Fix incompatibility with Reiser4 patched kernels

In ZFSOnLinux, our sources and build system are self contained such that
we do not need to make changes to the Linux kernel sources. Reiser4 on
the other hand exists solely as a kernel tree patch and opts to make
changes to the kernel rather than adapt to it. After Linux 4.1 made a
VFS change that replaced new_sync_read with do_sync_read, Reiser4's
maintainer decided to modify the kernel VFS to export the old function.
This caused our autotools check to misidentify the kernel API as
predating Linux 4.1 on kernels that have been patched with Reiser4
support, which breaks our build.

Reiser4 really should be patched to stop doing this, but lets modify our
check to be more strict to help the affected users of both filesystems.

Also, we were not checking the types of arguments and return value of
new_sync_read() and new_sync_write() . Lets fix that too.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Giuseppe Di Natale <dinatale2@llnl.gov>
Reviewed-by: George Melikov <mail@gmelikov.ru>
Signed-off-by: Richard Yao <ryao@gentoo.org>
Closes openzfs#6241 
Closes openzfs#7021

config/kernel-vfs-rw-iterate.m4

@@ -32,15 +32,23 @@ dnl #
 dnl # Linux 4.1 API
 dnl #
 AC_DEFUN([ZFS_AC_KERNEL_NEW_SYNC_READ],
-	[AC_MSG_CHECKING([whether new_sync_read() is available])
+	[AC_MSG_CHECKING([whether new_sync_read/write() are available])
 	ZFS_LINUX_TRY_COMPILE([
 		#include <linux/fs.h>
 	],[
-		new_sync_read(NULL, NULL, 0, NULL);
+		        ssize_t ret __attribute__ ((unused));
+			struct file *filp = NULL;
+			char __user *rbuf = NULL;
+			const char __user *wbuf = NULL;
+			size_t len = 0;
+			loff_t ppos;
+
+			ret = new_sync_read(filp, rbuf, len, &ppos);
+			ret = new_sync_write(filp, wbuf, len, &ppos);
 	],[
 		AC_MSG_RESULT(yes)
 		AC_DEFINE(HAVE_NEW_SYNC_READ, 1,
-			[new_sync_read() is available])
+			[new_sync_read()/new_sync_write() are available])
 	],[
 		AC_MSG_RESULT(no)
 	])