> +static int enter_ffu_mode(int *dev_fd)
> +{
> + int ret;
> + struct mmc_ioc_cmd cmd;
> + memset(&cmd, 0, sizeof(cmd));
>
> - /*
> - * send image chunk: blksz and blocks essentially do not matter, as
> - * long as the product is fw_size, but some hosts don't handle larger
> - * blksz well.
> - */
> - set_single_cmd(&multi_cmd->cmds[2], MMC_WRITE_MULTIPLE_BLOCK,
> 1, bytes / 512, arg);
> - mmc_ioc_cmd_set_data(multi_cmd->cmds[2], buf + offset);
> + fill_switch_cmd(&cmd, EXT_CSD_MODE_CONFIG, EXT_CSD_FFU_MODE);
> + ret = ioctl(*dev_fd, MMC_IOC_CMD, &cmd);
> + if (ret)
> + perror("enter FFU mode faled\n");
> +
> + return ret;
> +}
Thinking aloud if we need the enter phase at all, because:
1) The first cmd6 is always part of the sequence, and
2) is it even safe to send it as a single ioctl?
I guess the answer to 2) is yes, and having a start phase actually make sense:
it makes clearer of what the ffu process comprised of:
Start (enter ffu mode), repeat...., finish up.
I would even add a comment in the code (and in the commit log) for all modes other than 1,
emphasizing that from this point onward, the device is blocked in ffu mode.
Thanks,
Avri
