Messages in this thread | | | Date | Mon, 29 Jun 1998 00:33:08 +0200 | From | Erik Corry <> | Subject | Re: sendfile(2) idea (was: Thread implementations) |
| |
On Sun, Jun 28, 1998 at 09:57:30AM -0700, Linus Torvalds wrote: > > > On Sun, 28 Jun 1998, Oliver Xymoron wrote: > > > > Is it possible to optimize the "mmap()+write()" combination instead and > > avoid adding a new scheme? > > No. Think TLB and page table overhead. Think TLB coherency in SMP. Think > virtual caches.
The suggestion was to do lazy mmap, ie you map at an address that was going to fault anyway, don't alter the TLB, 'just' make copy_from_user recognise what's going on. It doesn't have to work from user space since the mmap is never touched from user space anyway.
> mmap() _really_ really isn't the solution to zero-copy.
You are probably right, but...
-- Erik Corry
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |