iptables 1.4.16.1 release
Subsystems (3)
Pablo Neira Ayuso
maintainerOn Mon, Oct 08, 2012 at 12:24:41AM +0200, Pablo Neira Ayuso wrote:
Hi!
The Netfilter project proudly presents:
iptables 1.4.16.1
This release fixes a major breakage introduced by:
commit cd2f9bdbb7f9b737e5d640aafeb78bcd8e3a7adf
Author: Jan Engelhardt <[email protected]> Date: Tue Sep 4 05:24:47 2012 +0200
iptables: support for target aliases
This is really unfortunate, it seems this patch has been pushed mainstream without sufficient testing. We are really sorry for the inconvenience. Please, don't use 1.4.16, this bug reders it completely useless.
See ChangeLog that comes attached to this email for more details.
You can download it from:
http://www.netfilter.org/projects/iptables/downloads.html
ftp://ftp.netfilter.org/pub/iptables/
Have fun!
Pablo Neira Ayuso (2):
iptables: fix standard target
bump version to 1.4.16.1
Pablo Neira Ayuso
maintainerQuoted 13 lines · Pablo Neira Ayuso
On Mon, Oct 08, 2012 at 12:24:41AM +0200, Pablo Neira Ayuso wrote:
Hi!The Netfilter project proudly presents:
iptables 1.4.16.1This release fixes a major breakage introduced by:
commit
cd2f9bdbb7f9b737e5d640aafeb78bcd8e3a7adfAuthor: Jan Engelhardt <[email protected]> Date: Tue Sep 4 05:24:47 2012 +0200
Quoted 8 lines
iptables: support for target aliases
This is really unfortunate, it seems this patch has been pushed mainstream without sufficient testing. We are really sorry for the inconvenience. Please, don't use 1.4.16, this bug reders it completely useless.
See ChangeLog that comes attached to this email for more details.
I got everything working back but the new target alias infrastructure while rushing to fix the breakage.
The patch attached should get the target aliasing working. This is not serious though, the only target that can benefit from this is feature is NOTRACK, and it will be removed in the upcoming 3.7, ie. there is no kernel available that can benefit from this feature.
iptables/ip6tables.c +1 -1
@@ -1286,7 +1286,7 @@ static void command_jump(struct iptables_command_state *cs) cs->target->t =xtables_calloc (1 , size); cs->target->t->u.target_size = size; -if (cs->target->real_name !=NULL ) +if (strcmp (cs->target->real_name,"standard" ) ==0 )strcpy (cs->target->t->u.user.name, cs->jumpto);else strcpy (cs->target->t->u.user.name, cs->target->real_name);
iptables/iptables.c +1 -1
@@ -1295,7 +1295,7 @@ static void command_jump(struct iptables_command_state *cs) cs->target->t =xtables_calloc (1 , size); cs->target->t->u.target_size = size; -if (cs->target->real_name !=NULL ) +if (strcmp (cs->target->real_name,"standard" ) ==0 )strcpy (cs->target->t->u.user.name, cs->jumpto);else strcpy (cs->target->t->u.user.name, cs->target->real_name);
Quoted 9 lines · Pablo Neira Ayuso
The Netfilter project proudly presents:
iptables 1.4.16.1iptables -I INPUT -j ACCEPT
says:
iptables: No chain/target/match by that name.
This also breaks iptables-restore, of course. Jan, you'll have to explain me how you have tested this.
This was tested by adding rules with different targets that had both aliases defined and those without.
./iptables/xtables-multi main4 -t raw -N foo ./iptables/xtables-multi main4 -t raw -A foo -j NOTRACK with kernels that had xt_CT and no xt_CT at all
./iptables/xtables-multi main4 -N foo ./iptables/xtables-multi main4 -A foo -m state --state NEW with kernels that had xt_conntrack.3, and xt_conntrack.3 removed (leaving only xt_conntrack.2)
./iptables/xtables-multi main4 -t raw -N bar ./iptables/xtables-multi main4 -t raw -A bar -j MARK --set-xmark 1 ./iptables/xtables-multi main4 -t raw -A foo -j bar
plus of course the "standard" (no pun intended) testsuite that we had so far:
# ./iptables/xtables-multi restore6 tests/options-most.rules WARNING: --localtz is being replaced by --kerneltz, since "local" is ambiguous. Note the kernel timezone has caveats - see manpage for details.
As you spotted, options-most.rules did not include -j <verdict>.
While v1.4.16-1-g2aaa7ec fixes -j verdict, it breaks NOTRACK in all instances. To reuse a line, "you'll have to explain me how you have tested this."
A patch to what I think should fly is posted as a reply hereto. Please give that a spin.
PATCH 1/1 iptables: restore NOTRACK functionality, target aliasing
Commit v1.4.16-1-g2aaa7ec is testing for real_name (not) being NULL which was always false (true). real_name was never NULL, so cs->jumpto would always be used, which rendered -j NOTRACK unusable, since the chosen real name.revision is for example NOTRACK.1, which does not exist at the kernel side.
# ./iptables/xtables-multi main4 -t raw -A foo -j NOTRACK dbg: Using NOTRACK.1 WARNING: The NOTRACK target is obsolete. Use CT instead. iptables: Protocol wrong type for socket.
To reasonably support the extra-special verdict names, make it so that real_name remains NULL when an extension defined no alias, which we can then use to determine whether the user entered an alias name (which needs to be followed) or not.
---
(This is a try at a resend since I have not gotten back this very email from majordomo as a list subscriber...)
Patch downloadable via
git://git.inai.de/iptables master
as well
iptables/ip6tables.c | 21 ++++++++++++++------- iptables/iptables.c | 23 +++++++++++++++-------- libxtables/xtables.c | 26 ++++++++++++++------------ 3 files changed, 43 insertions(+), 27 deletions(-)
iptables/ip6tables.c +14 -7
@@ -1286,15 +1286,19 @@ static void command_jump(struct iptables_command_state *cs) cs->target->t =xtables_calloc (1 , size); cs->target->t->u.target_size = size; -if (cs->target->real_name !=NULL ) +if (cs->target->real_name ==NULL ) { +/* +* Spooky action at a distance: Verdicts like ACCEPT, DROP, +* etc. are translated to the real name in libiptc instead. +*/ strcpy (cs->target->t->u.user.name, cs->jumpto); -else + }else {strcpy (cs->target->t->u.user.name, cs->target->real_name); - cs->target->t->u.user.revision = cs->target->revision; -if (cs->target->real_name != cs->target->name)fprintf (stderr,"WARNING: The %s target is obsolete. " "Use %s instead.\n" , cs->jumpto, cs->target->real_name); + } + cs->target->t->u.user.revision = cs->target->revision;xs_init_target (cs->target);if (cs->target->x6_options !=NULL )
@@ -1322,11 +1326,14 @@ static void command_match(struct iptables_command_state *cs) size =XT_ALIGN (sizeof (struct xt_entry_match )) + m->size; m->m =xtables_calloc (1 , size); m->m->u.match_size = size; -strcpy (m->m->u.user.name, m->real_name); - m->m->u.user.revision = m->revision; -if (m->real_name != m->name) +if (m->real_name ==NULL ) { +strcpy (m->m->u.user.name, m->name); + }else { +strcpy (m->m->u.user.name, m->real_name);fprintf (stderr,"WARNING: The %s match is obsolete. " "Use %s instead.\n" , m->name, m->real_name); + } + m->m->u.user.revision = m->revision;xs_init_match (m);if (m == m->next)
iptables/iptables.c +15 -8
@@ -1295,16 +1295,20 @@ static void command_jump(struct iptables_command_state *cs) cs->target->t =xtables_calloc (1 , size); cs->target->t->u.target_size = size; -if (cs->target->real_name !=NULL ) +if (cs->target->real_name ==NULL ) { +/* +* Spooky action at a distance: Verdicts like ACCEPT, DROP, +* etc. are translated to the real name in libiptc instead. +*/ strcpy (cs->target->t->u.user.name, cs->jumpto); -else -strcpy (cs->target->t->u.user.name, cs->target->real_name); - cs->target->t->u.user.revision = cs->target->revision; -if (cs->target->real_name != cs->target->name) + }else {/* Alias support for userspace side */ +strcpy (cs->target->t->u.user.name, cs->target->real_name);fprintf (stderr,"WARNING: The %s target is obsolete. " "Use %s instead.\n" , cs->jumpto, cs->target->real_name); + } + cs->target->t->u.user.revision = cs->target->revision;xs_init_target (cs->target);
@@ -1333,11 +1337,14 @@ static void command_match(struct iptables_command_state *cs) size =XT_ALIGN (sizeof (struct xt_entry_match )) + m->size; m->m =xtables_calloc (1 , size); m->m->u.match_size = size; -strcpy (m->m->u.user.name, m->real_name); - m->m->u.user.revision = m->revision; -if (m->real_name != m->name) +if (m->real_name ==NULL ) { +strcpy (m->m->u.user.name, m->name); + }else { +strcpy (m->m->u.user.name, m->real_name);fprintf (stderr,"WARNING: The %s match is obsolete. " "Use %s instead.\n" , m->name, m->real_name); + } + m->m->u.user.revision = m->revision;xs_init_match (m);if (m == m->next)
libxtables/xtables.c +14 -12
@@ -848,8 +848,6 @@ void xtables_register_match(struct xtables_match *me)exit (1 ); } -if (me->real_name ==NULL ) - me->real_name = me->name;if (me->x6_options !=NULL )xtables_option_metavalidate (me->name, me->x6_options);if (me->extra_opts !=NULL )
@@ -905,9 +903,9 @@ xtables_mt_prefer(bool a_alias, unsigned int a_rev, unsigned int a_fam,static int xtables_match_prefer (const struct xtables_match *a,const struct xtables_match *b) { -return xtables_mt_prefer (a->name != a->real_name, +return xtables_mt_prefer (a->real_name !=NULL , a->revision, a->family, - b->name != b->real_name, + b->real_name !=NULL , b->revision, b->family); }
@@ -919,15 +917,16 @@ static int xtables_target_prefer(const struct xtables_target *a,* xtables_register_*; the direct pointer comparison here is therefore * legitimate to detect an alias. */ -return xtables_mt_prefer (a->name != a->real_name, +return xtables_mt_prefer (a->real_name !=NULL , a->revision, a->family, - b->name != b->real_name, + b->real_name !=NULL , b->revision, b->family); }static void xtables_fully_register_pending_match (struct xtables_match *me) {struct xtables_match **i, *old; +const char *rn;int compare; old =xtables_find_match (me->name,XTF_DURING_LOAD ,NULL );
@@ -941,12 +940,14 @@ static void xtables_fully_register_pending_match(struct xtables_match *me) }/* Now we have two (or more) options, check compatibility. */ + rn = (old->real_name !=NULL ) ? old->real_name : old->name;if (compare >0 && -compatible_match_revision (old->real_name, old->revision)) +compatible_match_revision (rn, old->revision))return ;/* See if new match can be used. */ -if (!compatible_match_revision (me->real_name, me->revision)) + rn = (me->real_name !=NULL ) ? me->real_name : me->name; +if (!compatible_match_revision (rn, me->revision))return ;/* Delete old one. */
@@ -1005,8 +1006,6 @@ void xtables_register_target(struct xtables_target *me)exit (1 ); } -if (me->real_name ==NULL ) - me->real_name = me->name;if (me->x6_options !=NULL )xtables_option_metavalidate (me->name, me->x6_options);if (me->extra_opts !=NULL )
@@ -1024,6 +1023,7 @@ void xtables_register_target(struct xtables_target *me)static void xtables_fully_register_pending_target (struct xtables_target *me) {struct xtables_target *old; +const char *rn;int compare; old =xtables_find_target (me->name,XTF_DURING_LOAD );
@@ -1039,12 +1039,14 @@ static void xtables_fully_register_pending_target(struct xtables_target *me) }/* Now we have two (or more) options, check compatibility. */ + rn = (old->real_name !=NULL ) ? old->real_name : old->name;if (compare >0 && -compatible_target_revision (old->real_name, old->revision)) +compatible_target_revision (rn, old->revision))return ;/* See if new target can be used. */ -if (!compatible_target_revision (me->real_name, me->revision)) + rn = (me->real_name !=NULL ) ? me->real_name : me->name; +if (!compatible_target_revision (rn, me->revision))return ;/* Delete old one. */
Signature · 1 line
1.7.10.4
Pablo Neira Ayuso
maintainerRe: [PATCH] iptables: restore NOTRACK functionality, target aliasing
Quoted 11 lines · Jan Engelhardt
Commit v1.4.16-1-g2aaa7ec is testing for
real_name(not) being NULL which was always false (true).real_namewas never NULL, socs->jumptowould always be used, which rendered -j NOTRACK unusable, since the chosen real name.revision is for example NOTRACK.1, which does not exist at the kernel side.# ./iptables/xtables-multi main4 -t raw -A foo -j NOTRACK dbg: Using NOTRACK.1 WARNING: The NOTRACK target is obsolete. Use CT instead. iptables: Protocol wrong type for socket.
Quoted 4 lines
To reasonably support the extra-special verdict names, make it so that
real_nameremains NULL when an extension defined no alias, which we can then use to determine whether the user entered an alias name (which needs to be followed) or not.
I have applied this and made a new release.
I kindly told you. I don't want late patches to hit iptables if I'm about to release it, ie. close to when the Linux kernel comes out.
The reason was that chances to hit bugs and not noticing becomes higher. In other words, stick to conservative mode.
Let this serve as proof of it.
You disregarded my advice and now we have this shame, three releases in one day just because of rushing.
Re: [PATCH] iptables: restore NOTRACK functionality, target aliasing
Quoted 17 lines · Pablo Neira Ayuso
Quoted 11 lines · Jan Engelhardt
Commit v1.4.16-1-g2aaa7ec is testing for
real_name(not) being NULL which was always false (true).real_namewas never NULL, socs->jumptowould always be used, which rendered -j NOTRACK unusable, since the chosen real name.revision is for example NOTRACK.1, which does not exist at the kernel side.# ./iptables/xtables-multi main4 -t raw -A foo -j NOTRACK dbg: Using NOTRACK.1 WARNING: The NOTRACK target is obsolete. Use CT instead. iptables: Protocol wrong type for socket.Quoted 4 lines
To reasonably support the extra-special verdict names, make it so that
real_nameremains NULL when an extension defined no alias, which we can then use to determine whether the user entered an alias name (which needs to be followed) or not.
Quoted 6 lines
I have applied this and made a new release.
I kindly told you. I don't want late patches to hit iptables if I'm about to release it, ie. close to when the Linux kernel comes out. The reason was that chances to hit bugs and not noticing becomes higher. In other words, stick to conservative mode.
That is clear now, but was not when when the Alias Support patches went in. Sorry for that.
As for the documentation updates ("doc:" labeled patches in
http://marc.info/?l=netfilter-devel&m=134903982604088&w=2 ), you gave an OK, but then infuriated over them being merged, though documentation generally does not cause bugs.
Quoted 5 lines
Let this serve as proof of it.
You disregarded my advice and now we have this shame, three releases in one day just because of rushing.
I will heed your words, but I do also think that letting patches sit in git for longer may not necessarily buy more. For -j verdict, maybe. But in general, iptables firewalling & networking users do not use git so much, but they love tarballs, also because that gets spread out by distros in a glimpse.
Here is an instance of iproute2 where an issue also only got noticed after the tar got out:
http://archlinux.2023198.n4.nabble.com/ip-addr-show-doesn-t-show-my-ipv4-address-after-update-is-this-expected-behaviour-td4658792.html
http://forums.funtoo.org/viewtopic.php?id=1526
(this subsequently also got posted to netdev)
There is also that one NLM_F_MULTI issue in libnfnetlink, which slept for 2 months in git before becoming a tar.
If, and only if, the "next" branch is empty (which was the case for libnfnetlink), I do not think synchronization to the kernel release schedule makes sense. But that is just my opinion..
Something completely different:
There seems to be something happening in your git cherry-pick process causing the time to shift. Commit ed03bc396b861ae0302b5cc9b3826cf58ef6aa24 fromgit://git.inai.de/iptables has AuthorDate 2012-10-08 01:54:48+0200, but commit dd43527cb6bdf3d469100850ca10dcd2fb761304 in iptables has
2012-10-07 14:32:36+0000. I am not sure if your clock is off, since
the CommitDate looks accurate.