Hello, I wrote:
This patch updates the "busy" field to reflect the current status of the
endpoint. This fix is critical to enable the request queuing logic to
prevent
a queued request from overwriting/interfering in-flight request.
Frankly speaking, I'm not seeing how a newly queued request can
overwrite anything as it's only started being *at the head of the
request queue* (which it won't be unless it's queued onto an already
empty queue with the last request has been already done). Moreover,
ep->busy will have already been set by musb_g_givebck() iff the queue()
method is called from the request completion handler, so
musb_ep_restart() cannotr be called from there at this time. What this
is really fixing is your broken logic in the prior patch about which so
many spears have been broken already, if I don't mistake, and in this
case these patches should be merged.
But I do think that your whole approach to stalling/aborting in flight
request is wrong and that's what should be fixed instead.
Felipe, I'm seeing both the questionable patches already queued by you
despite of the patches being inconsistent and fixing the issues in an
inconsisnt way. Despite of this, I'd like to restate my request to Swaminathan
to provide a failure scenario that this patch is allegedly fixing.
WBR, Sergei
_______________________________________________
Davinci-linux-open-source mailing list
[email protected]
http://linux.davincidsp.com/mailman/listinfo/davinci-linux-open-source
