Re: [PATCH] patch-o-matic fixes

Re: [PATCH] patch-o-matic fixes

[Harald Welte]

[-- Attachment #1: Type: text/plain, Size: 1133 bytes --]

On Wed, Apr 30, 2003 at 10:53:32AM +0200, Willy Tarreau wrote:
> Hi Harald,
> 
> here are 2 patches so slightly modify 2 patches in the patch-o-matic :
> 
> 1) u32 classifier applies at a wrong level because it contains the full path,
>    which ends up creating a usr directory in the kernel tree

fixed in cvs now.

> 2) extra/tcp_window_tracking doesn't apply to 2.4.21-rc1 anymore because
>    of a small context change.

I don't want to change this before we have a cumulative 2.4.21.patch.
This is the only way in order to have patch-o-matic compatible with
old and new kernel versions.

Martin, can you please make a cumulative 2.4.21-rc1.patch, including all
netfilter changes that have been included up to rc1? 

> Cheers,
> Willy
-- 
- Harald Welte <laforge@netfilter.org>             http://www.netfilter.org/
============================================================================
  "Fragmentation is like classful addressing -- an interesting early
   architectural error that shows how much experimentation was going
   on while IP was being designed."                    -- Paul Vixie

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: [PATCH] patch-o-matic fixes

[Martin Josefsson]

On Wed, 2003-04-30 at 17:32, Harald Welte wrote:

> > 2) extra/tcp_window_tracking doesn't apply to 2.4.21-rc1 anymore because
> >    of a small context change.
> 
> I don't want to change this before we have a cumulative 2.4.21.patch.
> This is the only way in order to have patch-o-matic compatible with
> old and new kernel versions.
> 
> Martin, can you please make a cumulative 2.4.21-rc1.patch, including all
> netfilter changes that have been included up to rc1? 

It does apply IF you've applied pending/24_conntrack-nosysctl.patch
Which should have been done. We've had implicit dependencies on
submitted and pending stuff many times.

No need to make a cumulative -rc1 patch, I've managed to get it working
for .18 .19 .20 .21-rc1 and .21-bk

When .21 comes out we'll make a cumulative patch.

-- 
/Martin

Re: [PATCH] patch-o-matic fixes

[Willy Tarreau]

Hello Martin,

On Wed, Apr 30, 2003 at 05:52:01PM +0200, Martin Josefsson wrote:
 
> It does apply IF you've applied pending/24_conntrack-nosysctl.patch
> Which should have been done. We've had implicit dependencies on
> submitted and pending stuff many times.

That's not my case, I disabled 24_conntrack-nosysctl because of the dependancy.
Unless I have a broken tree, or a conflict is induced by other patches,
I'm certain that tcp_window_tracking on top of 2.4.21-bk doesn't apply
cleanly. Here's what I have (in fact, what 'runme' sees) :

                 submitted/01_2.4.19
                 submitted/02_2.4.20
                 submitted/03_ip_conntrack_proto_tcp-lockfix
                 submitted/04_newnat-udp-helper
                 submitted/05_REJECT-fwspotting-phrack60-fix
                 submitted/06_ftp-conntrack-msg-fix
                 submitted/07_ECN-tcpchecksum-littleendian-fix
                 submitted/08_ftp-conntrack-debugftp
                 submitted/08_mangle_input_noroutemeharder
                 submitted/09_icmp-match-all
                 submitted/10_confirm_fix
                 submitted/10_local-nat-expectfn
                 submitted/11_inner-icmp-translation-fix
                 submitted/13_ftp-conntrack-epsv-typo
                 submitted/14_hl-ipv6
                 submitted/14_ip6t_ext_hdr-ipv6
                 submitted/15_ahesp6-ipv6
                 submitted/16_frag6-ipv6
                 submitted/17_ipv6header-ipv6
                 submitted/18_opts6-ipv6
                 submitted/19_route6-ipv6
                 submitted/21_tftp-conntrack-nat
                 submitted/22_amanda-conntrack-nat
                 submitted/23_REJECT-headroom-tcprst
                 submitted/24_conntrack-modify-after-free-fix
                 submitted/25_ip_tables-comment-fix
                 submitted/26_ipt_unclean-ubit
                 submitted/27_no_version
                 submitted/28_compat_code_removal
                 submitted/29_mangle-bogus_cast-ipv6
                 submitted/30_ip6_route_me_harder-ipv6
                 submitted/ipt_ULOG-mac_len-fix
                 submitted/ipt_multiport-invfix
                 pending/06_early_drop-backwards
                 pending/20_iptables-proc
                 pending/25_natcore-nohelper
                 pending/26_morethan32cpufix
                 pending/27_include-nfc-order
                 pending/unused_var
                 base/HL-ipv6
                 base/IPV4OPTSSTRIP
                 base/NETLINK
                 base/NETMAP
                 base/REJECT-ipv6
                 base/SAME
                 base/TTL
                 base/condition
                 base/condition6-ipv6
                 base/fuzzy
                 base/fuzzy6-ipv6
                 base/iplimit
                 base/ipv4options
                 base/mport
                 base/nth
                 base/pool
                 base/psd
                 base/quota
                 base/random
                 base/realm
                 base/time
		 base/u32
                 extra/CLASSIFY
                 extra/CONNMARK
                 extra/ROUTE
                 extra/cuseeme-nat
                 extra/eggdrop-conntrack
                 extra/h323-conntrack-nat
                 extra/ip6_tables-proc-ipv6
                 extra/ipt_TARPIT
                 extra/mms-conntrack-nat
                 extra/nfnetlink-ctnetlink-0.11
                 extra/owner-socketlookup
                 extra/ownercmd-ipv6
                 extra/pptp-conntrack-nat
                 extra/quake3-conntrack
                 extra/recent
                 extra/rpc
                 extra/rsh
                 extra/string
                 extra/talk-conntrack-nat
                 extra/tcp-window-tracking

And those I exclude because of too complicated rejects, or incompatibilities :

		pending/12_conntrack_nat_dependancy
		pending/24_conntrack-nosysctl
		extra/ip_conntrack-timeouts

> No need to make a cumulative -rc1 patch, I've managed to get it working
> for .18 .19 .20 .21-rc1 and .21-bk

Ok, thanks for the info, I'll retry on a fresh kernel when I have time, and without
all other patches.

Regards,
Willy

Re: [PATCH] patch-o-matic fixes

[Martin Josefsson]

On Wed, 2003-04-30 at 18:28, Willy Tarreau wrote:
> Hello Martin,
> 
> On Wed, Apr 30, 2003 at 05:52:01PM +0200, Martin Josefsson wrote:
>  
> > It does apply IF you've applied pending/24_conntrack-nosysctl.patch
> > Which should have been done. We've had implicit dependencies on
> > submitted and pending stuff many times.
> 
> That's not my case, I disabled 24_conntrack-nosysctl because of the dependancy.
> Unless I have a broken tree, or a conflict is induced by other patches,
> I'm certain that tcp_window_tracking on top of 2.4.21-bk doesn't apply
> cleanly. Here's what I have (in fact, what 'runme' sees) :

Please get a current patch-o-matic from cvs and try again.

> And those I exclude because of too complicated rejects, or incompatibilities :
> 
> 		pending/12_conntrack_nat_dependancy

This patch doesn't even exist in p-o-m anymore.

> 		extra/ip_conntrack-timeouts

This one is incompatible with extra/tcp-window-tracking.patch

The dependencies in p-o-m won't allow you to try applying both these
patches at the same time (unless you use force)

-- 
/Martin

Re: [PATCH] patch-o-matic fixes

[Harald Welte]

[-- Attachment #1: Type: text/plain, Size: 797 bytes --]

On Wed, Apr 30, 2003 at 05:52:01PM +0200, Martin Josefsson wrote:

> It does apply IF you've applied pending/24_conntrack-nosysctl.patch
> Which should have been done. We've had implicit dependencies on
> submitted and pending stuff many times.
> 
> No need to make a cumulative -rc1 patch, I've managed to get it working
> for .18 .19 .20 .21-rc1 and .21-bk

great, thanks for pointing this out.

> /Martin
> 

-- 
- Harald Welte <laforge@netfilter.org>             http://www.netfilter.org/
============================================================================
  "Fragmentation is like classful addressing -- an interesting early
   architectural error that shows how much experimentation was going
   on while IP was being designed."                    -- Paul Vixie

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

Re: [PATCH] patch-o-matic fixes

[Martin Josefsson]

On Wed, 2003-04-30 at 22:20, Harald Welte wrote:
> On Wed, Apr 30, 2003 at 05:52:01PM +0200, Martin Josefsson wrote:
> 
> > It does apply IF you've applied pending/24_conntrack-nosysctl.patch
> > Which should have been done. We've had implicit dependencies on
> > submitted and pending stuff many times.
> > 
> > No need to make a cumulative -rc1 patch, I've managed to get it working
> > for .18 .19 .20 .21-rc1 and .21-bk
> 
> great, thanks for pointing this out.

I have a small script that tests p-o-m against .18 .19 .20
latest .21-{pre,rc} and .21-bk and reports any failures. It is still
missing some stuff and needs a good cleanup before I dare releasing it.

My plans are to implement compile-testing for both user and kernelspace
as well.

-- 
/Martin

Re: [PATCH] patch-o-matic fixes

[Willy Tarreau]

On Wed, Apr 30, 2003 at 06:44:26PM +0200, Martin Josefsson wrote:
> Please get a current patch-o-matic from cvs and try again.

I've just rechecked with a fresh 2.4.20 + patch-2.4.21-rc1 + cset-1.1141
from kernel.org, and a fresh p-o-m-20030429 from netfilter.org, and I can
only reproduce the problem if I don't enable 24_conntrack-nosysctl (as you
said). Once it's applied, tcp-window-tracking applies correctly. I don't
really know why I wasn't including it, because I had marked it as one to
avoid because of a conflict it would have caused in an earlier version, or
against a broken kernel... Strange...

> > 		pending/12_conntrack_nat_dependancy
> 
> This patch doesn't even exist in p-o-m anymore.

Yes, I know that but it's still listed here because I maintain a list of
patches to avoid, which makes frequent rebuilds far easier.

> > 		extra/ip_conntrack-timeouts
> 
> This one is incompatible with extra/tcp-window-tracking.patch
> 
> The dependencies in p-o-m won't allow you to try applying both these
> patches at the same time (unless you use force)

I know, but it's on my avoid-list too, because if I apply it by mistake,
I only see the conflict on the last one... tcp-window-tracking, so I
prefer drop it from the start.

Well, it seems that my list initiated to make my life easier fooled me
instead, this time !

Thanks for your help, Harald and Martin, and sorry for the noise.

Regards,
Willy