[PATCH][WAS:bcmai,axi] bcma: add Broadcom specific AMBA bus driver

Russell King - ARM Linux linux at arm.linux.org.uk
Sun May 8 10:50:27 UTC 2011


On Sun, May 08, 2011 at 12:37:15PM +0200, Rafał Miłecki wrote:
> 2011/5/8 Russell King - ARM Linux <linux at arm.linux.org.uk>:
> > On Sat, May 07, 2011 at 08:48:10PM +0200, Rafał Miłecki wrote:
> >> Really, what's wrong with that? Does it kill anyone's pet we print
> >> this? We also do:
> >> pr_err("Scanning failed because of wrong CID\n");
> >> return -1;
> >> While we could drop pr_err. Why to do this? Advanced used can always
> >> check what -1 means.
> >
> > And why return -1 when we have a system of error codes?  I _really_ wish
> > people would stop returning -1 for "some random error occurred".
> 
> You commented on imagined code, but we actually do sth similar in code.
> 
> I did this because:
> 1) I had no idea what err code would be valid for invalid EPROM layout
> (content). Nothing from include/asm-generic/errno-base.h sounds
> reasonable.
> 2) I wanted to use different error codes for different EPROM layout
> issues. Sometimes we don't get CIA block. Sometimes we don't get CIB
> block. Sometimes there is problem with master port (not found in EPROM
> when expected). They all would probably use the same errno.
> 
> Could you help me with this?

The problem is if you start using -1 and mixing it with stuff which does
return negative errno codes, you end up hitting one of two bugs:

1. you interpret -1 as being -EPERM when actually you meant something else.
2. you check the function's return value for -1 rather than < 0, and you
   unintentionally ignore valid -errno codes.

So it's normally far better to find something in the errno stuff which
approximates the error you have rather than using -1.  Eg, if something
is invalid and you can't find something which fits, -EINVAL is probably
a good idea.

If you can't access the eeprom because its not responding, maybe -EIO
or -ETIMEDOUT would be better than -1?

Maybe for CRC errors, or unexpected data -EILSEQ would be appropriate?

Maybe if something being requested isn't found, -ENOENT would be better
(may not be a file or directory, but it approximates the error as being
'error no entry').



More information about the devel mailing list