> >From looking at fs/block_dev.c, I understand that opening a block
> device with O_SYNC does indeed give you unbuffered devices. If this is
> the case, why is everybody claiming Linux does not support it?
O_SYNC on a block device simply means that a write won't return
until the data has been written to the device. It still goes
through the buffer cache and you still have to copy the data
between user memory and the buffer cache.
What heavy iron databases seem to want is the ability to schedule
all their own I/O using AIO in blocksized buffers with the kernel
side doing the I/O direct to/from the given buffer in user space.
No buffer copying. No kernel memory wasted in buffering. No read
ahead other than that specifically done by the program.
Mike
-- .----------------------------------------------------------------------. | Mike Jagdis | Internet: mailto:mike@roan.co.uk | | Roan Technology Ltd. | | | 54A Peach Street, Wokingham | Telephone: +44 118 989 0403 | | RG40 1XG, ENGLAND | Fax: +44 118 989 1195 | `----------------------------------------------------------------------'