Commit 0b46b8a718 (clocksource: arch_timer: Fix code to use physical
timers when requested) introduces the use of physical counters in the
ARM architected timer driver. However, he arm64 kernel uses CNTVCT in
VDSO. When booting in EL2, the kernel switches to the physical timers to
make things easier for KVM but it continues to use the virtual counter
both in user and kernel. While in such scenario CNTVCT == CNTPCT (since
CNTVOFF is initialised by the kernel to 0), we want to spot firmware
bugs corrupting CNTVOFF early (which would affect CNTVCT).
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Tested-by: Yingjoe Chen <yingjoe.chen@mediatek.com>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
* clocksource/physical-timers:
clocksource: arch_timer: Allow the device tree to specify uninitialized timer registers
clocksource: arch_timer: Fix code to use physical timers when requested
Some 32-bit (ARMv7) systems are architected like this:
* The firmware doesn't know and doesn't care about hypervisor mode and
we don't want to add the complexity of hypervisor there.
* The firmware isn't involved in SMP bringup or resume.
* The ARCH timer come up with an uninitialized offset (CNTVOFF)
between the virtual and physical counters. Each core gets a
different random offset.
* The device boots in "Secure SVC" mode.
* Nothing has touched the reset value of CNTHCTL.PL1PCEN or
CNTHCTL.PL1PCTEN (both default to 1 at reset)
On systems like the above, it doesn't make sense to use the virtual
counter. There's nobody managing the offset and each time a core goes
down and comes back up it will get reinitialized to some other random
value.
This adds an optional property which can inform the kernel of this
situation, and firmware is free to remove the property if it is going
to initialize the CNTVOFF registers when each CPU comes out of reset.
Currently, the best course of action in this case is to use the
physical timer, which is why it is important that CNTHCTL hasn't been
changed from its reset value and it's a reasonable assumption given
that the firmware has never entered HYP mode.
Note that it's been said that on ARMv8 systems the firmware and
kernel really can't be architected as described above. That means
using the physical timer like this really only makes sense for ARMv7
systems.
Signed-off-by: Doug Anderson <dianders@chromium.org>
Signed-off-by: Sonny Rao <sonnyrao@chromium.org>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Olof Johansson <olof@lixom.net>
This is a bug fix for using physical arch timers when
the arch_timer_use_virtual boolean is false. It restores the
arch_counter_get_cntpct() function after removal in
0d651e4e "clocksource: arch_timer: use virtual counters"
We need this on certain ARMv7 systems which are architected like this:
* The firmware doesn't know and doesn't care about hypervisor mode and
we don't want to add the complexity of hypervisor there.
* The firmware isn't involved in SMP bringup or resume.
* The ARCH timer come up with an uninitialized offset between the
virtual and physical counters. Each core gets a different random
offset.
* The device boots in "Secure SVC" mode.
* Nothing has touched the reset value of CNTHCTL.PL1PCEN or
CNTHCTL.PL1PCTEN (both default to 1 at reset)
One example of such as system is RK3288 where it is much simpler to
use the physical counter since there's nobody managing the offset and
each time a core goes down and comes back up it will get reinitialized
to some other random value.
Fixes: 0d651e4e65 ("clocksource: arch_timer: use virtual counters")
Cc: stable@vger.kernel.org
Signed-off-by: Sonny Rao <sonnyrao@chromium.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: Olof Johansson <olof@lixom.net>
Commit c387f07e62 (clocksource: arm_arch_timer: Discard unavailable
timers correctly) changed the way the driver makes sure both the memory
and system-register timers have been probed before finalizing the probing.
There is a interesting flaw in this logic that leads to this final step
never to be executed. Things seems to work pretty well until something
actually needs the data that is produced during this final stage.
For example, KVM explodes on the first run of a guest when executed on
a platform that has both memory and sysreg nodes (Juno, for example).
Just fix the damned logic, and enjoy booting VMs again.
Tested on a Juno system.
Cc: Sudeep Holla <sudeep.holla@arm.com>
Cc: Stephen Boyd <sboyd@codeaurora.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: Christoffer Dall <christoffer.dall@linaro.org>
Reported-by: Riku Voipio <riku.voipio@linaro.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Sudeep Holla <sudeep.holla@arm.com>
Tested-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
The arch_timer_evtstrm_enable hooks in arm and arm64 are substantially
similar, the only difference being a CONFIG_COMPAT-conditional section
which is relevant only for arm64. Copy the arm64 version to the
driver, removing the arch-specific hooks.
Signed-off-by: Nathan Lynch <nathan_lynch@mentor.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Will Deacon <will.deacon@arm.com>
The only difference between arm and arm64's implementations of
arch_counter_set_user_access is that 32-bit ARM does not enable user
access to the virtual counter. We want to enable this access for the
32-bit ARM VDSO, so copy the arm64 version to the driver itself, and
remove the arch-specific implementations.
Signed-off-by: Nathan Lynch <nathan_lynch@mentor.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Will Deacon <will.deacon@arm.com>
The arm and arm64 VDSOs need CP15 access to the architected counter.
If this is unavailable (which is allowed by ARM v7), indicate this by
changing the clocksource name to "arch_mem_counter" before registering
the clocksource.
Suggested by Stephen Boyd.
Signed-off-by: Nathan Lynch <nathan_lynch@mentor.com>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Will Deacon <will.deacon@arm.com>
Currently we wait until both cp15 and mem timers are probed if we
have both timer device nodes present in the device tree without
checking if the device is actually available. If one of the timer
device node present is disabled, the system locks up on the boot
as no timer gets registered.
This patch adds the check for the availability of the timer device
so that unavailable timers are discarded correctly. It also adds
the missing of_node_put.
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
ARM arch timers are tightly coupled with the CPU logic and lose context
on platform implementing HW power management when cores are powered
down at run-time. Marking the arch timers as C3STOP regardless of power
management capabilities causes issues on platforms with no power management,
since in that case the arch timers cannot possibly enter states where the
timer loses context at runtime and therefore can always be used as a high
resolution clockevent device.
In order to fix the C3STOP issue in a way compliant with how real HW
works, this patch adds a boolean property to the arch timer bindings
to define if the arch timer is managed by an always-on power domain.
This power domain is present on all ARM platforms to date, and manages
HW that must not be turned off, whatever the state of other HW
components (eg power controller). On platforms with no power management
capabilities, it is the only power domain present, which encompasses
and manages power supply for all HW components in the system.
If the timer is powered by the always-on power domain, the always-on
property must be present in the bindings which means that the timer cannot
be shutdown at runtime, so it is not a C3STOP clockevent device.
If the timer binding does not contain the always-on property, the timer is
assumed to be power-gateable, hence it must be defined as a C3STOP
clockevent device.
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: Magnus Damm <damm@opensource.se>
Cc: Marc Carino <marc.ceeeee@gmail.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Acked-by: Rob Herring <robh@kernel.org>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Set the CLOCK_EVT_FEAT_DYNIRQ flag on the memory mapped
clockevent so that we save power by waking up the CPU with the
next event when this timer is used in broadcast mode.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Commit:
65cd4f6 ("arch_timer: Move to generic sched_clock framework")
added code to register the arch_sys_counter in arch_timer_register(),
but it is already registered in arch_counter_register().
This results in the timer being added to the clocksource list twice,
therefore causing an infinite loop in the list.
Remove the duplicate registration and register the scheduler
clock after the original registration instead.
This fixes a hang during boot on Tegra114 (Cortex-A15).
[ While I've only tested this on Tegra114, I suspect the same hang
during boot happens for all processors that use this clock source. ]
Signed-off-by: Thierry Reding <treding@nvidia.com>
Acked-by: John Stultz <john.stultz@linaro.org>
Cc: Stephen Boyd <sboyd@codeaurora.org>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Stephen Warren <swarren@wwwdotorg.org>
Cc: linux-arm-kernel@lists.infradead.org
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: http://lkml.kernel.org/r/1381843911-31962-1-git-send-email-treding@nvidia.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Pull more timekeeping items for v3.13 from John Stultz:
* Small cleanup in the clocksource code.
* Fix for rtc-pl031 to let it work with alarmtimers.
* Move arm64 to using the generic sched_clock framework & resulting
cleanup in the generic sched_clock code.
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Register with the generic sched_clock framework now that it
supports 64 bits. This fixes two problems with the current
sched_clock support for machines using the architected timers.
First off, we don't subtract the start value from subsequent
sched_clock calls so we can potentially start off with
sched_clock returning gigantic numbers. Second, there is no
support for suspend/resume handling so problems such as discussed
in 6a4dae5 (ARM: 7565/1: sched: stop sched_clock() during
suspend, 2012-10-23) can happen without this patch. Finally, it
allows us to move the sched_clock setup into drivers clocksource
out of the arch ports.
Cc: Christopher Covington <cov@codeaurora.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: John Stultz <john.stultz@linaro.org>
Adds support to configure the rate and enable the event stream for architected
timer. The event streams can be used to impose a timeout on a wfe, to safeguard
against any programming error in case an expected event is not generated or
even to implement wfe-based timeouts for userspace locking implementations.
This feature can be disabled(enabled by default).
Since the timer control register is reset to zero on warm boot, CPU PM notifier
is added to save and restore the value.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
The ARM architected timers keep counting during suspend so we can
mark this clocksource with the CLOCK_SOURCE_SUSPEND_NONSTOP flag.
This flag will indicate that this clocksource can be used for
calculating suspend time and injecting sleep time into the
timekeeping core. This should be more accurate than using an
external RTC or architecture specific persistent clock.
Cc: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Few control settings done in architected timer as part of initialisation
can be lost when CPU enters deeper power states. They need to be
restored when the CPU is (warm)reset again.
This patch adds CPU PM notifiers to save the counter control register
when entering low power modes and restore it when CPU exits low power.
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Acked-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Sudeep KarkadaNagesha <sudeep.karkadanagesha@arm.com>
The ARM architected timer can generate events (used for waking up
CPUs executing the wfe instruction) at a frequency represented as a
power-of-2 divisor of the clock rate.
An event stream might be used:
- To implement wfe-based timeouts for userspace locking implementations.
- To impose a timeout on a wfe for safeguarding against any programming
error in case an expected event is not generated.
This patch computes the event stream frequency aiming for a period
of 100us between events. It uses ARM/ARM64 specific backends to configure
and enable the event stream.
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Will Deacon <will.deacon@arm.com>
[sudeep: moving ARM/ARM64 changes into separate patches
and adding Kconfig option]
Signed-off-by: Sudeep KarkadaNagesha <sudeep.karkadanagesha@arm.com>
* Support for memory mapped arch_timers
* Trivial fixes to the moxart timer code
* Documentation updates
Trivial conflicts in drivers/clocksource/arm_arch_timer.c. Fixed up
the newly added __cpuinit annotations as well.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Add support for the memory mapped timers by filling in the
read/write functions and adding some parsing code. Note that we
only register one clocksource, preferring the cp15 based
clocksource over the mmio one.
To keep things simple we register one global clockevent. This
covers the case of UP and SMP systems with only mmio hardware and
systems where the memory mapped timers are used as the broadcast
timer in low power modes.
The DT binding allows for per-CPU memory mapped timers in case we
want to support that in the future, but the code isn't added
here. We also don't do much for hypervisor support, although it
should be possible to support it by searching for at least two
frames where one frame has the virtual capability and then
updating KVM timers to support it.
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marc Zyngier <Marc.Zyngier@arm.com>
Cc: Rob Herring <robherring2@gmail.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
We're going to introduce support to read and write the memory
mapped timer registers in the next patch, so push the cp15
read/write functions one level deeper. This simplifies the next
patch and makes it clearer what's going on.
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marc Zyngier <Marc.Zyngier@arm.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
There isn't any reason why we don't pass the event here and we'll
need it in the near future for memory mapped arch timers anyway.
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marc Zyngier <Marc.Zyngier@arm.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Using an enum for the register we wish to access allows newer
compilers to determine if we've forgotten a case in our switch
statement. This allows us to remove the BUILD_BUG() instances in
the arm64 port, avoiding problems where optimizations may not
happen.
To try and force better code generation we're currently marking
the accessor functions as inline, but newer compilers can ignore
the inline keyword unless it's marked __always_inline. Luckily on
arm and arm64 inline is __always_inline, but let's make
everything __always_inline to be explicit.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Marc Zyngier <Marc.Zyngier@arm.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Acked-by: Mark Rutland <mark.rutland@arm.com>
The __cpuinit type of throwaway sections might have made sense
some time ago when RAM was more constrained, but now the savings
do not offset the cost and complications. For example, the fix in
commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time")
is a good example of the nasty type of bugs that can be created
with improper use of the various __init prefixes.
After a discussion on LKML[1] it was decided that cpuinit should go
the way of devinit and be phased out. Once all the users are gone,
we can then finally remove the macros themselves from linux/init.h.
This removes all the drivers/clocksource and drivers/irqchip uses of
the __cpuinit macros from all C files.
[1] https://lkml.org/lkml/2013/5/20/589
Cc: John Stultz <john.stultz@linaro.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Switching between reading the virtual or physical counters is
problematic, as some core code wants a view of time before we're fully
set up. Using a function pointer and switching the source after the
first read can make time appear to go backwards, and having a check in
the read function is an unfortunate block on what we want to be a fast
path.
Instead, this patch makes us always use the virtual counters. If we're a
guest, or don't have hyp mode, we'll use the virtual timers, and as such
don't care about CNTVOFF as long as it doesn't change in such a way as
to make time appear to travel backwards. As the guest will use the
virtual timers, a (potential) KVM host must use the physical timers
(which can wake up the host even if they fire while a guest is
executing), and hence a host must have CNTVOFF set to zero so as to have
a consistent view of time between the physical timers and virtual
counters.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
Cc: Rob Herring <rob.herring@calxeda.com>
Hot-plugging with CONFIG_DEBUG_PREEMPT=y on a device with arm
architected timers causes a slew of "using smp_processor_id() in
preemptible" warnings:
BUG: using smp_processor_id() in preemptible [00000000] code: sh/111
caller is arch_timer_cpu_notify+0x14/0xc8
This happens because sometimes the cpu notifier,
arch_timer_cpu_notify(), is called in preemptible context and
other times in non-preemptible context but we use this_cpu_ptr()
to retrieve the clockevent in all cases. We're only going to
actually use the pointer in non-preemptible context though, so
push the this_cpu_ptr() access down into the cases to force the
checks to occur only in non-preemptible contexts.
Cc: John Stultz <john.stultz@linaro.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Mark Rutland <mark.rutland@arm.com>
Acked-by: Marc Zyngier <Marc.Zyngier@arm.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Olof Johansson <olof@lixom.net>
This converts arm and arm64 to use CLKSRC_OF DT based initialization for
the arch timer. A new function arch_timer_arch_init is added to allow for
arch specific setup.
This has a side effect of enabling sched_clock on omap5 and exynos5. There
should not be any reason not to use the arch timers for sched_clock.
Signed-off-by: Rob Herring <rob.herring@calxeda.com>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Kukjin Kim <kgene.kim@samsung.com>
Cc: Tony Lindgren <tony@atomide.com>
Cc: Simon Horman <horms@verge.net.au>
Cc: Magnus Damm <magnus.damm@gmail.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-samsung-soc@vger.kernel.org
Cc: linux-omap@vger.kernel.org
Cc: linux-sh@vger.kernel.org
Acked-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
If we're booted in HYP mode, it is possible that we'll run some
kind of virtualized environment. In this case, it is a better to
switch to the physical timers, and leave the virtual timers to
guests.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
The arch_timer driver supports a superset of the functionality of the
arm_generic driver, and is not tied to a particular arch.
This patch moves arm64 to use the arch_timer driver, gaining additional
functionality in doing so, and removes the (now unused) arm_generic
driver. Timer-related hooks specific to arm64 are moved into
arch/arm64/kernel/time.c.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
The core functionality of the arch_timer driver is not directly tied to
anything under arch/arm, and can be split out.
This patch factors out the core of the arch_timer driver, so it can be
shared with other architectures. A couple of functions are added so
that architecture-specific code can interact with the driver without
needing to touch its internals.
The ARM_ARCH_TIMER config variable is moved out to
drivers/clocksource/Kconfig, existing uses in arch/arm are replaced with
HAVE_ARM_ARCH_TIMER, which selects it.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
