On Tue 2018-11-06 23:35:02, Sergey Senozhatsky wrote:

On (11/02/18 22:31), Tetsuo Handa wrote:

quoted

  (1) Call get_printk_buffer() and acquire "struct printk_buffer *".

  (2) Rewrite printk() calls in the following way. The "ptr" is
      "struct printk_buffer *" obtained in step (1).

      printk(fmt, ...)     => printk_buffered(ptr, fmt, ...)
      vprintk(fmt, args)   => vprintk_buffered(ptr, fmt, args)
      pr_emerg(fmt, ...)   => bpr_emerg(ptr, fmt, ...)
      pr_alert(fmt, ...)   => bpr_alert(ptr, fmt, ...)
      pr_crit(fmt, ...)    => bpr_crit(ptr, fmt, ...)
      pr_err(fmt, ...)     => bpr_err(ptr, fmt, ...)
      pr_warning(fmt, ...) => bpr_warning(ptr, fmt, ...)
      pr_warn(fmt, ...)    => bpr_warn(ptr, fmt, ...)
      pr_notice(fmt, ...)  => bpr_notice(ptr, fmt, ...)
      pr_info(fmt, ...)    => bpr_info(ptr, fmt, ...)
      pr_cont(fmt, ...)    => bpr_cont(ptr, fmt, ...)

  (3) Release "struct printk_buffer" by calling put_printk_buffer().

[..]

quoted

Since we want to remove "struct cont" eventually, we will try to remove
both "implicit printk() users who are expecting KERN_CONT behavior" and
"explicit pr_cont()/printk(KERN_CONT) users". Therefore, converting to
this API is recommended.

- The printk-fallback sounds like a hint that the existing 'cont' handling
  better stay in the kernel. I don't see how the existing 'cont' is
  significantly worse than
		bpr_warn(NULL, ...)->printk() // no 'cont' support
  I don't see why would we want to do it, sorry. I don't see "it takes 16
  printk-buffers to make a thing go right" as a sure thing.

I see it the following way:

   + mixed cont lines are very rare but they happen

   + 16 buffers are more than 1 so it could only be better [*]

   + the printk_buffer() code is self-contained and does not
     complicate the logic of the classic printk() code [**]


[*] A missing put_printk_buffer() might cause that we would get
    out of buffers. But the same problem is with locks,
    disabled preemption, disabled interrupts, seq_buffer,
    alloc/free. Such problems happen but they are rare.

    Also I do not expect that the same buffer would be shared
    between many functions. Therefore it should be easy
    to use it correctly.


[**] I admit that cont buffer implementation is much easier
     after removing the early flush to consoles but still...


Anyway, I do not think that both implementations are worth it.
We could keep both for some transition period but we should
remove the old one later.

A question.

How bad would it actually be to:

- Allocate seq_buf 512-bytes buffer (GFP_ATOMIC) just-in-time, when we
  need it.
    // How often systems cannot allocate a 512-byte buffer? //

- OK, assuming that systems around the world are so badly OOM like all the
  time and even kmalloc(512) is absolutely impossible, then have a fallback
  to the existing 'cont' handling; it just looks to me better than a plain
  printk()-fallback with removed 'cont' support.

This would prevent removing the fallback to struct cont. OOM is
one important scenario where continuous lines are used.

- Do not allocate seq_buf if we are in printk-safe or in printk-nmi mode.
  To avoid "buffering for the sake of buffering". IOW, when in printk-safe
  use printk-safe.

Sure, my plan is to add a helper function is_buffered_printk_context() or so
that would check printk_context. Then we could do the following in
vprintk_buffered()

	if (is_buffered_printk_context())
		vprintk_func(....);

It might be added on top of the current patchset. I opened this
problem once and it got lost. So I did not want to complicate
it at this moment.

Best Regards,
Petr