee18a87e40c0..e0319b8358ef 100644
--- a/drivers/net/can/usb/etas_es58x/es58x_fd.h
+++ b/drivers/net/can/usb/etas_es58x/es58x_fd.h

@@ -228,7 +228,7 @@ struct es58x_fd_urb_cmd {
                struct es58x_fd_tx_ack_msg tx_ack_msg;
                __le64 timestamp;
                __le32 rx_cmd_ret_le32;
-               u8 raw_msg[0];
+               u8 raw_msg[USHRT_MAX];
        } __packed;

        __le16 reserved_for_crc16_do_not_use;
--

2.30.2

raw_msg is part of a union so its maximum size is implicitly the
biggest size of the other member of that union:

Yup, understood. See below...

quoted

| struct es58x_fd_urb_cmd {
|     __le16 SOF;
|    u8 cmd_type;
|    u8 cmd_id;
|    u8 channel_idx;
|    __le16 msg_len;
|
|    union {
|        struct es58x_fd_tx_conf_msg tx_conf_msg;
|        u8 tx_can_msg_buf[ES58X_FD_TX_BULK_MAX * ES58X_FD_CANFD_TX_LEN];
|        u8 rx_can_msg_buf[ES58X_FD_RX_BULK_MAX * ES58X_FD_CANFD_RX_LEN];
|        struct es58x_fd_echo_msg echo_msg[ES58X_FD_ECHO_BULK_MAX];
|        struct es58x_fd_rx_event_msg rx_event_msg;
|        struct es58x_fd_tx_ack_msg tx_ack_msg;
|        __le64 timestamp;
|        __le32 rx_cmd_ret_le32;
|        u8 raw_msg[0];
|    } __packed;
|
|    __le16 reserved_for_crc16_do_not_use;
| } __packed;

ram_msg can then be used to manipulate the other fields at the byte level.
I am sorry but I fail to understand why this is an issue.

The issue is with using a 0-element array (these are being removed from
the kernel[1] so we can add -Warray-bounds). Normally in this situation I
would replace the 0-element array with a flexible array, but this
case is unusual in several ways:

- There is a trailing struct member (reserved_for_crc16_do_not_use),
  which is never accessed (good), and documented as "please never access
  this".

Yes. And FYI, this field is here so that
| sizeof(struct es58x_fd_urb_cmd)
returns the correct maximum size.

And, of course, because this structure will be sent to the
device, there is no possibility to reorder those fields.

- struct es58x_fd_urb_cmd is statically allocated (it is written into
  from the URB handler).

- The message lengths coming from the USB device are stored in a u16,
  which looked like it was possible to overflow the buffer.

In taking a closer look, I see that the URB command length is checked,
and the in-data length is checked as well, so the overflow concern
appears to be addressed.

quoted

Also, the proposed fix drastically increases the size of the structure.

Indeed. I will send a v2, now that I see that the overflow concern isn't
an issue.

Thanks for the explanation. That makes sense.

At the end, the only goal of raw_msg[] is to have a tag pointing
to the beginning of the union. It would be virtually identical to
something like:
|    u8 raw_msg[];
|    union {
|        /* ... */
|    } __packed ;

I had a look at your work and especially at your struct_group() macro.
Do you think it would make sense to introduce a union_group()?

Result would look like:

|    union_group_attr(urb_msg, __packed, /* raw_msg renamed to urb_msg */
|        struct es58x_fd_tx_conf_msg tx_conf_msg;
|        u8 tx_can_msg_buf[ES58X_FD_TX_BULK_MAX * ES58X_FD_CANFD_TX_LEN];
|        u8 rx_can_msg_buf[ES58X_FD_RX_BULK_MAX * ES58X_FD_CANFD_RX_LEN];
|        struct es58x_fd_echo_msg echo_msg[ES58X_FD_ECHO_BULK_MAX];
|        struct es58x_fd_rx_event_msg rx_event_msg;
|        struct es58x_fd_tx_ack_msg tx_ack_msg;
|        __le64 timestamp;
|        __le32 rx_cmd_ret_le32;
|    );

And I can then use urb_msg in place of the old raw_msg (might
need a bit of rework here and there but I can take care of it).

This is the most pretty way I can think of to remove this zero length array.
Keeping the raw_msg[] but with another size seems odd to me.

Or maybe I would be the only one using this feature in the full
tree? In that case, maybe it would make sense to keep the
union_group_attr() macro local to the etas_es58x driver?

Yours sincerely,
Vincent