Re: more locks..
Frank van der Linden <[email protected]>
| Newsgroups | gmane.os.netbsd.devel.smp |
|---|---|
| Message-ID | <[email protected]> |
On Sun, Feb 02, 2003 at 11:44:28PM +0100, Paul Kranenburg wrote: > I've looked into making several kernel modules MP-safer. Attached is > a set of diffs for the buffer cache (mostly vfs_bio.c) for your > inspection. With this patch, my 2-cpu machine can run the buffer cache > code concurrently, although I didn't push it very hard (just ffs & lfs > on a scsi disk). > > Note that the patch omits bunch of files that need trivial changes, such > as initing a buffer's simple lock. First of all, it's cool that you're looking into this. I haven't looked at the diff closely yet, but: how did you test this? You have to make a decision to (not) take the big lock somewhere, and be careful that you're not getting in to trouble because you're calling other subsystems (e.g UVM) that haven't been verified safe yet. - Frank