On Fri, May 26, 2017 at 1:40 PM, Arnd Bergmann <arnd@xxxxxxxx> wrote: > From: Kees Cook <keescook@xxxxxxxxxxxx> > > Commit f5f893c57e37ca730808cb2eee3820abd05e7507 upstream. > > Under SMAP/PAN/etc, we cannot write directly to userspace memory, so > this rearranges the test bytes to get written through copy_to_user(). > Additionally drops the bad copy_from_user() test that would trigger a > memcpy() against userspace on failure. > > [arnd: the test module was added in 3.14, and this backported patch > should apply cleanly on all version from 3.14 to 4.10. > The original patch was in 4.11 on top of a context change. > I saw the bug triggered with kselftest on a 4.4.y stable kernel] > > Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx> > Signed-off-by: Arnd Bergmann <arnd@xxxxxxxx> > --- > lib/test_user_copy.c | 21 +++++++++++++++++++-- > 1 file changed, 19 insertions(+), 2 deletions(-) > > diff --git a/lib/test_user_copy.c b/lib/test_user_copy.c > index 0ecef3e4690e..b16891d01112 100644 > --- a/lib/test_user_copy.c > +++ b/lib/test_user_copy.c > @@ -58,7 +58,9 @@ static int __init test_user_copy_init(void) > usermem = (char __user *)user_addr; > bad_usermem = (char *)user_addr; > > - /* Legitimate usage: none of these should fail. */ > + /* > + * Legitimate usage: none of these copies should fail. > + */ > ret |= test(copy_from_user(kmem, usermem, PAGE_SIZE), > "legitimate copy_from_user failed"); > ret |= test(copy_to_user(usermem, kmem, PAGE_SIZE), > @@ -68,19 +70,34 @@ static int __init test_user_copy_init(void) > ret |= test(put_user(value, (unsigned long __user *)usermem), > "legitimate put_user failed"); > > - /* Invalid usage: none of these should succeed. */ > + /* > + * Invalid usage: none of these copies should succeed. > + */ > + > + /* Reject kernel-to-kernel copies through copy_from_user(). */ > ret |= test(!copy_from_user(kmem, (char __user *)(kmem + PAGE_SIZE), > PAGE_SIZE), > "illegal all-kernel copy_from_user passed"); > + > +#if 0 > + /* > + * When running with SMAP/PAN/etc, this will Oops the kernel > + * due to the zeroing of userspace memory on failure. This needs > + * to be tested in LKDTM instead, since this test module does not > + * expect to explode. > + */ > ret |= test(!copy_from_user(bad_usermem, (char __user *)kmem, > PAGE_SIZE), > "illegal reversed copy_from_user passed"); > +#endif > ret |= test(!copy_to_user((char __user *)kmem, kmem + PAGE_SIZE, > PAGE_SIZE), > "illegal all-kernel copy_to_user passed"); > ret |= test(!copy_to_user((char __user *)kmem, bad_usermem, > PAGE_SIZE), > "illegal reversed copy_to_user passed"); > + > + value = 0x5a; I don't think this "value" bit wanted, but it should be harmless. -Kees > ret |= test(!get_user(value, (unsigned long __user *)kmem), > "illegal get_user passed"); > ret |= test(!put_user(value, (unsigned long __user *)kmem), > -- > 2.9.0 > > > _______________________________________________ > linux-arm-kernel mailing list > linux-arm-kernel@xxxxxxxxxxxxxxxxxxx > http://lists.infradead.org/mailman/listinfo/linux-arm-kernel -- Kees Cook Pixel Security