mbox series

[v6,00/14] serial: qcom-geni-serial: implement support for SE DMA

Message ID 20221229155030.418800-1-brgl@bgdev.pl (mailing list archive)
Headers show
Series serial: qcom-geni-serial: implement support for SE DMA | expand

Message

Bartosz Golaszewski Dec. 29, 2022, 3:50 p.m. UTC
From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>

The goal of this series is to update the qcom-geni-serial driver to use
the DMA mode of the QUPv3 serial engine. This is accomplished by the last
patch in the series. The previous ones contain either various tweaks,
reworks and refactoring or prepare the driver for adding DMA support.

More work will follow on the serial engine in order to reduce code
redundancy among its users and add support for SE DMA to the qcom GENI
SPI driver.

v5 -> v6:
- move patch 12/14 to the front of the series and mark it as a fix
- rebase on top of v6.2-rc1 (there were some updates to the driver)

v4 -> v5:
- split patch 8/13 into two: one for splitting out the chunk sending code
  and one for refactoring it (for easier review)
- when stopping uart: disable the irq first before stopping transfers in
  progress, otherwise we rist rescheduling additional transfers after
  interrupt
- make types even nore consistent in qcom_geni_serial_send_chunk_fifo()

v3 -> v4:
- don't assign NULL (even cast) to variables of type dma_addr_t
- refactor checking the SE_GENI_STATUS into two inline functions
- drop min_t() in favor of regular min() after adding some consistency to types
- check if the abort command was successful and emit a message if not when
  stopping TX in DMA mode

v2 -> v3:
- drop devres patches from the series

v1 -> v2:
- turn to_dev_uport() macro into a static inline function
- use CIRC_CNT_TO_END() and uart_xmit_advance() where applicable and don't
  handle xmit->tail directly
- drop sizeof() where BYTES_PER_FIFO_WORD can be used
- further refactor qcom_geni_serial_handle_tx_fifo()
- collect review tags

Bartosz Golaszewski (14):
  tty: serial: qcom-geni-serial: stop operations in progress at shutdown
  tty: serial: qcom-geni-serial: drop unneeded forward definitions
  tty: serial: qcom-geni-serial: remove unused symbols
  tty: serial: qcom-geni-serial: align #define values
  tty: serial: qcom-geni-serial: improve the to_dev_port() macro
  tty: serial: qcom-geni-serial: remove stray newlines
  tty: serial: qcom-geni-serial: refactor qcom_geni_serial_isr()
  tty: serial: qcom-geni-serial: remove unneeded tabs
  tty: serial: qcom-geni-serial: split out the FIFO tx code
  tty: serial: qcom-geni-serial: refactor
    qcom_geni_serial_send_chunk_fifo()
  tty: serial: qcom-geni-serial: drop the return value from handle_rx
  tty: serial: qcom-geni-serial: use of_device_id data
  soc: qcom-geni-se: add more symbol definitions
  tty: serial: qcom-geni-serial: add support for serial engine DMA

 drivers/tty/serial/qcom_geni_serial.c | 626 +++++++++++++++++---------
 include/linux/qcom-geni-se.h          |   3 +
 2 files changed, 413 insertions(+), 216 deletions(-)

Comments

Bartosz Golaszewski Jan. 12, 2023, 8:59 a.m. UTC | #1
On Thu, Dec 29, 2022 at 4:50 PM Bartosz Golaszewski <brgl@bgdev.pl> wrote:
>
> From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
>
> The goal of this series is to update the qcom-geni-serial driver to use
> the DMA mode of the QUPv3 serial engine. This is accomplished by the last
> patch in the series. The previous ones contain either various tweaks,
> reworks and refactoring or prepare the driver for adding DMA support.
>
> More work will follow on the serial engine in order to reduce code
> redundancy among its users and add support for SE DMA to the qcom GENI
> SPI driver.
>
> v5 -> v6:
> - move patch 12/14 to the front of the series and mark it as a fix
> - rebase on top of v6.2-rc1 (there were some updates to the driver)
>
> v4 -> v5:
> - split patch 8/13 into two: one for splitting out the chunk sending code
>   and one for refactoring it (for easier review)
> - when stopping uart: disable the irq first before stopping transfers in
>   progress, otherwise we rist rescheduling additional transfers after
>   interrupt
> - make types even nore consistent in qcom_geni_serial_send_chunk_fifo()
>
> v3 -> v4:
> - don't assign NULL (even cast) to variables of type dma_addr_t
> - refactor checking the SE_GENI_STATUS into two inline functions
> - drop min_t() in favor of regular min() after adding some consistency to types
> - check if the abort command was successful and emit a message if not when
>   stopping TX in DMA mode
>
> v2 -> v3:
> - drop devres patches from the series
>
> v1 -> v2:
> - turn to_dev_uport() macro into a static inline function
> - use CIRC_CNT_TO_END() and uart_xmit_advance() where applicable and don't
>   handle xmit->tail directly
> - drop sizeof() where BYTES_PER_FIFO_WORD can be used
> - further refactor qcom_geni_serial_handle_tx_fifo()
> - collect review tags
>
> Bartosz Golaszewski (14):
>   tty: serial: qcom-geni-serial: stop operations in progress at shutdown
>   tty: serial: qcom-geni-serial: drop unneeded forward definitions
>   tty: serial: qcom-geni-serial: remove unused symbols
>   tty: serial: qcom-geni-serial: align #define values
>   tty: serial: qcom-geni-serial: improve the to_dev_port() macro
>   tty: serial: qcom-geni-serial: remove stray newlines
>   tty: serial: qcom-geni-serial: refactor qcom_geni_serial_isr()
>   tty: serial: qcom-geni-serial: remove unneeded tabs
>   tty: serial: qcom-geni-serial: split out the FIFO tx code
>   tty: serial: qcom-geni-serial: refactor
>     qcom_geni_serial_send_chunk_fifo()
>   tty: serial: qcom-geni-serial: drop the return value from handle_rx
>   tty: serial: qcom-geni-serial: use of_device_id data
>   soc: qcom-geni-se: add more symbol definitions
>   tty: serial: qcom-geni-serial: add support for serial engine DMA
>
>  drivers/tty/serial/qcom_geni_serial.c | 626 +++++++++++++++++---------
>  include/linux/qcom-geni-se.h          |   3 +
>  2 files changed, 413 insertions(+), 216 deletions(-)
>
> --
> 2.37.2
>

It's been two weeks without any further comments, can this be picked
up now into the serial tree?

Bart
Greg KH Jan. 12, 2023, 9:41 a.m. UTC | #2
On Thu, Jan 12, 2023 at 09:59:51AM +0100, Bartosz Golaszewski wrote:
> On Thu, Dec 29, 2022 at 4:50 PM Bartosz Golaszewski <brgl@bgdev.pl> wrote:
> >
> > From: Bartosz Golaszewski <bartosz.golaszewski@linaro.org>
> >
> > The goal of this series is to update the qcom-geni-serial driver to use
> > the DMA mode of the QUPv3 serial engine. This is accomplished by the last
> > patch in the series. The previous ones contain either various tweaks,
> > reworks and refactoring or prepare the driver for adding DMA support.
> >
> > More work will follow on the serial engine in order to reduce code
> > redundancy among its users and add support for SE DMA to the qcom GENI
> > SPI driver.
> >
> > v5 -> v6:
> > - move patch 12/14 to the front of the series and mark it as a fix
> > - rebase on top of v6.2-rc1 (there were some updates to the driver)
> >
> > v4 -> v5:
> > - split patch 8/13 into two: one for splitting out the chunk sending code
> >   and one for refactoring it (for easier review)
> > - when stopping uart: disable the irq first before stopping transfers in
> >   progress, otherwise we rist rescheduling additional transfers after
> >   interrupt
> > - make types even nore consistent in qcom_geni_serial_send_chunk_fifo()
> >
> > v3 -> v4:
> > - don't assign NULL (even cast) to variables of type dma_addr_t
> > - refactor checking the SE_GENI_STATUS into two inline functions
> > - drop min_t() in favor of regular min() after adding some consistency to types
> > - check if the abort command was successful and emit a message if not when
> >   stopping TX in DMA mode
> >
> > v2 -> v3:
> > - drop devres patches from the series
> >
> > v1 -> v2:
> > - turn to_dev_uport() macro into a static inline function
> > - use CIRC_CNT_TO_END() and uart_xmit_advance() where applicable and don't
> >   handle xmit->tail directly
> > - drop sizeof() where BYTES_PER_FIFO_WORD can be used
> > - further refactor qcom_geni_serial_handle_tx_fifo()
> > - collect review tags
> >
> > Bartosz Golaszewski (14):
> >   tty: serial: qcom-geni-serial: stop operations in progress at shutdown
> >   tty: serial: qcom-geni-serial: drop unneeded forward definitions
> >   tty: serial: qcom-geni-serial: remove unused symbols
> >   tty: serial: qcom-geni-serial: align #define values
> >   tty: serial: qcom-geni-serial: improve the to_dev_port() macro
> >   tty: serial: qcom-geni-serial: remove stray newlines
> >   tty: serial: qcom-geni-serial: refactor qcom_geni_serial_isr()
> >   tty: serial: qcom-geni-serial: remove unneeded tabs
> >   tty: serial: qcom-geni-serial: split out the FIFO tx code
> >   tty: serial: qcom-geni-serial: refactor
> >     qcom_geni_serial_send_chunk_fifo()
> >   tty: serial: qcom-geni-serial: drop the return value from handle_rx
> >   tty: serial: qcom-geni-serial: use of_device_id data
> >   soc: qcom-geni-se: add more symbol definitions
> >   tty: serial: qcom-geni-serial: add support for serial engine DMA
> >
> >  drivers/tty/serial/qcom_geni_serial.c | 626 +++++++++++++++++---------
> >  include/linux/qcom-geni-se.h          |   3 +
> >  2 files changed, 413 insertions(+), 216 deletions(-)
> >
> > --
> > 2.37.2
> >
> 
> It's been two weeks without any further comments, can this be picked
> up now into the serial tree?

When I dig out of the huge hole that is my todo queue, yes:
	$ mdfrm -c ~/mail/todo/
	2560 messages in /home/gregkh/mail/todo/

You are in good company...