Re: [PATCH 2/2] docs: can: update SocketCAN documentation for CAN XL
From: Rakuram Eswaran <hidden>
Date: 2026-01-21 18:24:37
Also in:
linux-can, linux-doc
On Sun, 18 Jan 2026 at 23:53, Oliver Hartkopp [off-list ref] wrote:
On 18.01.26 15:41, Rakuram Eswaran wrote:quoted
On Tue, 13 Jan 2026 at 21:45, Oliver Hartkopp [off-list ref] wrote:quoted
quoted
quoted
quoted
The length of the two CAN(FD) frame structures define the maximum transfer unit (MTU) of the CAN(FD) network interface and skbuff data length. Two -definitions are specified for CAN specific MTUs in include/linux/can.h: +definitions are specified for CAN specific MTUs in include/uapi/linux/can.h:No. From the user perspective he has to include include/linux/can.h Better "MTUs in the linux/can.h include file:"Can I just incude linux/can.h or include/linux/can.h? As in the current document include/linux/can.h is used.If you look into the code you need e.g. #include <linux/can.h> #include <linux/can/raw.h> So I would name it the linux/can.h include file.
Hi Oliver, Ah, now I understand what I did wrong in my interpretation. I replaced all instances of include/linux/can.h to linux/can.h include file.
quoted
quoted
What about the PWM settings here? When TMS is "on" the PWM values can be automatically calculated or set manually. There's also no CAN XL TDC when TMS=on as the TDC is a mixed-mode requirement for non-TMS transceivers.Can I add the PWM settings under new heading (CAN XL PWM) or is it fine to keep the content under the same heading (CAN XL TDC)?Yes. I would propose a new section CAN XL TMS (Transceiver Mode Setting / PWM) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The Transceiver Mode Setting (TMS) switches the RX/TX lines between the CAN XL controller and the CAN XL transceiver into the "Fast Mode" by enabling a PWM protocol. In the Fast Mode a different sample point is used and the PWM ratio is calculated automatically or can be set manually.
Ok, will add this section. In the below article, I saw a table named "CAN transceiver modes signaled". Can I add them here or above content is fine? https://www.can-cia.org/can-knowledge/can-xl
quoted
quoted
There's one big difference between CC/FD and XL frames when you read/write it to CAN_RAW sockets: For CAN CC and CAN FD you write struct can(fd)_frame's with CAN_MTU resp. CANFD_MTU lengths - no matter about the data length (cf->len). When you read/write CAN XL frames you are reading and writing the CANXL_HDR_SIZE + the length of the data. So only in the case of writing 2048 byte data, you write 2060 bytes. The minimum size for read/write is CANXL_HDR_SIZE + CANXL_MIN_DLEN == 13Good point! I will add this information along with an example. I will go through your code and decide what to add. Does the example code should focus only on CAN XL frames or also on CC/FD frames?
Here, I added a updated content for PWM section and Application considerations
for CAN XL section.
CAN XL TMS (Transceiver Mode Setting / PWM)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The Transceiver Mode Setting (TMS) switches the RX/TX lines between the
CAN XL controller and the CAN SIC XL transceiver into the "Fast Mode" by
enabling a PWM protocol.
In the Fast Mode, a different sample point is used and the PWM ratio is
either calculated automatically or can be set manually.
Application considerations for CAN XL
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
For user space applications the following rules are important when
handling CAN XL:
- Use ``struct canxl_frame`` as basic data structure when CAN XL traffic
is expected.
- Set CANXL_XLF in ``canxl_frame.flags`` for all valid CAN XL frames.
- Ensure that undefined bits in ``canxl_frame.flags`` are kept at zero.
- Respect the configured device MTU; do not send frames larger than
the MTU announced by the kernel.
- For mixed-mode controllers, be prepared to handle Classical CAN,
CAN FD and CAN XL frames on the same interface and choose the frame
structure according to the socket/protocol semantics (e.g. dedicated
CAN XL APIs when available).
- There's one big difference between CC/FD and XL frames when you
read/write it to CAN_RAW sockets:
- For CAN CC and CAN FD, struct can(fd)_frame's with CAN_MTU/CANFD_MTU
lengths.
- Read/write of CAN XL frames you are reading and writing the
CANXL_HDR_SIZE + the length of the data.
- The minimum size for read/write is CANXL_HDR_SIZE + CANXL_MIN_DLEN == 13.
When the CAN_RAW socket enables the CAN XL support you have to deal with all kind of CAN frames. IMO is makes sense to show an example that deals with all three types of frames.quoted
quoted
Here is an example that I've been implemented recently that shows a good example how to handle CC/FD/XL frames, when they are all enabled on the CAN_RAW socket: https://github.com/hartkopp/can-utils/commit/bf0cae218af9b1c1f5eabad7f3704b88ab642e00 Feel free to pick the code for some example. But please do not reference the commit as it is in my private repo and not yet integrated in the official can-utils repo.
I had gone through the code and picked the below code for example. I add the
code here. Kindly let me know if I need to add anything extra.
/* CAN CC/FD/XL frame union */
union cfu {
struct can_frame cc;
struct canfd_frame fd;
struct canxl_frame xl;
};
struct sockaddr_can addr;
static union cfu cu;
// open CAN_RAW socket
s = socket(PF_CAN, SOCK_RAW, CAN_RAW);
/* try to switch the socket into CAN FD mode */
setsockopt(s, SOL_CAN_RAW, CAN_RAW_FD_FRAMES, &canfx_on, sizeof(canfx_on));
setsockopt(s, SOL_CAN_RAW, CAN_RAW_XL_FRAMES, &canfx_on, sizeof(canfx_on));
/* try to enable the CAN XL VCID pass through mode */
setsockopt(s, SOL_CAN_RAW, CAN_RAW_XL_VCID_OPTS, &vcid_opts, sizeof(vcid_opts));
addr.can_family = AF_CAN;
// create the sockaddr binding
if (bind(s, (struct sockaddr *)&addr, sizeof(addr)) < 0) {
perror("bind");
return 1;
}
// The minimum size for read/write is CANXL_HDR_SIZE + CANXL_MIN_DLEN == 13 bytes
// (12 bytes of header + minimum 1 byte data)
if (cu.xl.flags & CANXL_XLF) {
if (nbytes != (int)CANXL_HDR_SIZE + cu.xl.len) {
printf("nbytes = %d\n", nbytes);
fprintf(stderr, "read: no CAN XL frame\n");
return 1;
}
rx_id = cu.xl.prio & CANXL_PRIO_MASK; /* remove VCID value */
data = cu.xl.data;
framelen = cu.xl.len;
} else {
/* mark dual-use struct canfd_frame */
if (nbytes == CAN_MTU)
cu.fd.flags = 0;
else if (nbytes == CANFD_MTU)
cu.fd.flags |= CANFD_FDF;
else {
fprintf(stderr, "read: incomplete CAN CC/FD frame\n");
return 1;
}
rx_id = cu.fd.can_id;
data = cu.fd.data;
framelen = cu.fd.len;
}
Best regards, Oliver
Best Regards, Rakuram.