Pull in fixes from Daniel Lezcano:
- Fix the vt8500 timer leading to a system lock up when dealing with too
small delta (Roman Volkov)
- Select the CLKSRC_MMIO when the fsl_ftm_timer is enabled with COMPILE_TEST
(Daniel Lezcano)
- Prevent to compile timers using the 'iomem' API when the architecture has
not HAS_IOMEM set (Richard Weinberger)
The vt8500 clocksource driver declares itself as capable to handle the
minimum delay of 4 cycles by passing the value into
clockevents_config_and_register(). The vt8500_timer_set_next_event()
requires the passed cycles value to be at least 16. The impact is that
userspace hangs in nanosleep() calls with small delay intervals.
This problem is reproducible in Linux 4.2 starting from:
c6eb3f70d4 ('hrtimer: Get rid of hrtimer softirq')
From Russell King, more detailed explanation:
"It's a speciality of the StrongARM/PXA hardware. It takes a certain
number of OSCR cycles for the value written to hit the compare registers.
So, if a very small delta is written (eg, the compare register is written
with a value of OSCR + 1), the OSCR will have incremented past this value
before it hits the underlying hardware. The result is, that you end up
waiting a very long time for the OSCR to wrap before the event fires.
So, we introduce a check in set_next_event() to detect this and return
-ETIME if the calculated delta is too small, which causes the generic
clockevents code to retry after adding the min_delta specified in
clockevents_config_and_register() to the current time value.
min_delta must be sufficient that we don't re-trip the -ETIME check - if
we do, we will return -ETIME, forward the next event time, try to set it,
return -ETIME again, and basically lock the system up. So, min_delta
must be larger than the check inside set_next_event(). A factor of two
was chosen to ensure that this situation would never occur.
The PXA code worked on PXA systems for years, and I'd suggest no one
changes this mechanism without access to a wide range of PXA systems,
otherwise they're risking breakage."
Cc: stable@vger.kernel.org
Cc: Russell King <linux@arm.linux.org.uk>
Acked-by: Alexey Charkov <alchark@gmail.com>
Signed-off-by: Roman Volkov <rvolkov@v1ros.org>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Select CLKSRC_MMIO when FSL_FTM_TIMER is enabled. Otherwise it fails to
compile on i386 with COMPILE_TEST=y.
"
on i386:
when CLKSRC_MMIO is not enabled:
drivers/built-in.o: In function `ftm_timer_init':
fsl_ftm_timer.c:(.init.text+0x6842): undefined reference to `clocksource_mmio_readl_up'
fsl_ftm_timer.c:(.init.text+0x6855): undefined reference to `clocksource_mmio_init'
"
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Not every arch has io memory.
So, unbreak the build by fixing the dependencies.
Signed-off-by: Richard Weinberger <richard@nod.at>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Like it's already done in one place in the driver, convert the rest to use pr_*
macros instead of printk(KERN_LEVEL) calls.
While here, join strings to be one string for one line to make grep on them
easier.
There is no functional change.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: http://lkml.kernel.org/r/1451310085-113182-1-git-send-email-andriy.shevchenko@linux.intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Counter overflow detection use for overflow interrupt
Signed-off-by: Yoshinori Sato <ysato@users.sourceforge.jp>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
If by some reason timerclk is not available, both clockevent and
clocksource initializations correctly exit, but output of errno to
kernel log buffer may be confusing:
lpc32xx_clk_init: failed to map system control block registers
lpc32xx_clocksource_init: clock get failed (4294966779)
lpc32xx_clockevent_init: clock get failed (4294966779)
Use signed integer output in the correspondent pr_err() string formats:
lpc32xx_clocksource_init: clock get failed (-517)
lpc32xx_clockevent_init: clock get failed (-517)
Signed-off-by: Vladimir Zapolskiy <vz@mleia.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Now the System stall is observed on TI AM437x based board (am437x-gp-evm)
during resuming from System suspend when ARM Global timer is selected as
clocksource device (CPUIdle not enabled) - SysRq are working, but nothing
else.
The reason of stall is that ARM Global timer loses its contexts during
System suspend:
GT_CONTROL.TIMER_ENABLE = 0 (unbanked)
GT_COUNTERx = 0
Hence, update ARM Global timer driver to reflect above behaviour
- re-enable ARM Global timer on resume (GT_CONTROL.TIMER_ENABLE = 1)
if not enabled.
CC: Arnd Bergmann <arnd@arndb.de>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Felipe Balbi <balbi@ti.com>
Cc: Tony Lindgren <tony@atomide.com>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Reviewed-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Let's assume the counter value is 0xf0000000, the pistachio clocksource
read cycles function should return ~0x0fffffff but actually it returns
0xffffffff0fffffff.
That occurs because:
~(cycle_t)value is different from (cycle_t)~value.
unsigned long val = ~(unsigned long)0xf0000000;
40049a: 48 b8 ff ff ff 0f ff movabs $0xffffffff0fffffff,%rax
unsigned long val = (unsigned long)~0xf0000000;
40049a: 48 c7 45 f8 ff ff ff movq $0xfffffff,-0x8(%rbp)
We fix this issue by calculating bitwise-not counter, then cast to
cycle_t.
Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Use the relaxed version to improve performance. we measured time of
4096 rounds of gt_compare_set() spent on Marvell BG2Q:
before the patch: 3690648ns on average
after the patch: 1083023ns on average
improved by 70%!
Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
It seems gcc can automatically inline apbt_writel() for us, but
apbt_real isn't inlined. This patch makes them inline to get a trivial
performance improvement: 4096 rounds of __apbt_read_clocksource() call
spend time on Marvell BG4CT platform:
before the patch 1275240ns on average
after the patch 1263240ns on average
so we get 1% performance improvement.
Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
It's safe to use the relaxed version. From another side, the relaxed io
accessor macros are available on all architectures now, so we can use
the relaxed versions to get a trivial system performance improvement,
we measured time the following functions spent on Marvell BG4CT:
4096 rounds of __apbt_read_clocksource() call:
before the patch: 1263240ns on average
after the patch: 1250080ns on average
improved by 1%
4096 rounds of apbt_eoi() call:
before the patch: 1290960ns on average
after the patch: 1248240ns on average
4096 rounds of apbt_next_event() call:
before the patch: 3333660ns on average
after the patch: 1322040ns on average
improved by 60%!
Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
On Marvell BG4CT platform, we observed the __apbt_read_clocksource()
return wrong value: Let's assume the APBTMR_N_CURRENT_VALUE value is
0xf0000000, we got 0xffffffff0fffffff, but it should be 0xfffffff.
This issue should be common on all 64bit platforms. We fix the issue
by letting aptb_readl() return u32. apbt_writel() is also updated
to write u32 val rather than unsigned long.
Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
The current code to initialize, register and read the clocksource is
already factored out in mmio.c via the clocksource_mmio_init function.
Factor out the code with the clocksource_mmio_init function.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Add the COMPILE_TEST option so the drivers can be compiled on different
architecture with the 'allyesconfig' kernel configuration.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
For the sake of consistency, let rename all ctrl_out/in calls to the write/read
calls so we have the same API consistent with the other architectures hence
open the door for the increasing of the test compilation coverage.
The unsigned long coercive cast is removed because all variables are set to
the right type "void __iomem *".
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
The current Kconfig option is the H8300 arch option. In order to comply to the
current rule, let's create a specific option for the timer8 and select it
from the arch's Kconfig.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
The current code to initialize, register and read the clocksource is
already factored out in mmio.c via the clocksource_mmio_init function.
The only difference is the readl vs readl_relaxed.
Factor out the code with the clocksource_mmio_init function.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
The function irq_of_parse_and_map returns zero in case of failure.
Fix the return code test to check against zero.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
The current code retrieves the rate value when the timer is enabled which
occurs each time a timer is re-armed. Except if the clock frequency has changed
magically I don't see why this should be done each time.
Retrieve the clock rate value at init time only.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
The time framawork takes care of disabling the interrupts and takes a lock
to prevent races.
Remove the legacy code in the driver taking care of the races.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
The current code assumes the interrupt function is re-entrant.
That is not correct. An interrupt handler is never invoked concurrently. The
interrupt line is masked on all processors.
Remove the chewing flags in the code.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
The value returned in case of error for the 'irq_of_parse_and_map' function is
zero in case of error. Fix the check in the init code.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Specify the delta as parameter for the timer8_clock_event_start function
instead of using a macro to tell PERIODIC or ONESHOT.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
The dev_warn is using the platform driver which was removed in the previous
patch.
Let's replace dev_warn by pr_warn.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Remove some legacy code and replace it by the clksrc-of code.
Do some cleanup and code consolidation.
Signed-off-by: Yoshinori Sato <ysato@users.sourceforge.jp>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Implement an ARM delay timer to be used for udelay(). This allows us to
skip the delay loop calibration at boot on Marvell BG2, BG2Q, BG2CD
platforms. And after this patch, udelay() will be unaffected by CPU
frequency changes.
Note: Although in case there are several possible delay timers, we may
not select the "best" delay timer. Take one Marvell Berlin platform for
example: we have arch timer and dw-apb timer. The arch timer freq is
25MHZ while the dw-apb timer freq is 100MHZ, current selection would
choose the dw-apb timer. But the dw apb timer is on the APB bus while
arch timer sits in CPU, the cost of accessing the apb timer is higher
than the arch timer. We could introduce "rating" concept to delay
timer, but this approach "brings a lot of complexity and workarounds
in the code for a small benefit" as pointed out by Daniel.
Later, Arnd pointed out "However, we could argue that this actually
doesn't matter at all, because the entire point of the ndelay()/
udelay()/mdelay() functions is to waste CPU cycles doing not much at
all, so we can just as well waste them reading the timer register
than spinning on the CPU reading the arch timer more often.", so we
just simply register the dw apb base delay timer.
Signed-off-by: Jisheng Zhang <jszhang@marvell.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
In order to compile on all arch without error with 'allyesconfig' make
sure the platform selected the GENERIC_CLOCKEVENTS. Without this patch
the new added drivers will prevent the kernel to compile on PARISC.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Let the platform's Kconfig to select the clock instead of having a reverse
dependency from the driver to the platform options.
Add the COMPILE_TEST option for the compilation test coverage. Due to the
non portable 'delay' code, this driver is only compilable on ARM.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Tested-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Reviewed-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Reviewed-by: Chanwoo Choi <cw00.choi@samsung.com>
Let the platform's Kconfig to select the clock instead of having a reverse
dependency from the driver to the platform options.
Add the COMPILE_TEST option for the compilation test coverage.
This change is debatable as the option itself in the Kconfig allows to
select the driver for the platform or not. This change will make the prcmu
timer always selected.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Increase the compilation test coverage by adding the COMPILE_TEST option.
Due to the non portable code for the delay timer, this option is only
available for the ARM architecture.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Increase the compilation test coverage by adding the COMPILE_TEST option.
The driver depends on the common clock framework, thus the dependency added
on COMMON_CLK.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Thomas Gleixner
Roman Volkov
Daniel Lezcano
Richard Weinberger
Andy Shevchenko
Yoshinori Sato
Vladimir Zapolskiy
Grygorii Strashko
Jisheng Zhang