parisc: Fix gcc miscompilation in pa_memcpy()

When running the LTP testsuite one may hit this kernel BUG() with the
write06 testcase:

kernel BUG at mm/filemap.c:2023!
CPU: 1 PID: 8614 Comm: writev01 Not tainted 3.10.0-rc7-64bit-c3000+ #6
IASQ: 0000000000000000 0000000000000000 IAOQ: 00000000401e6e84 00000000401e6e88
 IIR: 03ffe01f    ISR: 0000000010340000  IOR: 000001fbe0380820
 CPU:        1   CR30: 00000000bef80000 CR31: ffffffffffffffff
 ORIG_R28: 00000000bdc192c0
 IAOQ[0]: iov_iter_advance+0x3c/0xc0
 IAOQ[1]: iov_iter_advance+0x40/0xc0
 RP(r2): generic_file_buffered_write+0x204/0x3f0
Backtrace:
 [<00000000401e764c>] generic_file_buffered_write+0x204/0x3f0
 [<00000000401eab24>] __generic_file_aio_write+0x244/0x448
 [<00000000401eadc0>] generic_file_aio_write+0x98/0x150
 [<000000004024f460>] do_sync_readv_writev+0xc0/0x130
 [<000000004025037c>] compat_do_readv_writev+0x12c/0x340
 [<00000000402505f8>] compat_writev+0x68/0xa0
 [<0000000040251d88>] compat_SyS_writev+0x98/0xf8

Reason for this crash is a gcc miscompilation in the fault handlers of
pa_memcpy() which return the fault address instead of the copied bytes.
Since this seems to be a generic problem with gcc-4.7.x (and below), it's
better to simplify the fault handlers in pa_memcpy to avoid this problem.

Here is a simple reproducer for the problem:

int main(int argc, char **argv)
{
	int fd, nbytes;
	struct iovec wr_iovec[] = {
		{ "TEST STRING                     ",32},
		{ (char*)0x40005000,32} }; // random memory.
	fd = open(DATA_FILE, O_RDWR | O_CREAT, 0666);
	nbytes = writev(fd, wr_iovec, 2);
	printf("return value = %d, errno %d (%s)\n",
		nbytes, errno, strerror(errno));
	return 0;
}

In addition, John David Anglin wrote:
There is no gcc PR as pa_memcpy is not legitimate C code. There is an
implicit assumption that certain variables will contain correct values
when an exception occurs and the code randomly jumps to one of the
exception blocks.  There is no guarantee of this.  If a PR was filed, it
would likely be marked as invalid.
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: John David Anglin <dave.anglin@bell.net>
Cc: <stable@vger.kernel.org> # 3.8+
Signed-off-by: Helge Deller <deller@gmx.de>

arch/parisc/lib/memcpy.c

@@ -2,6 +2,7 @@
  *    Optimized memory copy routines.
  *
  *    Copyright (C) 2004 Randolph Chung <tausq@debian.org>
+ *    Copyright (C) 2013 Helge Deller <deller@gmx.de>
  *
  *    This program is free software; you can redistribute it and/or modify
  *    it under the terms of the GNU General Public License as published by
@@ -153,17 +154,21 @@ static inline void prefetch_dst(const void *addr)
 #define prefetch_dst(addr) do { } while(0)
 #endif
 
+#define PA_MEMCPY_OK		0
+#define PA_MEMCPY_LOAD_ERROR	1
+#define PA_MEMCPY_STORE_ERROR	2
+
 /* Copy from a not-aligned src to an aligned dst, using shifts. Handles 4 words
  * per loop.  This code is derived from glibc. 
  */
-static inline unsigned long copy_dstaligned(unsigned long dst, unsigned long src, unsigned long len, unsigned long o_dst, unsigned long o_src, unsigned long o_len)
+static inline unsigned long copy_dstaligned(unsigned long dst,
+					unsigned long src, unsigned long len)
 {
 	/* gcc complains that a2 and a3 may be uninitialized, but actually
 	 * they cannot be.  Initialize a2/a3 to shut gcc up.
 	 */
 	register unsigned int a0, a1, a2 = 0, a3 = 0;
 	int sh_1, sh_2;
-	struct exception_data *d;
 
 	/* prefetch_src((const void *)src); */
 
@@ -197,7 +202,7 @@ static inline unsigned long copy_dstaligned(unsigned long dst, unsigned long src
 			goto do2;
 		case 0:
 			if (len == 0)
-				return 0;
+				return PA_MEMCPY_OK;
 			/* a3 = ((unsigned int *) src)[0];
 			   a0 = ((unsigned int *) src)[1]; */
 			ldw(s_space, 0, src, a3, cda_ldw_exc);
@@ -256,42 +261,35 @@ static inline unsigned long copy_dstaligned(unsigned long dst, unsigned long src
 	preserve_branch(handle_load_error);
 	preserve_branch(handle_store_error);
 
-	return 0;
+	return PA_MEMCPY_OK;
 
 handle_load_error:
 	__asm__ __volatile__ ("cda_ldw_exc:\n");
-	d = &__get_cpu_var(exception_data);
-	DPRINTF("cda_ldw_exc: o_len=%lu fault_addr=%lu o_src=%lu ret=%lu\n",
-		o_len, d->fault_addr, o_src, o_len - d->fault_addr + o_src);
-	return o_len * 4 - d->fault_addr + o_src;
+	return PA_MEMCPY_LOAD_ERROR;
 
 handle_store_error:
 	__asm__ __volatile__ ("cda_stw_exc:\n");
-	d = &__get_cpu_var(exception_data);
-	DPRINTF("cda_stw_exc: o_len=%lu fault_addr=%lu o_dst=%lu ret=%lu\n",
-		o_len, d->fault_addr, o_dst, o_len - d->fault_addr + o_dst);
-	return o_len * 4 - d->fault_addr + o_dst;
+	return PA_MEMCPY_STORE_ERROR;
 }
 
 
-/* Returns 0 for success, otherwise, returns number of bytes not transferred. */
-static unsigned long pa_memcpy(void *dstp, const void *srcp, unsigned long len)
+/* Returns PA_MEMCPY_OK, PA_MEMCPY_LOAD_ERROR or PA_MEMCPY_STORE_ERROR.
+ * In case of an access fault the faulty address can be read from the per_cpu
+ * exception data struct. */
+static unsigned long pa_memcpy_internal(void *dstp, const void *srcp,
+					unsigned long len)
 {
 	register unsigned long src, dst, t1, t2, t3;
 	register unsigned char *pcs, *pcd;
 	register unsigned int *pws, *pwd;
 	register double *pds, *pdd;
-	unsigned long ret = 0;
-	unsigned long o_dst, o_src, o_len;
-	struct exception_data *d;
+	unsigned long ret;
 
 	src = (unsigned long)srcp;
 	dst = (unsigned long)dstp;
 	pcs = (unsigned char *)srcp;
 	pcd = (unsigned char *)dstp;
 
-	o_dst = dst; o_src = src; o_len = len;
-
 	/* prefetch_src((const void *)srcp); */
 
 	if (len < THRESHOLD)
@@ -401,7 +399,7 @@ static unsigned long pa_memcpy(void *dstp, const void *srcp, unsigned long len)
 		len--;
 	}
 
-	return 0;
+	return PA_MEMCPY_OK;
 
 unaligned_copy:
 	/* possibly we are aligned on a word, but not on a double... */
@@ -438,8 +436,7 @@ static unsigned long pa_memcpy(void *dstp, const void *srcp, unsigned long len)
 		src = (unsigned long)pcs;
 	}
 
-	ret = copy_dstaligned(dst, src, len / sizeof(unsigned int), 
-		o_dst, o_src, o_len);
+	ret = copy_dstaligned(dst, src, len / sizeof(unsigned int));
 	if (ret)
 		return ret;
 
@@ -454,17 +451,41 @@ static unsigned long pa_memcpy(void *dstp, const void *srcp, unsigned long len)
 
 handle_load_error:
 	__asm__ __volatile__ ("pmc_load_exc:\n");
-	d = &__get_cpu_var(exception_data);
-	DPRINTF("pmc_load_exc: o_len=%lu fault_addr=%lu o_src=%lu ret=%lu\n",
-		o_len, d->fault_addr, o_src, o_len - d->fault_addr + o_src);
-	return o_len - d->fault_addr + o_src;
+	return PA_MEMCPY_LOAD_ERROR;
 
 handle_store_error:
 	__asm__ __volatile__ ("pmc_store_exc:\n");
+	return PA_MEMCPY_STORE_ERROR;
+}
+
+
+/* Returns 0 for success, otherwise, returns number of bytes not transferred. */
+static unsigned long pa_memcpy(void *dstp, const void *srcp, unsigned long len)
+{
+	unsigned long ret, fault_addr, reference;
+	struct exception_data *d;
+
+	ret = pa_memcpy_internal(dstp, srcp, len);
+	if (likely(ret == PA_MEMCPY_OK))
+		return 0;
+
+	/* if a load or store fault occured we can get the faulty addr */
 	d = &__get_cpu_var(exception_data);
-	DPRINTF("pmc_store_exc: o_len=%lu fault_addr=%lu o_dst=%lu ret=%lu\n",
-		o_len, d->fault_addr, o_dst, o_len - d->fault_addr + o_dst);
-	return o_len - d->fault_addr + o_dst;
+	fault_addr = d->fault_addr;
+
+	/* error in load or store? */
+	if (ret == PA_MEMCPY_LOAD_ERROR)
+		reference = (unsigned long) srcp;
+	else
+		reference = (unsigned long) dstp;
+
+	DPRINTF("pa_memcpy: fault type = %lu, len=%lu fault_addr=%lu ref=%lu\n",
+		ret, len, fault_addr, reference);
+
+	if (fault_addr >= reference)
+		return len - (fault_addr - reference);
+	else
+		return len;
 }
 
 #ifdef __KERNEL__