Under the assumption that this is related to harddrive (sata) I've tried changing fstab to not use the realtime function, but with no avail. The 2.6.25-7/8 kernels did not contribute to any improvement either.
However there's a major clue (ugly fix),
echo 1 > /proc/sys/vm/block_dump
this triggers repetitive debug writes to disk, thus works like a "keep-alive". The funny thing is that the freezes cease when the block_dump is enabled. If the type of freeze that you experience is akin of mine then this should work. But this is nothing I would recommend. Also, I don't have tested this long enough to be sure, but 1h+ without freeze is sure a great improvement in my case.
Under the assumption that this is related to harddrive (sata) I've tried changing fstab to not use the realtime function, but with no avail. The 2.6.25-7/8 kernels did not contribute to any improvement either.
However there's a major clue (ugly fix), vm/block_ dump
echo 1 > /proc/sys/
this triggers repetitive debug writes to disk, thus works like a "keep-alive". The funny thing is that the freezes cease when the block_dump is enabled. If the type of freeze that you experience is akin of mine then this should work. But this is nothing I would recommend. Also, I don't have tested this long enough to be sure, but 1h+ without freeze is sure a great improvement in my case.