dont try to prevent multiple concurrent runs of a devices start routine

by setting flags around the loop. there is a race which can prevent
necessary work being completed by any of the currently running instances
of xxstart.

the caveat with the removal of this code is because multiple xxstarts can
be running at the same time they can cause io reordering, but that is less
of a problem than no io.

found by and fix tested by claudio@
debugged with krw@ claudio@ beck@ deraadt@

1 files changed, 1 insertions, 14 deletions

diff --git a/sys/scsi/cd.c b/sys/scsi/cd.c
index 5b993c3077c..49ab9d6f73b 100644
--- a/sys/scsi/cd.c
+++ b/sys/scsi/cd.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: cd.c,v 1.157 2009/12/16 10:51:28 dlg Exp $	*/
+/*	$OpenBSD: cd.c,v 1.158 2010/01/09 21:12:06 dlg Exp $	*/
 /*	$NetBSD: cd.c,v 1.100 1997/04/02 02:29:30 mycroft Exp $	*/
 
 /*
@@ -104,7 +104,6 @@ struct cd_softc {
 #define	CDF_LABELLING	0x08		/* writing label */
 #define	CDF_ANCIENT	0x10		/* disk is ancient; for minphys */
 #define CDF_WAITING	0x100
-#define CDF_STARTING	0x200
 	struct scsi_link *sc_link;	/* contains our targ, lun, etc. */
 	struct cd_parms {
 		u_int32_t blksize;
@@ -611,15 +610,6 @@ cdstart(void *v)
 	 * Check if the device has room for another command
 	 */
 
-	mtx_enter(&sc->sc_start_mtx);
-	if (ISSET(sc->sc_flags, CDF_STARTING)) {
-		mtx_leave(&sc->sc_start_mtx);
-		return;
-	}
-
-	SET(sc->sc_flags, CDF_STARTING);
-	mtx_leave(&sc->sc_start_mtx);
-
 	CLR(sc->sc_flags, CDF_WAITING);
 	while (!ISSET(sc->sc_flags, CDF_WAITING) &&
 	    (bp = cd_buf_dequeue(sc)) != NULL) {
@@ -700,9 +690,6 @@ cdstart(void *v)
 
 		scsi_xs_exec(xs);
 	}
-	mtx_enter(&sc->sc_start_mtx);
-	CLR(sc->sc_flags, CDF_STARTING);
-	mtx_leave(&sc->sc_start_mtx);
 }
 
 void