Re: [PATCH] fix sys cpumap for > 352 NR_CPUS

[Paul Jackson <pj@sgi.com>]

> Honestly, I dislike the static check altogether ...

  The build time check was your idea in the first place, as I recall.
  I hadn't added it to my variants.  Apparently we agree not to add it.
  Ok.

> Because now you have silently truncated, which is much worse than

  I absolutely agree with your dislike of hidden intermittent failures.

  For a constant failure such as this, even if everyone misses the
  botch for the first few times that SGI boots a bazillion CPU system
  in the lab, it will get noticed soon enough.  This is in fact
  exactly what happened with the 99 char limit that was there now.

> 	len = cpumask_scnprintf(buf, PAGE_SIZE, mask);
> 	if (len == PAGE_SIZE)
> 		return -ENOMEM;

  That looks nice.

  If you want me to add the "if (len ..."  check, let me know.
  Or if you want to send Andrew a patch that adds it, I'll
  gladly support that.

  ==> Do note that I had to change the -1UL to PAGE_SIZE, in a
      patch to Andrew about 12 hours ago.  The *scnprintf()
      family of fine formatting functions suppresses all requested
      output if handed a length with the high bit set.

> That would be extremely unusual; we tend not to panic ...

  Yup - I think it was Greg who said the same thing.  Clearly this is
  not a panic.

  I was wrong to suggest panic'ing here.

> I'll do it; seems like we need negotiation with Greg-KH, too.

  Ok - have fun.

> I question anyone's ability to produce a perfectly balanced solution
> without any external input.

  Whatever ... my view of who was saying and doing what here doesn't
  entirely match yours.

  So be it.

-- 
                          I won't rest till it's the best ...
                          Programmer, Linux Scalability
                          Paul Jackson <pj@sgi.com> 1.650.933.1373