Re: [PATCH] HID: amd_sfh: Increase bitwidth to avoid shift-out-of-bounds

[Basavaraj Natikar]

On 7/4/2023 1:47 PM, Kai-Heng Feng wrote:
> UBSAN complains shift-out-of-bounds on amd_sfh:
> [    7.593412] input: SYNA3105:00 06CB:CEA3 Mouse as /devices/platform/AMDI0010:00/i2c-0/i2c-SYNA3105:00/0018:06CB:CEA3.0001/input/input11
> [    7.593541] input: SYNA3105:00 06CB:CEA3 Touchpad as /devices/platform/AMDI0010:00/i2c-0/i2c-SYNA3105:00/0018:06CB:CEA3.0001/input/input12
> [    7.593625] hid-multitouch 0018:06CB:CEA3.0001: input,hidraw0: I2C HID v1.00 Mouse [SYNA3105:00 06CB:CEA3] on i2c-SYNA3105:00
> [    7.664537] ================================================================================
> [    7.664540] UBSAN: shift-out-of-bounds in drivers/hid/amd-sfh-hid/sfh1_1/amd_sfh_desc.c:149:50
> [    7.664544] shift exponent 103 is too large for 64-bit type 'long unsigned int'
> [    7.664547] CPU: 5 PID: 124 Comm: kworker/5:1 Tainted: G        W          6.4.0-next-20230703 #1
> [    7.664549] Hardware name: HP HP EliteBook 835 13 inch G10 Notebook PC/8C10, BIOS V83 Ver. 01.01.09 06/05/2023
> [    7.664551] Workqueue: events amd_sfh_work_buffer [amd_sfh]
> [    7.664562] Call Trace:
> [    7.664564]  <TASK>
> [    7.664567]  dump_stack_lvl+0x48/0x70
> [    7.664576]  dump_stack+0x10/0x20
> [    7.664578]  __ubsan_handle_shift_out_of_bounds+0x156/0x310
> [    7.664584]  ? sched_clock_noinstr+0x9/0x10
> [    7.664589]  ? sched_clock_cpu+0x12/0x1d0
> [    7.664594]  ? raw_spin_rq_lock_nested+0x1e/0xa0
> [    7.664597]  ? psi_group_change+0x237/0x520
> [    7.664600]  float_to_int.cold+0x18/0xcf [amd_sfh]
> [    7.664606]  ? __pfx_get_input_rep+0x10/0x10 [amd_sfh]
> [    7.664611]  get_input_rep+0x241/0x340 [amd_sfh]
> [    7.664617]  amd_sfh_work_buffer+0x91/0x190 [amd_sfh]
> [    7.664622]  process_one_work+0x229/0x450
> [    7.664627]  worker_thread+0x50/0x3f0
> [    7.664629]  ? __pfx_worker_thread+0x10/0x10
> [    7.664632]  kthread+0xf4/0x130
> [    7.664635]  ? __pfx_kthread+0x10/0x10
> [    7.664638]  ret_from_fork+0x29/0x50
> [    7.664644]  </TASK>
> [    7.664652] ================================================================================
>
> So increase the variable bitwidth to solve the issue.
>
> Signed-off-by: Kai-Heng Feng <kai.heng.feng@xxxxxxxxxxxxx>
> ---
>  drivers/hid/amd-sfh-hid/sfh1_1/amd_sfh_desc.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/hid/amd-sfh-hid/sfh1_1/amd_sfh_desc.c b/drivers/hid/amd-sfh-hid/sfh1_1/amd_sfh_desc.c
> index 6f0d332ccf51..550639ef8bfe 100644
> --- a/drivers/hid/amd-sfh-hid/sfh1_1/amd_sfh_desc.c
> +++ b/drivers/hid/amd-sfh-hid/sfh1_1/amd_sfh_desc.c
> @@ -134,7 +134,8 @@ static void get_common_inputs(struct common_input_property *common, int report_i
>  
>  static int float_to_int(u32 float32)
>  {
> -	int fraction, shift, mantissa, sign, exp, zeropre;
> +	int fraction, shift, sign, exp, zeropre;
> +	s128 mantissa;
>  
>  	mantissa = float32 & GENMASK(22, 0);
>  	sign = (float32 & BIT(31)) ? -1 : 1;

Could you please check below patch which covers more scenario's and corner cases
for UBSAN: shift-out-of-bounds.
https://lore.kernel.org/all/20230707065722.9036-1-Basavaraj.Natikar@xxxxxxx/T/#m67e0b7aa6a288ea074b3bfa61e7f085f7e036172

Thanks,
--
Basavaraj