[PATCH] kvm-userspace: move x86 specific calls introduced by device assignement to x86 files

[PATCH] kvm-userspace: move x86 specific calls introduced by device assignement to x86 files

[Ehrhardt Christian]

The device asignment patches added the x86 specific ioperm in qemu-kvm.
This patch moves
  qemu-kvm.c:kvm_do_ioperm()
to
  qemu-kvm-x86.c:kvm_arch_do_ioperm()
The patch also changes the qemu-kvm header and the includes according to that.

Signed-off-by: Christian Ehrhardt <ehrhardt@linux.vnet.ibm.com>
---

[diffstat]
 qemu-kvm-x86.c |    7 +++++++
 qemu-kvm.c     |   11 +++--------
 qemu-kvm.h     |    3 +++
 3 files changed, 13 insertions(+), 8 deletions(-)

[diff]

diff --git a/qemu/qemu-kvm-x86.c b/qemu/qemu-kvm-x86.c
--- a/qemu/qemu-kvm-x86.c
+++ b/qemu/qemu-kvm-x86.c
@@ -11,6 +11,7 @@
 
 #include <string.h>
 #include "hw/hw.h"
+#include <sys/io.h>
 
 #include "qemu-kvm.h"
 #include <libkvm.h>
@@ -717,3 +718,9 @@
 	}
     }
 }
+
+void kvm_arch_do_ioperm(void *_data)
+{
+    struct ioperm_data *data = _data;
+    ioperm(data->start_port, data->num, data->turn_on);
+}
diff --git a/qemu/qemu-kvm.c b/qemu/qemu-kvm.c
--- a/qemu/qemu-kvm.c
+++ b/qemu/qemu-kvm.c
@@ -27,7 +27,6 @@
 #include <sys/utsname.h>
 #include <sys/syscall.h>
 #include <sys/mman.h>
-#include <sys/io.h>
 
 #define bool _Bool
 #define false 0
@@ -1049,14 +1048,10 @@
     return kvm_unregister_coalesced_mmio(kvm_context, addr, size);
 }
 
-static void kvm_do_ioperm(void *_data)
-{
-    struct ioperm_data *data = _data;
-    ioperm(data->start_port, data->num, data->turn_on);
-}
-
+#ifdef USE_KVM_DEVICE_ASSIGNMENT
 void kvm_ioperm(CPUState *env, void *data)
 {
     if (kvm_enabled() && qemu_system_ready)
-	on_vcpu(env, kvm_do_ioperm, data);
+	on_vcpu(env, kvm_arch_do_ioperm, data);
 }
+#endif
diff --git a/qemu/qemu-kvm.h b/qemu/qemu-kvm.h
--- a/qemu/qemu-kvm.h
+++ b/qemu/qemu-kvm.h
@@ -93,7 +93,10 @@
 
 void qemu_kvm_system_reset_request(void);
 
+#ifdef USE_KVM_DEVICE_ASSIGNMENT
 void kvm_ioperm(CPUState *env, void *data);
+void kvm_arch_do_ioperm(void *_data);
+#endif
 
 #ifdef TARGET_PPC
 int handle_powerpc_dcr_read(int vcpu, uint32_t dcrn, uint32_t *data);

Re: [PATCH] kvm-userspace: move x86 specific calls introduced by device assignement to x86 files

[Hollis Blanchard]

On Mon, 2008-11-03 at 18:16 +0100, Ehrhardt Christian wrote:
> The device asignment patches added the x86 specific ioperm in qemu-kvm.
> This patch moves
>   qemu-kvm.c:kvm_do_ioperm()
> to
>   qemu-kvm-x86.c:kvm_arch_do_ioperm()
> The patch also changes the qemu-kvm header and the includes according to that.

The important point here (which should probably be in the changelog) is
that this patch fixes a PowerPC build break caused by "[PATCH] kvm:
qemu: support for assigning host PCI devices to guests".

-- 
Hollis Blanchard
IBM Linux Technology Center

Re: [PATCH] kvm-userspace: move x86 specific calls introduced by device assignement to x86 files

[Avi Kivity]

Ehrhardt Christian wrote:
> The device asignment patches added the x86 specific ioperm in qemu-kvm.
> This patch moves
>   qemu-kvm.c:kvm_do_ioperm()
> to
>   qemu-kvm-x86.c:kvm_arch_do_ioperm()
> The patch also changes the qemu-kvm header and the includes according to that.
>
>   

Applied, thanks.

-- 
error compiling committee.c: too many arguments to function

regression tests? (was Re: [PATCH] kvm-userspace: move x86 specific calls introduced by device assignement to x86 files)

[Hollis Blanchard]

On Tue, 2008-11-04 at 12:11 +0200, Avi Kivity wrote:
> Ehrhardt Christian wrote:
> > The device asignment patches added the x86 specific ioperm in qemu-kvm.
> > This patch moves
> >   qemu-kvm.c:kvm_do_ioperm()
> > to
> >   qemu-kvm-x86.c:kvm_arch_do_ioperm()
> > The patch also changes the qemu-kvm header and the includes according to that.
> >
> >   
> 
> Applied, thanks.

Avi, do you have a quick regression test you run before pushing new
patches? If so, we'd like to help you add PowerPC to that, if only to
detect build breaks like this one.

-- 
Hollis Blanchard
IBM Linux Technology Center

Re: regression tests? (was Re: [PATCH] kvm-userspace: move x86 specific calls introduced by device assignement to x86 files)

[Avi Kivity]

Hollis Blanchard wrote:
> Avi, do you have a quick regression test you run before pushing new
> patches? If so, we'd like to help you add PowerPC to that, if only to
> detect build breaks like this one

I have a fairly slow regression test for x86.  I would really like to 
get at least build tests for the other archs.  I actually have a ppc box 
I could use for it, just need to set it up in some corner and start testing.

-- 
error compiling committee.c: too many arguments to function

Re: regression tests? (was Re: [PATCH] kvm-userspace: move x86 specific calls introduced by device assignement to x86 files)

[Hollis Blanchard]

On Tue, 2008-11-04 at 17:25 +0200, Avi Kivity wrote:
> Hollis Blanchard wrote:
> > Avi, do you have a quick regression test you run before pushing new
> > patches? If so, we'd like to help you add PowerPC to that, if only to
> > detect build breaks like this one
> 
> I have a fairly slow regression test for x86.  I would really like to 
> get at least build tests for the other archs.  I actually have a ppc box 
> I could use for it, just need to set it up in some corner and start testing.

Out of curiosity, do you run the slow x86 tests after every commit?
After every batch of commits? Weekly?

Please let us know if we can help set up the PowerPC stuff; it's pretty
important to us to avoid these sorts of problems in the future. We have
tried to write fairly comprehensive build directions at
http://kvm.qumranet.com/kvmwiki/PowerPC_Source and runtime information
at http://kvm.qumranet.com/kvmwiki/PowerPC_Run .

-- 
Hollis Blanchard
IBM Linux Technology Center

Re: regression tests? (was Re: [PATCH] kvm-userspace: move x86 specific calls introduced by device assignement to x86 files)

[Avi Kivity]

Hollis Blanchard wrote:
>> I have a fairly slow regression test for x86.  I would really like to 
>> get at least build tests for the other archs.  I actually have a ppc box 
>> I could use for it, just need to set it up in some corner and start testing.
>>     
>
> Out of curiosity, do you run the slow x86 tests after every commit?
> After every batch of commits? Weekly?
>   

Depends on the changes.  I certainly batch commits, and even pipeline 
(keep committing while the test is running, in the expectation that is 
succeeds).  I don't batch qemu merges as they are more likely to fail 
one test or another.

> Please let us know if we can help set up the PowerPC stuff; it's pretty
> important to us to avoid these sorts of problems in the future. We have
> tried to write fairly comprehensive build directions at
> http://kvm.qumranet.com/kvmwiki/PowerPC_Source and runtime information
> at http://kvm.qumranet.com/kvmwiki/PowerPC_Run .
>   

Thanks.  I have to start by unpacking the box...

-- 
error compiling committee.c: too many arguments to function