CoreSight

coresight@lists.linaro.org

7 participants
2454 discussions

[PATCH] opencsd: docs: update for latest kernel versions

by Andrea Brunato

This patch updates the documentation to better capture the current status with the latest kernel releases and defines new scripts for controlling strobing accordingly. Signed-off-by: Andrea Brunato <andrea.brunato(a)arm.com> --- HOWTO.md | 54 +++++++++--------- decoder/tests/auto-fdo/autofdo.md | 73 +++++++++++++++++++------ decoder/tests/auto-fdo/record.sh | 68 ----------------------- decoder/tests/auto-fdo/set_strobing.sh | 29 ++++++++++ decoder/tests/auto-fdo/show_strobing.sh | 6 ++ 5 files changed, 115 insertions(+), 115 deletions(-) delete mode 100644 decoder/tests/auto-fdo/record.sh create mode 100755 decoder/tests/auto-fdo/set_strobing.sh create mode 100755 decoder/tests/auto-fdo/show_strobing.sh diff --git a/HOWTO.md b/HOWTO.md index b16294a..ebf44eb 100644 --- a/HOWTO.md +++ b/HOWTO.md @@ -6,35 +6,31 @@ HOWTO - using the library with perf {#howto_perf} This HOWTO explains how to use the perf cmd line tools and the openCSD library to collect and extract program flow traces generated by the CoreSight IP blocks on a Linux system. The examples have been generated using -an aarch64 Juno-r0 platform. All information is considered accurate and tested -using the latest version of the library and the `master` branch on the -[perf-opencsd github repository][1]. +an aarch64 Juno-r0 platform. On Target Trace Acquisition - Perf Record ----------------------------------------- -All the enhancement to the Perf tools that support the new `cs_etm` pmu have -not been upstreamed yet. To get the required functionality branch -`perf-opencsd-master` needs to be downloaded to the target system where -traces are to be collected. This branch is a vanilla upstream kernel -supplemented with modifications to the CoreSight framework and drivers to be -usable by the Perf core. The remaining out of tree patches are being -upstreamed incrementally. - -From there compiling the perf tools with `make -C tools/perf CORESIGHT=1` will -yield a `perf` executable that will support CoreSight trace collection. Note -that if traces are to be decompressed *off* target, there is no need to download + +Compile the perf tool from the same kernel source code version you are using, +with `make -C tools/perf`. +This will yield a `perf` executable that will support CoreSight trace collection. +Note that if traces are to be decompressed *off* target, there is no need to download and compile the openCSD library (on the target). +If you are instead planning to use perf to record and decode the trace on the target, +compile the perf tool linking against the openCSD library, in the following way: +`make -C tools/perf VF=1 CORESIGHT=1` + + Before launching a trace run a sink that will collect trace data needs to be identified. All CoreSight blocks identified by the framework are registed in sysFS: linaro@linaro-nano:~$ ls /sys/bus/coresight/devices/ - 20010000.etf 20040000.main_funnel 22040000.etm 22140000.etm - 230c0000.A53_funnel 23240000.etm replicator@20020000 20030000.tpiu - 20070000.etr 220c0000.A57_funnel 23040000.etm 23140000.etm 23340000.etm + etm0 etm2 etm4 etm6 funnel0 funnel2 funnel4 stm0 tmc_etr0 + etm1 etm3 etm5 etm7 funnel1 funnel3 replicator0 tmc_etf0 CoreSight blocks are listed in the device tree for a specific system and @@ -43,7 +39,7 @@ the sink that will recieve trace data needs to be identified and given as an option on the perf command line. Once a sink has been identify trace collection can start. An easy and yet interesting example is the `uname` command: - linaro@linaro-nano:~/kernel$ ./tools/perf/perf record -e cs_etm/(a)20070000.etr/ --per-thread uname + linaro@linaro-nano:~/kernel$ ./tools/perf/perf record -e cs_etm/@tmc_etr0/ --per-thread uname This will generate a `perf.data` file where execution has been traced for both user and kernel space. To narrow the field to either user or kernel space the @@ -51,7 +47,7 @@ user and kernel space. To narrow the field to either user or kernel space the traces to user space: - linaro@linaro-nano:~/kernel$ ./tools/perf/perf record -vvv -e cs_etm/(a)20070000.etr/u --per-thread uname + linaro@linaro-nano:~/kernel$ ./tools/perf/perf record -vvv -e cs_etm/@tmc_etr0/u --per-thread uname Problems setting modules path maps, continuing anyway... ----------------------------------------------------------- perf_event_attr: @@ -131,9 +127,9 @@ falls within the specified range. Any work done by the CPU outside of that range will not be traced. Address range filters can be specified for both user and kernel space session: - perf record -e cs_etm/(a)20070000.etr/k --filter 'filter 0xffffff8008562d0c/0x48' --per-thread uname + perf record -e cs_etm/@tmc_etr0/k --filter 'filter 0xffffff8008562d0c/0x48' --per-thread uname - perf record -e cs_etm/(a)20070000.etr/u --filter 'filter 0x72c/0x40(a)/opt/lib/libcstest.so.1.0' --per-thread ./main + perf record -e cs_etm/@tmc_etr0/u --filter 'filter 0x72c/0x40(a)/opt/lib/libcstest.so.1.0' --per-thread ./main When dealing with kernel space trace addresses are typically taken in the 'System.map' file. In user space addresses are relocatable and can be @@ -171,20 +167,20 @@ equal to the start address. Incidentally traces stop being generated when the insruction pointer is equal to the stop address. Anything that happens between there to events is traced: - perf record -e cs_etm/(a)20070000.etr/k --filter 'start 0xffffff800856bc50,stop 0xffffff800856bcb0' --per-thread uname + perf record -e cs_etm/@tmc_etr0/k --filter 'start 0xffffff800856bc50,stop 0xffffff800856bcb0' --per-thread uname - perf record -vvv -e cs_etm/(a)20070000.etr/u --filter 'start 0x72c(a)/opt/lib/libcstest.so.1.0, \ + perf record -vvv -e cs_etm/@tmc_etr0/u --filter 'start 0x72c(a)/opt/lib/libcstest.so.1.0, \ stop 0x40082c@/home/linaro/main' \ - --per-thread ./main + --per-thread ./main **Limitation on address filters:** The only limitation on address filters is the amount of address comparator found on an implementation and the mutual exclusion between range and start stop filters. As such the following example would _not_ work: - perf record -e cs_etm/(a)20070000.etr/k --filter 'start 0xffffff800856bc50,stop 0xffffff800856bcb0, \ // start/stop + perf record -e cs_etm/@tmc_etr0/k --filter 'start 0xffffff800856bc50,stop 0xffffff800856bcb0, \ // start/stop filter 0x72c/0x40(a)/opt/lib/libcstest.so.1.0' \ // address range - --per-thread uname + --per-thread uname Additional Trace Options ------------------------ @@ -198,7 +194,7 @@ Presently this threshold is fixed at 256 cycles for `perf record`. Command line options in `perf record` to use these features are part of the options for the `cs_etm` event: - perf record -e cs_etm/timestamp,cycacc,(a)20070000.etr/ --per-thread uname + perf record -e cs_etm/timestamp,cycacc,@tmc_etr0/ --per-thread uname At current version, `perf record` and `perf script` do not use this additional information. @@ -248,7 +244,7 @@ The openCSD library is not part of the perf tools. It is available on [github][1] and needs to be compiled before the perf tools. Checkout the required branch/tag version into a local directory. - linaro@t430:~/linaro/coresight$ git clone -b v0.8 https://github.com/Linaro/OpenCSD.git my-opencsd + linaro@t430:~/linaro/coresight$ git clone https://github.com/Linaro/OpenCSD.git my-opencsd Cloning into 'OpenCSD'... remote: Counting objects: 2063, done. remote: Total 2063 (delta 0), reused 0 (delta 0), pack-reused 2063 @@ -629,7 +625,7 @@ Best regards, *The Linaro CoreSight Team* -------------------------------------- -[1]: https://github.com/Linaro/perf-opencsd "perf-opencsd Github" +[1]: https://github.com/Linaro/OpenCSD [2]: http://people.linaro.org/~mathieu.poirier/openCSD/uname.v4.user.sept20.tgz diff --git a/decoder/tests/auto-fdo/autofdo.md b/decoder/tests/auto-fdo/autofdo.md index b1f2241..d7d37b1 100644 --- a/decoder/tests/auto-fdo/autofdo.md +++ b/decoder/tests/auto-fdo/autofdo.md @@ -99,6 +99,8 @@ You can include these backports in your kernel by either merging the appropriate branch using git or generating patches (using `git format-patch`). +For 5.5 based kernel, the only patch which needs to be applied is the one enabling strobing - etm4x: `Enable strobing of ETM`. + For 4.9 based kernels, use the `coresight-4.9-etr-etm_strobe` branch: ``` @@ -129,7 +131,7 @@ git am /output/dir/*.patch # or patch -p1 /output/dir/*.patch if not using git The CoreSight trace drivers must also be enabled in the kernel configuration. This can be done using the configuration menu (`make -menuconfig`), selecting `Kernel hacking` / `CoreSight Tracing Support` and +menuconfig`), selecting `Kernel hacking` / `arm64 Debugging` /`CoreSight Tracing Support` and enabling all options, or by setting the following in the configuration file: @@ -165,11 +167,15 @@ CoreSight devices, you should find the devices in sysfs: ``` # ls /sys/bus/coresight/devices/ -28440000.etm 28540000.etm 28640000.etm 28740000.etm -28c03000.funnel 28c04000.etf 28c05000.replicator 28c06000.etr -28c07000.tpiu +etm0 etm2 etm4 etm6 funnel0 funnel2 funnel4 stm0 tmc_etr0 +etm1 etm3 etm5 etm7 funnel1 funnel3 replicator0 tmc_etf0 ``` +The naming convention for etm devices can be different according to the kernel version you're using. +For more information about the naming scheme, please check out the [Linux Kernel Documentation](https://www.kernel.org/doc/html/latest/trace/coresight/cores… + +If `/sys/bus/coresight/devices/` is empty, you may want to check out your Kernel configuration to make sure your .config file is including CoreSight dependencies, such as the clock. + ### Perf tools The perf tool is used to capture execution trace, configuring the trace @@ -180,9 +186,12 @@ Arm recommends to use the perf version corresponding to the kernel running on the target. This can be built from the same kernel sources with ``` -make -C tools/perf ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- +make -C tools/perf CORESIGHT=1 VF=1 ARCH=arm64 CROSS_COMPILE=aarch64-linux-gnu- ``` +When specifying CORESIGHT=1, perf will be built using the installed OpenCSD library. +If you are cross compiling, then additional setup is required to ensure the build process links against the correct version of the library. + If the post-processing (`perf inject`) of the captured data is not being done on the target, then the OpenCSD library is not required for this build of perf. @@ -193,13 +202,22 @@ also be restricted to user space or kernel space with 'u' or 'k' parameters. For example: ``` -perf record -e cs_etm/(a)28c06000.etr/u --per-thread -- /bin/ls +perf record -e cs_etm/@tmc_etr0/u --per-thread -- /bin/ls ``` -Will record the userspace execution of '/bin/ls' into the ETR located at -0x28c06000. Note the `--per-thread` option is required - perf currently -only supports trace of a single thread of execution. CPU wide trace is a -work in progresss. +Will record the userspace execution of '/bin/ls' using tmc_etr0 as sink. + +## Capturing modes + +You can trace a single-threaded program in two different ways: + +1. By specifying `--per-thread`, and in this case the CoreSight subsystem will +record only a trace relative to the given program. + +2. By NOT specifying `--per-thread`, and in this case CPU-wide tracing will +be enabled. In this scenario the trace will contain both the target program trace +and other workloads that were executing on the same CPU + ## Processing trace and profiles @@ -241,26 +259,42 @@ For example, a typical configuration is to use a window size of 5000 cycles and a period of 10000 - this will collect 5000 cycles of trace every 50M cycles. With these proof-of-concept patches, the strobe parameters are configured via sysfs - each ETM will have `strobe_window` and -`strobe_period` parameters in `/sys/bus/coresight/devices/NNNNNNNN.etm` and +`strobe_period` parameters in `/sys/bus/coresight/devices/<sink>` and these values will have to be written to each (In a future version, this -will be integrated into the drivers and perf tool). The `record.sh` -script in this directory [`<opencsd>/decoder/tests/auto-fdo`] automates this process. +will be integrated into the drivers and perf tool). +The `set_strobing.sh` script in this directory [`<opencsd>/decoder/tests/auto-fdo`] automates this process. To collect trace from an application using ETM strobing, run: ``` -taskset -c 0 ./record.sh --strobe 5000 10000 28c06000.etr ./my_application arg1 arg2 +sudo ./set_strobing.sh 5000 10000 +perf record -e cs_etm/@tmc_etr0/u --per-thread -- <your app>" ``` -The taskset command is used to ensure the process stays on the same CPU -during execution. - The raw trace can be examined using the `perf report` command: ``` perf report -D -i perf.data --stdio ``` +Perf needs to be built from your linux kernel version souce code repository against the OpenCSD library in order to be able to properly read ETM-gathered samples and post-process them. +If running `perf report` produces an error like: + +``` +0x1f8 [0x268]: failed to process type: 70 [Operation not permitted] +Error: +failed to process sample +``` +or + +``` +"file uses a more recent and unsupported ABI (8 bytes extra). incompatible file format". +``` + +You are probably using a perf version which is not using this library: please make sure to install this project in your system by either compiling it from [Source Code]( <https://github.com/Linaro/OpenCSD>) from v0.9.1 or later and compile perf using this library. +Otherwise, this project is packaged for debian (install the libopencsd0, libopencsd-dev packages). + + For example: ``` @@ -295,6 +329,8 @@ an embedded target). The `perf inject` command decodes the execution trace and generates periodic instruction samples, with branch histories: +!! Careful: if you are using a device different than the one used to collect the profiling data, +you'll need to run `perf buildid-cache` as described below. ``` perf inject -i perf.data -o inj.data --itrace=i100000il ``` @@ -393,7 +429,8 @@ clang -O2 -fprofile-sample-use=program.llvmprof -o program program.c The basic commands to run an application and create a compiler profile are: ``` -taskset -c 0 ./record.sh --strobe 5000 10000 28c06000.etr ./my_application arg1 arg2 +sudo ./set_strobing.sh 5000 10000 +perf record -e cs_etm/@tmc_etr0/u --per-thread -- <your app>" perf inject -i perf.data -o inj.data --itrace=i100000il create_llvm_prof -binary=/path/to/binary -profile=inj.data -out=program.llvmprof ``` diff --git a/decoder/tests/auto-fdo/record.sh b/decoder/tests/auto-fdo/record.sh deleted file mode 100644 index 16d4ba2..0000000 --- a/decoder/tests/auto-fdo/record.sh +++ /dev/null @@ -1,68 +0,0 @@ -#!/bin/sh - -BUFFER_ETF_A53=ec802000.etf -BUFFER_ETF_A73=ed002000.etf -BUFFER_ETF_SYS=ec036000.etf -BUFFER_ETR=ec033000.etr - -OUT_FILE=perf.data - -STROBE= - -while :; do - case $1 in - --strobe) - STROBE=y - WINDOW=$2 - PERIOD=$3 - shift 3 - ;; - - *) - break ;; - esac -done - -case $1 in - etr) - BUFFER=$BUFFER_ETR - ;; - - etf-sys) - BUFFER=$BUFFER_ETF_SYS - ;; - - "") - BUFFER=$BUFFER_ETR - ;; - - *) - BUFFER=$1 - ;; -esac - -shift 1 - -case $0 in - /*) F=$0 ;; - *) F=$(pwd)/$0 ;; -esac - -SCRIPT_DIR=$(dirname $F) - -if [ "$STROBE" ]; then - for e in /sys/bus/coresight/devices/*.etm/; do - printf "%x" $WINDOW | sudo tee $e/strobe_window > /dev/null - printf "%x" $PERIOD | sudo tee $e/strobe_period > /dev/null - done -fi - -PERF=$SCRIPT_DIR/perf - -export LD_LIBRARY_PATH=$SCRIPT_DIR:$LD_LIBRARY_PATH - -sudo LD_LIBRARY_PATH=$SCRIPT_DIR:$LD_LIBRARY_PATH $PERF record $PERF_ARGS -e cs_etm/@$BUFFER/u --per-thread "$@" - -sudo chown $(id -u):$(id -g) $OUT_FILE - - diff --git a/decoder/tests/auto-fdo/set_strobing.sh b/decoder/tests/auto-fdo/set_strobing.sh new file mode 100755 index 0000000..e11e62d --- /dev/null +++ b/decoder/tests/auto-fdo/set_strobing.sh @@ -0,0 +1,29 @@ +#!/bin/bash + +WINDOW=$1 +PERIOD=$2 + +if [[ -z $WINDOW ]] || [[ -z $PERIOD ]]; then + echo "Window or Period not specified!" + echo "Example usage: ./set_strobing.sh <WINDOW VALUE> <PERIOD VALUE>" + echo "Example usage: ./set_strobing.sh 5000 10000" + exit +fi + + +if [[ $EUID != 0 ]]; then + echo "Please run as root" + exit +fi + +for e in /sys/bus/coresight/devices/etm*/; do + printf "%x" $WINDOW | tee $e/strobe_window > /dev/null + printf "%x" $PERIOD | tee $e/strobe_period > /dev/null + echo "Strobing period for $e set to $((`cat $e/strobe_period`))" + echo "Strobing window for $e set to $((`cat $e/strobe_window`))" +done + +## Shows the user a simple usage example +echo ">> Done! <<" +echo "You can now run perf to trace your application, for example:" +echo "perf record -e cs_etm/@tmc_etr0/u -- <your app>" diff --git a/decoder/tests/auto-fdo/show_strobing.sh b/decoder/tests/auto-fdo/show_strobing.sh new file mode 100755 index 0000000..d80d84c --- /dev/null +++ b/decoder/tests/auto-fdo/show_strobing.sh @@ -0,0 +1,6 @@ +#!/bin/bash + +for e in /sys/bus/coresight/devices/etm*/; do + echo "Strobing period for $e is $((`cat $e/strobe_period`))" + echo "Strobing window for $e is $((`cat $e/strobe_window`))" +done -- 2.17.1 IMPORTANT NOTICE: The contents of this email and any attachments are confidential and may also be privileged. If you are not the intended recipient, please notify the sender immediately and do not disclose the contents to any other person, use it for any purpose, or store or copy the information in any medium. Thank you.

5 years, 5 months

Release v0.14.0 OpenCSD

by Mike Leach

OpenCSD v0.14.0 is now released. This update contains a re-work of the ETMv4 decoder to simplify, reduce re-entancy and enable speculative tracing support. Speculative tracing support has been verified by an architecture development team within ARM. Other bugfixes and minor enhancements as per the readme file. Documentation is updated to reflect the latest versions of the kernel drivers (5.x). v0.15.0-dev version also released. This contains preliminary support for Q elements. Regards Mike -- Mike Leach Principal Engineer, ARM Ltd. Manchester Design Centre. UK

5 years, 5 months

Fix your SALES problem in 15 Days!

by ZAWIA AGENCY

5 years, 5 months

[PATCH] docs: Document how to change buildid cache (a.k.a ~/.debug)

by Wojciech Żmuda

From: Wojciech Zmuda <wzmuda(a)n7space.com> Perf allows changing where the buildid cache directory is created. Mention it in the howto document. Signed-off-by: Wojciech Zmuda <wzmuda(a)n7space.com> --- HOWTO.md | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/HOWTO.md b/HOWTO.md index b16294a..a8b5ce9 100644 --- a/HOWTO.md +++ b/HOWTO.md @@ -633,4 +633,5 @@ Best regards, [2]: http://people.linaro.org/~mathieu.poirier/openCSD/uname.v4.user.sept20.tgz -[3]: Get in touch with us if you know a way to change this. +[3]: It can be changed with: perf-config: + perf config --system buildid.dir=/my/own/buildid/dir -- 2.11.0

5 years, 5 months

coresight perf issue, no samples

by Peng Fan

Hi All, I follow this guide to do test on i.MX8MP, https://elinux.org/images/b/b3/Hardware_Assisted_Tracing_on_ARM.pdf I could see dump with perf report --dump (attached), but when I use perf report --stdio, I met Error: The perf.data data has no samples! # To display the perf.data header info, please use --header/--header-only options. # The perf I use to dump data is not compiled with CORESIGHT support, but the host perf has been compiled with CORESIGHT support. Do you know what might cause this issue? Thanks, Peng.

5 years, 5 months

Two parallel perf sessions tracing processes on two CPUs to shared sink?

by Wojciech Żmuda

Hello, I'm exploring possibilities of tracing two concurrent programs pinned to two CPU cores, sharing the same sink. I tried to spawn two concurrent perf sessions but my results are not satisfying, so I wonder if such possibility exists at all. # taskset -c 1 ./progA # taskset -c 2 ./progB # perf record -e cs_etm/timestamp,@tmc_etr0/u --filter "filter symbolA @./progA" --per-thread --pid $progA_pid # perf record -e cs_etm/timestamp,@tmc_etr0/u --filter "filter symbolA @./progA" --per-thread --pid $progB_pid If progA and progB mostly sleep - I get trace data for both, which is fine. However If at least one of the programs gets more CPU-intensive (loops with arithmetic computations inside, no explicit sleeping nor waiting or IO), I get trace only for the more intensive one. If both are intensive, it seems random which one gets traced. This observations suggests ETR buffer overflow. However, if CPU-intensive versions of progA and progB are scheduled on the same CPU - this method seem to work. I would expect ETR buffer being insufficient in this scenario as well. To enlarge the ETR buffer I experimented with the -m flag, but I'm unable to use more than -m,16M. Independent RSZ polling shows that this value gets programmed, but anything above 16 MB makes TMC-ETR driver complain that cma_alloc failed to get that amount of memory. Anyway, enlarging buffer to 16MB doesn't seem to affect my issue. With bigger buffer my observations are identical. Those observations make me suspect that another technical obstacle might exists, beside possible buffer overflow. I also tried with CPU-wide mode and it seem to work: taskset -c 1 ./progA taskset -c 2 ./progB perf record -e cs_etm/timestamp,@tmc_etr0/u -C 2,3 but this approach is quite limited as filters don't work in CPU-wide mode and perf itself is also traced (which is weird, as I tried setting CPU affinity of perf-record with taskset as well - didn't help). To wrap-up: 1. Is it possible to trace two programs with two perf-record sessions at the same time, sharing a sink? 2. Is it possible to enlarge TMC-ETR buffer above 16MB? I guess SG mode might be an option here, but as I can't really modify my kernel and DT right now. Perhaps there's a possibility to make the kernel allocator work past the 16MB boundary? Thank you and best regards, Wojciech PS Sorry I don't proceed with the Coresight@Zynq MPSoC support I started some time ago. My access to the board is limited recently and it's hard to proceed with kernel development remotely. I hope to get back to it soon.

5 years, 5 months

[PATCH v5 0/5] perf cs-etm: Fix synthesizing instruction samples

by Leo Yan

This patch series is to address issues for synthesizing instruction samples, especially when the instruction sample period is small enough, the current logic cannot synthesize multiple instruction samples within one instruction range packet. Patch 0001 is to swap packets for instruction samples, so this allow option '--itrace=iNNN' can work well. Patch 0002 avoids to reset the last branches for every instruction sample; if reset the last branches for every time generating sample, the later samples in the same range packet cannot use the last branches anymore. Patch 0003 is the fixing for handling different instruction periods, especially for small sample period. Patch 0004 is an optimization for copying last branches; it only copies last branches once if the instruction samples share the same last branches. Patch 0005 is a minor fix for unsigned variable comparison to zero. This patch set has been rebased on the latest perf/core branch; and verified on Juno board with below commands: # perf script --itrace=i2 # perf script --itrace=i2il16 # perf inject --itrace=i2il16 -i perf.data -o perf.data.new # perf inject --itrace=i100il16 -i perf.data -o perf.data.new Changes from v4: * Added Mike's review tag for patch 03; * Added Mathieu's review tags for all patches. Changes from v3: * Refactored patch 0001 with new function cs_etm__packet_swap() (Mike); * Refined instruction sample generation flow with single while loop, which completely uses Mike's suggestions (Mike); * Added Mike's review tags for patch 01/02/04/05. Changes from v2: * Added patch 0001 which is to fix swapping packets for instruction samples; * Refined minor commit logs and comments; * Rebased on the latest perf/core branch. Changes from v1: * Rebased patch set on perf/core branch with latest commit 9fec3cd5fa4a ("perf map: Check if the map still has some refcounts on exit"). Leo Yan (5): perf cs-etm: Swap packets for instruction samples perf cs-etm: Continuously record last branch perf cs-etm: Correct synthesizing instruction samples perf cs-etm: Optimize copying last branches perf cs-etm: Fix unsigned variable comparison to zero tools/perf/util/cs-etm.c | 157 +++++++++++++++++++++++++++------------ 1 file changed, 111 insertions(+), 46 deletions(-) -- 2.17.1

5 years, 5 months

[PATCH 13/23] perf cs-etm: Fix unsigned variable comparison to zero

by Arnaldo Carvalho de Melo

From: Leo Yan <leo.yan(a)linaro.org> The variable 'offset' in function cs_etm__sample() is u64 type, it's not appropriate to check it with 'while (offset > 0)'; this patch changes to 'while (offset)'. Signed-off-by: Leo Yan <leo.yan(a)linaro.org> Reviewed-by: Mathieu Poirier <mathieu.poirier(a)linaro.org> Reviewed-by: Mike Leach <mike.leach(a)linaro.org> Cc: Alexander Shishkin <alexander.shishkin(a)linux.intel.com> Cc: Jiri Olsa <jolsa(a)redhat.com> Cc: Mark Rutland <mark.rutland(a)arm.com> Cc: Namhyung Kim <namhyung(a)kernel.org> Cc: Peter Zijlstra <peterz(a)infradead.org> Cc: Robert Walker <robert.walker(a)arm.com> Cc: Suzuki Poulouse <suzuki.poulose(a)arm.com> Cc: coresight ml <coresight(a)lists.linaro.org> Cc: linux-arm-kernel(a)lists.infradead.org Link: http://lore.kernel.org/lkml/20200219021811.20067-6-leo.yan@linaro.org Signed-off-by: Arnaldo Carvalho de Melo <acme(a)redhat.com> --- tools/perf/util/cs-etm.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tools/perf/util/cs-etm.c b/tools/perf/util/cs-etm.c index 87d9943177bc..62d2f9b9ce1b 100644 --- a/tools/perf/util/cs-etm.c +++ b/tools/perf/util/cs-etm.c @@ -962,7 +962,7 @@ static inline u64 cs_etm__instr_addr(struct cs_etm_queue *etmq, if (packet->isa == CS_ETM_ISA_T32) { u64 addr = packet->start_addr; - while (offset > 0) { + while (offset) { addr += cs_etm__t32_instr_size(etmq, trace_chan_id, addr); offset--; -- 2.21.1

5 years, 5 months

[PATCH 12/23] perf cs-etm: Optimize copying last branches

by Arnaldo Carvalho de Melo

From: Leo Yan <leo.yan(a)linaro.org> If an instruction range packet can generate multiple instruction samples, these samples share the same last branches; it's not necessary to copy the same last branches repeatedly for these samples within the same packet. This patch moves out the last branches copying from function cs_etm__synth_instruction_sample(), and execute it prior to generating instruction samples. Signed-off-by: Leo Yan <leo.yan(a)linaro.org> Reviewed-by: Mathieu Poirier <mathieu.poirier(a)linaro.org> Reviewed-by: Mike Leach <mike.leach(a)linaro.org> Cc: Alexander Shishkin <alexander.shishkin(a)linux.intel.com> Cc: Jiri Olsa <jolsa(a)redhat.com> Cc: Mark Rutland <mark.rutland(a)arm.com> Cc: Namhyung Kim <namhyung(a)kernel.org> Cc: Peter Zijlstra <peterz(a)infradead.org> Cc: Robert Walker <robert.walker(a)arm.com> Cc: Suzuki Poulouse <suzuki.poulose(a)arm.com> Cc: coresight ml <coresight(a)lists.linaro.org> Cc: linux-arm-kernel(a)lists.infradead.org Link: http://lore.kernel.org/lkml/20200219021811.20067-5-leo.yan@linaro.org Signed-off-by: Arnaldo Carvalho de Melo <acme(a)redhat.com> --- tools/perf/util/cs-etm.c | 22 +++++++++++++++++----- 1 file changed, 17 insertions(+), 5 deletions(-) diff --git a/tools/perf/util/cs-etm.c b/tools/perf/util/cs-etm.c index 1ddcc67e13dd..87d9943177bc 100644 --- a/tools/perf/util/cs-etm.c +++ b/tools/perf/util/cs-etm.c @@ -1151,10 +1151,8 @@ static int cs_etm__synth_instruction_sample(struct cs_etm_queue *etmq, cs_etm__copy_insn(etmq, tidq->trace_chan_id, tidq->packet, &sample); - if (etm->synth_opts.last_branch) { - cs_etm__copy_last_branch_rb(etmq, tidq); + if (etm->synth_opts.last_branch) sample.branch_stack = tidq->last_branch; - } if (etm->synth_opts.inject) { ret = cs_etm__inject_event(event, &sample, @@ -1431,6 +1429,10 @@ static int cs_etm__sample(struct cs_etm_queue *etmq, u64 offset = etm->instructions_sample_period - instrs_prev; u64 addr; + /* Prepare last branches for instruction sample */ + if (etm->synth_opts.last_branch) + cs_etm__copy_last_branch_rb(etmq, tidq); + while (tidq->period_instructions >= etm->instructions_sample_period) { /* @@ -1508,6 +1510,11 @@ static int cs_etm__flush(struct cs_etm_queue *etmq, if (etmq->etm->synth_opts.last_branch && tidq->prev_packet->sample_type == CS_ETM_RANGE) { + u64 addr; + + /* Prepare last branches for instruction sample */ + cs_etm__copy_last_branch_rb(etmq, tidq); + /* * Generate a last branch event for the branches left in the * circular buffer at the end of the trace. @@ -1515,7 +1522,7 @@ static int cs_etm__flush(struct cs_etm_queue *etmq, * Use the address of the end of the last reported execution * range */ - u64 addr = cs_etm__last_executed_instr(tidq->prev_packet); + addr = cs_etm__last_executed_instr(tidq->prev_packet); err = cs_etm__synth_instruction_sample( etmq, tidq, addr, @@ -1560,11 +1567,16 @@ static int cs_etm__end_block(struct cs_etm_queue *etmq, */ if (etmq->etm->synth_opts.last_branch && tidq->prev_packet->sample_type == CS_ETM_RANGE) { + u64 addr; + + /* Prepare last branches for instruction sample */ + cs_etm__copy_last_branch_rb(etmq, tidq); + /* * Use the address of the end of the last reported execution * range. */ - u64 addr = cs_etm__last_executed_instr(tidq->prev_packet); + addr = cs_etm__last_executed_instr(tidq->prev_packet); err = cs_etm__synth_instruction_sample( etmq, tidq, addr, -- 2.21.1

5 years, 5 months

[PATCH 11/23] perf cs-etm: Correct synthesizing instruction samples

by Arnaldo Carvalho de Melo

From: Leo Yan <leo.yan(a)linaro.org> When 'etm->instructions_sample_period' is less than 'tidq->period_instructions', the function cs_etm__sample() cannot handle this case properly with its logic. Let's see below flow as an example: - If we set itrace option '--itrace=i4', then function cs_etm__sample() has variables with initialized values: tidq->period_instructions = 0 etm->instructions_sample_period = 4 - When the first packet is coming: packet->instr_count = 10; the number of instructions executed in this packet is 10, thus update period_instructions as below: tidq->period_instructions = 0 + 10 = 10 instrs_over = 10 - 4 = 6 offset = 10 - 6 - 1 = 3 tidq->period_instructions = instrs_over = 6 - When the second packet is coming: packet->instr_count = 10; in the second pass, assume 10 instructions in the trace sample again: tidq->period_instructions = 6 + 10 = 16 instrs_over = 16 - 4 = 12 offset = 10 - 12 - 1 = -3 -> the negative value tidq->period_instructions = instrs_over = 12 So after handle these two packets, there have below issues: The first issue is that cs_etm__instr_addr() returns the address within the current trace sample of the instruction related to offset, so the offset is supposed to be always unsigned value. But in fact, function cs_etm__sample() might calculate a negative offset value (in handling the second packet, the offset is -3) and pass to cs_etm__instr_addr() with u64 type with a big positive integer. The second issue is it only synthesizes 2 samples for sample period = 4. In theory, every packet has 10 instructions so the two packets have total 20 instructions, 20 instructions should generate 5 samples (4 x 5 = 20). This is because cs_etm__sample() only calls once cs_etm__synth_instruction_sample() to generate instruction sample per range packet. This patch fixes the logic in function cs_etm__sample(); the basic idea for handling coming packet is: - To synthesize the first instruction sample, it combines the left instructions from the previous packet and the head of the new packet; then generate continuous samples with sample period; - At the tail of the new packet, if it has the rest instructions, these instructions will be left for the sequential sample. Suggested-by: Mike Leach <mike.leach(a)linaro.org> Signed-off-by: Leo Yan <leo.yan(a)linaro.org> Reviewed-by: Mathieu Poirier <mathieu.poirier(a)linaro.org> Reviewed-by: Mike Leach <mike.leach(a)linaro.org> Cc: Alexander Shishkin <alexander.shishkin(a)linux.intel.com> Cc: Jiri Olsa <jolsa(a)redhat.com> Cc: Mark Rutland <mark.rutland(a)arm.com> Cc: Namhyung Kim <namhyung(a)kernel.org> Cc: Peter Zijlstra <peterz(a)infradead.org> Cc: Robert Walker <robert.walker(a)arm.com> Cc: Suzuki Poulouse <suzuki.poulose(a)arm.com> Cc: coresight ml <coresight(a)lists.linaro.org> Cc: linux-arm-kernel(a)lists.infradead.org Link: http://lore.kernel.org/lkml/20200219021811.20067-4-leo.yan@linaro.org Signed-off-by: Arnaldo Carvalho de Melo <acme(a)redhat.com> --- tools/perf/util/cs-etm.c | 87 ++++++++++++++++++++++++++++++++-------- 1 file changed, 70 insertions(+), 17 deletions(-) diff --git a/tools/perf/util/cs-etm.c b/tools/perf/util/cs-etm.c index 2c4156c5ed09..1ddcc67e13dd 100644 --- a/tools/perf/util/cs-etm.c +++ b/tools/perf/util/cs-etm.c @@ -1358,9 +1358,12 @@ static int cs_etm__sample(struct cs_etm_queue *etmq, struct cs_etm_auxtrace *etm = etmq->etm; int ret; u8 trace_chan_id = tidq->trace_chan_id; - u64 instrs_executed = tidq->packet->instr_count; + u64 instrs_prev; - tidq->period_instructions += instrs_executed; + /* Get instructions remainder from previous packet */ + instrs_prev = tidq->period_instructions; + + tidq->period_instructions += tidq->packet->instr_count; /* * Record a branch when the last instruction in @@ -1378,26 +1381,76 @@ static int cs_etm__sample(struct cs_etm_queue *etmq, * TODO: allow period to be defined in cycles and clock time */ - /* Get number of instructions executed after the sample point */ - u64 instrs_over = tidq->period_instructions - - etm->instructions_sample_period; + /* + * Below diagram demonstrates the instruction samples + * generation flows: + * + * Instrs Instrs Instrs Instrs + * Sample(n) Sample(n+1) Sample(n+2) Sample(n+3) + * | | | | + * V V V V + * -------------------------------------------------- + * ^ ^ + * | | + * Period Period + * instructions(Pi) instructions(Pi') + * + * | | + * \---------------- -----------------/ + * V + * tidq->packet->instr_count + * + * Instrs Sample(n...) are the synthesised samples occurring + * every etm->instructions_sample_period instructions - as + * defined on the perf command line. Sample(n) is being the + * last sample before the current etm packet, n+1 to n+3 + * samples are generated from the current etm packet. + * + * tidq->packet->instr_count represents the number of + * instructions in the current etm packet. + * + * Period instructions (Pi) contains the the number of + * instructions executed after the sample point(n) from the + * previous etm packet. This will always be less than + * etm->instructions_sample_period. + * + * When generate new samples, it combines with two parts + * instructions, one is the tail of the old packet and another + * is the head of the new coming packet, to generate + * sample(n+1); sample(n+2) and sample(n+3) consume the + * instructions with sample period. After sample(n+3), the rest + * instructions will be used by later packet and it is assigned + * to tidq->period_instructions for next round calculation. + */ /* - * Calculate the address of the sampled instruction (-1 as - * sample is reported as though instruction has just been - * executed, but PC has not advanced to next instruction) + * Get the initial offset into the current packet instructions; + * entry conditions ensure that instrs_prev is less than + * etm->instructions_sample_period. */ - u64 offset = (instrs_executed - instrs_over - 1); - u64 addr = cs_etm__instr_addr(etmq, trace_chan_id, - tidq->packet, offset); + u64 offset = etm->instructions_sample_period - instrs_prev; + u64 addr; - ret = cs_etm__synth_instruction_sample( - etmq, tidq, addr, etm->instructions_sample_period); - if (ret) - return ret; + while (tidq->period_instructions >= + etm->instructions_sample_period) { + /* + * Calculate the address of the sampled instruction (-1 + * as sample is reported as though instruction has just + * been executed, but PC has not advanced to next + * instruction) + */ + addr = cs_etm__instr_addr(etmq, trace_chan_id, + tidq->packet, offset - 1); + ret = cs_etm__synth_instruction_sample( + etmq, tidq, addr, + etm->instructions_sample_period); + if (ret) + return ret; - /* Carry remaining instructions into next sample period */ - tidq->period_instructions = instrs_over; + offset += etm->instructions_sample_period; + tidq->period_instructions -= + etm->instructions_sample_period; + } } if (etm->sample_branches) { -- 2.21.1

5 years, 5 months

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

CoreSight