- 11 Feb, 2019 5 commits
-
-
Kan Liang authored
Clean up counter freezing quirk to use the new facility to check for min microcode revisions. Rename the counter freezing quirk related functions. Because other platforms, e.g. Goldmont, also needs to call the quirk. Only check the boot CPU, assuming models and features are consistent over all CPUs. Signed-off-by:
Kan Liang <kan.liang@linux.intel.com> Signed-off-by:
Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: David Ahern <dsahern@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> 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: bp@alien8.de Link: https://lkml.kernel.org/r/1549319013-4522-4-git-send-email-kan.liang@linux.intel.comSigned-off-by:
Ingo Molnar <mingo@kernel.org>
-
Kan Liang authored
Clean up SNB PEBS quirk to use the new facility to check for min microcode revisions. Only check the boot CPU, assuming models and features are consistent over all CPUs. Signed-off-by:
-
Andi Kleen authored
KVM added a workaround for PEBS events leaking into guests with commit: 26a4f3c0 ("perf/x86: disable PEBS on a guest entry.") This uses the VT entry/exit list to add an extra disable of the PEBS_ENABLE MSR. Intel also added a fix for this issue to microcode updates on Haswell/Broadwell/Skylake. It turns out using the MSR entry/exit list makes VM exits significantly slower. The list is only needed for disabling PEBS, because the GLOBAL_CTRL change gets optimized by KVM into changing the VMCS. Check for the microcode updates that have the microcode fix for leaking PEBS, and disable the extra entry/exit list entry for PEBS_ENABLE. In addition we always clear the GLOBAL_CTRL for the PEBS counter while running in the guest, which is enough to make them never fire at the wrong side of the host/guest transition. The overhead for VM exits with the filtering active with the patch is reduced from 8% to 4%. The microcode patch has already been merged into future platforms. This patch is one-off thing. The quirks is used here. For other old platforms which doesn't have microcode patch and quirks, extra disable of the PEBS_ENABLE MSR is still required. Signed-off-by:
Andi Kleen <ak@linux.intel.com> Signed-off-by:
-
Ingo Molnar authored
Signed-off-by: -
Kan Liang authored
For bug workarounds or checks, it is useful to check for specific microcode revisions. Add a new generic function to match the CPU with stepping. Add the other function to check the min microcode revisions for the matched CPU. A new table format is introduced to facilitate the quirk to fill the related information. This does not change the existing x86_cpu_id because it's an ABI shared with modules, and also has quite different requirements, as in no wildcards, but everything has to be matched exactly. Originally-by:
Thomas Gleixner <tglx@linutronix.de> Signed-off-by:
Borislav Petkov <bp@alien8.de> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: eranian@google.com Link: https://lkml.kernel.org/r/1549319013-4522-1-git-send-email-kan.liang@linux.intel.comSigned-off-by:
-
- 09 Feb, 2019 3 commits
-
-
Ingo Molnar authored
Merge tag 'perf-core-for-mingo-5.1-20190206' 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: Hardware tracing: Adrian Hunter: - Handle calls optimized into jumps to a different symbol in the thread stack routines used to process hardware traces (Adrian Hunter) Intel PT: Adrian Hunter: - Fix overlap calculation for padding. - Fix CYC timestamp calculation after OVF. - Packet splitting can only happen in 32-bit. - Add timestamp to auxtrace errors. ARM CoreSight: Leo Yan: - Add last instruction information in packet - Set sample flags for instruction range, exception and return packets and for a trace discontinuity. - Add exception number in exception packet - Change tuple from traceID-CPU# to traceID-metadata - Add traceID in packet Mathieu Poirier: - Add "sinks" group to PMU directory - Use event attributes to send sink information to kernel - Remove set_drv_config() API, no longer used. perf annotate: Jiri Olsa: - Delay symbol annotation to the resort phase, speeding up 'perf report' startup. perf record: Alexey Budankov: - Allow binding userspace buffers to NUMA nodes. Symbols: Adrian Hunter: - Fix calculating of symbol sizes when splitting kallsyms into maps for kcore processing. Vendor events: William Cohen: - Intel: Fix Load_Miss_Real_Latency on CLX Misc: Arnaldo Carvalho de Melo: - Streamline headers, removing includes when all that is needed are just forward declarations, fixup the fallout for cases where headers should have been explicitely included but were instead obtained indirectly, by sheer luck. - Add fallback versions for CPU_{OR,EQUAL}(), so that code using it continue to build on older systems where those were not yet introduced or in systems using some other libc than the GNU one where those helpers aren't present. Documentation: Changbin Du: - Add documentation for BPF event selection. Signed-off-by:
Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by:
-
Ingo Molnar authored
Signed-off-by: -
Ingo Molnar authored
Merge tag 'perf-urgent-for-mingo-5.0-20190205' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/urgent Pull perf/urgent fixes from Arnaldo Carvalho de Melo: perf trace: Arnaldo Carvalho de Melo: Fix handling of probe:vfs_getname when the probed routine is inlined in multiple places, fixing the collection of the 'filename' parameter in open syscalls. perf test: Gustavo A. R. Silva: Fix bitwise operator usage in evsel-tp-sched test, which made tat test always detect fields as signed. Jiri Olsa: Filter out hidden symbols from labels, added in systems where the annobin plugin is used, such as RHEL8, which, if left in place make the DWARF unwind 'perf test' to fail on PPC. Tony Jones: Fix 'perf_event_attr' tests when building with python3. perf mem/c2c: Ravi Bangoria: Fix perf_mem_events on PowerPC. tools headers UAPI: Arnaldo Carvalho de Melo: Sync linux/in.h copy from the kernel sources, silencing a perf build warning. Signed-off-by:
-
- 06 Feb, 2019 32 commits
-
-
Adrian Hunter authored
The timestamp can use useful to find part of a trace that has an error without outputting all of the trace e.g. using the itrace 's' option to skip initial number of events. Signed-off-by:
Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@redhat.com> Link: http://lkml.kernel.org/r/20190206103947.15750-6-adrian.hunter@intel.comSigned-off-by:
-
Adrian Hunter authored
Data is copied when the trace is stopped, so packets are never split between buffers except when processing if the buffer cannot fit in the address space which can only happen on 32-bit systems. Change the logic to reflect that. Signed-off-by:
-
Adrian Hunter authored
CYC packet timestamp calculation depends upon CBR which was being cleared upon overflow (OVF). That can cause errors due to failing to synchronize with sideband events. Even if a CBR change has been lost, the old CBR is still a better estimate than zero. So remove the clearing of CBR. Signed-off-by:
-
Adrian Hunter authored
Auxtrace records might have up to 7 bytes of padding appended. Adjust the overlap accordingly. Signed-off-by:
-
Adrian Hunter authored
Define auxtrace record alignment so that it can be referenced elsewhere. Note this is preparation for patch "perf intel-pt: Fix overlap calculation for padding" Signed-off-by:
-
Adrian Hunter authored
The compiler might optimize a call/ret combination by making it a jmp. However the thread-stack does not presently cater for that, so that such control flow is not visible in the call graph. Make it visible by recording on the stack a branch to the start of a different symbol. Note, that means when a ret pops the stack, all jmps must be popped off first. Example: $ cat jmp-to-fn.c __attribute__((noinline)) int bar(void) { return -1; } __attribute__((noinline)) int foo(void) { return bar() + 1; } int main() { return foo(); } $ gcc -ggdb3 -Wall -Wextra -O2 -o jmp-to-fn jmp-to-fn.c $ objdump -d jmp-to-fn <SNIP> 0000000000001040 <main>: 1040: 31 c0 xor %eax,%eax 1042: e9 09 01 00 00 jmpq 1150 <foo> <SNIP> 0000000000001140 <bar>: 1140: b8 ff ff ff ff mov $0xffffffff,%eax 1145: c3 retq <SNIP> 0000000000001150 <foo>: 1150: 31 c0 xor %eax,%eax 1152: e8 e9 ff ff ff callq 1140 <bar> 1157: 83 c0 01 add $0x1,%eax 115a: c3 retq <SNIP> $ perf record -o jmp-to-fn.perf.data -e intel_pt/cyc/u ./jmp-to-fn [ perf record: Woken up 1 times to write data ] [ perf record: Captured and wrote 0,017 MB jmp-to-fn.perf.data ] $ perf script -i jmp-to-fn.perf.data --itrace=be -s ~/libexec/perf-core/scripts/python/export-to-sqlite.py jmp-to-fn.db branches calls 2019-01-08 13:24:58.783069 Creating database... 2019-01-08 13:24:58.794650 Writing records... 2019-01-08 13:24:59.008050 Adding indexes 2019-01-08 13:24:59.015802 Done $ ~/libexec/perf-core/scripts/python/exported-sql-viewer.py jmp-to-fn.db Before: main -> bar After: main -> foo -> bar Committer testing: Install the python2-pyside package, then select these menu options on the GUI: "Reports" "Context sensitive callgraphs" Then go on expanding the symbols, to get, full picture when doing this on a fedora:29 with gcc version 8.2.1 20181215 (Red Hat 8.2.1-6) (GCC): jmp-to-fn PID:TID _start (ld-2.28.so) __libc_start_main main foo bar To verify that indeed, this fixes the problem. Signed-off-by:Jiri Olsa <jolsa@kernel.org> Link: http://lkml.kernel.org/r/20190109091835.5570-5-adrian.hunter@intel.comSigned-off-by:
-
Adrian Hunter authored
Make thread_stack__no_call_return() more readable by adding more local variables. Signed-off-by:
-
Adrian Hunter authored
If 'cp' is checked in thread_stack__push_cp() a number of error checks can be removed, reducing code size and improving readability. Signed-off-by:
-
Adrian Hunter authored
Kallsyms symbols do not have a size, so the size becomes the distance to the next symbol. Consequently the recently added trampoline symbols end up with large sizes because the trampolines are some distance from one another and the main kernel map. However, symbols that end outside their map can disrupt the symbol tree because, after mapping, it can appear incorrectly that they overlap other symbols. Add logic to truncate symbol size to the end of the corresponding map. Signed-off-by:
-
William Cohen authored
Fix incorrect event names for the Load_Miss_Real_Latency metric for Cascadelake server in the same manner as commit 91b2b970 for SKL/SKX. Signed-off-by:
William Cohen <wcohen@redhat.com> Reviewed-by:
-
Leo Yan authored
When return from exception, we need to distinguish if it's system call return or for other type exceptions for setting sample flags. Due to the exception return packet doesn't contain exception number, so we cannot decide sample flags based on exception number. On the other hand, the exception return packet is followed by an instruction range packet; this range packet deliveries the start address after exception handling, we can check if it is a SVC instruction just before the start address. If there has one SVC instruction is found ahead the return address, this means it's an exception return for system call; otherwise it is an normal return for other exceptions. This patch is to set sample flags for exception return packet, firstly it simply set sample flags as PERF_IP_FLAG_INTERRUPT for all exception returns since at this point it doesn't know what's exactly the exception type. We will defer to decide if it's an exception return for system call when the next instruction range packet comes, it checks if there has one SVC instruction prior to the start address and if so we will change sample flags to PERF_IP_FLAG_SYSCALLRET for system call return. Signed-off-by:
Leo Yan <leo.yan@linaro.org> Reviewed-by:
Mathieu Poirier <mathieu.poirier@linaro.org> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mike Leach <mike.leach@linaro.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Robert Walker <robert.walker@arm.com> Cc: Suzuki K Poulouse <suzuki.poulose@arm.com> Cc: coresight ml <coresight@lists.linaro.org> Cc: linux-arm-kernel@lists.infradead.org Link: http://lkml.kernel.org/r/20190129122842.32041-9-leo.yan@linaro.orgSigned-off-by:
-
Leo Yan authored
The exception taken and returning are typical flow for instruction jump but it needs to be handled with exception packets. This patch is to set sample flags for exception packet. Since the exception packet contains the exception number, according to the exception number this patch makes decision for belonging to which exception types. The decoder have defined different exception number for ETMv3 and ETMv4 separately, hence this patch needs firstly decide the ETM version by using the metadata magic number, and this patch adds helper function cs_etm__get_magic() for easily getting magic number. Based on different ETM version, the exception packet contains the exception number, according to the exception number this patch makes decision for the exception belonging to which exception types. In this patch, it introduces helper function cs_etm__is_svc_instr(); for ETMv4 CS_ETMV4_EXC_CALL covers SVC, SMC and HVC cases in the single exception number, thus need to use cs_etm__is_svc_instr() to decide an exception taken for system call. Signed-off-by:
Robert Walker <robert.walker@arm.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mike Leach <mike.leach@linaro.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Suzuki K Poulouse <suzuki.poulose@arm.com> Cc: coresight ml <coresight@lists.linaro.org> Cc: linux-arm-kernel@lists.infradead.org Link: http://lkml.kernel.org/r/20190129122842.32041-8-leo.yan@linaro.orgSigned-off-by:
-
Leo Yan authored
Add traceID in packet, thus we can use traceID to retrieve metadata pointer from traceID-metadata tuple. Signed-off-by:
-
Leo Yan authored
If packet processing wants to know the packet is bound with which ETM version, it needs to access metadata to decide that based on metadata magic number; but we cannot simply to use CPU logic ID number as index to access metadata sequential array, especially when system have hotplugged off CPUs, the metadata array are only allocated for online CPUs but not offline CPUs, so the CPU logic number doesn't match with its index in the array. This patch is to change tuple from traceID-CPU# to traceID-metadata, thus it can use the tuple to retrieve metadata pointer according to traceID. For safe accessing metadata fields, this patch provides helper function cs_etm__get_cpu() which is used to return CPU number according to traceID; cs_etm_decoder__buffer_packet() is the first consumer for this helper function. Signed-off-by:
-
Leo Yan authored
When an exception packet comes, it contains the information for exception number; the exception number indicates the exception types, so from it we can know if the exception is taken for interrupt, system call or other traps, etc. This patch simply adds a field in cs_etm_packet struct, it records exception number for exception packet that will then be used to properly identify exception types to the perf synthesize mechanic. Signed-off-by:
-
Leo Yan authored
In the middle of trace stream, it might be interrupted thus the trace data is not continuous, the trace stream firstly is ended for previous trace block and restarted for next block. To display related information for showing trace is restarted, this patch set sample flags for trace discontinuity: - If one discontinuity packet is coming, append flag PERF_IP_FLAG_TRACE_END to the previous packet to indicate the trace has been ended; - If one instruction packet is following discontinuity packet, this instruction packet is the first one packet to restarting trace. So set flag PERF_IP_FLAG_TRACE_START to discontinuity packet, this flag will be used to generate sample when connect with the sequential instruction packet. Signed-off-by:
-
Leo Yan authored
The perf sample data contains flags to indicate the hardware trace data is belonging to which type branch instruction, thus this can be used to print out the human readable string. Arm CoreSight ETM sample data is missed to set flags and it is always set to zeros, this results in perf tool skips to print string for instruction types. This patch is to set branch instruction flags for instruction range packet. Signed-off-by:
-
Leo Yan authored
Decoder provides last instruction related information, these information can be used for trace analysis; specifically we can get to know what kind of branch instruction has been executed, mainly the information are contained in three element fields: last_i_type: this is significant type for waypoint calculation, it indicates the last instruction is one of immediate branch instruction, indirect branch instruction, instruction barrier (ISB), or data barrier (DSB/DMB). last_i_subtype: this is used for instruction sub type, it can be branch with link, ARMv8 return instruction, ARMv8 eret instruction (return from exception), or ARMv7 instruction which could imply return (e.g. MOV PC, LR; POP { ,PC}). last_instr_cond: it indicates if the last instruction was conditional. But these three fields are not saved into cs_etm_packet struct, thus cs-etm layer don't know related information and cannot generate sample flags for branch instructions. This patch add corresponding three new fields in cs_etm_packet struct and save related value into the packet structure, it is preparation for supporting sample flags. Signed-off-by: -
Changbin Du authored
Add documentation for how to pass a BPF program as a perf event. Signed-off-by:
Changbin Du <changbin.du@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/r/20190201134651.12373-1-changbin.du@gmail.comSigned-off-by:
-
Jiri Olsa authored
Currently we make the annotation for the IPC column during the entry display, already outside of the progress bar scope, so it appears like 'perf report' is stuck. Move the annotation retrieval to the resort phase, so that all the data are ready for display. Signed-off-by:
-
Jiri Olsa authored
Add perf_evsel__output_resort_cb() so we have an interface with a callback for each hist entry. It will be used in the following patch. Signed-off-by:
-
Jiri Olsa authored
Add argument to hists__resort_cb_t so that we can pass data from upper layers to the callback function. It will be used in the following patches. Signed-off-by:
-
Arnaldo Carvalho de Melo authored
It prevents copy elision, generating this warning when building with fedora:rawhide's clang: clang version 7.0.1 (Fedora 7.0.1-2.fc30) Target: x86_64-unknown-linux-gnu Thread model: posix InstalledDir: /usr/bin Found candidate GCC installation: /usr/bin/../lib/gcc/x86_64-redhat-linux/9 Found candidate GCC installation: /usr/lib/gcc/x86_64-redhat-linux/9 Selected GCC installation: /usr/bin/../lib/gcc/x86_64-redhat-linux/9 Candidate multilib: .;@m64 Candidate multilib: 32;@m32 Selected multilib: .;@m64 $ make -C tools/perf CC=clang LIBCLANGLLVM=1 <SNIP> util/c++/clang.cpp: In function 'std::unique_ptr<llvm::SmallVectorImpl<char> > perf::getBPFObjectFromModule(llvm::Module*)': util/c++/clang.cpp:163:18: error: moving a local object in a return statement prevents copy elision [-Werror=pessimizing-move] 163 | return std::move(Buffer); | ~~~~~~~~~^~~~~~~~ util/c++/clang.cpp:163:18: note: remove 'std::move' call cc1plus: all warnings being treated as errors <SNIP> References: http://www.cplusplus.com/forum/general/186411/#msg908572 https://en.cppreference.com/w/cpp/language/return#Notes https://en.cppreference.com/w/cpp/language/copy_elision Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Luis Cláudio Gonçalves <lclaudio@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Wang Nan <wangnan0@huawei.com> Link: https://lkml.kernel.org/n/tip-lehqf5x5q96l0o8myhb6blz6@git.kernel.orgSigned-off-by: -
Alexey Budankov authored
Build node cpu masks for mmap data buffers. Apply node cpu masks to tool thread every time it references data buffers cross node or cross cpu. Signed-off-by:
Alexey Budankov <alexey.budankov@linux.intel.com> Reviewed-by:
Jiri Olsa <jolsa@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/b25e4ebc-078d-2c7b-216c-f0bed108d073@linux.intel.com [ Use cpu-set-sched.h to get the CPU_{EQUAL,OR}() fallbacks for older systems ] Signed-off-by:
-
Arnaldo Carvalho de Melo authored
From the glibc sources, so that we can keep the tooling buildable in older systems while using recent sched.h CPU_ macros. Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexey Budankov <alexey.budankov@linux.intel.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Jiri Olsa <jolsa@kernel.org> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Link: https://lkml.kernel.org/n/tip-hvm9ysmrjip75ebdzhzoh429@git.kernel.orgSigned-off-by:
-
Alexey Budankov authored
Allocate and bind AIO user space buffers to the memory nodes that mmap kernel buffers are bound to. Signed-off-by:
-
Alexey Budankov authored
Allocate affinity option and masks for mmap data buffers and record thread as well as initialize allocated objects. Signed-off-by:
-
Mathieu Poirier authored
CoreSight was the only client of the PMU's set_drv_config() API. Now that it is no longer needed by CoreSight remove it from the code base. Signed-off-by:
Suzuki K Poulouse <suzuki.poulose@arm.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com> Cc: Alexei Starovoitov <ast@kernel.org> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Mark Rutland <mark.rutland@arm.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Will Deacon <will.deacon@arm.com> Cc: linux-arm-kernel@lists.infradead.org Cc: linux-s390@vger.kernel.org Link: http://lkml.kernel.org/r/20190131184714.20388-8-mathieu.poirier@linaro.orgSigned-off-by:
-
Mathieu Poirier authored
Now that the event's config2 attribute is used to communicate sink selection to the kernel, remove the old set_drv_config() implementation since it is no longer needed. Signed-off-by:
-
Mathieu Poirier authored
The communication of sink information for a trace session doesn't work when more than on CPU is involved in the scenario due to the static nature of sysfs. As such communicate the sink information to each event by using the perf_event::attr:config2 attribute. The information sent to the kernel is an hash of the sink's name, which is unique in a system. Signed-off-by:
-
Mathieu Poirier authored
Move definition of EVENT_SOURCE_DEVICE_PATH to pmu.h so that it can be used by other files than pmu.c Signed-off-by:
-
Mathieu Poirier authored
This patch uses the information conveyed by perf_event::attr::config2 to select a sink to use for the session. That way a sink can easily be selected to be used by more than one source, something that isn't currently possible with the sysfs implementation. Signed-off-by:
-
