linux-kernel - Re: [RFC] globmatch() helper function

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20081218080027.27011.qmail@science.horizon.com>
Date:	Thu, 18 Dec 2008 03:00:27 -0500
From:	"George Spelvin" <linux@...izon.com>
To:	srostedt@...hat.com, peterz@...radead.org
Cc:	tj@...nel.org, linux@...izon.com, linux-kernel@...r.kernel.org,
	andi@...stfloor.org
Subject: Re: [RFC] globmatch() helper function

Steven Rostedt <srostedt@...hat.com> wrote:
> I need to look at your code (I would like a generalized glob feature for
> user input). Can you accomplish the same with using a loop instead of
> recursion?

Unfortunately, no.  Matching multiple * characters inherently requires
backtracking, and backtracking requires a stack.

I could make it a finite-depth explicit stack rather than using
the C stack implicitly, but it doesn't save much on reasonable
patterns.

But if you want it for user input, I can work at making it more
robust.  What are your sped needs?
0 - Runs once only, speed unimportant
1 - Runs several thousand times (counting all NxM matches), would like
    it to not be too slow.
2 - Run a thousand times per second, but it's a debug feature so
    a speed hit is acceptable.
3 - Common code path, maybe rearchitect to save time.
4 - Kernel fast path, every cycle counts!

There are quite a few things (with commensurate code size increases)
that can't help the big-O worst-case performance, but improve the average
case considerably.

One thing I don't do that's quite possible is to find the fixed-length
head and tail of the pattern (the parts before the first * and after the
last) and match those before dealing with the annoying middle.

That would avoid all backtracking in cases with a single *, and make a
good pre-filter before doing anything expensive in the middle.

H'm... I think there's an equivalent and more general version based on
the number of characters a pattern matches.  Any pattern's length is
N characters, plus a flag indicating if that's fixed or just a minimum.

When matching a *, compute the length of the following pattern and only
try matching cases which have the right length.

If we need to do a lot of matching against a single pattern, it's possible
to preprocess it to find the fixed-length spans and their lengths.
(The current blacklist application is exactly the opposite: matching
one string against a list of patterns.)
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/