From: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
To: jeff@garzik.org
Cc: netdev@vger.kernel.org, davem@davemloft.net,
Jesse Brandeburg <jesse.brandeburg@intel.com>,
Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Subject: [NET-NEXT PATCH 03/25] ixgb: fix bug in descriptor ring due to prefetch corruption
Date: Tue, 08 Jul 2008 15:51:17 -0700 [thread overview]
Message-ID: <20080708225117.6925.15563.stgit@localhost.localdomain> (raw)
In-Reply-To: <20080708224858.6925.29725.stgit@localhost.localdomain>
From: Jesse Brandeburg <jesse.brandeburg@intel.com>
there was one more bug hidden in the prefetch routines in ixgb hardware
that force us to remove it completely. Writebacks were being done on
descriptors with stale data due to internal hardware fifo corruption.
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
drivers/net/ixgb/ixgb_main.c | 27 ++++++---------------------
1 files changed, 6 insertions(+), 21 deletions(-)
diff --git a/drivers/net/ixgb/ixgb_main.c b/drivers/net/ixgb/ixgb_main.c
index 652c0ea..ecd5252 100644
--- a/drivers/net/ixgb/ixgb_main.c
+++ b/drivers/net/ixgb/ixgb_main.c
@@ -146,14 +146,6 @@ static int debug = DEFAULT_DEBUG_LEVEL_SHIFT;
module_param(debug, int, 0);
MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
-/* some defines for controlling descriptor fetches in h/w */
-#define RXDCTL_WTHRESH_DEFAULT 15 /* chip writes back at this many or RXT0 */
-#define RXDCTL_PTHRESH_DEFAULT 0 /* chip considers prefech below
- * this */
-#define RXDCTL_HTHRESH_DEFAULT 0 /* chip will only prefetch if tail
- * is pushed this many descriptors
- * from head */
-
/**
* ixgb_init_module - Driver Registration Routine
*
@@ -839,7 +831,6 @@ ixgb_configure_rx(struct ixgb_adapter *adapter)
struct ixgb_hw *hw = &adapter->hw;
u32 rctl;
u32 rxcsum;
- u32 rxdctl;
/* make sure receives are disabled while setting up the descriptors */
@@ -861,18 +852,12 @@ ixgb_configure_rx(struct ixgb_adapter *adapter)
IXGB_WRITE_REG(hw, RDH, 0);
IXGB_WRITE_REG(hw, RDT, 0);
- /* set up pre-fetching of receive buffers so we get some before we
- * run out (default hardware behavior is to run out before fetching
- * more). This sets up to fetch if HTHRESH rx descriptors are avail
- * and the descriptors in hw cache are below PTHRESH. This avoids
- * the hardware behavior of fetching <=512 descriptors in a single
- * burst that pre-empts all other activity, usually causing fifo
- * overflows. */
- /* use WTHRESH to burst write 16 descriptors or burst when RXT0 */
- rxdctl = RXDCTL_WTHRESH_DEFAULT << IXGB_RXDCTL_WTHRESH_SHIFT |
- RXDCTL_HTHRESH_DEFAULT << IXGB_RXDCTL_HTHRESH_SHIFT |
- RXDCTL_PTHRESH_DEFAULT << IXGB_RXDCTL_PTHRESH_SHIFT;
- IXGB_WRITE_REG(hw, RXDCTL, rxdctl);
+ /* due to the hardware errata with RXDCTL, we are unable to use any of
+ * the performance enhancing features of it without causing other
+ * subtle bugs, some of the bugs could include receive length
+ * corruption at high data rates (WTHRESH > 0) and/or receive
+ * descriptor ring irregularites (particularly in hardware cache) */
+ IXGB_WRITE_REG(hw, RXDCTL, 0);
/* Enable Receive Checksum Offload for TCP and UDP */
if (adapter->rx_csum) {
next prev parent reply other threads:[~2008-07-08 22:51 UTC|newest]
Thread overview: 27+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-07-08 22:51 [NET-NEXT PATCH 00/25] ixgb: update to latest Jeff Kirsher
2008-07-08 22:51 ` [NET-NEXT PATCH 01/25] ixgb: maybe stop tx port missed a piece Jeff Kirsher
2008-07-11 5:21 ` Jeff Garzik
2008-07-08 22:51 ` [NET-NEXT PATCH 02/25] ixgb: repeat 32 bit ioremap cleanup Jeff Kirsher
2008-07-08 22:51 ` Jeff Kirsher [this message]
2008-07-08 22:51 ` [NET-NEXT PATCH 04/25] ixgb: leave room for extra hardware memory usage Jeff Kirsher
2008-07-08 22:51 ` [NET-NEXT PATCH 05/25] ixgb: check down state before enable irq Jeff Kirsher
2008-07-08 22:51 ` [NET-NEXT PATCH 06/25] ixgb: don't allow too small MTU Jeff Kirsher
2008-07-08 22:51 ` [NET-NEXT PATCH 07/25] ixgb: move time stamp set before setting dma pointer Jeff Kirsher
2008-07-08 22:51 ` [NET-NEXT PATCH 08/25] ixgb: fix race on rx_buffer_len in mtu change Jeff Kirsher
2008-07-08 22:51 ` [NET-NEXT PATCH 09/25] ixgb: fix unload race with timers Jeff Kirsher
2008-07-08 22:51 ` [NET-NEXT PATCH 10/25] ixgb: remove lltx support and update tx routine Jeff Kirsher
2008-07-08 22:51 ` [NET-NEXT PATCH 11/25] ixgb: update readme text Jeff Kirsher
2008-07-08 22:52 ` [NET-NEXT PATCH 12/25] ixgb: add copybreak parameter Jeff Kirsher
2008-07-08 22:52 ` [NET-NEXT PATCH 13/25] ixgb: clean up un-necessary declarations Jeff Kirsher
2008-07-08 22:52 ` [NET-NEXT PATCH 14/25] ixgb: format all if( to be if ( Jeff Kirsher
2008-07-08 22:52 ` [NET-NEXT PATCH 15/25] ixgb: cleanup space after while Jeff Kirsher
2008-07-08 22:52 ` [NET-NEXT PATCH 16/25] ixgb: whitespace fixups Jeff Kirsher
2008-07-08 22:52 ` [NET-NEXT PATCH 17/25] ixgb: fix spelling errors Jeff Kirsher
2008-07-08 22:52 ` [NET-NEXT PATCH 18/25] ixgb: trivial fix space after for Jeff Kirsher
2008-07-08 22:52 ` [NET-NEXT PATCH 19/25] ixgb: cleanup checkpatch suggestions that are relevant Jeff Kirsher
2008-07-08 22:52 ` [NET-NEXT PATCH 20/25] ixgb: rx cleanup performance improvements Jeff Kirsher
2008-07-08 22:52 ` [NET-NEXT PATCH 21/25] ixgb: clean up assignments inside if statements Jeff Kirsher
2008-07-08 22:52 ` [NET-NEXT PATCH 22/25] ixgb: audit use of dev_kfree_skb_any Jeff Kirsher
2008-07-08 22:52 ` [NET-NEXT PATCH 23/25] ixgb: cleanup header Jeff Kirsher
2008-07-08 22:53 ` [NET-NEXT PATCH 24/25] ixgb: make NAPI the only option and the default Jeff Kirsher
2008-07-08 22:53 ` [NET-NEXT PATCH 25/25] ixgb: update copyright dates and versions Jeff Kirsher
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20080708225117.6925.15563.stgit@localhost.localdomain \
--to=jeffrey.t.kirsher@intel.com \
--cc=davem@davemloft.net \
--cc=jeff@garzik.org \
--cc=jesse.brandeburg@intel.com \
--cc=netdev@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).