• Contact: tech-kern
• Duration estimate: 2-3 months

NetBSD has a fixed, kernel-wide upper limit on transfer size, which is currently set to 64k on most port. This is too small to have good performances on modern IDE and SCSI hardware; on the other hand some devices can't do more than 64k, and in some case are even limited to less (the Xen virtual block device for example). Software RAID will also cause requests to be split in multiple smaller requests.

NetBSD would greatly benefit from a more intelligent buffer queue management between the block device drivers and the higher levels (the framework here currently only applies some selectable algorithm to sort the queue). This framework should be able to split buffers too large for a device into smaller ones, or aggregate multiple contiguous requests into a larger one. This will most probably require change to at last some block device drivers.