Hello, Dan, Thank you very much for helping check the driver, from the code logical view, this could happen. Just as you say, list_for_each_entry_safe() is a proper substitute. Then should I submit a fix on this or you have had one? Thanks! Best Regards, Jingchang > -----Original Message----- > From: Dan Carpenter [mailto:dan.carpenter@xxxxxxxxxx] > Sent: Thursday, February 20, 2014 7:30 PM > To: Lu Jingchang-B35083 > Cc: dmaengine@xxxxxxxxxxxxxxx > Subject: re: dma: Add Freescale eDMA engine driver support > > Hello Jingchang Lu, > > The patch d6be34fbd39b: "dma: Add Freescale eDMA engine driver support" > from Feb 18, 2014, leads to the following static checker > warning: > > drivers/dma/fsl-edma.c:732 fsl_edma_xlate() > error: we previously assumed 'chan' could be null (see line 737) > > drivers/dma/fsl-edma.c > 731 mutex_lock(&fsl_edma->fsl_edma_mutex); > 732 list_for_each_entry(chan, &fsl_edma->dma_dev.channels, > device_node) { > ^^^^^^^^^^^^^^^^^^^ > This will have a NULL dereference if ... > > 733 if (chan->client_count) > 734 continue; > 735 if ((chan->chan_id / DMAMUX_NR) == dma_spec- > >args[0]) { > 736 chan = dma_get_slave_channel(chan); > 737 if (chan) { > ^^^^ ... if "chan" is NULL here. > > 738 chan->device->privatecnt++; > 739 > fsl_edma_chan_mux(to_fsl_edma_chan(chan), > 740 dma_spec->args[1], true); > 741 mutex_unlock(&fsl_edma- > >fsl_edma_mutex); > 742 return chan; > 743 } > 744 } > 745 } > 746 mutex_unlock(&fsl_edma->fsl_edma_mutex); > 747 return NULL; > > It's most likely that list_for_each_entry_safe() was intended instead of > list_for_each_entry() but I'm not sure about this enough to make the > change myself. > > regards, > dan carpenter > ��.n��������+%������w��{.n��������)�)�� jg��� �����ݢj����G���� ���j:+v���w�m������w�������h�����٥
