linux-kernel - Re: [PATCH 2/2] platform/x86: firmware_attributes

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: <acf81e37-3b3a-48ef-8b7a-8e8c50a161a4@t-8ch.de>
Date: Tue, 7 Jan 2025 23:47:02 +0100
From: Thomas Weißschuh <linux@...ssschuh.net>
To: Mario Limonciello <mario.limonciello@....com>
Cc: Hans de Goede <hdegoede@...hat.com>, 
	Ilpo Järvinen <ilpo.jarvinen@...ux.intel.com>, Armin Wolf <W_Armin@....de>, platform-driver-x86@...r.kernel.org, 
	linux-kernel@...r.kernel.org, Joshua Grisham <josh@...huagrisham.com>
Subject: Re: [PATCH 2/2] platform/x86: firmware_attributes_class: Add test
 driver

On 2025-01-07 16:20:14-0600, Mario Limonciello wrote:
> On 1/7/2025 16:13, Thomas Weißschuh wrote:
> > On 2025-01-07 15:18:21-0600, Mario Limonciello wrote:
> > > On 1/7/2025 14:50, Thomas Weißschuh wrote:
> > > > On 2025-01-07 13:29:08-0600, Mario Limonciello wrote:
> > > > > On 1/7/2025 11:05, Thomas Weißschuh wrote:
> > > > > > The driver showcases the use of the new subsystem API.
> > > > > > 
> > > > > > Signed-off-by: Thomas Weißschuh <linux@...ssschuh.net>
> > > > > > ---
> > > > > >     drivers/platform/x86/Kconfig                    | 12 ++++
> > > > > >     drivers/platform/x86/Makefile                   |  1 +
> > > > > >     drivers/platform/x86/firmware_attributes_test.c | 78 +++++++++++++++++++++++++
> > > > > >     3 files changed, 91 insertions(+)
> > > > > > 
> > > > > > diff --git a/drivers/platform/x86/Kconfig b/drivers/platform/x86/Kconfig
> > > > > > index 0258dd879d64be389f4dd9bc309fe089f23cc798..2a0e828657d2f07074944d6c42dc204fc8825a42 100644
> > > > > > --- a/drivers/platform/x86/Kconfig
> > > > > > +++ b/drivers/platform/x86/Kconfig
> > > > > > @@ -1065,6 +1065,18 @@ source "drivers/platform/x86/x86-android-tablets/Kconfig"
> > > > > >     config FW_ATTR_CLASS
> > > > > >     	tristate
> > > > > > +config FW_ATTR_TEST
> > > > > > +	tristate "Firmware attribute test driver"
> > > > > > +	select FW_ATTR_CLASS
> > > > > > +	help
> > > > > > +	  This driver provides a test user of the firmware attribute subsystem.
> > > > > > +
> > > > > > +	  An instance is created at /sys/class/firmware-attributes/test/
> > > > > > +	  container various example attributes.
> > > > > > +
> > > > > > +	  To compile this driver as a module, choose M here: the module
> > > > > > +	  will be called firmware_attributes_test.
> > > > > > +
> > > > > 
> > > > > I think if you're going to be introducing a test driver it should be
> > > > > compliant to what's in sysfs-class-firmware-attributes so that when it's
> > > > > inevitably copy/pasted we end up with higher quality drivers.
> > > > > 
> > > > > That is you need at a minimum 'type', 'current_value', 'default_value',
> > > > > 'display_name' and 'display_name_language_code'.  Then individual types
> > > > > employ additional requirements.
> > > > > 
> > > > > I see 'type', 'current_value', and 'default_value, but I don't see
> > > > > 'display_name' or 'display_name_language_code' here.
> > > > > 
> > > > > Furthermore as this is a "string" attribute you're supposed to also have a
> > > > > "max_length" and "min_length".
> > > > 
> > > > Agreed that more examples are better.
> > > > 
> > > > But are these attributes really mandatory?
> > > > "This attribute is mandatory" is only specified for "type" and>
> > > "current_value".
> > > 
> > > Ah wow, I had thought they were, but you're right they're not!
> > > 
> > > > While "possible_values" certainly looks necessary for "enumeration",
> > > > "min_length" for "strings" does so much less.
> > > 
> > > Even if they're not mandatory, I think it's better to include them for the
> > > same copy/paste reason I mentioned though.
> > 
> > Thinking about it some more, which attributes should all be included?
> > Having all of them in there could motivate driver authors to implement
> > them all even it would mean filling in random values.
> > The provided examples can already be copied-and-pasted and slightly
> > adapted to add more attributes.
> > 
> > Also as for providing an even higher level interface. There exists a
> > fairly big feature matrix. For example the display_name_language_code:
> > * Is it used at all?
> > * Is it the same for all attributes implemented by the driver and the
> >    sysfs attribute can be reused for them all.
> > * Should the same handler logic be reused between different settings which
> >    only differ in their language code?
> > 
> > The answers seem to differ between each driver and having a uniform
> > high-level interface that can handle all cases would look horrible.
> > So I'd like to stick with the API provided currently for now and we can
> > revisit it if there are more drivers.
> > As mentioned before, the current API should be a decent baseline to
> > build on top of.
> > 
> > 
> > Thomas
> 
> How about just adding min_length and max_length?  Since you're adding a
> string attribute then it makes a good example of a string attribute.
> It should be pretty trivial; AFAICT it's 0 to PAGE_SIZE.

IMO attributes should only exist if they provide useful, driver-specific
information. The proposed values are exactly what I'd like to avoid being
copied around.
min_length=0 does not add any information.
max_length=PAGE_SIZE exposes an implementation detail of sysfs. If that
limit should become an issue for a driver they can switch to a bin_attribute.

I would prefer using get_random_u32() for both attributes, so it's clear
that they are only placeholders.

But I'm also open to getting outvoted.