On Wed 2017-08-30 14:37:34, Sergey Senozhatsky wrote:
> On (08/29/17 19:58), Joe Perches wrote:
> > > > 
> > > > Why?
> > > > 
> > > > What's wrong with a simple printk?
> > > > It'd still do a log_store.
> > > 
> > > sure, it will. but in separate logbuf entries, and between two
> > > consequent printk calls on the same CPU a lot of stuff can happen:
> > 
> > I think you don't quite understand how this would work.
> > The idea is that the entire concatenated bit would be emitted
> > in one go.
> 
> may be :)
> 
> I was thinking about the way to make it work in similar way with
> printk-safe/printk-nmi. basically seq buffer should hold both
> continuation and "normal" lines, IMHO. when we emit the buffer
> we do something like this
> 
>       /* Print line by line. */
>       while (c < end) {
>               if (*c == '\n') {
>                       printk_safe_flush_line(start, c - start + 1);
>                       start = ++c;
>                       header = true;
>                       continue;
>               }
> 
>               /* Handle continuous lines or missing new line. */
>               if ((c + 1 < end) && printk_get_level(c)) {
>                       if (header) {
>                               c = printk_skip_level(c);
>                               continue;
>                       }
> 
>                       printk_safe_flush_line(start, c - start);
>                       start = c++;
>                       header = true;
>                       continue;
>               }
> 
>               header = false;
>               c++;
>       }
> 
> except that instead of printk_safe_flush_line() we will call log_store()
> and the whole loop will be under logbuf_lock.
> 
> for that to work, we need API to require header/loglevel etc for every
> message. so the use case can look like this:
> 
>       init_printk_buffer(&buf);
>       print_line(&buf, KERN_ERR "Oops....\n");
> 
>       print_line(&buf, KERN_ERR "continuation line: foo");
>       print_line(&buf, KERN_CONT "bar");
>       print_line(&buf, KERN_CONT "baz\n");
>       ...
> 
>       print_line(&buf, KERN_ERR "....\n");
>       ...
>       print_line(&buf, KERN_ERR "--- end of oops ---\n");
>       emit_printk_buffer(&buf);
> 
> so that not only concatenated continuation lines will be handled,
> but also more complex things. like backtraces or whatever someone
> might want to handle.

For oopses... please don't. It is quite important that Oops goes out
"as soon as possible". I have seen oopses cut in half, etc... They are
still quite helpful.
                                                                        Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) 
http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Attachment: signature.asc
Description: Digital signature

Reply via email to