Re: [patch 1/5] Staging: VME Framework for the Linux Kernel

[Martyn Welch <martyn.welch@gefanuc.com>]

Emilio G. Cota wrote:
> Martyn Welch wrote:
>   
>> If the driver writer wants a "single shot" effect, there's little  
>> overhead in creating a one item link-list and executing it. However,  
>> without support for this mode the author is stuck with single shot  
>> transfers.  If the author wants to perform scatter gather type  
>> transfers, without this functionality in the API - which is supported by  
>> all the bridges I have seen - he won't be able to easily avail himself  
>> of this functionality and get "work done".
>>     
>
> Again, he (the driver writer) should know *nothing* about that.
>
> As for the scatter-gather case, what should happen is:
>
> User:		hey, there's a buffer I want to transfer. here's
> 		a pointer and size
>   

Buffer where? On VME or PCI, where are you transferring it to? The API I 
have will allow scattered locations on VME to be copied to PCI and 
visa-versa, scattered locations on PCI to be copied to VME and 
visa-versa. Even scattered locations on VME to scattered locations on 
PCI with a bit of effort.

> VME layer:	ok, passing it to the bridge
> VME bridge:	I know I'm over PCI, so I'll get the pages
> 		from memory and map the sg list over PCI, and
> 		then will tell the bridge to perform from there.
>
> That's just a particular example, but for the user (read driver
> writer) it's much simpler this way. Doesn't need to know about
> lists or whatever.
>   

But seems to fail to provide the flexibility that I outlined above.

Martyn
> E.
>   

-- 
Martyn Welch MEng MPhil MIET (Principal Software Engineer)   T:+44(0)1327322748
GE Fanuc Intelligent Platforms Ltd,        |Registered in England and Wales
Tove Valley Business Park, Towcester,      |(3828642) at 100 Barbirolli Square,
Northants, NN12 6PF, UK T:+44(0)1327359444 |Manchester,M2 3AB  VAT:GB 927559189