On Mar 04 Jonathan Woithe wrote:
On Sat, Mar 01, 2014 at 03:20:50PM +0100, Stefan Richter wrote:
Since FFADO applies 200 ms or more as FCP transaction timeout, shouldn't firewire-lib's fcp.c increase FCP_TIMEOUT_MS from 125 to 200 or more as well?
[...]
I've personally had nothing to do with devices utilising FCP transactions so unfortunately I don't really know. My feeling is that the authors of the respective FFADO drivers would not have applied an FCP timeout of 200 ms if there was no demonstrated need for it. Therefore in the absence of other evidence I would be assuming that there are devices which require the higher timeout allowed for in FFADO's streaming code. However, I don't know which specific devices these might be.
./libffado/config.h.in: #define IEEE1394SERVICE_FCP_RESPONSE_TIMEOUT_USEC 200000
was added in this FFADO revision: http://subversion.ffado.org/changeset/1371 Timestamp: 10/23/08 09:00:47 (5 years ago) Author: ppalmers Message: * implement our own code to do FCP transactions. the code from libavc had too much side-effects. * remove libavc1394 as a dependency * set the SPLIT_TIMEOUT value for the host controller such that late responses by the DM1x00 based devices are not discarded. Should fix the issues with FA-101 discovery. (re: #155, #162)
I will look through the libavc1394 source history.