Am 05.01.2011 21:07, Gilles Chanteperdrix wrote: > Jan Kiszka wrote: >> Am 05.01.2011 20:40, Stefan Kisdaroczi wrote: >>> Am Dienstag 04 Januar 2011, 20:49:59 schrieben Sie: >>>> Stefan Kisdaroczi wrote: >>>>> Hi, >>>>> >>>>> the attached patch adds pci support to the 16550A serial driver. I have >>>>> a Moxa CP-132 board and the first tests are looking good. If there is a >>>>> chance to get the patch applied I kindly request you to review it. I'll >>>>> add support for more Moxa boards if there is a interest to apply the >>>>> patch. >>>>> >>>>> thanks >>>>> Stefan >>>> Looks fine to me, however, nit-picking a bit, could we do this differently: >>> new patches attached. >>> >> >> Looks good to me as well except for a minor white space damage: >> >>> +static int __devinit rt_16550_pci_probe( struct pci_dev *pdev, >> ^^^^^^^ >> >> I have one further suggestion which can be done on top: The RTDM_IRQTYPE >> flags should be customized as well. So far we assume edge-type devices, >> but PCI is level-triggered. Requesting edge works, but is suboptimal. > > Well, on vanilla linux, the edge handler does not mask the interrupt > line before invoking the handler, so, using the edge handler for level > interrupts should not work and lock-up (except if IRQF_DISABLED is > passed, which should now be the default, OK).
IRQF_DISABLED is a (soon to be removed) nop these days, all handlers are called with interrupts disabled. > > Anyway, is a driver the right place to pass these flags? Since the > interrupt controller pins to which the PCI interrupts are wired really > depends on the board more than on the device driver. > I know. But can we always obtain the required information from Linux? If Xenomai could set the right types automatically, that would be much better of course. Jan
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Xenomai-help mailing list [email protected] https://mail.gna.org/listinfo/xenomai-help
