target/linux/s3c24xx/patches-2.6.24/1316-jffs2-choke-gc-thread.patch.patch

   1 From 9706327002caebe6633c93e605882ea37172ec57 Mon Sep 17 00:00:00 2001
   2 From: Andres Salomon <dilinger@debian.org>
   3 Date: Mon, 3 Nov 2008 01:08:25 +0000
   4 Subject: [PATCH] jffs2-choke-gc-thread.patch
   5
   6 I've noticed some pretty poor behavior on OLPC machines after bootup, when
   7 gdm/X are starting.  The GCD monopolizes the scheduler (which in turns means
   8 it gets to do more nand i/o), which results in processes taking much much
   9 longer than they should to start.
  10
  11 As an example, on an OLPC machine going from OFW to a usable X (via auto-login
  12 gdm) takes 2m 30s.  The majority of this time is consumed by the switch into
  13 graphical mode.  With this patch, we cut a full 60s off of bootup time.  After
  14 bootup, things are much snappier as well.
  15
  16 Note that we have seen a CRC node error with this patch that causes the machine
  17 to fail to boot, but we've also seen that problem without this patch.
  18
  19 Signed-off-by: Andres Salomon <dilinger@debian.org>
  20 ---
  21  fs/jffs2/background.c |   18 +++++++++++-------
  22  1 files changed, 11 insertions(+), 7 deletions(-)
  23
  24 diff --git a/fs/jffs2/background.c b/fs/jffs2/background.c
  25 index 8adebd3..f38d557 100644
  26 --- a/fs/jffs2/background.c
  27 +++ b/fs/jffs2/background.c
  28 @@ -95,13 +95,17 @@ static int jffs2_garbage_collect_thread(void *_c)
  29                         schedule();
  30                 }
  31
  32 -               /* This thread is purely an optimisation. But if it runs when
  33 -                  other things could be running, it actually makes things a
  34 -                  lot worse. Use yield() and put it at the back of the runqueue
  35 -                  every time. Especially during boot, pulling an inode in
  36 -                  with read_inode() is much preferable to having the GC thread
  37 -                  get there first. */
  38 -               yield();
  39 +               /* Problem - immediately after bootup, the GCD spends a lot
  40 +                * of time in places like jffs2_kill_fragtree(); so much so
  41 +                * that userspace processes (like gdm and X) are starved
  42 +                * despite plenty of cond_resched()s and renicing.  Yield()
  43 +                * doesn't help, either (presumably because userspace and GCD
  44 +                * are generally competing for a higher latency resource -
  45 +                * disk).
  46 +                * This forces the GCD to slow the hell down.   Pulling an
  47 +                * inode in with read_inode() is much preferable to having
  48 +                * the GC thread get there first. */
  49 +               schedule_timeout_interruptible(msecs_to_jiffies(50));
  50
  51                 /* Put_super will send a SIGKILL and then wait on the sem.
  52                  */
  53 --
  54 1.5.6.5
  55