05.08.2021 12:14, Linus Walleij пишет:
> On Thu, Aug 5, 2021 at 3:22 AM Dmitry Osipenko <digetx@xxxxxxxxx> wrote:
>> 04.08.2021 18:34, Linus Walleij пишет:
>>> +static void brcmf_fw_request_done_first(const struct firmware *fw, void *ctx)
>>> {
>>> struct brcmf_fw *fwctx = ctx;
>>> + struct brcmf_fw_item *first = &fwctx->req->items[0];
>>> int ret;
>>>
>>> + /* Something failed with the first firmware request, such as not
>>> + * getting the per-board firmware. Retry this, now using the less
>>> + * specific path for the first firmware item, i.e. without the board
>>> + * suffix.
>>> + */
>>> + if (!fw && !fwctx->tested_board_variant) {
>>> + fwctx->tested_board_variant = true;
>>> + ret = request_firmware_nowait(THIS_MODULE, true, first->path,
>>> + fwctx->dev, GFP_KERNEL, fwctx,
>>> + brcmf_fw_request_done_first);
>>> + return;
>>
>> The original code was proceeding on error. Is this a typo here?
>
> No, we are testing specifically for fw being NULL and in that case we issue
> a new request_firmware_nowait() call with ourselves as "done" callback,
> so we really need to return here.
>
> The worker will call the same function again after this but now
> tested_board_variant is true.
The worker won't call the same function if request_firmware_nowait() fails.
