Build breakage between kvm-userspace.git HEAD and 2.6.29-rc8

DISCUSSION Gregory Haskins 4 participants 10 messages Mar 13, 2009 – Mar 17, 2009 KVM

Share & tools

Gregory Haskins

author

2009-03-13 15:33 UTC #0 lore raw

Hi kvm-devs,
Been away from the kvm scene for a while...hope everyone is well.

FYI: There are a few build breakages between kvm-userspace.git HEAD (f29d402ab885d30bfa7350378cff911c9c17a226) when you compile against Linus' linux-2.6.git HEAD (v2.6.29-rc8-041b62374c7fedc11a8a1eeda2868612d3d1436c)

I am not sure if this is a supported configuration (e.g. perhaps you only support building kvm-userspace HEAD against Avi's kernel tree?). But in case this was unintentional, read on...

First error:

libkvm.c:702: error: ‘struct kvm_irq_level’ has no member named ‘status’ make[1]: *** [libkvm.o] Error 1

Which is attributable to:

commit ea1b668e7684dc43e9d198ba0b25fe47a0b2acd2

Author: Gleb Natapov <[email protected]>
Date:   Wed Feb 4 17:30:01 2009 +0200

kvm: qemu: handle IRQ status injection in userspace

Reverting this patch does allow the build to get past this error. Then I hit the next one:

device-assignment.o: In function `deassign_device':
/home/ghaskins/sandbox/git/fabric/kvm-userspace/qemu/hw/device-assignment.c:562:
undefined reference to `kvm_deassign_pci_device'

Reverting:

commit 1435f18213edee9fabf936f5d6be70054f1f2bd1

Author: Weidong Han <[email protected]>
Date:   Wed Feb 18 14:33:41 2009 +0800

kvm: qemu: fix hot remove assigned device

Followed by:

commit 71ae34caaa961330f53f40090cd180a74fcda5fc

Author: Weidong Han <[email protected]>
Date:   Wed Feb 18 14:28:30 2009 +0800

kvm: qemu: deassign device from guest

Does fix the build. I realize reversion is probably not the desirable long term solution, but at least it points to the proper places that need some backwards-compat love ;)

Regards,
-Greg

Avi Kivity

2009-03-15 09:23 UTC #1 lore raw

Quoted 11 lines · Gregory Haskins

Hi kvm-devs,
Been away from the kvm scene for a while...hope everyone is well.
FYI: There are a few build breakages between kvm-userspace.git HEAD (f29d402ab885d30bfa7350378cff911c9c17a226) when you compile against Linus' linux-2.6.git HEAD (v2.6.29-rc8-041b62374c7fedc11a8a1eeda2868612d3d1436c)
I am not sure if this is a supported configuration (e.g. perhaps you
only support building kvm-userspace HEAD against Avi's kernel tree?). But in case this was unintentional, read on...

It's a semi-supported configuration. We don't test it, but we'll fix reported bugs (or apply patches...).

Quoted 5 lines

First error:
libkvm.c:702: error: ‘struct kvm_irq_level’ has no member named ‘status’ make[1]: *** [libkvm.o] Error 1

Quoted 8 lines

Which is attributable to:
commit ea1b668e7684dc43e9d198ba0b25fe47a0b2acd2
Author: Gleb Natapov <[email protected]>
Date:   Wed Feb 4 17:30:01 2009 +0200
kvm: qemu: handle IRQ status injection in userspace

This is due to a missing #ifdef KVM_CAP_... somewhere.

Gleb, can you take a look?

[more breakage]

Let's take this one at a time.

Signature · 1 line

error compiling committee.c: too many arguments to function

Gleb Natapov

2009-03-15 10:23 UTC #2 lore raw

Quoted 28 lines · Avi Kivity

Quoted 12 lines · Gregory Haskins
Hi kvm-devs,
Been away from the kvm scene for a while...hope everyone is well.
FYI: There are a few build breakages between kvm-userspace.git HEAD (f29d402ab885d30bfa7350378cff911c9c17a226) when you compile against Linus' linux-2.6.git HEAD (v2.6.29-rc8-041b62374c7fedc11a8a1eeda2868612d3d1436c)
I am not sure if this is a supported configuration (e.g. perhaps you only support building kvm-userspace HEAD against Avi's kernel tree?). But in case this was unintentional, read on...
It's a semi-supported configuration. We don't test it, but we'll fix reported bugs (or apply patches...).
Quoted 11 lines
First error:
libkvm.c:702: error: ‘struct kvm_irq_level’ has no member named ‘status’ make[1]: *** [libkvm.o] Error 1
Which is attributable to:
commit ea1b668e7684dc43e9d198ba0b25fe47a0b2acd2
Author: Gleb Natapov <[email protected]>
Date:   Wed Feb 4 17:30:01 2009 +0200

Quoted 7 lines

Quoted 2 lines
kvm: qemu: handle IRQ status injection in userspace
This is due to a missing #ifdef KVM_CAP_... somewhere.
Gleb, can you take a look?

Does this help?

Signed-off-by Gleb Natapov <[email protected]>

libkvm/libkvm.c +6 -1

index 0ac1c28..3e6bef1 100644
@@ -667,9 +667,14 @@ int kvm_set_irq_level(kvm_context_t kvm, int irq, int level, int *status)
 	if (r == -1)
 		perror("kvm_set_irq_level");
 
-	if (status)
+	if (status) {
+#if defined(KVM_CAP_IRQ_INJECT_STATUS) && defined(KVM_IRQ_LINE_STATUS)
 		*status = (kvm->irqchip_inject_ioctl == KVM_IRQ_LINE) ?
 			1 : event.status;
+#else
+		*status = 1;
+#endif
+	}
 
 	return 1;
 }

--
Gleb.

Avi Kivity

2009-03-15 10:43 UTC #3 lore raw

Quoted 11 lines · Gleb Natapov

libkvm/libkvm.c +1 -1

index 0ac1c28..3e6bef1 100644
@@ -667,9 +667,14 @@ int kvm_set_irq_level(kvm_context_t kvm, int irq, int level, int *status)
 	if (r == -1)
 		perror("kvm_set_irq_level");
 
-	if (status)
+	if (status) {

libkvm/libkvm.c int kvm_set_irq_level(kvm_context_t kvm, int irq, int level, int *status) · line 680

Quoted 10 lines

libkvm/libkvm.c +5 -0

+#if defined(KVM_CAP_IRQ_INJECT_STATUS) && defined(KVM_IRQ_LINE_STATUS)
 		*status = (kvm->irqchip_inject_ioctl == KVM_IRQ_LINE) ?
 			1 : event.status;
+#else
+		*status = 1;
+#endif
+	}
 
 	return 1;
 }

Isn't #ifdef KVM_CAP_... sufficient? If it is defined it should imply everything needed for injection status is defined and works.

Signature · 1 line

error compiling committee.c: too many arguments to function

Gleb Natapov

2009-03-15 10:45 UTC #4 lore raw

Quoted 15 lines · Avi Kivity

Quoted 14 lines · Gleb Natapov

libkvm/libkvm.c +4 -0

index 0ac1c28..3e6bef1 100644
@@ -667,9 +667,14 @@ int kvm_set_irq_level(kvm_context_t kvm, int irq, int level, int *status)
 	if (r == -1)
 		perror("kvm_set_irq_level");
 -	if (status)
+	if (status) {
+#if defined(KVM_CAP_IRQ_INJECT_STATUS) && defined(KVM_IRQ_LINE_STATUS)
 		*status = (kvm->irqchip_inject_ioctl == KVM_IRQ_LINE) ?
 			1 : event.status;
+#else
+		*status = 1;

Quoted 9 lines

libkvm/libkvm.c int kvm_set_irq_level(kvm_context_t kvm, int irq, int level, int *status)
Quoted 5 lines
libkvm/libkvm.c +2 -0
+#endif
+	}
  	return 1;
 }
 
Isn't #ifdef KVM_CAP_... sufficient? If it is defined it should imply everything needed for injection status is defined and works.

It should be sufficient. Just extra care. Remove?

--
Gleb.

Avi Kivity

2009-03-15 12:19 UTC #5 lore raw

Quoted 5 lines · Gleb Natapov

Quoted 3 lines
Isn't #ifdef KVM_CAP_... sufficient? If it is defined it should imply everything needed for injection status is defined and works.
It should be sufficient. Just extra care. Remove?

I removed it locally and applied.

Signature · 1 line

error compiling committee.c: too many arguments to function

Han, Weidong

2009-03-16 09:27 UTC #6 lore raw

In reply to Gregory Haskins

Quoted 43 lines · Gregory Haskins

Hi kvm-devs,
Been away from the kvm scene for a while...hope everyone is well.
FYI: There are a few build breakages between kvm-userspace.git HEAD (f29d402ab885d30bfa7350378cff911c9c17a226) when you compile against Linus' linux-2.6.git HEAD (v2.6.29-rc8-041b62374c7fedc11a8a1eeda2868612d3d1436c)
I am not sure if this is a supported configuration (e.g. perhaps you only support building kvm-userspace HEAD against Avi's kernel tree?). But in case this was unintentional, read on...
First error:
libkvm.c:702: error: 'struct kvm_irq_level' has no member named 'status' make[1]: *** [libkvm.o] Error 1
Which is attributable to:
commit ea1b668e7684dc43e9d198ba0b25fe47a0b2acd2
Author: Gleb Natapov <[email protected]>
Date:   Wed Feb 4 17:30:01 2009 +0200
kvm: qemu: handle IRQ status injection in userspace
Reverting this patch does allow the build to get past this error. Then I hit the next one:
device-assignment.o: In function `deassign_device':
/home/ghaskins/sandbox/git/fabric/kvm-userspace/qemu/hw/device-assignment.c:562:
undefined reference to `kvm_deassign_pci_device'
Reverting:
commit 1435f18213edee9fabf936f5d6be70054f1f2bd1
Author: Weidong Han <[email protected]>
Date:   Wed Feb 18 14:33:41 2009 +0800
kvm: qemu: fix hot remove assigned device
Followed by:
commit 71ae34caaa961330f53f40090cd180a74fcda5fc

Quoted 9 lines

Author: Weidong Han <[email protected]>
Date:   Wed Feb 18 14:28:30 2009 +0800
kvm: qemu: deassign device from guest
Does fix the build. I realize reversion is probably not the desirable long term solution, but at least it points to the proper places that need some backwards-compat love ;)

kvm_deassign_pci_device is defined when KVM_CAP_DEVICE_DEASSIGNMENT is defined. Following patch should fix the breakage.

qemu/hw/device-assignment.c +2 -0

index 7c73210..6765683 100644
@@ -551,6 +551,7 @@ static int assign_irq(AssignedDevInfo *adev)
 
 static void deassign_device(AssignedDevInfo *adev)
 {
+#ifdef KVM_CAP_DEVICE_DEASSIGNMENT
     struct kvm_assigned_pci_dev assigned_dev_data;
     AssignedDevice *dev = adev->assigned_dev;
     int r;

@@ -563,6 +564,7 @@ static void deassign_device(AssignedDevInfo *adev)
     if (r < 0)
 	fprintf(stderr, "Failed to deassign device \"%s\" : %s\n",
                 adev->name, strerror(-r));
+#endif
 }
 
 void remove_assigned_device(AssignedDevInfo *adev)

Quoted 2 lines

Regards,
-Greg

Avi Kivity

2009-03-16 09:48 UTC #7 lore raw

Quoted 2 lines · Weidong

kvm_deassign_pci_device is defined when KVM_CAP_DEVICE_DEASSIGNMENT is defined. Following patch should fix the breakage.

Signoff please.

Signature · 1 line

error compiling committee.c: too many arguments to function

Han, Weidong

2009-03-17 01:03 UTC #8 lore raw

Quoted 6 lines · Avi Kivity

Quoted 3 lines · Weidong
kvm_deassign_pci_device is defined when KVM_CAP_DEVICE_DEASSIGNMENT is defined. Following patch should fix the breakage.
Signoff please.

Sorry for missing it. Signed-off-by: Weidong Han <[email protected]>

Regards,
Weidong

Avi Kivity

2009-03-17 09:36 UTC #9 lore raw

Quoted 4 lines · Weidong

Quoted 2 lines · Avi Kivity
Han, Weidong wrote:

Quoted 10 lines

Quoted 6 lines
Quoted 4 lines
kvm_deassign_pci_device is defined when KVM_CAP_DEVICE_DEASSIGNMENT is defined. Following patch should fix the breakage.
Signoff please.
Sorry for missing it. Signed-off-by: Weidong Han <[email protected]>

Thanks; applied.

Signature · 1 line

error compiling committee.c: too many arguments to function