Pull x86 asm updates from Ingo Molnar:
"This is another big update. Main changes are:
- lots of x86 system call (and other traps/exceptions) entry code
enhancements. In particular the complex parts of the 64-bit entry
code have been migrated to C code as well, and a number of dusty
corners have been refreshed. (Andy Lutomirski)
- vDSO special mapping robustification and general cleanups (Andy
Lutomirski)
- cpufeature refactoring, cleanups and speedups (Borislav Petkov)
- lots of other changes ..."
* 'x86-asm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (64 commits)
x86/cpufeature: Enable new AVX-512 features
x86/entry/traps: Show unhandled signal for i386 in do_trap()
x86/entry: Call enter_from_user_mode() with IRQs off
x86/entry/32: Change INT80 to be an interrupt gate
x86/entry: Improve system call entry comments
x86/entry: Remove TIF_SINGLESTEP entry work
x86/entry/32: Add and check a stack canary for the SYSENTER stack
x86/entry/32: Simplify and fix up the SYSENTER stack #DB/NMI fixup
x86/entry: Only allocate space for tss_struct::SYSENTER_stack if needed
x86/entry: Vastly simplify SYSENTER TF (single-step) handling
x86/entry/traps: Clear DR6 early in do_debug() and improve the comment
x86/entry/traps: Clear TIF_BLOCKSTEP on all debug exceptions
x86/entry/32: Restore FLAGS on SYSEXIT
x86/entry/32: Filter NT and speed up AC filtering in SYSENTER
x86/entry/compat: In SYSENTER, sink AC clearing below the existing FLAGS test
selftests/x86: In syscall_nt, test NT|TF as well
x86/asm-offsets: Remove PARAVIRT_enabled
x86/entry/32: Introduce and use X86_BUG_ESPFIX instead of paravirt_enabled
uprobes: __create_xol_area() must nullify xol_mapping.fault
x86/cpufeature: Create a new synthetic cpu capability for machine check recovery
...
Pull NOHZ updates from Ingo Molnar:
"NOHZ enhancements, by Frederic Weisbecker, which reorganizes/refactors
the NOHZ 'can the tick be stopped?' infrastructure and related code to
be data driven, and harmonizes the naming and handling of all the
various properties"
[ This makes the ugly "fetch_or()" macro that the scheduler used
internally a new generic helper, and does a bad job at it.
I'm pulling it, but I've asked Ingo and Frederic to get this
fixed up ]
* 'timers-nohz-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
sched-clock: Migrate to use new tick dependency mask model
posix-cpu-timers: Migrate to use new tick dependency mask model
sched: Migrate sched to use new tick dependency mask model
sched: Account rr tasks
perf: Migrate perf to use new tick dependency mask model
nohz: Use enum code for tick stop failure tracing message
nohz: New tick dependency mask
nohz: Implement wide kick on top of irq work
atomic: Export fetch_or()
Pull scheduler updates from Ingo Molnar:
"The main changes in this cycle are:
- Make schedstats a runtime tunable (disabled by default) and
optimize it via static keys.
As most distributions enable CONFIG_SCHEDSTATS=y due to its
instrumentation value, this is a nice performance enhancement.
(Mel Gorman)
- Implement 'simple waitqueues' (swait): these are just pure
waitqueues without any of the more complex features of full-blown
waitqueues (callbacks, wake flags, wake keys, etc.). Simple
waitqueues have less memory overhead and are faster.
Use simple waitqueues in the RCU code (in 4 different places) and
for handling KVM vCPU wakeups.
(Peter Zijlstra, Daniel Wagner, Thomas Gleixner, Paul Gortmaker,
Marcelo Tosatti)
- sched/numa enhancements (Rik van Riel)
- NOHZ performance enhancements (Rik van Riel)
- Various sched/deadline enhancements (Steven Rostedt)
- Various fixes (Peter Zijlstra)
- ... and a number of other fixes, cleanups and smaller enhancements"
* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (29 commits)
sched/cputime: Fix steal_account_process_tick() to always return jiffies
sched/deadline: Remove dl_new from struct sched_dl_entity
Revert "kbuild: Add option to turn incompatible pointer check into error"
sched/deadline: Remove superfluous call to switched_to_dl()
sched/debug: Fix preempt_disable_ip recording for preempt_disable()
sched, time: Switch VIRT_CPU_ACCOUNTING_GEN to jiffy granularity
time, acct: Drop irq save & restore from __acct_update_integrals()
acct, time: Change indentation in __acct_update_integrals()
sched, time: Remove non-power-of-two divides from __acct_update_integrals()
sched/rt: Kick RT bandwidth timer immediately on start up
sched/debug: Add deadline scheduler bandwidth ratio to /proc/sched_debug
sched/debug: Move sched_domain_sysctl to debug.c
sched/debug: Move the /sys/kernel/debug/sched_features file setup into debug.c
sched/rt: Fix PI handling vs. sched_setscheduler()
sched/core: Remove duplicated sched_group_set_shares() prototype
sched/fair: Consolidate nohz CPU load update code
sched/fair: Avoid using decay_load_missed() with a negative value
sched/deadline: Always calculate end of period on sched_yield()
sched/cgroup: Fix cgroup entity load tracking tear-down
rcu: Use simple wait queues where possible in rcutree
...
Pull perf updates from Ingo Molnar:
"Main kernel side changes:
- Big reorganization of the x86 perf support code. The old code grew
organically deep inside arch/x86/kernel/cpu/perf* and its naming
became somewhat messy.
The new location is under arch/x86/events/, using the following
cleaner hierarchy of source code files:
perf/x86: Move perf_event.c .................. => x86/events/core.c
perf/x86: Move perf_event_amd.c .............. => x86/events/amd/core.c
perf/x86: Move perf_event_amd_ibs.c .......... => x86/events/amd/ibs.c
perf/x86: Move perf_event_amd_iommu.[ch] ..... => x86/events/amd/iommu.[ch]
perf/x86: Move perf_event_amd_uncore.c ....... => x86/events/amd/uncore.c
perf/x86: Move perf_event_intel_bts.c ........ => x86/events/intel/bts.c
perf/x86: Move perf_event_intel.c ............ => x86/events/intel/core.c
perf/x86: Move perf_event_intel_cqm.c ........ => x86/events/intel/cqm.c
perf/x86: Move perf_event_intel_cstate.c ..... => x86/events/intel/cstate.c
perf/x86: Move perf_event_intel_ds.c ......... => x86/events/intel/ds.c
perf/x86: Move perf_event_intel_lbr.c ........ => x86/events/intel/lbr.c
perf/x86: Move perf_event_intel_pt.[ch] ...... => x86/events/intel/pt.[ch]
perf/x86: Move perf_event_intel_rapl.c ....... => x86/events/intel/rapl.c
perf/x86: Move perf_event_intel_uncore.[ch] .. => x86/events/intel/uncore.[ch]
perf/x86: Move perf_event_intel_uncore_nhmex.c => x86/events/intel/uncore_nmhex.c
perf/x86: Move perf_event_intel_uncore_snb.c => x86/events/intel/uncore_snb.c
perf/x86: Move perf_event_intel_uncore_snbep.c => x86/events/intel/uncore_snbep.c
perf/x86: Move perf_event_knc.c .............. => x86/events/intel/knc.c
perf/x86: Move perf_event_p4.c ............... => x86/events/intel/p4.c
perf/x86: Move perf_event_p6.c ............... => x86/events/intel/p6.c
perf/x86: Move perf_event_msr.c .............. => x86/events/msr.c
(Borislav Petkov)
- Update various x86 PMU constraint and hw support details (Stephane
Eranian)
- Optimize kprobes for BPF execution (Martin KaFai Lau)
- Rewrite, refactor and fix the Intel uncore PMU driver code (Thomas
Gleixner)
- Rewrite, refactor and fix the Intel RAPL PMU code (Thomas Gleixner)
- Various fixes and smaller cleanups.
There are lots of perf tooling updates as well. A few highlights:
perf report/top:
- Hierarchy histogram mode for 'perf top' and 'perf report',
showing multiple levels, one per --sort entry: (Namhyung Kim)
On a mostly idle system:
# perf top --hierarchy -s comm,dso
Then expand some levels and use 'P' to take a snapshot:
# cat perf.hist.0
- 92.32% perf
58.20% perf
22.29% libc-2.22.so
5.97% [kernel]
4.18% libelf-0.165.so
1.69% [unknown]
- 4.71% qemu-system-x86
3.10% [kernel]
1.60% qemu-system-x86_64 (deleted)
+ 2.97% swapper
#
- Add 'L' hotkey to dynamicly set the percent threshold for
histogram entries and callchains, i.e. dynamicly do what the
--percent-limit command line option to 'top' and 'report' does.
(Namhyung Kim)
perf mem:
- Allow specifying events via -e in 'perf mem record', also listing
what events can be specified via 'perf mem record -e list' (Jiri
Olsa)
perf record:
- Add 'perf record' --all-user/--all-kernel options, so that one
can tell that all the events in the command line should be
restricted to the user or kernel levels (Jiri Olsa), i.e.:
perf record -e cycles:u,instructions:u
is equivalent to:
perf record --all-user -e cycles,instructions
- Make 'perf record' collect CPU cache info in the perf.data file header:
$ perf record usleep 1
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.017 MB perf.data (7 samples) ]
$ perf report --header-only -I | tail -10 | head -8
# CPU cache info:
# L1 Data 32K [0-1]
# L1 Instruction 32K [0-1]
# L1 Data 32K [2-3]
# L1 Instruction 32K [2-3]
# L2 Unified 256K [0-1]
# L2 Unified 256K [2-3]
# L3 Unified 4096K [0-3]
Will be used in 'perf c2c' and eventually in 'perf diff' to
allow, for instance running the same workload in multiple
machines and then when using 'diff' show the hardware difference.
(Jiri Olsa)
- Improved support for Java, using the JVMTI agent library to do
jitdumps that then will be inserted in synthesized
PERF_RECORD_MMAP2 events via 'perf inject' pointed to synthesized
ELF files stored in ~/.debug and keyed with build-ids, to allow
symbol resolution and even annotation with source line info, see
the changeset comments to see how to use it (Stephane Eranian)
perf script/trace:
- Decode data_src values (e.g. perf.data files generated by 'perf
mem record') in 'perf script': (Jiri Olsa)
# perf script
perf 693 [1] 4.088652: 1 cpu/mem-loads,ldlat=30/P: ffff88007d0b0f40 68100142 L1 hit|SNP None|TLB L1 or L2 hit|LCK No <SNIP>
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
- Improve support to 'data_src', 'weight' and 'addr' fields in
'perf script' (Jiri Olsa)
- Handle empty print fmts in 'perf script -s' i.e. when running
python or perl scripts (Taeung Song)
perf stat:
- 'perf stat' now shows shadow metrics (insn per cycle, etc) in
interval mode too. E.g:
# perf stat -I 1000 -e instructions,cycles sleep 1
# time counts unit events
1.000215928 519,620 instructions # 0.69 insn per cycle
1.000215928 752,003 cycles
<SNIP>
- Port 'perf kvm stat' to PowerPC (Hemant Kumar)
- Implement CSV metrics output in 'perf stat' (Andi Kleen)
perf BPF support:
- Support converting data from bpf events in 'perf data' (Wang Nan)
- Print bpf-output events in 'perf script': (Wang Nan).
# perf record -e bpf-output/no-inherit,name=evt/ -e ./test_bpf_output_3.c/map:channel.event=evt/ usleep 1000
# perf script
usleep 4882 21384.532523: evt: ffffffff810e97d1 sys_nanosleep ([kernel.kallsyms])
BPF output: 0000: 52 61 69 73 65 20 61 20 Raise a
0008: 42 50 46 20 65 76 65 6e BPF even
0010: 74 21 00 00 t!..
BPF string: "Raise a BPF event!"
#
- Add API to set values of map entries in a BPF object, be it
individual map slots or ranges (Wang Nan)
- Introduce support for the 'bpf-output' event (Wang Nan)
- Add glue to read perf events in a BPF program (Wang Nan)
- Improve support for bpf-output events in 'perf trace' (Wang Nan)
... and tons of other changes as well - see the shortlog and git log
for details!"
* 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (342 commits)
perf stat: Add --metric-only support for -A
perf stat: Implement --metric-only mode
perf stat: Document CSV format in manpage
perf hists browser: Check sort keys before hot key actions
perf hists browser: Allow thread filtering for comm sort key
perf tools: Add sort__has_comm variable
perf tools: Recalc total periods using top-level entries in hierarchy
perf tools: Remove nr_sort_keys field
perf hists browser: Cleanup hist_browser__fprintf_hierarchy_entry()
perf tools: Remove hist_entry->fmt field
perf tools: Fix command line filters in hierarchy mode
perf tools: Add more sort entry check functions
perf tools: Fix hist_entry__filter() for hierarchy
perf jitdump: Build only on supported archs
tools lib traceevent: Add '~' operation within arg_num_eval()
perf tools: Omit unnecessary cast in perf_pmu__parse_scale
perf tools: Pass perf_hpp_list all the way through setup_sort_list
perf tools: Fix perf script python database export crash
perf jitdump: DWARF is also needed
perf bench mem: Prepare the x86-64 build for upstream memcpy_mcsafe() changes
...
Pull read-only kernel memory updates from Ingo Molnar:
"This tree adds two (security related) enhancements to the kernel's
handling of read-only kernel memory:
- extend read-only kernel memory to a new class of formerly writable
kernel data: 'post-init read-only memory' via the __ro_after_init
attribute, and mark the ARM and x86 vDSO as such read-only memory.
This kind of attribute can be used for data that requires a once
per bootup initialization sequence, but is otherwise never modified
after that point.
This feature was based on the work by PaX Team and Brad Spengler.
(by Kees Cook, the ARM vDSO bits by David Brown.)
- make CONFIG_DEBUG_RODATA always enabled on x86 and remove the
Kconfig option. This simplifies the kernel and also signals that
read-only memory is the default model and a first-class citizen.
(Kees Cook)"
* 'mm-readonly-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
ARM/vdso: Mark the vDSO code read-only after init
x86/vdso: Mark the vDSO code read-only after init
lkdtm: Verify that '__ro_after_init' works correctly
arch: Introduce post-init read-only memory
x86/mm: Always enable CONFIG_DEBUG_RODATA and remove the Kconfig option
mm/init: Add 'rodata=off' boot cmdline parameter to disable read-only kernel mappings
asm-generic: Consolidate mark_rodata_ro()
Pull locking changes from Ingo Molnar:
"Various updates:
- Futex scalability improvements: remove page lock use for shared
futex get_futex_key(), which speeds up 'perf bench futex hash'
benchmarks by over 40% on a 60-core Westmere. This makes anon-mem
shared futexes perform close to private futexes. (Mel Gorman)
- lockdep hash collision detection and fix (Alfredo Alvarez
Fernandez)
- lockdep testing enhancements (Alfredo Alvarez Fernandez)
- robustify lockdep init by using hlists (Andrew Morton, Andrey
Ryabinin)
- mutex and csd_lock micro-optimizations (Davidlohr Bueso)
- small x86 barriers tweaks (Michael S Tsirkin)
- qspinlock updates (Waiman Long)"
* 'locking-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (25 commits)
locking/csd_lock: Use smp_cond_acquire() in csd_lock_wait()
locking/csd_lock: Explicitly inline csd_lock*() helpers
futex: Replace barrier() in unqueue_me() with READ_ONCE()
locking/lockdep: Detect chain_key collisions
locking/lockdep: Prevent chain_key collisions
tools/lib/lockdep: Fix link creation warning
tools/lib/lockdep: Add tests for AA and ABBA locking
tools/lib/lockdep: Add userspace version of READ_ONCE()
tools/lib/lockdep: Fix the build on recent kernels
locking/qspinlock: Move __ARCH_SPIN_LOCK_UNLOCKED to qspinlock_types.h
locking/mutex: Allow next waiter lockless wakeup
locking/pvqspinlock: Enable slowpath locking count tracking
locking/qspinlock: Use smp_cond_acquire() in pending code
locking/pvqspinlock: Move lock stealing count tracking code into pv_queued_spin_steal_lock()
locking/mcs: Fix mcs_spin_lock() ordering
futex: Remove requirement for lock_page() in get_futex_key()
futex: Rename barrier references in ordering guarantees
locking/atomics: Update comment about READ_ONCE() and structures
locking/lockdep: Eliminate lockdep_init()
locking/lockdep: Convert hash tables to hlists
...
Pull ram resource handling changes from Ingo Molnar:
"Core kernel resource handling changes to support NVDIMM error
injection.
This tree introduces a new I/O resource type, IORESOURCE_SYSTEM_RAM,
for System RAM while keeping the current IORESOURCE_MEM type bit set
for all memory-mapped ranges (including System RAM) for backward
compatibility.
With this resource flag it no longer takes a strcmp() loop through the
resource tree to find "System RAM" resources.
The new resource type is then used to extend ACPI/APEI error injection
facility to also support NVDIMM"
* 'core-resources-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
ACPI/EINJ: Allow memory error injection to NVDIMM
resource: Kill walk_iomem_res()
x86/kexec: Remove walk_iomem_res() call with GART type
x86, kexec, nvdimm: Use walk_iomem_res_desc() for iomem search
resource: Add walk_iomem_res_desc()
memremap: Change region_intersects() to take @flags and @desc
arm/samsung: Change s3c_pm_run_res() to use System RAM type
resource: Change walk_system_ram() to use System RAM type
drivers: Initialize resource entry to zero
xen, mm: Set IORESOURCE_SYSTEM_RAM to System RAM
kexec: Set IORESOURCE_SYSTEM_RAM for System RAM
arch: Set IORESOURCE_SYSTEM_RAM flag for System RAM
ia64: Set System RAM type and descriptor
x86/e820: Set System RAM type and descriptor
resource: Add I/O resource descriptor
resource: Handle resource flags properly
resource: Add System RAM resource type
We can micro-optimize this call and mildly relax the
barrier requirements by relying on ctrl + rmb, keeping
the acquire semantics. In addition, this is pretty much
the now standard for busy-waiting under such restraints.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: dave@stgolabs.net
Link: http://lkml.kernel.org/r/1457574936-19065-3-git-send-email-dbueso@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
While the compiler tends to already to it for us (except for
csd_unlock), make it explicit. These helpers mainly deal with
the ->flags, are short-lived and can be called, for example,
from smp_call_function_many().
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: dave@stgolabs.net
Link: http://lkml.kernel.org/r/1457574936-19065-2-git-send-email-dbueso@suse.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
In memremap's helper function try_ram_remap(), we dereference a struct
page pointer that was derived from a PFN that is known to be covered by
a 'System RAM' iomem region, and is thus assumed to be a 'valid' PFN,
i.e., a PFN that has a struct page associated with it and is covered by
the kernel direct mapping.
However, the assumption that there is a 1:1 relation between the System
RAM iomem region and the kernel direct mapping is not universally valid
on all architectures, and on ARM and arm64, 'System RAM' may include
regions for which pfn_valid() returns false.
Generally speaking, both __va() and pfn_to_page() should only ever be
called on PFNs/physical addresses for which pfn_valid() returns true, so
add that check to try_ram_remap().
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Functions which the compiler has instrumented for KASAN place poison on
the stack shadow upon entry and remove this poision prior to returning.
In the case of CPU hotplug, CPUs exit the kernel a number of levels deep
in C code. Any instrumented functions on this critical path will leave
portions of the stack shadow poisoned.
When a CPU is subsequently brought back into the kernel via a different
path, depending on stackframe, layout calls to instrumented functions
may hit this stale poison, resulting in (spurious) KASAN splats to the
console.
To avoid this, clear any stale poison from the idle thread for a CPU
prior to bringing a CPU online.
Signed-off-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Reviewed-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Alexander Potapenko <glider@google.com>
Cc: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The check for whether we overlap "System RAM" needs to be done at
section granularity. For example a system with the following mapping:
100000000-37bffffff : System RAM
37c000000-837ffffff : Persistent Memory
...is unable to use devm_memremap_pages() as it would result in two
zones colliding within a given section.
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Reviewed-by: Toshi Kani <toshi.kani@hpe.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Given we have uninitialized list_heads being passed to list_add() it
will always be the case that those uninitialized values randomly trigger
the poison value. Especially since a list_add() operation will seed the
stack with the poison value for later stack allocations to trip over.
For example, see these two false positive reports:
list_add attempted on force-poisoned entry
WARNING: at lib/list_debug.c:34
[..]
NIP [c00000000043c390] __list_add+0xb0/0x150
LR [c00000000043c38c] __list_add+0xac/0x150
Call Trace:
__list_add+0xac/0x150 (unreliable)
__down+0x4c/0xf8
down+0x68/0x70
xfs_buf_lock+0x4c/0x150 [xfs]
list_add attempted on force-poisoned entry(0000000000000500),
new->next == d0000000059ecdb0, new->prev == 0000000000000500
WARNING: at lib/list_debug.c:33
[..]
NIP [c00000000042db78] __list_add+0xa8/0x140
LR [c00000000042db74] __list_add+0xa4/0x140
Call Trace:
__list_add+0xa4/0x140 (unreliable)
rwsem_down_read_failed+0x6c/0x1a0
down_read+0x58/0x60
xfs_log_commit_cil+0x7c/0x600 [xfs]
Fixes: commit 5c2c2587b1 ("mm, dax, pmem: introduce {get|put}_dev_pagemap() for dax-gup")
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reported-by: Eryu Guan <eguan@redhat.com>
Tested-by: Eryu Guan <eguan@redhat.com>
Cc: Ross Zwisler <ross.zwisler@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit e91467ecd1 ("bug in futex unqueue_me") introduced a barrier() in
unqueue_me() to prevent the compiler from rereading the lock pointer which
might change after a check for NULL.
Replace the barrier() with a READ_ONCE() for the following reasons:
1) READ_ONCE() is a weaker form of barrier() that affects only the specific
load operation, while barrier() is a general compiler level memory barrier.
READ_ONCE() was not available at the time when the barrier was added.
2) Aside of that READ_ONCE() is descriptive and self explainatory while a
barrier without comment is not clear to the casual reader.
No functional change.
[ tglx: Massaged changelog ]
Signed-off-by: Jianyu Zhan <nasa4836@gmail.com>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Darren Hart <dvhart@linux.intel.com>
Cc: dave@stgolabs.net
Cc: peterz@infradead.org
Cc: linux@rasmusvillemoes.dk
Cc: akpm@linux-foundation.org
Cc: fengguang.wu@intel.com
Cc: bigeasy@linutronix.de
Link: http://lkml.kernel.org/r/1457314344-5685-1-git-send-email-nasa4836@gmail.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Pull nohz enhancements from Frederic Weisbecker:
"Currently in nohz full configs, the tick dependency is checked
asynchronously by nohz code from interrupt and context switch for each
concerned subsystem with a set of function provided by these. Such
functions are made of many conditions and details that can be heavyweight
as they are called on fastpath: sched_can_stop_tick(),
posix_cpu_timer_can_stop_tick(), perf_event_can_stop_tick()...
Thomas suggested a few months ago to make that tick dependency check
synchronous. Instead of checking subsystems details from each interrupt
to guess if the tick can be stopped, every subsystem that may have a tick
dependency should set itself a flag specifying the state of that
dependency. This way we can verify if we can stop the tick with a single
lightweight mask check on fast path.
This conversion from a pull to a push model to implement tick dependency
is the core feature of this patchset that is split into:
* Nohz wide kick simplification
* Improve nohz tracing
* Introduce tick dependency mask
* Migrate scheduler, posix timers, perf events and sched clock tick
dependencies to the tick dependency mask."
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The callers of steal_account_process_tick() expect it to return
whether a jiffy should be considered stolen or not.
Currently the return value of steal_account_process_tick() is in
units of cputime, which vary between either jiffies or nsecs
depending on CONFIG_VIRT_CPU_ACCOUNTING_GEN.
If cputime has nsecs granularity and there is a tiny amount of
stolen time (a few nsecs, say) then we will consider the entire
tick stolen and will not account the tick on user/system/idle,
causing /proc/stats to show invalid data.
The fix is to change steal_account_process_tick() to accumulate
the stolen time and only account it once it's worth a jiffy.
(Thanks to Frederic Weisbecker for suggestions to fix a bug in my
first version of the patch.)
Signed-off-by: Chris Friesen <chris.friesen@windriver.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: <stable@vger.kernel.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/56DBBDB8.40305@mail.usask.ca
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The dl_new field of struct sched_dl_entity is currently used to
identify new deadline tasks, so that their deadline and runtime
can be properly initialised.
However, these tasks can be easily identified by checking if
their deadline is smaller than the current time when they switch
to SCHED_DEADLINE. So, dl_new can be removed by introducing this
check in switched_to_dl(); this allows to simplify the
SCHED_DEADLINE code.
Signed-off-by: Luca Abeni <luca.abeni@unitn.it>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Juri Lelli <juri.lelli@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1457350024-7825-2-git-send-email-luca.abeni@unitn.it
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The error path in perf_event_open() is such that asking for a sampling
event on a PMU that doesn't generate interrupts will end up in dropping
the perf_sched_count even though it hasn't been incremented for this
event yet.
Given a sufficient amount of these calls, we'll end up disabling
scheduler's jump label even though we'd still have active events in the
system, thereby facilitating the arrival of the infernal regions upon us.
I'm fixing this by moving account_event() inside perf_event_alloc().
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <stable@vger.kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vince Weaver <vincent.weaver@maine.edu>
Cc: vince@deater.net
Link: http://lkml.kernel.org/r/1456917854-29427-1-git-send-email-alexander.shishkin@linux.intel.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
a tasks "comm" field. But this prevented filtering on a comm field that
is within a trace event (like sched_migrate_task).
When trying to filter on when a program migrated, this change prevented
the filtering of the sched_migrate_task.
To fix this, the event fields are examined first, and then the extra fields
like "comm" and "cpu" are examined. Also, instead of testing to assign
the comm filter function based on the field's name, the generic comm field
is given a new filter type (FILTER_COMM). When this field is used to filter
the type is checked. The same is done for the cpu filter field.
Two new special filter types are added: "COMM" and "CPU". This allows users
to still filter the tasks comm for events that have "comm" as one of their
fields, in cases that users would like to filter sched_migrate_task on the
comm of the task that called the event, and not the comm of the task that
is being migrated.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQEcBAABAgAGBQJW2argAAoJEKKk/i67LK/8b78H/32nYPizDIsK/p2bL1mgbtMl
vrkcfb+maPOC7cjB+CdQmyV4EIVpSn06XFouYghGprdoVocVyBuIflxn0j3Gbymy
zLCg8lR70KTATTqst1wsWMbnh+UvAKNEiXj8jf2qcK2xhgalXMDwsTC4+LDlLugu
YAx89lmsjK1YpP/wIzMww2jQG+07Nhm9gHWXF2MC3egZ+sgYxARnfds0yTcGgS8o
dc/yJGZDCI44JMDNThcCFxNvsmoTa9tpm+JNe2YTht6KCympa+Ht9Jj9MMlD06cq
M5CqMQlok+mrVsW5LbJPCk1u83ynr6d/PcPQuT2nykRx8bGvKjA7AKMPaxw1Jz4=
=ixBz
-----END PGP SIGNATURE-----
Merge tag 'trace-fixes-v4.5-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace
Pull tracing fix from Steven Rostedt:
"A feature was added in 4.3 that allowed users to filter trace points
on a tasks "comm" field. But this prevented filtering on a comm field
that is within a trace event (like sched_migrate_task).
When trying to filter on when a program migrated, this change
prevented the filtering of the sched_migrate_task.
To fix this, the event fields are examined first, and then the extra
fields like "comm" and "cpu" are examined. Also, instead of testing
to assign the comm filter function based on the field's name, the
generic comm field is given a new filter type (FILTER_COMM). When
this field is used to filter the type is checked. The same is done
for the cpu filter field.
Two new special filter types are added: "COMM" and "CPU". This allows
users to still filter the tasks comm for events that have "comm" as
one of their fields, in cases that users would like to filter
sched_migrate_task on the comm of the task that called the event, and
not the comm of the task that is being migrated"
* tag 'trace-fixes-v4.5-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing: Do not have 'comm' filter override event 'comm' field
Commit 9f61668073 "tracing: Allow triggers to filter for CPU ids and
process names" added a 'comm' filter that will filter events based on the
current tasks struct 'comm'. But this now hides the ability to filter events
that have a 'comm' field too. For example, sched_migrate_task trace event.
That has a 'comm' field of the task to be migrated.
echo 'comm == "bash"' > events/sched_migrate_task/filter
will now filter all sched_migrate_task events for tasks named "bash" that
migrates other tasks (in interrupt context), instead of seeing when "bash"
itself gets migrated.
This fix requires a couple of changes.
1) Change the look up order for filter predicates to look at the events
fields before looking at the generic filters.
2) Instead of basing the filter function off of the "comm" name, have the
generic "comm" filter have its own filter_type (FILTER_COMM). Test
against the type instead of the name to assign the filter function.
3) Add a new "COMM" filter that works just like "comm" but will filter based
on the current task, even if the trace event contains a "comm" field.
Do the same for "cpu" field, adding a FILTER_CPU and a filter "CPU".
Cc: stable@vger.kernel.org # v4.3+
Fixes: 9f61668073 "tracing: Allow triggers to filter for CPU ids and process names"
Reported-by: Matt Fleming <matt@codeblueprint.co.uk>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Instead of checking sched_clock_stable from the nohz subsystem to verify
its tick dependency, migrate it to the new mask in order to include it
to the all-in-one check.
Reviewed-by: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Instead of providing asynchronous checks for the nohz subsystem to verify
posix cpu timers tick dependency, migrate the latter to the new mask.
In order to keep track of the running timers and expose the tick
dependency accordingly, we must probe the timers queuing and dequeuing
on threads and process lists.
Unfortunately it implies both task and signal level dependencies. We
should be able to further optimize this and merge all that on the task
level dependency, at the cost of a bit of complexity and may be overhead.
Reviewed-by: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Instead of providing asynchronous checks for the nohz subsystem to verify
sched tick dependency, migrate sched to the new mask.
Everytime a task is enqueued or dequeued, we evaluate the state of the
tick dependency on top of the policy of the tasks in the runqueue, by
order of priority:
SCHED_DEADLINE: Need the tick in order to periodically check for runtime
SCHED_FIFO : Don't need the tick (no round-robin)
SCHED_RR : Need the tick if more than 1 task of the same priority
for round robin (simplified with checking if more than
one SCHED_RR task no matter what priority).
SCHED_NORMAL : Need the tick if more than 1 task for round-robin.
We could optimize that further with one flag per sched policy on the tick
dependency mask and perform only the checks relevant to the policy
concerned by an enqueue/dequeue operation.
Since the checks aren't based on the current task anymore, we could get
rid of the task switch hook but it's still needed for posix cpu
timers.
Reviewed-by: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
In order to evaluate the scheduler tick dependency without probing
context switches, we need to know how much SCHED_RR and SCHED_FIFO tasks
are enqueued as those policies don't have the same preemption
requirements.
To prepare for that, let's account SCHED_RR tasks, we'll be able to
deduce SCHED_FIFO tasks as well from it and the total RT tasks in the
runqueue.
Reviewed-by: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Instead of providing asynchronous checks for the nohz subsystem to verify
perf event tick dependency, migrate perf to the new mask.
Perf needs the tick for two situations:
1) Freq events. We could set the tick dependency when those are
installed on a CPU context. But setting a global dependency on top of
the global freq events accounting is much easier. If people want that
to be optimized, we can still refine that on the per-CPU tick dependency
level. This patch dooesn't change the current behaviour anyway.
2) Throttled events: this is a per-cpu dependency.
Reviewed-by: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
The tick dependency is evaluated on every IRQ and context switch. This
consists is a batch of checks which determine whether it is safe to
stop the tick or not. These checks are often split in many details:
posix cpu timers, scheduler, sched clock, perf events.... each of which
are made of smaller details: posix cpu timer involves checking process
wide timers then thread wide timers. Perf involves checking freq events
then more per cpu details.
Checking these informations asynchronously every time we update the full
dynticks state bring avoidable overhead and a messy layout.
Let's introduce instead tick dependency masks: one for system wide
dependency (unstable sched clock, freq based perf events), one for CPU
wide dependency (sched, throttling perf events), and task/signal level
dependencies (posix cpu timers). The subsystems are responsible
for setting and clearing their dependency through a set of APIs that will
take care of concurrent dependency mask modifications and kick targets
to restart the relevant CPU tick whenever needed.
This new dependency engine stays beside the old one until all subsystems
having a tick dependency are converted to it.
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Reviewed-by: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Chris Metcalf <cmetcalf@ezchip.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Luiz Capitulino <lcapitulino@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Some tracepoint have multiple fields with the same name, "nr", the first
one is a unique syscall ID, the other is a syscall argument:
# cat /sys/kernel/debug/tracing/events/syscalls/sys_enter_io_getevents/format
name: sys_enter_io_getevents
ID: 747
format:
field:unsigned short common_type; offset:0; size:2; signed:0;
field:unsigned char common_flags; offset:2; size:1; signed:0;
field:unsigned char common_preempt_count; offset:3; size:1; signed:0;
field:int common_pid; offset:4; size:4; signed:1;
field:int nr; offset:8; size:4; signed:1;
field:aio_context_t ctx_id; offset:16; size:8; signed:0;
field:long min_nr; offset:24; size:8; signed:0;
field:long nr; offset:32; size:8; signed:0;
field:struct io_event * events; offset:40; size:8; signed:0;
field:struct timespec * timeout; offset:48; size:8; signed:0;
print fmt: "ctx_id: 0x%08lx, min_nr: 0x%08lx, nr: 0x%08lx, events: 0x%08lx, timeout: 0x%08lx", ((unsigned long)(REC->ctx_id)), ((unsigned long)(REC->min_nr)), ((unsigned long)(REC->nr)), ((unsigned long)(REC->events)), ((unsigned long)(REC->timeout))
#
Fix it by renaming the "/format" common tracepoint field "nr" to "__syscall_nr".
Signed-off-by: Taeung Song <treeze.taeung@gmail.com>
[ Do not rename the struct member, just the '/format' field name ]
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Lai Jiangshan <jiangshanlai@gmail.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20160226132301.3ae065a4@gandalf.local.home
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Add detection for chain_key collision under CONFIG_DEBUG_LOCKDEP.
When a collision is detected the problem is reported and all lock
debugging is turned off.
Tested using liblockdep and the added tests before and after
applying the fix, confirming both that the code added for the
detection correctly reports the problem and that the fix actually
fixes it.
Tested tweaking lockdep to generate false collisions and
verified that the problem is reported and that lock debugging is
turned off.
Also tested with lockdep's test suite after applying the patch:
[ 0.000000] Good, all 253 testcases passed! |
Signed-off-by: Alfredo Alvarez Fernandez <alfredoalvarezernandez@gmail.com>
Cc: Alfredo Alvarez Fernandez <alfredoalvarezfernandez@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: sasha.levin@oracle.com
Link: http://lkml.kernel.org/r/1455864533-7536-4-git-send-email-alfredoalvarezernandez@gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The chain_key hashing macro iterate_chain_key(key1, key2) does not
generate a new different value if both key1 and key2 are 0. In that
case the generated value is again 0. This can lead to collisions which
can result in lockdep not detecting deadlocks or circular
dependencies.
Avoid the problem by using class_idx (1-based) instead of class id
(0-based) as an input for the hashing macro 'key2' in
iterate_chain_key(key1, key2).
The use of class id created collisions in cases like the following:
1.- Consider an initial state in which no class has been acquired yet.
Under these circumstances an AA deadlock will not be detected by
lockdep:
lock [key1,key2]->new key (key1=old chain_key, key2=id)
--------------------------
A [0,0]->0
A [0,0]->0 (collision)
The newly generated chain_key collides with the one used before and as
a result the check for a deadlock is skipped
A simple test using liblockdep and a pthread mutex confirms the
problem: (omitting stack traces)
new class 0xe15038: 0x7ffc64950f20
acquire class [0xe15038] 0x7ffc64950f20
acquire class [0xe15038] 0x7ffc64950f20
hash chain already cached, key: 0000000000000000 tail class:
[0xe15038] 0x7ffc64950f20
2.- Consider an ABBA in 2 different tasks and no class yet acquired.
T1 [key1,key2]->new key T2[key1,key2]->new key
-- --
A [0,0]->0
B [0,1]->1
B [0,1]->1 (collision)
A
In this case the collision prevents lockdep from creating the new
dependency A->B. This in turn results in lockdep not detecting the
circular dependency when T2 acquires A.
Signed-off-by: Alfredo Alvarez Fernandez <alfredoalvarezernandez@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: sasha.levin@oracle.com
Link: http://lkml.kernel.org/r/1455147212-2389-4-git-send-email-alfredoalvarezernandez@gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Make use of wake-queues and enable the wakeup to occur after releasing the
wait_lock. This is similar to what we do with rtmutex top waiter,
slightly shortening the critical region and allow other waiters to
acquire the wait_lock sooner. In low contention cases it can also help
the recently woken waiter to find the wait_lock available (fastpath)
when it continues execution.
Reviewed-by: Waiman Long <Waiman.Long@hpe.com>
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Ding Tianhong <dingtianhong@huawei.com>
Cc: Jason Low <jason.low2@hp.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Paul E. McKenney <paulmck@us.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Waiman Long <waiman.long@hpe.com>
Cc: Will Deacon <Will.Deacon@arm.com>
Link: http://lkml.kernel.org/r/20160125022343.GA3322@linux-uzut.site
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This patch enables the tracking of the number of slowpath locking
operations performed. This can be used to compare against the number
of lock stealing operations to see what percentage of locks are stolen
versus acquired via the regular slowpath.
Signed-off-by: Waiman Long <Waiman.Long@hpe.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Douglas Hatch <doug.hatch@hpe.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Scott J Norton <scott.norton@hpe.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1449778666-13593-2-git-send-email-Waiman.Long@hpe.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The newly introduced smp_cond_acquire() was used to replace the
slowpath lock acquisition loop. Similarly, the new function can also
be applied to the pending bit locking loop. This patch uses the new
function in that loop.
Signed-off-by: Waiman Long <Waiman.Long@hpe.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Douglas Hatch <doug.hatch@hpe.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Scott J Norton <scott.norton@hpe.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1449778666-13593-1-git-send-email-Waiman.Long@hpe.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This patch moves the lock stealing count tracking code into
pv_queued_spin_steal_lock() instead of via a jacket function simplifying
the code.
Signed-off-by: Waiman Long <Waiman.Long@hpe.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Douglas Hatch <doug.hatch@hpe.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Scott J Norton <scott.norton@hpe.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1449778666-13593-3-git-send-email-Waiman.Long@hpe.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Similar to commit b4b29f9485 ("locking/osq: Fix ordering of node
initialisation in osq_lock") the use of xchg_acquire() is
fundamentally broken with MCS like constructs.
Furthermore, it turns out we rely on the global transitivity of this
operation because the unlock path observes the pointer with a
READ_ONCE(), not an smp_load_acquire().
This is non-critical because the MCS code isn't actually used and
mostly serves as documentation, a stepping stone to the more complex
things we've build on top of the idea.
Reported-by: Andrea Parri <parri.andrea@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Will Deacon <will.deacon@arm.com>
Fixes: 3552a07a9c ("locking/mcs: Use acquire/release semantics")
Signed-off-by: Ingo Molnar <mingo@kernel.org>
if (A || B) {
} else if (A && !B) {
}
If A we'll take the first branch, if !A we will not satisfy the second.
Therefore the second branch will never be taken.
Reported-by: luca abeni <luca.abeni@unitn.it>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Juri Lelli <juri.lelli@arm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20160225140149.GK6357@twins.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
The preempt_disable() invokes preempt_count_add() which saves the caller
in ->preempt_disable_ip. It uses CALLER_ADDR1 which does not look for
its caller but for the parent of the caller. Which means we get the correct
caller for something like spin_lock() unless the architectures inlines
those invocations. It is always wrong for preempt_disable() or
local_bh_disable().
This patch makes the function get_lock_parent_ip() which tries
CALLER_ADDR0,1,2 if the former is a locking function.
This seems to record the preempt_disable() caller properly for
preempt_disable() itself as well as for get_cpu_var() or
local_bh_disable().
Steven asked for the get_parent_ip() -> get_lock_parent_ip() rename.
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20160226135456.GB18244@linutronix.de
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When profiling syscall overhead on nohz-full kernels,
after removing __acct_update_integrals() from the profile,
native_sched_clock() remains as the top CPU user. This can be
reduced by moving VIRT_CPU_ACCOUNTING_GEN to jiffy granularity.
This will reduce timing accuracy on nohz_full CPUs to jiffy
based sampling, just like on normal CPUs. It results in
totally removing native_sched_clock from the profile, and
significantly speeding up the syscall entry and exit path,
as well as irq entry and exit, and KVM guest entry & exit.
Additionally, only call the more expensive functions (and
advance the seqlock) when jiffies actually changed.
This code relies on another CPU advancing jiffies when the
system is busy. On a nohz_full system, this is done by a
housekeeping CPU.
A microbenchmark calling an invalid syscall number 10 million
times in a row speeds up an additional 30% over the numbers
with just the previous patches, for a total speedup of about
40% over 4.4 and 4.5-rc1.
Run times for the microbenchmark:
4.4 3.8 seconds
4.5-rc1 3.7 seconds
4.5-rc1 + first patch 3.3 seconds
4.5-rc1 + first 3 patches 3.1 seconds
4.5-rc1 + all patches 2.3 seconds
A non-NOHZ_FULL cpu (not the housekeeping CPU):
all kernels 1.86 seconds
Signed-off-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: clark@redhat.com
Cc: eric.dumazet@gmail.com
Cc: fweisbec@gmail.com
Cc: luto@amacapital.net
Link: http://lkml.kernel.org/r/1455152907-18495-5-git-send-email-riel@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
It looks like all the call paths that lead to __acct_update_integrals()
already have irqs disabled, and __acct_update_integrals() does not need
to disable irqs itself.
This is very convenient since about half the CPU time left in this
function was spent in local_irq_save alone.
Performance of a microbenchmark that calls an invalid syscall
ten million times in a row on a nohz_full CPU improves 21% vs.
4.5-rc1 with both the removal of divisions from __acct_update_integrals()
and this patch, with runtime dropping from 3.7 to 2.9 seconds.
With these patches applied, the highest remaining cpu user in
the trace is native_sched_clock, which is addressed in the next
patch.
For testing purposes I stuck a WARN_ON(!irqs_disabled()) test
in __acct_update_integrals(). It did not trigger.
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: clark@redhat.com
Cc: eric.dumazet@gmail.com
Cc: fweisbec@gmail.com
Cc: luto@amacapital.net
Link: http://lkml.kernel.org/r/1455152907-18495-4-git-send-email-riel@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Change the indentation in __acct_update_integrals() to make the function
a little easier to read.
Suggested-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: clark@redhat.com
Cc: eric.dumazet@gmail.com
Cc: fweisbec@gmail.com
Cc: luto@amacapital.net
Link: http://lkml.kernel.org/r/1455152907-18495-3-git-send-email-riel@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
When running a microbenchmark calling an invalid syscall number
in a loop, on a nohz_full CPU, we spend a full 9% of our CPU
time in __acct_update_integrals().
This function converts cputime_t to jiffies, to a timeval, only to
convert the timeval back to microseconds before discarding it.
This patch leaves __acct_update_integrals() functionally equivalent,
but speeds things up by about 12%, with 10 million calls to an
invalid syscall number dropping from 3.7 to 3.25 seconds.
Signed-off-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: clark@redhat.com
Cc: eric.dumazet@gmail.com
Cc: fweisbec@gmail.com
Cc: luto@amacapital.net
Link: http://lkml.kernel.org/r/1455152907-18495-2-git-send-email-riel@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
I've been debugging why deadline tasks can cause the RT scheduler to
throttle, even when the deadline tasks are only taking up 50% of the
CPU and RT tasks are not even using 1% of the CPU. Here's what I found.
In order to keep a CPU from being hogged by RT tasks, the deadline
scheduler adds its run time (delta_exec) to the rt_time of the RT
bandwidth. That way, if the two use more than 95% of the CPU within one
second (default settings), the RT tasks are throttled to allow non RT
tasks to run.
Although the deadline tasks add their run time to the RT bandwidth, it
lets the RT tasks do the accounting. This is where the problem lies. If
a deadline task runs for a bit, and no RT tasks are running, then it
will continually add to the RT rt_time that is used to calculate how
much CPU the RT tasks use. But no RT period is in play, and this
accumulation of the runtime never gets reset.
When an RT task finally gets to run, and the watchdog goes off, it can
see that the RT task has used more than it should of, because the
deadline task added all this runtime to its rt_time. Then the RT task
that just woke up gets throttled for no good reason.
I also noticed that when an RT task is queued, it starts the timer to
account for overload and such. But that timer goes off one period
later, which may be too late and the extra rt_time will trigger a
throttle.
This is a quick work around to the problem. When a new RT task is
queued, the bandwidth timer is set to go off immediately. Then the
timer can clear out the extra time added to the rt_time while there was
no RT task running. This stops my tests from triggering the throttle,
and it will still throttle if an RT task runs too much, even while a
deadline task is running.
A better solution may be to subtract the bandwidth that the deadline
task uses from the rt_runtime, and add it back when its finished. Then
there wont be a need for runtime tracking of the time used by deadline
tasks.
I may play with that solution tomorrow.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <juri.lelli@gmail.com>
Cc: <williams@redhat.com>
Cc: Clark Williams
Cc: Daniel Bristot de Oliveira <bristot@redhat.com>
Cc: John Kacur <jkacur@redhat.com>
Cc: Juri Lelli
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20160216183746.349ec98b@gandalf.local.home
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Playing with SCHED_DEADLINE and cpusets, I found that I was unable to create
new SCHED_DEADLINE tasks, with the error of EBUSY as if the bandwidth was
already used up. I then realized there wa no way to see what bandwidth is
used by the runqueues to debug the issue.
By adding the dl_bw->bw and dl_bw->total_bw to the output of the deadline
info in /proc/sched_debug, this allows us to see what bandwidth has been
reserved and where a problem may exist.
For example, before the issue we see the ratio of the bandwidth:
# cat /proc/sys/kernel/sched_rt_runtime_us
950000
# cat /proc/sys/kernel/sched_rt_period_us
1000000
# grep dl /proc/sched_debug
dl_rq[0]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : 0
dl_rq[1]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : 0
dl_rq[2]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : 0
dl_rq[3]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : 0
dl_rq[4]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : 0
dl_rq[5]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : 0
dl_rq[6]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : 0
dl_rq[7]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : 0
Note: (950000 / 1000000) << 20 == 996147
After I played with cpusets and hit the issue, the result is now:
# grep dl /proc/sched_debug
dl_rq[0]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : -104857
dl_rq[1]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : 104857
dl_rq[2]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : 104857
dl_rq[3]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : 104857
dl_rq[4]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : -104857
dl_rq[5]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : -104857
dl_rq[6]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : -104857
dl_rq[7]:
.dl_nr_running : 0
.dl_bw->bw : 996147
.dl_bw->total_bw : -104857
This shows that there is definitely a problem as we should never have a
negative total bandwidth.
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Clark Williams <williams@redhat.com>
Cc: Juri Lelli <juri.lelli@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20160222212825.756849091@goodmis.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
