]> git.baikalelectronics.ru Git - kernel.git/commit
HID: core: Provide new max_buffer_size attribute to over-ride the default
authorLee Jones <lee@kernel.org>
Mon, 20 Mar 2023 13:06:31 +0000 (13:06 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 22 Mar 2023 12:28:10 +0000 (13:28 +0100)
commit78711769e45da3718053ab03cd9b53dd9da63dba
tree4368ad7f84fe8a6530e9e2af050add46e6935197
parent6f9d65fe1412f5248d82c64cc3f3ce189ce9d7c2
HID: core: Provide new max_buffer_size attribute to over-ride the default

commit b1a37ed00d7908a991c1d0f18a8cba3c2aa99bdc upstream.

Presently, when a report is processed, its proposed size, provided by
the user of the API (as Report Size * Report Count) is compared against
the subsystem default HID_MAX_BUFFER_SIZE (16k).  However, some
low-level HID drivers allocate a reduced amount of memory to their
buffers (e.g. UHID only allocates UHID_DATA_MAX (4k) buffers), rending
this check inadequate in some cases.

In these circumstances, if the received report ends up being smaller
than the proposed report size, the remainder of the buffer is zeroed.
That is, the space between sizeof(csize) (size of the current report)
and the rsize (size proposed i.e. Report Size * Report Count), which can
be handled up to HID_MAX_BUFFER_SIZE (16k).  Meaning that memset()
shoots straight past the end of the buffer boundary and starts zeroing
out in-use values, often resulting in calamity.

This patch introduces a new variable into 'struct hid_ll_driver' where
individual low-level drivers can over-ride the default maximum value of
HID_MAX_BUFFER_SIZE (16k) with something more sympathetic to the
interface.

Signed-off-by: Lee Jones <lee@kernel.org>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
[Lee: Backported to v5.10.y]
Signed-off-by: Lee Jones <lee@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/hid/hid-core.c
include/linux/hid.h