On Mon, Jul 26, 2010 at 06:33:57PM +0300, Antti Kantee wrote: > On Mon Jul 26 2010 at 15:22:17 +0000, Juergen Hannken-Illjes wrote: > > Module Name: src > > Committed By: hannken > > Date: Mon Jul 26 15:22:16 UTC 2010 > > > > Modified Files: > > src/sys/kern: vfs_subr.c > > > > Log Message: > > When both vget() and vrelel() call vn_lock() we know VI_XLOCK is clear. > > No need to use LK_INTERLOCK or LK_RETRY here. > > I'd like the commit message to include the essential bit about them > examining the return value and not blindly plowing on after vn_lock(). > Purely from the commit message I get the impression that this introduced > a race condition.
Commit message changed as: When both vget() and vrelel() call vn_lock() we know VI_XLOCK is clear. No need to use LK_INTERLOCK or LK_RETRY here. The return value of vn_lock() is already examined here. Ok: Antti Kantee <po...@netbsd.org> -- Juergen Hannken-Illjes - hann...@eis.cs.tu-bs.de - TU Braunschweig (Germany)