linux-kernel - Re: [RFC v2 2/3] pinctrl: Implementation of the generic scmi-pinctrl driver

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: <20230512121801.GA3494263@EPUAKYIW0A6A>
Date:   Fri, 12 May 2023 12:18:03 +0000
From:   Oleksii Moisieiev <Oleksii_Moisieiev@...m.com>
To:     Cristian Marussi <cristian.marussi@....com>
CC:     "andy.shevchenko@...il.com" <andy.shevchenko@...il.com>,
        "sudeep.holla@....com" <sudeep.holla@....com>,
        Linus Walleij <linus.walleij@...aro.org>,
        Peng Fan <peng.fan@....nxp.com>,
        Michal Simek <michal.simek@....com>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-arm-kernel@...ts.infradead.org" 
        <linux-arm-kernel@...ts.infradead.org>,
        "linux-gpio@...r.kernel.org" <linux-gpio@...r.kernel.org>
Subject: Re: [RFC v2 2/3] pinctrl: Implementation of the generic scmi-pinctrl
 driver

Hello Cristian,

On Fri, May 12, 2023 at 10:04:41AM +0100, Cristian Marussi wrote:
> On Thu, May 11, 2023 at 01:15:46PM +0000, Oleksii Moisieiev wrote:
> > Hello Andy,
> > 
> > On 05.05.23 23:35, andy.shevchenko@...il.com wrote:
> > > Wed, Apr 26, 2023 at 01:26:37PM +0000, Oleksii Moisieiev kirjoitti:
> > >> scmi-pinctrl driver implements pinctrl driver interface and using
> > >> SCMI protocol to redirect messages from pinctrl subsystem SDK to
> > >> SCP firmware, which does the changes in HW.
> > >>
> > >> This setup expects SCP firmware (or similar system, such as ATF)
> > >> to be installed on the platform, which implements pinctrl driver
> > >> for the specific platform.
> > >>
> > >> SCMI-Pinctrl driver should be configured from the device-tree and uses
> > >> generic device-tree mappings for the configuration.
> > > 
> > > ...
> > > 
> > >> +#include <linux/device.h>
> > >> +#include <linux/err.h>
> > > 
> > >> +#include <linux/of.h>
> > > 
> > > I do not see any user of this header. Do you?
> > > 
> > Yes, thanks. Removing
> > 
> > >> +#include <linux/module.h>
> > >> +#include <linux/seq_file.h>
> > >> +
> > >> +#include <linux/pinctrl/machine.h>
> > >> +#include <linux/pinctrl/pinconf.h>
> > >> +#include <linux/pinctrl/pinconf-generic.h>
> > >> +#include <linux/pinctrl/pinctrl.h>
> > >> +#include <linux/pinctrl/pinmux.h>
> > > 
> > >> +#include <linux/scmi_protocol.h>
> > >> +#include <linux/slab.h>
> > > 
> > > Please, move these two to the upper group of the generic headers.
> > > 
> > Thanks, fixed.
> > 
> > >> +struct scmi_pinctrl_funcs {
> > >> +	unsigned int num_groups;
> > >> +	const char **groups;
> > >> +};
> > > 
> > > Please, use struct pinfunction.
> > >
> > I can't use pincfunction here because it has the following groups 
> > definition:
> > const char * const *groups;
> > 
> > Which is meant to be constantly allocated.
> > So I when I try to gather list of groups in 
> > pinctrl_scmi_get_function_groups I will receive compilation error.
> > 
> 
> Maybe this is a further signal that we should re-evaluate the benefits of
> the lazy allocations you now perform during protocol initialization
> instead of querying and allocating statically all the info structs about
> existing resources.
> 
> Not saying that is necessarily bad, I understood your points about reducing
> the number of SCMI queries during boot and let pinctrl subsystem trigger only
> the strictly needed one, just saying maybe good to reason a bit more about this
> once V3 is posted. (i.e. I could bother you more :P ..)
> 
> Thanks,
> Cristian
> 
> P.S. [off-topic]: remember to use get_maintainer.pl as advised elsewhere
> to include proper maintainers (and their bots)

That's a good point to think about. Actually, functions are the only
thing that should be cached on pinctrl side. And we need it specifically
because groups in each function are presented by names, not selectors.
Maybe It's better to move this caching to pinctrl scmi driver. But, from
the other side - storing group names for each function is Linux Kernel
specific implementation and we probably don't want to add some specific
case to the Generic protocol driver.

I think I would leave it as in V3 so we can continue discussion.

Oleksii.