On Mon, 14 Sep 2009 18:03:56 -0700 ebiederm@xxxxxxxxxxxx (Eric W. Biederman) wrote: > Jesse Barnes <jbarnes@xxxxxxxxxxxxxxxx> writes: > > > On Thu, 10 Sep 2009 10:31:57 -0600 > > Bjorn Helgaas <bjorn.helgaas@xxxxxx> wrote: > > > >> On Monday 07 September 2009 04:40:22 am Eric W. Biederman wrote: > >> > > >> > What follows below is my alternate pcie hotplug driver. > >> > It is very stupid, very simple and very robust. > >> > > >> > This driver should work on any pcie hotplug bridge that > >> > only has support for the interrupt when the pcie link > >> > comes or down, and that sets the hotplug and the hotplug > >> > surprise bits. > >> > > >> > I wrote this because in my environment the pciehp driver > >> > totally fails and 500 lines of code are much easier to > >> > debug than 3000. > >> > > >> > Now that I have the code working I'm looking for the best > >> > path to get a driver I can use into the mainstream kernel. > >> > >> I think it'd be great to simplify pciehp, and pcielw looks > >> nice and clean. > >> > >> I think pciehp/acpiphp/pcielw are somewhat user-unfriendly > >> because (a) it's hard for a user to figure out which to use, > >> and (b) there's no nice way to autoload them because there's > >> nothing that connects them to a udev event. > >> > >> My personal opinion is that we shouldn't merge pcielw alongside > >> pciehp because it would make the user confusion worse and dilute > >> the already small testing pool. > >> > >> We could address the autoload issue by making pciehp/pcielw part > >> of the pcieport driver. That would simplify the code as well as > >> the user experience, but maybe there's some reason to keep them > >> separate. > >> > >> I think it'd be nice to have a series of evolutionary patches > >> to transform pciehp rather than replacing it wholesale. Otherwise, > >> bisection (one of the few tools we non-expert masses have) won't > >> be as useful. > > > > Agreed. Is that something you'd be interested in doing, Eric? > > Time permitting, I will take a stab at it. It looks like things are > pretty healthy on the development front so it should be doable. > > There is a fair amount of heavy lifting. Would it help if I tossed > up a git tree where I accumulate the changes as I make them. Then > you can just pull them as you have time? Yep, that's fine. It would help to post them to the mailing list at the same time to get extra review. Thanks, -- Jesse Barnes, Intel Open Source Technology Center -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
