[bug report] octeontx2-af: support for coalescing KPU profiles

[Dan Carpenter <dan.carpenter@oracle.com>]

Hello Harman Kalra,

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

The patch 11c730bfbf5b: "octeontx2-af: support for coalescing KPU 
profiles" from May 27, 2021, leads to the following Smatch complaint:

    drivers/net/ethernet/marvell/octeontx2/af/rvu_npc.c:1624 npc_load_kpu_profile_fwdb()
    warn: variable dereferenced before check 'rvu->kpu_prfl_addr' (see line 1619)

drivers/net/ethernet/marvell/octeontx2/af/rvu_npc.c
  1608  static int npc_load_kpu_profile_fwdb(struct rvu *rvu, const char *kpu_profile)
  1609  {
  1610          int ret = -EINVAL;
  1611          u64 prfl_sz;
  1612  
  1613          /* Setting up the mapping for NPC profile image */
  1614          ret = npc_fwdb_prfl_img_map(rvu, &rvu->kpu_prfl_addr, &prfl_sz);
  1615          if (ret < 0)
  1616                  goto done;
  1617  
  1618          /* Detect if profile is coalesced or single KPU profile and load */
  1619          ret = npc_fwdb_detect_load_prfl_img(rvu, prfl_sz, kpu_profile);

rvu->kpu_prfl_addr dereferenced inside function call

  1620          if (ret == 0)
  1621                  goto done;
  1622  
  1623          /* Cleaning up if KPU profile image from fwdata is not valid. */
  1624          if (rvu->kpu_prfl_addr) {
                    ^^^^^^^^^^^^^^^^^^
Smatch isn't supposed to warn about places where the NULL check is
*obviously* not required but I think I messed up something.  Anyway,
this NULL check doesn't make sense.

  1625                  iounmap(rvu->kpu_prfl_addr);
  1626                  rvu->kpu_prfl_addr = NULL;
  1627                  rvu->kpu_fwdata_sz = 0;
  1628                  rvu->kpu_fwdata = NULL;
  1629          }
  1630  
  1631  done:
  1632          return ret;
  1633  }

regards,
dan carpenter