Skip to contentSkip to conversation

Fix wrong memory limit calculation in memory-test

PATCH Superseded Guennadi Liakhovetski 3 participants 4 messages Feb 7, 2008 – Feb 14, 2008

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Fix wrong memory limit calculation in memory-test

If the length of the memory address range passed to the "mtest" command is not of the form 2^x - 1, not all address lines are tested. This bug is inherited from the original software at
http://www.netrino.com/Embedded-Systems/How-To/Memory-Test-Suite-C. Fix this.

---

common/cmd_mem.c +8 -8
index a994211..2b55e7e 100644
@@ -661,7 +661,7 @@ int do_mem_mtest (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
 	ulong	readback;
 
 #if defined(CFG_ALT_MEMTEST)
-	vu_long	addr_mask;
+	vu_long	len;
 	vu_long	offset;
 	vu_long	test_offset;
 	vu_long	pattern;
@@ -800,24 +800,24 @@ int do_mem_mtest (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
 		 * Returns:     0 if the test succeeds, 1 if the test fails.
 		 *
 		 * ## NOTE ##	Be sure to specify start and end
-		 *              addresses such that addr_mask has
+		 *              addresses such that len has
 		 *              lots of bits set. For example an
 		 *              address range of 01000000 02000000 is
 		 *              bad while a range of 01000000
 		 *              01ffffff is perfect.
 		 */
-		addr_mask = ((ulong)end - (ulong)start)/sizeof(vu_long);
+		len = ((ulong)end - (ulong)start)/sizeof(vu_long);
 		pattern = (vu_long) 0xaaaaaaaa;
 		anti_pattern = (vu_long) 0x55555555;
 
 		PRINTF("%s:%d: addr mask = 0x%.8lx\n",
 			__FUNCTION__, __LINE__,
-			addr_mask);
+			len);
 		/*
 		 * Write the default pattern at each of the
 		 * power-of-two offsets.
 		 */
-		for (offset = 1; (offset & addr_mask) != 0; offset <<= 1) {
+		for (offset = 1; offset < len; offset <<= 1) {
 			start[offset] = pattern;
 		}
 
@@ -827,7 +827,7 @@ int do_mem_mtest (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
 		test_offset = 0;
 		start[test_offset] = anti_pattern;
 
-		for (offset = 1; (offset & addr_mask) != 0; offset <<= 1) {
+		for (offset = 1; offset < len; offset <<= 1) {
 		    temp = start[offset];
 		    if (temp != pattern) {
 			printf ("\nFAILURE: Address bit stuck high @ 0x%.8lx:"
@@ -841,10 +841,10 @@ int do_mem_mtest (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
 		/*
 		 * Check for addr bits stuck low or shorted.
 		 */
-		for (test_offset = 1; (test_offset & addr_mask) != 0; test_offset <<= 1) {
+		for (test_offset = 1; test_offset < len; test_offset <<= 1) {
 		    start[test_offset] = anti_pattern;
 
-		    for (offset = 1; (offset & addr_mask) != 0; offset <<= 1) {
+		    for (offset = 1; offset < len; offset <<= 1) {
 			temp = start[offset];
 			if ((temp != pattern) && (offset != test_offset)) {
 			    printf ("\nFAILURE: Address bit stuck low or shorted @"
#1 lore raw

PATCH 1/1 Fix wrong memory limit calculation in memory-test

Quoted 24 lines · Guennadi Liakhovetski

If the length of the memory address range passed to the "mtest" command is not of the form 2^x - 1, not all address lines are tested. This bug is inherited from the original software at
http://www.netrino.com/Embedded-Systems/How-To/Memory-Test-Suite-C. Fix this.

---

common/cmd_mem.c +1 -1
index a994211..2b55e7e 100644
@@ -661,7 +661,7 @@ int do_mem_mtest (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
 	ulong	readback;
 
 #if defined(CFG_ALT_MEMTEST)
-	vu_long	addr_mask;
+	vu_long	len;
 	vu_long	offset;
 	vu_long	test_offset;
 	vu_long	pattern;

common/cmd_mem.c int do_mem_mtest (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[]) · line 801

Quoted 2 lines
common/cmd_mem.c +0 -0
@@ -800,24 +800,24 @@ int do_mem_mtest (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
 		 * Returns:     0 if the test succeeds, 1 if the test fails.
 		 *

The "## NOTE ##" doesn't really apply any more. Besides that, generally "start" and "end" are constrained by the system (hardware and software configuration) so "picking" them is a nebulous concept. I would simply delete the note.

Quoted 4 lines
common/cmd_mem.c +1 -1
 		 * ## NOTE ##	Be sure to specify start and end
-		 *              addresses such that addr_mask has
+		 *              addresses such that len has
 		 *              lots of bits set. For example an

common/cmd_mem.c int do_mem_mtest (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[]) · line 813

Quoted 10 lines
common/cmd_mem.c +1 -1
 		 *              address range of 01000000 02000000 is
 		 *              bad while a range of 01000000
 		 *              01ffffff is perfect.
 		 */
-		addr_mask = ((ulong)end - (ulong)start)/sizeof(vu_long);
+		len = ((ulong)end - (ulong)start)/sizeof(vu_long);
 		pattern = (vu_long) 0xaaaaaaaa;
 		anti_pattern = (vu_long) 0x55555555;
 
 		PRINTF("%s:%d: addr mask = 0x%.8lx\n",

s/addr mask/len/ in the printf() string

common/cmd_mem.c int do_mem_mtest (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[]) · line 815

Quoted 3 lines
common/cmd_mem.c +1 -1
 			__FUNCTION__, __LINE__,
-			addr_mask);
+			len);

Otherwise, good!

Thanks,
gvb

#2 lore raw

PATCH v2 1/1 Fix wrong memory limit calculation in memory-test

If the length of the memory address range passed to the "mtest" command is not of the form 2^x - 1, not all address lines are tested. This bug is inherited from the original software at
http://www.netrino.com/Embedded-Systems/How-To/Memory-Test-Suite-C. Fix this.

---

This is version 2 of the yesterday's patch, further taking into account suggestions by Jerry Van Baren.

Thanks
Guennadi

common/cmd_mem.c +8 -15
index a994211..7569851 100644
@@ -661,7 +661,7 @@ int do_mem_mtest (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
 	ulong	readback;
 
 #if defined(CFG_ALT_MEMTEST)
-	vu_long	addr_mask;
+	vu_long	len;
 	vu_long	offset;
 	vu_long	test_offset;
 	vu_long	pattern;
@@ -798,26 +798,19 @@ int do_mem_mtest (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
 		 *              all possible.
 		 *
 		 * Returns:     0 if the test succeeds, 1 if the test fails.
-		 *
-		 * ## NOTE ##	Be sure to specify start and end
-		 *              addresses such that addr_mask has
-		 *              lots of bits set. For example an
-		 *              address range of 01000000 02000000 is
-		 *              bad while a range of 01000000
-		 *              01ffffff is perfect.
 		 */
-		addr_mask = ((ulong)end - (ulong)start)/sizeof(vu_long);
+		len = ((ulong)end - (ulong)start)/sizeof(vu_long);
 		pattern = (vu_long) 0xaaaaaaaa;
 		anti_pattern = (vu_long) 0x55555555;
 
-		PRINTF("%s:%d: addr mask = 0x%.8lx\n",
+		PRINTF("%s:%d: length = 0x%.8lx\n",
 			__FUNCTION__, __LINE__,
-			addr_mask);
+			len);
 		/*
 		 * Write the default pattern at each of the
 		 * power-of-two offsets.
 		 */
-		for (offset = 1; (offset & addr_mask) != 0; offset <<= 1) {
+		for (offset = 1; offset < len; offset <<= 1) {
 			start[offset] = pattern;
 		}
 
@@ -827,7 +820,7 @@ int do_mem_mtest (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
 		test_offset = 0;
 		start[test_offset] = anti_pattern;
 
-		for (offset = 1; (offset & addr_mask) != 0; offset <<= 1) {
+		for (offset = 1; offset < len; offset <<= 1) {
 		    temp = start[offset];
 		    if (temp != pattern) {
 			printf ("\nFAILURE: Address bit stuck high @ 0x%.8lx:"
@@ -841,10 +834,10 @@ int do_mem_mtest (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
 		/*
 		 * Check for addr bits stuck low or shorted.
 		 */
-		for (test_offset = 1; (test_offset & addr_mask) != 0; test_offset <<= 1) {
+		for (test_offset = 1; test_offset < len; test_offset <<= 1) {
 		    start[test_offset] = anti_pattern;
 
-		    for (offset = 1; (offset & addr_mask) != 0; offset <<= 1) {
+		    for (offset = 1; offset < len; offset <<= 1) {
 			temp = start[offset];
 			if ((temp != pattern) && (offset != test_offset)) {
 			    printf ("\nFAILURE: Address bit stuck low or shorted @"
#3 lore raw

PATCH v2 1/1 Fix wrong memory limit calculation in memory-test

Quoted 7 lines · In message <[email protected]> you wrote:

If the length of the memory address range passed to the "mtest" command is not of the form 2^x - 1, not all address lines are tested. This bug is inherited from the original software at
http://www.netrino.com/Embedded-Systems/How-To/Memory-Test-Suite-C. Fix this.

Applied, thanks.

Best regards,

Wolfgang Denk

Signature · 6 lines
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
I wish Captain Vimes were here. He wouldn't have  known  what  to  do
either, but he's got a much better vocabulary to be baffled in.
                                 - Terry Pratchett, _Guards! Guards!_