[flashrom] [PATCH] Generate a test pattern

Carl-Daniel Hailfinger c-d.hailfinger.devel.2006 at gmx.net
Sun Nov 22 02:07:50 CET 2009


New version.

Add the ability to generate test patterns for write testing. This will
be useful once we create a --test function for flashrom.

The test patterns make it easy to find skipped and duplicated bytes, are
human readable, and the first 8 of them have block numbers to detect
aliasing or wraparounds. Current size limit for aliasing detection
is 16 MByte, but since neither LPC nor FWH nor SPI chips exist with
bigger sizes, this is reasonably safe.


Detailed documentation is available as source code comments above the
new function generate_testpattern().

Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006 at gmx.net>

Index: flashrom-testpattern/flashrom.c
===================================================================
--- flashrom-testpattern/flashrom.c	(Revision 768)
+++ flashrom-testpattern/flashrom.c	(Arbeitskopie)
@@ -407,6 +407,142 @@
 	return ret;
 }
 
+/* This function generates various test patterns useful for testing controller
+ * and chip communication as well as chip behaviour.
+ *
+ * If a byte can be written multiple times, each time keeping 0-bits at 0
+ * and changing 1-bits to 0 if the new value for that bit is 0, the effect
+ * is essentially an AND operation. That's also the reason why this function
+ * provides the result of AND between various patterns.
+ *
+ * Below is a list of patterns (and their block length).
+ * Pattern 0 is 05 15 25 35 45 55 65 75 85 95 a5 b5 c5 d5 e5 f5 (16 Bytes)
+ * Pattern 1 is 0a 1a 2a 3a 4a 5a 6a 7a 8a 9a aa ba ca da ea fa (16 Bytes)
+ * Pattern 2 is 50 51 52 53 54 55 56 57 58 59 5a 5b 5c 5d 5e 5f (16 Bytes)
+ * Pattern 3 is a0 a1 a2 a3 a4 a5 a6 a7 a8 a9 aa ab ac ad ae af (16 Bytes)
+ * Pattern 4 is 00 10 20 30 40 50 60 70 80 90 a0 b0 c0 d0 e0 f0 (16 Bytes)
+ * Pattern 5 is 00 01 02 03 04 05 06 07 08 09 0a 0b 0c 0d 0e 0f (16 Bytes)
+ * Pattern 6 is 00 (1 Byte)
+ * Pattern 7 is ff (1 Byte)
+ * Patterns 0-7 have a big-endian block number in the last 2 bytes of each 256
+ * byte block.
+ *
+ * Pattern 8 is 00 01 02 03 04 05 06 07 08 09 0a 0b 0c 0d 0e 0f 10 11... (256 B)
+ * Pattern 9 is ff fe fd fc fb fa f9 f8 f7 f6 f5 f4 f3 f2 f1 f0 ef ee... (256 B)
+ * Pattern 10 is 00 00 00 01 00 02 00 03 00 04... (128 kB big-endian counter)
+ * Pattern 11 is ff ff ff fe ff fd ff fc ff fb... (128 kB big-endian downwards)
+ * Pattern 12 is 00 (1 Byte)
+ * Pattern 13 is ff (1 Byte)
+ * Patterns 8-13 have no block number.
+ *
+ * Patterns 0-3 are created to detect and efficiently diagnose communication
+ * slips like missed bits or bytes and their repetitive nature gives good visual
+ * cues to the person inspecting the results. In addition, the following holds:
+ * AND Pattern 0/1 == Pattern 4
+ * AND Pattern 2/3 == Pattern 5
+ * AND Pattern 0/1/2/3 == AND Pattern 4/5 == Pattern 6
+ * A weakness of pattern 0-5 is the inability to detect swaps/copies between
+ * any two 16-byte blocks except for the last 16-byte block in a 256-byte bloc.
+ * They work perfectly for detecting any swaps/aliasing of blocks >= 256 bytes.
+ * 0x5 and 0xa were picked because they are 0101 and 1010 binary.
+ * Patterns 8-9 are best for detecting swaps/aliasing of blocks < 256 bytes.
+ * Besides that, they provide for bit testing of the last two bytes of every
+ * 256 byte block which contains the block number for patterns 0-6.
+ * Patterns 10-11 are special purpose for detecting subblock aliasing with
+ * block sizes >256 bytes (some Dataflash chips etc.)
+ * AND Pattern 8/9 == Pattern 12
+ * AND Pattern 10/11 == Pattern 12
+ * Pattern 13 is the completely erased state.
+ * None of the patterns can detect aliasing at boundaries which are a multiple
+ * of 16 MBytes (but such chips do not exist anyway for Parallel/LPC/FWH/SPI).
+ */
+int generate_testpattern(uint8_t *buf, uint32_t size, int variant)
+{
+	int i;
+
+	if (!buf) {
+		fprintf(stderr, "Invalid buffer!\n");
+		return 1;
+	}
+
+	switch (variant) {
+	case 0:
+		for (i = 0; i < size; i++)
+			buf[i] = (i & 0xf) << 4 | 0x5;
+		break;
+	case 1:
+		for (i = 0; i < size; i++)
+			buf[i] = (i & 0xf) << 4 | 0xa;
+		break;
+	case 2:
+		for (i = 0; i < size; i++)
+			buf[i] = 0x50 | (i & 0xf);
+		break;
+	case 3:
+		for (i = 0; i < size; i++)
+			buf[i] = 0xa0 | (i & 0xf);
+		break;
+	case 4:
+		for (i = 0; i < size; i++)
+			buf[i] = (i & 0xf) << 4;
+		break;
+	case 5:
+		for (i = 0; i < size; i++)
+			buf[i] = i & 0xf;
+		break;
+	case 6:
+		memset(buf, 0x00, size);
+		break;
+	case 7:
+		memset(buf, 0xff, size);
+		break;
+	case 8:
+		for (i = 0; i < size; i++)
+			buf[i] = i & 0xff;
+		break;
+	case 9:
+		for (i = 0; i < size; i++)
+			buf[i] = ~(i & 0xff);
+		break;
+	case 10:
+		for (i = 0; i < size % 2; i++) {
+			buf[i * 2] = (i >> 8) & 0xff;
+			buf[i * 2 + 1] = i & 0xff;
+		}
+		if (size & 0x1)
+			buf[i * 2] = (i >> 8) & 0xff;
+		break;
+	case 11:
+		for (i = 0; i < size % 2; i++) {
+			buf[i * 2] = ~((i >> 8) & 0xff);
+			buf[i * 2 + 1] = ~(i & 0xff);
+		}
+		if (size & 0x1)
+			buf[i * 2] = ~((i >> 8) & 0xff);
+		break;
+	case 12:
+		memset(buf, 0x00, size);
+		break;
+	case 13:
+		memset(buf, 0xff, size);
+		break;
+	}
+
+	if ((variant >= 0) && (variant <= 7)) {
+		/* Write block number in the last two bytes of each 256-byte
+		 * block, big endian for easier reading of the hexdump.
+		 * Note that this wraps around for chips larger than 2^24 bytes
+		 * (16 MB).
+		 */
+		for (i = 0; i < size / 256; i++) {
+			buf[i * 256 + 254] = (i >> 8) & 0xff;
+			buf[i * 256 + 255] = i & 0xff;
+		}
+	}
+
+	return 0;
+}
+
 int check_max_decode(enum chipbustype buses, uint32_t size)
 {
 	int limitexceeded = 0;


-- 
Developer quote of the month: 
"We are juggling too many chainsaws and flaming arrows and tigers."





More information about the flashrom mailing list