Skip to contentSkip to conversation

Marking legacy watchdog drivers as deprecated / obsolete

DISCUSSION Guenter Roeck 3 participants 4 messages Jul 29, 2019 – Aug 1, 2019

lore
Share & tools
lore
#0 lore raw

Hi,

we have recently seen a number of changes to legacy watchdog drivers, mostly surrounding the coding style used some 10+ years ago, but also fixing minor formatting or coding problems found by static analyzers. This slowly rises above the level of background noise.

Would it be acceptable to mark all those drivers as deprecated/obsolete, warn users that the driver should be converted to use the watchdog subsystem, and that it will otherwise be removed in a later Linux kernel version ? This would give us an idea which drivers are still in use, and it would enable us to remove the remaining drivers maybe 5 or 6 releases for now.

Thoughts ?

Guenter

#1 lore raw
Quoted 8 lines · Guenter Roeck

Hi,

we have recently seen a number of changes to legacy watchdog drivers, mostly surrounding the coding style used some 10+ years ago, but also fixing minor formatting or coding problems found by static analyzers. This slowly rises above the level of background noise.

Quoted 8 lines

Would it be acceptable to mark all those drivers as deprecated/obsolete, warn users that the driver should be converted to use the watchdog subsystem, and that it will otherwise be removed in a later Linux kernel version ? This would give us an idea which drivers are still in use, and it would enable us to remove the remaining drivers maybe 5 or 6 releases for now.

Thoughts ?

I don't think an automated approach across 61 drivers is likely to work well. About half of the drivers appear to be for specific SoCs, and removing the watchdog driver while keeping the rest of the SoC support would not be helpful, it just means we break one of the drivers for the last remaining users of an old SoC the next time they try to upgrade to a new kernel.

It would probably be helpful to go through the list and see if any of the drivers
are for platforms that are already gone. FWIW, here is the list of drivers that have their own .ioctl() method, taken from a patch I'm sending soon to add a .compat_ioctl handler:

arch/powerpc/platforms/52xx/mpc52xx_gpt.c
arch/um/drivers/harddog_kern.c
drivers/char/ipmi/ipmi_watchdog.c
drivers/hwmon/fschmd.c
drivers/rtc/rtc-ds1374.c
drivers/watchdog/acquirewdt.c
drivers/watchdog/advantechwdt.c
drivers/watchdog/alim1535_wdt.c
drivers/watchdog/alim7101_wdt.c
drivers/watchdog/ar7_wdt.c
drivers/watchdog/at91rm9200_wdt.c
drivers/watchdog/ath79_wdt.c
drivers/watchdog/bcm63xx_wdt.c
drivers/watchdog/cpu5wdt.c
drivers/watchdog/eurotechwdt.c
drivers/watchdog/f71808e_wdt.c
drivers/watchdog/gef_wdt.c
drivers/watchdog/geodewdt.c
drivers/watchdog/ib700wdt.c
drivers/watchdog/ibmasr.c
drivers/watchdog/indydog.c
drivers/watchdog/intel_scu_watchdog.c
drivers/watchdog/iop_wdt.c
drivers/watchdog/it8712f_wdt.c
drivers/watchdog/ixp4xx_wdt.c
drivers/watchdog/ks8695_wdt.c
drivers/watchdog/m54xx_wdt.c
drivers/watchdog/machzwd.c
drivers/watchdog/mixcomwd.c
drivers/watchdog/mtx-1_wdt.c
drivers/watchdog/mv64x60_wdt.c
drivers/watchdog/nuc900_wdt.c
drivers/watchdog/nv_tco.c
drivers/watchdog/pc87413_wdt.c
drivers/watchdog/pcwd.c
drivers/watchdog/pcwd_pci.c
drivers/watchdog/pcwd_usb.c
drivers/watchdog/pika_wdt.c
drivers/watchdog/pnx833x_wdt.c
drivers/watchdog/rc32434_wdt.c
drivers/watchdog/rdc321x_wdt.c
drivers/watchdog/riowd.c
drivers/watchdog/sa1100_wdt.c
drivers/watchdog/sb_wdog.c
drivers/watchdog/sbc60xxwdt.c
drivers/watchdog/sbc7240_wdt.c
drivers/watchdog/sbc_epx_c3.c
drivers/watchdog/sbc_fitpc2_wdt.c
drivers/watchdog/sc1200wdt.c
drivers/watchdog/sc520_wdt.c
drivers/watchdog/sch311x_wdt.c
drivers/watchdog/scx200_wdt.c
drivers/watchdog/smsc37b787_wdt.c
drivers/watchdog/w83877f_wdt.c
drivers/watchdog/w83977f_wdt.c
drivers/watchdog/wafer5823wdt.c
drivers/watchdog/wdrtas.c
drivers/watchdog/wdt.c
drivers/watchdog/wdt285.c
drivers/watchdog/wdt977.c
drivers/watchdog/wdt_pci.c

Arnd

#2 lore raw
Quoted 16 lines · Arnd Bergmann
Quoted 15 lines · Guenter Roeck

Hi,

we have recently seen a number of changes to legacy watchdog drivers, mostly surrounding the coding style used some 10+ years ago, but also fixing minor formatting or coding problems found by static analyzers. This slowly rises above the level of background noise.

Would it be acceptable to mark all those drivers as deprecated/obsolete, warn users that the driver should be converted to use the watchdog subsystem, and that it will otherwise be removed in a later Linux kernel version ? This would give us an idea which drivers are still in use, and it would enable us to remove the remaining drivers maybe 5 or 6 releases for now.

Quoted 9 lines
Quoted 1 line

Thoughts ?

I don't think an automated approach across 61 drivers is likely to work well. About half of the drivers appear to be for specific SoCs, and removing the watchdog driver while keeping the rest of the SoC support would not be helpful, it just means we break one of the drivers for the last remaining users of an old SoC the next time they try to upgrade to a new kernel.

The primary goal would be to identify drivers still in use, and to trigger efforts to convert those drivers to the new infrastructure. Removal of obsolete / unused drivers would be a separate decision, to be made at some point in the future, and individually for each driver. I specifically wasn't trying to suggest auto-removal.

Guenter

Quoted 69 lines

It would probably be helpful to go through the list and see if any of the drivers
are for platforms that are already gone. FWIW, here is the list of drivers that have their own .ioctl() method, taken from a patch I'm sending soon to add a .compat_ioctl handler:

arch/powerpc/platforms/52xx/mpc52xx_gpt.c
arch/um/drivers/harddog_kern.c
drivers/char/ipmi/ipmi_watchdog.c
drivers/hwmon/fschmd.c
drivers/rtc/rtc-ds1374.c
drivers/watchdog/acquirewdt.c
drivers/watchdog/advantechwdt.c
drivers/watchdog/alim1535_wdt.c
drivers/watchdog/alim7101_wdt.c
drivers/watchdog/ar7_wdt.c
drivers/watchdog/at91rm9200_wdt.c
drivers/watchdog/ath79_wdt.c
drivers/watchdog/bcm63xx_wdt.c
drivers/watchdog/cpu5wdt.c
drivers/watchdog/eurotechwdt.c
drivers/watchdog/f71808e_wdt.c
drivers/watchdog/gef_wdt.c
drivers/watchdog/geodewdt.c
drivers/watchdog/ib700wdt.c
drivers/watchdog/ibmasr.c
drivers/watchdog/indydog.c
drivers/watchdog/intel_scu_watchdog.c
drivers/watchdog/iop_wdt.c
drivers/watchdog/it8712f_wdt.c
drivers/watchdog/ixp4xx_wdt.c
drivers/watchdog/ks8695_wdt.c
drivers/watchdog/m54xx_wdt.c
drivers/watchdog/machzwd.c
drivers/watchdog/mixcomwd.c
drivers/watchdog/mtx-1_wdt.c
drivers/watchdog/mv64x60_wdt.c
drivers/watchdog/nuc900_wdt.c
drivers/watchdog/nv_tco.c
drivers/watchdog/pc87413_wdt.c
drivers/watchdog/pcwd.c
drivers/watchdog/pcwd_pci.c
drivers/watchdog/pcwd_usb.c
drivers/watchdog/pika_wdt.c
drivers/watchdog/pnx833x_wdt.c
drivers/watchdog/rc32434_wdt.c
drivers/watchdog/rdc321x_wdt.c
drivers/watchdog/riowd.c
drivers/watchdog/sa1100_wdt.c
drivers/watchdog/sb_wdog.c
drivers/watchdog/sbc60xxwdt.c
drivers/watchdog/sbc7240_wdt.c
drivers/watchdog/sbc_epx_c3.c
drivers/watchdog/sbc_fitpc2_wdt.c
drivers/watchdog/sc1200wdt.c
drivers/watchdog/sc520_wdt.c
drivers/watchdog/sch311x_wdt.c
drivers/watchdog/scx200_wdt.c
drivers/watchdog/smsc37b787_wdt.c
drivers/watchdog/w83877f_wdt.c
drivers/watchdog/w83977f_wdt.c
drivers/watchdog/wafer5823wdt.c
drivers/watchdog/wdrtas.c
drivers/watchdog/wdt.c
drivers/watchdog/wdt285.c
drivers/watchdog/wdt977.c
drivers/watchdog/wdt_pci.c

Arnd

#3 lore raw
Quoted 27 lines · On Tuesday 30 July 2019 17:57:37 Guenter Roeck wrote:
Quoted 16 lines · Arnd Bergmann
Quoted 15 lines · Guenter Roeck

Hi,

we have recently seen a number of changes to legacy watchdog drivers, mostly surrounding the coding style used some 10+ years ago, but also fixing minor formatting or coding problems found by static analyzers. This slowly rises above the level of background noise.

Would it be acceptable to mark all those drivers as deprecated/obsolete, warn users that the driver should be converted to use the watchdog subsystem, and that it will otherwise be removed in a later Linux kernel version ? This would give us an idea which drivers are still in use, and it would enable us to remove the remaining drivers maybe 5 or 6 releases for now.

Quoted 9 lines
Quoted 1 line

Thoughts ?

I don't think an automated approach across 61 drivers is likely to work well. About half of the drivers appear to be for specific SoCs, and removing the watchdog driver while keeping the rest of the SoC support would not be helpful, it just means we break one of the drivers for the last remaining users of an old SoC the next time they try to upgrade to a new kernel.

Quoted 78 lines

The primary goal would be to identify drivers still in use, and to trigger efforts to convert those drivers to the new infrastructure. Removal of obsolete / unused drivers would be a separate decision, to be made at some point in the future, and individually for each driver. I specifically wasn't trying to suggest auto-removal.

Guenter

Quoted 69 lines

It would probably be helpful to go through the list and see if any of the drivers
are for platforms that are already gone. FWIW, here is the list of drivers that have their own .ioctl() method, taken from a patch I'm sending soon to add a .compat_ioctl handler:

arch/powerpc/platforms/52xx/mpc52xx_gpt.c
arch/um/drivers/harddog_kern.c
drivers/char/ipmi/ipmi_watchdog.c
drivers/hwmon/fschmd.c
drivers/rtc/rtc-ds1374.c
drivers/watchdog/acquirewdt.c
drivers/watchdog/advantechwdt.c
drivers/watchdog/alim1535_wdt.c
drivers/watchdog/alim7101_wdt.c
drivers/watchdog/ar7_wdt.c
drivers/watchdog/at91rm9200_wdt.c
drivers/watchdog/ath79_wdt.c
drivers/watchdog/bcm63xx_wdt.c
drivers/watchdog/cpu5wdt.c
drivers/watchdog/eurotechwdt.c
drivers/watchdog/f71808e_wdt.c
drivers/watchdog/gef_wdt.c
drivers/watchdog/geodewdt.c
drivers/watchdog/ib700wdt.c
drivers/watchdog/ibmasr.c
drivers/watchdog/indydog.c
drivers/watchdog/intel_scu_watchdog.c
drivers/watchdog/iop_wdt.c
drivers/watchdog/it8712f_wdt.c
drivers/watchdog/ixp4xx_wdt.c
drivers/watchdog/ks8695_wdt.c
drivers/watchdog/m54xx_wdt.c
drivers/watchdog/machzwd.c
drivers/watchdog/mixcomwd.c
drivers/watchdog/mtx-1_wdt.c
drivers/watchdog/mv64x60_wdt.c
drivers/watchdog/nuc900_wdt.c
drivers/watchdog/nv_tco.c
drivers/watchdog/pc87413_wdt.c
drivers/watchdog/pcwd.c
drivers/watchdog/pcwd_pci.c
drivers/watchdog/pcwd_usb.c
drivers/watchdog/pika_wdt.c
drivers/watchdog/pnx833x_wdt.c
drivers/watchdog/rc32434_wdt.c
drivers/watchdog/rdc321x_wdt.c
drivers/watchdog/riowd.c
drivers/watchdog/sa1100_wdt.c
drivers/watchdog/sb_wdog.c
drivers/watchdog/sbc60xxwdt.c
drivers/watchdog/sbc7240_wdt.c
drivers/watchdog/sbc_epx_c3.c
drivers/watchdog/sbc_fitpc2_wdt.c
drivers/watchdog/sc1200wdt.c
drivers/watchdog/sc520_wdt.c
drivers/watchdog/sch311x_wdt.c
drivers/watchdog/scx200_wdt.c
drivers/watchdog/smsc37b787_wdt.c
drivers/watchdog/w83877f_wdt.c
drivers/watchdog/w83977f_wdt.c
drivers/watchdog/wafer5823wdt.c
drivers/watchdog/wdrtas.c
drivers/watchdog/wdt.c
drivers/watchdog/wdt285.c
drivers/watchdog/wdt977.c
drivers/watchdog/wdt_pci.c

Arnd

I have some older x86 boards so I probably could test some of them. Likely ALi chipset (alim1535_wdt.c and/or alim7101_wdt.c) and some super I/Os (it8712f_wdt.c, w83877f_wdt.c, w83977f_wdt.c).

Signature · 1 line
Ondrej Zary