linux-kernel - Re: [PATCH 2/4] ARM OMAP2+ GPMC: always program GPMCFCLKDIVIDER

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: <54E34F73.2080108@ti.com>
Date:	Tue, 17 Feb 2015 16:25:55 +0200
From:	Roger Quadros <rogerq@...com>
To:	Robert Abel <rabel@...-ec.uni-bielefeld.de>
CC:	<khilman@...prootsystems.com>, Tony Lindgren <tony@...mide.com>,
	<linux@....linux.org.uk>, <linux-omap@...r.kernel.org>,
	Linux Kernel Maling List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 2/4] ARM OMAP2+ GPMC: always program GPMCFCLKDIVIDER

On 17/02/15 16:06, Robert Abel wrote:
> Hi Roger,
> 
> On Tue, Feb 17, 2015 at 2:52 PM, Roger Quadros <rogerq@...com> wrote:
>> nobody stops the DT binding from specifying a large enough "gpmc,wait-monitoring-ns" value.
>> The driver must use that to scale the GPMC_CLK if it doesn't fit in the GPMC_FCLK.
>> This feature can come separately though. So for now I was suggesting to set the divisor to 1.
>> [...]
>> AFAIK "gpmc,sync-clk-ps" is not specified for asynchronous devices so it defaults to 0
>> in the driver.
> 
> As you have rightly pointed out, sync-clk-ps defaults to 0, i.e.
> divider 1. My solution would work for people /now/ with different
> gpmc,wait-monitoring-ns requirements. Of course, in general you're
> right, the driver could compute that on its own. However, this
> influences sampling behavior of the GPMC, which is somewhat strange
> anyway. Since I lack a proper test setup and time to experiment with
> the GPMC, I'd compromise on leaving sync-clk-ps default to 0, divider
> defaults to 1. If somebody feels up to implementing driver-side
> GPMC_CLK scaling, they might as well nix the dependency at that point
> in time. Right now, keeping the dependency seems more useful to users
> than killing it right away.

one more thing to note is that just specifying sync-clk-ps in DT is not enough for
asynchronous devices.

The driver doesn't set gpmc_t->sync_clk if "gpmc,sync-read" or "gpmc,sync-write"
was not set in the DT, which would be the case for asynchronous devices.

> 
>> What I'm stressing on is that there shouldn't be any dependency on "gpmc,sync-clk-ps" for
>> asynchronous devices. It also becomes easier to specify the wait-monitoring-ns as we don't need
>> to cross reference with "sync-clk-ps".
> 
> As an aside: There shouldn't be a dependency on the FCLK for
> synchronous accesses either. The GPMC driver is in a somewhat terrible
> state that synchronous protocols have to specify in ns, which get
> scaled by the startup FCLK period... So this wrongful dependency
> doesn't make my top ten, especially since it right now would fit a use
> case.

What is your proposal to make things better? And what is your use case that doesn't
work with existing setup?

cheers,
-roger
--
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/