[Intel-wired-lan] [PATCH S51 08/13] ice: port fix for chk_linearlize
Tony Nguyen
anthony.l.nguyen at intel.com
Thu Jul 30 00:19:17 UTC 2020
From: Kiran Patil <kiran.patil at intel.com>
This is a port of commit 248de22e638f ("i40e/i40evf: Account for frags
split over multiple descriptors in check linearize")
As part of testing workloads (read/write) using larger IO size (128K)
tx_timeout is observed and whenever it happens, it was due to
tx_linearize.
Signed-off-by: Kiran Patil <kiran.patil at intel.com>
---
drivers/net/ethernet/intel/ice/ice_txrx.c | 26 ++++++++++++++++++++---
1 file changed, 23 insertions(+), 3 deletions(-)
diff --git a/drivers/net/ethernet/intel/ice/ice_txrx.c b/drivers/net/ethernet/intel/ice/ice_txrx.c
index 53c67eeec2fa..77de8869e7ca 100644
--- a/drivers/net/ethernet/intel/ice/ice_txrx.c
+++ b/drivers/net/ethernet/intel/ice/ice_txrx.c
@@ -2292,10 +2292,30 @@ static bool __ice_chk_linearize(struct sk_buff *skb)
/* Walk through fragments adding latest fragment, testing it, and
* then removing stale fragments from the sum.
*/
- stale = &skb_shinfo(skb)->frags[0];
- for (;;) {
+ for (stale = &skb_shinfo(skb)->frags[0];; stale++) {
+ int stale_size = skb_frag_size(stale);
+
sum += skb_frag_size(frag++);
+ /* The stale fragment may present us with a smaller
+ * descriptor than the actual fragment size. To account
+ * for that we need to remove all the data on the front and
+ * figure out what the remainder would be in the last
+ * descriptor associated with the fragment.
+ */
+ if (stale_size > ICE_MAX_DATA_PER_TXD) {
+ int align_pad = -(skb_frag_off(stale)) &
+ (ICE_MAX_READ_REQ_SIZE - 1);
+
+ sum -= align_pad;
+ stale_size -= align_pad;
+
+ do {
+ sum -= ICE_MAX_DATA_PER_TXD_ALIGNED;
+ stale_size -= ICE_MAX_DATA_PER_TXD_ALIGNED;
+ } while (stale_size > ICE_MAX_DATA_PER_TXD);
+ }
+
/* if sum is negative we failed to make sufficient progress */
if (sum < 0)
return true;
@@ -2303,7 +2323,7 @@ static bool __ice_chk_linearize(struct sk_buff *skb)
if (!nr_frags--)
break;
- sum -= skb_frag_size(stale++);
+ sum -= stale_size;
}
return false;
--
2.20.1
More information about the Intel-wired-lan
mailing list