[U-Boot] [PATCH v2] dfu, nand, ubi: add partubi alt settings for updating ubi partition

[Heiko Schocher <hs@denx.de>]

Hello Scott,

Am 18.07.2013 00:35, schrieb Scott Wood:
> On 07/15/2013 11:54:09 PM, Heiko Schocher wrote:
>> updating an ubi partition needs a completely erased mtd partition,
>> see:
>> http://lists.infradead.org/pipermail/linux-mtd/2011-May/035416.html
>>
>> So, add partubi alt setting for the dfu_alt_info environment
>> variable to mark this partition as an ubi partition. In case we
>> update an ubi partition, we erase after flashing the image into the
>> partition, the remaining sektors.
>>
>> Signed-off-by: Heiko Schocher <hs@denx.de>
>> Cc: Pantelis Antoniou <panto@antoniou-consulting.com>
>> Cc: Tom Rini <trini@ti.com>
>> Cc: Lukasz Majewski <l.majewski@samsung.com>
>> Cc: Kyungmin Park <kyungmin.park@samsung.com>
>> Cc: Marek Vasut <marex@denx.de>
>> Cc: Wolfgang Denk <wd@denx.de>
>>
>> ---
>>
>> - This patch is also a good starting point to fix up updating ubi, as
>> we currently use "nand erase" for erasing the sektors. This is
>> not the prefered way for writing an ubi image, see:
>> http://www.linux-mtd.infradead.org/faq/ubi.html#L_flash_img
>>
>> This must be fixed ... we have no "ubiformat" in u-boot, or?
>
> The lack of erase counter preservation is a problem, but the part about ECC on erased pages is dealt with in U-Boot by the WITH_DROP_FFS flag.

Ah, ok!

>> - changes for v2:
>> - do not use spread = 1 for nand_erase_opts, to prevent
>> errormessage if there are bad blocks in the erase range.
>> ---
>> drivers/dfu/dfu.c | 30 +++++++++++++++++++++++++++++-
>> drivers/dfu/dfu_nand.c | 26 ++++++++++++++++++++++++++
>> include/dfu.h | 2 ++
>> 3 Dateien ge?ndert, 57 Zeilen hinzugef?gt(+), 1 Zeile entfernt(-)
>>
>> diff --git a/drivers/dfu/dfu.c b/drivers/dfu/dfu.c
>> index 0521752..7ba7026 100644
>> --- a/drivers/dfu/dfu.c
>> +++ b/drivers/dfu/dfu.c
>> @@ -23,6 +23,7 @@
>> #include <errno.h>
>> #include <malloc.h>
>> #include <mmc.h>
>> +#include <nand.h>
>> #include <fat.h>
>> #include <dfu.h>
>> #include <linux/list.h>
>> @@ -176,6 +177,34 @@ int dfu_write(struct dfu_entity *dfu, void *buf, int size, int blk_seq_num)
>> ret = dfu->flush_medium(dfu);
>> printf("\nDFU complete CRC32: 0x%08x\n", dfu->crc);
>>
>> + /* in case of ubi partition, erase rest of the partition */
>> + if (dfu->ubi == 1) {
>> + int ret;
>> + nand_info_t *nand;
>> + /* erase complete partition */
>> + nand_erase_options_t opts;
>> +
>> + if (nand_curr_device < 0 ||
>> + nand_curr_device >= CONFIG_SYS_MAX_NAND_DEVICE ||
>> + !nand_info[nand_curr_device].name) {
>> + printf("%s: invalid nand device\n", __func__);
>> + return -1;
>> + }
>> +
>> + nand = &nand_info[nand_curr_device];
>> +
>> + memset(&opts, 0, sizeof(opts));
>> + opts.offset = dfu->data.nand.start + dfu->offset +
>> + dfu->bad_skip;
>> + opts.length = dfu->data.nand.start +
>> + dfu->data.nand.size - opts.offset;
>> + ret = nand_erase_opts(nand, &opts);
>> + if (ret != 0) {
>> + printf("Failure erase: %d\n", ret);
>> + return ret;
>> + }
>> + }
>
> Instead of separately erasing the remainder of the partition, how about recognizing up front that it's UBI (or that a full partition erase is otherwise desired) and erasing the full partition then? Besides being cleaner, it would be easier to convert to
> an ubi-aware mechanism.

I tried this too, but if I erase the hole partition first (when
a dfu transfer is started) I got:

GADGET DRIVER: usb_dnl_dfu
dfu_write: Wrong sequence number! [7] [0]
dfu_write: Wrong sequence number! [7] [1]
dfu_write: Wrong sequence number! [7] [2]

So, this version needs more investigation ...

bye,
Heiko
-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany