From: Uwe Kleine-König <uwe@xxxxxxxxxxxxxxxxx> Hello, here comes v10 of this series. Changes compared to v9 sent with Message-Id: 20201018204022.910815-1-u.kleine-koenig@xxxxxxxxxxxxxx in October: - Bump date and kernel version in ABI doc - Fix double unlock in error path; found by Pavel - Don't stop the workqueue in ttyname_store() to fix error behaviour on memory allocation failure. Now it continues with the previous configuration instead of stopping. Also found by Pavel. Unaddressed review comments by Pavel are: - Unused assignment to len in ttyname_show This is wrong - "Poll every 100 msec... Hmm.... Okay, I guess?" Yes, I think there is no way around this given the trigger uses polling. There is no easy way to get notified instead. - "Are you sure about LED_ON [in the worker]? It should use current brightness selected by brightness file..." I found no consistent behaviour in other triggers. ledtrig-gpio implements a dedicated "desired_brightness" sysfs file, several use led_cdev->blink_brightness (via led_trigger_blink_oneshot), ledtrig-cpu uses led_trigger_event with LED_FULL. - "How is [the data initialized in ledtrig_tty_activate()] protected from concurrent access from sysfs?" I think there is no need to protect this. But I'm not sure I understood the question correctly, so please recheck and if needed point out the problem you see in more detail. Uwe Kleine-König (3): tty: rename tty_kopen() and add new function tty_kopen_shared() tty: new helper function tty_get_icount() leds: trigger: implement a tty trigger .../ABI/testing/sysfs-class-led-trigger-tty | 6 + drivers/accessibility/speakup/spk_ttyio.c | 2 +- drivers/leds/trigger/Kconfig | 9 + drivers/leds/trigger/Makefile | 1 + drivers/leds/trigger/ledtrig-tty.c | 188 ++++++++++++++++++ drivers/tty/tty_io.c | 85 ++++++-- include/linux/tty.h | 7 +- 7 files changed, 273 insertions(+), 25 deletions(-) create mode 100644 Documentation/ABI/testing/sysfs-class-led-trigger-tty create mode 100644 drivers/leds/trigger/ledtrig-tty.c -- 2.29.2
