Skip to contentSkip to conversation

drivers/staging/comedi/drivers/ni_pcimio.c:1694 pcimio_attach_pci() warn: 'dev->irq' was not release

DISCUSSION Fengguang Wu 2 participants 2 messages Sep 27, 2012 Kernel janitors

lore
Share & tools
lore

Fengguang Wu

author
#0 lore raw

Hi Bill, Ian,

FYI, this smatch warning probably exists from day 0 of the driver code:

+ drivers/staging/comedi/drivers/ni_pcimio.c:1694 pcimio_attach_pci() warn: 'dev->irq' was not released on error

vim +1694 drivers/staging/comedi/drivers/ni_pcimio.c
1678 dev->irq = mite_irq(devpriv->mite);
1679
1680 if (dev->irq = 0) {
1681 pr_warn("unknown irq (bad)\n");
1682 } else {
1683 pr_debug("( irq = %u )\n", dev->irq);
1684 ret = request_irq(dev->irq, ni_E_interrupt, NI_E_IRQ_FLAGS,
1685 DRV_NAME, dev);
1686 if (ret < 0) {
1687 pr_warn("irq not available\n");
1688 dev->irq = 0;
1689 }
1690 }
1691
1692 ret = ni_E_init(dev);
1693 if (ret < 0)

Quoted 1 line

1694 return ret;

1695
1696 dev->subdevices[NI_AI_SUBDEV].buf_change = &pcimio_ai_change;
1697 dev->subdevices[NI_AO_SUBDEV].buf_change = &pcimio_ao_change;
1698 dev->subdevices[NI_GPCT_SUBDEV(0)].buf_change = &pcimio_gpct0_change;
1699 dev->subdevices[NI_GPCT_SUBDEV(1)].buf_change = &pcimio_gpct1_change;
1700 dev->subdevices[NI_DIO_SUBDEV].buf_change = &pcimio_dio_change;
1701
1702 return ret;

---
0-DAY kernel build testing backend Open Source Technology Centre
Fengguang Wu, Yuanhan Liu Intel Corporation

#1 lore raw

Re: drivers/staging/comedi/drivers/ni_pcimio.c:1694 pcimio_attach_pci() warn: 'dev->irq' was not rel

Quoted 25 lines · Fengguang Wu

Hi Bill, Ian,

FYI, this smatch warning probably exists from day 0 of the driver code:

+ drivers/staging/comedi/drivers/ni_pcimio.c:1694 pcimio_attach_pci() warn: 'dev->irq' was not released on error

vim +1694 drivers/staging/comedi/drivers/ni_pcimio.c

   1678		dev->irq = mite_irq(devpriv->mite);
   1679	
   1680		if (dev->irq = 0) {
   1681			pr_warn("unknown irq (bad)\n");
   1682		} else {
   1683			pr_debug("( irq = %u )\n", dev->irq);
   1684			ret = request_irq(dev->irq, ni_E_interrupt, NI_E_IRQ_FLAGS,
   1685					  DRV_NAME, dev);
   1686			if (ret < 0) {
   1687				pr_warn("irq not available\n");
   1688				dev->irq = 0;
   1689			}
   1690		}
   1691	
   1692		ret = ni_E_init(dev);
   1693		if (ret < 0)
Quoted 1 line

1694 return ret;

Quoted 8 lines

1695

   1696		dev->subdevices[NI_AI_SUBDEV].buf_change = &pcimio_ai_change;
   1697		dev->subdevices[NI_AO_SUBDEV].buf_change = &pcimio_ao_change;
   1698		dev->subdevices[NI_GPCT_SUBDEV(0)].buf_change = &pcimio_gpct0_change;
   1699		dev->subdevices[NI_GPCT_SUBDEV(1)].buf_change = &pcimio_gpct1_change;
   1700		dev->subdevices[NI_DIO_SUBDEV].buf_change = &pcimio_dio_change;
   1701	
   1702		return ret;

It should be fine. pcimio_attach_pci() doesn't clean up after itself, but if it returns an error, pcimio_detach() will be called to clean up afterwards. pcimio_detach() always executes the following statement:

	if (dev->irq)
		free_irq(dev->irq, dev);

where dev is the same dev that was passed to pcimio_attach_pci().

Signature · 2 lines
-=( Ian Abbott @ MEV Ltd.    E-mail: <[email protected]>        )=-
-=( Tel: +44 (0)161 477 1898   FAX: +44 (0)161 718 3587         )=-