Thread (9 messages) 9 messages, 5 authors, 2016-03-16

Re: linux-next: build failure after merge of the aio tree

From: Andy Shevchenko <hidden>
Date: 2016-03-15 14:38:55
Also in: linux-next, lkml

On Tue, Mar 15, 2016 at 8:46 AM, Stephen Rothwell [off-list ref] wrote:
Hi Benjamin,

After merging the aio tree, today's linux-next build (powerpc
ppc44x_defconfig) failed like this:

fs/built-in.o: In function `aio_thread_op_foo_at':
aio.c:(.text+0x4dab4): undefined reference to `__get_user_bad'
aio.c:(.text+0x4daec): undefined reference to `__get_user_bad'
avr32 seems affected as well and the below solution is not suitable
(should be much more verbose in asm, I guess).
quoted hunk ↗ jump to hunk
Caused by commit

  150a0b4905f1 ("aio: add support for async openat()")

despite commit

  d2f7a973e11e ("aio: don't use __get_user() for 64 bit values")

This is due to a bug in the powerpc __get_user_check() macro (the return
value is defined to be "unsigned long" which is only 32 bits on a 32
bit platform).

I applied the patch below which seems to help (Michael, what do you
think?), but given Al's and Christoph's reactions, I am inclined to
remove the aio tree from tomorrow and maybe it can be revisited after
the merge window.

From: Stephen Rothwell <redacted>
Date: Tue, 15 Mar 2016 16:36:06 +1100
Subject: [PATCH] powerpc: fix get_user for 64 bit typs on 32 bit platforms

solution borrowed from the x86 uaccess.h

Signed-off-by: Stephen Rothwell <redacted>
---
 arch/powerpc/include/asm/uaccess.h | 9 ++++++---
 1 file changed, 6 insertions(+), 3 deletions(-)
diff --git a/arch/powerpc/include/asm/uaccess.h b/arch/powerpc/include/asm/uaccess.h
index b7c20f0b8fbe..52262b2f37fc 100644
--- a/arch/powerpc/include/asm/uaccess.h
+++ b/arch/powerpc/include/asm/uaccess.h
@@ -261,10 +261,13 @@ do {                                                              \
        }                                                       \
 } while (0)

+#define __inttype(x) \
+       __typeof__(__builtin_choose_expr(sizeof(x) > sizeof(0UL), 0ULL, 0UL))
+
 #define __get_user_nocheck(x, ptr, size)                       \
 ({                                                             \
        long __gu_err;                                          \
-       unsigned long __gu_val;                                 \
+       __inttype(*(ptr)) __gu_val;                             \
        __typeof__(*(ptr)) __user *__gu_addr = (ptr);   \
        __chk_user_ptr(ptr);                                    \
        if (!is_kernel_addr((unsigned long)__gu_addr))          \
@@ -277,7 +280,7 @@ do {                                                                \
 #define __get_user_check(x, ptr, size)                                 \
 ({                                                                     \
        long __gu_err = -EFAULT;                                        \
-       unsigned long  __gu_val = 0;                                    \
+       __inttype(*(ptr)) __gu_val = 0;                         \
        __typeof__(*(ptr)) __user *__gu_addr = (ptr);           \
        might_fault();                                                  \
        if (access_ok(VERIFY_READ, __gu_addr, (size)))                  \
@@ -289,7 +292,7 @@ do {                                                                \
 #define __get_user_nosleep(x, ptr, size)                       \
 ({                                                             \
        long __gu_err;                                          \
-       unsigned long __gu_val;                                 \
+       __inttype(*(ptr)) __gu_val;                             \
        __typeof__(*(ptr)) __user *__gu_addr = (ptr);   \
        __chk_user_ptr(ptr);                                    \
        __get_user_size(__gu_val, __gu_addr, (size), __gu_err); \
--
2.7.0

--
Cheers,
Stephen Rothwell


-- 
With Best Regards,
Andy Shevchenko
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help