linux-kernel - Re: [PATCH] power: supply: core: Ignore -EIO for uevent

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: <YwgdraBXTWk1DhCE@google.com>
Date:   Thu, 25 Aug 2022 18:11:09 -0700
From:   Brian Norris <briannorris@...omium.org>
To:     Sebastian Reichel <sebastian.reichel@...labora.com>
Cc:     linux-kernel@...r.kernel.org, linux-pm@...r.kernel.org
Subject: Re: [PATCH] power: supply: core: Ignore -EIO for uevent

Hi Sebastian,

Thanks for the response.

On Thu, Aug 25, 2022 at 04:02:43PM +0200, Sebastian Reichel wrote:
> > For uevents, we enumerate all properties. Some battery implementations
> > don't implement all standard properties, and may return -EIO for
> > properties that aren't recognized. This means we never report uevents
> > for such batteries.
> > 
> > It's better to ignore these errors and skip the property, as we do with
> > ENODATA and ENODEV.
> > 
> > Example battery implementation: Acer Chromebook Tab 10 (a.k.a. Google
> > Gru-Scarlet) has a virtual "SBS" battery implementation in its Embedded
> > Controller on top of an otherwise non-SBS battery.
> > 
> > Signed-off-by: Brian Norris <briannorris@...omium.org>
> > ---
> 
> -EIO means input/output error. If a driver is reporting that for an
> unimplemented feature it's a bug that should be fixed in the driver.
> Handling it here means userspace ABI changes for temporary issues.

I suppose I can agree with your last sentence.

But the first part is much easier said than done. This is sbs-battery.c,
on top of i2c-cros-ec-tunnel.c, talking to an EC (whose firmware is
pretty much unchangeable at this point), which implements a subset of
commands.

The intention is that i2c-cros-ec-tunnel.c will see something like a NAK
/ "invalid argument" response, and it converts that to ENXIO.
Unforunately, for reasons I have yet to figure out, it's very common for
retries (|i2c_retry_count|) to eventually yield an unexpected response
size, which i2c_smbus_xfer_emulated() treats as EIO; so this layer is
seeing EIO.

Anyway, I might be able to coax the i2c/sbs-battery driver to return
ENXIO instead. Would you consider that to be a better case to handle
here? "No such device or address" seems like an appropriate description
of a permanent error, and not a temporary IO error.

Brian