ULE status

[Ion-Mihai Tetcu]

On Sat, 12 Feb 2005 01:47:27 +0200
Ion-Mihai Tetcu <itetcu at people.tecnik93.com> wrote:

> On Tue, 8 Feb 2005 08:39:15 -0800
> Kris Kennaway <kris at obsecurity.org> wrote:
> 
> > On Tue, Feb 08, 2005 at 03:58:22PM +0200, Ion-Mihai Tetcu wrote:
> > > On Tue, 8 Feb 2005 14:45:17 +0200 (EET)
> > > "Viktor Ivanov" <v0rbiz at icon.bg> wrote:
> > > 
> 
>  [ ... ]
> 
> > > > On the other hand, using SCHED_ULE improves sound quality and general
> > > > system 'response' concerning GUI... don't know 'bout performance.
> > > 
> > > By any chance does it help with copying from ata disks on different
> > > controllers ? For me on large files this brings up "swap_pager:
> > > indefinite wait buffer" with 4BSD.
> > 
> > That doesn't sound like a scheduler problem, rather a hardware or ata
> > driver problem.  Did you try sos' new driver yet?
> 
> Well, I'm running my desktop with an ULE kernel now, snippets bellow:
> options         SCHED_ULE               # ULE scheduler
> options         ADAPTIVE_GIANT          # Giant mutex is adaptive.
> options         HZ=1200
> options         INVARIANTS              # Enable calls of extra sanity checking
> options         INVARIANT_SUPPORT       # Extra sanity checks of internal structures,
> #                                        required by INVARIANTS
> options         WITNESS                 # Enable checks to detect deadlocks and cycles
> options         WITNESS_SKIPSPIN        # Don't run witness on spinlocks for speed
> 
> and
>  # sysctl debug.witness
> debug.witness.skipspin: 1
> debug.witness.trace: 1
> debug.witness.kdb: 0
> debug.witness.watch: 0
> 
> 
> And while I don't see may improvements in general operation, at least
> this problem is gone. I still get a few seconds pauses in sound and
> switching between app in X is sluggish while copying large files, but at
> least I don't get swap pager complains and I'd say there's a 3x
> responsiveness improvement.
> 
> 
> I didn't test ata patches yet, just to take one step at the time.

But the problem is seems there somewhere, at least `vmstat -i 1` show
98-103% busy ad0 when copying from ad2 to ad0 and the system doesn't
swap or look loaded.

Mem:KB    REAL            VIRTUAL                     VN PAGER  SWAP PAGER
        Tot   Share      Tot    Share    Free                  in  out     in  out
Act  326880   32656   522764    64152   34144 count
All  481012   36440  3327348    73980        pages
                                                                           Interrupts
Proc:r  p  d  s  w    Csw  Trp  Sys  Int  Sof  Flt        cow    1931 total
     1  2  6102      5661  305 7295 3128  429   56 105476 wire        1: atkb
                                                   335032 act         3: sio1
10.1%Sys   2.3%Intr  5.4%User  0.0%Nice 82.2%Idl    30768 inact       4: sio0
|    |    |    |    |    |    |    |    |    |      12072 cache       6: fdc0
=====+>>>                                           22072 free    128 8: rtc
                                                          daefr   159 12: psm
Namei         Name-cache    Dir-cache                     prcfr       13: npx
    Calls     hits    %     hits    %                     react    57 14: ata
     1686     1686  100                                   pdwak    51 15: ata
                                       26 zfod            pdpgs   143 16: xl0
Disks   ad0   ad2                      26 ofod            intrn   189 22: pcm
KB/t    128   128                         %slo-z    61472 buf    1204 0: clk
tps      28    26                      12 tfree        41 dirtybuf
MB/s   3.46  3.21                                   35666 desiredvnodes
% busy   98     9                                    8916 numvnodes
                                                     7269 freevnodes



-- 
IOnut
Unregistered ;) FreeBSD "user"