[PATCH for-4.6 0/3] FreeBSD fixes

[PATCH for-4.6 0/3] FreeBSD fixes

[Roger Pau Monne]

Misc fixes for FreeBSD that affect libxl and the recently added 
xendriverdomain rc.d script.

[PATCH for-4.6 1/3] libxl: include sys/endian.h for FreeBSD

[Roger Pau Monne]

be64toh and friends are declared in sys/endian.h on FreeBSD, so include it
as part of libxl_osdeps.h.

Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Cc: Ian Jackson <ian.jackson@eu.citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
---
 tools/libxl/libxl_osdeps.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/tools/libxl/libxl_osdeps.h b/tools/libxl/libxl_osdeps.h
index 08eaf0c..b265df8 100644
--- a/tools/libxl/libxl_osdeps.h
+++ b/tools/libxl/libxl_osdeps.h
@@ -42,6 +42,7 @@
 #define SYSFS_PCIBACK_DRIVER   "/dev/null"
 #define NETBACK_NIC_NAME       "xnb%u.%d"
 #include <libutil.h>
+#include <sys/endian.h>
 #endif
 
 #ifndef SYSFS_PCIBACK_DRIVER
-- 
1.9.5 (Apple Git-50.3)


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: [PATCH for-4.6 1/3] libxl: include sys/endian.h for FreeBSD

[Wei Liu]

On Mon, Jul 20, 2015 at 04:55:00PM +0200, Roger Pau Monne wrote:
> be64toh and friends are declared in sys/endian.h on FreeBSD, so include it
> as part of libxl_osdeps.h.
> 
> Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
> Cc: Ian Jackson <ian.jackson@eu.citrix.com>
> Cc: Ian Campbell <ian.campbell@citrix.com>
> Cc: Wei Liu <wei.liu2@citrix.com>

Acked-by: Wei Liu <wei.liu2@citrix.com>

> ---
>  tools/libxl/libxl_osdeps.h | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/tools/libxl/libxl_osdeps.h b/tools/libxl/libxl_osdeps.h
> index 08eaf0c..b265df8 100644
> --- a/tools/libxl/libxl_osdeps.h
> +++ b/tools/libxl/libxl_osdeps.h
> @@ -42,6 +42,7 @@
>  #define SYSFS_PCIBACK_DRIVER   "/dev/null"
>  #define NETBACK_NIC_NAME       "xnb%u.%d"
>  #include <libutil.h>
> +#include <sys/endian.h>
>  #endif
>  
>  #ifndef SYSFS_PCIBACK_DRIVER
> -- 
> 1.9.5 (Apple Git-50.3)

Re: [PATCH for-4.6 1/3] libxl: include sys/endian.h for FreeBSD

[Ian Campbell]

On Tue, 2015-07-21 at 10:36 +0100, Wei Liu wrote:
> On Mon, Jul 20, 2015 at 04:55:00PM +0200, Roger Pau Monne wrote:
> > be64toh and friends are declared in sys/endian.h on FreeBSD, so 
> > include it
> > as part of libxl_osdeps.h.
> > 
> > Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
> > Cc: Ian Jackson <ian.jackson@eu.citrix.com>
> > Cc: Ian Campbell <ian.campbell@citrix.com>
> > Cc: Wei Liu <wei.liu2@citrix.com>
> 
> Acked-by: Wei Liu <wei.liu2@citrix.com>

Applied.

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

[PATCH for-4.6 2/3] libxl/psr: use Xen error codes when checking hypercall return values

[Roger Pau Monne]

We cannot use the systems errno values when checking return values from Xen,
because some OSes don't have the same set of errno definitions. Instead
use the definitions present in Xen public errno.h header.

Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Cc: Ian Jackson <ian.jackson@eu.citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
---
I have not checked if there are other places in libxl that need similar
treatment, I just came around this because FreeBSD doesn't have EBADSLT
defined.
---
 tools/libxl/libxl_internal.h |  1 +
 tools/libxl/libxl_psr.c      | 26 +++++++++++++-------------
 2 files changed, 14 insertions(+), 13 deletions(-)

diff --git a/tools/libxl/libxl_internal.h b/tools/libxl/libxl_internal.h
index 2b6b2a0..cf5db8a 100644
--- a/tools/libxl/libxl_internal.h
+++ b/tools/libxl/libxl_internal.h
@@ -55,6 +55,7 @@
 #include "xentoollog.h"
 
 #include <xen/io/xenbus.h>
+#include <xen/errno.h>
 
 #ifdef LIBXL_H
 # error libxl.h should be included via libxl_internal.h, not separately
diff --git a/tools/libxl/libxl_psr.c b/tools/libxl/libxl_psr.c
index 2a07777..5711c38 100644
--- a/tools/libxl/libxl_psr.c
+++ b/tools/libxl/libxl_psr.c
@@ -24,17 +24,17 @@ static void libxl__psr_log_err_msg(libxl__gc *gc, int err)
     char *msg;
 
     switch (err) {
-    case ENOSYS:
-    case EOPNOTSUPP:
+    case XEN_ENOSYS:
+    case XEN_EOPNOTSUPP:
         msg = "unsupported operation";
         break;
-    case ESRCH:
+    case XEN_ESRCH:
         msg = "invalid domain ID";
         break;
-    case EBADSLT:
+    case XEN_EBADSLT:
         msg = "socket is not supported";
         break;
-    case EFAULT:
+    case XEN_EFAULT:
         msg = "failed to exchange data with Xen";
         break;
     default:
@@ -50,16 +50,16 @@ static void libxl__psr_cmt_log_err_msg(libxl__gc *gc, int err)
     char *msg;
 
     switch (err) {
-    case ENODEV:
+    case XEN_ENODEV:
         msg = "CMT is not supported in this system";
         break;
-    case EEXIST:
+    case XEN_EEXIST:
         msg = "CMT is already attached to this domain";
         break;
-    case ENOENT:
+    case XEN_ENOENT:
         msg = "CMT is not attached to this domain";
         break;
-    case EUSERS:
+    case XEN_EUSERS:
         msg = "no free RMID available";
         break;
     default:
@@ -75,16 +75,16 @@ static void libxl__psr_cat_log_err_msg(libxl__gc *gc, int err)
     char *msg;
 
     switch (err) {
-    case ENODEV:
+    case XEN_ENODEV:
         msg = "CAT is not supported in this system";
         break;
-    case ENOENT:
+    case XEN_ENOENT:
         msg = "CAT is not enabled on the socket";
         break;
-    case EUSERS:
+    case XEN_EUSERS:
         msg = "no free COS available";
         break;
-    case EEXIST:
+    case XEN_EEXIST:
         msg = "The same CBM is already set to this domain";
         break;
 
-- 
1.9.5 (Apple Git-50.3)


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: [PATCH for-4.6 2/3] libxl/psr: use Xen error codes when checking hypercall return values

[Ian Jackson]

Roger Pau Monne writes ("[PATCH for-4.6 2/3] libxl/psr: use Xen error codes when checking hypercall return values"):
> We cannot use the systems errno values when checking return values from Xen,
> because some OSes don't have the same set of errno definitions. Instead
> use the definitions present in Xen public errno.h header.
> 
> Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
> Cc: Ian Jackson <ian.jackson@eu.citrix.com>
> Cc: Ian Campbell <ian.campbell@citrix.com>
> Cc: Wei Liu <wei.liu2@citrix.com>

Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>

> I have not checked if there are other places in libxl that need similar
> treatment, I just came around this because FreeBSD doesn't have EBADSLT
> defined.

Right.  There is probably more of this, but that's not a blocker for
this patch.

Maybe it would be worth mentioning this issue in CODING_STYLE.

Ian.

Re: [PATCH for-4.6 2/3] libxl/psr: use Xen error codes when checking hypercall return values

[Ian Jackson]

Ian Jackson writes ("Re: [PATCH for-4.6 2/3] libxl/psr: use Xen error codes when checking hypercall return values"):
> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>

Actually, I withdraw that.  I was confused about how things worked.
To summarise an irc discussion:

 * On NetBSD, hypercall numbers are translated by the privcmd driver.
   This is probably correct.  (Not sure if FreeBSD does the same.)

 * So the libxc API is that errno values are OS errno values.  Roger's
   patch is wrong.

 * A corrollary is that the hypervisor may not use errno values which
   are not in the toolstack OS.  Since the hypervisor is supposed to
   be fairly portable to different toolstacks, that means we shouldn't
   be using errno values which are not in POSIX.

 * Re EBADSLT in particular:
     - There is one return site in 4.5, which is in hvm_save_one,
       which is toolstack-onloy functionality.  We can therefore
       change it for 4.6.
     - The other sites where EBADSLT are generated in staging must
       also be fixed.

So the correct approach is:

 - Patch to change EBADSLT to something else, everywhere.
 - Patch to remove EBADSLT from hypervisor errno list (if it's not
   just been renamed/renumbered).
 - If FreeBSD privcmd does not translate errno numbers, make it do so.

Ian.

Re: [PATCH for-4.6 2/3] libxl/psr: use Xen error codes when checking hypercall return values

[Andrew Cooper]

On 21/07/15 12:36, Ian Jackson wrote:
> Ian Jackson writes ("Re: [PATCH for-4.6 2/3] libxl/psr: use Xen error codes when checking hypercall return values"):
>> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
> Actually, I withdraw that.  I was confused about how things worked.
> To summarise an irc discussion:
>
>  * On NetBSD, hypercall numbers are translated by the privcmd driver.
>    This is probably correct.  (Not sure if FreeBSD does the same.)
>
>  * So the libxc API is that errno values are OS errno values.  Roger's
>    patch is wrong.
>
>  * A corrollary is that the hypervisor may not use errno values which
>    are not in the toolstack OS.  Since the hypervisor is supposed to
>    be fairly portable to different toolstacks, that means we shouldn't
>    be using errno values which are not in POSIX.
>
>  * Re EBADSLT in particular:
>      - There is one return site in 4.5, which is in hvm_save_one,
>        which is toolstack-onloy functionality.  We can therefore
>        change it for 4.6.
>      - The other sites where EBADSLT are generated in staging must
>        also be fixed.
>
> So the correct approach is:
>
>  - Patch to change EBADSLT to something else, everywhere.
>  - Patch to remove EBADSLT from hypervisor errno list (if it's not
>    just been renamed/renumbered).
>  - If FreeBSD privcmd does not translate errno numbers, make it do so.

EBADSLT isn't the only example here.  I believe ENODATA is a similar
example.

If we are going to do this, we should prune all non-posix errnos out of
the hypervisor errno list.

~Andrew

Re: [PATCH for-4.6 2/3] libxl/psr: use Xen error codes when checking hypercall return values

[Ian Jackson]

Andrew Cooper writes ("Re: [Xen-devel] [PATCH for-4.6 2/3] libxl/psr: use Xen error codes when checking hypercall return values"):
> EBADSLT isn't the only example here.  I believe ENODATA is a similar
> example.

http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/errno.h.html

http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/sys/errno.h?rev=1.40&content-type=text/x-cvsweb-markup&only_with_tag=MAIN

But

http://fxr.watson.org/fxr/source/sys/errno.h

How annoying.

Ian.

Re: [PATCH for-4.6 2/3] libxl/psr: use Xen error codes when checking hypercall return values

[Ian Campbell]

On Tue, 2015-07-21 at 13:40 +0100, Ian Jackson wrote:
> Andrew Cooper writes ("Re: [Xen-devel] [PATCH for-4.6 2/3] libxl/psr: 
> use Xen error codes when checking hypercall return values"):
> > EBADSLT isn't the only example here.  I believe ENODATA is a 
> > similar
> > example.
> 
> http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/errno.h.html
> 
> http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/sys/errno.h?rev=1.40&cont
> ent-type=text/x-cvsweb-markup&only_with_tag=MAIN
> 
> But
> 
> http://fxr.watson.org/fxr/source/sys/errno.h
> 
> How annoying.

THat's just a FreeBSD bug then?

Re: [PATCH for-4.6 2/3] libxl/psr: use Xen error codes when checking hypercall return values

[Jan Beulich]

>>> On 21.07.15 at 14:47, <ian.campbell@citrix.com> wrote:
> On Tue, 2015-07-21 at 13:40 +0100, Ian Jackson wrote:
>> Andrew Cooper writes ("Re: [Xen-devel] [PATCH for-4.6 2/3] libxl/psr: 
>> use Xen error codes when checking hypercall return values"):
>> > EBADSLT isn't the only example here.  I believe ENODATA is a 
>> > similar
>> > example.
>> 
>> http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/errno.h.html 
>> 
>> http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/sys/errno.h?rev=1.40&cont 
>> ent-type=text/x-cvsweb-markup&only_with_tag=MAIN
>> 
>> But
>> 
>> http://fxr.watson.org/fxr/source/sys/errno.h 
>> 
>> How annoying.
> 
> THat's just a FreeBSD bug then?

Why? The OpenGroup page above clearly says that the two so
far identified values are an extension.

Jan

Re: [PATCH for-4.6 2/3] libxl/psr: use Xen error codes when checking hypercall return values

[Ian Campbell]

On Tue, 2015-07-21 at 07:14 -0600, Jan Beulich wrote:
> > 
> > > > On 21.07.15 at 14:47, <ian.campbell@citrix.com> wrote:
> > On Tue, 2015-07-21 at 13:40 +0100, Ian Jackson wrote:
> > > Andrew Cooper writes ("Re: [Xen-devel] [PATCH for-4.6 2/3] 
> > > libxl/psr: 
> > > use Xen error codes when checking hypercall return values"):
> > > > EBADSLT isn't the only example here.  I believe ENODATA is a 
> > > > similar
> > > > example.
> > > 
> > > http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/errno.h.
> > > html 
> > > 
> > > http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/sys/errno.h?rev=1.40&
> > > cont 
> > > ent-type=text/x-cvsweb-markup&only_with_tag=MAIN
> > > 
> > > But
> > > 
> > > http://fxr.watson.org/fxr/source/sys/errno.h 
> > > 
> > > How annoying.
> > 
> > THat's just a FreeBSD bug then?
> 
> Why? The OpenGroup page above clearly says that the two so
> far identified values are an extension.

I didn't realise that was what the XSR tag meant, sorry.

I note that OB is obsolete too FWIW...

I suppose it could be worse than "POSIX plus a couple of grandfathered
in exceptions (sorry)"...

Ian.

[PATCH for-4.6 3/3] hotplug/FreeBSD: fix xendriverdomain rc.d script

[Roger Pau Monne]

hotplugpath.sh by default is located in /usr/local/etc/xen/scripts on
FreeBSD. Instead of hardcoding it's location use the XEN_SCRIPT_DIR variable
like it's used on the xencommons rc.d script.

Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
Cc: Ian Jackson <ian.jackson@eu.citrix.com>
Cc: Ian Campbell <ian.campbell@citrix.com>
Cc: Wei Liu <wei.liu2@citrix.com>
---
 tools/hotplug/FreeBSD/rc.d/xendriverdomain.in | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tools/hotplug/FreeBSD/rc.d/xendriverdomain.in b/tools/hotplug/FreeBSD/rc.d/xendriverdomain.in
index 25e3edd..4063c06 100644
--- a/tools/hotplug/FreeBSD/rc.d/xendriverdomain.in
+++ b/tools/hotplug/FreeBSD/rc.d/xendriverdomain.in
@@ -7,7 +7,7 @@
 
 . /etc/rc.subr
 
-. /etc/xen/scripts/hotplugpath.sh
+. @XEN_SCRIPT_DIR@/hotplugpath.sh
 
 LD_LIBRARY_PATH="${libdir}"
 export LD_LIBRARY_PATH
-- 
1.9.5 (Apple Git-50.3)


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

Re: [PATCH for-4.6 3/3] hotplug/FreeBSD: fix xendriverdomain rc.d script

[Wei Liu]

On Mon, Jul 20, 2015 at 04:55:02PM +0200, Roger Pau Monne wrote:
> hotplugpath.sh by default is located in /usr/local/etc/xen/scripts on
> FreeBSD. Instead of hardcoding it's location use the XEN_SCRIPT_DIR variable
> like it's used on the xencommons rc.d script.
> 
> Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
> Cc: Ian Jackson <ian.jackson@eu.citrix.com>
> Cc: Ian Campbell <ian.campbell@citrix.com>
> Cc: Wei Liu <wei.liu2@citrix.com>

Acked-by: Wei Liu <wei.liu2@citrix.com>

> ---
>  tools/hotplug/FreeBSD/rc.d/xendriverdomain.in | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/tools/hotplug/FreeBSD/rc.d/xendriverdomain.in b/tools/hotplug/FreeBSD/rc.d/xendriverdomain.in
> index 25e3edd..4063c06 100644
> --- a/tools/hotplug/FreeBSD/rc.d/xendriverdomain.in
> +++ b/tools/hotplug/FreeBSD/rc.d/xendriverdomain.in
> @@ -7,7 +7,7 @@
>  
>  . /etc/rc.subr
>  
> -. /etc/xen/scripts/hotplugpath.sh
> +. @XEN_SCRIPT_DIR@/hotplugpath.sh
>  
>  LD_LIBRARY_PATH="${libdir}"
>  export LD_LIBRARY_PATH
> -- 
> 1.9.5 (Apple Git-50.3)

Re: [PATCH for-4.6 3/3] hotplug/FreeBSD: fix xendriverdomain rc.d script

[Ian Campbell]

On Tue, 2015-07-21 at 12:03 +0100, Wei Liu wrote:
> On Mon, Jul 20, 2015 at 04:55:02PM +0200, Roger Pau Monne wrote:
> > hotplugpath.sh by default is located in /usr/local/etc/xen/scripts 
> > on
> > FreeBSD. Instead of hardcoding it's location use the XEN_SCRIPT_DIR 
> > variable
> > like it's used on the xencommons rc.d script.
> > 
> > Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
> > Cc: Ian Jackson <ian.jackson@eu.citrix.com>
> > Cc: Ian Campbell <ian.campbell@citrix.com>
> > Cc: Wei Liu <wei.liu2@citrix.com>
> 
> Acked-by: Wei Liu <wei.liu2@citrix.com>

Me too, plus applied.

Thanks Roger, sorry for the breakage.


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel