| 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
| ||
|
Message-ID: <48489ea8-4eb0-4c08-84cf-1183054ac77c@gmx.de>
Date: Sat, 15 Feb 2025 19:37:11 +0100
From: Armin Wolf <W_Armin@....de>
To: Kurt Borja <kuurtb@...il.com>, jlee@...e.com, basak.sb2006@...il.com,
rayanmargham4@...il.com
Cc: hdegoede@...hat.com, ilpo.jarvinen@...ux.intel.com,
platform-driver-x86@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [RFC PATCH 1/3] platform/x86: acer-wmi: Fix setting of fan
behavior
Am 15.02.25 um 19:07 schrieb Kurt Borja:
> On Sat Feb 15, 2025 at 12:40 PM -05, Armin Wolf wrote:
>> Am 15.02.25 um 02:30 schrieb Kurt Borja:
>>
>>> Hi Armin,
>>>
>>> On Fri Feb 14, 2025 at 5:13 PM -05, Armin Wolf wrote:
>>>> After studying the linuwu_sense driver
>>>> (https://github.com/0x7375646F/Linuwu-Sense) i was able to understand
>>>> the meaning of the SetGamingFanBehavior() WMI method:
>>>>
>>>> - the first 16-bit are a bitmap of all fans affected by a fan behavior
>>>> change request.
>>>>
>>>> - the next 8 bits contain four fan mode fields (2-bit), each being
>>>> associated with a bit inside the fan bitmap.
>>>>
>>>> There are three fan modes: auto, turbo and custom.
>>>>
>>>> Use this newfound knowledge to fix the turbo fan handling by setting
>>>> the correct bits before calling SetGamingFanBehavior(). Also check
>>>> the result of the WMI method call and return an error should the ACPI
>>>> firmware signal failure.
>>>>
>>>> Signed-off-by: Armin Wolf <W_Armin@....de>
>>>> ---
>>>> drivers/platform/x86/acer-wmi.c | 75 +++++++++++++++++++++++----------
>>>> 1 file changed, 52 insertions(+), 23 deletions(-)
>>>>
>>>> --
>>>> 2.39.5
>>>>
>>>> diff --git a/drivers/platform/x86/acer-wmi.c b/drivers/platform/x86/acer-wmi.c
>>>> index 69336bd778ee..f20a882e3650 100644
>>>> --- a/drivers/platform/x86/acer-wmi.c
>>>> +++ b/drivers/platform/x86/acer-wmi.c
>>>> @@ -68,10 +68,19 @@ MODULE_LICENSE("GPL");
>>>> #define ACER_WMID_SET_GAMING_LED_METHODID 2
>>>> #define ACER_WMID_GET_GAMING_LED_METHODID 4
>>>> #define ACER_WMID_GET_GAMING_SYS_INFO_METHODID 5
>>>> -#define ACER_WMID_SET_GAMING_FAN_BEHAVIOR 14
>>>> +#define ACER_WMID_SET_GAMING_FAN_BEHAVIOR_METHODID 14
>>>> #define ACER_WMID_SET_GAMING_MISC_SETTING_METHODID 22
>>>> #define ACER_WMID_GET_GAMING_MISC_SETTING_METHODID 23
>>>>
>>>> +#define ACER_GAMING_FAN_BEHAVIOR_ID_MASK GENMASK_ULL(15, 0)
>>>> +#define ACER_GAMING_FAN_BEHAVIOR_SET_MODE_MASK GENMASK_ULL(23, 16)
>>>> +
>>>> +#define ACER_GAMING_FAN_BEHAVIOR_CPU BIT(0)
>>>> +#define ACER_GAMING_FAN_BEHAVIOR_GPU BIT(3)
>>>> +
>>>> +#define ACER_GAMING_FAN_BEHAVIOR_CPU_MODE_MASK GENMASK(1, 0)
>>>> +#define ACER_GAMING_FAN_BEHAVIOR_GPU_MODE_MASK GENMASK(7, 6)
>>>> +
>>>> #define ACER_GAMING_MISC_SETTING_STATUS_MASK GENMASK_ULL(7, 0)
>>>> #define ACER_GAMING_MISC_SETTING_INDEX_MASK GENMASK_ULL(7, 0)
>>>> #define ACER_GAMING_MISC_SETTING_VALUE_MASK GENMASK_ULL(15, 8)
>>>> @@ -121,6 +130,12 @@ enum acer_wmi_predator_v4_sensor_id {
>>>> ACER_WMID_SENSOR_GPU_TEMPERATURE = 0x0A,
>>>> };
>>>>
>>>> +enum acer_wmi_gaming_fan_mode {
>>>> + ACER_WMID_FAN_MODE_AUTO = 0x01,
>>>> + ACER_WMID_FAN_MODE_TURBO = 0x02,
>>>> + ACER_WMID_FAN_MODE_CUSTOM = 0x03,
>>>> +};
>>>> +
>>>> enum acer_wmi_predator_v4_oc {
>>>> ACER_WMID_OC_NORMAL = 0x0000,
>>>> ACER_WMID_OC_TURBO = 0x0002,
>>>> @@ -1565,9 +1580,6 @@ static acpi_status WMID_gaming_set_u64(u64 value, u32 cap)
>>>> case ACER_CAP_TURBO_LED:
>>>> method_id = ACER_WMID_SET_GAMING_LED_METHODID;
>>>> break;
>>>> - case ACER_CAP_TURBO_FAN:
>>>> - method_id = ACER_WMID_SET_GAMING_FAN_BEHAVIOR;
>>>> - break;
>>>> default:
>>>> return AE_BAD_PARAMETER;
>>>> }
>>>> @@ -1618,25 +1630,42 @@ static int WMID_gaming_get_sys_info(u32 command, u64 *out)
>>>> return 0;
>>>> }
>>>>
>>>> +static int WMID_gaming_set_fan_behavior(u16 fan_bitmap, u8 mode_bitmap)
>>>> +{
>>>> + acpi_status status;
>>>> + u64 input = 0;
>>>> + u64 result;
>>>> +
>>>> + input |= FIELD_PREP(ACER_GAMING_FAN_BEHAVIOR_ID_MASK, fan_bitmap);
>>>> + input |= FIELD_PREP(ACER_GAMING_FAN_BEHAVIOR_SET_MODE_MASK, mode_bitmap);
>>>> +
>>>> + status = WMI_gaming_execute_u64(ACER_WMID_SET_GAMING_FAN_BEHAVIOR_METHODID, input,
>>>> + &result);
>>>> + if (ACPI_FAILURE(status))
>>>> + return -EIO;
>>>> +
>>>> + /* TODO: Proper error handling */
>>>> + pr_notice("Fan behavior return status: %llu\n", result);
>>>> +
>>>> + return 0;
>>>> +}
>>>> +
>>>> static void WMID_gaming_set_fan_mode(u8 fan_mode)
>>>> {
>>>> - /* fan_mode = 1 is used for auto, fan_mode = 2 used for turbo*/
>>>> - u64 gpu_fan_config1 = 0, gpu_fan_config2 = 0;
>>>> - int i;
>>>> -
>>>> - if (quirks->cpu_fans > 0)
>>>> - gpu_fan_config2 |= 1;
>>>> - for (i = 0; i < (quirks->cpu_fans + quirks->gpu_fans); ++i)
>>>> - gpu_fan_config2 |= 1 << (i + 1);
>>> This was not replicated bellow. Just to be sure, are there no fans at
>>> BIT(1) and BIT(2)?
>> AFAIK the Acer OEM software support the following fans:
>>
>> - CPU (BIT(0))
>> - GPU 1 (BIT(3))
>> - GPU 2 (BIT(4), but untested)
>>
>> The other bits seem to be unused.
> Interesting.
>
> I ask because quirks->cpu_fans + quirks->gpu_fans is currently always 2,
> so the line two `-` lines I referenced make the following equivalent
> operation:
>
> fan_bitmap |= BIT(1);
> mode_bitmap |= FIELD_PREP(GENMASK(3, 2), fan_mode);
>
> fan_bitmap |= BIT(2);
> mode_bitmap |= FIELD_PREP(GENMASK(5, 4), fan_mode);
>
> So if any model has fans at BIT(1), BIT(2) this may cause regressions if
> the behavior is not mimicked.
>
> Am I missing something?
>
> Anyway, your explaination for how this method works makes a lot of
> sense, so it is weird that the original author of this is summing the
> number of fans and setting this bits in the first place.
>
I do not know why the original code assigned BIT(1) and BIT(2), but i know for sure
that no other (out-of-tree) implementation does this.
I suspect that some models just ignore those unused bits, while others return an error.
This might explain why the turbo fan functionality does not work on some machines.
Should we find a model in the future with more fans then we can extend the fan control code
as needed. For now the risk of regressions should be low since all models inside the whitelist
have only a single CPU and GPU fan.
Thanks,
Armin Wolf
Powered by blists - more mailing lists