Re: [PATCH v4] printk: Add monotonic, boottime, and realtime timestamps

[Prarit Bhargava]

On 08/07/2017 12:58 PM, Mark Salyzyn wrote:
> On 08/07/2017 08:52 AM, Prarit Bhargava wrote:
>> diff --git a/arch/arm/configs/aspeed_g4_defconfig
>> b/arch/arm/configs/aspeed_g4_defconfig
>> index cfc2465e8b77..5f3c50914e92 100644
>> --- a/arch/arm/configs/aspeed_g4_defconfig
>> +++ b/arch/arm/configs/aspeed_g4_defconfig
>> @@ -162,7 +162,7 @@ CONFIG_JFFS2_FS_XATTR=y
>>   CONFIG_UBIFS_FS=y
>>   CONFIG_SQUASHFS=y
>>   CONFIG_SQUASHFS_XZ=y
>> -CONFIG_PRINTK_TIME=y
>> +CONFIG_PRINTK_TIME_LOCAL=y
>>   CONFIG_DYNAMIC_DEBUG=y
>>   CONFIG_STRIP_ASM_SYMS=y
>>   CONFIG_DEBUG_FS=y
> Many have had misgivings, let me try another pass at this.
> 
> We (royal we) should really look into adjusting configuration parsing to allow
> an easy transition from boolean to selection ... I am sure this is not the first
> time bistate/tristate was moved to a number.
> 
> An idea? Maybe look into a way to deal with this to use something _other_ than
> CONFIG_PRINTK_TIME to hold the selection, and keep a (hidden/legacy?)
> CONFIG_PRINTK_TIME that when selected sets CONFIG_PRINTK_TIME_LOCAL, and switch
> to _not_ CONFIG_PRINTK_TIME_DISABLE as the internal mechanical replacement for
> it. I do not know how disruptive this will be, but is worth it if the codebase
> supports it, and legacy config retained?

I looked for one but couldn't find one.  The kernel is a big place, though, and
perhaps it already exists :/.

>> +
>> +static int printk_time_set(const char *val, const struct kernel_param *kp)
>> +{
>> +    char *param = strstrip((char *)val);
>> +    int _printk_time = -1;
>> +    int stamp;
>> +
>> +    if (strlen(param) == 1) {
>> +        /* Preserve legacy boolean settings */
>> +        if (!strcmp("0", param) || !strcmp("n", param) ||
> if strlen(param) == 1, then param[0] == '0' etc works fine and is KISS.
>> +    /*
>> +     * Only allow enabling and disabling of the current printk_time
>> +     * setting.  Changing it from one setting to another confuses
>> +     * userspace.
>> +     */
>> +    if (printk_time_setting == PRINTK_TIME_DISABLE) {
>> +        printk_time_setting = _printk_time;
>> +    } else if ((printk_time_setting != _printk_time) &&
>> +           (_printk_time != 0)) {
>> +        pr_warn("printk: timestamp can only be set to 0(disabled) or %s\n",
>> +            printk_time_str[printk_time_setting]);
>> +        return -EINVAL;
>> +    }
> I agree with the restriction in the general case.  However (as hinted at
> before() #ifdef CONFIG_PRINTK_TIME_RESTRICT (default y, or #ifndef
> CONFIG_PRINTK_TIME_DEBUG default n) around this will allow us users to choose if
> we are confused or not. I can see being able to change it on the fly as an
> option. Especially since we have /sys/module/printk/parameters/time.

Yeah, but I think that should be a later enhancement.

P.

> 
> -- Mark