fedora 14 kernel performance with ip forwarding workload
by Jesse Brandeburg
The other day I was running the stock fedora kernel on my ip
forwarding setup, to see what the performance was, and the performance
wasn't very good.
system is S5520HC dual socket 2.93GHz Xeon 5570 (Nehalem) with 3 quad
port 82580 adapters (12 ports). Traffic is bidirectional 64 byte
packets being forwarded and received on each port, basically port to
port routing. I am only using 12 flows currently.
The driver is igb, and I am using an affinity script that lines up
each pair of ports that are forwarding traffic into optimal
configurations for cache locality. I am also disabling
remote_node_defrag_ratio to stop cross node traffic.
With the fedora default kernel from F14 it appears that
CONFIG_NETFILTER=y means that I cannot unload all of netfilter even if
I stop iptables service.
perf showed netfilter being prominent, and removing it gives me much
higher throughput. Is there a reason CONFIG_NETFILTER=y ? Isn't it a
good thing to be able to disable netfilter if you want to?
Jesse
8 years, 9 months
Re: [kernel] Disable debugging options.
by Thorsten Leemhuis
On 18.06.2012 15:09, Josh Boyer wrote:
> commit 3b37beedf48825354c42e25f7001677320958d38
> Author: Josh Boyer <jwboyer(a)redhat.com>
> Date: Mon Jun 18 09:09:58 2012 -0400
>
> Disable debugging options.
>
> config-generic | 8 ++--
> config-nodebug | 112 ++++++++++++++++++++++++++--------------------------
> config-x86-generic | 2 +-
> kernel.spec | 9 +++-
> 4 files changed, 67 insertions(+), 64 deletions(-)
> ---
> diff --git a/config-generic b/config-generic
> index 6b1c651..08c28ba 100644
> --- a/config-generic
> +++ b/config-generic
> @@ -1446,13 +1446,13 @@ CONFIG_B43_SDIO=y
> CONFIG_B43_BCMA=y
> # CONFIG_B43_BCMA_EXTRA is not set
> CONFIG_B43_BCMA_PIO=y
> -CONFIG_B43_DEBUG=y
> +# CONFIG_B43_DEBUG is not set
> [...]
Just out of curiosity: Why is enabling and disabling the debug options
done via "make {no,}debug" in the git checkout and not by an conditional
within the spec file? That afaics makes it harder to switch the debug
options off if you only have the SRPM at hand.
Or am I missing something obvious here?
CU
knurd
11 years, 4 months
CONFIG_USB_UAS
by Gerd Hoffmann
Hi,
Any specific reason why CONFIG_USB_UAS is not enabled?
If not, can it be enabled please (rawhide & f17)?
thanks,
Gerd
11 years, 7 months
Kernel rebases
by Josh Boyer
Howdy,
Just a quick update on kernel rebases.
Justin rebased Rawhide to the 3.6 merge window git tree last week and
I'm continuing on with that until the merge window closes. Then we'll
pick up the RC kernels throughout, as usual. If there is a feature or
option you have been waiting for in 3.6, now is the time to speak up
and make sure it's set as you want.
F17 was rebased to the 3.5 kernel last week as well. The initial builds
are done and submitted to updates-testing today. F16 will likely follow
suit after 3.5.1 is out.
For both Rawhide and F17, the ARM configs could use some serious looking
at. We continue to have a lot of config options pop up when we do
stable rebases. Getting them set in Rawhide during the merge window
kernels is going to be the best bet to ensure things are working as
expected.
If you have questions/comments, feel free to let us know.
josh
11 years, 7 months
make modules_install error
by ivan ouyang
Hi, There
After I run: make modules make bzImages successfully, I tried to make
modules_install but could not go any further.
# make modules_install
rm:invalid option -- '3'
Try 'rm -- help' for more information
make : *** [_modinst_] Error 1
Could anyone please help me with the situation?
I just add something to the wireless-testing and compile it. The version of the
linux is 3.4.4-3.fc17.x86_64 #1
Thank you so much.
--
Best Regards
Xun Ouyang
11 years, 8 months
[rawhide PATCH] config: enable NVMe driver
by Kyle McMartin
Seems to build on i386 now, so it should probably be built as
a module.
--- a/config-generic
+++ b/config-generic
@@ -248,7 +248,7 @@ CONFIG_BLK_DEV_LOOP=y
CONFIG_BLK_DEV_LOOP_MIN_COUNT=8
CONFIG_BLK_DEV_CRYPTOLOOP=m
CONFIG_BLK_DEV_NBD=m
-# CONFIG_BLK_DEV_NVME is not set
+CONFIG_BLK_DEV_NVME=m
CONFIG_BLK_DEV_OSD=m
CONFIG_BLK_DEV_RAM=m
CONFIG_BLK_DEV_RAM_COUNT=16
11 years, 8 months
Configuring and building the kernel using clang
by Darryl L. Pierce
I'm trying to get up to speed with Linux kernel development and have a
few questions:
1. According to the book I'm reading (_Linux Kernel Development_ by
Robert Love) the current kernel's configuration can be found at
/proc/config.gz. But on Fedora, that's not the case, and it's in /boot/
as config-{VERSION}. How do I use this in the same way that I would have
used /proc/config.gz?
2. On the kernel-janitors mailing list I asked how to get started, and
was recommended to use clang to build the kernel and submit patches to
fix errors reported. Has anybody successfully built the kernel usin
clang packaged for Fedora? If so, how? I tried:
make cc=clang
and the build immediately failed on inline assembler code. I'd
appreciate a recommendation on how to approach building that way.
Thanks in advance.
--
Darryl L. Pierce <mcpierce(a)gmail.com>
http://mcpierce.multiply.com/
"What do you care what people think, Mr. Feynman?"
11 years, 8 months
Could we include this in our kernel? vfs: make O_PATH file descriptors usable for fchdir()
by Michal Hlavinka
Hi,
I was asked if we can add this patch to our kernel in Fedora.
It's for ability to open a directory with execute permission but without
read permission.
Michal
-------- Original Message --------
Michal, can you ask the Fedora Linux kernel people to backport this
patch, please (this is going to be needed by both ksh93 and KDE4)?
Josh
---------- Forwarded message ----------
From: Greg KH <gregkh(a)linuxfoundation.org>
Date: Thu, Jul 12, 2012 at 7:51 PM
Subject: [ 42/68] vfs: make O_PATH file descriptors usable for fchdir()
To: linux-kernel(a)vger.kernel.org, stable(a)vger.kernel.org
Cc: torvalds(a)linux-foundation.org, akpm(a)linux-foundation.org,
alan(a)lxorguk.ukuu.org.uk, Al Viro <viro(a)zeniv.linux.org.uk>
3.0-stable review patch. If anyone has any objections, please let me know.
------------------
From: Linus Torvalds <torvalds(a)linux-foundation.org>
commit 332a2e1244bd08b9e3ecd378028513396a004a24 upstream.
We already use them for openat() and friends, but fchdir() also wants to
be able to use O_PATH file descriptors. This should make it comparable
to the O_SEARCH of Solaris. In particular, O_PATH allows you to access
(not-quite-open) a directory you don't have read persmission to, only
execute permission.
Noticed during development of multithread support for ksh93.
Reported-by: ольга крыжановская <olga.kryzhanovska(a)gmail.com>
Cc: Al Viro <viro(a)zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds(a)linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh(a)linuxfoundation.org>
---
fs/open.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
--- a/fs/open.c
+++ b/fs/open.c
@@ -396,10 +396,10 @@ SYSCALL_DEFINE1(fchdir, unsigned int, fd
{
struct file *file;
struct inode *inode;
- int error;
+ int error, fput_needed;
error = -EBADF;
- file = fget(fd);
+ file = fget_raw_light(fd, &fput_needed);
if (!file)
goto out;
@@ -413,7 +413,7 @@ SYSCALL_DEFINE1(fchdir, unsigned int, fd
if (!error)
set_fs_pwd(current->fs, &file->f_path);
out_putf:
- fput(file);
+ fput_light(file, fput_needed);
out:
return error;
}
Josh
11 years, 8 months
Caution if 3.5 gets brought to F17
by Bruno Wolff III
Under the 3.5 kernel I need to have the hid_generic module loaded for my
USB keyboards to work. I didn't with the 3.4 kernels. This caused a problem
with being able to be able to enter the password for my encrypted rootfs.
dracut master now has hid_generic in the list of modules it needs to include.
This change will be in the next dracut release for rawhide. However, if 3.5
kernels do get used for F17, it would be good to get the change backported
to dracut used in F17 before the new kernel shows up.
11 years, 8 months
NX cannot be enabled: non-PAE kernel
by Michael Zintakis
I have compiled a PAE kernel and installed it on my system (it *is* a
PAE kernel, I am sure of it :-) ), but when I boot up I see the
following message: "Notice: NX (Execute Disable) protection cannot be
enabled: non-PAE kernel!"
I am certain I am using the right kernel and cannot explain why this
error occurs. Could it be that the BIOS may have the NX-bit disabled (if
so, how do I find out?) or is it something I am definitely doing wrong?
11 years, 8 months