[PATCH v6 07/10] ARM: OMAP2+: gpmc: generic timing calculation

Jon Hunter jon-hunter at ti.com
Mon Aug 27 16:30:13 EDT 2012


Hi Afzal,

On 08/27/2012 05:37 AM, Mohammed, Afzal wrote:
> On Thu, Aug 23, 2012 at 08:28:44, Hunter, Jon wrote:

[snip]

>> I understand that this is based upon how timings for onenand and tusb
>> are being calculated today, but I am not sure that this is the way to go
>> for all devices. Personally, I would like to see us get away from how
>> those devices are calculating timings for any new device.
> 
> You cannot do away with many of the those, as logically they
> are right. Eg. read data should be available at access time,
> assuming a zero data hold time, we can very well derive an
> expression as,

I am not saying that we do away with them for current devices, just
maintain them as is.

> read de-assertion time (oe_off) = access time plus 1 gpmc clock,
> 
> and this is what the existing calculations do, and also the
> generic routine. There could be other constraints, but this
> certainly should be one (I do realize that oe_off could be
> optimized to be less than access time, by relying on read
> hold time, then again effect of it would be in similar way
> for different peripherals, but let's forget about
> optimization in the beginning)
> 
>> In general, I am concerned that we are abstracting the timings further
>> from the actual register fields. For example, the timings parameter
>> "t_ceasu" is described "address setup to CS valid" which is not
>> incorrect but this value is really just programming the CSONTIME field
>> and so why not call this cs_on?
> 
> Timing fields of struct gpmc_device_timings are selected such
> that they should be bindable by DT. At least one of the peripheral
> datasheet has these fields.

Right, but these are not applicable to every device and so I worry this
could be confusing. However, more documentation may help clear this up.

> If user knows timings in terms of
> gpmc values, he can directly use struct gpmc_timings, but
> then all the values should be updated in struct gpmc_timings.
> User should not update some of the values in terms of
> peripheral timings, others in terms of gpmc timings, that
> would make things complex and does not seem the right way
> to me.
> 
> cs_on and other gpmc aware timings could be binded by DT, not as
> peripheral timing, but as gpmc timing.
> 
> Bindings for peripheral DT timings should be something that
> can be obtained from peripheral datasheet, here accidentally
> it is same as gpmc timings, but not most timings are this way.
> Also there could be other constraints that can come for cs_on,
> even though I have not come across any till now.
> 
>>
>> So although this may consolidate how the timings are calculated today, I
>> am concerned it will be confusing to add timings for a new device. At
>> least if I am calculating timings, I am taking the timing information
>> for the device and translating that to the how I need to program the
>> gpmc register fields.
> 
> If I am not wrong, GPMC IP has been present for around a
> decade, and so far I have not come across any generic time
> calculation method that can be applied to all peripherals.

Yes not an easy problem to solve :-(

> Getting to work same peripheral for a different gpmc
> frequency is another problem.
> 
> Here we are trying to generalize based on the understanding of
> gpmc & peripheral timings, as well as by a kind of reverse
> engineering without most of the hardware or datasheet. Think of
> this as an attempt to create one, let it evolve and become a
> robust one. If a new user want to add a peripheral, let him add
> support, input timings are selected such that it is found in
> peripheral datasheet, if that does not work, let him try
> to add any new timing field or alter existing expression
> as he deems to be proper and verify that it does not break
> others. And I can help provided I am not heavily loaded
> with other works.

So long as it is maintainable ;-)

> And at least for initial users, they are expected to have
> some grasp on how to calculate timings, such a user will
> not be much worried about your 3 concerns above, anyway as
> of now they need to have a good grasp on it.

I would consider myself to be an initial user and I am concerned,
doesn't that count?

An example, would be the following where you have 4 timing parameters
for access time. You need to dig through the code to understand how
these are being used.

+	u32     t_aa;		/* access time from ADV assertion */
+	u32     t_iaa;		/* initial access time */
+	u32     t_oe;		/* access time from OE assertion */
+	u32     t_ce;		/* access time from CS asertion */

> Meanwhile I will try to document more.

Yes more documentation is definitely needed.

Cheers
Jon



More information about the linux-arm-kernel mailing list