Quoting luigi:
In order to make the userland code fully 64-bit clean it may
be necessary to commit other changes that may or may not cause
a minor change in the ABI.
Reviewed by: luigi
has always done.
Technically, this is the wrong format, but it reduces the diffs in
-stable. Someday, when we get rid of ipfw1, I will put the port number
in the proper format both in kernel and userland.
MFC after: 3 days
(with re@ permission)
following Julian's good suggestion: since you can specify any match
pattern as an option, rules now have the following format:
[<proto> from <src> to <dst>] [options]
i.e. the first part is now entirely optional (and left there just
for compatibility with ipfw1 rulesets).
Add a "-c" flag to show/list rules in the compact form
(i.e. without the "ip from any to any" part) when possible.
The default is to include it so that scripts processing ipfw's
canonical output will still work.
Note that as part of this cleanup (and to remove ambiguity), MAC
fields now can only be specified in the options part.
Update the manpage to reflect the syntax.
Clarify the behaviour when a match is attempted on fields which
are not present in the packet, e.g. port numbers on non TCP/UDP
packets, and the "not" operator is specified. E.g.
ipfw add allow not src-port 80
will match also ICMP packets because they do not have port numbers, so
"src-port 80" will fail and "not src-port 80" will succeed. For such
cases it is advised to insert further options to prevent undesired results
(e.g. in the case above, "ipfw add allow proto tcp not src-port 80").
We definitely need to rewrite the parser using lex and yacc!
render the syntax less ambiguous.
Now rules can be in one of these two forms
<action> <protocol> from <src> to <dst> [options]
<action> MAC dst-mac src-mac mac-type [options]
however you can now specify MAC and IP header fields as options e.g.
ipfw add allow all from any to any mac-type arp
ipfw add allow all from any to any { dst-ip me or src-ip me }
which makes complex expressions a lot easier to write and parse.
The "all from any to any" part is there just for backward compatibility.
Manpage updated accordingly.
Implement the M_SKIP_FIREWALL bit in m_flags to avoid loops
for firewall-generated packets (the constant has to go in sys/mbuf.h).
Better comments on keepalive generation, and enforce dyn_rst_lifetime
and dyn_fin_lifetime to be less than dyn_keepalive_period.
Enforce limits (up to 64k) on the number of dynamic buckets, and
retry allocation with smaller sizes.
Raise default number of dynamic rules to 4096.
Improved handling of set of rules -- now you can atomically
enable/disable multiple sets, move rules from one set to another,
and swap sets.
sbin/ipfw/ipfw2.c:
userland support for "noerror" pipe attribute.
userland support for sets of rules.
minor improvements on rule parsing and printing.
sbin/ipfw/ipfw.8:
more documentation on ipfw2 extensions, differences from ipfw1
(so we can use the same manpage for both), stateful rules,
and some additional examples.
Feedback and more examples needed here.
with ipfw2 extensions and give examples of use of the new features.
This is just a preliminary commit, where i simply added the basic
syntax for the extensions, and clean up the page (e.g. by listing
things in alphabetical rather than random order).
I would appreciate feedback and possible corrections/extensions
by interested parties.
Still missing are a more detailed description of stateful rules
(with keepalives), interaction with of stateful rules and natd (don't do
that!), examples of use with the recently introduced rule sets.
There is an issue related to the MFC: RELENG_4 still has ipfw as a
default, and ipfw2 is optional. We have two options here: MFC this
page as ipfw(8) adding a large number of "SORRY NOT IN IPFW" notes,
or create a new ipfw2(8) manpage just for -stable users. I am all
for the first approach, but of course am listening to your comments.
The bugfix (ipfw2.c) makes the handling of port numbers with
a dash in the name, e.g. ftp-data, consistent with old ipfw:
use \\ before the - to consider it as part of the name and not
a range separator.
The new feature (all this description will go in the manpage):
each rule now belongs to one of 32 different sets, which can
be optionally specified in the following form:
ipfw add 100 set 23 allow ip from any to any
If "set N" is not specified, the rule belongs to set 0.
Individual sets can be disabled, enabled, and deleted with the commands:
ipfw disable set N
ipfw enable set N
ipfw delete set N
Enabling/disabling of a set is atomic. Rules belonging to a disabled
set are skipped during packet matching, and they are not listed
unless you use the '-S' flag in the show/list commands.
Note that dynamic rules, once created, are always active until
they expire or their parent rule is deleted.
Set 31 is reserved for the default rule and cannot be disabled.
All sets are enabled by default. The enable/disable status of the sets
can be shown with the command
ipfw show sets
Hopefully, this feature will make life easier to those who want to
have atomic ruleset addition/deletion/tests. Examples:
To add a set of rules atomically:
ipfw disable set 18
ipfw add ... set 18 ... # repeat as needed
ipfw enable set 18
To delete a set of rules atomically
ipfw disable set 18
ipfw delete set 18
ipfw enable set 18
To test a ruleset and disable it and regain control if something
goes wrong:
ipfw disable set 18
ipfw add ... set 18 ... # repeat as needed
ipfw enable set 18 ; echo "done "; sleep 30 && ipfw disable set 18
here if everything goes well, you press control-C before
the "sleep" terminates, and your ruleset will be left
active. Otherwise, e.g. if you cannot access your box,
the ruleset will be disabled after the sleep terminates.
I think there is only one more thing that one might want, namely
a command to assign all rules in set X to set Y, so one can
test a ruleset using the above mechanisms, and once it is
considered acceptable, make it part of an existing ruleset.
+ the header file contains two different opcodes (O_IPOPTS and O_IPOPT)
for what is the same thing, and sure enough i used one in the kernel
and the other one in userland. Be consistent!
+ "keep-state" and "limit" must be the last match pattern in a rule,
so no matter how you enter them move them to the end of the rule.
* accept "icmptype" as an alias for "icmptypes";
* remove an extra whitespace after "log" rules;
* print correctly the "limit" masks;
* correct a typo in parsing dummynet arguments (this caused a coredump);
* do not allow specifying both "check-state" and "limit", they are
(and have always been) mutually exclusive;
* remove an extra print of the rule before installing it;
* make stdout buffered -- otherwise, if you log its output with syslog,
you will see one entry for each printf(). Rather unpleasant.
fatal on alphas.
Fixed setting of WARNS. WARNS should never be set unconditionally, since
this breaks testing of different WARNS values by setting it at a higher
level (e.g., on the command line).
now it should support all the instructions of the old ipfw.
Fix some bugs in the user interface, /sbin/ipfw.
Please check this code against your rulesets, so i can fix the
remaining bugs (if any, i think they will be mostly in /sbin/ipfw).
Once we have done a bit of testing, this code is ready to be MFC'ed,
together with a bunch of other changes (glue to ipfw, and also the
removal of some global variables) which have been in -current for
a couple of weeks now.
MFC after: 7 days
This code makes use of variable-size kernel representation of rules
(exactly the same concept of BPF instructions, as used in the BSDI's
firewall), which makes firewall operation a lot faster, and the
code more readable and easier to extend and debug.
The interface with the rest of the system is unchanged, as witnessed
by this commit. The only extra kernel files that I am touching
are if_fw.h and ip_dummynet.c, which is quite tied to ipfw. In
userland I only had to touch those programs which manipulate the
internal representation of firewall rules).
The code is almost entirely new (and I believe I have written the
vast majority of those sections which were taken from the former
ip_fw.c), so rather than modifying the old ip_fw.c I decided to
create a new file, sys/netinet/ip_fw2.c . Same for the user
interface, which is in sbin/ipfw/ipfw2.c (it still compiles to
/sbin/ipfw). The old files are still there, and will be removed
in due time.
I have not renamed the header file because it would have required
touching a one-line change to a number of kernel files.
In terms of user interface, the new "ipfw" is supposed to accepts
the old syntax for ipfw rules (and produce the same output with
"ipfw show". Only a couple of the old options (out of some 30 of
them) has not been implemented, but they will be soon.
On the other hand, the new code has some very powerful extensions.
First, you can put "or" connectives between match fields (and soon
also between options), and write things like
ipfw add allow ip from { 1.2.3.4/27 or 5.6.7.8/30 } 10-23,25,1024-3000 to any
This should make rulesets slightly more compact (and lines longer!),
by condensing 2 or more of the old rules into single ones.
Also, as an example of how easy the rules can be extended, I have
implemented an 'address set' match pattern, where you can specify
an IP address in a format like this:
10.20.30.0/26{18,44,33,22,9}
which will match the set of hosts listed in braces belonging to the
subnet 10.20.30.0/26 . The match is done using a bitmap, so it is
essentially a constant time operation requiring a handful of CPU
instructions (and a very small amount of memmory -- for a full /24
subnet, the instruction only consumes 40 bytes).
Again, in this commit I have focused on functionality and tried
to minimize changes to the other parts of the system. Some performance
improvement can be achieved with minor changes to the interface of
ip_fw_chk_t. This will be done later when this code is settled.
The code is meant to compile unmodified on RELENG_4 (once the
PACKET_TAG_* changes have been merged), for this reason
you will see #ifdef __FreeBSD_version in a couple of places.
This should minimize errors when (hopefully soon) it will be time
to do the MFC.
fields as discussed in the commit to ip_fw.c:1.186
On top of this, a ton of non functional changes to clean up the code,
write functions to replace sections of code that were replicated
multiple times (e.g. the printing or matching of flags and options),
splitting long sections of inlined code into separate functions,
and the like.
I have tested the code quite a bit, but some typos (using one variable
in place of another) might have escaped.
The "embedded manpage" is a bit inconsistent, but i am leaving fixing
it for later. The current format makes no sense, it is over 40 lines
long and practically unreadable. We can either split it into sections
( ipfw -h options , ipfw -h pipe , ipfw -h queue ...)
or remove it altogether and refer to the manpage.
+ setting a bandwidth too large for a pipe (above 2Gbit/s) could
cause the internal representation (which is int) to wrap to a
negative number, causing an infinite loop in the kernel;
+ (see PR bin/35628): when configuring RED parameters for a queue,
the values are not passed to the kernel resulting in panics at
runtime (part of the problem here is also that the kernel does
not check for valid parameters being passed, but this will be
fixed in a separate commit).
These are both critical fixes which need to be merged into 4.6-RELEASE.
MFC after: 1 day
more on how ipfw(8) deals with tiny fragments. While we're at it, add
a quick log message to even let people know we dropped a packet. (Note
that the second FINE POINT is somewhat redundant given the first, but
since the code is there, leave the docs for it.)
MFC after: 1 day
time_to_xxx() and xxx_to_time() functions. e.g. _time_to_xxx()
instead of time_to_xxx(), to make it more obvious that these are
stopgap functions & placemarkers and not meant to create a defacto
standard. They will eventually be replaced when a real standard
comes out of committee.
reinserted by a userland process, will lose a number of packet
attributes, including their source interface. This may affect
the behavior of later rules, and while not strictly a BUG, may
cause unexpected behavior if not clearly documented. A similar
note for natd(8) might be desirable.
ipfirewall(4) to the IMPLEMENTATION NOTES section because it
considers kernel internals and may confuse newbies if placed
at the very beginning of the manpage (where it used to be previously.)
Not objected by: luigi
Fair Queueing) and RED (Random Early Detection) to both give the reader
a hint what they are and to make it easier to find out more information
about them.
