linux-kernel - [PATCH 00/12] i2c-hid: fixes for unnumbered reports and other improvements

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-next>] [day] [month] [year] [list]

Message-Id: <20220118072628.1617172-1-dmitry.torokhov@gmail.com>
Date:   Mon, 17 Jan 2022 23:26:16 -0800
From:   Dmitry Torokhov <dmitry.torokhov@...il.com>
To:     Jiri Kosina <jikos@...nel.org>,
        Benjamin Tissoires <benjamin.tissoires@...hat.com>
Cc:     Angela Czubak <acz@...ihalf.com>, linux-input@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: [PATCH 00/12] i2c-hid: fixes for unnumbered reports and other improvements

Hi,

This series came about after I reviewed Angela's patch that fixed issues
with incorrect handling of high-numbered reports (15 and above) in
i2c-hid driver:

- it appears to me that the driver did not handle unnumbered reports
  correctly as the kernel internally expects unnumbered reports to be
  still prepended with report number 0, but i2c_hid_get_raw_report() and
  i2c_hid_output_raw_report() only expected report ID to be present for
  numbered reports.

- the driver tried to consolidate both command handling and sending
  output reports in __i2c_hid_command() but the rules for different
  commands vary significantly and because of that the logic was hard to
  follow and it bled out from __i2c_hid_command() to callers. I decided
  to introduce a few simple helpers and have the data encoding for
  individual commands done at the call site. I believe this made it
  easier to validate the rules and the logic and allowed to remove
  special handling for the HID descriptor retrieval, among other things.

- the driver does too many copying of data; encoding the data for
  commands at the call site allowed to only copy data once into the
  transfer buffers.

I tested this on a couple of Chromebooks with i2c-hid touchscreens, but
it would be great if other folks tried it out as well.

Thanks.

Angela Czubak (1):
  HID: i2c-hid: fix handling numbered reports with IDs of 15 and above

Dmitry Torokhov (11):
  HID: i2c-hid: fix GET/SET_REPORT for unnumbered reports
  HID: i2c-hid: use "struct i2c_hid" as argument in most calls
  HID: i2c-hid: refactor reset command
  HID: i2c-hid: explicitly code setting and sending reports
  HID: i2c-hid: define i2c_hid_read_register() and use it
  HID: i2c-hid: create a helper for SET_POWER command
  HID: i2c-hid: convert i2c_hid_execute_reset() to use i2c_hid_xfer()
  HID: i2c-hid: rework i2c_hid_get_report() to use i2c_hid_xfer()
  HID: i2c-hid: use helpers to do endian conversion in i2c_hid_get_input()
  HID: i2c-hid: no longer need raw access to HID descriptor structure
  HID: i2c-hid: note that I2C xfer buffers are DMA-safe

 drivers/hid/i2c-hid/i2c-hid-core.c | 593 +++++++++++++++--------------
 1 file changed, 313 insertions(+), 280 deletions(-)

-- 
Dmitry