[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <e7e272e0-a7ad-1a97-a081-acf3bfa42b6f@arm.com>
Date: Tue, 17 Jul 2018 13:27:51 +0100
From: Robin Murphy <robin.murphy@....com>
To: Olof Johansson <olof@...om.net>,
Marc Zyngier <marc.zyngier@....com>
Cc: Catalin Marinas <catalin.marinas@....com>,
Will Deacon <will.deacon@....com>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Linux ARM Mailing List <linux-arm-kernel@...ts.infradead.org>
Subject: Re: [PATCH] arm64: cpuinfo: Include cleartext implementer and part
strings
On 16/07/18 15:34, Olof Johansson wrote:
[...]
> There's also a growing expectation of the system to behave more like
> x86, especially when it comes to trivial ways of detecting what kind
> of machine you are running on. On x86 it's been trivial to look at
> /proc/cpuinfo since it is provided by the platform there. Nobody wants
> to change their implementation from having to read a file to exec a
> program, dealing with errors, and parsing its output (sure, easier
> from scripts, but more awkward from real source).
The thing with the "be more like x86" argument is that it never actually
stands up to scrutiny. What is proposed here (again) is for the kernel
to decode the raw MIDR description of the *CPU core microarchitecture*
to a human-readable string. What does the nearest Intel box say about
its microarchitecture name?
cpu family : 6
model : 79
OK, I had to go and look up that that apparently implies "Broadwell", so
the x86 kernel is in fact no more helpful in that regard than arm64.
Furthermore, telling the user that they have have 4 "Cortex-A53" cores
vs. 4 "part: 0xd03" cores doesn't actually give them any more meaningful
information about their *system*, because that example reflects about
half the low-to-mid-range SoCs on the market today. What the "model
name" on x86 describes is a specific processor SKU, of which Arm has no
*architectural* equivalent, for reasons which should hopefully be
obvious from the fundamental differences in the business models driving
the respective architectures.
And yet whenever /proc/cpuinfo/ comes up, nobody ever seems to ask for
some kind of SoC identifier in there; it's always just decoding MIDR
into a microarchitecture name, which implies that all these people want
is to see a string for the sake of seeing a string, regardless of how
meaningful it actually might be (and yes, I do realise that in *some*
cases a CPU core is essentially unique to a particular SoC, but we have
to look at the big picture here).
In summary; when a small minority of users are complaining that oranges
aren't apples, the sensible response is not to start debating whether to
paint the oranges green or red.
Robin.
Powered by blists - more mailing lists