From: Tom Zanussi <tom.zanussi@linux.intel.com>
To: Richard Weinberger <richard.weinberger@gmail.com>
Cc: Andi Kleen <andi@firstfloor.org>,
"netdev@vger.kernel.org" <netdev@vger.kernel.org>,
LKML <linux-kernel@vger.kernel.org>
Subject: Re: RFC: A reduced Linux network stack for small systems
Date: Tue, 06 May 2014 08:34:06 -0500 [thread overview]
Message-ID: <1399383246.8767.99.camel@empanada> (raw)
In-Reply-To: <CAFLxGvww8t-PpuS3B=dFvn8k3xBydk3CgA=qoLyEbkKJZJ0xhQ@mail.gmail.com>
On Tue, 2014-05-06 at 09:25 +0200, Richard Weinberger wrote:
> On Tue, May 6, 2014 at 12:25 AM, Andi Kleen <andi@firstfloor.org> wrote:
> > There has been a lot of interest recently to run Linux on very small systems,
> > like Quark systems. These may have only 2-4MB memory. They are also limited
> > by flash space.
> >
> > One problem on these small system is the size of the network stack.
> > Currently enabling IPv4 costs about 400k in text, which is prohibitive on
> > a 2MB system, and very expensive with 4MB.
> >
> > There were proposals to instead use LWIP in user space. LWIP with
> > its socket interface comes in at a bit over 100k overhead per application.
> >
> > I maintain that the Linux network stack is actually not that bloated,
> > it just has a lot of features :-) The goal of this project was to
> > subset it in a sensible way so that the native kernel stack becomes
> > competitive with LWIP.
> >
> > It turns out that the standard stack has a couple of features that
> > are not really needed on client systems. Luckily it is also
> > relatively well modularized, so it becomes possible to stub
> > out these features at the edge.
> >
> > With removing these features we still have a powerful TCP/IP stack,
> > but one that fits better into small systems.
> >
> > It would have been prohibitive to ifdef every optional feature.
> > This patchkit relies heavily on LTO to effectively remove unused
> > code. This allows to disable features only at the module boundaries,
> > and rely on the compiler to drop unreferenced code and data.
> >
> > A few features have been also reimplemented in a simpler way.
> > And I shrank a number of data structures based on CONFIG_BASE_SMALL.
> >
> > With these changes I can get a fully featured network stack down
> > to about 170k with LTO. Without LTO there are also benefits,
> > but somewhat less.
> >
> > There are essentially three sensible configurations:
> > - Full featured like today.
> > - Client only subset, but still works with standard distribution userland.
> > Remove some obscure features like fastopen, make all tables smaller,
> > packet socket mmap code, use a simpler routing table, remove
> > high speed networking features like RPX, XPS, GRO offload.
> > Disable SNMP, TCP metrics
> > - Minimal subset for deeply embedded systems that can use special userland.
> > Remove rtnetlink (ioctl only), remove ethtool, raw sockets.
> >
> > Right now I'm using own Kconfigs for every removed features. I realize
> > this somewhat increases the compile test matrix. It would be possible
> > to hide some of the options and select them using higher level
> > configurations like the ones listed above. I haven't done this
> > in this version.
> >
> > At this point I'm mainly interested in review and comments.
> >
> > Git trees:
> >
> > git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc net/debloat
> > Main tree
> >
> > git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-misc net/debloat-3.14
> > 3.14 based tree.
> >
> > Thanks to Tom Zanussi for contributions and testing.
>
> What kind of userspace do you use on such a small system?
> It looks like you run kernels without procfs and netlink, so not even
> ps would work. :)
>
The microYocto 'distro' I have running with these net-diet patches
doesn't use a full procfs, but a pared-down version (CONFIG_PROCFS_MIN).
Keeping ps working is of course essential, and it does that (along with
a couple other things like /proc/filesystems and /proc/mounts I needed
to boot):
https://github.com/tzanussi/linux-yocto-micro-3.14/commit/68379432afcfa82ac695d9f02892fcf48ade5ae8
Anyway all the userspace and kernel bits are available for anyone who
wants to build it and try it out:
https://github.com/tzanussi/meta-galileo/blob/daisy/meta-galileo/README
It's very much a work-in-progress with a lot of rough edges, but it is a
fully functional system on real hardware (Galileo board/Quark processor)
with a usable shell (ps too!) and web server running on a kernel with
native networking and ~ 750k text size.
Tom
next prev parent reply other threads:[~2014-05-06 13:34 UTC|newest]
Thread overview: 124+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-05-05 22:25 RFC: A reduced Linux network stack for small systems Andi Kleen
2014-05-05 22:25 ` [PATCH 01/24] net, diet: Reduce netdev name hash table for CONFIG_BASE_SMALL Andi Kleen
2014-05-06 3:03 ` David Miller
2014-05-05 22:25 ` [PATCH 02/24] net, diet: Reduce NAPI " Andi Kleen
2014-05-05 22:25 ` [PATCH 03/24] net, diet: Decrease ip defrag hash tables and max length with BASE_SMALL Andi Kleen
2014-05-05 22:25 ` [PATCH 04/24] net, diet: Make FIB hash tables smaller with CONFIG_BASE_SMALL Andi Kleen
2014-05-06 3:10 ` David Miller
2014-05-05 22:25 ` [PATCH 05/24] net, diet: Default to small TCP hash tables on small kernels Andi Kleen
2014-05-06 3:12 ` David Miller
2014-05-05 22:25 ` [PATCH 06/24] net, diet: Use small UDP " Andi Kleen
2014-05-06 3:11 ` David Miller
2014-05-06 14:26 ` Eric Dumazet
2014-05-06 18:23 ` Andi Kleen
2014-05-06 19:25 ` Eric Dumazet
2014-05-05 22:25 ` [PATCH 07/24] net, diet: Make ping sockets optional Andi Kleen
2014-05-06 3:04 ` David Miller
2014-05-05 22:25 ` [PATCH 08/24] net, diet: Make TCP metrics optional Andi Kleen
2014-05-05 23:18 ` Yuchung Cheng
2014-05-05 23:28 ` Andi Kleen
2014-05-06 3:12 ` David Miller
2014-05-06 3:21 ` Andi Kleen
2014-05-06 3:23 ` David Miller
2014-05-06 4:39 ` Tom Zanussi
2014-05-07 13:59 ` One Thousand Gnomes
2014-05-07 17:20 ` David Miller
2014-05-07 22:19 ` Tim Bird
2014-05-07 22:52 ` David Miller
2014-05-06 15:57 ` josh
2014-05-06 15:59 ` David Miller
2014-05-06 16:41 ` josh
2014-05-06 17:12 ` Rick Jones
2014-05-06 18:09 ` josh
2014-05-06 17:16 ` David Miller
2014-05-06 17:55 ` josh
2014-05-06 18:33 ` Cong Wang
2014-05-06 18:48 ` Andi Kleen
2014-05-06 19:19 ` Richard Cochran
2014-05-06 20:44 ` David Miller
2014-05-06 21:08 ` josh
2014-05-06 21:11 ` David Miller
2014-05-06 22:50 ` josh
2014-05-06 23:27 ` Eric Dumazet
2014-06-14 20:33 ` Pavel Machek
2014-06-23 10:17 ` zhuyj
2014-05-06 19:14 ` Richard Cochran
2014-05-06 19:50 ` Andi Kleen
2014-05-06 20:07 ` Richard Cochran
2014-05-06 21:05 ` Andi Kleen
2014-05-06 23:29 ` Eric Dumazet
2014-05-07 4:33 ` Andi Kleen
2014-05-06 20:46 ` David Miller
2014-05-07 8:39 ` David Laight
2014-05-09 9:48 ` Pavel Machek
2014-05-06 16:39 ` Eric Dumazet
2014-05-06 16:45 ` josh
2014-05-06 17:03 ` Eric Dumazet
2014-05-06 17:30 ` josh
2014-05-06 17:17 ` David Miller
2014-05-06 17:21 ` josh
2014-05-06 17:25 ` David Miller
2014-05-06 20:06 ` Andi Kleen
2014-05-06 20:47 ` David Miller
2014-05-06 20:11 ` josh
2014-05-06 20:44 ` Andi Kleen
2014-05-06 17:14 ` David Miller
2014-05-06 18:32 ` Andi Kleen
2014-05-06 18:58 ` Tom Herbert
2014-05-06 19:37 ` josh
2014-05-06 19:57 ` Andi Kleen
2014-05-06 20:17 ` Eric Dumazet
2014-05-06 20:27 ` josh
2014-05-06 20:37 ` Andi Kleen
2014-05-06 20:48 ` David Miller
2014-05-07 13:35 ` One Thousand Gnomes
2014-05-13 16:22 ` Christer Weinigel
2014-05-09 7:38 ` Pavel Machek
2014-05-05 22:25 ` [PATCH 09/24] net, diet: Make ethtool optional Andi Kleen
2014-05-06 3:11 ` David Miller
2014-05-06 3:14 ` Andi Kleen
2014-05-06 3:22 ` David Miller
2014-05-05 22:25 ` [PATCH 10/24] net, diet: Make LPF filter optional Andi Kleen
2014-05-06 3:10 ` David Miller
2014-05-05 22:26 ` [PATCH 11/24] net, diet: Move rtnl_lock to separate file Andi Kleen
2014-05-05 22:26 ` [PATCH 12/24] net, diet: Make rtnetlink optional Andi Kleen
2014-05-06 3:08 ` David Miller
2014-05-06 3:11 ` Andi Kleen
2014-05-05 22:26 ` [PATCH 13/24] net, diet: Make GRO offload optional Andi Kleen
2014-05-06 3:01 ` David Miller
2014-05-06 3:03 ` Andi Kleen
2014-05-05 22:26 ` [PATCH 14/24] net, diet: Make MIB statistics collections depend on PROC_FS Andi Kleen
2014-05-06 3:05 ` David Miller
2014-05-05 22:26 ` [PATCH 15/24] net, diet: Make igmp and mcast ioctls depend on IP_MULTICAST Andi Kleen
2014-05-05 22:26 ` [PATCH 16/24] net, diet: Make TCP fastopen optional Andi Kleen
2014-05-06 3:06 ` David Miller
2014-05-05 22:26 ` [PATCH 17/24] net, diet: Make internal control sockets use UDP Andi Kleen
2014-05-05 22:26 ` [PATCH 18/24] net, diet: Make raw sockets optional Andi Kleen
2014-05-06 3:12 ` David Miller
2014-05-05 22:26 ` [PATCH 19/24] net, diet: Make RPS configurable Andi Kleen
2014-05-06 3:14 ` David Miller
2014-05-06 3:16 ` Andi Kleen
2014-05-06 8:32 ` Bjørn Mork
2014-05-05 22:26 ` [PATCH 20/24] net, diet: Make XPS configurable Andi Kleen
2014-05-05 22:26 ` [PATCH 21/24] net, diet: Make packet tpacket/mmap/fanout/rings optional Andi Kleen
2014-05-06 3:09 ` David Miller
2014-05-05 22:26 ` [PATCH 22/24] net, diet: Support simpler routing table Andi Kleen
2014-05-06 3:02 ` David Miller
2014-05-06 3:08 ` Andi Kleen
2014-05-06 3:21 ` David Miller
2014-05-05 22:26 ` [PATCH 23/24] net, diet: Add coccinelle script to convert drivers to ETHTOOL_OPS Andi Kleen
2014-05-06 9:27 ` Nicolas Palix
2014-05-06 15:05 ` David Miller
2014-05-07 0:18 ` Wilfried Klaebe
2014-05-05 22:26 ` [PATCH 24/24] net, diet: Convert all drivers to use SET_ETHTOOL_OPS Andi Kleen
2014-05-06 7:25 ` RFC: A reduced Linux network stack for small systems Richard Weinberger
2014-05-06 13:34 ` Tom Zanussi [this message]
2014-05-06 15:20 ` Alexei Starovoitov
2014-05-06 15:34 ` Tom Zanussi
2014-05-06 17:20 ` Alexei Starovoitov
2014-05-06 20:00 ` Andi Kleen
2014-05-06 21:05 ` Alexei Starovoitov
2014-05-09 7:39 ` Pavel Machek
2014-05-09 7:38 ` Pavel Machek
2014-05-06 9:18 ` David Laight
2014-05-07 13:20 ` One Thousand Gnomes
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1399383246.8767.99.camel@empanada \
--to=tom.zanussi@linux.intel.com \
--cc=andi@firstfloor.org \
--cc=linux-kernel@vger.kernel.org \
--cc=netdev@vger.kernel.org \
--cc=richard.weinberger@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).