On 20/01/2022 06.45, Andy Shevchenko wrote:

On Mon, Jan 17, 2022 at 4:30 PM Hector Martin [off-list ref] wrote:

quoted

This unbreaks support for USB devices, which do not have a board_type
to create an alt_path out of and thus were running into a NULL
dereference.

In v5.16 we have two call sites:

1.
  if (cur->type == BRCMF_FW_TYPE_NVRAM && fwctx->req->board_type) {
    ...
    alt_path = brcm_alt_fw_path(cur->path, fwctx->req->board_type);

2.
  alt_path = brcm_alt_fw_path(first->path, fwctx->req->board_type);
  if (alt_path) {
    ...

Looking at them I would rather expect to see (as a quick fix, the
better solution is to unify those call sites by splitting out a common
helper):

  if (fwctx->req->board_type) {
    alt_path = brcm_alt_fw_path(first->path, fwctx->req->board_type);
  else
    alt_path = NULL;
   ...

Since brcm_alt_fw_path can fail anyway, and its return value is already
NULL-checked, it makes sense to propagate the NULL board_path there
rather than doing it at all the callsites. That's a common pattern, e.g.
the entire DT API is designed to accept NULL nodes. That does mean that
the first callsite has a redundant NULL check, yes, but that doesn't hurt.

This is all going to change with subsequent patches anyway; the point of
this patch is just to fix the regression.

-- 
Hector Martin (marcan@marcan.st)
Public Key: https://mrcn.st/pub