Re: + drivers-ide-fix-error-return-bugs-interface.patch added to -mm tree

Re: + drivers-ide-fix-error-return-bugs-interface.patch added to -mm tree

[Alan Cox]

Ar Maw, 2006-10-10 am 13:47 -0700, ysgrifennodd akpm@osdl.org:
> This fixes the following error handling problems:
> 
> * The init_chipset API function is defined to return 'unsigned int', but
> 	- the caller code tests the return value for '< 0'
> 	- drivers sometimes return a negative value

Dropped for good. The core code doesn't support failing in init_chipset.
Instead I've adjusted via82cxxx to do the check earlier in init_one().
Also avoids an API change and all the noise.

> * cs5530: handle pci_set_mwi() failure with a printk()... shouldn't kill driver

NAK. Fix is to remove bogus must_check from pci_set_mwi (and some of the
other functions)

> * sc1200: handle pci_enable_device() failure during resume

NAK: if the pci_enable_device fails here the best we can do is attempt
to get it back and hope the pci_enable_device failure is bogus. This is
a can't happen case anyway so its a waste of memory. If you insist on
checking pci_enable_device returns everywhere then please move the
printk into the pci_enable_device function so that we don't bloat the
kernel with a load of pointless identically messages for an event that
never happen, one per driver.

Alan