Hi folks.
I'm still trying to push forward the coda kernel extension for MacOS, but this has shown to be more troublesome than I expected. I have been struggling quite a while with the problem of losing all my vnodes when copying lots of files, making the system hang or crash.
Loking at the FreeBSD code in coda_fsync, I see that whoever wrote this was unsure if it was wise to do an upcall into venus_fsync or not. The code goes like this:
/*
850 * We can expect fsync on any vnode at all if venus is pruging it.
851 * Venus can't very well answer the fsync request, now can it?
852 * Hopefully, it won't have to, because hopefully, venus preserves
853 * the (possibly untrue) invariant that it never purges an open
854 * vnode. Hopefully.
855 */
856 if (cp->c_flags & C_PURGING) {
857 return(0);
858 }
859
860 /* needs research */
861 return 0;
862 error = venus_fsync(vtomi(vp), &cp->c_fid, cred, p);
863
864 CODADEBUG(CODA_FSYNC, myprintf(("in fsync result %d\n",error)); );
865 return(error);
866 }
Note the return(0) before calling venus_fsync.
Can anybody explain this to me. Do we still need research ? Is is safe to do the upcall or not?
The linux code seems to do the upcall.
-- Christer
