As this isn't used at all in mmap.h but in evlist.h, so to cut down the
header dependency tree, move it to where it is used.
Also add mmap.h to the places using it but previously getting it
indirectly via evlist.h.
Add missing pthread.h to evlist.h, as it has a pthread_t struct member
and was getting the header via mmap.h.
Noticed while processing a Jiri's libperf batch touching mmap.h, where
almost everything gets rebuilt because evlist.h is so popular, so cut
down't this rebuild the world party.
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Link: https://lkml.kernel.org/n/tip-he0uljeftl0xfveh3d6vtode@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Compression is implemented using the functions from zstd.c. As the memory
to operate on the compression uses mmap->aio.data[] buffers. If Zstd
streaming compression API fails for some reason the data to be compressed
are just copied into the memory buffers using plain memcpy().
Compressed trace frame consists of an array of PERF_RECORD_COMPRESSED
records. Each element of the array is not longer that PERF_SAMPLE_MAX_SIZE
and consists of perf_event_header followed by the compressed chunk
that is decompressed on the loading stage.
perf_mmap__aio_push() is replaced by perf_mmap__push() which is now used
in the both serial and AIO streaming cases. perf_mmap__push() is extended
with positive return values to signify absence of data ready for
processing.
Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
Reviewed-by: Jiri Olsa <jolsa@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/77db2b2c-5d03-dbb0-aeac-c4dd92129ab9@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Implement a --mmap-flush option that specifies minimal number of bytes
that is extracted from mmaped kernel buffer to store into a trace. The
default option value is 1 byte what means every time trace writing
thread finds some new data in the mmaped buffer the data is extracted,
possibly compressed and written to a trace.
$ tools/perf/perf record --mmap-flush 1024 -e cycles -- matrix.gcc
$ tools/perf/perf record --aio --mmap-flush 1K -e cycles -- matrix.gcc
The option is independent from -z setting, doesn't vary with compression
level and can serve two purposes.
The first purpose is to increase the compression ratio of a trace data.
Larger data chunks are compressed more effectively so the implemented
option allows specifying data chunk size to compress. Also at some cases
executing more write syscalls with smaller data size can take longer
than executing less write syscalls with bigger data size due to syscall
overhead so extracting bigger data chunks specified by the option value
could additionally decrease runtime overhead.
The second purpose is to avoid self monitoring live-lock issue in system
wide (-a) profiling mode. Profiling in system wide mode with compression
(-a -z) can additionally induce data into the kernel buffers along with
the data from monitored processes. If performance data rate and volume
from the monitored processes is high then trace streaming and
compression activity in the tool is also high. High tool process
activity can lead to subtle live-lock effect when compression of single
new byte from some of mmaped kernel buffer leads to generation of the
next single byte at some mmaped buffer. So perf tool process ends up in
endless self monitoring.
Implemented synch parameter is the mean to force data move independently
from the specified flush threshold value. Despite the provided flush
value the tool needs capability to unconditionally drain memory buffers,
at least in the end of the collection.
Committer testing:
Running with the default value, i.e. as soon as there is something to
read go on consuming, we first write the synthesized events, small
chunks of about 128 bytes:
# perf trace -m 2048 --call-graph dwarf -e write -- perf record
<SNIP>
101.142 ( 0.004 ms): perf/25821 write(fd: 3</root/perf.data>, buf: 0x210db60, count: 120) = 120
__libc_write (/usr/lib64/libpthread-2.28.so)
ion (/home/acme/bin/perf)
record__write (inlined)
process_synthesized_event (/home/acme/bin/perf)
perf_tool__process_synth_event (inlined)
perf_event__synthesize_mmap_events (/home/acme/bin/perf)
Then we move to reading the mmap buffers consuming the events put there
by the kernel perf infrastructure:
107.561 ( 0.005 ms): perf/25821 write(fd: 3</root/perf.data>, buf: 0x7f1befc02000, count: 336) = 336
__libc_write (/usr/lib64/libpthread-2.28.so)
ion (/home/acme/bin/perf)
record__write (inlined)
record__pushfn (/home/acme/bin/perf)
perf_mmap__push (/home/acme/bin/perf)
record__mmap_read_evlist (inlined)
record__mmap_read_all (inlined)
__cmd_record (inlined)
cmd_record (/home/acme/bin/perf)
12919.953 ( 0.136 ms): perf/25821 write(fd: 3</root/perf.data>, buf: 0x7f1befc83150, count: 184984) = 184984
<SNIP same backtrace as in the 107.561 timestamp>
12920.094 ( 0.155 ms): perf/25821 write(fd: 3</root/perf.data>, buf: 0x7f1befc02150, count: 261816) = 261816
<SNIP same backtrace as in the 107.561 timestamp>
12920.253 ( 0.093 ms): perf/25821 write(fd: 3</root/perf.data>, buf: 0x7f1befb81120, count: 170832) = 170832
<SNIP same backtrace as in the 107.561 timestamp>
If we limit it to write only when more than 16MB are available for
reading, it throttles that to a quarter of the --mmap-pages set for
'perf record', which by default get to 528384 bytes, found out using
'record -v':
mmap flush: 132096
mmap size 528384B
With that in place all the writes coming from
record__mmap_read_evlist(), i.e. from the mmap buffers setup by the
kernel perf infrastructure were at least 132096 bytes long.
Trying with a bigger mmap size:
perf trace -e write perf record -v -m 2048 --mmap-flush 16M
74982.928 ( 2.471 ms): perf/26500 write(fd: 3</root/perf.data>, buf: 0x7ff94a6cc000, count: 3580888) = 3580888
74985.406 ( 2.353 ms): perf/26500 write(fd: 3</root/perf.data>, buf: 0x7ff949ecb000, count: 3453256) = 3453256
74987.764 ( 2.629 ms): perf/26500 write(fd: 3</root/perf.data>, buf: 0x7ff9496ca000, count: 3859232) = 3859232
74990.399 ( 2.341 ms): perf/26500 write(fd: 3</root/perf.data>, buf: 0x7ff948ec9000, count: 3769032) = 3769032
74992.744 ( 2.064 ms): perf/26500 write(fd: 3</root/perf.data>, buf: 0x7ff9486c8000, count: 3310520) = 3310520
74994.814 ( 2.619 ms): perf/26500 write(fd: 3</root/perf.data>, buf: 0x7ff947ec7000, count: 4194688) = 4194688
74997.439 ( 2.787 ms): perf/26500 write(fd: 3</root/perf.data>, buf: 0x7ff9476c6000, count: 4029760) = 4029760
Was again limited to a quarter of the mmap size:
mmap flush: 2098176
mmap size 8392704B
A warning about that would be good to have but can be added later,
something like:
"max flush is a quarter of the mmap size, if wanting to bump the mmap
flush further, bump the mmap size as well using -m/--mmap-pages"
Also rename the 'sync' parameters to 'synch' to keep tools/perf building
with older glibcs:
cc1: warnings being treated as errors
builtin-record.c: In function 'record__mmap_read_evlist':
builtin-record.c:775: warning: declaration of 'sync' shadows a global declaration
/usr/include/unistd.h:933: warning: shadowed declaration is here
builtin-record.c: In function 'record__mmap_read_all':
builtin-record.c:856: warning: declaration of 'sync' shadows a global declaration
/usr/include/unistd.h:933: warning: shadowed declaration is here
Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
Reviewed-by: Jiri Olsa <jolsa@kernel.org>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/f6600d72-ecfa-2eb7-7e51-f6954547d500@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Multi AIO trace writing allows caching more kernel data into userspace
memory postponing trace writing for the sake of overall profiling data
thruput increase. It could be seen as kernel data buffer extension into
userspace memory.
With an --aio option value different from 0 (default value is 1) the
tool has capability to cache more and more data into user space along
with delegating spill to AIO.
That allows avoiding to suspend at record__aio_sync() between calls of
record__mmap_read_evlist() and increases profiling data thruput at the
cost of userspace memory.
Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
Reviewed-by: Jiri Olsa <jolsa@redhat.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/050bb053-e7f3-aa83-fde7-f27ff90be7f6@linux.intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Pull networking updates from David Miller:
1) Add VF IPSEC offload support in ixgbe, from Shannon Nelson.
2) Add zero-copy AF_XDP support to i40e, from Björn Töpel.
3) All in-tree drivers are converted to {g,s}et_link_ksettings() so we
can get rid of the {g,s}et_settings ethtool callbacks, from Michal
Kubecek.
4) Add software timestamping to veth driver, from Michael Walle.
5) More work to make packet classifiers and actions lockless, from Vlad
Buslov.
6) Support sticky FDB entries in bridge, from Nikolay Aleksandrov.
7) Add ipv6 version of IP_MULTICAST_ALL sockopt, from Andre Naujoks.
8) Support batching of XDP buffers in vhost_net, from Jason Wang.
9) Add flow dissector BPF hook, from Petar Penkov.
10) i40e vf --> generic iavf conversion, from Jesse Brandeburg.
11) Add NLA_REJECT netlink attribute policy type, to signal when users
provide attributes in situations which don't make sense. From
Johannes Berg.
12) Switch TCP and fair-queue scheduler over to earliest departure time
model. From Eric Dumazet.
13) Improve guest receive performance by doing rx busy polling in tx
path of vhost networking driver, from Tonghao Zhang.
14) Add per-cgroup local storage to bpf
15) Add reference tracking to BPF, from Joe Stringer. The verifier can
now make sure that references taken to objects are properly released
by the program.
16) Support in-place encryption in TLS, from Vakul Garg.
17) Add new taprio packet scheduler, from Vinicius Costa Gomes.
18) Lots of selftests additions, too numerous to mention one by one here
but all of which are very much appreciated.
19) Support offloading of eBPF programs containing BPF to BPF calls in
nfp driver, frm Quentin Monnet.
20) Move dpaa2_ptp driver out of staging, from Yangbo Lu.
21) Lots of u32 classifier cleanups and simplifications, from Al Viro.
22) Add new strict versions of netlink message parsers, and enable them
for some situations. From David Ahern.
23) Evict neighbour entries on carrier down, also from David Ahern.
24) Support BPF sk_msg verdict programs with kTLS, from Daniel Borkmann
and John Fastabend.
25) Add support for filtering route dumps, from David Ahern.
26) New igc Intel driver for 2.5G parts, from Sasha Neftin et al.
27) Allow vxlan enslavement to bridges in mlxsw driver, from Ido
Schimmel.
28) Add queue and stack map types to eBPF, from Mauricio Vasquez B.
29) Add back byte-queue-limit support to r8169, with all the bug fixes
in other areas of the driver it works now! From Florian Westphal and
Heiner Kallweit.
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (2147 commits)
tcp: add tcp_reset_xmit_timer() helper
qed: Fix static checker warning
Revert "be2net: remove desc field from be_eq_obj"
Revert "net: simplify sock_poll_wait"
net: socionext: Reset tx queue in ndo_stop
net: socionext: Add dummy PHY register read in phy_write()
net: socionext: Stop PHY before resetting netsec
net: stmmac: Set OWN bit for jumbo frames
arm64: dts: stratix10: Support Ethernet Jumbo frame
tls: Add maintainers
net: ethernet: ti: cpsw: unsync mcast entries while switch promisc mode
octeontx2-af: Support for NIXLF's UCAST/PROMISC/ALLMULTI modes
octeontx2-af: Support for setting MAC address
octeontx2-af: Support for changing RSS algorithm
octeontx2-af: NIX Rx flowkey configuration for RSS
octeontx2-af: Install ucast and bcast pkt forwarding rules
octeontx2-af: Add LMAC channel info to NIXLF_ALLOC response
octeontx2-af: NPC MCAM and LDATA extract minimal configuration
octeontx2-af: Enable packet length and csum validation
octeontx2-af: Support for VTAG strip and capture
...
Currently, on x86-64, perf uses LFENCE and MFENCE (rmb() and mb(),
respectively) when processing events from the perf ring buffer which
is unnecessarily expensive as we can do more lightweight in particular
given this is critical fast-path in perf.
According to Peter rmb()/mb() were added back then via a94d342b9c
("tools/perf: Add required memory barriers") at a time where kernel
still supported chips that needed it, but nowadays support for these
has been ditched completely, therefore we can fix them up as well.
While for x86-64, replacing rmb() and mb() with smp_*() variants would
result in just a compiler barrier for the former and LOCK + ADD for
the latter (__sync_synchronize() uses slower MFENCE by the way), Peter
suggested we can use smp_{load_acquire,store_release}() instead for
architectures where its implementation doesn't resolve in slower smp_mb().
Thus, e.g. in x86-64 we would be able to avoid CPU barrier entirely due
to TSO. For architectures where the latter needs to use smp_mb() e.g.
on arm, we stick to cheaper smp_rmb() variant for fetching the head.
This work adds helpers ring_buffer_read_head() and ring_buffer_write_tail()
for tools infrastructure that either switches to smp_load_acquire() for
architectures where it is cheaper or uses READ_ONCE() + smp_rmb() barrier
for those where it's not in order to fetch the data_head from the perf
control page, and it uses smp_store_release() to write the data_tail.
Latter is smp_mb() + WRITE_ONCE() combination or a cheaper variant if
architecture allows for it. Those that rely on smp_rmb() and smp_mb() can
further improve performance in a follow up step by implementing the two
under tools/arch/*/include/asm/barrier.h such that they don't have to
fallback to rmb() and mb() in tools/include/asm/barrier.h.
Switch perf to use ring_buffer_read_head() and ring_buffer_write_tail()
so it can make use of the optimizations. Later, we convert libbpf as
well to use the same helpers.
Side note [0]: the topic has been raised of whether one could simply use
the C11 gcc builtins [1] for the smp_load_acquire() and smp_store_release()
instead:
__atomic_load_n(ptr, __ATOMIC_ACQUIRE);
__atomic_store_n(ptr, val, __ATOMIC_RELEASE);
Kernel and (presumably) tooling shipped along with the kernel has a
minimum requirement of being able to build with gcc-4.6 and the latter
does not have C11 builtins. While generally the C11 memory models don't
align with the kernel's, the C11 load-acquire and store-release alone
/could/ suffice, however. Issue is that this is implementation dependent
on how the load-acquire and store-release is done by the compiler and
the mapping of supported compilers must align to be compatible with the
kernel's implementation, and thus needs to be verified/tracked on a
case by case basis whether they match (unless an architecture uses them
also from kernel side). The implementations for smp_load_acquire() and
smp_store_release() in this patch have been adapted from the kernel side
ones to have a concrete and compatible mapping in place.
[0] http://patchwork.ozlabs.org/patch/985422/
[1] https://gcc.gnu.org/onlinedocs/gcc/_005f_005fatomic-Builtins.html
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Except for 'perf record', the other perf tools read events one by one
from the ring buffer using perf_mmap__read_forward(). But it only
supports non-overwrite mode.
Introduce perf_mmap__read_event() to support both non-overwrite and
overwrite mode.
Usage:
perf_mmap__read_init()
while(event = perf_mmap__read_event()) {
//process the event
perf_mmap__consume()
}
perf_mmap__read_done()
It cannot use perf_mmap__read_backward(). Because it always reads the
stale buffer which is already processed. Furthermore, the forward and
backward concepts have been removed. The perf_mmap__read_backward() will
be replaced and discarded later.
Signed-off-by: Kan Liang <kan.liang@intel.com>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Jin Yao <yao.jin@linux.intel.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Wang Nan <wangnan0@huawei.com>
Link: http://lkml.kernel.org/r/1516310792-208685-9-git-send-email-kan.liang@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
