On 08/29/2011 05:31 AM, andy pugh wrote: > On 29 August 2011 04:02, Sebastian Kuzminsky<s...@highlab.com> wrote: >> Whoops, the 5i21 has two IO Connectors named P1, and the 5i25 has no >> connectors. > Two connectors called P1 is wierd, but correct, as it is a single > connector sharing two 32 bit buffers. > I am confused by the second statement. Are you saying that physically > the 5i25 has no connectors, or that the patch doesn't set any up? > Neither appears to be correct to me, but I am probably missing a > subtle point, can you elucidate?
Never mind, I'm just being stupid. The patch sets up the 5i25 with connectors, I just missed it. >> There are a couple of other places where the pin number is converted to >> a port number and a pin-within-that-port number. hm2_set_pin_source() >> and hm2_set_pin_direction() come to mind. The code you added to >> hm2_print_pin_usage() seems to do it right, maybe it'd make sense to >> annotate the pins with their port and pin-within-port info at init >> time? Add it to the hm2_pin_t maybe? > That sounds like it might make sense. I am unfamiliar with that > section of the code (and the EMC2 machine is disconnected at the > moment) so can't immediately tell what would be involved. > >> Also, this commit broke the unit tests. One of the hm2-idrom test cases >> verifies that IOWidth is 24, and that's obviously no longer the case. > I am embarrassed to admit that I had never spotted that test > directory, so wasn't aware of the idrom test. The runtests script knows about it. If you build, then source emc-environment, then run runtests, it'll run all the unit tests. Or you can say "runtests $EMC2_HOME/tests/hm2-idrom" to run just that one. >> You removed that requirement in hm2_read_idrom() and hm2_print_idrom(). >> Maybe the llio should tell us what it's expected IOWidth is? It knows >> how big its connectors are, after all, and it seems like a good sanity >> check. > Yes, that probably does make sense. > Do you mean that an entry for port width should be added to the > hm2_lowlevel_io_t, or is there something already there that I haven't > spotted. > > To save duplication of effort, we probably ought to agree who is going > to make the changes before starting :-) I made a commit to 2.5 last night that implements this, and just merged it into master this morning. I did not do the "set per-pin port and pin-within-port information at init-time" work, if it's not done by the end of this week by someone else I'll do it. I think what's needed is to add "connector_number" and "pin_within_connector" to hm2_pin_t, and use the new logic that Peter added to hm2_print_pin_usage() to initialize those fields, maybe in hm2_configure_pins(). But I haven't looked very closely at it... BTW I tested v2.5.0-pre1-616-gd4e9c8f (which includes Peter's patch and my llio update) on my Bridgeport last night. It's using a 5i22 and it didn't break, so that's good... Man I wish we had a big hm2 test farm with all the AnyIO cards in it, and a bunch of loopback wiring harnesses and stuff... -- Sebastian Kuzminsky ------------------------------------------------------------------------------ EMC VNX: the world's simplest storage, starting under $10K The only unified storage solution that offers unified management Up to 160% more powerful than alternatives and 25% more efficient. Guaranteed. http://p.sf.net/sfu/emc-vnx-dev2dev _______________________________________________ Emc-developers mailing list Emc-developers@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-developers
