Project

General

Profile

Actions

Feature #9345

closed

look into using MSG_MORE and MSG_SENDPAGE_NOTLAST more

Added by Ilya Dryomov over 9 years ago. Updated about 5 years ago.

Status:
Resolved
Priority:
Normal
Assignee:
Category:
libceph
Target version:
-
% Done:

0%

Source:
other
Tags:
Backport:
Reviewed:
Affected Versions:

Description

Mike Christie pointed out the following:

In the current kernel we are doing tcp auto corking:

https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=f54b311142a92ea2e42598e347b84e1655caf8e3

(here is more info on corking
https://access.redhat.com/documentation/en-US/Red_Hat_Enterprise_MRG/2/html/Realtime_Reference_Guide/chap-Realtime_Reference_Guide-Sockets.html)

This can lead to similar perf issues if you are not using the MSG_MORE
and MSG_SENDPAGE_NOTLAST. I see the ceph kernel code is using the former
but not the latter.

Related issues 1 (0 open1 closed)

Related to Linux kernel client - Feature #9344: add a TCP_NODELAY optionResolved09/04/2014

Actions
Actions #1

Updated by Chaitanya Huilgol over 9 years ago

If I understand correctly, MSG_SENDPAGE_NOLAST is an internal flag and is set by pipe_to_sendpage. Should user be setting this?

Actions

Also available in: Atom PDF