linux-kernel - Re: [PATCH 1/2] hid: multitouch: add module paramter to override quirks

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 for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <CAO-hwJJkp0EPEaV_Ws4ZFLPk+RafcjVUX8QphwzqFJ+AMsQmzg@mail.gmail.com>
Date:   Thu, 21 Apr 2022 10:27:21 +0200
From:   Benjamin Tissoires <benjamin.tissoires@...hat.com>
To:     Tao Jin <tao-j@...look.com>
Cc:     linux-input@...r.kernel.org, Jiri Kosina <jikos@...nel.org>,
        linux-kernel@...r.kernel.org,
        Bingchen Gong <gongbingchen@...il.com>
Subject: Re: [PATCH 1/2] hid: multitouch: add module paramter to override quirks

Hi,

On Mon, Apr 18, 2022 at 5:18 AM Tao Jin <tao-j@...look.com> wrote:
>
> There is a sysfs interface to specify the quirks. However, some
> quirk bits such as HID_QUIRK_MULTI_INPUT are only read once during
> init/probe. Setting the quirks in sysfs does not make any change to the
> behaviors related to those bits. Also, it is hard to wait for udev to
> modify the quirks in case there is a rule given the current init and
> state machine structure.
>
> A simple kernel paramter is provided so that any time a custom quirk
> needs to be tested can be easily applied. This enables the users to test
> out which bits are indeed necessary with just a rmmod then
> insmod/modprobe cycle. Thus, new hardware support can be added very
> soon and unneccsary mildly costly quirks using mutex and timer can also
> be removed based on user's self-test.

I don't think the last sentence above is adding anything. I don't see
where the mutex and timers are related to this patch.

>
> Co-Developed-by: Bingchen Gong <gongbingchen@...il.com>
> Signed-off-by: Bingchen Gong <gongbingchen@...il.com>
> Signed-off-by: Tao Jin <tao-j@...look.com>
> ---
>  drivers/hid/hid-multitouch.c | 13 +++++++++++--
>  1 file changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/hid/hid-multitouch.c b/drivers/hid/hid-multitouch.c
> index 3ea57f3..c6d64f8 100644
> --- a/drivers/hid/hid-multitouch.c
> +++ b/drivers/hid/hid-multitouch.c
> @@ -398,6 +398,10 @@ static const struct mt_class mt_classes[] = {
>         { }
>  };
>
> +static int override_quirks = -1;
> +module_param(override_quirks, int, 0444);
> +MODULE_PARM_DESC(override_quirks, "Signed integer to override quirks in mtclass, must >= 0 to enable override.");
> +
>  static ssize_t mt_show_quirks(struct device *dev,
>                            struct device_attribute *attr,
>                            char *buf)
> @@ -1749,7 +1753,12 @@ static int mt_probe(struct hid_device *hdev, const struct hid_device_id *id)
>         if (id->group != HID_GROUP_MULTITOUCH_WIN_8)
>                 hdev->quirks |= HID_QUIRK_MULTI_INPUT;
>
> -       if (mtclass->quirks & MT_QUIRK_FORCE_MULTI_INPUT) {
> +       if (override_quirks >= 0) {
> +               hid_info(hdev, "overriding quirks with: %d(0x%x)", override_quirks, override_quirks);
> +               td->mtclass.quirks = override_quirks;
> +       }
> +
> +       if (td->mtclass.quirks & MT_QUIRK_FORCE_MULTI_INPUT) {
>                 hdev->quirks &= ~HID_QUIRK_INPUT_PER_APP;
>                 hdev->quirks |= HID_QUIRK_MULTI_INPUT;
>         }
> @@ -1760,7 +1769,7 @@ static int mt_probe(struct hid_device *hdev, const struct hid_device_id *id)
>         if (ret != 0)
>                 return ret;
>
> -       if (mtclass->quirks & MT_QUIRK_FIX_CONST_CONTACT_ID)
> +       if (td->mtclass.quirks & MT_QUIRK_FIX_CONST_CONTACT_ID)

The 2 changes "s/mtclass->quirks/td->mtclass.quirks/" should probably
be added as a separate patch (before this one). The reason is that
hid_parse might change the quirks values, and
MT_QUIRK_FIX_CONST_CONTACT_ID and MT_QUIRK_FORCE_MULTI_INPUT might be
different from the class definition at the end of probe, so
definitively, this is a bug (without consequences) that should be
addressed separately.

Cheers,
Benjamin

>                 mt_fix_const_fields(hdev, HID_DG_CONTACTID);
>
>         ret = hid_hw_start(hdev, HID_CONNECT_DEFAULT);
> --
> 2.35.1
>