netdev.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH net] tg3: Fix RSS ring refill race condition
@ 2012-03-22  1:38 Michael Chan
  2012-03-22  1:38 ` [PATCH net] cnic: Fix parity error code conflict Michael Chan
  2012-03-22  1:58 ` [PATCH net] tg3: Fix RSS ring refill race condition David Miller
  0 siblings, 2 replies; 4+ messages in thread
From: Michael Chan @ 2012-03-22  1:38 UTC (permalink / raw)
  To: davem; +Cc: netdev

The RSS feature in tg3 hardware has only one rx producer ring for all
RSS rings.  NAPI vector 1 is special and handles the refilling of the
rx producer ring on behalf of all RSS rings.  There is a race condition
between these RSS NAPIs and the NAPI[1].  If NAPI[1] finishes checking
for refill and then another RSS ring empties the rx producer ring
before NAPI[1] exits NAPI, the chip will be completely out of SKBs in
the rx producer ring.

We fix this by adding a flag tp->rx_refill and rely on napi_schedule()/
napi_complete() to help synchronize it to close the race condition.

Update driver version to 3.123.

Signed-off-by: Michael Chan <mchan@broadcom.com>
---
 drivers/net/ethernet/broadcom/tg3.c |   25 ++++++++++++++++++++++---
 drivers/net/ethernet/broadcom/tg3.h |    1 +
 2 files changed, 23 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/tg3.c b/drivers/net/ethernet/broadcom/tg3.c
index b065746..7b71387 100644
--- a/drivers/net/ethernet/broadcom/tg3.c
+++ b/drivers/net/ethernet/broadcom/tg3.c
@@ -89,10 +89,10 @@ static inline void _tg3_flag_clear(enum TG3_FLAGS flag, unsigned long *bits)
 
 #define DRV_MODULE_NAME		"tg3"
 #define TG3_MAJ_NUM			3
-#define TG3_MIN_NUM			122
+#define TG3_MIN_NUM			123
 #define DRV_MODULE_VERSION	\
 	__stringify(TG3_MAJ_NUM) "." __stringify(TG3_MIN_NUM)
-#define DRV_MODULE_RELDATE	"December 7, 2011"
+#define DRV_MODULE_RELDATE	"March 21, 2012"
 
 #define RESET_KIND_SHUTDOWN	0
 #define RESET_KIND_INIT		1
@@ -5953,8 +5953,10 @@ next_pkt_nopost:
 		tpr->rx_std_prod_idx = std_prod_idx & tp->rx_std_ring_mask;
 		tpr->rx_jmb_prod_idx = jmb_prod_idx & tp->rx_jmb_ring_mask;
 
-		if (tnapi != &tp->napi[1])
+		if (tnapi != &tp->napi[1]) {
+			tp->rx_refill = true;
 			napi_schedule(&tp->napi[1].napi);
+		}
 	}
 
 	return received;
@@ -6134,6 +6136,7 @@ static int tg3_poll_work(struct tg3_napi *tnapi, int work_done, int budget)
 		u32 std_prod_idx = dpr->rx_std_prod_idx;
 		u32 jmb_prod_idx = dpr->rx_jmb_prod_idx;
 
+		tp->rx_refill = false;
 		for (i = 1; i < tp->irq_cnt; i++)
 			err |= tg3_rx_prodring_xfer(tp, dpr,
 						    &tp->napi[i].prodring);
@@ -6197,9 +6200,25 @@ static int tg3_poll_msix(struct napi_struct *napi, int budget)
 		/* check for RX/TX work to do */
 		if (likely(sblk->idx[0].tx_consumer == tnapi->tx_cons &&
 			   *(tnapi->rx_rcb_prod_idx) == tnapi->rx_rcb_ptr)) {
+
+			/* This test here is not race free, but will reduce
+			 * the number of interrupts by looping again.
+			 */
+			if (tnapi == &tp->napi[1] && tp->rx_refill)
+				continue;
+
 			napi_complete(napi);
 			/* Reenable interrupts. */
 			tw32_mailbox(tnapi->int_mbox, tnapi->last_tag << 24);
+
+			/* This test here is synchronized by napi_schedule()
+			 * and napi_complete() to close the race condition.
+			 */
+			if (unlikely(tnapi == &tp->napi[1] && tp->rx_refill)) {
+				tw32(HOSTCC_MODE, tp->coalesce_mode |
+						  HOSTCC_MODE_ENABLE |
+						  tnapi->coal_now);
+			}
 			mmiowb();
 			break;
 		}
diff --git a/drivers/net/ethernet/broadcom/tg3.h b/drivers/net/ethernet/broadcom/tg3.h
index 66bcfca..93865f8 100644
--- a/drivers/net/ethernet/broadcom/tg3.h
+++ b/drivers/net/ethernet/broadcom/tg3.h
@@ -3007,6 +3007,7 @@ struct tg3 {
 	u32				rx_std_max_post;
 	u32				rx_offset;
 	u32				rx_pkt_map_sz;
+	bool				rx_refill;
 
 
 	/* begin "everything else" cacheline(s) section */
-- 
1.7.1

^ permalink raw reply related	[flat|nested] 4+ messages in thread

* [PATCH net] cnic: Fix parity error code conflict
  2012-03-22  1:38 [PATCH net] tg3: Fix RSS ring refill race condition Michael Chan
@ 2012-03-22  1:38 ` Michael Chan
  2012-03-22  1:58   ` David Miller
  2012-03-22  1:58 ` [PATCH net] tg3: Fix RSS ring refill race condition David Miller
  1 sibling, 1 reply; 4+ messages in thread
From: Michael Chan @ 2012-03-22  1:38 UTC (permalink / raw)
  To: davem; +Cc: netdev

The recently added parity error handling used an error code that was
already defined for a different error.  This could lead to bnx2x
firmware assert.  We need to fix this with new error codes that are
defined for parity error only.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Reviewed-by: Eddie Wai <eddie.wai@broadcom.com>
Reviewed-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>
---
 drivers/net/ethernet/broadcom/cnic.c      |   12 +++++++-----
 drivers/net/ethernet/broadcom/cnic_defs.h |   28 +---------------------------
 drivers/net/ethernet/broadcom/cnic_if.h   |    4 ++--
 drivers/scsi/bnx2fc/bnx2fc_constants.h    |    1 +
 drivers/scsi/bnx2i/57xx_iscsi_constants.h |    1 +
 5 files changed, 12 insertions(+), 34 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/cnic.c b/drivers/net/ethernet/broadcom/cnic.c
index 7b65716..c95e7b5 100644
--- a/drivers/net/ethernet/broadcom/cnic.c
+++ b/drivers/net/ethernet/broadcom/cnic.c
@@ -47,6 +47,7 @@
 #include "bnx2x/bnx2x_hsi.h"
 #include "../../../scsi/bnx2i/57xx_iscsi_constants.h"
 #include "../../../scsi/bnx2i/57xx_iscsi_hsi.h"
+#include "../../../scsi/bnx2fc/bnx2fc_constants.h"
 #include "cnic.h"
 #include "cnic_defs.h"
 
@@ -2547,7 +2548,7 @@ static void cnic_bnx2x_kwqe_err(struct cnic_dev *dev, struct kwqe *kwqe)
 		}
 		kcqe.kcqe_op_flag = kcqe_op << KCQE_FLAGS_OPCODE_SHIFT;
 		kcqe.kcqe_op_flag |= KCQE_FLAGS_LAYER_MASK_L5_FCOE;
-		kcqe.kcqe_info1 = FCOE_KCQE_COMPLETION_STATUS_NIC_ERROR;
+		kcqe.kcqe_info1 = FCOE_KCQE_COMPLETION_STATUS_PARITY_ERROR;
 		kcqe.kcqe_info2 = cid;
 		kcqe.kcqe_info0 = l5_cid;
 
@@ -2558,7 +2559,7 @@ static void cnic_bnx2x_kwqe_err(struct cnic_dev *dev, struct kwqe *kwqe)
 
 		kcqe.kcqe_op_flag = (opcode + 0x10) << KCQE_FLAGS_OPCODE_SHIFT;
 		kcqe.kcqe_op_flag |= KCQE_FLAGS_LAYER_MASK_L5_ISCSI;
-		kcqe.kcqe_info1 = ISCSI_KCQE_COMPLETION_STATUS_NIC_ERROR;
+		kcqe.kcqe_info1 = ISCSI_KCQE_COMPLETION_STATUS_PARITY_ERR;
 		kcqe.kcqe_info2 = cid;
 		cnic_get_l5_cid(cp, BNX2X_SW_CID(cid), &kcqe.kcqe_info0);
 
@@ -2577,7 +2578,7 @@ static void cnic_bnx2x_kwqe_err(struct cnic_dev *dev, struct kwqe *kwqe)
 
 		kcqe.kcqe_op_flag = (kcqe_op << KCQE_FLAGS_OPCODE_SHIFT) |
 				    KCQE_FLAGS_LAYER_MASK_L4;
-		l4kcqe->status = L4_KCQE_COMPLETION_STATUS_NIC_ERROR;
+		l4kcqe->status = L4_KCQE_COMPLETION_STATUS_PARITY_ERROR;
 		l4kcqe->cid = cid;
 		cnic_get_l5_cid(cp, BNX2X_SW_CID(cid), &l4kcqe->conn_id);
 	} else {
@@ -3933,7 +3934,8 @@ static void cnic_cm_process_kcqe(struct cnic_dev *dev, struct kcqe *kcqe)
 	case L4_KCQE_OPCODE_VALUE_CONNECT_COMPLETE:
 		if (l4kcqe->status == 0)
 			set_bit(SK_F_OFFLD_COMPLETE, &csk->flags);
-		else if (l4kcqe->status == L4_KCQE_COMPLETION_STATUS_NIC_ERROR)
+		else if (l4kcqe->status ==
+			 L4_KCQE_COMPLETION_STATUS_PARITY_ERROR)
 			set_bit(SK_F_HW_ERR, &csk->flags);
 
 		smp_mb__before_clear_bit();
@@ -3946,7 +3948,7 @@ static void cnic_cm_process_kcqe(struct cnic_dev *dev, struct kcqe *kcqe)
 	case L4_KCQE_OPCODE_VALUE_RESET_COMP:
 	case L5CM_RAMROD_CMD_ID_SEARCHER_DELETE:
 	case L5CM_RAMROD_CMD_ID_TERMINATE_OFFLOAD:
-		if (l4kcqe->status == L4_KCQE_COMPLETION_STATUS_NIC_ERROR)
+		if (l4kcqe->status == L4_KCQE_COMPLETION_STATUS_PARITY_ERROR)
 			set_bit(SK_F_HW_ERR, &csk->flags);
 
 		cp->close_conn(csk, opcode);
diff --git a/drivers/net/ethernet/broadcom/cnic_defs.h b/drivers/net/ethernet/broadcom/cnic_defs.h
index 06ca002..382c98b 100644
--- a/drivers/net/ethernet/broadcom/cnic_defs.h
+++ b/drivers/net/ethernet/broadcom/cnic_defs.h
@@ -35,16 +35,6 @@
 #define L5CM_RAMROD_CMD_ID_SEARCHER_DELETE	(L5CM_RAMROD_CMD_ID_BASE + 14)
 #define L5CM_RAMROD_CMD_ID_TERMINATE_OFFLOAD	(L5CM_RAMROD_CMD_ID_BASE + 15)
 
-#define FCOE_KCQE_OPCODE_INIT_FUNC			(0x10)
-#define FCOE_KCQE_OPCODE_DESTROY_FUNC			(0x11)
-#define FCOE_KCQE_OPCODE_STAT_FUNC			(0x12)
-#define FCOE_KCQE_OPCODE_OFFLOAD_CONN			(0x15)
-#define FCOE_KCQE_OPCODE_ENABLE_CONN			(0x16)
-#define FCOE_KCQE_OPCODE_DISABLE_CONN			(0x17)
-#define FCOE_KCQE_OPCODE_DESTROY_CONN			(0x18)
-#define FCOE_KCQE_OPCODE_CQ_EVENT_NOTIFICATION  (0x20)
-#define FCOE_KCQE_OPCODE_FCOE_ERROR				(0x21)
-
 #define FCOE_RAMROD_CMD_ID_INIT_FUNC		(FCOE_KCQE_OPCODE_INIT_FUNC)
 #define FCOE_RAMROD_CMD_ID_DESTROY_FUNC		(FCOE_KCQE_OPCODE_DESTROY_FUNC)
 #define FCOE_RAMROD_CMD_ID_STAT_FUNC		(FCOE_KCQE_OPCODE_STAT_FUNC)
@@ -54,23 +44,6 @@
 #define FCOE_RAMROD_CMD_ID_DESTROY_CONN		(FCOE_KCQE_OPCODE_DESTROY_CONN)
 #define FCOE_RAMROD_CMD_ID_TERMINATE_CONN	(0x81)
 
-#define FCOE_KWQE_OPCODE_INIT1                  (0)
-#define FCOE_KWQE_OPCODE_INIT2                  (1)
-#define FCOE_KWQE_OPCODE_INIT3                  (2)
-#define FCOE_KWQE_OPCODE_OFFLOAD_CONN1  (3)
-#define FCOE_KWQE_OPCODE_OFFLOAD_CONN2  (4)
-#define FCOE_KWQE_OPCODE_OFFLOAD_CONN3  (5)
-#define FCOE_KWQE_OPCODE_OFFLOAD_CONN4  (6)
-#define FCOE_KWQE_OPCODE_ENABLE_CONN	(7)
-#define FCOE_KWQE_OPCODE_DISABLE_CONN	(8)
-#define FCOE_KWQE_OPCODE_DESTROY_CONN	(9)
-#define FCOE_KWQE_OPCODE_DESTROY		(10)
-#define FCOE_KWQE_OPCODE_STAT			(11)
-
-#define FCOE_KCQE_COMPLETION_STATUS_ERROR	(0x1)
-#define FCOE_KCQE_COMPLETION_STATUS_CTX_ALLOC_FAILURE	(0x3)
-#define FCOE_KCQE_COMPLETION_STATUS_NIC_ERROR	(0x5)
-
 /* KCQ (kernel completion queue) response op codes */
 #define L4_KCQE_OPCODE_VALUE_CLOSE_COMP             (53)
 #define L4_KCQE_OPCODE_VALUE_RESET_COMP             (54)
@@ -87,6 +60,7 @@
 /* KCQ (kernel completion queue) completion status */
 #define L4_KCQE_COMPLETION_STATUS_SUCCESS           (0)
 #define L4_KCQE_COMPLETION_STATUS_NIC_ERROR         (4)
+#define L4_KCQE_COMPLETION_STATUS_PARITY_ERROR	    (0x81)
 #define L4_KCQE_COMPLETION_STATUS_TIMEOUT           (0x93)
 
 #define L4_KCQE_COMPLETION_STATUS_CTX_ALLOC_FAIL    (0x83)
diff --git a/drivers/net/ethernet/broadcom/cnic_if.h b/drivers/net/ethernet/broadcom/cnic_if.h
index 60deb84..289274e 100644
--- a/drivers/net/ethernet/broadcom/cnic_if.h
+++ b/drivers/net/ethernet/broadcom/cnic_if.h
@@ -12,8 +12,8 @@
 #ifndef CNIC_IF_H
 #define CNIC_IF_H
 
-#define CNIC_MODULE_VERSION	"2.5.9"
-#define CNIC_MODULE_RELDATE	"Feb 8, 2012"
+#define CNIC_MODULE_VERSION	"2.5.10"
+#define CNIC_MODULE_RELDATE	"March 21, 2012"
 
 #define CNIC_ULP_RDMA		0
 #define CNIC_ULP_ISCSI		1
diff --git a/drivers/scsi/bnx2fc/bnx2fc_constants.h b/drivers/scsi/bnx2fc/bnx2fc_constants.h
index c12702b..dad9924 100644
--- a/drivers/scsi/bnx2fc/bnx2fc_constants.h
+++ b/drivers/scsi/bnx2fc/bnx2fc_constants.h
@@ -47,6 +47,7 @@
 #define FCOE_KCQE_COMPLETION_STATUS_CTX_FREE_FAILURE	(0x4)
 #define FCOE_KCQE_COMPLETION_STATUS_NIC_ERROR			(0x5)
 #define FCOE_KCQE_COMPLETION_STATUS_WRONG_HSI_VERSION   (0x6)
+#define FCOE_KCQE_COMPLETION_STATUS_PARITY_ERROR	(0x81)
 
 /* CQE type */
 #define FCOE_PENDING_CQE_TYPE			0
diff --git a/drivers/scsi/bnx2i/57xx_iscsi_constants.h b/drivers/scsi/bnx2i/57xx_iscsi_constants.h
index 57515f1..495a841 100644
--- a/drivers/scsi/bnx2i/57xx_iscsi_constants.h
+++ b/drivers/scsi/bnx2i/57xx_iscsi_constants.h
@@ -122,6 +122,7 @@
 #define ISCSI_KCQE_COMPLETION_STATUS_LOM_ISCSI_NOT_ENABLED              (0x51)
 
 #define ISCSI_KCQE_COMPLETION_STATUS_CID_BUSY				(0x80)
+#define ISCSI_KCQE_COMPLETION_STATUS_PARITY_ERR                         (0x81)
 
 /* SQ/RQ/CQ DB structure sizes */
 #define ISCSI_SQ_DB_SIZE    (16)
-- 
1.7.1

^ permalink raw reply related	[flat|nested] 4+ messages in thread

* Re: [PATCH net] tg3: Fix RSS ring refill race condition
  2012-03-22  1:38 [PATCH net] tg3: Fix RSS ring refill race condition Michael Chan
  2012-03-22  1:38 ` [PATCH net] cnic: Fix parity error code conflict Michael Chan
@ 2012-03-22  1:58 ` David Miller
  1 sibling, 0 replies; 4+ messages in thread
From: David Miller @ 2012-03-22  1:58 UTC (permalink / raw)
  To: mchan; +Cc: netdev

From: "Michael Chan" <mchan@broadcom.com>
Date: Wed, 21 Mar 2012 17:38:33 -0800

> The RSS feature in tg3 hardware has only one rx producer ring for all
> RSS rings.  NAPI vector 1 is special and handles the refilling of the
> rx producer ring on behalf of all RSS rings.  There is a race condition
> between these RSS NAPIs and the NAPI[1].  If NAPI[1] finishes checking
> for refill and then another RSS ring empties the rx producer ring
> before NAPI[1] exits NAPI, the chip will be completely out of SKBs in
> the rx producer ring.
> 
> We fix this by adding a flag tp->rx_refill and rely on napi_schedule()/
> napi_complete() to help synchronize it to close the race condition.
> 
> Update driver version to 3.123.
> 
> Signed-off-by: Michael Chan <mchan@broadcom.com>

Applied.

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH net] cnic: Fix parity error code conflict
  2012-03-22  1:38 ` [PATCH net] cnic: Fix parity error code conflict Michael Chan
@ 2012-03-22  1:58   ` David Miller
  0 siblings, 0 replies; 4+ messages in thread
From: David Miller @ 2012-03-22  1:58 UTC (permalink / raw)
  To: mchan; +Cc: netdev

From: "Michael Chan" <mchan@broadcom.com>
Date: Wed, 21 Mar 2012 17:38:34 -0800

> The recently added parity error handling used an error code that was
> already defined for a different error.  This could lead to bnx2x
> firmware assert.  We need to fix this with new error codes that are
> defined for parity error only.
> 
> Signed-off-by: Michael Chan <mchan@broadcom.com>
> Reviewed-by: Eddie Wai <eddie.wai@broadcom.com>
> Reviewed-by: Bhanu Prakash Gollapudi <bprakash@broadcom.com>

Applied.

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2012-03-22  1:58 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2012-03-22  1:38 [PATCH net] tg3: Fix RSS ring refill race condition Michael Chan
2012-03-22  1:38 ` [PATCH net] cnic: Fix parity error code conflict Michael Chan
2012-03-22  1:58   ` David Miller
2012-03-22  1:58 ` [PATCH net] tg3: Fix RSS ring refill race condition David Miller

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).