This patch adds support for the SLC NAND controller inside the LPC32xx SoC.
Signed-off-by: Roland Stigge <[email protected]>
---
Applies to 3.4-rc6
This driver is ported from git.lpclinux.com 2.6.39.2-lpc, including DT and
dmaengine/pl080 support.
Documentation/devicetree/bindings/mtd/lpc32xx-slc.txt | 52
drivers/mtd/nand/Kconfig | 11
drivers/mtd/nand/Makefile | 1
drivers/mtd/nand/lpc32xx_nand_slc.c | 1067 ++++++++++++++++++
4 files changed, 1131 insertions(+)
--- /dev/null
+++ linux-2.6/Documentation/devicetree/bindings/mtd/lpc32xx-slc.txt
@@ -0,0 +1,52 @@
+NXP LPC32xx SoC NAND SLC controller
+
+Required properties:
+- compatible: "nxp,lpc3220-slc"
+- reg: Address and size of the controller
+- nand-on-flash-bbt: Use bad block table on flash
+- gpios: GPIO specification for NAND write protect
+
+The following required properties are very controller specific. See the LPC32xx
+User Manual:
+- nxp,wdr-clks: Delay before Ready signal is tested on write (W_RDY)
+- nxp,rdr-clks: Delay before Ready signal is tested on read (R_RDY)
+(The following values are specified in Hz, to make them independent of actual
+clock speed:)
+- nxp,wwidth: Write pulse width (W_WIDTH)
+- nxp,whold: Write hold time (W_HOLD)
+- nxp,wsetup: Write setup time (W_SETUP)
+- nxp,rwidth: Read pulse width (R_WIDTH)
+- nxp,rhold: Read hold time (R_HOLD)
+- nxp,rsetup: Read setup time (R_SETUP)
+
+Optional subnodes:
+- Partitions, see Documentation/devicetree/bindings/mtd/partition.txt
+
+Example:
+
+ slc: flash@20020000 {
+ compatible = "nxp,lpc3220-slc";
+ reg = <0x20020000 0x1000>;
+ #address-cells = <1>;
+ #size-cells = <1>;
+
+ nxp,wdr-clks = <14>;
+ nxp,wwidth = <40000000>;
+ nxp,whold = <100000000>;
+ nxp,wsetup = <100000000>;
+ nxp,rdr-clks = <14>;
+ nxp,rwidth = <40000000>;
+ nxp,rhold = <66666666>;
+ nxp,rsetup = <100000000>;
+ nand-on-flash-bbt;
+ gpios = <&gpio 5 19 1>; /* GPO_P3 19, active low */
+
+ mtd0@00000000 {
+ label = "phy3250-boot";
+ reg = <0x00000000 0x00064000>;
+ read-only;
+ };
+
+ ...
+
+ };
--- linux-2.6.orig/drivers/mtd/nand/Kconfig
+++ linux-2.6/drivers/mtd/nand/Kconfig
@@ -414,6 +414,17 @@ config MTD_NAND_PXA3xx
This enables the driver for the NAND flash device found on
PXA3xx processors
+config MTD_NAND_SLC_LPC32XX
+ bool "NXP LPC32xx SLC Controller"
+ depends on ARCH_LPC32XX
+ help
+ Enables support for NXP's LPC32XX SLC (i.e. for Single Level Cell
+ chips) NAND controller. This is the default for the PHYTEC 3250
+ reference board which contains a NAND256R3A2CZA6 chip.
+
+ Please check the actual NAND chip connected and its support
+ by the SLC NAND controller.
+
config MTD_NAND_CM_X270
tristate "Support for NAND Flash on CM-X270 modules"
depends on MACH_ARMCORE
--- linux-2.6.orig/drivers/mtd/nand/Makefile
+++ linux-2.6/drivers/mtd/nand/Makefile
@@ -40,6 +40,7 @@ obj-$(CONFIG_MTD_NAND_ORION) += orion_n
obj-$(CONFIG_MTD_NAND_FSL_ELBC) += fsl_elbc_nand.o
obj-$(CONFIG_MTD_NAND_FSL_IFC) += fsl_ifc_nand.o
obj-$(CONFIG_MTD_NAND_FSL_UPM) += fsl_upm.o
+obj-$(CONFIG_MTD_NAND_SLC_LPC32XX) += lpc32xx_nand_slc.o
obj-$(CONFIG_MTD_NAND_SH_FLCTL) += sh_flctl.o
obj-$(CONFIG_MTD_NAND_MXC) += mxc_nand.o
obj-$(CONFIG_MTD_NAND_SOCRATES) += socrates_nand.o
--- /dev/null
+++ linux-2.6/drivers/mtd/nand/lpc32xx_nand_slc.c
@@ -0,0 +1,1067 @@
+/*
+ * NXP LPC32XX NAND SLC driver
+ *
+ * Authors:
+ * Kevin Wells <[email protected]>
+ * Roland Stigge <[email protected]>
+ *
+ * Copyright (C) 2011 NXP Semiconductors
+ * Copyright (C) 2012 Roland Stigge
+ *
+ * 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 Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ */
+
+#include <linux/slab.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/mtd/mtd.h>
+#include <linux/mtd/nand.h>
+#include <linux/mtd/partitions.h>
+#include <linux/clk.h>
+#include <linux/err.h>
+#include <linux/delay.h>
+#include <linux/io.h>
+#include <linux/mm.h>
+#include <linux/dma-mapping.h>
+#include <linux/dmaengine.h>
+#include <linux/mtd/nand_ecc.h>
+#include <linux/gpio.h>
+#include <linux/of.h>
+#include <linux/of_mtd.h>
+#include <linux/of_gpio.h>
+#include <linux/amba/pl08x.h>
+
+#define LPC32XX_MODNAME "lpc32xx-nand"
+
+/**********************************************************************
+* SLC NAND controller register offsets
+**********************************************************************/
+
+#define SLC_DATA(x) (x + 0x000)
+#define SLC_ADDR(x) (x + 0x004)
+#define SLC_CMD(x) (x + 0x008)
+#define SLC_STOP(x) (x + 0x00C)
+#define SLC_CTRL(x) (x + 0x010)
+#define SLC_CFG(x) (x + 0x014)
+#define SLC_STAT(x) (x + 0x018)
+#define SLC_INT_STAT(x) (x + 0x01C)
+#define SLC_IEN(x) (x + 0x020)
+#define SLC_ISR(x) (x + 0x024)
+#define SLC_ICR(x) (x + 0x028)
+#define SLC_TAC(x) (x + 0x02C)
+#define SLC_TC(x) (x + 0x030)
+#define SLC_ECC(x) (x + 0x034)
+#define SLC_DMA_DATA(x) (x + 0x038)
+
+/**********************************************************************
+* slc_ctrl register definitions
+**********************************************************************/
+#define SLCCTRL_SW_RESET (1 << 2) /* Reset the NAND controller bit */
+#define SLCCTRL_ECC_CLEAR (1 << 1) /* Reset ECC bit */
+#define SLCCTRL_DMA_START (1 << 0) /* Start DMA channel bit */
+
+/**********************************************************************
+* slc_cfg register definitions
+**********************************************************************/
+#define SLCCFG_CE_LOW (1 << 5) /* Force CE low bit */
+#define SLCCFG_DMA_ECC (1 << 4) /* Enable DMA ECC bit */
+#define SLCCFG_ECC_EN (1 << 3) /* ECC enable bit */
+#define SLCCFG_DMA_BURST (1 << 2) /* DMA burst bit */
+#define SLCCFG_DMA_DIR (1 << 1) /* DMA write(0)/read(1) bit */
+#define SLCCFG_WIDTH (1 << 0) /* External device width, 0=8bit */
+
+/**********************************************************************
+* slc_stat register definitions
+**********************************************************************/
+#define SLCSTAT_DMA_FIFO (1 << 2) /* DMA FIFO has data bit */
+#define SLCSTAT_SLC_FIFO (1 << 1) /* SLC FIFO has data bit */
+#define SLCSTAT_NAND_READY (1 << 0) /* NAND device is ready bit */
+
+/**********************************************************************
+* slc_int_stat, slc_ien, slc_isr, and slc_icr register definitions
+**********************************************************************/
+#define SLCSTAT_INT_TC (1 << 1) /* Transfer count bit */
+#define SLCSTAT_INT_RDY_EN (1 << 0) /* Ready interrupt bit */
+
+/**********************************************************************
+* slc_tac register definitions
+**********************************************************************/
+/* Clock setting for RDY write sample wait time in 2*n clocks */
+#define SLCTAC_WDR(n) (((n) & 0xF) << 28)
+/* Write pulse width in clock cycles, 1 to 16 clocks */
+#define SLCTAC_WWIDTH(n) (((n) & 0xF) << 24)
+/* Write hold time of control and data signals, 1 to 16 clocks */
+#define SLCTAC_WHOLD(n) (((n) & 0xF) << 20)
+/* Write setup time of control and data signals, 1 to 16 clocks */
+#define SLCTAC_WSETUP(n) (((n) & 0xF) << 16)
+/* Clock setting for RDY read sample wait time in 2*n clocks */
+#define SLCTAC_RDR(n) (((n) & 0xF) << 12)
+/* Read pulse width in clock cycles, 1 to 16 clocks */
+#define SLCTAC_RWIDTH(n) (((n) & 0xF) << 8)
+/* Read hold time of control and data signals, 1 to 16 clocks */
+#define SLCTAC_RHOLD(n) (((n) & 0xF) << 4)
+/* Read setup time of control and data signals, 1 to 16 clocks */
+#define SLCTAC_RSETUP(n) (((n) & 0xF) << 0)
+
+/**********************************************************************
+* slc_ecc register definitions
+**********************************************************************/
+/* ECC line party fetch macro */
+#define SLCECC_TO_LINEPAR(n) (((n) >> 6) & 0x7FFF)
+#define SLCECC_TO_COLPAR(n) ((n) & 0x3F)
+
+/*
+ * DMA requires storage space for the DMA local buffer and the hardware ECC
+ * storage area. The DMA local buffer is only used if DMA mapping fails
+ * during runtime.
+ */
+#define LPC32XX_DMA_DATA_SIZE 4096
+#define LPC32XX_ECC_SAVE_SIZE ((4096 / 256) * 4)
+
+/* Number of bytes used for ECC stored in NAND per 256 bytes */
+#define LPC32XX_SLC_DEV_ECC_BYTES 3
+
+/*
+ * If the NAND base clock frequency can't be fetched, this frequency will be
+ * used instead as the base. This rate is used to setup the timing registers
+ * used for NAND accesses.
+ */
+#define LPC32XX_DEF_BUS_RATE 133250000
+
+/*
+ * This timeout is used for verifying the NAND buffer has commited its
+ * FIFO to memory or FLASH, or verifying the DMA transfer has completed.
+ * The timeout is used as a count for simple polled checks of the hardware.
+ * For most hardware, the actual timeouts are much lower than this, but
+ * very slow hardware may use most of this time.
+ */
+#define LPC32XX_DMA_SIMPLE_TIMEOUT 10000
+
+/*
+ * NAND ECC Layout for small page NAND devices
+ * Note: For large and huge page devices, the default layouts are used
+ */
+static struct nand_ecclayout lpc32xx_nand_oob_16 = {
+ .eccbytes = 6,
+ .eccpos = {10, 11, 12, 13, 14, 15},
+ .oobfree = {
+ { .offset = 0, .length = 4 },
+ { .offset = 6, .length = 4 },
+ },
+};
+
+static u8 bbt_pattern[] = {'B', 'b', 't', '0' };
+static u8 mirror_pattern[] = {'1', 't', 'b', 'B' };
+
+/*
+ * Small page FLASH BBT descriptors, marker at offset 0, version at offset 6
+ * Note: Large page devices used the default layout
+ */
+static struct nand_bbt_descr bbt_smallpage_main_descr = {
+ .options = NAND_BBT_LASTBLOCK | NAND_BBT_CREATE | NAND_BBT_WRITE
+ | NAND_BBT_2BIT | NAND_BBT_VERSION | NAND_BBT_PERCHIP,
+ .offs = 0,
+ .len = 4,
+ .veroffs = 6,
+ .maxblocks = 4,
+ .pattern = bbt_pattern
+};
+
+static struct nand_bbt_descr bbt_smallpage_mirror_descr = {
+ .options = NAND_BBT_LASTBLOCK | NAND_BBT_CREATE | NAND_BBT_WRITE
+ | NAND_BBT_2BIT | NAND_BBT_VERSION | NAND_BBT_PERCHIP,
+ .offs = 0,
+ .len = 4,
+ .veroffs = 6,
+ .maxblocks = 4,
+ .pattern = mirror_pattern
+};
+
+/*
+ * NAND platform configuration structure
+ */
+struct lpc32xx_nand_cfg_slc {
+ u32 wdr_clks;
+ u32 wwidth;
+ u32 whold;
+ u32 wsetup;
+ u32 rdr_clks;
+ u32 rwidth;
+ u32 rhold;
+ u32 rsetup;
+ bool use_bbt;
+ unsigned wp_gpio;
+ struct mtd_partition *parts;
+ unsigned num_parts;
+};
+
+struct lpc32xx_nand_host {
+ struct nand_chip nand_chip;
+ struct clk *clk;
+ struct mtd_info mtd;
+ void __iomem *io_base;
+ struct lpc32xx_nand_cfg_slc *ncfg;
+
+ struct completion comp;
+ struct dma_chan *dma_chan;
+ u32 dma_buf_len;
+ struct dma_slave_config dma_slave_config;
+ struct scatterlist sgl;
+
+ /*
+ * DMA and CPU addresses of ECC work area and data buffer
+ */
+ u32 *ecc_buf;
+ u8 *data_buf;
+ dma_addr_t io_base_dma;
+};
+
+static void lpc32xx_nand_setup(struct lpc32xx_nand_host *host)
+{
+ u32 clkrate, tmp;
+
+ /* Reset SLC controller */
+ writel(SLCCTRL_SW_RESET, SLC_CTRL(host->io_base));
+ udelay(1000);
+
+ /* Basic setup */
+ writel(0, SLC_CFG(host->io_base));
+ writel(0, SLC_IEN(host->io_base));
+ writel((SLCSTAT_INT_TC | SLCSTAT_INT_RDY_EN),
+ SLC_ICR(host->io_base));
+
+ /* Get base clock for SLC block */
+ clkrate = clk_get_rate(host->clk);
+ if (clkrate == 0)
+ clkrate = LPC32XX_DEF_BUS_RATE;
+
+ /* Compute clock setup values */
+ tmp = SLCTAC_WDR(host->ncfg->wdr_clks) |
+ SLCTAC_WWIDTH(1 + (clkrate / host->ncfg->wwidth)) |
+ SLCTAC_WHOLD(1 + (clkrate / host->ncfg->whold)) |
+ SLCTAC_WSETUP(1 + (clkrate / host->ncfg->wsetup)) |
+ SLCTAC_RDR(host->ncfg->rdr_clks) |
+ SLCTAC_RWIDTH(1 + (clkrate / host->ncfg->rwidth)) |
+ SLCTAC_RHOLD(1 + (clkrate / host->ncfg->rhold)) |
+ SLCTAC_RSETUP(1 + (clkrate / host->ncfg->rsetup));
+ writel(tmp, SLC_TAC(host->io_base));
+}
+
+/*
+ * Hardware specific access to control lines
+ */
+static void lpc32xx_nand_cmd_ctrl(struct mtd_info *mtd, int cmd,
+ unsigned int ctrl)
+{
+ u32 tmp;
+ struct nand_chip *chip = mtd->priv;
+ struct lpc32xx_nand_host *host = chip->priv;
+
+ /* Does CE state need to be changed? */
+ tmp = readl(SLC_CFG(host->io_base));
+ if (ctrl & NAND_NCE)
+ tmp |= SLCCFG_CE_LOW;
+ else
+ tmp &= ~SLCCFG_CE_LOW;
+ writel(tmp, SLC_CFG(host->io_base));
+
+ if (cmd != NAND_CMD_NONE) {
+ if (ctrl & NAND_CLE)
+ writel(cmd, SLC_CMD(host->io_base));
+ else
+ writel(cmd, SLC_ADDR(host->io_base));
+ }
+}
+
+/*
+ * Read the Device Ready pin
+ */
+static int lpc32xx_nand_device_ready(struct mtd_info *mtd)
+{
+ struct nand_chip *chip = mtd->priv;
+ struct lpc32xx_nand_host *host = chip->priv;
+ int rdy = 0;
+
+ if ((readl(SLC_STAT(host->io_base)) & SLCSTAT_NAND_READY) != 0)
+ rdy = 1;
+
+ return rdy;
+}
+
+/*
+ * Enable NAND write protect
+ */
+static void lpc32xx_wp_enable(struct lpc32xx_nand_host *host)
+{
+ gpio_set_value(host->ncfg->wp_gpio, 0);
+}
+
+/*
+ * Disable NAND write protect
+ */
+static void lpc32xx_wp_disable(struct lpc32xx_nand_host *host)
+{
+ gpio_set_value(host->ncfg->wp_gpio, 1);
+}
+
+/*
+ * Prepares SLC for transfers with H/W ECC enabled
+ */
+static void lpc32xx_nand_ecc_enable(struct mtd_info *mtd, int mode)
+{
+ /* Hardware ECC is enabled automatically in hardware as needed */
+}
+
+/*
+ * Calculates the ECC for the data
+ */
+static int lpc32xx_nand_ecc_calculate(struct mtd_info *mtd,
+ const unsigned char *buf,
+ unsigned char *code)
+{
+ /*
+ * ECC is calculated automatically in hardware during syndrome read
+ * and write operations, so it doesn't need to be calculated here.
+ */
+ return 0;
+}
+
+/*
+ * Read a single byte from NAND device
+ */
+static u8 lpc32xx_nand_read_byte(struct mtd_info *mtd)
+{
+ struct nand_chip *chip = mtd->priv;
+ struct lpc32xx_nand_host *host = chip->priv;
+
+ return (u8)readl(SLC_DATA(host->io_base));
+}
+
+/*
+ * Simple device read without ECC
+ */
+static void lpc32xx_nand_read_buf(struct mtd_info *mtd, u_char *buf, int len)
+{
+ struct nand_chip *chip = mtd->priv;
+ struct lpc32xx_nand_host *host = chip->priv;
+
+ /* Direct device read with no ECC */
+ while (len-- > 0)
+ *buf++ = (u8)readl(SLC_DATA(host->io_base));
+}
+
+/*
+ * Simple device write without ECC
+ */
+static void lpc32xx_nand_write_buf(struct mtd_info *mtd, const u8 *buf, int len)
+{
+ struct nand_chip *chip = mtd->priv;
+ struct lpc32xx_nand_host *host = chip->priv;
+
+ /* Direct device write with no ECC */
+ while (len-- > 0)
+ writel((u32)*buf++, SLC_DATA(host->io_base));
+}
+
+/*
+ * Verify data in buffer to data on device
+ */
+static int lpc32xx_verify_buf(struct mtd_info *mtd, const u8 *buf, int len)
+{
+ struct nand_chip *chip = mtd->priv;
+ struct lpc32xx_nand_host *host = chip->priv;
+ int i;
+
+ /* DATA register must be read as 32 bits or it will fail */
+ for (i = 0; i < len; i++) {
+ if (buf[i] != (u8)readl(SLC_DATA(host->io_base)))
+ return -EFAULT;
+ }
+
+ return 0;
+}
+
+/*
+ * Read the OOB data from the device without ECC using FIFO method
+ */
+static int lpc32xx_nand_read_oob_syndrome(struct mtd_info *mtd,
+ struct nand_chip *chip, int page,
+ int sndcmd)
+{
+ if (sndcmd) {
+ chip->cmdfunc(mtd, NAND_CMD_READOOB, 0, page);
+ sndcmd = 0;
+ }
+ chip->read_buf(mtd, chip->oob_poi, mtd->oobsize);
+
+ return sndcmd;
+}
+
+/*
+ * Write the OOB data to the device without ECC using FIFO method
+ */
+static int lpc32xx_nand_write_oob_syndrome(struct mtd_info *mtd,
+ struct nand_chip *chip, int page)
+{
+ int status;
+
+ chip->cmdfunc(mtd, NAND_CMD_SEQIN, mtd->writesize, page);
+ chip->write_buf(mtd, chip->oob_poi, mtd->oobsize);
+
+ /* Send command to program the OOB data */
+ chip->cmdfunc(mtd, NAND_CMD_PAGEPROG, -1, -1);
+
+ status = chip->waitfunc(mtd, chip);
+
+ return status & NAND_STATUS_FAIL ? -EIO : 0;
+}
+
+/*
+ * Fills in the ECC fields in the OOB buffer with the hardware generated ECC
+ */
+static void lpc32xx_slc_ecc_copy(u8 *spare, const u32 *ecc, int count)
+{
+ int i;
+
+ for (i = 0; i < (count * 3); i += 3) {
+ u32 ce = ecc[i / 3];
+ ce = ~(ce << 2) & 0xFFFFFF;
+ spare[i + 2] = (u8)(ce & 0xFF);
+ ce >>= 8;
+ spare[i + 1] = (u8)(ce & 0xFF);
+ ce >>= 8;
+ spare[i] = (u8)(ce & 0xFF);
+ }
+}
+
+static void lpc32xx_dma_complete_func(void *completion)
+{
+ complete(completion);
+}
+
+static int lpc32xx_xmit_dma(struct mtd_info *mtd, dma_addr_t dma,
+ void *mem, int len, enum dma_transfer_direction dir)
+{
+ struct nand_chip *chip = mtd->priv;
+ struct lpc32xx_nand_host *host = chip->priv;
+ struct dma_async_tx_descriptor *desc;
+ int flags = DMA_CTRL_ACK | DMA_PREP_INTERRUPT;
+ int res;
+ dma_cookie_t cookie;
+
+ host->dma_slave_config.direction = dir;
+ host->dma_slave_config.src_addr = dma;
+ host->dma_slave_config.dst_addr = dma;
+ host->dma_slave_config.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
+ host->dma_slave_config.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
+ host->dma_slave_config.src_maxburst = 4;
+ host->dma_slave_config.dst_maxburst = 4;
+ /* DMA controller does flow control: */
+ host->dma_slave_config.device_fc = false;
+ if (dmaengine_slave_config(host->dma_chan, &host->dma_slave_config)) {
+ dev_err(mtd->dev.parent, "Failed to setup DMA slave\n");
+ return -ENXIO;
+ }
+
+ sg_init_one(&host->sgl, mem, len);
+
+ res = dma_map_sg(host->dma_chan->device->dev, &host->sgl, 1, dir);
+ if (res != 1) {
+ dev_err(mtd->dev.parent, "Failed to map sg list\n");
+ return -ENXIO;
+ }
+ desc = dmaengine_prep_slave_sg(host->dma_chan, &host->sgl, 1, dir,
+ flags);
+ if (!desc) {
+ dev_err(mtd->dev.parent, "Failed to prepare slave sg\n");
+ goto out1;
+ }
+
+ init_completion(&host->comp);
+ desc->callback = lpc32xx_dma_complete_func;
+ desc->callback_param = &host->comp;
+
+ cookie = dmaengine_submit(desc);
+ if (dma_submit_error(cookie)) {
+ dev_err(mtd->dev.parent, "Failed to dmaengine_submit()\n");
+ goto out1;
+ }
+ dma_async_issue_pending(host->dma_chan);
+
+ wait_for_completion_timeout(&host->comp, msecs_to_jiffies(1000));
+
+ dma_sync_sg_for_cpu(host->dma_chan->device->dev, &host->sgl, 1, dir);
+
+ return 0;
+out1:
+ dma_unmap_sg(host->dma_chan->device->dev, &host->sgl, 1, dir);
+ return -ENXIO;
+}
+
+/*
+ * DMA read/write transfers with ECC support
+ */
+static int lpc32xx_xfer(struct mtd_info *mtd, u8 *buf, int eccsubpages,
+ int read)
+{
+ struct nand_chip *chip = mtd->priv;
+ struct lpc32xx_nand_host *host = chip->priv;
+ int i, timeout, status = 0;
+ int res;
+ enum dma_transfer_direction dir =
+ read ? DMA_DEV_TO_MEM : DMA_MEM_TO_DEV;
+ u8 *dma_buf;
+ bool dma_mapped;
+
+ if ((void *)buf <= high_memory) {
+ dma_buf = buf;
+ dma_mapped = true;
+ } else {
+ dma_buf = host->data_buf;
+ dma_mapped = false;
+ if (!read)
+ memcpy(host->data_buf, buf, mtd->writesize);
+ }
+
+ if (read) {
+ writel(readl(SLC_CFG(host->io_base)) |
+ SLCCFG_DMA_DIR | SLCCFG_ECC_EN | SLCCFG_DMA_ECC |
+ SLCCFG_DMA_BURST, SLC_CFG(host->io_base));
+ } else {
+ writel((readl(SLC_CFG(host->io_base)) |
+ SLCCFG_ECC_EN | SLCCFG_DMA_ECC | SLCCFG_DMA_BURST) &
+ ~SLCCFG_DMA_DIR,
+ SLC_CFG(host->io_base));
+ }
+
+ /* Clear initial ECC */
+ writel(SLCCTRL_ECC_CLEAR, SLC_CTRL(host->io_base));
+
+ /* Transfer size is data area only */
+ writel(mtd->writesize, SLC_TC(host->io_base));
+
+ /* Start transfer in the NAND controller */
+ writel(readl(SLC_CTRL(host->io_base)) | SLCCTRL_DMA_START,
+ SLC_CTRL(host->io_base));
+
+ for (i = 0; i < chip->ecc.steps; i++) {
+ /* Data */
+ res = lpc32xx_xmit_dma(mtd, SLC_DMA_DATA(host->io_base_dma),
+ dma_buf + i * chip->ecc.size,
+ mtd->writesize / chip->ecc.steps, dir);
+ if (res)
+ return res;
+
+ /* Always _read_ ECC */
+ if (i == chip->ecc.steps - 1)
+ break;
+ if (!read) /* ECC availability delayed on write */
+ udelay(10);
+ res = lpc32xx_xmit_dma(mtd, SLC_ECC(host->io_base_dma),
+ &host->ecc_buf[i], 4, DMA_DEV_TO_MEM);
+ if (res)
+ return res;
+ }
+
+ /*
+ * The DMA is finished, but the NAND controller may still have
+ * buffered data. Wait until all the data is sent.
+ */
+ timeout = LPC32XX_DMA_SIMPLE_TIMEOUT;
+ while ((readl(SLC_STAT(host->io_base)) & SLCSTAT_DMA_FIFO)
+ && (timeout > 0))
+ timeout--;
+ if (!timeout) {
+ dev_err(mtd->dev.parent, "FIFO held data too long\n");
+ status = -EIO;
+ }
+
+ /* Read last calculated ECC value */
+ if (!read)
+ udelay(10);
+ host->ecc_buf[chip->ecc.steps - 1] =
+ readl(SLC_ECC(host->io_base));
+
+ /* Flush DMA */
+ dmaengine_terminate_all(host->dma_chan);
+
+ if (readl(SLC_STAT(host->io_base)) & SLCSTAT_DMA_FIFO ||
+ readl(SLC_TC(host->io_base))) {
+ /* Something is left in the FIFO, something is wrong */
+ dev_err(mtd->dev.parent, "DMA FIFO failure\n");
+ status = -EIO;
+ }
+
+ /* Stop DMA & HW ECC */
+ writel(readl(SLC_CTRL(host->io_base)) & ~SLCCTRL_DMA_START,
+ SLC_CTRL(host->io_base));
+ writel(readl(SLC_CFG(host->io_base)) &
+ ~(SLCCFG_DMA_DIR | SLCCFG_ECC_EN | SLCCFG_DMA_ECC |
+ SLCCFG_DMA_BURST), SLC_CFG(host->io_base));
+
+ if (!dma_mapped && read)
+ memcpy(buf, host->data_buf, mtd->writesize);
+
+ return status;
+}
+
+/*
+ * Read the data and OOB data from the device, use ECC correction with the
+ * data, disable ECC for the OOB data
+ */
+static int lpc32xx_nand_read_page_syndrome(struct mtd_info *mtd,
+ struct nand_chip *chip, u8 *buf,
+ int page)
+{
+ struct lpc32xx_nand_host *host = chip->priv;
+ int stat, i, status;
+ u8 *oobecc, tmpecc[LPC32XX_ECC_SAVE_SIZE];
+
+ /* Issue read command */
+ chip->cmdfunc(mtd, NAND_CMD_READ0, 0, page);
+
+ /* Read data and oob, calculate ECC */
+ status = lpc32xx_xfer(mtd, buf, chip->ecc.steps, 1);
+
+ /* Get OOB data */
+ chip->read_buf(mtd, chip->oob_poi, mtd->oobsize);
+
+ /* Convert to stored ECC format */
+ lpc32xx_slc_ecc_copy(tmpecc, (u32 *) host->ecc_buf, chip->ecc.steps);
+
+ /* Pointer to ECC data retrieved from NAND spare area */
+ oobecc = chip->oob_poi + chip->ecc.layout->eccpos[0];
+
+ for (i = 0; i < chip->ecc.steps; i++) {
+ stat = chip->ecc.correct(mtd, buf, oobecc,
+ &tmpecc[i * chip->ecc.bytes]);
+ if (stat < 0)
+ mtd->ecc_stats.failed++;
+ else
+ mtd->ecc_stats.corrected += stat;
+
+ buf += chip->ecc.size;
+ oobecc += chip->ecc.bytes;
+ }
+
+ return status;
+}
+
+/*
+ * Read the data and OOB data from the device, no ECC correction with the
+ * data or OOB data
+ */
+static int lpc32xx_nand_read_page_raw_syndrome(struct mtd_info *mtd,
+ struct nand_chip *chip,
+ u8 *buf, int page)
+{
+ /* Issue read command */
+ chip->cmdfunc(mtd, NAND_CMD_READ0, 0, page);
+
+ /* Raw reads can just use the FIFO interface */
+ chip->read_buf(mtd, buf, chip->ecc.size * chip->ecc.steps);
+ chip->read_buf(mtd, chip->oob_poi, mtd->oobsize);
+
+ return 0;
+}
+
+/*
+ * Write the data and OOB data to the device, use ECC with the data,
+ * disable ECC for the OOB data
+ */
+static void lpc32xx_nand_write_page_syndrome(struct mtd_info *mtd,
+ struct nand_chip *chip,
+ const u8 *buf)
+{
+ struct lpc32xx_nand_host *host = chip->priv;
+ u8 *pb = chip->oob_poi + chip->ecc.layout->eccpos[0];
+
+ /* Write data, calculate ECC on outbound data */
+ lpc32xx_xfer(mtd, (u8 *)buf, chip->ecc.steps, 0);
+
+ /*
+ * The calculated ECC needs some manual work done to it before
+ * committing it to NAND. Process the calculated ECC and place
+ * the resultant values directly into the OOB buffer. */
+ lpc32xx_slc_ecc_copy(pb, (u32 *)host->ecc_buf, chip->ecc.steps);
+
+ /* Write ECC data to device */
+ chip->write_buf(mtd, chip->oob_poi, mtd->oobsize);
+}
+
+/*
+ * Write the data and OOB data to the device, no ECC correction with the
+ * data or OOB data
+ */
+static void lpc32xx_nand_write_page_raw_syndrome(struct mtd_info *mtd,
+ struct nand_chip *chip,
+ const u8 *buf)
+{
+ /* Raw writes can just use the FIFO interface */
+ chip->write_buf(mtd, buf, chip->ecc.size * chip->ecc.steps);
+ chip->write_buf(mtd, chip->oob_poi, mtd->oobsize);
+}
+
+static bool lpc32xx_dma_filter(struct dma_chan *chan, void *param)
+{
+ struct pl08x_dma_chan *ch =
+ container_of(chan, struct pl08x_dma_chan, chan);
+
+ /* In LPC32xx's PL080 DMA wiring, the NAND DMA signal is #1 */
+ if (ch->cd->min_signal == 1)
+ return true;
+ return false;
+}
+
+/*
+ * Get DMA channel and allocate DMA descriptors memory.
+ * Prepare DMA descriptors link lists
+ */
+static int lpc32xx_nand_dma_setup(struct lpc32xx_nand_host *host)
+{
+ struct mtd_info *mtd = &host->mtd;
+ dma_cap_mask_t mask;
+
+ dma_cap_zero(mask);
+ dma_cap_set(DMA_SLAVE, mask);
+ host->dma_chan = dma_request_channel(mask, lpc32xx_dma_filter, NULL);
+ if (!host->dma_chan) {
+ dev_err(mtd->dev.parent, "Failed to request DMA channel\n");
+ return -EBUSY;
+ }
+ if (dma_set_mask(host->dma_chan->device->dev, 0xFFFFFFFF)) {
+ dev_err(mtd->dev.parent, "Failed to set dma mask\n");
+ goto out1;
+ }
+
+ return 0;
+out1:
+ dma_release_channel(host->dma_chan);
+ return -ENXIO;
+}
+
+#ifdef CONFIG_OF
+static struct lpc32xx_nand_cfg_slc *lpc32xx_parse_dt(struct device *dev)
+{
+ struct lpc32xx_nand_cfg_slc *pdata;
+ struct device_node *np = dev->of_node;
+
+ pdata = devm_kzalloc(dev, sizeof(*pdata), GFP_KERNEL);
+ if (!pdata) {
+ dev_err(dev, "could not allocate memory for platform data\n");
+ return NULL;
+ }
+
+ of_property_read_u32(np, "nxp,wdr-clks", &pdata->wdr_clks);
+ of_property_read_u32(np, "nxp,wwidth", &pdata->wwidth);
+ of_property_read_u32(np, "nxp,whold", &pdata->whold);
+ of_property_read_u32(np, "nxp,wsetup", &pdata->wsetup);
+ of_property_read_u32(np, "nxp,rdr-clks", &pdata->rdr_clks);
+ of_property_read_u32(np, "nxp,rwidth", &pdata->rwidth);
+ of_property_read_u32(np, "nxp,rhold", &pdata->rhold);
+ of_property_read_u32(np, "nxp,rsetup", &pdata->rsetup);
+
+ if (!pdata->wdr_clks || !pdata->wwidth || !pdata->whold ||
+ !pdata->wsetup || !pdata->rdr_clks || !pdata->rwidth ||
+ !pdata->rhold || !pdata->rsetup) {
+ dev_err(dev, "chip parameters not specified correctly\n");
+ return NULL;
+ }
+
+ pdata->use_bbt = of_get_nand_on_flash_bbt(np);
+ pdata->wp_gpio = of_get_named_gpio_flags(np, "gpios", 0, NULL);
+
+ return pdata;
+}
+#else
+static struct lpc32xx_nand_cfg_slc *lpc32xx_parse_dt(struct device *dev)
+{
+ return NULL;
+}
+#endif
+
+/*
+ * Probe for NAND controller
+ */
+static int __devinit lpc32xx_nand_probe(struct platform_device *pdev)
+{
+ struct lpc32xx_nand_host *host;
+ struct mtd_info *mtd;
+ struct nand_chip *chip;
+ struct resource *rc;
+ struct mtd_part_parser_data ppdata = {};
+ int res;
+
+ rc = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+ if (rc == NULL) {
+ dev_err(&pdev->dev, "No memory resource found for device\n");
+ return -EBUSY;
+ }
+
+ /* Allocate memory for the device structure (and zero it) */
+ host = devm_kzalloc(&pdev->dev, sizeof(*host), GFP_KERNEL);
+ if (!host) {
+ dev_err(&pdev->dev, "failed to allocate device structure\n");
+ return -ENOMEM;
+ }
+ host->io_base_dma = rc->start;
+
+ host->io_base = devm_request_and_ioremap(&pdev->dev, rc);
+ if (host->io_base == NULL) {
+ dev_err(&pdev->dev, "ioremap failed\n");
+ return -ENOMEM;
+ }
+
+ if (pdev->dev.of_node)
+ host->ncfg = lpc32xx_parse_dt(&pdev->dev);
+ else
+ host->ncfg = pdev->dev.platform_data;
+ if (!host->ncfg) {
+ dev_err(&pdev->dev, "Missing platform data\n");
+ return -ENOENT;
+ }
+ if (gpio_request(host->ncfg->wp_gpio, "NAND WP")) {
+ dev_err(&pdev->dev, "GPIO not available\n");
+ return -EBUSY;
+ }
+ lpc32xx_wp_disable(host);
+
+ mtd = &host->mtd;
+ chip = &host->nand_chip;
+ chip->priv = host;
+ mtd->priv = chip;
+ mtd->owner = THIS_MODULE;
+ mtd->dev.parent = &pdev->dev;
+
+ /* Get NAND clock */
+ host->clk = clk_get(&pdev->dev, NULL);
+ if (IS_ERR(host->clk)) {
+ dev_err(&pdev->dev, "Clock failure\n");
+ res = -ENOENT;
+ goto err_exit1;
+ }
+ clk_enable(host->clk);
+
+ /* Set NAND IO addresses and command/ready functions */
+ chip->IO_ADDR_R = SLC_DATA(host->io_base);
+ chip->IO_ADDR_W = SLC_DATA(host->io_base);
+ chip->cmd_ctrl = lpc32xx_nand_cmd_ctrl;
+ chip->dev_ready = lpc32xx_nand_device_ready;
+ chip->chip_delay = 20; /* 20us command delay time */
+
+ /* Init NAND controller */
+ lpc32xx_nand_setup(host);
+
+ platform_set_drvdata(pdev, host);
+
+ /* NAND callbacks for LPC32xx SLC hardware */
+ chip->ecc.mode = NAND_ECC_HW_SYNDROME;
+ chip->read_byte = lpc32xx_nand_read_byte;
+ chip->read_buf = lpc32xx_nand_read_buf;
+ chip->write_buf = lpc32xx_nand_write_buf;
+ chip->ecc.read_page_raw = lpc32xx_nand_read_page_raw_syndrome;
+ chip->ecc.read_page = lpc32xx_nand_read_page_syndrome;
+ chip->ecc.write_page_raw = lpc32xx_nand_write_page_raw_syndrome;
+ chip->ecc.write_page = lpc32xx_nand_write_page_syndrome;
+ chip->ecc.write_oob = lpc32xx_nand_write_oob_syndrome;
+ chip->ecc.read_oob = lpc32xx_nand_read_oob_syndrome;
+ chip->ecc.calculate = lpc32xx_nand_ecc_calculate;
+ chip->ecc.correct = nand_correct_data;
+ chip->ecc.hwctl = lpc32xx_nand_ecc_enable;
+ chip->verify_buf = lpc32xx_verify_buf;
+
+ /*
+ * Allocate a large enough buffer for a single huge page plus
+ * extra space for the spare area and ECC storage area
+ */
+ host->dma_buf_len = LPC32XX_DMA_DATA_SIZE + LPC32XX_ECC_SAVE_SIZE;
+ host->data_buf = devm_kzalloc(&pdev->dev, host->dma_buf_len,
+ GFP_KERNEL);
+ if (host->data_buf == NULL) {
+ dev_err(&pdev->dev, "Error allocating memory\n");
+ res = -ENOMEM;
+ goto err_exit2;
+ }
+
+ res = lpc32xx_nand_dma_setup(host);
+ if (res) {
+ res = -EIO;
+ goto err_exit2;
+ }
+
+ /* Find NAND device */
+ if (nand_scan_ident(mtd, 1, NULL)) {
+ res = -ENXIO;
+ goto err_exit3;
+ }
+
+ /* OOB and ECC CPU and DMA work areas */
+ host->ecc_buf = (u32 *)(host->data_buf + LPC32XX_DMA_DATA_SIZE);
+
+ /*
+ * Small page FLASH has a unique OOB layout, but large and huge
+ * page FLASH use the standard layout. Small page FLASH uses a
+ * custom BBT marker layout.
+ */
+ if (mtd->writesize <= 512)
+ chip->ecc.layout = &lpc32xx_nand_oob_16;
+
+ /* These sizes remain the same regardless of page size */
+ chip->ecc.size = 256;
+ chip->ecc.bytes = LPC32XX_SLC_DEV_ECC_BYTES;
+ chip->ecc.prepad = chip->ecc.postpad = 0;
+
+ /* Avoid extra scan if using BBT, setup BBT support */
+ if (host->ncfg->use_bbt) {
+ chip->options |= NAND_SKIP_BBTSCAN;
+ chip->bbt_options |= NAND_BBT_USE_FLASH;
+
+ /*
+ * Use a custom BBT marker setup for small page FLASH that
+ * won't interfere with the ECC layout. Large and huge page
+ * FLASH use the standard layout.
+ */
+ if (mtd->writesize <= 512) {
+ chip->bbt_td = &bbt_smallpage_main_descr;
+ chip->bbt_md = &bbt_smallpage_mirror_descr;
+ }
+ }
+
+ /*
+ * Fills out all the uninitialized function pointers with the defaults
+ */
+ if (nand_scan_tail(mtd)) {
+ res = -ENXIO;
+ goto err_exit3;
+ }
+
+ /* Standard layout in FLASH for bad block tables */
+ if (host->ncfg->use_bbt) {
+ if (nand_default_bbt(mtd) < 0)
+ dev_err(&pdev->dev,
+ "Error initializing default bad block tables\n");
+ }
+
+ mtd->name = "nxp_lpc3220_slc";
+ ppdata.of_node = pdev->dev.of_node;
+ res = mtd_device_parse_register(mtd, NULL, &ppdata, host->ncfg->parts,
+ host->ncfg->num_parts);
+ if (!res)
+ return res;
+
+ nand_release(mtd);
+
+err_exit3:
+ dma_release_channel(host->dma_chan);
+err_exit2:
+ clk_disable(host->clk);
+ clk_put(host->clk);
+ platform_set_drvdata(pdev, NULL);
+err_exit1:
+ lpc32xx_wp_enable(host);
+ gpio_free(host->ncfg->wp_gpio);
+
+ return res;
+}
+
+/*
+ * Remove NAND device.
+ */
+static int __devexit lpc32xx_nand_remove(struct platform_device *pdev)
+{
+ u32 tmp;
+ struct lpc32xx_nand_host *host = platform_get_drvdata(pdev);
+ struct mtd_info *mtd = &host->mtd;
+
+ nand_release(mtd);
+ dma_release_channel(host->dma_chan);
+
+ /* Force CE high */
+ tmp = readl(SLC_CTRL(host->io_base));
+ tmp &= ~SLCCFG_CE_LOW;
+ writel(tmp, SLC_CTRL(host->io_base));
+
+ clk_disable(host->clk);
+ clk_put(host->clk);
+ platform_set_drvdata(pdev, NULL);
+ lpc32xx_wp_enable(host);
+ gpio_free(host->ncfg->wp_gpio);
+
+ return 0;
+}
+
+#ifdef CONFIG_PM
+static int lpc32xx_nand_resume(struct platform_device *pdev)
+{
+ struct lpc32xx_nand_host *host = platform_get_drvdata(pdev);
+
+ /* Re-enable NAND clock */
+ clk_enable(host->clk);
+
+ /* Fresh init of NAND controller */
+ lpc32xx_nand_setup(host);
+
+ /* Disable write protect */
+ lpc32xx_wp_disable(host);
+
+ return 0;
+}
+
+static int lpc32xx_nand_suspend(struct platform_device *pdev, pm_message_t pm)
+{
+ u32 tmp;
+ struct lpc32xx_nand_host *host = platform_get_drvdata(pdev);
+
+ /* Force CE high */
+ tmp = readl(SLC_CTRL(host->io_base));
+ tmp &= ~SLCCFG_CE_LOW;
+ writel(tmp, SLC_CTRL(host->io_base));
+
+ /* Enable write protect for safety */
+ lpc32xx_wp_enable(host);
+
+ /* Disable clock */
+ clk_disable(host->clk);
+
+ return 0;
+}
+
+#else
+#define lpc32xx_nand_resume NULL
+#define lpc32xx_nand_suspend NULL
+#endif
+
+#if defined(CONFIG_OF)
+static const struct of_device_id lpc32xx_nand_match[] = {
+ { .compatible = "nxp,lpc3220-slc" },
+ { /* sentinel */ },
+};
+MODULE_DEVICE_TABLE(of, lpc32xx_nand_match);
+#endif
+
+static struct platform_driver lpc32xx_nand_driver = {
+ .probe = lpc32xx_nand_probe,
+ .remove = __devexit_p(lpc32xx_nand_remove),
+ .resume = lpc32xx_nand_resume,
+ .suspend = lpc32xx_nand_suspend,
+ .driver = {
+ .name = LPC32XX_MODNAME,
+ .owner = THIS_MODULE,
+ .of_match_table = of_match_ptr(lpc32xx_nand_match),
+ },
+};
+
+module_platform_driver(lpc32xx_nand_driver);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Kevin Wells <[email protected]>");
+MODULE_AUTHOR("Roland Stigge <[email protected]>");
+MODULE_DESCRIPTION("NAND driver for the NXP LPC32XX SLC controller");
I am CCing few other guys who take care of several drivers which use
similar way of busy-waiting - probably you could change it?
Bastian: drivers/mtd/nand/sh_flctl.c
Lars-Peter: drivers/mtd/nand/jz4740_nand.c
Huang: drivers/mtd/nand/gpmi-nand/gpmi-lib.c
Lei Wen: drivers/mtd/nand/pxa3xx_nand.c
On Sat, 2012-05-12 at 15:29 +0200, Roland Stigge wrote:
> + /*
> + * The DMA is finished, but the NAND controller may still have
> + * buffered data. Wait until all the data is sent.
> + */
> + timeout = LPC32XX_DMA_SIMPLE_TIMEOUT;
> + while ((readl(SLC_STAT(host->io_base)) & SLCSTAT_DMA_FIFO)
> + && (timeout > 0))
> + timeout--;
> + if (!timeout) {
> + dev_err(mtd->dev.parent, "FIFO held data too long\n");
> + status = -EIO;
> + }
I know the MTD tree is full of this, but this is bad, I think. The
timeout should be time-backed, not CPU-cycles-backed.
I do not know the best way to do this, hopefully someone in the arm list
could suggest, but the following pattern is at least better:
/* Chip reaction time timeout in milliseconds */
#define LPC32XX_DMA_TIMEOUT 100
timeout = loops_per_jiffy * msecs_to_jiffies(LPC32XX_DMA_TIMEOUT);
while ((readl(...)) && timeout-- > 0)
cpu_relax();
if (!timeout)
error;
So basically I turned your hard-coded iterations count into a time-based
timeout. I also used cpu_relax() which is commonly used in tight-loops
like this. Here is a piece of documentation about cpu_relax():
"
The right way to perform a busy wait is:
while (my_variable != what_i_want)
cpu_relax();
The cpu_relax() call can lower CPU power consumption or yield to a
hyperthreaded twin processor; it also happens to serve as a compiler
barrier, so, once again, volatile is unnecessary. Of course, busy-
waiting is generally an anti-social act to begin with.
"
--
Best Regards,
Artem Bityutskiy
于 2012年05月15日 15:55, Artem Bityutskiy 写道:
> I am CCing few other guys who take care of several drivers which use
> similar way of busy-waiting - probably you could change it?
>
> Bastian: drivers/mtd/nand/sh_flctl.c
> Lars-Peter: drivers/mtd/nand/jz4740_nand.c
> Huang: drivers/mtd/nand/gpmi-nand/gpmi-lib.c
> Lei Wen: drivers/mtd/nand/pxa3xx_nand.c
>
> On Sat, 2012-05-12 at 15:29 +0200, Roland Stigge wrote:
>> + /*
>> + * The DMA is finished, but the NAND controller may still have
>> + * buffered data. Wait until all the data is sent.
When all the data is sent, is there an interrupt for this?
Best Regards
Huang Shijie
>> + */
>> + timeout = LPC32XX_DMA_SIMPLE_TIMEOUT;
>> + while ((readl(SLC_STAT(host->io_base))& SLCSTAT_DMA_FIFO)
>> +&& (timeout> 0))
>> + timeout--;
>> + if (!timeout) {
>> + dev_err(mtd->dev.parent, "FIFO held data too long\n");
>> + status = -EIO;
>> + }
> I know the MTD tree is full of this, but this is bad, I think. The
> timeout should be time-backed, not CPU-cycles-backed.
>
> I do not know the best way to do this, hopefully someone in the arm list
> could suggest, but the following pattern is at least better:
>
>
> /* Chip reaction time timeout in milliseconds */
> #define LPC32XX_DMA_TIMEOUT 100
>
> timeout = loops_per_jiffy * msecs_to_jiffies(LPC32XX_DMA_TIMEOUT);
>
> while ((readl(...))&& timeout--> 0)
> cpu_relax();
>
> if (!timeout)
> error;
>
>
> So basically I turned your hard-coded iterations count into a time-based
> timeout. I also used cpu_relax() which is commonly used in tight-loops
> like this. Here is a piece of documentation about cpu_relax():
>
> "
> The right way to perform a busy wait is:
>
> while (my_variable != what_i_want)
> cpu_relax();
>
> The cpu_relax() call can lower CPU power consumption or yield to a
> hyperthreaded twin processor; it also happens to serve as a compiler
> barrier, so, once again, volatile is unnecessary. Of course, busy-
> waiting is generally an anti-social act to begin with.
> "
>
Hi Artem and Huang,
thank you for your feedback!
On 05/15/2012 10:15 AM, Huang Shijie wrote:
>> On Sat, 2012-05-12 at 15:29 +0200, Roland Stigge wrote:
>>> + /*
>>> + * The DMA is finished, but the NAND controller may still have
>>> + * buffered data. Wait until all the data is sent.
> When all the data is sent, is there an interrupt for this?
Bad news is: No
Good news is: The previous DMA operation finished with an interrupt
which according to the manual should already corresponds to this
condition. Tests show that at this point of sampling:
>>> + timeout = LPC32XX_DMA_SIMPLE_TIMEOUT;
>>> + while ((readl(SLC_STAT(host->io_base))& SLCSTAT_DMA_FIFO)
>>> +&& (timeout> 0))
>>> + timeout--;
... the condition is always true and always just jumps over this loop,
at least with my hardware.
>> /* Chip reaction time timeout in milliseconds */
>> #define LPC32XX_DMA_TIMEOUT 100
>>
>> timeout = loops_per_jiffy * msecs_to_jiffies(LPC32XX_DMA_TIMEOUT);
>>
>> while ((readl(...))&& timeout--> 0)
>> cpu_relax();
As I understand loops_per_jiffy, this loop will take much longer than
the 100 ms you defined above?
Anyway, I will keep the loop for safety reasons, add an msleep() and add
a warning, should the loop be entered _at all_.
Maybe someone from NXP can give us more insight here? Maybe the
condition check isn't necessary anymore after I ported the driver to
dmaengine (this controller is always wired together with an amba-pl080
in the LPC32xx)?
Thanks in advance,
Roland
On Tue, 2012-05-15 at 15:20 +0200, Roland Stigge wrote:
> >> while ((readl(...))&& timeout--> 0)
> >> cpu_relax();
>
> As I understand loops_per_jiffy, this loop will take much longer than
> the 100 ms you defined above?
Not sure about much, but longer. The idea is that this is about the
error path so if we report -EIO with a slight delay - no problem.
--
Best Regards,
Artem Bityutskiy
On 05/15/2012 03:31 PM, Artem Bityutskiy wrote:
>> As I understand loops_per_jiffy, this loop will take much longer
>> than the 100 ms you defined above?
>
> Not sure about much, but longer. The idea is that this is about
> the error path so if we report -EIO with a slight delay - no
> problem.
Turned out that the condition (FIFO empty) is always true for me.
Keeping the check for safety reasons for now, doing the timeout with
msleep()s which should be (cpu-wise) "social" enough and are
unexpected anyway but do approximate the ms timeout more precisely.
Roland