linux-kernel - Re: [PATCH 02/10] hwmon: (amc6821) Make reading and writing fan speed limits consistent

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

Message-ID: <535f3381-2fa9-41a3-896a-8d5879546ac9@cherry.de>
Date: Mon, 1 Jul 2024 18:13:58 +0200
From: Quentin Schulz <quentin.schulz@...rry.de>
To: Guenter Roeck <linux@...ck-us.net>, linux-hwmon@...r.kernel.org
Cc: linux-kernel@...r.kernel.org, Farouk Bouabid <farouk.bouabid@...rry.de>
Subject: Re: [PATCH 02/10] hwmon: (amc6821) Make reading and writing fan speed
 limits consistent

Hi Guenter,

On 7/1/24 4:37 PM, Guenter Roeck wrote:
> On 7/1/24 07:11, Guenter Roeck wrote:
>> On 7/1/24 04:05, Quentin Schulz wrote:
>>> Hi Guenter,
>>>
>>> On 6/28/24 5:13 PM, Guenter Roeck wrote:
>>>> The default value of the maximum fan speed limit register is 0,
>>>> essentially translating to an unlimited fan speed. When reading
>>>> the limit, a value of 0 is reported in this case. However, writing
>>>> a value of 0 results in writing a value of 0xffff into the register,
>>>> which is inconsistent.
>>>>  > Signed-off-by: Guenter Roeck <linux@...ck-us.net>
>>>> ---
>>>>   drivers/hwmon/amc6821.c | 6 +++---
>>>>   1 file changed, 3 insertions(+), 3 deletions(-)
>>>>
>>>> diff --git a/drivers/hwmon/amc6821.c b/drivers/hwmon/amc6821.c
>>>> index 3c614a0bd192..e37257ae1a6b 100644
>>>> --- a/drivers/hwmon/amc6821.c
>>>> +++ b/drivers/hwmon/amc6821.c
>>>> @@ -601,7 +601,7 @@ static ssize_t fan_show(struct device *dev, 
>>>> struct device_attribute *devattr,
>>>>       struct amc6821_data *data = amc6821_update_device(dev);
>>>>       int ix = to_sensor_dev_attr(devattr)->index;
>>>>       if (0 == data->fan[ix])
>>>> -        return sprintf(buf, "0");
>>>> +        return sprintf(buf, "6000000");
>>>>       return sprintf(buf, "%d\n", (int)(6000000 / data->fan[ix]));
>>>>   }
>>>> @@ -625,10 +625,10 @@ static ssize_t fan_store(struct device *dev, 
>>>> struct device_attribute *attr,
>>>>       int ret = kstrtol(buf, 10, &val);
>>>>       if (ret)
>>>>           return ret;
>>>> -    val = 1 > val ? 0xFFFF : 6000000/val;
>>>> +    val = val < 1 ? 0xFFFF : 6000000 / val;
>>>>       mutex_lock(&data->update_lock);
>>>> -    data->fan[ix] = (u16) clamp_val(val, 1, 0xFFFF);
>>>> +    data->fan[ix] = (u16)clamp_val(val, 0, 0xFFFF);
>>>
>>> This is an unrelated change I believe and I would therefore have this 
>>> in its own commit with proper documentation in the commit log. Indeed:
>>>
>>> 1- Change in fan_show handles the default 0x0 register value (which 
>>> can only currently be achieved via the default value of the registers)
>>> 2- Allow (re-)setting unlimited fan speed by allowing the user to 
>>> pass 6000001+ instead of clamping it to 6000000 RPM.
>>>
>>
>> Both changes are related.
>>
>> The whole point of this commit is to report and permit consistent 
>> values when
>> the register value is 0. But you do have a point - reading it after my 
>> changes
>> returns 6000000, but writing the same value sets the register to 1. So 
>> I think
>> the proper change would be to display 6000001 as speed if the register 
>> value is
>> 0, and provide a more detailed explanation. Would that address your 
>> concerns ?
>>
> 
> Ah, never  mind, I'll do it differently:
> 
> - If the register value is 0, keep reporting 0.

Or...... maybe UINT_MAX?

> - If the value written is 0, write 0, otherwise limit the range to 
> 1..6000000
>    and write clamp_val(6000000 / val, 1, 0xffff)
> 

Mmmm... I'm a bit worried about the implication of writing 0 in 
TACH-Low-Limit, what is actually going to happen in that scenario? I 
assume **every** possible RPM returned by TACH-DATA will be deemed 
invalid/below the limit then? Reading `Fan Spin-Up` section, if FSPD bit 
from register 0x20 (which we don't write to yet I think?) is set to 0, a 
spin-up is started whenever the fan is detected to be running at too low 
speed. And we would also be getting an interrupt for that too-low event.

Basically, wondering if we shouldn't gate the writing of 0 to only the 
MAX setting?

> This minimizes user visibility of the changes, and also ensures that
> the reported fan speed is 0 if the register value is 0 when reading the fan
> speed.
> 

But didn't you say this means the fan is running at unknown 60 000 000+ 
RPMs? Do we really want to return 0 even if the fan is actually running? 
In which case max < current (possibly) but with no event happening 
(which I would expect, reading the datasheet).

Quentin