From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-3.8 required=3.0 tests=BAYES_00,DKIM_INVALID, DKIM_SIGNED,MAILING_LIST_MULTI,SPF_HELO_NONE,SPF_PASS autolearn=no autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 86663C433E2 for ; Fri, 10 Jul 2020 17:32:52 +0000 (UTC) Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPS id 5983F20725 for ; Fri, 10 Jul 2020 17:32:52 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=fail reason="signature verification failed" (1024-bit key) header.d=kernel.org header.i=@kernel.org header.b="KCfbDr1T" DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 5983F20725 Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=kernel.org Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=intel-gfx-bounces@lists.freedesktop.org Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 9166F6ECBD; Fri, 10 Jul 2020 17:32:51 +0000 (UTC) Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by gabe.freedesktop.org (Postfix) with ESMTPS id 4528D6EC96 for ; Fri, 10 Jul 2020 17:09:14 +0000 (UTC) Received: from quaco.ghostprotocols.net (unknown [179.97.37.151]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPSA id BB51920657; Fri, 10 Jul 2020 17:09:13 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=default; t=1594400954; bh=OpaE6wanvPY6oZYoZQ+5OUnMUqe0KlNUtH+bJtmG/Gc=; h=Date:From:To:Cc:Subject:References:In-Reply-To:From; b=KCfbDr1TEthee8alZ4uKm4NLesJMTPj5oLRDmr/1v3LAWfZKzyv/u5BBEV22aVgAm 61lHNT9AC/KhBT2L1whQiaNkLtfJS50U/F0wGXlbwRrno8pcuenKgFluD9eoR2Wy2O 1BU0mv7DrerCe/r9aHYhBntMne3BKtz1SQEuQ4f4= Received: by quaco.ghostprotocols.net (Postfix, from userid 1000) id E0F00405FF; Fri, 10 Jul 2020 14:09:11 -0300 (-03) Date: Fri, 10 Jul 2020 14:09:11 -0300 From: Arnaldo Carvalho de Melo To: Alexey Budankov Message-ID: <20200710170911.GD7487@kernel.org> References: <76718dc6-5483-5e2e-85b8-64e70306ee1f@linux.ibm.com> <7776fa40-6c65-2aa6-1322-eb3a01201000@linux.intel.com> MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: <7776fa40-6c65-2aa6-1322-eb3a01201000@linux.intel.com> X-Url: http://acmel.wordpress.com X-Mailman-Approved-At: Fri, 10 Jul 2020 17:32:50 +0000 Subject: Re: [Intel-gfx] [PATCH v8 00/12] Introduce CAP_PERFMON to secure system performance monitoring and observability X-BeenThere: intel-gfx@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Intel graphics driver community testing & development List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: Ravi Bangoria , Song Liu , Andi Kleen , linux-man@vger.kernel.org, "linux-doc@vger.kernel.org" , Peter Zijlstra , linux-kernel , "intel-gfx@lists.freedesktop.org" , Igor Lubashev , Alexei Starovoitov , Stephane Eranian , James Morris , "selinux@vger.kernel.org" , "linux-security-module@vger.kernel.org" , Ingo Molnar , Namhyung Kim , Thomas Gleixner , Jiri Olsa , Serge Hallyn Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Errors-To: intel-gfx-bounces@lists.freedesktop.org Sender: "Intel-gfx" Em Fri, Jul 10, 2020 at 05:30:50PM +0300, Alexey Budankov escreveu: > On 10.07.2020 16:31, Ravi Bangoria wrote: > >> Currently access to perf_events, i915_perf and other performance > >> monitoring and observability subsystems of the kernel is open only for > >> a privileged process [1] with CAP_SYS_ADMIN capability enabled in the > >> process effective set [2]. > >> This patch set introduces CAP_PERFMON capability designed to secure > >> system performance monitoring and observability operations so that > >> CAP_PERFMON would assist CAP_SYS_ADMIN capability in its governing role > >> for performance monitoring and observability subsystems of the kernel. = > > I'm seeing an issue with CAP_PERFMON when I try to record data for a > > specific target. I don't know whether this is sort of a regression or > > an expected behavior. = > Thanks for reporting and root causing this case. The behavior looks like > kind of expected since currently CAP_PERFMON takes over the related part > of CAP_SYS_ADMIN credentials only. Actually Perf security docs [1] say > that access control is also subject to CAP_SYS_PTRACE credentials. I think that stating that in the error message would be helpful, after all, who reads docs? 8-) I.e., this: $ ./perf stat ls =A0 Error: =A0 Access to performance monitoring and observability operations is limite= d. $ Could become: $ ./perf stat ls =A0 Error: =A0 Access to performance monitoring and observability operations is limite= d. Right now only CAP_PERFMON is granted, you may need CAP_SYS_PTRACE. $ - Arnaldo = > CAP_PERFMON could be used to extend and substitute ptrace_may_access() > check in perf_events subsystem to simplify user experience at least in > this specific case. > = > Alexei > = > [1] https://www.kernel.org/doc/html/latest/admin-guide/perf-security.html > = > > = > > Without setting CAP_PERFMON: > > = > > =A0 $ getcap ./perf > > =A0 $ ./perf stat -a ls > > =A0=A0=A0 Error: > > =A0=A0=A0 Access to performance monitoring and observability operations= is limited. > > =A0 $ ./perf stat ls > > =A0=A0=A0 Performance counter stats for 'ls': > > =A0=A0 =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 2.06 msec task-= clock:u=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.418 CPUs utiliz= ed > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 0=A0=A0=A0=A0= =A0 context-switches:u=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.000 K/sec > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 0=A0=A0=A0=A0= =A0 cpu-migrations:u=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.000 K/sec > > = > > With CAP_PERFMON: > > = > > =A0 $ getcap ./perf > > =A0=A0=A0 ./perf =3D cap_perfmon+ep > > =A0 $ ./perf stat -a ls > > =A0=A0=A0 Performance counter stats for 'system wide': > > =A0=A0 =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 142.42 msec cpu-clock= =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0 25.062 CPUs utiliz= ed > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 182=A0=A0=A0=A0=A0 = context-switches=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.001 M/sec > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 48=A0=A0=A0=A0= =A0 cpu-migrations=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.337 K/sec > > =A0 $ ./perf stat ls > > =A0=A0=A0 Error: > > =A0=A0=A0 Access to performance monitoring and observability operations= is limited. > > = > > Am I missing something silly? > > = > > Analysis: > > --------- > > A bit more analysis lead me to below kernel code fs/exec.c: > > = > > =A0 begin_new_exec() > > =A0 { > > =A0=A0=A0=A0=A0=A0=A0 ... > > =A0=A0=A0=A0=A0=A0=A0 if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NON= DUMP || > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 !(uid_eq(current_euid(), current_uid(= )) && > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 gid_eq(current_egid(), current_= gid()))) > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 set_dumpable(current->mm,= suid_dumpable); > > =A0=A0=A0=A0=A0=A0=A0 else > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 set_dumpable(current->mm,= SUID_DUMP_USER); > > = > > =A0=A0=A0=A0=A0=A0=A0 ... > > =A0=A0=A0=A0=A0=A0=A0 commit_creds(bprm->cred); > > =A0 } > > = > > When I execute './perf stat ls', it's going into else condition and thu= s sets > > dumpable flag as SUID_DUMP_USER. Then in commit_creds(): > > = > > =A0 int commit_creds(struct cred *new) > > =A0 { > > =A0=A0=A0=A0=A0=A0=A0 ... > > =A0=A0=A0=A0=A0=A0=A0 /* dumpability changes */ > > =A0=A0=A0=A0=A0=A0=A0 if (... > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 !cred_cap_issubset(old, new)) { > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 if (task->mm) > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 s= et_dumpable(task->mm, suid_dumpable); > > =A0 } > > = > > !cred_cap_issubset(old, new) fails for perf without any capability and = thus > > it doesn't execute set_dumpable(). Whereas that condition passes for pe= rf > > with CAP_PERFMON and thus it overwrites old value (SUID_DUMP_USER) with > > suid_dumpable in mm_flags. On an Ubuntu, suid_dumpable default value is > > SUID_DUMP_ROOT. On Fedora, it's SUID_DUMP_DISABLE. (/proc/sys/fs/suid_d= umpable). > > = > > Now while opening an event: > > = > > =A0 perf_event_open() > > =A0=A0=A0 ptrace_may_access() > > =A0=A0=A0=A0=A0 __ptrace_may_access() { > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 ... > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 if (mm && > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 ((get_dumpabl= e(mm) !=3D SUID_DUMP_USER) && > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 !ptrace_ha= s_cap(cred, mm->user_ns, mode))) > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 return -EPERM; > > =A0=A0=A0=A0=A0 } > > = > > This if condition passes for perf with CAP_PERFMON and thus it returns = -EPERM. > > But it fails for perf without CAP_PERFMON and thus it goes ahead and re= turns > > success. So opening an event fails when perf has CAP_PREFMON and tries = to open > > process specific event as normal user. > > = > > Workarounds: > > ------------ > > Based on above analysis, I found couple of workarounds (examples are on > > Ubuntu 18.04.4 powerpc): > > = > > Workaround1: > > Setting SUID_DUMP_USER as default (in /proc/sys/fs/suid_dumpable) solve= s the > > issue. > > = > > =A0 # echo 1 > /proc/sys/fs/suid_dumpable > > =A0 $ getcap ./perf > > =A0=A0=A0 ./perf =3D cap_perfmon+ep > > =A0 $ ./perf stat ls > > =A0=A0=A0 Performance counter stats for 'ls': > > =A0=A0 =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 1.47 msec task-= clock=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.806 CPUs ut= ilized > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 0=A0=A0=A0=A0= =A0 context-switches=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.000 K/sec > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 0=A0=A0=A0=A0= =A0 cpu-migrations=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.000 K/sec > > = > > Workaround2: > > Using CAP_SYS_PTRACE along with CAP_PERFMON solves the issue. > > = > > =A0 $ cat /proc/sys/fs/suid_dumpable > > =A0=A0=A0 2 > > =A0 # setcap "cap_perfmon,cap_sys_ptrace=3Dep" ./perf > > =A0 $ ./perf stat ls > > =A0=A0=A0 Performance counter stats for 'ls': > > =A0=A0 =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 1.41 msec task-= clock=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.826 CPUs ut= ilized > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 0=A0=A0=A0=A0= =A0 context-switches=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.000 K/sec > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 0=A0=A0=A0=A0= =A0 cpu-migrations=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.000 K/sec > > = > > Workaround3: > > Adding CAP_PERFMON to parent of perf (/bin/bash) also solves the issue. > > = > > =A0 $ cat /proc/sys/fs/suid_dumpable > > =A0=A0=A0 2 > > =A0 # setcap "cap_perfmon=3Dep" /bin/bash > > =A0 # setcap "cap_perfmon=3Dep" ./perf > > =A0 $ bash > > =A0 $ ./perf stat ls > > =A0=A0=A0 Performance counter stats for 'ls': > > =A0=A0 =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 1.47 msec task-= clock=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.806 CPUs ut= ilized > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 0=A0=A0=A0=A0= =A0 context-switches=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.000 K/sec > > =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 0=A0=A0=A0=A0= =A0 cpu-migrations=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0 #=A0=A0=A0 0.000 K/sec > > = > > - Ravi -- = - Arnaldo _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/intel-gfx