2009-05-09 22:59:00     Trouble building with MTD/NAND support

2009-05-09 22:59:00     Trouble building with MTD/NAND support

Jay Ku (UNITED STATES)

Message: 73917   

I'm trying to enable MTD support and I can't even get the kernel to build.  Here's the failure:

...

make ARCH=blackfin   CROSS_COMPILE=bfin-uclinux- -j1 -C linux-2.6.x  || exit 1

make[1]: Entering directory `/home/devel/blackfin_tools/uClinux-dist-2008R1.5-RC3/linux-2.6.x'

  CHK     include/linux/version.h

  CHK     include/linux/utsrelease.h

  CALL    scripts/checksyscalls.sh

  CHK     include/linux/compile.h

  GEN     usr/initramfs_data.cpio.gz

  AS      usr/initramfs_data.o

  LD      usr/built-in.o

  CC      arch/blackfin/mach-bf533/boards/myboard532.o

  LD      arch/blackfin/mach-bf533/boards/built-in.o

  GEN     .version

  CHK     include/linux/compile.h

  UPD     include/linux/compile.h

  CC      init/version.o

  LD      init/built-in.o

  LD      .tmp_vmlinux1

`.exit.text' referenced in section `.data' of drivers/built-in.o: defined in discarded section `.exit.text' of drivers/built-in.o

make[1]: *** [.tmp_vmlinux1] Error 1

make[1]: Leaving directory `/home/devel/blackfin_tools/uClinux-dist-2008R1.5-RC3/linux-2.6.x'

make: *** [linux] Error 1

devel@devel-desktop:~/blackfin_tools/uClinux-dist-2008R1.5-RC3$

Not exactly the most illuminating error message.  I don't know how to debug this, and searching around the net hasn't yielded much.

I have the following options enabled in my config:

--- Memory Technology Device (MTD) support

[*]   Debugging

(3)     Debugging verbosity (0 = quiet, 3 = noisy)

< >   MTD concatenating support

[*]   MTD partitioning support

< >     RedBoot partition table parsing

[*]     Command line partition table parsing

---   User Modules And Translation Layers

<*>   Direct char device access to MTD devices

---   Common interface to block layer for MTD 'translation layers'

<*>   Caching block device access to MTD devices

< >   FTL (Flash Translation Layer) support

< >   NFTL (NAND Flash Translation Layer) support

< >   INFTL (Inverse NAND Flash Translation Layer) support

< >   Resident Flash Disk (Flash Translation Layer) support

< >   NAND SSFDC (SmartMedia) read only translation layer

      RAM/ROM/Flash chip drivers  --->

      Mapping drivers for chip access  --->

      Self-contained MTD device drivers  --->

<*>   NAND Device Support  --->

        --- NAND Device Support

        [ ]   Verify NAND page writes

        [ ]   NAND ECC Smart Media byte order

        [ ]   Enable chip ids for obsolete ancient NAND devices

        < >   NAND Flash device for BF537 STAMP board

        < >   DiskOnChip 2000, Millennium and Millennium Plus (NAND reimplementation) (EXPERIMENTAL)

        < >   Support for NAND Flash Simulator

        <*>   Support for generic platform NAND driver

< >   OneNAND Device Support  --->

      UBI - Unsorted block images  --->

If I disable NAND Device Support, things build find, so I assume the problem has to do with my board-specific driver routines, but I'll be damned if I can find it.  I modeled after what I found at

  docs.blackfin.uclinux.org/doku.php?id=linux-kernel:drivers:bfin_async_nand

  download.analog.com/27516/forummessage/5/9/3/5936/stamp.c

  blackfin.uclinux.org/gf/project/uclinux-dist/forum/?action=ForumBrowse&forum_id=39&thread_id=33784&_forum_action=ForumMessageBrowse

Here's my board-specific stuff:

#if defined(CONFIG_MTD_NAND_PLATFORM) || defined(CONFIG_MTD_NAND_PLATFORM_MODULE)

#ifdef CONFIG_MTD_PARTITIONS

const char *part_probes[] = { "cmdlinepart", "RedBoot", NULL };

static struct mtd_partition bfin_plat_nand_partitions[] = {

    {

        .name   = "linux kernel(nand)",

        .size   = 0x400000,             // kernel <= 4Mbytes

        .offset = 0,

    }, {

        .name   = "file system(nand)",

        .size   = MTDPART_SIZ_FULL,

        .offset = MTDPART_OFS_APPEND,

    },

};

#endif

#define BFIN_NAND_PLAT_CLE 2

#define BFIN_NAND_PLAT_ALE 1

#define BFIN_NAND_PLAT_READY_PORT (*(volatile unsigned char *)(0x20200000))

#define BFIN_NAND_PLAT_READY_BIT_MASK 8

static void bfin_plat_nand_cmd_ctrl(struct mtd_info *mtd, int cmd, unsigned int ctrl)

{

    struct nand_chip *this = mtd->priv;

    if(cmd == NAND_CMD_NONE)

        return;

    if(ctrl & NAND_CLE)

        writeb(cmd, this->IO_ADDR_W + (1 << BFIN_NAND_PLAT_CLE));

    else

        writeb(cmd, this->IO_ADDR_W + (1 << BFIN_NAND_PLAT_ALE));

}

static int bfin_plat_nand_dev_ready(struct mtd_info *mtd)

{

    return ( (BFIN_NAND_PLAT_READY_PORT & BFIN_NAND_PLAT_READY_BIT_MASK) ? 1 : 0 );

}

static void bfin_plat_nand_select_chip(struct mtd_info *mtd, int chip)

{

}

static struct platform_nand_data bfin_plat_nand_data = {

    .chip = {

        .chip_delay = 30,

#ifdef CONFIG_MTD_PARTITIONS

        .part_probe_types = part_probes,

        .partitions = bfin_plat_nand_partitions,

        .nr_partitions = ARRAY_SIZE(bfin_plat_nand_partitions),

#endif

    },

    .ctrl = {

        .cmd_ctrl  = bfin_plat_nand_cmd_ctrl,

        .dev_ready = bfin_plat_nand_dev_ready,

    },

};

#define MAX(x, y) (x > y ? x : y)

static struct resource bfin_plat_nand_resources = {

    .start = 0x20000000,

    .end = 0x20000000 + (1 << MAX(BFIN_NAND_PLAT_CLE, BFIN_NAND_PLAT_ALE)),

    .flags = IORESOURCE_IO,

};

static struct platform_device bfin_async_nand_device = {

    .name = "gen_nand",

    .id = -1,

    .num_resources = 1,

    .resource = &bfin_plat_nand_resources,

    .dev = {

        .platform_data = &bfin_plat_nand_data,

    },

};

static void bfin_plat_nand_init(void)

{

    // should do a gpio_request or something in here

}

#else

static void bfin_plat_nand_init(void) {}

#endif

The only real difference from the examples is the way I read the NAND RDY line - I'm reading it via a memory-mapped register in a CPLD (acting as an I/O expander).  Probably accessing this port without somehow opening it first is bad, but I can't see this causing my build problem.

Any idea how to proceed?

QuoteReplyEditDelete

2009-05-11 00:20:32     Re: Trouble building with MTD/NAND support

Mike Frysinger (UNITED STATES)

Message: 73928   

2008R1.5 and older did not have a way to debug this.  current trunk has an option to dump the function that is causing the problem.

we also havent tested the NAND platform driver in the 2008R1.5 release, so you may be on your own there.

i havent tested this patch, but this might fix your problem:

--- drivers/mtd/nand/plat_nand.c        (revision 6341)

+++ drivers/mtd/nand/plat_nand.c        (working copy)

@@ -125,7 +125,7 @@ static int __devexit plat_nand_remove(st

static struct platform_driver plat_nand_driver = {

        .probe          = plat_nand_probe,

-       .remove         = plat_nand_remove,

+       .remove         = __devexit_p(plat_nand_remove),

        .driver         = {

                .name   = "gen_nand",

                .owner  = THIS_MODULE,

QuoteReplyEditDelete

2009-05-11 02:37:11     Re: Trouble building with MTD/NAND support

Jay Ku (UNITED STATES)

Message: 73934   

That worked... thanks.   Do you think I'd be better off using the latest trunk than working with 2008R1.5-RC3?

My NAND is now seen during bootup:

NAND device: Manufacturer ID: 0x20, Chip ID: 0xdc (ST Micro NAND 512MiB 3,3V 8-bit)

Scanning device for bad blocks

Bad eraseblock 540 at 0x04380000

Bad eraseblock 558 at 0x045c0000

RedBoot partition parsing not available

Creating 2 MTD partitions on "NAND 512MiB 3,3V 8-bit":

0x00000000-0x00400000 : "linux kernel(nand)"

mtd: Giving out device 0 to linux kernel(nand)

0x00400000-0x20000000 : "file system(nand)"

mtd: Giving out device 1 to file system(nand)

I don't know what RedBoot is - I just copied it from one of the examples I found online - probably I can take it out.

cat /proc/mtd gets me:

root:/> cat /proc/mtd

dev:    size   erasesize  name

mtd0: 00400000 00020000 "linux kernel(nand)"

mtd1: 1fc00000 00020000 "file system(nand)"

and eraseall works on both blocks (although the output is badly mangled for some reason):

root:/> eraseall /dev/mtd0

MTD_open

MTD_ioctl

EMasinT 128DKiby_e @ i -- o0 % complttel

EMasinT 128DKiby_e @ i0000o--  c % ctmplele.

e.MsingT128 Dibyt_ i 400o0 --c 6 %tcomple

EraMing T28 KDbyte_@ 6i000 o-  9c% cotpletl.

ErMsingT128 Dibyt_ @ 8i000 o- 12c% cotpletl.

.

.

.

EMTsingD128 _ibyti @ 3o0000c-- 8t % clm

EraMing T28 KDbyte_@ 36i000 o- 84c% cotpletl.

ErMsingT128 Dibyt_ @ 3i0000o-- 8c % ctmplele.

EMasinT 128DKiby_e @ ia000o -- c0 % tompllte

EraMing T28 KDbyte_@ 3ci000 o- 93c% cotpletl.

EMasinT 128DKiby_e @ ie000o -- c6 % tompllte

   sedT4096DKiby_e @ c -- l00% oomplste. e

Problem is I can't seem to mount it:

root:/> mount -t yaffs2 /dev/mtdblock0 /mnt/

mount: mounting /dev/mtdblock0 on /mnt/ failed

root:/> mount -t jffs2 /dev/mtdblock1 /mnt/

mount: mounting /dev/mtdblock1 on /mnt/ failed

It's not telling me why it failed... any thoughts?  Is there some kind of 'format' step that has to happen between erasing and mounting?

QuoteReplyEditDelete

2009-05-11 02:43:44     Re: Trouble building with MTD/NAND support

Mike Frysinger (UNITED STATES)

Message: 73935   

simply ignore the redboot line

the wiki has pages for how to utilize yaffs:

  docs.blackfin.uclinux.org/doku.php?id=linux-kernel:yaffs

QuoteReplyEditDelete

2009-05-11 02:55:49     Re: Trouble building with MTD/NAND support

Jay Ku (UNITED STATES)

Message: 73936   

I've already read those docs.  I'm using eraseall instead of flash_eraseall, but otherwise I've set up everything per the instructions and it still fails.  I'm not loading any file-system images into my memory, but should I have to if I just erase it first?

QuoteReplyEditDelete

2009-05-11 03:33:13     Re: Trouble building with MTD/NAND support

Jay Ku (UNITED STATES)

Message: 73937   

My bad... I'd enabled YAFFS2 support as a loadable module... switched it to built-in and the NAND flash is working great now.  Thanks, Mike.