Special Kernel Modification
Hi Jeff,
Quoted 2 lines · Jeff Dike
Ummm, how about
O_DIRECTinstead ofO_SYNC(or maybe as well, my googling hasn't been clear on whetherO_DIRECTbypasses the cache on writes as well)?
It does, but it has been deimplemented in the Linus tree and never put into the -ac tree. You would need -aa for working O_DIRECT or use a raw device.
Also they both have some restrictions on buffer alignment, but these would be already fulfilled if UML internally writes through the page cache (which it does of course)
-Andi