Re: pre-patch-2.0.31 try_to_free_page() messages
H.J. Lu (hjl@lucon.org)
Sun, 1 Jun 1997 10:29:55 -0700 (PDT)
>
> From: alan@lxorguk.ukuu.org.uk (Alan Cox)
> Date: Sun, 1 Jun 1997 15:47:49 +0100 (BST)
>
> > Is SMP supposed to be stable now?
>
> SMP should be the same from 2.0.18 upwards. The performance issues
> under heavy I/O load (losing serial chars, slipping clock) aren not going to
> change in the 2.0.x series. I do intend to put Leonards memory handling tweaks
> into the code however.
>
> I assume you're referring to my SMP deadlock patch?
>
> My 2.0.30 SMP deadlock patch has solved the problem 100% for my systems but
> some other people still reported problems. However, despite Ingo's posting a
> clever new deadlock detection patch, no one has reported tracking down any of
> the remaining deadlocks.
>
2.0.30 plus Leonard's patch has been running very well on my heavily
loaded SMP machine. Sometimes my load average is over 40. I was doing
heavy SCSI disk I/O on 4 SCSI drives at the same time. Also I was
building gcc, binutils, XFree86 on a NFS client machine accessing the
sources on my SMP machine which is my NFS server.
But with linux 2.1.39/2.1.40, my SMP machine crashed/locked up very
easily under my heavy load.
H.J.