Re: [PATCH 10/15] multipath: blacklist virtio-blk devices

Re: [PATCH 10/15] multipath: blacklist virtio-blk devices

[Xose Vazquez Perez]

Benjamin Marzinski wrote:

> virtio-blk devices don't report anything to use as a WWID, so
> multipath shouldn't waste it's time checking them.
> 
> Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
> ---
>  libmultipath/blacklist.c | 2 +-
>  multipath.conf.defaults  | 2 +-
>  2 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/libmultipath/blacklist.c b/libmultipath/blacklist.c
> index 05f1697..cea128c 100644
> --- a/libmultipath/blacklist.c
> +++ b/libmultipath/blacklist.c
> @@ -178,7 +178,7 @@ setup_default_blist (struct config * conf)
>  	if (store_ble(conf->blist_devnode, str, ORIGIN_DEFAULT))
>  		return 1;
>  
> -	str = STRDUP("^(td|hd)[a-z]");
> +	str = STRDUP("^(td|hd|vd)[a-z]");
                                      ^
                                      Missing "*" ?

Re: [PATCH 10/15] multipath: blacklist virtio-blk devices

[Christophe Varoqui]

[-- Attachment #1.1: Type: text/plain, Size: 1155 bytes --]

Hi Xose,

it sure wouldn't hurt, be as there is no trailing "$" I guess the pattern
does its job as-is.
Please let me know if I'm reading this wrong.

Best regards,
Christophe Varoqui
OpenSVC

On Sun, Mar 8, 2015 at 11:24 AM, Xose Vazquez Perez <xose.vazquez@gmail.com>
wrote:

> Benjamin Marzinski wrote:
>
> > virtio-blk devices don't report anything to use as a WWID, so
> > multipath shouldn't waste it's time checking them.
> >
> > Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
> > ---
> >  libmultipath/blacklist.c | 2 +-
> >  multipath.conf.defaults  | 2 +-
> >  2 files changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/libmultipath/blacklist.c b/libmultipath/blacklist.c
> > index 05f1697..cea128c 100644
> > --- a/libmultipath/blacklist.c
> > +++ b/libmultipath/blacklist.c
> > @@ -178,7 +178,7 @@ setup_default_blist (struct config * conf)
> >       if (store_ble(conf->blist_devnode, str, ORIGIN_DEFAULT))
> >               return 1;
> >
> > -     str = STRDUP("^(td|hd)[a-z]");
> > +     str = STRDUP("^(td|hd|vd)[a-z]");
>                                       ^
>                                       Missing "*" ?
>

[-- Attachment #1.2: Type: text/html, Size: 1847 bytes --]

[-- Attachment #2: Type: text/plain, Size: 0 bytes --]

Re: [PATCH 10/15] multipath: blacklist virtio-blk devices

[Benjamin Marzinski]

On Sun, Mar 08, 2015 at 11:33:38AM +0100, Christophe Varoqui wrote:
>    Hi Xose,
>    it sure wouldn't hurt, be as there is no trailing "$" I guess the pattern
>    does its job as-is.
>    Please let me know if I'm reading this wrong.

You're reading it right. What's there will blacklist devices like "vdbx"
just fine. We could add a "$" and fully specify the pattern to make sure
we don't catch any devices by accident, but I don't really see that
happening with device names, and I've never seen any complaints about
the existing devnode balcklists that don't have them.

-Ben

>    Best regards,
>    Christophe Varoqui
>    OpenSVC
>    On Sun, Mar 8, 2015 at 11:24 AM, Xose Vazquez Perez
>    <[1]xose.vazquez@gmail.com> wrote:
> 
>      Benjamin Marzinski wrote:
> 
>      > virtio-blk devices don't report anything to use as a WWID, so
>      > multipath shouldn't waste it's time checking them.
>      >
>      > Signed-off-by: Benjamin Marzinski <[2]bmarzins@redhat.com>
>      > ---
>      >  libmultipath/blacklist.c | 2 +-
>      >  multipath.conf.defaults  | 2 +-
>      >  2 files changed, 2 insertions(+), 2 deletions(-)
>      >
>      > diff --git a/libmultipath/blacklist.c b/libmultipath/blacklist.c
>      > index 05f1697..cea128c 100644
>      > --- a/libmultipath/blacklist.c
>      > +++ b/libmultipath/blacklist.c
>      > @@ -178,7 +178,7 @@ setup_default_blist (struct config * conf)
>      >       if (store_ble(conf->blist_devnode, str, ORIGIN_DEFAULT))
>      >               return 1;
>      >
>      > -     str = STRDUP("^(td|hd)[a-z]");
>      > +     str = STRDUP("^(td|hd|vd)[a-z]");
>                                            ^
>                                            Missing "*" ?
> 
> References
> 
>    Visible links
>    1. mailto:xose.vazquez@gmail.com
>    2. mailto:bmarzins@redhat.com

[PATCH 00/15] Multipath patch resync

[Benjamin Marzinski]

The first eight of these patches are resends of earlier patches, just
rebased with minor changes. The next two are minor changes, adding
multipath's -B option to mutipathd, and blacklisting vd.* devices.
The two after that allow multipath to read additional configuration
files in a configuration directory /etc/multipath/conf.d by default,
and then do some cleanup of the parser code that I noticed while
adding the configuration directories. The next patch adds two new
configuration parameters that allow multipathd to delay the
reintegration of paths that are failing repeatedly. The next one
fixes a hard to hit memory corruption bug, and the last one
updates the EMC builtin configuration to autodetect ALUA setups.

Benjamin Marzinski (15):
  correctly set partition delimiter on rename
  libmultipath: fix sysfs_get_size bug
  Revert "libmultipath: fixup strlcpy"
  libmultipath: refactor partmaps code
  multipathd: fix memory corruption issue
  multipathd: add deferred_remove support
  multipathd: don't buffer output with systemd
  libmutipath: allow blanks in device blacklist
  multipathd: read-only bindings
  multipath: blacklist virtio-blk devices
  multipath: enable configuration directory
  libmultipath: cleanup parser code
  multipathd: Add delayed path reintegration
  Fix doublee free of alias after map add failure
  libmultipath: autodetect ALUA CLARiiON configurations

 libmultipath/Makefile      |   6 +
 libmultipath/blacklist.c   |  18 ++-
 libmultipath/checkers.c    |   1 +
 libmultipath/checkers.h    |   9 ++
 libmultipath/config.c      |  69 ++++++++++-
 libmultipath/config.h      |  12 ++
 libmultipath/configure.c   |   3 +
 libmultipath/defaults.h    |   3 +
 libmultipath/devmapper.c   | 284 +++++++++++++++++++++++++++++----------------
 libmultipath/devmapper.h   |  11 +-
 libmultipath/dict.c        | 126 +++++++++++++++++---
 libmultipath/dict.h        |   1 +
 libmultipath/hwtable.c     |   2 +
 libmultipath/parser.c      | 173 ++++++++-------------------
 libmultipath/parser.h      |   9 +-
 libmultipath/print.c       |   2 +
 libmultipath/propsel.c     |  58 +++++++++
 libmultipath/propsel.h     |   3 +
 libmultipath/structs.h     |  17 +++
 libmultipath/structs_vec.c |   3 +-
 libmultipath/sysfs.c       |   2 +-
 libmultipath/util.c        |   3 +-
 multipath.conf.annotated   |  80 +++++++++++++
 multipath.conf.defaults    |   7 +-
 multipath/multipath.conf.5 |  54 +++++++++
 multipathd/main.c          |  86 ++++++++++----
 multipathd/multipathd.8    |   5 +
 27 files changed, 761 insertions(+), 286 deletions(-)

-- 
1.8.3.1

[PATCH 10/15] multipath: blacklist virtio-blk devices

[Benjamin Marzinski]

virtio-blk devices don't report anything to use as a WWID, so
multipath shouldn't waste it's time checking them.

Signed-off-by: Benjamin Marzinski <bmarzins@redhat.com>
---
 libmultipath/blacklist.c | 2 +-
 multipath.conf.defaults  | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/libmultipath/blacklist.c b/libmultipath/blacklist.c
index 05f1697..cea128c 100644
--- a/libmultipath/blacklist.c
+++ b/libmultipath/blacklist.c
@@ -178,7 +178,7 @@ setup_default_blist (struct config * conf)
 	if (store_ble(conf->blist_devnode, str, ORIGIN_DEFAULT))
 		return 1;
 
-	str = STRDUP("^(td|hd)[a-z]");
+	str = STRDUP("^(td|hd|vd)[a-z]");
 	if (!str)
 		return 1;
 	if (store_ble(conf->blist_devnode, str, ORIGIN_DEFAULT))
diff --git a/multipath.conf.defaults b/multipath.conf.defaults
index cf8cfdb..5e46fd8 100644
--- a/multipath.conf.defaults
+++ b/multipath.conf.defaults
@@ -29,7 +29,7 @@
 #}
 #blacklist {
 #	devnode "^(ram|raw|loop|fd|md|dm-|sr|scd|st)[0-9]*"
-#	devnode "^hd[a-z]"
+#	devnode "^(td|hd|vd)[a-z]"
 #	devnode "^dcssblk[0-9]*"
 #	devnode "^nvme.*"
 #	device {
-- 
1.8.3.1