[PATCH v2] xen: fix build without pci passthrough

Roger Pau Monne posted 1 patch 1 week ago
Failed in applying to current master (apply log)
hw/xen/xen-common.c | 4 ++++
hw/xen/xen_pt.h     | 6 ++++++
2 files changed, 10 insertions(+)

[PATCH v2] xen: fix build without pci passthrough

Posted by Roger Pau Monne 1 week ago
has_igd_gfx_passthru is only available when QEMU is built with
CONFIG_XEN_PCI_PASSTHROUGH, and hence shouldn't be used in common
code without checking if it's available.

Fixes: 46472d82322d0 ('xen: convert "-machine igd-passthru" to an accelerator property')
Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
---
Cc: Stefano Stabellini <sstabellini@kernel.org>
Cc: Anthony Perard <anthony.perard@citrix.com>
Cc: Paul Durrant <paul@xen.org>
Cc: xen-devel@lists.xenproject.org
---
Changes since v1:
 - Do not include osdep in header file.
 - Always add the setters/getters of igd-passthru, report an error
   when attempting to set igd-passthru without built in
   pci-passthrough support.
---
 hw/xen/xen-common.c | 4 ++++
 hw/xen/xen_pt.h     | 6 ++++++
 2 files changed, 10 insertions(+)

diff --git a/hw/xen/xen-common.c b/hw/xen/xen-common.c
index 70564cc952..d758770da0 100644
--- a/hw/xen/xen-common.c
+++ b/hw/xen/xen-common.c
@@ -134,7 +134,11 @@ static bool xen_get_igd_gfx_passthru(Object *obj, Error **errp)
 
 static void xen_set_igd_gfx_passthru(Object *obj, bool value, Error **errp)
 {
+#ifdef CONFIG_XEN_PCI_PASSTHROUGH
     has_igd_gfx_passthru = value;
+#else
+    error_setg(errp, "Xen PCI passthrough support not built in");
+#endif
 }
 
 static void xen_setup_post(MachineState *ms, AccelState *accel)
diff --git a/hw/xen/xen_pt.h b/hw/xen/xen_pt.h
index 179775db7b..7430235a27 100644
--- a/hw/xen/xen_pt.h
+++ b/hw/xen/xen_pt.h
@@ -322,7 +322,13 @@ extern void *pci_assign_dev_load_option_rom(PCIDevice *dev,
                                             unsigned int domain,
                                             unsigned int bus, unsigned int slot,
                                             unsigned int function);
+
+#ifdef CONFIG_XEN_PCI_PASSTHROUGH
 extern bool has_igd_gfx_passthru;
+#else
+# define has_igd_gfx_passthru false
+#endif
+
 static inline bool is_igd_vga_passthrough(XenHostPCIDevice *dev)
 {
     return (has_igd_gfx_passthru
-- 
2.26.2


Re: [PATCH v2] xen: fix build without pci passthrough

Posted by Anthony PERARD 1 week ago
On Tue, May 19, 2020 at 04:31:01PM +0200, Roger Pau Monne wrote:
> has_igd_gfx_passthru is only available when QEMU is built with
> CONFIG_XEN_PCI_PASSTHROUGH, and hence shouldn't be used in common
> code without checking if it's available.
> 
> Fixes: 46472d82322d0 ('xen: convert "-machine igd-passthru" to an accelerator property')
> Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
> ---
> Cc: Stefano Stabellini <sstabellini@kernel.org>
> Cc: Anthony Perard <anthony.perard@citrix.com>
> Cc: Paul Durrant <paul@xen.org>
> Cc: xen-devel@lists.xenproject.org
> ---
> Changes since v1:
>  - Do not include osdep in header file.
>  - Always add the setters/getters of igd-passthru, report an error
>    when attempting to set igd-passthru without built in
>    pci-passthrough support.
> ---
>  hw/xen/xen-common.c | 4 ++++
>  hw/xen/xen_pt.h     | 6 ++++++
>  2 files changed, 10 insertions(+)
> 
> diff --git a/hw/xen/xen-common.c b/hw/xen/xen-common.c
> index 70564cc952..d758770da0 100644
> --- a/hw/xen/xen-common.c
> +++ b/hw/xen/xen-common.c
> @@ -134,7 +134,11 @@ static bool xen_get_igd_gfx_passthru(Object *obj, Error **errp)
>  
>  static void xen_set_igd_gfx_passthru(Object *obj, bool value, Error **errp)
>  {
> +#ifdef CONFIG_XEN_PCI_PASSTHROUGH
>      has_igd_gfx_passthru = value;
> +#else
> +    error_setg(errp, "Xen PCI passthrough support not built in");
> +#endif
>  }
>  

There's an issue that I haven't thought about before.
CONFIG_XEN_PCI_PASSTHROUGH is never defined in xen-common.c. So
xen_set_igd_gfx_passthru will always return an error.

I'm not sure what to do about that yet, maybe change the way that
CONFIG_ is defined, or maybe have have the setter/getter in xen_pt.c
with a stub in stubs/ which would return an error. or maybe some other
way.

-- 
Anthony PERARD

Re: [PATCH v2] xen: fix build without pci passthrough

Posted by Roger Pau Monné 1 week ago
On Tue, May 19, 2020 at 04:52:58PM +0100, Anthony PERARD wrote:
> On Tue, May 19, 2020 at 04:31:01PM +0200, Roger Pau Monne wrote:
> > has_igd_gfx_passthru is only available when QEMU is built with
> > CONFIG_XEN_PCI_PASSTHROUGH, and hence shouldn't be used in common
> > code without checking if it's available.
> > 
> > Fixes: 46472d82322d0 ('xen: convert "-machine igd-passthru" to an accelerator property')
> > Signed-off-by: Roger Pau Monné <roger.pau@citrix.com>
> > ---
> > Cc: Stefano Stabellini <sstabellini@kernel.org>
> > Cc: Anthony Perard <anthony.perard@citrix.com>
> > Cc: Paul Durrant <paul@xen.org>
> > Cc: xen-devel@lists.xenproject.org
> > ---
> > Changes since v1:
> >  - Do not include osdep in header file.
> >  - Always add the setters/getters of igd-passthru, report an error
> >    when attempting to set igd-passthru without built in
> >    pci-passthrough support.
> > ---
> >  hw/xen/xen-common.c | 4 ++++
> >  hw/xen/xen_pt.h     | 6 ++++++
> >  2 files changed, 10 insertions(+)
> > 
> > diff --git a/hw/xen/xen-common.c b/hw/xen/xen-common.c
> > index 70564cc952..d758770da0 100644
> > --- a/hw/xen/xen-common.c
> > +++ b/hw/xen/xen-common.c
> > @@ -134,7 +134,11 @@ static bool xen_get_igd_gfx_passthru(Object *obj, Error **errp)
> >  
> >  static void xen_set_igd_gfx_passthru(Object *obj, bool value, Error **errp)
> >  {
> > +#ifdef CONFIG_XEN_PCI_PASSTHROUGH
> >      has_igd_gfx_passthru = value;
> > +#else
> > +    error_setg(errp, "Xen PCI passthrough support not built in");
> > +#endif
> >  }
> >  
> 
> There's an issue that I haven't thought about before.
> CONFIG_XEN_PCI_PASSTHROUGH is never defined in xen-common.c. So
> xen_set_igd_gfx_passthru will always return an error.
> 
> I'm not sure what to do about that yet, maybe change the way that
> CONFIG_ is defined, or maybe have have the setter/getter in xen_pt.c
> with a stub in stubs/ which would return an error. or maybe some other
> way.

Hm, I think making it available is OK? Would it make sense to set it
in $config_host_mak instead of $config_target_mak?

It's a host property, not a target one AFAICT, as the support in the
host determines whether PCI passthrough is supported or not.

Thanks, Roger.