linux-kernel - Re: [PATCH] Revert 9fc2105aeaaf56b0cf75296a84702d0f9e64437b to fix pyaudio (and probably more)

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <alpine.LFD.2.11.1501041526390.18844@knanqh.ubzr>
Date:	Sun, 4 Jan 2015 15:45:22 -0500 (EST)
From:	Nicolas Pitre <nicolas.pitre@...aro.org>
To:	Linus Torvalds <torvalds@...ux-foundation.org>
cc:	Pavel Machek <pavel@....cz>, Marc Zyngier <marc.zyngier@....com>,
	Russell King <rmk+kernel@....linux.org.uk>,
	kernel list <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] Revert 9fc2105aeaaf56b0cf75296a84702d0f9e64437b to fix
 pyaudio (and probably more)

On Sun, 4 Jan 2015, Linus Torvalds wrote:

> On Sun, Jan 4, 2015 at 12:03 PM, Nicolas Pitre <nicolas.pitre@...aro.org> wrote:
> >
> > NAK.
> 
> To quote the standard response for people who ignore regressions:
> 
>   "SHUT THE FUCK UP"
> 
> you cannot NAK regression fixes. Seriously.
> 
> I don't understand how people can't get this simple thing.

OK, let's look at the issue more deeply then.

We're talking about exporting a bogomips entry via /proc/cpuinfo.

On ARM the calibration loop is no longer performed because udelay() is 
based on a hardware counter these days.  Why? because the calibration 
was completely unreliable in the presence of CPU frequency scaling, and 
other modern hardware niceties.  

Still, for a while, we did export the calibration loop value for 
/proc/cpuinfo's sake.  Yet people complained because the exported value 
was "wrong".  Of course it is wrong as it is just impossible to get 
"right", unless you have antique hardware that is.

> No regressions. If user mode breaks, it is absolutely *never*
> acceptable to blame user mode.
> 
> Occasionally there may be major overriding reasons (security issue,
> whatever), but even then we bend over *backwards* trying to make sure
> breakage is minimal.

I'm all for ensuring breakage is minimal.  That'd imply settling on a 
good phony bogomips value that is no more broken than the real one was 
when it was exported.

> Christ people. Why does this even have to be discussed any more?

Because we're discussing a choice between two evils.  The actual 
regression happened when people upgraded their antique hardware and 
expected the bogomips to still work with new hardware.  Even if the 
$subject commit is reverted, that won't solve the wrong bogomips problem 
and different people will start complaining again.

As I mentioned earlier, putting the entry back is not a problem if we 
may find the best phony number to go along with it.

Nicolas
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/