* [PATCH v3 0/4] optimize and clean-up of OMAP NAND and ELM driver
@ 2013-11-02 9:46 Pekon Gupta
2013-11-02 9:46 ` [PATCH v3 1/4] mtd: nand: omap: optimized chip->ecc.correct() for H/W ECC schemes Pekon Gupta
` (3 more replies)
0 siblings, 4 replies; 11+ messages in thread
From: Pekon Gupta @ 2013-11-02 9:46 UTC (permalink / raw)
To: computersforpeace, dedekind1; +Cc: linux-mtd, Pekon Gupta, balbi
*changes from v2 (rebased)*
- rebased on [PATCH v11] mtd:nand:omap2: clean-up of supported ECC schemes
http://lists.infradead.org/pipermail/linux-mtd/2013-October/049410.html
*changes from v1 (minor)*
- rebased for newer version of previous patch-set
[PATCH v7 0/6] mtd:nand:omap2: clean-up of supported ECC schemes
http://lists.infradead.org/pipermail/linux-mtd/2013-October/048983.html
- code cleanup and formating updates
*original v1*
This patch series builds the base for adding support for newer H/W based ECC
schemes for NAND flash. This patch series
- cleans-up redundant code among various
- optimizes ecc calculation and correction paths
- improves scalability and readability, for adding newer ecc-schemes in future.
This patch affects following data-paths:
[PATCH 1/4] nand_chip->ecc.correct(): Detects and corrects ECC errors on read accesses.
[PATCH 2/4] nand_chip->ecc.calculate(): Calculates and fetches ECC syndrome from controller.
[PATCH 3/4] nand_chip->ecc.hwctl(): Configures H/W controller for Read/Write data acccesses.
[PATCH 4/4] ELM driver: add checks and updates code for scalability.
Pekon Gupta (4):
mtd: nand: omap: optimized chip->ecc.correct() for H/W ECC schemes
mtd: nand: omap: optimize chip->ecc.calculate() for H/W ECC schemes
mtd: nand: omap: optimize chip->ecc.hwctl() for H/W ECC schemes
mtd: devices: elm: add checks ELM H/W constrains, driver code cleanup
drivers/mtd/devices/elm.c | 122 ++++---
drivers/mtd/nand/omap2.c | 704 +++++++++++++-------------------------
include/linux/platform_data/elm.h | 6 +-
3 files changed, 316 insertions(+), 516 deletions(-)
--
1.8.1
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH v3 1/4] mtd: nand: omap: optimized chip->ecc.correct() for H/W ECC schemes
2013-11-02 9:46 [PATCH v3 0/4] optimize and clean-up of OMAP NAND and ELM driver Pekon Gupta
@ 2013-11-02 9:46 ` Pekon Gupta
2013-11-13 21:11 ` Brian Norris
2013-11-02 9:46 ` [PATCH v3 2/4] mtd: nand: omap: optimize chip->ecc.calculate() " Pekon Gupta
` (2 subsequent siblings)
3 siblings, 1 reply; 11+ messages in thread
From: Pekon Gupta @ 2013-11-02 9:46 UTC (permalink / raw)
To: computersforpeace, dedekind1; +Cc: linux-mtd, Pekon Gupta, balbi
chip->ecc.correct() is used for detecting and correcting bit-flips during read
operations. In omap2-nand driver this is done usingt following functions:
- omap_correct_data(): for H/W based HAM1_ECC schemes
(Un-Touched in current patch)
- omap_elm_correct_data(): for H/W based BCHx_ECC scheme
Current implementation of this function is not scalable for newer ECC
schemes because:
- It depends on a specific byte-position in OOB area (reserved as 0x00)
to differentiates between programmed-pages and erased-pages.
This reserved byte-position cannot be accomodated in all ECC schemes.
- Current code is not scalable for future ECC schemes due to tweaks for
BCH4_ECC and BCH8_ECC at multiple places.
- It checks for bit-flips in Erased-pages using check_erased_page().
This is over-work, as sanity of Erased-page can be verified by just
comparing them to a pre-defined ECC-syndrome for all_0xFF data.
This patch optimizes omap_elm_correct_data() in following ways:
(1) Removes dependency on specific reserved-byte (0x00) in OOB area,
instead Erased-page is identified by matching calc_ecc with a
pre-defined ECC syndrome of all(0xFF) data
(2) merges common code for BCH4_ECC and BCH8_ECC for scalability.
(3) handles incorrect elm_error_location beyond data+oob buffer.
(4) removes check_erased_page(): Bit-flips in erased-page are handled
in same way as for programmed-page
Signed-off-by: Pekon Gupta <pekon@ti.com>
---
drivers/mtd/nand/omap2.c | 247 ++++++++++++++---------------------------------
1 file changed, 74 insertions(+), 173 deletions(-)
diff --git a/drivers/mtd/nand/omap2.c b/drivers/mtd/nand/omap2.c
index ec40b8d..c946f22 100644
--- a/drivers/mtd/nand/omap2.c
+++ b/drivers/mtd/nand/omap2.c
@@ -160,6 +160,7 @@ struct omap_nand_info {
int gpmc_cs;
unsigned long phys_base;
unsigned long mem_size;
+ enum omap_ecc ecc_opt;
struct completion comp;
struct dma_chan *dma;
int gpmc_irq_fifo;
@@ -1291,219 +1292,118 @@ static int omap3_calculate_ecc_bch(struct mtd_info *mtd, const u_char *dat,
}
/**
- * erased_sector_bitflips - count bit flips
- * @data: data sector buffer
- * @oob: oob buffer
- * @info: omap_nand_info
- *
- * Check the bit flips in erased page falls below correctable level.
- * If falls below, report the page as erased with correctable bit
- * flip, else report as uncorrectable page.
- */
-static int erased_sector_bitflips(u_char *data, u_char *oob,
- struct omap_nand_info *info)
-{
- int flip_bits = 0, i;
-
- for (i = 0; i < info->nand.ecc.size; i++) {
- flip_bits += hweight8(~data[i]);
- if (flip_bits > info->nand.ecc.strength)
- return 0;
- }
-
- for (i = 0; i < info->nand.ecc.bytes - 1; i++) {
- flip_bits += hweight8(~oob[i]);
- if (flip_bits > info->nand.ecc.strength)
- return 0;
- }
-
- /*
- * Bit flips falls in correctable level.
- * Fill data area with 0xFF
- */
- if (flip_bits) {
- memset(data, 0xFF, info->nand.ecc.size);
- memset(oob, 0xFF, info->nand.ecc.bytes);
- }
-
- return flip_bits;
-}
-
-/**
* omap_elm_correct_data - corrects page data area in case error reported
* @mtd: MTD device structure
* @data: page data
* @read_ecc: ecc read from nand flash
- * @calc_ecc: ecc read from HW ECC registers
- *
- * Calculated ecc vector reported as zero in case of non-error pages.
- * In case of error/erased pages non-zero error vector is reported.
- * In case of non-zero ecc vector, check read_ecc at fixed offset
- * (x = 13/7 in case of BCH8/4 == 0) to find page programmed or not.
- * To handle bit flips in this data, count the number of 0's in
- * read_ecc[x] and check if it greater than 4. If it is less, it is
- * programmed page, else erased page.
- *
- * 1. If page is erased, check with standard ecc vector (ecc vector
- * for erased page to find any bit flip). If check fails, bit flip
- * is present in erased page. Count the bit flips in erased page and
- * if it falls under correctable level, report page with 0xFF and
- * update the correctable bit information.
- * 2. If error is reported on programmed page, update elm error
- * vector and correct the page with ELM error correction routine.
- *
+ * @calc_ecc: ecc calculated after reading Data and OOB regions from flash
+ * As calc_ecc is calculated over both main & oob, so calc_ecc would be
+ * non-zero only in following cases:
+ * - bit-flips in data or oob region
+ * - erase page, where no ECC is written in OOB area
+ * However, erased_pages can be differentiated from corrupted pages
+ * by comparing the calculated ECC with pre-defined syndrome ECC_of_ALL(0xFF)
+ * Bit-flips in erased-pages would also be caught by comparing, calc_ecc
+ * with ECC_of_ALL(0xFF)
*/
static int omap_elm_correct_data(struct mtd_info *mtd, u_char *data,
u_char *read_ecc, u_char *calc_ecc)
{
struct omap_nand_info *info = container_of(mtd, struct omap_nand_info,
mtd);
- int eccsteps = info->nand.ecc.steps;
- int i , j, stat = 0;
- int eccsize, eccflag, ecc_vector_size;
+ enum omap_ecc ecc_opt = info->ecc_opt;
+ struct nand_chip *chip = mtd->priv;
+ int eccsteps = chip->ecc.steps;
+ int eccsize = chip->ecc.size;
+ int eccbytes = chip->ecc.bytes;
+ int i , j, stat = 0, ret = 0, flag_read_ecc;
struct elm_errorvec err_vec[ERROR_VECTOR_MAX];
- u_char *ecc_vec = calc_ecc;
- u_char *spare_ecc = read_ecc;
- u_char *erased_ecc_vec;
- enum bch_ecc type;
+ u_char *ecc;
bool is_error_reported = false;
+ u32 bit_pos, byte_pos, error_max, pos;
/* Initialize elm error vector to zero */
memset(err_vec, 0, sizeof(err_vec));
- if (info->nand.ecc.strength == BCH8_MAX_ERROR) {
- type = BCH8_ECC;
- erased_ecc_vec = bch8_vector;
- } else {
- type = BCH4_ECC;
- erased_ecc_vec = bch4_vector;
- }
-
- ecc_vector_size = info->nand.ecc.bytes;
-
- /*
- * Remove extra byte padding for BCH8 RBL
- * compatibility and erased page handling
- */
- eccsize = ecc_vector_size - 1;
-
for (i = 0; i < eccsteps ; i++) {
- eccflag = 0; /* initialize eccflag */
-
- /*
- * Check any error reported,
- * In case of error, non zero ecc reported.
- */
-
- for (j = 0; (j < eccsize); j++) {
- if (calc_ecc[j] != 0) {
- eccflag = 1; /* non zero ecc, error present */
+ flag_read_ecc = 0;
+ ecc = calc_ecc + (i * eccbytes);
+ /* check calc_ecc */
+ for (j = 0; j < eccbytes; j++) {
+ if (*(ecc + j) != 0x00) {
+ flag_read_ecc = 1;
break;
}
}
-
- if (eccflag == 1) {
- /*
- * Set threshold to minimum of 4, half of ecc.strength/2
- * to allow max bit flip in byte to 4
- */
- unsigned int threshold = min_t(unsigned int, 4,
- info->nand.ecc.strength / 2);
-
- /*
- * Check data area is programmed by counting
- * number of 0's at fixed offset in spare area.
- * Checking count of 0's against threshold.
- * In case programmed page expects at least threshold
- * zeros in byte.
- * If zeros are less than threshold for programmed page/
- * zeros are more than threshold erased page, either
- * case page reported as uncorrectable.
- */
- if (hweight8(~read_ecc[eccsize]) >= threshold) {
- /*
- * Update elm error vector as
- * data area is programmed
- */
- err_vec[i].error_reported = true;
- is_error_reported = true;
- } else {
- /* Error reported in erased page */
- int bitflip_count;
- u_char *buf = &data[info->nand.ecc.size * i];
-
- if (memcmp(calc_ecc, erased_ecc_vec, eccsize)) {
- bitflip_count = erased_sector_bitflips(
- buf, read_ecc, info);
-
- if (bitflip_count)
- stat += bitflip_count;
- else
- return -EINVAL;
- }
+ /* check if its a erased-page */
+ if (flag_read_ecc) {
+ switch (ecc_opt) {
+ case OMAP_ECC_BCH8_CODE_HW:
+ if (memcmp(ecc, bch8_vector, eccbytes))
+ err_vec[i].error_reported = true;
+ break;
+ case OMAP_ECC_BCH4_CODE_HW:
+ if (memcmp(ecc, bch4_vector, eccbytes))
+ err_vec[i].error_reported = true;
+ break;
+ default:
+ pr_err("%s: invalid configuration",
+ DRIVER_NAME);
+ return -EINVAL;
}
}
-
- /* Update the ecc vector */
- calc_ecc += ecc_vector_size;
- read_ecc += ecc_vector_size;
+ /* page definitely has bit-flips */
+ if (err_vec[i].error_reported)
+ is_error_reported = true;
}
- /* Check if any error reported */
if (!is_error_reported)
return 0;
+ /* detect bit-flips using ELM module */
+ elm_decode_bch_error_page(info->elm_dev, calc_ecc, err_vec);
- /* Decode BCH error using ELM module */
- elm_decode_bch_error_page(info->elm_dev, ecc_vec, err_vec);
-
+ /* correct bit-flip */
for (i = 0; i < eccsteps; i++) {
- if (err_vec[i].error_reported) {
+ if (err_vec[i].error_uncorrectable) {
+ ret = -EBADMSG;
+ } else if (err_vec[i].error_reported) {
for (j = 0; j < err_vec[i].error_count; j++) {
- u32 bit_pos, byte_pos, error_max, pos;
-
- if (type == BCH8_ECC)
- error_max = BCH8_ECC_MAX;
- else
- error_max = BCH4_ECC_MAX;
-
- if (info->nand.ecc.strength == BCH8_MAX_ERROR)
- pos = err_vec[i].error_loc[j];
- else
+ switch (ecc_opt) {
+ case OMAP_ECC_BCH4_CODE_HW:
+ error_max = SECTOR_BYTES +
+ (eccbytes - 1);
/* Add 4 to take care 4 bit padding */
pos = err_vec[i].error_loc[j] +
- BCH4_BIT_PAD;
-
- /* Calculate bit position of error */
+ BCH4_BIT_PAD;
+ break;
+ case OMAP_ECC_BCH8_CODE_HW:
+ error_max = SECTOR_BYTES +
+ (eccbytes - 1);
+ pos = err_vec[i].error_loc[j];
+ break;
+ default:
+ return -EINVAL;
+ }
+ /* Calculate bit & byte bit-flip position */
bit_pos = pos % 8;
-
- /* Calculate byte position of error */
- byte_pos = (error_max - pos - 1) / 8;
-
- if (pos < error_max) {
- if (byte_pos < 512)
- data[byte_pos] ^= 1 << bit_pos;
- else
- spare_ecc[byte_pos - 512] ^=
+ byte_pos = error_max - (pos / 8) - 1;
+ if (byte_pos < SECTOR_BYTES)
+ data[byte_pos] ^= 1 << bit_pos;
+ else if (byte_pos < error_max)
+ read_ecc[byte_pos - SECTOR_BYTES] ^=
1 << bit_pos;
- }
- /* else, not interested to correct ecc */
+ else
+ ret = -EBADMSG;
}
}
-
/* Update number of correctable errors */
stat += err_vec[i].error_count;
-
/* Update page data with sector size */
- data += info->nand.ecc.size;
- spare_ecc += ecc_vector_size;
+ data += eccsize;
+ read_ecc += eccbytes;
}
- for (i = 0; i < eccsteps; i++)
- /* Return error if uncorrectable error present */
- if (err_vec[i].error_uncorrectable)
- return -EINVAL;
-
- return stat;
+ return (ret < 0) ? ret : stat;
}
/**
@@ -1656,6 +1556,7 @@ static int omap_nand_probe(struct platform_device *pdev)
info->gpmc_cs = pdata->cs;
info->reg = pdata->reg;
info->of_node = pdata->of_node;
+ info->ecc_opt = pdata->ecc_opt;
mtd = &info->mtd;
mtd->priv = &info->nand;
mtd->name = dev_name(&pdev->dev);
--
1.8.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH v3 2/4] mtd: nand: omap: optimize chip->ecc.calculate() for H/W ECC schemes
2013-11-02 9:46 [PATCH v3 0/4] optimize and clean-up of OMAP NAND and ELM driver Pekon Gupta
2013-11-02 9:46 ` [PATCH v3 1/4] mtd: nand: omap: optimized chip->ecc.correct() for H/W ECC schemes Pekon Gupta
@ 2013-11-02 9:46 ` Pekon Gupta
2013-11-13 21:45 ` Brian Norris
2013-11-02 9:46 ` [PATCH v3 3/4] mtd: nand: omap: optimize chip->ecc.hwctl() " Pekon Gupta
2013-11-02 9:46 ` [PATCH v3 4/4] mtd: devices: elm: add checks ELM H/W constrains, driver code cleanup Pekon Gupta
3 siblings, 1 reply; 11+ messages in thread
From: Pekon Gupta @ 2013-11-02 9:46 UTC (permalink / raw)
To: computersforpeace, dedekind1; +Cc: linux-mtd, Pekon Gupta, balbi
chip->ecc.calculate() is used for calculating and fetching of ECC syndrome by
processing the data passed during Read/Write accesses.
All H/W based ECC schemes supported in omap2-nand driver use GPMC controller
to calculate ECC syndrome. But each BCHx_ECC scheme implements its own function
to process and fetch ECC syndrom from GPMC controller.
This patch tries to merges the common code for different BCHx_ECC schemes into
single omap_calculate_ecc_bch(), And adds schemes specific post-possessing
after fetching ECC-syndrome. This removes redundant code and adds scalability
for future ECC-schemes. This patch:
- [un-touched] omap_calculate_ecc(): Used for HAM1_ECC
- [merged] omap3_calculate_ecc_bch4(): Used for BCH4_HW_DETECTION_SW
- [merged] omap3_calculate_ecc_bch8(): Used for BCH8_HW_DETECTION_SW
- [merged] omap3_calculate_ecc_bch(): Used for BCH4_HW and BCH8_HW
- [new] omap_calculate_ecc_bch(): Now used for all BCHx_ECC
Signed-off-by: Pekon Gupta <pekon@ti.com>
---
drivers/mtd/nand/omap2.c | 245 ++++++++++++++++++-----------------------------
1 file changed, 92 insertions(+), 153 deletions(-)
diff --git a/drivers/mtd/nand/omap2.c b/drivers/mtd/nand/omap2.c
index c946f22..1f59505 100644
--- a/drivers/mtd/nand/omap2.c
+++ b/drivers/mtd/nand/omap2.c
@@ -146,7 +146,11 @@ static u_char bch8_vector[] = {0xf3, 0xdb, 0x14, 0x16, 0x8b, 0xd2, 0xbe, 0xcc,
0xac, 0x6b, 0xff, 0x99, 0x7b};
static u_char bch4_vector[] = {0x00, 0x6b, 0x31, 0xdd, 0x41, 0xbc, 0x10};
#endif
-
+#if defined(CONFIG_MTD_NAND_ECC_BCH) || defined(CONFIG_MTD_NAND_OMAP_BCH)
+static u8 bch4_polynomial[] = {0x28, 0x13, 0xcc, 0x39, 0x96, 0xac, 0x7f};
+static u8 bch8_polynomial[] = {0xef, 0x51, 0x2e, 0x09, 0xed, 0x93, 0x9a, 0xc2,
+ 0x97, 0x79, 0xe5, 0x24, 0xb5};
+#endif
/* oob info generated runtime depending on ecc algorithm and layout selected */
static struct nand_ecclayout omap_oobinfo;
@@ -934,9 +938,11 @@ static int omap_calculate_ecc(struct mtd_info *mtd, const u_char *dat,
u32 val;
val = readl(info->reg.gpmc_ecc_config);
- if (((val >> ECC_CONFIG_CS_SHIFT) & ~CS_MASK) != info->gpmc_cs)
+ if (((val >> 1) & 0x07) != info->gpmc_cs) {
+ pr_err("%s: invalid ECC configuration for chip-select=%d",
+ DRIVER_NAME, info->gpmc_cs);
return -EINVAL;
-
+ }
/* read ecc result */
val = readl(info->reg.gpmc_ecc1_result);
*ecc_code++ = val; /* P128e, ..., P1e */
@@ -1125,172 +1131,105 @@ static void omap3_enable_hwecc_bch(struct mtd_info *mtd, int mode)
/* Clear ecc and enable bits */
writel(ECCCLEAR | ECC1, info->reg.gpmc_ecc_control);
}
-#endif
-
-#ifdef CONFIG_MTD_NAND_ECC_BCH
-/**
- * omap3_calculate_ecc_bch4 - Generate 7 bytes of ECC bytes
- * @mtd: MTD device structure
- * @dat: The pointer to data on which ecc is computed
- * @ecc_code: The ecc_code buffer
- */
-static int omap3_calculate_ecc_bch4(struct mtd_info *mtd, const u_char *dat,
- u_char *ecc_code)
-{
- struct omap_nand_info *info = container_of(mtd, struct omap_nand_info,
- mtd);
- unsigned long nsectors, val1, val2;
- int i;
-
- nsectors = ((readl(info->reg.gpmc_ecc_config) >> 4) & 0x7) + 1;
-
- for (i = 0; i < nsectors; i++) {
-
- /* Read hw-computed remainder */
- val1 = readl(info->reg.gpmc_bch_result0[i]);
- val2 = readl(info->reg.gpmc_bch_result1[i]);
-
- /*
- * Add constant polynomial to remainder, in order to get an ecc
- * sequence of 0xFFs for a buffer filled with 0xFFs; and
- * left-justify the resulting polynomial.
- */
- *ecc_code++ = 0x28 ^ ((val2 >> 12) & 0xFF);
- *ecc_code++ = 0x13 ^ ((val2 >> 4) & 0xFF);
- *ecc_code++ = 0xcc ^ (((val2 & 0xF) << 4)|((val1 >> 28) & 0xF));
- *ecc_code++ = 0x39 ^ ((val1 >> 20) & 0xFF);
- *ecc_code++ = 0x96 ^ ((val1 >> 12) & 0xFF);
- *ecc_code++ = 0xac ^ ((val1 >> 4) & 0xFF);
- *ecc_code++ = 0x7f ^ ((val1 & 0xF) << 4);
- }
-
- return 0;
-}
/**
- * omap3_calculate_ecc_bch8 - Generate 13 bytes of ECC bytes
- * @mtd: MTD device structure
- * @dat: The pointer to data on which ecc is computed
- * @ecc_code: The ecc_code buffer
- */
-static int omap3_calculate_ecc_bch8(struct mtd_info *mtd, const u_char *dat,
- u_char *ecc_code)
-{
- struct omap_nand_info *info = container_of(mtd, struct omap_nand_info,
- mtd);
- unsigned long nsectors, val1, val2, val3, val4;
- int i;
-
- nsectors = ((readl(info->reg.gpmc_ecc_config) >> 4) & 0x7) + 1;
-
- for (i = 0; i < nsectors; i++) {
-
- /* Read hw-computed remainder */
- val1 = readl(info->reg.gpmc_bch_result0[i]);
- val2 = readl(info->reg.gpmc_bch_result1[i]);
- val3 = readl(info->reg.gpmc_bch_result2[i]);
- val4 = readl(info->reg.gpmc_bch_result3[i]);
-
- /*
- * Add constant polynomial to remainder, in order to get an ecc
- * sequence of 0xFFs for a buffer filled with 0xFFs.
- */
- *ecc_code++ = 0xef ^ (val4 & 0xFF);
- *ecc_code++ = 0x51 ^ ((val3 >> 24) & 0xFF);
- *ecc_code++ = 0x2e ^ ((val3 >> 16) & 0xFF);
- *ecc_code++ = 0x09 ^ ((val3 >> 8) & 0xFF);
- *ecc_code++ = 0xed ^ (val3 & 0xFF);
- *ecc_code++ = 0x93 ^ ((val2 >> 24) & 0xFF);
- *ecc_code++ = 0x9a ^ ((val2 >> 16) & 0xFF);
- *ecc_code++ = 0xc2 ^ ((val2 >> 8) & 0xFF);
- *ecc_code++ = 0x97 ^ (val2 & 0xFF);
- *ecc_code++ = 0x79 ^ ((val1 >> 24) & 0xFF);
- *ecc_code++ = 0xe5 ^ ((val1 >> 16) & 0xFF);
- *ecc_code++ = 0x24 ^ ((val1 >> 8) & 0xFF);
- *ecc_code++ = 0xb5 ^ (val1 & 0xFF);
- }
-
- return 0;
-}
-#endif /* CONFIG_MTD_NAND_ECC_BCH */
-
-#ifdef CONFIG_MTD_NAND_OMAP_BCH
-/**
- * omap3_calculate_ecc_bch - Generate bytes of ECC bytes
+ * omap_calculate_ecc_bch - Generate bytes of ECC bytes
* @mtd: MTD device structure
* @dat: The pointer to data on which ecc is computed
* @ecc_code: The ecc_code buffer
*
* Support calculating of BCH4/8 ecc vectors for the page
*/
-static int omap3_calculate_ecc_bch(struct mtd_info *mtd, const u_char *dat,
- u_char *ecc_code)
+static int omap_calculate_ecc_bch(struct mtd_info *mtd, const u_char *dat,
+ u_char *ecc_calc)
{
struct omap_nand_info *info = container_of(mtd, struct omap_nand_info,
mtd);
- unsigned long nsectors, bch_val1, bch_val2, bch_val3, bch_val4;
- int i, eccbchtsel;
+ struct nand_chip *chip = mtd->priv;
+ enum omap_ecc ecc_opt = info->ecc_opt;
+ struct gpmc_nand_regs *gpmc_regs = &info->reg;
+ u32 eccbytes = chip->ecc.bytes;
+ u_char *ecc_ptr;
+ u32 nsectors;
+ int i, val;
- nsectors = ((readl(info->reg.gpmc_ecc_config) >> 4) & 0x7) + 1;
- /*
- * find BCH scheme used
- * 0 -> BCH4
- * 1 -> BCH8
- */
- eccbchtsel = ((readl(info->reg.gpmc_ecc_config) >> 12) & 0x3);
+ val = readl(info->reg.gpmc_ecc_config);
+ if (((val >> 1) & 0x07) != info->gpmc_cs) {
+ pr_err("%s: invalid ECC configuration for chip-select=%d",
+ DRIVER_NAME, info->gpmc_cs);
+ return -EINVAL;
+ }
+ nsectors = ((readl(gpmc_regs->gpmc_ecc_config) >> 4) & 0x7) + 1;
for (i = 0; i < nsectors; i++) {
-
- /* Read hw-computed remainder */
- bch_val1 = readl(info->reg.gpmc_bch_result0[i]);
- bch_val2 = readl(info->reg.gpmc_bch_result1[i]);
- if (eccbchtsel) {
- bch_val3 = readl(info->reg.gpmc_bch_result2[i]);
- bch_val4 = readl(info->reg.gpmc_bch_result3[i]);
+ ecc_ptr = ecc_calc;
+ switch (ecc_opt) {
+ case OMAP_ECC_BCH4_CODE_HW_DETECTION_SW:
+ case OMAP_ECC_BCH4_CODE_HW:
+ val = readl(gpmc_regs->gpmc_bch_result1[i]);
+ *(ecc_ptr++) = ((val >> 12) & 0xFF);
+ *(ecc_ptr++) = ((val >> 4) & 0xFF);
+ *(ecc_ptr) = ((val >> 0) << 4) & 0xF0;
+ val = readl(gpmc_regs->gpmc_bch_result0[i]);
+ *(ecc_ptr) = ((val >> 28) & 0x0F) | *(ecc_ptr);
+ ecc_ptr++;
+ *(ecc_ptr++) = ((val >> 20) & 0xFF);
+ *(ecc_ptr++) = ((val >> 12) & 0xFF);
+ *(ecc_ptr++) = ((val >> 4) & 0xFF);
+ *(ecc_ptr++) = ((val >> 0) << 4) & 0xF0;
+ break;
+ case OMAP_ECC_BCH8_CODE_HW_DETECTION_SW:
+ case OMAP_ECC_BCH8_CODE_HW:
+ val = readl(gpmc_regs->gpmc_bch_result3[i]);
+ *(ecc_ptr++) = ((val >> 0) & 0xFF);
+ val = readl(gpmc_regs->gpmc_bch_result2[i]);
+ *(ecc_ptr++) = ((val >> 24) & 0xFF);
+ *(ecc_ptr++) = ((val >> 16) & 0xFF);
+ *(ecc_ptr++) = ((val >> 8) & 0xFF);
+ *(ecc_ptr++) = ((val >> 0) & 0xFF);
+ val = readl(gpmc_regs->gpmc_bch_result1[i]);
+ *(ecc_ptr++) = ((val >> 24) & 0xFF);
+ *(ecc_ptr++) = ((val >> 16) & 0xFF);
+ *(ecc_ptr++) = ((val >> 8) & 0xFF);
+ *(ecc_ptr++) = ((val >> 0) & 0xFF);
+ val = readl(gpmc_regs->gpmc_bch_result0[i]);
+ *(ecc_ptr++) = ((val >> 24) & 0xFF);
+ *(ecc_ptr++) = ((val >> 16) & 0xFF);
+ *(ecc_ptr++) = ((val >> 8) & 0xFF);
+ *(ecc_ptr++) = ((val >> 0) & 0xFF);
+ break;
+ default:
+ return -EINVAL;
}
-
- if (eccbchtsel) {
- /* BCH8 ecc scheme */
- *ecc_code++ = (bch_val4 & 0xFF);
- *ecc_code++ = ((bch_val3 >> 24) & 0xFF);
- *ecc_code++ = ((bch_val3 >> 16) & 0xFF);
- *ecc_code++ = ((bch_val3 >> 8) & 0xFF);
- *ecc_code++ = (bch_val3 & 0xFF);
- *ecc_code++ = ((bch_val2 >> 24) & 0xFF);
- *ecc_code++ = ((bch_val2 >> 16) & 0xFF);
- *ecc_code++ = ((bch_val2 >> 8) & 0xFF);
- *ecc_code++ = (bch_val2 & 0xFF);
- *ecc_code++ = ((bch_val1 >> 24) & 0xFF);
- *ecc_code++ = ((bch_val1 >> 16) & 0xFF);
- *ecc_code++ = ((bch_val1 >> 8) & 0xFF);
- *ecc_code++ = (bch_val1 & 0xFF);
- /*
- * Setting 14th byte to zero to handle
- * erased page & maintain compatibility
- * with RBL
- */
- *ecc_code++ = 0x0;
- } else {
- /* BCH4 ecc scheme */
- *ecc_code++ = ((bch_val2 >> 12) & 0xFF);
- *ecc_code++ = ((bch_val2 >> 4) & 0xFF);
- *ecc_code++ = ((bch_val2 & 0xF) << 4) |
- ((bch_val1 >> 28) & 0xF);
- *ecc_code++ = ((bch_val1 >> 20) & 0xFF);
- *ecc_code++ = ((bch_val1 >> 12) & 0xFF);
- *ecc_code++ = ((bch_val1 >> 4) & 0xFF);
- *ecc_code++ = ((bch_val1 & 0xF) << 4);
- /*
- * Setting 8th byte to zero to handle
- * erased page
- */
- *ecc_code++ = 0x0;
+ /* ECC scheme specific syndrome customizations */
+ switch (ecc_opt) {
+ case OMAP_ECC_BCH4_CODE_HW_DETECTION_SW:
+ for (i = 0; i < eccbytes; i++)
+ *(ecc_calc + i) = *(ecc_calc + i) ^
+ bch4_polynomial[i];
+ break;
+ case OMAP_ECC_BCH4_CODE_HW:
+ *(ecc_ptr++) = 0x00;
+ break;
+ case OMAP_ECC_BCH8_CODE_HW_DETECTION_SW:
+ for (i = 0; i < eccbytes; i++)
+ *(ecc_calc + i) = *(ecc_calc + i) ^
+ bch8_polynomial[i];
+ break;
+ case OMAP_ECC_BCH8_CODE_HW:
+ *(ecc_ptr++) = 0x00;
+ break;
+ default:
+ return -EINVAL;
}
+ /* update pointer to next sector */
+ ecc_calc += eccbytes;
}
-
return 0;
}
+#endif /* CONFIG_MTD_NAND_ECC_BCH || CONFIG_MTD_NAND_OMAP_BCH */
+#ifdef CONFIG_MTD_NAND_OMAP_BCH
/**
* omap_elm_correct_data - corrects page data area in case error reported
* @mtd: MTD device structure
@@ -1749,7 +1688,7 @@ static int omap_nand_probe(struct platform_device *pdev)
nand_chip->ecc.strength = 4;
nand_chip->ecc.hwctl = omap3_enable_hwecc_bch;
nand_chip->ecc.correct = nand_bch_correct_data;
- nand_chip->ecc.calculate = omap3_calculate_ecc_bch4;
+ nand_chip->ecc.calculate = omap_calculate_ecc_bch;
/* define ECC layout */
ecclayout->eccbytes = nand_chip->ecc.bytes *
(mtd->writesize /
@@ -1783,7 +1722,7 @@ static int omap_nand_probe(struct platform_device *pdev)
nand_chip->ecc.strength = 4;
nand_chip->ecc.hwctl = omap3_enable_hwecc_bch;
nand_chip->ecc.correct = omap_elm_correct_data;
- nand_chip->ecc.calculate = omap3_calculate_ecc_bch;
+ nand_chip->ecc.calculate = omap_calculate_ecc_bch;
nand_chip->ecc.read_page = omap_read_page_bch;
nand_chip->ecc.write_page = omap_write_page_bch;
/* define ECC layout */
@@ -1815,7 +1754,7 @@ static int omap_nand_probe(struct platform_device *pdev)
nand_chip->ecc.strength = 8;
nand_chip->ecc.hwctl = omap3_enable_hwecc_bch;
nand_chip->ecc.correct = nand_bch_correct_data;
- nand_chip->ecc.calculate = omap3_calculate_ecc_bch8;
+ nand_chip->ecc.calculate = omap_calculate_ecc_bch;
/* define ECC layout */
ecclayout->eccbytes = nand_chip->ecc.bytes *
(mtd->writesize /
@@ -1850,7 +1789,7 @@ static int omap_nand_probe(struct platform_device *pdev)
nand_chip->ecc.strength = 8;
nand_chip->ecc.hwctl = omap3_enable_hwecc_bch;
nand_chip->ecc.correct = omap_elm_correct_data;
- nand_chip->ecc.calculate = omap3_calculate_ecc_bch;
+ nand_chip->ecc.calculate = omap_calculate_ecc_bch;
nand_chip->ecc.read_page = omap_read_page_bch;
nand_chip->ecc.write_page = omap_write_page_bch;
/* This ECC scheme requires ELM H/W block */
--
1.8.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH v3 3/4] mtd: nand: omap: optimize chip->ecc.hwctl() for H/W ECC schemes
2013-11-02 9:46 [PATCH v3 0/4] optimize and clean-up of OMAP NAND and ELM driver Pekon Gupta
2013-11-02 9:46 ` [PATCH v3 1/4] mtd: nand: omap: optimized chip->ecc.correct() for H/W ECC schemes Pekon Gupta
2013-11-02 9:46 ` [PATCH v3 2/4] mtd: nand: omap: optimize chip->ecc.calculate() " Pekon Gupta
@ 2013-11-02 9:46 ` Pekon Gupta
2013-11-13 22:03 ` Brian Norris
2013-11-02 9:46 ` [PATCH v3 4/4] mtd: devices: elm: add checks ELM H/W constrains, driver code cleanup Pekon Gupta
3 siblings, 1 reply; 11+ messages in thread
From: Pekon Gupta @ 2013-11-02 9:46 UTC (permalink / raw)
To: computersforpeace, dedekind1; +Cc: linux-mtd, Pekon Gupta, balbi
chip->ecc.hwctl() is used for preparing the H/W controller before read/write
NAND accesses (like assigning data-buf, enabling ECC scheme configs, etc.)
Though all ECC schemes in OMAP NAND driver use GPMC controller for generating
ECC syndrome (for both Read/Write accesses). But but in current code
HAM1_ECC and BCHx_ECC schemes implement individual function to achieve this.
This patch merges the GPMC configuration code for all ECC schemes into
single omap_enable_hwecc(), thus adding scalability for future ECC schemes.
omap_enable_hwecc() + omap3_enable_hwecc_bch() -> omap_enable_hwecc()
Signed-off-by: Pekon Gupta <pekon@ti.com>
---
drivers/mtd/nand/omap2.c | 212 +++++++++++++++++------------------------------
1 file changed, 74 insertions(+), 138 deletions(-)
diff --git a/drivers/mtd/nand/omap2.c b/drivers/mtd/nand/omap2.c
index 1f59505..3a99e29 100644
--- a/drivers/mtd/nand/omap2.c
+++ b/drivers/mtd/nand/omap2.c
@@ -33,6 +33,10 @@
#define DRIVER_NAME "omap2-nand"
#define OMAP_NAND_TIMEOUT_MS 5000
+#define GPMC_ECC_READ 0 /* Reset Hardware ECC for read */
+#define GPMC_ECC_WRITE 1 /* Reset Hardware ECC for write */
+#define GPMC_ECC_READSYN 2 /* Reset before syndrom is read back */
+
#define NAND_Ecc_P1e (1 << 0)
#define NAND_Ecc_P2e (1 << 1)
#define NAND_Ecc_P4e (1 << 2)
@@ -101,13 +105,9 @@
#define P4o_s(a) (TF(a & NAND_Ecc_P4o) << 1)
#define PREFETCH_CONFIG1_CS_SHIFT 24
-#define ECC_CONFIG_CS_SHIFT 1
#define CS_MASK 0x7
#define ENABLE_PREFETCH (0x1 << 7)
#define DMA_MPU_MODE_SHIFT 2
-#define ECCSIZE0_SHIFT 12
-#define ECCSIZE1_SHIFT 22
-#define ECC1RESULTSIZE 0x1
#define ECCCLEAR 0x100
#define ECC1 0x1
#define PREFETCH_FIFOTHRESHOLD_MAX 0x40
@@ -118,26 +118,9 @@
#define OMAP24XX_DMA_GPMC 4
-#define BCH8_MAX_ERROR 8 /* upto 8 bit correctable */
-#define BCH4_MAX_ERROR 4 /* upto 4 bit correctable */
-
#define SECTOR_BYTES 512
/* 4 bit padding to make byte aligned, 56 = 52 + 4 */
#define BCH4_BIT_PAD 4
-#define BCH8_ECC_MAX ((SECTOR_BYTES + BCH8_ECC_OOB_BYTES) * 8)
-#define BCH4_ECC_MAX ((SECTOR_BYTES + BCH4_ECC_OOB_BYTES) * 8)
-
-/* GPMC ecc engine settings for read */
-#define BCH_WRAPMODE_1 1 /* BCH wrap mode 1 */
-#define BCH8R_ECC_SIZE0 0x1a /* ecc_size0 = 26 */
-#define BCH8R_ECC_SIZE1 0x2 /* ecc_size1 = 2 */
-#define BCH4R_ECC_SIZE0 0xd /* ecc_size0 = 13 */
-#define BCH4R_ECC_SIZE1 0x3 /* ecc_size1 = 3 */
-
-/* GPMC ecc engine settings for write */
-#define BCH_WRAPMODE_6 6 /* BCH wrap mode 6 */
-#define BCH_ECC_SIZE0 0x0 /* ecc_size0 = 0, no oob protection */
-#define BCH_ECC_SIZE1 0x20 /* ecc_size1 = 32 */
#define BADBLOCK_MARKER_LENGTH 2
@@ -177,7 +160,6 @@ struct omap_nand_info {
int buf_len;
struct gpmc_nand_regs reg;
/* fields specific for BCHx_HW ECC scheme */
- bool is_elm_used;
struct device *elm_dev;
struct device_node *of_node;
};
@@ -938,7 +920,7 @@ static int omap_calculate_ecc(struct mtd_info *mtd, const u_char *dat,
u32 val;
val = readl(info->reg.gpmc_ecc_config);
- if (((val >> 1) & 0x07) != info->gpmc_cs) {
+ if (((val >> 1) & 0x7) != info->gpmc_cs) {
pr_err("%s: invalid ECC configuration for chip-select=%d",
DRIVER_NAME, info->gpmc_cs);
return -EINVAL;
@@ -954,47 +936,6 @@ static int omap_calculate_ecc(struct mtd_info *mtd, const u_char *dat,
}
/**
- * omap_enable_hwecc - This function enables the hardware ecc functionality
- * @mtd: MTD device structure
- * @mode: Read/Write mode
- */
-static void omap_enable_hwecc(struct mtd_info *mtd, int mode)
-{
- struct omap_nand_info *info = container_of(mtd, struct omap_nand_info,
- mtd);
- struct nand_chip *chip = mtd->priv;
- unsigned int dev_width = (chip->options & NAND_BUSWIDTH_16) ? 1 : 0;
- u32 val;
-
- /* clear ecc and enable bits */
- val = ECCCLEAR | ECC1;
- writel(val, info->reg.gpmc_ecc_control);
-
- /* program ecc and result sizes */
- val = ((((info->nand.ecc.size >> 1) - 1) << ECCSIZE1_SHIFT) |
- ECC1RESULTSIZE);
- writel(val, info->reg.gpmc_ecc_size_config);
-
- switch (mode) {
- case NAND_ECC_READ:
- case NAND_ECC_WRITE:
- writel(ECCCLEAR | ECC1, info->reg.gpmc_ecc_control);
- break;
- case NAND_ECC_READSYN:
- writel(ECCCLEAR, info->reg.gpmc_ecc_control);
- break;
- default:
- dev_info(&info->pdev->dev,
- "error: unrecognized Mode[%d]!\n", mode);
- break;
- }
-
- /* (ECC 16 or 8 bit col) | ( CS ) | ECC Enable */
- val = (dev_width << 7) | (info->gpmc_cs << 1) | (0x1);
- writel(val, info->reg.gpmc_ecc_config);
-}
-
-/**
* omap_wait - wait until the command is done
* @mtd: MTD device structure
* @chip: NAND Chip structure
@@ -1050,88 +991,85 @@ static int omap_dev_ready(struct mtd_info *mtd)
}
}
-#if defined(CONFIG_MTD_NAND_ECC_BCH) || defined(CONFIG_MTD_NAND_OMAP_BCH)
/**
- * omap3_enable_hwecc_bch - Program OMAP3 GPMC to perform BCH ECC correction
+ * omap_enable_hwecc - Configure OMAP GPMC to perform ECC calculation
* @mtd: MTD device structure
* @mode: Read/Write mode
- *
- * When using BCH, sector size is hardcoded to 512 bytes.
- * Using wrapping mode 6 both for reading and writing if ELM module not uses
- * for error correction.
- * On writing,
- * eccsize0 = 0 (no additional protected byte in spare area)
- * eccsize1 = 32 (skip 32 nibbles = 16 bytes per sector in spare area)
+ * Configurations for eccsize0, eccsize1, and bch_wrapmode are based on
+ * GPMC function spec:
+ * Section 4.6.3.2.3: Supported NAND page mappings and ECC schemes
*/
-static void omap3_enable_hwecc_bch(struct mtd_info *mtd, int mode)
+static void omap_enable_hwecc(struct mtd_info *mtd, int mode)
{
- int nerrors;
- unsigned int dev_width, nsectors;
struct omap_nand_info *info = container_of(mtd, struct omap_nand_info,
mtd);
struct nand_chip *chip = mtd->priv;
- u32 val, wr_mode;
- unsigned int ecc_size1, ecc_size0;
-
- /* Using wrapping mode 6 for writing */
- wr_mode = BCH_WRAPMODE_6;
-
- /*
- * ECC engine enabled for valid ecc_size0 nibbles
- * and disabled for ecc_size1 nibbles.
- */
- ecc_size0 = BCH_ECC_SIZE0;
- ecc_size1 = BCH_ECC_SIZE1;
-
- /* Perform ecc calculation on 512-byte sector */
- nsectors = 1;
-
- /* Update number of error correction */
- nerrors = info->nand.ecc.strength;
-
- /* Multi sector reading/writing for NAND flash with page size < 4096 */
- if (info->is_elm_used && (mtd->writesize <= 4096)) {
- if (mode == NAND_ECC_READ) {
- /* Using wrapping mode 1 for reading */
- wr_mode = BCH_WRAPMODE_1;
-
- /*
- * ECC engine enabled for ecc_size0 nibbles
- * and disabled for ecc_size1 nibbles.
- */
- ecc_size0 = (nerrors == 8) ?
- BCH8R_ECC_SIZE0 : BCH4R_ECC_SIZE0;
- ecc_size1 = (nerrors == 8) ?
- BCH8R_ECC_SIZE1 : BCH4R_ECC_SIZE1;
+ unsigned int dev_width = (chip->options & NAND_BUSWIDTH_16) ? 1 : 0;
+ unsigned int nsectors = (mtd->writesize / SECTOR_BYTES);
+ unsigned int ecc_algo = 0;
+ unsigned int bch_type = 0;
+ unsigned int eccsize1 = 0x00, eccsize0 = 0x00, bch_wrapmode = 0x00;
+ u32 ecc_size_config_val = 0;
+ u32 ecc_config_val = 0;
+
+ switch (info->ecc_opt) {
+ case OMAP_ECC_HAM1_CODE_HW:
+ ecc_algo = 0x0;
+ bch_wrapmode = 0x00;
+ eccsize0 = (chip->ecc.size >> 1) - 1;
+ eccsize1 = 0;
+ nsectors = 0;
+ break;
+ case OMAP_ECC_BCH4_CODE_HW_DETECTION_SW:
+ case OMAP_ECC_BCH4_CODE_HW:
+ ecc_algo = 0x1;
+ bch_type = 0x0;
+ if (mode == GPMC_ECC_READ) {
+ bch_wrapmode = 0x01;
+ eccsize0 = 13; /* ECC bits in nibbles per sector */
+ eccsize1 = 3; /* non-ECC bits in nibbles per sector */
+ } else if (mode == GPMC_ECC_WRITE) {
+ eccsize0 = 0; /* extra bits in nibbles per sector */
+ eccsize1 = 32; /* OOB bits in nibbles per sector */
+ bch_wrapmode = 0x06;
}
-
- /* Perform ecc calculation for one page (< 4096) */
- nsectors = info->nand.ecc.steps;
+ break;
+ case OMAP_ECC_BCH8_CODE_HW_DETECTION_SW:
+ case OMAP_ECC_BCH8_CODE_HW:
+ ecc_algo = 0x1;
+ bch_type = 0x1;
+ if (mode == GPMC_ECC_READ) {
+ bch_wrapmode = 0x01;
+ eccsize0 = 26; /* ECC bits in nibbles per sector */
+ eccsize1 = 2; /* non-ECC bits in nibbles per sector */
+ } else if (mode == GPMC_ECC_WRITE) {
+ bch_wrapmode = 0x01;
+ eccsize0 = 0; /* extra bits in nibbles per sector */
+ eccsize1 = 28; /* OOB bits in nibbles per sector */
+ }
+ break;
+ default:
+ pr_err("selected ECC scheme not supported or not enabled\n");
}
-
- writel(ECC1, info->reg.gpmc_ecc_control);
-
+ /* clear previous ecc result and enable engine */
+ writel(ECCCLEAR | ECC1, info->reg.gpmc_ecc_control);
/* Configure ecc size for BCH */
- val = (ecc_size1 << ECCSIZE1_SHIFT) | (ecc_size0 << ECCSIZE0_SHIFT);
- writel(val, info->reg.gpmc_ecc_size_config);
-
- dev_width = (chip->options & NAND_BUSWIDTH_16) ? 1 : 0;
-
- /* BCH configuration */
- val = ((1 << 16) | /* enable BCH */
- (((nerrors == 8) ? 1 : 0) << 12) | /* 8 or 4 bits */
- (wr_mode << 8) | /* wrap mode */
- (dev_width << 7) | /* bus width */
- (((nsectors-1) & 0x7) << 4) | /* number of sectors */
- (info->gpmc_cs << 1) | /* ECC CS */
- (0x1)); /* enable ECC */
-
- writel(val, info->reg.gpmc_ecc_config);
-
- /* Clear ecc and enable bits */
+ ecc_size_config_val = (eccsize1 << 22) | (eccsize0 << 12);
+ writel(ecc_size_config_val, info->reg.gpmc_ecc_size_config);
+ /* Configure device details for BCH engine */
+ ecc_config_val = ((ecc_algo << 16) | /* HAM1 | BCHx */
+ (bch_type << 12) | /* BCH4/BCH8/BCH16 */
+ (bch_wrapmode << 8) | /* wrap mode */
+ (dev_width << 7) | /* bus width */
+ (((nsectors-1) & 0x7) << 4) | /* number of sectors */
+ (info->gpmc_cs << 1) | /* ECC CS */
+ (0x1)); /* enable ECC */
+ writel(ecc_config_val, info->reg.gpmc_ecc_config);
+ /* Clear ecc and re-enable */
writel(ECCCLEAR | ECC1, info->reg.gpmc_ecc_control);
}
+#if defined(CONFIG_MTD_NAND_ECC_BCH) || defined(CONFIG_MTD_NAND_OMAP_BCH)
/**
* omap_calculate_ecc_bch - Generate bytes of ECC bytes
* @mtd: MTD device structure
@@ -1440,7 +1378,6 @@ static int is_elm_present(struct omap_nand_info *info,
struct device_node *elm_node, enum bch_ecc bch_type)
{
struct platform_device *pdev;
- info->is_elm_used = false;
/* check whether elm-id is passed via DT */
if (!elm_node) {
pr_err("nand: error: ELM DT node not found\n");
@@ -1456,7 +1393,6 @@ static int is_elm_present(struct omap_nand_info *info,
info->elm_dev = &pdev->dev;
if (elm_config(info->elm_dev, bch_type))
return -ENODEV;
- info->is_elm_used = true;
return 0;
}
#endif /* CONFIG_MTD_NAND_ECC_BCH */
@@ -1686,7 +1622,7 @@ static int omap_nand_probe(struct platform_device *pdev)
nand_chip->ecc.size = 512;
nand_chip->ecc.bytes = 7;
nand_chip->ecc.strength = 4;
- nand_chip->ecc.hwctl = omap3_enable_hwecc_bch;
+ nand_chip->ecc.hwctl = omap_enable_hwecc;
nand_chip->ecc.correct = nand_bch_correct_data;
nand_chip->ecc.calculate = omap_calculate_ecc_bch;
/* define ECC layout */
@@ -1720,7 +1656,7 @@ static int omap_nand_probe(struct platform_device *pdev)
/* 14th bit is kept reserved for ROM-code compatibility */
nand_chip->ecc.bytes = 7 + 1;
nand_chip->ecc.strength = 4;
- nand_chip->ecc.hwctl = omap3_enable_hwecc_bch;
+ nand_chip->ecc.hwctl = omap_enable_hwecc;
nand_chip->ecc.correct = omap_elm_correct_data;
nand_chip->ecc.calculate = omap_calculate_ecc_bch;
nand_chip->ecc.read_page = omap_read_page_bch;
@@ -1752,7 +1688,7 @@ static int omap_nand_probe(struct platform_device *pdev)
nand_chip->ecc.size = 512;
nand_chip->ecc.bytes = 13;
nand_chip->ecc.strength = 8;
- nand_chip->ecc.hwctl = omap3_enable_hwecc_bch;
+ nand_chip->ecc.hwctl = omap_enable_hwecc;
nand_chip->ecc.correct = nand_bch_correct_data;
nand_chip->ecc.calculate = omap_calculate_ecc_bch;
/* define ECC layout */
@@ -1787,7 +1723,7 @@ static int omap_nand_probe(struct platform_device *pdev)
/* 14th bit is kept reserved for ROM-code compatibility */
nand_chip->ecc.bytes = 13 + 1;
nand_chip->ecc.strength = 8;
- nand_chip->ecc.hwctl = omap3_enable_hwecc_bch;
+ nand_chip->ecc.hwctl = omap_enable_hwecc;
nand_chip->ecc.correct = omap_elm_correct_data;
nand_chip->ecc.calculate = omap_calculate_ecc_bch;
nand_chip->ecc.read_page = omap_read_page_bch;
--
1.8.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH v3 4/4] mtd: devices: elm: add checks ELM H/W constrains, driver code cleanup
2013-11-02 9:46 [PATCH v3 0/4] optimize and clean-up of OMAP NAND and ELM driver Pekon Gupta
` (2 preceding siblings ...)
2013-11-02 9:46 ` [PATCH v3 3/4] mtd: nand: omap: optimize chip->ecc.hwctl() " Pekon Gupta
@ 2013-11-02 9:46 ` Pekon Gupta
2013-11-13 22:37 ` Brian Norris
3 siblings, 1 reply; 11+ messages in thread
From: Pekon Gupta @ 2013-11-02 9:46 UTC (permalink / raw)
To: computersforpeace, dedekind1; +Cc: linux-mtd, Pekon Gupta, balbi
ELM H/W engine is used by BCHx_ECC schemes for detecting and locating bit-flips.
However, ELM H/W engine has some constrains like:
- ELM can decode errors in chunks of 512 data bytes only
- ELM can operate max upto 8 such buffers in parallel
This patch
- add checks for above constrains
- fixes ELM register configs based on number of info->eccsteps
- cleans-up elm_load_syndrome()
Signed-off-by: Pekon Gupta <pekon@ti.com>
---
drivers/mtd/devices/elm.c | 122 ++++++++++++++++++++++----------------
drivers/mtd/nand/omap2.c | 2 +-
include/linux/platform_data/elm.h | 6 +-
3 files changed, 77 insertions(+), 53 deletions(-)
diff --git a/drivers/mtd/devices/elm.c b/drivers/mtd/devices/elm.c
index d1dd6a3..2167384 100644
--- a/drivers/mtd/devices/elm.c
+++ b/drivers/mtd/devices/elm.c
@@ -22,8 +22,11 @@
#include <linux/of.h>
#include <linux/sched.h>
#include <linux/pm_runtime.h>
+#include <linux/mtd/mtd.h>
+#include <linux/mtd/nand.h>
#include <linux/platform_data/elm.h>
+#define DRIVER_NAME "omap-elm"
#define ELM_SYSCONFIG 0x010
#define ELM_IRQSTATUS 0x018
#define ELM_IRQENABLE 0x01c
@@ -82,8 +85,10 @@ struct elm_info {
void __iomem *elm_base;
struct completion elm_completion;
struct list_head list;
+ struct mtd_info *mtd;
enum bch_ecc bch_type;
struct elm_registers elm_regs;
+ int eccsteps;
};
static LIST_HEAD(elm_devices);
@@ -103,19 +108,42 @@ static u32 elm_read_reg(struct elm_info *info, int offset)
* @dev: ELM device
* @bch_type: Type of BCH ecc
*/
-int elm_config(struct device *dev, enum bch_ecc bch_type)
+int elm_config(struct device *dev, struct mtd_info *mtd,
+ enum bch_ecc bch_type)
{
u32 reg_val;
- struct elm_info *info = dev_get_drvdata(dev);
-
+ struct elm_info *info;
+ struct nand_chip *chip;
+ if (!dev) {
+ pr_err("%s: ELM device not found\n", DRIVER_NAME);
+ return -ENODEV;
+ }
+ info = dev_get_drvdata(dev);
if (!info) {
- dev_err(dev, "Unable to configure elm - device not probed?\n");
+ pr_err("%s: ELM device data not found\n", DRIVER_NAME);
return -ENODEV;
}
-
+ if (!mtd) {
+ pr_err("%s: MTD device not found\n", DRIVER_NAME);
+ return -ENODEV;
+ }
+ chip = mtd->priv;
+ /* ELM supports error correction in chunks of 512bytes of data only
+ * where each 512bytes of data has its own ECC syndrome */
+ if (chip->ecc.size != 512) {
+ pr_err("%s: invalid ecc_size configuration", DRIVER_NAME);
+ return -EINVAL;
+ }
+ if (mtd->writesize > 4096) {
+ pr_err("%s: page-size > 4096 is not supported", DRIVER_NAME);
+ return -EINVAL;
+ }
+ /* ELM eccsteps required to decode complete NAND page */
+ info->mtd = mtd;
+ info->bch_type = bch_type;
+ info->eccsteps = mtd->writesize / chip->ecc.size;
reg_val = (bch_type & ECC_BCH_LEVEL_MASK) | (ELM_ECC_SIZE << 16);
elm_write_reg(info, ELM_LOCATION_CONFIG, reg_val);
- info->bch_type = bch_type;
return 0;
}
@@ -152,55 +180,51 @@ static void elm_configure_page_mode(struct elm_info *info, int index,
* Load syndrome fragment registers with calculated ecc in reverse order.
*/
static void elm_load_syndrome(struct elm_info *info,
- struct elm_errorvec *err_vec, u8 *ecc)
+ struct elm_errorvec *err_vec, u8 *ecc_calc)
{
+ struct nand_chip *chip = info->mtd->priv;
+ unsigned int eccbytes = chip->ecc.bytes;
+ u8 *ecc = ecc_calc;
int i, offset;
u32 val;
- for (i = 0; i < ERROR_VECTOR_MAX; i++) {
-
+ for (i = 0; i < info->eccsteps; i++) {
/* Check error reported */
if (err_vec[i].error_reported) {
elm_configure_page_mode(info, i, true);
- offset = ELM_SYNDROME_FRAGMENT_0 +
- SYNDROME_FRAGMENT_REG_SIZE * i;
-
- /* BCH8 */
- if (info->bch_type) {
-
- /* syndrome fragment 0 = ecc[9-12B] */
- val = cpu_to_be32(*(u32 *) &ecc[9]);
- elm_write_reg(info, offset, val);
-
- /* syndrome fragment 1 = ecc[5-8B] */
- offset += 4;
- val = cpu_to_be32(*(u32 *) &ecc[5]);
- elm_write_reg(info, offset, val);
-
- /* syndrome fragment 2 = ecc[1-4B] */
- offset += 4;
- val = cpu_to_be32(*(u32 *) &ecc[1]);
- elm_write_reg(info, offset, val);
-
- /* syndrome fragment 3 = ecc[0B] */
- offset += 4;
- val = ecc[0];
- elm_write_reg(info, offset, val);
- } else {
- /* syndrome fragment 0 = ecc[20-52b] bits */
- val = (cpu_to_be32(*(u32 *) &ecc[3]) >> 4) |
- ((ecc[2] & 0xf) << 28);
- elm_write_reg(info, offset, val);
-
- /* syndrome fragment 1 = ecc[0-20b] bits */
- offset += 4;
- val = cpu_to_be32(*(u32 *) &ecc[0]) >> 12;
- elm_write_reg(info, offset, val);
+ offset = SYNDROME_FRAGMENT_REG_SIZE * i;
+ ecc = ecc_calc + (i * eccbytes);
+ switch (info->bch_type) {
+ case BCH4_ECC:
+ val = ((*(ecc + 6) >> 4) & 0x0F) |
+ *(ecc + 5) << 4 | *(ecc + 4) << 12 |
+ *(ecc + 3) << 20 | *(ecc + 2) << 28;
+ elm_write_reg(info, (ELM_SYNDROME_FRAGMENT_0 +
+ offset), cpu_to_le32(val));
+ val = ((*(ecc + 2) >> 4) & 0x0F) |
+ *(ecc + 1) << 4 | *(ecc + 0) << 12;
+ elm_write_reg(info, (ELM_SYNDROME_FRAGMENT_1 +
+ offset), cpu_to_le32(val));
+ break;
+ case BCH8_ECC:
+ val = *(ecc + 12) << 0 | *(ecc + 11) << 8 |
+ *(ecc + 10) << 16 | *(ecc + 9) << 24;
+ elm_write_reg(info, (ELM_SYNDROME_FRAGMENT_0 +
+ offset), cpu_to_le32(val));
+ val = *(ecc + 8) << 0 | *(ecc + 7) << 8 |
+ *(ecc + 6) << 16 | *(ecc + 5) << 24;
+ elm_write_reg(info, (ELM_SYNDROME_FRAGMENT_1 +
+ offset), cpu_to_le32(val));
+ val = *(ecc + 4) << 0 | *(ecc + 3) << 8 |
+ *(ecc + 2) << 16 | *(ecc + 1) << 24;
+ elm_write_reg(info, (ELM_SYNDROME_FRAGMENT_2 +
+ offset), cpu_to_le32(val));
+ val = *(ecc + 0) << 0 & 0x000000FF;
+ elm_write_reg(info, (ELM_SYNDROME_FRAGMENT_3 +
+ offset), cpu_to_le32(val));
+ break;
}
}
-
- /* Update ecc pointer with ecc byte size */
- ecc += info->bch_type ? BCH8_SIZE : BCH4_SIZE;
}
}
@@ -223,7 +247,7 @@ static void elm_start_processing(struct elm_info *info,
* Set syndrome vector valid, so that ELM module
* will process it for vectors error is reported
*/
- for (i = 0; i < ERROR_VECTOR_MAX; i++) {
+ for (i = 0; i < info->eccsteps; i++) {
if (err_vec[i].error_reported) {
offset = ELM_SYNDROME_FRAGMENT_6 +
SYNDROME_FRAGMENT_REG_SIZE * i;
@@ -252,7 +276,7 @@ static void elm_error_correction(struct elm_info *info,
int offset;
u32 reg_val;
- for (i = 0; i < ERROR_VECTOR_MAX; i++) {
+ for (i = 0; i < info->eccsteps; i++) {
/* Check error reported */
if (err_vec[i].error_reported) {
@@ -263,14 +287,12 @@ static void elm_error_correction(struct elm_info *info,
if (reg_val & ECC_CORRECTABLE_MASK) {
offset = ELM_ERROR_LOCATION_0 +
ERROR_LOCATION_SIZE * i;
-
/* Read count of correctable errors */
err_vec[i].error_count = reg_val &
ECC_NB_ERRORS_MASK;
/* Update the error locations in error vector */
for (j = 0; j < err_vec[i].error_count; j++) {
-
reg_val = elm_read_reg(info, offset);
err_vec[i].error_loc[j] = reg_val &
ECC_ERROR_LOCATION_MASK;
diff --git a/drivers/mtd/nand/omap2.c b/drivers/mtd/nand/omap2.c
index 3a99e29..037e9e7 100644
--- a/drivers/mtd/nand/omap2.c
+++ b/drivers/mtd/nand/omap2.c
@@ -1391,7 +1391,7 @@ static int is_elm_present(struct omap_nand_info *info,
}
/* ELM module available, now configure it */
info->elm_dev = &pdev->dev;
- if (elm_config(info->elm_dev, bch_type))
+ if (elm_config(info->elm_dev, &info->mtd, bch_type))
return -ENODEV;
return 0;
}
diff --git a/include/linux/platform_data/elm.h b/include/linux/platform_data/elm.h
index bf0a83b..d16465b 100644
--- a/include/linux/platform_data/elm.h
+++ b/include/linux/platform_data/elm.h
@@ -25,6 +25,7 @@ enum bch_ecc {
/* ELM support 8 error syndrome process */
#define ERROR_VECTOR_MAX 8
+#define ELM_MAX_DETECTABLE_ERRORS 16
#define BCH8_ECC_OOB_BYTES 13
#define BCH4_ECC_OOB_BYTES 7
@@ -45,10 +46,11 @@ struct elm_errorvec {
bool error_reported;
bool error_uncorrectable;
int error_count;
- int error_loc[ERROR_VECTOR_MAX];
+ int error_loc[ELM_MAX_DETECTABLE_ERRORS];
};
void elm_decode_bch_error_page(struct device *dev, u8 *ecc_calc,
struct elm_errorvec *err_vec);
-int elm_config(struct device *dev, enum bch_ecc bch_type);
+int elm_config(struct device *dev, struct mtd_info *mtd,
+ enum bch_ecc bch_type);
#endif /* __ELM_H */
--
1.8.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH v3 1/4] mtd: nand: omap: optimized chip->ecc.correct() for H/W ECC schemes
2013-11-02 9:46 ` [PATCH v3 1/4] mtd: nand: omap: optimized chip->ecc.correct() for H/W ECC schemes Pekon Gupta
@ 2013-11-13 21:11 ` Brian Norris
2013-11-14 5:13 ` Gupta, Pekon
0 siblings, 1 reply; 11+ messages in thread
From: Brian Norris @ 2013-11-13 21:11 UTC (permalink / raw)
To: Pekon Gupta; +Cc: Philip Avinash, Ivan Djelic, linux-mtd, balbi, dedekind1
+ Philip and Ivan, who did relevant work
On Sat, Nov 02, 2013 at 03:16:13PM +0530, Pekon Gupta wrote:
> chip->ecc.correct() is used for detecting and correcting bit-flips during read
> operations. In omap2-nand driver this is done usingt following functions:
>
> - omap_correct_data(): for H/W based HAM1_ECC schemes
> (Un-Touched in current patch)
>
> - omap_elm_correct_data(): for H/W based BCHx_ECC scheme
> Current implementation of this function is not scalable for newer ECC
> schemes because:
> - It depends on a specific byte-position in OOB area (reserved as 0x00)
> to differentiates between programmed-pages and erased-pages.
> This reserved byte-position cannot be accomodated in all ECC schemes.
> - Current code is not scalable for future ECC schemes due to tweaks for
> BCH4_ECC and BCH8_ECC at multiple places.
> - It checks for bit-flips in Erased-pages using check_erased_page().
> This is over-work, as sanity of Erased-page can be verified by just
> comparing them to a pre-defined ECC-syndrome for all_0xFF data.
>
> This patch optimizes omap_elm_correct_data() in following ways:
> (1) Removes dependency on specific reserved-byte (0x00) in OOB area,
> instead Erased-page is identified by matching calc_ecc with a
> pre-defined ECC syndrome of all(0xFF) data
> (2) merges common code for BCH4_ECC and BCH8_ECC for scalability.
> (3) handles incorrect elm_error_location beyond data+oob buffer.
> (4) removes check_erased_page(): Bit-flips in erased-page are handled
> in same way as for programmed-page
Are you sure that this new scheme is equivalent to the old? i.e., is
this backwards compatible? Are you totally dropping the "reserved byte"
marker? Is this compatible with your bootloaders? I don't have a very
clear mental picture of exactly what your driver is doing to check
programmed/erased pages, and I'm certainly not familiar with your
bootloader/ROM code.
>
> Signed-off-by: Pekon Gupta <pekon@ti.com>
> ---
> drivers/mtd/nand/omap2.c | 247 ++++++++++++++---------------------------------
> 1 file changed, 74 insertions(+), 173 deletions(-)
>
> diff --git a/drivers/mtd/nand/omap2.c b/drivers/mtd/nand/omap2.c
> index ec40b8d..c946f22 100644
> --- a/drivers/mtd/nand/omap2.c
> +++ b/drivers/mtd/nand/omap2.c
> @@ -160,6 +160,7 @@ struct omap_nand_info {
> int gpmc_cs;
> unsigned long phys_base;
> unsigned long mem_size;
> + enum omap_ecc ecc_opt;
> struct completion comp;
> struct dma_chan *dma;
> int gpmc_irq_fifo;
> @@ -1291,219 +1292,118 @@ static int omap3_calculate_ecc_bch(struct mtd_info *mtd, const u_char *dat,
> }
>
> /**
> - * erased_sector_bitflips - count bit flips
> - * @data: data sector buffer
> - * @oob: oob buffer
> - * @info: omap_nand_info
> - *
> - * Check the bit flips in erased page falls below correctable level.
> - * If falls below, report the page as erased with correctable bit
> - * flip, else report as uncorrectable page.
> - */
> -static int erased_sector_bitflips(u_char *data, u_char *oob,
> - struct omap_nand_info *info)
> -{
> - int flip_bits = 0, i;
> -
> - for (i = 0; i < info->nand.ecc.size; i++) {
> - flip_bits += hweight8(~data[i]);
> - if (flip_bits > info->nand.ecc.strength)
> - return 0;
> - }
> -
> - for (i = 0; i < info->nand.ecc.bytes - 1; i++) {
> - flip_bits += hweight8(~oob[i]);
> - if (flip_bits > info->nand.ecc.strength)
> - return 0;
> - }
> -
> - /*
> - * Bit flips falls in correctable level.
> - * Fill data area with 0xFF
> - */
> - if (flip_bits) {
> - memset(data, 0xFF, info->nand.ecc.size);
> - memset(oob, 0xFF, info->nand.ecc.bytes);
> - }
> -
> - return flip_bits;
> -}
> -
> -/**
> * omap_elm_correct_data - corrects page data area in case error reported
> * @mtd: MTD device structure
> * @data: page data
> * @read_ecc: ecc read from nand flash
> - * @calc_ecc: ecc read from HW ECC registers
> - *
> - * Calculated ecc vector reported as zero in case of non-error pages.
> - * In case of error/erased pages non-zero error vector is reported.
> - * In case of non-zero ecc vector, check read_ecc at fixed offset
> - * (x = 13/7 in case of BCH8/4 == 0) to find page programmed or not.
> - * To handle bit flips in this data, count the number of 0's in
> - * read_ecc[x] and check if it greater than 4. If it is less, it is
> - * programmed page, else erased page.
> - *
> - * 1. If page is erased, check with standard ecc vector (ecc vector
> - * for erased page to find any bit flip). If check fails, bit flip
> - * is present in erased page. Count the bit flips in erased page and
> - * if it falls under correctable level, report page with 0xFF and
> - * update the correctable bit information.
> - * 2. If error is reported on programmed page, update elm error
> - * vector and correct the page with ELM error correction routine.
> - *
> + * @calc_ecc: ecc calculated after reading Data and OOB regions from flash
> + * As calc_ecc is calculated over both main & oob, so calc_ecc would be
> + * non-zero only in following cases:
> + * - bit-flips in data or oob region
> + * - erase page, where no ECC is written in OOB area
> + * However, erased_pages can be differentiated from corrupted pages
> + * by comparing the calculated ECC with pre-defined syndrome ECC_of_ALL(0xFF)
> + * Bit-flips in erased-pages would also be caught by comparing, calc_ecc
> + * with ECC_of_ALL(0xFF)
> */
> static int omap_elm_correct_data(struct mtd_info *mtd, u_char *data,
> u_char *read_ecc, u_char *calc_ecc)
> {
> struct omap_nand_info *info = container_of(mtd, struct omap_nand_info,
> mtd);
> - int eccsteps = info->nand.ecc.steps;
> - int i , j, stat = 0;
> - int eccsize, eccflag, ecc_vector_size;
> + enum omap_ecc ecc_opt = info->ecc_opt;
> + struct nand_chip *chip = mtd->priv;
> + int eccsteps = chip->ecc.steps;
> + int eccsize = chip->ecc.size;
> + int eccbytes = chip->ecc.bytes;
> + int i , j, stat = 0, ret = 0, flag_read_ecc;
> struct elm_errorvec err_vec[ERROR_VECTOR_MAX];
> - u_char *ecc_vec = calc_ecc;
> - u_char *spare_ecc = read_ecc;
> - u_char *erased_ecc_vec;
> - enum bch_ecc type;
> + u_char *ecc;
> bool is_error_reported = false;
> + u32 bit_pos, byte_pos, error_max, pos;
>
> /* Initialize elm error vector to zero */
> memset(err_vec, 0, sizeof(err_vec));
>
> - if (info->nand.ecc.strength == BCH8_MAX_ERROR) {
> - type = BCH8_ECC;
> - erased_ecc_vec = bch8_vector;
> - } else {
> - type = BCH4_ECC;
> - erased_ecc_vec = bch4_vector;
> - }
> -
> - ecc_vector_size = info->nand.ecc.bytes;
> -
> - /*
> - * Remove extra byte padding for BCH8 RBL
> - * compatibility and erased page handling
> - */
> - eccsize = ecc_vector_size - 1;
> -
> for (i = 0; i < eccsteps ; i++) {
> - eccflag = 0; /* initialize eccflag */
> -
> - /*
> - * Check any error reported,
> - * In case of error, non zero ecc reported.
> - */
> -
> - for (j = 0; (j < eccsize); j++) {
> - if (calc_ecc[j] != 0) {
> - eccflag = 1; /* non zero ecc, error present */
> + flag_read_ecc = 0;
> + ecc = calc_ecc + (i * eccbytes);
> + /* check calc_ecc */
> + for (j = 0; j < eccbytes; j++) {
> + if (*(ecc + j) != 0x00) {
> + flag_read_ecc = 1;
> break;
> }
> }
> -
> - if (eccflag == 1) {
> - /*
> - * Set threshold to minimum of 4, half of ecc.strength/2
> - * to allow max bit flip in byte to 4
> - */
> - unsigned int threshold = min_t(unsigned int, 4,
> - info->nand.ecc.strength / 2);
> -
> - /*
> - * Check data area is programmed by counting
> - * number of 0's at fixed offset in spare area.
> - * Checking count of 0's against threshold.
> - * In case programmed page expects at least threshold
> - * zeros in byte.
> - * If zeros are less than threshold for programmed page/
> - * zeros are more than threshold erased page, either
> - * case page reported as uncorrectable.
> - */
> - if (hweight8(~read_ecc[eccsize]) >= threshold) {
> - /*
> - * Update elm error vector as
> - * data area is programmed
> - */
> - err_vec[i].error_reported = true;
> - is_error_reported = true;
> - } else {
> - /* Error reported in erased page */
> - int bitflip_count;
> - u_char *buf = &data[info->nand.ecc.size * i];
> -
> - if (memcmp(calc_ecc, erased_ecc_vec, eccsize)) {
> - bitflip_count = erased_sector_bitflips(
> - buf, read_ecc, info);
> -
> - if (bitflip_count)
> - stat += bitflip_count;
> - else
> - return -EINVAL;
> - }
> + /* check if its a erased-page */
> + if (flag_read_ecc) {
> + switch (ecc_opt) {
> + case OMAP_ECC_BCH8_CODE_HW:
> + if (memcmp(ecc, bch8_vector, eccbytes))
> + err_vec[i].error_reported = true;
> + break;
> + case OMAP_ECC_BCH4_CODE_HW:
> + if (memcmp(ecc, bch4_vector, eccbytes))
> + err_vec[i].error_reported = true;
> + break;
> + default:
> + pr_err("%s: invalid configuration",
> + DRIVER_NAME);
> + return -EINVAL;
> }
> }
> -
> - /* Update the ecc vector */
> - calc_ecc += ecc_vector_size;
> - read_ecc += ecc_vector_size;
> + /* page definitely has bit-flips */
> + if (err_vec[i].error_reported)
> + is_error_reported = true;
> }
>
> - /* Check if any error reported */
> if (!is_error_reported)
> return 0;
> + /* detect bit-flips using ELM module */
> + elm_decode_bch_error_page(info->elm_dev, calc_ecc, err_vec);
>
> - /* Decode BCH error using ELM module */
> - elm_decode_bch_error_page(info->elm_dev, ecc_vec, err_vec);
> -
> + /* correct bit-flip */
> for (i = 0; i < eccsteps; i++) {
> - if (err_vec[i].error_reported) {
> + if (err_vec[i].error_uncorrectable) {
> + ret = -EBADMSG;
> + } else if (err_vec[i].error_reported) {
> for (j = 0; j < err_vec[i].error_count; j++) {
> - u32 bit_pos, byte_pos, error_max, pos;
> -
> - if (type == BCH8_ECC)
> - error_max = BCH8_ECC_MAX;
> - else
> - error_max = BCH4_ECC_MAX;
> -
> - if (info->nand.ecc.strength == BCH8_MAX_ERROR)
> - pos = err_vec[i].error_loc[j];
> - else
> + switch (ecc_opt) {
> + case OMAP_ECC_BCH4_CODE_HW:
> + error_max = SECTOR_BYTES +
> + (eccbytes - 1);
> /* Add 4 to take care 4 bit padding */
> pos = err_vec[i].error_loc[j] +
> - BCH4_BIT_PAD;
> -
> - /* Calculate bit position of error */
> + BCH4_BIT_PAD;
> + break;
> + case OMAP_ECC_BCH8_CODE_HW:
> + error_max = SECTOR_BYTES +
> + (eccbytes - 1);
> + pos = err_vec[i].error_loc[j];
> + break;
> + default:
> + return -EINVAL;
> + }
> + /* Calculate bit & byte bit-flip position */
> bit_pos = pos % 8;
> -
> - /* Calculate byte position of error */
> - byte_pos = (error_max - pos - 1) / 8;
> -
> - if (pos < error_max) {
> - if (byte_pos < 512)
> - data[byte_pos] ^= 1 << bit_pos;
> - else
> - spare_ecc[byte_pos - 512] ^=
> + byte_pos = error_max - (pos / 8) - 1;
> + if (byte_pos < SECTOR_BYTES)
> + data[byte_pos] ^= 1 << bit_pos;
> + else if (byte_pos < error_max)
> + read_ecc[byte_pos - SECTOR_BYTES] ^=
> 1 << bit_pos;
> - }
> - /* else, not interested to correct ecc */
> + else
> + ret = -EBADMSG;
> }
> }
> -
> /* Update number of correctable errors */
> stat += err_vec[i].error_count;
> -
> /* Update page data with sector size */
> - data += info->nand.ecc.size;
> - spare_ecc += ecc_vector_size;
> + data += eccsize;
> + read_ecc += eccbytes;
> }
>
> - for (i = 0; i < eccsteps; i++)
> - /* Return error if uncorrectable error present */
> - if (err_vec[i].error_uncorrectable)
> - return -EINVAL;
> -
> - return stat;
> + return (ret < 0) ? ret : stat;
> }
>
> /**
> @@ -1656,6 +1556,7 @@ static int omap_nand_probe(struct platform_device *pdev)
> info->gpmc_cs = pdata->cs;
> info->reg = pdata->reg;
> info->of_node = pdata->of_node;
> + info->ecc_opt = pdata->ecc_opt;
Now that you have this 'info->ecc_opt', you might want to replace the
later uses of 'pdata->ecc_opt' in this function. (Not necessary for the
current patch.)
> mtd = &info->mtd;
> mtd->priv = &info->nand;
> mtd->name = dev_name(&pdev->dev);
Brian
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH v3 2/4] mtd: nand: omap: optimize chip->ecc.calculate() for H/W ECC schemes
2013-11-02 9:46 ` [PATCH v3 2/4] mtd: nand: omap: optimize chip->ecc.calculate() " Pekon Gupta
@ 2013-11-13 21:45 ` Brian Norris
0 siblings, 0 replies; 11+ messages in thread
From: Brian Norris @ 2013-11-13 21:45 UTC (permalink / raw)
To: Pekon Gupta; +Cc: linux-mtd, balbi, dedekind1
Hi Pekon,
On Sat, Nov 02, 2013 at 03:16:14PM +0530, Pekon Gupta wrote:
> chip->ecc.calculate() is used for calculating and fetching of ECC syndrome by
> processing the data passed during Read/Write accesses.
>
> All H/W based ECC schemes supported in omap2-nand driver use GPMC controller
> to calculate ECC syndrome. But each BCHx_ECC scheme implements its own function
> to process and fetch ECC syndrom from GPMC controller.
>
> This patch tries to merges the common code for different BCHx_ECC schemes into
> single omap_calculate_ecc_bch(), And adds schemes specific post-possessing
> after fetching ECC-syndrome. This removes redundant code and adds scalability
> for future ECC-schemes. This patch:
> - [un-touched] omap_calculate_ecc(): Used for HAM1_ECC
> - [merged] omap3_calculate_ecc_bch4(): Used for BCH4_HW_DETECTION_SW
> - [merged] omap3_calculate_ecc_bch8(): Used for BCH8_HW_DETECTION_SW
> - [merged] omap3_calculate_ecc_bch(): Used for BCH4_HW and BCH8_HW
> - [new] omap_calculate_ecc_bch(): Now used for all BCHx_ECC
This patch looks pretty good. Good job simplifying some things. It still
looks like there is some more repetition that could be shortened, but
that may be a little more difficult of a task.
>
> Signed-off-by: Pekon Gupta <pekon@ti.com>
> ---
> drivers/mtd/nand/omap2.c | 245 ++++++++++++++++++-----------------------------
> 1 file changed, 92 insertions(+), 153 deletions(-)
>
> diff --git a/drivers/mtd/nand/omap2.c b/drivers/mtd/nand/omap2.c
> index c946f22..1f59505 100644
> --- a/drivers/mtd/nand/omap2.c
> +++ b/drivers/mtd/nand/omap2.c
> @@ -146,7 +146,11 @@ static u_char bch8_vector[] = {0xf3, 0xdb, 0x14, 0x16, 0x8b, 0xd2, 0xbe, 0xcc,
> 0xac, 0x6b, 0xff, 0x99, 0x7b};
> static u_char bch4_vector[] = {0x00, 0x6b, 0x31, 0xdd, 0x41, 0xbc, 0x10};
> #endif
> -
> +#if defined(CONFIG_MTD_NAND_ECC_BCH) || defined(CONFIG_MTD_NAND_OMAP_BCH)
> +static u8 bch4_polynomial[] = {0x28, 0x13, 0xcc, 0x39, 0x96, 0xac, 0x7f};
> +static u8 bch8_polynomial[] = {0xef, 0x51, 0x2e, 0x09, 0xed, 0x93, 0x9a, 0xc2,
> + 0x97, 0x79, 0xe5, 0x24, 0xb5};
> +#endif
> /* oob info generated runtime depending on ecc algorithm and layout selected */
> static struct nand_ecclayout omap_oobinfo;
>
> @@ -934,9 +938,11 @@ static int omap_calculate_ecc(struct mtd_info *mtd, const u_char *dat,
> u32 val;
>
> val = readl(info->reg.gpmc_ecc_config);
> - if (((val >> ECC_CONFIG_CS_SHIFT) & ~CS_MASK) != info->gpmc_cs)
> + if (((val >> 1) & 0x07) != info->gpmc_cs) {
Why are you dropping the macros in favor of magic numbers? You do this
in a few places.
> + pr_err("%s: invalid ECC configuration for chip-select=%d",
> + DRIVER_NAME, info->gpmc_cs);
> return -EINVAL;
> -
> + }
> /* read ecc result */
> val = readl(info->reg.gpmc_ecc1_result);
> *ecc_code++ = val; /* P128e, ..., P1e */
> @@ -1125,172 +1131,105 @@ static void omap3_enable_hwecc_bch(struct mtd_info *mtd, int mode)
...
>
> - nsectors = ((readl(info->reg.gpmc_ecc_config) >> 4) & 0x7) + 1;
> - /*
> - * find BCH scheme used
> - * 0 -> BCH4
> - * 1 -> BCH8
> - */
> - eccbchtsel = ((readl(info->reg.gpmc_ecc_config) >> 12) & 0x3);
> + val = readl(info->reg.gpmc_ecc_config);
> + if (((val >> 1) & 0x07) != info->gpmc_cs) {
This seems to be the same construct from earlier. Why not use the macros
that explain the shift and mask?
> + pr_err("%s: invalid ECC configuration for chip-select=%d",
> + DRIVER_NAME, info->gpmc_cs);
> + return -EINVAL;
> + }
> + nsectors = ((readl(gpmc_regs->gpmc_ecc_config) >> 4) & 0x7) + 1;
>
> for (i = 0; i < nsectors; i++) {
...
Brian
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH v3 3/4] mtd: nand: omap: optimize chip->ecc.hwctl() for H/W ECC schemes
2013-11-02 9:46 ` [PATCH v3 3/4] mtd: nand: omap: optimize chip->ecc.hwctl() " Pekon Gupta
@ 2013-11-13 22:03 ` Brian Norris
2013-11-14 19:01 ` Ezequiel Garcia
0 siblings, 1 reply; 11+ messages in thread
From: Brian Norris @ 2013-11-13 22:03 UTC (permalink / raw)
To: Pekon Gupta; +Cc: linux-mtd, Ezequiel Garcia, balbi, dedekind1
On Sat, Nov 02, 2013 at 03:16:15PM +0530, Pekon Gupta wrote:
> chip->ecc.hwctl() is used for preparing the H/W controller before read/write
> NAND accesses (like assigning data-buf, enabling ECC scheme configs, etc.)
>
> Though all ECC schemes in OMAP NAND driver use GPMC controller for generating
> ECC syndrome (for both Read/Write accesses). But but in current code
s/But but/but/
:)
> HAM1_ECC and BCHx_ECC schemes implement individual function to achieve this.
> This patch merges the GPMC configuration code for all ECC schemes into
> single omap_enable_hwecc(), thus adding scalability for future ECC schemes.
>
> omap_enable_hwecc() + omap3_enable_hwecc_bch() -> omap_enable_hwecc()
>
> Signed-off-by: Pekon Gupta <pekon@ti.com>
> ---
> drivers/mtd/nand/omap2.c | 212 +++++++++++++++++------------------------------
> 1 file changed, 74 insertions(+), 138 deletions(-)
>
> diff --git a/drivers/mtd/nand/omap2.c b/drivers/mtd/nand/omap2.c
> index 1f59505..3a99e29 100644
> --- a/drivers/mtd/nand/omap2.c
> +++ b/drivers/mtd/nand/omap2.c
> @@ -33,6 +33,10 @@
> #define DRIVER_NAME "omap2-nand"
> #define OMAP_NAND_TIMEOUT_MS 5000
>
> +#define GPMC_ECC_READ 0 /* Reset Hardware ECC for read */
> +#define GPMC_ECC_WRITE 1 /* Reset Hardware ECC for write */
> +#define GPMC_ECC_READSYN 2 /* Reset before syndrom is read back */
> +
> #define NAND_Ecc_P1e (1 << 0)
> #define NAND_Ecc_P2e (1 << 1)
> #define NAND_Ecc_P4e (1 << 2)
> @@ -101,13 +105,9 @@
> #define P4o_s(a) (TF(a & NAND_Ecc_P4o) << 1)
>
> #define PREFETCH_CONFIG1_CS_SHIFT 24
> -#define ECC_CONFIG_CS_SHIFT 1
You stopped using this macro in a different patch, so this removal is
unrelated to this patch. But I'm not sure the code is better without
these macros anyway... I'll leave it to your discretion, I guess.
> #define CS_MASK 0x7
> #define ENABLE_PREFETCH (0x1 << 7)
> #define DMA_MPU_MODE_SHIFT 2
> -#define ECCSIZE0_SHIFT 12
> -#define ECCSIZE1_SHIFT 22
> -#define ECC1RESULTSIZE 0x1
> #define ECCCLEAR 0x100
> #define ECC1 0x1
> #define PREFETCH_FIFOTHRESHOLD_MAX 0x40
...
The rest looks OK but very hard for me to verify. Do we have any
testers? Any platforms to check for regressions?
Brian
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH v3 4/4] mtd: devices: elm: add checks ELM H/W constrains, driver code cleanup
2013-11-02 9:46 ` [PATCH v3 4/4] mtd: devices: elm: add checks ELM H/W constrains, driver code cleanup Pekon Gupta
@ 2013-11-13 22:37 ` Brian Norris
0 siblings, 0 replies; 11+ messages in thread
From: Brian Norris @ 2013-11-13 22:37 UTC (permalink / raw)
To: Pekon Gupta; +Cc: linux-mtd, Ezequiel Garcia, balbi, dedekind1
On Sat, Nov 02, 2013 at 03:16:16PM +0530, Pekon Gupta wrote:
> ELM H/W engine is used by BCHx_ECC schemes for detecting and locating bit-flips.
> However, ELM H/W engine has some constrains like:
> - ELM can decode errors in chunks of 512 data bytes only
> - ELM can operate max upto 8 such buffers in parallel
>
> This patch
> - add checks for above constrains
> - fixes ELM register configs based on number of info->eccsteps
> - cleans-up elm_load_syndrome()
>
> Signed-off-by: Pekon Gupta <pekon@ti.com>
> ---
> drivers/mtd/devices/elm.c | 122 ++++++++++++++++++++++----------------
> drivers/mtd/nand/omap2.c | 2 +-
> include/linux/platform_data/elm.h | 6 +-
> 3 files changed, 77 insertions(+), 53 deletions(-)
>
> diff --git a/drivers/mtd/devices/elm.c b/drivers/mtd/devices/elm.c
> index d1dd6a3..2167384 100644
> --- a/drivers/mtd/devices/elm.c
> +++ b/drivers/mtd/devices/elm.c
> @@ -22,8 +22,11 @@
> #include <linux/of.h>
> #include <linux/sched.h>
> #include <linux/pm_runtime.h>
> +#include <linux/mtd/mtd.h>
> +#include <linux/mtd/nand.h>
> #include <linux/platform_data/elm.h>
>
> +#define DRIVER_NAME "omap-elm"
The driver is actually named "elm", in the platform_driver struct.
Should you use the same name?
> #define ELM_SYSCONFIG 0x010
> #define ELM_IRQSTATUS 0x018
> #define ELM_IRQENABLE 0x01c
> @@ -82,8 +85,10 @@ struct elm_info {
> void __iomem *elm_base;
> struct completion elm_completion;
> struct list_head list;
> + struct mtd_info *mtd;
> enum bch_ecc bch_type;
> struct elm_registers elm_regs;
> + int eccsteps;
> };
>
> static LIST_HEAD(elm_devices);
> @@ -103,19 +108,42 @@ static u32 elm_read_reg(struct elm_info *info, int offset)
> * @dev: ELM device
> * @bch_type: Type of BCH ecc
> */
> -int elm_config(struct device *dev, enum bch_ecc bch_type)
> +int elm_config(struct device *dev, struct mtd_info *mtd,
> + enum bch_ecc bch_type)
> {
> u32 reg_val;
> - struct elm_info *info = dev_get_drvdata(dev);
> -
> + struct elm_info *info;
> + struct nand_chip *chip;
> + if (!dev) {
> + pr_err("%s: ELM device not found\n", DRIVER_NAME);
How about defining pr_fmt(), so you don't have to repeat the DRIVER_NAME
boiler-plate?
> + return -ENODEV;
> + }
> + info = dev_get_drvdata(dev);
> if (!info) {
> - dev_err(dev, "Unable to configure elm - device not probed?\n");
> + pr_err("%s: ELM device data not found\n", DRIVER_NAME);
> return -ENODEV;
> }
> -
> + if (!mtd) {
> + pr_err("%s: MTD device not found\n", DRIVER_NAME);
> + return -ENODEV;
> + }
> + chip = mtd->priv;
> + /* ELM supports error correction in chunks of 512bytes of data only
> + * where each 512bytes of data has its own ECC syndrome */
> + if (chip->ecc.size != 512) {
> + pr_err("%s: invalid ecc_size configuration", DRIVER_NAME);
> + return -EINVAL;
> + }
> + if (mtd->writesize > 4096) {
> + pr_err("%s: page-size > 4096 is not supported", DRIVER_NAME);
> + return -EINVAL;
> + }
> + /* ELM eccsteps required to decode complete NAND page */
> + info->mtd = mtd;
> + info->bch_type = bch_type;
> + info->eccsteps = mtd->writesize / chip->ecc.size;
Isn't this the same as chip->ecc.steps? Do we need to recalculate it
here?
> reg_val = (bch_type & ECC_BCH_LEVEL_MASK) | (ELM_ECC_SIZE << 16);
> elm_write_reg(info, ELM_LOCATION_CONFIG, reg_val);
> - info->bch_type = bch_type;
>
> return 0;
> }
> @@ -152,55 +180,51 @@ static void elm_configure_page_mode(struct elm_info *info, int index,
> * Load syndrome fragment registers with calculated ecc in reverse order.
> */
> static void elm_load_syndrome(struct elm_info *info,
> - struct elm_errorvec *err_vec, u8 *ecc)
> + struct elm_errorvec *err_vec, u8 *ecc_calc)
> {
> + struct nand_chip *chip = info->mtd->priv;
> + unsigned int eccbytes = chip->ecc.bytes;
> + u8 *ecc = ecc_calc;
> int i, offset;
> u32 val;
>
> - for (i = 0; i < ERROR_VECTOR_MAX; i++) {
> -
> + for (i = 0; i < info->eccsteps; i++) {
> /* Check error reported */
> if (err_vec[i].error_reported) {
> elm_configure_page_mode(info, i, true);
> - offset = ELM_SYNDROME_FRAGMENT_0 +
> - SYNDROME_FRAGMENT_REG_SIZE * i;
> -
> - /* BCH8 */
> - if (info->bch_type) {
> -
> - /* syndrome fragment 0 = ecc[9-12B] */
> - val = cpu_to_be32(*(u32 *) &ecc[9]);
> - elm_write_reg(info, offset, val);
> -
> - /* syndrome fragment 1 = ecc[5-8B] */
> - offset += 4;
> - val = cpu_to_be32(*(u32 *) &ecc[5]);
> - elm_write_reg(info, offset, val);
> -
> - /* syndrome fragment 2 = ecc[1-4B] */
> - offset += 4;
> - val = cpu_to_be32(*(u32 *) &ecc[1]);
> - elm_write_reg(info, offset, val);
> -
> - /* syndrome fragment 3 = ecc[0B] */
> - offset += 4;
> - val = ecc[0];
> - elm_write_reg(info, offset, val);
> - } else {
> - /* syndrome fragment 0 = ecc[20-52b] bits */
> - val = (cpu_to_be32(*(u32 *) &ecc[3]) >> 4) |
> - ((ecc[2] & 0xf) << 28);
> - elm_write_reg(info, offset, val);
> -
> - /* syndrome fragment 1 = ecc[0-20b] bits */
> - offset += 4;
> - val = cpu_to_be32(*(u32 *) &ecc[0]) >> 12;
> - elm_write_reg(info, offset, val);
> + offset = SYNDROME_FRAGMENT_REG_SIZE * i;
> + ecc = ecc_calc + (i * eccbytes);
> + switch (info->bch_type) {
> + case BCH4_ECC:
> + val = ((*(ecc + 6) >> 4) & 0x0F) |
> + *(ecc + 5) << 4 | *(ecc + 4) << 12 |
> + *(ecc + 3) << 20 | *(ecc + 2) << 28;
> + elm_write_reg(info, (ELM_SYNDROME_FRAGMENT_0 +
> + offset), cpu_to_le32(val));
> + val = ((*(ecc + 2) >> 4) & 0x0F) |
> + *(ecc + 1) << 4 | *(ecc + 0) << 12;
> + elm_write_reg(info, (ELM_SYNDROME_FRAGMENT_1 +
> + offset), cpu_to_le32(val));
> + break;
> + case BCH8_ECC:
> + val = *(ecc + 12) << 0 | *(ecc + 11) << 8 |
> + *(ecc + 10) << 16 | *(ecc + 9) << 24;
> + elm_write_reg(info, (ELM_SYNDROME_FRAGMENT_0 +
> + offset), cpu_to_le32(val));
> + val = *(ecc + 8) << 0 | *(ecc + 7) << 8 |
> + *(ecc + 6) << 16 | *(ecc + 5) << 24;
> + elm_write_reg(info, (ELM_SYNDROME_FRAGMENT_1 +
> + offset), cpu_to_le32(val));
> + val = *(ecc + 4) << 0 | *(ecc + 3) << 8 |
> + *(ecc + 2) << 16 | *(ecc + 1) << 24;
> + elm_write_reg(info, (ELM_SYNDROME_FRAGMENT_2 +
> + offset), cpu_to_le32(val));
> + val = *(ecc + 0) << 0 & 0x000000FF;
> + elm_write_reg(info, (ELM_SYNDROME_FRAGMENT_3 +
> + offset), cpu_to_le32(val));
> + break;
That's some "interesting" shifting logic. But I think it looks OK...
> }
> }
> -
> - /* Update ecc pointer with ecc byte size */
> - ecc += info->bch_type ? BCH8_SIZE : BCH4_SIZE;
> }
> }
>
> @@ -223,7 +247,7 @@ static void elm_start_processing(struct elm_info *info,
> * Set syndrome vector valid, so that ELM module
> * will process it for vectors error is reported
> */
> - for (i = 0; i < ERROR_VECTOR_MAX; i++) {
> + for (i = 0; i < info->eccsteps; i++) {
> if (err_vec[i].error_reported) {
> offset = ELM_SYNDROME_FRAGMENT_6 +
> SYNDROME_FRAGMENT_REG_SIZE * i;
> @@ -252,7 +276,7 @@ static void elm_error_correction(struct elm_info *info,
> int offset;
> u32 reg_val;
>
> - for (i = 0; i < ERROR_VECTOR_MAX; i++) {
> + for (i = 0; i < info->eccsteps; i++) {
>
> /* Check error reported */
> if (err_vec[i].error_reported) {
> @@ -263,14 +287,12 @@ static void elm_error_correction(struct elm_info *info,
> if (reg_val & ECC_CORRECTABLE_MASK) {
> offset = ELM_ERROR_LOCATION_0 +
> ERROR_LOCATION_SIZE * i;
> -
Random whitespace change? Might not belong in this patch.
> /* Read count of correctable errors */
> err_vec[i].error_count = reg_val &
> ECC_NB_ERRORS_MASK;
>
> /* Update the error locations in error vector */
> for (j = 0; j < err_vec[i].error_count; j++) {
> -
Random whitespace change?
> reg_val = elm_read_reg(info, offset);
> err_vec[i].error_loc[j] = reg_val &
> ECC_ERROR_LOCATION_MASK;
> diff --git a/include/linux/platform_data/elm.h b/include/linux/platform_data/elm.h
> index bf0a83b..d16465b 100644
> --- a/include/linux/platform_data/elm.h
> +++ b/include/linux/platform_data/elm.h
> @@ -25,6 +25,7 @@ enum bch_ecc {
>
> /* ELM support 8 error syndrome process */
> #define ERROR_VECTOR_MAX 8
> +#define ELM_MAX_DETECTABLE_ERRORS 16
Why 16? Isn't 8 the actual max, with 4K page and 512B sector?
>
> #define BCH8_ECC_OOB_BYTES 13
> #define BCH4_ECC_OOB_BYTES 7
Brian
^ permalink raw reply [flat|nested] 11+ messages in thread
* RE: [PATCH v3 1/4] mtd: nand: omap: optimized chip->ecc.correct() for H/W ECC schemes
2013-11-13 21:11 ` Brian Norris
@ 2013-11-14 5:13 ` Gupta, Pekon
0 siblings, 0 replies; 11+ messages in thread
From: Gupta, Pekon @ 2013-11-14 5:13 UTC (permalink / raw)
To: Brian Norris, Ivan Djelic, avinashphilipk
Cc: linux-mtd, Balbi, Felipe, dedekind1
> From: Brian Norris [mailto:computersforpeace@gmail.com]
> + Philip and Ivan, who did relevant work
>
Including Avinash Philip's new email-id... <avinashphilipk@gmail.com>
> > On Sat, Nov 02, 2013 at 03:16:13PM +0530, Pekon Gupta wrote:
> > chip->ecc.correct() is used for detecting and correcting bit-flips during read
> > operations. In omap2-nand driver this is done usingt following functions:
> >
> > - omap_correct_data(): for H/W based HAM1_ECC schemes
> > (Un-Touched in current patch)
> >
> > - omap_elm_correct_data(): for H/W based BCHx_ECC scheme
> > Current implementation of this function is not scalable for newer ECC
> > schemes because:
> > - It depends on a specific byte-position in OOB area (reserved as
> 0x00)
> > to differentiates between programmed-pages and erased-pages.
> > This reserved byte-position cannot be accomodated in all ECC
> schemes.
> > - Current code is not scalable for future ECC schemes due to tweaks
> for
> > BCH4_ECC and BCH8_ECC at multiple places.
> > - It checks for bit-flips in Erased-pages using check_erased_page().
> > This is over-work, as sanity of Erased-page can be verified by just
> > comparing them to a pre-defined ECC-syndrome for all_0xFF data.
> >
> > This patch optimizes omap_elm_correct_data() in following ways:
> > (1) Removes dependency on specific reserved-byte (0x00) in OOB
> area,
> > instead Erased-page is identified by matching calc_ecc with a
> > pre-defined ECC syndrome of all(0xFF) data
> > (2) merges common code for BCH4_ECC and BCH8_ECC for scalability.
> > (3) handles incorrect elm_error_location beyond data+oob buffer.
> > (4) removes check_erased_page(): Bit-flips in erased-page are
> handled
> > in same way as for programmed-page
>
> Are you sure that this new scheme is equivalent to the old? i.e., is
> this backwards compatible?
>
Yes, this is backward compatible. I'm also testing it using old boot-loaders.
With this patch NAND boots (ROM-> boot-loader (SPL + u-boot) -> kernel).
[...]
> Are you totally dropping the "reserved byte" marker?
>
No, The "reserved byte" is part of ecc.layout which is defined in ROM code.
(please refer AM335x TRM [1] (b) Figure which explains ECC layout)
I'm just replacing below check in which driver determines whether page is
programmed or not based on number of '0' bits present in "reserved" byte.
> > - /*
> > - * Check data area is programmed by counting
> > - * number of 0's at fixed offset in spare area.
> > - * Checking count of 0's against threshold.
> > - * In case programmed page expects at least threshold
> > - * zeros in byte.
> > - * If zeros are less than threshold for programmed page/
> > - * zeros are more than threshold erased page, either
> > - * case page reported as uncorrectable.
> > - */
> > - if (hweight8(~read_ecc[eccsize]) >= threshold) {
> > - /*
> > - * Update elm error vector as
> > - * data area is programmed
> > - */
> > - err_vec[i].error_reported = true;
> > - is_error_reported = true;
> > - } else {
> > - /* Error reported in erased page */
> > - int bitflip_count;
> > - u_char *buf = &data[info->nand.ecc.size * i];
> > -
> > - if (memcmp(calc_ecc, erased_ecc_vec, eccsize)) {
> > - bitflip_count = erased_sector_bitflips(
> > - buf, read_ecc, info);
Also, based on following discussion, erased pages also need to be checked
for bit-flips.
https://lkml.org/lkml/2012/10/5/288
So instead of depending on "reserved" byte marker in OOB area,
I'm identifying the erased page based
if ECC_of(read_data + read_ecc) == ECC_of( all 0xff * (ecc.size + ecc.bytes))
then page is erased.
[...]
> Is this compatible with your bootloaders? I don't have a very
> clear mental picture of exactly what your driver is doing to check
> programmed/erased pages, and I'm certainly not familiar with your
> bootloader/ROM code.
>
Pointing to AM335x TRM which I'm referring for this ..
- http://www.ti.com/product/am3359
- [1] latest AM3359 TRM http://www.ti.com/litv/pdf/spruh73i
(a) Section: 26.1.7.4 NAND
This section explains ROM code behavior for booting from NAND.
ROM code also fetches device geometry from ONFI params.
In addition there is a separate boot mode (called NANDI2C) where
device parameters can be fetched from on-board EEPROM.
(b) Figure 26-15. ECC Data Mapping for 2 KB Page and 8b BCH Encoding
This figure explains ecc layout used by ROM code, which is what
I'm using as reference for BCH8_HW algorithm.
with regards, pekon
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH v3 3/4] mtd: nand: omap: optimize chip->ecc.hwctl() for H/W ECC schemes
2013-11-13 22:03 ` Brian Norris
@ 2013-11-14 19:01 ` Ezequiel Garcia
0 siblings, 0 replies; 11+ messages in thread
From: Ezequiel Garcia @ 2013-11-14 19:01 UTC (permalink / raw)
To: Brian Norris; +Cc: linux-mtd, Pekon Gupta, balbi, dedekind1
On Wed, Nov 13, 2013 at 02:03:30PM -0800, Brian Norris wrote:
> On Sat, Nov 02, 2013 at 03:16:15PM +0530, Pekon Gupta wrote:
> > chip->ecc.hwctl() is used for preparing the H/W controller before read/write
> > NAND accesses (like assigning data-buf, enabling ECC scheme configs, etc.)
> >
> > Though all ECC schemes in OMAP NAND driver use GPMC controller for generating
> > ECC syndrome (for both Read/Write accesses). But but in current code
>
> s/But but/but/
>
> :)
>
> > HAM1_ECC and BCHx_ECC schemes implement individual function to achieve this.
> > This patch merges the GPMC configuration code for all ECC schemes into
> > single omap_enable_hwecc(), thus adding scalability for future ECC schemes.
> >
> > omap_enable_hwecc() + omap3_enable_hwecc_bch() -> omap_enable_hwecc()
> >
> > Signed-off-by: Pekon Gupta <pekon@ti.com>
> > ---
> > drivers/mtd/nand/omap2.c | 212 +++++++++++++++++------------------------------
> > 1 file changed, 74 insertions(+), 138 deletions(-)
> >
> > diff --git a/drivers/mtd/nand/omap2.c b/drivers/mtd/nand/omap2.c
> > index 1f59505..3a99e29 100644
> > --- a/drivers/mtd/nand/omap2.c
> > +++ b/drivers/mtd/nand/omap2.c
> > @@ -33,6 +33,10 @@
> > #define DRIVER_NAME "omap2-nand"
> > #define OMAP_NAND_TIMEOUT_MS 5000
> >
> > +#define GPMC_ECC_READ 0 /* Reset Hardware ECC for read */
> > +#define GPMC_ECC_WRITE 1 /* Reset Hardware ECC for write */
> > +#define GPMC_ECC_READSYN 2 /* Reset before syndrom is read back */
> > +
> > #define NAND_Ecc_P1e (1 << 0)
> > #define NAND_Ecc_P2e (1 << 1)
> > #define NAND_Ecc_P4e (1 << 2)
> > @@ -101,13 +105,9 @@
> > #define P4o_s(a) (TF(a & NAND_Ecc_P4o) << 1)
> >
> > #define PREFETCH_CONFIG1_CS_SHIFT 24
> > -#define ECC_CONFIG_CS_SHIFT 1
>
> You stopped using this macro in a different patch, so this removal is
> unrelated to this patch. But I'm not sure the code is better without
> these macros anyway... I'll leave it to your discretion, I guess.
>
> > #define CS_MASK 0x7
> > #define ENABLE_PREFETCH (0x1 << 7)
> > #define DMA_MPU_MODE_SHIFT 2
> > -#define ECCSIZE0_SHIFT 12
> > -#define ECCSIZE1_SHIFT 22
> > -#define ECC1RESULTSIZE 0x1
> > #define ECCCLEAR 0x100
> > #define ECC1 0x1
> > #define PREFETCH_FIFOTHRESHOLD_MAX 0x40
>
> ...
>
> The rest looks OK but very hard for me to verify. Do we have any
> testers? Any platforms to check for regressions?
>
I have 8-bit and 16-bit NAND modules to test this.
Pekon: Are you sending a new version of the whole patchset?
Please Cc me and I'll see about testing it.
--
Ezequiel García, Free Electrons
Embedded Linux, Kernel and Android Engineering
http://free-electrons.com
^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2013-11-14 19:01 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2013-11-02 9:46 [PATCH v3 0/4] optimize and clean-up of OMAP NAND and ELM driver Pekon Gupta
2013-11-02 9:46 ` [PATCH v3 1/4] mtd: nand: omap: optimized chip->ecc.correct() for H/W ECC schemes Pekon Gupta
2013-11-13 21:11 ` Brian Norris
2013-11-14 5:13 ` Gupta, Pekon
2013-11-02 9:46 ` [PATCH v3 2/4] mtd: nand: omap: optimize chip->ecc.calculate() " Pekon Gupta
2013-11-13 21:45 ` Brian Norris
2013-11-02 9:46 ` [PATCH v3 3/4] mtd: nand: omap: optimize chip->ecc.hwctl() " Pekon Gupta
2013-11-13 22:03 ` Brian Norris
2013-11-14 19:01 ` Ezequiel Garcia
2013-11-02 9:46 ` [PATCH v3 4/4] mtd: devices: elm: add checks ELM H/W constrains, driver code cleanup Pekon Gupta
2013-11-13 22:37 ` Brian Norris
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.