Re: linux-next: Tree for January 12 (cifs vs. staging)

["Steve French" <smfrench@gmail.com>]

On Mon, Jan 12, 2009 at 2:21 PM, Greg KH <gregkh@suse.de> wrote:
> On Mon, Jan 12, 2009 at 02:13:29PM -0600, Steve French wrote:
>> On Mon, Jan 12, 2009 at 1:32 PM, Greg KH <gregkh@suse.de> wrote:
>> > On Mon, Jan 12, 2009 at 01:23:07PM -0600, Steve French wrote:
>> >> Is there some magic compile option alternative (other than renaming
>> >> the function, which may be ok, since others probably don't depend on
>> >> it)?  MD5Init is used outside fs/cifs/md5.c so can't be static
>> >
>> > No, you should rename it to cifs_md5init to show that it is only for the
>> > cifs module to use.  Try not to polute the global namespace with generic
>> > function names.
>>
>> Looks like cifs has the following related functions which I can rename
>> if that makes it easier:
>> 001a2e9 R_386_PC32        MD5Final
>> 0001a302 R_386_PC32        MD5Update
>> 0001a38c R_386_PC32        MD5Init
>> 0001a39f R_386_PC32        MD5Update
>
> That would make life easier for you as well :)
>
>> Any others that I missed ...
>
> Not that I see at the moment, but you might want to look at all of your
> global symbol names to make sure they start with "cifs_".

I had gone through the cifs.ko module's namespace a few times years
ago to make sure that most of the names were unlikely to cause
problems, but it is a moving target (functions get added and static
ones get moved etc. over time), and keeping functions names reasonably
short but still descriptive can be helpful (as long as they are likely
to be unique).   Looking at it now, the vast majority have something
smb/cifs specific in the function name (e.g. cifs or SMB or less
commonly: lanman or trans2 or ntlm etc. which are also likely to be
unique)


-- 
Thanks,

Steve