Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Vince Weaver <vincent.weaver@maine.edu>
Date:   Fri, 17 Jun 2022 16:11:34 -0400 (EDT)
To:     linux-kernel@vger.kernel.org, linux-perf-users@vger.kernel.org
cc:     Mark Rutland <mark.rutland@arm.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Arnaldo Carvalho de Melo <acme@kernel.org>,
        Namhyung Kim <namhyung@kernel.org>,
        Will Deacon <will@kernel.org>,
        Alexander Shishkin <alexander.shishkin@linux.intel.com>,
        Ingo Molnar <mingo@redhat.com>
Message-ID: <db2031ec-cf4-4323-9d3-8262849add3d@maine.edu>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Precedence: bulk

Hello

Is the perf_mmap__read_self() interface as found in tools/lib/perf/mmap.c
the "official" interface for accessing the perf mmap info from
userspace?

Are the READ_ONCE() calls required?  If they are left out, will accessing 
the mmap interface potentially fail?  Has this ever been seen in practice?

Part of why I am asking is both
	tools/lib/perf/mmap.c
and
	tools/linux/compiler.h (which defines READ_ONCE)
have SPDX headers indicating they are GPL-2.0 licensed, which means it 
seems like it would not be possible to easily use the perf mmap interface 
from BSD-licensed code.  Would it be possible to get those two files 
re-licensed?

The (BSD-licensed) PAPI is currently using a mmap reading interface 
based on early documentation for the feature, but it isn't 100% the same 
as the version from libperf (and isn't using READ_ONCE).  Life would be 
easier if we could use the perf version of the code because then we would 
have one less variable to deal with when trying to track down issues.

Vince Weaver
vincent.weaver@maine.edu