On 2008-05-27 at 17:01:23 [+0200], Axel Dörfler <axeld@xxxxxxxxxxxxxxxx> wrote: > Ingo Weinhold <ingo_weinhold@xxxxxx> wrote: > > Come to think of it, we need to disable the file cache for the swap > > file or > > have it operate in a special mode. Axel, do you have anything special > > in mind? > > While O_NOCACHE has not yet been implemented, it's also thought for > this case. > I haven't thought it through, but basically, the whole file map has to > be cached upfront then, so that no allocations will be necessary > anymore when touching the swap file. > I may add a call to the file cache API that triggers just that, so that > the file system can then use it on open() when O_NOCACHE has been > specified. Sounds good. We should probably also prevent other open()s from succeeding. I guess locking the node should do the trick. CU, Ingo