Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
> Regarding the OOPS. By now I do not understand what is going on > here... > > there is a call to wake_up() in _drbd_process_ee. But why should > a page-fault happen within this wake_up() call ??? now, we introduced this typo in the new drbd_get_ee, al_wait instead of ee_wait, finish_wait using the wrong spinlock, and you can have two threads modifying the waitq lists at the same time. if dereferencing LIST_POISON[12] involves a page-fault, then that might be the root of the problem :-/ so if that is the case, current CVS might already help. > While looking at the description of your setup, I thought that > you might be the first one who tested DRBD-0.7 on a machine > that actually uses HIGHMEM ? > > Could you try to reproduce this with HIGHMEM disabled ? yes please. Lars Ellenberg