[PATCH] kvm-vfio: do not use module_init

[PATCH] kvm-vfio: do not use module_init

[Paolo Bonzini]

/me got confused between the kernel and QEMU.  In the kernel, you can
only have one module_init function, and it will prevent unloading the
module unless you also have the corresponding module_exit function.

So, commit 80ce1639727e (KVM: VFIO: register kvm_device_ops dynamically,
2014-09-02) broke unloading of the kvm module, by adding a module_init
function and no module_exit.

Repair it by making kvm_vfio_ops_init weak, and checking it in
kvm_init.

Cc: Will Deacon <will.deacon@arm.com>
Cc: Gleb Natapov <gleb@kernel.org>
Cc: Alex Williamson <Alex.Williamson@redhat.com>
Fixes: 80ce1639727e9d38729c34f162378508c307ca25
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 virt/kvm/kvm_main.c | 6 ++++++
 virt/kvm/vfio.c     | 4 ++--
 virt/kvm/vfio.h     | 6 ++++++
 3 files changed, 14 insertions(+), 2 deletions(-)
 create mode 100644 virt/kvm/vfio.h

diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index db57363cc287..69f90fb7899f 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -57,6 +57,7 @@
 
 #include "coalesced_mmio.h"
 #include "async_pf.h"
+#include "vfio.h"
 
 #define CREATE_TRACE_POINTS
 #include <trace/events/kvm.h>
@@ -3226,6 +3227,11 @@ int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
 		goto out_undebugfs;
 	}
 
+	if (kvm_vfio_ops_init) {
+		r = kvm_vfio_ops_init();
+		WARN_ON(r);
+	}
+
 	return 0;
 
 out_undebugfs:
diff --git a/virt/kvm/vfio.c b/virt/kvm/vfio.c
index bb11b36ee8a2..8f7418186d7d 100644
--- a/virt/kvm/vfio.c
+++ b/virt/kvm/vfio.c
@@ -18,6 +18,7 @@
 #include <linux/slab.h>
 #include <linux/uaccess.h>
 #include <linux/vfio.h>
+#include "vfio.h"
 
 struct kvm_vfio_group {
 	struct list_head node;
@@ -278,8 +279,7 @@ static int kvm_vfio_create(struct kvm_device *dev, u32 type)
 	return 0;
 }
 
-static int __init kvm_vfio_ops_init(void)
+int __weak kvm_vfio_ops_init(void)
 {
 	return kvm_register_device_ops(&kvm_vfio_ops, KVM_DEV_TYPE_VFIO);
 }
-module_init(kvm_vfio_ops_init);
diff --git a/virt/kvm/vfio.h b/virt/kvm/vfio.h
new file mode 100644
index 000000000000..f70b384d0ef8
--- /dev/null
+++ b/virt/kvm/vfio.h
@@ -0,0 +1,6 @@
+#ifndef __KVM_VFIO_H
+#define __KVM_VFIO_H
+
+int __weak kvm_vfio_ops_init(void);
+
+#endif
-- 
1.8.3.1

Re: [PATCH] kvm-vfio: do not use module_init

[Will Deacon]

On Wed, Sep 24, 2014 at 12:29:09PM +0100, Paolo Bonzini wrote:
> /me got confused between the kernel and QEMU.  In the kernel, you can
> only have one module_init function, and it will prevent unloading the
> module unless you also have the corresponding module_exit function.

Happy for you to take the blame, but I think this one's my fault!

> So, commit 80ce1639727e (KVM: VFIO: register kvm_device_ops dynamically,
> 2014-09-02) broke unloading of the kvm module, by adding a module_init
> function and no module_exit.

I forget kvm builds as a module for other architectures (ie. not arm/arm64).

> Repair it by making kvm_vfio_ops_init weak, and checking it in
> kvm_init.

Hehe, if only there was a kconfig option for kvm-vfio.c...

Will

Re: [PATCH] kvm-vfio: do not use module_init

[Paolo Bonzini]

Il 24/09/2014 13:44, Will Deacon ha scritto:
> On Wed, Sep 24, 2014 at 12:29:09PM +0100, Paolo Bonzini wrote:
>> /me got confused between the kernel and QEMU.  In the kernel, you can
>> only have one module_init function, and it will prevent unloading the
>> module unless you also have the corresponding module_exit function.
> 
> Happy for you to take the blame, but I think this one's my fault!

That's why you were CCed! ;)

>> So, commit 80ce1639727e (KVM: VFIO: register kvm_device_ops dynamically,
>> 2014-09-02) broke unloading of the kvm module, by adding a module_init
>> function and no module_exit.
> 
> I forget kvm builds as a module for other architectures (ie. not arm/arm64).
> 
>> Repair it by making kvm_vfio_ops_init weak, and checking it in
>> kvm_init.
> 
> Hehe, if only there was a kconfig option for kvm-vfio.c...

Yeah, I was tempted to put it back.  What do you think?

Paolo

Re: [PATCH] kvm-vfio: do not use module_init

[Will Deacon]

On Wed, Sep 24, 2014 at 12:45:53PM +0100, Paolo Bonzini wrote:
> Il 24/09/2014 13:44, Will Deacon ha scritto:
> > On Wed, Sep 24, 2014 at 12:29:09PM +0100, Paolo Bonzini wrote:
> >> /me got confused between the kernel and QEMU.  In the kernel, you can
> >> only have one module_init function, and it will prevent unloading the
> >> module unless you also have the corresponding module_exit function.
> > 
> > Happy for you to take the blame, but I think this one's my fault!
> 
> That's why you were CCed! ;)
> 
> >> So, commit 80ce1639727e (KVM: VFIO: register kvm_device_ops dynamically,
> >> 2014-09-02) broke unloading of the kvm module, by adding a module_init
> >> function and no module_exit.
> > 
> > I forget kvm builds as a module for other architectures (ie. not arm/arm64).
> > 
> >> Repair it by making kvm_vfio_ops_init weak, and checking it in
> >> kvm_init.
> > 
> > Hehe, if only there was a kconfig option for kvm-vfio.c...
> 
> Yeah, I was tempted to put it back.  What do you think?

I think it's nicer than the __weak symbol and I don't see a downside to
having the option (the reason for removing it was lack of users, which
lasted a couple of hours).

Will