- 02 Apr, 2015 18 commits
-
-
Alexander Shishkin authored
For pmus that wish to write data to ring buffer's AUX area, provide perf_aux_output_{begin,end}() calls to initiate/commit data writes, similarly to perf_output_{begin,end}. These also use the same output handle structure. Also, similarly to software counterparts, these will direct inherited events' output to parents' ring buffers. After the perf_aux_output_begin() returns successfully, handle->size is set to the maximum amount of data that can be written wrt aux_tail pointer, so that no data that the user hasn't seen will be overwritten, therefore this should always be called before hardware writing is enabled. On success, this will return the pointer to pmu driver's private structure allocated for this aux area by pmu::setup_aux. Same pointer can also be retrieved using perf_get_aux() while hardware writing is enabled. PMU driver should pass the actual amount of data written as a parameter to perf_aux_output_end(). All hardware writes should be completed and visible before this one is called. Additionally, perf_aux_output_skip() will adjust output handle and aux_head in case some part of the buffer has to be skipped over to maintain hardware's alignment constraints. Nested writers are forbidden and guards are in place to catch such attempts. Signed-off-by:Alexander Shishkin <alexander.shishkin@linux.intel.com> Signed-off-by:
Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Borislav Petkov <bp@alien8.de> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Kaixu Xia <kaixu.xia@linaro.org> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Mike Galbraith <efault@gmx.de> Cc: Paul Mackerras <paulus@samba.org> Cc: Robert Richter <rric@kernel.org> Cc: Stephane Eranian <eranian@google.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: acme@infradead.org Cc: adrian.hunter@intel.com Cc: kan.liang@intel.com Cc: markus.t.metzger@intel.com Cc: mathieu.poirier@linaro.org Link: http://lkml.kernel.org/r/1421237903-181015-8-git-send-email-alexander.shishkin@linux.intel.comSigned-off-by:
Ingo Molnar <mingo@kernel.org>
-
Alexander Shishkin authored
When there's new data in the AUX space, output a record indicating its offset and size and a set of flags, such as PERF_AUX_FLAG_TRUNCATED, to mean the described data was truncated to fit in the ring buffer. Signed-off-by:
-
Alexander Shishkin authored
Usually, pmus that do, for example, instruction tracing, would only ever be able to have one event per task per cpu (or per perf_event_context). For such pmus it makes sense to disallow creating conflicting events early on, so as to provide consistent behavior for the user. This patch adds a pmu capability that indicates such constraint on event creation. Signed-off-by:
-
Alexander Shishkin authored
For pmus that don't support scatter-gather for AUX data in hardware, it might still make sense to implement software double buffering to avoid losing data while the user is reading data out. For this purpose, add a pmu capability that guarantees multiple high-order chunks for AUX buffer, so that the pmu driver can do switchover tricks. To make use of this feature, add PERF_PMU_CAP_AUX_SW_DOUBLEBUF to your pmu's capability mask. This will make the ring buffer AUX allocation code ensure that the biggest high order allocation for the aux buffer pages is no bigger than half of the total requested buffer size, thus making sure that the buffer has at least two high order allocations. Signed-off-by:
-
Alexander Shishkin authored
Some pmus (such as BTS or Intel PT without multiple-entry ToPA capability) don't support scatter-gather and will prefer larger contiguous areas for their output regions. This patch adds a new pmu capability to request higher order allocations. Signed-off-by:
-
Peter Zijlstra authored
This patch introduces "AUX space" in the perf mmap buffer, intended for exporting high bandwidth data streams to userspace, such as instruction flow traces. AUX space is a ring buffer, defined by aux_{offset,size} fields in the user_page structure, and read/write pointers aux_{head,tail}, which abide by the same rules as data_* counterparts of the main perf buffer. In order to allocate/mmap AUX, userspace needs to set up aux_offset to such an offset that will be greater than data_offset+data_size and aux_size to be the desired buffer size. Both need to be page aligned. Then, same aux_offset and aux_size should be passed to mmap() call and if everything adds up, you should have an AUX buffer as a result. Pages that are mapped into this buffer also come out of user's mlock rlimit plus perf_event_mlock_kb allowance. Signed-off-by: -
Alexander Shishkin authored
Currently, the actual perf ring buffer is one page into the mmap area, following the user page and the userspace follows this convention. This patch adds data_{offset,size} fields to user_page that can be used by userspace instead for locating perf data in the mmap area. This is also helpful when mapping existing or shared buffers if their size is not known in advance. Right now, it is made to follow the existing convention that data_offset == PAGE_SIZE and data_offset + data_size == mmap_size. Signed-off-by: -
Ingo Molnar authored
So bpf_tracing.o depends on CONFIG_BPF_SYSCALL - but that's not its only dependency, it also depends on the tracing infrastructure and on kprobes, without which it will fail to build with: In file included from kernel/trace/bpf_trace.c:14:0: kernel/trace/trace.h: In function ‘trace_test_and_set_recursion’: kernel/trace/trace.h:491:28: error: ‘struct task_struct’ has no member named ‘trace_recursion’ unsigned int val = current->trace_recursion; [...] It took quite some time to trigger this build failure, because right now BPF_SYSCALL is very obscure, depends on CONFIG_EXPERT. So also make BPF_SYSCALL more configurable, not just under CONFIG_EXPERT. If BPF_SYSCALL, tracing and kprobes are enabled then enable the bpf_tracing gateway as well. We might want to make this an interactive option later on, although I'd not complicate it unnecessarily: enabling BPF_SYSCALL is enough of an indicator that the user wants BPF support. Cc: Alexei Starovoitov <ast@plumgrid.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: David S. Miller <davem@davemloft.net> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: linux-kernel@vger.kernel.org Signed-off-by: -
Alexei Starovoitov authored
One BPF program attaches to kmem_cache_alloc_node() and remembers all allocated objects in the map. Another program attaches to kmem_cache_free() and deletes corresponding object from the map. User space walks the map every second and prints any objects which are older than 1 second. Usage: $ sudo tracex4 Then start few long living processes. The 'tracex4' will print something like this: obj 0xffff880465928000 is 13sec old was allocated at ip ffffffff8105dc32 obj 0xffff88043181c280 is 13sec old was allocated at ip ffffffff8105dc32 obj 0xffff880465848000 is 8sec old was allocated at ip ffffffff8105dc32 obj 0xffff8804338bc280 is 15sec old was allocated at ip ffffffff8105dc32 $ addr2line -fispe vmlinux ffffffff8105dc32 do_fork at fork.c:1665 As soon as processes exit the memory is reclaimed and 'tracex4' prints nothing. Similar experiment can be done with the __kmalloc()/kfree() pair. Signed-off-by:
Alexei Starovoitov <ast@plumgrid.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: David S. Miller <davem@davemloft.net> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Steven Rostedt <rostedt@goodmis.org> Link: http://lkml.kernel.org/r/1427312966-8434-10-git-send-email-ast@plumgrid.comSigned-off-by:
-
Alexei Starovoitov authored
BPF C program attaches to blk_mq_start_request()/blk_update_request() kprobe events to calculate IO latency. For every completed block IO event it computes the time delta in nsec and records in a histogram map: map[log10(delta)*10]++ User space reads this histogram map every 2 seconds and prints it as a 'heatmap' using gray shades of text terminal. Black spaces have many events and white spaces have very few events. Left most space is the smallest latency, right most space is the largest latency in the range. Usage: $ sudo ./tracex3 and do 'sudo dd if=/dev/sda of=/dev/null' in other terminal. Observe IO latencies and how different activity (like 'make kernel') affects it. Similar experiments can be done for network transmit latencies, syscalls, etc. '-t' flag prints the heatmap using normal ascii characters: $ sudo ./tracex3 -t heatmap of IO latency # - many events with this latency - few events |1us |10us |100us |1ms |10ms |100ms |1s |10s *ooo. *O.#. # 221 . *# . # 125 .. .o#*.. # 55 . . . . .#O # 37 .# # 175 .#*. # 37 # # 199 . . *#*. # 55 *#..* # 42 # # 266 ...***Oo#*OO**o#* . # 629 # # 271 . .#o* o.*o* # 221 . . o* *#O.. # 50 Signed-off-by: -
Alexei Starovoitov authored
this example has two probes in one C file that attach to different kprove events and use two different maps. 1st probe is x64 specific equivalent of dropmon. It attaches to kfree_skb, retrevies 'ip' address of kfree_skb() caller and counts number of packet drops at that 'ip' address. User space prints 'location - count' map every second. 2nd probe attaches to kprobe:sys_write and computes a histogram of different write sizes Usage: $ sudo tracex2 location 0xffffffff81695995 count 1 location 0xffffffff816d0da9 count 2 location 0xffffffff81695995 count 2 location 0xffffffff816d0da9 count 2 location 0xffffffff81695995 count 3 location 0xffffffff816d0da9 count 2 557145+0 records in 557145+0 records out 285258240 bytes (285 MB) copied, 1.02379 s, 279 MB/s syscall write() stats byte_size : count distribution 1 -> 1 : 3 | | 2 -> 3 : 0 | | 4 -> 7 : 0 | | 8 -> 15 : 0 | | 16 -> 31 : 2 | | 32 -> 63 : 3 | | 64 -> 127 : 1 | | 128 -> 255 : 1 | | 256 -> 511 : 0 | | 512 -> 1023 : 1118968 |************************************* | Ctrl-C at any time. Kernel will auto cleanup maps and programs $ addr2line -ape ./bld_x64/vmlinux 0xffffffff81695995 0xffffffff816d0da9 0xffffffff81695995: ./bld_x64/../net/ipv4/icmp.c:1038 0xffffffff816d0da9: ./bld_x64/../net/unix/af_unix.c:1231 Signed-off-by:
-
Alexei Starovoitov authored
tracex1_kern.c - C program compiled into BPF. It attaches to kprobe:netif_receive_skb() When skb->dev->name == "lo", it prints sample debug message into trace_pipe via bpf_trace_printk() helper function. tracex1_user.c - corresponding user space component that: - loads BPF program via bpf() syscall - opens kprobes:netif_receive_skb event via perf_event_open() syscall - attaches the program to event via ioctl(event_fd, PERF_EVENT_IOC_SET_BPF, prog_fd); - prints from trace_pipe Note, this BPF program is non-portable. It must be recompiled with current kernel headers. kprobe is not a stable ABI and BPF+kprobe scripts may no longer be meaningful when kernel internals change. No matter in what way the kernel changes, neither the kprobe, nor the BPF program can ever crash or corrupt the kernel, assuming the kprobes, perf and BPF subsystem has no bugs. The verifier will detect that the program is using bpf_trace_printk() and the kernel will print 'this is a DEBUG kernel' warning banner, which means that bpf_trace_printk() should be used for debugging of the BPF program only. Usage: $ sudo tracex1 ping-19826 [000] d.s2 63103.382648: : skb ffff880466b1ca00 len 84 ping-19826 [000] d.s2 63103.382684: : skb ffff880466b1d300 len 84 ping-19826 [000] d.s2 63104.382533: : skb ffff880466b1ca00 len 84 ping-19826 [000] d.s2 63104.382594: : skb ffff880466b1d300 len 84 Signed-off-by: -
Alexei Starovoitov authored
Debugging of BPF programs needs some form of printk from the program, so let programs call limited trace_printk() with %d %u %x %p modifiers only. Similar to kernel modules, during program load verifier checks whether program is calling bpf_trace_printk() and if so, kernel allocates trace_printk buffers and emits big 'this is debug only' banner. Signed-off-by:
Steven Rostedt <rostedt@goodmis.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: David S. Miller <davem@davemloft.net> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1427312966-8434-6-git-send-email-ast@plumgrid.comSigned-off-by:
-
Alexei Starovoitov authored
bpf_ktime_get_ns() is used by programs to compute time delta between events or as a timestamp Signed-off-by:
-
Alexei Starovoitov authored
BPF programs, attached to kprobes, provide a safe way to execute user-defined BPF byte-code programs without being able to crash or hang the kernel in any way. The BPF engine makes sure that such programs have a finite execution time and that they cannot break out of their sandbox. The user interface is to attach to a kprobe via the perf syscall: struct perf_event_attr attr = { .type = PERF_TYPE_TRACEPOINT, .config = event_id, ... }; event_fd = perf_event_open(&attr,...); ioctl(event_fd, PERF_EVENT_IOC_SET_BPF, prog_fd); 'prog_fd' is a file descriptor associated with BPF program previously loaded. 'event_id' is an ID of the kprobe created. Closing 'event_fd': close(event_fd); ... automatically detaches BPF program from it. BPF programs can call in-kernel helper functions to: - lookup/update/delete elements in maps - probe_read - wraper of probe_kernel_read() used to access any kernel data structures BPF programs receive 'struct pt_regs *' as an input ('struct pt_regs' is architecture dependent) and return 0 to ignore the event and 1 to store kprobe event into the ring buffer. Note, kprobes are a fundamentally _not_ a stable kernel ABI, so BPF programs attached to kprobes must be recompiled for every kernel version and user must supply correct LINUX_VERSION_CODE in attr.kern_version during bpf_prog_load() call. Signed-off-by:Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arnaldo Carvalho de Melo <acme@infradead.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Daniel Borkmann <daniel@iogearbox.net> Cc: David S. Miller <davem@davemloft.net> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/r/1427312966-8434-4-git-send-email-ast@plumgrid.comSigned-off-by:
-
Alexei Starovoitov authored
add TRACE_EVENT_FL_KPROBE flag to differentiate kprobe type of tracepoints, since bpf programs can only be attached to kprobe type of PERF_TYPE_TRACEPOINT perf events. Signed-off-by:
-
Daniel Borkmann authored
Socket filter code and other subsystems with upcoming eBPF support should not need to deal with the fact that we have CONFIG_BPF_SYSCALL defined or not. Having the bpf syscall as a config option is a nice thing and I'd expect it to stay that way for expert users (I presume one day the default setting of it might change, though), but code making use of it should not care if it's actually enabled or not. Instead, hide this via header files and let the rest deal with it. Signed-off-by:
Daniel Borkmann <daniel@iogearbox.net> Signed-off-by:
-
Ingo Molnar authored
This WIP branch is now ready to be merged. Signed-off-by:
-
- 01 Apr, 2015 1 commit
-
-
Ingo Molnar authored
Merge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core Pull perf/core improvements and fixes from Arnaldo Carvalho de Melo: User visible changes: - Fix 'perf script' pipe mode segfault, by always initializing ordered_events in perf_session__new(). (Arnaldo Carvalho de Melo) - Fix ppid for synthesized fork events. (David Ahern) - Fix kernel symbol resolution of callchains in S/390 by remembering the cpumode. (David Hildenbrand) Infrastructure changes: - Disable libbabeltrace check by default in the build system. (Jiri Olsa) Signed-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by:
-
- 31 Mar, 2015 5 commits
-
-
Arnaldo Carvalho de Melo authored
As these can be obtained from the ordered_events pointer, via container_of, reducing the cross section of ordered_samples. These were added to ordered_samples in: commit b7b61cbe Author: Arnaldo Carvalho de Melo <acme@redhat.com> Date: Tue Mar 3 11:58:45 2015 -0300 perf ordered_events: Shorten function signatures By keeping pointers to machines, evlist and tool in ordered_events. But that was more a transitional patch while moving stuff out from perf_session.c to ordered_events.c and possibly not even needed by then, as we could use the container_of() method and instead of having the nr_unordered_samples stats in events_stats, we can have it in ordered_samples. Based-on-a-patch-by:
Jiri Olsa <jolsa@kernel.org> Acked-by:
-
Arnaldo Carvalho de Melo authored
Even when it is not used to actually reorder events, some of its fields are used, like session->ordered_events->tool, to shorten function signatures where tool, for instance, was being passed, as the tool is needed for the ordered_events code, we need it there and might as well use it for other perf_session needs. This fixes a problem where 'perf script' had some condition that made session->ordered_events not to be initialized even with its script->tool ordered_events related flags asking for it to be, which looks like another bug and needs to be investigated further. Always initializing session->ordered_events at least leaves the current assumptions in place, so do it now. Reported-by:
David Ahern <dsahern@gmail.com> Reviewed-by:
-
David Ahern authored
363b785f added synthesized fork events and set a thread's parent id to itself. Since we are already processing /proc/<pid>/status the ppid can be determined properly. Make it so. Signed-off-by:
Don Zickus <dzickus@redhat.com> Acked-by:
-
David Ahern authored
Rather than parsing /proc/pid/status file one line at a time, read it into a buffer in one shot and search for all strings in one pass. tgid conversion also simplified -- removing the isspace walk. As noted by Arnaldo those are not needed for atoi == strtol calls. Signed-off-by:
-
David Hildenbrand authored
Commit 2e77784b ("perf callchain: Move cpumode resolve code to add_callchain_ip") promised "No change in behavior.". As this commit breaks callchains on s390x (symbols not getting resolved, observed when profiling the kernel), this statement is wrong. The cpumode must be kept when iterating over all ips, otherwise the default (PERF_RECORD_MISC_USER) will be used by error. Signed-off-by:
David Hildenbrand <dahi@linux.vnet.ibm.com> Acked-by:
-
- 30 Mar, 2015 1 commit
-
-
Jiri Olsa authored
Disabling libbabeltrace check by default and replacing the NO_LIBBABELTRACE make variable with LIBBABELTRACE. Users wanting the libbabeltrace feature need to build via: $ make LIBBABELTRACE=1 The reason for this is that the libababeltrace interface we use (version 1.3) hasn't been packaged/released yet, thus the failing feature check only slows down build and confuses other (non CTF) developers. Requested-by:
-
- 27 Mar, 2015 15 commits
-
-
Peter Zijlstra authored
While thinking on the whole clock discussion it occurred to me we have two distinct uses of time: 1) the tracking of event/ctx/cgroup enabled/running/stopped times which includes the self-monitoring support in struct perf_event_mmap_page. 2) the actual timestamps visible in the data records. And we've been conflating them. The first is all about tracking time deltas, nobody should really care in what time base that happens, its all relative information, as long as its internally consistent it works. The second however is what people are worried about when having to merge their data with external sources. And here we have the discussion on MONOTONIC vs MONOTONIC_RAW etc.. Where MONOTONIC is good for correlating between machines (static offset), MONOTNIC_RAW is required for correlating against a fixed rate hardware clock. This means configurability; now 1) makes that hard because it needs to be internally consistent across groups of unrelated events; which is why we had to have a global perf_clock(). However, for 2) it doesn't really matter, perf itself doesn't care what it writes into the buffer. The below patch makes the distinction between these two cases by adding perf_event_clock() which is used for the second case. It further makes this configurable on a per-event basis, but adds a few sanity checks such that we cannot combine events with different clocks in confusing ways. And since we then have per-event configurability we might as well retain the 'legacy' behaviour as a default. Signed-off-by: -
Ingo Molnar authored
Signed-off-by: -
Ingo Molnar authored
An upcoming patch will depend on tai_ns() and NMI-safe ktime_get_raw_fast(), so merge timers/core here in a separate topic branch until it's all cooked and timers/core is merged upstream. Signed-off-by: -
Peter Zijlstra authored
While looking at some fuzzer output I noticed that we do not hold any locks on leader->ctx and therefore the sibling_list iteration is unsafe. Acquire the relevant ctx->mutex before calling into the pmu specific code. Signed-off-by:
-
David Ahern authored
perf_pmu_disable() is called before pmu->add() and perf_pmu_enable() is called afterwards. No need to call these inside of x86_pmu_add() as well. Signed-off-by:
-
Ingo Molnar authored
Signed-off-by: -
Ingo Molnar authored
Signed-off-by: -
Peter Zijlstra authored
Because it was the only clock for which we didn't have a _ns() accessor yet. Signed-off-by:
-
Peter Zijlstra authored
Add the NMI safe CLOCK_MONOTONIC_RAW accessor.. Signed-off-by:
John Stultz <john.stultz@linaro.org> Cc: Thomas Gleixner <tglx@linutronix.de> Link: http://lkml.kernel.org/r/20150319093400.562746929@infradead.orgSigned-off-by:
-
Peter Zijlstra authored
In preparation for more tk_fast instances, remove all hard-coded tk_fast_mono references. Signed-off-by:
-
Peter Zijlstra authored
Introduce tkr_raw and make use of it. base_raw -> tkr_raw.base clock->{mult,shift} -> tkr_raw.{mult.shift} Kill timekeeping_get_ns_raw() in favour of timekeeping_get_ns(&tkr_raw), this removes all mono_raw special casing. Duplicate the updates to tkr_mono.cycle_last into tkr_raw.cycle_last, both need the same value. Signed-off-by: -
Peter Zijlstra authored
In preparation of adding another tkr field, rename this one to tkr_mono. Also rename tk_read_base::base_mono to tk_read_base::base, since the structure is not specific to CLOCK_MONOTONIC and the mono name got added to the tk_read_base instance. Lots of trivial churn. Signed-off-by:
-
Andi Kleen authored
On Broadwell INST_RETIRED.ALL cannot be used with any period that doesn't have the lowest 6 bits cleared. And the period should not be smaller than 128. This is erratum BDM11 and BDM55: http://www.intel.com/content/dam/www/public/us/en/documents/specification-updates/5th-gen-core-family-spec-update.pdf BDM11: When using a period < 100; we may get incorrect PEBS/PMI interrupts and/or an invalid counter state. BDM55: When bit0-5 of the period are !0 we may get redundant PEBS records on overflow. Add a new callback to enforce this, and set it for Broadwell. How does this handle the case when an app requests a specific period with some of the bottom bits set? Short answer: Any useful instruction sampling period needs to be 4-6 orders of magnitude larger than 128, as an PMI every 128 instructions would instantly overwhelm the system and be throttled. So the +-64 error from this is really small compared to the period, much smaller than normal system jitter. Long answer (by Peterz): IFF we guarantee perf_event_attr::sample_period >= 128. Suppose we start out with sample_period=192; then we'll set period_left to 192, we'll end up with left = 128 (we truncate the lower bits). We get an interrupt, find that period_left = 64 (>0 so we return 0 and don't get an overflow handler), up that to 128. Then we trigger again, at n=256. Then we find period_left = -64 (<=0 so we return 1 and do get an overflow). We increment with sample_period so we get left = 128. We fire again, at n=384, period_left = 0 (<=0 so we return 1 and get an overflow). And on and on. So while the individual interrupts are 'wrong' we get then with interval=256,128 in exactly the right ratio to average out at 192. And this works for everything >=128. So the num_samples*fixed_period thing is still entirely correct +- 127, which is good enough I'd say, as you already have that error anyhow. So no need to 'fix' the tools, al we need to do is refuse to create INST_RETIRED:ALL events with sample_period < 128. Signed-off-by:
Andi Kleen <ak@linux.intel.com> [ Updated comments and changelog a bit. ] Signed-off-by:
-
Andi Kleen authored
Add Broadwell support for Broadwell to perf. The basic support is very similar to Haswell. We use the new cache event list added for Haswell earlier. The only differences are a few bits related to remote nodes. To avoid an extra, mostly identical, table these are patched up in the initialization code. The constraint list has one new event that needs to be handled over Haswell. Includes code and testing from Kan Liang. Signed-off-by:
-
Andi Kleen authored
Haswell offcore events are quite different from Sandy Bridge. Add a new table to handle Haswell properly. Note that the offcore bits listed in the SDM are not quite correct (this is currently being fixed). An uptodate list of bits is in the patch. The basic setup is similar to Sandy Bridge. The prefetch columns have been removed, as prefetch counting is not very reliable on Haswell. One L1 event that is not in the event list anymore has been also removed. - data reads do not include code reads (comparable to earlier Sandy Bridge tables) - data counts include speculative execution (except L1 write, dtlb, bpu) - remote node access includes both remote memory, remote cache, remote mmio. - prefetches are not included in the counts for consistency (different from Sandy Bridge, which includes prefetches in the remote node) Signed-off-by:
-
