Messages in this thread | | | Date | Tue, 30 Nov 2010 11:51:06 +1300 | From | Ryan Mallon <> | Subject | Re: [PATCH] Repalce strncmp by memcmp |
| |
On 11/30/2010 11:26 AM, Steven Rostedt wrote: > On Mon, 2010-11-29 at 17:18 -0500, Steven Rostedt wrote: > >> Um, do you realize that the kernel does not always use the same memcmp >> as gcc. >> > > Note, I'm not against the change, because in 99% of the cases, memcmp() > can and will be faster, and we don't need to worry about these strange > cases.
It can still break things in subtle ways. Lots of the replacements are of the form:
if (strncmp(string, "foo", 3) == 0)
Which can only be replaced with memcmp if the minimum length of string is _always_ 3. This may be true for some callsites (with careful audit), but in general I doubt it is and it will lead to subtle bugs.
I hardly think it is worth auditing a bunch of strncmp calls to ensure that the minimum length of the checked string is always n in order to remove a single instruction. Making such a change will also introduce subtle bugs if the rules for the string ever change, eg. a change is made to allow string = "".
~Ryan
-- Bluewater Systems Ltd - ARM Technology Solution Centre
Ryan Mallon 5 Amuri Park, 404 Barbadoes St ryan@bluewatersys.com PO Box 13 889, Christchurch 8013 http://www.bluewatersys.com New Zealand Phone: +64 3 3779127 Freecall: Australia 1800 148 751 Fax: +64 3 3779135 USA 1800 261 2934
| |