nand_do_write_ops: attempt to write non page aligned data

[Scott Wood]

On Fri, 2015-02-27 at 12:56 +0000, Instigate | Narine Hovhannisyan
wrote:
> Hello Everyone,
> 
> 
> I have a question about writing data to NAND flash device and will appreciate
> your comments about it.
> 
> Our device has NAND Flash type memory and we are able to mount the flash (as
> /dev/mtd4) with jffs2.  
> We are able to erase the flash using flash_eraseall without any errors. 
> But when we try to write data on it, we get the following error:
> 
> > cat uImage > /dev/mtd4
> nand_do_write_ops: attempt to write non page aligned data
> cat: write error: Invalid argument
> 
> The goal is to write the uBoot image to the mtdchar device directly without mounting it as file system. 
> 
> In our case the last block of the data is not aligned to page size (4096).
> As I understand I need to pad the image with some garbage to align it to 4K
> border. And if I add padding to the image (e.g. with dd), the command works
> fine.
> 
> However I did take a look at kernel sources and found the following lines:
> 
> drivers/mtd/nand/nand_base.c:
> 
> 
>     /* Reject writes, whichare not page aligned */
>     if (NOTALIGNED(to) || NOTALIGNED(ops->len)) {
>         pr_notice("%s: attempt to write non page aligned data\n", __func__);
> 
>         return -EINVAL;
>     }
> 
> that's the place, where I assume the error comes from.
> 
> Below the check I see the code which handles the partial page write 
> (fills the page with 0xFFs and then writes the data).
> 
> 
>     /* Partial page write? */    if (unlikely(column || writelen < (mtd->writesize - 1))) {
>         cached = 0;
>         bytes = min_t(int, bytes - column, (int) writelen);
>         chip->pagebuf = -1;
>         memset(chip->buffers->databuf, 0xff, mtd->writesize);
>         memcpy(&chip->buffers->databuf[column], buf, bytes);
>         wbuf = chip->buffers->databuf;
>     }
> 
> So I'd expect that this part should handle the partial page write. Also I tried
> to remove the above check of NOTALIGNED(ops->len) data and it worked fine.
> 
> Can please someone describe the intention of the check and handling of partial
> page write? Can I just remove the NOTALIGNED(ops->len) check and leave the
> padding to kernel? What is the risk in this case?

FWIW, in U-Boot we removed the check on "ops->len" and only error out if
the starting address is unaligned.

-Scott