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

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'

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

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).

-- 
With Best Regards,
Andy Shevchenko

On Tue, Mar 15, 2016 at 05:46:34PM +1100, Stephen Rothwell 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'

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).

m68k allmodconfig and all defs of m32r fails while building next-20160315.

regards
sudip

On Tue, Mar 15, 2016 at 04:19:02PM +0000, Sudip Mukherjee wrote:

On Tue, Mar 15, 2016 at 05:46:34PM +1100, Stephen Rothwell wrote:

quoted

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'

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).

m68k allmodconfig and all defs of m32r fails while building next-20160315.

regards
sudip

I've removed everything from the aio-next.git tree for now.  Will revisit 
after the merge window.

		-ben
-- 
"Thought is the essence of where you are now."

On Tuesday 15 March 2016 16:38:51 Andy Shevchenko wrote:

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

quoted

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).

ARM as well, at least for ARMv7-M.

	Arnd

On Tuesday 15 March 2016 12:22:28 Benjamin LaHaise wrote:

On Tue, Mar 15, 2016 at 04:19:02PM +0000, Sudip Mukherjee wrote:

quoted

On Tue, Mar 15, 2016 at 05:46:34PM +1100, Stephen Rothwell wrote:

quoted

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'

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).

m68k allmodconfig and all defs of m32r fails while building next-20160315.

regards
sudip

I've removed everything from the aio-next.git tree for now.  Will revisit 
after the merge window.

I've also sent a patch that fixes the link error on ARM and that should
work on all other architectures too.

	Arnd

On Wed, Mar 16, 2016 at 12:02 AM, Arnd Bergmann [off-list ref] wrote:

On Tuesday 15 March 2016 12:22:28 Benjamin LaHaise wrote:

quoted

On Tue, Mar 15, 2016 at 04:19:02PM +0000, Sudip Mukherjee wrote:

quoted

On Tue, Mar 15, 2016 at 05:46:34PM +1100, Stephen Rothwell wrote:

quoted

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'

Caused by commit

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

despite commit

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

quoted

I've removed everything from the aio-next.git tree for now.  Will revisit
after the merge window.

I think it is the best solution right now.

I've also sent a patch that fixes the link error on ARM and that should
work on all other architectures too.

In case of avr32 signalfd_read() fails. Does your patch help with it as well?

P.S. Bisecting shows same culprit: 150a0b4905f1 ("aio: add support for
async openat()")

-- 
With Best Regards,
Andy Shevchenko

On Wednesday 16 March 2016 13:12:36 Andy Shevchenko wrote:

quoted

I've also sent a patch that fixes the link error on ARM and that should
work on all other architectures too.

In case of avr32 signalfd_read() fails. Does your patch help with it as well?

P.S. Bisecting shows same culprit: 150a0b4905f1 ("aio: add support for
async openat()")

I don't know. What is the symptom on avr32? My patch only removes the
get_user() instances on 64-bit values and replaces them with a
single copy_from_user() call.

	Arnd

On Wed, Mar 16, 2016 at 02:59:38PM +0100, Arnd Bergmann wrote:

On Wednesday 16 March 2016 13:12:36 Andy Shevchenko wrote:

quoted

quoted

I've also sent a patch that fixes the link error on ARM and that should
work on all other architectures too.

In case of avr32 signalfd_read() fails. Does your patch help with it as well?

P.S. Bisecting shows same culprit: 150a0b4905f1 ("aio: add support for
async openat()")

I don't know. What is the symptom on avr32? My patch only removes the
get_user() instances on 64-bit values and replaces them with a
single copy_from_user() call.

Which is the wrong fix.  Arch code should be able to handle 64 bit values 
in all the get/put_user() variants.  We use 64 bit variables all over the 
place in interfaces to userspace.

		-ben