[PATCH] x86/kvm: restrict kvm user region memory size

[PATCH] x86/kvm: restrict kvm user region memory size

[Sabyrzhan Tasbolatov]

syzbot found WARNING in memslot_rmap_alloc[1] when
struct kvm_userspace_memory_region .memory_size is bigger than
0x40000000000, which is 4GB, e.g. KMALLOC_MAX_SIZE * 100 * PAGE_SIZE.

Here is the PoC to trigger the warning:

    struct kvm_userspace_memory_region mem = {
        .slot = 0,
        .guest_phys_addr = 0,
        /* + 0x100 extra to trigger kmalloc WARNING */
        .memory_size = 0x40000000000 + 0x100,
        .userspace_addr = 0,
    };

    ioctl(kvm_fd, KVM_SET_USER_MEMORY_REGION, &mem);

I couldn't find any relevant max constant to restrict unsigned long npages.
There might be another solution with chunking big portions of pages, but
there is already KVM_MAX_HUGEPAGE_LEVEL, though warning happens in
memslot_rmap_alloc() when level = 1, base_gfn = 0, e.g.
on the very first KVM_NR_PAGE_SIZES iteration.

This is, seems, valid for early Linux versions as well. Can't tell which is
exactly can be considered for git bisect.
Here is Commit d89cc617b954af ("KVM: Push rmap into kvm_arch_memory_slot")
for example, Linux 3.7.

[1]
Call Trace:
 kvmalloc include/linux/mm.h:806 [inline]
 kvmalloc_array include/linux/mm.h:824 [inline]
 kvcalloc include/linux/mm.h:829 [inline]
 memslot_rmap_alloc+0xf6/0x310 arch/x86/kvm/x86.c:11320
 kvm_alloc_memslot_metadata arch/x86/kvm/x86.c:11388 [inline]
 kvm_arch_prepare_memory_region+0x48d/0x610 arch/x86/kvm/x86.c:11462
 kvm_set_memslot+0xfe/0x1700 arch/x86/kvm/../../../virt/kvm/kvm_main.c:1505
 ...
 kvm_set_memory_region arch/x86/kvm/../../../virt/kvm/kvm_main.c:1689
 kvm_vm_ioctl_set_memory_region arch/x86/kvm/../../../virt/kvm/kvm_main.c

Reported-by: syzbot+e0de2333cbf95ea473e8@syzkaller.appspotmail.com
Signed-off-by: Sabyrzhan Tasbolatov <snovitoll@gmail.com>
---
 arch/x86/kvm/mmu/page_track.c | 3 +++
 arch/x86/kvm/x86.c            | 3 +++
 2 files changed, 6 insertions(+)

diff --git a/arch/x86/kvm/mmu/page_track.c b/arch/x86/kvm/mmu/page_track.c
index 21427e84a82e..e790bb341680 100644
--- a/arch/x86/kvm/mmu/page_track.c
+++ b/arch/x86/kvm/mmu/page_track.c
@@ -35,6 +35,9 @@ int kvm_page_track_create_memslot(struct kvm_memory_slot *slot,
 	int  i;
 
 	for (i = 0; i < KVM_PAGE_TRACK_MAX; i++) {
+		if (npages > KMALLOC_MAX_SIZE)
+			return -ENOMEM;
+
 		slot->arch.gfn_track[i] =
 			kvcalloc(npages, sizeof(*slot->arch.gfn_track[i]),
 				 GFP_KERNEL_ACCOUNT);
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index aabd3a2ec1bc..2bad607976a9 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -11394,6 +11394,9 @@ static int memslot_rmap_alloc(struct kvm_memory_slot *slot,
 
 		WARN_ON(slot->arch.rmap[i]);
 
+		if (lpages > KMALLOC_MAX_SIZE)
+			return -ENOMEM;
+
 		slot->arch.rmap[i] = kvcalloc(lpages, sz, GFP_KERNEL_ACCOUNT);
 		if (!slot->arch.rmap[i]) {
 			memslot_rmap_free(slot);
-- 
2.25.1

Re: [PATCH] x86/kvm: restrict kvm user region memory size

[Paolo Bonzini]

On 14/10/21 14:01, Sabyrzhan Tasbolatov wrote:
> syzbot found WARNING in memslot_rmap_alloc[1] when
> struct kvm_userspace_memory_region .memory_size is bigger than
> 0x40000000000, which is 4GB, e.g. KMALLOC_MAX_SIZE * 100 * PAGE_SIZE.
> 
> Here is the PoC to trigger the warning:
> 
>      struct kvm_userspace_memory_region mem = {
>          .slot = 0,
>          .guest_phys_addr = 0,
>          /* + 0x100 extra to trigger kmalloc WARNING */
>          .memory_size = 0x40000000000 + 0x100,
>          .userspace_addr = 0,
>      };
> 
>      ioctl(kvm_fd, KVM_SET_USER_MEMORY_REGION, &mem);
> 
> I couldn't find any relevant max constant to restrict unsigned long npages.
> There might be another solution with chunking big portions of pages, but
> there is already KVM_MAX_HUGEPAGE_LEVEL, though warning happens in
> memslot_rmap_alloc() when level = 1, base_gfn = 0, e.g.
> on the very first KVM_NR_PAGE_SIZES iteration.
> 
> This is, seems, valid for early Linux versions as well. Can't tell which is
> exactly can be considered for git bisect.
> Here is Commit d89cc617b954af ("KVM: Push rmap into kvm_arch_memory_slot")
> for example, Linux 3.7.

The warning is bogus in this case.  See the discussion in 
https://lkml.org/lkml/2021/9/7/669.  The right fix is simply to use 
vmalloc instead of kmalloc.

I'm woefully behind on my KVM maintainer duties, but this is on my todo 
list.

Paolo

> [1]
> Call Trace:
>   kvmalloc include/linux/mm.h:806 [inline]
>   kvmalloc_array include/linux/mm.h:824 [inline]
>   kvcalloc include/linux/mm.h:829 [inline]
>   memslot_rmap_alloc+0xf6/0x310 arch/x86/kvm/x86.c:11320
>   kvm_alloc_memslot_metadata arch/x86/kvm/x86.c:11388 [inline]
>   kvm_arch_prepare_memory_region+0x48d/0x610 arch/x86/kvm/x86.c:11462
>   kvm_set_memslot+0xfe/0x1700 arch/x86/kvm/../../../virt/kvm/kvm_main.c:1505
>   ...
>   kvm_set_memory_region arch/x86/kvm/../../../virt/kvm/kvm_main.c:1689
>   kvm_vm_ioctl_set_memory_region arch/x86/kvm/../../../virt/kvm/kvm_main.c
> 
> Reported-by: syzbot+e0de2333cbf95ea473e8@syzkaller.appspotmail.com
> Signed-off-by: Sabyrzhan Tasbolatov <snovitoll@gmail.com>
> ---
>   arch/x86/kvm/mmu/page_track.c | 3 +++
>   arch/x86/kvm/x86.c            | 3 +++
>   2 files changed, 6 insertions(+)
> 
> diff --git a/arch/x86/kvm/mmu/page_track.c b/arch/x86/kvm/mmu/page_track.c
> index 21427e84a82e..e790bb341680 100644
> --- a/arch/x86/kvm/mmu/page_track.c
> +++ b/arch/x86/kvm/mmu/page_track.c
> @@ -35,6 +35,9 @@ int kvm_page_track_create_memslot(struct kvm_memory_slot *slot,
>   	int  i;
>   
>   	for (i = 0; i < KVM_PAGE_TRACK_MAX; i++) {
> +		if (npages > KMALLOC_MAX_SIZE)
> +			return -ENOMEM;
> +
>   		slot->arch.gfn_track[i] =
>   			kvcalloc(npages, sizeof(*slot->arch.gfn_track[i]),
>   				 GFP_KERNEL_ACCOUNT);
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index aabd3a2ec1bc..2bad607976a9 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -11394,6 +11394,9 @@ static int memslot_rmap_alloc(struct kvm_memory_slot *slot,
>   
>   		WARN_ON(slot->arch.rmap[i]);
>   
> +		if (lpages > KMALLOC_MAX_SIZE)
> +			return -ENOMEM;
> +
>   		slot->arch.rmap[i] = kvcalloc(lpages, sz, GFP_KERNEL_ACCOUNT);
>   		if (!slot->arch.rmap[i]) {
>   			memslot_rmap_free(slot);
>