From: Hugh Dickins Date: Sat, 28 Oct 2006 17:38:43 +0000 (-0700) Subject: [PATCH] hugetlb: fix prio_tree unit X-Git-Tag: v2.6.19-rc4~50 X-Git-Url: http://git.kernel.org/gitweb.cgi?p=linux%2Fkernel%2Fgit%2Ftorvalds%2Flinux-2.6.git;a=commitdiff_plain;h=856fc29505556cf263f3dcda2533cf3766c14ab6 [PATCH] hugetlb: fix prio_tree unit hugetlb_vmtruncate_list was misconverted to prio_tree: its prio_tree is in units of PAGE_SIZE (PAGE_CACHE_SIZE) like any other, not HPAGE_SIZE (whereas its radix_tree is kept in units of HPAGE_SIZE, otherwise slots would be absurdly sparse). At first I thought the error benign, just calling __unmap_hugepage_range on more vmas than necessary; but on 32-bit machines, when the prio_tree is searched correctly, it happens to ensure the v_offset calculation won't overflow. As it stood, when truncating at or beyond 4GB, it was liable to discard pages COWed from lower offsets; or even to clear pmd entries of preceding vmas, triggering exit_mmap's BUG_ON(nr_ptes). Signed-off-by: Hugh Dickins Cc: Adam Litke Cc: David Gibson Cc: "Chen, Kenneth W" Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- diff --git a/fs/hugetlbfs/inode.c b/fs/hugetlbfs/inode.c index 0b23b96..0bea6a6 100644 --- a/fs/hugetlbfs/inode.c +++ b/fs/hugetlbfs/inode.c @@ -271,26 +271,24 @@ static void hugetlbfs_drop_inode(struct inode *inode) hugetlbfs_forget_inode(inode); } -/* - * h_pgoff is in HPAGE_SIZE units. - * vma->vm_pgoff is in PAGE_SIZE units. - */ static inline void -hugetlb_vmtruncate_list(struct prio_tree_root *root, unsigned long h_pgoff) +hugetlb_vmtruncate_list(struct prio_tree_root *root, pgoff_t pgoff) { struct vm_area_struct *vma; struct prio_tree_iter iter; - vma_prio_tree_foreach(vma, &iter, root, h_pgoff, ULONG_MAX) { - unsigned long h_vm_pgoff; + vma_prio_tree_foreach(vma, &iter, root, pgoff, ULONG_MAX) { unsigned long v_offset; - h_vm_pgoff = vma->vm_pgoff >> (HPAGE_SHIFT - PAGE_SHIFT); - v_offset = (h_pgoff - h_vm_pgoff) << HPAGE_SHIFT; /* - * Is this VMA fully outside the truncation point? + * Can the expression below overflow on 32-bit arches? + * No, because the prio_tree returns us only those vmas + * which overlap the truncated area starting at pgoff, + * and no vma on a 32-bit arch can span beyond the 4GB. */ - if (h_vm_pgoff >= h_pgoff) + if (vma->vm_pgoff < pgoff) + v_offset = (pgoff - vma->vm_pgoff) << PAGE_SHIFT; + else v_offset = 0; __unmap_hugepage_range(vma, @@ -303,14 +301,14 @@ hugetlb_vmtruncate_list(struct prio_tree_root *root, unsigned long h_pgoff) */ static int hugetlb_vmtruncate(struct inode *inode, loff_t offset) { - unsigned long pgoff; + pgoff_t pgoff; struct address_space *mapping = inode->i_mapping; if (offset > inode->i_size) return -EINVAL; BUG_ON(offset & ~HPAGE_MASK); - pgoff = offset >> HPAGE_SHIFT; + pgoff = offset >> PAGE_SHIFT; inode->i_size = offset; spin_lock(&mapping->i_mmap_lock);