linux-kernel - Re: [PATCH v1 RFC 6/6] crypto: implement KFuzzTest targets for PKCS7 and RSA parsing

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] [day] [month] [year] [list]

Message-ID: <CALrw=nGo5CfZseNwM88uqoTDwfmuD7BgXaijpCU-7qefx8+BZA@mail.gmail.com>
Date: Tue, 19 Aug 2025 12:41:00 +0100
From: Ignat Korchagin <ignat@...udflare.com>
To: Marco Elver <elver@...gle.com>
Cc: Eric Biggers <ebiggers@...nel.org>, Ethan Graham <ethan.w.s.graham@...il.com>, 
	ethangraham@...gle.com, glider@...gle.com, andreyknvl@...il.com, 
	brendan.higgins@...ux.dev, davidgow@...gle.com, dvyukov@...gle.com, 
	jannh@...gle.com, rmoar@...gle.com, shuah@...nel.org, tarasmadan@...gle.com, 
	kasan-dev@...glegroups.com, kunit-dev@...glegroups.com, 
	linux-kernel@...r.kernel.org, linux-mm@...ck.org, 
	David Howells <dhowells@...hat.com>, Lukas Wunner <lukas@...ner.de>, 
	Herbert Xu <herbert@...dor.apana.org.au>, "David S. Miller" <davem@...emloft.net>, 
	"open list:HARDWARE RANDOM NUMBER GENERATOR CORE" <linux-crypto@...r.kernel.org>
Subject: Re: [PATCH v1 RFC 6/6] crypto: implement KFuzzTest targets for PKCS7
 and RSA parsing

On Tue, Aug 19, 2025 at 11:08 AM Marco Elver <elver@...gle.com> wrote:
>
> On Fri, 15 Aug 2025 at 15:00, Ignat Korchagin <ignat@...udflare.com> wrote:
> >
> > On Fri, Aug 15, 2025 at 2:18 AM Eric Biggers <ebiggers@...nel.org> wrote:
> > >
> > > On Thu, Aug 14, 2025 at 04:28:13PM +0100, Ignat Korchagin wrote:
> > > > Not sure if it has been mentioned elsewhere, but one thing I already
> > > > don't like about it is that these definitions "pollute" the actual
> > > > source files. Might not be such a big deal here, but kernel source
> > > > files for core subsystems tend to become quite large and complex
> > > > already, so not a great idea to make them even larger and harder to
> > > > follow with fuzz definitions.
> > > >
> > > > As far as I'm aware, for the same reason KUnit [1] is not that popular
> > > > (or at least less popular than other approaches, like selftests [2]).
> > > > Is it possible to make it that these definitions live in separate
> > > > files or even closer to selftests?
> > >
> > > That's not the impression I get.  KUnit suites are normally defined in
> > > separate files, and KUnit seems to be increasing in popularity.
> >
> > Great! Either I was wrong from the start or it changed and I haven't
> > looked there recently.
> >
> > > KFuzzTest can use separate files too, it looks like?
> > >
> > > Would it make any sense for fuzz tests to be a special type of KUnit
> > > test, instead of a separate framework?
> >
> > I think so, if possible. There is always some hurdles adopting new
> > framework, but if it would be a new feature of an existing one (either
> > KUnit or selftests - whatever fits better semantically), the existing
> > users of that framework are more likely to pick it up.
>
> The dependency would be in name only (i.e. "branding"). Right now it's
> possible to use KFuzzTest without the KUnit dependency. So there is
> technical merit to decouple.

Probably strong (Kbuild) dependency is not what I was thinking about,
rather just semantical similarity. That is, if I "learned" KUnit -
KFuzzTest is easy to pick up for me.

> Would sufficient documentation, and perhaps suggesting separate files
> to be the canonical way of defining KFuzzTests, improve the situation?

Probably.

> For example something like:
> For subsystem foo.c, define a KFuzzTest in foo_kfuzz.c, and then in
> the Makfile add "obj-$(CONFIG_KFUZZTEST) += foo_kfuzz.o".
> Alternatively, to test internal static functions, place the KFuzzTest
> harness in a file foo_kfuzz.h, and include at the bottom of foo.c.

Having includes at the bottom of the file feels weird and "leaks"
kfuzz tests into the sources. Perhaps we can somehow rely on the fact
that kernel is a flat address space and you can always get the address
of a symbol (even if static - similar to how eBPF kprobes do it)? Or
have a bit more complex Kbuild configuration: for example
"foo_kfuzz.c" would include "foo.c" (although including .c files also
feels weird). If CONFIG_KFUZZTEST is disabled, Kbuild just includes
"foo.o", if enabled we include "foo_kfuzz.o" (which includes foo.c as
a source).

Ignat

> Alex, Ethan, and KUnit folks: What's your preference?