| 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: <dd80ee77-3b71-ee47-2744-36b09e0ec372@roeck-us.net>
Date: Tue, 10 Jan 2023 10:22:54 -0800
From: Guenter Roeck <linux@...ck-us.net>
To: Jon Cormier <jcormier@...ticallink.com>
Cc: linux-hwmon@...r.kernel.org, Jean Delvare <jdelvare@...e.com>,
linux-kernel@...r.kernel.org, devicetree@...r.kernel.org,
Rob Herring <robh+dt@...nel.org>,
Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
Bob Duke <bduke@...ticallink.com>,
John Pruitt <jpruitt@...ticallink.com>,
Dan Vincelette <dvincelette@...ticallink.com>
Subject: Re: [PATCH v3 3/5] hwmon: ltc2945: Handle error case in
ltc2945_value_store
On 1/10/23 10:19, Jon Cormier wrote:
> On Mon, Jan 9, 2023 at 7:04 PM Guenter Roeck <linux@...ck-us.net> wrote:
>>
>> On 1/9/23 15:35, Jonathan Cormier wrote:
>>> ltc2945_val_to_reg errors were not being handled
>>> which would have resulted in register being set to
>>> 0 (clamped) instead of being left alone.
>>>
>>> Change reg_to_val and val_to_reg to return values
>>> via parameters to make it more obvious when an
>>> error case isn't handled. Also to allow
>>> the regval type to be the correct sign in prep for
>>> next commits.
>>>
>>
>> Sorry, I don't see that as reason or argument for such invasive changes.
>> As far as I can see, a two-liner to check the return value of val_to_reg()
>> should have been sufficient. Most of the rest, such as splitting
>> the return value into two elements, is POV and just adds additional code
>> and complexity for zero gain.
> I can do that. However, you had also mentioned changing the return
> type to match what the calling function was expecting, an unsigned
> long. But I can't do that since error codes are negative so it would
> be a signed long which would lose precision and seemingly defeat the
> point of matching the variable type the caller wants. I could make it
> a signed long long but that still doesn't match. So it seemed saner
> to just return the error and the value separately, that way the
> function declaration was explicit about the types it wanted/returned,
> and less room for error. Would love to know your preferred solution.
>
That is only true if the upper bit is actually ever set in that signed long.
Which means I'll have to verify if "would lose precision" is actually
a correct statement.
Guenter
>>
>> Guenter
>>
>>> Fixes: 6700ce035f83 ("hwmon: Driver for Linear Technologies LTC2945")
>>>
>>> Signed-off-by: Jonathan Cormier <jcormier@...ticallink.com>
>>> ---
>>> drivers/hwmon/ltc2945.c | 30 ++++++++++++++++++------------
>>> 1 file changed, 18 insertions(+), 12 deletions(-)
>>>
>>> diff --git a/drivers/hwmon/ltc2945.c b/drivers/hwmon/ltc2945.c
>>> index 9af3e3821152..c66acf8d2124 100644
>>> --- a/drivers/hwmon/ltc2945.c
>>> +++ b/drivers/hwmon/ltc2945.c
>>> @@ -70,12 +70,12 @@ static inline bool is_power_reg(u8 reg)
>>> }
>>>
>>> /* Return the value from the given register in uW, mV, or mA */
>>> -static long long ltc2945_reg_to_val(struct device *dev, u8 reg)
>>> +static int ltc2945_reg_to_val(struct device *dev, u8 reg, u64 *regval)
>>> {
>>> struct regmap *regmap = dev_get_drvdata(dev);
>>> unsigned int control;
>>> u8 buf[3];
>>> - long long val;
>>> + u64 val;
>>> int ret;
>>>
>>> ret = regmap_bulk_read(regmap, reg, buf,
>>> @@ -148,11 +148,12 @@ static long long ltc2945_reg_to_val(struct device *dev, u8 reg)
>>> default:
>>> return -EINVAL;
>>> }
>>> - return val;
>>> + *regval = val;
>>> + return 0;
>>> }
>>>
>>> static int ltc2945_val_to_reg(struct device *dev, u8 reg,
>>> - unsigned long val)
>>> + unsigned long val, unsigned long *regval)
>>> {
>>> struct regmap *regmap = dev_get_drvdata(dev);
>>> unsigned int control;
>>> @@ -220,19 +221,21 @@ static int ltc2945_val_to_reg(struct device *dev, u8 reg,
>>> default:
>>> return -EINVAL;
>>> }
>>> - return val;
>>> + *regval = val;
>>> + return 0;
>>> }
>>>
>>> static ssize_t ltc2945_value_show(struct device *dev,
>>> struct device_attribute *da, char *buf)
>>> {
>>> struct sensor_device_attribute *attr = to_sensor_dev_attr(da);
>>> - long long value;
>>> + int ret;
>>> + u64 value;
>>>
>>> - value = ltc2945_reg_to_val(dev, attr->index);
>>> - if (value < 0)
>>> - return value;
>>> - return sysfs_emit(buf, "%lld\n", value);
>>> + ret = ltc2945_reg_to_val(dev, attr->index, &value);
>>> + if (ret < 0)
>>> + return ret;
>>> + return sysfs_emit(buf, "%llu\n", value);
>>> }
>>>
>>> static ssize_t ltc2945_value_store(struct device *dev,
>>> @@ -245,7 +248,7 @@ static ssize_t ltc2945_value_store(struct device *dev,
>>> unsigned long val;
>>> u8 regbuf[3];
>>> int num_regs;
>>> - int regval;
>>> + unsigned long regval;
>>> int ret;
>>>
>>> ret = kstrtoul(buf, 10, &val);
>>> @@ -253,7 +256,10 @@ static ssize_t ltc2945_value_store(struct device *dev,
>>> return ret;
>>>
>>> /* convert to register value, then clamp and write result */
>>> - regval = ltc2945_val_to_reg(dev, reg, val);
>>> + ret = ltc2945_val_to_reg(dev, reg, val, ®val);
>>> + if (ret < 0)
>>> + return ret;
>>> +
>>> if (is_power_reg(reg)) {
>>> regval = clamp_val(regval, 0, 0xffffff);
>>> regbuf[0] = regval >> 16;
>>
>
>
> --
> Jonathan Cormier
> Software Engineer
>
> Voice: 315.425.4045 x222
>
>
>
> http://www.CriticalLink.com
> 6712 Brooklawn Parkway, Syracuse, NY 13211
Powered by blists - more mailing lists