Booting a DEBUG_PREEMPT enabled kernel on a CCN-based system
results in the following splat:
[...]
arm-ccn e8000000.ccn: No access to interrupts, using timer.
BUG: using smp_processor_id() in preemptible [00000000] code: swapper/0/1
caller is debug_smp_processor_id+0x1c/0x28
CPU: 1 PID: 1 Comm: swapper/0 Not tainted 4.13.0 #6111
Hardware name: AMD Seattle/Seattle, BIOS 17:08:23 Jun 26 2017
Call trace:
[<ffff000008089e78>] dump_backtrace+0x0/0x278
[<ffff00000808a22c>] show_stack+0x24/0x30
[<ffff000008bc3bc4>] dump_stack+0x8c/0xb0
[<ffff00000852b534>] check_preemption_disabled+0xfc/0x100
[<ffff00000852b554>] debug_smp_processor_id+0x1c/0x28
[<ffff000008551bd8>] arm_ccn_probe+0x358/0x4f0
[...]
as we use smp_processor_id() in the wrong context.
Turn this into a get_cpu()/put_cpu() that extends over the CPU hotplug
registration, making sure that we don't race against a CPU down operation.
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Cc: stable@vger.kernel.org # 4.2+
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Use 'devm_kasprintf()' to simplify the code.
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Acked-by: Scott Branden <scott.branden@broadcom.com>
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
attribute_groups are not supposed to change at runtime. All functions
working with attribute_groups provided by <linux/sysfs.h> work with const
attribute_group. So mark the non-const structs as const.
File size before:
text data bss dec hex filename
9074 5592 416 15082 3aea drivers/bus/arm-ccn.o
File size After adding 'const':
text data bss dec hex filename
9327 5336 416 15079 3ae7 drivers/bus/arm-ccn.o
Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
When the PMU driver is built as a module, the perf expects the
pmu->module to be valid, so that the driver is prevented from
being unloaded while it is in use. Fix the CCN pmu driver to
fill in this field.
Fixes: a33b0daab7 ("bus: ARM CCN PMU driver")
Cc: Pawel Moll <pawel.moll@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Add compatible string for the ARM CCN-502 interconnect
Signed-off-by: Velibor Markovski <velibor.markovski@broadcom.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Scott Branden <scott.branden@broadcom.com>
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Multiplications for the size determination of memory allocations
indicated that array data structures should be processed.
Thus use the corresponding function "devm_kcalloc".
This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
If the driver is built as a module, autoload won't work because the module
alias information is not filled. So user-space can't match the registered
device with the corresponding module.
Export the module alias information using the MODULE_DEVICE_TABLE() macro.
Before this patch:
$ modinfo drivers/bus/arm-ccn.ko | grep alias
$
After this patch:
$ modinfo drivers/bus/arm-ccn.ko | grep alias
alias: of:N*T*Carm,ccn-504C*
alias: of:N*T*Carm,ccn-504
Signed-off-by: Javier Martinez Canillas <javier@osg.samsung.com>
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
When the state names got added a script was used to add the extra argument
to the calls. The script basically converted the state constant to a
string, but the cleanup to convert these strings into meaningful ones did
not happen.
Replace all the useless strings with 'subsys/xxx/yyy:state' strings which
are used in all the other places already.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Sebastian Siewior <bigeasy@linutronix.de>
Link: http://lkml.kernel.org/r/20161221192112.085444152@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
In case the driver registration fails, the hotplug callback is leaked.
Not fatal, because it's never invoked as there are no instances registered,
but wrong nevertheless.
Fixes: fdc15a36d8 ("bus/arm-ccn: Convert to hotplug statemachine")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Pawel Moll <pawel.moll@arm.com>
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will.deacon@arm.com>
Pull CPU hotplug updates from Thomas Gleixner:
"Yet another batch of cpu hotplug core updates and conversions:
- Provide core infrastructure for multi instance drivers so the
drivers do not have to keep custom lists.
- Convert custom lists to the new infrastructure. The block-mq custom
list conversion comes through the block tree and makes the diffstat
tip over to more lines removed than added.
- Handle unbalanced hotplug enable/disable calls more gracefully.
- Remove the obsolete CPU_STARTING/DYING notifier support.
- Convert another batch of notifier users.
The relayfs changes which conflicted with the conversion have been
shipped to me by Andrew.
The remaining lot is targeted for 4.10 so that we finally can remove
the rest of the notifiers"
* 'smp-hotplug-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (46 commits)
cpufreq: Fix up conversion to hotplug state machine
blk/mq: Reserve hotplug states for block multiqueue
x86/apic/uv: Convert to hotplug state machine
s390/mm/pfault: Convert to hotplug state machine
mips/loongson/smp: Convert to hotplug state machine
mips/octeon/smp: Convert to hotplug state machine
fault-injection/cpu: Convert to hotplug state machine
padata: Convert to hotplug state machine
cpufreq: Convert to hotplug state machine
ACPI/processor: Convert to hotplug state machine
virtio scsi: Convert to hotplug state machine
oprofile/timer: Convert to hotplug state machine
block/softirq: Convert to hotplug state machine
lib/irq_poll: Convert to hotplug state machine
x86/microcode: Convert to hotplug state machine
sh/SH-X3 SMP: Convert to hotplug state machine
ia64/mca: Convert to hotplug state machine
ARM/OMAP/wakeupgen: Convert to hotplug state machine
ARM/shmobile: Convert to hotplug state machine
arm64/FP/SIMD: Convert to hotplug state machine
...
The CCN PMU driver leaves the counting logic always enabled, and thus
events are enabled while groups are manipulated. As each event is
stopped and read individually, this leads to arbitrary skew across event
groups, which can be seen if counting several identical events.
To avoid this, implement pmu_{enable,disable} callbacks to stop and
start all counters atomically around event manipulation. As the counters
are now stopped, we cannot poll the cycle counter to wait for events to
drain from the bus. However, as the counters are stopped and the events
will not be read regardless, we can simply allow the bus to drain
naturally.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
The CCN PMU driver has a single hrtimer, used to simulate a periodic
interrupt on systems where the overflow interrupt is not possible to
use. The hrtimer is started when any event is started, and cancelled when
any event is stopped. Thus, stopping a single event is sufficient to
disable to hrtimer, and overflows (of other events) may be lost.
To avoid this, this patch reworks the hrtimer start/cancel to only occur
when the first event is added to a PMU, and the last event removed,
making use of the existing bitmap counting active events.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Currently the IRQ core is permitted to make the CCN PMU IRQ handler
threaded, and will allow userspace to change the CPU affinity of the
interrupt behind our back. Both of these could violate our
synchronisation requirements with the core perf code, which relies upon
strict CPU affinity and disabling of interrupts to guarantee mutual
exclusion in some cases.
As with the CPU PMU drivers, we should request the interrupt with
IRQF_NOBALANCING and IRQF_NO_THREAD, to avoid these issues.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Pawel Moll <pawel.moll@arm.com>
Reviewed-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
CCN PMUs have no knowledge into VM-related origins of the memory
traffic, therefore can't handle requests for host-only or guest-only
events.
Added appropriate exclusions (they should have been there from the
beginning). This required changing the error code returned, as the
userspace tool only re-negotiates the options (exclude_guest is true by
default) only for EINVAL.
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
XP can provide events from two sources: watchpoints, observing traffic
on device ports and PMU looking at internal buses.
Unfortunately the sysfs definition of the PMU events was requiring
port number (instead of bus number) and direction (the buses are
unidirectional), as these fields were shared with the watchpoint event.
Although it does not introduce a major problem (port can be used as
bus alias and direction is simply ignored for XP PMU events), it's
better to fix it now, before external tools start depending on this
behaviour.
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
The code setting XP watchpoint comparator and mask registers should, in
order to be fully compliant with specification, zero one or more most
significant bits of each field. In both L cases it means zeroing bit 63.
The bitmask doing this was wrong, though, zeroing bit 60 instead.
Fortunately, due to a lucky coincidence, this turned out to be fairly
innocent with the existing hardware.
Fixed now.
Cc: stable@vger.kernel.org # 3.17+
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Fuzzing the CCN perf driver revealed a small but definitely dangerous
mistake in the event setup code. When a cycle counter is requested, the
driver should not reconfigure the events bus at all, otherwise it will
corrupt (in most but the simplest cases) its configuration and may end
up accessing XP array out of its bounds and corrupting control
registers.
Reported-by: Mark Rutland <mark.rutland@arm.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Tested-by: Mark Rutland <mark.rutland@arm.com>
Cc: stable@vger.kernel.org # 3.17+
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
The "Miscellaneous Node" fell through cracks of node initialisation,
as its ID is shared with HN-I.
This patch treats MN as a special case (which it is), adding separate
validation check for it and pre-defining the node ID in relevant events
descriptions. That way one can simply run:
# perf stat -a -e ccn/mn_ecbarrier/ <workload>
Additionally, direction in the MN pseudo-events XP watchpoint
definitions is corrected to be "TX" (1) as they are defined from the
crosspoint point of view (thus barriers are transmitted from XP to MN).
Cc: stable@vger.kernel.org # 3.17+
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Install the callbacks via the state machine and let the core invoke
the callbacks on the already online CPUs.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Anna-Maria Gleixner <anna-maria@linutronix.de>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Pawel Moll <pawel.moll@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: rt@linutronix.de
Link: http://lkml.kernel.org/r/20160713153334.768498577@linutronix.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
arm-ccn driver uses irq_set_affinity, which is not exported and
hence cannot be built as a module, eventhough we have all the
bits ready. This patch makes use of the exported helper
irq_set_affinity_hint() instead. Also, the __free_irq expects
the affinity_hint to be NULL when we free the irq. So set the
affinity_hint to NULL at clean up.
Now that we can build it as a module, update the Kconfig to
reflect the change.
Cc: Will Deacon <will.deacon@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Acked-by: Pawel Moll <pawel.moll@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
When PMU context is migrating between CPUs, interrupt affinity is set as
well. Only this should not happen when the CCN interrupt is not being
used at all (the driver is using a hrtimer tick instead).
Fixed now.
Cc: <stable@vger.kernel.org> # 4.2+
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
When migrating events the driver picks another cpu using
cpumask_any_but() function, which returns value >= nr_cpu_ids
when there is none available, not a negative value as the code
assumed. Fixed now.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
hrtimer_start() will no longer defer already expired timers to the
softirq in 4.2, and the __hrtimer_start_range_ns() function is
getting removed, causing build errors when both the tip tree and
the arm-ccn changes are merged.
This changes the code back to using hrtimer_start, which will
do the right thing after this branch gets merged with the
timers update from tip.
As pointed out after a discussion on the mailing list, the result will
not be worse than the what was there before you pulled my updates, as
the code was using normal hrtimer_start(). It's just when I realised
that it should be pinned I looked at what x86 uncore pmu is doing and
shamelessly (and probably a bit mindlessly) copied the "do not wakeup"
version from there.
[arnd: update commit message]
Reported-by: Mark Brown <mark.brown@arm.com>
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
To make events rotation possible, they should be allocated when event
is being ->added(), not during initialisation. This patch moves the
respective code.
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Groups must not mix events from different PMUs (software events are
allowed). Unfortunately the core does not ensures that, so it is
necessary to validate the group at the PMU driver level.
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Since 688d4dfcdd "perf tools: Support
parsing parameterized events" the perf userspace tools understands
"argument=?" syntax in the events file, making sure that required
arguments are provided by the user and not defaulting to 0, causing
confusion.
This patch adds the required arguments lists for CCN events.
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
This patch adds a "cpumask" attribute to CCN's event_source class sysfs
directory. Perf user tool uses it to restrict events to the
processor(s) enumerated in this mask.
This patch provides a single CPU mask, making it possible to run "-a"
perf session (previously it would request the same CCN event, for
example cycle counter, on each available core and most likely fail).
Initially the mask is set to the CPU that happened to probe the driver,
but it will be changed when it is hot-un-plugged (active events are
migrated to another CPU then).
Example:
Performance counter stats for 'system wide':
CPU0 2968148 cycles
CPU1 2236736 cycles
CPU2 1797968 cycles
CPU3 1831715 cycles
CPU1 1201850868 ccn/cycles/
1.001241383 seconds time elapsed
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Events defined as watchpoints on nodes must have their config values
converted so that they apply to the respective node's XP. The
function setting new values was using wrong mask for the "port" field,
resulting in corrupted value. Fixed now.
Cc: stable@vger.kernel.org # 3.17+
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
These are changes for drivers that are intimately tied to some SoC
and for some reason could not get merged through the respective
subsystem maintainer tree.
Most of the new code is for the Keystone Navigator driver, which is
new base support that is going to be needed for their hardware
accelerated network driver and other units.
Most of the commits are for moving old code around from at91 and omap
for things that are done in device drivers nowadays.
- at91: move reset, poweroff, memory and clocksource code into drivers
directories
- socfpga: add edac driver (through arm-soc, as requested by Boris)
- omap: move omap-intc code to drivers/irqchip
- sunxi: added an RTC driver for sun6i
- omap: mailbox driver related changes
- keystone: support for the "Navigator" component
- versatile: new reboot, led and soc drivers
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)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=POJD
-----END PGP SIGNATURE-----
Merge tag 'drivers-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc
Pull ARM SoC driver updates from Arnd Bergmann:
"These are changes for drivers that are intimately tied to some SoC and
for some reason could not get merged through the respective subsystem
maintainer tree.
Most of the new code is for the Keystone Navigator driver, which is
new base support that is going to be needed for their hardware
accelerated network driver and other units.
Most of the commits are for moving old code around from at91 and omap
for things that are done in device drivers nowadays.
- at91: move reset, poweroff, memory and clocksource code into
drivers directories
- socfpga: add edac driver (through arm-soc, as requested by Boris)
- omap: move omap-intc code to drivers/irqchip
- sunxi: added an RTC driver for sun6i
- omap: mailbox driver related changes
- keystone: support for the "Navigator" component
- versatile: new reboot, led and soc drivers"
* tag 'drivers-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc: (92 commits)
bus: arm-ccn: Fix spurious warning message
leds: add device tree bindings for register bit LEDs
soc: add driver for the ARM RealView
power: reset: driver for the Versatile syscon reboot
leds: add a driver for syscon-based LEDs
drivers/soc: ti: fix build break with modules
MAINTAINERS: Add Keystone Multicore Navigator drivers entry
soc: ti: add Keystone Navigator DMA support
Documentation: dt: soc: add Keystone Navigator DMA bindings
soc: ti: add Keystone Navigator QMSS driver
Documentation: dt: soc: add Keystone Navigator QMSS bindings
rtc: sunxi: Depend on platforms sun4i/sun7i that actually have the rtc
rtc: sun6i: Add sun6i RTC driver
irqchip: omap-intc: remove unnecessary comments
irqchip: omap-intc: correct maximum number or MIR registers
irqchip: omap-intc: enable TURBO idle mode
irqchip: omap-intc: enable IP protection
irqchip: omap-intc: remove unnecesary of_address_to_resource() call
irqchip: omap-intc: comment style cleanup
irqchip: omap-intc: minor improvement to omap_irq_pending()
...
Because CCN's cycle counter always runs, it will generate
an interrupt on overflow even if the relevant perf event
was not requested, causing a spurious warning message.
Fixed now by warning on only normal counter unwanted
overflows. Also cleaning the overflow mask at init now,
not to warn on event previously requested by firmware.
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
The function cleaning up an initialized event
was called from the "event_del" handler, instead
of being used as the "destroy" callback. In case of
events group allocation this caused NULL pointer
dereference (as events are added and deleted
multiple times then). Fixed now.
Signed-off-by: Pawel Moll <mail@pawelmoll.com>
Signed-off-by: Kevin Hilman <khilman@linaro.org>
A message warning a user about wrong vc value was printing
out port instead.
Reported-by: Drew Richardson <drew.richardson@arm.com>
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Signed-off-by: Olof Johansson <olof@lixom.net>
The bitfield allocation function returns error condition
as a negative value, but in two cases its result
was assigned to an unsigned member of the hw_perf_event
structure, thus the error would not be ever detected.
Fixed by using an intermediate, signed variable.
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Signed-off-by: Olof Johansson <olof@lixom.net>
Driver providing perf backend for ARM Cache Coherent Network
interconnect. Supports counting all hardware events and crosspoint
watchpoints.
Currently works with CCN-504 only, although there should be
no changes required for CCN-508 (just impossible to test it now).
Signed-off-by: Pawel Moll <pawel.moll@arm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Kim Phillips