| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
We will use it to count how many addresses are in the entry->ip[] array,
excluding PERF_CONTEXT_{KERNEL,USER,etc} entries, so that we can really
return the number of entries specified by the user via the relevant
sysctl, kernel.perf_event_max_contexts, or via the per event
perf_event_attr.sample_max_stack knob.
This way we keep the perf_sample->ip_callchain->nr meaning, that is the
number of entries, be it real addresses or PERF_CONTEXT_ entries, while
honouring the max_stack knobs, i.e. the end result will be max_stack
entries if we have at least that many entries in a given stack trace.
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/n/tip-s8teto51tdqvlfhefndtat9r@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
|
|
This makes perf_callchain_{user,kernel}() receive the max stack
as context for the perf_callchain_entry, instead of accessing
the global sysctl_perf_event_max_stack.
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Brendan Gregg <brendan.d.gregg@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Milian Wolff <milian.wolff@kdab.com>
Cc: Namhyung Kim <namhyung@kernel.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: Wang Nan <wangnan0@huawei.com>
Cc: Zefan Li <lizefan@huawei.com>
Link: http://lkml.kernel.org/n/tip-kolmn1yo40p7jhswxwrc7rrd@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
|
|
The default remains 127, which is good for most cases, and not even hit
most of the time, but then for some cases, as reported by Brendan, 1024+
deep frames are appearing on the radar for things like groovy, ruby.
And in some workloads putting a _lower_ cap on this may make sense. One
that is per event still needs to be put in place tho.
The new file is:
# cat /proc/sys/kernel/perf_event_max_stack
127
Chaging it:
# echo 256 > /proc/sys/kernel/perf_event_max_stack
# cat /proc/sys/kernel/perf_event_max_stack
256
But as soon as there is some event using callchains we get:
# echo 512 > /proc/sys/kernel/perf_event_max_stack
-bash: echo: write error: Device or resource busy
#
Because we only allocate the callchain percpu data structures when there
is a user, which allows for changing the max easily, its just a matter
of having no callchain users at that point.
Reported-and-Tested-by: Brendan Gregg <brendan.d.gregg@gmail.com>
Reviewed-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: David Ahern <dsahern@gmail.com>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: He Kuang <hekuang@huawei.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Milian Wolff <milian.wolff@kdab.com>
Cc: Namhyung Kim <namhyung@kernel.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: Wang Nan <wangnan0@huawei.com>
Cc: Zefan Li <lizefan@huawei.com>
Link: http://lkml.kernel.org/r/20160426002928.GB16708@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
|
|
There have been several reports of random processes being killed with
a bus error or segfault during userspace stack walking in perf. One
of the root causes of this problem is an asynchronous modification to
thread_info fault_address and fault_code, which stems from a perf
counter interrupt arriving during kernel processing of a "benign"
fault, such as a TSB miss. Since perf_callchain_user() invokes
copy_from_user() to read user stacks, a fault is not only possible,
but probable. Validity checks on the stack address merely cover up the
problem and reduce its frequency.
The solution here is to save and restore fault_address and fault_code
in perf_callchain_user() so that the benign fault handler is not
disturbed by a perf interrupt.
Signed-off-by: Rob Gardner <rob.gardner@oracle.com>
Signed-off-by: Dave Aldridge <david.j.aldridge@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When an interrupt (such as a perf counter interrupt) is delivered
while executing in user space, the trap entry code puts ASI_AIUS in
%asi so that copy_from_user() and copy_to_user() will access the
correct memory. But if a perf counter interrupt is delivered while the
cpu is already executing in kernel space, then the trap entry code
will put ASI_P in %asi, and this will prevent copy_from_user() from
reading any useful stack data in either of the perf_callchain_user_X
functions, and thus no user callgraph data will be collected for this
sample period. An additional problem is that a fault is guaranteed
to occur, and though it will be silently covered up, it wastes time
and could perturb state.
In perf_callchain_user(), we ensure that %asi contains ASI_AIUS
because we know for a fact that the subsequent calls to
copy_from_user() are intended to read the user's stack.
[ Use get_fs()/set_fs() -DaveM ]
Signed-off-by: Rob Gardner <rob.gardner@oracle.com>
Signed-off-by: Dave Aldridge <david.j.aldridge@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
There were still a number of references to my old Red Hat email
address in the kernel source. Remove these while keeping the
Red Hat copyright notices intact.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Mike Galbraith <efault@gmx.de>
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>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
We currently use PERF_EVENT_TXN flag to determine if we are in the middle
of a transaction. If in a transaction, we defer the schedulability checks
from pmu->add() operation to the pmu->commit() operation.
Now that we have "transaction types" (PERF_PMU_TXN_ADD, PERF_PMU_TXN_READ)
we can use the type to determine if we are in a transaction and drop the
PERF_EVENT_TXN flag.
When PERF_EVENT_TXN is dropped, the cpuhw->group_flag on some architectures
becomes unused, so drop that field as well.
This is an extension of the Powerpc patch from Peter Zijlstra to s390,
Sparc and x86 architectures.
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
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>
Link: http://lkml.kernel.org/r/1441336073-22750-11-git-send-email-sukadev@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
Currently, the PMU interface allows reading only one counter at a time.
But some PMUs like the 24x7 counters in Power, support reading several
counters at once. To leveage this functionality, extend the transaction
interface to support a "transaction type".
The first type, PERF_PMU_TXN_ADD, refers to the existing transactions,
i.e. used to _schedule_ all the events on the PMU as a group. A second
transaction type, PERF_PMU_TXN_READ, will be used in a follow-on patch,
by the 24x7 counters to read several counters at once.
Extend the transaction interfaces to the PMU to accept a 'txn_flags'
parameter and use this parameter to ignore any transactions that are
not of type PERF_PMU_TXN_ADD.
Thanks to Peter Zijlstra for his input.
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
[peterz: s390 compile fix]
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Michael Ellerman <mpe@ellerman.id.au>
Cc: Arnaldo Carvalho de Melo <acme@kernel.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>
Link: http://lkml.kernel.org/r/1441336073-22750-3-git-send-email-sukadev@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
In ->commit_txn() 'cpuc' is already initialized when it is
declared, so we can remove the duplicate assignment.
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Michael Ellerman <mpe@ellerman.id.au>
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: sparclinux@vger.kernel.org
Link: http://lkml.kernel.org/r/1441336073-22750-2-git-send-email-sukadev@linux.vnet.ibm.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
perf walks userspace callchains by following frame pointers. Use the
UREG_FP macro to make it clearer that the %fp is being used.
Signed-off-by: David Ahern <david.ahern@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Processes are getting killed (sigbus or segv) while walking userspace
callchains when using perf. In some instances I have seen ufp = 0x7ff
which does not seem like a proper stack address.
This patch adds a function to run validity checks against the address
before attempting the copy_from_user. The checks are copied from the
x86 version as a start point with the addition of a 4-byte alignment
check.
Signed-off-by: David Ahern <david.ahern@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Page faults generated walking userspace stacks can call schedule to switch
out the task. When collecting callchains for scheduler tracepoints this
causes a deadlock as the tracepoints can be hit with the runqueue lock held:
[ 8138.159054] WARNING: CPU: 758 PID: 12488 at /opt/dahern/linux.git/arch/sparc/kernel/nmi.c:80 perfctr_irq+0x1f8/0x2b4()
[ 8138.203152] Watchdog detected hard LOCKUP on cpu 758
[ 8138.410969] CPU: 758 PID: 12488 Comm: perf Not tainted 4.0.0-rc6+ #6
[ 8138.437146] Call Trace:
[ 8138.447193] [000000000045cdd4] warn_slowpath_common+0x7c/0xa0
[ 8138.471238] [000000000045ce90] warn_slowpath_fmt+0x30/0x40
[ 8138.494189] [0000000000983e38] perfctr_irq+0x1f8/0x2b4
[ 8138.515716] [00000000004209f4] tl0_irq15+0x14/0x20
[ 8138.535791] [00000000009839ec] _raw_spin_trylock_bh+0x68/0x108
[ 8138.560180] [0000000000980018] __schedule+0xcc/0x710
[ 8138.580981] [00000000009806dc] preempt_schedule_common+0x10/0x3c
[ 8138.606082] [000000000098077c] _cond_resched+0x34/0x44
[ 8138.627603] [0000000000565990] kmem_cache_alloc_node+0x24/0x1a0
[ 8138.652345] [0000000000450b60] tsb_grow+0xac/0x488
[ 8138.672429] [0000000000985040] do_sparc64_fault+0x4dc/0x6e4
[ 8138.695736] [0000000000407c2c] sparc64_realfault_common+0x10/0x20
[ 8138.721202] [00000000006f2e24] NG4copy_from_user+0xa4/0x3c0
[ 8138.744510] [000000000044f900] perf_callchain_user+0x5c/0x6c
[ 8138.768182] [0000000000517b5c] perf_callchain+0x16c/0x19c
[ 8138.790774] [0000000000515f84] perf_prepare_sample+0x68/0x218
[ 8138.814801] ---[ end trace 42ca6294b1ff7573 ]---
As with PowerPC (b59a1bfcc240, "powerpc/perf: Disable pagefaults during
callchain stack read") disable pagefaults while walking userspace stacks.
Signed-off-by: David Ahern <david.ahern@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
They both work equally well, and the M7 implementation is
simpler and cheaper (less register writes).
With help from David Ahern.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The M7 processor has a different hypervisor group id and different PCR fast
trap values. PIC read/write functions and PCR bit fields are the same as
the T4 so those are reused.
Signed-off-by: David Ahern <david.ahern@oracle.com>
Acked-by: Bob Picco <bob.picco@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Currently perf-stat (aka, counting mode) does not work:
$ perf stat ls
...
Performance counter stats for 'ls':
1.585665 task-clock (msec) # 0.580 CPUs utilized
24 context-switches # 0.015 M/sec
0 cpu-migrations # 0.000 K/sec
86 page-faults # 0.054 M/sec
<not supported> cycles
<not supported> stalled-cycles-frontend
<not supported> stalled-cycles-backend
<not supported> instructions
<not supported> branches
<not supported> branch-misses
0.002735100 seconds time elapsed
The reason is that state is never reset (stays with PERF_HES_UPTODATE set).
Add a call to sparc_pmu_enable_event during the added_event handling.
Clean up the encoding since pmu_start calls sparc_pmu_enable_event which
does the same. Passing PERF_EF_RELOAD to sparc_pmu_start means the call
to sparc_perf_event_set_period can be removed as well.
With this patch:
$ perf stat ls
...
Performance counter stats for 'ls':
1.552890 task-clock (msec) # 0.552 CPUs utilized
24 context-switches # 0.015 M/sec
0 cpu-migrations # 0.000 K/sec
86 page-faults # 0.055 M/sec
5,748,997 cycles # 3.702 GHz
<not supported> stalled-cycles-frontend:HG
<not supported> stalled-cycles-backend:HG
1,684,362 instructions:HG # 0.29 insns per cycle
295,133 branches:HG # 190.054 M/sec
28,007 branch-misses:HG # 9.49% of all branches
0.002815665 seconds time elapsed
Signed-off-by: David Ahern <david.ahern@oracle.com>
Acked-by: Bob Picco <bob.picco@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
perf_pmu_disable is called by core perf code before pmu->del and the
enable function is called by core perf code afterwards. No need to
call again within sparc_pmu_del.
Ditto for pmu->add and sparc_pmu_add.
Signed-off-by: David Ahern <david.ahern@oracle.com>
Acked-by: Bob Picco <bob.picco@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu
Pull percpu consistent-ops changes from Tejun Heo:
"Way back, before the current percpu allocator was implemented, static
and dynamic percpu memory areas were allocated and handled separately
and had their own accessors. The distinction has been gone for many
years now; however, the now duplicate two sets of accessors remained
with the pointer based ones - this_cpu_*() - evolving various other
operations over time. During the process, we also accumulated other
inconsistent operations.
This pull request contains Christoph's patches to clean up the
duplicate accessor situation. __get_cpu_var() uses are replaced with
with this_cpu_ptr() and __this_cpu_ptr() with raw_cpu_ptr().
Unfortunately, the former sometimes is tricky thanks to C being a bit
messy with the distinction between lvalues and pointers, which led to
a rather ugly solution for cpumask_var_t involving the introduction of
this_cpu_cpumask_var_ptr().
This converts most of the uses but not all. Christoph will follow up
with the remaining conversions in this merge window and hopefully
remove the obsolete accessors"
* 'for-3.18-consistent-ops' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/percpu: (38 commits)
irqchip: Properly fetch the per cpu offset
percpu: Resolve ambiguities in __get_cpu_var/cpumask_var_t -fix
ia64: sn_nodepda cannot be assigned to after this_cpu conversion. Use __this_cpu_write.
percpu: Resolve ambiguities in __get_cpu_var/cpumask_var_t
Revert "powerpc: Replace __get_cpu_var uses"
percpu: Remove __this_cpu_ptr
clocksource: Replace __this_cpu_ptr with raw_cpu_ptr
sparc: Replace __get_cpu_var uses
avr32: Replace __get_cpu_var with __this_cpu_write
blackfin: Replace __get_cpu_var uses
tile: Use this_cpu_ptr() for hardware counters
tile: Replace __get_cpu_var uses
powerpc: Replace __get_cpu_var uses
alpha: Replace __get_cpu_var
ia64: Replace __get_cpu_var uses
s390: cio driver &__get_cpu_var replacements
s390: Replace __get_cpu_var uses
mips: Replace __get_cpu_var uses
MIPS: Replace __get_cpu_var uses in FPU emulator.
arm: Replace __this_cpu_ptr with raw_cpu_ptr
...
|
|
The T5 (niagara5) has different PCR related HV fast trap values and a new
HV API Group. This patch utilizes these and shares when possible with niagara4.
We use the same sparc_pmu niagara4_pmu. Should there be new effort to
obtain the MCU perf statistics then this would have to be changed.
Cc: sparclinux@vger.kernel.org
Signed-off-by: Bob Picco <bob.picco@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
__get_cpu_var() is used for multiple purposes in the kernel source. One of
them is address calculation via the form &__get_cpu_var(x). This calculates
the address for the instance of the percpu variable of the current processor
based on an offset.
Other use cases are for storing and retrieving data from the current
processors percpu area. __get_cpu_var() can be used as an lvalue when
writing data or on the right side of an assignment.
__get_cpu_var() is defined as :
#define __get_cpu_var(var) (*this_cpu_ptr(&(var)))
__get_cpu_var() always only does an address determination. However, store
and retrieve operations could use a segment prefix (or global register on
other platforms) to avoid the address calculation.
this_cpu_write() and this_cpu_read() can directly take an offset into a
percpu area and use optimized assembly code to read and write per cpu
variables.
This patch converts __get_cpu_var into either an explicit address
calculation using this_cpu_ptr() or into a use of this_cpu operations that
use the offset. Thereby address calculations are avoided and less registers
are used when code is generated.
At the end of the patch set all uses of __get_cpu_var have been removed so
the macro is removed too.
The patch set includes passes over all arches as well. Once these operations
are used throughout then specialized macros can be defined in non -x86
arches as well in order to optimize per cpu access by f.e. using a global
register that may be set to the per cpu base.
Transformations done to __get_cpu_var()
1. Determine the address of the percpu instance of the current processor.
DEFINE_PER_CPU(int, y);
int *x = &__get_cpu_var(y);
Converts to
int *x = this_cpu_ptr(&y);
2. Same as #1 but this time an array structure is involved.
DEFINE_PER_CPU(int, y[20]);
int *x = __get_cpu_var(y);
Converts to
int *x = this_cpu_ptr(y);
3. Retrieve the content of the current processors instance of a per cpu
variable.
DEFINE_PER_CPU(int, y);
int x = __get_cpu_var(y)
Converts to
int x = __this_cpu_read(y);
4. Retrieve the content of a percpu struct
DEFINE_PER_CPU(struct mystruct, y);
struct mystruct x = __get_cpu_var(y);
Converts to
memcpy(&x, this_cpu_ptr(&y), sizeof(x));
5. Assignment to a per cpu variable
DEFINE_PER_CPU(int, y)
__get_cpu_var(y) = x;
Converts to
__this_cpu_write(y, x);
6. Increment/Decrement etc of a per cpu variable
DEFINE_PER_CPU(int, y);
__get_cpu_var(y)++
Converts to
__this_cpu_inc(y)
Cc: sparclinux@vger.kernel.org
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
Christopher reports that perf_event_print_debug() can crash in uniprocessor
builds. The crash is due to pcr_ops being NULL.
This happens because pcr_arch_init() is only invoked by smp_cpus_done() which
only executes in SMP builds.
init_hw_perf_events() is closely intertwined with pcr_ops being setup properly,
therefore:
1) Call pcr_arch_init() early on from init_hw_perf_events(), instead of
from smp_cpus_done().
2) Do not hook up a PMU type if pcr_ops is NULL after pcr_arch_init().
3) Move init_hw_perf_events to a later initcall so that it we will be
sure to invoke pcr_arch_init() after all cpus are brought up.
Finally, guard the one naked sequence of pcr_ops dereferences in
__global_pmu_self() with an appropriate NULL check.
Reported-by: Christopher Alexander Tobias Schulze <cat.schulze@alice-dsl.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Fix following sparse warnings:
kernel/perf_event.c:113:1: warning: symbol 'cpu_hw_events' was not declared. Should it be static?
kernel/perf_event.c:1156:6: warning: symbol 'perf_event_grab_pmc' was not declared. Should it be static?
kernel/perf_event.c:1172:6: warning: symbol 'perf_event_release_pmc' was not declared. Should it be static?
kernel/perf_event.c:1672:12: warning: symbol 'init_hw_perf_events' was not declared. Should it be static?
kernel/perf_event.c:1749:52: warning: incorrect type in argument 2 (different address spaces)
kernel/perf_event.c:1772:60: warning: incorrect type in argument 2 (different address spaces)
kernel/perf_event.c:1779:60: warning: incorrect type in argument 2 (different address spaces)
Define the functions static as they are not used outside this file.
Fix it so copy_from_user are supplied with pointers annotated _user
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The Montgomery Multiply, Montgomery Square, and Multiple-Precision
Multiply instructions work by loading a combination of the floating
point and multiple register windows worth of integer registers
with the inputs.
These values are 64-bit. But for 32-bit userland processes we only
save the low 32-bits of each integer register during a register spill.
This is because the register window save area is in the user stack and
has a fixed layout.
Therefore, the only way to use these instruction in 32-bit mode is to
perform the following sequence:
1) Load the top-32bits of a choosen integer register with a sentinel,
say "-1". This will be in the outer-most register window.
The idea is that we're trying to see if the outer-most register
window gets spilled, and thus the 64-bit values were truncated.
2) Load all the inputs for the montmul/montsqr/mpmul instruction,
down to the inner-most register window.
3) Execute the opcode.
4) Traverse back up to the outer-most register window.
5) Check the sentinel, if it's still "-1" store the results.
Otherwise retry the entire sequence.
This retry is extremely troublesome. If you're just unlucky and an
interrupt or other trap happens, it'll push that outer-most window to
the stack and clear the sentinel when we restore it.
We could retry forever and never make forward progress if interrupts
arrive at a fast enough rate (consider perf events as one example).
So we have do limited retries and fallback to software which is
extremely non-deterministic.
Luckily it's very straightforward to provide a mechanism to let
32-bit applications use a 64-bit stack. Stacks in 64-bit mode are
biased by 2047 bytes, which means that the lowest bit is set in the
actual %sp register value.
So if we see bit zero set in a 32-bit application's stack we treat
it like a 64-bit stack.
Runtime detection of such a facility is tricky, and cumbersome at
best. For example, just trying to use a biased stack and seeing if it
works is hard to recover from (the signal handler will need to use an
alt stack, plus something along the lines of longjmp). Therefore, we
add a system call to report a bitmask of arch specific features like
this in a cheap and less hairy way.
With help from Andy Polyakov.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
There was a serious disconnect in the logic happening in
sparc_pmu_disable_event() vs. sparc_pmu_enable_event().
Event disable is implemented by programming a NOP event into the PCR.
However, event enable was not reversing this operation. Instead, it
was setting the User/Priv/Hypervisor trace enable bits.
That's not sparc_pmu_enable_event()'s job, that's what
sparc_pmu_enable() and sparc_pmu_disable() do .
The intent of sparc_pmu_enable_event() is clear, since it first clear
out the event type encoding field. So fix this by OR'ing in the event
encoding rather than the trace enable bits.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
If the MM is not active, only report the top-level PC. Do not try to
access the address space.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Describe how we support two types of PMU setups, one with a single control
register and two counters stored in a single register, and another with
one control register per counter and each counter living in it's own
register.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When cpuc->n_events is zero, we actually don't do anything and we just
write the cpuc->pcr[0] value as-is without any modifications.
The "pcr = 0;" assignment there was just useless and confusing.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make the per-cpu pcr save area an array instead of one u64.
Describe how many PCR and PIC registers the chip has in the sparc_pmu
descriptor.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Now specified in sparc_pmu descriptor.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Starting with SPARC-T4 we have a seperate PCR control register
for each performance counter, and there are absolutely no
restrictions on what events can run on which counters.
Add flags that we can use to elide the conflict and dependency
logic used to handle older chips.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
And, like for the PCR, allow indexing of different PIC register
numbers.
This also removes all of the non-__KERNEL__ bits from asm/perfctr.h,
nothing kernel side should include it any more.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
SPARC-T4 and later have multiple PCR registers, one for each
PIC counter.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
We always need to pass the last sample period to
perf_sample_data_init(), otherwise the event distribution will be
wrong. Thus, modifiyng the function interface with the required period
as argument. So basically a pattern like this:
perf_sample_data_init(&data, ~0ULL);
data.period = event->hw.last_period;
will now be like that:
perf_sample_data_init(&data, ~0ULL, event->hw.last_period);
Avoids unininitialized data.period and simplifies code.
Signed-off-by: Robert Richter <robert.richter@amd.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1333390758-10893-3-git-send-email-robert.richter@amd.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
|
|
Disintegrate asm/system.h for Sparc.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: sparclinux@vger.kernel.org
|
|
PERF_SAMPLE_BRANCH_* is disabled for:
- SW events (sw counters, tracepoints)
- HW breakpoints
- ALL but Intel x86 architecture
- AMD64 processors
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1328826068-11713-10-git-send-email-eranian@google.com
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
The cpu compatible string we look for is "SPARC-T3".
As far as memset/memcpy optimizations go, we treat this chip the same
as Niagara-T2/T2+. Use cache initializing stores for memset, and use
perfetch, FPU block loads, cache initializing stores, and block stores
for copies.
We use the Niagara-T2 perf support, since T3 is a close relative in
this regard. Later we'll add support for the new events T3 can
report, plus enable T3's new "sample" mode.
For now I haven't added any new ELF hwcap flags. We probably need
to add a couple, for example:
T2 and T3 both support the population count instruction in hardware.
T3 supports VIS3 instructions, including support (finally) for
partitioned shift. One can also now move directly between float
and integer registers.
T3 supports instructions meant to help with Galois Field and other HPC
calculations, such as XOR multiply. Also there are "OP and negate"
instructions, for example "fnmul" which is multiply-and-negate.
T3 recognizes the transactional memory opcodes, however since
transactional memory isn't supported: 1) 'commit' behaves as a NOP and
2) 'chkpt' always branches 3) 'rdcps' returns all zeros and 4) 'wrcps'
behaves as a NOP.
So we'll need about 3 new elf capability flags in the end to represent
all of these things.
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
This allows us to move duplicated code in <asm/atomic.h>
(atomic_inc_not_zero() for now) to <linux/atomic.h>
Signed-off-by: Arun Sharma <asharma@fb.com>
Reviewed-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Add a NODE level to the generic cache events which is used to measure
local vs remote memory accesses. Like all other cache events, an
ACCESS is HIT+MISS, if there is no way to distinguish between reads
and writes do reads only etc..
The below needs filling out for !x86 (which I filled out with
unsupported events).
I'm fairly sure ARM can leave it like that since it doesn't strike me as
an architecture that even has NUMA support. SH might have something since
it does appear to have some NUMA bits.
Sparc64, PowerPC and MIPS certainly want a good look there since they
clearly are NUMA capable.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: David Miller <davem@davemloft.net>
Cc: Anton Blanchard <anton@samba.org>
Cc: David Daney <ddaney@caviumnetworks.com>
Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1303508226.4865.8.camel@laptop
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
The nmi parameter indicated if we could do wakeups from the current
context, if not, we would set some state and self-IPI and let the
resulting interrupt do the wakeup.
For the various event classes:
- hardware: nmi=0; PMI is in fact an NMI or we run irq_work_run from
the PMI-tail (ARM etc.)
- tracepoint: nmi=0; since tracepoint could be from NMI context.
- software: nmi=[0,1]; some, like the schedule thing cannot
perform wakeups, and hence need 0.
As one can see, there is very little nmi=1 usage, and the down-side of
not using it is that on some platforms some software events can have a
jiffy delay in wakeup (when arch_irq_work_raise isn't implemented).
The up-side however is that we can remove the nmi parameter and save a
bunch of conditionals in fast paths.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Michael Cree <mcree@orcon.net.nz>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com>
Cc: Anton Blanchard <anton@samba.org>
Cc: Eric B Munson <emunson@mgebm.net>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: David S. Miller <davem@davemloft.net>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jason Wessel <jason.wessel@windriver.com>
Cc: Don Zickus <dzickus@redhat.com>
Link: http://lkml.kernel.org/n/tip-agjev8eu666tvknpb3iaj0fg@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
We have all the cpu related info in cpu.c - so move
the remaining functions to support /proc/cpuinfo to this file.
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Fixes generated by 'codespell' and manually reviewed.
Signed-off-by: Lucas De Marchi <lucas.demarchi@profusion.mobi>
|
|
Extend the perf_pmu_register() interface to allow for named and
dynamic pmu types.
Because we need to support the existing static types we cannot use
dynamic types for everything, hence provide a type argument.
If we want to enumerate the PMUs they need a name, provide one.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <20101117222056.259707703@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
Fix a typo in:
004417a6d468: perf, arch: Cleanup perf-pmu init vs lockup-detector
Which caused a build failure on Sparc, reported by Stephen Rothwell.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
The perf hardware pmu got initialized at various points in the boot,
some before early_initcall() some after (notably arch_initcall).
The problem is that the NMI lockup detector is ran from early_initcall()
and expects the hardware pmu to be present.
Sanitize this by moving all architecture hardware pmu implementations to
initialize at early_initcall() and move the lockup detector to an explicit
initcall right after that.
Cc: paulus <paulus@samba.org>
Cc: davem <davem@davemloft.net>
Cc: Michael Cree <mcree@orcon.net.nz>
Cc: Deng-Cheng Zhu <dengcheng.zhu@gmail.com>
Acked-by: Paul Mundt <lethal@linux-sh.org>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1290707759.2145.119.camel@laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
Conflicts:
arch/sparc/kernel/perf_event.c
Merge reason: Resolve the conflict.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
Encoding is "(encoding << 16) | pic_mask"
Signed-off-by: David S. Miller <davem@davemloft.net>
|
