On Fri, Dec 03, 2021 at 08:01:11AM +0000, Martin Wilck wrote: > On Thu, 2021-12-02 at 19:34 -0600, Benjamin Marzinski wrote: > > There are reasons that dm_get_info() can fail other than the > > multipath > > device not existing. If there is an existing dm_info structure and > > the > > call fails, don't free it. > > > > Signed-off-by: Benjamin Marzinski <bmarzins@xxxxxxxxxx> > > This is ok, but I think it'd be better to simply embed the dmi struct > in struct multipath and thus avoid having allocate and free it > separately. Either that, or just use a struct dm_info on stack in > dm_get_info() and just copy the properties we actually look at, i.e. > read_only, major, minor, suspended, to respective fields in struct > multipath. I'd prefer embedding the complete dmi, as it's not that > large and contains other interesting fields, too (for example, we might > be able to leverage the event_nr field to avoid duplicate work between > uevent and dmevent handlers). > > We would need to be able to figure out if the dmi is valid, but that's > rather easy, we could simply set major = minor = 0. Makes sense. I can change this. -Ben > > Regards > Martin -- dm-devel mailing list dm-devel@xxxxxxxxxx https://listman.redhat.com/mailman/listinfo/dm-devel
