In preparation for making the clockevents core NTP correction aware,
all clockevent device drivers must set ->min_delta_ticks and
->max_delta_ticks rather than ->min_delta_ns and ->max_delta_ns: a
clockevent device's rate is going to change dynamically and thus, the
ratio of ns to ticks ceases to stay invariant.
Make the sh_cmt clockevent driver initialize these fields properly.
This patch alone doesn't introduce any change in functionality as the
clockevents core still looks exclusively at the (untouched) ->min_delta_ns
and ->max_delta_ns. As soon as this has changed, a followup patch will
purge the initialization of ->min_delta_ns and ->max_delta_ns from this
driver.
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: Richard Cochran <richardcochran@gmail.com>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Stephen Boyd <sboyd@codeaurora.org>
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: Nicolai Stange <nicstange@gmail.com>
Signed-off-by: John Stultz <john.stultz@linaro.org>
With the upcoming NTP correction related rate adjustments to be implemented
in the clockevents core, the latter needs to get informed about every rate
change of a clockevent device made after its registration.
Currently, sh_cmt violates this requirement in that it registers its
clockevent device with a dummy rate and sets its final ->mult and ->shift
values from its ->set_state_oneshot() and ->set_state_periodic() functions
respectively.
This patch moves the setting of the clockevent device's ->mult and ->shift
values to before its registration.
Note that there has been some back and forth regarding this question with
respect to the clocksource also provided by this driver:
commit f4d7c3565c ("clocksource: sh_cmt: compute mult and shift before
registration")
moves the rate determination from the clocksource's ->enable() function to
before its registration. OTOH, the later
commit 3593f5fe40 ("clocksource: sh_cmt: __clocksource_updatefreq_hz()
update")
basically reverts this, saying
"Without this patch the old code uses clocksource_register() together
with a hack that assumes a never changing clock rate."
However, I checked all current sh_cmt users in arch/sh as well as in
arch/arm/mach-shmobile carefully and right now, none of them changes any
rate in any clock tree relevant to sh_cmt after their respective
time_init(). Since all sh_cmt instances are created after time_init(), none
of them should ever observe any clock rate changes.
What's more, both, a clocksource as well as a clockevent device, can
immediately get selected for use at their registration and thus, enabled
at this point already. So it's probably safer to assume a "never changing
clock rate" here.
- Move the struct sh_cmt_channel's ->rate member to struct sh_cmt_device:
it's a property of the underlying clock which is in turn specific to
the sh_cmt_device.
- Determine the ->rate value in sh_cmt_setup() at device probing rather
than at first usage.
- Set the clockevent device's ->mult and ->shift values right before its
registration.
- Although not strictly necessary for the upcoming clockevent core changes,
set the clocksource's rate at its registration for consistency.
Signed-off-by: Nicolai Stange <nicstange@gmail.com>
Signed-off-by: John Stultz <john.stultz@linaro.org>
There is no point in having an extra type for extra confusion. u64 is
unambiguous.
Conversion was done with the following coccinelle script:
@rem@
@@
-typedef u64 cycle_t;
@fix@
typedef cycle_t;
@@
-cycle_t
+u64
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: John Stultz <john.stultz@linaro.org>
Memory for cmt struct is allocated by kzalloc() in sh_cmt_setup.
Signed-off-by: Alexey Klimov <alexey.klimov@linaro.org>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Pull timer updates from Thomas Gleixner:
"Rather large, but nothing exiting:
- new range check for settimeofday() to prevent that boot time
becomes negative.
- fix for file time rounding
- a few simplifications of the hrtimer code
- fix for the proc/timerlist code so the output of clock realtime
timers is accurate
- more y2038 work
- tree wide conversion of clockevent drivers to the new callbacks"
* 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (88 commits)
hrtimer: Handle failure of tick_init_highres() gracefully
hrtimer: Unconfuse switch_hrtimer_base() a bit
hrtimer: Simplify get_target_base() by returning current base
hrtimer: Drop return code of hrtimer_switch_to_hres()
time: Introduce timespec64_to_jiffies()/jiffies_to_timespec64()
time: Introduce current_kernel_time64()
time: Introduce struct itimerspec64
time: Add the common weak version of update_persistent_clock()
time: Always make sure wall_to_monotonic isn't positive
time: Fix nanosecond file time rounding in timespec_trunc()
timer_list: Add the base offset so remaining nsecs are accurate for non monotonic timers
cris/time: Migrate to new 'set-state' interface
kernel: broadcast-hrtimer: Migrate to new 'set-state' interface
xtensa/time: Migrate to new 'set-state' interface
unicore/time: Migrate to new 'set-state' interface
um/time: Migrate to new 'set-state' interface
sparc/time: Migrate to new 'set-state' interface
sh/localtimer: Migrate to new 'set-state' interface
score/time: Migrate to new 'set-state' interface
s390/time: Migrate to new 'set-state' interface
...
Since the removal of the r8a7740 legacy SoC code in commit
44d88c754e ("ARM: shmobile: Remove legacy SoC code for R-Mobile
A1"), all former users of the "sh-cmt-48-gen2" platform device name are
only supported in generic DT-only ARM multi-platform builds. The driver
doesn't need to match platform devices by name anymore, hence remove the
corresponding platform_device_id entry.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Simon Horman <horms+renesas@verge.net.au>
Migrate sh_cmt driver to the new 'set-state' interface provided by
clockevents core, the earlier 'set-mode' interface is marked obsolete
now.
This also enables us to implement callbacks for new states of clockevent
devices, for example: ONESHOT_STOPPED.
Cc: Magnus Damm <damm+renesas@opensource.se>
Cc: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Since commit 914d7d1484 ("ARM: shmobile: r8a73a4: Remove legacy
code"), all former users of the "sh-cmt-48-gen2" platform device name
are only supported in generic DT-only ARM multi-platform builds. The
driver doesn't need to match platform devices by name anymore, hence
remove the corresponding platform_device_id entry.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Simon Horman <horms+renesas@verge.net.au>
Since commit 59b89af1d5 ("ARM: shmobile: sh7372: Remove Legacy C
SoC code"), there are no more users left of the "sh-cmt-32-fast"
platform device name. Hence remove the corresponding platform_device_id
entry from the driver.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Simon Horman <horms+renesas@verge.net.au>
Currently the sh_cmt clocksource timer is disabled or enabled
unconditionally on clocksource suspend resp. resume, even if a
better clocksource is present (e.g. arch_sys_counter) and the
sh_cmt clocksource is not enabled.
As sh_cmt is a syscore device when its timer is enabled, this
may lead to a genpd.prepared_count imbalance in the presence of
PM Domains, which may cause a lock-up during reboot after s2ram.
During suspend:
- pm_genpd_prepare() is called for all non-syscore devices (incl.
sh_cmt), increasing genpd.prepared_count for each device,
- clocksource.suspend() is called for all clocksource devices,
- sh_cmt_clocksource_suspend() calls sh_cmt_stop(), which is a no-op
as the clocksource was not enabled.
During resume:
- clocksource.resume() is called for all clocksource devices,
- sh_cmt_clocksource_resume() calls sh_cmt_start(), which enables the
clocksource timer, and turns sh_cmt into a syscore device,
- pm_genpd_complete() is called for all non-syscore devices (excl.
sh_cmt now!), decreasing genpd.prepared_count for each device but
sh_cmt.
Now genpd.prepared_count of the PM Domain containing sh_cmt is
still 1 instead of zero. On subsequent suspend/resume cycles,
sh_cmt is still a syscore device, hence it's skipped for
pm_genpd_{prepare,complete}(), keeping the imbalance of
genpd.prepared_count at 1.
During reboot:
- platform_drv_shutdown() is called for any platform device that has
a driver with a .shutdown() method (only rcar-dmac on R-Car Gen2),
- platform_drv_shutdown() calls dev_pm_domain_detach(), which
calls genpd_dev_pm_detach(),
- genpd_dev_pm_detach() keeps calling pm_genpd_remove_device() until
it doesn't return -EAGAIN[*],
- If the device is part of the same PM Domain as sh_cmt,
pm_genpd_remove_device() always fails with -EAGAIN due to
genpd.prepared_count > 0.
- Infinite loop in genpd_dev_pm_detach()[*].
[*] Commit 93af5e9354 ("PM / Domains: Avoid infinite loops in
attach/detach code") already limited the number of loop iterations,
avoiding the lock-up.
To fix this, only disable or enable the clocksource timer on
clocksource suspend resp. resume if the clocksource was enabled.
This was tested on r8a7791/koelsch with the CPG Clock Domain:
- using arch_sys_counter as the clocksource, which is the default, and
which showed the problem,
- using sh_cmt as a clocksource ("echo ffca0000.timer > \
/sys/devices/system/clocksource/clocksource0/current_clocksource"),
which behaves the same as before.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1438875126-12596-2-git-send-email-daniel.lezcano@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Ingo requested this function be renamed to improve readability,
so I've renamed __clocksource_updatefreq_scale() as well as the
__clocksource_updatefreq_hz/khz() functions to avoid
squishedtogethernames.
This touches some of the sh clocksources, which I've not tested.
The arch/arm/plat-omap change is just a comment change for
consistency.
Signed-off-by: John Stultz <john.stultz@linaro.org>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: Dave Jones <davej@codemonkey.org.uk>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Richard Cochran <richardcochran@gmail.com>
Cc: Stephen Boyd <sboyd@codeaurora.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1426133800-29329-13-git-send-email-john.stultz@linaro.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Document DT bindings and parse them in the CMT driver.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Tested-by: Simon Horman <horms+renesas@verge.net.au>
The global spinlock is used to protect the shared start/stop register.
Now that all CMT channels are handled by a single device instance, use a
per-device spinlock.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Tested-by: Simon Horman <horms+renesas@verge.net.au>
Now that all platforms have switched to the new-style platform data,
drop support for the legacy version.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Tested-by: Simon Horman <horms+renesas@verge.net.au>
The site-specific OOM messages are unnecessary, because they
duplicate the MM subsystem generic OOM message.
[dlezcano] : refreshed against latest modifications: kmalloc -> kzalloc
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Clock sources don't need an IRQ, request the IRQ only for channels used
as clock event devices.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Do not include the paragraph about writing to the Free Software
Foundation's mailing address from the sample GPL notice. The FSF has
changed addresses in the past, and may do so again. Linux already
includes a copy of the GPL.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
The sh_cmt driver gets the CMT functional clock using a connection ID of
"cmt_fck". While all SH SoCs create clock lookup entries with a NULL
device ID and a "cmt_fck" connection ID, the ARM SoCs use the device ID
only with a NULL connection ID. This works on legacy platforms but will
break on ARM with DT boot.
Fix the situation by using a connection ID of "fck" in the non-legacy
platform data case. Clock lookup entries will be renamed to use the
device ID as well as the connection ID as platforms get moved to new
platform data. The legacy code will eventually be dropped, leaving us
with device ID based clock lookup, compatible with DT boot.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
CMT hardware devices can support multiple channels, with global
registers and per-channel registers. The sh_cmt driver currently models
the hardware with one Linux device per channel. This model makes it
difficult to handle global registers in a clean way.
Add support for a new model that uses one Linux device per timer with
multiple channels per device. This requires changes to platform data,
add new channel configuration fields.
Support for the legacy model is kept and will be removed after all
platforms switch to the new model.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
All boards use or should use a clock source rating of 125 for the CMT,
hardcode it in the driver.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
All boards use or should use a clock event rating of 125 for the CMT,
hardcode it in the driver.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
The CMT is a global timer not restricted to a single CPU. It has a lower
rating than the TMU or ARM architected timer, but is still useful on
systems where the other timers are stopped during CPU sleep.
When multiple timers are available the timers core selects which timer
to use based on timer ratings.
On SMP systems where timer broadcasting is required, one dummy timer is
instantiated per CPU with a rating of 100. On those systems the CMT
timer has a rating of 80, which makes the dummy timer selected by
default on all CPUs. The CMT is then available, and will be used as a
broadcast timer.
On UP systems no dummy timer is instantiated. The CMT timer has a rating
of 125 on those systems and is used directly as a clock event device for
CPU0 without broadcasting.
The CMT rating shouldn't depend on whether we boot a UP or SMP system.
We can't raise the CMT rating to 125 on SMP systems. This would select
CMT as the clock event device for CPU0 as its rating is higher than the
dummy timer rating, and would leave the system without a broadcast
timer. We could instead lower the rating to 80 on all systems, but that
wouldn't reflect reality as ratings between 1 and 99 are documented as
"unfit for real use".
We should raise the rating above 99 and still have the CMT selected as a
broadcast timer. This can be done by changing the cpumask from
cpumask_of(0) to cpu_possible_mask. In that case the timer selection
logic will prefer the previously probed and already selected dummy timer
for all CPUs based on the fact that already selected per-cpu timers are
preferred over new global timers, regardless of their respective
ratings. This also better reflects reality, as the CMT is not tied to
the boot CPU.
Ideally the timer selection logic should realize that the CMT needs to
be used as a broadcast timer on SMP systems as no other broadcast timer
is available, regardless of the cpumask and rating.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Create a new sh_cmt_info structure to hold static information about the
device model and reference that structure from the sh_cmt_device
structure.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
The channel memory base is channel-specific, add it to the channel
structure in preparation for support of multiple channels per device.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
The mapbase variable points to the mapped base address of the channel,
rename it to mapbase_sh. mapbase_str points to the mapped base address
of the CMT device, rename it to mapbase.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Move the channel setup code from sh_cmt_setup to a new
sh_cmt_setup_channel function and call it from sh_cmt_setup.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Channel data is private as well, rename priv to device to make the
distrinction between the core device and the channels clearer.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Create a new sh_cmt_channel structure to hold the channel-specific
field in preparation for multiple channels per device support.
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
The driver claims it needs to register an interrupt handler too early
for request_irq(). This might have been true in the past, but the only
meaningful difference between request_irq() and setup_irq() today is an
additional kzalloc() call in request_irq(). As the driver calls
kmalloc() itself we know that the slab allocator is available, we can
thus switch to request_irq().
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Prepare the clock at probe time, as there is no other appropriate place
in the driver where we're allowed to sleep.
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
This patch removes the use of the IRQF_DISABLED flag
It's a NOOP since 2.6.35 and it will be removed one day.
[dlezcano] : slightly changed the changelog
Signed-off-by: Michael Opdenacker <michael.opdenacker@free-electrons.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Add support for CMT hardware with 32-bit control and counter
registers, as found on r8a73a4 and r8a7790. To use the CMT
with 32-bit hardware a second I/O memory resource needs to
point out the CMSTR register and it needs to be 32 bit wide.
Signed-off-by: Magnus Damm <damm@opensource.se>
Reviewed-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
The reason for this is to ensure that CMT is probed earlier
than with its previous initcall level, module init.
This came up as a problem with using kzm9g-reference which does
not make use of early timers or devices. In that scenario initialisation
of SDHI and MMCIF both stall on msleep() calls due to the absence
of a initialised clock source.
Boot tested on: armadillo800eva, mackerel and kzm9g
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Add a comment about different register layouts
supported by the CMT driver.
Signed-off-by: Magnus Damm <damm@opensource.se>
Acked-by: John Stultz <john.stultz@linaro.org>
Tested-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
This patch adds control register callbacks for the CMT
driver. At this point only 16-bit access is supported
but in the future this will be updated to allow 32-bit
access as well.
Signed-off-by: Magnus Damm <damm@opensource.se>
Acked-by: John Stultz <john.stultz@linaro.org>
Tested-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Break out the CMCNT and CMCOR register access code
into separate 16-bit and 32-bit functions that are
hooked into callbacks at init time. This reduces
the amount of software calculations happening at
runtime.
Signed-off-by: Magnus Damm <damm@opensource.se>
Acked-by: John Stultz <john.stultz@linaro.org>
Tested-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Update hardware register access code for CMSTR and CMCSR
from using sh_cmt_read() and sh_cmt_write() to make use
of 16-bit register access functions such as sh_cmt_read16()
and sh_cmt_write16(). Also update sh_cmt_read() and
sh_cmt_write() now when the special cases are gone.
This patch moves us one step closer to the goal of separating
counter register access functions from control control register
functions.
Signed-off-by: Magnus Damm <damm@opensource.se>
Acked-by: John Stultz <john.stultz@linaro.org>
Tested-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Cleanup the use of platform_set_drvdata() to reduce code size
Signed-off-by: Shinya Kuribayashi <shinya.kuribayashi.px@renesas.com>
Signed-off-by: Magnus Damm <damm@opensource.se>
Acked-by: John Stultz <john.stultz@linaro.org>
Tested-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Introduce sh_cmt_read_cmstr/cmcsr/cmcnt() and
sh_cmt_write_cmstr/cmcsr/cmcnt/cmcor() to in the
future allow us to split counter registers from
control registers and reduce code complexity by
removing sh_cmt_read() and sh_cmt_write().
Signed-off-by: Magnus Damm <damm@opensource.se>
Acked-by: John Stultz <john.stultz@linaro.org>
Tested-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Move the setup of spinlock and max_match_value to sh_cmt_setup().
There's no need to defer those steps until sh_cmt_register().
Signed-off-by: Shinya Kuribayashi <shinya.kuribayashi.px@renesas.com>
Signed-off-by: Magnus Damm <damm@opensource.se>
Acked-by: John Stultz <john.stultz@linaro.org>
Tested-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Make sure clk_put() is called in case of failure in sh_cmt_setup().
Signed-off-by: Shinya Kuribayashi <shinya.kuribayashi.px@renesas.com>
Signed-off-by: Magnus Damm <damm@opensource.se>
Acked-by: John Stultz <john.stultz@linaro.org>
Tested-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
CONFIG_HOTPLUG is going away as an option. As a result, the __dev*
markings need to be removed.
This change removes the use of __devinit, __devexit_p, __devinitdata,
__devinitconst, and __devexit from these drivers.
Based on patches originally written by Bill Pemberton, but redone by me
in order to handle some of the coding style issues better, by hand.
Cc: Bill Pemberton <wfp5p@virginia.edu>
Cc: John Stultz <johnstul@us.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Modify the SH CMT clock source/clock event device driver to support
runtime PM at a basic level (i.e. device clocks can be disabled and
enabled, but domain power must be on, because the devices have to
be marked as "irq safe").
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Magnus Damm <damm@opensource.se>
