user/sven/linux.git/include/uapi/linux/perf_event.h, branch v4.9.309

perf core: Per event callchain limit

2016-05-30T15:41:44Z

Additionally to being able to control the system wide maximum depth via /proc/sys/kernel/perf_event_max_stack, now we are able to ask for different depths per event, using perf_event_attr.sample_max_stack for that. This uses an u16 hole at the end of perf_event_attr, that, when perf_event_attr.sample_type has the PERF_SAMPLE_CALLCHAIN, if sample_max_stack is zero, means use perf_event_max_stack, otherwise it'll be bounds checked under callchain_mutex. Cc: Adrian Hunter Cc: Alexander Shishkin Cc: Alexei Starovoitov Cc: Brendan Gregg Cc: David Ahern Cc: Frederic Weisbecker Cc: He Kuang Cc: Jiri Olsa Cc: Linus Torvalds Cc: Masami Hiramatsu Cc: Milian Wolff Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Cc: Wang Nan Cc: Zefan Li Link: http://lkml.kernel.org/n/tip-kolmn1yo40p7jhswxwrc7rrd@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo

perf core: Separate accounting of contexts and real addresses in a stack trace

2016-05-17T02:11:53Z

The perf_sample->ip_callchain->nr value includes all the entries in the ip_callchain->ip[] array, real addresses and PERF_CONTEXT_{KERNEL,USER,etc}, while what the user expects is that what is in the kernel.perf_event_max_stack sysctl or in the upcoming per event perf_event_attr.sample_max_stack knob be honoured in terms of IP addresses in the stack trace. So allocate a bunch of extra entries for contexts, and do the accounting via perf_callchain_entry_ctx struct members. A new sysctl, kernel.perf_event_max_contexts_per_stack is also introduced for investigating possible bugs in the callchain implementation by some arch. Cc: Adrian Hunter Cc: Alexander Shishkin Cc: Alexei Starovoitov Cc: Brendan Gregg Cc: David Ahern Cc: Frederic Weisbecker Cc: He Kuang Cc: Jiri Olsa Cc: Masami Hiramatsu Cc: Milian Wolff Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Cc: Wang Nan Cc: Zefan Li Link: http://lkml.kernel.org/n/tip-3b4wnqk340c4sg4gwkfdi9yk@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo

perf/core: Add ::write_backward attribute to perf event

2016-04-23T12:12:39Z

This patch introduces 'write_backward' bit to perf_event_attr, which controls the direction of a ring buffer. After set, the corresponding ring buffer is written from end to beginning. This feature is design to support reading from overwritable ring buffer. Ring buffer can be created by mapping a perf event fd. Kernel puts event records into ring buffer, user tooling like perf fetch them from address returned by mmap(). To prevent racing between kernel and tooling, they communicate to each other through 'head' and 'tail' pointers. Kernel maintains 'head' pointer, points it to the next free area (tail of the last record). Tooling maintains 'tail' pointer, points it to the tail of last consumed record (record has already been fetched). Kernel determines the available space in a ring buffer using these two pointers to avoid overwrite unfetched records. By mapping without 'PROT_WRITE', an overwritable ring buffer is created. Different from normal ring buffer, tooling is unable to maintain 'tail' pointer because writing is forbidden. Therefore, for this type of ring buffers, kernel overwrite old records unconditionally, works like flight recorder. This feature would be useful if reading from overwritable ring buffer were as easy as reading from normal ring buffer. However, there's an obscure problem. The following figure demonstrates a full overwritable ring buffer. In this figure, the 'head' pointer points to the end of last record, and a long record 'E' is pending. For a normal ring buffer, a 'tail' pointer would have pointed to position (X), so kernel knows there's no more space in the ring buffer. However, for an overwritable ring buffer, kernel ignore the 'tail' pointer. (X) head . | . V +------+-------+----------+------+---+ |A....A|B.....B|C........C|D....D| | +------+-------+----------+------+---+ Record 'A' is overwritten by event 'E': head | V +--+---+-------+----------+------+---+ |.E|..A|B.....B|C........C|D....D|E..| +--+---+-------+----------+------+---+ Now tooling decides to read from this ring buffer. However, none of these two natural positions, 'head' and the start of this ring buffer, are pointing to the head of a record. Even the full ring buffer can be accessed by tooling, it is unable to find a position to start decoding. The first attempt tries to solve this problem AFAIK can be found from [1]. It makes kernel to maintain 'tail' pointer: updates it when ring buffer is half full. However, this approach introduces overhead to fast path. Test result shows a 1% overhead [2]. In addition, this method utilizes no more tham 50% records. Another attempt can be found from [3], which allows putting the size of an event at the end of each record. This approach allows tooling to find records in a backward manner from 'head' pointer by reading size of a record from its tail. However, because of alignment requirement, it needs 8 bytes to record the size of a record, which is a huge waste. Its performance is also not good, because more data need to be written. This approach also introduces some extra branch instructions to fast path. 'write_backward' is a better solution to this problem. Following figure demonstrates the state of the overwritable ring buffer when 'write_backward' is set before overwriting: head | V +---+------+----------+-------+------+ | |D....D|C........C|B.....B|A....A| +---+------+----------+-------+------+ and after overwriting: head | V +---+------+----------+-------+---+--+ |..E|D....D|C........C|B.....B|A..|E.| +---+------+----------+-------+---+--+ In each situation, 'head' points to the beginning of the newest record. From this record, tooling can iterate over the full ring buffer and fetch records one by one. The only limitation that needs to be considered is back-to-back reading. Due to the non-deterministic of user programs, it is impossible to ensure the ring buffer keeps stable during reading. Consider an extreme situation: tooling is scheduled out after reading record 'D', then a burst of events come, eat up the whole ring buffer (one or multiple rounds). When the tooling process comes back, reading after 'D' is incorrect now. To prevent this problem, we need to find a way to ensure the ring buffer is stable during reading. ioctl(PERF_EVENT_IOC_PAUSE_OUTPUT) is suggested because its overhead is lower than ioctl(PERF_EVENT_IOC_ENABLE). By carefully verifying 'header' pointer, reader can avoid pausing the ring-buffer. For example: /* A union of all possible events */ union perf_event event; p = head = perf_mmap__read_head(); while (true) { /* copy header of next event */ fetch(&event.header, p, sizeof(event.header)); /* read 'head' pointer */ head = perf_mmap__read_head(); /* check overwritten: is the header good? */ if (!verify(sizeof(event.header), p, head)) break; /* copy the whole event */ fetch(&event, p, event.header.size); /* read 'head' pointer again */ head = perf_mmap__read_head(); /* is the whole event good? */ if (!verify(event.header.size, p, head)) break; p += event.header.size; } However, the overhead is high because: a) In-place decoding is not safe. Copying-verifying-decoding is required. b) Fetching 'head' pointer requires additional synchronization. (From Alexei Starovoitov: Even when this trick works, pause is needed for more than stability of reading. When we collect the events into overwrite buffer we're waiting for some other trigger (like all cpu utilization spike or just one cpu running and all others are idle) and when it happens the buffer has valuable info from the past. At this point new events are no longer interesting and buffer should be paused, events read and unpaused until next trigger comes.) This patch utilizes event's default overflow_handler introduced previously. perf_event_output_backward() is created as the default overflow handler for backward ring buffers. To avoid extra overhead to fast path, original perf_event_output() becomes __perf_event_output() and marked '__always_inline'. In theory, there's no extra overhead introduced to fast path. Performance testing: Calling 3000000 times of 'close(-1)', use gettimeofday() to check duration. Use 'perf record -o /dev/null -e raw_syscalls:*' to capture system calls. In ns. Testing environment: CPU : Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz Kernel : v4.5.0 MEAN STDVAR BASE 800214.950 2853.083 PRE1 2253846.700 9997.014 PRE2 2257495.540 8516.293 POST 2250896.100 8933.921 Where 'BASE' is pure performance without capturing. 'PRE1' is test result of pure 'v4.5.0' kernel. 'PRE2' is test result before this patch. 'POST' is test result after this patch. See [4] for the detailed experimental setup. Considering the stdvar, this patch doesn't introduce performance overhead to the fast path. [1] http://lkml.iu.edu/hypermail/linux/kernel/1304.1/04584.html [2] http://lkml.iu.edu/hypermail/linux/kernel/1307.1/00535.html [3] http://lkml.iu.edu/hypermail/linux/kernel/1512.0/01265.html [4] http://lkml.kernel.org/g/56F89DCD.1040202@huawei.com Signed-off-by: Wang Nan Signed-off-by: Peter Zijlstra (Intel) Acked-by: Alexei Starovoitov Cc: Cc: Cc: Alexander Shishkin Cc: Arnaldo Carvalho de Melo Cc: Brendan Gregg Cc: He Kuang Cc: Jiri Olsa Cc: Jiri Olsa Cc: Linus Torvalds Cc: Masami Hiramatsu Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Cc: Zefan Li Link: http://lkml.kernel.org/r/1459865478-53413-1-git-send-email-wangnan0@huawei.com [ Fixed the changelog some more. ] Signed-off-by: Ingo Molnar Signed-off-by: Ingo Molnar

perf/ring_buffer: Introduce new ioctl options to pause and resume the ring-buffer

2016-03-31T08:30:45Z

Add new ioctl() to pause/resume ring-buffer output. In some situations we want to read from the ring-buffer only when we ensure nothing can write to the ring-buffer during reading. Without this patch we have to turn off all events attached to this ring-buffer to achieve this. This patch is a prerequisite to enable overwrite support for the perf ring-buffer support. Following commits will introduce new methods support reading from overwrite ring buffer. Before reading, caller must ensure the ring buffer is frozen, or the reading is unreliable. Signed-off-by: Wang Nan Signed-off-by: Peter Zijlstra (Intel) Cc: Cc: Alexander Shishkin Cc: Alexei Starovoitov Cc: Arnaldo Carvalho de Melo Cc: Brendan Gregg Cc: He Kuang Cc: Jiri Olsa Cc: Jiri Olsa Cc: Linus Torvalds Cc: Masami Hiramatsu Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Cc: Zefan Li Link: http://lkml.kernel.org/r/1459147292-239310-2-git-send-email-wangnan0@huawei.com Signed-off-by: Ingo Molnar

perf/x86: Add option to disable reading branch flags/cycles

2015-11-23T08:58:25Z

With LBRv5 reading the extra LBR flags like mispredict, TSX, cycles is not free anymore, as it has moved to a separate MSR. For callstack mode we don't need any of this information; so we can avoid the unnecessary MSR read. Add flags to the perf interface where perf record can request not collecting this information. Add branch_sample_type flags for CYCLES and FLAGS. It's a bit unusual for branch_sample_types to be negative (disable), not positive (enable), but since the legacy ABI reported the flags we need some form of explicit disabling to avoid breaking the ABI. After we have the flags the x86 perf code can keep track if any users need the flags. If noone needs it the information is not collected. This cuts down the cost of LBR callstack on Skylake significantly. Profiling a kernel build with LBR call stack the average run time of the PMI handler drops by 43%. Signed-off-by: Andi Kleen Signed-off-by: Peter Zijlstra (Intel) Cc: Arnaldo Carvalho de Melo Cc: Jiri Olsa Cc: Linus Torvalds Cc: Mike Galbraith Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Cc: acme@kernel.org Cc: jolsa@kernel.org Link: http://lkml.kernel.org/r/1445366797-30894-2-git-send-email-andi@firstfloor.org Signed-off-by: Ingo Molnar

Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next

2015-11-04T17:41:05Z

Pull networking updates from David Miller: Changes of note: 1) Allow to schedule ICMP packets in IPVS, from Alex Gartrell. 2) Provide FIB table ID in ipv4 route dumps just as ipv6 does, from David Ahern. 3) Allow the user to ask for the statistics to be filtered out of ipv4/ipv6 address netlink dumps. From Sowmini Varadhan. 4) More work to pass the network namespace context around deep into various packet path APIs, starting with the netfilter hooks. From Eric W Biederman. 5) Add layer 2 TX/RX checksum offloading to qeth driver, from Thomas Richter. 6) Use usec resolution for SYN/ACK RTTs in TCP, from Yuchung Cheng. 7) Support Very High Throughput in wireless MESH code, from Bob Copeland. 8) Allow setting the ageing_time in switchdev/rocker. From Scott Feldman. 9) Properly autoload L2TP type modules, from Stephen Hemminger. 10) Fix and enable offload features by default in 8139cp driver, from David Woodhouse. 11) Support both ipv4 and ipv6 sockets in a single vxlan device, from Jiri Benc. 12) Fix CWND limiting of thin streams in TCP, from Bendik Rønning Opstad. 13) Fix IPSEC flowcache overflows on large systems, from Steffen Klassert. 14) Convert bridging to track VLANs using rhashtable entries rather than a bitmap. From Nikolay Aleksandrov. 15) Make TCP listener handling completely lockless, this is a major accomplishment. Incoming request sockets now live in the established hash table just like any other socket too. From Eric Dumazet. 15) Provide more bridging attributes to netlink, from Nikolay Aleksandrov. 16) Use hash based algorithm for ipv4 multipath routing, this was very long overdue. From Peter Nørlund. 17) Several y2038 cures, mostly avoiding timespec. From Arnd Bergmann. 18) Allow non-root execution of EBPF programs, from Alexei Starovoitov. 19) Support SO_INCOMING_CPU as setsockopt, from Eric Dumazet. This influences the port binding selection logic used by SO_REUSEPORT. 20) Add ipv6 support to VRF, from David Ahern. 21) Add support for Mellanox Spectrum switch ASIC, from Jiri Pirko. 22) Add rtl8xxxu Realtek wireless driver, from Jes Sorensen. 23) Implement RACK loss recovery in TCP, from Yuchung Cheng. 24) Support multipath routes in MPLS, from Roopa Prabhu. 25) Fix POLLOUT notification for listening sockets in AF_UNIX, from Eric Dumazet. 26) Add new QED Qlogic river, from Yuval Mintz, Manish Chopra, and Sudarsana Kalluru. 27) Don't fetch timestamps on AF_UNIX sockets, from Hannes Frederic Sowa. 28) Support ipv6 geneve tunnels, from John W Linville. 29) Add flood control support to switchdev layer, from Ido Schimmel. 30) Fix CHECKSUM_PARTIAL handling of potentially fragmented frames, from Hannes Frederic Sowa. 31) Support persistent maps and progs in bpf, from Daniel Borkmann. * git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1790 commits) sh_eth: use DMA barriers switchdev: respect SKIP_EOPNOTSUPP flag in case there is no recursion net: sched: kill dead code in sch_choke.c irda: Delete an unnecessary check before the function call "irlmp_unregister_service" net: dsa: mv88e6xxx: include DSA ports in VLANs net: dsa: mv88e6xxx: disable SA learning for DSA and CPU ports net/core: fix for_each_netdev_feature vlan: Invoke driver vlan hooks only if device is present arcnet/com20020: add LEDS_CLASS dependency bpf, verifier: annotate verbose printer with __printf dp83640: Only wait for timestamps for packets with timestamping enabled. ptp: Change ptp_class to a proper bitmask dp83640: Prune rx timestamp list before reading from it dp83640: Delay scheduled work. dp83640: Include hash in timestamp/packet matching ipv6: fix tunnel error handling net/mlx5e: Fix LSO vlan insertion net/mlx5e: Re-eanble client vlan TX acceleration net/mlx5e: Return error in case mlx5e_set_features() fails net/mlx5e: Don't allow more than max supported channels ...

bpf: introduce bpf_perf_event_output() helper

2015-10-22T13:42:15Z

This helper is used to send raw data from eBPF program into special PERF_TYPE_SOFTWARE/PERF_COUNT_SW_BPF_OUTPUT perf_event. User space needs to perf_event_open() it (either for one or all cpus) and store FD into perf_event_array (similar to bpf_perf_event_read() helper) before eBPF program can send data into it. Today the programs triggered by kprobe collect the data and either store it into the maps or print it via bpf_trace_printk() where latter is the debug facility and not suitable to stream the data. This new helper replaces such bpf_trace_printk() usage and allows programs to have dedicated channel into user space for post-processing of the raw data collected. Signed-off-by: Alexei Starovoitov Signed-off-by: David S. Miller

perf: Add PERF_SAMPLE_BRANCH_CALL

2015-10-20T08:30:53Z

Add a new branch sample type to cover only call branches (function calls). The current ANY_CALL included direct, indirect calls and far jumps. We want to be able to differentiate indirect from direct calls. Therefore we introduce PERF_SAMPLE_BRANCH_CALL. The implementation is up to each architecture. Signed-off-by: Stephane Eranian Signed-off-by: Peter Zijlstra (Intel) Cc: Arnaldo Carvalho de Melo Cc: David Ahern Cc: Jiri Olsa Cc: Linus Torvalds Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: Vince Weaver Cc: khandual@linux.vnet.ibm.com Link: http://lkml.kernel.org/r/1444720151-10275-2-git-send-email-eranian@google.com Signed-off-by: Ingo Molnar

perf/x86: Fix time_shift in perf_event_mmap_page

2015-10-20T08:30:52Z

Commit: b20112edeadf ("perf/x86: Improve accuracy of perf/sched clock") allowed the time_shift value in perf_event_mmap_page to be as much as 32. Unfortunately the documented algorithms for using time_shift have it shifting an integer, whereas to work correctly with the value 32, the type must be u64. In the case of perf tools, Intel PT decodes correctly but the timestamps that are output (for example by perf script) have lost 32-bits of granularity so they look like they are not changing at all. Fix by limiting the shift to 31 and adjusting the multiplier accordingly. Also update the documentation of perf_event_mmap_page so that new code based on it will be more future-proof. Signed-off-by: Adrian Hunter Signed-off-by: Peter Zijlstra (Intel) Cc: Andy Lutomirski Cc: Arnaldo Carvalho de Melo Cc: Arnaldo Carvalho de Melo Cc: David Ahern Cc: Jiri Olsa Cc: Linus Torvalds Cc: Namhyung Kim Cc: Peter Zijlstra Cc: Stephane Eranian Cc: Thomas Gleixner Cc: Vince Weaver Fixes: b20112edeadf ("perf/x86: Improve accuracy of perf/sched clock") Link: http://lkml.kernel.org/r/1445001845-13688-2-git-send-email-adrian.hunter@intel.com Signed-off-by: Ingo Molnar

perf: Add cycles to branch_info

2015-08-04T08:16:57Z

Intel Skylake supports reporting the time in cycles a branch in the LBR took, to give a rough indication of the basic block performance. Export the cycle information in the branch_info structure. This can be done by just reusing some currently zero padding. This is just the generic header change. The architecture still needs to fill it in. There's no attempt to convert to real time, as we really want cycles here. Signed-off-by: Andi Kleen Signed-off-by: Peter Zijlstra (Intel) Cc: Linus Torvalds Cc: Peter Zijlstra Cc: Thomas Gleixner Cc: eranian@google.com Link: http://lkml.kernel.org/r/1431285767-27027-5-git-send-email-andi@firstfloor.org Signed-off-by: Ingo Molnar