Re: [PATCH 5.14 018/334] nbd: add the check to prevent overflow in __nbd_ioctl()

[Nick Desaulniers]

On Mon, Sep 13, 2021 at 11:39 AM Nick Desaulniers
<ndesaulniers@xxxxxxxxxx> wrote:
>
> On Mon, Sep 13, 2021 at 10:58 AM Greg Kroah-Hartman
> <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> >
> > On Mon, Sep 13, 2021 at 09:52:33PM +0530, Naresh Kamboju wrote:
> > > [PATCH 00/10] raise minimum GCC version to 5.1
> > > https://lore.kernel.org/lkml/20210910234047.1019925-1-ndesaulniers@xxxxxxxxxx/
> >
> > Has anyone submitted a fix for this upstream yet?  I can't seem to find
> > one :(
>
> That lore link has a series to address this, though that's maybe
> something we don't want to backport to stable.
>
> I thought about this all weekend; I think I might be able to work
> around the one concern I had with my other approach, using
> __builtin_choose_expr().
>
> There's an issue with my alternative approach
> (https://gist.github.com/nickdesaulniers/2479818f4983bbf2d688cebbab435863)
> with declaring the local variable z in div_64() since either operand
> could be 64b, which result in an unwanted truncation if the dividend
> is 32b (or less, and divisor is 64b). I think (what I realized this
> weekend) is that we might be able to replace the `if` with
> `__builtin_choose_expr`, then have that whole expression be the final
> statement and thus the "return value" of the statement expression.

Christ...that...works? Though, did Linus just merge my patches for gcc 5.1?

Anyways, I'll send something like this for stable:
---

diff --git a/include/linux/math64.h b/include/linux/math64.h
index 2928f03d6d46..e9ab8c25f8d3 100644
--- a/include/linux/math64.h
+++ b/include/linux/math64.h
@@ -11,6 +11,9 @@

 #define div64_long(x, y) div64_s64((x), (y))
 #define div64_ul(x, y)   div64_u64((x), (y))
+#ifndef is_signed_type
+#define is_signed_type(type)       (((type)(-1)) < (type)1)
+#endif

 /**
  * div_u64_rem - unsigned 64bit divide with 32bit divisor with remainder
@@ -112,6 +115,15 @@ extern s64 div64_s64(s64 dividend, s64 divisor);

 #endif /* BITS_PER_LONG */

+#define div64_x64(dividend, divisor) ({                        \
+       BUILD_BUG_ON_MSG(sizeof(dividend) < sizeof(u64),\
+                        "prefer div_x64");             \
+       __builtin_choose_expr(                          \
+               is_signed_type(typeof(dividend)),       \
+               div64_s64(dividend, divisor),           \
+               div64_u64(dividend, divisor));          \
+})
+
 /**
  * div_u64 - unsigned 64bit divide with 32bit divisor
  * @dividend: unsigned 64bit dividend
@@ -142,6 +154,28 @@ static inline s64 div_s64(s64 dividend, s32 divisor)
 }
 #endif

+#define div_x64(dividend, divisor) ({                  \
+       BUILD_BUG_ON_MSG(sizeof(dividend) > sizeof(u32),\
+                        "prefer div64_x64");           \
+       __builtin_choose_expr(                          \
+               is_signed_type(typeof(dividend)),       \
+               div_s64(dividend, divisor),             \
+               div_u64(dividend, divisor));            \
+})
+
+// TODO: what if divisor is 128b?
+#define div_64(dividend, divisor) ({
         \
+       __builtin_choose_expr(
         \
+               __builtin_types_compatible_p(typeof(dividend), s64) ||
         \
+               __builtin_types_compatible_p(typeof(dividend), u64),
         \
+               __builtin_choose_expr(
         \
+                       __builtin_types_compatible_p(typeof(divisor),
s64) ||   \
+                       __builtin_types_compatible_p(typeof(divisor),
u64),     \
+                       div64_x64(dividend, divisor),
         \
+                       div_x64(dividend, divisor)),
         \
+               dividend / divisor);
         \
+})
+
 u32 iter_div_u64_rem(u64 dividend, u32 divisor, u64 *remainder);

 #ifndef mul_u32_u32
---
-- 
Thanks,
~Nick Desaulniers