[BUGFIX] kvm, Fix a race condition for usage of is_hwpoison_address

[BUGFIX] kvm, Fix a race condition for usage of is_hwpoison_address

[Huang Ying]

is_hwpoison_address accesses the page table, so the caller must hold
current->mm->mmap_sem in read mode. So fix its usage in hav_to_pfn of
kvm accordingly.

Comments on is_hwpoison_address are added to remind other users.

Reported-by: Avi Kivity <avi@redhat.com>
Signed-off-by: Huang Ying <ying.huang@intel.com>
---
 mm/memory-failure.c |    3 +++
 virt/kvm/kvm_main.c |    3 ++-
 2 files changed, 5 insertions(+), 1 deletion(-)

--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -1298,6 +1298,9 @@ done:
 	return ret;
 }
 
+/*
+ * The caller must hold current->mm->mmap_sem in read mode.
+ */
 int is_hwpoison_address(unsigned long addr)
 {
 	pgd_t *pgdp;
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -955,12 +955,13 @@ static pfn_t hva_to_pfn(struct kvm *kvm,
 	if (unlikely(npages != 1)) {
 		struct vm_area_struct *vma;
 
+		down_read(&current->mm->mmap_sem);
 		if (is_hwpoison_address(addr)) {
+			up_read(&current->mm->mmap_sem);
 			get_page(hwpoison_page);
 			return page_to_pfn(hwpoison_page);
 		}
 
-		down_read(&current->mm->mmap_sem);
 		vma = find_vma(current->mm, addr);
 
 		if (vma == NULL || addr < vma->vm_start ||

Re: [BUGFIX] kvm, Fix a race condition for usage of is_hwpoison_address

[Avi Kivity]

On 06/22/2010 09:23 AM, Huang Ying wrote:
> is_hwpoison_address accesses the page table, so the caller must hold
> current->mm->mmap_sem in read mode. So fix its usage in hav_to_pfn of
> kvm accordingly.
>
>    

Applied, thanks.

-- 
error compiling committee.c: too many arguments to function