Merge Energy Model and power capping updates for 5.16-rc1:
- Add support for inefficient operating performance points to the
Energy Model and modify cpufreq to use them properly (Vincent
Donnefort).
- Rearrange the DTPM framework code to simplify it and make it easier
to follow (Daniel Lezcano).
- Fix power intialization in DTPM (Daniel Lezcano).
- Add CPU load consideration when estimating the instaneous power
consumption in DTPM (Daniel Lezcano).
* pm-em:
cpufreq: mediatek-hw: Fix cpufreq_table_find_index_dl() call
PM: EM: Mark inefficiencies in CPUFreq
cpufreq: Use CPUFREQ_RELATION_E in DVFS governors
cpufreq: Introducing CPUFREQ_RELATION_E
cpufreq: Add an interface to mark inefficient frequencies
cpufreq: Make policy min/max hard requirements
PM: EM: Allow skipping inefficient states
PM: EM: Extend em_perf_domain with a flag field
PM: EM: Mark inefficient states
PM: EM: Fix inefficient states detection
* powercap:
powercap/drivers/dtpm: Fix power limit initialization
powercap/drivers/dtpm: Scale the power with the load
powercap/drivers/dtpm: Use container_of instead of a private data field
powercap/drivers/dtpm: Simplify the dtpm table
powercap/drivers/dtpm: Encapsulate even more the code
In order to increase the self-encapsulation of the dtpm generic code,
the following changes are adding a power update ops to the dtpm
ops. That allows the generic code to call directly the dtpm backend
function to update the power values.
The power update function does compute the power characteristics when
the function is invoked. In the case of the CPUs, the power
consumption depends on the number of online CPUs. The online CPUs mask
is not up to date at CPUHP_AP_ONLINE_DYN state in the tear down
callback. That is the reason why the online / offline are at separate
state. As there is already an existing state for DTPM, this one is
only moved to the DEAD state, so there is no addition of new state
with these changes. The dtpm node is not removed when the cpu is
unplugged.
That simplifies the code for the next changes and results in a more
self-encapsulated code.
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Reviewed-by: Lukasz Luba <lukasz.luba@arm.com>
Link: https://lore.kernel.org/r/20210312130411.29833-1-daniel.lezcano@linaro.org
The node demotion order needs to be updated during CPU hotplug. Because
whether a NUMA node has CPU may influence the demotion order. The
update function should be called during CPU online/offline after the
node_states[N_CPU] has been updated. That is done in
CPUHP_AP_ONLINE_DYN during CPU online and in CPUHP_MM_VMSTAT_DEAD during
CPU offline. But in commit 884a6e5d1f ("mm/migrate: update node
demotion order on hotplug events"), the function to update node demotion
order is called in CPUHP_AP_ONLINE_DYN during CPU online/offline. This
doesn't satisfy the order requirement.
For example, there are 4 CPUs (P0, P1, P2, P3) in 2 sockets (P0, P1 in S0
and P2, P3 in S1), the demotion order is
- S0 -> NUMA_NO_NODE
- S1 -> NUMA_NO_NODE
After P2 and P3 is offlined, because S1 has no CPU now, the demotion
order should have been changed to
- S0 -> S1
- S1 -> NO_NODE
but it isn't changed, because the order updating callback for CPU
hotplug doesn't see the new nodemask. After that, if P1 is offlined,
the demotion order is changed to the expected order as above.
So in this patch, we added CPUHP_AP_MM_DEMOTION_ONLINE and
CPUHP_MM_DEMOTION_DEAD to be called after CPUHP_AP_ONLINE_DYN and
CPUHP_MM_VMSTAT_DEAD during CPU online and offline, and register the
update function on them.
Link: https://lkml.kernel.org/r/20210929060351.7293-1-ying.huang@intel.com
Fixes: 884a6e5d1f ("mm/migrate: update node demotion order on hotplug events")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Wei Xu <weixugc@google.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Keith Busch <kbusch@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull xfs updates from Darrick Wong:
"There's a lot in this cycle.
Starting with bug fixes: To avoid livelocks between the logging code
and the quota code, we've disabled the ability of quotaoff to turn off
quota accounting. (Admins can still disable quota enforcement, but
truly turning off accounting requires a remount.) We've tried to do
this in a careful enough way that there shouldn't be any user visible
effects aside from quotaoff no longer randomly hanging the system.
We've also fixed some bugs in runtime log behavior that could trip up
log recovery if (otherwise unrelated) transactions manage to start and
commit concurrently; some bugs in the GETFSMAP ioctl where we would
incorrectly restrict the range of records output if the two xfs
devices are of different sizes; a bug that resulted in fallocate
funshare failing unnecessarily; and broken behavior in the xfs inode
cache when DONTCACHE is in play.
As for new features: we now batch inode inactivations in percpu
background threads, which sharply decreases frontend thread wait time
when performing file deletions and should improve overall directory
tree deletion times. This eliminates both the problem where closing an
unlinked file (especially on a frozen fs) can stall for a long time,
and should also ease complaints about direct reclaim bogging down on
unlinked file cleanup.
Starting with this release, we've enabled pipelining of the XFS log.
On workloads with high rates of metadata updates to different shards
of the filesystem, multiple threads can be used to format committed
log updates into log checkpoints.
Lastly, with this release, two new features have graduated to
supported status: inode btree counters (for faster mounts), and
support for dates beyond Y2038. Expect these to be enabled by default
in a future release of xfsprogs.
Summary:
- Fix a potential log livelock on busy filesystems when there's so
much work going on that we can't finish a quotaoff before filling
up the log by removing the ability to disable quota accounting.
- Introduce the ability to use per-CPU data structures in XFS so that
we can do a better job of maintaining CPU locality for certain
operations.
- Defer inode inactivation work to per-CPU lists, which will help us
batch that processing. Deletions of large sparse files will
*appear* to run faster, but all that means is that we've moved the
work to the backend.
- Drop the EXPERIMENTAL warnings from the y2038+ support and the
inode btree counters, since it's been nearly a year and no
complaints have come in.
- Remove more of our bespoke kmem* variants in favor of using the
standard Linux calls.
- Prepare for the addition of log incompat features in upcoming
cycles by actually adding code to support this.
- Small cleanups of the xattr code in preparation for landing support
for full logging of extended attribute updates in a future cycle.
- Replace the various log shutdown state and flag code all over xfs
with a single atomic bit flag.
- Fix a serious log recovery bug where log item replay can be skipped
based on the start lsn of a transaction even though the transaction
commit lsn is the key data point for that by enforcing start lsns
to appear in the log in the same order as commit lsns.
- Enable pipelining in the code that pushes log items to disk.
- Drop ->writepage.
- Fix some bugs in GETFSMAP where the last fsmap record reported for
a device could extend beyond the end of the device, and a separate
bug where query keys for one device could be applied to another.
- Don't let GETFSMAP query functions edit their input parameters.
- Small cleanups to the scrub code's handling of perag structures.
- Small cleanups to the incore inode tree walk code.
- Constify btree function parameters that aren't changed, so that
there will never again be confusion about range query functions
changing their input parameters.
- Standardize the format and names of tracepoint data attributes.
- Clean up all the mount state and feature flags to use wrapped
bitset functions instead of inconsistently open-coded flag checks.
- Fix some confusion between xfs_buf hash table key variable vs.
block number.
- Fix a mis-interaction with iomap where we reported shared delalloc
cow fork extents to iomap, which would cause the iomap unshare
operation to return IO errors unnecessarily.
- Fix DONTCACHE behavior"
* tag 'xfs-5.15-merge-6' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux: (103 commits)
xfs: fix I_DONTCACHE
xfs: only set IOMAP_F_SHARED when providing a srcmap to a write
xfs: fix perag structure refcounting error when scrub fails
xfs: rename buffer cache index variable b_bn
xfs: convert bp->b_bn references to xfs_buf_daddr()
xfs: introduce xfs_buf_daddr()
xfs: kill xfs_sb_version_has_v3inode()
xfs: introduce xfs_sb_is_v5 helper
xfs: remove unused xfs_sb_version_has wrappers
xfs: convert xfs_sb_version_has checks to use mount features
xfs: convert scrub to use mount-based feature checks
xfs: open code sb verifier feature checks
xfs: convert xfs_fs_geometry to use mount feature checks
xfs: replace XFS_FORCED_SHUTDOWN with xfs_is_shutdown
xfs: convert remaining mount flags to state flags
xfs: convert mount flags to features
xfs: consolidate mount option features in m_features
xfs: replace xfs_sb_version checks with feature flag checks
xfs: reflect sb features in xfs_mount
xfs: rework attr2 feature and mount options
...
Pull support for struct bio recycling from Jens Axboe:
"This adds bio recycling support for polled IO, allowing quick reuse of
a bio for high IOPS scenarios via a percpu bio_set list.
It's good for almost a 10% improvement in performance, bumping our
per-core IO limit from ~3.2M IOPS to ~3.5M IOPS"
* tag 'io_uring-bio-cache.5-2021-08-30' of git://git.kernel.dk/linux-block:
bio: improve kerneldoc documentation for bio_alloc_kiocb()
block: provide bio_clear_hipri() helper
block: use the percpu bio cache in __blkdev_direct_IO
io_uring: enable use of bio alloc cache
block: clear BIO_PERCPU_CACHE flag if polling isn't supported
bio: add allocation cache abstraction
fs: add kiocb alloc cache flag
bio: optimize initialization of a bio
Add a per-cpu bio_set cache for bio allocations, enabling us to quickly
recycle them instead of going through the slab allocator. This cache
isn't IRQ safe, and hence is only really suitable for polled IO.
Very simple - keeps a count of bio's in the cache, and maintains a max
of 512 with a slack of 64. If we get above max + slack, we drop slack
number of bio's.
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
We need to move to per-cpu state for both deferred inode
inactivation and CIL tracking, but to do that we
need to handle CPUs being removed from the system by the hot-plug
code. Introduce generic XFS infrastructure to handle CPU hotplug
events that is set up at module init time and torn down at module
exit time.
Initially, we only need CPU dead notifications, so we only set
up a callback for these notifications. The infrastructure can be
updated in future for other CPU hotplug state machine notifications
easily if ever needed.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
[djwong: rearrange some macros, fix function prototypes]
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Commit cf6acb8bdb ("s390/cpumf: Add support for complete counter set extraction")
allows access to the CPU Measurement Counter Facility via character
device /dev/hwctr. The access was exclusive via this device or
via perf_event_open() system call. Only one path at a time was
permitted. The CPU Measurement Counter Facility device driver blocked
access to other processes.
This patch removes this restriction and allows concurrent access to
the CPU Measurement Counter Facility from multiple processes at the same
time via perf_event_open() SVC and via /dev/hwctr device. The access
via /dev/hwctr device is still exclusive, only one process is allowed to
access this device.
This patch
- moves the /dev/hwctr device access from file perf_cpum_cf_diag.c.
to file perf_cpum_cf.c.
- use only one trace buffer .../s390dbf/cpum_cf.
- remove cfset_csd structure and includes its members it into the
structure cpu_cf_events. This results in one data structure and
simplifies the access.
- rework function familiy ctr_set_enable, ctr_set_disable, ctr_set_start
and ctr_set_stop which operate on a counter set number.
Now they operate on a counter set bit mask.
- move CF_DIAG event functionality to file perf_cpum_cf.c. It now
contains the complete functionality of the CPU Measurement Counter
Facility:
- Performance measurement support for counters using perf stat.
- Support for complete counter set extraction with device /dev/hwctr.
- Support for counter set extraction event CF_DIAG attached to
samples using perf record.
- removes file perf_cpum_cf_diag.c
Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Reviewed-by: Sumanth Korikkar <sumanthk@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
The PCP high watermark is based on the number of online CPUs so the
watermarks must be adjusted during CPU hotplug. At the time of
hot-remove, the number of online CPUs is already adjusted but during
hot-add, a delta needs to be applied to update PCP to the correct value.
After this patch is applied, the high watermarks are adjusted correctly.
# grep high: /proc/zoneinfo | tail -1
high: 649
# echo 0 > /sys/devices/system/cpu/cpu4/online
# grep high: /proc/zoneinfo | tail -1
high: 664
# echo 1 > /sys/devices/system/cpu/cpu4/online
# grep high: /proc/zoneinfo | tail -1
high: 649
Link: https://lkml.kernel.org/r/20210525080119.5455-4-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull dmaengine updates from Vinod Koul:
"New drivers/devices:
- Support for QCOM SM8150 GPI DMA
Updates:
- Big pile of idxd updates including support for performance
monitoring
- Support in dw-edma for interleaved dma
- Support for synchronize() in Xilinx driver"
* tag 'dmaengine-5.13-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/dmaengine: (42 commits)
dmaengine: idxd: Enable IDXD performance monitor support
dmaengine: idxd: Add IDXD performance monitor support
dmaengine: idxd: remove MSIX masking for interrupt handlers
dmaengine: idxd: device cmd should use dedicated lock
dmaengine: idxd: support reporting of halt interrupt
dmaengine: idxd: enable SVA feature for IOMMU
dmaengine: idxd: convert sprintf() to sysfs_emit() for all usages
dmaengine: idxd: add interrupt handle request and release support
dmaengine: idxd: add support for readonly config mode
dmaengine: idxd: add percpu_ref to descriptor submission path
dmaengine: idxd: remove detection of device type
dmaengine: idxd: iax bus removal
dmaengine: idxd: fix cdev setup and free device lifetime issues
dmaengine: idxd: fix group conf_dev lifetime
dmaengine: idxd: fix engine conf_dev lifetime
dmaengine: idxd: fix wq conf_dev 'struct device' lifetime
dmaengine: idxd: fix idxd conf_dev 'struct device' lifetime
dmaengine: idxd: use ida for device instance enumeration
dmaengine: idxd: removal of pcim managed mmio mapping
dmaengine: idxd: cleanup pci interrupt vector allocation management
...
Pull iommu updates from Joerg Roedel:
- Big cleanup of almost unsused parts of the IOMMU API by Christoph
Hellwig. This mostly affects the Freescale PAMU driver.
- New IOMMU driver for Unisoc SOCs
- ARM SMMU Updates from Will:
- Drop vestigial PREFETCH_ADDR support (SMMUv3)
- Elide TLB sync logic for empty gather (SMMUv3)
- Fix "Service Failure Mode" handling (SMMUv3)
- New Qualcomm compatible string (SMMUv2)
- Removal of the AMD IOMMU performance counter writeable check on AMD.
It caused long boot delays on some machines and is only needed to
work around an errata on some older (possibly pre-production) chips.
If someone is still hit by this hardware issue anyway the performance
counters will just return 0.
- Support for targeted invalidations in the AMD IOMMU driver. Before
that the driver only invalidated a single 4k page or the whole IO/TLB
for an address space. This has been extended now and is mostly useful
for emulated AMD IOMMUs.
- Several fixes for the Shared Virtual Memory support in the Intel VT-d
driver
- Mediatek drivers can now be built as modules
- Re-introduction of the forcedac boot option which got lost when
converting the Intel VT-d driver to the common dma-iommu
implementation.
- Extension of the IOMMU device registration interface and support
iommu_ops to be const again when drivers are built as modules.
* tag 'iommu-updates-v5.13' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu: (84 commits)
iommu: Streamline registration interface
iommu: Statically set module owner
iommu/mediatek-v1: Add error handle for mtk_iommu_probe
iommu/mediatek-v1: Avoid build fail when build as module
iommu/mediatek: Always enable the clk on resume
iommu/fsl-pamu: Fix uninitialized variable warning
iommu/vt-d: Force to flush iotlb before creating superpage
iommu/amd: Put newline after closing bracket in warning
iommu/vt-d: Fix an error handling path in 'intel_prepare_irq_remapping()'
iommu/vt-d: Fix build error of pasid_enable_wpe() with !X86
iommu/amd: Remove performance counter pre-initialization test
Revert "iommu/amd: Fix performance counter initialization"
iommu/amd: Remove duplicate check of devid
iommu/exynos: Remove unneeded local variable initialization
iommu/amd: Page-specific invalidations for more than one page
iommu/arm-smmu-v3: Remove the unused fields for PREFETCH_CONFIG command
iommu/vt-d: Avoid unnecessary cache flush in pasid entry teardown
iommu/vt-d: Invalidate PASID cache when root/context entry changed
iommu/vt-d: Remove WO permissions on second-level paging entries
iommu/vt-d: Report the right page fault address
...
Pull ARM Apple M1 platform support from Arnd Bergmann:
"The Apple M1 is the processor used it all current generation Apple
Macintosh computers. Support for this platform so far is rudimentary,
but it boots and can use framebuffer and serial console over a special
USB cable.
Support for several essential on-chip devices (USB, PCIe, IOMMU, NVMe)
is work in progress but was not ready in time.
A very detailed description of what works is in the commit message of
commit 1bb2fd3880 ("Merge tag 'm1-soc-bringup-v5' [..]") and on the
AsahiLinux wiki"
Link: https://lore.kernel.org/linux-arm-kernel/bdb18e9f-fcd7-1e31-2224-19c0e5090706@marcan.st/
* tag 'arm-apple-m1-5.13' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc:
asm-generic/io.h: Unbork ioremap_np() declaration
arm64: apple: Add initial Apple Mac mini (M1, 2020) devicetree
dt-bindings: display: Add apple,simple-framebuffer
arm64: Kconfig: Introduce CONFIG_ARCH_APPLE
irqchip/apple-aic: Add support for the Apple Interrupt Controller
dt-bindings: interrupt-controller: Add DT bindings for apple-aic
arm64: Move ICH_ sysreg bits from arm-gic-v3.h to sysreg.h
of/address: Add infrastructure to declare MMIO as non-posted
asm-generic/io.h: implement pci_remap_cfgspace using ioremap_np
arm64: Implement ioremap_np() to map MMIO as nGnRnE
docs: driver-api: device-io: Document ioremap() variants & access funcs
docs: driver-api: device-io: Document I/O access functions
asm-generic/io.h: Add a non-posted variant of ioremap()
arm64: arch_timer: Implement support for interrupt-names
dt-bindings: timer: arm,arch_timer: Add interrupt-names support
arm64: cputype: Add CPU implementor & types for the Apple M1 cores
dt-bindings: arm: cpus: Add apple,firestorm & icestorm compatibles
dt-bindings: arm: apple: Add bindings for Apple ARM platforms
dt-bindings: vendor-prefixes: Add apple prefix
Pull arm64 updates from Catalin Marinas:
- MTE asynchronous support for KASan. Previously only synchronous
(slower) mode was supported. Asynchronous is faster but does not
allow precise identification of the illegal access.
- Run kernel mode SIMD with softirqs disabled. This allows using NEON
in softirq context for crypto performance improvements. The
conditional yield support is modified to take softirqs into account
and reduce the latency.
- Preparatory patches for Apple M1: handle CPUs that only have the VHE
mode available (host kernel running at EL2), add FIQ support.
- arm64 perf updates: support for HiSilicon PA and SLLC PMU drivers,
new functions for the HiSilicon HHA and L3C PMU, cleanups.
- Re-introduce support for execute-only user permissions but only when
the EPAN (Enhanced Privileged Access Never) architecture feature is
available.
- Disable fine-grained traps at boot and improve the documented boot
requirements.
- Support CONFIG_KASAN_VMALLOC on arm64 (only with KASAN_GENERIC).
- Add hierarchical eXecute Never permissions for all page tables.
- Add arm64 prctl(PR_PAC_{SET,GET}_ENABLED_KEYS) allowing user programs
to control which PAC keys are enabled in a particular task.
- arm64 kselftests for BTI and some improvements to the MTE tests.
- Minor improvements to the compat vdso and sigpage.
- Miscellaneous cleanups.
* tag 'arm64-upstream' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux: (86 commits)
arm64/sve: Add compile time checks for SVE hooks in generic functions
arm64/kernel/probes: Use BUG_ON instead of if condition followed by BUG.
arm64: pac: Optimize kernel entry/exit key installation code paths
arm64: Introduce prctl(PR_PAC_{SET,GET}_ENABLED_KEYS)
arm64: mte: make the per-task SCTLR_EL1 field usable elsewhere
arm64/sve: Remove redundant system_supports_sve() tests
arm64: fpsimd: run kernel mode NEON with softirqs disabled
arm64: assembler: introduce wxN aliases for wN registers
arm64: assembler: remove conditional NEON yield macros
kasan, arm64: tests supports for HW_TAGS async mode
arm64: mte: Report async tag faults before suspend
arm64: mte: Enable async tag check fault
arm64: mte: Conditionally compile mte_enable_kernel_*()
arm64: mte: Enable TCO in functions that can read beyond buffer limits
kasan: Add report for async mode
arm64: mte: Drop arch_enable_tagging()
kasan: Add KASAN mode kernel parameter
arm64: mte: Add asynchronous mode support
arm64: Get rid of CONFIG_ARM64_VHE
arm64: Cope with CPUs stuck in VHE mode
...
Implement the IDXD performance monitor capability (named 'perfmon' in
the DSA (Data Streaming Accelerator) spec [1]), which supports the
collection of information about key events occurring during DSA and
IAX (Intel Analytics Accelerator) device execution, to assist in
performance tuning and debugging.
The idxd perfmon support is implemented as part of the IDXD driver and
interfaces with the Linux perf framework. It has several features in
common with the existing uncore pmu support:
- it does not support sampling
- does not support per-thread counting
However it also has some unique features not present in the core and
uncore support:
- all general-purpose counters are identical, thus no event constraints
- operation is always system-wide
While the core perf subsystem assumes that all counters are by default
per-cpu, the uncore pmus are socket-scoped and use a cpu mask to
restrict counting to one cpu from each socket. IDXD counters use a
similar strategy but expand the scope even further; since IDXD
counters are system-wide and can be read from any cpu, the IDXD perf
driver picks a single cpu to do the work (with cpu hotplug notifiers
to choose a different cpu if the chosen one is taken off-line).
More specifically, the perf userspace tool by default opens a counter
for each cpu for an event. However, if it finds a cpumask file
associated with the pmu under sysfs, as is the case with the uncore
pmus, it will open counters only on the cpus specified by the cpumask.
Since perfmon only needs to open a single counter per event for a
given IDXD device, the perfmon driver will create a sysfs cpumask file
for the device and insert the first cpu of the system into it. When a
user uses perf to open an event, perf will open a single counter on
the cpu specified by the cpu mask. This amounts to the default
system-wide rather than per-cpu counting mentioned previously for
perfmon pmu events. In order to keep the cpu mask up-to-date, the
driver implements cpu hotplug support for multiple devices, as IDXD
usually enumerates and registers more than one idxd device.
The perfmon driver implements basic perfmon hardware capability
discovery and configuration, and is initialized by the IDXD driver's
probe function. During initialization, the driver retrieves the total
number of supported performance counters, the pmu ID, and the device
type from idxd device, and registers itself under the Linux perf
framework.
The perf userspace tool can be used to monitor single or multiple
events depending on the given configuration, as well as event groups,
which are also supported by the perfmon driver. The user configures
events using the perf tool command-line interface by specifying the
event and corresponding event category, along with an optional set of
filters that can be used to restrict counting to specific work queues,
traffic classes, page and transfer sizes, and engines (See [1] for
specifics).
With the configuration specified by the user, the perf tool issues a
system call passing that information to the kernel, which uses it to
initialize the specified event(s). The event(s) are opened and
started, and following termination of the perf command, they're
stopped. At that point, the perfmon driver will read the latest count
for the event(s), calculate the difference between the latest counter
values and previously tracked counter values, and display the final
incremental count as the event count for the cycle. An overflow
handler registered on the IDXD irq path is used to account for counter
overflows, which are signaled by an overflow interrupt.
Below are a couple of examples of perf usage for monitoring DSA events.
The following monitors all events in the 'engine' category. Becuuse
no filters are specified, this captures all engine events for the
workload, which in this case is 19 iterations of the work generated by
the kernel dmatest module.
Details describing the events can be found in Appendix D of [1],
Performance Monitoring Events, but briefly they are:
event 0x1: total input data processed, in 32-byte units
event 0x2: total data written, in 32-byte units
event 0x4: number of work descriptors that read the source
event 0x8: number of work descriptors that write the destination
event 0x10: number of work descriptors dispatched from batch descriptors
event 0x20: number of work descriptors dispatched from work queues
# perf stat -e dsa0/event=0x1,event_category=0x1/,
dsa0/event=0x2,event_category=0x1/,
dsa0/event=0x4,event_category=0x1/,
dsa0/event=0x8,event_category=0x1/,
dsa0/event=0x10,event_category=0x1/,
dsa0/event=0x20,event_category=0x1/
modprobe dmatest channel=dma0chan0 timeout=2000
iterations=19 run=1 wait=1
Performance counter stats for 'system wide':
5,332 dsa0/event=0x1,event_category=0x1/
5,327 dsa0/event=0x2,event_category=0x1/
19 dsa0/event=0x4,event_category=0x1/
19 dsa0/event=0x8,event_category=0x1/
0 dsa0/event=0x10,event_category=0x1/
19 dsa0/event=0x20,event_category=0x1/
21.977436186 seconds time elapsed
The command below illustrates filter usage with a simple example. It
specifies that MEM_MOVE operations should be counted for the DSA
device dsa0 (event 0x8 corresponds to the EV_MEM_MOVE event - Number
of Memory Move Descriptors, which is part of event category 0x3 -
Operations. The detailed category and event IDs are available in
Appendix D, Performance Monitoring Events, of [1]). In addition to
the event and event category, a number of filters are also specified
(the detailed filter values are available in Chapter 6.4 (Filter
Support) of [1]), which will restrict counting to only those events
that meet all of the filter criteria. In this case, the filters
specify that only MEM_MOVE operations that are serviced by work queue
wq0 and specifically engine number engine0 and traffic class tc0
having sizes between 0 and 4k and page size of between 0 and 1G result
in a counter hit; anything else will be filtered out and not appear in
the final count. Note that filters are optional - any filter not
specified is assumed to be all ones and will pass anything.
# perf stat -e dsa0/filter_wq=0x1,filter_tc=0x1,filter_sz=0x7,
filter_eng=0x1,event=0x8,event_category=0x3/
modprobe dmatest channel=dma0chan0 timeout=2000
iterations=19 run=1 wait=1
Performance counter stats for 'system wide':
19 dsa0/filter_wq=0x1,filter_tc=0x1,filter_sz=0x7,
filter_eng=0x1,event=0x8,event_category=0x3/
21.865914091 seconds time elapsed
The output above reflects that the unspecified workload resulted in
the counting of 19 MEM_MOVE operation events that met the filter
criteria.
[1]: https://software.intel.com/content/www/us/en/develop/download/intel-data-streaming-accelerator-preliminary-architecture-specification.html
[ Based on work originally by Jing Lin. ]
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Reviewed-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Tom Zanussi <tom.zanussi@linux.intel.com>
Link: https://lore.kernel.org/r/0c5080a7d541904c4ad42b848c76a1ce056ddac7.1619276133.git.zanussi@kernel.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
There is a timer wrap issue on dra7 for the ARM architected timer.
In a typical clock configuration the timer fails to wrap after 388 days.
To work around the issue, we need to use timer-ti-dm percpu timers instead.
Let's configure dmtimer3 and 4 as percpu timers by default, and warn about
the issue if the dtb is not configured properly.
Let's do this as a single patch so it can be backported to v5.8 and later
kernels easily. Note that this patch depends on earlier timer-ti-dm
systimer posted mode fixes, and a preparatory clockevent patch
"clocksource/drivers/timer-ti-dm: Prepare to handle dra7 timer wrap issue".
For more information, please see the errata for "AM572x Sitara Processors
Silicon Revisions 1.1, 2.0":
https://www.ti.com/lit/er/sprz429m/sprz429m.pdf
The concept is based on earlier reference patches done by Tero Kristo and
Keerthy.
Cc: Keerthy <j-keerthy@ti.com>
Cc: Tero Kristo <kristo@kernel.org>
Signed-off-by: Tony Lindgren <tony@atomide.com>
Signed-off-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Link: https://lore.kernel.org/r/20210323074326.28302-3-tony@atomide.com
This is the root interrupt controller used on Apple ARM SoCs such as the
M1. This irqchip driver performs multiple functions:
* Handles both IRQs and FIQs
* Drives the AIC peripheral itself (which handles IRQs)
* Dispatches FIQs to downstream hard-wired clients (currently the ARM
timer).
* Implements a virtual IPI multiplexer to funnel multiple Linux IPIs
into a single hardware IPI
Reviewed-by: Marc Zyngier <maz@kernel.org>
Acked-by: Will Deacon <will@kernel.org>
Signed-off-by: Hector Martin <marcan@marcan.st>
On HiSilicon Hip09 platform, there is a PA (Protocol Adapter) module on
each chip SICL (Super I/O Cluster) which incorporates three Hydra interface
and facilitates the cache coherency between the dies on the chip. While PA
uncore PMU model is the same as other Hip09 PMU modules and many PMU events
are supported. Let's support the PMU driver using the HiSilicon uncore PMU
framework.
PA PMU supports the following filter functions:
* tracetag_en: allows user to count events according to tt_req or
tt_core set in L3C PMU. It's the same as other PMUs.
* srcid_cmd & srcid_msk: allows user to filter statistics that come from
specific CCL/ICL by configuration source ID.
* tgtid_cmd & tgtid_msk: it is the similar function to srcid_cmd &
srcid_msk. Both are used to check where the data comes from or go to.
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: John Garry <john.garry@huawei.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: John Garry <john.garry@huawei.com>
Co-developed-by: Qi Liu <liuqi115@huawei.com>
Signed-off-by: Qi Liu <liuqi115@huawei.com>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Link: https://lore.kernel.org/r/1615186237-22263-9-git-send-email-zhangshaokun@hisilicon.com
Signed-off-by: Will Deacon <will@kernel.org>
HiSilicon's Hip09 is comprised by multi-dies that can be connected by SLLC
module (Skyros Link Layer Controller), its has separate PMU registers which
the driver can program it freely and interrupt is supported to handle
counter overflow. Let's support its driver under the framework of HiSilicon
uncore PMU driver.
SLLC PMU supports the following filter functions:
* tracetag_en: allows user to count data according to tt_req or
tt_core set in L3C PMU.
* srcid_cmd & srcid_msk: allows user to filter statistics that come from
specific CCL/ICL by configuration source ID.
* tgtid_hi & tgtid_lo: it also supports event statistics that these
operations will go to the CCL/ICL by configuration target ID or
target ID range. It's the same as source ID with 11-bit width in
the SoC. More introduction is added in documentation:
Documentation/admin-guide/perf/hisi-pmu.rst
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: John Garry <john.garry@huawei.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Reviewed-by: John Garry <john.garry@huawei.com>
Co-developed-by: Qi Liu <liuqi115@huawei.com>
Signed-off-by: Qi Liu <liuqi115@huawei.com>
Signed-off-by: Shaokun Zhang <zhangshaokun@hisilicon.com>
Link: https://lore.kernel.org/r/1615186237-22263-8-git-send-email-zhangshaokun@hisilicon.com
Signed-off-by: Will Deacon <will@kernel.org>
Add support to the CPU Measurement counter facility device driver
to extract complete counter sets per CPU and per counter set from user
space. This includes a new device named /dev/hwctr and support
for the device driver functions open, close and ioctl. Other
functions are not supported.
The ioctl command supports 3 subcommands:
S390_HWCTR_START: enables counter sets on a list of CPUs.
S390_HWCTR_STOP: disables counter sets on a list of CPUs.
S390_HWCTR_READ: reads counter sets on a list of CPUs.
The ioctl(..., S390_HWCTR_READ, ...) is the only subcommand which
returns data. It requires member data_bytes to be positive and
indicates the maximum amount of data available to store counter set
data. The other ioctl() subcommands do not use this member and it
should be set to zero.
The S390_HWCTR_READ subcommand returns the following data:
The cpuset data is flattened using the following scheme, stored in member
data:
0x0 0x8 0xc 0x10 0x10 0x18 0x20 0x28 0xU-1
+---------+-----+---------+-----+---------+-----+-----+------+------+
| no_cpus | cpu | no_sets | set | no_cnts | cv1 | cv2 | .... | cv_n |
+---------+-----+---------+-----+---------+-----+-----+------+------+
0xU 0xU+4 0xU+8 0xU+10 0xV-1
+-----+---------+-----+-----+------+------+
| set | no_cnts | cv1 | cv2 | .... | cv_n |
+-----+---------+-----+-----+------+------+
0xV 0xV+4 0xV+8 0xV+c
+-----+---------+-----+---------+-----+-----+------+------+
| cpu | no_sets | set | no_cnts | cv1 | cv2 | .... | cv_n |
+-----+---------+-----+---------+-----+-----+------+------+
U and V denote arbitrary hexadezimal addresses.
The first integer represents the number of CPUs data was extracted
from. This is followed by CPU number and number of counter sets extracted.
Both are two integer values. This is followed by the set identifer
and number of counters extracted. Both are two integer values. This is
followed by the counter values, each element is eight bytes in size.
The S390_HWCTR_READ ioctl subcommand is also limited to one call per
minute. This ensures that an application does not read out the
counter sets too often and reduces the overall CPU performance.
The complete counter set extraction is an expensive operation.
Reviewed-by: Sumanth Korikkar <sumanthk@linux.ibm.com>
Signed-off-by: Thomas Richter <tmricht@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Current perf init will failed with:
[ 1.452433] csky-pmu: probe of soc:pmu failed with error -16
This patch fix it up with adding CPUHP_AP_PERF_CSKY_ONLINE in
cpuhotplug.h.
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
