netdev - [PATCH net-next 09/14] bnxt_en: reduce timeout on initial HWRM calls

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 25 Apr 2018 20:40:52 -0400
From:   Michael Chan <michael.chan@...adcom.com>
To:     davem@...emloft.net
Cc:     netdev@...r.kernel.org, Andy Gospodarek <gospo@...adcom.com>
Subject: [PATCH net-next 09/14] bnxt_en: reduce timeout on initial HWRM calls

From: Andy Gospodarek <gospo@...adcom.com>

Testing with DIM enabled on older kernels indicated that firmware calls
were slower than expected.  More detailed analysis indicated that the
default 25us delay was higher than necessary.  Reducing the time spend in
usleep_range() for the first several calls would reduce the overall
latency of firmware calls on newer Intel processors.

Signed-off-by: Andy Gospodarek <gospo@...adcom.com>
Signed-off-by: Michael Chan <michael.chan@...adcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 26 +++++++++++++++++++++++---
 drivers/net/ethernet/broadcom/bnxt/bnxt.h |  6 ++++++
 2 files changed, 29 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index a221a10..ff9a5cd 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -3495,15 +3495,29 @@ static int bnxt_hwrm_do_send_msg(struct bnxt *bp, void *msg, u32 msg_len,
 
 	if (!timeout)
 		timeout = DFLT_HWRM_CMD_TIMEOUT;
+	/* convert timeout to usec */
+	timeout *= 1000;
 
 	i = 0;
-	tmo_count = timeout * 40;
+	/* Short timeout for the first few iterations:
+	 * number of loops = number of loops for short timeout +
+	 * number of loops for standard timeout.
+	 */
+	tmo_count = HWRM_SHORT_TIMEOUT_COUNTER;
+	timeout = timeout - HWRM_SHORT_MIN_TIMEOUT * HWRM_SHORT_TIMEOUT_COUNTER;
+	tmo_count += DIV_ROUND_UP(timeout, HWRM_MIN_TIMEOUT);
 	resp_len = bp->hwrm_cmd_resp_addr + HWRM_RESP_LEN_OFFSET;
 	if (intr_process) {
 		/* Wait until hwrm response cmpl interrupt is processed */
 		while (bp->hwrm_intr_seq_id != HWRM_SEQ_ID_INVALID &&
 		       i++ < tmo_count) {
-			usleep_range(25, 40);
+			/* on first few passes, just barely sleep */
+			if (i < HWRM_SHORT_TIMEOUT_COUNTER)
+				usleep_range(HWRM_SHORT_MIN_TIMEOUT,
+					     HWRM_SHORT_MAX_TIMEOUT);
+			else
+				usleep_range(HWRM_MIN_TIMEOUT,
+					     HWRM_MAX_TIMEOUT);
 		}
 
 		if (bp->hwrm_intr_seq_id != HWRM_SEQ_ID_INVALID) {
@@ -3521,7 +3535,13 @@ static int bnxt_hwrm_do_send_msg(struct bnxt *bp, void *msg, u32 msg_len,
 			      HWRM_RESP_LEN_SFT;
 			if (len)
 				break;
-			usleep_range(25, 40);
+			/* on first few passes, just barely sleep */
+			if (i < DFLT_HWRM_CMD_TIMEOUT)
+				usleep_range(HWRM_SHORT_MIN_TIMEOUT,
+					     HWRM_SHORT_MAX_TIMEOUT);
+			else
+				usleep_range(HWRM_MIN_TIMEOUT,
+					     HWRM_MAX_TIMEOUT);
 		}
 
 		if (i >= tmo_count) {
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index 057f8a2..7fa4a45 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -532,6 +532,12 @@ struct rx_tpa_end_cmp_ext {
 #define BNXT_HWRM_REQ_MAX_SIZE		128
 #define BNXT_HWRM_REQS_PER_PAGE		(BNXT_PAGE_SIZE /	\
 					 BNXT_HWRM_REQ_MAX_SIZE)
+#define HWRM_SHORT_MIN_TIMEOUT		3
+#define HWRM_SHORT_MAX_TIMEOUT		10
+#define HWRM_SHORT_TIMEOUT_COUNTER	5
+
+#define HWRM_MIN_TIMEOUT		25
+#define HWRM_MAX_TIMEOUT		40
 
 #define BNXT_RX_EVENT	1
 #define BNXT_AGG_EVENT	2
-- 
1.8.3.1