Re: Suggest make 'user_access_begin()' do 'access_ok()' to stable kernel

[Miles Chen <miles.chen@mediatek.com>]

[-- Attachment #1: Type: text/plain, Size: 3355 bytes --]

On Wed, 2020-06-10 at 20:02 +0200, Greg KH wrote:
> On Thu, Jun 11, 2020 at 01:58:20AM +0800, Miles Chen wrote:
> > Hi,
> > 
> > I suggest to include the commit: 594cc251fdd0 make 'user_access_begin()'
> > do 'access_ok()' for CVE-2018-20669.
> > 
> > stable version to apply to: kernel-4.14.y and kernel-4.19.y.
> > 
> > 
> > From the discussion below, I checked the latest kernel and found that we
> > should also apply other 4 patches. (total 5 patches)
> > https://lkml.org/lkml/2020/5/12/943
> > 
> > 
> > patch list:
> > commit ab10ae1c3bef lib: Reduce user_access_begin() boundaries in
> > strncpy_from_user() and strnlen_user()
> > commit 6e693b3ffecb x86: uaccess: Inhibit speculation past access_ok()
> > in user_access_begin()
> > commit 9cb2feb4d21d arch/openrisc: Fix issues with access_ok()
> > commit 94bd8a05cd4d Fix 'acccess_ok()' on alpha and SH
> > commit 594cc251fdd0 make 'user_access_begin()' do 'access_ok()'
> > 
> > 
> > Where only commit 6e693b3ffecb does not need backport modifications.
> > I attach my backport patches in this email.
> > 
> > I merged the patches with kernel-4.19.127 and kernel-4.14.183 without
> > conflicts.
> > Build with arm64 defconfig and bootup on arm64 QEMU environment.
> > 
> > cheers,
> > Miles
> 
> > From ac351de9ddd86ef717a3f89236dc5f6b2a108cc7 Mon Sep 17 00:00:00 2001
> > From: Linus Torvalds <torvalds@linux-foundation.org>
> > Date: Fri, 4 Jan 2019 12:56:09 -0800
> > Subject: [PATCH] BACKPORT: make 'user_access_begin()' do 'access_ok()'
> > 
> > upstream commit 594cc251fdd0 ("make 'user_access_begin()' do 'access_ok()'")
> > 
> > Originally, the rule used to be that you'd have to do access_ok()
> > separately, and then user_access_begin() before actually doing the
> > direct (optimized) user access.
> > 
> > But experience has shown that people then decide not to do access_ok()
> > at all, and instead rely on it being implied by other operations or
> > similar.  Which makes it very hard to verify that the access has
> > actually been range-checked.
> > 
> > If you use the unsafe direct user accesses, hardware features (either
> > SMAP - Supervisor Mode Access Protection - on x86, or PAN - Privileged
> > Access Never - on ARM) do force you to use user_access_begin().  But
> > nothing really forces the range check.
> > 
> > By putting the range check into user_access_begin(), we actually force
> > people to do the right thing (tm), and the range check vill be visible
> > near the actual accesses.  We have way too long a history of people
> > trying to avoid them.
> > 
> > Bug: 135368228
> > Change-Id: I4ca0e4566ea080fa148c5e768bb1a0b6f7201c01
> > Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
> 
> No need for "Bug:" or "Change-Id:" for patches for stable trees.
> 
> Also, can you please sign off on these as well?
> 
> Can you fix that up and resend?  I'll be glad to queue them up then.
> 
> thanks,

Remove the "Bug/Change-Id" from
0001-BACKPORT-make-user_access_begin-do-access_ok.patch.

Actually, I got the patch from
https://android-review.googlesource.com/c/kernel/common/+/1114632
Todd backported the patch but there is no Todd's signed-off-by in his
patch. Should I add "Signed-off-by: Todd Kjos <tkjos@google.com>" as
well?

cheers
Miles
> 
> greg k-h


[-- Attachment #2: 0001-BACKPORT-make-user_access_begin-do-access_ok.patch --]
[-- Type: text/x-patch, Size: 8031 bytes --]

From ac351de9ddd86ef717a3f89236dc5f6b2a108cc7 Mon Sep 17 00:00:00 2001
From: Linus Torvalds <torvalds@linux-foundation.org>
Date: Fri, 4 Jan 2019 12:56:09 -0800
Subject: [PATCH] BACKPORT: make 'user_access_begin()' do 'access_ok()'

upstream commit 594cc251fdd0 ("make 'user_access_begin()' do 'access_ok()'")

Originally, the rule used to be that you'd have to do access_ok()
separately, and then user_access_begin() before actually doing the
direct (optimized) user access.

But experience has shown that people then decide not to do access_ok()
at all, and instead rely on it being implied by other operations or
similar.  Which makes it very hard to verify that the access has
actually been range-checked.

If you use the unsafe direct user accesses, hardware features (either
SMAP - Supervisor Mode Access Protection - on x86, or PAN - Privileged
Access Never - on ARM) do force you to use user_access_begin().  But
nothing really forces the range check.

By putting the range check into user_access_begin(), we actually force
people to do the right thing (tm), and the range check vill be visible
near the actual accesses.  We have way too long a history of people
trying to avoid them.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Miles Chen <miles.chen@mediatek.com>
---
 arch/x86/include/asm/uaccess.h             | 12 +++++++++++-
 drivers/gpu/drm/i915/i915_gem_execbuffer.c | 16 ++++++++++++++--
 include/linux/uaccess.h                    |  2 +-
 kernel/compat.c                            |  6 ++----
 kernel/exit.c                              |  6 ++----
 lib/strncpy_from_user.c                    |  9 +++++----
 lib/strnlen_user.c                         |  9 +++++----
 7 files changed, 40 insertions(+), 20 deletions(-)

diff --git a/arch/x86/include/asm/uaccess.h b/arch/x86/include/asm/uaccess.h
index 4111edb3188e..4128e6fdfcc4 100644
--- a/arch/x86/include/asm/uaccess.h
+++ b/arch/x86/include/asm/uaccess.h
@@ -709,7 +709,17 @@ extern struct movsl_mask {
  * checking before using them, but you have to surround them with the
  * user_access_begin/end() pair.
  */
-#define user_access_begin()	__uaccess_begin()
+static __must_check inline bool user_access_begin(int type,
+						  const void __user *ptr,
+						  size_t len)
+{
+	if (unlikely(!access_ok(type, ptr, len)))
+		return 0;
+	__uaccess_begin();
+	return 1;
+}
+
+#define user_access_begin(a, b, c)	user_access_begin(a, b, c)
 #define user_access_end()	__uaccess_end()
 
 #define unsafe_put_user(x, ptr, err_label)					\
diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 679bbae52945..04188e5168a2 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -1602,7 +1602,9 @@ static int eb_copy_relocations(const struct i915_execbuffer *eb)
 		 * happened we would make the mistake of assuming that the
 		 * relocations were valid.
 		 */
-		user_access_begin();
+		if (!user_access_begin(VERIFY_WRITE, urelocs, size))
+			goto end_user;
+
 		for (copied = 0; copied < nreloc; copied++)
 			unsafe_put_user(-1,
 					&urelocs[copied].presumed_offset,
@@ -2601,7 +2603,17 @@ i915_gem_execbuffer2_ioctl(struct drm_device *dev, void *data,
 		unsigned int i;
 
 		/* Copy the new buffer offsets back to the user's exec list. */
-		user_access_begin();
+		/*
+		 * Note: count * sizeof(*user_exec_list) does not overflow,
+		 * because we checked 'count' in check_buffer_count().
+		 *
+		 * And this range already got effectively checked earlier
+		 * when we did the "copy_from_user()" above.
+		 */
+		if (!user_access_begin(VERIFY_WRITE, user_exec_list,
+				       count * sizeof(*user_exec_list)))
+			goto end_user;
+
 		for (i = 0; i < args->buffer_count; i++) {
 			if (!(exec2_list[i].offset & UPDATE))
 				continue;
diff --git a/include/linux/uaccess.h b/include/linux/uaccess.h
index efe79c1cdd47..d55b68b113de 100644
--- a/include/linux/uaccess.h
+++ b/include/linux/uaccess.h
@@ -267,7 +267,7 @@ extern long strncpy_from_unsafe(char *dst, const void *unsafe_addr, long count);
 	probe_kernel_read(&retval, addr, sizeof(retval))
 
 #ifndef user_access_begin
-#define user_access_begin() do { } while (0)
+#define user_access_begin(type, ptr, len) access_ok(type, ptr, len)
 #define user_access_end() do { } while (0)
 #define unsafe_get_user(x, ptr, err) do { if (unlikely(__get_user(x, ptr))) goto err; } while (0)
 #define unsafe_put_user(x, ptr, err) do { if (unlikely(__put_user(x, ptr))) goto err; } while (0)
diff --git a/kernel/compat.c b/kernel/compat.c
index 8e40efc2928a..e4548a9e9c52 100644
--- a/kernel/compat.c
+++ b/kernel/compat.c
@@ -354,10 +354,9 @@ long compat_get_bitmap(unsigned long *mask, const compat_ulong_t __user *umask,
 	bitmap_size = ALIGN(bitmap_size, BITS_PER_COMPAT_LONG);
 	nr_compat_longs = BITS_TO_COMPAT_LONGS(bitmap_size);
 
-	if (!access_ok(VERIFY_READ, umask, bitmap_size / 8))
+	if (!user_access_begin(VERIFY_READ, umask, bitmap_size / 8))
 		return -EFAULT;
 
-	user_access_begin();
 	while (nr_compat_longs > 1) {
 		compat_ulong_t l1, l2;
 		unsafe_get_user(l1, umask++, Efault);
@@ -384,10 +383,9 @@ long compat_put_bitmap(compat_ulong_t __user *umask, unsigned long *mask,
 	bitmap_size = ALIGN(bitmap_size, BITS_PER_COMPAT_LONG);
 	nr_compat_longs = BITS_TO_COMPAT_LONGS(bitmap_size);
 
-	if (!access_ok(VERIFY_WRITE, umask, bitmap_size / 8))
+	if (!user_access_begin(VERIFY_WRITE, umask, bitmap_size / 8))
 		return -EFAULT;
 
-	user_access_begin();
 	while (nr_compat_longs > 1) {
 		unsigned long m = *mask++;
 		unsafe_put_user((compat_ulong_t)m, umask++, Efault);
diff --git a/kernel/exit.c b/kernel/exit.c
index e6d8ba1cd2e2..213ba451dcb3 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -1614,10 +1614,9 @@ SYSCALL_DEFINE5(waitid, int, which, pid_t, upid, struct siginfo __user *,
 	if (!infop)
 		return err;
 
-	if (!access_ok(VERIFY_WRITE, infop, sizeof(*infop)))
+	if (!user_access_begin(VERIFY_WRITE, infop, sizeof(*infop)))
 		return -EFAULT;
 
-	user_access_begin();
 	unsafe_put_user(signo, &infop->si_signo, Efault);
 	unsafe_put_user(0, &infop->si_errno, Efault);
 	unsafe_put_user(info.cause, &infop->si_code, Efault);
@@ -1742,10 +1741,9 @@ COMPAT_SYSCALL_DEFINE5(waitid,
 	if (!infop)
 		return err;
 
-	if (!access_ok(VERIFY_WRITE, infop, sizeof(*infop)))
+	if (!user_access_begin(VERIFY_WRITE, infop, sizeof(*infop)))
 		return -EFAULT;
 
-	user_access_begin();
 	unsafe_put_user(signo, &infop->si_signo, Efault);
 	unsafe_put_user(0, &infop->si_errno, Efault);
 	unsafe_put_user(info.cause, &infop->si_code, Efault);
diff --git a/lib/strncpy_from_user.c b/lib/strncpy_from_user.c
index e304b54c9c7d..b8570a11776d 100644
--- a/lib/strncpy_from_user.c
+++ b/lib/strncpy_from_user.c
@@ -115,10 +115,11 @@ long strncpy_from_user(char *dst, const char __user *src, long count)
 
 		kasan_check_write(dst, count);
 		check_object_size(dst, count, false);
-		user_access_begin();
-		retval = do_strncpy_from_user(dst, src, count, max);
-		user_access_end();
-		return retval;
+		if (user_access_begin(VERIFY_READ, src, max)) {
+			retval = do_strncpy_from_user(dst, src, count, max);
+			user_access_end();
+			return retval;
+		}
 	}
 	return -EFAULT;
 }
diff --git a/lib/strnlen_user.c b/lib/strnlen_user.c
index 184f80f7bacf..f5fa5b266ea2 100644
--- a/lib/strnlen_user.c
+++ b/lib/strnlen_user.c
@@ -114,10 +114,11 @@ long strnlen_user(const char __user *str, long count)
 		unsigned long max = max_addr - src_addr;
 		long retval;
 
-		user_access_begin();
-		retval = do_strnlen_user(str, count, max);
-		user_access_end();
-		return retval;
+		if (user_access_begin(VERIFY_READ, str, max)) {
+			retval = do_strnlen_user(str, count, max);
+			user_access_end();
+			return retval;
+		}
 	}
 	return 0;
 }
-- 
2.18.0