Re: [PATCH] can: uapi: can.h: mark union inside struct can_frame packed

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 





On 3/23/21 4:24 PM, Marc Kleine-Budde wrote:
On 22.03.2021 17:27:38, Oliver Hartkopp wrote:

On 22.03.21 11:28, Marc Kleine-Budde wrote:
In commit ea7800565a12 ("can: add optional DLC element to Classical
CAN frame structure") the struct can_frame::can_dlc was put into an
anonymous union with another u8 variable.

For various reasons some members in struct can_frame and canfd_frame
including the first 8 byes of data are expected to have the same
memory layout. This is enforced by a BUILD_BUG_ON check in af_can.c.

Since the above mentioned commit this check fails on at least one compiler
(arm-linux-gnueabi-gcc (GCC) 9.3.0). Rong Chen analyzed the problem
and found that the union in struct can_frame takes 4 bytes instead of
the expected 1:

| struct can_frame {
|          canid_t                    can_id;               /* 0     4 */
|          union {
|                  __u8               len;                  /* 4     1 */
|                  __u8               can_dlc;              /* 4     1 */
|          };                                               /* 4     4 */
|          __u8                       __pad;                /* 8     1 */
|          __u8                       __res0;               /* 9     1 */
|          __u8                       len8_dlc;             /* 10     1 */
|
|          /* XXX 5 bytes hole, try to pack */
|
|          __u8                       data[8]
| __attribute__((__aligned__(8))); /*    16     8 */
|
|          /* size: 24, cachelines: 1, members: 6 */
|          /* sum members: 19, holes: 1, sum holes: 5 */
|          /* forced alignments: 1, forced holes: 1, sum forced holes: 5 */
|          /* last cacheline: 24 bytes */
| } __attribute__((__aligned__(8)));

Marking the union as packed fixes the problem.
Is this a proper answer to this issue?
With the affected compiler, yes. This patch makes the union 1 byte long,
as expected. This effectively fixes compiling.

I'm glad that the build bug triggered, which avoids having a broken
running kernel.

Shouldn't this problem cause the developer to update the compiler?
The question is, are the other silent corruptions with a similar union
somewhere? Maybe we should escalate this problem to the linux-arm-kernel
ML.

Rong Chen, can I download the compiler you're using somewhere?

Hi Marc,

I tried the below two compilers , and detail is at https://lore.kernel.org/linux-can/f8075a19-10e1-abf9-6d59-1a46454b74b1@xxxxxxxxxxxx/T/#m1eff42a8f6c24ebf066fa74a826a0ca068457ed8

https://download.01.org/0day-ci/cross-package/gcc-9.3.0-nolibc/x86_64-gcc-9.3.0-nolibc_arm-linux-gnueabi.tar.xz

http://cdn.kernel.org/pub/tools/crosstool/files/bin/x86_64/10.1.0/x86_64-gcc-10.1.0-nolibc-arm-linux-gnueabi.tar.xz

Best Regards,
Rong Chen


https://lore.kernel.org/linux-can/f8075a19-10e1-abf9-6d59-1a46454b74b1@xxxxxxxxxxxx/T/#u
regards,
Marc





[Index of Archives]     [Automotive Discussions]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Security]     [Bugtraq]     [Linux]     [Linux OMAP]     [Linux MIPS]     [eCos]     [Asterisk Internet PBX]     [Linux API]     [CAN Bus]

  Powered by Linux