On Mon, 2013-09-30 at 11:44 +0200, Javier Lopez wrote: > Original implementation uses sysfs to get dev index from > dev name. Due the changes on netns and sysfs iw is broken > if using multiple network namespaces. iw works properly > if using it from the main namespace, but it won't work if > using from the new namespace. > > Kernel commit 3ff195b0, "sysfs: Implement sysfs tagged > directory support" patch, added a filtering mechanism > to sysfs, allowing sysfs directories to look different > depending on the context where they are being observed. > > When an interface is moved to another namespace, the > interface dissapears from sysfs structure. In order > to recover access to the directory a solution is to > remount sysfs from the correct context. This will force > the user to remount sysfs before using iw from a > different namespace. > > To avoid this issue we can use nl80211 (using > NL80211_CMD_GET_WIPHY command) this returns the list of > phys, then process the list, find the device and return > the device index. This seems a bit workaround-ish/hack-ish to me. Why would a remount be necessary? Can't sysfs look at the process tags when determining the access? Should we maybe do something in our sysfs code in the kernel for this? johannes -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
