From: Alexey Budankov Subject: Re: [RFC 0/5] perf: Per PMU access controls (paranoid setting) References: <20180919122751.12439-1-tvrtko.ursulin@linux.intel.com> Message-ID: Date: Mon, 1 Oct 2018 09:27:07 +0300 MIME-Version: 1.0 In-Reply-To: Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit To: Jann Horn , Kees Cook Cc: tvrtko.ursulin@linux.intel.com, Thomas Gleixner , tursulin@ursulin.net, kernel list , Peter Zijlstra , the arch/x86 maintainers , "H . Peter Anvin" , acme@kernel.org, alexander.shishkin@linux.intel.com, jolsa@redhat.com, namhyung@kernel.org, maddy@linux.vnet.ibm.com, Andi Kleen , kernel-hardening@lists.openwall.com List-ID: Hello Jann and Kees, On 29.09.2018 1:02, Jann Horn wrote: > Ah, I guess the answer is "0", since you want to see data about what > other users are doing. > > Does the i915 PMU expose sampling events, counting events, or both? > The thing about sampling events is that they AFAIK always let the user > pick arbitrary data to collect - like register contents, or userspace > stack memory -, and independent of the performance counter being > monitored, this kind of access should not be permitted to other > contexts. (But it might be that I misunderstand how perf works - I'm > not super familiar with its API.) > Currently *core* paranoid >= 1 (per-process mode) prevents simultaneous sampling on CPU events (perf record) and reading of uncore HW counters (perf stat -I), because uncore counters count system wide and that is allowed only when *core* paranoid <= 0. Uncore counts collected simultaneously with CPU event samples can be correlated using timestamps taken from some common system clock e.g. CLOCK_MONOTONIC_RAW. Could it be secure enough to still allow reading of system wide uncore HW counters when sampling of CPU events is limited to specific processes by *core* paranoid >= 1? Thanks, Alexey