From: Peter Zijlstra Subject: Re: [PATCH 2/9] perf/core: Add PERF_SAMPLE_CGROUP feature Date: Fri, 20 Dec 2019 17:48:02 +0100 Message-ID: <20191220164802.GO2827@hirez.programming.kicks-ass.net> References: <20191220043253.3278951-1-namhyung@kernel.org> <20191220043253.3278951-3-namhyung@kernel.org> <20191220152324.GG2914998@devbig004.ftw2.facebook.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20191220152324.GG2914998@devbig004.ftw2.facebook.com> Sender: linux-kernel-owner@vger.kernel.org To: Tejun Heo Cc: Namhyung Kim , Ingo Molnar , Arnaldo Carvalho de Melo , Jiri Olsa , Alexander Shishkin , Mark Rutland , Stephane Eranian , LKML , linux-perf-users@vger.kernel.org, Li Zefan , Johannes Weiner List-Id: linux-perf-users.vger.kernel.org On Fri, Dec 20, 2019 at 07:23:24AM -0800, Tejun Heo wrote: > On Fri, Dec 20, 2019 at 01:32:46PM +0900, Namhyung Kim wrote: > > The PERF_SAMPLE_CGROUP bit is to save (perf_event) cgroup information > > in the sample. It will add a 64-bit id to identify current cgroup and > > it's the file handle in the cgroup file system. Userspace should use > > this information with PERF_RECORD_CGROUP event to match which cgroup > > it belongs. > > You don't need PERF_RECORD_CGROUP for that. Something like the > following should work. > > struct { > struct file_handle fh; > char stor[MAX_HANDLE_SZ]; > } fh_store; > struct file_handle *fh = &fh_store; > > fh->handle_type = 0xfe; // FILEID_KERNFS > fh->handle_bytes = sizeof(u64); > *(u64 *)fh->f_handle = cgrp_id; > > mnt_fd = open('/sys/fs/cgroup', O_RDONLY); > fd = open_by_handle_at(mnt_fd, fh, O_RDONLY); > > snprintf(proc_path, PATH_MAX, "/proc/self/fd/%d", fd); > readlink(proc_path, cgrp_path, PATH_MAX); That assumes the cgroup is still in existence, which might not be the case I presume.