linux-kernel - Re: [PATCH] regulator: qcom_spmi: Fix math of spmi_regulator_set_voltage_time

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: <871848c2-d600-fb20-1d5d-c196ea5aba44@linaro.org>
Date:   Wed, 19 Jun 2019 21:43:57 +0200
From:   Jorge Ramirez <jorge.ramirez-ortiz@...aro.org>
To:     Bjorn Andersson <bjorn.andersson@...aro.org>,
        Jeffrey Hugo <jeffrey.l.hugo@...il.com>
Cc:     lgirdwood@...il.com, broonie@...nel.org,
        linux-arm-msm@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] regulator: qcom_spmi: Fix math of
 spmi_regulator_set_voltage_time_sel

On 6/19/19 21:05, Bjorn Andersson wrote:
> On Wed 19 Jun 11:56 PDT 2019, Jeffrey Hugo wrote:
> 
>> spmi_regulator_set_voltage_time_sel() calculates the amount of delay
>> needed as the result of setting a new voltage.  Essentially this is the
>> absolute difference of the old and new voltages, divided by the slew rate.
>>
>> The implementation of spmi_regulator_set_voltage_time_sel() is wrong.
>>
>> It attempts to calculate the difference in voltages by using the
>> difference in selectors and multiplying by the voltage step between
>> selectors.  This ignores the possibility that the old and new selectors
>> might be from different ranges, which have different step values.  Also,
>> the difference between the selectors may encapsulate N ranges inbetween,
>> so a summation of each selector change from old to new would be needed.
>>
>> Lets avoid all of that complexity, and just get the actual voltage
>> represented by both the old and new selector, and use those to directly
>> compute the voltage delta.  This is more straight forward, and has the
>> side benifit of avoiding issues with regulator implementations that don't
>> have hardware register support to get the current configured range.
>>
>> Fixes: e92a4047419c ("regulator: Add QCOM SPMI regulator driver")
>> Reported-by: Bjorn Andersson <bjorn.andersson@...aro.org>
>> Reported-by: Jorge Ramirez-Ortiz <jorge.ramirez-ortiz@...aro.org>
>> Signed-off-by: Jeffrey Hugo <jeffrey.l.hugo@...il.com>
> 
> Reviewed-by: Bjorn Andersson <bjorn.andersson@...aro.org>

Tested on EVB-4000 using the cpufreq patchset that I still need to
repost v3  [1]

Tested-by: Jorge Ramirez-Ortiz <jorge.ramirez-ortiz@...aro.org>


[1] https://patchwork.kernel.org/cover/10784383/

> 
>> ---
>>  drivers/regulator/qcom_spmi-regulator.c | 8 ++------
>>  1 file changed, 2 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/regulator/qcom_spmi-regulator.c b/drivers/regulator/qcom_spmi-regulator.c
>> index 13f83be50076..877df33e0246 100644
>> --- a/drivers/regulator/qcom_spmi-regulator.c
>> +++ b/drivers/regulator/qcom_spmi-regulator.c
>> @@ -813,14 +813,10 @@ static int spmi_regulator_set_voltage_time_sel(struct regulator_dev *rdev,
>>  		unsigned int old_selector, unsigned int new_selector)
>>  {
>>  	struct spmi_regulator *vreg = rdev_get_drvdata(rdev);
>> -	const struct spmi_voltage_range *range;
>>  	int diff_uV;
>>  
>> -	range = spmi_regulator_find_range(vreg);
>> -	if (!range)
>> -		return -EINVAL;
>> -
>> -	diff_uV = abs(new_selector - old_selector) * range->step_uV;
>> +	diff_uV = abs(spmi_regulator_common_list_voltage(rdev, new_selector) -
>> +		      spmi_regulator_common_list_voltage(rdev, old_selector));
>>  
>>  	return DIV_ROUND_UP(diff_uV, vreg->slew_rate);
>>  }
>> -- 
>> 2.17.1
>>
>