On 2015-12-07 at 07:32 "'Davide Libenzi' via Akaros"
<[email protected]> wrote:
> No, qpread() does not discard. Like pread(), simply reads at offset.

In that case, let's not use the queue stuff for this.  All we need is a
buffer and length, and hang it off the aux.

> For #arch/interrupts device, is create response at open(), free at
> close(), and use at read() (using qpread()).
> The lseek() is a layer above that, which simply changes the pointer
> with which read(2) are coming to the device.

Dan and I were chatting about this, and building the response one time
on open() sounds like the best approach.  The downside is that users
will need to close and open a new chan if they want new data.  That
sounds like a reasonable limitation for synthesized data/tables.

> I think it is OK to require that writes are atomic. The user should
> send in the whole buffer.

Yeah, me too.  That's the current state of our write commands anyways.
Also, some devices just ignore the offset for writes (e.g. ipwrite()).

Barret

-- 
You received this message because you are subscribed to the Google Groups 
"Akaros" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to