MFp4(simokawa_sbp)

Improve if_fwe performance.
- Simplify mbuf handling by using bulkxfer.
	Now, it uses mbuf clusters for RX buffer as usual ethernet drivers.
- Recycle struct xfer buffer and don't call malloc at runtime.
- Count input and output errors.
- Handle a mbuf chain longer than 6 correctly.
- Increase queue length.

1 files changed, 7 insertions, 6 deletions

diff --git a/sys/dev/firewire/fwdev.c b/sys/dev/firewire/fwdev.c
index fe79bc7..e590edc 100644
--- a/sys/dev/firewire/fwdev.c
+++ b/sys/dev/firewire/fwdev.c
@@ -67,6 +67,7 @@ static	d_mmap_t	fw_mmap;
 
 struct cdevsw firewire_cdevsw = 
 {
+#if __FreeBSD_version >= 500104
 	.d_open =	fw_open,
 	.d_close =	fw_close,
 	.d_read =	fw_read,
@@ -77,6 +78,10 @@ struct cdevsw firewire_cdevsw =
 	.d_name =	"fw",
 	.d_maj =	CDEV_MAJOR,
 	.d_flags =	D_MEM
+#else
+	fw_open, fw_close, fw_read, fw_write, fw_ioctl,
+	fw_poll, fw_mmap, nostrategy, "fw", CDEV_MAJOR, nodump, nopsize, D_MEM
+#endif
 };
 
 static int
@@ -707,18 +712,14 @@ fw_ioctl (dev_t dev, u_long cmd, caddr_t data, int flag, fw_proc *td)
 
 		for(i = 0 ; i < sc->fc->ir[sub]->bnchunk; i++){
 			ir->bulkxfer[i].buf =
-				ir->buf +
-				i * sc->fc->ir[sub]->bnpacket *
-			  	sc->fc->ir[sub]->psize;
+				ir->buf + i * ir->bnpacket * ir->psize;
 			STAILQ_INSERT_TAIL(&ir->stfree,
 					&ir->bulkxfer[i], link);
 			ir->bulkxfer[i].npacket = ir->bnpacket;
 		}
 		for(i = 0 ; i < sc->fc->it[sub]->bnchunk; i++){
 			it->bulkxfer[i].buf =
-				it->buf +
-				i * sc->fc->it[sub]->bnpacket *
-			  	sc->fc->it[sub]->psize;
+				it->buf + i * it->bnpacket * it->psize;
 			STAILQ_INSERT_TAIL(&it->stfree,
 					&it->bulkxfer[i], link);
 			it->bulkxfer[i].npacket = it->bnpacket;