Re: [RFC PATCH] printk: Change timestamp to triplet as mono, boot and real

[Petr Mladek]

On Thu 2020-08-13 20:31:55, Sergey Senozhatsky wrote:
> On (20/08/13 12:22), Petr Mladek wrote:
> >
> >   + would take more space (prefix + text vs. binary representation)
> 
> Dict buffer is allocated regardless of how we use it, and only printks
> from drivers/* (dev_printk*) add dict payload. It might be the case
> that on some (if not most) systems dict pages are not used 90% of times
> (if not 99%).
> 
> >   + not reliable because dict is currently dropped when no space
> 
> Well, in the perfect world this is a problem, but "maybe not having
> alternative timestamps sometimes" can be OK approach for people who
> wants to use those triplet timestamps.

I am afraid the people would see missing timestamps as a bug.
They want it because they want to corelate kernel and userspace logs.
The timestamps from realtime clock are supposed to make it
straightforward.

More importantly. I do not see a reasonable way how to handle it.
No timestamp is worse than timestamps from non-synchronized clock.
Mixing timestamps from different clocks (as fallback) would cause
a lot of confusion (more harm than good).

> But, in general, how real this problem is? What I sae so far (on my boxes)
> was that printk messages are longer than dict payload.

Yeah, I think that dict are less used on normal systems. But it might
be very different when a driver gets more verbose during debugging
and the amount of dev_printk() calls is much higher.

Best Regards,
Petr