Messages in this thread | | | Date | Mon, 14 Sep 2015 09:37:16 +0800 | From | "Wangnan (F)" <> | Subject | Re: [RFC 00/13] perf_env/CPU socket reorg/fixes |
| |
On 2015/9/11 21:36, Arnaldo Carvalho de Melo wrote: > Em Fri, Sep 11, 2015 at 10:30:52AM -0300, Arnaldo Carvalho de Melo escreveu: >> Em Fri, Sep 11, 2015 at 10:29:37AM -0300, Arnaldo Carvalho de Melo escreveu: >>> Em Fri, Sep 11, 2015 at 10:03:39AM -0300, Arnaldo Carvalho de Melo escreveu: >>>> Em Fri, Sep 11, 2015 at 08:20:54PM +0800, Wangnan (F) escreveu: >>>>> I have tested patch 1 to 10. They looks good to me except patch 4/13. Please >>>> Ok, I'll take that as a Tested-by: you for 1-10 with 4/13 having the >>>> checks added, ok? >>> >>>>> see my email in that thread. >>> >>>> I add those checks. >>> Ok, below is the diff for adding the checks. The get_{core,socket}_id >>> functions should be moved to tools/lib/api/cpu.[ch], using the same >>> interface as cpu__get_max_freq(&value), using the int return value to >>> propagate the precise error, etc. Will do it in a follow up patch. >> Humm, but then, what happens if a CPU is offline? I'm checking it now... > # cat /sys/devices/system/cpu/cpu3/topology/core_id > 3 > # cat > /sys/devices/system/cpu/cpu3/topology/physical_package_id > 0 > # echo 0 > /sys/devices/system/cpu/cpu3/online > # cat > /sys/devices/system/cpu/cpu3/topology/physical_package_id > cat: /sys/devices/system/cpu/cpu3/topology/physical_package_id: No such file or directory > # cat /sys/devices/system/cpu/cpu3/topology/core_id > cat: /sys/devices/system/cpu/cpu3/topology/core_id: No such file or directory > # > > So we shouldn't check the result, right? We could further validate it by > checking: > > # cat /sys/devices/system/cpu/cpu3/online > 0 > # > > But assuming that not being able to access it means it is offline looks > almost reasonable, if not strictly correct, so I'm removing the tests > and will revisit this when I move those functions to > tools/lib/api/cpu.[ch].
I highly suspect that setting 'online' to 0 is not the only reason of the removal of topology directory. Testing the existance of the two files should be better.
Thank you.
> - Arnaldo > >>> diff --git a/tools/perf/util/env.c b/tools/perf/util/env.c >>> index 6af4f7c36820..2e4cad84197b 100644 >>> --- a/tools/perf/util/env.c >>> +++ b/tools/perf/util/env.c >>> @@ -60,7 +60,7 @@ out_enomem: >>> >>> int perf_env__read_cpu_topology_map(struct perf_env *env) >>> { >>> - int cpu, nr_cpus; >>> + int cpu, nr_cpus, err; >>> >>> if (env->cpu != NULL) >>> return 0; >>> @@ -77,10 +77,17 @@ int perf_env__read_cpu_topology_map(struct perf_env *env) >>> return -ENOMEM; >>> >>> for (cpu = 0; cpu < nr_cpus; ++cpu) { >>> - env->cpu[cpu].core_id = cpu_map__get_core_id(cpu); >>> - env->cpu[cpu].socket_id = cpu_map__get_socket_id(cpu); >>> + err = env->cpu[cpu].core_id = cpu_map__get_core_id(cpu); >>> + if (err < 0) >>> + goto out_free; >>> + err = env->cpu[cpu].socket_id = cpu_map__get_socket_id(cpu); >>> + if (err < 0) >>> + goto out_free; >>> } >>> >>> - env->nr_cpus_avail = nr_cpus; >>> return 0; >>> + >>> +out_free: >>> + zfree(&env->cpu); >>> + return err; >>> }
| |