On Wed, 23 Feb 2000, William Montgomery wrote:
>
> On Tue, 22 Feb 2000, Andrea Arcangeli wrote:
>
> > On Tue, 22 Feb 2000, Ingo Molnar wrote:
> >
> > >[..] do a 'b_count++; schedule(); b_count--; goto repeat;'
> > >and everything should be fine. [..]
> >
> > unless browsing the whole list takes more than the task timeslice (around
> > 10msec with a reniced +20 task). With 2giga in buffer cache 10msec of
> > latnecy are realistic. If that will happen what you suggested above will
> > cause the task to hang forever in R state.
> >
> > Also after the b_count-- you can process the bh. You should "goto repeat"
> > only after processing the bh. Just like what I am doing in 2.3.47 in
> > invalidate_buffers/set_blocksize.
> >
> I fixed the buffer.c code according to Ingo and removed a
> conditional_schedule from mark_buffer_clean in fs.h as spotted by Andrea.
> I have been testing for over 18hrs and no stuck processes yet.
>
> Now to make more robust, there appear to be 2 remaining problems:
> 1) goto repeat only after processing the bh in buffer.c.
> 2) fix inode bug.
>
> Correct?
>
Don't you also have to check whether the lowlatency goal has still been
met?
john alvord
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Wed Feb 23 2000 - 21:00:33 EST