On Sun, 5 Jan 2003 [EMAIL PROTECTED] wrote: > In message <[EMAIL PROTECTED]>, Bruce Evans writes: > >On Sun, 5 Jan 2003 [EMAIL PROTECTED] wrote: > >> This is not tested with DEVFS I take it ? > > > >It doesn't affect devfs because devfs doesn't go through ufs. It goes > >straight to the default vnodeop table so it gets std* since it doesn't > >override them. > > Uhm, no. DEVFS only goes to the default vector for directories, for > devices it goes to spec_vnoperate.
Hmm, that means that the spec_vnoperate() overrides actually worked, so devfs has never had locking and fixing specfs might break devfs :-). But devfs didn't have the bug either. This is presumably ufs stays out of its way in another way: it doesn't use ffs_vget(), so the vnode is not bogusly locked initially. > >I'm getting some other panics. One while writing this was > >"bwrite: buffer is not busy???". > > Yes, I'm hunting that one atm, but havn't found a way to reproduce. > > Did you get any complaints about the wrong strategy for the wrong > type of node before this panic ? I didn't notice one for that, but there is a panic for running wine which seems to be easy to reproduce and the message occurred just before that for at least the second of 2 panics in 2 attempts to run wine here. It was something simple involving vop_stdgetpages ...-> ffsext_strategy ... ffs presumably avoids this path by having a specialized getpages. > Do you have DEBUG_VFS_LOCKS in your kernel ? No. Also no INVARIANTS and the like. A profiling kernel (with profiling not running) seemed to panic faster. Bruce To Unsubscribe: send mail to [EMAIL PROTECTED] with "unsubscribe freebsd-current" in the body of the message
