[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAMuHMdVig-cm7KPi76ow9-xyZ55z4YgYrx6QuO9V5A0ni7HxDA@mail.gmail.com>
Date: Wed, 20 Mar 2024 09:08:42 +0100
From: Geert Uytterhoeven <geert@...ux-m68k.org>
To: Bart Van Assche <bvanassche@....org>
Cc: Mike Christie <michael.christie@...cle.com>,
"James E . J . Bottomley" <jejb@...ux.ibm.com>, "Martin K . Petersen" <martin.petersen@...cle.com>,
Lukas Bulwahn <lukas.bulwahn@...il.com>, linux-scsi@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH] scsi: core: Make scsi_lib KUnit tests modular for real
Hoi Bart,
CC linux-kselftest@...r.kernel.org
On Tue, Mar 19, 2024 at 6:01 PM Bart Van Assche <bvanassche@....org> wrote:
> On 3/19/24 09:10, Geert Uytterhoeven wrote:
> > On Tue, Mar 19, 2024 at 5:03 PM Bart Van Assche <bvanassche@...org> wrote:
> >> On 3/19/24 05:02, Geert Uytterhoeven wrote:
> >> kernel module? What are the advantages compared to the current approach?
> >> That information is missing from the patch description.
> >
> > SCSI_LIB_KUNIT_TEST is already tristate, so the original author must
> > have meant it to be modular. Or perhaps he just copied it from
> > (most/all) other tests ;-)
> >
> > Anyway, I find it very useful to be able to do "modprobe kunit" and
> > "modprobe <test>" to run a test when I feel the need to do so.
>
> Why to run hardware-independent kunit tests on the target system instead
> of on the host? Isn't it much more convenient when developing embedded
> software to run kunit tests on the host using UML? The script I use to
Because test results may differ between target and host?
It's not uncommon for supposedly hardware-independent tests to behave
differently on different architectures and platforms, due to subtle
differences in word size, endianness, alignment rules, CPU topology, ...
> run SCSI kunit tests is available below. And if there is a desire to run
> SCSI tests on the target system, how about adding triggers in sysfs for
> running kunit tests? The (GPL v2) Samsung smartphone kernel supports
> this but I have not yet checked whether their implementation is
> appropriate for the upstream kernel.
That would require all tests to be built-in, reducing the amount of memory
(if any remains at all) available to the real application.
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@...ux-m68korg
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
Powered by blists - more mailing lists