attaching UBI while getting read errors from NAND driver fails.

[Artem Bityutskiy]

On Tue, 2010-04-20 at 20:44 +0200, Sebastian Andrzej Siewior wrote:
> Attaching empty nand with a block which contains a RS-Error which can't
> be fixed:
> 
> | UBI: attaching mtd9 to ubi0
> | UBI: physical eraseblock size:   131072 bytes (128 KiB)
> | UBI: logical eraseblock size:    129024 bytes
> | UBI: smallest flash I/O unit:    2048
> | UBI: sub-page size:              512
> | UBI: VID header offset:          512 (aligned 512)
> | UBI: data offset:                2048
> | UBI error: ubi_io_read: error -74 while reading 64 bytes from PEB 3399:0, read 64 bytes
> | Call Trace:
> | [cfbd5c60] [c0008558] show_stack+0x48/0x19c (unreliable)
> | [cfbd5ca0] [c01a71e8] ubi_io_read+0x188/0x288
> | [cfbd5cf0] [c01a76e8] ubi_io_read_ec_hdr+0x74/0x2a4
> | [cfbd5d20] [c01abe9c] ubi_scan+0x178/0x10b4
> | [cfbd5d80] [c01a1464] ubi_attach_mtd_dev+0x67c/0xe44
> | [cfbd5e80] [c01a1fc8] ctrl_cdev_ioctl+0x178/0x210
> | [cfbd5ec0] [c008711c] do_ioctl+0x3c/0xc4
> | [cfbd5ee0] [c0087224] vfs_ioctl+0x80/0x448
> | [cfbd5f10] [c008762c] sys_ioctl+0x40/0x88
> | [cfbd5f40] [c000f960] ret_from_syscall+0x0/0x38
> | UBI error: ubi_read_volume_table: the layout volume was not found
> | UBI error: ubi_attach_mtd_dev: failed to attach by scanning, error -22
> 
> ubi_io_read() calls ubi->mtd->read() which ends in nand_do_read_ops() at
> some point. The NAND controller was not able to read the page and
> incremented mtd->ecc_stats.failed. The tail of nand_do_read_ops():
> 
> |      if (mtd->ecc_stats.failed - stats.failed)
> |                return -EBADMSG;
> 
> So we return with -EBADMSG as you can see in the error message. This
> error is then converted by ubi_io_read() into -EIO:
> |                /*
> |                 * The driver should never return -EBADMSG if it failed to read
> |                 * all the requested data. But some buggy drivers might do
> |                 * this, so we change it to -EIO.
> |                 */
> |                if (read != len && err == -EBADMSG) {
> |                        ubi_assert(0);
> |                        err = -EIO;
> |                }
> 
> At this point we leave ubi_scan() and the mtd partition remains
> unattached. No really sure how we ended up with -EINVAL in
> ubi_attach_mtd_dev().
> 
> I guess this is not wanted, is it?
> 
> The following patch led to:
> | UBI: empty MTD device detected
> | UBI: create volume table (copy #1)
> | UBI: create volume table (copy #2)
> | UBI: attached mtd9 to ubi0
> | UBI: MTD device name:            "ubi"
> | UBI: MTD device size:            512 MiB
> | UBI: number of good PEBs:        4092
> | UBI: number of bad PEBs:         4
> | UBI: max. allowed volumes:       128
> | UBI: wear-leveling threshold:    4096
> | UBI: number of internal volumes: 1
> | UBI: number of user volumes:     0
> | UBI: available PEBs:             4048
> | UBI: total number of reserved PEBs: 44
> | UBI: number of PEBs reserved for bad PEB handling: 40
> | UBI: max/mean erase counter: 0/0
> | UBI: image sequence number: 0
> | UBI: background thread "ubi_bgt0d" started, PID 1986
> 
> ubimkvol gave me:
> | UBI error: ubi_io_read: error -74 while reading 64 bytes from PEB 3399:0, read 64 bytes
> | nand_erase: Failed erase, page 0x000751c0
> | nand_erase: Failed erase, page 0x000751c0
> | nand_erase: Failed erase, page 0x000751c0
> | nand_erase: Failed erase, page 0x000751c0
> | UBI error: do_sync_erase: cannot erase PEB 3399, error -5
> | UBI error: erase_worker: failed to erase PEB 3399, error -5
> | UBI: mark PEB 3399 as bad
> | UBI: 39 PEBs left in the reserve
> 
> This looks good so far I guess. Is there much I break with patch? Should
> this case be handled at another level?

Well, I think the idea was that the MTD layer tries to read all the
requested data despite of a possible unrecoverable ECC error in the
middle. So if it returns -EBADMSG, we know that the read buffer contains
all the data anyway, even though some of the data may be incorrect.

So, I'd suggest changing MTD and making it return read == len in that
case.

> @@ -756,7 +756,8 @@ static int process_eb(struct ubi_device *ubi, struct ubi_scan_info *si,
>  		bitflips = 1;
>  	}
>  
> -	si->is_empty = 0;
> +	if (err != UBI_IO_BAD_EC_HDR)
> +		si->is_empty = 0;
>  
>  	if (!ec_corr) {
>  		int image_seq;
> @@ -827,6 +828,7 @@ static int process_eb(struct ubi_device *ubi, struct ubi_scan_info *si,
>  			return err;
>  		goto adjust_mean_ec;
>  	}
> +	si->is_empty = 0;
>  
>  	vol_id = be32_to_cpu(vidh->vol_id);
>  	if (vol_id > UBI_MAX_VOLUMES && vol_id != UBI_LAYOUT_VOLUME_ID) {

This part looks right.

-- 
Best Regards,
Artem Bityutskiy (Артём Битюцкий)