On 19.12.21 10:38, Dario Binacchi wrote:
> The series was created to optimize the use of the C_CAN controller
> (maximum number of message objects, use of FIFO, ...) to get the
> best possible RX/TX throughput from hardware.
>
> To emphasize the purpose of the series I also added the patch
> which adds multi-message support to `sendmsg ', rather than'
> submitting it separately. In fact, it too has contributed to
> increasing transmission throughput.
>
> The patches prior to the one adding the BOSCH C_CAN controller driver,
> are a backporting of commits from the linux kernel that also impacted
> on the driver itself and I found it necessary to add to the series.
>
> The latest patch, as described in its commit, modified Stephen
> Battazzo's patch [1] with bug fixes, backporting of commits from the
> linux kernel, extension of message objects to 64 and improvements in
> FIFO handling. This series was in fact preceded by the creation of
> patches which I sent to the Linux kernel and which were then accepted:
>
> 387da6bc7a82 ("can: c_can: cache frames to operate as a true FIFO")
> 28e86e9ab522 ("can: c_can: support tx ring algorithm")
> a54cdbba9dee ("can: c_can: exit c_can_do_tx() early if no frames have been
> sent")
> 5064e40596f4 ("can: c_can: remove struct c_can_priv::priv field")
> 06fc143b2ede dt-bindings: net: can: c_can: convert to json-schema
> 132f2d45fb23 ("can: c_can: add support to 64 message objects")
> 13831ce69c77 ("can: c_can: prepare to up the message objects number")
> fcbded019855 ("can: c_can: use 32-bit write to set arbitration register")
> eddf67115040 ("can: c_can: add a comment about IF_RX interface's use")
> c8a6b44388cb ("can: c_can: fix indentation")
> f65735c203d5 ("can: c_can: remove unused code")
>
> The patch was tested on a custom board mounting an AM335x SOC for the
> next configurations:
> - CONFIG_XENO_DRIVERS_CAN=m, CONFIG_XENO_DRIVERS_CAN_VIRT=m
> - CONFIG_XENO_DRIVERS_CAN=y, CONFIG_XENO_DRIVERS_CAN_VIRT=m
> - CONFIG_XENO_DRIVERS_CAN=y, CONFIG_XENO_DRIVERS_CAN_VIRT=y
>
> PCI support has not been developed.
>
> [1] http://xenomai.org/pipermail/xenomai/2015-July/034690.html
>
> Changes in v5:
> - Add the `drivers/can: add corectl to retrieve configuration' patch to
> the series.
> - smokey_modprobe(): check the `name' parameter.
> - smokey_modprobe(): return `err' instead of 0 in case of no error.
> - smokey_rmmod(): set smokey_modules[midx] to NULL after freeing the
> pointed string.
> - Check _CC_COBALT_GET_CAN_CONFIG through cobalt_corectl call.
> - Don't check modprobe return code. In case the modules are builtin the
> test would be improperly terminated with an error, since the modprobe
> does not check if a module is builtin.
>
> Changes in v4:
> - Add the patch to the series.
> - Load/unload the CAN drivers modules before/after running the test.
> - Add setup/teardown function.
>
> Changes in v3:
> - Add the patch to the series.
>
> Changes in v2:
> - Expand the commit message to better describe the added API.
> - Drop 'default n' Kconfig
> - Enable the driver for Xenomai CI builds
> - Drop commented code
> - Drop power managemet (pm) code
>
> Dario Binacchi (8):
> drivers/can: add corectl to retrieve configuration
> lib/smokey: add smokey_{modprobe,rmmod} helpers
> testsuite/smokey: add virtual CAN testsuite
> drivers/can: add multi message support to sendmsg
> drivers/can: add CAN_ERR_CRTL_ACTIVE info status
> drivers/can: add len field to can frame structures
> drivers/can: add a minimal support to ethtool API
> drivers/can: add support for Bosch C_CAN controller
>
> .gitlab-ci.yml | 2 +
> configure.ac | 1 +
> include/cobalt/uapi/corectl.h | 2 +
> include/rtdm/uapi/can.h | 6 +-
> include/smokey/smokey.h | 6 +-
> kernel/drivers/can/Kconfig | 1 +
> kernel/drivers/can/Makefile | 6 +-
> kernel/drivers/can/README | 1 +
> kernel/drivers/can/c_can/Kconfig | 15 +
> kernel/drivers/can/c_can/Makefile | 8 +
> kernel/drivers/can/c_can/rtcan_c_can.c | 1298 +++++++++++++++++
> kernel/drivers/can/c_can/rtcan_c_can.h | 255 ++++
> .../drivers/can/c_can/rtcan_c_can_ethtool.c | 47 +
> .../drivers/can/c_can/rtcan_c_can_platform.c | 461 ++++++
> kernel/drivers/can/corectl.c | 41 +
> kernel/drivers/can/rtcan_dev.h | 4 +-
> kernel/drivers/can/rtcan_ethtool.c | 75 +
> kernel/drivers/can/rtcan_ethtool.h | 24 +
> kernel/drivers/can/rtcan_module.c | 6 +
> kernel/drivers/can/rtcan_raw.c | 239 +--
> kernel/drivers/can/rtcan_raw_dev.c | 20 +
> kernel/drivers/can/rtcan_socket.h | 5 +-
> lib/smokey/helpers.c | 103 ++
> testsuite/smokey/Makefile.am | 2 +
> testsuite/smokey/can-virt/Makefile.am | 8 +
> testsuite/smokey/can-virt/virt.c | 370 +++++
> testsuite/smokey/net_common/setup.c | 68 +-
> 27 files changed, 2890 insertions(+), 184 deletions(-)
> create mode 100644 kernel/drivers/can/c_can/Kconfig
> create mode 100644 kernel/drivers/can/c_can/Makefile
> create mode 100644 kernel/drivers/can/c_can/rtcan_c_can.c
> create mode 100644 kernel/drivers/can/c_can/rtcan_c_can.h
> create mode 100644 kernel/drivers/can/c_can/rtcan_c_can_ethtool.c
> create mode 100644 kernel/drivers/can/c_can/rtcan_c_can_platform.c
> create mode 100644 kernel/drivers/can/corectl.c
> create mode 100644 kernel/drivers/can/rtcan_ethtool.c
> create mode 100644 kernel/drivers/can/rtcan_ethtool.h
> create mode 100644 testsuite/smokey/can-virt/Makefile.am
> create mode 100644 testsuite/smokey/can-virt/virt.c
>
Thanks, applied with the changes mentioned in patch 1.
Jan
--
Siemens AG, Technology
Competence Center Embedded Linux
