* [PATCH v9 0/3] MTD: at91: Add PMECC support for at91 nand flash driver
@ 2012-05-26 13:24 ` Josh Wu
0 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-26 13:24 UTC (permalink / raw)
To: linux-mtd, linux-arm-kernel, dedekind1
Cc: hongxu.cn, nicolas.ferre, ivan.djelic, plagnioj, Josh Wu
Code is based on 3.4-rc2
Changes since v8,
use _relaxed read/write in most place. use writel in operations of Control Register since it needs memory barrier.
allocate the data for PMECC computation.
add pmecc prefix for related variable/functions.
modify code according to J.C's suggestion. except:
>> + for (i = 2; i <= 2 * host->cap; i += 2) {
> manage the j in the for loop
since that will change to:
+ for (i = 2, j = 1; i <= 2 * host->cap; i += 2, j = i / 2) {
it is not as simple as original one.
Changes since v7,
add time out for PMECC status reading.
modify the oobfree[0].offset to 2.
fix coding style.
Changes since v6,
split into 3 patches.
remove of_flat_dt_is_compatible() function. use additional dt parameter "has-pmecc".
refine the error handling code.
refine original atmel_nand_init_params() function.
Changes since v5,
add has_pmecc field to replace cpu_has_pmecc() function. Use compatible check in when proble.
simplify the pmecc_get_ecc_bytes() function.
Changes since v4,
fix typo and checkpatch warnings.
fix according to JC's suggestion. replace cpu_is_xxx() with DT
modify dt binding atmel nand document to add pmecc support.
tested in sam9263 without break hw ecc.
add ecc.strength.
Josh Wu (3):
extract the hw ecc function.
add DT variables.
add pmecc support
.../devicetree/bindings/mtd/atmel-nand.txt | 6 +
drivers/mtd/nand/atmel_nand.c | 939 ++++++++++++++++++--
drivers/mtd/nand/atmel_nand_ecc.h | 127 ++-
3 files changed, 997 insertions(+), 75 deletions(-)
--
1.7.10
^ permalink raw reply [flat|nested] 24+ messages in thread
* [PATCH v9 0/3] MTD: at91: Add PMECC support for at91 nand flash driver
@ 2012-05-26 13:24 ` Josh Wu
0 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-26 13:24 UTC (permalink / raw)
To: linux-arm-kernel
Code is based on 3.4-rc2
Changes since v8,
use _relaxed read/write in most place. use writel in operations of Control Register since it needs memory barrier.
allocate the data for PMECC computation.
add pmecc prefix for related variable/functions.
modify code according to J.C's suggestion. except:
>> + for (i = 2; i <= 2 * host->cap; i += 2) {
> manage the j in the for loop
since that will change to:
+ for (i = 2, j = 1; i <= 2 * host->cap; i += 2, j = i / 2) {
it is not as simple as original one.
Changes since v7,
add time out for PMECC status reading.
modify the oobfree[0].offset to 2.
fix coding style.
Changes since v6,
split into 3 patches.
remove of_flat_dt_is_compatible() function. use additional dt parameter "has-pmecc".
refine the error handling code.
refine original atmel_nand_init_params() function.
Changes since v5,
add has_pmecc field to replace cpu_has_pmecc() function. Use compatible check in when proble.
simplify the pmecc_get_ecc_bytes() function.
Changes since v4,
fix typo and checkpatch warnings.
fix according to JC's suggestion. replace cpu_is_xxx() with DT
modify dt binding atmel nand document to add pmecc support.
tested in sam9263 without break hw ecc.
add ecc.strength.
Josh Wu (3):
extract the hw ecc function.
add DT variables.
add pmecc support
.../devicetree/bindings/mtd/atmel-nand.txt | 6 +
drivers/mtd/nand/atmel_nand.c | 939 ++++++++++++++++++--
drivers/mtd/nand/atmel_nand_ecc.h | 127 ++-
3 files changed, 997 insertions(+), 75 deletions(-)
--
1.7.10
^ permalink raw reply [flat|nested] 24+ messages in thread
* [PATCH v9 1/3] MTD: at91: extract hw ecc initialization to one function and use relaxed read/write
2012-05-26 13:24 ` Josh Wu
@ 2012-05-26 13:24 ` Josh Wu
-1 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-26 13:24 UTC (permalink / raw)
To: linux-mtd, linux-arm-kernel, dedekind1
Cc: hongxu.cn, nicolas.ferre, ivan.djelic, plagnioj, Josh Wu
use _relaxed read/write in most place. And use writel in operations of Control Register since it needs memory barrier.
Signed-off-by: Hong Xu <hong.xu@atmel.com>
Signed-off-by: Josh Wu <josh.wu@atmel.com>
---
drivers/mtd/nand/atmel_nand.c | 158 ++++++++++++++++++++-----------------
drivers/mtd/nand/atmel_nand_ecc.h | 11 ++-
2 files changed, 94 insertions(+), 75 deletions(-)
diff --git a/drivers/mtd/nand/atmel_nand.c b/drivers/mtd/nand/atmel_nand.c
index 2165576..ba61153 100644
--- a/drivers/mtd/nand/atmel_nand.c
+++ b/drivers/mtd/nand/atmel_nand.c
@@ -42,20 +42,15 @@
#include <mach/cpu.h>
+/* Hardware ECC registers */
+#include "atmel_nand_ecc.h"
+
static int use_dma = 1;
module_param(use_dma, int, 0);
static int on_flash_bbt = 0;
module_param(on_flash_bbt, int, 0);
-/* Register access macros */
-#define ecc_readl(add, reg) \
- __raw_readl(add + ATMEL_ECC_##reg)
-#define ecc_writel(add, reg, value) \
- __raw_writel((value), add + ATMEL_ECC_##reg)
-
-#include "atmel_nand_ecc.h" /* Hardware ECC registers */
-
/* oob layout for large page size
* bad block info is on bytes 0 and 1
* the bytes have to be consecutives to avoid
@@ -304,13 +299,13 @@ static int atmel_nand_calculate(struct mtd_info *mtd,
unsigned int ecc_value;
/* get the first 2 ECC bytes */
- ecc_value = ecc_readl(host->ecc, PR);
+ ecc_value = ecc_readl_relaxed(host->ecc, PR);
ecc_code[0] = ecc_value & 0xFF;
ecc_code[1] = (ecc_value >> 8) & 0xFF;
/* get the last 2 ECC bytes */
- ecc_value = ecc_readl(host->ecc, NPR) & ATMEL_ECC_NPARITY;
+ ecc_value = ecc_readl_relaxed(host->ecc, NPR) & ATMEL_ECC_NPARITY;
ecc_code[2] = ecc_value & 0xFF;
ecc_code[3] = (ecc_value >> 8) & 0xFF;
@@ -406,16 +401,16 @@ static int atmel_nand_correct(struct mtd_info *mtd, u_char *dat,
unsigned int ecc_word, ecc_bit;
/* get the status from the Status Register */
- ecc_status = ecc_readl(host->ecc, SR);
+ ecc_status = ecc_readl_relaxed(host->ecc, SR);
/* if there's no error */
if (likely(!(ecc_status & ATMEL_ECC_RECERR)))
return 0;
/* get error bit offset (4 bits) */
- ecc_bit = ecc_readl(host->ecc, PR) & ATMEL_ECC_BITADDR;
+ ecc_bit = ecc_readl_relaxed(host->ecc, PR) & ATMEL_ECC_BITADDR;
/* get word address (12 bits) */
- ecc_word = ecc_readl(host->ecc, PR) & ATMEL_ECC_WORDADDR;
+ ecc_word = ecc_readl_relaxed(host->ecc, PR) & ATMEL_ECC_WORDADDR;
ecc_word >>= 4;
/* if there are multiple errors */
@@ -523,6 +518,76 @@ static int __devinit atmel_of_init_port(struct atmel_nand_host *host,
}
#endif
+static int __init atmel_hw_nand_init_params(struct platform_device *pdev,
+ struct atmel_nand_host *host)
+{
+ struct resource *regs;
+ struct mtd_info *mtd;
+ struct nand_chip *nand_chip;
+
+ nand_chip = &host->nand_chip;
+ mtd = &host->mtd;
+
+ nand_chip->ecc.mode = NAND_ECC_SOFT;
+ regs = platform_get_resource(pdev, IORESOURCE_MEM, 1);
+ if (!regs) {
+ dev_err(host->dev,
+ "Can't get I/O resource regs, use software ECC\n");
+ return 0;
+ }
+
+ host->ecc = ioremap(regs->start, resource_size(regs));
+ if (host->ecc == NULL) {
+ dev_err(host->dev, "ioremap failed\n");
+ return -EIO;
+ }
+
+ nand_chip->ecc.mode = NAND_ECC_HW;
+ nand_chip->ecc.calculate = atmel_nand_calculate;
+ nand_chip->ecc.correct = atmel_nand_correct;
+ nand_chip->ecc.hwctl = atmel_nand_hwctl;
+ nand_chip->ecc.read_page = atmel_nand_read_page;
+ nand_chip->ecc.bytes = 4;
+ nand_chip->ecc.strength = 1;
+
+ /* ECC is calculated for the whole page (1 step) */
+ nand_chip->ecc.size = mtd->writesize;
+
+ /* set ECC page size and oob layout */
+ switch (mtd->writesize) {
+ case 512:
+ nand_chip->ecc.layout = &atmel_oobinfo_small;
+ ecc_writel_relaxed(host->ecc, MR, ATMEL_ECC_PAGESIZE_528);
+ break;
+ case 1024:
+ nand_chip->ecc.layout = &atmel_oobinfo_large;
+ ecc_writel_relaxed(host->ecc, MR, ATMEL_ECC_PAGESIZE_1056);
+ break;
+ case 2048:
+ nand_chip->ecc.layout = &atmel_oobinfo_large;
+ ecc_writel_relaxed(host->ecc, MR, ATMEL_ECC_PAGESIZE_2112);
+ break;
+ case 4096:
+ nand_chip->ecc.layout = &atmel_oobinfo_large;
+ ecc_writel_relaxed(host->ecc, MR, ATMEL_ECC_PAGESIZE_4224);
+ break;
+ default:
+ /* page size not handled by HW ECC */
+ /* switching back to soft ECC */
+ nand_chip->ecc.mode = NAND_ECC_SOFT;
+ nand_chip->ecc.calculate = NULL;
+ nand_chip->ecc.correct = NULL;
+ nand_chip->ecc.hwctl = NULL;
+ nand_chip->ecc.read_page = NULL;
+ nand_chip->ecc.postpad = 0;
+ nand_chip->ecc.prepad = 0;
+ nand_chip->ecc.bytes = 0;
+ break;
+ }
+
+ return 0;
+}
+
/*
* Probe for the NAND device.
*/
@@ -531,7 +596,6 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
struct atmel_nand_host *host;
struct mtd_info *mtd;
struct nand_chip *nand_chip;
- struct resource *regs;
struct resource *mem;
struct mtd_part_parser_data ppdata = {};
int res;
@@ -583,29 +647,6 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
nand_chip->dev_ready = atmel_nand_device_ready;
nand_chip->ecc.mode = host->board.ecc_mode;
-
- regs = platform_get_resource(pdev, IORESOURCE_MEM, 1);
- if (!regs && nand_chip->ecc.mode == NAND_ECC_HW) {
- printk(KERN_ERR "atmel_nand: can't get I/O resource "
- "regs\nFalling back on software ECC\n");
- nand_chip->ecc.mode = NAND_ECC_SOFT;
- }
-
- if (nand_chip->ecc.mode == NAND_ECC_HW) {
- host->ecc = ioremap(regs->start, resource_size(regs));
- if (host->ecc == NULL) {
- printk(KERN_ERR "atmel_nand: ioremap failed\n");
- res = -EIO;
- goto err_ecc_ioremap;
- }
- nand_chip->ecc.calculate = atmel_nand_calculate;
- nand_chip->ecc.correct = atmel_nand_correct;
- nand_chip->ecc.hwctl = atmel_nand_hwctl;
- nand_chip->ecc.read_page = atmel_nand_read_page;
- nand_chip->ecc.bytes = 4;
- nand_chip->ecc.strength = 1;
- }
-
nand_chip->chip_delay = 20; /* 20us command delay time */
if (host->board.bus_width_16) /* 16-bit bus width */
@@ -657,40 +698,9 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
}
if (nand_chip->ecc.mode == NAND_ECC_HW) {
- /* ECC is calculated for the whole page (1 step) */
- nand_chip->ecc.size = mtd->writesize;
-
- /* set ECC page size and oob layout */
- switch (mtd->writesize) {
- case 512:
- nand_chip->ecc.layout = &atmel_oobinfo_small;
- ecc_writel(host->ecc, MR, ATMEL_ECC_PAGESIZE_528);
- break;
- case 1024:
- nand_chip->ecc.layout = &atmel_oobinfo_large;
- ecc_writel(host->ecc, MR, ATMEL_ECC_PAGESIZE_1056);
- break;
- case 2048:
- nand_chip->ecc.layout = &atmel_oobinfo_large;
- ecc_writel(host->ecc, MR, ATMEL_ECC_PAGESIZE_2112);
- break;
- case 4096:
- nand_chip->ecc.layout = &atmel_oobinfo_large;
- ecc_writel(host->ecc, MR, ATMEL_ECC_PAGESIZE_4224);
- break;
- default:
- /* page size not handled by HW ECC */
- /* switching back to soft ECC */
- nand_chip->ecc.mode = NAND_ECC_SOFT;
- nand_chip->ecc.calculate = NULL;
- nand_chip->ecc.correct = NULL;
- nand_chip->ecc.hwctl = NULL;
- nand_chip->ecc.read_page = NULL;
- nand_chip->ecc.postpad = 0;
- nand_chip->ecc.prepad = 0;
- nand_chip->ecc.bytes = 0;
- break;
- }
+ res = atmel_hw_nand_init_params(pdev, host);
+ if (res != 0)
+ goto err_hw_ecc;
}
/* second phase scan */
@@ -707,15 +717,15 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
return res;
err_scan_tail:
+ if (host->ecc)
+ iounmap(host->ecc);
+err_hw_ecc:
err_scan_ident:
err_no_card:
atmel_nand_disable(host);
platform_set_drvdata(pdev, NULL);
if (host->dma_chan)
dma_release_channel(host->dma_chan);
- if (host->ecc)
- iounmap(host->ecc);
-err_ecc_ioremap:
iounmap(host->io_base);
err_nand_ioremap:
kfree(host);
diff --git a/drivers/mtd/nand/atmel_nand_ecc.h b/drivers/mtd/nand/atmel_nand_ecc.h
index 578c776..5cbff11 100644
--- a/drivers/mtd/nand/atmel_nand_ecc.h
+++ b/drivers/mtd/nand/atmel_nand_ecc.h
@@ -3,7 +3,7 @@
* Based on AT91SAM9260 datasheet revision B.
*
* Copyright (C) 2007 Andrew Victor
- * Copyright (C) 2007 Atmel Corporation.
+ * Copyright (C) 2007 - 2012 Atmel Corporation.
*
* 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 the
@@ -36,4 +36,13 @@
#define ATMEL_ECC_NPR 0x10 /* NParity register */
#define ATMEL_ECC_NPARITY (0xffff << 0) /* NParity */
+/* Relaxed version of Register Access Macros */
+#define ecc_readl_relaxed(add, reg) \
+ readl_relaxed(add + ATMEL_ECC_##reg)
+#define ecc_writel_relaxed(add, reg, value) \
+ writel_relaxed((value), add + ATMEL_ECC_##reg)
+/* Register Write Macros with memory barries, use to write control regiser */
+#define ecc_writel(add, reg, value) \
+ writel((value), add + ATMEL_ECC_##reg)
+
#endif
--
1.7.10
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH v9 1/3] MTD: at91: extract hw ecc initialization to one function and use relaxed read/write
@ 2012-05-26 13:24 ` Josh Wu
0 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-26 13:24 UTC (permalink / raw)
To: linux-arm-kernel
use _relaxed read/write in most place. And use writel in operations of Control Register since it needs memory barrier.
Signed-off-by: Hong Xu <hong.xu@atmel.com>
Signed-off-by: Josh Wu <josh.wu@atmel.com>
---
drivers/mtd/nand/atmel_nand.c | 158 ++++++++++++++++++++-----------------
drivers/mtd/nand/atmel_nand_ecc.h | 11 ++-
2 files changed, 94 insertions(+), 75 deletions(-)
diff --git a/drivers/mtd/nand/atmel_nand.c b/drivers/mtd/nand/atmel_nand.c
index 2165576..ba61153 100644
--- a/drivers/mtd/nand/atmel_nand.c
+++ b/drivers/mtd/nand/atmel_nand.c
@@ -42,20 +42,15 @@
#include <mach/cpu.h>
+/* Hardware ECC registers */
+#include "atmel_nand_ecc.h"
+
static int use_dma = 1;
module_param(use_dma, int, 0);
static int on_flash_bbt = 0;
module_param(on_flash_bbt, int, 0);
-/* Register access macros */
-#define ecc_readl(add, reg) \
- __raw_readl(add + ATMEL_ECC_##reg)
-#define ecc_writel(add, reg, value) \
- __raw_writel((value), add + ATMEL_ECC_##reg)
-
-#include "atmel_nand_ecc.h" /* Hardware ECC registers */
-
/* oob layout for large page size
* bad block info is on bytes 0 and 1
* the bytes have to be consecutives to avoid
@@ -304,13 +299,13 @@ static int atmel_nand_calculate(struct mtd_info *mtd,
unsigned int ecc_value;
/* get the first 2 ECC bytes */
- ecc_value = ecc_readl(host->ecc, PR);
+ ecc_value = ecc_readl_relaxed(host->ecc, PR);
ecc_code[0] = ecc_value & 0xFF;
ecc_code[1] = (ecc_value >> 8) & 0xFF;
/* get the last 2 ECC bytes */
- ecc_value = ecc_readl(host->ecc, NPR) & ATMEL_ECC_NPARITY;
+ ecc_value = ecc_readl_relaxed(host->ecc, NPR) & ATMEL_ECC_NPARITY;
ecc_code[2] = ecc_value & 0xFF;
ecc_code[3] = (ecc_value >> 8) & 0xFF;
@@ -406,16 +401,16 @@ static int atmel_nand_correct(struct mtd_info *mtd, u_char *dat,
unsigned int ecc_word, ecc_bit;
/* get the status from the Status Register */
- ecc_status = ecc_readl(host->ecc, SR);
+ ecc_status = ecc_readl_relaxed(host->ecc, SR);
/* if there's no error */
if (likely(!(ecc_status & ATMEL_ECC_RECERR)))
return 0;
/* get error bit offset (4 bits) */
- ecc_bit = ecc_readl(host->ecc, PR) & ATMEL_ECC_BITADDR;
+ ecc_bit = ecc_readl_relaxed(host->ecc, PR) & ATMEL_ECC_BITADDR;
/* get word address (12 bits) */
- ecc_word = ecc_readl(host->ecc, PR) & ATMEL_ECC_WORDADDR;
+ ecc_word = ecc_readl_relaxed(host->ecc, PR) & ATMEL_ECC_WORDADDR;
ecc_word >>= 4;
/* if there are multiple errors */
@@ -523,6 +518,76 @@ static int __devinit atmel_of_init_port(struct atmel_nand_host *host,
}
#endif
+static int __init atmel_hw_nand_init_params(struct platform_device *pdev,
+ struct atmel_nand_host *host)
+{
+ struct resource *regs;
+ struct mtd_info *mtd;
+ struct nand_chip *nand_chip;
+
+ nand_chip = &host->nand_chip;
+ mtd = &host->mtd;
+
+ nand_chip->ecc.mode = NAND_ECC_SOFT;
+ regs = platform_get_resource(pdev, IORESOURCE_MEM, 1);
+ if (!regs) {
+ dev_err(host->dev,
+ "Can't get I/O resource regs, use software ECC\n");
+ return 0;
+ }
+
+ host->ecc = ioremap(regs->start, resource_size(regs));
+ if (host->ecc == NULL) {
+ dev_err(host->dev, "ioremap failed\n");
+ return -EIO;
+ }
+
+ nand_chip->ecc.mode = NAND_ECC_HW;
+ nand_chip->ecc.calculate = atmel_nand_calculate;
+ nand_chip->ecc.correct = atmel_nand_correct;
+ nand_chip->ecc.hwctl = atmel_nand_hwctl;
+ nand_chip->ecc.read_page = atmel_nand_read_page;
+ nand_chip->ecc.bytes = 4;
+ nand_chip->ecc.strength = 1;
+
+ /* ECC is calculated for the whole page (1 step) */
+ nand_chip->ecc.size = mtd->writesize;
+
+ /* set ECC page size and oob layout */
+ switch (mtd->writesize) {
+ case 512:
+ nand_chip->ecc.layout = &atmel_oobinfo_small;
+ ecc_writel_relaxed(host->ecc, MR, ATMEL_ECC_PAGESIZE_528);
+ break;
+ case 1024:
+ nand_chip->ecc.layout = &atmel_oobinfo_large;
+ ecc_writel_relaxed(host->ecc, MR, ATMEL_ECC_PAGESIZE_1056);
+ break;
+ case 2048:
+ nand_chip->ecc.layout = &atmel_oobinfo_large;
+ ecc_writel_relaxed(host->ecc, MR, ATMEL_ECC_PAGESIZE_2112);
+ break;
+ case 4096:
+ nand_chip->ecc.layout = &atmel_oobinfo_large;
+ ecc_writel_relaxed(host->ecc, MR, ATMEL_ECC_PAGESIZE_4224);
+ break;
+ default:
+ /* page size not handled by HW ECC */
+ /* switching back to soft ECC */
+ nand_chip->ecc.mode = NAND_ECC_SOFT;
+ nand_chip->ecc.calculate = NULL;
+ nand_chip->ecc.correct = NULL;
+ nand_chip->ecc.hwctl = NULL;
+ nand_chip->ecc.read_page = NULL;
+ nand_chip->ecc.postpad = 0;
+ nand_chip->ecc.prepad = 0;
+ nand_chip->ecc.bytes = 0;
+ break;
+ }
+
+ return 0;
+}
+
/*
* Probe for the NAND device.
*/
@@ -531,7 +596,6 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
struct atmel_nand_host *host;
struct mtd_info *mtd;
struct nand_chip *nand_chip;
- struct resource *regs;
struct resource *mem;
struct mtd_part_parser_data ppdata = {};
int res;
@@ -583,29 +647,6 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
nand_chip->dev_ready = atmel_nand_device_ready;
nand_chip->ecc.mode = host->board.ecc_mode;
-
- regs = platform_get_resource(pdev, IORESOURCE_MEM, 1);
- if (!regs && nand_chip->ecc.mode == NAND_ECC_HW) {
- printk(KERN_ERR "atmel_nand: can't get I/O resource "
- "regs\nFalling back on software ECC\n");
- nand_chip->ecc.mode = NAND_ECC_SOFT;
- }
-
- if (nand_chip->ecc.mode == NAND_ECC_HW) {
- host->ecc = ioremap(regs->start, resource_size(regs));
- if (host->ecc == NULL) {
- printk(KERN_ERR "atmel_nand: ioremap failed\n");
- res = -EIO;
- goto err_ecc_ioremap;
- }
- nand_chip->ecc.calculate = atmel_nand_calculate;
- nand_chip->ecc.correct = atmel_nand_correct;
- nand_chip->ecc.hwctl = atmel_nand_hwctl;
- nand_chip->ecc.read_page = atmel_nand_read_page;
- nand_chip->ecc.bytes = 4;
- nand_chip->ecc.strength = 1;
- }
-
nand_chip->chip_delay = 20; /* 20us command delay time */
if (host->board.bus_width_16) /* 16-bit bus width */
@@ -657,40 +698,9 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
}
if (nand_chip->ecc.mode == NAND_ECC_HW) {
- /* ECC is calculated for the whole page (1 step) */
- nand_chip->ecc.size = mtd->writesize;
-
- /* set ECC page size and oob layout */
- switch (mtd->writesize) {
- case 512:
- nand_chip->ecc.layout = &atmel_oobinfo_small;
- ecc_writel(host->ecc, MR, ATMEL_ECC_PAGESIZE_528);
- break;
- case 1024:
- nand_chip->ecc.layout = &atmel_oobinfo_large;
- ecc_writel(host->ecc, MR, ATMEL_ECC_PAGESIZE_1056);
- break;
- case 2048:
- nand_chip->ecc.layout = &atmel_oobinfo_large;
- ecc_writel(host->ecc, MR, ATMEL_ECC_PAGESIZE_2112);
- break;
- case 4096:
- nand_chip->ecc.layout = &atmel_oobinfo_large;
- ecc_writel(host->ecc, MR, ATMEL_ECC_PAGESIZE_4224);
- break;
- default:
- /* page size not handled by HW ECC */
- /* switching back to soft ECC */
- nand_chip->ecc.mode = NAND_ECC_SOFT;
- nand_chip->ecc.calculate = NULL;
- nand_chip->ecc.correct = NULL;
- nand_chip->ecc.hwctl = NULL;
- nand_chip->ecc.read_page = NULL;
- nand_chip->ecc.postpad = 0;
- nand_chip->ecc.prepad = 0;
- nand_chip->ecc.bytes = 0;
- break;
- }
+ res = atmel_hw_nand_init_params(pdev, host);
+ if (res != 0)
+ goto err_hw_ecc;
}
/* second phase scan */
@@ -707,15 +717,15 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
return res;
err_scan_tail:
+ if (host->ecc)
+ iounmap(host->ecc);
+err_hw_ecc:
err_scan_ident:
err_no_card:
atmel_nand_disable(host);
platform_set_drvdata(pdev, NULL);
if (host->dma_chan)
dma_release_channel(host->dma_chan);
- if (host->ecc)
- iounmap(host->ecc);
-err_ecc_ioremap:
iounmap(host->io_base);
err_nand_ioremap:
kfree(host);
diff --git a/drivers/mtd/nand/atmel_nand_ecc.h b/drivers/mtd/nand/atmel_nand_ecc.h
index 578c776..5cbff11 100644
--- a/drivers/mtd/nand/atmel_nand_ecc.h
+++ b/drivers/mtd/nand/atmel_nand_ecc.h
@@ -3,7 +3,7 @@
* Based on AT91SAM9260 datasheet revision B.
*
* Copyright (C) 2007 Andrew Victor
- * Copyright (C) 2007 Atmel Corporation.
+ * Copyright (C) 2007 - 2012 Atmel Corporation.
*
* 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 the
@@ -36,4 +36,13 @@
#define ATMEL_ECC_NPR 0x10 /* NParity register */
#define ATMEL_ECC_NPARITY (0xffff << 0) /* NParity */
+/* Relaxed version of Register Access Macros */
+#define ecc_readl_relaxed(add, reg) \
+ readl_relaxed(add + ATMEL_ECC_##reg)
+#define ecc_writel_relaxed(add, reg, value) \
+ writel_relaxed((value), add + ATMEL_ECC_##reg)
+/* Register Write Macros with memory barries, use to write control regiser */
+#define ecc_writel(add, reg, value) \
+ writel((value), add + ATMEL_ECC_##reg)
+
#endif
--
1.7.10
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH v9 2/3] MTD: at91: add dt parameters for PMECC
2012-05-26 13:24 ` Josh Wu
@ 2012-05-26 13:24 ` Josh Wu
-1 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-26 13:24 UTC (permalink / raw)
To: linux-mtd, linux-arm-kernel, dedekind1
Cc: hongxu.cn, nicolas.ferre, ivan.djelic, plagnioj, Josh Wu
Signed-off-by: Hong Xu <hong.xu@atmel.com>
Signed-off-by: Josh Wu <josh.wu@atmel.com>
---
.../devicetree/bindings/mtd/atmel-nand.txt | 6 ++++++
drivers/mtd/nand/atmel_nand.c | 22 ++++++++++++++++++++
2 files changed, 28 insertions(+)
diff --git a/Documentation/devicetree/bindings/mtd/atmel-nand.txt b/Documentation/devicetree/bindings/mtd/atmel-nand.txt
index a200695..e442236 100644
--- a/Documentation/devicetree/bindings/mtd/atmel-nand.txt
+++ b/Documentation/devicetree/bindings/mtd/atmel-nand.txt
@@ -16,6 +16,12 @@ Optional properties:
- nand-ecc-mode : String, operation mode of the NAND ecc mode, soft by default.
Supported values are: "none", "soft", "hw", "hw_syndrome", "hw_oob_first",
"soft_bch".
+- atmel,has-pmecc : boolean to enable Programmable Multibit ECC hardware.
+ Only supported by at91sam9x5 or later sam9 product.
+- atmel,pmecc-cap : error correct capability for Programmable Multibit ECC
+ Controller. Supported values are: 2, 4, 8, 12, 24.
+- atmel,pmecc-sector-size : sector size for ECC computation. Supported values
+ are: 512, 1024.
- nand-bus-width : 8 or 16 bus width if not present 8
- nand-on-flash-bbt: boolean to enable on flash bbt option if not present false
diff --git a/drivers/mtd/nand/atmel_nand.c b/drivers/mtd/nand/atmel_nand.c
index ba61153..9a9bfbf 100644
--- a/drivers/mtd/nand/atmel_nand.c
+++ b/drivers/mtd/nand/atmel_nand.c
@@ -88,6 +88,10 @@ struct atmel_nand_host {
struct completion comp;
struct dma_chan *dma_chan;
+
+ bool has_pmecc;
+ u8 pmecc_corr_cap;
+ u16 pmecc_sector_size;
};
static int cpu_has_dma(void)
@@ -508,6 +512,24 @@ static int __devinit atmel_of_init_port(struct atmel_nand_host *host,
board->enable_pin = of_get_gpio(np, 1);
board->det_pin = of_get_gpio(np, 2);
+ host->has_pmecc = of_property_read_bool(np, "atmel,has-pmecc");
+
+ if (!(board->ecc_mode == NAND_ECC_HW) || !host->has_pmecc)
+ return 0;
+
+ /* Use PMECC */
+ if (of_property_read_u32(np, "atmel,pmecc-cap", &val) != 0) {
+ dev_err(host->dev, "Cannot decide PMECC Capability\n");
+ return -EINVAL;
+ }
+ host->pmecc_corr_cap = (u8)val;
+
+ if (of_property_read_u32(np, "atmel,pmecc-sector-size", &val) != 0) {
+ dev_err(host->dev, "Cannot decide PMECC Sector Size\n");
+ return -EINVAL;
+ }
+ host->pmecc_sector_size = (u16)val;
+
return 0;
}
#else
--
1.7.10
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH v9 2/3] MTD: at91: add dt parameters for PMECC
@ 2012-05-26 13:24 ` Josh Wu
0 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-26 13:24 UTC (permalink / raw)
To: linux-arm-kernel
Signed-off-by: Hong Xu <hong.xu@atmel.com>
Signed-off-by: Josh Wu <josh.wu@atmel.com>
---
.../devicetree/bindings/mtd/atmel-nand.txt | 6 ++++++
drivers/mtd/nand/atmel_nand.c | 22 ++++++++++++++++++++
2 files changed, 28 insertions(+)
diff --git a/Documentation/devicetree/bindings/mtd/atmel-nand.txt b/Documentation/devicetree/bindings/mtd/atmel-nand.txt
index a200695..e442236 100644
--- a/Documentation/devicetree/bindings/mtd/atmel-nand.txt
+++ b/Documentation/devicetree/bindings/mtd/atmel-nand.txt
@@ -16,6 +16,12 @@ Optional properties:
- nand-ecc-mode : String, operation mode of the NAND ecc mode, soft by default.
Supported values are: "none", "soft", "hw", "hw_syndrome", "hw_oob_first",
"soft_bch".
+- atmel,has-pmecc : boolean to enable Programmable Multibit ECC hardware.
+ Only supported by at91sam9x5 or later sam9 product.
+- atmel,pmecc-cap : error correct capability for Programmable Multibit ECC
+ Controller. Supported values are: 2, 4, 8, 12, 24.
+- atmel,pmecc-sector-size : sector size for ECC computation. Supported values
+ are: 512, 1024.
- nand-bus-width : 8 or 16 bus width if not present 8
- nand-on-flash-bbt: boolean to enable on flash bbt option if not present false
diff --git a/drivers/mtd/nand/atmel_nand.c b/drivers/mtd/nand/atmel_nand.c
index ba61153..9a9bfbf 100644
--- a/drivers/mtd/nand/atmel_nand.c
+++ b/drivers/mtd/nand/atmel_nand.c
@@ -88,6 +88,10 @@ struct atmel_nand_host {
struct completion comp;
struct dma_chan *dma_chan;
+
+ bool has_pmecc;
+ u8 pmecc_corr_cap;
+ u16 pmecc_sector_size;
};
static int cpu_has_dma(void)
@@ -508,6 +512,24 @@ static int __devinit atmel_of_init_port(struct atmel_nand_host *host,
board->enable_pin = of_get_gpio(np, 1);
board->det_pin = of_get_gpio(np, 2);
+ host->has_pmecc = of_property_read_bool(np, "atmel,has-pmecc");
+
+ if (!(board->ecc_mode == NAND_ECC_HW) || !host->has_pmecc)
+ return 0;
+
+ /* Use PMECC */
+ if (of_property_read_u32(np, "atmel,pmecc-cap", &val) != 0) {
+ dev_err(host->dev, "Cannot decide PMECC Capability\n");
+ return -EINVAL;
+ }
+ host->pmecc_corr_cap = (u8)val;
+
+ if (of_property_read_u32(np, "atmel,pmecc-sector-size", &val) != 0) {
+ dev_err(host->dev, "Cannot decide PMECC Sector Size\n");
+ return -EINVAL;
+ }
+ host->pmecc_sector_size = (u16)val;
+
return 0;
}
#else
--
1.7.10
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH v9 3/3] MTD: at91: atmel_nand: Update driver to support Programmable Multibit ECC controller
2012-05-26 13:24 ` Josh Wu
@ 2012-05-26 13:24 ` Josh Wu
-1 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-26 13:24 UTC (permalink / raw)
To: linux-mtd, linux-arm-kernel, dedekind1
Cc: hongxu.cn, nicolas.ferre, ivan.djelic, plagnioj, Josh Wu
The Programmable Multibit ECC (PMECC) controller is a programmable binary
BCH(Bose, Chaudhuri and Hocquenghem) encoder and decoder. This controller
can be used to support both SLC and MLC NAND Flash devices. It supports to
generate ECC to correct 2, 4, 8, 12 or 24 bits of error per sector of data.
To use this driver, the user needs to pass in the correction capability and
the sector size.
This driver has been tested on AT91SAM9X5-EK and AT91SAM9N12-EK with JFFS2,
YAFFS2, UBIFS and mtd-utils.
Signed-off-by: Hong Xu <hong.xu@atmel.com>
Signed-off-by: Josh Wu <josh.wu@atmel.com>
---
drivers/mtd/nand/atmel_nand.c | 761 ++++++++++++++++++++++++++++++++++++-
drivers/mtd/nand/atmel_nand_ecc.h | 116 ++++++
2 files changed, 876 insertions(+), 1 deletion(-)
diff --git a/drivers/mtd/nand/atmel_nand.c b/drivers/mtd/nand/atmel_nand.c
index 9a9bfbf..ddcf1ed 100644
--- a/drivers/mtd/nand/atmel_nand.c
+++ b/drivers/mtd/nand/atmel_nand.c
@@ -15,6 +15,8 @@
* (u-boot-1.1.5/board/atmel/at91sam9263ek/nand.c)
* (C) Copyright 2006 ATMEL Rousset, Lacressonniere Nicolas
*
+ * Add Programmable Multibit ECC support for various AT91 SoC
+ * (C) Copyright 2012 ATMEL, Hong Xu
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
@@ -77,6 +79,21 @@ static struct nand_ecclayout atmel_oobinfo_small = {
},
};
+/* a structure includes datas for PMECC computation */
+struct atmel_pmecc_data {
+ int16_t partial_syn[2 * PMECC_MAX_ERROR_NB + 1];
+ int16_t si[2 * PMECC_MAX_ERROR_NB + 1];
+
+ /* Sigma table */
+ int16_t smu[PMECC_MAX_ERROR_NB + 2][2 * PMECC_MAX_ERROR_NB + 1];
+ /* polynomal order */
+ int16_t lmu[PMECC_MAX_ERROR_NB + 1];
+
+ int mu[PMECC_MAX_ERROR_NB + 1];
+ int dmu[PMECC_MAX_ERROR_NB + 1];
+ int delta[PMECC_MAX_ERROR_NB + 1];
+};
+
struct atmel_nand_host {
struct nand_chip nand_chip;
struct mtd_info mtd;
@@ -92,8 +109,25 @@ struct atmel_nand_host {
bool has_pmecc;
u8 pmecc_corr_cap;
u16 pmecc_sector_size;
+
+ int pmecc_bytes_per_sector;
+ int pmecc_sector_number;
+ int pmecc_degree; /* Degree of remainders */
+ int pmecc_cw_len; /* Length of codeword */
+
+ void __iomem *pmerrloc_base;
+ void __iomem *pmecc_rom_base;
+
+ /* lookup table for alpha_to and index_of */
+ void __iomem *pmecc_alpha_to;
+ void __iomem *pmecc_index_of;
+
+ /* data for pmecc computation */
+ struct atmel_pmecc_data *pmecc_data;
};
+static struct nand_ecclayout atmel_pmecc_oobinfo;
+
static int cpu_has_dma(void)
{
return cpu_is_at91sam9rl() || cpu_is_at91sam9g45();
@@ -287,6 +321,708 @@ static void atmel_write_buf(struct mtd_info *mtd, const u8 *buf, int len)
}
/*
+ * Return number of ecc bytes per sector according to sector size and
+ * correction capability
+ *
+ * Following table shows what at91 PMECC supported:
+ * Correction Capability Sector_512_bytes Sector_1024_bytes
+ * ===================== ================ =================
+ * 2-bits 4-bytes 4-bytes
+ * 4-bits 7-bytes 7-bytes
+ * 8-bits 13-bytes 14-bytes
+ * 12-bits 20-bytes 21-bytes
+ * 24-bits 39-bytes 42-bytes
+ */
+static int pmecc_get_ecc_bytes(int cap, int sector_size)
+{
+ int m = 12 + sector_size / 512;
+ return (m * cap + 7) / 8;
+}
+
+static void pmecc_config_ecc_layout(struct nand_ecclayout *layout, int oobsize,
+ int ecc_len)
+{
+ int i;
+
+ layout->eccbytes = ecc_len;
+
+ /* ECC will occupy the last ecc_len bytes continuously */
+ for (i = 0; i < ecc_len; i++)
+ layout->eccpos[i] = oobsize - ecc_len + i;
+
+ layout->oobfree[0].offset = 2;
+ layout->oobfree[0].length =
+ oobsize - ecc_len - layout->oobfree[0].offset;
+}
+
+static void __iomem *pmecc_get_alpha_to(struct atmel_nand_host *host)
+{
+ void __iomem *p;
+
+ switch (host->pmecc_sector_size) {
+ case 512:
+ p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_512 +
+ PMECC_LOOKUP_TABLE_SIZE_512 * sizeof(int16_t);
+ break;
+ case 1024:
+ p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_1024 +
+ PMECC_LOOKUP_TABLE_SIZE_1024 * sizeof(int16_t);
+ break;
+ default:
+ BUG();
+ }
+
+ return p;
+}
+
+static void __iomem *pmecc_get_index_of(struct atmel_nand_host *host)
+{
+ void __iomem *p;
+
+ switch (host->pmecc_sector_size) {
+ case 512:
+ p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_512;
+ break;
+ case 1024:
+ p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_1024;
+ break;
+ default:
+ BUG();
+ }
+
+ return p;
+}
+
+static void pmecc_gen_syndrome(struct mtd_info *mtd, int sector)
+{
+ int i;
+ uint32_t value;
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+
+ /* Fill odd syndromes */
+ for (i = 0; i < host->pmecc_corr_cap; i++) {
+ value = pmecc_readl_rem_relaxed(host->ecc, sector, i / 2);
+ value = (i & 1) ? (value & 0xffff0000) >> 16 : value & 0xffff;
+ host->pmecc_data->partial_syn[(2 * i) + 1] = (int16_t)value;
+ }
+}
+
+static void pmecc_substitute(struct mtd_info *mtd)
+{
+ int16_t *si;
+ int i, j;
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+ int16_t __iomem *alpha_to = host->pmecc_alpha_to;
+ int16_t __iomem *index_of = host->pmecc_index_of;
+ int16_t *partial_syn = host->pmecc_data->partial_syn;
+
+ /* si[] is a table that holds the current syndrome value,
+ * an element of that table belongs to the field
+ */
+ si = host->pmecc_data->si;
+
+ for (i = 1; i < 2 * PMECC_MAX_ERROR_NB; i++)
+ si[i] = 0;
+
+ /* Computation 2t syndromes based on S(x) */
+ /* Odd syndromes */
+ for (i = 1; i < 2 * host->pmecc_corr_cap; i += 2) {
+ si[i] = 0;
+ for (j = 0; j < host->pmecc_degree; j++) {
+ if (partial_syn[i] & ((unsigned short)0x1 << j))
+ si[i] = readw_relaxed(alpha_to + i * j) ^ si[i];
+ }
+ }
+ /* Even syndrome = (Odd syndrome) ** 2 */
+ for (i = 2; i <= 2 * host->pmecc_corr_cap; i += 2) {
+ j = i / 2;
+ if (si[j] == 0)
+ si[i] = 0;
+ else {
+ int16_t tmp;
+ tmp = readw_relaxed(index_of + si[j]);
+ tmp = (tmp * 2) % host->pmecc_cw_len;
+ si[i] = readw_relaxed(alpha_to + tmp);
+ }
+ }
+
+ return;
+}
+
+static void pmecc_get_sigma(struct mtd_info *mtd)
+{
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+
+ int i, j, k;
+ uint32_t dmu_0_count, tmp;
+ int16_t (*smu)[2 * PMECC_MAX_ERROR_NB + 1];
+ int16_t *lmu = host->pmecc_data->lmu;
+ int16_t *si = host->pmecc_data->si;
+ int *mu = host->pmecc_data->mu;
+ int *dmu = host->pmecc_data->dmu; /* Discrepancy */
+ int *delta = host->pmecc_data->delta; /* Delta order */
+ int cw_len = host->pmecc_cw_len;
+ int16_t cap = host->pmecc_corr_cap;
+
+ int16_t __iomem *index_of = host->pmecc_index_of;
+ int16_t __iomem *alpha_to = host->pmecc_alpha_to;
+
+ /* index of largest delta */
+ int ro;
+ int largest;
+ int diff;
+
+ dmu_0_count = 0;
+ smu = host->pmecc_data->smu;
+
+ /* First Row */
+
+ /* Mu */
+ mu[0] = -1;
+
+ memset(&smu[0][0], 0,
+ sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
+ smu[0][0] = 1;
+
+ /* discrepancy set to 1 */
+ dmu[0] = 1;
+ /* polynom order set to 0 */
+ lmu[0] = 0;
+ delta[0] = (mu[0] * 2 - lmu[0]) >> 1;
+
+ /* Second Row */
+
+ /* Mu */
+ mu[1] = 0;
+ /* Sigma(x) set to 1 */
+ memset(&smu[1][0], 0,
+ sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
+ smu[1][0] = 1;
+
+ /* discrepancy set to S1 */
+ dmu[1] = si[1];
+
+ /* polynom order set to 0 */
+ lmu[1] = 0;
+
+ delta[1] = (mu[1] * 2 - lmu[1]) >> 1;
+
+ /* Init the Sigma(x) last row */
+ memset(&smu[cap + 1][0], 0,
+ sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
+
+ for (i = 1; i <= cap; i++) {
+ mu[i+1] = i << 1;
+ /* Begin Computing Sigma (Mu+1) and L(mu) */
+ /* check if discrepancy is set to 0 */
+ if (dmu[i] == 0) {
+ dmu_0_count++;
+
+ tmp = ((cap - (lmu[i] >> 1) - 1) / 2);
+ if ((cap - (lmu[i] >> 1) - 1) & 0x1)
+ tmp += 2;
+ else
+ tmp += 1;
+
+ if (dmu_0_count == tmp) {
+ for (j = 0; j <= (lmu[i] >> 1) + 1; j++)
+ smu[cap + 1][j] = smu[i][j];
+ lmu[cap + 1] = lmu[i];
+ return;
+ }
+
+ /* copy polynom */
+ for (j = 0; j <= lmu[i] >> 1; j++)
+ smu[i + 1][j] = smu[i][j];
+
+ /* copy previous polynom order to the next */
+ lmu[i + 1] = lmu[i];
+ } else {
+ ro = 0;
+ largest = -1;
+ /* find largest delta with dmu != 0 */
+ for (j = 0; j < i; j++) {
+ if ((dmu[j]) && (delta[j] > largest)) {
+ largest = delta[j];
+ ro = j;
+ }
+ }
+
+ /* compute difference */
+ diff = (mu[i] - mu[ro]);
+
+ /* Compute degree of the new smu polynomial */
+ if ((lmu[i] >> 1) > ((lmu[ro] >> 1) + diff))
+ lmu[i + 1] = lmu[i];
+ else
+ lmu[i + 1] = ((lmu[ro] >> 1) + diff) * 2;
+
+ /* Init smu[i+1] with 0 */
+ for (k = 0; k < (2 * PMECC_MAX_ERROR_NB + 1); k++)
+ smu[i+1][k] = 0;
+
+ /* Compute smu[i+1] */
+ for (k = 0; k <= lmu[ro] >> 1; k++) {
+ int16_t a, b, c;
+
+ if (!(smu[ro][k] && dmu[i]))
+ continue;
+ a = readw_relaxed(index_of + dmu[i]);
+ b = readw_relaxed(index_of + dmu[ro]);
+ c = readw_relaxed(index_of + smu[ro][k]);
+ tmp = a + (cw_len - b) + c;
+ a = readw_relaxed(alpha_to + tmp % cw_len);
+ smu[i + 1][k + diff] = a;
+ }
+
+ for (k = 0; k <= lmu[i] >> 1; k++)
+ smu[i + 1][k] ^= smu[i][k];
+ }
+
+ /* End Computing Sigma (Mu+1) and L(mu) */
+ /* In either case compute delta */
+ delta[i + 1] = (mu[i + 1] * 2 - lmu[i + 1]) >> 1;
+
+ /* Do not compute discrepancy for the last iteration */
+ if (i >= cap)
+ continue;
+
+ for (k = 0 ; k <= (lmu[i + 1] >> 1); k++) {
+ tmp = 2 * (i - 1);
+ if (k == 0)
+ dmu[i + 1] = si[tmp + 3];
+ else if (smu[i+1][k] && si[tmp + 3 - k]) {
+ int16_t a, b, c;
+ a = readw_relaxed(index_of + smu[i + 1][k]);
+ b = si[2 * (i - 1) + 3 - k];
+ c = readw_relaxed(index_of + b);
+ tmp = a + c;
+ tmp %= cw_len;
+ dmu[i + 1] = readw_relaxed(alpha_to + tmp) ^
+ dmu[i + 1];
+ }
+ }
+ }
+
+ return;
+}
+
+static int pmecc_err_location(struct mtd_info *mtd)
+{
+ int i;
+ int err_nbr; /* number of error */
+ int roots_nbr; /* number of roots */
+ int sector_size;
+ uint32_t val;
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+ int timeout_count = 0;
+ int cap = host->pmecc_corr_cap;
+
+ err_nbr = 0;
+ sector_size = host->pmecc_sector_size;
+
+ pmerrloc_writel(host->pmerrloc_base, ELDIS, PMERRLOC_DISABLE);
+
+ for (i = 0; i <= host->pmecc_data->lmu[cap + 1] >> 1; i++) {
+ pmerrloc_writel_sigma_relaxed(host->pmerrloc_base, i,
+ host->pmecc_data->smu[cap + 1][i]);
+ err_nbr++;
+ }
+
+ val = (err_nbr - 1) << 16;
+ if (sector_size == 1024)
+ val |= 1;
+
+ pmerrloc_writel(host->pmerrloc_base, ELCFG, val);
+ pmerrloc_writel(host->pmerrloc_base, ELEN,
+ sector_size * 8 + host->pmecc_degree * cap);
+
+ while (!(pmerrloc_readl_relaxed(host->pmerrloc_base, ELISR)
+ & PMERRLOC_CALC_DONE)) {
+ if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT))
+ return -1; /* Time out */
+ cpu_relax();
+ }
+
+ roots_nbr = (pmerrloc_readl_relaxed(host->pmerrloc_base, ELISR)
+ & PMERRLOC_ERR_NUM_MASK) >> 8;
+ /* Number of roots == degree of smu hence <= cap */
+ if (roots_nbr == host->pmecc_data->lmu[cap + 1] >> 1)
+ return err_nbr - 1;
+
+ /* Number of roots does not match the degree of smu
+ * unable to correct error */
+ return -1;
+}
+
+static void pmecc_correct_data(struct mtd_info *mtd, uint8_t *buf,
+ int extra_bytes, int err_nbr)
+{
+ int i = 0;
+ int byte_pos, bit_pos;
+ int sector_size, ecc_size;
+ uint32_t tmp;
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+
+ sector_size = host->pmecc_sector_size;
+ ecc_size = nand_chip->ecc.bytes;
+
+ while (err_nbr) {
+ tmp = pmerrloc_readl_el_relaxed(host->pmerrloc_base, i) - 1;
+ byte_pos = tmp / 8;
+ bit_pos = tmp % 8;
+ dev_info(host->dev, "PMECC correction, byte_pos: %d bit_pos: %d\n",
+ byte_pos, bit_pos);
+
+ if (byte_pos < (sector_size + extra_bytes)) {
+ tmp = sector_size +
+ pmecc_readl_relaxed(host->ecc, SADDR);
+
+ if (byte_pos < tmp)
+ *(buf + byte_pos) ^= (1 << bit_pos);
+ else
+ *(buf + byte_pos + ecc_size) ^= (1 << bit_pos);
+ }
+
+ i++;
+ err_nbr--;
+ }
+
+ return;
+}
+
+static int pmecc_correction(struct mtd_info *mtd, u32 pmecc_stat, uint8_t *buf,
+ u8 *ecc)
+{
+ int i, err_nbr;
+ uint8_t *buf_pos;
+ int eccbytes;
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+
+ eccbytes = nand_chip->ecc.bytes;
+ for (i = 0; i < eccbytes; i++)
+ if (ecc[i] != 0xff)
+ goto normal_check;
+ /* Erased page, return OK */
+ return 0;
+
+normal_check:
+ for (i = 0; i < host->pmecc_sector_number; i++) {
+ err_nbr = 0;
+ if (pmecc_stat & 0x1) {
+ buf_pos = buf + i * host->pmecc_sector_size;
+
+ pmecc_gen_syndrome(mtd, i);
+ pmecc_substitute(mtd);
+ pmecc_get_sigma(mtd);
+
+ err_nbr = pmecc_err_location(mtd);
+ if (err_nbr == -1) {
+ dev_err(host->dev, "PMECC: Too many errors\n");
+ mtd->ecc_stats.failed++;
+ return -EIO;
+ } else {
+ pmecc_correct_data(mtd, buf_pos, 0, err_nbr);
+ mtd->ecc_stats.corrected += err_nbr;
+ }
+ }
+ pmecc_stat >>= 1;
+ }
+
+ return 0;
+}
+
+static int atmel_nand_pmecc_read_page(struct mtd_info *mtd,
+ struct nand_chip *chip, uint8_t *buf, int page)
+{
+ uint32_t stat;
+ int timeout_count = 0;
+ int eccsize = chip->ecc.size;
+ uint8_t *oob = chip->oob_poi;
+ struct atmel_nand_host *host = chip->priv;
+ uint32_t *eccpos = chip->ecc.layout->eccpos;
+
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
+ pmecc_writel(host->ecc, CFG, (pmecc_readl_relaxed(host->ecc, CFG)
+ & ~PMECC_CFG_WRITE_OP) | PMECC_CFG_AUTO_ENABLE);
+
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DATA);
+
+ chip->read_buf(mtd, buf, eccsize);
+ chip->read_buf(mtd, oob, mtd->oobsize);
+
+ while ((pmecc_readl_relaxed(host->ecc, SR) & PMECC_SR_BUSY)) {
+ if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT))
+ return -EIO; /* Time out */
+ cpu_relax();
+ }
+
+ stat = pmecc_readl_relaxed(host->ecc, ISR);
+ if (stat != 0) {
+ if (pmecc_correction(mtd, stat, buf, &oob[eccpos[0]]) != 0)
+ return -EIO;
+ }
+
+ return 0;
+}
+
+static void atmel_nand_pmecc_write_page(struct mtd_info *mtd,
+ struct nand_chip *chip, const uint8_t *buf)
+{
+ int i, j;
+ int timeout_count = 0;
+ struct atmel_nand_host *host = chip->priv;
+ uint32_t *eccpos = chip->ecc.layout->eccpos;
+
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
+
+ pmecc_writel(host->ecc, CFG, (pmecc_readl_relaxed(host->ecc, CFG) |
+ PMECC_CFG_WRITE_OP) & ~PMECC_CFG_AUTO_ENABLE);
+
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DATA);
+
+ chip->write_buf(mtd, (u8 *)buf, mtd->writesize);
+
+ while ((pmecc_readl_relaxed(host->ecc, SR) & PMECC_SR_BUSY)) {
+ if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT)) {
+ dev_err(host->dev, "PMECC: Timeout to get ECC value.\n");
+ return; /* Time out */
+ }
+ cpu_relax();
+ }
+
+ for (i = 0; i < host->pmecc_sector_number; i++) {
+ for (j = 0; j < host->pmecc_bytes_per_sector; j++) {
+ int pos;
+
+ pos = i * host->pmecc_bytes_per_sector + j;
+ chip->oob_poi[eccpos[pos]] =
+ pmecc_readb_ecc_relaxed(host->ecc, i, j);
+ }
+ }
+ chip->write_buf(mtd, chip->oob_poi, mtd->oobsize);
+
+ return;
+}
+
+static void atmel_pmecc_core_init(struct mtd_info *mtd)
+{
+ uint32_t val = 0;
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+ struct nand_ecclayout *ecc_layout;
+
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
+
+ switch (host->pmecc_corr_cap) {
+ case 2:
+ val = PMECC_CFG_BCH_ERR2;
+ break;
+ case 4:
+ val = PMECC_CFG_BCH_ERR4;
+ break;
+ case 8:
+ val = PMECC_CFG_BCH_ERR8;
+ break;
+ case 12:
+ val = PMECC_CFG_BCH_ERR12;
+ break;
+ case 24:
+ val = PMECC_CFG_BCH_ERR24;
+ break;
+ }
+
+ if (host->pmecc_sector_size == 512)
+ val |= PMECC_CFG_SECTOR512;
+ else if (host->pmecc_sector_size == 1024)
+ val |= PMECC_CFG_SECTOR1024;
+
+ switch (host->pmecc_sector_number) {
+ case 1:
+ val |= PMECC_CFG_PAGE_1SECTOR;
+ break;
+ case 2:
+ val |= PMECC_CFG_PAGE_2SECTORS;
+ break;
+ case 4:
+ val |= PMECC_CFG_PAGE_4SECTORS;
+ break;
+ case 8:
+ val |= PMECC_CFG_PAGE_8SECTORS;
+ break;
+ }
+
+ val |= (PMECC_CFG_READ_OP | PMECC_CFG_SPARE_DISABLE
+ | PMECC_CFG_AUTO_DISABLE);
+ pmecc_writel(host->ecc, CFG, val);
+
+ ecc_layout = nand_chip->ecc.layout;
+ pmecc_writel(host->ecc, SAREA, mtd->oobsize - 1);
+ pmecc_writel(host->ecc, SADDR, ecc_layout->eccpos[0]);
+ pmecc_writel(host->ecc, EADDR,
+ ecc_layout->eccpos[ecc_layout->eccbytes - 1]);
+ /* See datasheet about PMECC Clock Control Register */
+ pmecc_writel(host->ecc, CLK, 2);
+ pmecc_writel(host->ecc, IDR, 0xff);
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
+}
+
+static int __init atmel_pmecc_nand_init_params(struct platform_device *pdev,
+ struct atmel_nand_host *host)
+{
+ int cap, sector_size, err_no;
+ struct mtd_info *mtd;
+ struct nand_chip *nand_chip;
+ struct resource *regs;
+ struct resource *regs_pmerr, *regs_rom;
+
+ cap = host->pmecc_corr_cap;
+ sector_size = host->pmecc_sector_size;
+ dev_info(host->dev, "Initialize PMECC params, cap: %d, sector: %d\n",
+ cap, sector_size);
+
+ /* Sanity check */
+ if ((sector_size != 512) && (sector_size != 1024)) {
+ dev_err(host->dev,
+ "Unsupported PMECC sector size: %d; should be 512 or 1024 bytes\n",
+ sector_size);
+ return -EINVAL;
+ }
+ if ((cap != 2) && (cap != 4) && (cap != 8) && (cap != 12) &&
+ (cap != 24)) {
+ dev_err(host->dev,
+ "Unsupported PMECC correction capability, should be 2, 4, 8, 12 or 24\n");
+ return -EINVAL;
+ }
+
+ nand_chip = &host->nand_chip;
+ mtd = &host->mtd;
+
+ nand_chip->ecc.mode = NAND_ECC_SOFT; /* By default */
+
+ regs = platform_get_resource(pdev, IORESOURCE_MEM, 1);
+ if (!regs) {
+ dev_warn(host->dev,
+ "Can't get I/O resource regs, rolling back on software ECC\n");
+ return 0;
+ }
+
+ host->ecc = ioremap(regs->start, resource_size(regs));
+ if (host->ecc == NULL) {
+ dev_err(host->dev, "ioremap failed\n");
+ err_no = -EIO;
+ goto err_pmecc_ioremap;
+ }
+
+ regs_pmerr = platform_get_resource(pdev, IORESOURCE_MEM, 2);
+ regs_rom = platform_get_resource(pdev, IORESOURCE_MEM, 3);
+ if (regs_pmerr && regs_rom) {
+ host->pmerrloc_base = ioremap(regs_pmerr->start,
+ resource_size(regs_pmerr));
+ host->pmecc_rom_base = ioremap(regs_rom->start,
+ resource_size(regs_rom));
+
+ if (host->pmerrloc_base && host->pmecc_rom_base) {
+ nand_chip->ecc.mode = NAND_ECC_HW;
+ nand_chip->ecc.read_page =
+ atmel_nand_pmecc_read_page;
+ nand_chip->ecc.write_page =
+ atmel_nand_pmecc_write_page;
+ } else {
+ dev_err(host->dev,
+ "Can not get I/O resource for PMECC controller!\n");
+ err_no = -EIO;
+ goto err_pmloc_ioremap;
+ }
+ }
+
+ /* ECC is calculated for the whole page (1 step) */
+ nand_chip->ecc.size = mtd->writesize;
+
+ /* set ECC page size and oob layout */
+ switch (mtd->writesize) {
+ case 2048:
+ host->pmecc_degree = PMECC_GF_DIMENSION_13;
+ host->pmecc_cw_len = (1 << host->pmecc_degree) - 1;
+ host->pmecc_corr_cap = cap;
+ host->pmecc_sector_number = mtd->writesize / sector_size;
+ host->pmecc_bytes_per_sector = pmecc_get_ecc_bytes(
+ cap, sector_size);
+ host->pmecc_alpha_to = pmecc_get_alpha_to(host);
+ host->pmecc_index_of = pmecc_get_index_of(host);
+
+ nand_chip->ecc.steps = 1;
+ nand_chip->ecc.strength = cap;
+ nand_chip->ecc.bytes = host->pmecc_bytes_per_sector *
+ host->pmecc_sector_number;
+ if (nand_chip->ecc.bytes > mtd->oobsize - 2) {
+ dev_err(host->dev, "No room for ECC bytes\n");
+ err_no = -EINVAL;
+ goto err;
+ }
+ pmecc_config_ecc_layout(&atmel_pmecc_oobinfo,
+ mtd->oobsize,
+ nand_chip->ecc.bytes);
+ nand_chip->ecc.layout = &atmel_pmecc_oobinfo;
+ break;
+ case 512:
+ case 1024:
+ case 4096:
+ /* TODO */
+ dev_warn(host->dev,
+ "Unsupported page size for PMECC, use Software ECC\n");
+ default:
+ /* page size not handled by HW ECC */
+ /* switching back to soft ECC */
+ nand_chip->ecc.mode = NAND_ECC_SOFT;
+ nand_chip->ecc.calculate = NULL;
+ nand_chip->ecc.correct = NULL;
+ nand_chip->ecc.hwctl = NULL;
+ nand_chip->ecc.read_page = NULL;
+ nand_chip->ecc.write_page = NULL;
+ nand_chip->ecc.postpad = 0;
+ nand_chip->ecc.prepad = 0;
+ nand_chip->ecc.bytes = 0;
+ err_no = 0;
+ goto err;
+ }
+
+ /* Allocate data for PMECC computation */
+ host->pmecc_data = kzalloc(sizeof(struct atmel_pmecc_data), GFP_KERNEL);
+ if (!host->pmecc_data) {
+ dev_err(host->dev,
+ "Cannot allocate memory for PMECC computation!\n");
+ err_no = -ENOMEM;
+ goto err;
+ }
+
+ atmel_pmecc_core_init(mtd);
+
+ return 0;
+
+err:
+err_pmloc_ioremap:
+ iounmap(host->ecc);
+ if (host->pmerrloc_base)
+ iounmap(host->pmerrloc_base);
+ if (host->pmecc_rom_base)
+ iounmap(host->pmecc_rom_base);
+err_pmecc_ioremap:
+ return err_no;
+}
+
+/*
* Calculate HW ECC
*
* function called after a write
@@ -720,7 +1456,11 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
}
if (nand_chip->ecc.mode == NAND_ECC_HW) {
- res = atmel_hw_nand_init_params(pdev, host);
+ if (host->has_pmecc)
+ res = atmel_pmecc_nand_init_params(pdev, host);
+ else
+ res = atmel_hw_nand_init_params(pdev, host);
+
if (res != 0)
goto err_hw_ecc;
}
@@ -741,6 +1481,12 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
err_scan_tail:
if (host->ecc)
iounmap(host->ecc);
+ if (host->has_pmecc) {
+ if (host->pmerrloc_base)
+ iounmap(host->pmerrloc_base);
+ if (host->pmecc_rom_base)
+ iounmap(host->pmecc_rom_base);
+ }
err_hw_ecc:
err_scan_ident:
err_no_card:
@@ -766,6 +1512,19 @@ static int __exit atmel_nand_remove(struct platform_device *pdev)
atmel_nand_disable(host);
+ if (host->has_pmecc) {
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
+ if (host->pmerrloc_base) {
+ pmerrloc_writel(host->pmerrloc_base, ELDIS,
+ PMERRLOC_DISABLE);
+ iounmap(host->pmerrloc_base);
+ }
+ if (host->pmecc_rom_base)
+ iounmap(host->pmecc_rom_base);
+
+ kfree(host->pmecc_data);
+ }
+
if (host->ecc)
iounmap(host->ecc);
diff --git a/drivers/mtd/nand/atmel_nand_ecc.h b/drivers/mtd/nand/atmel_nand_ecc.h
index 5cbff11..bd404bf 100644
--- a/drivers/mtd/nand/atmel_nand_ecc.h
+++ b/drivers/mtd/nand/atmel_nand_ecc.h
@@ -45,4 +45,120 @@
#define ecc_writel(add, reg, value) \
writel((value), add + ATMEL_ECC_##reg)
+/* PMECC Register Definitions */
+#define ATMEL_PMECC_CFG 0x000 /* Configuration Register */
+#define PMECC_CFG_BCH_ERR2 (0 << 0)
+#define PMECC_CFG_BCH_ERR4 (1 << 0)
+#define PMECC_CFG_BCH_ERR8 (2 << 0)
+#define PMECC_CFG_BCH_ERR12 (3 << 0)
+#define PMECC_CFG_BCH_ERR24 (4 << 0)
+
+#define PMECC_CFG_SECTOR512 (0 << 4)
+#define PMECC_CFG_SECTOR1024 (1 << 4)
+
+#define PMECC_CFG_PAGE_1SECTOR (0 << 8)
+#define PMECC_CFG_PAGE_2SECTORS (1 << 8)
+#define PMECC_CFG_PAGE_4SECTORS (2 << 8)
+#define PMECC_CFG_PAGE_8SECTORS (3 << 8)
+
+#define PMECC_CFG_READ_OP (0 << 12)
+#define PMECC_CFG_WRITE_OP (1 << 12)
+
+#define PMECC_CFG_SPARE_ENABLE (1 << 16)
+#define PMECC_CFG_SPARE_DISABLE (0 << 16)
+
+#define PMECC_CFG_AUTO_ENABLE (1 << 20)
+#define PMECC_CFG_AUTO_DISABLE (0 << 20)
+
+#define ATMEL_PMECC_SAREA 0x004 /* Spare area size */
+#define ATMEL_PMECC_SADDR 0x008 /* PMECC starting address */
+#define ATMEL_PMECC_EADDR 0x00c /* PMECC ending address */
+#define ATMEL_PMECC_CLK 0x010 /* PMECC clock control */
+#define PMECC_CLK_133MHZ (2 << 0)
+
+#define ATMEL_PMECC_CTRL 0x014 /* PMECC control register */
+#define PMECC_CTRL_RST (1 << 0)
+#define PMECC_CTRL_DATA (1 << 1)
+#define PMECC_CTRL_USER (1 << 2)
+#define PMECC_CTRL_ENABLE (1 << 4)
+#define PMECC_CTRL_DISABLE (1 << 5)
+
+#define ATMEL_PMECC_SR 0x018 /* PMECC status register */
+#define PMECC_SR_BUSY (1 << 0)
+#define PMECC_SR_ENABLE (1 << 4)
+
+#define ATMEL_PMECC_IER 0x01c /* PMECC interrupt enable */
+#define PMECC_IER_ENABLE (1 << 0)
+#define ATMEL_PMECC_IDR 0x020 /* PMECC interrupt disable */
+#define PMECC_IER_DISABLE (1 << 0)
+#define ATMEL_PMECC_IMR 0x024 /* PMECC interrupt mask */
+#define PMECC_IER_MASK (1 << 0)
+#define ATMEL_PMECC_ISR 0x028 /* PMECC interrupt status */
+#define ATMEL_PMECC_ECCx 0x040 /* PMECC ECC x */
+#define ATMEL_PMECC_REMx 0x240 /* PMECC REM x */
+
+/* PMERRLOC Register Definitions */
+#define ATMEL_PMERRLOC_ELCFG 0x000 /* Error location config */
+#define PMERRLOC_ELCFG_SECTOR_512 (0 << 0)
+#define PMERRLOC_ELCFG_SECTOR_1024 (1 << 0)
+#define PMERRLOC_ELCFG_NUM_ERRORS(n) ((n) << 16)
+
+#define ATMEL_PMERRLOC_ELPRIM 0x004 /* Error location primitive */
+#define ATMEL_PMERRLOC_ELEN 0x008 /* Error location enable */
+#define ATMEL_PMERRLOC_ELDIS 0x00c /* Error location disable */
+#define PMERRLOC_DISABLE (1 << 0)
+
+#define ATMEL_PMERRLOC_ELSR 0x010 /* Error location status */
+#define PMERRLOC_ELSR_BUSY (1 << 0)
+#define ATMEL_PMERRLOC_ELIER 0x014 /* Error location int enable */
+#define ATMEL_PMERRLOC_ELIDR 0x018 /* Error location int disable */
+#define ATMEL_PMERRLOC_ELIMR 0x01c /* Error location int mask */
+#define ATMEL_PMERRLOC_ELISR 0x020 /* Error location int status */
+#define PMERRLOC_ERR_NUM_MASK (0x1f << 8)
+#define PMERRLOC_CALC_DONE (1 << 0)
+#define ATMEL_PMERRLOC_SIGMAx 0x028 /* Error location SIGMA x */
+#define ATMEL_PMERRLOC_ELx 0x08c /* Error location x */
+
+/* Register access macros for PMECC */
+#define pmecc_readl_relaxed(addr, reg) \
+ readl_relaxed((addr) + ATMEL_PMECC_##reg)
+
+#define pmecc_writel(addr, reg, value) \
+ writel((value), (addr) + ATMEL_PMECC_##reg)
+
+#define pmecc_readb_ecc_relaxed(addr, sector, n) \
+ readb_relaxed((addr) + ATMEL_PMECC_ECCx + ((sector) * 0x40) + (n))
+
+#define pmecc_readl_rem_relaxed(addr, sector, n) \
+ readl_relaxed((addr) + ATMEL_PMECC_REMx + ((sector) * 0x40) + ((n) * 4))
+
+#define pmerrloc_readl_relaxed(addr, reg) \
+ readl_relaxed((addr) + ATMEL_PMERRLOC_##reg)
+
+#define pmerrloc_writel(addr, reg, value) \
+ writel((value), (addr) + ATMEL_PMERRLOC_##reg)
+
+#define pmerrloc_writel_sigma_relaxed(addr, n, value) \
+ writel_relaxed((value), (addr) + ATMEL_PMERRLOC_SIGMAx + ((n) * 4))
+
+#define pmerrloc_readl_sigma_relaxed(addr, n) \
+ readl_relaxed((addr) + ATMEL_PMERRLOC_SIGMAx + ((n) * 4))
+
+#define pmerrloc_readl_el_relaxed(addr, n) \
+ readl_relaxed((addr) + ATMEL_PMERRLOC_ELx + ((n) * 4))
+
+/* Galois field dimension */
+#define PMECC_GF_DIMENSION_13 13
+#define PMECC_GF_DIMENSION_14 14
+
+#define PMECC_MAX_ERROR_NB 25
+#define PMECC_MAX_ECC_BYTES 42
+#define PMECC_MAX_NB_SECTOR 8
+#define PMECC_MAX_TIMEOUT_COUNT 100
+
+#define PMECC_LOOKUP_TABLE_OFFSET_512 0x8000
+#define PMECC_LOOKUP_TABLE_SIZE_512 0x2000
+#define PMECC_LOOKUP_TABLE_OFFSET_1024 0x10000
+#define PMECC_LOOKUP_TABLE_SIZE_1024 0x4000
+
#endif
--
1.7.10
^ permalink raw reply related [flat|nested] 24+ messages in thread
* [PATCH v9 3/3] MTD: at91: atmel_nand: Update driver to support Programmable Multibit ECC controller
@ 2012-05-26 13:24 ` Josh Wu
0 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-26 13:24 UTC (permalink / raw)
To: linux-arm-kernel
The Programmable Multibit ECC (PMECC) controller is a programmable binary
BCH(Bose, Chaudhuri and Hocquenghem) encoder and decoder. This controller
can be used to support both SLC and MLC NAND Flash devices. It supports to
generate ECC to correct 2, 4, 8, 12 or 24 bits of error per sector of data.
To use this driver, the user needs to pass in the correction capability and
the sector size.
This driver has been tested on AT91SAM9X5-EK and AT91SAM9N12-EK with JFFS2,
YAFFS2, UBIFS and mtd-utils.
Signed-off-by: Hong Xu <hong.xu@atmel.com>
Signed-off-by: Josh Wu <josh.wu@atmel.com>
---
drivers/mtd/nand/atmel_nand.c | 761 ++++++++++++++++++++++++++++++++++++-
drivers/mtd/nand/atmel_nand_ecc.h | 116 ++++++
2 files changed, 876 insertions(+), 1 deletion(-)
diff --git a/drivers/mtd/nand/atmel_nand.c b/drivers/mtd/nand/atmel_nand.c
index 9a9bfbf..ddcf1ed 100644
--- a/drivers/mtd/nand/atmel_nand.c
+++ b/drivers/mtd/nand/atmel_nand.c
@@ -15,6 +15,8 @@
* (u-boot-1.1.5/board/atmel/at91sam9263ek/nand.c)
* (C) Copyright 2006 ATMEL Rousset, Lacressonniere Nicolas
*
+ * Add Programmable Multibit ECC support for various AT91 SoC
+ * (C) Copyright 2012 ATMEL, Hong Xu
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
@@ -77,6 +79,21 @@ static struct nand_ecclayout atmel_oobinfo_small = {
},
};
+/* a structure includes datas for PMECC computation */
+struct atmel_pmecc_data {
+ int16_t partial_syn[2 * PMECC_MAX_ERROR_NB + 1];
+ int16_t si[2 * PMECC_MAX_ERROR_NB + 1];
+
+ /* Sigma table */
+ int16_t smu[PMECC_MAX_ERROR_NB + 2][2 * PMECC_MAX_ERROR_NB + 1];
+ /* polynomal order */
+ int16_t lmu[PMECC_MAX_ERROR_NB + 1];
+
+ int mu[PMECC_MAX_ERROR_NB + 1];
+ int dmu[PMECC_MAX_ERROR_NB + 1];
+ int delta[PMECC_MAX_ERROR_NB + 1];
+};
+
struct atmel_nand_host {
struct nand_chip nand_chip;
struct mtd_info mtd;
@@ -92,8 +109,25 @@ struct atmel_nand_host {
bool has_pmecc;
u8 pmecc_corr_cap;
u16 pmecc_sector_size;
+
+ int pmecc_bytes_per_sector;
+ int pmecc_sector_number;
+ int pmecc_degree; /* Degree of remainders */
+ int pmecc_cw_len; /* Length of codeword */
+
+ void __iomem *pmerrloc_base;
+ void __iomem *pmecc_rom_base;
+
+ /* lookup table for alpha_to and index_of */
+ void __iomem *pmecc_alpha_to;
+ void __iomem *pmecc_index_of;
+
+ /* data for pmecc computation */
+ struct atmel_pmecc_data *pmecc_data;
};
+static struct nand_ecclayout atmel_pmecc_oobinfo;
+
static int cpu_has_dma(void)
{
return cpu_is_at91sam9rl() || cpu_is_at91sam9g45();
@@ -287,6 +321,708 @@ static void atmel_write_buf(struct mtd_info *mtd, const u8 *buf, int len)
}
/*
+ * Return number of ecc bytes per sector according to sector size and
+ * correction capability
+ *
+ * Following table shows what at91 PMECC supported:
+ * Correction Capability Sector_512_bytes Sector_1024_bytes
+ * ===================== ================ =================
+ * 2-bits 4-bytes 4-bytes
+ * 4-bits 7-bytes 7-bytes
+ * 8-bits 13-bytes 14-bytes
+ * 12-bits 20-bytes 21-bytes
+ * 24-bits 39-bytes 42-bytes
+ */
+static int pmecc_get_ecc_bytes(int cap, int sector_size)
+{
+ int m = 12 + sector_size / 512;
+ return (m * cap + 7) / 8;
+}
+
+static void pmecc_config_ecc_layout(struct nand_ecclayout *layout, int oobsize,
+ int ecc_len)
+{
+ int i;
+
+ layout->eccbytes = ecc_len;
+
+ /* ECC will occupy the last ecc_len bytes continuously */
+ for (i = 0; i < ecc_len; i++)
+ layout->eccpos[i] = oobsize - ecc_len + i;
+
+ layout->oobfree[0].offset = 2;
+ layout->oobfree[0].length =
+ oobsize - ecc_len - layout->oobfree[0].offset;
+}
+
+static void __iomem *pmecc_get_alpha_to(struct atmel_nand_host *host)
+{
+ void __iomem *p;
+
+ switch (host->pmecc_sector_size) {
+ case 512:
+ p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_512 +
+ PMECC_LOOKUP_TABLE_SIZE_512 * sizeof(int16_t);
+ break;
+ case 1024:
+ p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_1024 +
+ PMECC_LOOKUP_TABLE_SIZE_1024 * sizeof(int16_t);
+ break;
+ default:
+ BUG();
+ }
+
+ return p;
+}
+
+static void __iomem *pmecc_get_index_of(struct atmel_nand_host *host)
+{
+ void __iomem *p;
+
+ switch (host->pmecc_sector_size) {
+ case 512:
+ p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_512;
+ break;
+ case 1024:
+ p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_1024;
+ break;
+ default:
+ BUG();
+ }
+
+ return p;
+}
+
+static void pmecc_gen_syndrome(struct mtd_info *mtd, int sector)
+{
+ int i;
+ uint32_t value;
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+
+ /* Fill odd syndromes */
+ for (i = 0; i < host->pmecc_corr_cap; i++) {
+ value = pmecc_readl_rem_relaxed(host->ecc, sector, i / 2);
+ value = (i & 1) ? (value & 0xffff0000) >> 16 : value & 0xffff;
+ host->pmecc_data->partial_syn[(2 * i) + 1] = (int16_t)value;
+ }
+}
+
+static void pmecc_substitute(struct mtd_info *mtd)
+{
+ int16_t *si;
+ int i, j;
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+ int16_t __iomem *alpha_to = host->pmecc_alpha_to;
+ int16_t __iomem *index_of = host->pmecc_index_of;
+ int16_t *partial_syn = host->pmecc_data->partial_syn;
+
+ /* si[] is a table that holds the current syndrome value,
+ * an element of that table belongs to the field
+ */
+ si = host->pmecc_data->si;
+
+ for (i = 1; i < 2 * PMECC_MAX_ERROR_NB; i++)
+ si[i] = 0;
+
+ /* Computation 2t syndromes based on S(x) */
+ /* Odd syndromes */
+ for (i = 1; i < 2 * host->pmecc_corr_cap; i += 2) {
+ si[i] = 0;
+ for (j = 0; j < host->pmecc_degree; j++) {
+ if (partial_syn[i] & ((unsigned short)0x1 << j))
+ si[i] = readw_relaxed(alpha_to + i * j) ^ si[i];
+ }
+ }
+ /* Even syndrome = (Odd syndrome) ** 2 */
+ for (i = 2; i <= 2 * host->pmecc_corr_cap; i += 2) {
+ j = i / 2;
+ if (si[j] == 0)
+ si[i] = 0;
+ else {
+ int16_t tmp;
+ tmp = readw_relaxed(index_of + si[j]);
+ tmp = (tmp * 2) % host->pmecc_cw_len;
+ si[i] = readw_relaxed(alpha_to + tmp);
+ }
+ }
+
+ return;
+}
+
+static void pmecc_get_sigma(struct mtd_info *mtd)
+{
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+
+ int i, j, k;
+ uint32_t dmu_0_count, tmp;
+ int16_t (*smu)[2 * PMECC_MAX_ERROR_NB + 1];
+ int16_t *lmu = host->pmecc_data->lmu;
+ int16_t *si = host->pmecc_data->si;
+ int *mu = host->pmecc_data->mu;
+ int *dmu = host->pmecc_data->dmu; /* Discrepancy */
+ int *delta = host->pmecc_data->delta; /* Delta order */
+ int cw_len = host->pmecc_cw_len;
+ int16_t cap = host->pmecc_corr_cap;
+
+ int16_t __iomem *index_of = host->pmecc_index_of;
+ int16_t __iomem *alpha_to = host->pmecc_alpha_to;
+
+ /* index of largest delta */
+ int ro;
+ int largest;
+ int diff;
+
+ dmu_0_count = 0;
+ smu = host->pmecc_data->smu;
+
+ /* First Row */
+
+ /* Mu */
+ mu[0] = -1;
+
+ memset(&smu[0][0], 0,
+ sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
+ smu[0][0] = 1;
+
+ /* discrepancy set to 1 */
+ dmu[0] = 1;
+ /* polynom order set to 0 */
+ lmu[0] = 0;
+ delta[0] = (mu[0] * 2 - lmu[0]) >> 1;
+
+ /* Second Row */
+
+ /* Mu */
+ mu[1] = 0;
+ /* Sigma(x) set to 1 */
+ memset(&smu[1][0], 0,
+ sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
+ smu[1][0] = 1;
+
+ /* discrepancy set to S1 */
+ dmu[1] = si[1];
+
+ /* polynom order set to 0 */
+ lmu[1] = 0;
+
+ delta[1] = (mu[1] * 2 - lmu[1]) >> 1;
+
+ /* Init the Sigma(x) last row */
+ memset(&smu[cap + 1][0], 0,
+ sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
+
+ for (i = 1; i <= cap; i++) {
+ mu[i+1] = i << 1;
+ /* Begin Computing Sigma (Mu+1) and L(mu) */
+ /* check if discrepancy is set to 0 */
+ if (dmu[i] == 0) {
+ dmu_0_count++;
+
+ tmp = ((cap - (lmu[i] >> 1) - 1) / 2);
+ if ((cap - (lmu[i] >> 1) - 1) & 0x1)
+ tmp += 2;
+ else
+ tmp += 1;
+
+ if (dmu_0_count == tmp) {
+ for (j = 0; j <= (lmu[i] >> 1) + 1; j++)
+ smu[cap + 1][j] = smu[i][j];
+ lmu[cap + 1] = lmu[i];
+ return;
+ }
+
+ /* copy polynom */
+ for (j = 0; j <= lmu[i] >> 1; j++)
+ smu[i + 1][j] = smu[i][j];
+
+ /* copy previous polynom order to the next */
+ lmu[i + 1] = lmu[i];
+ } else {
+ ro = 0;
+ largest = -1;
+ /* find largest delta with dmu != 0 */
+ for (j = 0; j < i; j++) {
+ if ((dmu[j]) && (delta[j] > largest)) {
+ largest = delta[j];
+ ro = j;
+ }
+ }
+
+ /* compute difference */
+ diff = (mu[i] - mu[ro]);
+
+ /* Compute degree of the new smu polynomial */
+ if ((lmu[i] >> 1) > ((lmu[ro] >> 1) + diff))
+ lmu[i + 1] = lmu[i];
+ else
+ lmu[i + 1] = ((lmu[ro] >> 1) + diff) * 2;
+
+ /* Init smu[i+1] with 0 */
+ for (k = 0; k < (2 * PMECC_MAX_ERROR_NB + 1); k++)
+ smu[i+1][k] = 0;
+
+ /* Compute smu[i+1] */
+ for (k = 0; k <= lmu[ro] >> 1; k++) {
+ int16_t a, b, c;
+
+ if (!(smu[ro][k] && dmu[i]))
+ continue;
+ a = readw_relaxed(index_of + dmu[i]);
+ b = readw_relaxed(index_of + dmu[ro]);
+ c = readw_relaxed(index_of + smu[ro][k]);
+ tmp = a + (cw_len - b) + c;
+ a = readw_relaxed(alpha_to + tmp % cw_len);
+ smu[i + 1][k + diff] = a;
+ }
+
+ for (k = 0; k <= lmu[i] >> 1; k++)
+ smu[i + 1][k] ^= smu[i][k];
+ }
+
+ /* End Computing Sigma (Mu+1) and L(mu) */
+ /* In either case compute delta */
+ delta[i + 1] = (mu[i + 1] * 2 - lmu[i + 1]) >> 1;
+
+ /* Do not compute discrepancy for the last iteration */
+ if (i >= cap)
+ continue;
+
+ for (k = 0 ; k <= (lmu[i + 1] >> 1); k++) {
+ tmp = 2 * (i - 1);
+ if (k == 0)
+ dmu[i + 1] = si[tmp + 3];
+ else if (smu[i+1][k] && si[tmp + 3 - k]) {
+ int16_t a, b, c;
+ a = readw_relaxed(index_of + smu[i + 1][k]);
+ b = si[2 * (i - 1) + 3 - k];
+ c = readw_relaxed(index_of + b);
+ tmp = a + c;
+ tmp %= cw_len;
+ dmu[i + 1] = readw_relaxed(alpha_to + tmp) ^
+ dmu[i + 1];
+ }
+ }
+ }
+
+ return;
+}
+
+static int pmecc_err_location(struct mtd_info *mtd)
+{
+ int i;
+ int err_nbr; /* number of error */
+ int roots_nbr; /* number of roots */
+ int sector_size;
+ uint32_t val;
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+ int timeout_count = 0;
+ int cap = host->pmecc_corr_cap;
+
+ err_nbr = 0;
+ sector_size = host->pmecc_sector_size;
+
+ pmerrloc_writel(host->pmerrloc_base, ELDIS, PMERRLOC_DISABLE);
+
+ for (i = 0; i <= host->pmecc_data->lmu[cap + 1] >> 1; i++) {
+ pmerrloc_writel_sigma_relaxed(host->pmerrloc_base, i,
+ host->pmecc_data->smu[cap + 1][i]);
+ err_nbr++;
+ }
+
+ val = (err_nbr - 1) << 16;
+ if (sector_size == 1024)
+ val |= 1;
+
+ pmerrloc_writel(host->pmerrloc_base, ELCFG, val);
+ pmerrloc_writel(host->pmerrloc_base, ELEN,
+ sector_size * 8 + host->pmecc_degree * cap);
+
+ while (!(pmerrloc_readl_relaxed(host->pmerrloc_base, ELISR)
+ & PMERRLOC_CALC_DONE)) {
+ if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT))
+ return -1; /* Time out */
+ cpu_relax();
+ }
+
+ roots_nbr = (pmerrloc_readl_relaxed(host->pmerrloc_base, ELISR)
+ & PMERRLOC_ERR_NUM_MASK) >> 8;
+ /* Number of roots == degree of smu hence <= cap */
+ if (roots_nbr == host->pmecc_data->lmu[cap + 1] >> 1)
+ return err_nbr - 1;
+
+ /* Number of roots does not match the degree of smu
+ * unable to correct error */
+ return -1;
+}
+
+static void pmecc_correct_data(struct mtd_info *mtd, uint8_t *buf,
+ int extra_bytes, int err_nbr)
+{
+ int i = 0;
+ int byte_pos, bit_pos;
+ int sector_size, ecc_size;
+ uint32_t tmp;
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+
+ sector_size = host->pmecc_sector_size;
+ ecc_size = nand_chip->ecc.bytes;
+
+ while (err_nbr) {
+ tmp = pmerrloc_readl_el_relaxed(host->pmerrloc_base, i) - 1;
+ byte_pos = tmp / 8;
+ bit_pos = tmp % 8;
+ dev_info(host->dev, "PMECC correction, byte_pos: %d bit_pos: %d\n",
+ byte_pos, bit_pos);
+
+ if (byte_pos < (sector_size + extra_bytes)) {
+ tmp = sector_size +
+ pmecc_readl_relaxed(host->ecc, SADDR);
+
+ if (byte_pos < tmp)
+ *(buf + byte_pos) ^= (1 << bit_pos);
+ else
+ *(buf + byte_pos + ecc_size) ^= (1 << bit_pos);
+ }
+
+ i++;
+ err_nbr--;
+ }
+
+ return;
+}
+
+static int pmecc_correction(struct mtd_info *mtd, u32 pmecc_stat, uint8_t *buf,
+ u8 *ecc)
+{
+ int i, err_nbr;
+ uint8_t *buf_pos;
+ int eccbytes;
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+
+ eccbytes = nand_chip->ecc.bytes;
+ for (i = 0; i < eccbytes; i++)
+ if (ecc[i] != 0xff)
+ goto normal_check;
+ /* Erased page, return OK */
+ return 0;
+
+normal_check:
+ for (i = 0; i < host->pmecc_sector_number; i++) {
+ err_nbr = 0;
+ if (pmecc_stat & 0x1) {
+ buf_pos = buf + i * host->pmecc_sector_size;
+
+ pmecc_gen_syndrome(mtd, i);
+ pmecc_substitute(mtd);
+ pmecc_get_sigma(mtd);
+
+ err_nbr = pmecc_err_location(mtd);
+ if (err_nbr == -1) {
+ dev_err(host->dev, "PMECC: Too many errors\n");
+ mtd->ecc_stats.failed++;
+ return -EIO;
+ } else {
+ pmecc_correct_data(mtd, buf_pos, 0, err_nbr);
+ mtd->ecc_stats.corrected += err_nbr;
+ }
+ }
+ pmecc_stat >>= 1;
+ }
+
+ return 0;
+}
+
+static int atmel_nand_pmecc_read_page(struct mtd_info *mtd,
+ struct nand_chip *chip, uint8_t *buf, int page)
+{
+ uint32_t stat;
+ int timeout_count = 0;
+ int eccsize = chip->ecc.size;
+ uint8_t *oob = chip->oob_poi;
+ struct atmel_nand_host *host = chip->priv;
+ uint32_t *eccpos = chip->ecc.layout->eccpos;
+
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
+ pmecc_writel(host->ecc, CFG, (pmecc_readl_relaxed(host->ecc, CFG)
+ & ~PMECC_CFG_WRITE_OP) | PMECC_CFG_AUTO_ENABLE);
+
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DATA);
+
+ chip->read_buf(mtd, buf, eccsize);
+ chip->read_buf(mtd, oob, mtd->oobsize);
+
+ while ((pmecc_readl_relaxed(host->ecc, SR) & PMECC_SR_BUSY)) {
+ if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT))
+ return -EIO; /* Time out */
+ cpu_relax();
+ }
+
+ stat = pmecc_readl_relaxed(host->ecc, ISR);
+ if (stat != 0) {
+ if (pmecc_correction(mtd, stat, buf, &oob[eccpos[0]]) != 0)
+ return -EIO;
+ }
+
+ return 0;
+}
+
+static void atmel_nand_pmecc_write_page(struct mtd_info *mtd,
+ struct nand_chip *chip, const uint8_t *buf)
+{
+ int i, j;
+ int timeout_count = 0;
+ struct atmel_nand_host *host = chip->priv;
+ uint32_t *eccpos = chip->ecc.layout->eccpos;
+
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
+
+ pmecc_writel(host->ecc, CFG, (pmecc_readl_relaxed(host->ecc, CFG) |
+ PMECC_CFG_WRITE_OP) & ~PMECC_CFG_AUTO_ENABLE);
+
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DATA);
+
+ chip->write_buf(mtd, (u8 *)buf, mtd->writesize);
+
+ while ((pmecc_readl_relaxed(host->ecc, SR) & PMECC_SR_BUSY)) {
+ if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT)) {
+ dev_err(host->dev, "PMECC: Timeout to get ECC value.\n");
+ return; /* Time out */
+ }
+ cpu_relax();
+ }
+
+ for (i = 0; i < host->pmecc_sector_number; i++) {
+ for (j = 0; j < host->pmecc_bytes_per_sector; j++) {
+ int pos;
+
+ pos = i * host->pmecc_bytes_per_sector + j;
+ chip->oob_poi[eccpos[pos]] =
+ pmecc_readb_ecc_relaxed(host->ecc, i, j);
+ }
+ }
+ chip->write_buf(mtd, chip->oob_poi, mtd->oobsize);
+
+ return;
+}
+
+static void atmel_pmecc_core_init(struct mtd_info *mtd)
+{
+ uint32_t val = 0;
+ struct nand_chip *nand_chip = mtd->priv;
+ struct atmel_nand_host *host = nand_chip->priv;
+ struct nand_ecclayout *ecc_layout;
+
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
+
+ switch (host->pmecc_corr_cap) {
+ case 2:
+ val = PMECC_CFG_BCH_ERR2;
+ break;
+ case 4:
+ val = PMECC_CFG_BCH_ERR4;
+ break;
+ case 8:
+ val = PMECC_CFG_BCH_ERR8;
+ break;
+ case 12:
+ val = PMECC_CFG_BCH_ERR12;
+ break;
+ case 24:
+ val = PMECC_CFG_BCH_ERR24;
+ break;
+ }
+
+ if (host->pmecc_sector_size == 512)
+ val |= PMECC_CFG_SECTOR512;
+ else if (host->pmecc_sector_size == 1024)
+ val |= PMECC_CFG_SECTOR1024;
+
+ switch (host->pmecc_sector_number) {
+ case 1:
+ val |= PMECC_CFG_PAGE_1SECTOR;
+ break;
+ case 2:
+ val |= PMECC_CFG_PAGE_2SECTORS;
+ break;
+ case 4:
+ val |= PMECC_CFG_PAGE_4SECTORS;
+ break;
+ case 8:
+ val |= PMECC_CFG_PAGE_8SECTORS;
+ break;
+ }
+
+ val |= (PMECC_CFG_READ_OP | PMECC_CFG_SPARE_DISABLE
+ | PMECC_CFG_AUTO_DISABLE);
+ pmecc_writel(host->ecc, CFG, val);
+
+ ecc_layout = nand_chip->ecc.layout;
+ pmecc_writel(host->ecc, SAREA, mtd->oobsize - 1);
+ pmecc_writel(host->ecc, SADDR, ecc_layout->eccpos[0]);
+ pmecc_writel(host->ecc, EADDR,
+ ecc_layout->eccpos[ecc_layout->eccbytes - 1]);
+ /* See datasheet about PMECC Clock Control Register */
+ pmecc_writel(host->ecc, CLK, 2);
+ pmecc_writel(host->ecc, IDR, 0xff);
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
+}
+
+static int __init atmel_pmecc_nand_init_params(struct platform_device *pdev,
+ struct atmel_nand_host *host)
+{
+ int cap, sector_size, err_no;
+ struct mtd_info *mtd;
+ struct nand_chip *nand_chip;
+ struct resource *regs;
+ struct resource *regs_pmerr, *regs_rom;
+
+ cap = host->pmecc_corr_cap;
+ sector_size = host->pmecc_sector_size;
+ dev_info(host->dev, "Initialize PMECC params, cap: %d, sector: %d\n",
+ cap, sector_size);
+
+ /* Sanity check */
+ if ((sector_size != 512) && (sector_size != 1024)) {
+ dev_err(host->dev,
+ "Unsupported PMECC sector size: %d; should be 512 or 1024 bytes\n",
+ sector_size);
+ return -EINVAL;
+ }
+ if ((cap != 2) && (cap != 4) && (cap != 8) && (cap != 12) &&
+ (cap != 24)) {
+ dev_err(host->dev,
+ "Unsupported PMECC correction capability, should be 2, 4, 8, 12 or 24\n");
+ return -EINVAL;
+ }
+
+ nand_chip = &host->nand_chip;
+ mtd = &host->mtd;
+
+ nand_chip->ecc.mode = NAND_ECC_SOFT; /* By default */
+
+ regs = platform_get_resource(pdev, IORESOURCE_MEM, 1);
+ if (!regs) {
+ dev_warn(host->dev,
+ "Can't get I/O resource regs, rolling back on software ECC\n");
+ return 0;
+ }
+
+ host->ecc = ioremap(regs->start, resource_size(regs));
+ if (host->ecc == NULL) {
+ dev_err(host->dev, "ioremap failed\n");
+ err_no = -EIO;
+ goto err_pmecc_ioremap;
+ }
+
+ regs_pmerr = platform_get_resource(pdev, IORESOURCE_MEM, 2);
+ regs_rom = platform_get_resource(pdev, IORESOURCE_MEM, 3);
+ if (regs_pmerr && regs_rom) {
+ host->pmerrloc_base = ioremap(regs_pmerr->start,
+ resource_size(regs_pmerr));
+ host->pmecc_rom_base = ioremap(regs_rom->start,
+ resource_size(regs_rom));
+
+ if (host->pmerrloc_base && host->pmecc_rom_base) {
+ nand_chip->ecc.mode = NAND_ECC_HW;
+ nand_chip->ecc.read_page =
+ atmel_nand_pmecc_read_page;
+ nand_chip->ecc.write_page =
+ atmel_nand_pmecc_write_page;
+ } else {
+ dev_err(host->dev,
+ "Can not get I/O resource for PMECC controller!\n");
+ err_no = -EIO;
+ goto err_pmloc_ioremap;
+ }
+ }
+
+ /* ECC is calculated for the whole page (1 step) */
+ nand_chip->ecc.size = mtd->writesize;
+
+ /* set ECC page size and oob layout */
+ switch (mtd->writesize) {
+ case 2048:
+ host->pmecc_degree = PMECC_GF_DIMENSION_13;
+ host->pmecc_cw_len = (1 << host->pmecc_degree) - 1;
+ host->pmecc_corr_cap = cap;
+ host->pmecc_sector_number = mtd->writesize / sector_size;
+ host->pmecc_bytes_per_sector = pmecc_get_ecc_bytes(
+ cap, sector_size);
+ host->pmecc_alpha_to = pmecc_get_alpha_to(host);
+ host->pmecc_index_of = pmecc_get_index_of(host);
+
+ nand_chip->ecc.steps = 1;
+ nand_chip->ecc.strength = cap;
+ nand_chip->ecc.bytes = host->pmecc_bytes_per_sector *
+ host->pmecc_sector_number;
+ if (nand_chip->ecc.bytes > mtd->oobsize - 2) {
+ dev_err(host->dev, "No room for ECC bytes\n");
+ err_no = -EINVAL;
+ goto err;
+ }
+ pmecc_config_ecc_layout(&atmel_pmecc_oobinfo,
+ mtd->oobsize,
+ nand_chip->ecc.bytes);
+ nand_chip->ecc.layout = &atmel_pmecc_oobinfo;
+ break;
+ case 512:
+ case 1024:
+ case 4096:
+ /* TODO */
+ dev_warn(host->dev,
+ "Unsupported page size for PMECC, use Software ECC\n");
+ default:
+ /* page size not handled by HW ECC */
+ /* switching back to soft ECC */
+ nand_chip->ecc.mode = NAND_ECC_SOFT;
+ nand_chip->ecc.calculate = NULL;
+ nand_chip->ecc.correct = NULL;
+ nand_chip->ecc.hwctl = NULL;
+ nand_chip->ecc.read_page = NULL;
+ nand_chip->ecc.write_page = NULL;
+ nand_chip->ecc.postpad = 0;
+ nand_chip->ecc.prepad = 0;
+ nand_chip->ecc.bytes = 0;
+ err_no = 0;
+ goto err;
+ }
+
+ /* Allocate data for PMECC computation */
+ host->pmecc_data = kzalloc(sizeof(struct atmel_pmecc_data), GFP_KERNEL);
+ if (!host->pmecc_data) {
+ dev_err(host->dev,
+ "Cannot allocate memory for PMECC computation!\n");
+ err_no = -ENOMEM;
+ goto err;
+ }
+
+ atmel_pmecc_core_init(mtd);
+
+ return 0;
+
+err:
+err_pmloc_ioremap:
+ iounmap(host->ecc);
+ if (host->pmerrloc_base)
+ iounmap(host->pmerrloc_base);
+ if (host->pmecc_rom_base)
+ iounmap(host->pmecc_rom_base);
+err_pmecc_ioremap:
+ return err_no;
+}
+
+/*
* Calculate HW ECC
*
* function called after a write
@@ -720,7 +1456,11 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
}
if (nand_chip->ecc.mode == NAND_ECC_HW) {
- res = atmel_hw_nand_init_params(pdev, host);
+ if (host->has_pmecc)
+ res = atmel_pmecc_nand_init_params(pdev, host);
+ else
+ res = atmel_hw_nand_init_params(pdev, host);
+
if (res != 0)
goto err_hw_ecc;
}
@@ -741,6 +1481,12 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
err_scan_tail:
if (host->ecc)
iounmap(host->ecc);
+ if (host->has_pmecc) {
+ if (host->pmerrloc_base)
+ iounmap(host->pmerrloc_base);
+ if (host->pmecc_rom_base)
+ iounmap(host->pmecc_rom_base);
+ }
err_hw_ecc:
err_scan_ident:
err_no_card:
@@ -766,6 +1512,19 @@ static int __exit atmel_nand_remove(struct platform_device *pdev)
atmel_nand_disable(host);
+ if (host->has_pmecc) {
+ pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
+ if (host->pmerrloc_base) {
+ pmerrloc_writel(host->pmerrloc_base, ELDIS,
+ PMERRLOC_DISABLE);
+ iounmap(host->pmerrloc_base);
+ }
+ if (host->pmecc_rom_base)
+ iounmap(host->pmecc_rom_base);
+
+ kfree(host->pmecc_data);
+ }
+
if (host->ecc)
iounmap(host->ecc);
diff --git a/drivers/mtd/nand/atmel_nand_ecc.h b/drivers/mtd/nand/atmel_nand_ecc.h
index 5cbff11..bd404bf 100644
--- a/drivers/mtd/nand/atmel_nand_ecc.h
+++ b/drivers/mtd/nand/atmel_nand_ecc.h
@@ -45,4 +45,120 @@
#define ecc_writel(add, reg, value) \
writel((value), add + ATMEL_ECC_##reg)
+/* PMECC Register Definitions */
+#define ATMEL_PMECC_CFG 0x000 /* Configuration Register */
+#define PMECC_CFG_BCH_ERR2 (0 << 0)
+#define PMECC_CFG_BCH_ERR4 (1 << 0)
+#define PMECC_CFG_BCH_ERR8 (2 << 0)
+#define PMECC_CFG_BCH_ERR12 (3 << 0)
+#define PMECC_CFG_BCH_ERR24 (4 << 0)
+
+#define PMECC_CFG_SECTOR512 (0 << 4)
+#define PMECC_CFG_SECTOR1024 (1 << 4)
+
+#define PMECC_CFG_PAGE_1SECTOR (0 << 8)
+#define PMECC_CFG_PAGE_2SECTORS (1 << 8)
+#define PMECC_CFG_PAGE_4SECTORS (2 << 8)
+#define PMECC_CFG_PAGE_8SECTORS (3 << 8)
+
+#define PMECC_CFG_READ_OP (0 << 12)
+#define PMECC_CFG_WRITE_OP (1 << 12)
+
+#define PMECC_CFG_SPARE_ENABLE (1 << 16)
+#define PMECC_CFG_SPARE_DISABLE (0 << 16)
+
+#define PMECC_CFG_AUTO_ENABLE (1 << 20)
+#define PMECC_CFG_AUTO_DISABLE (0 << 20)
+
+#define ATMEL_PMECC_SAREA 0x004 /* Spare area size */
+#define ATMEL_PMECC_SADDR 0x008 /* PMECC starting address */
+#define ATMEL_PMECC_EADDR 0x00c /* PMECC ending address */
+#define ATMEL_PMECC_CLK 0x010 /* PMECC clock control */
+#define PMECC_CLK_133MHZ (2 << 0)
+
+#define ATMEL_PMECC_CTRL 0x014 /* PMECC control register */
+#define PMECC_CTRL_RST (1 << 0)
+#define PMECC_CTRL_DATA (1 << 1)
+#define PMECC_CTRL_USER (1 << 2)
+#define PMECC_CTRL_ENABLE (1 << 4)
+#define PMECC_CTRL_DISABLE (1 << 5)
+
+#define ATMEL_PMECC_SR 0x018 /* PMECC status register */
+#define PMECC_SR_BUSY (1 << 0)
+#define PMECC_SR_ENABLE (1 << 4)
+
+#define ATMEL_PMECC_IER 0x01c /* PMECC interrupt enable */
+#define PMECC_IER_ENABLE (1 << 0)
+#define ATMEL_PMECC_IDR 0x020 /* PMECC interrupt disable */
+#define PMECC_IER_DISABLE (1 << 0)
+#define ATMEL_PMECC_IMR 0x024 /* PMECC interrupt mask */
+#define PMECC_IER_MASK (1 << 0)
+#define ATMEL_PMECC_ISR 0x028 /* PMECC interrupt status */
+#define ATMEL_PMECC_ECCx 0x040 /* PMECC ECC x */
+#define ATMEL_PMECC_REMx 0x240 /* PMECC REM x */
+
+/* PMERRLOC Register Definitions */
+#define ATMEL_PMERRLOC_ELCFG 0x000 /* Error location config */
+#define PMERRLOC_ELCFG_SECTOR_512 (0 << 0)
+#define PMERRLOC_ELCFG_SECTOR_1024 (1 << 0)
+#define PMERRLOC_ELCFG_NUM_ERRORS(n) ((n) << 16)
+
+#define ATMEL_PMERRLOC_ELPRIM 0x004 /* Error location primitive */
+#define ATMEL_PMERRLOC_ELEN 0x008 /* Error location enable */
+#define ATMEL_PMERRLOC_ELDIS 0x00c /* Error location disable */
+#define PMERRLOC_DISABLE (1 << 0)
+
+#define ATMEL_PMERRLOC_ELSR 0x010 /* Error location status */
+#define PMERRLOC_ELSR_BUSY (1 << 0)
+#define ATMEL_PMERRLOC_ELIER 0x014 /* Error location int enable */
+#define ATMEL_PMERRLOC_ELIDR 0x018 /* Error location int disable */
+#define ATMEL_PMERRLOC_ELIMR 0x01c /* Error location int mask */
+#define ATMEL_PMERRLOC_ELISR 0x020 /* Error location int status */
+#define PMERRLOC_ERR_NUM_MASK (0x1f << 8)
+#define PMERRLOC_CALC_DONE (1 << 0)
+#define ATMEL_PMERRLOC_SIGMAx 0x028 /* Error location SIGMA x */
+#define ATMEL_PMERRLOC_ELx 0x08c /* Error location x */
+
+/* Register access macros for PMECC */
+#define pmecc_readl_relaxed(addr, reg) \
+ readl_relaxed((addr) + ATMEL_PMECC_##reg)
+
+#define pmecc_writel(addr, reg, value) \
+ writel((value), (addr) + ATMEL_PMECC_##reg)
+
+#define pmecc_readb_ecc_relaxed(addr, sector, n) \
+ readb_relaxed((addr) + ATMEL_PMECC_ECCx + ((sector) * 0x40) + (n))
+
+#define pmecc_readl_rem_relaxed(addr, sector, n) \
+ readl_relaxed((addr) + ATMEL_PMECC_REMx + ((sector) * 0x40) + ((n) * 4))
+
+#define pmerrloc_readl_relaxed(addr, reg) \
+ readl_relaxed((addr) + ATMEL_PMERRLOC_##reg)
+
+#define pmerrloc_writel(addr, reg, value) \
+ writel((value), (addr) + ATMEL_PMERRLOC_##reg)
+
+#define pmerrloc_writel_sigma_relaxed(addr, n, value) \
+ writel_relaxed((value), (addr) + ATMEL_PMERRLOC_SIGMAx + ((n) * 4))
+
+#define pmerrloc_readl_sigma_relaxed(addr, n) \
+ readl_relaxed((addr) + ATMEL_PMERRLOC_SIGMAx + ((n) * 4))
+
+#define pmerrloc_readl_el_relaxed(addr, n) \
+ readl_relaxed((addr) + ATMEL_PMERRLOC_ELx + ((n) * 4))
+
+/* Galois field dimension */
+#define PMECC_GF_DIMENSION_13 13
+#define PMECC_GF_DIMENSION_14 14
+
+#define PMECC_MAX_ERROR_NB 25
+#define PMECC_MAX_ECC_BYTES 42
+#define PMECC_MAX_NB_SECTOR 8
+#define PMECC_MAX_TIMEOUT_COUNT 100
+
+#define PMECC_LOOKUP_TABLE_OFFSET_512 0x8000
+#define PMECC_LOOKUP_TABLE_SIZE_512 0x2000
+#define PMECC_LOOKUP_TABLE_OFFSET_1024 0x10000
+#define PMECC_LOOKUP_TABLE_SIZE_1024 0x4000
+
#endif
--
1.7.10
^ permalink raw reply related [flat|nested] 24+ messages in thread
* Re: [PATCH v9 1/3] MTD: at91: extract hw ecc initialization to one function and use relaxed read/write
2012-05-26 13:24 ` Josh Wu
@ 2012-05-27 12:44 ` Artem Bityutskiy
-1 siblings, 0 replies; 24+ messages in thread
From: Artem Bityutskiy @ 2012-05-27 12:44 UTC (permalink / raw)
To: Josh Wu
Cc: hongxu.cn, nicolas.ferre, linux-mtd, ivan.djelic, plagnioj,
linux-arm-kernel
[-- Attachment #1: Type: text/plain, Size: 762 bytes --]
On Sat, 2012-05-26 at 21:24 +0800, Josh Wu wrote:
> use _relaxed read/write in most place. And use writel in operations of Control Register since it needs memory barrier.
>
> Signed-off-by: Hong Xu <hong.xu@atmel.com>
> Signed-off-by: Josh Wu <josh.wu@atmel.com>
This should be split on 2 or even 3 parts:
1. You move definitions and things around.
2. You separate out the init stuff into a function
3. You start using _relaxed helpers.
And you should provide better commit messages. You should explain why
you do things - just cleanup? improvement? fix? preparation for
something?
I think this patch is too big, commit message is poor, patch does more
things than documented in the commit message.
--
Best Regards,
Artem Bityutskiy
[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 836 bytes --]
^ permalink raw reply [flat|nested] 24+ messages in thread
* [PATCH v9 1/3] MTD: at91: extract hw ecc initialization to one function and use relaxed read/write
@ 2012-05-27 12:44 ` Artem Bityutskiy
0 siblings, 0 replies; 24+ messages in thread
From: Artem Bityutskiy @ 2012-05-27 12:44 UTC (permalink / raw)
To: linux-arm-kernel
On Sat, 2012-05-26 at 21:24 +0800, Josh Wu wrote:
> use _relaxed read/write in most place. And use writel in operations of Control Register since it needs memory barrier.
>
> Signed-off-by: Hong Xu <hong.xu@atmel.com>
> Signed-off-by: Josh Wu <josh.wu@atmel.com>
This should be split on 2 or even 3 parts:
1. You move definitions and things around.
2. You separate out the init stuff into a function
3. You start using _relaxed helpers.
And you should provide better commit messages. You should explain why
you do things - just cleanup? improvement? fix? preparation for
something?
I think this patch is too big, commit message is poor, patch does more
things than documented in the commit message.
--
Best Regards,
Artem Bityutskiy
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20120527/5b38f766/attachment.sig>
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v9 3/3] MTD: at91: atmel_nand: Update driver to support Programmable Multibit ECC controller
2012-05-26 13:24 ` Josh Wu
@ 2012-05-27 12:50 ` Artem Bityutskiy
-1 siblings, 0 replies; 24+ messages in thread
From: Artem Bityutskiy @ 2012-05-27 12:50 UTC (permalink / raw)
To: Josh Wu
Cc: hongxu.cn, nicolas.ferre, linux-mtd, ivan.djelic, plagnioj,
linux-arm-kernel
[-- Attachment #1: Type: text/plain, Size: 892 bytes --]
On Sat, 2012-05-26 at 21:24 +0800, Josh Wu wrote:
> + while ((pmecc_readl_relaxed(host->ecc, SR) & PMECC_SR_BUSY)) {
> + if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT)) {
> + dev_err(host->dev, "PMECC: Timeout to get ECC value.\n");
> + return; /* Time out */
How this error is communicated then up the the user?
> + }
> + cpu_relax();
> + }
I see this pattern all over the place - why people consider it reliable?
Is this code guaranteed to run on the same CPU?
Why not to use loops_per_jiffie * msecs_to_jiffies(TIMOUT) instead to
calculate how many iterations to do? Yes, due to HW register reading and
cpu_relax() the real timeout will be larger, but this is about error
anyway, so it does not hurt to iterate longer?
--
Best Regards,
Artem Bityutskiy
[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 836 bytes --]
^ permalink raw reply [flat|nested] 24+ messages in thread
* [PATCH v9 3/3] MTD: at91: atmel_nand: Update driver to support Programmable Multibit ECC controller
@ 2012-05-27 12:50 ` Artem Bityutskiy
0 siblings, 0 replies; 24+ messages in thread
From: Artem Bityutskiy @ 2012-05-27 12:50 UTC (permalink / raw)
To: linux-arm-kernel
On Sat, 2012-05-26 at 21:24 +0800, Josh Wu wrote:
> + while ((pmecc_readl_relaxed(host->ecc, SR) & PMECC_SR_BUSY)) {
> + if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT)) {
> + dev_err(host->dev, "PMECC: Timeout to get ECC value.\n");
> + return; /* Time out */
How this error is communicated then up the the user?
> + }
> + cpu_relax();
> + }
I see this pattern all over the place - why people consider it reliable?
Is this code guaranteed to run on the same CPU?
Why not to use loops_per_jiffie * msecs_to_jiffies(TIMOUT) instead to
calculate how many iterations to do? Yes, due to HW register reading and
cpu_relax() the real timeout will be larger, but this is about error
anyway, so it does not hurt to iterate longer?
--
Best Regards,
Artem Bityutskiy
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20120527/f05d8118/attachment.sig>
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v9 0/3] MTD: at91: Add PMECC support for at91 nand flash driver
2012-05-26 13:24 ` Josh Wu
@ 2012-05-28 6:39 ` Jean-Christophe PLAGNIOL-VILLARD
-1 siblings, 0 replies; 24+ messages in thread
From: Jean-Christophe PLAGNIOL-VILLARD @ 2012-05-28 6:39 UTC (permalink / raw)
To: Josh Wu
Cc: hongxu.cn, dedekind1, nicolas.ferre, linux-mtd, ivan.djelic,
linux-arm-kernel
On 21:24 Sat 26 May , Josh Wu wrote:
> Code is based on 3.4-rc2
>
> Changes since v8,
> use _relaxed read/write in most place. use writel in operations of Control Register since it needs memory barrier.
> allocate the data for PMECC computation.
> add pmecc prefix for related variable/functions.
> modify code according to J.C's suggestion. except:
> >> + for (i = 2; i <= 2 * host->cap; i += 2) {
> > manage the j in the for loop
> since that will change to:
> + for (i = 2, j = 1; i <= 2 * host->cap; i += 2, j = i / 2) {
why you insist on doing j = i / 2
you can simplify it by j++
or loop on j and do i = j << 1;
it's more simple and faster
Best Regards,
J.
^ permalink raw reply [flat|nested] 24+ messages in thread
* [PATCH v9 0/3] MTD: at91: Add PMECC support for at91 nand flash driver
@ 2012-05-28 6:39 ` Jean-Christophe PLAGNIOL-VILLARD
0 siblings, 0 replies; 24+ messages in thread
From: Jean-Christophe PLAGNIOL-VILLARD @ 2012-05-28 6:39 UTC (permalink / raw)
To: linux-arm-kernel
On 21:24 Sat 26 May , Josh Wu wrote:
> Code is based on 3.4-rc2
>
> Changes since v8,
> use _relaxed read/write in most place. use writel in operations of Control Register since it needs memory barrier.
> allocate the data for PMECC computation.
> add pmecc prefix for related variable/functions.
> modify code according to J.C's suggestion. except:
> >> + for (i = 2; i <= 2 * host->cap; i += 2) {
> > manage the j in the for loop
> since that will change to:
> + for (i = 2, j = 1; i <= 2 * host->cap; i += 2, j = i / 2) {
why you insist on doing j = i / 2
you can simplify it by j++
or loop on j and do i = j << 1;
it's more simple and faster
Best Regards,
J.
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v9 3/3] MTD: at91: atmel_nand: Update driver to support Programmable Multibit ECC controller
2012-05-26 13:24 ` Josh Wu
@ 2012-05-28 6:58 ` Jean-Christophe PLAGNIOL-VILLARD
-1 siblings, 0 replies; 24+ messages in thread
From: Jean-Christophe PLAGNIOL-VILLARD @ 2012-05-28 6:58 UTC (permalink / raw)
To: Josh Wu
Cc: hongxu.cn, dedekind1, nicolas.ferre, linux-mtd, ivan.djelic,
linux-arm-kernel
On 21:24 Sat 26 May , Josh Wu wrote:
> The Programmable Multibit ECC (PMECC) controller is a programmable binary
> BCH(Bose, Chaudhuri and Hocquenghem) encoder and decoder. This controller
> can be used to support both SLC and MLC NAND Flash devices. It supports to
> generate ECC to correct 2, 4, 8, 12 or 24 bits of error per sector of data.
>
> To use this driver, the user needs to pass in the correction capability and
> the sector size.
>
> This driver has been tested on AT91SAM9X5-EK and AT91SAM9N12-EK with JFFS2,
> YAFFS2, UBIFS and mtd-utils.
>
> Signed-off-by: Hong Xu <hong.xu@atmel.com>
> Signed-off-by: Josh Wu <josh.wu@atmel.com>
> ---
> drivers/mtd/nand/atmel_nand.c | 761 ++++++++++++++++++++++++++++++++++++-
> drivers/mtd/nand/atmel_nand_ecc.h | 116 ++++++
> 2 files changed, 876 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/mtd/nand/atmel_nand.c b/drivers/mtd/nand/atmel_nand.c
> index 9a9bfbf..ddcf1ed 100644
> --- a/drivers/mtd/nand/atmel_nand.c
> +++ b/drivers/mtd/nand/atmel_nand.c
> @@ -15,6 +15,8 @@
> * (u-boot-1.1.5/board/atmel/at91sam9263ek/nand.c)
> * (C) Copyright 2006 ATMEL Rousset, Lacressonniere Nicolas
> *
> + * Add Programmable Multibit ECC support for various AT91 SoC
> + * (C) Copyright 2012 ATMEL, Hong Xu
> *
> * This program is free software; you can redistribute it and/or modify
> * it under the terms of the GNU General Public License version 2 as
> @@ -77,6 +79,21 @@ static struct nand_ecclayout atmel_oobinfo_small = {
> },
> };
>
> +/* a structure includes datas for PMECC computation */
> +struct atmel_pmecc_data {
> + int16_t partial_syn[2 * PMECC_MAX_ERROR_NB + 1];
> + int16_t si[2 * PMECC_MAX_ERROR_NB + 1];
> +
> + /* Sigma table */
> + int16_t smu[PMECC_MAX_ERROR_NB + 2][2 * PMECC_MAX_ERROR_NB + 1];
you still hardcode the array in the struct
and if the pmecc evolve we will have to touch again
please allocate them
> + /* polynomal order */
> + int16_t lmu[PMECC_MAX_ERROR_NB + 1];
> +
> + int mu[PMECC_MAX_ERROR_NB + 1];
> + int dmu[PMECC_MAX_ERROR_NB + 1];
> + int delta[PMECC_MAX_ERROR_NB + 1];
> +};
> +
> struct atmel_nand_host {
> struct nand_chip nand_chip;
> struct mtd_info mtd;
> @@ -92,8 +109,25 @@ struct atmel_nand_host {
> bool has_pmecc;
> u8 pmecc_corr_cap;
> u16 pmecc_sector_size;
> +
> + int pmecc_bytes_per_sector;
> + int pmecc_sector_number;
> + int pmecc_degree; /* Degree of remainders */
> + int pmecc_cw_len; /* Length of codeword */
> +
> + void __iomem *pmerrloc_base;
> + void __iomem *pmecc_rom_base;
> +
> + /* lookup table for alpha_to and index_of */
> + void __iomem *pmecc_alpha_to;
> + void __iomem *pmecc_index_of;
> +
> + /* data for pmecc computation */
> + struct atmel_pmecc_data *pmecc_data;
> };
>
> +static struct nand_ecclayout atmel_pmecc_oobinfo;
> +
> static int cpu_has_dma(void)
> {
> return cpu_is_at91sam9rl() || cpu_is_at91sam9g45();
> @@ -287,6 +321,708 @@ static void atmel_write_buf(struct mtd_info *mtd, const u8 *buf, int len)
> }
>
> /*
> + * Return number of ecc bytes per sector according to sector size and
> + * correction capability
> + *
> + * Following table shows what at91 PMECC supported:
> + * Correction Capability Sector_512_bytes Sector_1024_bytes
> + * ===================== ================ =================
> + * 2-bits 4-bytes 4-bytes
> + * 4-bits 7-bytes 7-bytes
> + * 8-bits 13-bytes 14-bytes
> + * 12-bits 20-bytes 21-bytes
> + * 24-bits 39-bytes 42-bytes
> + */
> +static int pmecc_get_ecc_bytes(int cap, int sector_size)
> +{
> + int m = 12 + sector_size / 512;
> + return (m * cap + 7) / 8;
> +}
> +
> +static void pmecc_config_ecc_layout(struct nand_ecclayout *layout, int oobsize,
> + int ecc_len)
> +{
> + int i;
> +
> + layout->eccbytes = ecc_len;
> +
> + /* ECC will occupy the last ecc_len bytes continuously */
> + for (i = 0; i < ecc_len; i++)
> + layout->eccpos[i] = oobsize - ecc_len + i;
> +
> + layout->oobfree[0].offset = 2;
> + layout->oobfree[0].length =
> + oobsize - ecc_len - layout->oobfree[0].offset;
> +}
> +
> +static void __iomem *pmecc_get_alpha_to(struct atmel_nand_host *host)
> +{
> + void __iomem *p;
> +
> + switch (host->pmecc_sector_size) {
> + case 512:
> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_512 +
> + PMECC_LOOKUP_TABLE_SIZE_512 * sizeof(int16_t);
> + break;
> + case 1024:
> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_1024 +
> + PMECC_LOOKUP_TABLE_SIZE_1024 * sizeof(int16_t);
> + break;
> + default:
> + BUG();
> + }
> +
> + return p;
> +}
> +
> +static void __iomem *pmecc_get_index_of(struct atmel_nand_host *host)
this is a __dev_init function plese check the other too
btw you need to use __dev_init and not __init
> +{
> + void __iomem *p;
> +
> + switch (host->pmecc_sector_size) {
> + case 512:
> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_512;
> + break;
> + case 1024:
> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_1024;
> + break;
> + default:
> + BUG();
> + }
> +
> + return p;
> +}
> +
> +static void pmecc_gen_syndrome(struct mtd_info *mtd, int sector)
> +{
> + int i;
> + uint32_t value;
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> +
> + /* Fill odd syndromes */
> + for (i = 0; i < host->pmecc_corr_cap; i++) {
> + value = pmecc_readl_rem_relaxed(host->ecc, sector, i / 2);
> + value = (i & 1) ? (value & 0xffff0000) >> 16 : value & 0xffff;
simplify by
if (i & 1)
val >>= 16;
value &= 0xffff;
> + host->pmecc_data->partial_syn[(2 * i) + 1] = (int16_t)value;
> + }
> +}
> +
> +static void pmecc_substitute(struct mtd_info *mtd)
> +{
> + int16_t *si;
> + int i, j;
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> + int16_t __iomem *alpha_to = host->pmecc_alpha_to;
> + int16_t __iomem *index_of = host->pmecc_index_of;
> + int16_t *partial_syn = host->pmecc_data->partial_syn;
> +
> + /* si[] is a table that holds the current syndrome value,
> + * an element of that table belongs to the field
> + */
> + si = host->pmecc_data->si;
> +
> + for (i = 1; i < 2 * PMECC_MAX_ERROR_NB; i++)
> + si[i] = 0;
please use memset
> +
> + /* Computation 2t syndromes based on S(x) */
> + /* Odd syndromes */
> + for (i = 1; i < 2 * host->pmecc_corr_cap; i += 2) {
> + si[i] = 0;
shy this you already init the array at 0 before
> + for (j = 0; j < host->pmecc_degree; j++) {
> + if (partial_syn[i] & ((unsigned short)0x1 << j))
> + si[i] = readw_relaxed(alpha_to + i * j) ^ si[i];
> + }
> + }
> + /* Even syndrome = (Odd syndrome) ** 2 */
> + for (i = 2; i <= 2 * host->pmecc_corr_cap; i += 2) {
> + j = i / 2;
> + if (si[j] == 0)
here if {
} else {
}
> + si[i] = 0;
> + else {
> + int16_t tmp;
missing blank line
> + tmp = readw_relaxed(index_of + si[j]);
> + tmp = (tmp * 2) % host->pmecc_cw_len;
> + si[i] = readw_relaxed(alpha_to + tmp);
> + }
> + }
> +
> + return;
> +}
> +
> +static void pmecc_get_sigma(struct mtd_info *mtd)
> +{
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> +
> + int i, j, k;
> + uint32_t dmu_0_count, tmp;
> + int16_t (*smu)[2 * PMECC_MAX_ERROR_NB + 1];
> + int16_t *lmu = host->pmecc_data->lmu;
> + int16_t *si = host->pmecc_data->si;
> + int *mu = host->pmecc_data->mu;
> + int *dmu = host->pmecc_data->dmu; /* Discrepancy */
> + int *delta = host->pmecc_data->delta; /* Delta order */
> + int cw_len = host->pmecc_cw_len;
> + int16_t cap = host->pmecc_corr_cap;
> +
> + int16_t __iomem *index_of = host->pmecc_index_of;
> + int16_t __iomem *alpha_to = host->pmecc_alpha_to;
> +
> + /* index of largest delta */
> + int ro;
> + int largest;
> + int diff;
> +
> + dmu_0_count = 0;
> + smu = host->pmecc_data->smu;
> +
> + /* First Row */
> +
> + /* Mu */
> + mu[0] = -1;
> +
> + memset(&smu[0][0], 0,
> + sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
> + smu[0][0] = 1;
> +
> + /* discrepancy set to 1 */
> + dmu[0] = 1;
> + /* polynom order set to 0 */
> + lmu[0] = 0;
> + delta[0] = (mu[0] * 2 - lmu[0]) >> 1;
> +
> + /* Second Row */
> +
> + /* Mu */
> + mu[1] = 0;
> + /* Sigma(x) set to 1 */
> + memset(&smu[1][0], 0,
> + sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
> + smu[1][0] = 1;
> +
> + /* discrepancy set to S1 */
> + dmu[1] = si[1];
> +
> + /* polynom order set to 0 */
> + lmu[1] = 0;
> +
> + delta[1] = (mu[1] * 2 - lmu[1]) >> 1;
> +
> + /* Init the Sigma(x) last row */
> + memset(&smu[cap + 1][0], 0,
> + sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
> +
> + for (i = 1; i <= cap; i++) {
> + mu[i+1] = i << 1;
> + /* Begin Computing Sigma (Mu+1) and L(mu) */
> + /* check if discrepancy is set to 0 */
> + if (dmu[i] == 0) {
> + dmu_0_count++;
> +
> + tmp = ((cap - (lmu[i] >> 1) - 1) / 2);
> + if ((cap - (lmu[i] >> 1) - 1) & 0x1)
> + tmp += 2;
> + else
> + tmp += 1;
> +
> + if (dmu_0_count == tmp) {
> + for (j = 0; j <= (lmu[i] >> 1) + 1; j++)
> + smu[cap + 1][j] = smu[i][j];
> + lmu[cap + 1] = lmu[i];
> + return;
> + }
> +
> + /* copy polynom */
> + for (j = 0; j <= lmu[i] >> 1; j++)
> + smu[i + 1][j] = smu[i][j];
> +
> + /* copy previous polynom order to the next */
> + lmu[i + 1] = lmu[i];
> + } else {
> + ro = 0;
> + largest = -1;
> + /* find largest delta with dmu != 0 */
> + for (j = 0; j < i; j++) {
> + if ((dmu[j]) && (delta[j] > largest)) {
> + largest = delta[j];
> + ro = j;
> + }
> + }
> +
> + /* compute difference */
> + diff = (mu[i] - mu[ro]);
> +
> + /* Compute degree of the new smu polynomial */
> + if ((lmu[i] >> 1) > ((lmu[ro] >> 1) + diff))
> + lmu[i + 1] = lmu[i];
> + else
> + lmu[i + 1] = ((lmu[ro] >> 1) + diff) * 2;
> +
> + /* Init smu[i+1] with 0 */
> + for (k = 0; k < (2 * PMECC_MAX_ERROR_NB + 1); k++)
> + smu[i+1][k] = 0;
> +
> + /* Compute smu[i+1] */
> + for (k = 0; k <= lmu[ro] >> 1; k++) {
> + int16_t a, b, c;
> +
> + if (!(smu[ro][k] && dmu[i]))
> + continue;
> + a = readw_relaxed(index_of + dmu[i]);
> + b = readw_relaxed(index_of + dmu[ro]);
> + c = readw_relaxed(index_of + smu[ro][k]);
> + tmp = a + (cw_len - b) + c;
> + a = readw_relaxed(alpha_to + tmp % cw_len);
> + smu[i + 1][k + diff] = a;
> + }
> +
> + for (k = 0; k <= lmu[i] >> 1; k++)
> + smu[i + 1][k] ^= smu[i][k];
> + }
> +
> + /* End Computing Sigma (Mu+1) and L(mu) */
> + /* In either case compute delta */
> + delta[i + 1] = (mu[i + 1] * 2 - lmu[i + 1]) >> 1;
> +
> + /* Do not compute discrepancy for the last iteration */
> + if (i >= cap)
> + continue;
> +
> + for (k = 0 ; k <= (lmu[i + 1] >> 1); k++) {
> + tmp = 2 * (i - 1);
> + if (k == 0)
> + dmu[i + 1] = si[tmp + 3];
> + else if (smu[i+1][k] && si[tmp + 3 - k]) {
> + int16_t a, b, c;
> + a = readw_relaxed(index_of + smu[i + 1][k]);
> + b = si[2 * (i - 1) + 3 - k];
> + c = readw_relaxed(index_of + b);
> + tmp = a + c;
> + tmp %= cw_len;
> + dmu[i + 1] = readw_relaxed(alpha_to + tmp) ^
> + dmu[i + 1];
> + }
> + }
> + }
> +
> + return;
> +}
> +
> +static int pmecc_err_location(struct mtd_info *mtd)
> +{
> + int i;
> + int err_nbr; /* number of error */
> + int roots_nbr; /* number of roots */
> + int sector_size;
> + uint32_t val;
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> + int timeout_count = 0;
> + int cap = host->pmecc_corr_cap;
> +
> + err_nbr = 0;
> + sector_size = host->pmecc_sector_size;
> +
> + pmerrloc_writel(host->pmerrloc_base, ELDIS, PMERRLOC_DISABLE);
> +
> + for (i = 0; i <= host->pmecc_data->lmu[cap + 1] >> 1; i++) {
> + pmerrloc_writel_sigma_relaxed(host->pmerrloc_base, i,
> + host->pmecc_data->smu[cap + 1][i]);
> + err_nbr++;
> + }
> +
> + val = (err_nbr - 1) << 16;
> + if (sector_size == 1024)
> + val |= 1;
> +
> + pmerrloc_writel(host->pmerrloc_base, ELCFG, val);
> + pmerrloc_writel(host->pmerrloc_base, ELEN,
> + sector_size * 8 + host->pmecc_degree * cap);
> +
> + while (!(pmerrloc_readl_relaxed(host->pmerrloc_base, ELISR)
> + & PMERRLOC_CALC_DONE)) {
> + if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT))
> + return -1; /* Time out */
> + cpu_relax();
> + }
> +
> + roots_nbr = (pmerrloc_readl_relaxed(host->pmerrloc_base, ELISR)
> + & PMERRLOC_ERR_NUM_MASK) >> 8;
> + /* Number of roots == degree of smu hence <= cap */
> + if (roots_nbr == host->pmecc_data->lmu[cap + 1] >> 1)
> + return err_nbr - 1;
> +
> + /* Number of roots does not match the degree of smu
> + * unable to correct error */
> + return -1;
> +}
> +
> +static void pmecc_correct_data(struct mtd_info *mtd, uint8_t *buf,
> + int extra_bytes, int err_nbr)
> +{
> + int i = 0;
> + int byte_pos, bit_pos;
> + int sector_size, ecc_size;
> + uint32_t tmp;
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> +
> + sector_size = host->pmecc_sector_size;
> + ecc_size = nand_chip->ecc.bytes;
> +
> + while (err_nbr) {
> + tmp = pmerrloc_readl_el_relaxed(host->pmerrloc_base, i) - 1;
> + byte_pos = tmp / 8;
> + bit_pos = tmp % 8;
> + dev_info(host->dev, "PMECC correction, byte_pos: %d bit_pos: %d\n",
> + byte_pos, bit_pos);
> +
> + if (byte_pos < (sector_size + extra_bytes)) {
> + tmp = sector_size +
> + pmecc_readl_relaxed(host->ecc, SADDR);
> +
> + if (byte_pos < tmp)
> + *(buf + byte_pos) ^= (1 << bit_pos);
> + else
> + *(buf + byte_pos + ecc_size) ^= (1 << bit_pos);
> + }
> +
> + i++;
> + err_nbr--;
> + }
> +
> + return;
> +}
> +
> +static int pmecc_correction(struct mtd_info *mtd, u32 pmecc_stat, uint8_t *buf,
> + u8 *ecc)
> +{
> + int i, err_nbr;
> + uint8_t *buf_pos;
> + int eccbytes;
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> +
> + eccbytes = nand_chip->ecc.bytes;
> + for (i = 0; i < eccbytes; i++)
> + if (ecc[i] != 0xff)
> + goto normal_check;
> + /* Erased page, return OK */
> + return 0;
> +
> +normal_check:
> + for (i = 0; i < host->pmecc_sector_number; i++) {
> + err_nbr = 0;
> + if (pmecc_stat & 0x1) {
> + buf_pos = buf + i * host->pmecc_sector_size;
> +
> + pmecc_gen_syndrome(mtd, i);
> + pmecc_substitute(mtd);
> + pmecc_get_sigma(mtd);
> +
> + err_nbr = pmecc_err_location(mtd);
> + if (err_nbr == -1) {
> + dev_err(host->dev, "PMECC: Too many errors\n");
> + mtd->ecc_stats.failed++;
> + return -EIO;
> + } else {
> + pmecc_correct_data(mtd, buf_pos, 0, err_nbr);
> + mtd->ecc_stats.corrected += err_nbr;
> + }
> + }
> + pmecc_stat >>= 1;
> + }
> +
> + return 0;
> +}
> +
> +static int atmel_nand_pmecc_read_page(struct mtd_info *mtd,
> + struct nand_chip *chip, uint8_t *buf, int page)
> +{
> + uint32_t stat;
> + int timeout_count = 0;
> + int eccsize = chip->ecc.size;
> + uint8_t *oob = chip->oob_poi;
> + struct atmel_nand_host *host = chip->priv;
> + uint32_t *eccpos = chip->ecc.layout->eccpos;
> +
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
> + pmecc_writel(host->ecc, CFG, (pmecc_readl_relaxed(host->ecc, CFG)
> + & ~PMECC_CFG_WRITE_OP) | PMECC_CFG_AUTO_ENABLE);
> +
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DATA);
> +
> + chip->read_buf(mtd, buf, eccsize);
> + chip->read_buf(mtd, oob, mtd->oobsize);
> +
> + while ((pmecc_readl_relaxed(host->ecc, SR) & PMECC_SR_BUSY)) {
> + if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT))
> + return -EIO; /* Time out */
> + cpu_relax();
> + }
> +
> + stat = pmecc_readl_relaxed(host->ecc, ISR);
> + if (stat != 0) {
> + if (pmecc_correction(mtd, stat, buf, &oob[eccpos[0]]) != 0)
> + return -EIO;
> + }
> +
> + return 0;
> +}
> +
> +static void atmel_nand_pmecc_write_page(struct mtd_info *mtd,
> + struct nand_chip *chip, const uint8_t *buf)
> +{
> + int i, j;
> + int timeout_count = 0;
> + struct atmel_nand_host *host = chip->priv;
> + uint32_t *eccpos = chip->ecc.layout->eccpos;
> +
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
> +
> + pmecc_writel(host->ecc, CFG, (pmecc_readl_relaxed(host->ecc, CFG) |
> + PMECC_CFG_WRITE_OP) & ~PMECC_CFG_AUTO_ENABLE);
> +
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DATA);
> +
> + chip->write_buf(mtd, (u8 *)buf, mtd->writesize);
> +
> + while ((pmecc_readl_relaxed(host->ecc, SR) & PMECC_SR_BUSY)) {
> + if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT)) {
> + dev_err(host->dev, "PMECC: Timeout to get ECC value.\n");
> + return; /* Time out */
> + }
> + cpu_relax();
> + }
> +
> + for (i = 0; i < host->pmecc_sector_number; i++) {
> + for (j = 0; j < host->pmecc_bytes_per_sector; j++) {
> + int pos;
> +
> + pos = i * host->pmecc_bytes_per_sector + j;
> + chip->oob_poi[eccpos[pos]] =
> + pmecc_readb_ecc_relaxed(host->ecc, i, j);
> + }
> + }
> + chip->write_buf(mtd, chip->oob_poi, mtd->oobsize);
> +
> + return;
> +}
> +
> +static void atmel_pmecc_core_init(struct mtd_info *mtd)
> +{
> + uint32_t val = 0;
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> + struct nand_ecclayout *ecc_layout;
> +
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
> +
> + switch (host->pmecc_corr_cap) {
> + case 2:
> + val = PMECC_CFG_BCH_ERR2;
> + break;
> + case 4:
> + val = PMECC_CFG_BCH_ERR4;
> + break;
> + case 8:
> + val = PMECC_CFG_BCH_ERR8;
> + break;
> + case 12:
> + val = PMECC_CFG_BCH_ERR12;
> + break;
> + case 24:
> + val = PMECC_CFG_BCH_ERR24;
> + break;
> + }
> +
> + if (host->pmecc_sector_size == 512)
> + val |= PMECC_CFG_SECTOR512;
> + else if (host->pmecc_sector_size == 1024)
> + val |= PMECC_CFG_SECTOR1024;
> +
> + switch (host->pmecc_sector_number) {
> + case 1:
> + val |= PMECC_CFG_PAGE_1SECTOR;
> + break;
> + case 2:
> + val |= PMECC_CFG_PAGE_2SECTORS;
> + break;
> + case 4:
> + val |= PMECC_CFG_PAGE_4SECTORS;
> + break;
> + case 8:
> + val |= PMECC_CFG_PAGE_8SECTORS;
> + break;
> + }
> +
> + val |= (PMECC_CFG_READ_OP | PMECC_CFG_SPARE_DISABLE
> + | PMECC_CFG_AUTO_DISABLE);
> + pmecc_writel(host->ecc, CFG, val);
> +
> + ecc_layout = nand_chip->ecc.layout;
> + pmecc_writel(host->ecc, SAREA, mtd->oobsize - 1);
> + pmecc_writel(host->ecc, SADDR, ecc_layout->eccpos[0]);
> + pmecc_writel(host->ecc, EADDR,
> + ecc_layout->eccpos[ecc_layout->eccbytes - 1]);
> + /* See datasheet about PMECC Clock Control Register */
> + pmecc_writel(host->ecc, CLK, 2);
> + pmecc_writel(host->ecc, IDR, 0xff);
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
> +}
> +
> +static int __init atmel_pmecc_nand_init_params(struct platform_device *pdev,
> + struct atmel_nand_host *host)
> +{
> + int cap, sector_size, err_no;
> + struct mtd_info *mtd;
> + struct nand_chip *nand_chip;
> + struct resource *regs;
> + struct resource *regs_pmerr, *regs_rom;
> +
> + cap = host->pmecc_corr_cap;
> + sector_size = host->pmecc_sector_size;
> + dev_info(host->dev, "Initialize PMECC params, cap: %d, sector: %d\n",
> + cap, sector_size);
> +
> + /* Sanity check */
> + if ((sector_size != 512) && (sector_size != 1024)) {
> + dev_err(host->dev,
> + "Unsupported PMECC sector size: %d; should be 512 or 1024 bytes\n",
> + sector_size);
> + return -EINVAL;
> + }
> + if ((cap != 2) && (cap != 4) && (cap != 8) && (cap != 12) &&
> + (cap != 24)) {
> + dev_err(host->dev,
> + "Unsupported PMECC correction capability, should be 2, 4, 8, 12 or 24\n");
> + return -EINVAL;
> + }
> +
> + nand_chip = &host->nand_chip;
> + mtd = &host->mtd;
> +
> + nand_chip->ecc.mode = NAND_ECC_SOFT; /* By default */
> +
> + regs = platform_get_resource(pdev, IORESOURCE_MEM, 1);
> + if (!regs) {
> + dev_warn(host->dev,
> + "Can't get I/O resource regs, rolling back on software ECC\n");
> + return 0;
> + }
> +
> + host->ecc = ioremap(regs->start, resource_size(regs));
> + if (host->ecc == NULL) {
> + dev_err(host->dev, "ioremap failed\n");
> + err_no = -EIO;
> + goto err_pmecc_ioremap;
> + }
> +
> + regs_pmerr = platform_get_resource(pdev, IORESOURCE_MEM, 2);
> + regs_rom = platform_get_resource(pdev, IORESOURCE_MEM, 3);
> + if (regs_pmerr && regs_rom) {
> + host->pmerrloc_base = ioremap(regs_pmerr->start,
> + resource_size(regs_pmerr));
> + host->pmecc_rom_base = ioremap(regs_rom->start,
> + resource_size(regs_rom));
> +
> + if (host->pmerrloc_base && host->pmecc_rom_base) {
> + nand_chip->ecc.mode = NAND_ECC_HW;
> + nand_chip->ecc.read_page =
> + atmel_nand_pmecc_read_page;
> + nand_chip->ecc.write_page =
> + atmel_nand_pmecc_write_page;
> + } else {
> + dev_err(host->dev,
> + "Can not get I/O resource for PMECC controller!\n");
> + err_no = -EIO;
> + goto err_pmloc_ioremap;
> + }
> + }
> +
> + /* ECC is calculated for the whole page (1 step) */
> + nand_chip->ecc.size = mtd->writesize;
> +
> + /* set ECC page size and oob layout */
> + switch (mtd->writesize) {
> + case 2048:
> + host->pmecc_degree = PMECC_GF_DIMENSION_13;
> + host->pmecc_cw_len = (1 << host->pmecc_degree) - 1;
> + host->pmecc_corr_cap = cap;
> + host->pmecc_sector_number = mtd->writesize / sector_size;
> + host->pmecc_bytes_per_sector = pmecc_get_ecc_bytes(
> + cap, sector_size);
> + host->pmecc_alpha_to = pmecc_get_alpha_to(host);
> + host->pmecc_index_of = pmecc_get_index_of(host);
> +
> + nand_chip->ecc.steps = 1;
> + nand_chip->ecc.strength = cap;
> + nand_chip->ecc.bytes = host->pmecc_bytes_per_sector *
> + host->pmecc_sector_number;
> + if (nand_chip->ecc.bytes > mtd->oobsize - 2) {
> + dev_err(host->dev, "No room for ECC bytes\n");
> + err_no = -EINVAL;
> + goto err;
> + }
> + pmecc_config_ecc_layout(&atmel_pmecc_oobinfo,
> + mtd->oobsize,
> + nand_chip->ecc.bytes);
> + nand_chip->ecc.layout = &atmel_pmecc_oobinfo;
> + break;
> + case 512:
> + case 1024:
> + case 4096:
> + /* TODO */
> + dev_warn(host->dev,
> + "Unsupported page size for PMECC, use Software ECC\n");
> + default:
> + /* page size not handled by HW ECC */
> + /* switching back to soft ECC */
> + nand_chip->ecc.mode = NAND_ECC_SOFT;
> + nand_chip->ecc.calculate = NULL;
> + nand_chip->ecc.correct = NULL;
> + nand_chip->ecc.hwctl = NULL;
> + nand_chip->ecc.read_page = NULL;
> + nand_chip->ecc.write_page = NULL;
> + nand_chip->ecc.postpad = 0;
> + nand_chip->ecc.prepad = 0;
> + nand_chip->ecc.bytes = 0;
> + err_no = 0;
> + goto err;
> + }
> +
> + /* Allocate data for PMECC computation */
> + host->pmecc_data = kzalloc(sizeof(struct atmel_pmecc_data), GFP_KERNEL);
why do you always allocate the pmecc_data?
you need to allocate it only if you use it
Best Regards,
J.
> + if (!host->pmecc_data) {
> + dev_err(host->dev,
> + "Cannot allocate memory for PMECC computation!\n");
> + err_no = -ENOMEM;
> + goto err;
> + }
> +
> + atmel_pmecc_core_init(mtd);
> +
> + return 0;
> +
> +err:
> +err_pmloc_ioremap:
> + iounmap(host->ecc);
> + if (host->pmerrloc_base)
> + iounmap(host->pmerrloc_base);
> + if (host->pmecc_rom_base)
> + iounmap(host->pmecc_rom_base);
> +err_pmecc_ioremap:
> + return err_no;
> +}
> +
> +/*
> * Calculate HW ECC
> *
> * function called after a write
> @@ -720,7 +1456,11 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
> }
>
> if (nand_chip->ecc.mode == NAND_ECC_HW) {
> - res = atmel_hw_nand_init_params(pdev, host);
> + if (host->has_pmecc)
> + res = atmel_pmecc_nand_init_params(pdev, host);
> + else
> + res = atmel_hw_nand_init_params(pdev, host);
> +
> if (res != 0)
> goto err_hw_ecc;
> }
> @@ -741,6 +1481,12 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
> err_scan_tail:
> if (host->ecc)
> iounmap(host->ecc);
> + if (host->has_pmecc) {
no need to check if you have teh pmecc
if it's no the case pmerrloc_base will be NULLo
don't you need to disable it in the error path?
Best Regards,
J.
^ permalink raw reply [flat|nested] 24+ messages in thread
* [PATCH v9 3/3] MTD: at91: atmel_nand: Update driver to support Programmable Multibit ECC controller
@ 2012-05-28 6:58 ` Jean-Christophe PLAGNIOL-VILLARD
0 siblings, 0 replies; 24+ messages in thread
From: Jean-Christophe PLAGNIOL-VILLARD @ 2012-05-28 6:58 UTC (permalink / raw)
To: linux-arm-kernel
On 21:24 Sat 26 May , Josh Wu wrote:
> The Programmable Multibit ECC (PMECC) controller is a programmable binary
> BCH(Bose, Chaudhuri and Hocquenghem) encoder and decoder. This controller
> can be used to support both SLC and MLC NAND Flash devices. It supports to
> generate ECC to correct 2, 4, 8, 12 or 24 bits of error per sector of data.
>
> To use this driver, the user needs to pass in the correction capability and
> the sector size.
>
> This driver has been tested on AT91SAM9X5-EK and AT91SAM9N12-EK with JFFS2,
> YAFFS2, UBIFS and mtd-utils.
>
> Signed-off-by: Hong Xu <hong.xu@atmel.com>
> Signed-off-by: Josh Wu <josh.wu@atmel.com>
> ---
> drivers/mtd/nand/atmel_nand.c | 761 ++++++++++++++++++++++++++++++++++++-
> drivers/mtd/nand/atmel_nand_ecc.h | 116 ++++++
> 2 files changed, 876 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/mtd/nand/atmel_nand.c b/drivers/mtd/nand/atmel_nand.c
> index 9a9bfbf..ddcf1ed 100644
> --- a/drivers/mtd/nand/atmel_nand.c
> +++ b/drivers/mtd/nand/atmel_nand.c
> @@ -15,6 +15,8 @@
> * (u-boot-1.1.5/board/atmel/at91sam9263ek/nand.c)
> * (C) Copyright 2006 ATMEL Rousset, Lacressonniere Nicolas
> *
> + * Add Programmable Multibit ECC support for various AT91 SoC
> + * (C) Copyright 2012 ATMEL, Hong Xu
> *
> * This program is free software; you can redistribute it and/or modify
> * it under the terms of the GNU General Public License version 2 as
> @@ -77,6 +79,21 @@ static struct nand_ecclayout atmel_oobinfo_small = {
> },
> };
>
> +/* a structure includes datas for PMECC computation */
> +struct atmel_pmecc_data {
> + int16_t partial_syn[2 * PMECC_MAX_ERROR_NB + 1];
> + int16_t si[2 * PMECC_MAX_ERROR_NB + 1];
> +
> + /* Sigma table */
> + int16_t smu[PMECC_MAX_ERROR_NB + 2][2 * PMECC_MAX_ERROR_NB + 1];
you still hardcode the array in the struct
and if the pmecc evolve we will have to touch again
please allocate them
> + /* polynomal order */
> + int16_t lmu[PMECC_MAX_ERROR_NB + 1];
> +
> + int mu[PMECC_MAX_ERROR_NB + 1];
> + int dmu[PMECC_MAX_ERROR_NB + 1];
> + int delta[PMECC_MAX_ERROR_NB + 1];
> +};
> +
> struct atmel_nand_host {
> struct nand_chip nand_chip;
> struct mtd_info mtd;
> @@ -92,8 +109,25 @@ struct atmel_nand_host {
> bool has_pmecc;
> u8 pmecc_corr_cap;
> u16 pmecc_sector_size;
> +
> + int pmecc_bytes_per_sector;
> + int pmecc_sector_number;
> + int pmecc_degree; /* Degree of remainders */
> + int pmecc_cw_len; /* Length of codeword */
> +
> + void __iomem *pmerrloc_base;
> + void __iomem *pmecc_rom_base;
> +
> + /* lookup table for alpha_to and index_of */
> + void __iomem *pmecc_alpha_to;
> + void __iomem *pmecc_index_of;
> +
> + /* data for pmecc computation */
> + struct atmel_pmecc_data *pmecc_data;
> };
>
> +static struct nand_ecclayout atmel_pmecc_oobinfo;
> +
> static int cpu_has_dma(void)
> {
> return cpu_is_at91sam9rl() || cpu_is_at91sam9g45();
> @@ -287,6 +321,708 @@ static void atmel_write_buf(struct mtd_info *mtd, const u8 *buf, int len)
> }
>
> /*
> + * Return number of ecc bytes per sector according to sector size and
> + * correction capability
> + *
> + * Following table shows what at91 PMECC supported:
> + * Correction Capability Sector_512_bytes Sector_1024_bytes
> + * ===================== ================ =================
> + * 2-bits 4-bytes 4-bytes
> + * 4-bits 7-bytes 7-bytes
> + * 8-bits 13-bytes 14-bytes
> + * 12-bits 20-bytes 21-bytes
> + * 24-bits 39-bytes 42-bytes
> + */
> +static int pmecc_get_ecc_bytes(int cap, int sector_size)
> +{
> + int m = 12 + sector_size / 512;
> + return (m * cap + 7) / 8;
> +}
> +
> +static void pmecc_config_ecc_layout(struct nand_ecclayout *layout, int oobsize,
> + int ecc_len)
> +{
> + int i;
> +
> + layout->eccbytes = ecc_len;
> +
> + /* ECC will occupy the last ecc_len bytes continuously */
> + for (i = 0; i < ecc_len; i++)
> + layout->eccpos[i] = oobsize - ecc_len + i;
> +
> + layout->oobfree[0].offset = 2;
> + layout->oobfree[0].length =
> + oobsize - ecc_len - layout->oobfree[0].offset;
> +}
> +
> +static void __iomem *pmecc_get_alpha_to(struct atmel_nand_host *host)
> +{
> + void __iomem *p;
> +
> + switch (host->pmecc_sector_size) {
> + case 512:
> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_512 +
> + PMECC_LOOKUP_TABLE_SIZE_512 * sizeof(int16_t);
> + break;
> + case 1024:
> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_1024 +
> + PMECC_LOOKUP_TABLE_SIZE_1024 * sizeof(int16_t);
> + break;
> + default:
> + BUG();
> + }
> +
> + return p;
> +}
> +
> +static void __iomem *pmecc_get_index_of(struct atmel_nand_host *host)
this is a __dev_init function plese check the other too
btw you need to use __dev_init and not __init
> +{
> + void __iomem *p;
> +
> + switch (host->pmecc_sector_size) {
> + case 512:
> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_512;
> + break;
> + case 1024:
> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_1024;
> + break;
> + default:
> + BUG();
> + }
> +
> + return p;
> +}
> +
> +static void pmecc_gen_syndrome(struct mtd_info *mtd, int sector)
> +{
> + int i;
> + uint32_t value;
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> +
> + /* Fill odd syndromes */
> + for (i = 0; i < host->pmecc_corr_cap; i++) {
> + value = pmecc_readl_rem_relaxed(host->ecc, sector, i / 2);
> + value = (i & 1) ? (value & 0xffff0000) >> 16 : value & 0xffff;
simplify by
if (i & 1)
val >>= 16;
value &= 0xffff;
> + host->pmecc_data->partial_syn[(2 * i) + 1] = (int16_t)value;
> + }
> +}
> +
> +static void pmecc_substitute(struct mtd_info *mtd)
> +{
> + int16_t *si;
> + int i, j;
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> + int16_t __iomem *alpha_to = host->pmecc_alpha_to;
> + int16_t __iomem *index_of = host->pmecc_index_of;
> + int16_t *partial_syn = host->pmecc_data->partial_syn;
> +
> + /* si[] is a table that holds the current syndrome value,
> + * an element of that table belongs to the field
> + */
> + si = host->pmecc_data->si;
> +
> + for (i = 1; i < 2 * PMECC_MAX_ERROR_NB; i++)
> + si[i] = 0;
please use memset
> +
> + /* Computation 2t syndromes based on S(x) */
> + /* Odd syndromes */
> + for (i = 1; i < 2 * host->pmecc_corr_cap; i += 2) {
> + si[i] = 0;
shy this you already init the array at 0 before
> + for (j = 0; j < host->pmecc_degree; j++) {
> + if (partial_syn[i] & ((unsigned short)0x1 << j))
> + si[i] = readw_relaxed(alpha_to + i * j) ^ si[i];
> + }
> + }
> + /* Even syndrome = (Odd syndrome) ** 2 */
> + for (i = 2; i <= 2 * host->pmecc_corr_cap; i += 2) {
> + j = i / 2;
> + if (si[j] == 0)
here if {
} else {
}
> + si[i] = 0;
> + else {
> + int16_t tmp;
missing blank line
> + tmp = readw_relaxed(index_of + si[j]);
> + tmp = (tmp * 2) % host->pmecc_cw_len;
> + si[i] = readw_relaxed(alpha_to + tmp);
> + }
> + }
> +
> + return;
> +}
> +
> +static void pmecc_get_sigma(struct mtd_info *mtd)
> +{
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> +
> + int i, j, k;
> + uint32_t dmu_0_count, tmp;
> + int16_t (*smu)[2 * PMECC_MAX_ERROR_NB + 1];
> + int16_t *lmu = host->pmecc_data->lmu;
> + int16_t *si = host->pmecc_data->si;
> + int *mu = host->pmecc_data->mu;
> + int *dmu = host->pmecc_data->dmu; /* Discrepancy */
> + int *delta = host->pmecc_data->delta; /* Delta order */
> + int cw_len = host->pmecc_cw_len;
> + int16_t cap = host->pmecc_corr_cap;
> +
> + int16_t __iomem *index_of = host->pmecc_index_of;
> + int16_t __iomem *alpha_to = host->pmecc_alpha_to;
> +
> + /* index of largest delta */
> + int ro;
> + int largest;
> + int diff;
> +
> + dmu_0_count = 0;
> + smu = host->pmecc_data->smu;
> +
> + /* First Row */
> +
> + /* Mu */
> + mu[0] = -1;
> +
> + memset(&smu[0][0], 0,
> + sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
> + smu[0][0] = 1;
> +
> + /* discrepancy set to 1 */
> + dmu[0] = 1;
> + /* polynom order set to 0 */
> + lmu[0] = 0;
> + delta[0] = (mu[0] * 2 - lmu[0]) >> 1;
> +
> + /* Second Row */
> +
> + /* Mu */
> + mu[1] = 0;
> + /* Sigma(x) set to 1 */
> + memset(&smu[1][0], 0,
> + sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
> + smu[1][0] = 1;
> +
> + /* discrepancy set to S1 */
> + dmu[1] = si[1];
> +
> + /* polynom order set to 0 */
> + lmu[1] = 0;
> +
> + delta[1] = (mu[1] * 2 - lmu[1]) >> 1;
> +
> + /* Init the Sigma(x) last row */
> + memset(&smu[cap + 1][0], 0,
> + sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
> +
> + for (i = 1; i <= cap; i++) {
> + mu[i+1] = i << 1;
> + /* Begin Computing Sigma (Mu+1) and L(mu) */
> + /* check if discrepancy is set to 0 */
> + if (dmu[i] == 0) {
> + dmu_0_count++;
> +
> + tmp = ((cap - (lmu[i] >> 1) - 1) / 2);
> + if ((cap - (lmu[i] >> 1) - 1) & 0x1)
> + tmp += 2;
> + else
> + tmp += 1;
> +
> + if (dmu_0_count == tmp) {
> + for (j = 0; j <= (lmu[i] >> 1) + 1; j++)
> + smu[cap + 1][j] = smu[i][j];
> + lmu[cap + 1] = lmu[i];
> + return;
> + }
> +
> + /* copy polynom */
> + for (j = 0; j <= lmu[i] >> 1; j++)
> + smu[i + 1][j] = smu[i][j];
> +
> + /* copy previous polynom order to the next */
> + lmu[i + 1] = lmu[i];
> + } else {
> + ro = 0;
> + largest = -1;
> + /* find largest delta with dmu != 0 */
> + for (j = 0; j < i; j++) {
> + if ((dmu[j]) && (delta[j] > largest)) {
> + largest = delta[j];
> + ro = j;
> + }
> + }
> +
> + /* compute difference */
> + diff = (mu[i] - mu[ro]);
> +
> + /* Compute degree of the new smu polynomial */
> + if ((lmu[i] >> 1) > ((lmu[ro] >> 1) + diff))
> + lmu[i + 1] = lmu[i];
> + else
> + lmu[i + 1] = ((lmu[ro] >> 1) + diff) * 2;
> +
> + /* Init smu[i+1] with 0 */
> + for (k = 0; k < (2 * PMECC_MAX_ERROR_NB + 1); k++)
> + smu[i+1][k] = 0;
> +
> + /* Compute smu[i+1] */
> + for (k = 0; k <= lmu[ro] >> 1; k++) {
> + int16_t a, b, c;
> +
> + if (!(smu[ro][k] && dmu[i]))
> + continue;
> + a = readw_relaxed(index_of + dmu[i]);
> + b = readw_relaxed(index_of + dmu[ro]);
> + c = readw_relaxed(index_of + smu[ro][k]);
> + tmp = a + (cw_len - b) + c;
> + a = readw_relaxed(alpha_to + tmp % cw_len);
> + smu[i + 1][k + diff] = a;
> + }
> +
> + for (k = 0; k <= lmu[i] >> 1; k++)
> + smu[i + 1][k] ^= smu[i][k];
> + }
> +
> + /* End Computing Sigma (Mu+1) and L(mu) */
> + /* In either case compute delta */
> + delta[i + 1] = (mu[i + 1] * 2 - lmu[i + 1]) >> 1;
> +
> + /* Do not compute discrepancy for the last iteration */
> + if (i >= cap)
> + continue;
> +
> + for (k = 0 ; k <= (lmu[i + 1] >> 1); k++) {
> + tmp = 2 * (i - 1);
> + if (k == 0)
> + dmu[i + 1] = si[tmp + 3];
> + else if (smu[i+1][k] && si[tmp + 3 - k]) {
> + int16_t a, b, c;
> + a = readw_relaxed(index_of + smu[i + 1][k]);
> + b = si[2 * (i - 1) + 3 - k];
> + c = readw_relaxed(index_of + b);
> + tmp = a + c;
> + tmp %= cw_len;
> + dmu[i + 1] = readw_relaxed(alpha_to + tmp) ^
> + dmu[i + 1];
> + }
> + }
> + }
> +
> + return;
> +}
> +
> +static int pmecc_err_location(struct mtd_info *mtd)
> +{
> + int i;
> + int err_nbr; /* number of error */
> + int roots_nbr; /* number of roots */
> + int sector_size;
> + uint32_t val;
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> + int timeout_count = 0;
> + int cap = host->pmecc_corr_cap;
> +
> + err_nbr = 0;
> + sector_size = host->pmecc_sector_size;
> +
> + pmerrloc_writel(host->pmerrloc_base, ELDIS, PMERRLOC_DISABLE);
> +
> + for (i = 0; i <= host->pmecc_data->lmu[cap + 1] >> 1; i++) {
> + pmerrloc_writel_sigma_relaxed(host->pmerrloc_base, i,
> + host->pmecc_data->smu[cap + 1][i]);
> + err_nbr++;
> + }
> +
> + val = (err_nbr - 1) << 16;
> + if (sector_size == 1024)
> + val |= 1;
> +
> + pmerrloc_writel(host->pmerrloc_base, ELCFG, val);
> + pmerrloc_writel(host->pmerrloc_base, ELEN,
> + sector_size * 8 + host->pmecc_degree * cap);
> +
> + while (!(pmerrloc_readl_relaxed(host->pmerrloc_base, ELISR)
> + & PMERRLOC_CALC_DONE)) {
> + if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT))
> + return -1; /* Time out */
> + cpu_relax();
> + }
> +
> + roots_nbr = (pmerrloc_readl_relaxed(host->pmerrloc_base, ELISR)
> + & PMERRLOC_ERR_NUM_MASK) >> 8;
> + /* Number of roots == degree of smu hence <= cap */
> + if (roots_nbr == host->pmecc_data->lmu[cap + 1] >> 1)
> + return err_nbr - 1;
> +
> + /* Number of roots does not match the degree of smu
> + * unable to correct error */
> + return -1;
> +}
> +
> +static void pmecc_correct_data(struct mtd_info *mtd, uint8_t *buf,
> + int extra_bytes, int err_nbr)
> +{
> + int i = 0;
> + int byte_pos, bit_pos;
> + int sector_size, ecc_size;
> + uint32_t tmp;
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> +
> + sector_size = host->pmecc_sector_size;
> + ecc_size = nand_chip->ecc.bytes;
> +
> + while (err_nbr) {
> + tmp = pmerrloc_readl_el_relaxed(host->pmerrloc_base, i) - 1;
> + byte_pos = tmp / 8;
> + bit_pos = tmp % 8;
> + dev_info(host->dev, "PMECC correction, byte_pos: %d bit_pos: %d\n",
> + byte_pos, bit_pos);
> +
> + if (byte_pos < (sector_size + extra_bytes)) {
> + tmp = sector_size +
> + pmecc_readl_relaxed(host->ecc, SADDR);
> +
> + if (byte_pos < tmp)
> + *(buf + byte_pos) ^= (1 << bit_pos);
> + else
> + *(buf + byte_pos + ecc_size) ^= (1 << bit_pos);
> + }
> +
> + i++;
> + err_nbr--;
> + }
> +
> + return;
> +}
> +
> +static int pmecc_correction(struct mtd_info *mtd, u32 pmecc_stat, uint8_t *buf,
> + u8 *ecc)
> +{
> + int i, err_nbr;
> + uint8_t *buf_pos;
> + int eccbytes;
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> +
> + eccbytes = nand_chip->ecc.bytes;
> + for (i = 0; i < eccbytes; i++)
> + if (ecc[i] != 0xff)
> + goto normal_check;
> + /* Erased page, return OK */
> + return 0;
> +
> +normal_check:
> + for (i = 0; i < host->pmecc_sector_number; i++) {
> + err_nbr = 0;
> + if (pmecc_stat & 0x1) {
> + buf_pos = buf + i * host->pmecc_sector_size;
> +
> + pmecc_gen_syndrome(mtd, i);
> + pmecc_substitute(mtd);
> + pmecc_get_sigma(mtd);
> +
> + err_nbr = pmecc_err_location(mtd);
> + if (err_nbr == -1) {
> + dev_err(host->dev, "PMECC: Too many errors\n");
> + mtd->ecc_stats.failed++;
> + return -EIO;
> + } else {
> + pmecc_correct_data(mtd, buf_pos, 0, err_nbr);
> + mtd->ecc_stats.corrected += err_nbr;
> + }
> + }
> + pmecc_stat >>= 1;
> + }
> +
> + return 0;
> +}
> +
> +static int atmel_nand_pmecc_read_page(struct mtd_info *mtd,
> + struct nand_chip *chip, uint8_t *buf, int page)
> +{
> + uint32_t stat;
> + int timeout_count = 0;
> + int eccsize = chip->ecc.size;
> + uint8_t *oob = chip->oob_poi;
> + struct atmel_nand_host *host = chip->priv;
> + uint32_t *eccpos = chip->ecc.layout->eccpos;
> +
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
> + pmecc_writel(host->ecc, CFG, (pmecc_readl_relaxed(host->ecc, CFG)
> + & ~PMECC_CFG_WRITE_OP) | PMECC_CFG_AUTO_ENABLE);
> +
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DATA);
> +
> + chip->read_buf(mtd, buf, eccsize);
> + chip->read_buf(mtd, oob, mtd->oobsize);
> +
> + while ((pmecc_readl_relaxed(host->ecc, SR) & PMECC_SR_BUSY)) {
> + if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT))
> + return -EIO; /* Time out */
> + cpu_relax();
> + }
> +
> + stat = pmecc_readl_relaxed(host->ecc, ISR);
> + if (stat != 0) {
> + if (pmecc_correction(mtd, stat, buf, &oob[eccpos[0]]) != 0)
> + return -EIO;
> + }
> +
> + return 0;
> +}
> +
> +static void atmel_nand_pmecc_write_page(struct mtd_info *mtd,
> + struct nand_chip *chip, const uint8_t *buf)
> +{
> + int i, j;
> + int timeout_count = 0;
> + struct atmel_nand_host *host = chip->priv;
> + uint32_t *eccpos = chip->ecc.layout->eccpos;
> +
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
> +
> + pmecc_writel(host->ecc, CFG, (pmecc_readl_relaxed(host->ecc, CFG) |
> + PMECC_CFG_WRITE_OP) & ~PMECC_CFG_AUTO_ENABLE);
> +
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DATA);
> +
> + chip->write_buf(mtd, (u8 *)buf, mtd->writesize);
> +
> + while ((pmecc_readl_relaxed(host->ecc, SR) & PMECC_SR_BUSY)) {
> + if (unlikely(timeout_count++ > PMECC_MAX_TIMEOUT_COUNT)) {
> + dev_err(host->dev, "PMECC: Timeout to get ECC value.\n");
> + return; /* Time out */
> + }
> + cpu_relax();
> + }
> +
> + for (i = 0; i < host->pmecc_sector_number; i++) {
> + for (j = 0; j < host->pmecc_bytes_per_sector; j++) {
> + int pos;
> +
> + pos = i * host->pmecc_bytes_per_sector + j;
> + chip->oob_poi[eccpos[pos]] =
> + pmecc_readb_ecc_relaxed(host->ecc, i, j);
> + }
> + }
> + chip->write_buf(mtd, chip->oob_poi, mtd->oobsize);
> +
> + return;
> +}
> +
> +static void atmel_pmecc_core_init(struct mtd_info *mtd)
> +{
> + uint32_t val = 0;
> + struct nand_chip *nand_chip = mtd->priv;
> + struct atmel_nand_host *host = nand_chip->priv;
> + struct nand_ecclayout *ecc_layout;
> +
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
> +
> + switch (host->pmecc_corr_cap) {
> + case 2:
> + val = PMECC_CFG_BCH_ERR2;
> + break;
> + case 4:
> + val = PMECC_CFG_BCH_ERR4;
> + break;
> + case 8:
> + val = PMECC_CFG_BCH_ERR8;
> + break;
> + case 12:
> + val = PMECC_CFG_BCH_ERR12;
> + break;
> + case 24:
> + val = PMECC_CFG_BCH_ERR24;
> + break;
> + }
> +
> + if (host->pmecc_sector_size == 512)
> + val |= PMECC_CFG_SECTOR512;
> + else if (host->pmecc_sector_size == 1024)
> + val |= PMECC_CFG_SECTOR1024;
> +
> + switch (host->pmecc_sector_number) {
> + case 1:
> + val |= PMECC_CFG_PAGE_1SECTOR;
> + break;
> + case 2:
> + val |= PMECC_CFG_PAGE_2SECTORS;
> + break;
> + case 4:
> + val |= PMECC_CFG_PAGE_4SECTORS;
> + break;
> + case 8:
> + val |= PMECC_CFG_PAGE_8SECTORS;
> + break;
> + }
> +
> + val |= (PMECC_CFG_READ_OP | PMECC_CFG_SPARE_DISABLE
> + | PMECC_CFG_AUTO_DISABLE);
> + pmecc_writel(host->ecc, CFG, val);
> +
> + ecc_layout = nand_chip->ecc.layout;
> + pmecc_writel(host->ecc, SAREA, mtd->oobsize - 1);
> + pmecc_writel(host->ecc, SADDR, ecc_layout->eccpos[0]);
> + pmecc_writel(host->ecc, EADDR,
> + ecc_layout->eccpos[ecc_layout->eccbytes - 1]);
> + /* See datasheet about PMECC Clock Control Register */
> + pmecc_writel(host->ecc, CLK, 2);
> + pmecc_writel(host->ecc, IDR, 0xff);
> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
> +}
> +
> +static int __init atmel_pmecc_nand_init_params(struct platform_device *pdev,
> + struct atmel_nand_host *host)
> +{
> + int cap, sector_size, err_no;
> + struct mtd_info *mtd;
> + struct nand_chip *nand_chip;
> + struct resource *regs;
> + struct resource *regs_pmerr, *regs_rom;
> +
> + cap = host->pmecc_corr_cap;
> + sector_size = host->pmecc_sector_size;
> + dev_info(host->dev, "Initialize PMECC params, cap: %d, sector: %d\n",
> + cap, sector_size);
> +
> + /* Sanity check */
> + if ((sector_size != 512) && (sector_size != 1024)) {
> + dev_err(host->dev,
> + "Unsupported PMECC sector size: %d; should be 512 or 1024 bytes\n",
> + sector_size);
> + return -EINVAL;
> + }
> + if ((cap != 2) && (cap != 4) && (cap != 8) && (cap != 12) &&
> + (cap != 24)) {
> + dev_err(host->dev,
> + "Unsupported PMECC correction capability, should be 2, 4, 8, 12 or 24\n");
> + return -EINVAL;
> + }
> +
> + nand_chip = &host->nand_chip;
> + mtd = &host->mtd;
> +
> + nand_chip->ecc.mode = NAND_ECC_SOFT; /* By default */
> +
> + regs = platform_get_resource(pdev, IORESOURCE_MEM, 1);
> + if (!regs) {
> + dev_warn(host->dev,
> + "Can't get I/O resource regs, rolling back on software ECC\n");
> + return 0;
> + }
> +
> + host->ecc = ioremap(regs->start, resource_size(regs));
> + if (host->ecc == NULL) {
> + dev_err(host->dev, "ioremap failed\n");
> + err_no = -EIO;
> + goto err_pmecc_ioremap;
> + }
> +
> + regs_pmerr = platform_get_resource(pdev, IORESOURCE_MEM, 2);
> + regs_rom = platform_get_resource(pdev, IORESOURCE_MEM, 3);
> + if (regs_pmerr && regs_rom) {
> + host->pmerrloc_base = ioremap(regs_pmerr->start,
> + resource_size(regs_pmerr));
> + host->pmecc_rom_base = ioremap(regs_rom->start,
> + resource_size(regs_rom));
> +
> + if (host->pmerrloc_base && host->pmecc_rom_base) {
> + nand_chip->ecc.mode = NAND_ECC_HW;
> + nand_chip->ecc.read_page =
> + atmel_nand_pmecc_read_page;
> + nand_chip->ecc.write_page =
> + atmel_nand_pmecc_write_page;
> + } else {
> + dev_err(host->dev,
> + "Can not get I/O resource for PMECC controller!\n");
> + err_no = -EIO;
> + goto err_pmloc_ioremap;
> + }
> + }
> +
> + /* ECC is calculated for the whole page (1 step) */
> + nand_chip->ecc.size = mtd->writesize;
> +
> + /* set ECC page size and oob layout */
> + switch (mtd->writesize) {
> + case 2048:
> + host->pmecc_degree = PMECC_GF_DIMENSION_13;
> + host->pmecc_cw_len = (1 << host->pmecc_degree) - 1;
> + host->pmecc_corr_cap = cap;
> + host->pmecc_sector_number = mtd->writesize / sector_size;
> + host->pmecc_bytes_per_sector = pmecc_get_ecc_bytes(
> + cap, sector_size);
> + host->pmecc_alpha_to = pmecc_get_alpha_to(host);
> + host->pmecc_index_of = pmecc_get_index_of(host);
> +
> + nand_chip->ecc.steps = 1;
> + nand_chip->ecc.strength = cap;
> + nand_chip->ecc.bytes = host->pmecc_bytes_per_sector *
> + host->pmecc_sector_number;
> + if (nand_chip->ecc.bytes > mtd->oobsize - 2) {
> + dev_err(host->dev, "No room for ECC bytes\n");
> + err_no = -EINVAL;
> + goto err;
> + }
> + pmecc_config_ecc_layout(&atmel_pmecc_oobinfo,
> + mtd->oobsize,
> + nand_chip->ecc.bytes);
> + nand_chip->ecc.layout = &atmel_pmecc_oobinfo;
> + break;
> + case 512:
> + case 1024:
> + case 4096:
> + /* TODO */
> + dev_warn(host->dev,
> + "Unsupported page size for PMECC, use Software ECC\n");
> + default:
> + /* page size not handled by HW ECC */
> + /* switching back to soft ECC */
> + nand_chip->ecc.mode = NAND_ECC_SOFT;
> + nand_chip->ecc.calculate = NULL;
> + nand_chip->ecc.correct = NULL;
> + nand_chip->ecc.hwctl = NULL;
> + nand_chip->ecc.read_page = NULL;
> + nand_chip->ecc.write_page = NULL;
> + nand_chip->ecc.postpad = 0;
> + nand_chip->ecc.prepad = 0;
> + nand_chip->ecc.bytes = 0;
> + err_no = 0;
> + goto err;
> + }
> +
> + /* Allocate data for PMECC computation */
> + host->pmecc_data = kzalloc(sizeof(struct atmel_pmecc_data), GFP_KERNEL);
why do you always allocate the pmecc_data?
you need to allocate it only if you use it
Best Regards,
J.
> + if (!host->pmecc_data) {
> + dev_err(host->dev,
> + "Cannot allocate memory for PMECC computation!\n");
> + err_no = -ENOMEM;
> + goto err;
> + }
> +
> + atmel_pmecc_core_init(mtd);
> +
> + return 0;
> +
> +err:
> +err_pmloc_ioremap:
> + iounmap(host->ecc);
> + if (host->pmerrloc_base)
> + iounmap(host->pmerrloc_base);
> + if (host->pmecc_rom_base)
> + iounmap(host->pmecc_rom_base);
> +err_pmecc_ioremap:
> + return err_no;
> +}
> +
> +/*
> * Calculate HW ECC
> *
> * function called after a write
> @@ -720,7 +1456,11 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
> }
>
> if (nand_chip->ecc.mode == NAND_ECC_HW) {
> - res = atmel_hw_nand_init_params(pdev, host);
> + if (host->has_pmecc)
> + res = atmel_pmecc_nand_init_params(pdev, host);
> + else
> + res = atmel_hw_nand_init_params(pdev, host);
> +
> if (res != 0)
> goto err_hw_ecc;
> }
> @@ -741,6 +1481,12 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
> err_scan_tail:
> if (host->ecc)
> iounmap(host->ecc);
> + if (host->has_pmecc) {
no need to check if you have teh pmecc
if it's no the case pmerrloc_base will be NULLo
don't you need to disable it in the error path?
Best Regards,
J.
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v9 3/3] MTD: at91: atmel_nand: Update driver to support Programmable Multibit ECC controller
2012-05-28 6:58 ` Jean-Christophe PLAGNIOL-VILLARD
@ 2012-05-28 8:34 ` Josh Wu
-1 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-28 8:34 UTC (permalink / raw)
To: Jean-Christophe PLAGNIOL-VILLARD
Cc: hongxu.cn, dedekind1, nicolas.ferre, linux-mtd, ivan.djelic,
linux-arm-kernel
On 5/28/2012 2:58 PM, Jean-Christophe PLAGNIOL-VILLARD wrote:
> On 21:24 Sat 26 May , Josh Wu wrote:
>> The Programmable Multibit ECC (PMECC) controller is a programmable binary
>> BCH(Bose, Chaudhuri and Hocquenghem) encoder and decoder. This controller
>> can be used to support both SLC and MLC NAND Flash devices. It supports to
>> generate ECC to correct 2, 4, 8, 12 or 24 bits of error per sector of data.
>>
>> To use this driver, the user needs to pass in the correction capability and
>> the sector size.
>>
>> This driver has been tested on AT91SAM9X5-EK and AT91SAM9N12-EK with JFFS2,
>> YAFFS2, UBIFS and mtd-utils.
>>
>> Signed-off-by: Hong Xu<hong.xu@atmel.com>
>> Signed-off-by: Josh Wu<josh.wu@atmel.com>
>> ---
>> drivers/mtd/nand/atmel_nand.c | 761 ++++++++++++++++++++++++++++++++++++-
>> drivers/mtd/nand/atmel_nand_ecc.h | 116 ++++++
>> 2 files changed, 876 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/mtd/nand/atmel_nand.c b/drivers/mtd/nand/atmel_nand.c
>> index 9a9bfbf..ddcf1ed 100644
>> --- a/drivers/mtd/nand/atmel_nand.c
>> +++ b/drivers/mtd/nand/atmel_nand.c
>> @@ -15,6 +15,8 @@
>> * (u-boot-1.1.5/board/atmel/at91sam9263ek/nand.c)
>> * (C) Copyright 2006 ATMEL Rousset, Lacressonniere Nicolas
>> *
>> + * Add Programmable Multibit ECC support for various AT91 SoC
>> + * (C) Copyright 2012 ATMEL, Hong Xu
>> *
>> * This program is free software; you can redistribute it and/or modify
>> * it under the terms of the GNU General Public License version 2 as
>> @@ -77,6 +79,21 @@ static struct nand_ecclayout atmel_oobinfo_small = {
>> },
>> };
>>
>> +/* a structure includes datas for PMECC computation */
>> +struct atmel_pmecc_data {
>> + int16_t partial_syn[2 * PMECC_MAX_ERROR_NB + 1];
>> + int16_t si[2 * PMECC_MAX_ERROR_NB + 1];
>> +
>> + /* Sigma table */
>> + int16_t smu[PMECC_MAX_ERROR_NB + 2][2 * PMECC_MAX_ERROR_NB + 1];
> you still hardcode the array in the struct
>
> and if the pmecc evolve we will have to touch again
> please allocate them
ok.
>> + /* polynomal order */
>> + int16_t lmu[PMECC_MAX_ERROR_NB + 1];
>> +
>> + int mu[PMECC_MAX_ERROR_NB + 1];
>> + int dmu[PMECC_MAX_ERROR_NB + 1];
>> + int delta[PMECC_MAX_ERROR_NB + 1];
>> +};
>> +
>> struct atmel_nand_host {
>> struct nand_chip nand_chip;
>> struct mtd_info mtd;
>> @@ -92,8 +109,25 @@ struct atmel_nand_host {
>> bool has_pmecc;
>> u8 pmecc_corr_cap;
>> u16 pmecc_sector_size;
>> +
>> + int pmecc_bytes_per_sector;
>> + int pmecc_sector_number;
>> + int pmecc_degree; /* Degree of remainders */
>> + int pmecc_cw_len; /* Length of codeword */
>> +
>> + void __iomem *pmerrloc_base;
>> + void __iomem *pmecc_rom_base;
>> +
>> + /* lookup table for alpha_to and index_of */
>> + void __iomem *pmecc_alpha_to;
>> + void __iomem *pmecc_index_of;
>> +
>> + /* data for pmecc computation */
>> + struct atmel_pmecc_data *pmecc_data;
>> };
>>
>> +static struct nand_ecclayout atmel_pmecc_oobinfo;
>> +
>> static int cpu_has_dma(void)
>> {
>> return cpu_is_at91sam9rl() || cpu_is_at91sam9g45();
>> @@ -287,6 +321,708 @@ static void atmel_write_buf(struct mtd_info *mtd, const u8 *buf, int len)
>> }
>>
>> /*
>> + * Return number of ecc bytes per sector according to sector size and
>> + * correction capability
>> + *
>> + * Following table shows what at91 PMECC supported:
>> + * Correction Capability Sector_512_bytes Sector_1024_bytes
>> + * ===================== ================ =================
>> + * 2-bits 4-bytes 4-bytes
>> + * 4-bits 7-bytes 7-bytes
>> + * 8-bits 13-bytes 14-bytes
>> + * 12-bits 20-bytes 21-bytes
>> + * 24-bits 39-bytes 42-bytes
>> + */
>> +static int pmecc_get_ecc_bytes(int cap, int sector_size)
>> +{
>> + int m = 12 + sector_size / 512;
>> + return (m * cap + 7) / 8;
>> +}
>> +
>> +static void pmecc_config_ecc_layout(struct nand_ecclayout *layout, int oobsize,
>> + int ecc_len)
>> +{
>> + int i;
>> +
>> + layout->eccbytes = ecc_len;
>> +
>> + /* ECC will occupy the last ecc_len bytes continuously */
>> + for (i = 0; i< ecc_len; i++)
>> + layout->eccpos[i] = oobsize - ecc_len + i;
>> +
>> + layout->oobfree[0].offset = 2;
>> + layout->oobfree[0].length =
>> + oobsize - ecc_len - layout->oobfree[0].offset;
>> +}
>> +
>> +static void __iomem *pmecc_get_alpha_to(struct atmel_nand_host *host)
>> +{
>> + void __iomem *p;
>> +
>> + switch (host->pmecc_sector_size) {
>> + case 512:
>> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_512 +
>> + PMECC_LOOKUP_TABLE_SIZE_512 * sizeof(int16_t);
>> + break;
>> + case 1024:
>> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_1024 +
>> + PMECC_LOOKUP_TABLE_SIZE_1024 * sizeof(int16_t);
>> + break;
>> + default:
>> + BUG();
>> + }
>> +
>> + return p;
>> +}
>> +
>> +static void __iomem *pmecc_get_index_of(struct atmel_nand_host *host)
> this is a __dev_init function plese check the other too
>
> btw you need to use __dev_init and not __init
Here I use pmecc_get_index_of() function to get a lookup table base
which is in ROM.
I'm not clear about how should I need __dev_init here. Can you give more
information?
>> +{
>> + void __iomem *p;
>> +
>> + switch (host->pmecc_sector_size) {
>> + case 512:
>> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_512;
>> + break;
>> + case 1024:
>> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_1024;
>> + break;
>> + default:
>> + BUG();
>> + }
>> +
>> + return p;
>> +}
>> +
>> +static void pmecc_gen_syndrome(struct mtd_info *mtd, int sector)
>> +{
>> + int i;
>> + uint32_t value;
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> +
>> + /* Fill odd syndromes */
>> + for (i = 0; i< host->pmecc_corr_cap; i++) {
>> + value = pmecc_readl_rem_relaxed(host->ecc, sector, i / 2);
>> + value = (i& 1) ? (value& 0xffff0000)>> 16 : value& 0xffff;
> simplify by
> if (i& 1)
> val>>= 16;
> value&= 0xffff;
I'll fix it.
>
>> + host->pmecc_data->partial_syn[(2 * i) + 1] = (int16_t)value;
>> + }
>> +}
>> +
>> +static void pmecc_substitute(struct mtd_info *mtd)
>> +{
>> + int16_t *si;
>> + int i, j;
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> + int16_t __iomem *alpha_to = host->pmecc_alpha_to;
>> + int16_t __iomem *index_of = host->pmecc_index_of;
>> + int16_t *partial_syn = host->pmecc_data->partial_syn;
>> +
>> + /* si[] is a table that holds the current syndrome value,
>> + * an element of that table belongs to the field
>> + */
>> + si = host->pmecc_data->si;
>> +
>> + for (i = 1; i< 2 * PMECC_MAX_ERROR_NB; i++)
>> + si[i] = 0;
> please use memset
ok
>> +
>> + /* Computation 2t syndromes based on S(x) */
>> + /* Odd syndromes */
>> + for (i = 1; i< 2 * host->pmecc_corr_cap; i += 2) {
>> + si[i] = 0;
> shy this you already init the array at 0 before
yes, I'll fix it.
>> + for (j = 0; j< host->pmecc_degree; j++) {
>> + if (partial_syn[i]& ((unsigned short)0x1<< j))
>> + si[i] = readw_relaxed(alpha_to + i * j) ^ si[i];
>> + }
>> + }
>> + /* Even syndrome = (Odd syndrome) ** 2 */
>> + for (i = 2; i<= 2 * host->pmecc_corr_cap; i += 2) {
>> + j = i / 2;
>> + if (si[j] == 0)
> here if {
> } else {
> }
ok
>> + si[i] = 0;
>> + else {
>> + int16_t tmp;
> missing blank line
ok.
>> + tmp = readw_relaxed(index_of + si[j]);
>> + tmp = (tmp * 2) % host->pmecc_cw_len;
>> + si[i] = readw_relaxed(alpha_to + tmp);
>> + }
>> + }
>> +
>> + return;
>> +}
>> +
>> +static void pmecc_get_sigma(struct mtd_info *mtd)
>> +{
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> +
>> + int i, j, k;
>> + uint32_t dmu_0_count, tmp;
>> + int16_t (*smu)[2 * PMECC_MAX_ERROR_NB + 1];
>> + int16_t *lmu = host->pmecc_data->lmu;
>> + int16_t *si = host->pmecc_data->si;
>> + int *mu = host->pmecc_data->mu;
>> + int *dmu = host->pmecc_data->dmu; /* Discrepancy */
>> + int *delta = host->pmecc_data->delta; /* Delta order */
>> + int cw_len = host->pmecc_cw_len;
>> + int16_t cap = host->pmecc_corr_cap;
>> +
>> + int16_t __iomem *index_of = host->pmecc_index_of;
>> + int16_t __iomem *alpha_to = host->pmecc_alpha_to;
>> +
>> + /* index of largest delta */
>> + int ro;
>> + int largest;
>> + int diff;
>> +
>> + dmu_0_count = 0;
>> + smu = host->pmecc_data->smu;
>> +
>> + /* First Row */
>> +
>> + /* Mu */
>> + mu[0] = -1;
>> +
>> + memset(&smu[0][0], 0,
>> + sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
>> + smu[0][0] = 1;
>> +
>> + /* discrepancy set to 1 */
>> + dmu[0] = 1;
>> + /* polynom order set to 0 */
>> + lmu[0] = 0;
>> + delta[0] = (mu[0] * 2 - lmu[0])>> 1;
>> +
>> + /* Second Row */
>> +
>> + /* Mu */
>> + mu[1] = 0;
>> + /* Sigma(x) set to 1 */
>> + memset(&smu[1][0], 0,
>> + sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
>> + smu[1][0] = 1;
>> +
>> + /* discrepancy set to S1 */
>> + dmu[1] = si[1];
>> +
>> + /* polynom order set to 0 */
>> + lmu[1] = 0;
>> +
>> + delta[1] = (mu[1] * 2 - lmu[1])>> 1;
>> +
>> + /* Init the Sigma(x) last row */
>> + memset(&smu[cap + 1][0], 0,
>> + sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
>> +
>> + for (i = 1; i<= cap; i++) {
>> + mu[i+1] = i<< 1;
>> + /* Begin Computing Sigma (Mu+1) and L(mu) */
>> + /* check if discrepancy is set to 0 */
>> + if (dmu[i] == 0) {
>> + dmu_0_count++;
>> +
>> + tmp = ((cap - (lmu[i]>> 1) - 1) / 2);
>> + if ((cap - (lmu[i]>> 1) - 1)& 0x1)
>> + tmp += 2;
>> + else
>> + tmp += 1;
>> +
>> + if (dmu_0_count == tmp) {
>> + for (j = 0; j<= (lmu[i]>> 1) + 1; j++)
>> + smu[cap + 1][j] = smu[i][j];
>> + lmu[cap + 1] = lmu[i];
>> + return;
>> + }
>> +
>> + /* copy polynom */
>> + for (j = 0; j<= lmu[i]>> 1; j++)
>> + smu[i + 1][j] = smu[i][j];
>> +
>> + /* copy previous polynom order to the next */
>> + lmu[i + 1] = lmu[i];
>> + } else {
>> + ro = 0;
>> + largest = -1;
>> + /* find largest delta with dmu != 0 */
>> + for (j = 0; j< i; j++) {
>> + if ((dmu[j])&& (delta[j]> largest)) {
>> + largest = delta[j];
>> + ro = j;
>> + }
>> + }
>> +
>> + /* compute difference */
>> + diff = (mu[i] - mu[ro]);
>> +
>> + /* Compute degree of the new smu polynomial */
>> + if ((lmu[i]>> 1)> ((lmu[ro]>> 1) + diff))
>> + lmu[i + 1] = lmu[i];
>> + else
>> + lmu[i + 1] = ((lmu[ro]>> 1) + diff) * 2;
>> +
>> + /* Init smu[i+1] with 0 */
>> + for (k = 0; k< (2 * PMECC_MAX_ERROR_NB + 1); k++)
>> + smu[i+1][k] = 0;
>> +
>> + /* Compute smu[i+1] */
>> + for (k = 0; k<= lmu[ro]>> 1; k++) {
>> + int16_t a, b, c;
>> +
>> + if (!(smu[ro][k]&& dmu[i]))
>> + continue;
>> + a = readw_relaxed(index_of + dmu[i]);
>> + b = readw_relaxed(index_of + dmu[ro]);
>> + c = readw_relaxed(index_of + smu[ro][k]);
>> + tmp = a + (cw_len - b) + c;
>> + a = readw_relaxed(alpha_to + tmp % cw_len);
>> + smu[i + 1][k + diff] = a;
>> + }
>> +
>> + for (k = 0; k<= lmu[i]>> 1; k++)
>> + smu[i + 1][k] ^= smu[i][k];
>> + }
>> +
>> + /* End Computing Sigma (Mu+1) and L(mu) */
>> + /* In either case compute delta */
>> + delta[i + 1] = (mu[i + 1] * 2 - lmu[i + 1])>> 1;
>> +
>> + /* Do not compute discrepancy for the last iteration */
>> + if (i>= cap)
>> + continue;
>> +
>> + for (k = 0 ; k<= (lmu[i + 1]>> 1); k++) {
>> + tmp = 2 * (i - 1);
>> + if (k == 0)
>> + dmu[i + 1] = si[tmp + 3];
>> + else if (smu[i+1][k]&& si[tmp + 3 - k]) {
>> + int16_t a, b, c;
>> + a = readw_relaxed(index_of + smu[i + 1][k]);
>> + b = si[2 * (i - 1) + 3 - k];
>> + c = readw_relaxed(index_of + b);
>> + tmp = a + c;
>> + tmp %= cw_len;
>> + dmu[i + 1] = readw_relaxed(alpha_to + tmp) ^
>> + dmu[i + 1];
>> + }
>> + }
>> + }
>> +
>> + return;
>> +}
>> +
>> +static int pmecc_err_location(struct mtd_info *mtd)
>> +{
>> + int i;
>> + int err_nbr; /* number of error */
>> + int roots_nbr; /* number of roots */
>> + int sector_size;
>> + uint32_t val;
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> + int timeout_count = 0;
>> + int cap = host->pmecc_corr_cap;
>> +
>> + err_nbr = 0;
>> + sector_size = host->pmecc_sector_size;
>> +
>> + pmerrloc_writel(host->pmerrloc_base, ELDIS, PMERRLOC_DISABLE);
>> +
>> + for (i = 0; i<= host->pmecc_data->lmu[cap + 1]>> 1; i++) {
>> + pmerrloc_writel_sigma_relaxed(host->pmerrloc_base, i,
>> + host->pmecc_data->smu[cap + 1][i]);
>> + err_nbr++;
>> + }
>> +
>> + val = (err_nbr - 1)<< 16;
>> + if (sector_size == 1024)
>> + val |= 1;
>> +
>> + pmerrloc_writel(host->pmerrloc_base, ELCFG, val);
>> + pmerrloc_writel(host->pmerrloc_base, ELEN,
>> + sector_size * 8 + host->pmecc_degree * cap);
>> +
>> + while (!(pmerrloc_readl_relaxed(host->pmerrloc_base, ELISR)
>> + & PMERRLOC_CALC_DONE)) {
>> + if (unlikely(timeout_count++> PMECC_MAX_TIMEOUT_COUNT))
>> + return -1; /* Time out */
>> + cpu_relax();
>> + }
>> +
>> + roots_nbr = (pmerrloc_readl_relaxed(host->pmerrloc_base, ELISR)
>> + & PMERRLOC_ERR_NUM_MASK)>> 8;
>> + /* Number of roots == degree of smu hence<= cap */
>> + if (roots_nbr == host->pmecc_data->lmu[cap + 1]>> 1)
>> + return err_nbr - 1;
>> +
>> + /* Number of roots does not match the degree of smu
>> + * unable to correct error */
>> + return -1;
>> +}
>> +
>> +static void pmecc_correct_data(struct mtd_info *mtd, uint8_t *buf,
>> + int extra_bytes, int err_nbr)
>> +{
>> + int i = 0;
>> + int byte_pos, bit_pos;
>> + int sector_size, ecc_size;
>> + uint32_t tmp;
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> +
>> + sector_size = host->pmecc_sector_size;
>> + ecc_size = nand_chip->ecc.bytes;
>> +
>> + while (err_nbr) {
>> + tmp = pmerrloc_readl_el_relaxed(host->pmerrloc_base, i) - 1;
>> + byte_pos = tmp / 8;
>> + bit_pos = tmp % 8;
>> + dev_info(host->dev, "PMECC correction, byte_pos: %d bit_pos: %d\n",
>> + byte_pos, bit_pos);
>> +
>> + if (byte_pos< (sector_size + extra_bytes)) {
>> + tmp = sector_size +
>> + pmecc_readl_relaxed(host->ecc, SADDR);
>> +
>> + if (byte_pos< tmp)
>> + *(buf + byte_pos) ^= (1<< bit_pos);
>> + else
>> + *(buf + byte_pos + ecc_size) ^= (1<< bit_pos);
>> + }
>> +
>> + i++;
>> + err_nbr--;
>> + }
>> +
>> + return;
>> +}
>> +
>> +static int pmecc_correction(struct mtd_info *mtd, u32 pmecc_stat, uint8_t *buf,
>> + u8 *ecc)
>> +{
>> + int i, err_nbr;
>> + uint8_t *buf_pos;
>> + int eccbytes;
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> +
>> + eccbytes = nand_chip->ecc.bytes;
>> + for (i = 0; i< eccbytes; i++)
>> + if (ecc[i] != 0xff)
>> + goto normal_check;
>> + /* Erased page, return OK */
>> + return 0;
>> +
>> +normal_check:
>> + for (i = 0; i< host->pmecc_sector_number; i++) {
>> + err_nbr = 0;
>> + if (pmecc_stat& 0x1) {
>> + buf_pos = buf + i * host->pmecc_sector_size;
>> +
>> + pmecc_gen_syndrome(mtd, i);
>> + pmecc_substitute(mtd);
>> + pmecc_get_sigma(mtd);
>> +
>> + err_nbr = pmecc_err_location(mtd);
>> + if (err_nbr == -1) {
>> + dev_err(host->dev, "PMECC: Too many errors\n");
>> + mtd->ecc_stats.failed++;
>> + return -EIO;
>> + } else {
>> + pmecc_correct_data(mtd, buf_pos, 0, err_nbr);
>> + mtd->ecc_stats.corrected += err_nbr;
>> + }
>> + }
>> + pmecc_stat>>= 1;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +static int atmel_nand_pmecc_read_page(struct mtd_info *mtd,
>> + struct nand_chip *chip, uint8_t *buf, int page)
>> +{
>> + uint32_t stat;
>> + int timeout_count = 0;
>> + int eccsize = chip->ecc.size;
>> + uint8_t *oob = chip->oob_poi;
>> + struct atmel_nand_host *host = chip->priv;
>> + uint32_t *eccpos = chip->ecc.layout->eccpos;
>> +
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
>> + pmecc_writel(host->ecc, CFG, (pmecc_readl_relaxed(host->ecc, CFG)
>> + & ~PMECC_CFG_WRITE_OP) | PMECC_CFG_AUTO_ENABLE);
>> +
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DATA);
>> +
>> + chip->read_buf(mtd, buf, eccsize);
>> + chip->read_buf(mtd, oob, mtd->oobsize);
>> +
>> + while ((pmecc_readl_relaxed(host->ecc, SR)& PMECC_SR_BUSY)) {
>> + if (unlikely(timeout_count++> PMECC_MAX_TIMEOUT_COUNT))
>> + return -EIO; /* Time out */
>> + cpu_relax();
>> + }
>> +
>> + stat = pmecc_readl_relaxed(host->ecc, ISR);
>> + if (stat != 0) {
>> + if (pmecc_correction(mtd, stat, buf,&oob[eccpos[0]]) != 0)
>> + return -EIO;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +static void atmel_nand_pmecc_write_page(struct mtd_info *mtd,
>> + struct nand_chip *chip, const uint8_t *buf)
>> +{
>> + int i, j;
>> + int timeout_count = 0;
>> + struct atmel_nand_host *host = chip->priv;
>> + uint32_t *eccpos = chip->ecc.layout->eccpos;
>> +
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
>> +
>> + pmecc_writel(host->ecc, CFG, (pmecc_readl_relaxed(host->ecc, CFG) |
>> + PMECC_CFG_WRITE_OP)& ~PMECC_CFG_AUTO_ENABLE);
>> +
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DATA);
>> +
>> + chip->write_buf(mtd, (u8 *)buf, mtd->writesize);
>> +
>> + while ((pmecc_readl_relaxed(host->ecc, SR)& PMECC_SR_BUSY)) {
>> + if (unlikely(timeout_count++> PMECC_MAX_TIMEOUT_COUNT)) {
>> + dev_err(host->dev, "PMECC: Timeout to get ECC value.\n");
>> + return; /* Time out */
>> + }
>> + cpu_relax();
>> + }
>> +
>> + for (i = 0; i< host->pmecc_sector_number; i++) {
>> + for (j = 0; j< host->pmecc_bytes_per_sector; j++) {
>> + int pos;
>> +
>> + pos = i * host->pmecc_bytes_per_sector + j;
>> + chip->oob_poi[eccpos[pos]] =
>> + pmecc_readb_ecc_relaxed(host->ecc, i, j);
>> + }
>> + }
>> + chip->write_buf(mtd, chip->oob_poi, mtd->oobsize);
>> +
>> + return;
>> +}
>> +
>> +static void atmel_pmecc_core_init(struct mtd_info *mtd)
>> +{
>> + uint32_t val = 0;
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> + struct nand_ecclayout *ecc_layout;
>> +
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
>> +
>> + switch (host->pmecc_corr_cap) {
>> + case 2:
>> + val = PMECC_CFG_BCH_ERR2;
>> + break;
>> + case 4:
>> + val = PMECC_CFG_BCH_ERR4;
>> + break;
>> + case 8:
>> + val = PMECC_CFG_BCH_ERR8;
>> + break;
>> + case 12:
>> + val = PMECC_CFG_BCH_ERR12;
>> + break;
>> + case 24:
>> + val = PMECC_CFG_BCH_ERR24;
>> + break;
>> + }
>> +
>> + if (host->pmecc_sector_size == 512)
>> + val |= PMECC_CFG_SECTOR512;
>> + else if (host->pmecc_sector_size == 1024)
>> + val |= PMECC_CFG_SECTOR1024;
>> +
>> + switch (host->pmecc_sector_number) {
>> + case 1:
>> + val |= PMECC_CFG_PAGE_1SECTOR;
>> + break;
>> + case 2:
>> + val |= PMECC_CFG_PAGE_2SECTORS;
>> + break;
>> + case 4:
>> + val |= PMECC_CFG_PAGE_4SECTORS;
>> + break;
>> + case 8:
>> + val |= PMECC_CFG_PAGE_8SECTORS;
>> + break;
>> + }
>> +
>> + val |= (PMECC_CFG_READ_OP | PMECC_CFG_SPARE_DISABLE
>> + | PMECC_CFG_AUTO_DISABLE);
>> + pmecc_writel(host->ecc, CFG, val);
>> +
>> + ecc_layout = nand_chip->ecc.layout;
>> + pmecc_writel(host->ecc, SAREA, mtd->oobsize - 1);
>> + pmecc_writel(host->ecc, SADDR, ecc_layout->eccpos[0]);
>> + pmecc_writel(host->ecc, EADDR,
>> + ecc_layout->eccpos[ecc_layout->eccbytes - 1]);
>> + /* See datasheet about PMECC Clock Control Register */
>> + pmecc_writel(host->ecc, CLK, 2);
>> + pmecc_writel(host->ecc, IDR, 0xff);
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
>> +}
>> +
>> +static int __init atmel_pmecc_nand_init_params(struct platform_device *pdev,
>> + struct atmel_nand_host *host)
>> +{
>> + int cap, sector_size, err_no;
>> + struct mtd_info *mtd;
>> + struct nand_chip *nand_chip;
>> + struct resource *regs;
>> + struct resource *regs_pmerr, *regs_rom;
>> +
>> + cap = host->pmecc_corr_cap;
>> + sector_size = host->pmecc_sector_size;
>> + dev_info(host->dev, "Initialize PMECC params, cap: %d, sector: %d\n",
>> + cap, sector_size);
>> +
>> + /* Sanity check */
>> + if ((sector_size != 512)&& (sector_size != 1024)) {
>> + dev_err(host->dev,
>> + "Unsupported PMECC sector size: %d; should be 512 or 1024 bytes\n",
>> + sector_size);
>> + return -EINVAL;
>> + }
>> + if ((cap != 2)&& (cap != 4)&& (cap != 8)&& (cap != 12)&&
>> + (cap != 24)) {
>> + dev_err(host->dev,
>> + "Unsupported PMECC correction capability, should be 2, 4, 8, 12 or 24\n");
>> + return -EINVAL;
>> + }
>> +
>> + nand_chip =&host->nand_chip;
>> + mtd =&host->mtd;
>> +
>> + nand_chip->ecc.mode = NAND_ECC_SOFT; /* By default */
>> +
>> + regs = platform_get_resource(pdev, IORESOURCE_MEM, 1);
>> + if (!regs) {
>> + dev_warn(host->dev,
>> + "Can't get I/O resource regs, rolling back on software ECC\n");
>> + return 0;
>> + }
>> +
>> + host->ecc = ioremap(regs->start, resource_size(regs));
>> + if (host->ecc == NULL) {
>> + dev_err(host->dev, "ioremap failed\n");
>> + err_no = -EIO;
>> + goto err_pmecc_ioremap;
>> + }
>> +
>> + regs_pmerr = platform_get_resource(pdev, IORESOURCE_MEM, 2);
>> + regs_rom = platform_get_resource(pdev, IORESOURCE_MEM, 3);
>> + if (regs_pmerr&& regs_rom) {
>> + host->pmerrloc_base = ioremap(regs_pmerr->start,
>> + resource_size(regs_pmerr));
>> + host->pmecc_rom_base = ioremap(regs_rom->start,
>> + resource_size(regs_rom));
>> +
>> + if (host->pmerrloc_base&& host->pmecc_rom_base) {
>> + nand_chip->ecc.mode = NAND_ECC_HW;
>> + nand_chip->ecc.read_page =
>> + atmel_nand_pmecc_read_page;
>> + nand_chip->ecc.write_page =
>> + atmel_nand_pmecc_write_page;
>> + } else {
>> + dev_err(host->dev,
>> + "Can not get I/O resource for PMECC controller!\n");
>> + err_no = -EIO;
>> + goto err_pmloc_ioremap;
>> + }
>> + }
>> +
>> + /* ECC is calculated for the whole page (1 step) */
>> + nand_chip->ecc.size = mtd->writesize;
>> +
>> + /* set ECC page size and oob layout */
>> + switch (mtd->writesize) {
>> + case 2048:
>> + host->pmecc_degree = PMECC_GF_DIMENSION_13;
>> + host->pmecc_cw_len = (1<< host->pmecc_degree) - 1;
>> + host->pmecc_corr_cap = cap;
>> + host->pmecc_sector_number = mtd->writesize / sector_size;
>> + host->pmecc_bytes_per_sector = pmecc_get_ecc_bytes(
>> + cap, sector_size);
>> + host->pmecc_alpha_to = pmecc_get_alpha_to(host);
>> + host->pmecc_index_of = pmecc_get_index_of(host);
>> +
>> + nand_chip->ecc.steps = 1;
>> + nand_chip->ecc.strength = cap;
>> + nand_chip->ecc.bytes = host->pmecc_bytes_per_sector *
>> + host->pmecc_sector_number;
>> + if (nand_chip->ecc.bytes> mtd->oobsize - 2) {
>> + dev_err(host->dev, "No room for ECC bytes\n");
>> + err_no = -EINVAL;
>> + goto err;
>> + }
>> + pmecc_config_ecc_layout(&atmel_pmecc_oobinfo,
>> + mtd->oobsize,
>> + nand_chip->ecc.bytes);
>> + nand_chip->ecc.layout =&atmel_pmecc_oobinfo;
>> + break;
>> + case 512:
>> + case 1024:
>> + case 4096:
>> + /* TODO */
>> + dev_warn(host->dev,
>> + "Unsupported page size for PMECC, use Software ECC\n");
>> + default:
>> + /* page size not handled by HW ECC */
>> + /* switching back to soft ECC */
>> + nand_chip->ecc.mode = NAND_ECC_SOFT;
>> + nand_chip->ecc.calculate = NULL;
>> + nand_chip->ecc.correct = NULL;
>> + nand_chip->ecc.hwctl = NULL;
>> + nand_chip->ecc.read_page = NULL;
>> + nand_chip->ecc.write_page = NULL;
>> + nand_chip->ecc.postpad = 0;
>> + nand_chip->ecc.prepad = 0;
>> + nand_chip->ecc.bytes = 0;
>> + err_no = 0;
>> + goto err;
>> + }
>> +
>> + /* Allocate data for PMECC computation */
>> + host->pmecc_data = kzalloc(sizeof(struct atmel_pmecc_data), GFP_KERNEL);
> why do you always allocate the pmecc_data?
>
> you need to allocate it only if you use it
If not using PMECC, the code will never come to here. So I think it's
right time to allocate all the computation data here.
>
> Best Regards,
> J.
>> + if (!host->pmecc_data) {
>> + dev_err(host->dev,
>> + "Cannot allocate memory for PMECC computation!\n");
>> + err_no = -ENOMEM;
>> + goto err;
>> + }
>> +
>> + atmel_pmecc_core_init(mtd);
>> +
>> + return 0;
>> +
>> +err:
>> +err_pmloc_ioremap:
>> + iounmap(host->ecc);
>> + if (host->pmerrloc_base)
>> + iounmap(host->pmerrloc_base);
>> + if (host->pmecc_rom_base)
>> + iounmap(host->pmecc_rom_base);
>> +err_pmecc_ioremap:
>> + return err_no;
>> +}
>> +
>> +/*
>> * Calculate HW ECC
>> *
>> * function called after a write
>> @@ -720,7 +1456,11 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
>> }
>>
>> if (nand_chip->ecc.mode == NAND_ECC_HW) {
>> - res = atmel_hw_nand_init_params(pdev, host);
>> + if (host->has_pmecc)
>> + res = atmel_pmecc_nand_init_params(pdev, host);
>> + else
>> + res = atmel_hw_nand_init_params(pdev, host);
>> +
>> if (res != 0)
>> goto err_hw_ecc;
>> }
>> @@ -741,6 +1481,12 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
>> err_scan_tail:
>> if (host->ecc)
>> iounmap(host->ecc);
>> + if (host->has_pmecc) {
> no need to check if you have teh pmecc
>
> if it's no the case pmerrloc_base will be NULLo
>
> don't you need to disable it in the error path?
right, I should disable the hardware in this error path.
>
> Best Regards,
> J.
Best Regards,
Josh Wu
^ permalink raw reply [flat|nested] 24+ messages in thread
* [PATCH v9 3/3] MTD: at91: atmel_nand: Update driver to support Programmable Multibit ECC controller
@ 2012-05-28 8:34 ` Josh Wu
0 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-28 8:34 UTC (permalink / raw)
To: linux-arm-kernel
On 5/28/2012 2:58 PM, Jean-Christophe PLAGNIOL-VILLARD wrote:
> On 21:24 Sat 26 May , Josh Wu wrote:
>> The Programmable Multibit ECC (PMECC) controller is a programmable binary
>> BCH(Bose, Chaudhuri and Hocquenghem) encoder and decoder. This controller
>> can be used to support both SLC and MLC NAND Flash devices. It supports to
>> generate ECC to correct 2, 4, 8, 12 or 24 bits of error per sector of data.
>>
>> To use this driver, the user needs to pass in the correction capability and
>> the sector size.
>>
>> This driver has been tested on AT91SAM9X5-EK and AT91SAM9N12-EK with JFFS2,
>> YAFFS2, UBIFS and mtd-utils.
>>
>> Signed-off-by: Hong Xu<hong.xu@atmel.com>
>> Signed-off-by: Josh Wu<josh.wu@atmel.com>
>> ---
>> drivers/mtd/nand/atmel_nand.c | 761 ++++++++++++++++++++++++++++++++++++-
>> drivers/mtd/nand/atmel_nand_ecc.h | 116 ++++++
>> 2 files changed, 876 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/mtd/nand/atmel_nand.c b/drivers/mtd/nand/atmel_nand.c
>> index 9a9bfbf..ddcf1ed 100644
>> --- a/drivers/mtd/nand/atmel_nand.c
>> +++ b/drivers/mtd/nand/atmel_nand.c
>> @@ -15,6 +15,8 @@
>> * (u-boot-1.1.5/board/atmel/at91sam9263ek/nand.c)
>> * (C) Copyright 2006 ATMEL Rousset, Lacressonniere Nicolas
>> *
>> + * Add Programmable Multibit ECC support for various AT91 SoC
>> + * (C) Copyright 2012 ATMEL, Hong Xu
>> *
>> * This program is free software; you can redistribute it and/or modify
>> * it under the terms of the GNU General Public License version 2 as
>> @@ -77,6 +79,21 @@ static struct nand_ecclayout atmel_oobinfo_small = {
>> },
>> };
>>
>> +/* a structure includes datas for PMECC computation */
>> +struct atmel_pmecc_data {
>> + int16_t partial_syn[2 * PMECC_MAX_ERROR_NB + 1];
>> + int16_t si[2 * PMECC_MAX_ERROR_NB + 1];
>> +
>> + /* Sigma table */
>> + int16_t smu[PMECC_MAX_ERROR_NB + 2][2 * PMECC_MAX_ERROR_NB + 1];
> you still hardcode the array in the struct
>
> and if the pmecc evolve we will have to touch again
> please allocate them
ok.
>> + /* polynomal order */
>> + int16_t lmu[PMECC_MAX_ERROR_NB + 1];
>> +
>> + int mu[PMECC_MAX_ERROR_NB + 1];
>> + int dmu[PMECC_MAX_ERROR_NB + 1];
>> + int delta[PMECC_MAX_ERROR_NB + 1];
>> +};
>> +
>> struct atmel_nand_host {
>> struct nand_chip nand_chip;
>> struct mtd_info mtd;
>> @@ -92,8 +109,25 @@ struct atmel_nand_host {
>> bool has_pmecc;
>> u8 pmecc_corr_cap;
>> u16 pmecc_sector_size;
>> +
>> + int pmecc_bytes_per_sector;
>> + int pmecc_sector_number;
>> + int pmecc_degree; /* Degree of remainders */
>> + int pmecc_cw_len; /* Length of codeword */
>> +
>> + void __iomem *pmerrloc_base;
>> + void __iomem *pmecc_rom_base;
>> +
>> + /* lookup table for alpha_to and index_of */
>> + void __iomem *pmecc_alpha_to;
>> + void __iomem *pmecc_index_of;
>> +
>> + /* data for pmecc computation */
>> + struct atmel_pmecc_data *pmecc_data;
>> };
>>
>> +static struct nand_ecclayout atmel_pmecc_oobinfo;
>> +
>> static int cpu_has_dma(void)
>> {
>> return cpu_is_at91sam9rl() || cpu_is_at91sam9g45();
>> @@ -287,6 +321,708 @@ static void atmel_write_buf(struct mtd_info *mtd, const u8 *buf, int len)
>> }
>>
>> /*
>> + * Return number of ecc bytes per sector according to sector size and
>> + * correction capability
>> + *
>> + * Following table shows what at91 PMECC supported:
>> + * Correction Capability Sector_512_bytes Sector_1024_bytes
>> + * ===================== ================ =================
>> + * 2-bits 4-bytes 4-bytes
>> + * 4-bits 7-bytes 7-bytes
>> + * 8-bits 13-bytes 14-bytes
>> + * 12-bits 20-bytes 21-bytes
>> + * 24-bits 39-bytes 42-bytes
>> + */
>> +static int pmecc_get_ecc_bytes(int cap, int sector_size)
>> +{
>> + int m = 12 + sector_size / 512;
>> + return (m * cap + 7) / 8;
>> +}
>> +
>> +static void pmecc_config_ecc_layout(struct nand_ecclayout *layout, int oobsize,
>> + int ecc_len)
>> +{
>> + int i;
>> +
>> + layout->eccbytes = ecc_len;
>> +
>> + /* ECC will occupy the last ecc_len bytes continuously */
>> + for (i = 0; i< ecc_len; i++)
>> + layout->eccpos[i] = oobsize - ecc_len + i;
>> +
>> + layout->oobfree[0].offset = 2;
>> + layout->oobfree[0].length =
>> + oobsize - ecc_len - layout->oobfree[0].offset;
>> +}
>> +
>> +static void __iomem *pmecc_get_alpha_to(struct atmel_nand_host *host)
>> +{
>> + void __iomem *p;
>> +
>> + switch (host->pmecc_sector_size) {
>> + case 512:
>> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_512 +
>> + PMECC_LOOKUP_TABLE_SIZE_512 * sizeof(int16_t);
>> + break;
>> + case 1024:
>> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_1024 +
>> + PMECC_LOOKUP_TABLE_SIZE_1024 * sizeof(int16_t);
>> + break;
>> + default:
>> + BUG();
>> + }
>> +
>> + return p;
>> +}
>> +
>> +static void __iomem *pmecc_get_index_of(struct atmel_nand_host *host)
> this is a __dev_init function plese check the other too
>
> btw you need to use __dev_init and not __init
Here I use pmecc_get_index_of() function to get a lookup table base
which is in ROM.
I'm not clear about how should I need __dev_init here. Can you give more
information?
>> +{
>> + void __iomem *p;
>> +
>> + switch (host->pmecc_sector_size) {
>> + case 512:
>> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_512;
>> + break;
>> + case 1024:
>> + p = host->pmecc_rom_base + PMECC_LOOKUP_TABLE_OFFSET_1024;
>> + break;
>> + default:
>> + BUG();
>> + }
>> +
>> + return p;
>> +}
>> +
>> +static void pmecc_gen_syndrome(struct mtd_info *mtd, int sector)
>> +{
>> + int i;
>> + uint32_t value;
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> +
>> + /* Fill odd syndromes */
>> + for (i = 0; i< host->pmecc_corr_cap; i++) {
>> + value = pmecc_readl_rem_relaxed(host->ecc, sector, i / 2);
>> + value = (i& 1) ? (value& 0xffff0000)>> 16 : value& 0xffff;
> simplify by
> if (i& 1)
> val>>= 16;
> value&= 0xffff;
I'll fix it.
>
>> + host->pmecc_data->partial_syn[(2 * i) + 1] = (int16_t)value;
>> + }
>> +}
>> +
>> +static void pmecc_substitute(struct mtd_info *mtd)
>> +{
>> + int16_t *si;
>> + int i, j;
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> + int16_t __iomem *alpha_to = host->pmecc_alpha_to;
>> + int16_t __iomem *index_of = host->pmecc_index_of;
>> + int16_t *partial_syn = host->pmecc_data->partial_syn;
>> +
>> + /* si[] is a table that holds the current syndrome value,
>> + * an element of that table belongs to the field
>> + */
>> + si = host->pmecc_data->si;
>> +
>> + for (i = 1; i< 2 * PMECC_MAX_ERROR_NB; i++)
>> + si[i] = 0;
> please use memset
ok
>> +
>> + /* Computation 2t syndromes based on S(x) */
>> + /* Odd syndromes */
>> + for (i = 1; i< 2 * host->pmecc_corr_cap; i += 2) {
>> + si[i] = 0;
> shy this you already init the array at 0 before
yes, I'll fix it.
>> + for (j = 0; j< host->pmecc_degree; j++) {
>> + if (partial_syn[i]& ((unsigned short)0x1<< j))
>> + si[i] = readw_relaxed(alpha_to + i * j) ^ si[i];
>> + }
>> + }
>> + /* Even syndrome = (Odd syndrome) ** 2 */
>> + for (i = 2; i<= 2 * host->pmecc_corr_cap; i += 2) {
>> + j = i / 2;
>> + if (si[j] == 0)
> here if {
> } else {
> }
ok
>> + si[i] = 0;
>> + else {
>> + int16_t tmp;
> missing blank line
ok.
>> + tmp = readw_relaxed(index_of + si[j]);
>> + tmp = (tmp * 2) % host->pmecc_cw_len;
>> + si[i] = readw_relaxed(alpha_to + tmp);
>> + }
>> + }
>> +
>> + return;
>> +}
>> +
>> +static void pmecc_get_sigma(struct mtd_info *mtd)
>> +{
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> +
>> + int i, j, k;
>> + uint32_t dmu_0_count, tmp;
>> + int16_t (*smu)[2 * PMECC_MAX_ERROR_NB + 1];
>> + int16_t *lmu = host->pmecc_data->lmu;
>> + int16_t *si = host->pmecc_data->si;
>> + int *mu = host->pmecc_data->mu;
>> + int *dmu = host->pmecc_data->dmu; /* Discrepancy */
>> + int *delta = host->pmecc_data->delta; /* Delta order */
>> + int cw_len = host->pmecc_cw_len;
>> + int16_t cap = host->pmecc_corr_cap;
>> +
>> + int16_t __iomem *index_of = host->pmecc_index_of;
>> + int16_t __iomem *alpha_to = host->pmecc_alpha_to;
>> +
>> + /* index of largest delta */
>> + int ro;
>> + int largest;
>> + int diff;
>> +
>> + dmu_0_count = 0;
>> + smu = host->pmecc_data->smu;
>> +
>> + /* First Row */
>> +
>> + /* Mu */
>> + mu[0] = -1;
>> +
>> + memset(&smu[0][0], 0,
>> + sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
>> + smu[0][0] = 1;
>> +
>> + /* discrepancy set to 1 */
>> + dmu[0] = 1;
>> + /* polynom order set to 0 */
>> + lmu[0] = 0;
>> + delta[0] = (mu[0] * 2 - lmu[0])>> 1;
>> +
>> + /* Second Row */
>> +
>> + /* Mu */
>> + mu[1] = 0;
>> + /* Sigma(x) set to 1 */
>> + memset(&smu[1][0], 0,
>> + sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
>> + smu[1][0] = 1;
>> +
>> + /* discrepancy set to S1 */
>> + dmu[1] = si[1];
>> +
>> + /* polynom order set to 0 */
>> + lmu[1] = 0;
>> +
>> + delta[1] = (mu[1] * 2 - lmu[1])>> 1;
>> +
>> + /* Init the Sigma(x) last row */
>> + memset(&smu[cap + 1][0], 0,
>> + sizeof(int16_t) * (2 * PMECC_MAX_ERROR_NB + 1));
>> +
>> + for (i = 1; i<= cap; i++) {
>> + mu[i+1] = i<< 1;
>> + /* Begin Computing Sigma (Mu+1) and L(mu) */
>> + /* check if discrepancy is set to 0 */
>> + if (dmu[i] == 0) {
>> + dmu_0_count++;
>> +
>> + tmp = ((cap - (lmu[i]>> 1) - 1) / 2);
>> + if ((cap - (lmu[i]>> 1) - 1)& 0x1)
>> + tmp += 2;
>> + else
>> + tmp += 1;
>> +
>> + if (dmu_0_count == tmp) {
>> + for (j = 0; j<= (lmu[i]>> 1) + 1; j++)
>> + smu[cap + 1][j] = smu[i][j];
>> + lmu[cap + 1] = lmu[i];
>> + return;
>> + }
>> +
>> + /* copy polynom */
>> + for (j = 0; j<= lmu[i]>> 1; j++)
>> + smu[i + 1][j] = smu[i][j];
>> +
>> + /* copy previous polynom order to the next */
>> + lmu[i + 1] = lmu[i];
>> + } else {
>> + ro = 0;
>> + largest = -1;
>> + /* find largest delta with dmu != 0 */
>> + for (j = 0; j< i; j++) {
>> + if ((dmu[j])&& (delta[j]> largest)) {
>> + largest = delta[j];
>> + ro = j;
>> + }
>> + }
>> +
>> + /* compute difference */
>> + diff = (mu[i] - mu[ro]);
>> +
>> + /* Compute degree of the new smu polynomial */
>> + if ((lmu[i]>> 1)> ((lmu[ro]>> 1) + diff))
>> + lmu[i + 1] = lmu[i];
>> + else
>> + lmu[i + 1] = ((lmu[ro]>> 1) + diff) * 2;
>> +
>> + /* Init smu[i+1] with 0 */
>> + for (k = 0; k< (2 * PMECC_MAX_ERROR_NB + 1); k++)
>> + smu[i+1][k] = 0;
>> +
>> + /* Compute smu[i+1] */
>> + for (k = 0; k<= lmu[ro]>> 1; k++) {
>> + int16_t a, b, c;
>> +
>> + if (!(smu[ro][k]&& dmu[i]))
>> + continue;
>> + a = readw_relaxed(index_of + dmu[i]);
>> + b = readw_relaxed(index_of + dmu[ro]);
>> + c = readw_relaxed(index_of + smu[ro][k]);
>> + tmp = a + (cw_len - b) + c;
>> + a = readw_relaxed(alpha_to + tmp % cw_len);
>> + smu[i + 1][k + diff] = a;
>> + }
>> +
>> + for (k = 0; k<= lmu[i]>> 1; k++)
>> + smu[i + 1][k] ^= smu[i][k];
>> + }
>> +
>> + /* End Computing Sigma (Mu+1) and L(mu) */
>> + /* In either case compute delta */
>> + delta[i + 1] = (mu[i + 1] * 2 - lmu[i + 1])>> 1;
>> +
>> + /* Do not compute discrepancy for the last iteration */
>> + if (i>= cap)
>> + continue;
>> +
>> + for (k = 0 ; k<= (lmu[i + 1]>> 1); k++) {
>> + tmp = 2 * (i - 1);
>> + if (k == 0)
>> + dmu[i + 1] = si[tmp + 3];
>> + else if (smu[i+1][k]&& si[tmp + 3 - k]) {
>> + int16_t a, b, c;
>> + a = readw_relaxed(index_of + smu[i + 1][k]);
>> + b = si[2 * (i - 1) + 3 - k];
>> + c = readw_relaxed(index_of + b);
>> + tmp = a + c;
>> + tmp %= cw_len;
>> + dmu[i + 1] = readw_relaxed(alpha_to + tmp) ^
>> + dmu[i + 1];
>> + }
>> + }
>> + }
>> +
>> + return;
>> +}
>> +
>> +static int pmecc_err_location(struct mtd_info *mtd)
>> +{
>> + int i;
>> + int err_nbr; /* number of error */
>> + int roots_nbr; /* number of roots */
>> + int sector_size;
>> + uint32_t val;
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> + int timeout_count = 0;
>> + int cap = host->pmecc_corr_cap;
>> +
>> + err_nbr = 0;
>> + sector_size = host->pmecc_sector_size;
>> +
>> + pmerrloc_writel(host->pmerrloc_base, ELDIS, PMERRLOC_DISABLE);
>> +
>> + for (i = 0; i<= host->pmecc_data->lmu[cap + 1]>> 1; i++) {
>> + pmerrloc_writel_sigma_relaxed(host->pmerrloc_base, i,
>> + host->pmecc_data->smu[cap + 1][i]);
>> + err_nbr++;
>> + }
>> +
>> + val = (err_nbr - 1)<< 16;
>> + if (sector_size == 1024)
>> + val |= 1;
>> +
>> + pmerrloc_writel(host->pmerrloc_base, ELCFG, val);
>> + pmerrloc_writel(host->pmerrloc_base, ELEN,
>> + sector_size * 8 + host->pmecc_degree * cap);
>> +
>> + while (!(pmerrloc_readl_relaxed(host->pmerrloc_base, ELISR)
>> + & PMERRLOC_CALC_DONE)) {
>> + if (unlikely(timeout_count++> PMECC_MAX_TIMEOUT_COUNT))
>> + return -1; /* Time out */
>> + cpu_relax();
>> + }
>> +
>> + roots_nbr = (pmerrloc_readl_relaxed(host->pmerrloc_base, ELISR)
>> + & PMERRLOC_ERR_NUM_MASK)>> 8;
>> + /* Number of roots == degree of smu hence<= cap */
>> + if (roots_nbr == host->pmecc_data->lmu[cap + 1]>> 1)
>> + return err_nbr - 1;
>> +
>> + /* Number of roots does not match the degree of smu
>> + * unable to correct error */
>> + return -1;
>> +}
>> +
>> +static void pmecc_correct_data(struct mtd_info *mtd, uint8_t *buf,
>> + int extra_bytes, int err_nbr)
>> +{
>> + int i = 0;
>> + int byte_pos, bit_pos;
>> + int sector_size, ecc_size;
>> + uint32_t tmp;
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> +
>> + sector_size = host->pmecc_sector_size;
>> + ecc_size = nand_chip->ecc.bytes;
>> +
>> + while (err_nbr) {
>> + tmp = pmerrloc_readl_el_relaxed(host->pmerrloc_base, i) - 1;
>> + byte_pos = tmp / 8;
>> + bit_pos = tmp % 8;
>> + dev_info(host->dev, "PMECC correction, byte_pos: %d bit_pos: %d\n",
>> + byte_pos, bit_pos);
>> +
>> + if (byte_pos< (sector_size + extra_bytes)) {
>> + tmp = sector_size +
>> + pmecc_readl_relaxed(host->ecc, SADDR);
>> +
>> + if (byte_pos< tmp)
>> + *(buf + byte_pos) ^= (1<< bit_pos);
>> + else
>> + *(buf + byte_pos + ecc_size) ^= (1<< bit_pos);
>> + }
>> +
>> + i++;
>> + err_nbr--;
>> + }
>> +
>> + return;
>> +}
>> +
>> +static int pmecc_correction(struct mtd_info *mtd, u32 pmecc_stat, uint8_t *buf,
>> + u8 *ecc)
>> +{
>> + int i, err_nbr;
>> + uint8_t *buf_pos;
>> + int eccbytes;
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> +
>> + eccbytes = nand_chip->ecc.bytes;
>> + for (i = 0; i< eccbytes; i++)
>> + if (ecc[i] != 0xff)
>> + goto normal_check;
>> + /* Erased page, return OK */
>> + return 0;
>> +
>> +normal_check:
>> + for (i = 0; i< host->pmecc_sector_number; i++) {
>> + err_nbr = 0;
>> + if (pmecc_stat& 0x1) {
>> + buf_pos = buf + i * host->pmecc_sector_size;
>> +
>> + pmecc_gen_syndrome(mtd, i);
>> + pmecc_substitute(mtd);
>> + pmecc_get_sigma(mtd);
>> +
>> + err_nbr = pmecc_err_location(mtd);
>> + if (err_nbr == -1) {
>> + dev_err(host->dev, "PMECC: Too many errors\n");
>> + mtd->ecc_stats.failed++;
>> + return -EIO;
>> + } else {
>> + pmecc_correct_data(mtd, buf_pos, 0, err_nbr);
>> + mtd->ecc_stats.corrected += err_nbr;
>> + }
>> + }
>> + pmecc_stat>>= 1;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +static int atmel_nand_pmecc_read_page(struct mtd_info *mtd,
>> + struct nand_chip *chip, uint8_t *buf, int page)
>> +{
>> + uint32_t stat;
>> + int timeout_count = 0;
>> + int eccsize = chip->ecc.size;
>> + uint8_t *oob = chip->oob_poi;
>> + struct atmel_nand_host *host = chip->priv;
>> + uint32_t *eccpos = chip->ecc.layout->eccpos;
>> +
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
>> + pmecc_writel(host->ecc, CFG, (pmecc_readl_relaxed(host->ecc, CFG)
>> + & ~PMECC_CFG_WRITE_OP) | PMECC_CFG_AUTO_ENABLE);
>> +
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DATA);
>> +
>> + chip->read_buf(mtd, buf, eccsize);
>> + chip->read_buf(mtd, oob, mtd->oobsize);
>> +
>> + while ((pmecc_readl_relaxed(host->ecc, SR)& PMECC_SR_BUSY)) {
>> + if (unlikely(timeout_count++> PMECC_MAX_TIMEOUT_COUNT))
>> + return -EIO; /* Time out */
>> + cpu_relax();
>> + }
>> +
>> + stat = pmecc_readl_relaxed(host->ecc, ISR);
>> + if (stat != 0) {
>> + if (pmecc_correction(mtd, stat, buf,&oob[eccpos[0]]) != 0)
>> + return -EIO;
>> + }
>> +
>> + return 0;
>> +}
>> +
>> +static void atmel_nand_pmecc_write_page(struct mtd_info *mtd,
>> + struct nand_chip *chip, const uint8_t *buf)
>> +{
>> + int i, j;
>> + int timeout_count = 0;
>> + struct atmel_nand_host *host = chip->priv;
>> + uint32_t *eccpos = chip->ecc.layout->eccpos;
>> +
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
>> +
>> + pmecc_writel(host->ecc, CFG, (pmecc_readl_relaxed(host->ecc, CFG) |
>> + PMECC_CFG_WRITE_OP)& ~PMECC_CFG_AUTO_ENABLE);
>> +
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DATA);
>> +
>> + chip->write_buf(mtd, (u8 *)buf, mtd->writesize);
>> +
>> + while ((pmecc_readl_relaxed(host->ecc, SR)& PMECC_SR_BUSY)) {
>> + if (unlikely(timeout_count++> PMECC_MAX_TIMEOUT_COUNT)) {
>> + dev_err(host->dev, "PMECC: Timeout to get ECC value.\n");
>> + return; /* Time out */
>> + }
>> + cpu_relax();
>> + }
>> +
>> + for (i = 0; i< host->pmecc_sector_number; i++) {
>> + for (j = 0; j< host->pmecc_bytes_per_sector; j++) {
>> + int pos;
>> +
>> + pos = i * host->pmecc_bytes_per_sector + j;
>> + chip->oob_poi[eccpos[pos]] =
>> + pmecc_readb_ecc_relaxed(host->ecc, i, j);
>> + }
>> + }
>> + chip->write_buf(mtd, chip->oob_poi, mtd->oobsize);
>> +
>> + return;
>> +}
>> +
>> +static void atmel_pmecc_core_init(struct mtd_info *mtd)
>> +{
>> + uint32_t val = 0;
>> + struct nand_chip *nand_chip = mtd->priv;
>> + struct atmel_nand_host *host = nand_chip->priv;
>> + struct nand_ecclayout *ecc_layout;
>> +
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_RST);
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_DISABLE);
>> +
>> + switch (host->pmecc_corr_cap) {
>> + case 2:
>> + val = PMECC_CFG_BCH_ERR2;
>> + break;
>> + case 4:
>> + val = PMECC_CFG_BCH_ERR4;
>> + break;
>> + case 8:
>> + val = PMECC_CFG_BCH_ERR8;
>> + break;
>> + case 12:
>> + val = PMECC_CFG_BCH_ERR12;
>> + break;
>> + case 24:
>> + val = PMECC_CFG_BCH_ERR24;
>> + break;
>> + }
>> +
>> + if (host->pmecc_sector_size == 512)
>> + val |= PMECC_CFG_SECTOR512;
>> + else if (host->pmecc_sector_size == 1024)
>> + val |= PMECC_CFG_SECTOR1024;
>> +
>> + switch (host->pmecc_sector_number) {
>> + case 1:
>> + val |= PMECC_CFG_PAGE_1SECTOR;
>> + break;
>> + case 2:
>> + val |= PMECC_CFG_PAGE_2SECTORS;
>> + break;
>> + case 4:
>> + val |= PMECC_CFG_PAGE_4SECTORS;
>> + break;
>> + case 8:
>> + val |= PMECC_CFG_PAGE_8SECTORS;
>> + break;
>> + }
>> +
>> + val |= (PMECC_CFG_READ_OP | PMECC_CFG_SPARE_DISABLE
>> + | PMECC_CFG_AUTO_DISABLE);
>> + pmecc_writel(host->ecc, CFG, val);
>> +
>> + ecc_layout = nand_chip->ecc.layout;
>> + pmecc_writel(host->ecc, SAREA, mtd->oobsize - 1);
>> + pmecc_writel(host->ecc, SADDR, ecc_layout->eccpos[0]);
>> + pmecc_writel(host->ecc, EADDR,
>> + ecc_layout->eccpos[ecc_layout->eccbytes - 1]);
>> + /* See datasheet about PMECC Clock Control Register */
>> + pmecc_writel(host->ecc, CLK, 2);
>> + pmecc_writel(host->ecc, IDR, 0xff);
>> + pmecc_writel(host->ecc, CTRL, PMECC_CTRL_ENABLE);
>> +}
>> +
>> +static int __init atmel_pmecc_nand_init_params(struct platform_device *pdev,
>> + struct atmel_nand_host *host)
>> +{
>> + int cap, sector_size, err_no;
>> + struct mtd_info *mtd;
>> + struct nand_chip *nand_chip;
>> + struct resource *regs;
>> + struct resource *regs_pmerr, *regs_rom;
>> +
>> + cap = host->pmecc_corr_cap;
>> + sector_size = host->pmecc_sector_size;
>> + dev_info(host->dev, "Initialize PMECC params, cap: %d, sector: %d\n",
>> + cap, sector_size);
>> +
>> + /* Sanity check */
>> + if ((sector_size != 512)&& (sector_size != 1024)) {
>> + dev_err(host->dev,
>> + "Unsupported PMECC sector size: %d; should be 512 or 1024 bytes\n",
>> + sector_size);
>> + return -EINVAL;
>> + }
>> + if ((cap != 2)&& (cap != 4)&& (cap != 8)&& (cap != 12)&&
>> + (cap != 24)) {
>> + dev_err(host->dev,
>> + "Unsupported PMECC correction capability, should be 2, 4, 8, 12 or 24\n");
>> + return -EINVAL;
>> + }
>> +
>> + nand_chip =&host->nand_chip;
>> + mtd =&host->mtd;
>> +
>> + nand_chip->ecc.mode = NAND_ECC_SOFT; /* By default */
>> +
>> + regs = platform_get_resource(pdev, IORESOURCE_MEM, 1);
>> + if (!regs) {
>> + dev_warn(host->dev,
>> + "Can't get I/O resource regs, rolling back on software ECC\n");
>> + return 0;
>> + }
>> +
>> + host->ecc = ioremap(regs->start, resource_size(regs));
>> + if (host->ecc == NULL) {
>> + dev_err(host->dev, "ioremap failed\n");
>> + err_no = -EIO;
>> + goto err_pmecc_ioremap;
>> + }
>> +
>> + regs_pmerr = platform_get_resource(pdev, IORESOURCE_MEM, 2);
>> + regs_rom = platform_get_resource(pdev, IORESOURCE_MEM, 3);
>> + if (regs_pmerr&& regs_rom) {
>> + host->pmerrloc_base = ioremap(regs_pmerr->start,
>> + resource_size(regs_pmerr));
>> + host->pmecc_rom_base = ioremap(regs_rom->start,
>> + resource_size(regs_rom));
>> +
>> + if (host->pmerrloc_base&& host->pmecc_rom_base) {
>> + nand_chip->ecc.mode = NAND_ECC_HW;
>> + nand_chip->ecc.read_page =
>> + atmel_nand_pmecc_read_page;
>> + nand_chip->ecc.write_page =
>> + atmel_nand_pmecc_write_page;
>> + } else {
>> + dev_err(host->dev,
>> + "Can not get I/O resource for PMECC controller!\n");
>> + err_no = -EIO;
>> + goto err_pmloc_ioremap;
>> + }
>> + }
>> +
>> + /* ECC is calculated for the whole page (1 step) */
>> + nand_chip->ecc.size = mtd->writesize;
>> +
>> + /* set ECC page size and oob layout */
>> + switch (mtd->writesize) {
>> + case 2048:
>> + host->pmecc_degree = PMECC_GF_DIMENSION_13;
>> + host->pmecc_cw_len = (1<< host->pmecc_degree) - 1;
>> + host->pmecc_corr_cap = cap;
>> + host->pmecc_sector_number = mtd->writesize / sector_size;
>> + host->pmecc_bytes_per_sector = pmecc_get_ecc_bytes(
>> + cap, sector_size);
>> + host->pmecc_alpha_to = pmecc_get_alpha_to(host);
>> + host->pmecc_index_of = pmecc_get_index_of(host);
>> +
>> + nand_chip->ecc.steps = 1;
>> + nand_chip->ecc.strength = cap;
>> + nand_chip->ecc.bytes = host->pmecc_bytes_per_sector *
>> + host->pmecc_sector_number;
>> + if (nand_chip->ecc.bytes> mtd->oobsize - 2) {
>> + dev_err(host->dev, "No room for ECC bytes\n");
>> + err_no = -EINVAL;
>> + goto err;
>> + }
>> + pmecc_config_ecc_layout(&atmel_pmecc_oobinfo,
>> + mtd->oobsize,
>> + nand_chip->ecc.bytes);
>> + nand_chip->ecc.layout =&atmel_pmecc_oobinfo;
>> + break;
>> + case 512:
>> + case 1024:
>> + case 4096:
>> + /* TODO */
>> + dev_warn(host->dev,
>> + "Unsupported page size for PMECC, use Software ECC\n");
>> + default:
>> + /* page size not handled by HW ECC */
>> + /* switching back to soft ECC */
>> + nand_chip->ecc.mode = NAND_ECC_SOFT;
>> + nand_chip->ecc.calculate = NULL;
>> + nand_chip->ecc.correct = NULL;
>> + nand_chip->ecc.hwctl = NULL;
>> + nand_chip->ecc.read_page = NULL;
>> + nand_chip->ecc.write_page = NULL;
>> + nand_chip->ecc.postpad = 0;
>> + nand_chip->ecc.prepad = 0;
>> + nand_chip->ecc.bytes = 0;
>> + err_no = 0;
>> + goto err;
>> + }
>> +
>> + /* Allocate data for PMECC computation */
>> + host->pmecc_data = kzalloc(sizeof(struct atmel_pmecc_data), GFP_KERNEL);
> why do you always allocate the pmecc_data?
>
> you need to allocate it only if you use it
If not using PMECC, the code will never come to here. So I think it's
right time to allocate all the computation data here.
>
> Best Regards,
> J.
>> + if (!host->pmecc_data) {
>> + dev_err(host->dev,
>> + "Cannot allocate memory for PMECC computation!\n");
>> + err_no = -ENOMEM;
>> + goto err;
>> + }
>> +
>> + atmel_pmecc_core_init(mtd);
>> +
>> + return 0;
>> +
>> +err:
>> +err_pmloc_ioremap:
>> + iounmap(host->ecc);
>> + if (host->pmerrloc_base)
>> + iounmap(host->pmerrloc_base);
>> + if (host->pmecc_rom_base)
>> + iounmap(host->pmecc_rom_base);
>> +err_pmecc_ioremap:
>> + return err_no;
>> +}
>> +
>> +/*
>> * Calculate HW ECC
>> *
>> * function called after a write
>> @@ -720,7 +1456,11 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
>> }
>>
>> if (nand_chip->ecc.mode == NAND_ECC_HW) {
>> - res = atmel_hw_nand_init_params(pdev, host);
>> + if (host->has_pmecc)
>> + res = atmel_pmecc_nand_init_params(pdev, host);
>> + else
>> + res = atmel_hw_nand_init_params(pdev, host);
>> +
>> if (res != 0)
>> goto err_hw_ecc;
>> }
>> @@ -741,6 +1481,12 @@ static int __init atmel_nand_probe(struct platform_device *pdev)
>> err_scan_tail:
>> if (host->ecc)
>> iounmap(host->ecc);
>> + if (host->has_pmecc) {
> no need to check if you have teh pmecc
>
> if it's no the case pmerrloc_base will be NULLo
>
> don't you need to disable it in the error path?
right, I should disable the hardware in this error path.
>
> Best Regards,
> J.
Best Regards,
Josh Wu
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v9 3/3] MTD: at91: atmel_nand: Update driver to support Programmable Multibit ECC controller
2012-05-27 12:50 ` Artem Bityutskiy
@ 2012-05-28 8:43 ` Josh Wu
-1 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-28 8:43 UTC (permalink / raw)
To: Artem Bityutskiy
Cc: hongxu.cn, nicolas.ferre, linux-mtd, ivan.djelic, plagnioj,
linux-arm-kernel
On 5/27/2012 8:50 PM, Artem Bityutskiy wrote:
> On Sat, 2012-05-26 at 21:24 +0800, Josh Wu wrote:
>> + while ((pmecc_readl_relaxed(host->ecc, SR)& PMECC_SR_BUSY)) {
>> + if (unlikely(timeout_count++> PMECC_MAX_TIMEOUT_COUNT)) {
>> + dev_err(host->dev, "PMECC: Timeout to get ECC value.\n");
>> + return; /* Time out */
> How this error is communicated then up the the user?
If this error happened, that should mean the PMECC is not configurated
correctly. I think I only I can do is that prompt to user and then call
BUG() here.
>
>> + }
>> + cpu_relax();
>> + }
> I see this pattern all over the place - why people consider it reliable?
> Is this code guaranteed to run on the same CPU?
>
> Why not to use loops_per_jiffie * msecs_to_jiffies(TIMOUT) instead to
> calculate how many iterations to do? Yes, due to HW register reading and
> cpu_relax() the real timeout will be larger, but this is about error
> anyway, so it does not hurt to iterate longer?
>
I will fix that in next version. thanks.
Best Regards,
Josh Wu
^ permalink raw reply [flat|nested] 24+ messages in thread
* [PATCH v9 3/3] MTD: at91: atmel_nand: Update driver to support Programmable Multibit ECC controller
@ 2012-05-28 8:43 ` Josh Wu
0 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-28 8:43 UTC (permalink / raw)
To: linux-arm-kernel
On 5/27/2012 8:50 PM, Artem Bityutskiy wrote:
> On Sat, 2012-05-26 at 21:24 +0800, Josh Wu wrote:
>> + while ((pmecc_readl_relaxed(host->ecc, SR)& PMECC_SR_BUSY)) {
>> + if (unlikely(timeout_count++> PMECC_MAX_TIMEOUT_COUNT)) {
>> + dev_err(host->dev, "PMECC: Timeout to get ECC value.\n");
>> + return; /* Time out */
> How this error is communicated then up the the user?
If this error happened, that should mean the PMECC is not configurated
correctly. I think I only I can do is that prompt to user and then call
BUG() here.
>
>> + }
>> + cpu_relax();
>> + }
> I see this pattern all over the place - why people consider it reliable?
> Is this code guaranteed to run on the same CPU?
>
> Why not to use loops_per_jiffie * msecs_to_jiffies(TIMOUT) instead to
> calculate how many iterations to do? Yes, due to HW register reading and
> cpu_relax() the real timeout will be larger, but this is about error
> anyway, so it does not hurt to iterate longer?
>
I will fix that in next version. thanks.
Best Regards,
Josh Wu
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v9 1/3] MTD: at91: extract hw ecc initialization to one function and use relaxed read/write
2012-05-27 12:44 ` Artem Bityutskiy
@ 2012-05-28 8:50 ` Josh Wu
-1 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-28 8:50 UTC (permalink / raw)
To: Artem Bityutskiy
Cc: hongxu.cn, nicolas.ferre, linux-mtd, ivan.djelic, plagnioj,
linux-arm-kernel
On 5/27/2012 8:44 PM, Artem Bityutskiy wrote:
> On Sat, 2012-05-26 at 21:24 +0800, Josh Wu wrote:
>> use _relaxed read/write in most place. And use writel in operations of Control Register since it needs memory barrier.
>>
>> Signed-off-by: Hong Xu<hong.xu@atmel.com>
>> Signed-off-by: Josh Wu<josh.wu@atmel.com>
> This should be split on 2 or even 3 parts:
>
> 1. You move definitions and things around.
> 2. You separate out the init stuff into a function
> 3. You start using _relaxed helpers.
>
> And you should provide better commit messages. You should explain why
> you do things - just cleanup? improvement? fix? preparation for
> something?
My goal here is to prepare for the PMECC support.
Point 1 is consistent for the later PMECC patch.
Point 2 is prepare for PMECC support.
So I think I will not add the _relaxed helpers for HW ECC operation in
this patch series. Since that is out of the goal of this patch series.
>
> I think this patch is too big, commit message is poor, patch does more
> things than documented in the commit message.
>
Indeed the commit message is poor. I will fix it.
Best Regards,
Josh Wu
^ permalink raw reply [flat|nested] 24+ messages in thread
* [PATCH v9 1/3] MTD: at91: extract hw ecc initialization to one function and use relaxed read/write
@ 2012-05-28 8:50 ` Josh Wu
0 siblings, 0 replies; 24+ messages in thread
From: Josh Wu @ 2012-05-28 8:50 UTC (permalink / raw)
To: linux-arm-kernel
On 5/27/2012 8:44 PM, Artem Bityutskiy wrote:
> On Sat, 2012-05-26 at 21:24 +0800, Josh Wu wrote:
>> use _relaxed read/write in most place. And use writel in operations of Control Register since it needs memory barrier.
>>
>> Signed-off-by: Hong Xu<hong.xu@atmel.com>
>> Signed-off-by: Josh Wu<josh.wu@atmel.com>
> This should be split on 2 or even 3 parts:
>
> 1. You move definitions and things around.
> 2. You separate out the init stuff into a function
> 3. You start using _relaxed helpers.
>
> And you should provide better commit messages. You should explain why
> you do things - just cleanup? improvement? fix? preparation for
> something?
My goal here is to prepare for the PMECC support.
Point 1 is consistent for the later PMECC patch.
Point 2 is prepare for PMECC support.
So I think I will not add the _relaxed helpers for HW ECC operation in
this patch series. Since that is out of the goal of this patch series.
>
> I think this patch is too big, commit message is poor, patch does more
> things than documented in the commit message.
>
Indeed the commit message is poor. I will fix it.
Best Regards,
Josh Wu
^ permalink raw reply [flat|nested] 24+ messages in thread
* Re: [PATCH v9 3/3] MTD: at91: atmel_nand: Update driver to support Programmable Multibit ECC controller
2012-05-28 8:34 ` Josh Wu
@ 2012-05-29 16:01 ` Jean-Christophe PLAGNIOL-VILLARD
-1 siblings, 0 replies; 24+ messages in thread
From: Jean-Christophe PLAGNIOL-VILLARD @ 2012-05-29 16:01 UTC (permalink / raw)
To: Josh Wu
Cc: hongxu.cn, dedekind1, nicolas.ferre, linux-mtd, ivan.djelic,
linux-arm-kernel
> >>+
> >>+static void __iomem *pmecc_get_index_of(struct atmel_nand_host *host)
> >this is a __dev_init function plese check the other too
> >
> >btw you need to use __dev_init and not __init
>
> Here I use pmecc_get_index_of() function to get a lookup table base
> which is in ROM.
> I'm not clear about how should I need __dev_init here. Can you give
> more information?
it is the section where the function is going to be stored
pmecc_get_index_of is use at probe time so it's __devinit section
as it's supposed to be for the probe function too
Best Regards,
J.
^ permalink raw reply [flat|nested] 24+ messages in thread
* [PATCH v9 3/3] MTD: at91: atmel_nand: Update driver to support Programmable Multibit ECC controller
@ 2012-05-29 16:01 ` Jean-Christophe PLAGNIOL-VILLARD
0 siblings, 0 replies; 24+ messages in thread
From: Jean-Christophe PLAGNIOL-VILLARD @ 2012-05-29 16:01 UTC (permalink / raw)
To: linux-arm-kernel
> >>+
> >>+static void __iomem *pmecc_get_index_of(struct atmel_nand_host *host)
> >this is a __dev_init function plese check the other too
> >
> >btw you need to use __dev_init and not __init
>
> Here I use pmecc_get_index_of() function to get a lookup table base
> which is in ROM.
> I'm not clear about how should I need __dev_init here. Can you give
> more information?
it is the section where the function is going to be stored
pmecc_get_index_of is use at probe time so it's __devinit section
as it's supposed to be for the probe function too
Best Regards,
J.
^ permalink raw reply [flat|nested] 24+ messages in thread
end of thread, other threads:[~2012-05-29 16:01 UTC | newest]
Thread overview: 24+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2012-05-26 13:24 [PATCH v9 0/3] MTD: at91: Add PMECC support for at91 nand flash driver Josh Wu
2012-05-26 13:24 ` Josh Wu
2012-05-26 13:24 ` [PATCH v9 1/3] MTD: at91: extract hw ecc initialization to one function and use relaxed read/write Josh Wu
2012-05-26 13:24 ` Josh Wu
2012-05-27 12:44 ` Artem Bityutskiy
2012-05-27 12:44 ` Artem Bityutskiy
2012-05-28 8:50 ` Josh Wu
2012-05-28 8:50 ` Josh Wu
2012-05-26 13:24 ` [PATCH v9 2/3] MTD: at91: add dt parameters for PMECC Josh Wu
2012-05-26 13:24 ` Josh Wu
2012-05-26 13:24 ` [PATCH v9 3/3] MTD: at91: atmel_nand: Update driver to support Programmable Multibit ECC controller Josh Wu
2012-05-26 13:24 ` Josh Wu
2012-05-27 12:50 ` Artem Bityutskiy
2012-05-27 12:50 ` Artem Bityutskiy
2012-05-28 8:43 ` Josh Wu
2012-05-28 8:43 ` Josh Wu
2012-05-28 6:58 ` Jean-Christophe PLAGNIOL-VILLARD
2012-05-28 6:58 ` Jean-Christophe PLAGNIOL-VILLARD
2012-05-28 8:34 ` Josh Wu
2012-05-28 8:34 ` Josh Wu
2012-05-29 16:01 ` Jean-Christophe PLAGNIOL-VILLARD
2012-05-29 16:01 ` Jean-Christophe PLAGNIOL-VILLARD
2012-05-28 6:39 ` [PATCH v9 0/3] MTD: at91: Add PMECC support for at91 nand flash driver Jean-Christophe PLAGNIOL-VILLARD
2012-05-28 6:39 ` Jean-Christophe PLAGNIOL-VILLARD
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.