Messages in this thread | | | Date | Tue, 6 Mar 2018 17:10:49 +0900 | From | Sergey Senozhatsky <> | Subject | Re: Would you help to tell why async printk solution was not taken to upstream kernel ? |
| |
On (03/05/18 22:16), Steven Rostedt wrote: > > Yes. My point was that "CPU can print one full buffer max" is not > > guaranteed and not exactly true. There are ways for CPUs to break > > that O(logbuf) boundary. > > Yes, when printk or the consoles have a bug, it can be greater than > O(logbuf).
OK. Now the question is - what is "a bug" in this case? Are those printk-s really a bug? Consoles are very complex, with dependencies on timers, networking, etc. having them appending more messages to the logbuf is not very cool, but at the time the kernel does not BUG_ON(), nor panic(); it moves on. It's printk()->console_unlock() that turns it into lockup->panic(). Is the bug really in the consoles then?
-ss
| |