pdflush eating a lot of CPU on heavy NFS I/O

[Brent Cook]

Hi,

  Running any kernel from the 2.6.6-rc* series (and a few previous
-mm*'s), the pdflush process starts using near 100% CPU indefinitely after
a few minutes of initial NFS traffic, as far as I can tell.

To trigger this, I can either compile a kernel with the source residing on
an NFS share, or compile something bigger, like KDE.

I get the same results running on a PIII with an i815 chipset with
ReiserFS 3, or a newer nForce2 board with an Athlon XP on ext3, so I don't
think it has to do with the IDE chipsets or filesystems. pdflush has
something to do with writing back FS data, and NFS is just the common
factor between systems that experience this problem. pdflush just seems to
hang when the system is heavily loaded and eat up all CPU resources even
when the system is otherwise idle. Renice failes to reschedule pdflush.

 This didn't seem to be a problem with 2.6.5 or 2.4. Is there something I
can do to control pdflush or to provide more information?

Thanks
 - Brent


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/