pm80xx: Print SAS address of IO failed device.

Hello Anand Kumar Santhanam,

This is a semi-automatic email about new static checker warnings.

The patch cb269c26ed02: "[SCSI] pm80xx: Print SAS address of IO failed device." from Sep 17, 2013, leads to the following Smatch complaint:

drivers/scsi/pm8001/pm8001_hwi.c:2343 mpi_sata_completion()

	 error: we previously assumed 't->dev' could be null (see line 2319)

drivers/scsi/pm8001/pm8001_hwi.c
2318 if (t) {
2319 if (t->dev && (t->dev->lldd_dev))

                            ^^^^^

Exisiting check.

2320 pm8001_dev = t->dev->lldd_dev;
2321 } else {
2322 PM8001_FAIL_DBG(pm8001_ha,
2323 pm8001_printk("task null\n"));
2324 return;
2325 }
2326
2327 if ((pm8001_dev && !(pm8001_dev->id & NCQ_READ_LOG_FLAG))

  2328			&& unlikely(!t || !t->lldd_task || !t->dev)) {
                                                           ^^^^^^^

Existing check.

2329 PM8001_FAIL_DBG(pm8001_ha,
2330 pm8001_printk("task or dev null\n"));
2331 return;
2332 }
2333
2334 ts = &t->task_status;
2335 if (!ts) {
2336 PM8001_FAIL_DBG(pm8001_ha,
2337 pm8001_printk("ts null\n"));
2338 return;
2339 }
2340 /* Print sas address of IO failed device */
2341 if ((status != IO_SUCCESS) && (status != IO_OVERFLOW) &&
2342 (status != IO_UNDERFLOW)) {
2343 if (!((t->dev->parent) &&

                               ^^^^^^^^

Patch introduces new unchecked dereference.

  2344				(DEV_IS_EXPANDER(t->dev->parent->dev_type)))) {
  2345				for (i = 0 , j = 4; j <= 7 && i <= 3; i++ , j++)

regards,
dan carpenter