On Sat, Jun 15, 2013 at 01:26:10PM +0200, Luigi Rizzo wrote: > With "indirect buffers" in netmap/vale, i can eliminate the A->B copy, > and do A->C with a copyin in the kernel of the host. > But the per-packet budget is minuscule, and i am afraid that doing > an unconditional vslock() on each buffer is going to be too expensive > (and then i should also unwire the page ?
Using vslock() causes enormous user map fragmentation and should be avoided. The cost is basically the same as the page fault, since vslock() is a wrapper around vm_map_wire(), which just calls page fault handler for each address in region. On the other hand, I do not understand how any kernel-side lock would prevent usermode from unmapping a buffer while other thread in kernel mode access it. IMO you would be much better served with either with proc_rwmem(), as discussed before, or direct use of vm_fault_quick_hold_pages() and uiomove_fromphys() afterward. vm_fault_quick_hold_pages() also makes you code immune against usermode doing any unmap or mapping over the passed buffer.
pgphaJZJIuvh5.pgp
Description: PGP signature
