All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] multipath: better check for daemon mode.
@ 2011-07-25 18:47 Benjamin Marzinski
  2011-09-01  2:46 ` Benjamin Marzinski
  0 siblings, 1 reply; 5+ messages in thread
From: Benjamin Marzinski @ 2011-07-25 18:47 UTC (permalink / raw)
  To: device-mapper development

Switching to conf->daemon to check if we are in daemon mode.

Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
---
 libmultipath/dmparser.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Index: multipath-tools-110713/libmultipath/dmparser.c
===================================================================
--- multipath-tools-110713.orig/libmultipath/dmparser.c
+++ multipath-tools-110713/libmultipath/dmparser.c
@@ -13,6 +13,7 @@
 #include "structs.h"
 #include "util.h"
 #include "debug.h"
+#include "config.h"
 
 #define WORD_SIZE 64
 
@@ -330,7 +331,7 @@ disassemble_map (vector pathvec, char * 
 				strncpy(pp->dev_t, word, BLK_DEV_SIZE);
 
 				/* Only call this in multipath client mode */
-				if (!mpp->waiter && store_path(pathvec, pp))
+				if (!conf->daemon && store_path(pathvec, pp))
 					goto out1;
 			}
 			FREE(word);

^ permalink raw reply	[flat|nested] 5+ messages in thread
* Re: [PATCH] multipath: better check for daemon mode.
  2011-07-25 18:47 [PATCH] multipath: better check for daemon mode Benjamin Marzinski
@ 2011-09-01  2:46 ` Benjamin Marzinski
  2011-09-29  3:38   ` Benjamin Marzinski
  0 siblings, 1 reply; 5+ messages in thread
From: Benjamin Marzinski @ 2011-09-01  2:46 UTC (permalink / raw)
  To: device-mapper development; +Cc: Christophe Varoqui

Christophe, would it be possible to get this merged?

Thanks,
-Ben

On Mon, Jul 25, 2011 at 01:47:52PM -0500, Benjamin Marzinski wrote:
> Switching to conf->daemon to check if we are in daemon mode.
> 
> Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
> ---
>  libmultipath/dmparser.c |    3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> Index: multipath-tools-110713/libmultipath/dmparser.c
> ===================================================================
> --- multipath-tools-110713.orig/libmultipath/dmparser.c
> +++ multipath-tools-110713/libmultipath/dmparser.c
> @@ -13,6 +13,7 @@
>  #include "structs.h"
>  #include "util.h"
>  #include "debug.h"
> +#include "config.h"
>  
>  #define WORD_SIZE 64
>  
> @@ -330,7 +331,7 @@ disassemble_map (vector pathvec, char * 
>  				strncpy(pp->dev_t, word, BLK_DEV_SIZE);
>  
>  				/* Only call this in multipath client mode */
> -				if (!mpp->waiter && store_path(pathvec, pp))
> +				if (!conf->daemon && store_path(pathvec, pp))
>  					goto out1;
>  			}
>  			FREE(word);
> 
> --
> dm-devel mailing list
> dm-devel@redhat.com
> https://www.redhat.com/mailman/listinfo/dm-devel

^ permalink raw reply	[flat|nested] 5+ messages in thread
* Re: [PATCH] multipath: better check for daemon mode.
  2011-09-01  2:46 ` Benjamin Marzinski
@ 2011-09-29  3:38   ` Benjamin Marzinski
  2011-09-30  5:58     ` Christophe Varoqui
  0 siblings, 1 reply; 5+ messages in thread
From: Benjamin Marzinski @ 2011-09-29  3:38 UTC (permalink / raw)
  To: device-mapper development; +Cc: Christophe Varoqui

On Wed, Aug 31, 2011 at 09:46:32PM -0500, Benjamin Marzinski wrote:
> Christophe, would it be possible to get this merged?

Is there some objection to merging this?  It's not just a cosmetic
change.  For multipath devices added via ev_add_map(), disassemble_map()
will be called before the waiter thread is started, and this check will
allow any unknown paths in the device to be stored.  However, paths
added this way never get properly setup like they do when they are added
via ev_add_path(), or at configuration time.

So say you had a multipath device, "mpatha", which should include the
path "sda", except that the path was mistakenly blacklisted. If you
unblacklist the path and then run multipath, without first reconfiguring
multipathd, it will segfault, since it adds sda, but never initializes
all the necessary information.

We could solve this problem by making sure that paths added through
disassemble_map() go through the same initialization code that paths
added through ev_add_path(), but I don't see the necessity. If
multipathd isn't monitoring a path, it's either because that path was
blacklisted, or because that path was manually removed.  While it's
possible to create a multipath device using blacklisted paths via
dmsetup, multipathd still shouldn't be monitoring those paths.  The same
goes with paths that were manually removed via the multipathd command
line interface.

And if, for some reason, multipathd wasn't monitoring a path it should
be, the best thing to do is to reconfigure multipathd.  Code that
watched multipath.conf and automatically reconfigured multipathd when it
changed would be a nice addition though.

Thanks,
-Ben

> Thanks,
> -Ben
> 
> On Mon, Jul 25, 2011 at 01:47:52PM -0500, Benjamin Marzinski wrote:
> > Switching to conf->daemon to check if we are in daemon mode.
> > 
> > Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
> > ---
> >  libmultipath/dmparser.c |    3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> > 
> > Index: multipath-tools-110713/libmultipath/dmparser.c
> > ===================================================================
> > --- multipath-tools-110713.orig/libmultipath/dmparser.c
> > +++ multipath-tools-110713/libmultipath/dmparser.c
> > @@ -13,6 +13,7 @@
> >  #include "structs.h"
> >  #include "util.h"
> >  #include "debug.h"
> > +#include "config.h"
> >  
> >  #define WORD_SIZE 64
> >  
> > @@ -330,7 +331,7 @@ disassemble_map (vector pathvec, char * 
> >  				strncpy(pp->dev_t, word, BLK_DEV_SIZE);
> >  
> >  				/* Only call this in multipath client mode */
> > -				if (!mpp->waiter && store_path(pathvec, pp))
> > +				if (!conf->daemon && store_path(pathvec, pp))
> >  					goto out1;
> >  			}
> >  			FREE(word);
> > 
> > --
> > dm-devel mailing list
> > dm-devel@redhat.com
> > https://www.redhat.com/mailman/listinfo/dm-devel
> 
> --
> dm-devel mailing list
> dm-devel@redhat.com
> https://www.redhat.com/mailman/listinfo/dm-devel

^ permalink raw reply	[flat|nested] 5+ messages in thread
* Re: [PATCH] multipath: better check for daemon mode.
  2011-09-29  3:38   ` Benjamin Marzinski
@ 2011-09-30  5:58     ` Christophe Varoqui
  2011-09-30 15:08       ` Benjamin Marzinski
  0 siblings, 1 reply; 5+ messages in thread
From: Christophe Varoqui @ 2011-09-30  5:58 UTC (permalink / raw)
  To: Benjamin Marzinski; +Cc: dm-devel

On mer., 2011-09-28 at 22:38 -0500, Benjamin Marzinski wrote:
> On Wed, Aug 31, 2011 at 09:46:32PM -0500, Benjamin Marzinski wrote:
> > Christophe, would it be possible to get this merged?
> 
> Is there some objection to merging this?  It's not just a cosmetic
> change.  For multipath devices added via ev_add_map(), disassemble_map()
> will be called before the waiter thread is started, and this check will
> allow any unknown paths in the device to be stored.  However, paths
> added this way never get properly setup like they do when they are added
> via ev_add_path(), or at configuration time.
> 
> So say you had a multipath device, "mpatha", which should include the
> path "sda", except that the path was mistakenly blacklisted. If you
> unblacklist the path and then run multipath, without first reconfiguring
> multipathd, it will segfault, since it adds sda, but never initializes
> all the necessary information.
> 
> We could solve this problem by making sure that paths added through
> disassemble_map() go through the same initialization code that paths
> added through ev_add_path(), but I don't see the necessity. If
> multipathd isn't monitoring a path, it's either because that path was
> blacklisted, or because that path was manually removed.  While it's
> possible to create a multipath device using blacklisted paths via
> dmsetup, multipathd still shouldn't be monitoring those paths.  The same
> goes with paths that were manually removed via the multipathd command
> line interface.
> 
> And if, for some reason, multipathd wasn't monitoring a path it should
> be, the best thing to do is to reconfigure multipathd.  Code that
> watched multipath.conf and automatically reconfigured multipathd when it
> changed would be a nice addition though.
> 

The multipath daemon should not segfault for sure. I agree the
blacklisted paths can be left unmonitored even if they are part of
configured multipath.

But I wouldn't do the multipath.conf changes auto loading. Admins tend
to consider for granted daemons configuration files can be changed
without impact on the running daemon.

Regards,
cvaroqui
 
> Thanks,
> -Ben
> 
> > Thanks,
> > -Ben
> > 
> > On Mon, Jul 25, 2011 at 01:47:52PM -0500, Benjamin Marzinski wrote:
> > > Switching to conf->daemon to check if we are in daemon mode.
> > > 
> > > Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
> > > ---
> > >  libmultipath/dmparser.c |    3 ++-
> > >  1 file changed, 2 insertions(+), 1 deletion(-)
> > > 
> > > Index: multipath-tools-110713/libmultipath/dmparser.c
> > > ===================================================================
> > > --- multipath-tools-110713.orig/libmultipath/dmparser.c
> > > +++ multipath-tools-110713/libmultipath/dmparser.c
> > > @@ -13,6 +13,7 @@
> > >  #include "structs.h"
> > >  #include "util.h"
> > >  #include "debug.h"
> > > +#include "config.h"
> > >  
> > >  #define WORD_SIZE 64
> > >  
> > > @@ -330,7 +331,7 @@ disassemble_map (vector pathvec, char * 
> > >  				strncpy(pp->dev_t, word, BLK_DEV_SIZE);
> > >  
> > >  				/* Only call this in multipath client mode */
> > > -				if (!mpp->waiter && store_path(pathvec, pp))
> > > +				if (!conf->daemon && store_path(pathvec, pp))
> > >  					goto out1;
> > >  			}
> > >  			FREE(word);
> > > 
> > > --
> > > dm-devel mailing list
> > > dm-devel@redhat.com
> > > https://www.redhat.com/mailman/listinfo/dm-devel
> > 
> > --
> > dm-devel mailing list
> > dm-devel@redhat.com
> > https://www.redhat.com/mailman/listinfo/dm-devel

^ permalink raw reply	[flat|nested] 5+ messages in thread
* Re: [PATCH] multipath: better check for daemon mode.
  2011-09-30  5:58     ` Christophe Varoqui
@ 2011-09-30 15:08       ` Benjamin Marzinski
  0 siblings, 0 replies; 5+ messages in thread
From: Benjamin Marzinski @ 2011-09-30 15:08 UTC (permalink / raw)
  To: christophe.varoqui; +Cc: dm-devel

On Fri, Sep 30, 2011 at 07:58:27AM +0200, Christophe Varoqui wrote:
> On mer., 2011-09-28 at 22:38 -0500, Benjamin Marzinski wrote:
> > On Wed, Aug 31, 2011 at 09:46:32PM -0500, Benjamin Marzinski wrote:
> > > Christophe, would it be possible to get this merged?
> > 
> > Is there some objection to merging this?  It's not just a cosmetic
> > change.  For multipath devices added via ev_add_map(), disassemble_map()
> > will be called before the waiter thread is started, and this check will
> > allow any unknown paths in the device to be stored.  However, paths
> > added this way never get properly setup like they do when they are added
> > via ev_add_path(), or at configuration time.
> > 
> > So say you had a multipath device, "mpatha", which should include the
> > path "sda", except that the path was mistakenly blacklisted. If you
> > unblacklist the path and then run multipath, without first reconfiguring
> > multipathd, it will segfault, since it adds sda, but never initializes
> > all the necessary information.
> > 
> > We could solve this problem by making sure that paths added through
> > disassemble_map() go through the same initialization code that paths
> > added through ev_add_path(), but I don't see the necessity. If
> > multipathd isn't monitoring a path, it's either because that path was
> > blacklisted, or because that path was manually removed.  While it's
> > possible to create a multipath device using blacklisted paths via
> > dmsetup, multipathd still shouldn't be monitoring those paths.  The same
> > goes with paths that were manually removed via the multipathd command
> > line interface.
> > 
> > And if, for some reason, multipathd wasn't monitoring a path it should
> > be, the best thing to do is to reconfigure multipathd.  Code that
> > watched multipath.conf and automatically reconfigured multipathd when it
> > changed would be a nice addition though.
> > 
> 
> The multipath daemon should not segfault for sure. I agree the
> blacklisted paths can be left unmonitored even if they are part of
> configured multipath.
> 
> But I wouldn't do the multipath.conf changes auto loading. Admins tend
> to consider for granted daemons configuration files can be changed
> without impact on the running daemon.

Fine with me.  Nobody's ever asked me for this. But we occassionaly get
bug reports that end up being caused by admins forgetting or not knowing
to reconfigure multipathd after they change the config file.

-Ben
 
> Regards,
> cvaroqui
>  
> > Thanks,
> > -Ben
> > 
> > > Thanks,
> > > -Ben
> > > 
> > > On Mon, Jul 25, 2011 at 01:47:52PM -0500, Benjamin Marzinski wrote:
> > > > Switching to conf->daemon to check if we are in daemon mode.
> > > > 
> > > > Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
> > > > ---
> > > >  libmultipath/dmparser.c |    3 ++-
> > > >  1 file changed, 2 insertions(+), 1 deletion(-)
> > > > 
> > > > Index: multipath-tools-110713/libmultipath/dmparser.c
> > > > ===================================================================
> > > > --- multipath-tools-110713.orig/libmultipath/dmparser.c
> > > > +++ multipath-tools-110713/libmultipath/dmparser.c
> > > > @@ -13,6 +13,7 @@
> > > >  #include "structs.h"
> > > >  #include "util.h"
> > > >  #include "debug.h"
> > > > +#include "config.h"
> > > >  
> > > >  #define WORD_SIZE 64
> > > >  
> > > > @@ -330,7 +331,7 @@ disassemble_map (vector pathvec, char * 
> > > >  				strncpy(pp->dev_t, word, BLK_DEV_SIZE);
> > > >  
> > > >  				/* Only call this in multipath client mode */
> > > > -				if (!mpp->waiter && store_path(pathvec, pp))
> > > > +				if (!conf->daemon && store_path(pathvec, pp))
> > > >  					goto out1;
> > > >  			}
> > > >  			FREE(word);
> > > > 
> > > > --
> > > > dm-devel mailing list
> > > > dm-devel@redhat.com
> > > > https://www.redhat.com/mailman/listinfo/dm-devel
> > > 
> > > --
> > > dm-devel mailing list
> > > dm-devel@redhat.com
> > > https://www.redhat.com/mailman/listinfo/dm-devel
> 

^ permalink raw reply	[flat|nested] 5+ messages in thread
end of thread, other threads:[~2011-09-30 15:08 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-07-25 18:47 [PATCH] multipath: better check for daemon mode Benjamin Marzinski
2011-09-01  2:46 ` Benjamin Marzinski
2011-09-29  3:38   ` Benjamin Marzinski
2011-09-30  5:58     ` Christophe Varoqui
2011-09-30 15:08       ` Benjamin Marzinski

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.