[tty-next,v3,0/6] convert 8250 to nbcon

Message

John Ogness Oct. 25, 2024, 10:57 a.m. UTC

This is v3 of a series to convert the 8250 driver to an NBCON
console, providing both threaded and atomic printing
implementations. v2 of this series is here [0], which also
contains additional background information about NBCON consoles
in general in the cover letter.

To test this version I acquired real hardware (TI AM3358
BeagleBone Black) and tested the following modes:

RS232
- no flow control
- software flow control
  (UPF_SOFT_FLOW, UPSTAT_AUTOXOFF)
- hardware flow control
  (UPF_HARD_FLOW, UPSTAT_AUTOCTS, UPSTAT_AUTORTS)
- software emulated hardware flow control
  (UPF_CONS_FLOW, UPSTAT_CTS_ENABLE)

RS485
- with SER_RS485_RX_DURING_TX
- without SER_RS485_RX_DURING_TX

The tests focussed on kernel logging in various combinations of
normal, warning, and panic situations. Although not related to
the console printing code changes, the tests also included
using a getty/login session on the console.

Note that this UART (TI16750) supports a 64-byte TX-FIFO, which
is used in all console printing modes except for the software
emulated hardware flow control.

Here are the changes since v2:

- For RS485 start/stop TX, specify if called in console
  context.

- For RS485 start/stop TX, when in console context, do not
  disable/enable interrupts.

- Relocate modem_status_handler() to avoid unused static
  function for some configs.

- Move LSR_THRE waiting into a new
  serial8250_console_wait_putchar() function.

- For serial8250_console_fifo_write(), use
  serial8250_console_putchar() for writing. This allows newline
  tracking for FIFO mode as well.

- For serial8250_console_fifo_write(), allow 10ms timeout for
  each byte written.

- Use FIFO mode for thread and atomic modes when available.

- Introduce serial8250_console_byte_write() to handle writing
  when not using the FIFO mode.

- Consolidate thread and atomic callbacks. Now the only
  difference is modem control: For atomic, called as irq_work.
  For thread, called direct.

John Ogness

[0] https://lore.kernel.org/lkml/[e-mail]

John Ogness (6):
  serial: 8250: Adjust the timeout for FIFO mode
  serial: 8250: Use high-level write function for FIFO
  serial: 8250: Split out rx stop/start code into helpers
  serial: 8250: Specify console context for rs485_start/stop_tx
  serial: 8250: Switch to nbcon console
  serial: 8250: Revert "drop lockdep annotation from
    serial8250_clear_IER()"

 drivers/tty/serial/8250/8250.h            |   4 +-
 drivers/tty/serial/8250/8250_bcm2835aux.c |   4 +-
 drivers/tty/serial/8250/8250_core.c       |  35 ++-
 drivers/tty/serial/8250/8250_omap.c       |   2 +-
 drivers/tty/serial/8250/8250_port.c       | 267 +++++++++++++++++-----
 include/linux/serial_8250.h               |  11 +-
 6 files changed, 251 insertions(+), 72 deletions(-)


base-commit: 44059790a5cb9258ae6137387e4c39b717fd2ced

Comments

Andy Shevchenko Oct. 25, 2024, 1:58 p.m. UTC | #1

On Fri, Oct 25, 2024 at 01:03:22PM +0206, John Ogness wrote:
> This is v3 of a series to convert the 8250 driver to an NBCON
> console, providing both threaded and atomic printing
> implementations. v2 of this series is here [0], which also
> contains additional background information about NBCON consoles
> in general in the cover letter.
> 
> To test this version I acquired real hardware (TI AM3358
> BeagleBone Black) and tested the following modes:
> 
> RS232
> - no flow control
> - software flow control
>   (UPF_SOFT_FLOW, UPSTAT_AUTOXOFF)
> - hardware flow control
>   (UPF_HARD_FLOW, UPSTAT_AUTOCTS, UPSTAT_AUTORTS)
> - software emulated hardware flow control
>   (UPF_CONS_FLOW, UPSTAT_CTS_ENABLE)
> 
> RS485
> - with SER_RS485_RX_DURING_TX
> - without SER_RS485_RX_DURING_TX
> 
> The tests focussed on kernel logging in various combinations of
> normal, warning, and panic situations. Although not related to
> the console printing code changes, the tests also included
> using a getty/login session on the console.
> 
> Note that this UART (TI16750) supports a 64-byte TX-FIFO, which
> is used in all console printing modes except for the software
> emulated hardware flow control.

Thank you for the update.

I am going to review some patches at some point, but what I want to say here
is that if you have a new functions to utilise something, please also check
if the rest of 8250*.c may have an advantage of. It would reduce churn in case
if your series already exports APIs or provides inliners for such cases.