Skip to contentSkip to conversation

mtd: nand: omap: optimize and clean-up of OMAP NAND driver

PATCH Superseded Pekon Gupta 5 participants 19 messages Sep 30, 2013 – Oct 17, 2013

Subsystems (3)

v7 of 11 Superseded Latest is v11 →

lore
Share & tools
lore
#0 lore raw

v7 cover mtd: nand: omap: optimize and clean-up of OMAP NAND driver

*changes in v7*
[PATCH 1/5]
- omap_gpmc.c: fix: free bytes in OOB (ecclayout->oobfree[0].length)
- omap_gpmc.c: cleanup: redundant code added in previous patch versions
- am335x_evm.h: cleanup: redundant code added in previous patch versions
- tricorder.h: fix: CONFIG_NAND_OMAP_ECCSCHEME
[PATCH 2/5] removed: re-configuration of gpmc.config1[dev_width] added in
previous version of patch
[PATCH 3/5] <no update>
[PATCH 4/5] <no update>
[PATCH 5/5] minor fix: missing '$' in ${loadaddr}

*changes in v6*
[PATCH 1/5] incorporated feedbacks from Scott Wood <[email protected]>
- renamed CONFIG_SYS_NAND_ECCSCHEME to CONFIG_NAND_OMAP_ECCSCHEME
- updated omap_select_ecc_scheme() to handle error conditions without depending on caller.
- renamed OMAP_ECC_HAM1_CODE_HW_ROMCODE to OMAP_ECC_HAM1_CODE_HW to keep it naming compatible to linux kernel
- updated doc/README.nand and doc/README.omap3
[PATCH 2/5] minor code clean-up
[PATCH 3/5] minor code clean-up
[PATCH 4/5] <no update>
[PATCH 5/5] <moved CONFIG_NAND_OMAP_ECCSCHEME description to README.nand>

*changes in v5*
This version of patch is tested on am335x-evm with x8 NAND device, and boots SPL and u-boot from NAND
[PATCH 1/5]
- re-added omap_read_page_bch(): needed proper sequence of while reading DATA and ECC from NAND page, so that calc_ecc generated from GPMC is understood by ELM.
- added check to see if NAND OOB can accomodate ECC for entire page
[PATCH 2/5] fixed device-width in GPMC_CONFIG1_X to support x16 devices [PATCH 3/5] code clean-up for OMAP_ECC_BCH8_CODE_HW_DETECTION_SW mode [PATCH 4/5]
- fixed omap_correct_data_bch() for correcting bit-flips using ELM
- code-cleanup + added omap_reverse_list()
[PATCH 5/5] incorporated feedbacks from Peter Korsgaard <[email protected]>

*changes in v4*
[PATCH 1/5]
- removed omap_read_page_bch(): chip->ecc.read_page uses default API nand_read_page_hwecc() in nand_base.c
- updated tricorder.h: added new CONFIGS for ECCSCHEME & ONFI_DETECTION
- converted printf("ECC-SCHEME") to debug("ECC-SCHEME")
[PATCH 2/5] minor code clean-up
[PATCH 3/5] <no update>
[PATCH 4/5] <no update>
[PATCH 5/5] updated README as per feedbacks from trini at ti.com

*changes in v3*
[PATCH 1/5] (complete change)
- ecc-scheme is selection is controller by s/w, not CONFIG_NAND_xx
- added omap_select_ecc_scheme(), as common function to handle all ecc-scheme related configurations for both board_nand_init() & omap_nand_switch_ecc().
- removed un-used defines from asm/arch-am33xx/omap_gpmc.h
- updated doc/REAME.nand
[PATCH 2/5] removed un-used defines from asm/omap_gpmc.h [PATCH 3/5] removed omap_calculate_ecc_bch_sw() and omap_calculate_ecc()
and merged their logic into omap_calculate_ecc_bch()
[PATCH 4/5] updated return error-code in-case of invalid conditions [PATCH 5/5] (new) for board/ti/am335x/README

*changes in v2*
- added documentation for CONFIG_NAND_OMAP_xx in doc/README.nand
- added CONFIG_BCH along with CONFIG_NAND_OMAP_ECC_BCH8_CODE_HW_DETECTION_SW to include software library lib/bch.c
- fixed board_nand_init() and omap_enable_hwecc()

*original v1*
This patch series updates BCH8_ECC schemes in mtd/nand/omap_gpmc.c driver
- adds scalability for higher ECC schemes in future.
- removes CONFIG_AM335x and it makes it generic for all platforms.
- optimizes read_data paths

Pekon Gupta (5):
[PATCH 1/5] mtd: nand: omap: enable BCH ECC scheme using ELM for generic platform [PATCH 2/5] mtd: nand: omap: optimize chip->ecc.hwctl() for H/W ECC schemes [PATCH 3/5] mtd: nand: omap: optimize chip->ecc.calculate() for H/W ECC schemes [PATCH 4/5] mtd: nand: omap: optimized chip->ecc.correct() for H/W ECC schemes [PATCH 5/5] board/ti/am335x/README: update for NAND boot

 arch/arm/include/asm/arch-am33xx/omap_gpmc.h |  52 +-
 arch/arm/include/asm/omap_gpmc.h             |   7 -
 board/ti/am335x/README                       |  53 +-
 doc/README.nand                              |  30 +
 doc/README.omap3                             |   4 +-
 drivers/mtd/nand/omap_gpmc.c                 | 822 +++++++++++----------------
 include/configs/am335x_evm.h                 |  11 +-
 include/configs/tricorder.h                  |   3 +-
 8 files changed, 416 insertions(+), 566 deletions(-)
Signature · 1 line
1.8.1
#1 lore raw

PATCH v7 1/5 mtd: nand: omap: enable BCH ECC scheme using ELM for generic platform

BCH8_ECC scheme implemented in omap_gpmc.c driver has following favours

+-----------------------------------+-----------------+-----------------+
|ECC Scheme                         | ECC Calculation | Error Detection |
+-----------------------------------+-----------------+-----------------+
|OMAP_ECC_BCH8_CODE_HW              |GPMC             |ELM H/W engine   |
|OMAP_ECC_BCH8_CODE_HW_DETECTION_SW |GPMC             |S/W BCH library  |
+-----------------------------------+-----------------+-----------------+

Current implementation limits the BCH8_CODE_HW only for AM33xx device family. (using CONFIG_AM33XX). However, other SoC families (like TI81xx) also have ELM hardware module, and can support ECC error detection using ELM.

This patch
- removes CONFIG_AM33xx
Thus this driver can be reused by all devices having ELM h/w engine.
- removes CONFIG_NAND_OMAP_BCH8 and adds new CONFIG_NAND_OMAP_ECCSCHEME CONFIG_NAND_OMAP_ECCSCHEME determines ecc-scheme to be used refer doc/README.nand and board/ti/am335x/README
- adds omap_select_ecc_scheme()
A common function to handle ecc-scheme related configurations. This can be used both during device-probe and via user-space u-boot commads to change ecc-scheme. During device probe ecc-scheme is selected based on CONFIG_NAND_OMAP_ECCSCHEME.
- enables CONFIG_BCH
S/W library (lib/bch.c) required by OMAP_ECC_BCH8_CODE_HW_DETECTION_SW is enabled by CONFIG_BCH.
- enables CONFIG_SYS_NAND_ONFI_DETECTION
for auto-detection of ONFI compliant NAND devices
- updates following README doc

	doc/README.nand
	board/ti/am335x/README
	doc/README.omap3
---
 arch/arm/include/asm/arch-am33xx/omap_gpmc.h |  52 ++---
 doc/README.nand                              |  30 +++
 doc/README.omap3                             |   4 +-
 drivers/mtd/nand/omap_gpmc.c                 | 287 +++++++++++++++------------
 include/configs/am335x_evm.h                 |  11 +-
 include/configs/tricorder.h                  |   3 +-
 6 files changed, 216 insertions(+), 171 deletions(-)
arch/arm/include/asm/arch-am33xx/omap_gpmc.h +15 -37
index 00ad1d0..62f608e 100644
@@ -7,42 +7,20 @@
 #ifndef __ASM_ARCH_OMAP_GPMC_H
 #define __ASM_ARCH_OMAP_GPMC_H
 
-/* These GPMC_NAND_HW_BCHx_ECC_LAYOUT defines are based on AM33xx ELM */
-#define GPMC_NAND_HW_BCH4_ECC_LAYOUT {\
-	.eccbytes = 32,\
-	.eccpos = {2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,\
-				16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27,\
-				28, 29, 30, 31, 32, 33},\
-	.oobfree = {\
-		{.offset = 34,\
-		 .length = 30 } } \
-}
+enum omap_ecc {
+	/* 1-bit  ECC calculation by Software, Error detection by Software */
+	OMAP_ECC_HAM1_CODE_SW = 1, /* avoid un-initialized int can be 0x0 */
+	/* 1-bit  ECC calculation by GPMC, Error detection by Software */
+	/* ECC layout compatible to legacy ROMCODE. */
+	OMAP_ECC_HAM1_CODE_HW,
+	/* 4-bit  ECC calculation by GPMC, Error detection by Software */
+	OMAP_ECC_BCH4_CODE_HW_DETECTION_SW,
+	/* 4-bit  ECC calculation by GPMC, Error detection by ELM */
+	OMAP_ECC_BCH4_CODE_HW,
+	/* 8-bit  ECC calculation by GPMC, Error detection by Software */
+	OMAP_ECC_BCH8_CODE_HW_DETECTION_SW,
+	/* 8-bit  ECC calculation by GPMC, Error detection by ELM */
+	OMAP_ECC_BCH8_CODE_HW,
+};
 
-#define GPMC_NAND_HW_BCH8_ECC_LAYOUT {\
-	.eccbytes = 56,\
-	.eccpos = {2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,\
-				16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27,\
-				28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39,\
-				40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51,\
-				52, 53, 54, 55, 56, 57},\
-	.oobfree = {\
-		{.offset = 58,\
-		 .length = 6 } } \
-}
-
-#define GPMC_NAND_HW_BCH16_ECC_LAYOUT {\
-	.eccbytes = 104,\
-	.eccpos = {2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,\
-				16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27,\
-				28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39,\
-				40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51,\
-				52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63,\
-				64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75,\
-				76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87,\
-				88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99,\
-				100, 101, 102, 103, 104, 105},\
-	.oobfree = {\
-		{.offset = 106,\
-		 .length = 8 } } \
-}
 #endif /* __ASM_ARCH_OMAP_GPMC_H */
doc/README.nand +30 -0
index 913e9b5..f9c921e 100644
@@ -169,6 +169,36 @@ Configuration Options:
       Please convert your driver even if you don't need the extra
       flexibility, so that one day we can eliminate the old mechanism.
 
+   CONFIG_SYS_NAND_ONFI_DETECTION
+	Enables detection of ONFI compliant devices during probe.
+	And fetching device parameters flashed on device, by parsing
+	ONFI parameter page.
+
+   CONFIG_BCH
+	Enables software based BCH ECC algorithm present in lib/bch.c
+	This is used by SoC platforms which do not have built-in ELM
+	hardware engine required for BCH ECC correction.
+
+Platform specific options
+=========================
+
+   CONFIG_NAND_OMAP_ECCSCHEME
+	On OMAP platforms, this specifies NAND ECC scheme.
+	1 - HAM1_SW	1-bit Hamming code using software library
+			(for legacy devices only)
+	2 - HAM1_HW	1-bit Hamming code using GPMC hardware engine
+			(for legacy devices only)
+	3 - BCH4_SW	4-bit BCH code (unsupported)
+	4 - BCH4_HW	4-bit BCH code (unsupported)
+	5 - BCH8_SW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using software library.
+			- requires CONFIG_BCH to enable software BCH library
+			(For legacy device which do not have ELM h/w engine)
+	6 - BCH8_HW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using ELM hardware engine.
+
 NOTE:
 =====
 
doc/README.omap3 +2 -2
index 1fbe79d..fd85b22 100644
@@ -161,8 +161,8 @@ BCH8
 
 To enable hardware assisted BCH8 (8-bit BCH [Bose, Chaudhuri, Hocquenghem]) on
 OMAP3 devices we can use the BCH library in lib/bch.c. To do so add CONFIG_BCH
-to enable the library and CONFIG_NAND_OMAP_BCH8 to to enable hardware assisted
-syndrom generation to your board config.
+to enable the library and set CONFIG_NAND_OMAP_ECCSCHEME=5 (as per README.nand)
+to enable BCH8_SW ecc scheme.
 The NAND OOB layout is the same as in linux kernel, if the linux kernel BCH8
 implementation for OMAP3 works for you so the u-boot version should also.
 When you require the SPL to read with BCH8 there are two more configs to
drivers/mtd/nand/omap_gpmc.c +162 -125

411 lines not on this page · raw message

include/configs/am335x_evm.h +5 -6
index 3de30fc..994e235 100644
@@ -216,11 +216,11 @@
 
 #ifdef CONFIG_NAND
 #define CONFIG_SYS_NAND_5_ADDR_CYCLE
-#define CONFIG_SYS_NAND_PAGE_COUNT	(CONFIG_SYS_NAND_BLOCK_SIZE / \
-					 CONFIG_SYS_NAND_PAGE_SIZE)
+#define CONFIG_SYS_NAND_BLOCK_SIZE	131072
 #define CONFIG_SYS_NAND_PAGE_SIZE	2048
 #define CONFIG_SYS_NAND_OOBSIZE		64
-#define CONFIG_SYS_NAND_BLOCK_SIZE	(128*1024)
+#define CONFIG_SYS_NAND_PAGE_COUNT	(CONFIG_SYS_NAND_BLOCK_SIZE / \
+					 CONFIG_SYS_NAND_PAGE_SIZE)
 #define CONFIG_SYS_NAND_BAD_BLOCK_POS	NAND_LARGE_BADBLOCK_POS
 #define CONFIG_SYS_NAND_ECCPOS		{ 2, 3, 4, 5, 6, 7, 8, 9, \
 					 10, 11, 12, 13, 14, 15, 16, 17, \
@@ -232,9 +232,8 @@
 
 #define CONFIG_SYS_NAND_ECCSIZE		512
 #define CONFIG_SYS_NAND_ECCBYTES	14
-
-#define CONFIG_SYS_NAND_U_BOOT_START	CONFIG_SYS_TEXT_BASE
-#define CONFIG_SYS_NAND_U_BOOT_OFFS	0x80000
+#define CONFIG_SYS_NAND_ONFI_DETECTION
+#define CONFIG_NAND_OMAP_ECCSCHEME	6
 #endif
 #endif
 
include/configs/tricorder.h +2 -1
index a9b2714..27db618 100644
@@ -110,7 +110,6 @@
 
 #define CONFIG_SYS_MAX_NAND_DEVICE	1		/* Max number of NAND */
 							/* devices */
-#define CONFIG_NAND_OMAP_BCH8
 #define CONFIG_BCH
 
 /* commands to include */
@@ -298,6 +297,8 @@
 
 #define CONFIG_SYS_NAND_ECCSIZE		512
 #define CONFIG_SYS_NAND_ECCBYTES	13
+#define CONFIG_NAND_OMAP_ECCSCHEME	5
+#define CONFIG_SYS_NAND_ONFI_DETECTION
 
 #define CONFIG_SYS_NAND_U_BOOT_START	CONFIG_SYS_TEXT_BASE
 
Signature · 1 line
1.8.1
#2 lore raw

PATCH v7 1/5 mtd: nand: omap: enable BCH ECC scheme using ELM for generic platform

Hi Pekon,

Quoted 8 lines · Pekon Gupta

BCH8_ECC scheme implemented in omap_gpmc.c driver has following favours

+-----------------------------------+-----------------+-----------------+
|ECC Scheme                         | ECC Calculation | Error Detection |
+-----------------------------------+-----------------+-----------------+
|OMAP_ECC_BCH8_CODE_HW              |GPMC             |ELM H/W engine   |
|OMAP_ECC_BCH8_CODE_HW_DETECTION_SW |GPMC             |S/W BCH library  |
+-----------------------------------+-----------------+-----------------+
Quoted 7 lines

Current implementation limits the BCH8_CODE_HW only for AM33xx device family. (using CONFIG_AM33XX). However, other SoC families (like TI81xx) also have ELM hardware module, and can support ECC error detection using ELM.

This patch
- removes CONFIG_AM33xx
Thus this driver can be reused by all devices having ELM h/w engine.

Do you have an overview, which TI devices actually have this ELM h/w engine integrated? Especially if its integrated in the OMAP3530 SoC? I'm currently scanning the users manual and can't find any reference to this IP block.

Thanks,
Stefan

#3 lore raw

PATCH v7 1/5 mtd: nand: omap: enable BCH ECC scheme using ELM for generic platform

In reply to Pekon Gupta
Quoted 9 lines · Pekon Gupta
doc/README.nand +9 -0
+Platform specific options
+=========================
+
+   CONFIG_NAND_OMAP_ECCSCHEME
+	On OMAP platforms, this specifies NAND ECC scheme.
+	1 - HAM1_SW	1-bit Hamming code using software library
+			(for legacy devices only)
+	2 - HAM1_HW	1-bit Hamming code using GPMC hardware engine
+			(for legacy devices only)

doc/README.nand Configuration Options: · line 200

Quoted 10 lines
doc/README.nand +10 -0
+	3 - BCH4_SW	4-bit BCH code (unsupported)
+	4 - BCH4_HW	4-bit BCH code (unsupported)
+	5 - BCH8_SW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using software library.
+			- requires CONFIG_BCH to enable software BCH library
+			(For legacy device which do not have ELM h/w engine)
+	6 - BCH8_HW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using ELM hardware engine.

You should document the symbols, not the numbers that happen to be assigned to them.

drivers/mtd/nand/omap_gpmc.c static void __maybe_unused omap_free_bch(struct mtd_info *mtd) · line 760

Quoted 9 lines
drivers/mtd/nand/omap_gpmc.c +9 -0
+/**
+ * omap_select_ecc_scheme - configures driver for particular ecc-scheme
+ * @nand: NAND chip device structure
+ * @ecc_scheme: ecc scheme to configure
+ * @pagesize: number of main-area bytes per page of NAND device
+ * @oobsize: number of OOB/spare bytes per page of NAND device
+ */
+static int omap_select_ecc_scheme(struct nand_chip *nand, int ecc_scheme,
+		unsigned int pagesize, unsigned int oobsize) {

s/int ecc_scheme/enum omap_ecc ecc_scheme/

Quoted 86 lines
drivers/mtd/nand/omap_gpmc.c +86 -0
+	struct nand_bch_priv	*bch		= nand->priv;
+	struct nand_ecclayout	*ecclayout	= nand->ecc.layout;
+	int i;
+
+	switch (ecc_scheme) {
+	case OMAP_ECC_HAM1_CODE_SW:
+		debug("nand: selected OMAP_ECC_HAM1_CODE_SW\n");
+		nand->ecc.mode		= NAND_ECC_SOFT;
+		nand->ecc.layout	= NULL;
+		nand->ecc.size		= 0;
+		nand->ecc.strength	= 1;
+		bch->ecc_scheme		= OMAP_ECC_HAM1_CODE_SW;
+		break;
+	case OMAP_ECC_HAM1_CODE_HW:
+		debug("nand: selected OMAP_ECC_HAM1_CODE_HW\n");
+		nand->ecc.mode		= NAND_ECC_HW;
+		nand->ecc.strength	= 1;
+		nand->ecc.size		= 512;
+		nand->ecc.bytes		= 3;
+		nand->ecc.hwctl		= omap_enable_hwecc;
+		nand->ecc.correct	= omap_correct_data;
+		nand->ecc.calculate	= omap_calculate_ecc;
+		/* define ecc-layout */
+		ecclayout->eccbytes	= nand->ecc.bytes *
+						(pagesize / SECTOR_BYTES);
+		for (i = 0; i < ecclayout->eccbytes; i++)
+			ecclayout->eccpos[i] = i + BADBLOCK_MARKER_LENGTH;
+		ecclayout->oobfree[0].offset = i + BADBLOCK_MARKER_LENGTH;
+		ecclayout->oobfree[0].length = oobsize - ecclayout->eccbytes -
+						BADBLOCK_MARKER_LENGTH;
+		bch->ecc_scheme		= OMAP_ECC_HAM1_CODE_HW;
+		break;
+#ifdef CONFIG_BCH
+	case OMAP_ECC_BCH8_CODE_HW_DETECTION_SW:
+		debug("nand: selected OMAP_ECC_BCH8_CODE_HW_DETECTION_SW\n");
+		nand->ecc.mode		= NAND_ECC_HW;
+		nand->ecc.strength	= 8;
+		nand->ecc.size		= 512;
+		nand->ecc.bytes		= 13;
+		nand->ecc.hwctl		= omap_enable_ecc_bch;
+		nand->ecc.correct	= omap_correct_data_bch_sw;
+		nand->ecc.calculate	= omap_calculate_ecc_bch_sw;
+		/* BCH SW library is used for error detection */
+		bch_priv.control	= init_bch(13, 8, 0x201b);
+		if (!bch_priv.control) {
+			printf("nand: error: could not init_bch()\n");
+			return -ENODEV;
+		}
+		bch_priv.type		= ECC_BCH8;
+		/* define ecc-layout */
+		ecclayout->eccbytes	= nand->ecc.bytes *
+						(pagesize / SECTOR_BYTES);
+		for (i = 0; i < ecclayout->eccbytes; i++)
+			ecclayout->eccpos[i] = i + (oobsize -
+						ecclayout->eccbytes);
+		ecclayout->oobfree[0].offset = BADBLOCK_MARKER_LENGTH;
+		ecclayout->oobfree[0].length = oobsize - ecclayout->eccbytes -
+						BADBLOCK_MARKER_LENGTH;
+		omap_hwecc_init_bch(nand, NAND_ECC_READ);
+		bch->ecc_scheme		= OMAP_ECC_BCH8_CODE_HW_DETECTION_SW;
+		break;
+#endif
+	case OMAP_ECC_BCH8_CODE_HW:
+		debug("nand: selected OMAP_ECC_BCH8_CODE_HW\n");
+		nand->ecc.mode		= NAND_ECC_HW;
+		nand->ecc.strength	= 8;
+		nand->ecc.size		= 512;
+		nand->ecc.bytes		= 14;
+		nand->ecc.hwctl		= omap_enable_ecc_bch;
+		nand->ecc.correct	= omap_correct_data_bch;
+		nand->ecc.calculate	= omap_calculate_ecc_bch;
+		nand->ecc.read_page	= omap_read_page_bch;
+		/* ELM is used for ECC error detection */
+		elm_init();
+		bch_priv.type		= ECC_BCH8;
+		/* define ecc-layout */
+		ecclayout->eccbytes	= nand->ecc.bytes *
+						(pagesize / SECTOR_BYTES);
+		for (i = 0; i < ecclayout->eccbytes; i++)
+			ecclayout->eccpos[i] = i + BADBLOCK_MARKER_LENGTH;
+		ecclayout->oobfree[0].offset = i + BADBLOCK_MARKER_LENGTH;
+		ecclayout->oobfree[0].length = oobsize - ecclayout->eccbytes -
+						BADBLOCK_MARKER_LENGTH;
+		bch->ecc_scheme		= OMAP_ECC_BCH8_CODE_HW;
+		break;
+	default:

drivers/mtd/nand/omap_gpmc.c static void __maybe_unused omap_free_bch(struct mtd_info *mtd) · line 856

Quoted 10 lines
drivers/mtd/nand/omap_gpmc.c +10 -0
+		debug("nand: error: ecc scheme not enabled or supported\n");
+		return -EINVAL;
+	}
+
+	/* check if NAND spare/OOB has enough bytes to accomodate ecclayout */
+	if ((ecclayout->eccbytes + BADBLOCK_MARKER_LENGTH) > oobsize) {
+		printf("nand: error: insufficient OOB bytes. require=%d\n", (
+				ecclayout->eccbytes + BADBLOCK_MARKER_LENGTH));
+		return -EINVAL;
+	}

Check this before you make any changes to the current ECC setup.

drivers/mtd/nand/omap_gpmc.c static void __maybe_unused omap_free_bch(struct mtd_info *mtd) · line 910

Quoted 9 lines
drivers/mtd/nand/omap_gpmc.c +8 -0
+		err = omap_select_ecc_scheme(nand, OMAP_ECC_HAM1_CODE_SW,
+					mtd->writesize, mtd->oobsize);
+	}
+	if (err) {
+		printf("nand: error: could not switch ecc, reverting\n");
+		omap_select_ecc_scheme(nand, bch->ecc_scheme,
+				       mtd->writesize, mtd->oobsize);
+		return -EINVAL;
 	}

You won't need to "revert" here if omap_select_ecc_scheme doesn't damage anything in error cases.

-Scott

#4 lore raw

PATCH v7 1/5 mtd: nand: omap: enable BCH ECC scheme using ELM for generic platform

Hi,
Please see the replies inline..

Quoted 18 lines

From: Scott Wood [mailto:scottwood at freescale.com]

Quoted 6 lines · Pekon Gupta
doc/README.nand +5 -0
+Platform specific options
+=========================
+
+   CONFIG_NAND_OMAP_ECCSCHEME
+	On OMAP platforms, this specifies NAND ECC scheme.

doc/README.nand Configuration Options:

Quoted 10 lines
doc/README.nand +10 -0
+	1 - HAM1_SW	1-bit Hamming code using software library
+			(for legacy devices only)
+	2 - HAM1_HW	1-bit Hamming code using GPMC hardware engine
+			(for legacy devices only)
+	3 - BCH4_SW	4-bit BCH code (unsupported)
+	4 - BCH4_HW	4-bit BCH code (unsupported)
+	5 - BCH8_SW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using software library.
+			- requires CONFIG_BCH to enable software BCH

library

Quoted 9 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+			(For legacy device which do not have ELM h/w

engine)

doc/README.nand Configuration Options: · line 200

Quoted 3 lines
doc/README.nand +3 -0
+	6 - BCH8_HW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using ELM hardware engine.

You should document the symbols, not the numbers that happen to be assigned to them.

Sorry din't get you. This is based on your below feedback
http://lists.denx.de/pipermail/u-boot/2013-September/162773.html

Example: "6 - BCH8_HW" means 8-bit BCH ECC scheme using h/w engine. It is this number is what user needs to specify in include/configs/*.h Any other internal symbol like "OMAP_ECC_BCH8_CODE_SW" should not be exposed to user, user-interface should remain constant. This is similar to DT binding approach used in linux. Internal symbols are not exposed to users.

Quoted 9 lines

drivers/mtd/nand/omap_gpmc.c static void __maybe_unused omap_free_bch(struct mtd_info *mtd)

Quoted 8 lines
drivers/mtd/nand/omap_gpmc.c +8 -0
+/**
+ * omap_select_ecc_scheme - configures driver for particular ecc-scheme
+ * @nand: NAND chip device structure
+ * @ecc_scheme: ecc scheme to configure
+ * @pagesize: number of main-area bytes per page of NAND device
+ * @oobsize: number of OOB/spare bytes per page of NAND device
+ */
+static int omap_select_ecc_scheme(struct nand_chip *nand, int

ecc_scheme,

Quoted 4 lines

drivers/mtd/nand/omap_gpmc.c static void __maybe_unused omap_free_bch(struct mtd_info *mtd) · line 760

Quoted 1 line
drivers/mtd/nand/omap_gpmc.c +1 -0
+		unsigned int pagesize, unsigned int oobsize) {

s/int ecc_scheme/enum omap_ecc ecc_scheme/

If this is only the cosmetic change, may be I'll take it separately in another patch. Also 'omap_select_ecc_scheme()' has default statement in last, which gracefully handles all un-supported ecc-schemes.

[snip]

Quoted 4 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+	/* check if NAND spare/OOB has enough bytes to accomodate

ecclayout */

Review

Quoted 1 line
(quoted diff) +1 -0
+	if ((ecclayout->eccbytes + BADBLOCK_MARKER_LENGTH) > oobsize)

{

Quoted 8 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+		printf("nand: error: insufficient OOB bytes. require=%d\n", (
+				ecclayout->eccbytes +

BADBLOCK_MARKER_LENGTH));

Review

Quoted 2 lines
(quoted diff) +2 -0
+		return -EINVAL;
+	}

Check this before you make any changes to the current ECC setup.

'ecclayout->eccbytes' depends on ECC scheme selected, therefore this check cannot be done before selecting ECC scheme first.

Quoted 14 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+		err = omap_select_ecc_scheme(nand,

OMAP_ECC_HAM1_CODE_SW,

drivers/mtd/nand/omap_gpmc.c static void __maybe_unused omap_free_bch(struct mtd_info *mtd) · line 910

Quoted 8 lines
drivers/mtd/nand/omap_gpmc.c +7 -0
+					mtd->writesize, mtd->oobsize);
+	}
+	if (err) {
+		printf("nand: error: could not switch ecc, reverting\n");
+		omap_select_ecc_scheme(nand, bch->ecc_scheme,
+				       mtd->writesize, mtd->oobsize);
+		return -EINVAL;
 	}

You won't need to "revert" here if omap_select_ecc_scheme doesn't damage anything in error cases.

There are two cases when ECC selection could fail half-way:
(1) when if(ecclayout->eccbytes + BADBLOCK > oobsize), as this check can only be done after selecting ECC scheme so it can fail

(2) if 'bch_priv.control = init_bch(13, 8, 0x201b);' fails.
This check is also is done during ecc-scheme selection.

In both above cases, nand_chip.ecc information is half-configured so this reverting back to old ecc-scheme is essential.

with regards, pekon

#5 lore raw

PATCH v7 1/5 mtd: nand: omap: enable BCH ECC scheme using ELM for generic platform

Quoted 30 lines · Pekon

Hi,
Please see the replies inline..

Quoted 18 lines

From: Scott Wood [mailto:scottwood at freescale.com]

Quoted 6 lines · Pekon Gupta
doc/README.nand +5 -0
+Platform specific options
+=========================
+
+   CONFIG_NAND_OMAP_ECCSCHEME
+	On OMAP platforms, this specifies NAND ECC scheme.

doc/README.nand Configuration Options:

Quoted 10 lines
doc/README.nand +10 -0
+	1 - HAM1_SW	1-bit Hamming code using software library
+			(for legacy devices only)
+	2 - HAM1_HW	1-bit Hamming code using GPMC hardware engine
+			(for legacy devices only)
+	3 - BCH4_SW	4-bit BCH code (unsupported)
+	4 - BCH4_HW	4-bit BCH code (unsupported)
+	5 - BCH8_SW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using software library.
+			- requires CONFIG_BCH to enable software BCH

library

Quoted 9 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+			(For legacy device which do not have ELM h/w

engine)

doc/README.nand Configuration Options: · line 200

Quoted 3 lines
doc/README.nand +3 -0
+	6 - BCH8_HW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using ELM hardware engine.

You should document the symbols, not the numbers that happen to be assigned to them.

Quoted 7 lines

Sorry din't get you. This is based on your below feedback
http://lists.denx.de/pipermail/u-boot/2013-September/162773.html

Example: "6 - BCH8_HW" means 8-bit BCH ECC scheme using h/w engine. It is this number is what user needs to specify in include/configs/*.h Any other internal symbol like "OMAP_ECC_BCH8_CODE_SW" should not be exposed to user, user-interface should remain constant.

I disagree. The user should specify OMAP_ECC_BCH8_CODE_SW. It's more greppable, and it's more likely that the number will change than that the name will.

Quoted 2 lines

This is similar to DT binding approach used in linux. Internal symbols are not
exposed to users.

This is a shortcoming of device trees that we should not emulate in C code.

Quoted 26 lines
Quoted 9 lines

drivers/mtd/nand/omap_gpmc.c static void __maybe_unused omap_free_bch(struct mtd_info *mtd)

Quoted 8 lines
drivers/mtd/nand/omap_gpmc.c +8 -0
+/**
+ * omap_select_ecc_scheme - configures driver for particular ecc-scheme
+ * @nand: NAND chip device structure
+ * @ecc_scheme: ecc scheme to configure
+ * @pagesize: number of main-area bytes per page of NAND device
+ * @oobsize: number of OOB/spare bytes per page of NAND device
+ */
+static int omap_select_ecc_scheme(struct nand_chip *nand, int

ecc_scheme,

Quoted 4 lines

drivers/mtd/nand/omap_gpmc.c static void __maybe_unused omap_free_bch(struct mtd_info *mtd) · line 760

Quoted 1 line
drivers/mtd/nand/omap_gpmc.c +1 -0
+		unsigned int pagesize, unsigned int oobsize) {

s/int ecc_scheme/enum omap_ecc ecc_scheme/

If this is only the cosmetic change, may be I'll take it separately in another patch. Also 'omap_select_ecc_scheme()' has default statement in last, which gracefully handles all un-supported ecc-schemes.

[snip]

Quoted 7 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+	/* check if NAND spare/OOB has enough bytes to accomodate

ecclayout */

Review

Quoted 1 line
(quoted diff) +1 -0
+	if ((ecclayout->eccbytes + BADBLOCK_MARKER_LENGTH) > oobsize)

{

Review

Quoted 2 lines
(quoted diff) +2 -0
+		printf("nand: error: insufficient OOB bytes. require=%d\n", (
+				ecclayout->eccbytes +

BADBLOCK_MARKER_LENGTH));

Quoted 7 lines
Quoted 5 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+		return -EINVAL;
+	}

Check this before you make any changes to the current ECC setup.

'ecclayout->eccbytes' depends on ECC scheme selected, therefore this check cannot be done before selecting ECC scheme first.

It certainly can be done before you make any changes to the ECC struct.

You could either do it in each individual case, or you could have two separate switch statements with this code in between them.

Quoted 18 lines
Quoted 14 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+		err = omap_select_ecc_scheme(nand,

OMAP_ECC_HAM1_CODE_SW,

drivers/mtd/nand/omap_gpmc.c static void __maybe_unused omap_free_bch(struct mtd_info *mtd) · line 910

Quoted 8 lines
drivers/mtd/nand/omap_gpmc.c +7 -0
+					mtd->writesize, mtd->oobsize);
+	}
+	if (err) {
+		printf("nand: error: could not switch ecc, reverting\n");
+		omap_select_ecc_scheme(nand, bch->ecc_scheme,
+				       mtd->writesize, mtd->oobsize);
+		return -EINVAL;
 	}

You won't need to "revert" here if omap_select_ecc_scheme doesn't damage anything in error cases.

There are two cases when ECC selection could fail half-way:
(1) when if(ecclayout->eccbytes + BADBLOCK > oobsize), as this check can only be done after selecting ECC scheme so it can fail

Quoted 5 lines

(2) if 'bch_priv.control = init_bch(13, 8, 0x201b);' fails.
This check is also is done during ecc-scheme selection.

In both above cases, nand_chip.ecc information is half-configured so this reverting back to old ecc-scheme is essential.

If you really can't avoid the damage in error cases, move the cleanup code into omap_select_ecc_scheme().

-Scott

#6 lore raw

PATCH v7 1/5 mtd: nand: omap: enable BCH ECC scheme using ELM for generic platform

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Quoted 34 lines · Scott Wood
Quoted 33 lines · Pekon

Hi,
Please see the replies inline..

Quoted 18 lines

From: Scott Wood [mailto:scottwood at freescale.com]

Quoted 6 lines · Pekon Gupta
doc/README.nand +5 -0
+Platform specific options
+=========================
+
+   CONFIG_NAND_OMAP_ECCSCHEME
+	On OMAP platforms, this specifies NAND ECC scheme.

doc/README.nand Configuration Options:

Quoted 10 lines
doc/README.nand +10 -0
+	1 - HAM1_SW	1-bit Hamming code using software library
+			(for legacy devices only)
+	2 - HAM1_HW	1-bit Hamming code using GPMC hardware engine
+			(for legacy devices only)
+	3 - BCH4_SW	4-bit BCH code (unsupported)
+	4 - BCH4_HW	4-bit BCH code (unsupported)
+	5 - BCH8_SW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using software library.
+			- requires CONFIG_BCH to enable software BCH

library

Quoted 9 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+			(For legacy device which do not have ELM h/w

engine)

doc/README.nand Configuration Options: · line 200

Quoted 3 lines
doc/README.nand +3 -0
+	6 - BCH8_HW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using ELM hardware engine.

You should document the symbols, not the numbers that happen to be assigned to them.

Sorry din't get you. This is based on your below feedback
http://lists.denx.de/pipermail/u-boot/2013-September/162773.html

Quoted 8 lines
Quoted 4 lines

Example: "6 - BCH8_HW" means 8-bit BCH ECC scheme using h/w engine. It is this number is what user needs to specify in include/configs/*.h Any other internal symbol like "OMAP_ECC_BCH8_CODE_SW" should not be exposed to user, user-interface should remain constant.

I disagree. The user should specify OMAP_ECC_BCH8_CODE_SW. It's more greppable, and it's more likely that the number will change than that the name will.

I agree with Scott, enum (if it's not already one) and use the name so it's clear what you're picking.

- --
Tom
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQIcBAEBAgAGBQJSU2cEAAoJENk4IS6UOR1WnOQP/30AInl+d0QVU6QHq+DsGC/q Wh/yfXbN/EkWVPEvpASS5gftTSd+UjLSzUSWca+TcQxUVdzD64H1u4j8An2cSwJp vQuX1QZzMWA+g6D+u75BKMs/cIXb6rwX9A5is5R/k5cU4mew+ecLmBeRLcTENfpM +ROD7XfSJsXYff+dLhpOeVWC6ELzjgmHWajoqpdZNefrMtsrRHOeLd9ccbqYxoNZ Td56vwMYfk8f+BBRmGlFRdAf+xN+9ssou1CipK6E/H0QHnexRkuPINRdmPD4/p0o yP+opvOUq8XMYiO76HOGb37tMJvjgziEcZUy4S3Q0gbhRlC9EXz1D0qxs6ti1EFs 9M/Wsl8bsKx4rgdf12hj8iebFxvTdT1lXejJGNk9ThHqsHE0pl+K1cyouZ/GL0NA sk/UCPe7DL3RIAtkUy+pLT7A3nOKIhqJY1kTsPN24qTF7UVlatWov6C33d42jDzp YH5CkdP1FRffkRj1in1ex0CiVB/3h3SuI4kA+hceWdx0rGYRT/PRKQA3qSulU1p4 lx3jtzt07fSflPdeX9PxW5k9CqAmXut4rw5USU1OAdpRKANzp1FPXcKXjEVd2uQ1 zkfgUB4MeR6xs6HE6q5xBbaaRhy9zJX415ZG10NtKUXCrVBIoif51EQQmlvehgcy ufn39bBOmsjkXYXzJI8W
=tQHU
-----END PGP SIGNATURE-----

#7 lore raw

PATCH v7 1/5 mtd: nand: omap: enable BCH ECC scheme using ELM for generic platform

In reply to Scott Wood
Quoted 35 lines

From: Scott Wood [mailto:scottwood at freescale.com]

Quoted 34 lines · Pekon

Hi,
Please see the replies inline..

Quoted 18 lines

From: Scott Wood [mailto:scottwood at freescale.com]

Quoted 6 lines · Pekon Gupta
doc/README.nand +5 -0
+Platform specific options
+=========================
+
+   CONFIG_NAND_OMAP_ECCSCHEME
+	On OMAP platforms, this specifies NAND ECC scheme.

doc/README.nand Configuration Options:

Quoted 10 lines
doc/README.nand +10 -0
+	1 - HAM1_SW	1-bit Hamming code using software library
+			(for legacy devices only)
+	2 - HAM1_HW	1-bit Hamming code using GPMC hardware engine
+			(for legacy devices only)
+	3 - BCH4_SW	4-bit BCH code (unsupported)
+	4 - BCH4_HW	4-bit BCH code (unsupported)
+	5 - BCH8_SW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using software library.
+			- requires CONFIG_BCH to enable software BCH

library

Quoted 9 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+			(For legacy device which do not have ELM h/w

engine)

doc/README.nand Configuration Options: · line 200

Quoted 3 lines
doc/README.nand +3 -0
+	6 - BCH8_HW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using ELM hardware engine.

You should document the symbols, not the numbers that happen to be assigned to them.

Sorry din't get you. This is based on your below feedback
http://lists.denx.de/pipermail/u-boot/2013-September/162773.html

Quoted 9 lines
Quoted 4 lines

Example: "6 - BCH8_HW" means 8-bit BCH ECC scheme using h/w engine. It is this number is what user needs to specify in include/configs/*.h Any other internal symbol like "OMAP_ECC_BCH8_CODE_SW" should not be exposed to user, user-interface should remain constant.

I disagree. The user should specify OMAP_ECC_BCH8_CODE_SW. It's more greppable, and it's more likely that the number will change than that the name will.

Ok, accepted.

Quoted 6 lines
Quoted 2 lines

This is similar to DT binding approach used in linux. Internal symbols are not
exposed to users.

This is a shortcoming of device trees that we should not emulate in C code.

Don't know if its short-coming of device-tree or difference between u-boot and kernel ideologies. But use of internal symbols (enums) in DT was opposed by one of the kernel maintainers. Refer below:
http://lists.infradead.org/pipermail/linux-mtd/2013-May/047030.html

Quoted 32 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

+/**
+ * omap_select_ecc_scheme - configures driver for particular ecc-

scheme

Quoted 29 lines
Quoted 11 lines
Quoted 6 lines

+ * @nand: NAND chip device structure
+ * @ecc_scheme: ecc scheme to configure
+ * @pagesize: number of main-area bytes per page of NAND device
+ * @oobsize: number of OOB/spare bytes per page of NAND device
+ */
+static int omap_select_ecc_scheme(struct nand_chip *nand, int

ecc_scheme,

drivers/mtd/nand/omap_gpmc.c static void __maybe_unused omap_free_bch(struct mtd_info *mtd) · line 760

Quoted 1 line
drivers/mtd/nand/omap_gpmc.c +1 -0
+		unsigned int pagesize, unsigned int oobsize) {

s/int ecc_scheme/enum omap_ecc ecc_scheme/

If this is only the cosmetic change, may be I'll take it separately in another patch. Also 'omap_select_ecc_scheme()' has default statement in last, which gracefully handles all un-supported ecc-schemes.

[snip]

Quoted 12 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+	/* check if NAND spare/OOB has enough bytes to accomodate

ecclayout */

Review

Quoted 1 line
(quoted diff) +1 -0
+	if ((ecclayout->eccbytes + BADBLOCK_MARKER_LENGTH) > oobsize)

{

Review

Quoted 2 lines
(quoted diff) +2 -0
+		printf("nand: error: insufficient OOB bytes. require=%d\n", (
+				ecclayout->eccbytes +

BADBLOCK_MARKER_LENGTH));

Review

Quoted 2 lines
(quoted diff) +2 -0
+		return -EINVAL;
+	}

Check this before you make any changes to the current ECC setup.

Quoted 8 lines
Quoted 2 lines

'ecclayout->eccbytes' depends on ECC scheme selected, therefore this check cannot be done before selecting ECC scheme first.

It certainly can be done before you make any changes to the ECC struct.

You could either do it in each individual case, or you could have two separate switch statements with this code in between them.

Having 2 switch statements would clutter the code. Isn't it ?
(1) first switch-case would do a dry-run to check the ecc requirements.
(2) second switch-case would do actual assignments.

And omap_select_ecc_scheme() would be called only once in whole lifetime, because:
(1) omap devices have deprecated the dynamic switching of ecc-schemes.
(2) with falcon-mode where SPL directly loads kernel, most driver configs would be statically derived from CONFIGS_xx.

Anyways I would take the changes if you wish so..
But request you to please provide comments on all the patches, before I send next revision. This would help me consolidate all changes.

Quoted 18 lines
Quoted 18 lines
Quoted 14 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+		err = omap_select_ecc_scheme(nand,

OMAP_ECC_HAM1_CODE_SW,

drivers/mtd/nand/omap_gpmc.c static void __maybe_unused omap_free_bch(struct mtd_info *mtd) · line 910

Quoted 8 lines
drivers/mtd/nand/omap_gpmc.c +7 -0
+					mtd->writesize, mtd->oobsize);
+	}
+	if (err) {
+		printf("nand: error: could not switch ecc, reverting\n");
+		omap_select_ecc_scheme(nand, bch->ecc_scheme,
+				       mtd->writesize, mtd->oobsize);
+		return -EINVAL;
 	}

You won't need to "revert" here if omap_select_ecc_scheme doesn't damage anything in error cases.

There are two cases when ECC selection could fail half-way:
(1) when if(ecclayout->eccbytes + BADBLOCK > oobsize), as this check can only be done after selecting ECC scheme so it can fail

Quoted 9 lines
Quoted 5 lines

(2) if 'bch_priv.control = init_bch(13, 8, 0x201b);' fails.
This check is also is done during ecc-scheme selection.

In both above cases, nand_chip.ecc information is half-configured so this reverting back to old ecc-scheme is essential.

If you really can't avoid the damage in error cases, move the cleanup code into omap_select_ecc_scheme().

As described above, this has dependency on ecc information So again I have to go by 2 switch statement approach which would clutter the code. I'm okay if you want it so..

But request you to review and provide comments to other patches of this series, OR else atleast Ack/Nak them. This way my effort on re-working and testing the patches again and again would reduce.

So, should I wait for review of other remaining patches before sending next version v8 of this series ?

with regards, pekon

#8 lore raw

PATCH v7 1/5 mtd: nand: omap: enable BCH ECC scheme using ELM for generic platform

Quoted 50 lines · Pekon
Quoted 35 lines

From: Scott Wood [mailto:scottwood at freescale.com]

Quoted 34 lines · Pekon

Hi,
Please see the replies inline..

Quoted 18 lines

From: Scott Wood [mailto:scottwood at freescale.com]

Quoted 6 lines · Pekon Gupta
doc/README.nand +5 -0
+Platform specific options
+=========================
+
+   CONFIG_NAND_OMAP_ECCSCHEME
+	On OMAP platforms, this specifies NAND ECC scheme.

doc/README.nand Configuration Options:

Quoted 10 lines
doc/README.nand +10 -0
+	1 - HAM1_SW	1-bit Hamming code using software library
+			(for legacy devices only)
+	2 - HAM1_HW	1-bit Hamming code using GPMC hardware engine
+			(for legacy devices only)
+	3 - BCH4_SW	4-bit BCH code (unsupported)
+	4 - BCH4_HW	4-bit BCH code (unsupported)
+	5 - BCH8_SW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using software library.
+			- requires CONFIG_BCH to enable software BCH

library

Quoted 9 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+			(For legacy device which do not have ELM h/w

engine)

doc/README.nand Configuration Options: · line 200

Quoted 3 lines
doc/README.nand +3 -0
+	6 - BCH8_HW	8-bit BCH code with
+			- ecc calculation using GPMC hardware engine,
+			- error detection using ELM hardware engine.

You should document the symbols, not the numbers that happen to be assigned to them.

Sorry din't get you. This is based on your below feedback
http://lists.denx.de/pipermail/u-boot/2013-September/162773.html

Quoted 9 lines
Quoted 4 lines

Example: "6 - BCH8_HW" means 8-bit BCH ECC scheme using h/w engine. It is this number is what user needs to specify in include/configs/*.h Any other internal symbol like "OMAP_ECC_BCH8_CODE_SW" should not be exposed to user, user-interface should remain constant.

I disagree. The user should specify OMAP_ECC_BCH8_CODE_SW. It's more greppable, and it's more likely that the number will change than that the name will.

Ok, accepted.

Quoted 3 lines
Quoted 2 lines

This is similar to DT binding approach used in linux. Internal symbols are not
exposed to users.

Quoted 7 lines
Quoted 3 lines

This is a shortcoming of device trees that we should not emulate in C code.

Don't know if its short-coming of device-tree or difference between u-boot and kernel ideologies. But use of internal symbols (enums) in DT was opposed by one of the kernel maintainers. Refer below:
http://lists.infradead.org/pipermail/linux-mtd/2013-May/047030.html

It's a different situation. The device tree is supposed to be stable ABI. This isn't. The device tree is also supposed to be OS-independent and thus taking Linux identifiers as-is is frowned upon (as the Linux identifiers may change, plus device tree properties use a different style for naming).

It's not a general difference between U-Boot and Linux.

Quoted 35 lines
Quoted 35 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

+/**
+ * omap_select_ecc_scheme - configures driver for particular ecc-

scheme

Quoted 31 lines
Quoted 11 lines
Quoted 6 lines

+ * @nand: NAND chip device structure
+ * @ecc_scheme: ecc scheme to configure
+ * @pagesize: number of main-area bytes per page of NAND device
+ * @oobsize: number of OOB/spare bytes per page of NAND device
+ */
+static int omap_select_ecc_scheme(struct nand_chip *nand, int

ecc_scheme,

drivers/mtd/nand/omap_gpmc.c static void __maybe_unused omap_free_bch(struct mtd_info *mtd) · line 760

Quoted 1 line
drivers/mtd/nand/omap_gpmc.c +1 -0
+		unsigned int pagesize, unsigned int oobsize) {

s/int ecc_scheme/enum omap_ecc ecc_scheme/

If this is only the cosmetic change, may be I'll take it separately in another patch. Also 'omap_select_ecc_scheme()' has default statement in last, which gracefully handles all un-supported ecc-schemes.

[snip]

Quoted 4 lines

Review

Quoted 1 line
(quoted diff) +1 -0
+	/* check if NAND spare/OOB has enough bytes to accomodate

ecclayout */

Review

Quoted 1 line
(quoted diff) +1 -0
+	if ((ecclayout->eccbytes + BADBLOCK_MARKER_LENGTH) > oobsize)

{

Quoted 8 lines

Review

Quoted 2 lines
(quoted diff) +2 -0
+		printf("nand: error: insufficient OOB bytes. require=%d\n", (
+				ecclayout->eccbytes +

BADBLOCK_MARKER_LENGTH));

Review

Quoted 2 lines
(quoted diff) +2 -0
+		return -EINVAL;
+	}

Check this before you make any changes to the current ECC setup.

'ecclayout->eccbytes' depends on ECC scheme selected, therefore this check cannot be done before selecting ECC scheme first.

Quoted 8 lines
Quoted 5 lines

It certainly can be done before you make any changes to the ECC struct.

You could either do it in each individual case, or you could have two separate switch statements with this code in between them.

Having 2 switch statements would clutter the code. Isn't it ?
(1) first switch-case would do a dry-run to check the ecc requirements.
(2) second switch-case would do actual assignments.

It wouldn't be that bad. It would eliminate the need for recovery code and substantially reduce the complexity of a (likely little-tested, at least long term) error path.

Quoted 9 lines

And omap_select_ecc_scheme() would be called only once in whole lifetime, because:
(1) omap devices have deprecated the dynamic switching of ecc-schemes.
(2) with falcon-mode where SPL directly loads kernel, most driver configs would be statically derived from CONFIGS_xx.

Anyways I would take the changes if you wish so..
But request you to please provide comments on all the patches, before I send next revision. This would help me consolidate all changes.

I'm working on it, but I also have a lot of other patches to review, and doing nothing but patch reviews all day long can be very draining (not to mention starving other tasks that need to be done), and reviewing patches for hardware I'm not familiar with is even worse. I realize that the delays can be frustrating from your end, but I only have so much time and energy to expend. Is there nobody else who's familiar with this driver that can help review? Looking at the commit history no single person stands out as a logical maintainer for this driver. It would be nice to designate someone.

Quoted 2 lines

So, should I wait for review of other remaining patches before sending next version v8 of this series ?

Yes.

-Scott

#9 lore raw

PATCH v7 1/5 mtd: nand: omap: enable BCH ECC scheme using ELM for generic platform

Quoted 15 lines · Scott Wood
Quoted 3 lines · Pekon

Anyways I would take the changes if you wish so..
But request you to please provide comments on all the patches, before I send next revision. This would help me consolidate all changes.

I'm working on it, but I also have a lot of other patches to review, and doing nothing but patch reviews all day long can be very draining (not to mention starving other tasks that need to be done), and reviewing patches for hardware I'm not familiar with is even worse. I realize that the delays can be frustrating from your end, but I only have so much time and energy to expend. Is there nobody else who's familiar with this driver that can help review? Looking at the commit history no single person stands out as a logical maintainer for this driver. It would be nice to designate someone.

Quoted 4 lines
Quoted 2 lines

So, should I wait for review of other remaining patches before sending next version v8 of this series ?

Yes.

OK, I've scanned the rest of the patches and given some minor feedback.

-Scott

#10 lore raw

PATCH v7 1/5 mtd: nand: omap: enable BCH ECC scheme using ELM for generic platform

Hi,

Quoted 6 lines · Scott Wood
Quoted 4 lines · Pekon

Anyways I would take the changes if you wish so..
But request you to please provide comments on all the patches, before I send next revision. This would help me consolidate all changes.

Quoted 7 lines

I'm working on it, but I also have a lot of other patches to review, and doing nothing but patch reviews all day long can be very draining (not to mention starving other tasks that need to be done), and reviewing patches for hardware I'm not familiar with is even worse. I realize that the delays can be frustrating from your end, but I only have so much time and energy to expend.

Firstly my apologies for being bit aggressive in getting patches reviewed. Yes, I was bit irritated as patch re-submissions were taking much of my energy and time as well, while lot of other things are still pending on me to cleanup this driver and add functionality to it.

And I understand the patch reviewing is tough because reviewer needs to first understand the driver & hardware context, and then reply from a broader point of view of generic frame-work.

Quoted 5 lines

Is there nobody else who's familiar
with this driver that can help review? Looking at the commit history no single person stands out as a logical maintainer for this driver. It would be nice to designate someone.

Currently I'm the only supporting NAND driver in both kernel and u-boot from omap side. And therefore, I plan to get both u-boot and kernel nand drivers cleaned-up and optimized, so that it becomes scalable for future enhancements.
Aim is to get most of end-users to boot from mainline, so that they are self dependent and up-to-date to new frameworks.

Once I'm done with my current commitments, I can help you in getting some enhancements for generic framework.
For example, following patch adds checks for automatic bus-width detection in generic NAND framework.
http://lists.denx.de/pipermail/u-boot/2013-September/163882.html

Hope such things help reduce some of your load.
And thanks again for the reviews, I'll submit next revision of patch soon.

with regards, pekon

#11 lore raw

PATCH v7 2/5 mtd: nand: omap: optimize chip->ecc.hwctl() for H/W ECC schemes

In reply to Pekon Gupta

chip->ecc.hwctl() is used for preparing the H/W controller before read/write NAND accesses (like assigning data-buf, enabling ECC scheme configs, etc.)

Though all ECC schemes in OMAP NAND driver use GPMC controller for generating ECC syndrome (for both Read/Write accesses). But but in current code HAM1_ECC and BCHx_ECC schemes implement individual function to achieve this. This patch
(1) removes omap_hwecc_init() and omap_hwecc_init_bch()
as chip->ecc.hwctl will re-initializeGPMC before every read/write call. omap_hwecc_init_bch() -> omap_enable_ecc_bch()

(2) merges the GPMC configuration code for all ECC schemes into single omap_enable_hwecc(), thus adding scalability for future ECC schemes. omap_enable_hwecc() + omap_enable_ecc_bch() -> omap_enable_hwecc()

---
 arch/arm/include/asm/omap_gpmc.h |   7 --
 drivers/mtd/nand/omap_gpmc.c     | 192 +++++++++++----------------------------
 2 files changed, 54 insertions(+), 145 deletions(-)
arch/arm/include/asm/omap_gpmc.h +0 -7
index dd40cb6..9660842 100644
@@ -14,13 +14,6 @@
 #define GPMC_BUF_EMPTY	0
 #define GPMC_BUF_FULL	1
 
-#define ECCCLEAR	(0x1 << 8)
-#define ECCRESULTREG1	(0x1 << 0)
-#define ECCSIZE512BYTE	0xFF
-#define ECCSIZE1	(ECCSIZE512BYTE << 22)
-#define ECCSIZE0	(ECCSIZE512BYTE << 12)
-#define ECCSIZE0SEL	(0x000 << 0)
-
 /* Generic ECC Layouts */
 /* Large Page x8 NAND device Layout */
 #ifdef GPMC_NAND_ECC_LP_x8_LAYOUT
drivers/mtd/nand/omap_gpmc.c +54 -138

243 lines not on this page · raw message

Signature · 1 line
1.8.1
#12 lore raw

PATCH v7 3/5 mtd: nand: omap: optimize chip->ecc.calculate() for H/W ECC schemes

In reply to Pekon Gupta

chip->ecc.calculate() is used for calculating and fetching of ECC syndrome by processing the data passed during Read/Write accesses.

All H/W based ECC schemes use GPMC controller to calculate ECC syndrome. But each BCHx_ECC scheme has its own implemetation of post-processing and fetching ECC syndrome from GPMC controller.

This patch updates OMAP_ECC_BCH8_CODE_HW ECC scheme in following way:
- merges multiple chip->calculate API for different ECC schemes omap_calculate_ecc() + omap_calculate_ecc_bch() + omap_calculate_ecc_bch_sw()

				==> omap_calculate_ecc()

- removes omap_ecc_disable() and instead uses it as inline.

---
 drivers/mtd/nand/omap_gpmc.c | 220 ++++++++++++-------------------------------
 1 file changed, 59 insertions(+), 161 deletions(-)
drivers/mtd/nand/omap_gpmc.c +59 -161
index 67998ef..76be758 100644
@@ -21,7 +21,10 @@
 #define SECTOR_BYTES		512
 #define ECCCLEAR		(0x1 << 8)
 #define ECCRESULTREG1		(0x1 << 0)
-
+#ifdef CONFIG_BCH
+static u8  bch8_polynomial[] = {0xef, 0x51, 0x2e, 0x09, 0xed, 0x93, 0x9a, 0xc2,
+				0x97, 0x79, 0xe5, 0x24, 0xb5};
+#endif
 static uint8_t cs;
 static __maybe_unused struct nand_ecclayout omap_ecclayout;
 
@@ -143,42 +146,6 @@ static int __maybe_unused omap_correct_data(struct mtd_info *mtd, uint8_t *dat,
 }
 
 /*
- *  omap_calculate_ecc - Generate non-inverted ECC bytes.
- *
- *  Using noninverted ECC can be considered ugly since writing a blank
- *  page ie. padding will clear the ECC bytes. This is no problem as
- *  long nobody is trying to write data on the seemingly unused page.
- *  Reading an erased page will produce an ECC mismatch between
- *  generated and read ECC bytes that has to be dealt with separately.
- *  E.g. if page is 0xFF (fresh erased), and if HW ECC engine within GPMC
- *  is used, the result of read will be 0x0 while the ECC offsets of the
- *  spare area will be 0xFF which will result in an ECC mismatch.
- *  @mtd:	MTD structure
- *  @dat:	unused
- *  @ecc_code:	ecc_code buffer
- */
-static int __maybe_unused omap_calculate_ecc(struct mtd_info *mtd,
-		const uint8_t *dat, uint8_t *ecc_code)
-{
-	u_int32_t val;
-
-	/* Start Reading from HW ECC1_Result = 0x200 */
-	val = readl(&gpmc_cfg->ecc1_result);
-
-	ecc_code[0] = val & 0xFF;
-	ecc_code[1] = (val >> 16) & 0xFF;
-	ecc_code[2] = ((val >> 8) & 0x0F) | ((val >> 20) & 0xF0);
-
-	/*
-	 * Stop reading anymore ECC vals and clear old results
-	 * enable will be called if more reads are required
-	 */
-	writel(0x000, &gpmc_cfg->ecc_config);
-
-	return 0;
-}
-
-/*
  * Generic BCH interface
  */
 struct nand_bch_priv {
@@ -194,12 +161,7 @@ struct nand_bch_priv {
 #define ECC_BCH8	1
 #define ECC_BCH16	2
 
-/* GPMC ecc engine settings */
-#define BCH_WRAPMODE_1		1	/* BCH wrap mode 1 */
-#define BCH_WRAPMODE_6		6	/* BCH wrap mode 6 */
-
 /* BCH nibbles for diff bch levels */
-#define NAND_ECC_HW_BCH ((uint8_t)(NAND_ECC_HW_OOB_FIRST) + 1)
 #define ECC_BCH4_NIBBLES	13
 #define ECC_BCH8_NIBBLES	26
 #define ECC_BCH16_NIBBLES	52
@@ -211,7 +173,6 @@ struct nand_bch_priv {
  * When some users with other BCH strength will exists this have to change!
  */
 static __maybe_unused struct nand_bch_priv bch_priv = {
-	.mode = NAND_ECC_HW_BCH,
 	.type = ECC_BCH8,
 	.nibbles = ECC_BCH8_NIBBLES,
 	.control = NULL
@@ -280,54 +241,71 @@ static void omap_enable_hwecc(struct mtd_info *mtd, int32_t mode)
 }
 
 /*
- * omap_ecc_disable - Disable H/W ECC calculation
- *
- * @mtd:	MTD device structure
- */
-static void __maybe_unused omap_ecc_disable(struct mtd_info *mtd)
-{
-	writel((readl(&gpmc_cfg->ecc_config) & ~0x1), &gpmc_cfg->ecc_config);
-}
-
-/*
- * BCH support using ELM module
- */
-/*
- * omap_read_bch8_result - Read BCH result for BCH8 level
+ *  omap_calculate_ecc - Read ECC result
+ *  @mtd:	MTD structure
+ *  @dat:	unused
+ *  @ecc_code:	ecc_code buffer
  *
- * @mtd:	MTD device structure
- * @big_endian:	When set read register 3 first
- * @ecc_code:	Read syndrome from BCH result registers
+ *  Using noninverted ECC can be considered ugly since writing a blank
+ *  page ie. padding will clear the ECC bytes. This is no problem as
+ *  long nobody is trying to write data on the seemingly unused page.
+ *  Reading an erased page will produce an ECC mismatch between
+ *  generated and read ECC bytes that has to be dealt with separately.
+ *  E.g. if page is 0xFF (fresh erased), and if HW ECC engine within GPMC
+ *  is used, the result of read will be 0x0 while the ECC offsets of the
+ *  spare area will be 0xFF which will result in an ECC mismatch.
  */
-static void omap_read_bch8_result(struct mtd_info *mtd, uint8_t big_endian,
+static int omap_calculate_ecc(struct mtd_info *mtd, const uint8_t *dat,
 				uint8_t *ecc_code)
 {
-	uint32_t *ptr;
+	struct nand_chip *chip = mtd->priv;
+	struct nand_bch_priv *bch = chip->priv;
+	uint32_t *ptr, val = 0;
 	int8_t i = 0, j;
 
-	if (big_endian) {
+	switch (bch->ecc_scheme) {
+	case OMAP_ECC_HAM1_CODE_HW:
+		val = readl(&gpmc_cfg->ecc1_result);
+		ecc_code[0] = val & 0xFF;
+		ecc_code[1] = (val >> 16) & 0xFF;
+		ecc_code[2] = ((val >> 8) & 0x0F) | ((val >> 20) & 0xF0);
+		break;
+	case OMAP_ECC_BCH8_CODE_HW_DETECTION_SW:
+	case OMAP_ECC_BCH8_CODE_HW:
 		ptr = &gpmc_cfg->bch_result_0_3[0].bch_result_x[3];
-		ecc_code[i++] = readl(ptr) & 0xFF;
+		val = readl(ptr);
+		ecc_code[i++] = (val >>  0) & 0xFF;
 		ptr--;
 		for (j = 0; j < 3; j++) {
-			ecc_code[i++] = (readl(ptr) >> 24) & 0xFF;
-			ecc_code[i++] = (readl(ptr) >> 16) & 0xFF;
-			ecc_code[i++] = (readl(ptr) >>  8) & 0xFF;
-			ecc_code[i++] = readl(ptr) & 0xFF;
+			val = readl(ptr);
+			ecc_code[i++] = (val >> 24) & 0xFF;
+			ecc_code[i++] = (val >> 16) & 0xFF;
+			ecc_code[i++] = (val >>  8) & 0xFF;
+			ecc_code[i++] = (val >>  0) & 0xFF;
 			ptr--;
 		}
-	} else {
-		ptr = &gpmc_cfg->bch_result_0_3[0].bch_result_x[0];
-		for (j = 0; j < 3; j++) {
-			ecc_code[i++] = readl(ptr) & 0xFF;
-			ecc_code[i++] = (readl(ptr) >>  8) & 0xFF;
-			ecc_code[i++] = (readl(ptr) >> 16) & 0xFF;
-			ecc_code[i++] = (readl(ptr) >> 24) & 0xFF;
-			ptr++;
-		}
-		ecc_code[i++] = readl(ptr) & 0xFF;
-		ecc_code[i++] = 0;	/* 14th byte is always zero */
+		break;
+	default:
+		return -EINVAL;
+	}
+	/* ECC scheme specific syndrome customizations */
+	switch (bch->ecc_scheme) {
+	case OMAP_ECC_HAM1_CODE_HW:
+		break;
+	case OMAP_ECC_BCH8_CODE_HW_DETECTION_SW:
+#ifdef CONFIG_BCH
+		for (i = 0; i < chip->ecc.bytes; i++)
+			*(ecc_code + i) = *(ecc_code + i) ^
+						bch8_polynomial[i];
+#endif
+		break;
+	case OMAP_ECC_BCH8_CODE_HW:
+		ecc_code[chip->ecc.bytes - 1] = 0x00;
+		break;
+	default:
+		return -EINVAL;
 	}
+	return 0;
 }
 
 /*
@@ -366,35 +344,6 @@ static void omap_rotate_ecc_bch(struct mtd_info *mtd, uint8_t *calc_ecc,
 }
 
 /*
- *  omap_calculate_ecc_bch - Read BCH ECC result
- *
- *  @mtd:	MTD structure
- *  @dat:	unused
- *  @ecc_code:	ecc_code buffer
- */
-static int omap_calculate_ecc_bch(struct mtd_info *mtd, const uint8_t *dat,
-				uint8_t *ecc_code)
-{
-	struct nand_chip *chip = mtd->priv;
-	struct nand_bch_priv *bch = chip->priv;
-	uint8_t big_endian = 1;
-	int8_t ret = 0;
-
-	if (bch->type == ECC_BCH8)
-		omap_read_bch8_result(mtd, big_endian, ecc_code);
-	else /* BCH4 and BCH16 currently not supported */
-		ret = -1;
-
-	/*
-	 * Stop reading anymore ECC vals and clear old results
-	 * enable will be called if more reads are required
-	 */
-	omap_ecc_disable(mtd);
-
-	return ret;
-}
-
-/*
  * omap_fix_errors_bch - Correct bch error in the data
  *
  * @mtd:	MTD device structure
@@ -549,57 +498,6 @@ static int omap_read_page_bch(struct mtd_info *mtd, struct nand_chip *chip,
  * OMAP3 BCH8 support (with BCH library)
  */
 #ifdef CONFIG_BCH
-/*
- *  omap_calculate_ecc_bch_sw - Read BCH ECC result
- *
- *  @mtd:	MTD device structure
- *  @dat:	The pointer to data on which ecc is computed (unused here)
- *  @ecc:	The ECC output buffer
- */
-static int omap_calculate_ecc_bch_sw(struct mtd_info *mtd, const uint8_t *dat,
-				uint8_t *ecc)
-{
-	int ret = 0;
-	size_t i;
-	unsigned long nsectors, val1, val2, val3, val4;
-
-	nsectors = ((readl(&gpmc_cfg->ecc_config) >> 4) & 0x7) + 1;
-
-	for (i = 0; i < nsectors; i++) {
-		/* Read hw-computed remainder */
-		val1 = readl(&gpmc_cfg->bch_result_0_3[i].bch_result_x[0]);
-		val2 = readl(&gpmc_cfg->bch_result_0_3[i].bch_result_x[1]);
-		val3 = readl(&gpmc_cfg->bch_result_0_3[i].bch_result_x[2]);
-		val4 = readl(&gpmc_cfg->bch_result_0_3[i].bch_result_x[3]);
-
-		/*
-		 * Add constant polynomial to remainder, in order to get an ecc
-		 * sequence of 0xFFs for a buffer filled with 0xFFs.
-		 */
-		*ecc++ = 0xef ^ (val4 & 0xFF);
-		*ecc++ = 0x51 ^ ((val3 >> 24) & 0xFF);
-		*ecc++ = 0x2e ^ ((val3 >> 16) & 0xFF);
-		*ecc++ = 0x09 ^ ((val3 >> 8) & 0xFF);
-		*ecc++ = 0xed ^ (val3 & 0xFF);
-		*ecc++ = 0x93 ^ ((val2 >> 24) & 0xFF);
-		*ecc++ = 0x9a ^ ((val2 >> 16) & 0xFF);
-		*ecc++ = 0xc2 ^ ((val2 >> 8) & 0xFF);
-		*ecc++ = 0x97 ^ (val2 & 0xFF);
-		*ecc++ = 0x79 ^ ((val1 >> 24) & 0xFF);
-		*ecc++ = 0xe5 ^ ((val1 >> 16) & 0xFF);
-		*ecc++ = 0x24 ^ ((val1 >> 8) & 0xFF);
-		*ecc++ = 0xb5 ^ (val1 & 0xFF);
-	}
-
-	/*
-	 * Stop reading anymore ECC vals and clear old results
-	 * enable will be called if more reads are required
-	 */
-	omap_ecc_disable(mtd);
-
-	return ret;
-}
-
 /**
  * omap_correct_data_bch_sw - Decode received data and correct errors
  * @mtd: MTD device structure
@@ -716,7 +614,7 @@ static int omap_select_ecc_scheme(struct nand_chip *nand, int ecc_scheme,
 		nand->ecc.bytes		= 13;
 		nand->ecc.hwctl		= omap_enable_hwecc;
 		nand->ecc.correct	= omap_correct_data_bch_sw;
-		nand->ecc.calculate	= omap_calculate_ecc_bch_sw;
+		nand->ecc.calculate	= omap_calculate_ecc;
 		/* BCH SW library is used for error detection */
 		bch_priv.control	= init_bch(13, 8, 0x201b);
 		if (!bch_priv.control) {
@@ -744,7 +642,7 @@ static int omap_select_ecc_scheme(struct nand_chip *nand, int ecc_scheme,
 		nand->ecc.bytes		= 14;
 		nand->ecc.hwctl		= omap_enable_hwecc;
 		nand->ecc.correct	= omap_correct_data_bch;
-		nand->ecc.calculate	= omap_calculate_ecc_bch;
+		nand->ecc.calculate	= omap_calculate_ecc;
 		nand->ecc.read_page	= omap_read_page_bch;
 		/* ELM is used for ECC error detection */
 		elm_init();
Signature · 1 line
1.8.1
#13 lore raw

PATCH v7 3/5 mtd: nand: omap: optimize chip->ecc.calculate() for H/W ECC schemes

Quoted 7 lines · Pekon Gupta
drivers/mtd/nand/omap_gpmc.c +7 -0
+	/* ECC scheme specific syndrome customizations */
+	switch (bch->ecc_scheme) {
+	case OMAP_ECC_HAM1_CODE_HW:
+		break;
+	case OMAP_ECC_BCH8_CODE_HW_DETECTION_SW:
+#ifdef CONFIG_BCH
+		for (i = 0; i < chip->ecc.bytes; i++)

drivers/mtd/nand/omap_gpmc.c static void omap_enable_hwecc(struct mtd_info *mtd, int32_t mode) · line 307

Quoted 10 lines
drivers/mtd/nand/omap_gpmc.c +9 -0
+			*(ecc_code + i) = *(ecc_code + i) ^
+						bch8_polynomial[i];
+#endif
+		break;
+	case OMAP_ECC_BCH8_CODE_HW:
+		ecc_code[chip->ecc.bytes - 1] = 0x00;
+		break;
+	default:
+		return -EINVAL;
 	}

Shouldn't "case OMAP_ECC_BCH8_CODE_HW_DETECTION_SW:" be inside the "#ifdef CONFIG_BCH" so that you get -EINVAL if the implementation isn't there?

-Scott

#14 lore raw

PATCH v7 4/5 mtd: nand: omap: optimized chip->ecc.correct() for H/W ECC schemes

In reply to Pekon Gupta

chip->ecc.correct() is used for detecting and correcting bit-flips during read operations. In omap-nand driver it implemented as:
(a) omap_correct_data(): for h/w based ECC_HAM1 scheme
(b) omap_correct_data_bch() + CONFIG_NAND_OMAP_ECC_BCH8_CODE_HW_DETECTION_SW for ECC_BCH8 scheme using GPMC and software lib/bch.c
(c) omap_correct_data_bch() + CONFIG_NAND_OMAP_ECC_BCH8_CODE_HW for ECC_BCH8 scheme using GPMC and ELM

This patch updates (c)
- checks for calc_ecc[]==0x00 so that error_correction is not required for known good pages.
- adds scalability for other ECC_BCHx scheme by merging following omap_rotate_ecc_bch() + omap_fix_errors_bch() => omap_correct_data_bch()
- fixing logic for bit-flip correction based on error_loc[count]

---
 drivers/mtd/nand/omap_gpmc.c | 159 +++++++++++++++++++------------------------
 1 file changed, 69 insertions(+), 90 deletions(-)
drivers/mtd/nand/omap_gpmc.c +69 -90
index 76be758..24e6f8f 100644
@@ -21,6 +21,9 @@
 #define SECTOR_BYTES		512
 #define ECCCLEAR		(0x1 << 8)
 #define ECCRESULTREG1		(0x1 << 0)
+/* 4 bit padding to make byte aligned, 56 = 52 + 4 */
+#define BCH4_BIT_PAD		4
+
 #ifdef CONFIG_BCH
 static u8  bch8_polynomial[] = {0xef, 0x51, 0x2e, 0x09, 0xed, 0x93, 0x9a, 0xc2,
 				0x97, 0x79, 0xe5, 0x24, 0xb5};
@@ -179,6 +182,23 @@ static __maybe_unused struct nand_bch_priv bch_priv = {
 };
 
 /*
+ * omap_reverse_list - re-orders list elements in reverse order [internal]
+ * @list:	pointer to start of list
+ * @length:	length of list
+*/
+void omap_reverse_list(u8 *list, unsigned int length)
+{
+	unsigned int i, j;
+	unsigned int half_length = length / 2;
+	u8 tmp;
+	for (i = 0, j = length - 1; i < half_length; i++, j--) {
+		tmp = list[i];
+		list[i] = list[j];
+		list[j] = tmp;
+	}
+}
+
+/*
  * omap_enable_hwecc - configures GPMC as per ECC scheme before read/write
  * @mtd:	MTD device structure
  * @mode:	Read/Write mode
@@ -309,77 +329,6 @@ static int omap_calculate_ecc(struct mtd_info *mtd, const uint8_t *dat,
 }
 
 /*
- * omap_rotate_ecc_bch - Rotate the syndrome bytes
- *
- * @mtd:	MTD device structure
- * @calc_ecc:	ECC read from ECC registers
- * @syndrome:	Rotated syndrome will be retuned in this array
- *
- */
-static void omap_rotate_ecc_bch(struct mtd_info *mtd, uint8_t *calc_ecc,
-		uint8_t *syndrome)
-{
-	struct nand_chip *chip = mtd->priv;
-	struct nand_bch_priv *bch = chip->priv;
-	uint8_t n_bytes = 0;
-	int8_t i, j;
-
-	switch (bch->type) {
-	case ECC_BCH4:
-		n_bytes = 8;
-		break;
-
-	case ECC_BCH16:
-		n_bytes = 28;
-		break;
-
-	case ECC_BCH8:
-	default:
-		n_bytes = 13;
-		break;
-	}
-
-	for (i = 0, j = (n_bytes-1); i < n_bytes; i++, j--)
-		syndrome[i] =  calc_ecc[j];
-}
-
-/*
- * omap_fix_errors_bch - Correct bch error in the data
- *
- * @mtd:	MTD device structure
- * @data:	Data read from flash
- * @error_count:Number of errors in data
- * @error_loc:	Locations of errors in the data
- *
- */
-static void omap_fix_errors_bch(struct mtd_info *mtd, uint8_t *data,
-		uint32_t error_count, uint32_t *error_loc)
-{
-	struct nand_chip *chip = mtd->priv;
-	struct nand_bch_priv *bch = chip->priv;
-	uint8_t count = 0;
-	uint32_t error_byte_pos;
-	uint32_t error_bit_mask;
-	uint32_t last_bit = (bch->nibbles * 4) - 1;
-
-	/* Flip all bits as specified by the error location array. */
-	/* FOR( each found error location flip the bit ) */
-	for (count = 0; count < error_count; count++) {
-		if (error_loc[count] > last_bit) {
-			/* Remove the ECC spare bits from correction. */
-			error_loc[count] -= (last_bit + 1);
-			/* Offset bit in data region */
-			error_byte_pos = ((512 * 8) -
-					(error_loc[count]) - 1) / 8;
-			/* Error Bit mask */
-			error_bit_mask = 0x1 << (error_loc[count] % 8);
-			/* Toggle the error bit to make the correction. */
-			data[error_byte_pos] ^= error_bit_mask;
-		}
-	}
-}
-
-/*
  * omap_correct_data_bch - Compares the ecc read from nand spare area
  * with ECC registers values and corrects one bit error if it has occured
  *
@@ -395,40 +344,70 @@ static int omap_correct_data_bch(struct mtd_info *mtd, uint8_t *dat,
 {
 	struct nand_chip *chip = mtd->priv;
 	struct nand_bch_priv *bch = chip->priv;
-	uint8_t syndrome[28];
-	uint32_t error_count = 0;
+	uint32_t eccbytes = chip->ecc.bytes;
+	uint32_t error_count = 0, error_max;
 	uint32_t error_loc[8];
-	uint32_t i, ecc_flag;
+	uint32_t i, ecc_flag = 0;
+	uint8_t count, err = 0;
+	uint32_t byte_pos, bit_pos;
+
+	/* check calculated ecc */
+	for (i = 0; i < chip->ecc.bytes && !ecc_flag; i++)
+		if (calc_ecc[i] != 0x00)
+			ecc_flag = 1;
+	if (!ecc_flag)
+		return 0;
 
+	/* check for whether its a erased-page */
 	ecc_flag = 0;
-	for (i = 0; i < chip->ecc.bytes; i++)
+	for (i = 0; i < chip->ecc.bytes && !ecc_flag; i++)
 		if (read_ecc[i] != 0xff)
 			ecc_flag = 1;
-
 	if (!ecc_flag)
 		return 0;
 
-	elm_reset();
-	elm_config((enum bch_level)(bch->type));
-
 	/*
 	 * while reading ECC result we read it in big endian.
 	 * Hence while loading to ELM we have rotate to get the right endian.
 	 */
-	omap_rotate_ecc_bch(mtd, calc_ecc, syndrome);
-
+	switch (bch->ecc_scheme) {
+	case OMAP_ECC_BCH8_CODE_HW:
+		omap_reverse_list(calc_ecc, eccbytes - 1);
+		break;
+	default:
+		return -EINVAL;
+	}
 	/* use elm module to check for errors */
-	if (elm_check_error(syndrome, bch->nibbles, &error_count,
-				error_loc) != 0) {
-		printf("ECC: uncorrectable.\n");
-		return -1;
+	elm_config((enum bch_level)(bch->type));
+	if (elm_check_error(calc_ecc, bch->nibbles, &error_count, error_loc)) {
+		printf("nand: error: uncorrectable ECC errors\n");
+		return -EINVAL;
 	}
-
 	/* correct bch error */
-	if (error_count > 0)
-		omap_fix_errors_bch(mtd, dat, error_count, error_loc);
-
-	return 0;
+	for (count = 0; count < error_count; count++) {
+		switch (bch->type) {
+		case ECC_BCH8:
+			/* 14th byte in ECC is reserved to match ROM layout */
+			error_max = SECTOR_BYTES + (eccbytes - 1);
+			break;
+		default:
+			return -EINVAL;
+		}
+		byte_pos = error_max - (error_loc[count] / 8) - 1;
+		bit_pos  = error_loc[count] % 8;
+		if (byte_pos < SECTOR_BYTES) {
+			dat[byte_pos] ^= 1 << bit_pos;
+			printf("nand: bit-flip corrected @data=%d\n", byte_pos);
+		} else if (byte_pos < error_max) {
+			read_ecc[byte_pos - SECTOR_BYTES] = 1 << bit_pos;
+			printf("nand: bit-flip corrected @oob=%d\n", byte_pos -
+								SECTOR_BYTES);
+		} else {
+			err = -EBADMSG;
+			printf("nand: error: invalid bit-flip location\n");
+		}
+	}
+	return (err) ? err : error_count;
 }
 
 /**
Signature · 1 line
1.8.1
#15 lore raw

PATCH v7 4/5 mtd: nand: omap: optimized chip->ecc.correct() for H/W ECC schemes

drivers/mtd/nand/omap_gpmc.c static int omap_correct_data_bch(struct mtd_info *mtd, uint8_t *dat, · line 357

Quoted 4 lines · Pekon Gupta
drivers/mtd/nand/omap_gpmc.c +4 -0
+	/* check calculated ecc */
+	for (i = 0; i < chip->ecc.bytes && !ecc_flag; i++)
+		if (calc_ecc[i] != 0x00)
+			ecc_flag = 1;

U-Boot code style wants braces around multi-line bodies.

-Scott

#16 lore raw

PATCH v7 5/5 board/ti/am335x/README: update for NAND boot

In reply to Pekon Gupta

NAND boot mode on AM335x EVM has been verified, and steps to use it has been documented and update in this README

---
 board/ti/am335x/README | 53 ++++++++++++++++++++++++++++++++++----------------
 1 file changed, 36 insertions(+), 17 deletions(-)
board/ti/am335x/README +36 -17
index 2a30ab8..947305b 100644
@@ -46,23 +46,42 @@ NAND
 The AM335x GP EVM ships with a 256MiB NAND available in most profiles.  In
 this example to program the NAND we assume that an SD card has been
 inserted with the files to write in the first SD slot and that mtdparts
-have been configured correctly for the board.  As a time saving measure we
-load MLO into memory in one location, copy it into the three locatations
-that the ROM checks for additional valid copies, then load U-Boot into
-memory.  We then write that whole section of memory to NAND.
-
-U-Boot # mmc rescan
-U-Boot # env default -f -a
-U-Boot # nand erase.chip
-U-Boot # saveenv
-U-Boot # load mmc 0 81000000 MLO
-U-Boot # cp.b 81000000 81020000 20000
-U-Boot # cp.b 81000000 81040000 20000
-U-Boot # cp.b 81000000 81060000 20000
-U-Boot # load mmc 0 81080000 u-boot.img
-U-Boot # nand write 81000000 0 260000
-U-Boot # load mmc 0 ${loadaddr} uImage
-U-Boot # nand write ${loadaddr} kernel 500000
+have been configured correctly for the board. All images are first loaded
+into memory, then written to NAND.
+
+Step-1: Building u-boot for NAND boot
+	Set following CONFIGxx options for NAND device.
+	CONFIG_SYS_NAND_PAGE_SIZE	number of main bytes in NAND page
+	CONFIG_SYS_NAND_OOBSIZE		number of OOB bytes in NAND page
+	CONFIG_SYS_NAND_BLOCK_SIZE	number of bytes in NAND erase-block
+	CONFIG_SYS_NAND_ECCPOS		ECC map for NAND page
+	CONFIG_NAND_OMAP_ECCSCHEME	(refer doc/README.nand)
+
+Step-2: Flashing NAND via MMC/SD
+	# select BOOTSEL to MMC/SD boot and boot from MMC/SD card
+	U-Boot # mmc rescan
+	# erase flash
+	U-Boot # nand erase.chip
+	U-Boot # env default -f -a
+	U-Boot # saveenv
+	# flash MLO. Redundant copies of MLO are kept for failsafe
+	U-Boot # load mmc 0 0x82000000 MLO
+	U-Boot # nand write 0x82000000 0x00000 0x20000
+	U-Boot # nand write 0x82000000 0x20000 0x20000
+	U-Boot # nand write 0x82000000 0x40000 0x20000
+	U-Boot # nand write 0x82000000 0x60000 0x20000
+	# flash u-boot.img
+	U-Boot # load mmc 0 0x82000000 u-boot.img
+	U-Boot # nand write 0x82000000 0x80000 0x60000
+	# flash kernel image
+	U-Boot # load mmc 0 0x82000000 uImage
+	U-Boot # nand write 0x82000000 ${nandsrcaddr} ${nandimgsize}
+	# flash filesystem image
+	U-Boot # load mmc 0 0x82000000 filesystem.img
+	U-Boot # nand write 0x82000000 ${loadaddress} 0x300000
+
+Step-3: Set BOOTSEL pin to select NAND boot, and POR the device.
+	The device should boot from images flashed on NAND device.
 
 NOR
 ===
Signature · 1 line
1.8.1
#17 lore raw

v7 cover mtd: nand: omap: optimize and clean-up of OMAP NAND driver

In reply to Pekon Gupta

Hi Scott, Tom,

Quoted 7 lines

Subject: [PATCH v7 0/5] mtd: nand: omap: optimize and clean-up of OMAP NAND driver

*changes in v7*
[PATCH 1/5]
- omap_gpmc.c: fix: free bytes in OOB (ecclayout-

Quoted 1 line

oobfree[0].length)

Quoted 13 lines

- omap_gpmc.c: cleanup: redundant code added in previous patch
versions
- am335x_evm.h: cleanup: redundant code added in previous patch
versions
- tricorder.h: fix: CONFIG_NAND_OMAP_ECCSCHEME
[PATCH 2/5] removed: re-configuration of gpmc.config1[dev_width] added in
previous version of patch
[PATCH 3/5] <no update>
[PATCH 4/5] <no update>
[PATCH 5/5] minor fix: missing '$' in ${loadaddr}

Request you to please accept this patch series if everything is ok ? This one is more of driver clean-up and restructuring.

And is independent of another patch-series.. [PATCH v2 0/4] enable support for x16 NAND devices

with regards, pekon

#18 lore raw

v7 cover mtd: nand: omap: optimize and clean-up of OMAP NAND driver

In reply to Pekon Gupta
Quoted 11 lines · Pekon Gupta

*changes in v7*
[PATCH 1/5]
- omap_gpmc.c: fix: free bytes in OOB (ecclayout->oobfree[0].length)
- omap_gpmc.c: cleanup: redundant code added in previous patch versions
- am335x_evm.h: cleanup: redundant code added in previous patch versions
- tricorder.h: fix: CONFIG_NAND_OMAP_ECCSCHEME
[PATCH 2/5] removed: re-configuration of gpmc.config1[dev_width] added in
previous version of patch
[PATCH 3/5] <no update>
[PATCH 4/5] <no update>
[PATCH 5/5] minor fix: missing '$' in ${loadaddr}

OK, with this series applied, please build all of the OMAP platforms, and fix the various build problems that are introduced by this.

I see problems with: am3517_crane am335x_boneblack rut igep0020_nand omap3_zoom2 omap3_overo dig297 omap3_evm omap3_logic igep0033 tricorder mcx igep0030_nand twister pxm2 omap3_evm_quick_nand am43xx_evm mt_ventoux ti816x_evm am335x_evm_spiboot pcm051 dxr2 eco5pk devkit8000 am3517_evm cm_t35 omap3_beagle omap3_zoom1 omap3_pandora

Signature · 8 lines
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: Digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20131017/41a589af/attachment.pgp>