On Mon, 2019-08-19 at 20:32 +0000, Martin Wilck wrote: > On Fri, 2019-08-16 at 15:12 -0500, Benjamin Marzinski wrote: > > On Thu, Aug 15, 2019 at 02:46:54PM +0000, Martin Wilck wrote: > > > From: Martin Wilck <mwilck@xxxxxxxx> > > > > > > This new configuration parameter can be used to selectively > > > enable foreign libraries. The value is a regular expression, > > > against which foreign library names such as "nvme" are matched. > > > By setting this to a value that matches no foreign library > > > (e.g. "^$" or "NONE"), foreign code is completely disabled. > > > By default, all available foreign libraries are loaded. > > > > > > > This will stop the foreign libraries from even claiming devices. > > Won't > > this mean that pathinfo() will now treat these devices as > > multipathable > > paths, since is_claimed_by_foreign() will return false? > > IMO this won't happen, because we ignore the native multipath path > devices anyway as they're "hidden", and we ignore native NVMe > multipath > maps because they have "nvme-subsystem" subsystem rather than "nvme" > (commit b18ed66). But I'll double-check again. Confirmed. With nvme_core.multipath=Y and the nvme library disabled,no NVMe-related output is printed. Ben, with that in mind, would you ACK this patch? Martin -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
