Bug 120262 - slow outbound performance of the r8169 driver
Summary: slow outbound performance of the r8169 driver
Alias: None
Product: Red Hat Enterprise Linux 3
Classification: Red Hat
Component: kernel   
(Show other bugs)
Version: 3.0
Hardware: All
OS: Linux
Target Milestone: ---
Assignee: John W. Linville
QA Contact: Brian Brock
Depends On:
Blocks: 118637 123574
TreeView+ depends on / blocked
Reported: 2004-04-07 13:38 UTC by Bastien Nocera
Modified: 2007-11-30 22:07 UTC (History)
3 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Last Closed: 2004-12-20 20:55:01 UTC
Type: ---
Regression: ---
Mount Type: ---
Documentation: ---
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---

Attachments (Terms of Use)
patch for TX IRQ handling (1.41 KB, patch)
2004-04-08 09:55 UTC, Bastien Nocera
no flags Details | Diff
r8169-1.4-patch (1.64 KB, text/plain)
2004-05-14 16:08 UTC, Bastien Nocera
no flags Details
r8169-backport.patch (40.04 KB, patch)
2004-06-16 19:27 UTC, John W. Linville
no flags Details | Diff

External Trackers
Tracker ID Priority Status Summary Last Updated
Red Hat Product Errata RHBA-2004:550 normal SHIPPED_LIVE Updated kernel packages available for Red Hat Enterprise Linux 3 Update 4 2004-12-20 05:00:00 UTC

Description Bastien Nocera 2004-04-07 13:38:27 UTC
The problem is similar to:

With /home/users/ being an NFS share:
# time dd if=/dev/zero of=/home/users/tmp/testfile_wr bs=16k count=16384
16384+0 enregistrements lus.
16384+0 enregistrements ecrits.

real    2m0.944s
user    0m0.010s
sys     0m0.660s

# time dd if=/home/users/tmp/testfile_rd of=/dev/null bs=16k
16384+0 enregistrements lus.
16384+0 enregistrements ecrits.

real    0m26.131s
user    0m0.020s
sys     0m0.300s

Using another network card (Realtek 8139) shows decent outbound speed
(around 36 seconds with the 8139 card).
The machine is on a 10/100 network, with the cards set at 100Mbps.

This is with kernel 2.4.21-9.EL and 2.4.21-9.0.1.EL

Comment 1 Bastien Nocera 2004-04-08 09:55:54 UTC
Created attachment 99232 [details]
patch for TX IRQ handling

Patch provided by the customer, Olivier Lepape <ol@m2000.fr>

Comment 2 Don Howard 2004-04-28 22:24:21 UTC
I can confirm that the above patch builds against the latest RHEL 
devel kernel 2.4.21-14.   

Comment 3 Don Howard 2004-04-28 22:42:53 UTC
Also, the customer's patch is is very similar to what is in current 
upstream 2.4. 

Comment 5 Bastien Nocera 2004-05-14 16:08:46 UTC
Created attachment 100228 [details]

Updated patch, fixes a potential deadlock.

Comment 8 John W. Linville 2004-06-16 19:27:57 UTC
Created attachment 101195 [details]

Backport from upstream 2.4.27-pre5 kernel...

Comment 12 John W. Linville 2004-06-25 14:29:41 UTC
Testing w/ backported driver shows greatly improved results here...

[dhcp59-161]:> time dd if=/dev/zero of=testfile_wr bs=16k count=16386
16386+0 records in
16386+0 records out

real    0m23.320s
user    0m0.030s
sys     0m0.450s

[dhcp59-161]:> time dd if=testfile_rd of=/dev/null bs=16k
16386+0 records in
16386+0 records out

real    0m9.217s
user    0m0.020s
sys     0m0.100s

Comment 16 Ernie Petrides 2004-08-31 04:11:18 UTC
A fix for this problem has just been committed to the RHEL3 U4
patch pool this evening (in kernel version 2.4.21-20.1.EL).

Comment 17 John Flanagan 2004-12-20 20:55:01 UTC
An errata has been issued which should help the problem 
described in this bug report. This report is therefore being 
closed with a resolution of ERRATA. For more information
on the solution and/or where to find the updated files, 
please follow the link below. You may reopen this bug report 
if the solution does not work for you.


Note You need to log in before you can comment on or make changes to this bug.