Re: Can we drop __must_check from driver_for_each_device()?

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Wed, Jul 31, 2013 at 11:35:13PM +0200, Paul Bolle wrote:
> Greg,
> 
> 0) Summary: I think __must_check can be dropped from
> driver_for_each_device(). Do you agree?

No.

> 1) Commit 4a7fb6363f ("add __must_check to device management code")
> added __must_check to driver_for_each_device(), seven years ago. But of
> the seventeen current users of that function just one actually seems to
> care about its return value:
> 
> - dead code:
>   drivers/mtd/onenand/omap2.c:omap2_onenand_rephase()
> 
> - callback always returns zero:
>   drivers/infiniband/hw/qib/qib_init.c:qib_notify_dca()
>     (see all three possible values of f_notify_dca())
>   drivers/media/pci/cx25821/cx25821-alsa.c:cx25821_audio_fini()
>   drivers/media/pci/cx25821/cx25821-alsa.c:cx25821_alsa_init()
>   drivers/net/can/usb/peak_usb/pcan_usb_core.c:peak_usb_exit()
>   drivers/net/ethernet/intel/igb/igb_main.c:igb_notify_dca()
>   drivers/net/ethernet/intel/ixgbe/ixgbe_main.c:ixgbe_notify_dca()
>   drivers/net/ethernet/myricom/myri10ge/myri10ge.c:myri10ge_notify_dca()
> 
> - return value of callback ignored:
>   drivers/iommu/omap-iommu.c:omap_foreach_iommu_device()
>     (when called by iommu_debugfs_exit())
>   drivers/isdn/hardware/mISDN/hfcpci.c:hfcpci_softirq()
>   drivers/media/pci/cx18/cx18-alsa-main.c:cx18_alsa_exit()
>   drivers/media/pci/ivtv/ivtv-alsa-main.c:ivtv_alsa_exit()
>   drivers/media/pci/ivtv/ivtvfb.c:ivtvfb_init()
>   drivers/media/pci/ivtv/ivtvfb.c:ivtvfb_cleanup()
>   drivers/media/platform/s5p-tv/mixer_video.c:find_and_register_subdev()
>   drivers/net/wireless/brcm80211/brcmfmac/usb.c:brcmf_usb_exit()
> 
> - return value actually used:
>   drivers/iommu/omap-iommu.c:omap_foreach_iommu_device()
>     (when called by iommu_debug_init())
> 
> 2) Please note that if the callback always returns zero,
> driver_for_each_device() can still return -EINVAL, but only if it was
> provided a NULL "drv" (a struct device_driver). It sure seems odd to do
> so. Can that actually happen?

Possibly.

> 3) So to me it looks the __must_check attribute of
> driver_for_each_device() can be dropped. Do you agree?

Nope, it should be making people think about the return value of the
function.  If they use it or not might be a problem, but I would argue
that those call-sites must be fixed, as you point out above.

Is this somehow causing a problem that removing the marking would solve
for you?

thanks,

greg k-h