It seems Mike Smith wrote: > > > > No, it does not work at all with that maximum set. I'm using the disk to > > hold > > an FFS filesystem, and mounting it, BTW. See, in the old driver, it limited > > the transfer, but it ALSO queued the rest of the transfer. AFD doesn't, and > > I can tell this because I get random parts of kernel memory interspersed > > with > > my files (that's the corruption), so I know all of the buffer is not being > > filled. You need to queue the rest of the transfer too. > > I wrote the original code for this in wfd, and it requeues the transfer > for just this reason; the layer above doesn't retry the block operation > for the remaining residual. I'm not sure this is an error; it's too > long since I looked at the code, but I think there are some bad things > that can happen (or were suggested might happen) if you persist in > retrying the physical I/O.
Well, the code uses d_maxio in many places, just not all regrettably :( That would be the way to solve this... -Søren To Unsubscribe: send mail to majord...@freebsd.org with "unsubscribe freebsd-current" in the body of the message
