Description of problem: perf_event_tests：huge_group_start失败，Unexpected error at 1021 Argument list too long，日志如下： ./tests/corner_cases/huge_group_start data size=65544 Unexpected error at 1021 Argument list too long Testing start of max event group... FAILED Version-Release number of selected component (if applicable): perf -v perf version 5.10.134-13_rc1.an8.aarch64 版本信息： cat /etc/os-release NAME="Anolis OS" VERSION="8.6" ID="anolis" ID_LIKE="rhel fedora centos" VERSION_ID="8.6" PLATFORM_ID="platform:an8" PRETTY_NAME="Anolis OS 8.6" ANSI_COLOR="0;31" HOME_URL="https://openanolis.cn/" 内核信息： uname -r 5.10.134-13_rc1.an8.aarch64 cpu信息： lscpu Architecture: aarch64 Byte Order: Little Endian CPU(s): 8 On-line CPU(s) list: 0-7 Thread(s) per core: 1 Core(s) per socket: 8 Socket(s): 1 NUMA node(s): 1 Vendor ID: ARM BIOS Vendor ID: Alibaba Cloud Model: 0 BIOS Model name: virt-rhel7.6.0 Stepping: r0p0 CPU max MHz: 2750.0000 CPU min MHz: 2750.0000 BogoMIPS: 100.00 L1d cache: 64K L1i cache: 64K L2 cache: 1024K L3 cache: 65536K NUMA node0 CPU(s): 0-7 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 sm3 sm4 asimddp sha512 sve asimdfhm dit uscat ilrcpc flagm ssbs sb dcpodp sve2 sveaes svepmull svebitperm svesha3 svesm4 flagm2 frint svei8mm svebf16 i8mm bf16 dgh 内存信息： free -h total used free shared buff/cache available Mem: 30Gi 295Mi 23Gi 0.0Ki 6.9Gi 29Gi Swap: 0B 0B 0B How reproducible: 必现 Steps to Reproduce: git clone git://github.com/deater/perf_event_testscd perf_event_tests cd perf_event_tests make && make install cd /lkp/benchmarks/perf_event_tests/ ##huge_group_start ./tests/corner_cases/huge_group_start data size=65544 Unexpected error at 1021 Argument list too long Testing start of max event group... FAILED Actual results: 用例执行fail Expected results: 用例执行pass Additional info:
git clone git://github.com/deater/perf_event_testscd perf_event_tests 下载地址不对，修改为下列地址： http://gitlab-sp.alibaba-inc.com/AKTF/perf_event_tests.git
现象： ./tests/corner_cases/huge_group_start data size=65544 Unexpected error at 1021 Argument list too long Testing start of max event group... 分析源码： // sys_perf_event_open: if (!perf_event_validate_size(event)) { err = -E2BIG; goto err_locked; } 其中，E2BIG定义为： #define E2BIG 7 /* Argument list too long */ 具体看： // perf_event_validate_size __perf_event_read_size(event, event->group_leader->nr_siblings + 1); //... /* * Sum the lot; should not exceed the 64k limit we have on records. * Conservative limit to allow for callchains and other variable fields. */ if (event->read_size + event->header_size + event->id_header_size + sizeof(struct perf_event_header) >= 16*1024) return false; 这里在1021个event，超出限制。 增加打印： [ 667.626187] read_size=56, header_size=0, id_header_size=0, sizeof=8 [ 667.626203] read_size=56, header_size=0, id_header_size=0, sizeof=8 [ 667.626264] read_size=72, header_size=0, id_header_size=0, sizeof=8 [ 667.626277] read_size=88, header_size=0, id_header_size=0, sizeof=8 [ 667.626289] read_size=104, header_size=0, id_header_size=0, sizeof=8 ... [ 667.634312] read_size=16312, header_size=0, id_header_size=0, sizeof=8 [ 667.634326] read_size=16328, header_size=0, id_header_size=0, sizeof=8 [ 667.634339] read_size=16344, header_size=0, id_header_size=0, sizeof=8 [ 667.634350] read_size=16360, header_size=0, id_header_size=0, sizeof=8 [ 667.634365] read_size=16376, header_size=0, id_header_size=0, sizeof=8 // __perf_event_read_size static void __perf_event_read_size(struct perf_event *event, int nr_siblings) { int entry = sizeof(u64); /* value */ int size = 0; int nr = 1; if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) // valid size += sizeof(u64); if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) // valid size += sizeof(u64); if (event->attr.read_format & PERF_FORMAT_ID) // valid entry += sizeof(u64); if (event->attr.read_format & PERF_FORMAT_LOST) entry += sizeof(u64); if (event->attr.read_format & PERF_FORMAT_GROUP) { // valid nr += nr_siblings; size += sizeof(u64); } size += entry * nr; event->read_size = size; } read_size计算方法：event->read_size = size + entry * nr; 根据测试用例： pe.read_format=PERF_FORMAT_GROUP|PERF_FORMAT_ID|PERF_FORMAT_TOTAL_TIME_RUNNING|PERF_FORMAT_TOTAL_TIME_ENABLED; https://github.com/deater/perf_event_tests/blob/master/tests/corner_cases/huge_group_start.c#L104 size=8+8+8=24 entry=8+8=16 nr_siblings初始值为1 nr初始值为1+1=2 第1个event read_size=24+16*2=56 第1021个event read_size=24+16*1022=16376，加上sizeof(struct perf_event_header)=8, 16376+8=16384>= 16*1024，perf_event_validate_size返回false，符合预期。 在upstream Linux 6.0内核测试，现象相同。
[root@localhost.localdomain /root] #uname -a Linux localhost.localdomain 5.10.112-11.an8.aarch64 #1 SMP Tue May 24 15:54:43 CST 2022 aarch64 aarch64 aarch64 GNU/Linux [root@localhost.localdomain /root] #./huge_group_start data size=65544 Unexpected error at 1021 Argument list too long Testing start of max event group... FAILED 5.10.112-11.an8.aarch64存在同样的问题。
(In reply to XueShuai from comment #2) > 现象： > ./tests/corner_cases/huge_group_start > data size=65544 > Unexpected error at 1021 Argument list too long > Testing start of max event group... > > 分析源码： > // sys_perf_event_open: > if (!perf_event_validate_size(event)) { > err = -E2BIG; > goto err_locked; > } > > 其中，E2BIG定义为： > #define E2BIG 7 /* Argument list too long */ > > 具体看： > // perf_event_validate_size > __perf_event_read_size(event, event->group_leader->nr_siblings + 1); > //... > /* > * Sum the lot; should not exceed the 64k limit we have on records. > * Conservative limit to allow for callchains and other variable fields. > */ > if (event->read_size + event->header_size + > event->id_header_size + sizeof(struct perf_event_header) >= 16*1024) > return false; > > 这里在1021个event，超出限制。 > > 增加打印： > [ 667.626187] read_size=56, header_size=0, id_header_size=0, sizeof=8 > [ 667.626203] read_size=56, header_size=0, id_header_size=0, sizeof=8 > [ 667.626264] read_size=72, header_size=0, id_header_size=0, sizeof=8 > [ 667.626277] read_size=88, header_size=0, id_header_size=0, sizeof=8 > [ 667.626289] read_size=104, header_size=0, id_header_size=0, sizeof=8 > ... > [ 667.634312] read_size=16312, header_size=0, id_header_size=0, sizeof=8 > [ 667.634326] read_size=16328, header_size=0, id_header_size=0, sizeof=8 > [ 667.634339] read_size=16344, header_size=0, id_header_size=0, sizeof=8 > [ 667.634350] read_size=16360, header_size=0, id_header_size=0, sizeof=8 > [ 667.634365] read_size=16376, header_size=0, id_header_size=0, sizeof=8 > > > // __perf_event_read_size > static void __perf_event_read_size(struct perf_event *event, int nr_siblings) > { > int entry = sizeof(u64); /* value */ > int size = 0; > int nr = 1; > > if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_ENABLED) // valid > size += sizeof(u64); > > if (event->attr.read_format & PERF_FORMAT_TOTAL_TIME_RUNNING) // valid > size += sizeof(u64); > > if (event->attr.read_format & PERF_FORMAT_ID) // valid > entry += sizeof(u64); > > if (event->attr.read_format & PERF_FORMAT_LOST) > entry += sizeof(u64); > > if (event->attr.read_format & PERF_FORMAT_GROUP) { // valid > nr += nr_siblings; > size += sizeof(u64); > } > > size += entry * nr; > event->read_size = size; > } > > read_size计算方法：event->read_size = size + entry * nr; > > 根据测试用例： > pe. > read_format=PERF_FORMAT_GROUP|PERF_FORMAT_ID|PERF_FORMAT_TOTAL_TIME_RUNNING|P > ERF_FORMAT_TOTAL_TIME_ENABLED; > https://github.com/deater/perf_event_tests/blob/master/tests/corner_cases/ > huge_group_start.c#L104 > > > size=8+8+8=24 > entry=8+8=16 > nr_siblings初始值为1 > nr初始值为1+1=2 > > 第1个event read_size=24+16*2=56 > 第1021个event read_size=24+16*1022=16376，加上sizeof(struct perf_event_header)=8, > 16376+8=16384>= 16*1024，perf_event_validate_size返回false，符合预期。 > > 在upstream Linux 6.0内核测试，现象相同。 @yunhe123 反馈nightly机器（鲲鹏）上跑的是PASSED，测试结果如下： #./huge_group_start data size=65544 Ran out of file descriptors at 1021 Too many open files Trying to start 1021 events! Trying to read 65536 bytes into data Read 16360 bytes Testing start of max event group... PASSED 从Too many open files可以看到，因为打开的fd超过限制，导致测试了1021个event。 通过ulimit解除fd限制，也会出现本问题。 #ulimit -n 2048 #./huge_group_start data size=65544 Unexpected error at 1021 Argument list too long Testing start of max event group... FAILED 建议：测试时指定max limit。 #./huge_group_start 1021 data size=65544 Trying to start 1021 events! Trying to read 65536 bytes into data Read 16360 bytes Testing start of max event group... PASSED 结论：根据上面的分析，现象符合预期，置为invaild。
*** Bug 5604 has been marked as a duplicate of this bug. ***