Just experimenting with PCIe Fabric and running into unkown issues. Thanks for feedback, I will update when I find a good way to solve this. Thanks, Kallol On Mon, Aug 1, 2016 at 11:54 AM, Alex Williamson <alex.williamson@xxxxxxxxxx> wrote: > On Mon, 1 Aug 2016 11:29:58 -0700 > Kallol Biswas <kallol@xxxxxxxxxxxxxx> wrote: > >> Hi, >> I have exposed a realtek PF via an additional switch to a host CPU. >> >> The host CPU can now view the device via two paths. One, in regular >> path and another via the additional switch. The device has two B:D.Fs >> assigned to it. >> >> How can I tell the Realtek driver to configure the device only via new path? >> >> Bind/unbind may not work, as I do not want to initialize the device >> via the regular path. > > In general this sounds like a horrible idea, how can the OS know that > these two paths reach the same device? What about the BAR resources of > the device, can they be valid for both paths? The kernel is going > to try to allocate separate resources for each path. Even if you make > the rtl driver only attach once, you better try to pick the BDF that the > kernel scanned and programmed resources for most recently. > > If I had to hack something together for this, my first thought would be > to use pci-stub to claim both devices via the vendor:device IDs. Then > use a boot script that would "echo 1 > remove" for both devices in > sysfs, then hopefully find a "rescan" in sysfs far enough downstream > via the path you want that you can rediscover only the desired BDF, > then bind that device to the rtl driver. (I hope you don't intend to > ship any products this way). Thanks, > > Alex -- 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
