Clean up legacy clock tick users
Subsystems (30)
cover Clean up legacy clock tick users
When I created the patch removing CONFIG_ARCH_GETTIMEOFFSET, I also had a look at CONFIG_GENERIC_CLOCKEVENTS, which is selected by most, but not all, platforms today, each of the ones that lack it doing the timer tick slightly differently.
The cleanups here make the old platforms a bit more consistent, in multiple ways:
- rather than selecting GENERIC_CLOCKEVENTS on modern platforms, select LEGACY_TIMER_TICK on the old ones.
- Hide some more of the internal implementation and only provide a single common entry point for the timer tick, which also makes the behavior more consistent.
- Remove the m68k and arm specific infrastructure pieces and call the common helper directly from each timer interrupt function.
I tested the series on m68k with the qemu q800 target platform, both with the legacy_timer_tick() implementation and after converting that to a periodic clockevent driver.
Arnd
Arnd Bergmann (13):
timekeeping: add CONFIG_LEGACY_TIMER_TICK
ia64: convert to legacy_timer_tick
ARM: rpc: use legacy_timer_tick
parisc: use legacy_timer_tick
m68k: coldfire: use legacy_timer_tick()
m68k: split heartbeat out of timer function
m68k: sun3/sun3c: use legacy_timer_tick
m68k: m68328: use legacy_timer_tick()
m68k: change remaining timers to legacy_timer_tick
m68k: remove timer_interrupt() function
timekeeping: remove xtime_update
timekeeping: default GENERIC_CLOCKEVENTS to enabled [RFC] m68k: mac: convert to generic clockevent
.../time/clockevents/arch-support.txt | 8 ++-- arch/alpha/Kconfig | 1 - arch/arc/Kconfig | 1 - arch/arm/Kconfig | 13 +----- arch/arm/include/asm/mach/time.h | 2 - arch/arm/kernel/time.c | 14 ------ arch/arm/mach-rpc/time.c | 2 +- arch/arm64/Kconfig | 1 - arch/arm64/Kconfig.platforms | 1 - arch/c6x/Kconfig | 1 - arch/csky/Kconfig | 1 - arch/h8300/Kconfig | 1 - arch/hexagon/Kconfig | 1 - arch/ia64/Kconfig | 1 + arch/ia64/kernel/time.c | 36 ++++++--------- arch/m68k/68000/timers.c | 7 ++- arch/m68k/Kconfig.cpu | 34 +++++++++++--- arch/m68k/Kconfig.machine | 15 +++++++ arch/m68k/amiga/config.c | 11 +++-- arch/m68k/apollo/config.c | 11 +++-- arch/m68k/atari/config.c | 2 +- arch/m68k/atari/time.c | 9 ++-- arch/m68k/bvme6000/config.c | 9 ++-- arch/m68k/coldfire/Makefile | 32 ++++++++------ arch/m68k/coldfire/pit.c | 2 +- arch/m68k/coldfire/sltimers.c | 8 ++-- arch/m68k/coldfire/timers.c | 8 ++-- arch/m68k/hp300/time.c | 8 ++-- arch/m68k/hp300/time.h | 2 +- arch/m68k/include/asm/machdep.h | 12 +++-- arch/m68k/kernel/setup_mm.c | 2 +- arch/m68k/kernel/setup_no.c | 2 +- arch/m68k/kernel/time.c | 18 ++------ arch/m68k/mac/config.c | 6 +-- arch/m68k/mac/via.c | 44 +++++++++++++++---- arch/m68k/mvme147/config.c | 9 ++-- arch/m68k/mvme16x/config.c | 9 ++-- arch/m68k/q40/config.c | 2 +- arch/m68k/q40/q40ints.c | 10 ++--- arch/m68k/sun3/config.c | 4 +- arch/m68k/sun3/sun3ints.c | 3 +- arch/m68k/sun3x/time.c | 5 +-- arch/m68k/sun3x/time.h | 2 +- arch/microblaze/Kconfig | 1 - arch/mips/Kconfig | 1 - arch/nds32/Kconfig | 1 - arch/nios2/Kconfig | 1 - arch/openrisc/Kconfig | 1 - arch/parisc/Kconfig | 2 +- arch/parisc/kernel/time.c | 9 ++-- arch/powerpc/Kconfig | 1 - arch/riscv/Kconfig | 1 - arch/s390/Kconfig | 1 - arch/sh/Kconfig | 1 - arch/sparc/Kconfig | 1 - arch/um/Kconfig | 1 - arch/x86/Kconfig | 1 - arch/xtensa/Kconfig | 1 - include/linux/timekeeping.h | 5 +-- kernel/time/Kconfig | 9 +++- kernel/time/Makefile | 1 + kernel/time/tick-legacy.c | 37 ++++++++++++++++ kernel/time/timekeeping.c | 16 ------- kernel/time/timekeeping.h | 1 + 64 files changed, 242 insertions(+), 221 deletions(-) create mode 100644 kernel/time/tick-legacy.c
Signature · 1 line
2.27.0
PATCH 1/13 timekeeping: add CONFIG_LEGACY_TIMER_TICK
All platforms that currently do not use generic clockevents roughly call the same set of functions in their timer interrupts: xtime_update(), update_process_times() and profile_tick(), sometimes in a different sequence.
Add a helper function that performs all three of them, to make the callers more uniform and simplify the interface.
--- include/linux/timekeeping.h | 1 + kernel/time/Kconfig | 7 +++++++ kernel/time/Makefile | 1 + kernel/time/tick-legacy.c | 19 +++++++++++++++++++ 4 files changed, 28 insertions(+) create mode 100644 kernel/time/tick-legacy.c
include/linux/timekeeping.h +1 -0
@@ -12,6 +12,7 @@ extern int timekeeping_suspended;/* Architecture timer tick functions: */ extern void update_process_times (int user);extern void xtime_update (unsigned long ticks); +extern void legacy_timer_tick (unsigned long ticks);/* * Get and set timeofday
kernel/time/Kconfig +7 -0
@@ -61,6 +61,13 @@ config POSIX_CPU_TIMERS_TASK_WORK bool default y if POSIX_TIMERS && HAVE_POSIX_CPU_TIMERS_TASK_WORK +config LEGACY_TIMER_TICK + bool + help + The legacy timer tick helper is used by platforms that + lack support for the generic clockevent framework. + New platforms should use generic clockevents instead. + if GENERIC_CLOCKEVENTS menu "Timers subsystem"
kernel/time/Makefile +1 -0
@@ -16,6 +16,7 @@ ifeq ($(CONFIG_GENERIC_CLOCKEVENTS_BROADCAST),y) endif obj-$(CONFIG_GENERIC_SCHED_CLOCK) += sched_clock.o obj-$(CONFIG_TICK_ONESHOT) += tick-oneshot.o tick-sched.o +obj-$(CONFIG_LEGACY_TIMER_TICK) += tick-legacy.o obj-$(CONFIG_HAVE_GENERIC_VDSO) += vsyscall.o obj-$(CONFIG_DEBUG_FS) += timekeeping_debug.o obj-$(CONFIG_TEST_UDELAY) += test_udelay.o
kernel/time/tick-legacy.c new file +19 -0
@@ -0,0 +1,19 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Timer tick function for architectures that lack generic clockevents, +* consolidated here from m68k/ia64/parisc/arm. +*/ + +#include <linux/irq.h> +#include <linux/profile.h> +#include <linux/timekeeper_internal.h> + +#include "tick-internal.h" + +void legacy_timer_tick (unsigned long ticks) +{ +if (ticks) +xtime_update (ticks); +update_process_times (user_mode (get_irq_regs ())); +profile_tick (CPU_PROFILING ); +}
Signature · 1 line
2.27.0
Re: [PATCH 01/13] timekeeping: add CONFIG_LEGACY_TIMER_TICK
Hi Arnd,
Quoted 75 lines · Arnd Bergmann
All platforms that currently do not use generic clockevents roughly call the same set of functions in their timer interrupts:
xtime_update(),update_process_times()andprofile_tick(), sometimes in a different sequence.Add a helper function that performs all three of them, to make the callers more uniform and simplify the interface.
Signed-off-by Arnd Bergmann <[email protected]>--- include/linux/timekeeping.h | 1 + kernel/time/Kconfig | 7 +++++++ kernel/time/Makefile | 1 + kernel/time/tick-legacy.c | 19 +++++++++++++++++++ 4 files changed, 28 insertions(+) create mode 100644 kernel/time/tick-legacy.cinclude/linux/timekeeping.h +1
-0@@ -12,6 +12,7 @@ extern int timekeeping_suspended;/* Architecture timer tick functions: */ extern void update_process_times (int user);extern void xtime_update (unsigned long ticks); +extern void legacy_timer_tick (unsigned long ticks);/* * Get and set timeofday kernel/time/Kconfig +7
-0@@ -61,6 +61,13 @@ config POSIX_CPU_TIMERS_TASK_WORK bool default y if POSIX_TIMERS && HAVE_POSIX_CPU_TIMERS_TASK_WORK +config LEGACY_TIMER_TICK + bool + help + The legacy timer tick helper is used by platforms that + lack support for the generic clockevent framework. + New platforms should use generic clockevents instead. + if GENERIC_CLOCKEVENTS menu "Timers subsystem"kernel/time/Makefile +1
-0@@ -16,6 +16,7 @@ ifeq ($(CONFIG_GENERIC_CLOCKEVENTS_BROADCAST),y) endif obj-$(CONFIG_GENERIC_SCHED_CLOCK) += sched_clock.o obj-$(CONFIG_TICK_ONESHOT) += tick-oneshot.o tick-sched.o +obj-$(CONFIG_LEGACY_TIMER_TICK) += tick-legacy.o obj-$(CONFIG_HAVE_GENERIC_VDSO) += vsyscall.o obj-$(CONFIG_DEBUG_FS) += timekeeping_debug.o obj-$(CONFIG_TEST_UDELAY) += test_udelay.okernel/time/tick-legacy.c new file +9
-0@@ -0,0 +1,19 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* Timer tick function for architectures that lack generic clockevents, +* consolidated here from m68k/ia64/parisc/arm. +*/ + +#include <linux/irq.h> +#include <linux/profile.h> +#include <linux/timekeeper_internal.h>
kernel/time/tick-legacy.c line 19
Quoted 10 lines
kernel/time/tick-legacy.c new file +10
-0+ +#include "tick-internal.h" + +void legacy_timer_tick (unsigned long ticks) +{ +if (ticks) +xtime_update (ticks); +update_process_times (user_mode (get_irq_regs ())); +profile_tick (CPU_PROFILING ); +}
It's good to see this code refactored in this way because, as well as de-duplication, it reveals the logic that's common to the relevant platforms and may shed some light on the need for that logic.
Yet it's not clear to me that the clockevents framework is able to replace that logic on all of the affected hardware. I suppose it remains to be seen.
I hate to quibble about naming, but you seem to be using "legacy" here to mean "deprecated" (?) Is it a good idea to prepend such adjectives to symbol names?
IMO, the term "legacy" is redundant in this context. That term covers a large portion of kernel code, a large number of hardware features in current silicon, a large portion of the userspace ABI, a large number of production Linux systems, probably all "Unix" systems, etc.
As a corollary, cutting edge ("non-legacy") code is often kept out of open source projects by the owners of the intellectual property rights.
Re: [PATCH 01/13] timekeeping: add CONFIG_LEGACY_TIMER_TICK
Quoted 9 lines · Finn Thain
On Thu, 8 Oct 2020, Arnd Bergmann wrote:
It's good to see this code refactored in this way because, as well as de-duplication, it reveals the logic that's common to the relevant platforms and may shed some light on the need for that logic.
Yet it's not clear to me that the clockevents framework is able to replace that logic on all of the affected hardware. I suppose it remains to be seen.
I suspect that the change I did for one platform in patch 13/13 could be duplicated for all 16 platforms, adding lots of trivial clockevent drivers that only support periodic ticks, but any platform that can instead support oneshot timers should probably do that, or it won't provide any better behavior.
What do others think we should do here?
Quoted 2 lines
As a corollary, cutting edge ("non-legacy") code is often kept out of open source projects by the owners of the intellectual property rights.
I'm happy to change the name in any way if you have a suggestion that the clock event maintainers (Daniel and Thomas) like.
Arnd
Re: [PATCH 01/13] timekeeping: add CONFIG_LEGACY_TIMER_TICK
In reply to Arnd BergmannQuoted 9 lines · Arnd Bergmann
All platforms that currently do not use generic clockevents roughly call the same set of functions in their timer interrupts:
xtime_update(),update_process_times()andprofile_tick(), sometimes in a different sequence.Add a helper function that performs all three of them, to make the callers more uniform and simplify the interface.
Signed-off-by Arnd Bergmann <[email protected]>
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
PATCH 2/13 ia64: convert to legacy_timer_tick
In reply to Arnd Bergmannia64 is the only architecture that calls xtime_update() in a loop, once for each jiffie that has passed since the last event.
Before commit 3171a0305d62 ("[PATCH] simplify update_times (avoid jiffies/jiffies_64 aliasing problem)") in 2006, it could not actually do this any differently, but now it seems simpler to just pass the number of jiffies that passed in the meantime.
While this loses the ability process interrupts in the middle of the timer tick by calling local_irq_enable(), doing so is fairly peculiar anyway and it seems better to just do what everyone else does here.
--- arch/ia64/Kconfig | 1 + arch/ia64/kernel/time.c | 36 +++++++++++++----------------------- 2 files changed, 14 insertions(+), 23 deletions(-)
arch/ia64/Kconfig +1 -0
@@ -46,6 +46,7 @@ config IA64 select ARCH_THREAD_STACK_ALLOCATOR select ARCH_CLOCKSOURCE_DATA select GENERIC_TIME_VSYSCALL + select LEGACY_TIMER_TICK select SWIOTLB select SYSCTL_ARCH_UNALIGN_NO_WARN select HAVE_MOD_ARCH_SPECIFIC
arch/ia64/kernel/time.c +13 -23
@@ -161,39 +161,29 @@ void vtime_account_idle(struct task_struct *tsk)static irqreturn_t timer_interrupt (int irq,void *dev_id) { -unsigned long new_itm; +unsigned long cur_itm, new_itm, ticks;if (cpu_is_offline (smp_processor_id ())) {return IRQ_HANDLED ; } new_itm = local_cpu_data->itm_next; + cur_itm =ia64_get_itc (); -if (!time_after (ia64_get_itc (), new_itm)) +if (!time_after (cur_itm, new_itm)) {printk (KERN_ERR "Oops: timer tick before it's due (itc=%lx,itm=%lx)\n" , -ia64_get_itc (), new_itm); - -profile_tick (CPU_PROFILING ); - -while (1 ) { -update_process_times (user_mode (get_irq_regs ())); - - new_itm += local_cpu_data->itm_delta; - -if (smp_processor_id () = time_keeper_id) -xtime_update (1 ); - - local_cpu_data->itm_next = new_itm; + cur_itm, new_itm); + ticks =1 ; + }else { + ticks =DIV_ROUND_UP (cur_itm - new_itm, + local_cpu_data->itm_delta); + new_itm += ticks * local_cpu_data->itm_delta; + } -if (time_after (new_itm,ia64_get_itc ())) -break ; +if (smp_processor_id () != time_keeper_id) + ticks =0 ; -/* -* Allow IPIs to interrupt the timer loop. -*/ -local_irq_enable (); -local_irq_disable (); - } +legacy_timer_tick (ticks); do { /*
Signature · 1 line
2.27.0
PATCH 3/13 ARM: rpc: use legacy_timer_tick
In reply to Arnd Bergmannrpc is the only user of the timer_tick() function now, and can just call the newly added generic version instead.
--- arch/arm/Kconfig | 1 + arch/arm/include/asm/mach/time.h | 2 -- arch/arm/kernel/time.c | 14 -------------- arch/arm/mach-rpc/time.c | 2 +- 4 files changed, 2 insertions(+), 17 deletions(-)
arch/arm/Kconfig +1 -0
@@ -467,6 +467,7 @@ config ARCH_RPC select HAVE_IDE select HAVE_PATA_PLATFORM select ISA_DMA_API + select LEGACY_TIMER_TICK select NEED_MACH_IO_H select NEED_MACH_MEMORY_H select NO_IOPORT_MAP
arch/arm/include/asm/mach/time.h +0 -2
@@ -7,8 +7,6 @@#ifndef __ASM_ARM_MACH_TIME_H#define __ASM_ARM_MACH_TIME_H -extern void timer_tick (void ); -typedef void (*clock_access_fn )(struct timespec64 *);extern int register_persistent_clock (clock_access_fn read_persistent);
arch/arm/kernel/time.c +0 -14
@@ -60,20 +60,6 @@ unsigned long profile_pc(struct pt_regs *regs)EXPORT_SYMBOL (profile_pc);#endif -#ifndef CONFIG_GENERIC_CLOCKEVENTS -/* -* Kernel system timer support. -*/ -void timer_tick (void ) -{ -profile_tick (CPU_PROFILING ); -xtime_update (1 ); -#ifndef CONFIG_SMP -update_process_times (user_mode (get_irq_regs ())); -#endif -} -#endif -static void dummy_clock_access (struct timespec64 *ts) { ts->tv_sec =0 ;
arch/arm/mach-rpc/time.c +1 -1
@@ -81,7 +81,7 @@ static irqreturn_t ioc_timer_interrupt(int irq, void *dev_id) { ioc_time += RPC_LATCH; -timer_tick (); +legacy_timer_tick (1 );return IRQ_HANDLED ; }
Signature · 1 line
2.27.0
PATCH 4/13 parisc: use legacy_timer_tick
In reply to Arnd Bergmannparisc has selected CONFIG_GENERIC_CLOCKEVENTS since commit 43b1f6abd590 ("parisc: Switch to generic sched_clock implementation"), but does not appear to actually be using it, and instead calls the low-level timekeeping functions directly.
Remove the GENERIC_CLOCKEVENTS select again, and instead convert to the newly added legacy_timer_tick() helper.
--- Documentation/features/time/clockevents/arch-support.txt | 2 +- arch/parisc/Kconfig | 2 +- arch/parisc/kernel/time.c | 9 +++------ 3 files changed, 5 insertions(+), 8 deletions(-)
Documentation/features/time/clockevents/arch-support.txt +1 -1
@@ -21,7 +21,7 @@ | nds32: | ok | | nios2: | ok | | openrisc: | ok | - | parisc: | ok | + | parisc: | TODO | | powerpc: | ok | | riscv: | ok | | s390: | ok |
arch/parisc/Kconfig +1 -1
@@ -52,7 +52,7 @@ config PARISC select HAVE_REGS_AND_STACK_ACCESS_API select GENERIC_SCHED_CLOCK select HAVE_UNSTABLE_SCHED_CLOCK if SMP - select GENERIC_CLOCKEVENTS + select LEGACY_TIMER_TICK select CPU_NO_EFFICIENT_FFS select NEED_DMA_MAP_STATE select NEED_SG_DMA_LENGTH
arch/parisc/kernel/time.c +3 -6
@@ -70,8 +70,6 @@ irqreturn_t __irq_entry timer_interrupt(int irq, void *dev_id)/* gcc can optimize for "read-only" case with a local clocktick */ unsigned long cpt = clocktick; -profile_tick (CPU_PROFILING ); -/* Initialize next_tick to the old expected tick time. */ next_tick = cpuinfo->it_value;
@@ -86,10 +84,9 @@ irqreturn_t __irq_entry timer_interrupt(int irq, void *dev_id) cpuinfo->it_value = next_tick;/* Go do system house keeping. */ -if (cpu =0 ) -xtime_update (ticks_elapsed); - -update_process_times (user_mode (get_irq_regs ())); +if (cpu !=0 ) + ticks_elapsed =0 ; +legacy_timer_tick (ticks_elapsed);/* Skip clockticks on purpose if we know we would miss those. * The new CR16 must be "later" than current CR16 otherwise
Signature · 1 line
2.27.0
PATCH 5/13 m68k: coldfire: use legacy_timer_tick()
In reply to Arnd BergmannReplace the indirect function calls in the timer code with direct calls to the newly added legacy_timer_tick() helper for those that have not yet been converted to generic clockevents.
This makes the timer code a little more self-contained.
--- arch/m68k/Kconfig.cpu | 35 +++++++++++++++++++++++++++++------ arch/m68k/coldfire/Makefile | 32 ++++++++++++++++++-------------- arch/m68k/coldfire/sltimers.c | 6 ++---- arch/m68k/coldfire/timers.c | 6 ++---- 4 files changed, 51 insertions(+), 28 deletions(-)
arch/m68k/Kconfig.cpu +29 -6
@@ -137,6 +137,7 @@ config M5206 bool "MCF5206" depends on !MMU select COLDFIRE_SW_A7 + select COLDFIRE_TIMERS select HAVE_MBAR select CPU_NO_EFFICIENT_FFS help
@@ -146,6 +147,7 @@ config M5206e bool "MCF5206e" depends on !MMU select COLDFIRE_SW_A7 + select COLDFIRE_TIMERS select HAVE_MBAR select CPU_NO_EFFICIENT_FFS help
@@ -154,7 +156,7 @@ config M5206e config M520x bool "MCF520x" depends on !MMU - select GENERIC_CLOCKEVENTS + select COLDFIRE_PIT_TIMER select HAVE_CACHE_SPLIT help Freescale Coldfire 5207/5208 processor support.
@@ -162,7 +164,7 @@ config M520x config M523x bool "MCF523x" depends on !MMU - select GENERIC_CLOCKEVENTS + select COLDFIRE_PIT_TIMER select HAVE_CACHE_SPLIT select HAVE_IPSBAR help
@@ -172,6 +174,7 @@ config M5249 bool "MCF5249" depends on !MMU select COLDFIRE_SW_A7 + select COLDFIRE_TIMERS select HAVE_MBAR select CPU_NO_EFFICIENT_FFS help
@@ -181,6 +184,7 @@ config M525x bool "MCF525x" depends on !MMU select COLDFIRE_SW_A7 + select COLDFIRE_TIMERS select HAVE_MBAR select CPU_NO_EFFICIENT_FFS help
@@ -189,10 +193,10 @@ config M525x config M5271 bool "MCF5271" depends on !MMU + select COLDFIRE_PIT_TIMER select M527x select HAVE_CACHE_SPLIT select HAVE_IPSBAR - select GENERIC_CLOCKEVENTS help Freescale (Motorola) ColdFire 5270/5271 processor support.
@@ -200,6 +204,7 @@ config M5272 bool "MCF5272" depends on !MMU select COLDFIRE_SW_A7 + select COLDFIRE_TIMERS select HAVE_MBAR select CPU_NO_EFFICIENT_FFS help
@@ -208,17 +213,17 @@ config M5272 config M5275 bool "MCF5275" depends on !MMU + select COLDFIRE_PIT_TIMER select M527x select HAVE_CACHE_SPLIT select HAVE_IPSBAR - select GENERIC_CLOCKEVENTS help Freescale (Motorola) ColdFire 5274/5275 processor support. config M528x bool "MCF528x" depends on !MMU - select GENERIC_CLOCKEVENTS + select COLDFIRE_PIT_TIMER select HAVE_CACHE_SPLIT select HAVE_IPSBAR help
@@ -227,6 +232,7 @@ config M528x config M5307 bool "MCF5307" depends on !MMU + select COLDFIRE_TIMERS select COLDFIRE_SW_A7 select HAVE_CACHE_CB select HAVE_MBAR
@@ -237,6 +243,7 @@ config M5307 config M532x bool "MCF532x" depends on !MMU + select COLDFIRE_TIMERS select M53xx select HAVE_CACHE_CB help
@@ -245,6 +252,7 @@ config M532x config M537x bool "MCF537x" depends on !MMU + select COLDFIRE_TIMERS select M53xx select HAVE_CACHE_CB help
@@ -254,6 +262,7 @@ config M5407 bool "MCF5407" depends on !MMU select COLDFIRE_SW_A7 + select COLDFIRE_TIMERS select HAVE_CACHE_CB select HAVE_MBAR select CPU_NO_EFFICIENT_FFS
@@ -263,6 +272,7 @@ config M5407 config M547x bool "MCF547x" select M54xx + select COLDFIRE_SLTIMERS select MMU_COLDFIRE if MMU select FPU if MMU select HAVE_CACHE_CB
@@ -273,6 +283,7 @@ config M547x config M548x bool "MCF548x" + select COLDFIRE_SLTIMERS select MMU_COLDFIRE if MMU select FPU if MMU select M54xx
@@ -284,8 +295,8 @@ config M548x config M5441x bool "MCF5441x" + select COLDFIRE_PIT_TIMER select MMU_COLDFIRE if MMU - select GENERIC_CLOCKEVENTS select HAVE_CACHE_CB help Freescale Coldfire 54410/54415/54416/54417/54418 processor support.
@@ -302,6 +313,18 @@ config M54xx select HAVE_PCI bool +config COLDFIRE_PIT_TIMER + bool + select GENERIC_CLOCKEVENTS + +config COLDFIRE_TIMERS + bool + select LEGACY_TIMER_TICK + +config COLDFIRE_SLTIMERS + bool + select LEGACY_TIMER_TICK + endif # COLDFIRE
arch/m68k/coldfire/Makefile +18 -14
@@ -16,20 +16,24 @@ asflags-$(CONFIG_FULLDEBUG) := -DDEBUGGER_COMPATIBLE_CACHE=1 obj-$(CONFIG_COLDFIRE) += cache.o clk.o device.o dma.o entry.o vectors.o -obj-$(CONFIG_M5206) += m5206.o timers.o intc.o reset.o -obj-$(CONFIG_M5206e) += m5206.o timers.o intc.o reset.o -obj-$(CONFIG_M520x) += m520x.o pit.o intc-simr.o reset.o -obj-$(CONFIG_M523x) += m523x.o pit.o dma_timer.o intc-2.o reset.o -obj-$(CONFIG_M5249) += m5249.o timers.o intc.o intc-5249.o reset.o -obj-$(CONFIG_M525x) += m525x.o timers.o intc.o intc-525x.o reset.o -obj-$(CONFIG_M527x) += m527x.o pit.o intc-2.o reset.o -obj-$(CONFIG_M5272) += m5272.o intc-5272.o timers.o -obj-$(CONFIG_M528x) += m528x.o pit.o intc-2.o reset.o -obj-$(CONFIG_M5307) += m5307.o timers.o intc.o reset.o -obj-$(CONFIG_M53xx) += m53xx.o timers.o intc-simr.o reset.o -obj-$(CONFIG_M5407) += m5407.o timers.o intc.o reset.o -obj-$(CONFIG_M54xx) += m54xx.o sltimers.o intc-2.o -obj-$(CONFIG_M5441x) += m5441x.o pit.o intc-simr.o reset.o +obj-$(CONFIG_M5206) += m5206.o intc.o reset.o +obj-$(CONFIG_M5206e) += m5206.o intc.o reset.o +obj-$(CONFIG_M520x) += m520x.o intc-simr.o reset.o +obj-$(CONFIG_M523x) += m523x.o dma_timer.o intc-2.o reset.o +obj-$(CONFIG_M5249) += m5249.o intc.o intc-5249.o reset.o +obj-$(CONFIG_M525x) += m525x.o intc.o intc-525x.o reset.o +obj-$(CONFIG_M527x) += m527x.o intc-2.o reset.o +obj-$(CONFIG_M5272) += m5272.o intc-5272.o +obj-$(CONFIG_M528x) += m528x.o intc-2.o reset.o +obj-$(CONFIG_M5307) += m5307.o intc.o reset.o +obj-$(CONFIG_M53xx) += m53xx.o intc-simr.o reset.o +obj-$(CONFIG_M5407) += m5407.o intc.o reset.o +obj-$(CONFIG_M54xx) += m54xx.o intc-2.o +obj-$(CONFIG_M5441x) += m5441x.o intc-simr.o reset.o + +obj-$(CONFIG_COLDFIRE_PIT_TIMER) += pit.o +obj-$(CONFIG_COLDFIRE_TIMERS) += timers.o +obj-$(CONFIG_COLDFIRE_SLTIMERS) += sltimers.o obj-$(CONFIG_NETtel) += nettel.o obj-$(CONFIG_CLEOPATRA) += nettel.o
arch/m68k/coldfire/sltimers.c +2 -4
@@ -83,14 +83,13 @@ void mcfslt_profile_init(void)static u32 mcfslt_cycles_per_jiffy;static u32 mcfslt_cnt; -static irq_handler_t timer_interrupt; -static irqreturn_t mcfslt_tick (int irq,void *dummy) {/* Reset Slice Timer 0 */ __raw_writel (MCFSLT_SSR_BE |MCFSLT_SSR_TE ,TA (MCFSLT_SSR )); mcfslt_cnt += mcfslt_cycles_per_jiffy; -return timer_interrupt (irq, dummy); +legacy_timer_tick (1 ); +return IRQ_HANDLED ; }static u64 mcfslt_read_clk (struct clocksource *cs)
@@ -136,7 +135,6 @@ void hw_timer_init(irq_handler_t handler)/* initialize mcfslt_cnt knowing that slice timers count down */ mcfslt_cnt = mcfslt_cycles_per_jiffy; - timer_interrupt = handler; r =request_irq (MCF_IRQ_TIMER , mcfslt_tick,IRQF_TIMER ,"timer" ,NULL );if (r) { pr_err("Failed to request irq %d (timer): %pe\n", MCF_IRQ_TIMER,
arch/m68k/coldfire/timers.c +2 -4
@@ -48,8 +48,6 @@ void coldfire_profile_init(void);static u32 mcftmr_cycles_per_jiffy;static u32 mcftmr_cnt; -static irq_handler_t timer_interrupt; -/***************************************************************************/ static void init_timer_irq (void )
@@ -77,7 +75,8 @@ static irqreturn_t mcftmr_tick(int irq, void *dummy)__raw_writeb (MCFTIMER_TER_CAP |MCFTIMER_TER_REF ,TA (MCFTIMER_TER )); mcftmr_cnt += mcftmr_cycles_per_jiffy; -return timer_interrupt (irq, dummy); +legacy_timer_tick (1 ); +return IRQ_HANDLED ; }/***************************************************************************/
@@ -126,7 +125,6 @@ void hw_timer_init(irq_handler_t handler)clocksource_register_hz (&mcftmr_clk,FREQ ); - timer_interrupt = handler;init_timer_irq (); r =request_irq (MCF_IRQ_TIMER , mcftmr_tick,IRQF_TIMER ,"timer" ,NULL );if (r) {
Signature · 1 line
2.27.0
Re: [PATCH 05/13] m68k: coldfire: use legacy_timer_tick()
Hi Arnd,
Quoted 8 lines · Arnd Bergmann
Replace the indirect function calls in the timer code with direct calls to the newly added
legacy_timer_tick()helper for those that have not yet been converted to generic clockevents.This makes the timer code a little more self-contained.
Signed-off-by Arnd Bergmann <[email protected]>
I tested this series on a couple of different ColdFire parts (5208 and 5475) and under QEMU emulating the 5208. All checked out good, all worked as expected. So for the ColdFire changes:
Regards
Greg
Quoted 282 lines
--- arch/m68k/Kconfig.cpu | 35 +++++++++++++++++++++++++++++------ arch/m68k/coldfire/Makefile | 32 ++++++++++++++++++-------------- arch/m68k/coldfire/sltimers.c | 6 ++---- arch/m68k/coldfire/timers.c | 6 ++---- 4 files changed, 51 insertions(+), 28 deletions(-)arch/m68k/Kconfig.cpu +29
-6@@ -137,6 +137,7 @@ config M5206 bool "MCF5206" depends on !MMU select COLDFIRE_SW_A7 + select COLDFIRE_TIMERS select HAVE_MBAR select CPU_NO_EFFICIENT_FFS help@@ -146,6 +147,7 @@ config M5206e bool "MCF5206e" depends on !MMU select COLDFIRE_SW_A7 + select COLDFIRE_TIMERS select HAVE_MBAR select CPU_NO_EFFICIENT_FFS help@@ -154,7 +156,7 @@ config M5206e config M520x bool "MCF520x" depends on !MMU - select GENERIC_CLOCKEVENTS + select COLDFIRE_PIT_TIMER select HAVE_CACHE_SPLIT help Freescale Coldfire 5207/5208 processor support.@@ -162,7 +164,7 @@ config M520x config M523x bool "MCF523x" depends on !MMU - select GENERIC_CLOCKEVENTS + select COLDFIRE_PIT_TIMER select HAVE_CACHE_SPLIT select HAVE_IPSBAR help@@ -172,6 +174,7 @@ config M5249 bool "MCF5249" depends on !MMU select COLDFIRE_SW_A7 + select COLDFIRE_TIMERS select HAVE_MBAR select CPU_NO_EFFICIENT_FFS help@@ -181,6 +184,7 @@ config M525x bool "MCF525x" depends on !MMU select COLDFIRE_SW_A7 + select COLDFIRE_TIMERS select HAVE_MBAR select CPU_NO_EFFICIENT_FFS help@@ -189,10 +193,10 @@ config M525x config M5271 bool "MCF5271" depends on !MMU + select COLDFIRE_PIT_TIMER select M527x select HAVE_CACHE_SPLIT select HAVE_IPSBAR - select GENERIC_CLOCKEVENTS help Freescale (Motorola) ColdFire 5270/5271 processor support.@@ -200,6 +204,7 @@ config M5272 bool "MCF5272" depends on !MMU select COLDFIRE_SW_A7 + select COLDFIRE_TIMERS select HAVE_MBAR select CPU_NO_EFFICIENT_FFS help@@ -208,17 +213,17 @@ config M5272 config M5275 bool "MCF5275" depends on !MMU + select COLDFIRE_PIT_TIMER select M527x select HAVE_CACHE_SPLIT select HAVE_IPSBAR - select GENERIC_CLOCKEVENTS help Freescale (Motorola) ColdFire 5274/5275 processor support. config M528x bool "MCF528x" depends on !MMU - select GENERIC_CLOCKEVENTS + select COLDFIRE_PIT_TIMER select HAVE_CACHE_SPLIT select HAVE_IPSBAR help@@ -227,6 +232,7 @@ config M528x config M5307 bool "MCF5307" depends on !MMU + select COLDFIRE_TIMERS select COLDFIRE_SW_A7 select HAVE_CACHE_CB select HAVE_MBAR@@ -237,6 +243,7 @@ config M5307 config M532x bool "MCF532x" depends on !MMU + select COLDFIRE_TIMERS select M53xx select HAVE_CACHE_CB help@@ -245,6 +252,7 @@ config M532x config M537x bool "MCF537x" depends on !MMU + select COLDFIRE_TIMERS select M53xx select HAVE_CACHE_CB help@@ -254,6 +262,7 @@ config M5407 bool "MCF5407" depends on !MMU select COLDFIRE_SW_A7 + select COLDFIRE_TIMERS select HAVE_CACHE_CB select HAVE_MBAR select CPU_NO_EFFICIENT_FFS@@ -263,6 +272,7 @@ config M5407 config M547x bool "MCF547x" select M54xx + select COLDFIRE_SLTIMERS select MMU_COLDFIRE if MMU select FPU if MMU select HAVE_CACHE_CB@@ -273,6 +283,7 @@ config M547x config M548x bool "MCF548x" + select COLDFIRE_SLTIMERS select MMU_COLDFIRE if MMU select FPU if MMU select M54xx@@ -284,8 +295,8 @@ config M548x config M5441x bool "MCF5441x" + select COLDFIRE_PIT_TIMER select MMU_COLDFIRE if MMU - select GENERIC_CLOCKEVENTS select HAVE_CACHE_CB help Freescale Coldfire 54410/54415/54416/54417/54418 processor support.@@ -302,6 +313,18 @@ config M54xx select HAVE_PCI bool +config COLDFIRE_PIT_TIMER + bool + select GENERIC_CLOCKEVENTS + +config COLDFIRE_TIMERS + bool + select LEGACY_TIMER_TICK + +config COLDFIRE_SLTIMERS + bool + select LEGACY_TIMER_TICK + endif # COLDFIREarch/m68k/coldfire/Makefile +18
-14@@ -16,20 +16,24 @@ asflags-$(CONFIG_FULLDEBUG) := -DDEBUGGER_COMPATIBLE_CACHE=1 obj-$(CONFIG_COLDFIRE) += cache.o clk.o device.o dma.o entry.o vectors.o -obj-$(CONFIG_M5206) += m5206.o timers.o intc.o reset.o -obj-$(CONFIG_M5206e) += m5206.o timers.o intc.o reset.o -obj-$(CONFIG_M520x) += m520x.o pit.o intc-simr.o reset.o -obj-$(CONFIG_M523x) += m523x.o pit.o dma_timer.o intc-2.o reset.o -obj-$(CONFIG_M5249) += m5249.o timers.o intc.o intc-5249.o reset.o -obj-$(CONFIG_M525x) += m525x.o timers.o intc.o intc-525x.o reset.o -obj-$(CONFIG_M527x) += m527x.o pit.o intc-2.o reset.o -obj-$(CONFIG_M5272) += m5272.o intc-5272.o timers.o -obj-$(CONFIG_M528x) += m528x.o pit.o intc-2.o reset.o -obj-$(CONFIG_M5307) += m5307.o timers.o intc.o reset.o -obj-$(CONFIG_M53xx) += m53xx.o timers.o intc-simr.o reset.o -obj-$(CONFIG_M5407) += m5407.o timers.o intc.o reset.o -obj-$(CONFIG_M54xx) += m54xx.o sltimers.o intc-2.o -obj-$(CONFIG_M5441x) += m5441x.o pit.o intc-simr.o reset.o +obj-$(CONFIG_M5206) += m5206.o intc.o reset.o +obj-$(CONFIG_M5206e) += m5206.o intc.o reset.o +obj-$(CONFIG_M520x) += m520x.o intc-simr.o reset.o +obj-$(CONFIG_M523x) += m523x.o dma_timer.o intc-2.o reset.o +obj-$(CONFIG_M5249) += m5249.o intc.o intc-5249.o reset.o +obj-$(CONFIG_M525x) += m525x.o intc.o intc-525x.o reset.o +obj-$(CONFIG_M527x) += m527x.o intc-2.o reset.o +obj-$(CONFIG_M5272) += m5272.o intc-5272.o +obj-$(CONFIG_M528x) += m528x.o intc-2.o reset.o +obj-$(CONFIG_M5307) += m5307.o intc.o reset.o +obj-$(CONFIG_M53xx) += m53xx.o intc-simr.o reset.o +obj-$(CONFIG_M5407) += m5407.o intc.o reset.o +obj-$(CONFIG_M54xx) += m54xx.o intc-2.o +obj-$(CONFIG_M5441x) += m5441x.o intc-simr.o reset.o + +obj-$(CONFIG_COLDFIRE_PIT_TIMER) += pit.o +obj-$(CONFIG_COLDFIRE_TIMERS) += timers.o +obj-$(CONFIG_COLDFIRE_SLTIMERS) += sltimers.o obj-$(CONFIG_NETtel) += nettel.o obj-$(CONFIG_CLEOPATRA) += nettel.oarch/m68k/coldfire/sltimers.c +2
-4@@ -83,14 +83,13 @@ void mcfslt_profile_init(void)static u32 mcfslt_cycles_per_jiffy;static u32 mcfslt_cnt; -static irq_handler_t timer_interrupt; -static irqreturn_t mcfslt_tick (int irq,void *dummy) {/* Reset Slice Timer 0 */ __raw_writel (MCFSLT_SSR_BE |MCFSLT_SSR_TE ,TA (MCFSLT_SSR )); mcfslt_cnt += mcfslt_cycles_per_jiffy; -return timer_interrupt (irq, dummy); +legacy_timer_tick (1 ); +return IRQ_HANDLED ; }static u64 mcfslt_read_clk (struct clocksource *cs)@@ -136,7 +135,6 @@ void hw_timer_init(irq_handler_t handler)/* initialize mcfslt_cnt knowing that slice timers count down */ mcfslt_cnt = mcfslt_cycles_per_jiffy; - timer_interrupt = handler; r =request_irq (MCF_IRQ_TIMER , mcfslt_tick,IRQF_TIMER ,"timer" ,NULL );if (r) { pr_err("Failed to request irq %d (timer): %pe\n", MCF_IRQ_TIMER,arch/m68k/coldfire/timers.c +2
-4@@ -48,8 +48,6 @@ void coldfire_profile_init(void);static u32 mcftmr_cycles_per_jiffy;static u32 mcftmr_cnt; -static irq_handler_t timer_interrupt; -/***************************************************************************/ static void init_timer_irq (void )@@ -77,7 +75,8 @@ static irqreturn_t mcftmr_tick(int irq, void *dummy)__raw_writeb (MCFTIMER_TER_CAP |MCFTIMER_TER_REF ,TA (MCFTIMER_TER )); mcftmr_cnt += mcftmr_cycles_per_jiffy; -return timer_interrupt (irq, dummy); +legacy_timer_tick (1 ); +return IRQ_HANDLED ; }/***************************************************************************/ @@ -126,7 +125,6 @@ void hw_timer_init(irq_handler_t handler)clocksource_register_hz (&mcftmr_clk,FREQ ); - timer_interrupt = handler;init_timer_irq (); r =request_irq (MCF_IRQ_TIMER , mcftmr_tick,IRQF_TIMER ,"timer" ,NULL );if (r) {
Re: [PATCH 05/13] m68k: coldfire: use legacy_timer_tick()
Quoted 11 lines · Greg Ungerer
Hi Arnd,
Quoted 7 lines · Arnd Bergmann
Replace the indirect function calls in the timer code with direct calls to the newly added
legacy_timer_tick()helper for those that have not yet been converted to generic clockevents.This makes the timer code a little more self-contained.
Quoted 7 lines
Quoted 1 line
Signed-off-by Arnd Bergmann <[email protected]>I tested this series on a couple of different ColdFire parts (5208 and 5475) and under QEMU emulating the 5208. All checked out good, all worked as expected. So for the ColdFire changes:
Tested-by Greg Ungerer <[email protected]>
Awesome, thanks for testing!
Arnd
PATCH 6/13 m68k: split heartbeat out of timer function
In reply to Arnd BergmannThe heartbeat functionality is mostly separate from the actual timer interrupt handling, and it is only used on five platforms.
Split it out into a separate function and call that directly from the timer irq on those platforms.
--- arch/m68k/amiga/config.c | 1 + arch/m68k/apollo/config.c | 1 + arch/m68k/atari/time.c | 2 ++ arch/m68k/hp300/time.c | 1 + arch/m68k/include/asm/machdep.h | 7 +++++++ arch/m68k/kernel/time.c | 7 +++++-- arch/m68k/q40/q40ints.c | 2 ++ 7 files changed, 19 insertions(+), 2 deletions(-)
arch/m68k/amiga/config.c +1 -0
@@ -480,6 +480,7 @@ static irqreturn_t ciab_timer_handler(int irq, void *dev_id) clk_total += jiffy_ticks; clk_offset =0 ;timer_routine (0 ,NULL ); +timer_heartbeat ();return IRQ_HANDLED ; }
arch/m68k/apollo/config.c +1 -0
@@ -173,6 +173,7 @@ irqreturn_t dn_timer_int(int irq, void *dev_id)volatile unsigned char x;timer_handler (irq, dev_id); +timer_heartbeat (); x = *(volatile unsigned char *)(apollo_timer +3 ); x = *(volatile unsigned char *)(apollo_timer +5 );
arch/m68k/atari/time.c +2 -0
@@ -21,6 +21,7 @@#include <linux/export.h> #include <asm/atariints.h> +#include <asm/machdep.h> DEFINE_SPINLOCK (rtc_lock);EXPORT_SYMBOL_GPL (rtc_lock);
@@ -49,6 +50,7 @@ static irqreturn_t mfp_timer_c_handler(int irq, void *dev_id) }while (last_timer_count =1 ); clk_total +=INT_TICKS ;timer_routine (0 ,NULL ); +timer_heartbeat ();local_irq_restore (flags);return IRQ_HANDLED ;
arch/m68k/hp300/time.c +1 -0
@@ -65,6 +65,7 @@ static irqreturn_t hp300_tick(int irq, void *dev_id) clk_total +=INTVAL ; clk_offset =0 ;timer_routine (0 ,NULL ); +timer_heartbeat ();local_irq_restore (flags);/* Turn off the network and SCSI leds */
arch/m68k/include/asm/machdep.h +7 -0
@@ -35,6 +35,13 @@ extern void (*mach_beep) (unsigned int, unsigned int);/* Hardware clock functions */ extern void hw_timer_init (irq_handler_t handler);extern unsigned long hw_timer_offset (void ); +#ifdef CONFIG_HEARTBEAT +extern void timer_heartbeat (void ); +#else +static inline void timer_heartbeat (void ) +{ +} +#endif extern void config_BSP (char *command,int len);
arch/m68k/kernel/time.c +5 -2
@@ -45,8 +45,12 @@ static irqreturn_t timer_interrupt(int irq, void *dummy)xtime_update (1 );update_process_times (user_mode (get_irq_regs ()));profile_tick (CPU_PROFILING ); +return IRQ_HANDLED ; +}#ifdef CONFIG_HEARTBEAT +void timer_heartbeat (void ) +{/* use power LED as a heartbeat instead -- much more useful for debugging -- based on the version for PReP by Cort */ /* acts like an actual heart beat -- ie thump-thump-pause... */
@@ -68,9 +72,8 @@ static irqreturn_t timer_interrupt(int irq, void *dummy) dist = period / 4; } } -#endif /* CONFIG_HEARTBEAT */ -return IRQ_HANDLED ; } +#endif /* CONFIG_HEARTBEAT */ #ifdef CONFIG_M68KCLASSIC #if !IS_BUILTIN (CONFIG_RTC_DRV_GENERIC )
arch/m68k/q40/q40ints.c +2 -0
@@ -17,6 +17,7 @@#include <linux/interrupt.h> #include <linux/irq.h> +#include <asm/machdep.h> #include <asm/ptrace.h> #include <asm/traps.h>
@@ -144,6 +145,7 @@ static irqreturn_t q40_timer_int(int irq, void *dev_id)local_irq_save (flags);timer_routine (0 ,NULL ); +timer_heartbeat ();local_irq_restore (flags); }return IRQ_HANDLED ;
Signature · 1 line
2.27.0
Re: [PATCH 06/13] m68k: split heartbeat out of timer function
Quoted 8 lines · Arnd Bergmann
The heartbeat functionality is mostly separate from the actual timer interrupt handling, and it is only used on five platforms.
Split it out into a separate function and call that directly from the timer irq on those platforms.
Signed-off-by Arnd Bergmann <[email protected]>
And finally (for Amiga):
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
PATCH 7/13 m68k: sun3/sun3c: use legacy_timer_tick
In reply to Arnd BergmannThese two are different from all other machines:
* sun3 does not call timer_routine() but open-codes it except for the profile_tick() call that appears to be unintentionally missing.
* sun3x has a commented-out timer irq handler but no functional timer tick I could find.
Change both to calling the new legacy_timer_tick here, which includes the call to profile_tick() but does not fix sun3x as that is still commented out.
--- arch/m68k/Kconfig.machine | 2 ++ arch/m68k/sun3/sun3ints.c | 3 +-- arch/m68k/sun3x/time.c | 3 +-- 3 files changed, 4 insertions(+), 4 deletions(-)
arch/m68k/Kconfig.machine +2 -0
@@ -99,6 +99,7 @@ config HP300 config SUN3X bool "Sun3x support" depends on MMU + select LEGACY_TIMER_TICK select MMU_MOTOROLA if MMU select M68030 help
@@ -126,6 +127,7 @@ config SUN3 depends on MMU depends on !MMU_MOTOROLA select MMU_SUN3 if MMU + select LEGACY_TIMER_TICK select NO_DMA select M68020 help
arch/m68k/sun3/sun3ints.c +1 -2
@@ -73,8 +73,7 @@ static irqreturn_t sun3_int5(int irq, void *dev_id)#ifdef CONFIG_SUN3 intersil_clear ();#endif -xtime_update (1 ); -update_process_times (user_mode (get_irq_regs ())); +legacy_timer_tick (1 ); cnt =kstat_irqs_cpu (irq,0 );if (!(cnt %20 ))sun3_leds (led_pattern[cnt %160 /20 ]);
arch/m68k/sun3x/time.c +1 -2
@@ -77,14 +77,13 @@ int sun3x_hwclk(int set, struct rtc_time *t)#if 0 static irqreturn_t sun3x_timer_tick (int irq,void *dev_id) { -irq_handler_t timer_routine = dev_id;unsigned long flags;local_irq_save (flags);/* Clear the pending interrupt - pulse the enable line low */ disable_irq (5 );enable_irq (5 ); -timer_routine (0 ,NULL ); +legacy_timer_tick (1 );local_irq_restore (flags);return IRQ_HANDLED ;
Signature · 1 line
2.27.0
Re: [PATCH 07/13] m68k: sun3/sun3c: use legacy_timer_tick
Quoted 6 lines · Arnd Bergmann
These two are different from all other machines:
* sun3 does not call
timer_routine()but open-codes it except for theprofile_tick()call that appears to be unintentionally missing.
Quoted 8 lines
* sun3x has a commented-out timer irq handler but no functional timer tick I could find.
Change both to calling the new
legacy_timer_tickhere, which includes the call toprofile_tick()but does not fix sun3x as that is still commented out.Signed-off-by Arnd Bergmann <[email protected]>
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
PATCH 8/13 m68k: m68328: use legacy_timer_tick()
In reply to Arnd BergmannA couple of machines share the m68328 timer code that is based on calling timer_interrupt(). Change these to the new and slightly more generic legacy_timer_tick() helper.
--- arch/m68k/68000/timers.c | 5 ++--- arch/m68k/Kconfig.machine | 4 ++++ 2 files changed, 6 insertions(+), 3 deletions(-)
arch/m68k/68000/timers.c +2 -3
@@ -52,7 +52,6 @@#endif static u32 m68328_tick_cnt; -static irq_handler_t timer_interrupt;/***************************************************************************/
@@ -62,7 +61,8 @@ static irqreturn_t hw_tick(int irq, void *dummy)TSTAT &=0 ; m68328_tick_cnt +=TICKS_PER_JIFFY ; -return timer_interrupt (irq, dummy); +legacy_timer_tick (1 ); +return IRQ_HANDLED ; }/***************************************************************************/
@@ -113,7 +113,6 @@ void hw_timer_init(irq_handler_t handler)/* Enable timer 1 */ TCTL |=TCTL_TEN ;clocksource_register_hz (&m68328_clk,TICKS_PER_JIFFY *HZ ); - timer_interrupt = handler; }/***************************************************************************/
arch/m68k/Kconfig.machine +4 -0
@@ -146,6 +146,7 @@ config PILOT config PILOT3 bool "Pilot 1000/5000, PalmPilot Personal/Pro, or PalmIII support" depends on M68328 + select LEGACY_TIMER_TICK select PILOT help Support for the Palm Pilot 1000/5000, Personal/Pro and PalmIII.
@@ -159,18 +160,21 @@ config XCOPILOT_BUGS config UCSIMM bool "uCsimm module support" depends on M68EZ328 + select LEGACY_TIMER_TICK help Support for the Arcturus Networks uCsimm module. config UCDIMM bool "uDsimm module support" depends on M68VZ328 + select LEGACY_TIMER_TICK help Support for the Arcturus Networks uDsimm module. config DRAGEN2 bool "DragenEngine II board support" depends on M68VZ328 + select LEGACY_TIMER_TICK help Support for the DragenEngine II board.
Signature · 1 line
2.27.0
Re: [PATCH 08/13] m68k: m68328: use legacy_timer_tick()
Hi Arnd,
Quoted 6 lines · Arnd Bergmann
A couple of machines share the m68328 timer code that is based on calling
timer_interrupt(). Change these to the new and slightly more genericlegacy_timer_tick()helper.Signed-off-by Arnd Bergmann <[email protected]>
Thanks for your patch!
Quoted 21 lines
arch/m68k/Kconfig.machine +2
-0@@ -146,6 +146,7 @@ config PILOT config PILOT3 bool "Pilot 1000/5000, PalmPilot Personal/Pro, or PalmIII support" depends on M68328 + select LEGACY_TIMER_TICK select PILOT help Support for the Palm Pilot 1000/5000, Personal/Pro and PalmIII.@@ -159,18 +160,21 @@ config XCOPILOT_BUGS config UCSIMM bool "uCsimm module support" depends on M68EZ328 + select LEGACY_TIMER_TICK help Support for the Arcturus Networks uCsimm module. config UCDIMM bool "uDsimm module support" depends on M68VZ328
arch/m68k/Kconfig.machine config XCOPILOT_BUGS · line 179
Quoted 10 lines
arch/m68k/Kconfig.machine +2
-0+ select LEGACY_TIMER_TICK help Support for the Arcturus Networks uDsimm module. config DRAGEN2 bool "DragenEngine II board support" depends on M68VZ328 + select LEGACY_TIMER_TICK help Support for the DragenEngine II board.
Given this feature is SoC-specific, not platform-specific, perhaps it makes sense to move the selects to the M68{,EZ,VZ}328 symbols?
Regardless:
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
Re: [PATCH 08/13] m68k: m68328: use legacy_timer_tick()
Quoted 6 lines · Geert Uytterhoeven
Given this feature is SoC-specific, not platform-specific, perhaps it makes sense to move the selects to the M68{,EZ,VZ}328 symbols?
Regardless:
Reviewed-by Geert Uytterhoeven <[email protected]>
Ok, folded in the change blow, using one less line. I couldn't figure out whether
it should just be part of the CONFIG_M68000 instead, which doesn't appear to have any machine support by itself. The dragonball CPU configuration looks really odd, because you have to build exactly one of M68{,EZ,VZ}328 into the kernel to get a successful compilation, while Kconfig allows many other combinations.
Arnd
arch/m68k/Kconfig.cpu +3 -0
@@ -104,6 +104,7 @@ config M68060 config M68328 bool "MC68328" depends on !MMU + select LEGACY_TIMER_TICK select M68000 help Motorola 68328 processor support.
@@ -111,6 +112,7 @@ config M68328 config M68EZ328 bool "MC68EZ328" depends on !MMU + select LEGACY_TIMER_TICK select M68000 help Motorola 68EX328 processor support.
@@ -118,6 +120,7 @@ config M68EZ328 config M68VZ328 bool "MC68VZ328" depends on !MMU + select LEGACY_TIMER_TICK select M68000 help Motorola 68VZ328 processor support.
arch/m68k/Kconfig.machine +0 -4
@@ -146,7 +146,6 @@ config PILOT config PILOT3 bool "Pilot 1000/5000, PalmPilot Personal/Pro, or PalmIII support" depends on M68328 - select LEGACY_TIMER_TICK select PILOT help Support for the Palm Pilot 1000/5000, Personal/Pro and PalmIII.
@@ -160,21 +159,18 @@ config XCOPILOT_BUGS config UCSIMM bool "uCsimm module support" depends on M68EZ328 - select LEGACY_TIMER_TICK help Support for the Arcturus Networks uCsimm module. config UCDIMM bool "uDsimm module support" depends on M68VZ328 - select LEGACY_TIMER_TICK help Support for the Arcturus Networks uDsimm module. config DRAGEN2 bool "DragenEngine II board support" depends on M68VZ328 - select LEGACY_TIMER_TICK help Support for the DragenEngine II board.
Geert Uytterhoeven
maintainerRe: [PATCH 08/13] m68k: m68328: use legacy_timer_tick()
Hi Arnd,
Quoted 8 lines · Arnd Bergmann
Quoted 5 lines · Geert Uytterhoeven
Given this feature is SoC-specific, not platform-specific, perhaps it makes sense to move the selects to the M68{,EZ,VZ}328 symbols?
Regardless:
Reviewed-by Geert Uytterhoeven <[email protected]>Ok, folded in the change blow, using one less line. I couldn't figure
Thanks, looks good.
Quoted 2 lines
out whether
it should just be part of theCONFIG_M68000instead, which doesn't appear
It must definitely not be selected by CONFIG_M68000, as the plain MC68000 is a CPU, not an SoC, and does not have the timer.
Quoted 4 lines
to have any machine support by itself. The dragonball CPU configuration looks really odd, because you have to build exactly one of M68{,EZ,VZ}328 into the kernel to get a successful compilation, while
Kconfigallows many other combinations.
While CONFIG_M68000 could be used for a "pure" MC68000-based machine, I believe we don't support any yet.
M68{,EZ,VZ}328 select M68000 as they are SoCs containing a 68000 CPU core.
Other m68k SoCs have a CPU32 core, which is a simplified 68020 CPU core (hmm, what happened to 68360 support? Oh, removed in 2016, so nothing selects CPU32 anymore).
Gr{oetje,eeting}s,
Geert
Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected] In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds
PATCH 9/13 m68k: change remaining timers to legacy_timer_tick
In reply to Arnd BergmannThere are nine more machines that each have their own timer interrupt calling the m68k timer_interrupt() function through an indirect pointer.
This function is now the same as legacy_timer_tick, so just call that directly and select the corresponding Kconfig symbol.
--- arch/m68k/Kconfig.machine | 9 +++++++++ arch/m68k/amiga/config.c | 6 ++---- arch/m68k/apollo/config.c | 6 ++---- arch/m68k/atari/time.c | 5 ++--- arch/m68k/bvme6000/config.c | 5 ++--- arch/m68k/hp300/time.c | 5 ++--- arch/m68k/mac/via.c | 6 ++---- arch/m68k/mvme147/config.c | 5 ++--- arch/m68k/mvme16x/config.c | 5 ++--- arch/m68k/q40/q40ints.c | 6 ++---- 10 files changed, 27 insertions(+), 31 deletions(-)
arch/m68k/Kconfig.machine +9 -0
@@ -7,6 +7,7 @@ config AMIGA bool "Amiga support" depends on MMU select MMU_MOTOROLA if MMU + select LEGACY_TIMER_TICK help This option enables support for the Amiga series of computers. If you plan to use this kernel on an Amiga, say Y here and browse the
@@ -17,6 +18,7 @@ config ATARI depends on MMU select MMU_MOTOROLA if MMU select HAVE_ARCH_NVRAM_OPS + select LEGACY_TIMER_TICK help This option enables support for the 68000-based Atari series of computers (including the TT, Falcon and Medusa). If you plan to use
@@ -28,6 +30,7 @@ config MAC depends on MMU select MMU_MOTOROLA if MMU select HAVE_ARCH_NVRAM_OPS + select LEGACY_TIMER_TICK help This option enables support for the Apple Macintosh series of computers (yes, there is experimental support now, at least for part
@@ -40,6 +43,7 @@ config APOLLO bool "Apollo support" depends on MMU select MMU_MOTOROLA if MMU + select LEGACY_TIMER_TICK help Say Y here if you want to run Linux on an MC680x0-based Apollo Domain workstation such as the DN3500.
@@ -58,6 +62,7 @@ config MVME147 bool "MVME147 support" depends on MMU depends on VME + select LEGACY_TIMER_TICK help Say Y to include support for early Motorola VME boards. This will build a kernel which can run on MVME147 single-board computers. If
@@ -68,6 +73,7 @@ config MVME16x bool "MVME162, 166 and 167 support" depends on MMU depends on VME + select LEGACY_TIMER_TICK help Say Y to include support for Motorola VME boards. This will build a kernel which can run on MVME162, MVME166, MVME167, MVME172, and
@@ -79,6 +85,7 @@ config BVME6000 bool "BVME4000 and BVME6000 support" depends on MMU depends on VME + select LEGACY_TIMER_TICK help Say Y to include support for VME boards from BVM Ltd. This will build a kernel which can run on BVME4000 and BVME6000 boards. If
@@ -89,6 +96,7 @@ config HP300 bool "HP9000/300 and HP9000/400 support" depends on MMU select MMU_MOTOROLA if MMU + select LEGACY_TIMER_TICK help This option enables support for the HP9000/300 and HP9000/400 series of workstations. Support for these machines is still somewhat
@@ -115,6 +123,7 @@ config Q40 bool "Q40/Q60 support" depends on MMU select MMU_MOTOROLA if MMU + select LEGACY_TIMER_TICK help The Q40 is a Motorola 68040-based successor to the Sinclair QL manufactured in Germany. There is an official Q40 home page at
arch/m68k/amiga/config.c +2 -4
@@ -475,11 +475,9 @@ static u32 clk_total, clk_offset;static irqreturn_t ciab_timer_handler (int irq,void *dev_id) { -irq_handler_t timer_routine = dev_id; - clk_total += jiffy_ticks; clk_offset =0 ; -timer_routine (0 ,NULL ); +legacy_timer_tick (1 );timer_heartbeat ();return IRQ_HANDLED ;
@@ -504,7 +502,7 @@ static void __init amiga_sched_init(irq_handler_t timer_routine)* SCSI code. We'll have to take a look at this later */ if (request_irq (IRQ_AMIGA_CIAB_TA , ciab_timer_handler,IRQF_TIMER , -"timer" , timer_routine)) +"timer" ,NULL ))pr_err ("Couldn't register timer interrupt\n" );/* start timer */ ciab.cra |=0x11 ;
arch/m68k/apollo/config.c +2 -4
@@ -168,11 +168,9 @@ void __init config_apollo(void)irqreturn_t dn_timer_int (int irq,void *dev_id) { -irq_handler_t timer_handler = dev_id; -volatile unsigned char x; -timer_handler (irq, dev_id); +legacy_timer_tick (1 );timer_heartbeat (); x = *(volatile unsigned char *)(apollo_timer +3 );
@@ -199,7 +197,7 @@ void dn_sched_init(irq_handler_t timer_routine) *(volatile unsigned char *)(apollo_timer + 0x3));#endif -if (request_irq (IRQ_APOLLO , dn_timer_int,0 ,"time" , timer_routine)) +if (request_irq (IRQ_APOLLO , dn_timer_int,0 ,"time" ,NULL ))pr_err ("Couldn't register timer interrupt\n" ); }
arch/m68k/atari/time.c +2 -3
@@ -41,7 +41,6 @@ static u8 last_timer_count;static irqreturn_t mfp_timer_c_handler (int irq,void *dev_id) { -irq_handler_t timer_routine = dev_id;unsigned long flags;local_irq_save (flags);
@@ -49,7 +48,7 @@ static irqreturn_t mfp_timer_c_handler(int irq, void *dev_id) last_timer_count = st_mfp.tim_dt_c; }while (last_timer_count =1 ); clk_total +=INT_TICKS ; -timer_routine (0 ,NULL ); +legacy_timer_tick (1 );timer_heartbeat ();local_irq_restore (flags);
@@ -65,7 +64,7 @@ atari_sched_init(irq_handler_t timer_routine) st_mfp.tim_ct_cd = (st_mfp.tim_ct_cd &15 ) |0x60 ;/* install interrupt service routine for MFP Timer C */ if (request_irq (IRQ_MFP_TIMC , mfp_timer_c_handler,IRQF_TIMER ,"timer" , - timer_routine)) +NULL ))pr_err ("Couldn't register timer interrupt\n" );clocksource_register_hz (&atari_clk,INT_CLK );
arch/m68k/bvme6000/config.c +2 -3
@@ -165,7 +165,6 @@ static u32 clk_total, clk_offset;static irqreturn_t bvme6000_timer_int (int irq,void *dev_id) { -irq_handler_t timer_routine = dev_id;unsigned long flags;volatile RtcPtr_t rtc = (RtcPtr_t )BVME_RTC_BASE ;unsigned char msr;
@@ -175,7 +174,7 @@ static irqreturn_t bvme6000_timer_int (int irq, void *dev_id) rtc->msr = msr |0x20 ;/* Ack the interrupt */ clk_total +=RTC_TIMER_CYCLES ; clk_offset =0 ; -timer_routine (0 ,NULL ); +legacy_timer_tick (1 );local_irq_restore (flags);return IRQ_HANDLED ;
@@ -198,7 +197,7 @@ void bvme6000_sched_init (irq_handler_t timer_routine) rtc->msr =0 ;/* Ensure timer registers accessible */ if (request_irq (BVME_IRQ_RTC , bvme6000_timer_int,IRQF_TIMER ,"timer" , - timer_routine)) +NULL ))panic ("Couldn't register timer int" ); rtc->t1cr_omr =0x04 ;/* Mode 2, ext clk */
arch/m68k/hp300/time.c +2 -3
@@ -55,7 +55,6 @@ static u32 clk_total, clk_offset;static irqreturn_t hp300_tick (int irq,void *dev_id) { -irq_handler_t timer_routine = dev_id;unsigned long flags;unsigned long tmp;
@@ -64,7 +63,7 @@ static irqreturn_t hp300_tick(int irq, void *dev_id) asmvolatile ("movpw %1@(5),%0" :"=d" (tmp) :"a" (CLOCKBASE )); clk_total +=INTVAL ; clk_offset =0 ; -timer_routine (0 ,NULL ); +legacy_timer_tick (1 );timer_heartbeat ();local_irq_restore (flags);
@@ -106,7 +105,7 @@ void __init hp300_sched_init(irq_handler_t vector) asmvolatile (" movpw %0,%1@(5)" : :"d" (INTVAL ),"a" (CLOCKBASE )); -if (request_irq (IRQ_AUTO_6 , hp300_tick,IRQF_TIMER ,"timer tick" , vector)) +if (request_irq (IRQ_AUTO_6 , hp300_tick,IRQF_TIMER ,"timer tick" ,NULL ))pr_err ("Couldn't register timer interrupt\n" );out_8 (CLOCKBASE +CLKCR2 ,0x1 );/* select CR1 */
arch/m68k/mac/via.c +2 -4
@@ -602,11 +602,9 @@ static u32 clk_total, clk_offset;static irqreturn_t via_timer_handler (int irq,void *dev_id) { -irq_handler_t timer_routine = dev_id; - clk_total +=VIA_TIMER_CYCLES ; clk_offset =0 ; -timer_routine (0 ,NULL ); +legacy_timer_tick (1 );return IRQ_HANDLED ; }
@@ -614,7 +612,7 @@ static irqreturn_t via_timer_handler(int irq, void *dev_id)void __init via_init_clock (irq_handler_t timer_routine) {if (request_irq (IRQ_MAC_TIMER_1 , via_timer_handler,IRQF_TIMER ,"timer" , - timer_routine)) { +NULL )) {pr_err ("Couldn't register %s interrupt\n" ,"timer" );return ; }
arch/m68k/mvme147/config.c +2 -3
@@ -112,14 +112,13 @@ static u32 clk_total;static irqreturn_t mvme147_timer_int (int irq,void *dev_id) { -irq_handler_t timer_routine = dev_id;unsigned long flags;local_irq_save (flags); m147_pcc->t1_int_cntrl =PCC_TIMER_INT_CLR ; m147_pcc->t1_cntrl =PCC_TIMER_CLR_OVF ; clk_total +=PCC_TIMER_CYCLES ; -timer_routine (0 ,NULL ); +legacy_timer_tick (1 );local_irq_restore (flags);return IRQ_HANDLED ;
@@ -129,7 +128,7 @@ static irqreturn_t mvme147_timer_int (int irq, void *dev_id)void mvme147_sched_init (irq_handler_t timer_routine) {if (request_irq (PCC_IRQ_TIMER1 , mvme147_timer_int,IRQF_TIMER , -"timer 1" , timer_routine)) +"timer 1" ,NULL ))pr_err ("Couldn't register timer interrupt\n" );/* Init the clock with a value */
arch/m68k/mvme16x/config.c +2 -3
@@ -372,14 +372,13 @@ static u32 clk_total;static irqreturn_t mvme16x_timer_int (int irq,void *dev_id) { -irq_handler_t timer_routine = dev_id;unsigned long flags;local_irq_save (flags);out_8 (PCCTIC1 ,in_8 (PCCTIC1 ) |PCCTIC1_INT_CLR );out_8 (PCCTOVR1 ,PCCTOVR1_OVR_CLR ); clk_total +=PCC_TIMER_CYCLES ; -timer_routine (0 ,NULL ); +legacy_timer_tick (1 );local_irq_restore (flags);return IRQ_HANDLED ;
@@ -396,7 +395,7 @@ void mvme16x_sched_init (irq_handler_t timer_routine)out_8 (PCCTOVR1 ,in_8 (PCCTOVR1 ) |PCCTOVR1_TIC_EN |PCCTOVR1_COC_EN );out_8 (PCCTIC1 ,PCCTIC1_INT_EN |6 );if (request_irq (MVME16x_IRQ_TIMER, mvme16x_timer_int,IRQF_TIMER ,"timer" , - timer_routine)) +NULL ))panic ("Couldn't register timer int" );clocksource_register_hz (&mvme16x_clk,PCC_TIMER_CLOCK_FREQ );
arch/m68k/q40/q40ints.c +2 -4
@@ -130,8 +130,6 @@ void q40_mksound(unsigned int hz, unsigned int ticks)static irqreturn_t q40_timer_int (int irq,void *dev_id) { -irq_handler_t timer_routine = dev_id; - ql_ticks = ql_ticks ?0 :1 ;if (sound_ticks) {unsigned char sval=(sound_ticks &1 ) ?128 -SVOL :128 +SVOL ;
@@ -144,7 +142,7 @@ static irqreturn_t q40_timer_int(int irq, void *dev_id)unsigned long flags;local_irq_save (flags); -timer_routine (0 ,NULL ); +legacy_timer_tick (1 );timer_heartbeat ();local_irq_restore (flags); }
@@ -157,7 +155,7 @@ void q40_sched_init (irq_handler_t timer_routine) timer_irq =Q40_IRQ_FRAME ; -if (request_irq (timer_irq, q40_timer_int,0 ,"timer" , timer_routine)) +if (request_irq (timer_irq, q40_timer_int,0 ,"timer" ,NULL ))panic ("Couldn't register timer int" );master_outb (-1 ,FRAME_CLEAR_REG );
Signature · 1 line
2.27.0
Re: [PATCH 09/13] m68k: change remaining timers to legacy_timer_tick
Quoted 7 lines · Arnd Bergmann
There are nine more machines that each have their own timer interrupt calling the m68k
timer_interrupt()function through an indirect pointer.This function is now the same as
legacy_timer_tick, so just call that directly and select the correspondingKconfigsymbol.Signed-off-by Arnd Bergmann <[email protected]>
And finally (for Amiga and Atari/ARAnyM):
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
PATCH 10/13 m68k: remove timer_interrupt() function
In reply to Arnd BergmannThis gets passed to a number of init functions, but is ignored everywhere, so remove the function and change the mach_sched_init callback to take no arguments.
--- arch/m68k/68000/timers.c | 2 +- arch/m68k/amiga/config.c | 4 ++-- arch/m68k/apollo/config.c | 4 ++-- arch/m68k/atari/config.c | 2 +- arch/m68k/atari/time.c | 2 +- arch/m68k/bvme6000/config.c | 4 ++-- arch/m68k/coldfire/pit.c | 2 +- arch/m68k/coldfire/sltimers.c | 2 +- arch/m68k/coldfire/timers.c | 2 +- arch/m68k/hp300/time.c | 2 +- arch/m68k/hp300/time.h | 2 +- arch/m68k/include/asm/machdep.h | 5 ++--- arch/m68k/kernel/setup_mm.c | 2 +- arch/m68k/kernel/setup_no.c | 2 +- arch/m68k/kernel/time.c | 15 +-------------- arch/m68k/mac/config.c | 6 +++--- arch/m68k/mac/via.c | 2 +- arch/m68k/mvme147/config.c | 4 ++-- arch/m68k/mvme16x/config.c | 4 ++-- arch/m68k/q40/config.c | 2 +- arch/m68k/q40/q40ints.c | 2 +- arch/m68k/sun3/config.c | 4 ++-- arch/m68k/sun3x/time.c | 2 +- arch/m68k/sun3x/time.h | 2 +- 24 files changed, 33 insertions(+), 47 deletions(-)
arch/m68k/68000/timers.c +1 -1
@@ -91,7 +91,7 @@ static struct clocksource m68328_clk = {/***************************************************************************/ -void hw_timer_init (irq_handler_t handler) +void hw_timer_init (void ) {int ret;
arch/m68k/amiga/config.c +2 -2
@@ -92,7 +92,7 @@ static char *amiga_models[] __initdata = {static char amiga_model_name[13 ] ="Amiga " ; -static void amiga_sched_init (irq_handler_t handler); +static void amiga_sched_init (void );static void amiga_get_model (char *model);static void amiga_get_hardware_list (struct seq_file *m);extern void amiga_mksound (unsigned int count,unsigned int ticks);
@@ -483,7 +483,7 @@ static irqreturn_t ciab_timer_handler(int irq, void *dev_id)return IRQ_HANDLED ; } -static void __init amiga_sched_init (irq_handler_t timer_routine) +static void __init amiga_sched_init (void ) {static struct resource sched_res = { .name ="timer" , .start =0x00bfd400 , .end =0x00bfd5ff ,
arch/m68k/apollo/config.c +2 -2
@@ -26,7 +26,7 @@ u_long cpuctrl_physaddr;u_long timer_physaddr;u_long apollo_model; -extern void dn_sched_init (irq_handler_t handler); +extern void dn_sched_init (void );extern void dn_init_IRQ (void );extern int dn_dummy_hwclk (int ,struct rtc_time *);extern void dn_dummy_reset (void );
@@ -179,7 +179,7 @@ irqreturn_t dn_timer_int(int irq, void *dev_id)return IRQ_HANDLED ; } -void dn_sched_init (irq_handler_t timer_routine) +void dn_sched_init (void ) {/* program timer 1 */ *(volatile unsigned char *)(apollo_timer +3 ) =0x01 ;
arch/m68k/atari/config.c +1 -1
@@ -77,7 +77,7 @@ static void atari_heartbeat(int on);#endif /* atari specific timer functions (in time.c) */ -extern void atari_sched_init (irq_handler_t ); +extern void atari_sched_init (void );extern int atari_mste_hwclk (int ,struct rtc_time *);extern int atari_tt_hwclk (int ,struct rtc_time *);
arch/m68k/atari/time.c +1 -1
@@ -56,7 +56,7 @@ static irqreturn_t mfp_timer_c_handler(int irq, void *dev_id) }void __init -atari_sched_init (irq_handler_t timer_routine) +atari_sched_init (void ) {/* set Timer C data Register */ st_mfp.tim_dt_c =INT_TICKS ;
arch/m68k/bvme6000/config.c +2 -2
@@ -38,7 +38,7 @@#include <asm/bvme6000hw.h> static void bvme6000_get_model (char *model); -extern void bvme6000_sched_init (irq_handler_t handler); +extern void bvme6000_sched_init (void );extern int bvme6000_hwclk (int ,struct rtc_time *);extern void bvme6000_reset (void );void bvme6000_set_vectors (void );
@@ -189,7 +189,7 @@ static irqreturn_t bvme6000_timer_int (int irq, void *dev_id)* so divide by 8 to get the microsecond result. */ -void bvme6000_sched_init (irq_handler_t timer_routine) +void bvme6000_sched_init (void ) {volatile RtcPtr_t rtc = (RtcPtr_t )BVME_RTC_BASE ;unsigned char msr = rtc->msr &0xc0 ;
arch/m68k/coldfire/pit.c +1 -1
@@ -136,7 +136,7 @@ static struct clocksource pit_clk = {/***************************************************************************/ -void hw_timer_init (irq_handler_t handler) +void hw_timer_init (void ) {int ret;
arch/m68k/coldfire/sltimers.c +1 -1
@@ -118,7 +118,7 @@ static struct clocksource mcfslt_clk = { .flags = CLOCK_SOURCE_IS_CONTINUOUS, }; -void hw_timer_init (irq_handler_t handler) +void hw_timer_init (void ) {int r;
arch/m68k/coldfire/timers.c +1 -1
@@ -107,7 +107,7 @@ static struct clocksource mcftmr_clk = {/***************************************************************************/ -void hw_timer_init (irq_handler_t handler) +void hw_timer_init (void ) {int r;
arch/m68k/hp300/time.c +1 -1
@@ -98,7 +98,7 @@ static u64 hp300_read_clk(struct clocksource *cs)return ticks; } -void __init hp300_sched_init (irq_handler_t vector) +void __init hp300_sched_init (void ) {out_8 (CLOCKBASE +CLKCR2 ,0x1 );/* select CR1 */ out_8 (CLOCKBASE +CLKCR1 ,0x1 );/* reset */
arch/m68k/hp300/time.h +1 -1
@@ -1 +1 @@ -extern void hp300_sched_init (irq_handler_t vector); +extern void hp300_sched_init (void );
arch/m68k/include/asm/machdep.h +2 -3
@@ -12,7 +12,7 @@ struct rtc_time;struct rtc_pll_info ;struct buffer_head ; -extern void (*mach_sched_init) (irq_handler_t handler); +extern void (*mach_sched_init) (void );/* machine dependent irq functions */ extern void (*mach_init_IRQ) (void );extern void (*mach_get_model) (char *model);
@@ -33,8 +33,7 @@ extern void (*mach_l2_flush) (int);extern void (*mach_beep) (unsigned int ,unsigned int );/* Hardware clock functions */ -extern void hw_timer_init (irq_handler_t handler); -extern unsigned long hw_timer_offset (void ); +extern void hw_timer_init (void );#ifdef CONFIG_HEARTBEAT extern void timer_heartbeat (void );#else
arch/m68k/kernel/setup_mm.c +1 -1
@@ -82,7 +82,7 @@ static struct m68k_mem_info m68k_ramdisk __initdata;static char m68k_command_line[CL_SIZE ]__initdata ; -void (*mach_sched_init) (irq_handler_t handler)__initdata =NULL ; +void (*mach_sched_init) (void )__initdata =NULL ;/* machine dependent irq functions */ void (*mach_init_IRQ) (void )__initdata =NULL ;void (*mach_get_model) (char *model);
arch/m68k/kernel/setup_no.c +1 -1
@@ -49,7 +49,7 @@ EXPORT_SYMBOL(memory_end);char __initdata command_line[COMMAND_LINE_SIZE ];/* machine dependent timer functions */ -void (*mach_sched_init)(irq_handler_t handler)__initdata =NULL ; +void (*mach_sched_init)(void )__initdata =NULL ;int (*mach_hwclk) (int ,struct rtc_time *);/* machine dependent reboot functions */
arch/m68k/kernel/time.c +1 -14
@@ -35,19 +35,6 @@unsigned long (*mach_random_get_entropy)(void );EXPORT_SYMBOL_GPL (mach_random_get_entropy); - -/* -* timer_interrupt() needs to keep up the real-time clock, -* as well as call the "xtime_update()" routine every clocktick -*/ -static irqreturn_t timer_interrupt (int irq,void *dummy) -{ -xtime_update (1 ); -update_process_times (user_mode (get_irq_regs ())); -profile_tick (CPU_PROFILING ); -return IRQ_HANDLED ; -} -#ifdef CONFIG_HEARTBEAT void timer_heartbeat (void ) {
@@ -157,5 +144,5 @@ module_init(rtc_init);void __init time_init (void ) { -mach_sched_init (timer_interrupt); +mach_sched_init (); }
arch/m68k/mac/config.c +3 -3
@@ -58,7 +58,7 @@ extern int mac_hwclk(int, struct rtc_time *);extern void iop_preinit (void );extern void iop_init (void );extern void via_init (void ); -extern void via_init_clock (irq_handler_t func); +extern void via_init_clock (void );extern void oss_init (void );extern void psc_init (void );extern void baboon_init (void );
@@ -69,9 +69,9 @@ static void mac_get_model(char *str);static void mac_identify (void );static void mac_report_hardware (void ); -static void __init mac_sched_init (irq_handler_t vector) +static void __init mac_sched_init (void ) { -via_init_clock (vector); +via_init_clock (); }/*
arch/m68k/mac/via.c +1 -1
@@ -609,7 +609,7 @@ static irqreturn_t via_timer_handler(int irq, void *dev_id)return IRQ_HANDLED ; } -void __init via_init_clock (irq_handler_t timer_routine) +void __init via_init_clock (void ) {if (request_irq (IRQ_MAC_TIMER_1 , via_timer_handler,IRQF_TIMER ,"timer" ,NULL )) {
arch/m68k/mvme147/config.c +2 -2
@@ -37,7 +37,7 @@static void mvme147_get_model (char *model); -extern void mvme147_sched_init (irq_handler_t handler); +extern void mvme147_sched_init (void );extern int mvme147_hwclk (int ,struct rtc_time *);extern void mvme147_reset (void );
@@ -125,7 +125,7 @@ static irqreturn_t mvme147_timer_int (int irq, void *dev_id) } -void mvme147_sched_init (irq_handler_t timer_routine) +void mvme147_sched_init (void ) {if (request_irq (PCC_IRQ_TIMER1 , mvme147_timer_int,IRQF_TIMER ,"timer 1" ,NULL ))
arch/m68k/mvme16x/config.c +2 -2
@@ -43,7 +43,7 @@ extern t_bdid mvme_bdid;static MK48T08ptr_t volatile rtc = (MK48T08ptr_t )MVME_RTC_BASE ;static void mvme16x_get_model (char *model); -extern void mvme16x_sched_init (irq_handler_t handler); +extern void mvme16x_sched_init (void );extern int mvme16x_hwclk (int ,struct rtc_time *);extern void mvme16x_reset (void );
@@ -384,7 +384,7 @@ static irqreturn_t mvme16x_timer_int (int irq, void *dev_id)return IRQ_HANDLED ; } -void mvme16x_sched_init (irq_handler_t timer_routine) +void mvme16x_sched_init (void ) {uint16_t brdno =be16_to_cpu (mvme_bdid.brdno);int irq;
arch/m68k/q40/config.c +1 -1
@@ -37,7 +37,7 @@extern void q40_init_IRQ (void );static void q40_get_model (char *model); -extern void q40_sched_init (irq_handler_t handler); +extern void q40_sched_init (void );static int q40_hwclk (int ,struct rtc_time *);static unsigned int q40_get_ss (void );
arch/m68k/q40/q40ints.c +1 -1
@@ -149,7 +149,7 @@ static irqreturn_t q40_timer_int(int irq, void *dev_id)return IRQ_HANDLED ; } -void q40_sched_init (irq_handler_t timer_routine) +void q40_sched_init (void ) {int timer_irq;
arch/m68k/sun3/config.c +2 -2
@@ -36,7 +36,7 @@char sun3_reserved_pmeg[SUN3_PMEGS_NUM ]; -static void sun3_sched_init (irq_handler_t handler); +static void sun3_sched_init (void );extern void sun3_get_model (char * model);extern int sun3_hwclk (int set,struct rtc_time *t);
@@ -151,7 +151,7 @@ void __init config_sun3(void)sun3_bootmem_alloc (memory_start, memory_end); } -static void __init sun3_sched_init (irq_handler_t timer_routine) +static void __init sun3_sched_init (void ) {sun3_disable_interrupts (); intersil_clock->cmd_reg=(INTERSIL_RUN |INTERSIL_INT_DISABLE |INTERSIL_24H_MODE );
arch/m68k/sun3x/time.c +1 -1
@@ -90,7 +90,7 @@ static irqreturn_t sun3x_timer_tick(int irq, void *dev_id) }#endif -void __init sun3x_sched_init (irq_handler_t vector) +void __init sun3x_sched_init (void ) {sun3_disable_interrupts ();
arch/m68k/sun3x/time.h +1 -1
@@ -3,7 +3,7 @@#define SUN3X_TIME_H extern int sun3x_hwclk (int set,struct rtc_time *t); -void sun3x_sched_init (irq_handler_t vector); +void sun3x_sched_init (void );struct mostek_dt {volatile unsigned char csr;
Signature · 1 line
2.27.0
Re: [PATCH 10/13] m68k: remove timer_interrupt() function
Hi Arnd,
Quoted 5 lines · Arnd Bergmann
This gets passed to a number of init functions, but is ignored everywhere, so remove the function and change the
mach_sched_initcallback to take no arguments.Signed-off-by Arnd Bergmann <[email protected]>
And finally (for Amiga and Atari/ARAnyM):
arch/m68k/include/asm/machdep.h extern void (*mach_l2_flush) (int); · line 37
Quoted 8 lines
arch/m68k/include/asm/machdep.h +0
-2@@ -33,8 +33,7 @@ extern void (*mach_l2_flush) (int);extern void (*mach_beep) (unsigned int ,unsigned int );/* Hardware clock functions */ -extern void hw_timer_init (irq_handler_t handler); -extern unsigned long hw_timer_offset (void );
Yeah, this never-used prototype, which entered through the nommu merge, can go. But that's not really related to this patch.
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
PATCH 11/13 timekeeping: remove xtime_update
In reply to Arnd BergmannThere are no more users of xtime_update aside from legacy_timer_tick(), so fold it into that function and remove the declaration.
update_process_times() is now only called inside of the kernel/time/ code, so the declaration can be moved there.
--- include/linux/timekeeping.h | 4 +--- kernel/time/tick-legacy.c | 22 ++++++++++++++++++++-- kernel/time/timekeeping.c | 16 ---------------- kernel/time/timekeeping.h | 1 + 4 files changed, 22 insertions(+), 21 deletions(-)
include/linux/timekeeping.h +1 -3
@@ -9,9 +9,7 @@void timekeeping_init (void );extern int timekeeping_suspended; -/* Architecture timer tick functions: */ -extern void update_process_times (int user); -extern void xtime_update (unsigned long ticks); +/* Architecture timer tick functions */ extern void legacy_timer_tick (unsigned long ticks);/*
kernel/time/tick-legacy.c +20 -2
@@ -10,10 +10,28 @@#include "tick-internal.h" +/** +* legacy_timer_tick() - advances the timekeeping infrastructure +* @ticks: number of ticks, that have elapsed since the last call. +* +* This is used by platforms that have not been converted to +* generic clockevents. +* +* If 'ticks' is zero, the CPU is not handling timekeeping, so +* only perform process accounting and profiling. +* +* Must be called with interrupts disabled. +*/ void legacy_timer_tick (unsigned long ticks) { -if (ticks) -xtime_update (ticks); +if (ticks) { +raw_spin_lock (&jiffies_lock); +write_seqcount_begin (&jiffies_seq); +do_timer (ticks); +write_seqcount_end (&jiffies_seq); +raw_spin_unlock (&jiffies_lock); +update_wall_time (); + }update_process_times (user_mode (get_irq_regs ()));profile_tick (CPU_PROFILING ); }
kernel/time/timekeeping.c +0 -16
@@ -2461,19 +2461,3 @@ void hardpps(const struct timespec64 *phase_ts, const struct timespec64 *raw_ts) }EXPORT_SYMBOL (hardpps);#endif /* CONFIG_NTP_PPS */ - -/** -* xtime_update() - advances the timekeeping infrastructure -* @ticks: number of ticks, that have elapsed since the last call. -* -* Must be called with interrupts disabled. -*/ -void xtime_update (unsigned long ticks) -{ -raw_spin_lock (&jiffies_lock); -write_seqcount_begin (&jiffies_seq); -do_timer (ticks); -write_seqcount_end (&jiffies_seq); -raw_spin_unlock (&jiffies_lock); -update_wall_time (); -}
kernel/time/timekeeping.h +1 -0
@@ -22,6 +22,7 @@ static inline int sched_clock_suspend(void) { return 0; }static inline void sched_clock_resume (void ) { }#endif +extern void update_process_times (int user);extern void do_timer (unsigned long ticks);extern void update_wall_time (void );
Signature · 1 line
2.27.0
Re: [PATCH 11/13] timekeeping: remove xtime_update
Hi Arnd,
Quoted 7 lines · Arnd Bergmann
There are no more users of
xtime_updateaside fromlegacy_timer_tick(), so fold it into that function and remove the declaration.
update_process_times()is now only called inside of thekernel/time/code, so the declaration can be moved there.Signed-off-by Arnd Bergmann <[email protected]>
Thanks for your patch!
The comment about xtime_update() in arch/ia64/kernel/time.c needs an update.
Does the comment about update_process_times() in arch/openrisc/kernel/time.c needs an update, too?
For Amiga and Atari/ARAnyM:
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
Re: [PATCH 11/13] timekeeping: remove xtime_update
Quoted 7 lines · Geert Uytterhoeven
Quoted 6 lines · Arnd Bergmann
There are no more users of
xtime_updateaside fromlegacy_timer_tick(), so fold it into that function and remove the declaration.
update_process_times()is now only called inside of thekernel/time/code, so the declaration can be moved there.
Quoted 8 lines
Quoted 1 line
Signed-off-by Arnd Bergmann <[email protected]>Thanks for your patch!
Reviewed-by Geert Uytterhoeven <[email protected]>The comment about
xtime_update()inarch/ia64/kernel/time.cneeds an update.
I think the correct action for ia64 would be to make it a proper clockevent driver with oneshot support, and remove the rest of this logic.
I could try to rewrite the comment, but I tried not to touch that part since I don't understand the logic behind it. Maybe the ia64 maintainers can comment here why it even tries to skip a timer tick. Is there a danger of ending up with the timer irq permanently disabled if the timer_interrupt() function returns with the itm register in the past, or is this simply about not having too many interrupts in a row?
Quoted 2 lines
Does the comment about
update_process_times()inarch/openrisc/kernel/time.cneeds an update, too?
I think that one is still technically correct.
Arnd
Re: [PATCH 11/13] timekeeping: remove xtime_update
Quoted 8 lines · Arnd Bergmann
Quoted 3 lines · Geert Uytterhoeven
On Thu, Oct 8, 2020 at 5:48 PM Arnd Bergmann <[email protected]> wrote:
The comment aboutxtime_update()inarch/ia64/kernel/time.cneeds an update.I think the correct action for ia64 would be to make it a proper clockevent driver with oneshot support, and remove the rest of this logic.
Correct action would be to remove all of arch/ia64 :)
Quoted 7 lines
I could try to rewrite the comment, but I tried not to touch that part since I don't understand the logic behind it. Maybe the ia64 maintainers can comment here why it even tries to skip a timer tick. Is there a danger of ending up with the timer irq permanently disabled if the
timer_interrupt()function returns with the itm register in the past, or is this simply about not having too many interrupts in a row?
There was a comment in the initial ia64 code:
* There is a race condition here: to be on the "safe"
* side, we process timer ticks until itm.next is
* ahead of the itc by at least half the timer
* interval. This should give us enough time to set
* the new itm value without losing a timer tick.
The ITM (Interval Timer Match) register is raising an interrupt when the ITM value matches the ITC (Interval Timer Counter) register. If the new counter is already past the match then the timer interrupt will happen once ITC wrapped around and reaches the match value again. Might take ages for a 64bit counter to do that. :)
IIRC, PXA had the same problem and HPET definitely has it as well. Seems Intel patented the concept of broken timers, but at least they listened when they proposed to implement the TSC deadline timer on x86 in the exact same way.
See hpet_clkevt_set_next_event() for the gory details how to handle that correctly.
Thanks,
tglx
PATCH 12/13 timekeeping: default GENERIC_CLOCKEVENTS to enabled
In reply to Arnd BergmannAlmost all machines use GENERIC_CLOCKEVENTS, so it feels wrong to require each one to select that symbol manually.
Instead, enable it whenever CONFIG_LEGACY_TIMER_TICK is disabled as a simplification. It should be possible to select both GENERIC_CLOCKEVENTS and LEGACY_TIMER_TICK from an architecture now and decide at runtime between the two.
For the clockevents arch-support.txt file, this means that additional architectures are marked as TODO when they have at least one machine that still uses LEGACY_TIMER_TICK, rather than being marked 'ok' when at least one machine has been converted. This means that both m68k and arm (for riscpc) revert to TODO.
At this point, we could just always enable CONFIG_GENERIC_CLOCKEVENTS rather than leaving it off when not needed. I built an m68k defconfig kernel (using gcc-10.1.0) and found that this would add around 5.5KB in kernel image size:
text data bss dec hex filename
3861936 1092236 196656 5150828 4e986c obj-m68k/vmlinux-no-clockevent
3866201 1093832 196184 5156217 4ead79 obj-m68k/vmlinux-clockevent
On Arm (MACH_RPC), that difference appears to be twice as large, around 11KB on top of an 6MB vmlinux.
--- .../features/time/clockevents/arch-support.txt | 6 +++--- arch/alpha/Kconfig | 1 - arch/arc/Kconfig | 1 - arch/arm/Kconfig | 12 ------------ arch/arm64/Kconfig | 1 - arch/arm64/Kconfig.platforms | 1 - arch/c6x/Kconfig | 1 - arch/csky/Kconfig | 1 - arch/h8300/Kconfig | 1 - arch/hexagon/Kconfig | 1 - arch/m68k/Kconfig.cpu | 1 - arch/microblaze/Kconfig | 1 - arch/mips/Kconfig | 1 - arch/nds32/Kconfig | 1 - arch/nios2/Kconfig | 1 - arch/openrisc/Kconfig | 1 - arch/powerpc/Kconfig | 1 - arch/riscv/Kconfig | 1 - arch/s390/Kconfig | 1 - arch/sh/Kconfig | 1 - arch/sparc/Kconfig | 1 - arch/um/Kconfig | 1 - arch/x86/Kconfig | 1 - arch/xtensa/Kconfig | 1 - kernel/time/Kconfig | 2 +- 25 files changed, 4 insertions(+), 38 deletions(-)
Documentation/features/time/clockevents/arch-support.txt +3 -3
@@ -1,6 +1,6 @@ # # Feature name: clockevents -# Kconfig: GENERIC_CLOCKEVENTS +# Kconfig: !LEGACY_TIMER_TICK # description: arch support generic clock events # -----------------------
@@ -8,14 +8,14 @@ ----------------------- | alpha: | ok | | arc: | ok | - | arm: | ok | + | arm: | TODO | | arm64: | ok | | c6x: | ok | | csky: | ok | | h8300: | ok | | hexagon: | ok | | ia64: | TODO | - | m68k: | ok | + | m68k: | TODO | | microblaze: | ok | | mips: | ok | | nds32: | ok |
arch/alpha/Kconfig +0 -1
@@ -27,7 +27,6 @@ config ALPHA select ARCH_WANT_IPC_PARSE_VERSION select ARCH_HAVE_NMI_SAFE_CMPXCHG select AUDIT_ARCH - select GENERIC_CLOCKEVENTS select GENERIC_CPU_VULNERABILITIES select GENERIC_SMP_IDLE_THREAD select GENERIC_STRNCPY_FROM_USER
arch/arc/Kconfig +0 -1
@@ -19,7 +19,6 @@ config ARC select COMMON_CLK select DMA_DIRECT_REMAP select GENERIC_ATOMIC64 if !ISA_ARCV2 || !(ARC_HAS_LL64 && ARC_HAS_LLSC) - select GENERIC_CLOCKEVENTS select GENERIC_FIND_FIRST_BIT # for now, we don't need GENERIC_IRQ_PROBE, CONFIG_GENERIC_IRQ_CHIP select GENERIC_IRQ_SHOW
arch/arm/Kconfig +0 -12
@@ -322,7 +322,6 @@ config ARCH_MULTIPLATFORM select AUTO_ZRELADDR select TIMER_OF select COMMON_CLK - select GENERIC_CLOCKEVENTS select GENERIC_IRQ_MULTI_HANDLER select HAVE_PCI select PCI_DOMAINS_GENERIC if PCI
@@ -337,7 +336,6 @@ config ARM_SINGLE_ARMV7M select TIMER_OF select COMMON_CLK select CPU_V7M - select GENERIC_CLOCKEVENTS select NO_IOPORT_MAP select SPARSE_IRQ select USE_OF
@@ -366,7 +364,6 @@ config ARCH_EP93XX select CLKDEV_LOOKUP select CLKSRC_MMIO select CPU_ARM920T - select GENERIC_CLOCKEVENTS select GPIOLIB select HAVE_LEGACY_CLK help
@@ -376,7 +373,6 @@ config ARCH_FOOTBRIDGE bool "FootBridge" select CPU_SA110 select FOOTBRIDGE - select GENERIC_CLOCKEVENTS select HAVE_IDE select NEED_MACH_IO_H if !MMU select NEED_MACH_MEMORY_H
@@ -404,7 +400,6 @@ config ARCH_IXP4XX select ARCH_SUPPORTS_BIG_ENDIAN select CPU_XSCALE select DMABOUNCE if PCI - select GENERIC_CLOCKEVENTS select GENERIC_IRQ_MULTI_HANDLER select GPIO_IXP4XX select GPIOLIB
@@ -420,7 +415,6 @@ config ARCH_IXP4XX config ARCH_DOVE bool "Marvell Dove" select CPU_PJ4 - select GENERIC_CLOCKEVENTS select GENERIC_IRQ_MULTI_HANDLER select GPIOLIB select HAVE_PCI
@@ -444,7 +438,6 @@ config ARCH_PXA select CLKSRC_MMIO select TIMER_OF select CPU_XSCALE if !CPU_XSC3 - select GENERIC_CLOCKEVENTS select GENERIC_IRQ_MULTI_HANDLER select GPIO_PXA select GPIOLIB
@@ -485,7 +478,6 @@ config ARCH_SA1100 select COMMON_CLK select CPU_FREQ select CPU_SA1100 - select GENERIC_CLOCKEVENTS select GENERIC_IRQ_MULTI_HANDLER select GPIOLIB select HAVE_IDE
@@ -500,7 +492,6 @@ config ARCH_S3C24XX bool "Samsung S3C24XX SoCs" select ATAGS select CLKSRC_SAMSUNG_PWM - select GENERIC_CLOCKEVENTS select GPIO_SAMSUNG select GPIOLIB select GENERIC_IRQ_MULTI_HANDLER
@@ -524,7 +515,6 @@ config ARCH_OMAP1 select ARCH_OMAP select CLKDEV_LOOKUP select CLKSRC_MMIO - select GENERIC_CLOCKEVENTS select GENERIC_IRQ_CHIP select GENERIC_IRQ_MULTI_HANDLER select GPIOLIB
@@ -787,7 +777,6 @@ config ARCH_ACORN config PLAT_IOP bool - select GENERIC_CLOCKEVENTS config PLAT_ORION bool
@@ -1178,7 +1167,6 @@ config HAVE_SMP config SMP bool "Symmetric Multi-Processing" depends on CPU_V6K || CPU_V7 - depends on GENERIC_CLOCKEVENTS depends on HAVE_SMP depends on MMU || ARM_MPU select IRQ_WORK
arch/arm64/Kconfig +0 -1
@@ -101,7 +101,6 @@ config ARM64 select FRAME_POINTER select GENERIC_ALLOCATOR select GENERIC_ARCH_TOPOLOGY - select GENERIC_CLOCKEVENTS select GENERIC_CLOCKEVENTS_BROADCAST select GENERIC_CPU_AUTOPROBE select GENERIC_CPU_VULNERABILITIES
arch/arm64/Kconfig.platforms +0 -1
@@ -256,7 +256,6 @@ config ARCH_TEGRA select ARM_GIC_PM select CLKSRC_MMIO select TIMER_OF - select GENERIC_CLOCKEVENTS select GPIOLIB select PINCTRL select PM
arch/c6x/Kconfig +0 -1
@@ -19,7 +19,6 @@ config C6X select IRQ_DOMAIN select OF select OF_EARLY_FLATTREE - select GENERIC_CLOCKEVENTS select MODULES_USE_ELF_RELA select MMU_GATHER_NO_RANGE if MMU select SET_FS
arch/csky/Kconfig +0 -1
@@ -28,7 +28,6 @@ config CSKY select GENERIC_LIB_UCMPDI2 select GENERIC_ALLOCATOR select GENERIC_ATOMIC64 - select GENERIC_CLOCKEVENTS select GENERIC_CPU_DEVICES select GENERIC_IRQ_CHIP select GENERIC_IRQ_PROBE
arch/h8300/Kconfig +0 -1
@@ -12,7 +12,6 @@ config H8300 select FRAME_POINTER select GENERIC_CPU_DEVICES select MODULES_USE_ELF_RELA - select GENERIC_CLOCKEVENTS select COMMON_CLK select ARCH_WANT_FRAME_POINTERS select OF
arch/hexagon/Kconfig +0 -1
@@ -27,7 +27,6 @@ config HEXAGON select GENERIC_IOMAP select GENERIC_SMP_IDLE_THREAD select STACKTRACE_SUPPORT - select GENERIC_CLOCKEVENTS select GENERIC_CLOCKEVENTS_BROADCAST select MODULES_USE_ELF_RELA select GENERIC_CPU_DEVICES
arch/m68k/Kconfig.cpu +0 -1
@@ -315,7 +315,6 @@ config M54xx config COLDFIRE_PIT_TIMER bool - select GENERIC_CLOCKEVENTS config COLDFIRE_TIMERS bool
arch/microblaze/Kconfig +0 -1
@@ -17,7 +17,6 @@ config MICROBLAZE select COMMON_CLK select DMA_DIRECT_REMAP if MMU select GENERIC_ATOMIC64 - select GENERIC_CLOCKEVENTS select GENERIC_CPU_DEVICES select GENERIC_IDLE_POLL_SETUP select GENERIC_IRQ_PROBE
arch/mips/Kconfig +0 -1
@@ -21,7 +21,6 @@ config MIPS select CPU_NO_EFFICIENT_FFS if (TARGET_ISA_REV < 1) select CPU_PM if CPU_IDLE select GENERIC_ATOMIC64 if !64BIT - select GENERIC_CLOCKEVENTS select GENERIC_CMOS_UPDATE select GENERIC_CPU_AUTOPROBE select GENERIC_GETTIMEOFDAY
arch/nds32/Kconfig +0 -1
@@ -17,7 +17,6 @@ config NDS32 select DMA_DIRECT_REMAP select GENERIC_ATOMIC64 select GENERIC_CPU_DEVICES - select GENERIC_CLOCKEVENTS select GENERIC_IRQ_CHIP select GENERIC_IRQ_SHOW select GENERIC_IOREMAP
arch/nios2/Kconfig +0 -1
@@ -10,7 +10,6 @@ config NIOS2 select COMMON_CLK select TIMER_OF select GENERIC_ATOMIC64 - select GENERIC_CLOCKEVENTS select GENERIC_CPU_DEVICES select GENERIC_IRQ_PROBE select GENERIC_IRQ_SHOW
arch/openrisc/Kconfig +0 -1
@@ -24,7 +24,6 @@ config OPENRISC select GENERIC_CPU_DEVICES select HAVE_UID16 select GENERIC_ATOMIC64 - select GENERIC_CLOCKEVENTS select GENERIC_CLOCKEVENTS_BROADCAST select GENERIC_STRNCPY_FROM_USER select GENERIC_STRNLEN_USER
arch/powerpc/Kconfig +0 -1
@@ -163,7 +163,6 @@ config PPC select EDAC_ATOMIC_SCRUB select EDAC_SUPPORT select GENERIC_ATOMIC64 if PPC32 - select GENERIC_CLOCKEVENTS select GENERIC_CLOCKEVENTS_BROADCAST if SMP select GENERIC_CMOS_UPDATE select GENERIC_CPU_AUTOPROBE
arch/riscv/Kconfig +0 -1
@@ -37,7 +37,6 @@ config RISCV select EDAC_SUPPORT select GENERIC_ARCH_TOPOLOGY if SMP select GENERIC_ATOMIC64 if !64BIT - select GENERIC_CLOCKEVENTS select GENERIC_EARLY_IOREMAP select GENERIC_GETTIMEOFDAY if HAVE_GENERIC_VDSO select GENERIC_IOREMAP
arch/s390/Kconfig +0 -1
@@ -116,7 +116,6 @@ config S390 select CLONE_BACKWARDS2 select DMA_OPS if PCI select DYNAMIC_FTRACE if FUNCTION_TRACER - select GENERIC_CLOCKEVENTS select GENERIC_CPU_AUTOPROBE select GENERIC_CPU_VULNERABILITIES select GENERIC_FIND_FIRST_BIT
arch/sh/Kconfig +0 -1
@@ -16,7 +16,6 @@ config SUPERH select CPU_NO_EFFICIENT_FFS select DMA_DECLARE_COHERENT select GENERIC_ATOMIC64 - select GENERIC_CLOCKEVENTS select GENERIC_CMOS_UPDATE if SH_SH03 || SH_DREAMCAST select GENERIC_IDLE_POLL_SETUP select GENERIC_IRQ_SHOW
arch/sparc/Kconfig +0 -1
@@ -39,7 +39,6 @@ config SPARC select HAVE_EBPF_JIT if SPARC64 select HAVE_DEBUG_BUGVERBOSE select GENERIC_SMP_IDLE_THREAD - select GENERIC_CLOCKEVENTS select GENERIC_STRNCPY_FROM_USER select GENERIC_STRNLEN_USER select MODULES_USE_ELF_RELA
arch/um/Kconfig +0 -1
@@ -17,7 +17,6 @@ config UML select NO_DMA select GENERIC_IRQ_SHOW select GENERIC_CPU_DEVICES - select GENERIC_CLOCKEVENTS select HAVE_GCC_PLUGINS select SET_FS select TTY # Needed for line.c
arch/x86/Kconfig +0 -1
@@ -108,7 +108,6 @@ config X86 select DCACHE_WORD_ACCESS select EDAC_ATOMIC_SCRUB select EDAC_SUPPORT - select GENERIC_CLOCKEVENTS select GENERIC_CLOCKEVENTS_BROADCAST if X86_64 || (X86_32 && X86_LOCAL_APIC) select GENERIC_CLOCKEVENTS_MIN_ADJUST select GENERIC_CMOS_UPDATE
arch/xtensa/Kconfig +0 -1
@@ -16,7 +16,6 @@ config XTENSA select COMMON_CLK select DMA_REMAP if MMU select GENERIC_ATOMIC64 - select GENERIC_CLOCKEVENTS select GENERIC_IRQ_SHOW select GENERIC_PCI_IOMAP select GENERIC_SCHED_CLOCK
kernel/time/Kconfig +1 -1
@@ -32,7 +32,7 @@ config ARCH_USES_GETTIMEOFFSET # The generic clock events infrastructure config GENERIC_CLOCKEVENTS - bool + def_bool !LEGACY_TIMER_TICK # Architecture can handle broadcast in a driver-agnostic way config ARCH_HAS_TICK_BROADCAST
Signature · 1 line
2.27.0
Re: [PATCH 12/13] timekeeping: default GENERIC_CLOCKEVENTS to enabled
Quoted 19 lines · Arnd Bergmann
Almost all machines use
GENERIC_CLOCKEVENTS, so it feels wrong to require each one to select that symbol manually.Instead, enable it whenever
CONFIG_LEGACY_TIMER_TICKis disabled as a simplification. It should be possible to select bothGENERIC_CLOCKEVENTSandLEGACY_TIMER_TICKfrom an architecture now and decide at runtime between the two.For the clockevents arch-support.txt file, this means that additional architectures are marked as TODO when they have at least one machine that still uses
LEGACY_TIMER_TICK, rather than being marked 'ok' when at least one machine has been converted. This means that both m68k and arm (for riscpc) revert to TODO.At this point, we could just always enable
CONFIG_GENERIC_CLOCKEVENTSrather than leaving it off when not needed. I built an m68k defconfig kernel (using gcc-10.1.0) and found that this would add around 5.5KB in kernel image size:
Quoted 8 lines
text data bss dec hex filename
3861936 1092236 196656 5150828 4e986c obj-m68k/vmlinux-no-clockevent 3866201 1093832 196184 5156217 4ead79 obj-m68k/vmlinux-clockeventOn Arm (
MACH_RPC), that difference appears to be twice as large, around 11KB on top of an 6MB vmlinux.Signed-off-by Arnd Bergmann <[email protected]>
Quoted 1 line
arch/m68k/Kconfig.cpu | 1 -
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
RFC 13/13 m68k: mac: convert to generic clockevent
In reply to Arnd BergmannNow that the infrastructure allows kernels to have both legacy timer ticks and clockevent drivers in the same image, start by moving one platform to generic clockevents.
As qemu only supports the q800 platform among the classic m68k, use that as an example.
I also tried adding oneshot mode, which was successful but broke the clocksource. It's probably not hard to make it work properly, but this is where I've stopped.
---
I have never tried implementing a clockevent or clocksource driver in the past, so this is really just an experiment and I expect I got something wrong.
arch/m68k/Kconfig.machine | 2 +- arch/m68k/mac/via.c | 44 ++++++++++++++++++++++++++++++++------- 2 files changed, 37 insertions(+), 9 deletions(-)
arch/m68k/Kconfig.machine +1 -1
@@ -30,7 +30,7 @@ config MAC depends on MMU select MMU_MOTOROLA if MMU select HAVE_ARCH_NVRAM_OPS - select LEGACY_TIMER_TICK + select GENERIC_CLOCKEVENTS help This option enables support for the Apple Macintosh series of computers (yes, there is experimental support now, at least for part
arch/m68k/mac/via.c +36 -8
69 lines not on this page · raw message
Signature · 1 line
2.27.0
Re: [RFC 13/13] m68k: mac: convert to generic clockevent
Hi Arnd,
Perhaps patch 13 does not belong in this series (?).
All m68k platforms will need conversion before the TODO can be removed from Documentation/features/time/clockevents/arch-support.txt.
On m68k, HZ is fixed at 100. Without addressing that, would there be any benefit from adopting GENERIC_CLOCKEVENTS as per this RFC patch?
Quoted 7 lines · Arnd Bergmann
Now that the infrastructure allows kernels to have both legacy timer ticks and clockevent drivers in the same image, start by moving one platform to generic clockevents.
As qemu only supports the q800 platform among the classic m68k, use that as an example.
Correct VIA emulation is suprisingly difficult, so this kind of work should be tested on real hardware.
I say that because when I did the clocksource conversion for m68k I ran into a bug in QEMU (since fixed) and also because I once worked on some of the bugs in the emulated VIA device used in MAME/MESS.
Quoted 4 lines
I also tried adding oneshot mode, which was successful but broke the clocksource. It's probably not hard to make it work properly, but this is where I've stopped.
I'm not so sure that one timer is able to support both a clocksource driver and a clockevent driver. In some cases we may have to drop the clocksource driver (i.e. fall back on the jiffies clocksource).
Anyway, even on Macs with only one VIA chip we still have two timers. So I think we should try to use Timer 1 as a freerunning clocksource and Timer 2 as a oneshot clock event. This may result in better accuracy and simpler code. This may require some experimentation though.
Quoted 6 lines
Signed-off-by Arnd Bergmann <[email protected]>---
I have never tried implementing a clockevent or clocksource driver in the past, so this is really just an experiment and I expect I got something wrong.
Writing clockevent drivers is new to me too. I'm still trying to discover what the implications might be if the only available clockevent device offers oneshot mode or periodic mode but not both.
Re: [RFC 13/13] m68k: mac: convert to generic clockevent
Quoted 7 lines · Finn Thain
Hi Arnd,
Perhaps patch 13 does not belong in this series (?).
All m68k platforms will need conversion before the TODO can be removed from
Documentation/features/time/clockevents/arch-support.txt.
Yes, correct. I marked this patch as RFC instead of PATCH, as I'm just trying to find out where it should be headed. I would hope the other patches can just get merged.
Quoted 2 lines
On m68k, HZ is fixed at 100. Without addressing that, would there be any benefit from adopting
GENERIC_CLOCKEVENTSas per this RFC patch?
I don't think so, I mainly did it to see if there is a problem with mixing the two modes, and I couldn't find any. The behavior seems unchanged before and after my patch, the main difference being a few extra kilobytes in kernel .text for the generic clockevents code.
Quoted 10 lines · Arnd Bergmann
Quoted 8 lines
Now that the infrastructure allows kernels to have both legacy timer ticks and clockevent drivers in the same image, start by moving one platform to generic clockevents.
As qemu only supports the q800 platform among the classic m68k, use that as an example.
Quoted 6 lines
Correct VIA emulation is suprisingly difficult, so this kind of work should be tested on real hardware.
I say that because when I did the clocksource conversion for m68k I ran into a bug in QEMU (since fixed) and also because I once worked on some of the bugs in the emulated VIA device used in MAME/MESS.
Good point, though I would be surprised if anything went wrong with this patch on real hardware but not in emulation, as all the register-level interactions with the timer are the same.
Adding oneshot mode is a completely different matter though, that clearly needs to be tested on real hardware.
Quoted 5 lines
Quoted 4 lines
I also tried adding oneshot mode, which was successful but broke the clocksource. It's probably not hard to make it work properly, but this is where I've stopped.
Quoted 8 lines
I'm not so sure that one timer is able to support both a clocksource driver and a clockevent driver. In some cases we may have to drop the clocksource driver (i.e. fall back on the jiffies clocksource).
Anyway, even on Macs with only one VIA chip we still have two timers. So I think we should try to use Timer 1 as a freerunning clocksource and Timer 2 as a oneshot clock event. This may result in better accuracy and simpler code. This may require some experimentation though.
Ah, good. This is partly what I had been hoping for, as my patch can be used as a starting point for that if you want to give it a go.
Arnd
Re: [RFC 13/13] m68k: mac: convert to generic clockevent
Quoted 9 lines · Arnd Bergmann
Quoted 4 lines
Perhaps patch 13 does not belong in this series (?).
All m68k platforms will need conversion before the TODO can be removed from
Documentation/features/time/clockevents/arch-support.txt.Yes, correct. I marked this patch as RFC instead of PATCH, as I'm just trying to find out where it should be headed. I would hope the other patches can just get merged.
I wonder whether we can improve support for your proposed configuration i.e. a system with no oneshot clockevent device.
The 16 platforms you identified are not all in that category but I suspect that there are others which are (though they don't appear in this series because they already use GENERIC_CLOCKEVENTS).
One useful optimization would be some way to elide oneshot clockevent support (perhaps with the help of Link Time Optimization).
Quoted 8 lines
Quoted 2 lines
On m68k, HZ is fixed at 100. Without addressing that, would there be any benefit from adopting
GENERIC_CLOCKEVENTSas per this RFC patch?I don't think so, I mainly did it to see if there is a problem with mixing the two modes, and I couldn't find any. The behavior seems unchanged before and after my patch, the main difference being a few extra kilobytes in kernel .text for the generic clockevents code.
I think that is a good reason to convert all m68k platforms at once and to elide some of the dead code.
Quoted 13 lines
Quoted 13 lines · Arnd Bergmann
Quoted 8 lines
Now that the infrastructure allows kernels to have both legacy timer ticks and clockevent drivers in the same image, start by moving one platform to generic clockevents.
As qemu only supports the q800 platform among the classic m68k, use that as an example.
Correct VIA emulation is suprisingly difficult, so this kind of work should be tested on real hardware.
Quoted 8 lines
Quoted 3 lines
I say that because when I did the clocksource conversion for m68k I ran into a bug in QEMU (since fixed) and also because I once worked on some of the bugs in the emulated VIA device used in MAME/MESS.
Good point, though I would be surprised if anything went wrong with this patch on real hardware but not in emulation, as all the register-level interactions with the timer are the same.
On the subject of register accesses, via1[ACR] is shared with ADB drivers, so this patch probably has to protect those accesses with local_irq_save/restore or local_irq_disable/enable. (I can't be sure of the contexts in which .set_state_shutdown and .set_state_periodic methods are called.)
Quoted 3 lines
Adding oneshot mode is a completely different matter though, that clearly needs to be tested on real hardware.
Right, and many emulators trade-off timing accuracy for performance which makes them unsuitable for testing invasive changes of that sort.
Quoted 9 lines
Quoted 9 lines
Quoted 4 lines
I also tried adding oneshot mode, which was successful but broke the clocksource. It's probably not hard to make it work properly, but this is where I've stopped.
I'm not so sure that one timer is able to support both a clocksource driver and a clockevent driver. In some cases we may have to drop the clocksource driver (i.e. fall back on the jiffies clocksource).
Quoted 9 lines
Quoted 5 lines
Anyway, even on Macs with only one VIA chip we still have two timers. So I think we should try to use Timer 1 as a freerunning clocksource and Timer 2 as a oneshot clock event. This may result in better accuracy and simpler code. This may require some experimentation though.
Ah, good. This is partly what I had been hoping for, as my patch can be used as a starting point for that if you want to give it a go.
After looking at the chip documentation I don't think it's viable to use the hardware timers in the way I proposed. A VIA register access requires at least one full VIA clock cycle (about 1.3 us) which means register accesses themselves cause timing delays. They also make clocksource reads expensive.
I think this rules out oneshot clockevent devices because if the system offered such a device it would preferentially get used as a tick device.
So I think your approach (periodic clockevent device driven by the existing periodic tick interrupt) is best for this platform due to simplicity (not much code) and performance (good accuracy, no additional overhead).
I suspect the same approach would work equally well on other platforms too (even though they are probably be capable of oneshot clockevent devices).
Quoted 2 lines
Arnd
A message that is not in the archive (sent off-list or not yet received): CAK8P3a2ymv79j1edtJ983-VgjtxvT_6co7V0VRnHzcneW+0ZtA@mail.gmail.com
Re: [RFC 13/13] m68k: mac: convert to generic clockevent
Quoted 21 lines · Arnd Bergmann
Quoted 20 lines · Finn Thain
Quoted 9 lines · Arnd Bergmann
Quoted 4 lines
Perhaps patch 13 does not belong in this series (?).
All m68k platforms will need conversion before the TODO can be removed from
Documentation/features/time/clockevents/arch-support.txt.Yes, correct. I marked this patch as RFC instead of PATCH, as I'm just trying to find out where it should be headed. I would hope the other patches can just get merged.
I wonder whether we can improve support for your proposed configuration i.e. a system with no oneshot clockevent device.
The 16 platforms you identified are not all in that category but I suspect that there are others which are (though they don't appear in this series because they already use
GENERIC_CLOCKEVENTS).
Quoted 7 lines
Quoted 2 lines
One useful optimization would be some way to elide oneshot clockevent support (perhaps with the help of Link Time Optimization).
I think this already happens if one picks
CONFIG_HZ_PERIODICwhile disablingHIGH_RES_TIMERS. In that case,CONFIG_TICK_ONESHOTremains disabled.
That configuration still produces the same 5 KiB of bloat. I see that kernel/time/Kconfig has this --
# Core internal switch. Selected by NO_HZ_COMMON / HIGH_RES_TIMERS. This is
# only related to the tick functionality. Oneshot clockevent devices
# are supported independent of this.
config TICK_ONESHOT
boolBut my question was really about both kinds of dead code (oneshot device support and oneshot tick support). Anyway, after playing with the code for a bit I don't see any easy way to reduce the growth in text size.
Quoted 16 lines
...
Quoted 6 lines
After looking at the chip documentation I don't think it's viable to use the hardware timers in the way I proposed. A VIA register access requires at least one full VIA clock cycle (about 1.3 us) which means register accesses themselves cause timing delays. They also make clocksource reads expensive.
Quoted 8 lines
I think this rules out oneshot clockevent devices because if the system offered such a device it would preferentially get used as a tick device.
So I think your approach (periodic clockevent device driven by the existing periodic tick interrupt) is best for this platform due to simplicity (not much code) and performance (good accuracy, no additional overhead).
Quoted 5 lines
Yes, makes sense. I think the one remaining problem with the periodic mode in this driver is that it can drop timer ticks when interrupts are disabled for too long, while in oneshot mode there may be a way to know how much time has passed since the last tick as long as the counter does not overflow.
Is there any benefit from adopting a oneshot tick (rather than periodic) if no clocksource is consulted when calculating the next interval? (I'm assuming NO_HZ is not in use, for reasons discussed below.)
Quoted 3 lines
I would agree that despite this oneshot mode is probably worse overall for timekeeping if the register accesses introduce systematic errors.
It probably has to be tried. But consulting a VIA clocksource on every tick would be expensive on this platform, so if that was the only way to avoid cumulative errors, I'd probably just stick with the periodic tick.
Quoted 6 lines
...
The arm/rpc timer seems to be roughly in the same category as most of the m68k ones or the i8253 counter on a PC. It's possible that some of them could use the same logic asdrivers/clocksource/i8253.oas long as there is any hardware oneshot mode.
There appear to be 15 platforms in that category. 4 have no clocksource besides the jiffies clocksource, meaning there's no practical alternative to using a periodic tick, like you did in your RFC patch:
arch/m68k/apollo/config.carch/m68k/q40/q40ints.carch/m68k/sun3/sun3ints.carch/m68k/sun3x/time.c
The other 11 platforms in that category also have 'synthetic' clocksources derived from a timer reload interrupt. In 3 cases, the clocksource read method does not (or can not) check for a pending counter reload interrupt. For these also, I see no practical alternative to the approach you've taken in your RFC patch:
arch/m68k/68000/timers.carch/m68k/atari/time.carch/m68k/coldfire/timers.c
That leaves 8 platforms that have reliable clocksource devices which should be able to provide an accurate reading even in the presence of a dropped tick (due to drivers disabling interrupts for too long):
arch/arm/mach-rpc/time.carch/m68k/amiga/config.carch/m68k/bvme6000/config.carch/m68k/coldfire/sltimers.carch/m68k/hp300/time.carch/m68k/mac/via.carch/m68k/mvme147/config.carch/m68k/mvme16x/config.c
But is there any reason to adopt a oneshot tick on any of these platforms, if NO_HZ won't eliminate the timer interrupt that's needed to run a 'synthetic' clocksource?
Re: [RFC 13/13] m68k: mac: convert to generic clockevent
In reply to Finn ThainQuoted 8 lines · Finn Thain
Quoted 2 lines · Arnd Bergmann
Quoted 1 line · Finn Thain
On Sat, 10 Oct 2020, Arnd Bergmann wrote:
That configuration still produces the same 5 KiB of bloat. I see that
kernel/time/Kconfighas this --# Core internal switch. Selected by
NO_HZ_COMMON/HIGH_RES_TIMERS. This is
Quoted 8 lines
# only related to the tick functionality. Oneshot clockevent devices # are supported independent of this. config TICK_ONESHOT boolBut my question was really about both kinds of dead code (oneshot device support and oneshot tick support). Anyway, after playing with the code for a bit I don't see any easy way to reduce the growth in text size.
Did you look more deeply into where those 5KB are? Is this just the code in kernel/time/{clockevents,tick-common}.c and the added platform specific bits, or is there something more? I suppose the sysfs interface and the clockevents_update_freq() logic are not really needed on m68k, but it wouldn't make much sense to split those out either.
How does the 5KB bloat compare to the average bloat we get from one release to the next? Geert has been collecting statistics for this.
Quoted 9 lines
Quoted 5 lines
Yes, makes sense. I think the one remaining problem with the periodic mode in this driver is that it can drop timer ticks when interrupts are disabled for too long, while in oneshot mode there may be a way to know how much time has passed since the last tick as long as the counter does not overflow.
Is there any benefit from adopting a oneshot tick (rather than periodic) if no clocksource is consulted when calculating the next interval? (I'm assuming
NO_HZis not in use, for reasons discussed below.)
If the clocksource does not set CLOCK_SOURCE_IS_CONTINOUS, the kernel will keep using periodic timers and not allow hrtimers.
Quoted 7 lines
Quoted 3 lines
I would agree that despite this oneshot mode is probably worse overall for timekeeping if the register accesses introduce systematic errors.
It probably has to be tried. But consulting a VIA clocksource on every tick would be expensive on this platform, so if that was the only way to avoid cumulative errors, I'd probably just stick with the periodic tick.
I'm sure there is a tradeoff somewhere. Without hrtimers, some events will take longer when they have to wait for the next tick, and using NO_HZ_FULL can help help make things faster on some workloads.
Quoted 6 lines
Quoted 5 lines
The arm/rpc timer seems to be roughly in the same category as most of the m68k ones or the i8253 counter on a PC. It's possible that some of them could use the same logic as
drivers/clocksource/i8253.oas long as there is any hardware oneshot mode.
Quoted 8 lines
There appear to be 15 platforms in that category. 4 have no clocksource besides the jiffies clocksource, meaning there's no practical alternative to using a periodic tick, like you did in your RFC patch:
arch/m68k/apollo/config.carch/m68k/q40/q40ints.carch/m68k/sun3/sun3ints.carch/m68k/sun3x/time.c
Do any of these have users? I'm fairly sure sun3x has never worked in mainline, sun3 seems to still need the same few patches it did 20 years ago. I couldn't find
much about Linux on Apollo or q40, the information on the web for either of them seems to all be for linux-2.4 kernels.
Quoted 9 lines
The other 11 platforms in that category also have 'synthetic' clocksources derived from a timer reload interrupt. In 3 cases, the clocksource read method does not (or can not) check for a pending counter reload interrupt. For these also, I see no practical alternative to the approach you've taken in your RFC patch:
arch/m68k/68000/timers.carch/m68k/atari/time.carch/m68k/coldfire/timers.c
Agreed. It's possible there is a way, but I don't see one either.
Quoted 13 lines
That leaves 8 platforms that have reliable clocksource devices which should be able to provide an accurate reading even in the presence of a dropped tick (due to drivers disabling interrupts for too long):
arch/arm/mach-rpc/time.carch/m68k/amiga/config.carch/m68k/bvme6000/config.carch/m68k/coldfire/sltimers.carch/m68k/hp300/time.carch/m68k/mac/via.carch/m68k/mvme147/config.carch/m68k/mvme16x/config.c
Quoted 3 lines
But is there any reason to adopt a oneshot tick on any of these platforms, if
NO_HZwon't eliminate the timer interrupt that's needed to run a 'synthetic' clocksource?
I would expect that these could be changed to behave more like drivers/clocksource/i8253.c, which does support a clocksource in oneshot mode.
Arnd
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel
Geert Uytterhoeven
maintainerRe: [RFC 13/13] m68k: mac: convert to generic clockevent
Hi Arnd,
Quoted 25 lines · Arnd Bergmann
Quoted 8 lines · Finn Thain
Quoted 2 lines · Arnd Bergmann
Quoted 1 line · Finn Thain
On Sat, 10 Oct 2020, Arnd Bergmann wrote:
That configuration still produces the same 5 KiB of bloat. I see that
kernel/time/Kconfighas this --# Core internal switch. Selected by
NO_HZ_COMMON/HIGH_RES_TIMERS. This isQuoted 8 lines
# only related to the tick functionality. Oneshot clockevent devices # are supported independent of this. config TICK_ONESHOT boolBut my question was really about both kinds of dead code (oneshot device support and oneshot tick support). Anyway, after playing with the code for a bit I don't see any easy way to reduce the growth in text size.
Did you look more deeply into where those 5KB are? Is this just the code in
kernel/time/{clockevents,tick-common}.cand the added platform specific bits, or is there something more? I suppose the sysfs interface and theclockevents_update_freq()logic are not really needed on m68k, but it wouldn't make much sense to split those out either.
Quoted 3 lines
How does the 5KB bloat compare to the average bloat we get from one release to the next? Geert has been collecting statistics for this.
It would be a fair share of the typical increase of ca. 30 KiB per kernel release. Still, it would be lost in the noise of the increase for v5.10-rc1:
add/remove: 1200/455 grow/shrink: 1419/821 up/down: 468970/-93714 (375256)
Function old new delta
_printk_rb_static_infos - 180224 +180224
write_buf 8192 32768 +24576
_printk_rb_static_descs - 24576 +24576
HUF_decompress4X4_usingDTable_internal - 5664 +5664
HUF_decompress4X2_usingDTable_internal - 5006 +5006
__ext4_ioctl - 4774 +4774
sock_ops_convert_ctx_access 3840 8462 +4622
ZSTD_decompressSequences - 3100 +3100
Quoted 14 lines
Quoted 5 lines
Quoted 4 lines
The arm/rpc timer seems to be roughly in the same category as most of the m68k ones or the i8253 counter on a PC. It's possible that some of them could use the same logic as
drivers/clocksource/i8253.oas long as there is any hardware oneshot mode.Quoted 8 lines
There appear to be 15 platforms in that category. 4 have no clocksource besides the jiffies clocksource, meaning there's no practical alternative to using a periodic tick, like you did in your RFC patch:
arch/m68k/apollo/config.carch/m68k/q40/q40ints.carch/m68k/sun3/sun3ints.carch/m68k/sun3x/time.c
Quoted 5 lines
Do any of these have users? I'm fairly sure sun3x has never worked in mainline, sun3 seems to still need the same few patches it did 20 years ago. I couldn't find
much about Linux on Apollo or q40, the information on the web for either of them seems to all be for linux-2.4 kernels.
They probably don't have any users.
AFAIK, the only users are the usual triumvirate of amiga/atari/mac (and perhaps the fleet of VME boards in the Australian navy? ;)
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
Geert Uytterhoeven
maintainerRe: [RFC 13/13] m68k: mac: convert to generic clockevent
Quoted 34 lines · Geert Uytterhoeven
Quoted 25 lines · Arnd Bergmann
Quoted 8 lines · Finn Thain
Quoted 2 lines · Arnd Bergmann
Quoted 1 line · Finn Thain
On Sat, 10 Oct 2020, Arnd Bergmann wrote:
That configuration still produces the same 5 KiB of bloat. I see that
kernel/time/Kconfighas this --# Core internal switch. Selected by
NO_HZ_COMMON/HIGH_RES_TIMERS. This isQuoted 8 lines
# only related to the tick functionality. Oneshot clockevent devices # are supported independent of this. config TICK_ONESHOT boolBut my question was really about both kinds of dead code (oneshot device support and oneshot tick support). Anyway, after playing with the code for a bit I don't see any easy way to reduce the growth in text size.
Did you look more deeply into where those 5KB are? Is this just the code in
kernel/time/{clockevents,tick-common}.cand the added platform specific bits, or is there something more? I suppose the sysfs interface and theclockevents_update_freq()logic are not really needed on m68k, but it wouldn't make much sense to split those out either.Quoted 3 lines
How does the 5KB bloat compare to the average bloat we get from one release to the next? Geert has been collecting statistics for this.
It would be a fair share of the typical increase of ca. 30 KiB per kernel release. Still, it would be lost in the noise of the increase for v5.10-rc1:
Quoted 10 lines
add/remove: 1200/455 grow/shrink: 1419/821 up/down: 468970/-93714 (375256)
Function old new delta
_printk_rb_static_infos- 180224 +180224
write_buf8192 32768 +24576
_printk_rb_static_descs- 24576 +24576
HUF_decompress4X4_usingDTable_internal- 5664 +5664
HUF_decompress4X2_usingDTable_internal- 5006 +5006
__ext4_ioctl- 4774 +4774
sock_ops_convert_ctx_access3840 8462 +4622
ZSTD_decompressSequences- 3100 +3100
FTR, 3.9 KiB reclaimed by upgrading from gcc 8.4.0 in Ubuntu 18.04LTS to gcc 9.3.0 in Ubuntu 20.04LTS.
Gr{oetje,eeting}s,
Geert
Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected] In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that. -- Linus Torvalds
Re: [RFC 13/13] m68k: mac: convert to generic clockevent
In reply to Geert UytterhoevenQuoted 16 lines · Geert Uytterhoeven
Quoted 16 lines
Quoted 15 lines
Quoted 5 lines
The arm/rpc timer seems to be roughly in the same category as most of the m68k ones or the i8253 counter on a PC. It's possible that some of them could use the same logic as
drivers/clocksource/i8253.oas long as there is any hardware oneshot mode.There appear to be 15 platforms in that category. 4 have no clocksource besides the jiffies clocksource, meaning there's no practical alternative to using a periodic tick, like you did in your RFC patch:
arch/m68k/apollo/config.carch/m68k/q40/q40ints.carch/m68k/sun3/sun3ints.carch/m68k/sun3x/time.c
Quoted 7 lines
Quoted 5 lines
Do any of these have users? I'm fairly sure sun3x has never worked in mainline, sun3 seems to still need the same few patches it did 20 years ago. I couldn't find much about Linux on Apollo or q40, the information on the web for either of them seems to all be for linux-2.4 kernels.
They probably don't have any users.
I have access to several Sun 3 machines but no time to work on that port, unfortunately.
Are these 4 platforms (those with no clocksource besides the "jiffies" clocksource) the only reason for CONFIG_TIME_LOW_RES on m68k?
Sam Creasey
maintainerRe: [RFC 13/13] m68k: mac: convert to generic clockevent
Quoted 24 lines · Finn Thain
Quoted 22 lines · Geert Uytterhoeven
Quoted 21 lines
Quoted 11 lines
Quoted 5 lines
The arm/rpc timer seems to be roughly in the same category as most of the m68k ones or the i8253 counter on a PC. It's possible that some of them could use the same logic as
drivers/clocksource/i8253.oas long as there is any hardware oneshot mode.There appear to be 15 platforms in that category. 4 have no clocksource besides the jiffies clocksource, meaning there's no practical alternative to using a periodic tick, like you did in your RFC patch:
Quoted 4 lines
arch/m68k/apollo/config.carch/m68k/q40/q40ints.carch/m68k/sun3/sun3ints.carch/m68k/sun3x/time.cDo any of these have users? I'm fairly sure sun3x has never worked in mainline, sun3 seems to still need the same few patches it did 20 years ago. I couldn't find much about Linux on Apollo or q40, the information on the web for either of them seems to all be for linux-2.4 kernels.
Quoted 7 lines
Quoted 1 line
They probably don't have any users.
I have access to several Sun 3 machines but no time to work on that port, unfortunately.
Are these 4 platforms (those with no clocksource besides the "jiffies" clocksource) the only reason for
CONFIG_TIME_LOW_RESon m68k?
Sun3x was probably at least as close (if not closer) than Sun3 to working in mainline back in the day, but unfortunately I'm in the same place as Finn... I've got the hardware, but time is harder to come by.
-- Sam
Re: [RFC 13/13] m68k: mac: convert to generic clockevent
In reply to Arnd BergmannQuoted 14 lines · Arnd Bergmann
Quoted 13 lines · Finn Thain
Quoted 2 lines · Arnd Bergmann
Quoted 1 line · Finn Thain
On Sat, 10 Oct 2020, Arnd Bergmann wrote:
That configuration still produces the same 5 KiB of bloat. I see that
kernel/time/Kconfighas this --# Core internal switch. Selected by
NO_HZ_COMMON/HIGH_RES_TIMERS. This is# only related to the tick functionality. Oneshot clockevent devices # are supported independent of this. config TICK_ONESHOT bool
Quoted 8 lines
Quoted 4 lines
But my question was really about both kinds of dead code (oneshot device support and oneshot tick support). Anyway, after playing with the code for a bit I don't see any easy way to reduce the growth in text size.
Did you look more deeply into where those 5KB are? Is this just the code in
kernel/time/{clockevents,tick-common}.cand the added platform specific bits, or is there something more?
What I did was to list the relevant functions using scripts/bloat-o-meter and tried stubbing out some code related to oneshot clockevent devices. I didn't find any low fruit and don't plan to pursue that without the help of LTO.
Quoted 7 lines
I suppose the sysfs interface and the
clockevents_update_freq()logic are not really needed on m68k, but it wouldn't make much sense to split those out either.How does the 5KB bloat compare to the average bloat we get from one release to the next? Geert has been collecting statistics for this.
Perhaps that 5 KB is justified by gaining the hrtimers feature... hard to say; it's never been available on these platforms. I can see the value in it though.
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines
Yes, makes sense. I think the one remaining problem with the periodic mode in this driver is that it can drop timer ticks when interrupts are disabled for too long, while in oneshot mode there may be a way to know how much time has passed since the last tick as long as the counter does not overflow.
Quoted 8 lines
Quoted 4 lines
Is there any benefit from adopting a oneshot tick (rather than periodic) if no clocksource is consulted when calculating the next interval? (I'm assuming
NO_HZis not in use, for reasons discussed below.)If the clocksource does not set
CLOCK_SOURCE_IS_CONTINOUS, the kernel will keep using periodic timers and not allow hrtimers.
IIUC, when HIGH_RES_TIMERS=y, the kernel will enable hrtimers only if the platform provides both a continuous clocksource device and a oneshot clockevent device. However, the "jiffies" clocksource does not set CLOCK_SOURCE_IS_CONTINOUS and neither does the one in arch/arm/mach-rpc/time.c.
When HIGH_RES_TIMERS=n and NO_HZ_COMMON=n (which is presently the case for all platforms with GENERIC_CLOCKEVENTS=n) there's no use for a oneshot clockevent device, right?
It seems likely that NO_HZ_COMMON=n because the clocksources on these platforms produce a periodic interrupt regardless (e.g. clocksource/i8253.c, arm/rpc, m68k platform timers etc.).
Finally, HIGH_RES_TIMERS=n seems likely if the only clocksource is unreliable (e.g. because it loses time due to interrupt priorities). There may be a few of platforms in this category (Mac, Atari?).
Quoted 5 lines
Quoted 5 lines
Quoted 4 lines
I would agree that despite this oneshot mode is probably worse overall for timekeeping if the register accesses introduce systematic errors.
Quoted 9 lines
Quoted 4 lines
It probably has to be tried. But consulting a VIA clocksource on every tick would be expensive on this platform, so if that was the only way to avoid cumulative errors, I'd probably just stick with the periodic tick.
I'm sure there is a tradeoff somewhere. Without hrtimers, some events will take longer when they have to wait for the next tick, and using
NO_HZ_FULLcan help help make things faster on some workloads.
Yes, such a tradeoff is discussed in drivers/iio/adc/ep93xx_adc.c.
But OTOH, Documentation/timers/timers-howto.rst says,
On slower systems, (embedded, OR perhaps a speed-stepped PC!) the overhead of setting up the hrtimers for usleep *may* not be worth it
I guess it has to be tried.
Quoted 7 lines
...
Quoted 6 lines
The other 11 platforms in that category also have 'synthetic' clocksources derived from a timer reload interrupt. In 3 cases, the clocksource read method does not (or can not) check for a pending counter reload interrupt. For these also, I see no practical alternative to the approach you've taken in your RFC patch:
Quoted 6 lines
Quoted 3 lines
arch/m68k/68000/timers.carch/m68k/atari/time.carch/m68k/coldfire/timers.cAgreed. It's possible there is a way, but I don't see one either.
For arch/m68k/68000/timers.c, I suppose we may be able to check for the TMR1 bit in the Interrupt Status Register at 0xFFFFF30C or the COMP bit in the Timer Status Register at 0xFFFFF60A. But testing that patch could be difficult.
I expect that equivalent flags are available in Coldfire registers, making it possible to improve upon mcftmr_read_clk() and m68328_read_clk() if need be -- that is, if it turns out that the clocksource interrupt was subject to higher priority IRQs that would slow down the clocksource or defeat its monotonicity.
The other difficulty is a lack of hardware timers. There's only one timer on MC68EZ328. On Atari, for now only Timer C is available though Michael has said that it would be possible to free up Timer D. Some Coldfire chips have only 2 timers and the second timer seems to be allocated to profiling.
Quoted 13 lines
Quoted 13 lines
That leaves 8 platforms that have reliable clocksource devices which should be able to provide an accurate reading even in the presence of a dropped tick (due to drivers disabling interrupts for too long):
arch/arm/mach-rpc/time.carch/m68k/amiga/config.carch/m68k/bvme6000/config.carch/m68k/coldfire/sltimers.carch/m68k/hp300/time.carch/m68k/mac/via.carch/m68k/mvme147/config.carch/m68k/mvme16x/config.c
Quoted 8 lines
Quoted 3 lines
But is there any reason to adopt a oneshot tick on any of these platforms, if
NO_HZwon't eliminate the timer interrupt that's needed to run a 'synthetic' clocksource?I would expect that these could be changed to behave more like
drivers/clocksource/i8253.c, which does support a clocksource in oneshot mode.
I think you meant to write, "... support a clockevent device in oneshot mode". I would agree.
Since Macs do have a spare hardware timer, I will attempt to convert arch/m68k/mac/via.c to a oneshot clockevent, using your GENERIC_CLOCKEVENTS patch series as a basis, and see what effect that has in the NO_HZ_COMMON=n, HIGH_RES_TIMERS=y case.
Quoted 2 lines
Arnd
Greg Ungerer
maintainerRe: [RFC 13/13] m68k: mac: convert to generic clockevent
Quoted 123 lines · Finn Thain
Quoted 14 lines · Arnd Bergmann
Quoted 13 lines · Finn Thain
Quoted 2 lines · Arnd Bergmann
Quoted 1 line · Finn Thain
On Sat, 10 Oct 2020, Arnd Bergmann wrote:
That configuration still produces the same 5 KiB of bloat. I see that
kernel/time/Kconfighas this --# Core internal switch. Selected by
NO_HZ_COMMON/HIGH_RES_TIMERS. This is# only related to the tick functionality. Oneshot clockevent devices # are supported independent of this. config TICK_ONESHOT boolQuoted 8 lines
Quoted 4 lines
But my question was really about both kinds of dead code (oneshot device support and oneshot tick support). Anyway, after playing with the code for a bit I don't see any easy way to reduce the growth in text size.
Did you look more deeply into where those 5KB are? Is this just the code in
kernel/time/{clockevents,tick-common}.cand the added platform specific bits, or is there something more?What I did was to list the relevant functions using
scripts/bloat-o-meterand tried stubbing out some code related to oneshot clockevent devices. I didn't find any low fruit and don't plan to pursue that without the help of LTO.Quoted 7 lines
I suppose the sysfs interface and the
clockevents_update_freq()logic are not really needed on m68k, but it wouldn't make much sense to split those out either.How does the 5KB bloat compare to the average bloat we get from one release to the next? Geert has been collecting statistics for this.
Perhaps that 5 KB is justified by gaining the hrtimers feature... hard to say; it's never been available on these platforms. I can see the value in it though.
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines
Yes, makes sense. I think the one remaining problem with the periodic mode in this driver is that it can drop timer ticks when interrupts are disabled for too long, while in oneshot mode there may be a way to know how much time has passed since the last tick as long as the counter does not overflow.
Quoted 8 lines
Quoted 4 lines
Is there any benefit from adopting a oneshot tick (rather than periodic) if no clocksource is consulted when calculating the next interval? (I'm assuming
NO_HZis not in use, for reasons discussed below.)If the clocksource does not set
CLOCK_SOURCE_IS_CONTINOUS, the kernel will keep using periodic timers and not allow hrtimers.IIUC, when HIGH_RES_TIMERS=y, the kernel will enable hrtimers only if the platform provides both a continuous clocksource device and a oneshot clockevent device. However, the "jiffies" clocksource does not set
CLOCK_SOURCE_IS_CONTINOUSand neither does the one inarch/arm/mach-rpc/time.c.When HIGH_RES_TIMERS=n and NO_HZ_COMMON=n (which is presently the case for all platforms with GENERIC_CLOCKEVENTS=n) there's no use for a oneshot clockevent device, right?
It seems likely that NO_HZ_COMMON=n because the clocksources on these platforms produce a periodic interrupt regardless (e.g.
clocksource/i8253.c, arm/rpc, m68k platform timers etc.).Finally, HIGH_RES_TIMERS=n seems likely if the only clocksource is unreliable (e.g. because it loses time due to interrupt priorities). There may be a few of platforms in this category (Mac, Atari?).
Quoted 5 lines
Quoted 5 lines
Quoted 4 lines
I would agree that despite this oneshot mode is probably worse overall for timekeeping if the register accesses introduce systematic errors.
Quoted 9 lines
Quoted 4 lines
It probably has to be tried. But consulting a VIA clocksource on every tick would be expensive on this platform, so if that was the only way to avoid cumulative errors, I'd probably just stick with the periodic tick.
I'm sure there is a tradeoff somewhere. Without hrtimers, some events will take longer when they have to wait for the next tick, and using
NO_HZ_FULLcan help help make things faster on some workloads.Yes, such a tradeoff is discussed in
drivers/iio/adc/ep93xx_adc.c.But OTOH,
Documentation/timers/timers-howto.rstsays,On slower systems, (embedded, OR perhaps a speed-stepped PC!) the overhead of setting up the hrtimers for usleep *may* not be worth it
I guess it has to be tried.
Quoted 7 lines
...
Quoted 6 lines
The other 11 platforms in that category also have 'synthetic' clocksources derived from a timer reload interrupt. In 3 cases, the clocksource read method does not (or can not) check for a pending counter reload interrupt. For these also, I see no practical alternative to the approach you've taken in your RFC patch:
Quoted 6 lines
Quoted 3 lines
arch/m68k/68000/timers.carch/m68k/atari/time.carch/m68k/coldfire/timers.cAgreed. It's possible there is a way, but I don't see one either.
For
arch/m68k/68000/timers.c, I suppose we may be able to check for the TMR1 bit in the Interrupt Status Register at0xFFFFF30Cor the COMP bit in the Timer Status Register at0xFFFFF60A. But testing that patch could be difficult.I expect that equivalent flags are available in Coldfire registers, making it possible to improve upon
mcftmr_read_clk()andm68328_read_clk()if need be -- that is, if it turns out that the clocksource interrupt was subject to higher priority IRQs that would slow down the clocksource or defeat its monotonicity.
Quoted 5 lines
The other difficulty is a lack of hardware timers. There's only one timer on MC68EZ328. On Atari, for now only Timer C is available though Michael has said that it would be possible to free up Timer D. Some Coldfire chips have only 2 timers and the second timer seems to be allocated to profiling.
FWIW, I would have no problem with ditching the profiling clock, and using both on ColdFire platforms that have this. I doubt anyone has used that profiling setup in a _very_ long time.
Regards
Greg
Re: [RFC 13/13] m68k: mac: convert to generic clockevent
Quoted 33 lines · Greg Ungerer
Quoted 27 lines
Quoted 9 lines
...
Quoted 7 lines
The other 11 platforms in that category also have 'synthetic' clocksources derived from a timer reload interrupt. In 3 cases, the clocksource read method does not (or can not) check for a pending counter reload interrupt. For these also, I see no practical alternative to the approach you've taken in your RFC patch:
Quoted 6 lines
Quoted 3 lines
arch/m68k/68000/timers.carch/m68k/atari/time.carch/m68k/coldfire/timers.cAgreed. It's possible there is a way, but I don't see one either.
For
arch/m68k/68000/timers.c, I suppose we may be able to check for the TMR1 bit in the Interrupt Status Register at0xFFFFF30Cor the COMP bit in the Timer Status Register at0xFFFFF60A. But testing that patch could be difficult.I expect that equivalent flags are available in Coldfire registers, making it possible to improve upon
mcftmr_read_clk()andm68328_read_clk()if need be -- that is, if it turns out that the clocksource interrupt was subject to higher priority IRQs that would slow down the clocksource or defeat its monotonicity.Quoted 5 lines
The other difficulty is a lack of hardware timers. There's only one timer on MC68EZ328. On Atari, for now only Timer C is available though Michael has said that it would be possible to free up Timer D. Some Coldfire chips have only 2 timers and the second timer seems to be allocated to profiling.
Quoted 4 lines
FWIW, I would have no problem with ditching the profiling clock, and using both on ColdFire platforms that have this. I doubt anyone has used that profiling setup in a _very_ long time.
If we ditched the Coldfire profiling clock, it would be possible to dedicate one hardware timer to the clocksource device and the other to the (oneshot) clockevent device.
That's a win if it means that the clocksource can use the full counter width (making timer interrupts less frequent and timer interrupt latency less problematic) and run at higher frequency (making the clocksource more precise).
Also, note that hrtimers won't work with a periodic clockevent device (as in Arnd's RFC patch). If you want hrtimers, I think you'll need both hardware timers or else re-implement the existing synthetic clocksource using the same oneshot timer driving a new oneshot clockevent device.
Please note that the lack of a spare hardware timer is a separate issue to the failure of mcftmr_read_clk() and m68328_read_clk() to check the timer reload interrupt flag (which may make those clocksources needlessly susceptible to issues caused by timer interrupt latency...).
Quoted 15 lines · Arnd Bergmann
When I created the patch removing
CONFIG_ARCH_GETTIMEOFFSET, I also had a look atCONFIG_GENERIC_CLOCKEVENTS, which is selected by most, but not all, platforms today, each of the ones that lack it doing the timer tick slightly differently.The cleanups here make the old platforms a bit more consistent, in multiple ways:
- rather than selecting
GENERIC_CLOCKEVENTSon modern platforms, selectLEGACY_TIMER_TICKon the old ones.- Hide some more of the internal implementation and only provide a single common entry point for the timer tick, which also makes the behavior more consistent.
Quoted 7 lines
- Remove the m68k and arm specific infrastructure pieces and call the common helper directly from each timer interrupt function.
I tested the series on m68k with the qemu q800 target platform, both with the
legacy_timer_tick()implementation and after converting that to a periodic clockevent driver.
I am a big fan of this patch series:
Maybe the RFC patch can be kept in the sidelines but the rest should be merged ASAP in my opinion, the kernel just looks so much better after this.
Yours,
Linus Walleij