On Wed, May 28, 2025 at 01:42:25PM -0300, Jason Gunthorpe wrote:
> On Wed, May 28, 2025 at 05:47:19PM +0530, Aneesh Kumar K.V wrote:
>
> > +#if IS_ENABLED(CONFIG_KVM)
> > +#include <linux/kvm_host.h>
> > +
> > +static void viommu_get_kvm_safe(struct iommufd_viommu *viommu, struct kvm *kvm)
> > +{
> > + void (*put_fn)(struct kvm *kvm);
> > + bool (*get_fn)(struct kvm *kvm);
> > + bool ret;
> > +
> > + if (!kvm)
> > + return;
> > +
> > + put_fn = symbol_get(kvm_put_kvm);
> > + if (WARN_ON(!put_fn))
> > + return;
> > +
> > + get_fn = symbol_get(kvm_get_kvm_safe);
> > + if (WARN_ON(!get_fn)) {
> > + symbol_put(kvm_put_kvm);
> > + return;
> > + }
> > +
> > + ret = get_fn(kvm);
> > + symbol_put(kvm_get_kvm_safe);
> > + if (!ret) {
> > + symbol_put(kvm_put_kvm);
> > + return;
> > + }
> > +
> > + viommu->put_kvm = put_fn;
> > + viommu->kvm = kvm;
> > +}
>
> Shameer was working on something like this too
>
> I would probably split just the viommu kvm stuff into one patch so you
> two can share it.
>
> > @@ -68,10 +121,32 @@ int iommufd_viommu_alloc_ioctl(struct iommufd_ucmd *ucmd)
> > */
> > viommu->iommu_dev = __iommu_get_iommu_dev(idev->dev);
> >
> > + /* get the kvm details if specified. */
> > + if (cmd->kvm_vm_fd) {
>
> Pedantically a 0 fd is still valid, you should add a flag to indicate
> if the KVM is being supplied.
Did I miss something? Shameer's patch passed in struct kvm* through
iommufd_device_bind() then to viommu, and has your Reviewed-by. I'm a
little confused...
https://lore.kernel.org/all/20250319232848.GD126678@xxxxxxxx/
Thanks,
Yilun
