Re: [PATCH] /dev/random: Insufficient of entropy on many architectures

[John Stultz <john.stultz@linaro.org>]

On 09/10/2013 12:47 PM, Stephan Mueller wrote:
> Am Dienstag, 10. September 2013, 12:38:56 schrieb John Stultz:
>
> Hi John,
>
>> On Tue, Sep 10, 2013 at 4:31 AM, Stephan Mueller <smueller@chronox.de> 
> wrote:
>>> Hi,
>>>
>>> /dev/random uses the get_cycles() function to obtain entropy in
>>> addition to jiffies and the event value of hardware events.
>>>
>>> Typically the high-resolution timer of get_cycles delivers the
>>> majority of entropy, because the event value is quite deterministic
>>> and jiffies are very coarse.
>> [snip]
>>
>>> The following patch uses the clocksource clock for a time value in
>>> case get_cycles returns 0. As clocksource may not be available
>>> during boot time, a flag is introduced which allows random.c to
>>> check the availability of clocksource.
>> [snip]
>>
>>> diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
>>> index 48b9fff..75b1613 100644
>>> --- a/kernel/time/timekeeping.c
>>> +++ b/kernel/time/timekeeping.c
>>> @@ -35,6 +35,7 @@ static struct timekeeper timekeeper;
>>>
>>>  static DEFINE_RAW_SPINLOCK(timekeeper_lock);
>>>  static seqcount_t timekeeper_seq;
>>>  static struct timekeeper shadow_timekeeper;
>>>
>>> +static bool timekeeper_enabled = 0;
>>>
>>>  /* flag for if timekeeping is suspended */
>>>  int __read_mostly timekeeping_suspended;
>>>
>>> @@ -833,8 +834,15 @@ void __init timekeeping_init(void)
>>>
>>>         write_seqcount_end(&timekeeper_seq);
>>>         raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
>>>
>>> +       timekeeper_enabled = 1;
>>>
>>>  }
>> So the end of timekeeping_init() may not be what you want here. This
>> only means we've started up the timekeping core with only the default
>> clocksource (with only few exceptions, this is almost always jiffies).
>> Then as clocksource drivers are initialized, they are registered and
>> the timekeeping core will switch over to the best available
>> clocksource.  Also, to avoid the churn at boot of switching to every
>> clocksource registered, we queue them up and wait until fs_init time
>> to switch to whatever is the best available then.
>>
>> So its likely with this patch that the systems all still end up using
>> jiffies for their clocksource at least until fs_init time.
> Thank you for the explanation. Is there any trigger that is fired at 
> fs_init time that one can read?
Check out clocksource_done_booting(), as that's probably where you'd
want to add something similar. Now this doesn't ensure non-jiffies
clocksources are installed, just that we're far enough into boot that we
assume most of the system's clocksources have been registered.

timekeeping_valid_for_hres() might be a better check for what you're
looking for, since that means some free-running clocksource has been
installed.

thanks
-john