| Message ID | 20180117074939.7795-8-mwilck@suse.com (mailing list archive) |
|---|---|
| State | Not Applicable, archived |
| Delegated to: | christophe varoqui |
| Headers | show |
On 01/16/2018 11:49 PM, Martin Wilck wrote: > multipathd can't deal with other devices anyway. Proceeding further > with events for other devices just generates log noise. > > Based on an idea from Ritika Srivastava <ritika.srivastava@oracle.com>. > ("multipath-tools: Skip CHANGE uevent for non-mpath devices"). > > Signed-off-by: Martin Wilck <mwilck@suse.com> > --- > multipathd/main.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/multipathd/main.c b/multipathd/main.c > index ff3ecb640487..26632291657f 100644 > --- a/multipathd/main.c > +++ b/multipathd/main.c > @@ -1121,7 +1121,7 @@ uev_trigger (struct uevent * uev, void * trigger_data) > * Add events are ignored here as the tables > * are not fully initialised then. > */ > - if (!strncmp(uev->kernel, "dm-", 3)) { > + if (!strncmp(uev->kernel, "dm-", 3) && uevent_is_mpath(uev)) { > if (!strncmp(uev->action, "change", 6)) { > r = uev_add_map(uev, vecs); > Hi Martin, Thank you for the updated patch. With this patch, the error "uevent trigger error" would not be encountered when removing lvm snapshots. However, when the uevent is for a 'dm' device but not for a multipath device, then should we just return from this check in uev_trigger()? With this patch, uev_update_path()/uev_remove_path() would still be called which would generate further log. Should we avoid these function calls too?
On Mon, 2018-01-29 at 14:25 -0800, Ritika Srivastava wrote: > On 01/16/2018 11:49 PM, Martin Wilck wrote: > > multipathd can't deal with other devices anyway. Proceeding further > > with events for other devices just generates log noise. > > > > Based on an idea from Ritika Srivastava <ritika.srivastava@oracle.c > > om>. > > ("multipath-tools: Skip CHANGE uevent for non-mpath devices"). > > > > Signed-off-by: Martin Wilck <mwilck@suse.com> > > --- > > multipathd/main.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/multipathd/main.c b/multipathd/main.c > > index ff3ecb640487..26632291657f 100644 > > --- a/multipathd/main.c > > +++ b/multipathd/main.c > > @@ -1121,7 +1121,7 @@ uev_trigger (struct uevent * uev, void * > > trigger_data) > > * Add events are ignored here as the tables > > * are not fully initialised then. > > */ > > - if (!strncmp(uev->kernel, "dm-", 3)) { > > + if (!strncmp(uev->kernel, "dm-", 3) && > > uevent_is_mpath(uev)) { > > if (!strncmp(uev->action, "change", 6)) { > > r = uev_add_map(uev, vecs); > > > > Hi Martin, > Thank you for the updated patch. > With this patch, the error "uevent trigger error" would not be > encountered when removing lvm snapshots. > > However, when the uevent is for a 'dm' device but not for a > multipath > device, then should we just return from this check in uev_trigger()? > With this patch, uev_update_path()/uev_remove_path() would still be > called which would generate further log. > Should we avoid these function calls too? You are right, thanks for pointing that out. I'll post an update. Martin
diff --git a/multipathd/main.c b/multipathd/main.c index ff3ecb640487..26632291657f 100644 --- a/multipathd/main.c +++ b/multipathd/main.c @@ -1121,7 +1121,7 @@ uev_trigger (struct uevent * uev, void * trigger_data) * Add events are ignored here as the tables * are not fully initialised then. */ - if (!strncmp(uev->kernel, "dm-", 3)) { + if (!strncmp(uev->kernel, "dm-", 3) && uevent_is_mpath(uev)) { if (!strncmp(uev->action, "change", 6)) { r = uev_add_map(uev, vecs);
multipathd can't deal with other devices anyway. Proceeding further with events for other devices just generates log noise. Based on an idea from Ritika Srivastava <ritika.srivastava@oracle.com>. ("multipath-tools: Skip CHANGE uevent for non-mpath devices"). Signed-off-by: Martin Wilck <mwilck@suse.com> --- multipathd/main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)