[ When we renamed the files, then all the old warnings showed up as
new warnings again. - dan ]
Hello Ajay Singh,
The patch 9bc061e88054: "staging: wilc1000: added support to
dynamically add/remove interfaces" from Jun 26, 2019, leads to the
following static checker warning:
drivers/staging/wilc1000/wlan.c:497 wilc_wlan_handle_txq()
warn: missing error code here? 'wilc_wlan_txq_get_first()' failed.
drivers/staging/wilc1000/wlan.c
474 int wilc_wlan_handle_txq(struct wilc *wilc, u32 *txq_count)
475 {
476 int i, entries = 0;
477 u32 sum;
478 u32 reg;
479 u32 offset = 0;
480 int vmm_sz = 0;
481 struct txq_entry_t *tqe;
482 int ret = 0;
483 int counter;
484 int timeout;
485 u32 vmm_table[WILC_VMM_TBL_SIZE];
486 const struct wilc_hif_func *func;
487 u8 *txb = wilc->tx_buffer;
488 struct net_device *dev;
489 struct wilc_vif *vif;
490
491 if (wilc->quit)
492 goto out;
^^^^^^^^
One of my coding hints is that "goto out;" is always wrong. In the
best case the name is too ambiguous so it doesn't tell what the goto
does. But quite often the goto does too much. For example, it could do
kfree(foo->bar); where foo is NULL so it's a NULL dereference. Always
always distrust code which does a goto out.
In this case we are unlocking a lock but we aren't holding the lock.
Is this a success path? That's complicated to say and becomes even
more complicated when we review the rest of this function. It appears
that this function returns -ENOBUFS or random meaningless nonsense.
493
494 mutex_lock(&wilc->txq_add_to_head_cs);
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Lock.
495 tqe = wilc_wlan_txq_get_first(wilc);
496 if (!tqe)
497 goto out;
^^^^^^^^^
In this case, Smatch complains that maybe the error code is wrong.
Are we setting "*txq_count" to the correct value? It's hard to say.
498 dev = tqe->vif->ndev;
499 wilc_wlan_txq_filter_dup_tcp_ack(dev);
500 i = 0;
501 sum = 0;
502 do {
503 if (tqe && (i < (WILC_VMM_TBL_SIZE - 1))) {
504 if (tqe->type == WILC_CFG_PKT)
505 vmm_sz = ETH_CONFIG_PKT_HDR_OFFSET;
506
507 else if (tqe->type == WILC_NET_PKT)
508 vmm_sz = ETH_ETHERNET_HDR_OFFSET;
[ snip ]
666 ret = func->hif_clear_int_ext(wilc, ENABLE_TX_VMM);
667 if (!ret)
668 goto out_release_bus;
^^^^^^^^^^^^^^^^^^^^
These functions return 1 on success and 0 on failure. We should set the
error code here. There are several other similar places in this
function where we return zero on error.
669
670 ret = func->hif_block_tx_ext(wilc, 0, txb, offset);
671
672 out_release_bus:
673 release_bus(wilc, WILC_BUS_RELEASE_ALLOW_SLEEP);
674
675 out:
676 mutex_unlock(&wilc->txq_add_to_head_cs);
677
678 *txq_count = wilc->txq_entries;
679 return ret;
680 }
regards,
dan carpenter
_______________________________________________
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxx
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel
