| Message ID | 20250303171013.3548775-4-tabba@google.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | KVM: Mapping guest_memfd backed memory at the host for software protected VMs | expand |
On Mon, Mar 03, 2025 at 05:10:07PM +0000, Fuad Tabba wrote: > Add support for mmap() and fault() for guest_memfd backed memory > in the host for VMs that support in-place conversion between > shared and private. To that end, this patch adds the ability to > check whether the VM type supports in-place conversion, and only > allows mapping its memory if that's the case. > > Also add the KVM capability KVM_CAP_GMEM_SHARED_MEM, which > indicates that the VM supports shared memory in guest_memfd, or > that the host can create VMs that support shared memory. > Supporting shared memory implies that memory can be mapped when > shared with the host. > > This is controlled by the KVM_GMEM_SHARED_MEM configuration > option. > > Signed-off-by: Fuad Tabba <tabba@google.com> > --- > include/linux/kvm_host.h | 11 ++++ > include/uapi/linux/kvm.h | 1 + > virt/kvm/guest_memfd.c | 105 +++++++++++++++++++++++++++++++++++++++ > virt/kvm/kvm_main.c | 4 ++ > 4 files changed, 121 insertions(+) > > diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h > index 7788e3625f6d..2d025b8ee20e 100644 > --- a/include/linux/kvm_host.h > +++ b/include/linux/kvm_host.h > @@ -728,6 +728,17 @@ static inline bool kvm_arch_has_private_mem(struct kvm *kvm) > } > #endif > > +/* > + * Arch code must define kvm_arch_gmem_supports_shared_mem if support for > + * private memory is enabled and it supports in-place shared/private conversion. > + */ > +#if !defined(kvm_arch_gmem_supports_shared_mem) && !IS_ENABLED(CONFIG_KVM_PRIVATE_MEM) Hm. Do we expect any caller for !CONFIG_KVM_PRIVATE_MEM? > +static inline bool kvm_arch_gmem_supports_shared_mem(struct kvm *kvm) > +{ > + return false; > +} > +#endif > + > #ifndef kvm_arch_has_readonly_mem > static inline bool kvm_arch_has_readonly_mem(struct kvm *kvm) > { ... > diff --git a/virt/kvm/guest_memfd.c b/virt/kvm/guest_memfd.c > index b2aa6bf24d3a..4291956b51ae 100644 > --- a/virt/kvm/guest_memfd.c > +++ b/virt/kvm/guest_memfd.c > @@ -312,7 +312,112 @@ static pgoff_t kvm_gmem_get_index(struct kvm_memory_slot *slot, gfn_t gfn) > return gfn - slot->base_gfn + slot->gmem.pgoff; > } > > +#ifdef CONFIG_KVM_GMEM_SHARED_MEM > +static bool kvm_gmem_offset_is_shared(struct file *file, pgoff_t index) > +{ > + struct kvm_gmem *gmem = file->private_data; > + > + /* For now, VMs that support shared memory share all their memory. */ > + return kvm_arch_gmem_supports_shared_mem(gmem->kvm); > +} > + > +static vm_fault_t kvm_gmem_fault(struct vm_fault *vmf) > +{ > + struct inode *inode = file_inode(vmf->vma->vm_file); > + struct folio *folio; > + vm_fault_t ret = VM_FAULT_LOCKED; > + > + filemap_invalidate_lock_shared(inode->i_mapping); > + > + folio = kvm_gmem_get_folio(inode, vmf->pgoff); > + if (IS_ERR(folio)) { > + switch (PTR_ERR(folio)) { > + case -EAGAIN: > + ret = VM_FAULT_RETRY; > + break; > + case -ENOMEM: > + ret = VM_FAULT_OOM; > + break; > + default: > + ret = VM_FAULT_SIGBUS; > + break; > + } > + goto out_filemap; > + } > + > + if (folio_test_hwpoison(folio)) { > + ret = VM_FAULT_HWPOISON; > + goto out_folio; > + } > + > + /* Must be called with folio lock held, i.e., after kvm_gmem_get_folio() */ If this is a requirement, it would be cleaner to rename the function and pass down the folio and check the lock state inside.
Hi Kirill, On Tue, 4 Mar 2025 at 08:58, Kirill A. Shutemov <kirill.shutemov@linux.intel.com> wrote: > > On Mon, Mar 03, 2025 at 05:10:07PM +0000, Fuad Tabba wrote: > > Add support for mmap() and fault() for guest_memfd backed memory > > in the host for VMs that support in-place conversion between > > shared and private. To that end, this patch adds the ability to > > check whether the VM type supports in-place conversion, and only > > allows mapping its memory if that's the case. > > > > Also add the KVM capability KVM_CAP_GMEM_SHARED_MEM, which > > indicates that the VM supports shared memory in guest_memfd, or > > that the host can create VMs that support shared memory. > > Supporting shared memory implies that memory can be mapped when > > shared with the host. > > > > This is controlled by the KVM_GMEM_SHARED_MEM configuration > > option. > > > > Signed-off-by: Fuad Tabba <tabba@google.com> > > --- > > include/linux/kvm_host.h | 11 ++++ > > include/uapi/linux/kvm.h | 1 + > > virt/kvm/guest_memfd.c | 105 +++++++++++++++++++++++++++++++++++++++ > > virt/kvm/kvm_main.c | 4 ++ > > 4 files changed, 121 insertions(+) > > > > diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h > > index 7788e3625f6d..2d025b8ee20e 100644 > > --- a/include/linux/kvm_host.h > > +++ b/include/linux/kvm_host.h > > @@ -728,6 +728,17 @@ static inline bool kvm_arch_has_private_mem(struct kvm *kvm) > > } > > #endif > > > > +/* > > + * Arch code must define kvm_arch_gmem_supports_shared_mem if support for > > + * private memory is enabled and it supports in-place shared/private conversion. > > + */ > > +#if !defined(kvm_arch_gmem_supports_shared_mem) && !IS_ENABLED(CONFIG_KVM_PRIVATE_MEM) > > Hm. Do we expect any caller for !CONFIG_KVM_PRIVATE_MEM? > > > +static inline bool kvm_arch_gmem_supports_shared_mem(struct kvm *kvm) > > +{ > > + return false; > > +} > > +#endif > > + > > #ifndef kvm_arch_has_readonly_mem > > static inline bool kvm_arch_has_readonly_mem(struct kvm *kvm) > > { > > ... > > > diff --git a/virt/kvm/guest_memfd.c b/virt/kvm/guest_memfd.c > > index b2aa6bf24d3a..4291956b51ae 100644 > > --- a/virt/kvm/guest_memfd.c > > +++ b/virt/kvm/guest_memfd.c > > @@ -312,7 +312,112 @@ static pgoff_t kvm_gmem_get_index(struct kvm_memory_slot *slot, gfn_t gfn) > > return gfn - slot->base_gfn + slot->gmem.pgoff; > > } > > > > +#ifdef CONFIG_KVM_GMEM_SHARED_MEM > > +static bool kvm_gmem_offset_is_shared(struct file *file, pgoff_t index) > > +{ > > + struct kvm_gmem *gmem = file->private_data; > > + > > + /* For now, VMs that support shared memory share all their memory. */ > > + return kvm_arch_gmem_supports_shared_mem(gmem->kvm); > > +} > > + > > +static vm_fault_t kvm_gmem_fault(struct vm_fault *vmf) > > +{ > > + struct inode *inode = file_inode(vmf->vma->vm_file); > > + struct folio *folio; > > + vm_fault_t ret = VM_FAULT_LOCKED; > > + > > + filemap_invalidate_lock_shared(inode->i_mapping); > > + > > + folio = kvm_gmem_get_folio(inode, vmf->pgoff); > > + if (IS_ERR(folio)) { > > + switch (PTR_ERR(folio)) { > > + case -EAGAIN: > > + ret = VM_FAULT_RETRY; > > + break; > > + case -ENOMEM: > > + ret = VM_FAULT_OOM; > > + break; > > + default: > > + ret = VM_FAULT_SIGBUS; > > + break; > > + } > > + goto out_filemap; > > + } > > + > > + if (folio_test_hwpoison(folio)) { > > + ret = VM_FAULT_HWPOISON; > > + goto out_folio; > > + } > > + > > + /* Must be called with folio lock held, i.e., after kvm_gmem_get_folio() */ > > If this is a requirement, it would be cleaner to rename the function and > pass down the folio and check the lock state inside. Will do. Thanks, /fuad > -- > Kiryl Shutsemau / Kirill A. Shutemov
On Tue, Mar 04, 2025 at 09:27:06AM +0000, Fuad Tabba wrote: > Hi Kirill, > > On Tue, 4 Mar 2025 at 08:58, Kirill A. Shutemov > <kirill.shutemov@linux.intel.com> wrote: > > > > On Mon, Mar 03, 2025 at 05:10:07PM +0000, Fuad Tabba wrote: > > > Add support for mmap() and fault() for guest_memfd backed memory > > > in the host for VMs that support in-place conversion between > > > shared and private. To that end, this patch adds the ability to > > > check whether the VM type supports in-place conversion, and only > > > allows mapping its memory if that's the case. > > > > > > Also add the KVM capability KVM_CAP_GMEM_SHARED_MEM, which > > > indicates that the VM supports shared memory in guest_memfd, or > > > that the host can create VMs that support shared memory. > > > Supporting shared memory implies that memory can be mapped when > > > shared with the host. > > > > > > This is controlled by the KVM_GMEM_SHARED_MEM configuration > > > option. > > > > > > Signed-off-by: Fuad Tabba <tabba@google.com> > > > --- > > > include/linux/kvm_host.h | 11 ++++ > > > include/uapi/linux/kvm.h | 1 + > > > virt/kvm/guest_memfd.c | 105 +++++++++++++++++++++++++++++++++++++++ > > > virt/kvm/kvm_main.c | 4 ++ > > > 4 files changed, 121 insertions(+) > > > > > > diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h > > > index 7788e3625f6d..2d025b8ee20e 100644 > > > --- a/include/linux/kvm_host.h > > > +++ b/include/linux/kvm_host.h > > > @@ -728,6 +728,17 @@ static inline bool kvm_arch_has_private_mem(struct kvm *kvm) > > > } > > > #endif > > > > > > +/* > > > + * Arch code must define kvm_arch_gmem_supports_shared_mem if support for > > > + * private memory is enabled and it supports in-place shared/private conversion. > > > + */ > > > +#if !defined(kvm_arch_gmem_supports_shared_mem) && !IS_ENABLED(CONFIG_KVM_PRIVATE_MEM) > > > > Hm. Do we expect any caller for !CONFIG_KVM_PRIVATE_MEM? I think you missed this.
On Tue, 4 Mar 2025 at 09:46, Kirill A. Shutemov <kirill.shutemov@linux.intel.com> wrote: > > On Tue, Mar 04, 2025 at 09:27:06AM +0000, Fuad Tabba wrote: > > Hi Kirill, > > > > On Tue, 4 Mar 2025 at 08:58, Kirill A. Shutemov > > <kirill.shutemov@linux.intel.com> wrote: > > > > > > On Mon, Mar 03, 2025 at 05:10:07PM +0000, Fuad Tabba wrote: > > > > Add support for mmap() and fault() for guest_memfd backed memory > > > > in the host for VMs that support in-place conversion between > > > > shared and private. To that end, this patch adds the ability to > > > > check whether the VM type supports in-place conversion, and only > > > > allows mapping its memory if that's the case. > > > > > > > > Also add the KVM capability KVM_CAP_GMEM_SHARED_MEM, which > > > > indicates that the VM supports shared memory in guest_memfd, or > > > > that the host can create VMs that support shared memory. > > > > Supporting shared memory implies that memory can be mapped when > > > > shared with the host. > > > > > > > > This is controlled by the KVM_GMEM_SHARED_MEM configuration > > > > option. > > > > > > > > Signed-off-by: Fuad Tabba <tabba@google.com> > > > > --- > > > > include/linux/kvm_host.h | 11 ++++ > > > > include/uapi/linux/kvm.h | 1 + > > > > virt/kvm/guest_memfd.c | 105 +++++++++++++++++++++++++++++++++++++++ > > > > virt/kvm/kvm_main.c | 4 ++ > > > > 4 files changed, 121 insertions(+) > > > > > > > > diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h > > > > index 7788e3625f6d..2d025b8ee20e 100644 > > > > --- a/include/linux/kvm_host.h > > > > +++ b/include/linux/kvm_host.h > > > > @@ -728,6 +728,17 @@ static inline bool kvm_arch_has_private_mem(struct kvm *kvm) > > > > } > > > > #endif > > > > > > > > +/* > > > > + * Arch code must define kvm_arch_gmem_supports_shared_mem if support for > > > > + * private memory is enabled and it supports in-place shared/private conversion. > > > > + */ > > > > +#if !defined(kvm_arch_gmem_supports_shared_mem) && !IS_ENABLED(CONFIG_KVM_PRIVATE_MEM) > > > > > > Hm. Do we expect any caller for !CONFIG_KVM_PRIVATE_MEM? > > I think you missed this. Yes I did :) You're right (reading between the lines of where you're getting at that is). This should be +#if !defined(kvm_arch_gmem_supports_shared_mem) && !IS_ENABLED(CONFIG_KVM_GMEM_SHARED_MEM) I'll fix (if you agree with my understanding of your comment). Cheers, /fuad > -- > Kiryl Shutsemau / Kirill A. Shutemov
Hi Ackerley, On Thu, 6 Mar 2025 at 00:02, Ackerley Tng <ackerleytng@google.com> wrote: > > Fuad Tabba <tabba@google.com> writes: > > > <snip> > > > > +static vm_fault_t kvm_gmem_fault(struct vm_fault *vmf) > > +{ > > + struct inode *inode = file_inode(vmf->vma->vm_file); > > + struct folio *folio; > > + vm_fault_t ret = VM_FAULT_LOCKED; > > + > > + filemap_invalidate_lock_shared(inode->i_mapping); > > + > > + folio = kvm_gmem_get_folio(inode, vmf->pgoff); > > + if (IS_ERR(folio)) { > > + switch (PTR_ERR(folio)) { > > + case -EAGAIN: > > + ret = VM_FAULT_RETRY; > > + break; > > + case -ENOMEM: > > + ret = VM_FAULT_OOM; > > + break; > > + default: > > + ret = VM_FAULT_SIGBUS; > > + break; > > + } > > + goto out_filemap; > > + } > > + > > + if (folio_test_hwpoison(folio)) { > > + ret = VM_FAULT_HWPOISON; > > + goto out_folio; > > + } > > + > > + /* Must be called with folio lock held, i.e., after kvm_gmem_get_folio() */ > > + if (!kvm_gmem_offset_is_shared(vmf->vma->vm_file, vmf->pgoff)) { > > + ret = VM_FAULT_SIGBUS; > > + goto out_folio; > > + } > > + > > + /* > > + * Only private folios are marked as "guestmem" so far, and we never > > + * expect private folios at this point. > > + */ > > I think this is not quite accurate. > > Based on my understanding and kvm_gmem_handle_folio_put() in this other > patch [1], only pages *in transition* from shared to private state are > marked "guestmem", although it is true that no private folios or folios > marked guestmem are expected here. Technically, pages in transition are private as far as the host is concerned. This doesn't say that _all_ private pages are marked as guestmem. It says that only private pages are marked as guestmem. It could be private and _not_ be marked as guestmem :) I probably should rephrase something along the lines of, "no shared folios would be marked as guestmem". How does that sound? Thanks, /fuad > > + if (WARN_ON_ONCE(folio_test_guestmem(folio))) { > > + ret = VM_FAULT_SIGBUS; > > + goto out_folio; > > + } > > + > > + /* No support for huge pages. */ > > + if (WARN_ON_ONCE(folio_test_large(folio))) { > > + ret = VM_FAULT_SIGBUS; > > + goto out_folio; > > + } > > + > > + if (!folio_test_uptodate(folio)) { > > + clear_highpage(folio_page(folio, 0)); > > + kvm_gmem_mark_prepared(folio); > > + } > > + > > + vmf->page = folio_file_page(folio, vmf->pgoff); > > + > > +out_folio: > > + if (ret != VM_FAULT_LOCKED) { > > + folio_unlock(folio); > > + folio_put(folio); > > + } > > + > > +out_filemap: > > + filemap_invalidate_unlock_shared(inode->i_mapping); > > + > > + return ret; > > +} > > > > <snip> > > [1] https://lore.kernel.org/all/20250117163001.2326672-7-tabba@google.com/
On Mon, Mar 3, 2025 at 9:10 AM Fuad Tabba <tabba@google.com> wrote: > +static vm_fault_t kvm_gmem_fault(struct vm_fault *vmf) > +{ > + struct inode *inode = file_inode(vmf->vma->vm_file); > + struct folio *folio; > + vm_fault_t ret = VM_FAULT_LOCKED; > + > + filemap_invalidate_lock_shared(inode->i_mapping); > + > + folio = kvm_gmem_get_folio(inode, vmf->pgoff); > + if (IS_ERR(folio)) { > + switch (PTR_ERR(folio)) { > + case -EAGAIN: > + ret = VM_FAULT_RETRY; > + break; > + case -ENOMEM: > + ret = VM_FAULT_OOM; > + break; > + default: > + ret = VM_FAULT_SIGBUS; > + break; Tiny nit-pick: This smells almost like an open-coded vmf_error(). For the non-EAGAIN case, can we just call vmf_error()? > + } > + goto out_filemap; > + } > + > + if (folio_test_hwpoison(folio)) { > + ret = VM_FAULT_HWPOISON; > + goto out_folio; > + } > + > + /* Must be called with folio lock held, i.e., after kvm_gmem_get_folio() */ > + if (!kvm_gmem_offset_is_shared(vmf->vma->vm_file, vmf->pgoff)) { > + ret = VM_FAULT_SIGBUS; > + goto out_folio; > + }
Hi James, On Fri, 7 Mar 2025 at 01:53, James Houghton <jthoughton@google.com> wrote: > > On Mon, Mar 3, 2025 at 9:10 AM Fuad Tabba <tabba@google.com> wrote: > > +static vm_fault_t kvm_gmem_fault(struct vm_fault *vmf) > > +{ > > + struct inode *inode = file_inode(vmf->vma->vm_file); > > + struct folio *folio; > > + vm_fault_t ret = VM_FAULT_LOCKED; > > + > > + filemap_invalidate_lock_shared(inode->i_mapping); > > + > > + folio = kvm_gmem_get_folio(inode, vmf->pgoff); > > + if (IS_ERR(folio)) { > > + switch (PTR_ERR(folio)) { > > + case -EAGAIN: > > + ret = VM_FAULT_RETRY; > > + break; > > + case -ENOMEM: > > + ret = VM_FAULT_OOM; > > + break; > > + default: > > + ret = VM_FAULT_SIGBUS; > > + break; > > Tiny nit-pick: This smells almost like an open-coded vmf_error(). For > the non-EAGAIN case, can we just call vmf_error()? I wasn't aware of that. I will fix this on the respin, thanks for pointing it out. That said, any idea why vmf_error() doesn't convert -EAGAIN to VM_FAULT_RETRY? Cheers, /fuad > > + } > > + goto out_filemap; > > + } > > + > > + if (folio_test_hwpoison(folio)) { > > + ret = VM_FAULT_HWPOISON; > > + goto out_folio; > > + } > > + > > + /* Must be called with folio lock held, i.e., after kvm_gmem_get_folio() */ > > + if (!kvm_gmem_offset_is_shared(vmf->vma->vm_file, vmf->pgoff)) { > > + ret = VM_FAULT_SIGBUS; > > + goto out_folio; > > + }
Hi Ackerley, On Thu, 6 Mar 2025 at 22:46, Ackerley Tng <ackerleytng@google.com> wrote: > > Fuad Tabba <tabba@google.com> writes: > > > Add support for mmap() and fault() for guest_memfd backed memory > > in the host for VMs that support in-place conversion between > > shared and private. To that end, this patch adds the ability to > > check whether the VM type supports in-place conversion, and only > > allows mapping its memory if that's the case. > > > > Also add the KVM capability KVM_CAP_GMEM_SHARED_MEM, which > > indicates that the VM supports shared memory in guest_memfd, or > > that the host can create VMs that support shared memory. > > Supporting shared memory implies that memory can be mapped when > > shared with the host. > > > > This is controlled by the KVM_GMEM_SHARED_MEM configuration > > option. > > > > Signed-off-by: Fuad Tabba <tabba@google.com> > > --- > > include/linux/kvm_host.h | 11 ++++ > > include/uapi/linux/kvm.h | 1 + > > virt/kvm/guest_memfd.c | 105 +++++++++++++++++++++++++++++++++++++++ > > virt/kvm/kvm_main.c | 4 ++ > > 4 files changed, 121 insertions(+) > > > > diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h > > index 7788e3625f6d..2d025b8ee20e 100644 > > --- a/include/linux/kvm_host.h > > +++ b/include/linux/kvm_host.h > > @@ -728,6 +728,17 @@ static inline bool kvm_arch_has_private_mem(struct kvm *kvm) > > } > > #endif > > > > +/* > > + * Arch code must define kvm_arch_gmem_supports_shared_mem if support for > > + * private memory is enabled and it supports in-place shared/private conversion. > > + */ > > +#if !defined(kvm_arch_gmem_supports_shared_mem) && !IS_ENABLED(CONFIG_KVM_PRIVATE_MEM) > > Is this a copypasta error? I'm wondering if this should be > !IS_ENABLED(CONFIG_KVM_GMEM_SHARED_MEM). Yes. Kirill had pointed that out as well. I will fix it. > Also, would you consider defining a __weak function to be overridden by > different architectures, or would weak symbols not be inline-able? I have no strong opinion, but I think that it should follow the same pattern as kvm_arch_has_private_mem(). Cheers, /fuad > > +static inline bool kvm_arch_gmem_supports_shared_mem(struct kvm *kvm) > > +{ > > + return false; > > +} > > +#endif > > + > > > > <snip>
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h index 7788e3625f6d..2d025b8ee20e 100644 --- a/include/linux/kvm_host.h +++ b/include/linux/kvm_host.h @@ -728,6 +728,17 @@ static inline bool kvm_arch_has_private_mem(struct kvm *kvm) } #endif +/* + * Arch code must define kvm_arch_gmem_supports_shared_mem if support for + * private memory is enabled and it supports in-place shared/private conversion. + */ +#if !defined(kvm_arch_gmem_supports_shared_mem) && !IS_ENABLED(CONFIG_KVM_PRIVATE_MEM) +static inline bool kvm_arch_gmem_supports_shared_mem(struct kvm *kvm) +{ + return false; +} +#endif + #ifndef kvm_arch_has_readonly_mem static inline bool kvm_arch_has_readonly_mem(struct kvm *kvm) { diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h index 45e6d8fca9b9..117937a895da 100644 --- a/include/uapi/linux/kvm.h +++ b/include/uapi/linux/kvm.h @@ -929,6 +929,7 @@ struct kvm_enable_cap { #define KVM_CAP_PRE_FAULT_MEMORY 236 #define KVM_CAP_X86_APIC_BUS_CYCLES_NS 237 #define KVM_CAP_X86_GUEST_MODE 238 +#define KVM_CAP_GMEM_SHARED_MEM 239 struct kvm_irq_routing_irqchip { __u32 irqchip; diff --git a/virt/kvm/guest_memfd.c b/virt/kvm/guest_memfd.c index b2aa6bf24d3a..4291956b51ae 100644 --- a/virt/kvm/guest_memfd.c +++ b/virt/kvm/guest_memfd.c @@ -312,7 +312,112 @@ static pgoff_t kvm_gmem_get_index(struct kvm_memory_slot *slot, gfn_t gfn) return gfn - slot->base_gfn + slot->gmem.pgoff; } +#ifdef CONFIG_KVM_GMEM_SHARED_MEM +static bool kvm_gmem_offset_is_shared(struct file *file, pgoff_t index) +{ + struct kvm_gmem *gmem = file->private_data; + + /* For now, VMs that support shared memory share all their memory. */ + return kvm_arch_gmem_supports_shared_mem(gmem->kvm); +} + +static vm_fault_t kvm_gmem_fault(struct vm_fault *vmf) +{ + struct inode *inode = file_inode(vmf->vma->vm_file); + struct folio *folio; + vm_fault_t ret = VM_FAULT_LOCKED; + + filemap_invalidate_lock_shared(inode->i_mapping); + + folio = kvm_gmem_get_folio(inode, vmf->pgoff); + if (IS_ERR(folio)) { + switch (PTR_ERR(folio)) { + case -EAGAIN: + ret = VM_FAULT_RETRY; + break; + case -ENOMEM: + ret = VM_FAULT_OOM; + break; + default: + ret = VM_FAULT_SIGBUS; + break; + } + goto out_filemap; + } + + if (folio_test_hwpoison(folio)) { + ret = VM_FAULT_HWPOISON; + goto out_folio; + } + + /* Must be called with folio lock held, i.e., after kvm_gmem_get_folio() */ + if (!kvm_gmem_offset_is_shared(vmf->vma->vm_file, vmf->pgoff)) { + ret = VM_FAULT_SIGBUS; + goto out_folio; + } + + /* + * Only private folios are marked as "guestmem" so far, and we never + * expect private folios at this point. + */ + if (WARN_ON_ONCE(folio_test_guestmem(folio))) { + ret = VM_FAULT_SIGBUS; + goto out_folio; + } + + /* No support for huge pages. */ + if (WARN_ON_ONCE(folio_test_large(folio))) { + ret = VM_FAULT_SIGBUS; + goto out_folio; + } + + if (!folio_test_uptodate(folio)) { + clear_highpage(folio_page(folio, 0)); + kvm_gmem_mark_prepared(folio); + } + + vmf->page = folio_file_page(folio, vmf->pgoff); + +out_folio: + if (ret != VM_FAULT_LOCKED) { + folio_unlock(folio); + folio_put(folio); + } + +out_filemap: + filemap_invalidate_unlock_shared(inode->i_mapping); + + return ret; +} + +static const struct vm_operations_struct kvm_gmem_vm_ops = { + .fault = kvm_gmem_fault, +}; + +static int kvm_gmem_mmap(struct file *file, struct vm_area_struct *vma) +{ + struct kvm_gmem *gmem = file->private_data; + + if (!kvm_arch_gmem_supports_shared_mem(gmem->kvm)) + return -ENODEV; + + if ((vma->vm_flags & (VM_SHARED | VM_MAYSHARE)) != + (VM_SHARED | VM_MAYSHARE)) { + return -EINVAL; + } + + file_accessed(file); + vm_flags_set(vma, VM_DONTDUMP); + vma->vm_ops = &kvm_gmem_vm_ops; + + return 0; +} +#else +#define kvm_gmem_mmap NULL +#endif /* CONFIG_KVM_GMEM_SHARED_MEM */ + static struct file_operations kvm_gmem_fops = { + .mmap = kvm_gmem_mmap, .open = generic_file_open, .release = kvm_gmem_release, .fallocate = kvm_gmem_fallocate, diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c index ba0327e2d0d3..38f0f402ea46 100644 --- a/virt/kvm/kvm_main.c +++ b/virt/kvm/kvm_main.c @@ -4830,6 +4830,10 @@ static int kvm_vm_ioctl_check_extension_generic(struct kvm *kvm, long arg) #ifdef CONFIG_KVM_PRIVATE_MEM case KVM_CAP_GUEST_MEMFD: return !kvm || kvm_arch_has_private_mem(kvm); +#endif +#ifdef CONFIG_KVM_GMEM_SHARED_MEM + case KVM_CAP_GMEM_SHARED_MEM: + return !kvm || kvm_arch_gmem_supports_shared_mem(kvm); #endif default: break;
Add support for mmap() and fault() for guest_memfd backed memory in the host for VMs that support in-place conversion between shared and private. To that end, this patch adds the ability to check whether the VM type supports in-place conversion, and only allows mapping its memory if that's the case. Also add the KVM capability KVM_CAP_GMEM_SHARED_MEM, which indicates that the VM supports shared memory in guest_memfd, or that the host can create VMs that support shared memory. Supporting shared memory implies that memory can be mapped when shared with the host. This is controlled by the KVM_GMEM_SHARED_MEM configuration option. Signed-off-by: Fuad Tabba <tabba@google.com> --- include/linux/kvm_host.h | 11 ++++ include/uapi/linux/kvm.h | 1 + virt/kvm/guest_memfd.c | 105 +++++++++++++++++++++++++++++++++++++++ virt/kvm/kvm_main.c | 4 ++ 4 files changed, 121 insertions(+)