Re: [PATCH v2 1/2] mm: add probe_user_read()

On Tue, Jan 8, 2019 at 1:11 PM Christophe Leroy [off-list ref] wrote:

Le 08/01/2019 à 20:48, Andrew Morton a écrit :

quoted

On Tue,  8 Jan 2019 07:37:44 +0000 (UTC) Christophe Leroy [off-list ref] wrote:

quoted

In powerpc code, there are several places implementing safe
access to user data. This is sometimes implemented using
probe_kernel_address() with additional access_ok() verification,
sometimes with get_user() enclosed in a pagefault_disable()/enable()
pair, etc. :
     show_user_instructions()
     bad_stack_expansion()
     p9_hmi_special_emu()
     fsl_pci_mcheck_exception()
     read_user_stack_64()
     read_user_stack_32() on PPC64
     read_user_stack_32() on PPC32
     power_pmu_bhrb_to()

In the same spirit as probe_kernel_read(), this patch adds
probe_user_read().

probe_user_read() does the same as probe_kernel_read() but
first checks that it is really a user address.

...

--- a/include/linux/uaccess.h
+++ b/include/linux/uaccess.h

@@ -263,6 +263,40 @@ extern long strncpy_from_unsafe(char *dst, const void *unsafe_addr, long count);
  #define probe_kernel_address(addr, retval)         \
     probe_kernel_read(&retval, addr, sizeof(retval))

+/**
+ * probe_user_read(): safely attempt to read from a user location
+ * @dst: pointer to the buffer that shall take the data
+ * @src: address to read from
+ * @size: size of the data chunk
+ *
+ * Returns: 0 on success, -EFAULT on error.
+ *
+ * Safely read from address @src to the buffer at @dst.  If a kernel fault
+ * happens, handle that and return -EFAULT.
+ *
+ * We ensure that the copy_from_user is executed in atomic context so that
+ * do_page_fault() doesn't attempt to take mmap_sem.  This makes
+ * probe_user_read() suitable for use within regions where the caller
+ * already holds mmap_sem, or other locks which nest inside mmap_sem.
+ */
+
+#ifndef probe_user_read
+static __always_inline long probe_user_read(void *dst, const void __user *src,
+                                        size_t size)
+{
+    long ret;
+
+    if (!access_ok(src, size))
+            return -EFAULT;
+
+    pagefault_disable();
+    ret = __copy_from_user_inatomic(dst, src, size);
+    pagefault_enable();
+
+    return ret ? -EFAULT : 0;
+}
+#endif

Why was the __always_inline needed?

This function is pretty large.  Why is it inlined?

Kees told to do that way, see https://patchwork.ozlabs.org/patch/986848/

Yeah, I'd like to make sure we can plumb the size checks down into the
user copy primitives.

-- 
Kees Cook