IB/ipath: HW workaround for case where chip can send but not receive

Workaround a QLE7140 problem that in rare cases causes flow control
problems after link recovery by forcing a link retrain after recovery.
A module parameter is provided to control the behavior in case it causes
problems.

Signed-off-by: Dave Olson <dave.olson@qlogic.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>

3 files changed, 24 insertions, 0 deletions

diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c
index 5605f4f27521..2cad7335681b 100644
--- a/drivers/infiniband/hw/ipath/ipath_driver.c
+++ b/drivers/infiniband/hw/ipath/ipath_driver.c
@@ -82,6 +82,10 @@ module_param_named(hol_timeout_ms, ipath_hol_timeout_ms, uint, S_IRUGO);
 MODULE_PARM_DESC(hol_timeout_ms,
 	"duration of user app suspension after link failure");
 
+unsigned ipath_linkrecovery = 1;
+module_param_named(linkrecovery, ipath_linkrecovery, uint, S_IWUSR | S_IRUGO);
+MODULE_PARM_DESC(linkrecovery, "enable workaround for link recovery issue");
+
 MODULE_LICENSE("GPL");
 MODULE_AUTHOR("QLogic <support@pathscale.com>");
 MODULE_DESCRIPTION("QLogic InfiniPath driver");
diff --git a/drivers/infiniband/hw/ipath/ipath_intr.c b/drivers/infiniband/hw/ipath/ipath_intr.c
index dde5dfc9fcf5..ed82ecbb02da 100644
--- a/drivers/infiniband/hw/ipath/ipath_intr.c
+++ b/drivers/infiniband/hw/ipath/ipath_intr.c
@@ -366,6 +366,22 @@ static void handle_e_ibstatuschanged(struct ipath_devdata *dd,
 	dd->ipath_ibpollcnt = 0; /* not poll*, now */
 	ipath_stats.sps_iblink++;
 
+	if (ibstate != init && dd->ipath_lastlinkrecov && ipath_linkrecovery) {
+		u64 linkrecov;
+		linkrecov = ipath_snap_cntr(dd,
+			dd->ipath_cregs->cr_iblinkerrrecovcnt);
+		if (linkrecov != dd->ipath_lastlinkrecov) {
+			ipath_dbg("IB linkrecov up %Lx (%s %s) recov %Lu\n",
+				ibcs, ib_linkstate(dd, ibcs),
+				ipath_ibcstatus_str[ltstate],
+				linkrecov);
+			/* and no more until active again */
+			dd->ipath_lastlinkrecov = 0;
+			ipath_set_linkstate(dd, IPATH_IB_LINKDOWN);
+			goto skip_ibchange;
+		}
+	}
+
 	if (ibstate == init || ibstate == arm || ibstate == active) {
 		*dd->ipath_statusp &= ~IPATH_STATUS_IB_NOCABLE;
 		if (ibstate == init || ibstate == arm) {
@@ -392,6 +408,8 @@ static void handle_e_ibstatuschanged(struct ipath_devdata *dd,
 				IPATH_NOCABLE);
 			ipath_hol_down(dd);
 		} else {  /* active */
+			dd->ipath_lastlinkrecov = ipath_snap_cntr(dd,
+				dd->ipath_cregs->cr_iblinkerrrecovcnt);
 			*dd->ipath_statusp |=
 				IPATH_STATUS_IB_READY | IPATH_STATUS_IB_CONF;
 			dd->ipath_flags |= IPATH_LINKACTIVE;
diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h
index 960d5b7e7865..b8b81cb745b9 100644
--- a/drivers/infiniband/hw/ipath/ipath_kernel.h
+++ b/drivers/infiniband/hw/ipath/ipath_kernel.h
@@ -309,6 +309,7 @@ struct ipath_devdata {
 	ipath_err_t ipath_lasthwerror;
 	/* errors masked because they occur too fast */
 	ipath_err_t ipath_maskederrs;
+	u64 ipath_lastlinkrecov; /* link recoveries at last ACTIVE */
 	/* time in jiffies at which to re-enable maskederrs */
 	unsigned long ipath_unmasktime;
 	/* count of egrfull errors, combined for all ports */
@@ -1099,6 +1100,7 @@ dma_addr_t ipath_map_single(struct pci_dev *, void *, size_t, int);
 #endif
 
 extern unsigned ipath_debug; /* debugging bit mask */
+extern unsigned ipath_linkrecovery;
 extern unsigned ipath_mtu4096;
 
 #define IPATH_MAX_PARITY_ATTEMPTS 10000 /* max times to try recovery */