linux-kernel - Re: [PATCH] drivers/block/xsysace - replace in(out)_8/in(out)_be16/in(out)

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: <1360672254.2035.8.camel@pasglop>
Date:	Tue, 12 Feb 2013 23:30:54 +1100
From:	Benjamin Herrenschmidt <benh@...nel.crashing.org>
To:	Michal Simek <monstr@...str.eu>
Cc:	Arnd Bergmann <arnd@...db.de>,
	Geert Uytterhoeven <geert@...ux-m68k.org>,
	Grant Likely <grant.likely@...retlab.ca>,
	Alexey Brodkin <Alexey.Brodkin@...opsys.com>,
	Vineet Gupta <Vineet.Gupta1@...opsys.com>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Alan Cox <alan@...rguk.ukuu.org.uk>,
	dahinds@...rs.sourceforge.net
Subject: Re: [PATCH] drivers/block/xsysace - replace
 in(out)_8/in(out)_be16/in(out)_le16 with generic iowrite(read)8/16(be)

On Tue, 2013-02-12 at 11:11 +0100, Michal Simek wrote:
> For high performance IPs using accessors functions is still
> problematic
> because there will be performance regression it means that
> from my point of view there still should be any option to "setup"
> proper endians for the driver and it can't be setup at run-time.
> 
> Sure the question is if drivers like this should be in the mainline.
> But if yes, there should be an option to do it in acceptable way.

No. High performance IP should rely on DMA essentially and be less
affected if not at all. Endianess should affect the control path, not
the data path (unless you got your wiring wrong again :-)

If you want an Kconfig option for an "optimal" version, put the ifdef
trickery in the accessors and make them inline.

We do something like that iirc with OHCI where the type of endianness
needed is a pair of select's and the conditional only happens when both
are set. You can probably do better with feature bits & a mask of
"possible" features relying on the compiler optimisations rather than
the preprocessor.

Cheers,
Ben.

--
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/