apic: fix kvm build on UP without IOAPIC

PATCH Applied Michael S. Tsirkin 5 participants 9 messages Jul 1, 2012 – Jul 9, 2012

Subsystems (3)

x86 KVM Linux-next tree

lore

Share & tools

lore

Michael S. Tsirkin

author

2012-07-01 15:05 UTC #0 lore raw

PATCH 1/1 apic: fix kvm build on UP without IOAPIC

On UP i386, when APIC is disabled
# CONFIG_X86_UP_APIC is not set
# CONFIG_PCI_IOAPIC is not set

code looking at apicdrivers never has any effect but it still gets compiled in. In particular, this causes build failures with kvm, but it generally bloats the kernel unnecessarily.

Fix by defining both __apicdrivers and __apicdrivers_end
to be NULL when CONFIG_X86_LOCAL_APIC is unset: I verified that as the result any loop scanning __apicdrivers gets optimized out by the compiler.

Warning: a .config with apic disabled doesn't seem to boot for me (even without this patch). Still verifying why, meanwhile this patch is compile-tested only.

Signed-off-by Michael S. Tsirkin <[email protected]>

---

Note: if this patch makes sense, can x86 maintainers
please ACK applying it through the kvm tree, since that is where we see the issue that it addresses?
Avi, Marcelo, maybe you can carry this in kvm/linux-next as a temporary measure so that linux-next builds?

 arch/x86/include/asm/apic.h | 5 +++++
 1 file changed, 5 insertions(+)

arch/x86/include/asm/apic.h +5 -0

index eaff479..aa5b2ee 100644
@@ -417,7 +417,12 @@ extern struct apic *apic;
 	__aligned(sizeof(struct apic *))				\
 	__section(.apicdrivers) = { &sym1, &sym2 }
 
+#ifdef CONFIG_X86_LOCAL_APIC
 extern struct apic *__apicdrivers[], *__apicdrivers_end[];
+#else
+#define __apicdrivers ((struct apic **)NULL)
+#define __apicdrivers_end ((struct apic **)NULL)
+#endif
 
 /*
  * APIC functionality to boot other CPUs - only used on SMP:

Signature · 1 line

MST

Randy Dunlap

Acked-by on v1 1/1

2012-07-01 16:23 UTC #1 lore raw

Quoted 14 lines · Michael S. Tsirkin

On UP i386, when APIC is disabled
# CONFIG_X86_UP_APIC is not set
# CONFIG_PCI_IOAPIC is not set
code looking at apicdrivers never has any effect but it still gets compiled in. In particular, this causes build failures with kvm, but it generally bloats the kernel unnecessarily.
Fix by defining both __apicdrivers and __apicdrivers_end
to be NULL when CONFIG_X86_LOCAL_APIC is unset: I verified that as the result any loop scanning __apicdrivers gets optimized out by the compiler.

Quoted 5 lines

Warning: a .config with apic disabled doesn't seem to boot for me (even without this patch). Still verifying why, meanwhile this patch is compile-tested only.
Signed-off-by Michael S. Tsirkin <[email protected]>

Reported-by Randy Dunlap <[email protected]>

Acked-by Randy Dunlap <[email protected]>

Thanks.

Quoted 28 lines

---
Note: if this patch makes sense, can x86 maintainers
please ACK applying it through the kvm tree, since that is where we see the issue that it addresses?
Avi, Marcelo, maybe you can carry this in kvm/linux-next as a temporary measure so that linux-next builds?
 arch/x86/include/asm/apic.h | 5 +++++
 1 file changed, 5 insertions(+)
arch/x86/include/asm/apic.h +5 -0
index eaff479..aa5b2ee 100644
@@ -417,7 +417,12 @@ extern struct apic *apic;
 	__aligned(sizeof(struct apic *))				\
 	__section(.apicdrivers) = { &sym1, &sym2 }
 
+#ifdef CONFIG_X86_LOCAL_APIC
 extern struct apic *__apicdrivers[], *__apicdrivers_end[];
+#else
+#define __apicdrivers ((struct apic **)NULL)
+#define __apicdrivers_end ((struct apic **)NULL)
+#endif
 
 /*
  * APIC functionality to boot other CPUs - only used on SMP:

Signature · 1 line

~Randy

H. Peter Anvin

reviewer Acked-by on v1 1/1

2012-07-01 17:10 UTC #2 lore raw

In reply to Michael S. Tsirkin

Quoted 18 lines · Michael S. Tsirkin

On UP i386, when APIC is disabled
# CONFIG_X86_UP_APIC is not set
# CONFIG_PCI_IOAPIC is not set
code looking at apicdrivers never has any effect but it still gets compiled in. In particular, this causes build failures with kvm, but it generally bloats the kernel unnecessarily.
Fix by defining both __apicdrivers and __apicdrivers_end
to be NULL when CONFIG_X86_LOCAL_APIC is unset: I verified that as the result any loop scanning __apicdrivers gets optimized out by the compiler.
Warning: a .config with apic disabled doesn't seem to boot for me (even without this patch). Still verifying why, meanwhile this patch is compile-tested only.

Quoted 9 lines

Signed-off-by Michael S. Tsirkin <[email protected]>
---
Note: if this patch makes sense, can x86 maintainers
please ACK applying it through the kvm tree, since that is where we see the issue that it addresses?
Avi, Marcelo, maybe you can carry this in kvm/linux-next as a temporary measure so that linux-next builds?

I'm not happy about that as a workflow, but since you guys have an immediate problem I guess we can do that.

Acked-by H. Peter Anvin <[email protected]>

-hpa

Signature · 2 lines

H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.

Ingo Molnar

2012-07-06 11:13 UTC #3 lore raw

Quoted 29 lines · H. Peter Anvin

Quoted 18 lines · Michael S. Tsirkin
On UP i386, when APIC is disabled
# CONFIG_X86_UP_APIC is not set
# CONFIG_PCI_IOAPIC is not set
code looking at apicdrivers never has any effect but it still gets compiled in. In particular, this causes build failures with kvm, but it generally bloats the kernel unnecessarily.
Fix by defining both __apicdrivers and __apicdrivers_end
to be NULL when CONFIG_X86_LOCAL_APIC is unset: I verified that as the result any loop scanning __apicdrivers gets optimized out by the compiler.
Warning: a .config with apic disabled doesn't seem to boot for me (even without this patch). Still verifying why, meanwhile this patch is compile-tested only.
Quoted 9 lines
Signed-off-by Michael S. Tsirkin <[email protected]>
---
Note: if this patch makes sense, can x86 maintainers
please ACK applying it through the kvm tree, since that is where we see the issue that it addresses?
Avi, Marcelo, maybe you can carry this in kvm/linux-next as a temporary measure so that linux-next builds?

Quoted 2 lines

I'm not happy about that as a workflow, but since you guys have an immediate problem I guess we can do that.

I'm rather unhappy about this workflow - we've got quite a few apic bits in the x86 tree this cycle as well and need extra external interaction, not.

Which KVM tree commit caused this, could someone please give a lkml link or quote it here? It's not referenced in the fix patch either.

Thanks,

Ingo

Marcelo Tosatti

2012-07-06 11:32 UTC #4 lore raw

Quoted 39 lines · Ingo Molnar

Quoted 29 lines · H. Peter Anvin
Quoted 18 lines · Michael S. Tsirkin
On UP i386, when APIC is disabled
# CONFIG_X86_UP_APIC is not set
# CONFIG_PCI_IOAPIC is not set
code looking at apicdrivers never has any effect but it still gets compiled in. In particular, this causes build failures with kvm, but it generally bloats the kernel unnecessarily.
Fix by defining both __apicdrivers and __apicdrivers_end
to be NULL when CONFIG_X86_LOCAL_APIC is unset: I verified that as the result any loop scanning __apicdrivers gets optimized out by the compiler.
Warning: a .config with apic disabled doesn't seem to boot for me (even without this patch). Still verifying why, meanwhile this patch is compile-tested only.
Quoted 9 lines
Signed-off-by Michael S. Tsirkin <[email protected]>
---
Note: if this patch makes sense, can x86 maintainers
please ACK applying it through the kvm tree, since that is where we see the issue that it addresses?
Avi, Marcelo, maybe you can carry this in kvm/linux-next as a temporary measure so that linux-next builds?
Quoted 2 lines
I'm not happy about that as a workflow, but since you guys have an immediate problem I guess we can do that.
I'm rather unhappy about this workflow - we've got quite a few apic bits in the x86 tree this cycle as well and need extra external interaction, not.

Quoted 7 lines

Which KVM tree commit caused this, could someone please give a lkml link or quote it here? It's not referenced in the fix patch either.
Thanks,
Ingo

This tree (kvm.git next):

http://git.kernel.org/?p=virt/kvm/kvm.git;a=shortlog;h=refs/heads/next

Introduced by this commit:

http://git.kernel.org/?p=virt/kvm/kvm.git;a=commit;h=ab9cf4996bb989983e73da894b8dd0239aa2c3c2

Ingo Molnar

2012-07-06 14:12 UTC #5 lore raw

Quoted 48 lines · Marcelo Tosatti

Quoted 39 lines · Ingo Molnar
Quoted 29 lines · H. Peter Anvin
Quoted 18 lines · Michael S. Tsirkin
On UP i386, when APIC is disabled
# CONFIG_X86_UP_APIC is not set
# CONFIG_PCI_IOAPIC is not set
code looking at apicdrivers never has any effect but it still gets compiled in. In particular, this causes build failures with kvm, but it generally bloats the kernel unnecessarily.
Fix by defining both __apicdrivers and __apicdrivers_end
to be NULL when CONFIG_X86_LOCAL_APIC is unset: I verified that as the result any loop scanning __apicdrivers gets optimized out by the compiler.
Warning: a .config with apic disabled doesn't seem to boot for me (even without this patch). Still verifying why, meanwhile this patch is compile-tested only.
Quoted 9 lines
Signed-off-by Michael S. Tsirkin <[email protected]>
---
Note: if this patch makes sense, can x86 maintainers
please ACK applying it through the kvm tree, since that is where we see the issue that it addresses?
Avi, Marcelo, maybe you can carry this in kvm/linux-next as a temporary measure so that linux-next builds?
Quoted 2 lines
I'm not happy about that as a workflow, but since you guys have an immediate problem I guess we can do that.
I'm rather unhappy about this workflow - we've got quite a few apic bits in the x86 tree this cycle as well and need extra external interaction, not.
Quoted 7 lines
Which KVM tree commit caused this, could someone please give a lkml link or quote it here? It's not referenced in the fix patch either.
Thanks,
Ingo

Quoted 7 lines

This tree (kvm.git next):
http://git.kernel.org/?p=virt/kvm/kvm.git;a=shortlog;h=refs/heads/next
Introduced by this commit:
http://git.kernel.org/?p=virt/kvm/kvm.git;a=commit;h=ab9cf4996bb989983e73da894b8dd0239aa2c3c2

This bit:

Review

Quoted 10 lines

(quoted diff) +10 -0

+	if (kvm_para_has_feature(KVM_FEATURE_PV_EOI)) {
+		struct apic **drv;
+
+		for (drv = __apicdrivers; drv < __apicdrivers_end; drv++) {
+			/* Should happen once for each apic */
+			WARN_ON((*drv)->eoi_write == kvm_guest_apic_eoi_write);
+			(*drv)->eoi_write = kvm_guest_apic_eoi_write;
+		}
+	}
+

is rather disgusting I have to say.

WTH is the KVM code meddling with core x86 apic driver data structures directly? At minimum factor this out and create a proper apic.c function which is EXPORT_SYMBOL_GPL() exported or so...

Thanks,

Ingo

Michael S. Tsirkin

author

2012-07-08 20:56 UTC #6 lore raw

Quoted 74 lines · Ingo Molnar

Quoted 48 lines · Marcelo Tosatti
Quoted 39 lines · Ingo Molnar
Quoted 29 lines · H. Peter Anvin
Quoted 18 lines · Michael S. Tsirkin
On UP i386, when APIC is disabled
# CONFIG_X86_UP_APIC is not set
# CONFIG_PCI_IOAPIC is not set
code looking at apicdrivers never has any effect but it still gets compiled in. In particular, this causes build failures with kvm, but it generally bloats the kernel unnecessarily.
Fix by defining both __apicdrivers and __apicdrivers_end
to be NULL when CONFIG_X86_LOCAL_APIC is unset: I verified that as the result any loop scanning __apicdrivers gets optimized out by the compiler.
Warning: a .config with apic disabled doesn't seem to boot for me (even without this patch). Still verifying why, meanwhile this patch is compile-tested only.
Quoted 9 lines
Signed-off-by Michael S. Tsirkin <[email protected]>
---
Note: if this patch makes sense, can x86 maintainers
please ACK applying it through the kvm tree, since that is where we see the issue that it addresses?
Avi, Marcelo, maybe you can carry this in kvm/linux-next as a temporary measure so that linux-next builds?
Quoted 2 lines
I'm not happy about that as a workflow, but since you guys have an immediate problem I guess we can do that.
I'm rather unhappy about this workflow - we've got quite a few apic bits in the x86 tree this cycle as well and need extra external interaction, not.
Quoted 7 lines
Which KVM tree commit caused this, could someone please give a lkml link or quote it here? It's not referenced in the fix patch either.
Thanks,
Ingo
Quoted 7 lines
This tree (kvm.git next):
http://git.kernel.org/?p=virt/kvm/kvm.git;a=shortlog;h=refs/heads/next
Introduced by this commit:
http://git.kernel.org/?p=virt/kvm/kvm.git;a=commit;h=ab9cf4996bb989983e73da894b8dd0239aa2c3c2
This bit:
Review
Quoted 10 lines
(quoted diff) +10 -0
+	if (kvm_para_has_feature(KVM_FEATURE_PV_EOI)) {
+		struct apic **drv;
+
+		for (drv = __apicdrivers; drv < __apicdrivers_end; drv++) {
+			/* Should happen once for each apic */
+			WARN_ON((*drv)->eoi_write == kvm_guest_apic_eoi_write);
+			(*drv)->eoi_write = kvm_guest_apic_eoi_write;
+		}
+	}
+
is rather disgusting I have to say.

Quoted 8 lines

WTH is the KVM code meddling with core x86 apic driver data structures directly? At minimum factor this out and create a proper apic.c function which is EXPORT_SYMBOL_GPL() exported or so...
Thanks,
Ingo

OK, so apic_set_eoi_write()?

Signature · 1 line

MST

Ingo Molnar

2012-07-09 09:24 UTC #7 lore raw

Quoted 80 lines · Michael S. Tsirkin

Quoted 79 lines · Ingo Molnar
Quoted 48 lines · Marcelo Tosatti
Quoted 39 lines · Ingo Molnar
Quoted 29 lines · H. Peter Anvin
Quoted 18 lines · Michael S. Tsirkin
On UP i386, when APIC is disabled
# CONFIG_X86_UP_APIC is not set
# CONFIG_PCI_IOAPIC is not set
code looking at apicdrivers never has any effect but it still gets compiled in. In particular, this causes build failures with kvm, but it generally bloats the kernel unnecessarily.
Fix by defining both __apicdrivers and __apicdrivers_end
to be NULL when CONFIG_X86_LOCAL_APIC is unset: I verified that as the result any loop scanning __apicdrivers gets optimized out by the compiler.
Warning: a .config with apic disabled doesn't seem to boot for me (even without this patch). Still verifying why, meanwhile this patch is compile-tested only.
Quoted 9 lines
Signed-off-by Michael S. Tsirkin <[email protected]>
---
Note: if this patch makes sense, can x86 maintainers
please ACK applying it through the kvm tree, since that is where we see the issue that it addresses?
Avi, Marcelo, maybe you can carry this in kvm/linux-next as a temporary measure so that linux-next builds?
Quoted 2 lines
I'm not happy about that as a workflow, but since you guys have an immediate problem I guess we can do that.
I'm rather unhappy about this workflow - we've got quite a few apic bits in the x86 tree this cycle as well and need extra external interaction, not.
Quoted 7 lines
Which KVM tree commit caused this, could someone please give a lkml link or quote it here? It's not referenced in the fix patch either.
Thanks,
Ingo
Quoted 7 lines
This tree (kvm.git next):
http://git.kernel.org/?p=virt/kvm/kvm.git;a=shortlog;h=refs/heads/next
Introduced by this commit:
http://git.kernel.org/?p=virt/kvm/kvm.git;a=commit;h=ab9cf4996bb989983e73da894b8dd0239aa2c3c2
This bit:
Review
Quoted 10 lines
(quoted diff) +10 -0
+	if (kvm_para_has_feature(KVM_FEATURE_PV_EOI)) {
+		struct apic **drv;
+
+		for (drv = __apicdrivers; drv < __apicdrivers_end; drv++) {
+			/* Should happen once for each apic */
+			WARN_ON((*drv)->eoi_write == kvm_guest_apic_eoi_write);
+			(*drv)->eoi_write = kvm_guest_apic_eoi_write;
+		}
+	}
+
is rather disgusting I have to say.
WTH is the KVM code meddling with core x86 apic driver data structures directly? At minimum factor this out and create a proper apic.c function which is EXPORT_SYMBOL_GPL() exported or so...

Quoted 5 lines

Quoted 3 lines
Thanks,
Ingo
OK, so apic_set_eoi_write()?

Yes, with a changelog comment analyzing the design decisions and locking here - what happens if actual APIC driver use races with this update on SMP, why is it all safe, etc?

Thanks,

Ingo

Marcelo Tosatti

Applied

2012-07-03 18:55 UTC #8 lore raw

In reply to Michael S. Tsirkin

Quoted 18 lines · Michael S. Tsirkin

On UP i386, when APIC is disabled
# CONFIG_X86_UP_APIC is not set
# CONFIG_PCI_IOAPIC is not set
code looking at apicdrivers never has any effect but it still gets compiled in. In particular, this causes build failures with kvm, but it generally bloats the kernel unnecessarily.
Fix by defining both __apicdrivers and __apicdrivers_end
to be NULL when CONFIG_X86_LOCAL_APIC is unset: I verified that as the result any loop scanning __apicdrivers gets optimized out by the compiler.
Warning: a .config with apic disabled doesn't seem to boot for me (even without this patch). Still verifying why, meanwhile this patch is compile-tested only.

Quoted 8 lines

Signed-off-by Michael S. Tsirkin <[email protected]>
---
Note: if this patch makes sense, can x86 maintainers
please ACK applying it through the kvm tree, since that is where we see the issue that it addresses?
Avi, Marcelo, maybe you can carry this in kvm/linux-next as a temporary measure so that linux-next builds?

Applied, thanks.