Hey, On Sat, Aug 14, 2010 at 08:59:11AM +0900, Komuro wrote: > >@@ -248,9 +248,8 @@ static int fmvj18x_probe(struct pcmcia_device *link) > > lp->base = NULL; > > > > /* The io structure describes IO port mapping */ > >- link->io.NumPorts1 = 32; > >- link->io.Attributes1 = IO_DATA_PATH_WIDTH_AUTO; > >- link->io.IOAddrLines = 5; > >+ link->resource[0]->end = 32; > >+ link->resource[0]->flags |= IO_DATA_PATH_WIDTH_AUTO; > > > > > Do you have any reason to move "link->io_lines = 5" from fmvj18x_probe > to fmvj18x_config?
Yes -- I want to have it close to pcmcia_request_io(), especially as I wish to introduce an "autoconfig" feature for the next kernel (2.6.37). A preliminary patch for that can already be found in the "pcmcia-test" branch, accessible e.g. via gitweb: http://git.kernel.org/?p=linux/kernel/git/brodo/pcmcia-2.6.git;a=commitdiff;h=1fefe76a0a620ffc54c4084f4c04dd4d2ac05392 After this patch, a driver may have all the fields filled out automatically by the PCMCIA core, but can then override things -- e.g. by setting link->io_lines = 5 . But this override must happen between the setting of the values (within pcmcia_loop_config(), obviously) and the use of the values (in pcmcia_request_io()). That's the reason for the move. Best, Dominik _______________________________________________ Linux PCMCIA reimplementation list http://lists.infradead.org/mailman/listinfo/linux-pcmcia