Re: [PATCH] I/O space write barrier

[Jeremy Higdon <jeremy@sgi.com>]

Here's the qla1280 patch to go with Jesse's mmiowb patch.
Comments are verbose to satisfy a request from Mr. Bottomley.

signed-off-by: Jeremy Higdon  <jeremy@sgi.com>

===== drivers/scsi/qla1280.c 1.65 vs edited =====
--- 1.65/drivers/scsi/qla1280.c	2004-07-28 20:59:10 -07:00
+++ edited/drivers/scsi/qla1280.c	2004-09-29 23:43:30 -07:00
@@ -3397,8 +3397,22 @@
 		"qla1280_64bit_start_scsi: Wakeup RISC for pending command\n");
 	sp->flags |= SRB_SENT;
 	ha->actthreads++;
+
+	/*
+	 * Update request index to mailbox4 (Request Queue In).
+	 * The mmiowb() ensures that this write is ordered with writes by other
+	 * CPUs.  Without the mmiowb(), it is possible for the following:
+	 *    CPUA posts write of index 5 to mailbox4
+	 *    CPUA releases host lock
+	 *    CPUB acquires host lock
+	 *    CPUB posts write of index 6 to mailbox4
+	 *    On PCI bus, order reverses and write of 6 posts, then index 5,
+	 *       causing chip to issue full queue of stale commands
+	 * The mmiowb() prevents future writes from crossing the barrier.
+	 * See Documentation/DocBook/deviceiobook.tmpl for more information.
+	 */
 	WRT_REG_WORD(&reg->mailbox4, ha->req_ring_index);
-	(void) RD_REG_WORD(&reg->mailbox4); /* PCI posted write flush */
+	mmiowb();
 
  out:
 	if (status)
@@ -3665,8 +3679,22 @@
 		"for pending command\n");
 	sp->flags |= SRB_SENT;
 	ha->actthreads++;
+
+	/*
+	 * Update request index to mailbox4 (Request Queue In).
+	 * The mmiowb() ensures that this write is ordered with writes by other
+	 * CPUs.  Without the mmiowb(), it is possible for the following:
+	 *    CPUA posts write of index 5 to mailbox4
+	 *    CPUA releases host lock
+	 *    CPUB acquires host lock
+	 *    CPUB posts write of index 6 to mailbox4
+	 *    On PCI bus, order reverses and write of 6 posts, then index 5,
+	 *       causing chip to issue full queue of stale commands
+	 * The mmiowb() prevents future writes from crossing the barrier.
+	 * See Documentation/DocBook/deviceiobook.tmpl for more information.
+	 */
 	WRT_REG_WORD(&reg->mailbox4, ha->req_ring_index);
-	(void) RD_REG_WORD(&reg->mailbox4); /* PCI posted write flush */
+	mmiowb();
 
 out:
 	if (status)
@@ -3776,9 +3804,21 @@
 	} else
 		ha->request_ring_ptr++;
 
-	/* Set chip new ring index. */
+	/*
+	 * Update request index to mailbox4 (Request Queue In).
+	 * The mmiowb() ensures that this write is ordered with writes by other
+	 * CPUs.  Without the mmiowb(), it is possible for the following:
+	 *    CPUA posts write of index 5 to mailbox4
+	 *    CPUA releases host lock
+	 *    CPUB acquires host lock
+	 *    CPUB posts write of index 6 to mailbox4
+	 *    On PCI bus, order reverses and write of 6 posts, then index 5,
+	 *       causing chip to issue full queue of stale commands
+	 * The mmiowb() prevents future writes from crossing the barrier.
+	 * See Documentation/DocBook/deviceiobook.tmpl for more information.
+	 */
 	WRT_REG_WORD(&reg->mailbox4, ha->req_ring_index);
-	(void) RD_REG_WORD(&reg->mailbox4); /* PCI posted write flush */
+	mmiowb();
 
 	LEAVE("qla1280_isp_cmd");
 }