On Tue. 27 Jul. 2022 at 17:50, Marc Kleine-Budde <mkl@xxxxxxxxxxxxxx> wrote: > On 27.07.2022 17:29:25, Vincent MAILHOL wrote: > > On Wed. 27 Jul. 2022 at 17:10, Marc Kleine-Budde <mkl@xxxxxxxxxxxxxx> wrote: > > > On 26.07.2022 19:24:54, Vincent Mailhol wrote: > > > > Currently, userland has no method to query which timestamping features > > > > are supported by the peak_usb driver (aside maybe of getting RX > > > > messages and obseverse whever or not hardware timestamps stay at > > > > zero). > > > > > > > > The canonical way for a network driver to advertise what kind of > > > > timestamping it supports is to implement > > > > ethtool_ops::get_ts_info(). Here, we use the CAN specific > > > > can_ethtool_op_get_ts_info_hwts() function to achieve this. > > > > > > > > In addition, the driver currently does not support the hardware > > > > timestamps ioctls. According to [1], SIOCSHWTSTAMP is "must" and > > > > SIOCGHWTSTAMP is "should". This patch fills up that gap by > > > > implementing net_device_ops::ndo_eth_ioctl() using the CAN specific > > > > function can_eth_ioctl_hwts(). > > > > > > > > [1] kernel doc Timestamping, section 3.1: "Hardware Timestamping > > > > Implementation: Device Drivers" > > > > Link: https://docs.kernel.org/networking/timestamping.html#hardware-timestamping-implementation-device-drivers > > > > > > > > CC: Stephane Grosjean <s.grosjean@xxxxxxxxxxxxxxx> > > > > Signed-off-by: Vincent Mailhol <mailhol.vincent@xxxxxxxxxx> > > > > --- > > > > > > > > Hi Stéphane, as far as I understand, the pcan_usb and the pacn_usb_pro > > > ^^^^ > > > pcan > > > > > > > support hardware timestamps but the pcan_usb_fd doesn't. If not the > > > > case, let me know. > > > > > > > > This is not tested. If you find any issue or if you want to modify, > > > > feel free to pick up that patch and resend it. > > > > > > I have a: > > > > > > | Bus 002 Device 009: ID 0c72:0012 PEAK System PCAN-USB FD > > > > > > It supports hardware RX timestamps (Debian kernel 5.18.0-2-amd64) only: > > > > > > | (1970-01-01 01:00:00.000000) peakfd0 TX - - 002 [1] 01 > > > | (1970-01-01 02:17:09.473817) peakfd0 RX - - 002 [1] 3C > > > | (1970-01-01 01:00:00.000000) peakfd0 TX - - 002 [1] 02 > > > | (1970-01-01 02:17:09.673980) peakfd0 RX - - 002 [1] 3D > > > > Thanks for the confirmation. So this means that all Peak hardware > > supports the hardware timestamping. > > The PCAN-USB FD only support RX hardware timestamping, not TX. > > This will greatly simplify the > > logic. No need to have two different struct ethtool_ops. I will > > prepare a v4. > > We have a peak_pciefd card: > > | 01:00.0 Network controller: PEAK-System Technik GmbH Device 0018 (rev 01) > > Only RX HW timestamps (Debian kernel 5.16.0-0.bpo.4-amd64): > > | (1970-01-01 01:00:00.000000) can0_iobus TX - - 601 [8] 40 0C 21 01 00 00 00 00 > | (1970-02-16 18:25:06.810100) can0_iobus RX - - 581 [8] 53 0C 21 01 00 00 00 00 > | (1970-01-01 01:00:00.000000) can0_iobus TX - - 602 [8] 40 0C 21 01 00 00 00 00 > | (1970-02-16 18:25:06.819380) can0_iobus RX - - 582 [8] 53 0C 21 01 00 00 00 00 O_o This comes as a surprise. I read your first message too quick and did not get the nuance that only *RX* hardware timestamp was supported. After investigation, I found this message from Stéphane: | Those hw tx timestamps are not supported by the peak_usb | driver: your socket-can application will receive "local | software" TX timestamps only, corresponding roughly to the time | the CAN frame has been given to the USB core of the kernel. | | However, this is possible with our pcan driver: you can get | an "echoed" message with a HW TX timestamp for each message | your application sends. Source: https://forum.peak-system.com/viewtopic.php?t=5572#p12913 So the hardware does support it, it is just that the mainstream driver does not implement it. Yours sincerely, Vincent Mailhol
