Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20200110134337.1752000-1-arnd@arndb.de>
In-Reply-To: <20200110134337.1752000-1-arnd@arndb.de>
From:   Brendan Higgins <brendanhiggins@google.com>
Date:   Mon, 13 Jan 2020 18:12:52 -0800
Message-ID: <CAFd5g47+_oyqsS0o0kQ+CaLNtjqbvOmQc-n0Ch1jAT6P6RSFiw@mail.gmail.com>
Subject: Re: [PATCH] [RFC] kunit: move binary assertion out of line
To:     Arnd Bergmann <arnd@arndb.de>
Cc:     Kees Cook <keescook@chromium.org>,
        Ard Biesheuvel <ard.biesheuvel@linaro.org>,
        Stephen Boyd <sboyd@kernel.org>,
        Dmitry Torokhov <dmitry.torokhov@gmail.com>,
        "Rafael J. Wysocki" <rjw@rjwysocki.net>,
        "open list:KERNEL SELFTEST FRAMEWORK" 
        <linux-kselftest@vger.kernel.org>,
        KUnit Development <kunit-dev@googlegroups.com>,
        Shuah Khan <skhan@linuxfoundation.org>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Logan Gunthorpe <logang@deltatee.com>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Fri, Jan 10, 2020 at 5:43 AM Arnd Bergmann <arnd@arndb.de> wrote:
>
> In combination with the structleak gcc plugin, kunit can lead to excessive
> stack usage when each assertion adds another structure to the stack from
> of the calling function:
>
> base/test/property-entry-test.c:99:1: error: the frame size of 3032 bytes is larger than 2048 bytes [-Werror=frame-larger-than=]
>
> As most assertions are binary, change those over to a direct function
> call that does not have this problem.  This can probably be improved
> further, I just went for a straightforward conversion, but a function
> call with 12 fixed arguments plus varargs it not great either.

Yeah, I am not exactly excited by maintaining such a set of functions.

I don't think anyone wants to go with the heap allocation route.

Along the lines of the union/single copy idea[1]. What if we just put
a union of all the assertion types in the kunit struct? One is already
allocated for every test case and we only need one assertion object
for each test case at a time, so I imagine that sould work.

I will start messing around with the idea. Still, it sounds like we
are down to either reducing the number of instances of this struct
that get created per test case, or we need to remove it entirely (as
you have done here).

Cheers