summaryrefslogtreecommitdiff
path: root/firmware/link
diff options
context:
space:
mode:
authorlaforge <laforge@6dc7ffe9-61d6-0310-9af1-9938baff3ed1>2006-09-22 21:21:28 +0000
committerlaforge <laforge@6dc7ffe9-61d6-0310-9af1-9938baff3ed1>2006-09-22 21:21:28 +0000
commit238c6c3c0ac40dcdef82c430fae52cad34ed5a10 (patch)
tree75f5529c9c3c906b04dc78ccad227efc12c4d3a8 /firmware/link
parent3b1d90fae2efb52b0b3225be2d4a3dfde22adbbb (diff)
fix completely broken logic of processing pending outbound packets. We now do
all the req_ctx state changes from within pcd_handler.c and only expose the refill_ep() functions. It's questionable whether we need to keep calling those functions all the time either. Once kicked off, the tx completion interrupts of the UDC will refill the pipe as long as there is data available. So realistically speaking, we'd only need to trigger tx once (and keep triggering until no EBUSY or other error is signalled anymore). Some form of state machine would be fine for doing this, one that registers a timer with the core OS timer (once we have something equivalent to 'struct timer_list'. git-svn-id: https://svn.openpcd.org:2342/trunk@217 6dc7ffe9-61d6-0310-9af1-9938baff3ed1
Diffstat (limited to 'firmware/link')
0 files changed, 0 insertions, 0 deletions
personal git repositories of Harald Welte. Your mileage may vary