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  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]
Date:   Thu, 13 Apr 2017 22:09:43 +0200
From:   Carlo Caione <carlo@...lessm.com>
To:     Darren Hart <dvhart@...radead.org>
Cc:     Carlo Caione <carlo@...one.org>, andy@...radead.org,
        platform-driver-x86@...r.kernel.org, linux-kernel@...r.kernel.org,
        Linux Upstreaming Team <linux@...lessm.com>
Subject: Re: [PATCH 2/2] hp-wmi: Fix detection for dock and tablet mode

On Thu, Apr 13, 2017 at 8:21 PM, Darren Hart <dvhart@...radead.org> wrote:
> On Sun, Apr 09, 2017 at 03:56:08PM +0200, Carlo Caione wrote:
>> From: Carlo Caione <carlo@...lessm.com>

/cut
>> @@ -644,6 +646,7 @@ static int __init hp_wmi_input_setup(void)
>>  {
>>       acpi_status status;
>>       int err;
>> +     int val;
>>
>>       hp_wmi_input_dev = input_allocate_device();
>>       if (!hp_wmi_input_dev)
>> @@ -654,17 +657,26 @@ static int __init hp_wmi_input_setup(void)
>>       hp_wmi_input_dev->id.bustype = BUS_HOST;
>>
>>       __set_bit(EV_SW, hp_wmi_input_dev->evbit);
>> -     __set_bit(SW_DOCK, hp_wmi_input_dev->swbit);
>> -     __set_bit(SW_TABLET_MODE, hp_wmi_input_dev->swbit);
>> +
>> +     /* Dock */
>> +     val = hp_wmi_dock_state();
>> +     if (!(val < 0)) {
>> +             __set_bit(SW_DOCK, hp_wmi_input_dev->swbit);
>> +             input_report_switch(hp_wmi_input_dev, SW_DOCK, val);
>> +     }
>
> In general, these are fine and can go in. I did want to get your opinion on one
> thought though.
>
> This adds some complexity to deal with what appears to be an unknown failure
> mode (the query fails, we don't know why, so we don't set the bit on the input
> dev for that feature). Since we don't know why it fails, can we be confident it
> will always fail?

That's not exactly true, at least for the firmware I have on the
laptop I'm working on.

For this hardware (can we assume for all the HP models?) when the WMI
calls returns the value of 0x04, that means that the query
(HPWMI_HARDWARE_QUERY in this case) is not implemented at all in the
SSDT.
In general reading the disassembled AML code when the WMI query fails
and returns a positive value this can be:
- 0x04: Query ID is unknown / not implemented but valid
- 0x02: Wrong signature
- 0x05: Wrong / invalid query number (?)

The problem here is that: (1) this is my personal interpretation of
the AML code obtained by disassembling the SSDT and (2) we cannot be
sure that this is the same on all the HP firmwares around.
For sure in general in all the cases I extracted from the SSDT table
on this hardware if the call failed the first time all the chances are
that it is going to fail also in the future.

In [1] is the SSDT table, the WMI method is WMAA if you want to check
my interpretation.

> Could it succeed at init here, but then fail later and leave
> us in the same situation we are in now?

I think that this is really unlikely

> If so, have you considered just returning 0 on error and using a WARN_ONCE print
> statement to report the error? This would simplify a lot of this logic that
> you're adding in here to handle something we could just report and ignore.

Yes, I thought to report just 0 but in that case we are advertising to
userspace fake capabilities for the hardware, like dockability or
laptop mode that in most cases are not even implemented on the
hardware (like on this laptop).

> That being said, your version avoids the input_report_switch() in the event of a
> failure at init. In practice, I don't know if this is worth the added
> complexity.
>
> Your thoughts?

AFAICT we can fail in hp_wmi_perform_query (as written in the comment
to the function):
1) with -EINVAL if the query was not successful or the output buffer
size exceeds buffersize. In this case I don't see how the next calls
could be successful.
2) with a positive error code returned from the WMI method. Given my
interpretation of this positive code reported before I don't see why
we should fail only on init and not on all the subsequent calls

So I'm still convinced that my implementation is correct and that
probably adding complexity on top is not really worth it. But of
course this is your call as maintainer :)

Thanks,

[1] https://paste.fedoraproject.org/paste/bBnqUlazz1tAjKsJKq7NHl5M1UNdIGYhyRLivL9gydE=

-- 
Carlo Caione  |  +39.340.80.30.096  |  Endless

Powered by blists - more mailing lists