Hello Bjorn Andersson,
The patch 7a20917d30fb: "device property: Add helper to match
multiple connections" from Apr 22, 2022 (linux-next), leads to the
following Smatch static checker warning:
drivers/base/property.c:1374 fwnode_devcon_matches()
warn: 'ret' is an error pointer or valid
drivers/base/property.c
1353 static unsigned int fwnode_devcon_matches(const struct fwnode_handle *fwnode,
1354 const char *con_id, void *data,
1355 devcon_match_fn_t match,
1356 void **matches,
1357 unsigned int matches_len)
1358 {
1359 struct fwnode_handle *node;
1360 unsigned int count = 0;
1361 unsigned int i;
1362 void *ret;
1363
1364 for (i = 0; ; i++) {
1365 if (matches && count >= matches_len)
1366 break;
1367
1368 node = fwnode_find_reference(fwnode, con_id, i);
1369 if (IS_ERR(node))
1370 break;
1371
1372 ret = match(node, NULL, data);
This is hard to analyze. Smatch is doing something wrong but it still
ends up looking basically like Smatch is right?
According to Smatch on my config these are the match() functions
usb_role_switch_match(), typec_retimer_match(), typec_switch_match()
and typec_mux_match(). They can all return ERR_PTR(-EPROBE_DEFER).
1373 fwnode_handle_put(node);
--> 1374 if (ret) {
1375 if (matches)
1376 matches[count] = ret;
^^^^^^^^^^^^^^^^^^^^^
Do we want to record that here or treat it as an error?
1377 count++;
1378 }
1379 }
1380
1381 return count;
1382 }
regards,
dan carpenter
