Skip to contentSkip to conversation

remove napi in the tx path

PATCH In review Ajit Khaparde 2 participants 4 messages May 19, 2009

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 remove napi in the tx path

David, I have spun this patch again and against the net-2.6 tree. This patch is based on a certain feedback that we got and hence it is equally urgent as the previous patch for this to get in the net-2.6 tree.
Please consider. I hope this applies cleanly!

Thanks
Ajit

---
 drivers/net/benet/be.h      |    3 +-
 drivers/net/benet/be_main.c |  132 ++++++++++++++++++++++++-------------------
 2 files changed, 74 insertions(+), 61 deletions(-)
drivers/net/benet/be.h +1 -2
index b4bb06f..6d149a4 100644
@@ -159,8 +159,6 @@ struct be_eq_obj {
 	u16 min_eqd;		/* in usecs */
 	u16 max_eqd;		/* in usecs */
 	u16 cur_eqd;		/* in usecs */
-
-	struct napi_struct napi;
 };
 
 struct be_tx_obj {
@@ -179,6 +177,7 @@ struct be_rx_page_info {
 };
 
 struct be_rx_obj {
+	struct napi_struct napi;
 	struct be_queue_info q;
 	struct be_queue_info cq;
 	struct be_rx_page_info page_info_tbl[RX_Q_LEN];
drivers/net/benet/be_main.c +73 -59

215 lines not on this page · raw message

Signature · 1 line
1.5.4.3
maintainer
#1 lore raw

From: Ajit Khaparde <[email protected]>
Date: Tue, 19 May 2009 07:06:20 +0530

Quoted 5 lines

David, I have spun this patch again and against the net-2.6 tree. This patch is based on a certain feedback that we got and hence it is equally urgent as the previous patch for this to get in the net-2.6 tree.
Please consider. I hope this applies cleanly!

This patch does not meet 2.6.30 criteria this late in the release process. I will only apply it to net-next-2.6

maintainer
#2 lore raw
In reply to Ajit Khaparde

Actually, I'm not applying this patch at all.

It's garbage.

First of all, your commit message is way too terse. Why did you do these things, what are the reasons? You don't explain anything. If it's for performance reasons, you give no information about that.

Second of all, the change does two things. One of which you don't even mention in your subject line, all of this LRO stuff.

Third of all, module parameters ARE NOT THE WAY TO CONTROL LRO! We have an ethtool command to turn it on and off, via the ethtool_ops->{set,get}_flags() methods. So this part of your change is totally unacceptable.

Your patch submissions, I have to say, are horrible. Your patches don't apply cleanly to the net trees, your commit messages are one-liners at best, and your patches do multiple unrelated things at the same time.

#3 lore raw

David,

Sorry for clubbing the two unrelated items in one patch.
The patch apply failed because the previous patch has not been applied. The module parameter for lro was being added on a request from a distro. I will drop this change and submit another patch for doing TX completion in interrupt context instead of through NAPI.

Thanks
Ajit

Quoted 26 lines · David Miller

Actually, I'm not applying this patch at all.

It's garbage.

First of all, your commit message is way too terse. Why did you do these things, what are the reasons? You don't explain anything. If it's for performance reasons, you give no information about that.

Second of all, the change does two things. One of which you don't even mention in your subject line, all of this LRO stuff.

Third of all, module parameters ARE NOT THE WAY TO CONTROL LRO! We have an ethtool command to turn it on and off, via the ethtool_ops->{set,get}_flags() methods. So this part of your change is totally unacceptable.

Your patch submissions, I have to say, are horrible. Your patches don't apply cleanly to the net trees, your commit messages are one-liners at best, and your patches do multiple unrelated things at the same time.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html