On Monday 26 January 2004 18:58, Alan Cox wrote:
On Mon, Jan 26, 2004 at 11:07:06AM -0600, Steven Pritchard wrote:
> dd_rescue handles that by using a large block size until it hits an
> error, then retrying with a small block size.
> I've had to use it a few times. It works really well.
The file system may still do smart stuff like merging requests. Also only
usually used for reading from block devices (disks).
raw I/O to disks really lets you get down to the media block size not
fs block size (eg losing 512 bytes not 4k). Thats normally only an issue
with floppies because the 'real' blocksize of a hard disk is now typically
Does the kernel do read-ahead in the block layer or in vfs?