linux-kernel - Re: [PATCH] Turn rdmsr, rdtsc into inline functions, clarify names

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: <1154832963.29151.21.camel@localhost.localdomain>
Date:	Sun, 06 Aug 2006 12:56:03 +1000
From:	Rusty Russell <rusty@...tcorp.com.au>
To:	Andi Kleen <ak@....de>
Cc:	lkml - Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] Turn rdmsr, rdtsc into inline functions, clarify names

On Sun, 2006-08-06 at 04:38 +0200, Andi Kleen wrote:
> On Sat, Aug 05, 2006 at 07:47:41PM +1000, Rusty Russell wrote:
> > [Andi, sorry, x86_64 part untested, so sending straight to you]
> > 
> > rdmsr and rdtsc are macros, altering their arguments directly.  An
> > inline function would offer decent typechecking, but needs to take
> > pointer args.  The comment notes that gcc produces better code with
> 
> I think I prefer the macro variant actually. Sorry. It just looks
> better without the &s.

Hi Andi,

	Please reconsider.  This isn't about being pretty, it's about not
having hidden side-effects, and having typechecking.

> We don't care very much about the code quality here because
> rdmsr/wrmsr are always very slow in microcode anyways and tend
> to synchronize the CPUs.

Agreed, but comment about it above the macros made me wary, so I checked
it.  No significant code difference with gcc >= 4.0, at least.

> If you feel a need to clean up I would suggest you convert more
> users over to the ll variants which take a single 64bit value
> instead of two 32bit ones.

You mean the l and ll variants?  The 64 bit variants are rdmsrl and
rdtscll, not to be confused with rdtscl, which returns the lower 32
bits.  This confusion caused the x86_64 bug in gameport.c which the
patch comment mentioned (at least, seems to be a bug to me).

See why I want to fix these names?

So if you would prefer u64 rdtsc64(), u32 rdtsc_low(), u64 rdmsr64(int
msr), u32 rdmsr_low(int msr), I can convert everyone to that, although
it's a more invasive change...

Thanks,
Rusty.
-- 
Help! Save Australia from the worst of the DMCA: http://linux.org.au/law

-
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/