mm: fix vma_is_anonymous() false-positives
authorKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Thu, 26 Jul 2018 23:37:35 +0000 (16:37 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 27 Jul 2018 02:38:03 +0000 (19:38 -0700)
vma_is_anonymous() relies on ->vm_ops being NULL to detect anonymous
VMA.  This is unreliable as ->mmap may not set ->vm_ops.

False-positive vma_is_anonymous() may lead to crashes:

next ffff8801ce5e7040 prev ffff8801d20eca50 mm ffff88019c1e13c0
prot 27 anon_vma ffff88019680cdd8 vm_ops 0000000000000000
pgoff 0 file ffff8801b2ec2d00 private_data 0000000000000000
flags: 0xff(read|write|exec|shared|mayread|maywrite|mayexec|mayshare)
------------[ cut here ]------------
kernel BUG at mm/memory.c:1422!
invalid opcode: 0000 [#1] SMP KASAN
CPU: 0 PID: 18486 Comm: syz-executor3 Not tainted 4.18.0-rc3+ #136
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google
01/01/2011
RIP: 0010:zap_pmd_range mm/memory.c:1421 [inline]
RIP: 0010:zap_pud_range mm/memory.c:1466 [inline]
RIP: 0010:zap_p4d_range mm/memory.c:1487 [inline]
RIP: 0010:unmap_page_range+0x1c18/0x2220 mm/memory.c:1508
Call Trace:
 unmap_single_vma+0x1a0/0x310 mm/memory.c:1553
 zap_page_range_single+0x3cc/0x580 mm/memory.c:1644
 unmap_mapping_range_vma mm/memory.c:2792 [inline]
 unmap_mapping_range_tree mm/memory.c:2813 [inline]
 unmap_mapping_pages+0x3a7/0x5b0 mm/memory.c:2845
 unmap_mapping_range+0x48/0x60 mm/memory.c:2880
 truncate_pagecache+0x54/0x90 mm/truncate.c:800
 truncate_setsize+0x70/0xb0 mm/truncate.c:826
 simple_setattr+0xe9/0x110 fs/libfs.c:409
 notify_change+0xf13/0x10f0 fs/attr.c:335
 do_truncate+0x1ac/0x2b0 fs/open.c:63
 do_sys_ftruncate+0x492/0x560 fs/open.c:205
 __do_sys_ftruncate fs/open.c:215 [inline]
 __se_sys_ftruncate fs/open.c:213 [inline]
 __x64_sys_ftruncate+0x59/0x80 fs/open.c:213
 do_syscall_64+0x1b9/0x820 arch/x86/entry/common.c:290
 entry_SYSCALL_64_after_hwframe+0x49/0xbe

Reproducer:

#include <stdio.h>
#include <stddef.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/ioctl.h>
#include <sys/mman.h>
#include <unistd.h>
#include <fcntl.h>

#define KCOV_INIT_TRACE _IOR('c', 1, unsigned long)
#define KCOV_ENABLE _IO('c', 100)
#define KCOV_DISABLE _IO('c', 101)
#define COVER_SIZE (1024<<10)

#define KCOV_TRACE_PC  0
#define KCOV_TRACE_CMP 1

int main(int argc, char **argv)
{
int fd;
unsigned long *cover;

system("mount -t debugfs none /sys/kernel/debug");
fd = open("/sys/kernel/debug/kcov", O_RDWR);
ioctl(fd, KCOV_INIT_TRACE, COVER_SIZE);
cover = mmap(NULL, COVER_SIZE * sizeof(unsigned long),
PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
munmap(cover, COVER_SIZE * sizeof(unsigned long));
cover = mmap(NULL, COVER_SIZE * sizeof(unsigned long),
PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, 0);
memset(cover, 0, COVER_SIZE * sizeof(unsigned long));
ftruncate(fd, 3UL << 20);
return 0;
}

This can be fixed by assigning anonymous VMAs own vm_ops and not relying
on it being NULL.

If ->mmap() failed to set ->vm_ops, mmap_region() will set it to
dummy_vm_ops.  This way we will have non-NULL ->vm_ops for all VMAs.

Link: http://lkml.kernel.org/r/20180724121139.62570-4-kirill.shutemov@linux.intel.com
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reported-by: syzbot+3f84280d52be9b7083cc@syzkaller.appspotmail.com
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
drivers/char/mem.c
fs/exec.c
include/linux/mm.h
mm/mmap.c
mm/nommu.c

index ffeb60d..df66a9d 100644 (file)
@@ -708,6 +708,7 @@ static int mmap_zero(struct file *file, struct vm_area_struct *vma)
 #endif
        if (vma->vm_flags & VM_SHARED)
                return shmem_zero_setup(vma);
+       vma_set_anonymous(vma);
        return 0;
 }
 
index 72e961a..bdd0eac 100644 (file)
--- a/fs/exec.c
+++ b/fs/exec.c
@@ -293,6 +293,7 @@ static int __bprm_mm_init(struct linux_binprm *bprm)
        bprm->vma = vma = vm_area_alloc(mm);
        if (!vma)
                return -ENOMEM;
+       vma_set_anonymous(vma);
 
        if (down_write_killable(&mm->mmap_sem)) {
                err = -EINTR;
index 31540f1..7ba6d35 100644 (file)
@@ -454,10 +454,18 @@ struct vm_operations_struct {
 
 static inline void vma_init(struct vm_area_struct *vma, struct mm_struct *mm)
 {
+       static const struct vm_operations_struct dummy_vm_ops = {};
+
        vma->vm_mm = mm;
+       vma->vm_ops = &dummy_vm_ops;
        INIT_LIST_HEAD(&vma->anon_vma_chain);
 }
 
+static inline void vma_set_anonymous(struct vm_area_struct *vma)
+{
+       vma->vm_ops = NULL;
+}
+
 struct mmu_gather;
 struct inode;
 
index ff1944d..17bbf4d 100644 (file)
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -1778,6 +1778,8 @@ unsigned long mmap_region(struct file *file, unsigned long addr,
                error = shmem_zero_setup(vma);
                if (error)
                        goto free_vma;
+       } else {
+               vma_set_anonymous(vma);
        }
 
        vma_link(mm, vma, prev, rb_link, rb_parent);
@@ -2983,6 +2985,7 @@ static int do_brk_flags(unsigned long addr, unsigned long len, unsigned long fla
                return -ENOMEM;
        }
 
+       vma_set_anonymous(vma);
        vma->vm_start = addr;
        vma->vm_end = addr + len;
        vma->vm_pgoff = pgoff;
index 1d22fdb..9fc9e43 100644 (file)
@@ -1145,6 +1145,8 @@ static int do_mmap_private(struct vm_area_struct *vma,
                if (ret < len)
                        memset(base + ret, 0, len - ret);
 
+       } else {
+               vma_set_anonymous(vma);
        }
 
        return 0;