and change it to a void function.
We use a compressed structure for TCPS_TIME_WAIT to save memory. Any late
late segments arriving for such a connection is handled directly in the TW
code.
- Locks were not being unlocked when an invalid size chunk is
sent in.
- When a notification comes in, we cannot use it to look up
the fragment interleave stream information since its not
on a stream.
state tcp_debug, tcp_debx. Acquire and drop as required in tcp_trace().
Move to ANSI C function header, correct prototype types so that short TCP
state is no longer promoted to int unnecessarily.
Add comments.
MFC after: 3 weeks
it was full and a collision occured, then we would leave
a inp locked. Also fixes a missing inp unlock if IPSEC was
on and it failed during the attach. Bug found by Weongyo Jeong.
on each socket buffer with the socket buffer's mutex. This sleep lock is
used to serialize I/O on sockets in order to prevent I/O interlacing.
This change replaces the custom sleep lock with an sx(9) lock, which
results in marginally better performance, better handling of contention
during simultaneous socket I/O across multiple threads, and a cleaner
separation between the different layers of locking in socket buffers.
Specifically, the socket buffer mutex is now solely responsible for
serializing simultaneous operation on the socket buffer data structure,
and not for I/O serialization.
While here, fix two historic bugs:
(1) a bug allowing I/O to be occasionally interlaced during long I/O
operations (discovere by Isilon).
(2) a bug in which failed non-blocking acquisition of the socket buffer
I/O serialization lock might be ignored (discovered by sam).
SCTP portion of this patch submitted by rrs.
set/clear it but would not do it. Now we will.
- Moved to latest socket api for extended sndrcv info struct.
- Moved to support all new levels of fragment interleave (0-2).
- Codenomicon security test updates - length checks and such.
- Bug in stream reset (2 actually).
- setpeerprimary could unlock a null pointer, fixed.
- Added a flag in the pcb so netstat can see if we are listening easier.
Obtained from: (some of the Listen changes from Weongyo Jeong)
consistent with the naming of other structure field members, and
reducing improper grep matches. Clean up and comment structure
fields in structure definition.
set/clear it but would not do it. Now we will.
- Moved to latest socket api for extended sndrcv info struct.
- Moved to support all new levels of fragment interleave.
from the incoming SYN handling section of tcp_input().
Enforcement of the accept queue limits is done by sonewconn() after the
3WHS is completed. It is not necessary to have an earlier check before a
connection request enters the SYN cache awaiting the full handshake. It
rather limits the effectiveness of the syncache by preventing legit and
illegit connections from entering it and having them shaken out before we
hit the real limit which may have vanished by then.
Change return value of syncache_add() to void. No status communication
is required.
when the ACK is invalid and doesn't belong to any registered connection,
either in syncache or through SYN cookies. True but a NULL struct socket
is returned when the 3WHS completed but the socket could not be created
due to insufficient resources or limits reached.
For both cases an RST is sent back in tcp_input().
A logic error leading to a panic is fixed where syncache_expand() would
free the mbuf on socket allocation failure but tcp_input() later supplies
it to tcp_dropwithreset() to issue a RST to the peer.
Reported by: kris (the panic)
- Removed free-oqueue cache.
- Fix counter for sq entries
- Increased the amount of information retained
on ASOC_TSN logging on the association.
- Made it so with the ASOC_TSN logging on
sending or recieving an abort we dump the log.
- Went through and added invariant's around some
panic's that needed them.
- decrements went to atomic_subtact_int instead of add -1
- Removed residual count increment that threw off a
strm oq count.
- Tracks and complaints if we don't have a LAST fragment and
clean up the sp structure.
- Track a new stat that counts number of abandoned msgs that
happen if you close without reading.
- Fix lookup of frag point to be aware of a 0 assoc-id.
Reviewed by: gnn
to free the oldest entry in the current bucket row. The global
entry limit may be smaller than the bucket rows and their limit
combined however. Thus only try to free a syncache entry if we
found one in this bucket row.
Reported by: kris
"zone", which is generally not present in zone names. This reduces the
incidence of line-wrapping in "vmstat -z " using 80-column displays.
MFC after: 3 days
- Fix bug that prevented EEOR mode from working
and simplified the can_we_split code in the process.
- Reduce lock contention for the tcb_send_lock. I did
this especially for EEOR mode, still need to look at
why I need a lock when removing from the tailq and the
->next is NOT null. A lock fixes it but it implies a
bug yet exists.
- Activated Andre's proposed changes to better use the mbuf
infrastructure.
- Fixed places that were not using the aloc macro's to take
advantage of the per assoc cache.
- Adds ifdef fix so any logging will enable stat_logging to
get the right data structures in place (suggested by Max Laier).
- name change of prefered -> preferred
- CMT fast recover code added.
- Comment fixes in CMT.
- We were not giving a reason of cant_start_asoc per socket api
if we failed to get init/or/cookie to bring up an assoc. Change
so we don't just give a generic "comm lost" but look at actual
states of dying assoc.
- change "crc32" arguments to "crc32c" to silence strict/noisy
compiler warnings when crc32() is also declared
- A few minor tweaks to get the portable stuff truely portable
for sctp6_usrreq.c :-D
- one-2-one style vrf match problem.
- window recovery would leave chks marked for retran
during window probes on the sent queue. This would then
cause an out-of-order problem and assure that the flight
size "problem" would occur.
- Solves a flight size logging issue that caused rwnd
overruns, flight size off as well as false retransmissions.g
- Macroize the up and down of flight size.
- Fix a ECNE bug in its counting.
- The strict_sacks options was causing aborts when window probing
was active, fix to make strict sacks a bit smarter about what
the next unsent TSN is.
- Fixes a one-2-one wakeup bug found by Martin Kulas.
- If-defed out form, Andre's copy routines pending his
commit of at least m_last().. need to adjust for 6.2 as
well.. since m_last won't exist.
Reviewed by: gnn
directly to a merged model where only one callout, the next to fire,
is registered.
Instead of callout_reset(9) and callout_stop(9) the new function
tcp_timer_activate() is used which then internally manages the callout.
The single new callout is a mutex callout on inpcb simplifying the
locking a bit.
tcp_timer() is the called function which handles all race conditions
in one place and then dispatches the individual timer functions.
Reviewed by: rwatson (earlier version)
StartMediaTx message before an OpnRcvChnAck message was received.
Reviewed by: glebius
Approved by: glebius (mentor)
MFC after: 3 days
Found with: Coverity Prevent(tm)
CID: 498
tcp_input() to syncache_socket() where it belongs and the majority
of it already happens.
The "tp->snd_up = tp->snd_una" is removed as it is done with the
tcp_sendseqinit() macro a few lines earlier.
they have been reported back to the userland as being in 1970.
Add boot time to the timestamp to give the time in the scale of the 'current'
real timescale. Not perfect if you change the time a lot but good enough
to keep all the rules correct relative to each other correct in terms
of time relative to "now".
- fixed a refcount bug in the new ifa structures.
- use vrf's from default stcb or inp whenever possible.
- Address limits raised to account for a full IP fragmented
packet (1000 addresses).
- flight size correcting updated to include one message only
and to handle case where the peer does not cumack the
next segment aka lists 1/1 in sack blocks..
- Various bad init/init-ack handling could cause a panic
since we tried to unlock the destroyed mutex. Fixes
so we properly exit when we need to destroy an assoc.
(Found by Cisco DevTest team :D)
- name rename in src-addr-selection from pass to sifa.
- route structure typedef'd to allow different platforms
and updated into sctp_os_bsd file.
- Max retransmissions a chunk can be made added.
Reviewed by: gnn
incorrect, non-bundlable fragmentation.
- Added min residual to better control split points for
both how big a msg must be as well as how much needs
to be left over.
- With our new algo in place, we need to implicitly
set "end of msg" on the sp-> structure otherwise we
end up with "hung" associations.
- Room reserved up front in IP header by pushing IP
header to back of mbuf.
- Fix so FR's peg count of retransmissions needed.
- Fix so an unlucky chunk that never gets across
will kill the assoc via the kill timer and send an
abort too.
- Fix bug in sctp_input which can result in a crash.
- Do not strip off IP options anymore.
- Clean up sctp_calculate_rto().
- Get rid of unused sysctl.
- Fixed so we discard all M-Cast
- Fixed so port check done AFTER checksum
- Fixed bug in fragmentation code that prevented
us from fragmenting a small complete message when
we needed to.
- Window probes were not marked back to unsent and
flight adjusted when a sack came in with no
window change or accepting of the probe data.
We now fix this with having a mark on the net and
the chunk so we can clear it out when the sack arrives
forcing it to retran just like it was "new" this
improves the handling of window probes, which were
dropped by the receiver.
- Tighten AUTH protocol error checks during INIT/INIT-ACK exchange
* Join the IPv4 all-hosts multicast group 224.0.0.1 once only;
that is, when an IPv4 address is first configured on an interface.
* Do not join it for subsequent IPv4 addresses as this violates IGMP.
* Be sure to leave the group when all IPv4 addresses have been removed
from the interface.
* Add two DIAGNOSTIC printfs related to the issue.
Further care and attention is needed in this area; it is suggested that
netinet's attachment to the ifnet structure be compartmentalized and
non-implicit.
Bug found by: andre
MFC after: 1 month
function may be called without any TCP SACK option blocks present. Protect
iteration over SACK option blocks by checking for SACK options present flag
first.
Bug reported by: wkoszek, keramida, Nicolas Blais
explaining that some more locking is needed. The routing pieces are done,
but there is an interlocking issue between optionally compiled code and
mandatory code.
Spotted by: kris
o tcp_input() now handles TCP segment sanity checks and preparations
including the INPCB lookup and syncache.
o tcp_do_segment() handles all data and ACK processing and is IPv4/v6
agnostic.
Change all KASSERT() messages to ("%s: ", __func__).
The changes in this commit are primarily of mechanical nature and no
functional changes besides the function split are made.
Discussed with: rwatson
instance expiry of the ARP entries. Since we no longer abuse the IPv4
radix head lock, we can now enter arp_rtrequest() with a lock held on
an arbitrary rt_entry.
Reviewed by: bms
udp_sendspace, to avoid a situation where jumbograms (datagrams > 9KB)
are unnecessarily fragmented.
A common use case for this is OSPF link-state database synchronization
during adjacency bringup on a high speed network with a large MTU.
It is not possible to auto-tune this setting until a socket is bound to
a given interface, and because the laddr part of the inpcb tuple may be
overridden, it makes no sense to do so. Applications may request a larger
socket buffer size by using the SO_SENDBUF and SO_RECVBUF socket options.
Certain applications such as Quagga ospfd do not probe for interface MTU
and therefore do not increase SO_SENDBUF in this use case.
XORP is not affected by this problem as it preemptively uses SO_SENDBUF
and SO_RECVBUF to account for any possible additional latency in XRL IPC.
PR: kern/108375
Requested by: Vladimir Ivanov
MFC after: 1 week
shutdown which caused extra abort from peer.
- RTT time calculation was not being done in
express sack handling since it refered to an unused
variable (rto_pending). Removed variable.
- socket buffer high water access macro-ized.
structures. Detect when ifnet instances are detached from the network
stack and perform appropriate cleanup to prevent memory leaks.
This has been implemented in such a way as to be backwards ABI compatible.
Kernel consumers are changed to use if_delmulti_ifma(); in_delmulti()
is unable to detect interface removal by design, as it performs searches
on structures which are removed with the interface.
With this architectural change, the panics FreeBSD users have experienced
with carp and pfsync should be resolved.
Obtained from: p4 branch bms_netdev
Reviewed by: andre
Sponsored by: Garance A Drosehn
Idea from: NetBSD
MFC after: 1 month
on a per VRF basis (BSD has only one VRF currently).
Hash table is sized to 16 but may need to be adjusted
for machines with large numbers of addresses.
Reviewed by: gnn
- SB_CLEAR macro defined and used for sb clearing.
- Fix for CMT express_sack_handling did not do proper
pseudo-cumack updates.
- Get rid of extraneous function that was never used ip_2_ip6_hdr()
- Fixed source address selection bug (initialization problem).
- Source address selection debug added.
- *ip is not initialized in the case of inet6 connection, but ip->ip_len is
being changed anyway
Now the question is, why does it think an ipv4 connection is an ipv6 connection?
xemacs still doesn't work over X11 forwarding, but the kernel no longer panics.
and syncache_respond() into its own generic function tcp_addoptions().
tcp_addoptions() is alignment agnostic and does optimal packing in all cases.
In struct tcpopt rename to_requested_s_scale to just to_wscale.
Add a comment with quote from RFC1323: "The Window field in a SYN (i.e.,
a <SYN> or <SYN,ACK>) segment itself is never scaled."
Reviewed by: silby, mohans, julian
Sponsored by: TCP/IP Optimization Fundraise 2005
- moved away from ifn/ifa access to sctp_ifa/sctp_ifn
built and managed by the add-ip code.
- cleaned up add-ip code to use the iterator
- made iterator be a thread, which enables auto-asconf now.
- rewrote and cleaned up source address selection (also
made it use new structures).
- Fixed a couple of memory leaks.
- DACK now settable as to how many packets to delay as
well as time.
- connectx() to latest socket API, new associd arg.
- Fixed issue with revoking and loosing potential to
send when we inflate the flight size. We now inflate
the cwnd too and deflate it later when the revoked
chunk is sent or acked.
- Got rid of some temp debug code
- src addr selection moved to a common file (sctp_output.c)
- Support for simple VRF's (we have support for multi-vfr
via compile switch that is scrubbed from BSD but we won't
need multi-vrf until we first get VRF :-D)
- Rest of mib work for address information now done
- Limit number of addresses in INIT/INIT-ACK to
a #def (30).
Reviewed by: gnn
* To use this option with a UDP socket, it must be bound to a local port,
and INADDR_ANY, to disallow possible collisions with existing udp inpcbs
bound to the same port on other interfaces at send time.
* If the socket is bound to INADDR_ANY, specifying IP_SENDSRCADDR with
INADDR_ANY will be rejected as it is ambiguous.
* If the socket is bound to an address other than INADDR_ANY, specifying
IP_SENDSRCADDR with INADDR_ANY will be disallowed by in_pcbbind_setup().
Reviewed by: silence on -net
Tested with: src/tools/regression/netinet/ipbroadcast
MFC after: 4 days
in the field. In one situation, one end of the TCP connection sends
a back-to-back RST packet, with delayed ack, the last_ack_sent variable
has not been update yet. When tcp_insecure_rst is turned off, the code
treats the RST as invalid because last_ack_sent instead of rcv_nxt is
compared against th_seq. Apparently there is some kind of firewall that
sits in between the two ends and that RST packet is the only RST
packet received. With short lived HTTP connections, the symptom is
a large accumulation of connections over a short period of time .
The +/-(1) factor is to take care of implementations out there that
generate RST packets with these types of sequence numbers. This
behavior has also been observed in live environments.
Reviewed by: silby, Mike Karels
MFC after: 1 week
been set at the socket layer, in our somewhat convoluted IPv4 source
selection logic in ip_output().
IP_ONESBCAST is actually a special case of SO_DONTROUTE, as 255.255.255.255
must always be delivered on a local link with a TTL of 1.
If IP_ONESBCAST has been set at the socket layer, also perform destination
interface lookup for point-to-point interfaces based on the destination
address of the link; previously it was not possible to use the option with
such interfaces; also, the destination/broadcast address fields map to the
same field within struct ifnet, which doesn't help matters.
One more valid fix going forward for these issues is to treat 255.255.255.255
as a destination in its own right in the forwarding trie. Other
implementations do this. It fits with the use of multiple paths, though
it then becomes necessary to specify interface preference.
This hack will eventually go away when that comes to pass.
Reviewed by: andre
MFC after: 1 week
potential issues where the peer does not close, potentially leaving
thousands of connections in FIN_WAIT_2. This is controlled by a new sysctl
fast_finwait2_recycle, which is disabled by default.
Reviewed by: gnn, silby.
It is built in the same module as IPv4 multicast forwarding, i.e. ip_mroute.ko,
if and only if IPv6 support is enabled for loadable modules.
Export IPv6 forwarding structs to userland netstat(1) via sysctl(9).
- ZONE get now also take a type cast so it does the
cast like mtod does.
- New macro SCTP_LIST_EMPTY, which in bsd is just
LIST_EMPTY
- Removal of const in some of the static hmac functions
(not needed)
- Store length changes to allow for new fields in auth
- Auth code updated to current draft (this should be the
RFC version we think).
- use uint8_t instead of u_char in LOOPBACK address comparison
- Some u_int32_t converted to uint32_t (in crc code)
- A bug was found in the mib counts for ordered/unordered
count, this was fixed (was referencing a freed mbuf).
- SCTP_ASOCLOG_OF_TSNS added (code will probably disappear
after my testing completes. It allows us to keep a
small log on each assoc of the last 40 TSN's in/out and
stream assignment. It is NOT in options and so is only
good for private builds.
- Some CMT changes in prep for Jana fixing his problem
with reneging when CMT is enabled (Concurrent Multipath
Transfer = CMT).
- Some missing mib stats added.
- Correction to number of open assoc's count in mib
- Correction to os_bsd.h to get right sha2 macros
- Add of special AUTH_04 flags so you can compile the code
with the old format (in case the peer does not yet support
the latest auth code).
- Nonce sum was incorrectly being set in when ecn_nonce was
NOT on.
- LOR in listen with implicit bind found and fixed.
- Moved away from using mbuf's for socket options to using
just data pointers. The mbufs were used to harmonize
NetBSD code since both Net and Open used this method. We
have decided to move away from that and more conform to
FreeBSD style (which makes more sense).
- Very very nasty bug found in some of my "debug" code. The
cookie_how collision case tracking had an endless loop in
it if you got a second retransmission of a cookie collision
case. This would lock up a CPU .. ugly..
- auth function goes to using size_t instead of int which
conforms to socketapi better
- Found the nasty bug that happens after 9 days of testing.. you
get the data chunk, deliver it and due to the reference to a ch->
that every now and then has been deleted (depending on the postion
in the mbuf) you have an invalid ch->ch.flags.. and thus you don't
advance the stream sequence number.. so you block the stream
permanently. The fix is to make local variables of these guys
and set them up before you have any chance of trimming the
mbuf.
- style fix in sctp_util.h, not sure how this got bad maybe in
the last patch? (aka it may not be in the real source).
- Found interesting bug when using the extended snd/rcv info where
we would get an error on receiving with this. Thats because
it was NOT padded to the same size as the snd_rcv info. We
increase (add the pad) so the two structs are the same size
in sctp_uio.h
- In sctp_usrreq.c one of the most common things we did for
socket options was to cast the pointer and validate the size.
This as been macro-ized to help make the code more readable.
- in sctputil.c two things, the socketapi class found a missing
flag type (the next msg is a notification) and a missing
scope recovery was also fixed.
Reviewed by: gnn
IGMPMSG_WHOLEPKT notifications to the userland PIM routing daemon,
as an optimization to mitigate the effects of high multicast
forwarding load.
This is an experimental change, therefore it must be explicitly enabled by
setting the sysctl/tunable net.inet.pim.squelch_wholepkt to a non-zero value.
The tunable may be set from the loader or from within the kernel environment
when loading ip_mroute.ko as a module.
Submitted by: edrt <edrt at citiz.net>
See also: http://mailman.icsi.berkeley.edu/pipermail/xorp-users/2005-June/000639.html
Make PIM dynamically loadable by using encap_attach_func().
PIM may now be loaded into a GENERIC kernel.
Tested with: ports/net/pimdd && tcpreplay && wireshark
Reviewed by: Pavlin Radoslavov
by the token bucket filter will result in EINVAL being returned.
If you want to rate-limit traffic in future, use ALTQ or dummynet; this
isn't a general purpose QoS engine.
Preserve the now unused fields in struct vif so as to avoid having to
recompile netstat(1) and other tools.
Reviewed by: Pavlin Radslavov, Bill Fenner
never used them; with mrouted, their functionality may be replaced by
explicitly configuring gif(4) instances and specifying them with the
'phyint' keyword.
Bump __FreeBSD_version to 700030, and update UPDATING.
A doc update is forthcoming.
Discussed on: net
Reviewed by: fenner
MFC after: 3 months
socket option TCP_INFO.
Note that the units used in the original Linux API are in microseconds,
so use a 64-bit mantissa to convert FreeBSD's internal measurements
from struct tcpcb from ticks.
multicast memberships, when interface is detached. Thus, when
an underlying interface is detached, we do not need to free
our multicast memberships.
Reviewed by: bms
Normally the socket buffers are static (either derived from global
defaults or set with setsockopt) and do not adapt to real network
conditions. Two things happen: a) your socket buffers are too small
and you can't reach the full potential of the network between both
hosts; b) your socket buffers are too big and you waste a lot of
kernel memory for data just sitting around.
With automatic TCP send and receive socket buffers we can start with a
small buffer and quickly grow it in parallel with the TCP congestion
window to match real network conditions.
FreeBSD has a default 32K send socket buffer. This supports a maximal
transfer rate of only slightly more than 2Mbit/s on a 100ms RTT
trans-continental link. Or at 200ms just above 1Mbit/s. With TCP send
buffer auto scaling and the default values below it supports 20Mbit/s
at 100ms and 10Mbit/s at 200ms. That's an improvement of factor 10, or
1000%. For the receive side it looks slightly better with a default of
64K buffer size.
New sysctls are:
net.inet.tcp.sendbuf_auto=1 (enabled)
net.inet.tcp.sendbuf_inc=8192 (8K, step size)
net.inet.tcp.sendbuf_max=262144 (256K, growth limit)
net.inet.tcp.recvbuf_auto=1 (enabled)
net.inet.tcp.recvbuf_inc=16384 (16K, step size)
net.inet.tcp.recvbuf_max=262144 (256K, growth limit)
Tested by: many (on HEAD and RELENG_6)
Approved by: re
MFC after: 1 month
upper-bounding it to the size of the initial socket buffer lower-bound it
to the smallest MSS we accept. Ideally we'd use the actual MSS information
here but it is not available yet.
For socket buffer auto sizing to be effective we need room to grow the
receive window. The window scale shift is determined at connection setup
and can't be changed afterwards. The previous, original, method effectively
just did a power of two roundup of the socket buffer size at connection
setup severely limiting the headroom for larger socket buffers.
Tested by: many (as part of the socket buffer auto sizing patch)
MFC after: 1 month
This is not a functional change.
IN_LINKLOCAL() tests if an address falls within the IPv4 link-local prefix.
IN_PRIVATE() tests if an address falls within an RFC 1918 private prefix.
IN_LOCAL_GROUP() tests if an address falls within the statically assigned
link-local multicast scope specified in RFC 2365.
IN_ANY_LOCAL() tests for either of IN_LINKLOCAL() or IN_LOCAL_GROUP().
As with the existing macros in the FreeBSD netinet stack, comparisons
are performed in host-byte order.
See also: RFC 1918, RFC 2365, RFC 3927
Obtained from: NetBSD (dyoung@)
MFC after: 2 weeks
carp_clone_destroy() we are on a safe side, we don't need to
unlock the cif, that can me already non-existent at this point.
Reported by: Anton Yuzhaninov <citrin rambler-co.ru>
- Finally all splxx() are removed
- Count error fixed in mapping array which might
cause a wrong cumack generation.
- Invariants around panic for case D + printf when no invariants.
- one-to-one model race condition fixed by using
a pre-formed connection and then completing the
work so accept won't happen on a non-formed
association.
- Some additional paranoia checks in sctp_output.
- Locks that were missing in the accept code.
Approved by: gnn
- Added a short time wait (not used yet) constant
- Corrected the type of the crc32c table (it was
unsigned long and really is a uint32_t
- Got rid of the user of MHeaders until they
are truely needed by lower layers.
- Fixed an initialization problem in the readq structure
(ordering was off).
- Found yet another collision bug when the random number
generator returns two numbers on one side (during a collision)
that are the same. Also added some tracking of cookies
that will go away when we know that we have the last collision
bug gone.
- Fixed an init bug for book_size_scale, that was causing
Early FR code to run when it should not.
- Fixed a flight size tracking bug that was associated with
Early FR but due to above bug also effected all FR's
- Fixed it so Max Burst also will apply to Fast Retransmit.
- Fixed a bug in the temporary logging code that allowed a
static log array overflow
- hashinit_flags is now used.
- Two last mcopym's were converted to the macro sctp_m_copym that
has always been used by all other places
- macro sctp_m_copym was converted to upper case.
- We now validate sinfo_flags on input (we did not before).
- Fixed a bug that prevented a user from sending data and immediately
shuting down with one send operation.
- Moved to use hashdestroy instead of free() in our macros.
- Fixed an init problem in our timed_wait vtag where we
did not fully initialize our time-wait blocks.
- Timer stops were re-positioned.
- A pcb cleanup method was added, however this probably will
not be used in BSD.. unless we make module loadable protocols
- I think this fixes the mysterious timer bug.. it was a
ordering of locks problem in the way we did timers. It
now conforms to the timeout(9) manual (except for the
_drain part, we had to do this a different way due
to locks).
- Fixed error return code so we get either CONNREUSED or CONNRESET
depending on where one is in progression
- Purged an unused clone macro.
- Fixed a read erro code issue where we were NOT getting the proper
error when the connection was reset.
- Purged an unused clone macro.
- Fixed a read erro code issue where we were NOT getting the proper
error when the connection was reset.
Approved by: gnn
members right. However, it also said it was aligned(1), which meant
that gcc generated really bad code. Mark this as aligned(4). This
makes things a little faster on arm (a couple percent), but also saves
about 30k on the size of the kernel for arm.
I talked about doing this with bde, but didn't check with him before
the commit, so I'm hesitant say 'reviewed by: bde'.
mbuf. First moves toward being able to cope better with having layer 2 (or
other encapsulation data) before the IP header in the packet being examined.
More commits to come to round out this functionality. This commit should
have no practical effect but clears the way for what is coming.
Revirewed by: luigi, yar
MFC After: 2 weeks
With the second (and last) part of my previous Summer of Code work, we get:
-ipfw's in kernel nat
-redirect_* and LSNAT support
General information about nat syntax and some examples are available
in the ipfw (8) man page. The redirect and LSNAT syntax are identical
to natd, so please refer to natd (8) man page.
To enable in kernel nat in rc.conf, two options were added:
o firewall_nat_enable: equivalent to natd_enable
o firewall_nat_interface: equivalent to natd_interface
Remember to set net.inet.ip.fw.one_pass to 0, if you want the packet
to continue being checked by the firewall ruleset after being
(de)aliased.
NOTA BENE: due to some problems with libalias architecture, in kernel
nat won't work with TSO enabled nic, thus you have to disable TSO via
ifconfig (ifconfig foo0 -tso).
Approved by: glebius (mentor)
access plus timers. This makes the code
more portable and able to change out the
mbuf or timer system used more easily ;-)
b) removal of all use of pkt-hdr's until only
the places we need them (before ip_output routines).
c) remove a bunch of code not needed due to <b> aka
worrying about pkthdr's :-)
d) There was one last reorder problem it looks where
if a restart occur's and we release and relock (at
the point where we setup our alias vtag) we would
end up possibly getting the wrong TSN in place. The
code that fixed the TSN's just needed to be shifted
around BEFORE the release of the lock.. also code that
set the state (since this also could contribute).
Approved by: gnn
o fixed a comment
o made in kernel libalias a bit less verbose (disabled automatic
logging everytime a new link is added or deleted)
Approved by: glebius (mentor)
2) Fix all "magic numbers" to be constants.
3) A collision case that would generate two associations to
the same peer due to a missing lock is fixed.
4) Added tracking of where timers are stopped.
Approved by: gnn
kernel. This LOR snuck in with some of the recent syncache changes. To
fix this, the inpcb handling was changed:
- Hang a MAC label off the syncache object
- When the syncache entry is initially created, we pickup the PCB lock
is held because we extract information from it while initializing the
syncache entry. While we do this, copy the MAC label associated with
the PCB and use it for the syncache entry.
- When the packet is transmitted, copy the label from the syncache entry
to the mbuf so it can be processed by security policies which analyze
mbuf labels.
This change required that the MAC framework be extended to support the
label copy operations from the PCB to the syncache entry, and then from
the syncache entry to the mbuf.
These functions really should be referencing the syncache structure instead
of the label. However, due to some of the complexities associated with
exposing this syncache structure we operate directly on it's label pointer.
This should be OK since we aren't making any access control decisions within
this code directly, we are merely allocating and copying label storage so
we can properly initialize mbuf labels for any packets the syncache code
might create.
This also has a nice side effect of caching. Prior to this change, the
PCB would be looked up/locked for each packet transmitted. Now the label
is cached at the time the syncache entry is initialized.
Submitted by: andre [1]
Discussed with: rwatson
[1] andre submitted the tcp_syncache.c changes
for printing/logging ipv6 addresses.
The caller now has to hand in a sufficiently large buffer as first
argument.
This is the "+ one more change" missed in the original commit.
Noticed by: tinderbox
Pointy hat to: me (#1)
In ip6_sprintf no longer use and return one of eight static buffers
for printing/logging ipv6 addresses.
The caller now has to hand in a sufficiently large buffer as first
argument.
Fixing the IP accounting issue, if we plan to do so, needs to be better
thought out; the 'fix' introduces a hash lookup and a possible kernel panic.
Reported by: Mark Tinguely
marked INP_DROPPED or INP_TIMEWAIT:
o return ECONNRESET instead of EINVAL for close, disconnect, shutdown,
rcvd, rcvoob, and send operations
o return ECONNABORTED instead of EINVAL for accept
These changes should reduce confusion in applications since EINVAL is
normally interpreted to mean an invalid file descriptor. This change
does not conflict with POSIX or other standards I checked. The return
of EINVAL has always been possible but rare; it's become more common
with recent changes to the socket/inpcb handling and with finer-grained
locking and preemption.
Note: there are other instances of EINVAL for this state that were
left unchanged; they should be reviewed.
Reviewed by: rwatson, andre, ru
MFC after: 1 month
We are not yet aware of the protocol internals but this way
SCTP traffic over v6 will not be discarded.
Reported by: Peter Lei via rrs
Tested by: Peter Lei <peterlei cisco.com>
not being aquired. This meant that when we cleanup
the outbound we may have one in transit to be
added with the old sequence number. This is bad
since then we loose a message :(
Also the report_outbound needed to have the right
lock when its called which it did not.. I added
the lock with of course a flag since we want to
have the lock before we call it in the restart
case.
This also fixed the FIX ME case where, in the cookie
collision case, we mark for retransmit any that
were bundled with the cookie that was dropped.
This also means changes to the output routine
so we can assure getting the COOKIE-ACK sent
BEFORE we retransmit the Data.
Approved by: gnn
a colliding INIT. This if fine except when we have
data outstanding... we basically reset it to the
previous value it was.. so then we end up assigning
the same TSN to two different data chunks.
This patch:
1) Finds a missing lock for when we change the stream
numbers during COOKIE and INIT-ACK processing.. we
were NOT locking the send_buffer.. which COULD cause
problems (found by inspection looking for <2>)
2) Fixes a case during a colliding INIT where we incorrectly
reset the sending Sequence thus in some cases duplicately
assigning a TSN.
3) Additional enhancments to logging so we can see strm/tsn in
the receiver AND new tracking to watch what the sender
is doing with TSN and STRM seq's.
Approved by: gnn
We were calling select_a_tag() inside sctp_send_initate_ack().
During collision cases we have a stcb and thus a SCTP_LOCK. When
we call select_a_tag it (below it) locks the INFO lock. We now
1) pre-select the nonce-tie-tags in sctputil.c during setup of
a tcb.
2) In the other case where we have to select tags, we unlock after
incr the ref cnt (so assoc won't go away0 and then do the
tag selection followed by a relock and decr the refcnt.
Approved by: gnn
reset comes in we need to calculate the length and
therefore the number of listed streams (if any) based
on the TLV type. Otherwise if we get a retran we could
in theory panic by sending a notification to a user with
a incorrect list and thus no memory listing the streams.
Found in IOS by devtest :-)
Approved by: gnn
copy's were incorrect and so was the locking.
-A bug was also found that would create a race and
panic when an abort arrived on a socket being read
from.
-Also fix the reader to get MSG_TRUNC when a partial
delivery is aborted.
-Also addresses a couple of coverity caught error path
memory leaks and a couple of other valid complaints
Approved by: gnn
patch was prepared and committed to priv(9) calls. Add XXX comments
as, in each case, the semantics appear to differ from the TCP/UDP
versions of the calls with respect to jail, and because cr_canseecred()
is not used to validate the query.
Obtained from: TrustedBSD Project
to wakeup on close of the sender. It basically moves
the return (when the asoc has a reader/writer) further
down and gets the wakeup and assoc appending (of the
PD-API event) moved up before the return. It also
moves the flag set right before the return so we can
assure only once adding the PD-API events.
Approved by: gnn
specific privilege names to a broad range of privileges. These may
require some future tweaking.
Sponsored by: nCircle Network Security, Inc.
Obtained from: TrustedBSD Project
Discussed on: arch@
Reviewed (at least in part) by: mlaier, jmg, pjd, bde, ceri,
Alex Lyashkov <umka at sevcity dot net>,
Skip Ford <skip dot ford at verizon dot net>,
Antoine Brodin <antoine dot brodin at laposte dot net>
This also moves two 16 bit int's to become 32 bit
values so we do not have to use atomic_add_16.
Most of the changes are %p, casts and other various
nasty's that were in the orignal code base. With this
commit my machine will now do a build universe.. however
I as yet have not tested on a 64bit machine .. it may not work :-(
inserted a few to the new files.. but I falied to
add the #include <sys/cdef.h>
Which causes a compile error.. sorry about that... got it
now :-)
Approved by:gnn
work is not just mine, but it is also the works of Peter Lei
and Michael Tuexen. They both are my two key other developers
working on the project.. and they need ata-boy's too:
****
peterlei@cisco.comtuexen@fh-muenster.de
****
I did do a make sysent which updated the
syscall's and sysproto.. I hope that is correct... without
it you don't build since we have new syscalls for SCTP :-0
So go out and look at the NOTES, add
option SCTP (make sure inet and inet6 are present too)
and play with SCTP.
I will see about comitting some test tools I have after I
figure out where I should place them. I also have a
lib (libsctp.a) that adds some of the missing socketapi
functions that I need to put into lib's.. I will talk
to George about this :-)
There may still be some 64 bit issues in here, none of
us have a 64 bit processor to test with yet.. Michael
may have a MAC but thats another beast too..
If you have a mac and want to use SCTP contact Michael
he maintains a web site with a loadable module with
this code :-)
Reviewed by: gnn
Approved by: gnn
- Pay respect to net.isr.direct: use netisr_dispatch() instead of ip_input()
Reviewed by: glebius, rwatson
- purge_flow_set():
- Do not leak memory while purging queues which are not bound to pipe.
- style(9) cleanup
MFC after: 2 months
net.inet.ip.dummynet.curr_time
net.inet.ip.dummynet.searches
net.inet.ip.dummynet.search_steps
to SYSCTL_LONG nodes. It will prevent frequent wrap around on 64bit archs.
- Implement simple mechanics for dummynet(4) internal time correction.
Under certain circumstances (system high load, dummynet lock contention, etc)
dummynet's tick counter can be significantly slower than it should be.
(I've observed up to 25% difference on one of my production servers).
Since this counter used for packet scheduling, it's accuracy is vital for
precise bandwidth limitation.
Introduce new sysctl nodes:
net.inet.ip.dummynet.
tick_lost - number of ticks coalesced by taskqueue thread.
tick_adjustment - number of time corrections done.
tick_diff - adjusted vs non-adjusted tick counter difference
tick_delta - last vs 'standard' tick differnece (usec).
tick_delta_sum - accumulated (and not corrected yet) time
difference (usec).
Reviewed by: glebius
MFC after: 2 month
begun with a repo-copy of mac.h to mac_framework.h. sys/mac.h now
contains the userspace and user<->kernel API and definitions, with all
in-kernel interfaces moved to mac_framework.h, which is now included
across most of the kernel instead.
This change is the first step in a larger cleanup and sweep of MAC
Framework interfaces in the kernel, and will not be MFC'd.
Obtained from: TrustedBSD Project
Sponsored by: SPARTA
entries' by src:port and dst:port pairs. IPv6 part is non-functional
as ``limit'' does not support IPv6 flows.
PR: kern/103967
Submitted by: based on Bruce Campbell patch
MFC after: 1 month
in ip6_output. In case this fails handle the error directly and log it[1].
In addition permit CARP over v6 in ip_fw2.
PR: kern/98622
Similar patch by: suz
Discussed with: glebius [1]
Tested by: Paul.Dekkers surfnet.nl, Philippe.Pegon crc.u-strasbg.fr
MFC after: 3 days
scale it to min(ephemeral port range / 2, maxsockets / 5) so that people
with large gobs of memory and/or large maxsockets settings will not
exhaust their entire ephemeral port range with sockets in the TIME_WAIT
state during periods of heavy load.
Those who wish to tweak the size of the TIME_WAIT zone can still do so with
net.inet.tcp.maxtcptw.
Reviewed by: glebius, ru
of its internal state to ignore the failed send and try again a bit later.
If the error is EPERM the packet got blocked by the local firewall and the
revert may cause the session to get stuck and retry indefinitely. This way
we treat it like a packet loss and let the retransmit timer and timeouts
do their work over time.
The correct behavior is to drop a connection that gets an EPERM error.
However this _may_ introduce some POLA problems and a two commit approach
was chosen.
Discussed with: glebius
PR: kern/25986
PR: kern/102653
the MROUTER is running, the system would panic as described in the PR.
The fix in the PR is a good start, however, the other state associated
with the multicast forwarding cache has to be freed in order to avoid
leaking memory and other possible panics.
More care and attention is needed in this area.
PR: kern/82882
MFC after: 1 week
goes away. Without this change, it leaks in_multi (and often ether_multi
state) if many clonable interfaces are created and destroyed in quick
succession.
The concept of this fix is borrowed from KAME. Detailed information about
this behaviour, as well as test cases, are available in the PR.
PR: kern/78227
MFC after: 1 week
With the first part of my previous Summer of Code work, we get:
-made libalias modular:
-support for 'particular' protocols (like ftp/irc/etcetc) is no more
hardcoded inside libalias, but it's available through external
modules loadable at runtime
-modules are available both in kernel (/boot/kernel/alias_*.ko) and
user land (/lib/libalias_*)
-protocols/applications modularized are: cuseeme, ftp, irc, nbt, pptp,
skinny and smedia
-added logging support for kernel side
-cleanup
After a buildworld, do a 'mergemaster -i' to install the file libalias.conf
in /etc or manually copy it.
During startup (and after every HUP signal) user land applications running
the new libalias will try to read a file in /etc called libalias.conf:
that file contains the list of modules to load.
User land applications affected by this commit are ppp and natd:
if libalias.conf is present in /etc you won't notice any difference.
The only kernel land bit affected by this commit is ng_nat:
if you are using ng_nat, and it doesn't correctly handle
ftp/irc/etcetc sessions anymore, remember to kldload
the correspondent module (i.e. kldload alias_ftp).
General information and details about the inner working are available
in the libalias man page under the section 'MODULAR ARCHITECTURE
(AND ipfw(4) SUPPORT)'.
NOTA BENE: this commit affects _ONLY_ libalias, ipfw in-kernel nat
support will be part of the next libalias-related commit.
Approved by: glebius
Reviewed by: glebius, ru
the VRRPv2 advertisements will originate from the wrong source address.
This only affects kernels compiled with MROUTING and after the MRT_INIT
ioctl() has been issued.
Set imo_multicast_vif in carp's softc to the invalid value -1 after it is
zeroed by softc allocation, to stop the ip_output() path looking up the
incorrect source address thinking a vif is set.
PR: kern/100532
Submitted by: Bohus Plucinsky
MFC after: 1 week
the header field for possible later IPSEC SPD lookup, even
when the kernel is built without 'options INET6'.
PR: kern/57760
MFC after: 1 week
Submitted by: Joachim Schueth
functionality:
- Remove a rwlock aquisition/release per generated syncookie. Locking
is now integrated with the bucket row locking of syncache itself and
syncookies no longer add any additional lock overhead.
- Syncookie secrets are different for and stored per syncache buck row.
Secrets expire after 16 seconds and are reseeded on-demand.
- The computational overhead for syncookie generation and verification
is one MD5 hash computation as before.
- Syncache can be turned off and run with syncookies only by setting the
sysctl net.inet.tcp.syncookies_only=1.
This implementation extends the orginal idea and first implementation
of FreeBSD by using not only the initial sequence number field to store
information but also the timestamp field if present. This way we can
keep track of the entire state we need to know to recreate the session in
its original form. Almost all TCP speakers implement RFC1323 timestamps
these days. For those that do not we still have to live with the known
shortcomings of the ISN only SYN cookies. The use of the timestamp field
causes the timestamps to be randomized if syncookies are enabled.
The idea of SYN cookies is to encode and include all necessary information
about the connection setup state within the SYN-ACK we send back and thus
to get along without keeping any local state until the ACK to the SYN-ACK
arrives (if ever). Everything we need to know should be available from
the information we encoded in the SYN-ACK.
A detailed description of the inner working of the syncookies mechanism
is included in the comments in tcp_syncache.c.
Reviewed by: silby (slightly earlier version)
Sponsored by: TCP/IP Optimization Fundraise 2005
exists to allow the mandatory access control policy to properly initialize
mbufs generated by the firewall. An example where this might happen is keep
alive packets, or ICMP error packets in response to other packets.
This takes care of kernel panics associated with un-initialize mbuf labels
when the firewall generates packets.
[1] I modified this patch from it's original version, the initial patch
introduced a number of entry points which were programmatically
equivalent. So I introduced only one. Instead, we should leverage
mac_create_mbuf_netlayer() which is used for similar situations,
an example being icmp_error()
This will minimize the impact associated with the MFC
Submitted by: mlaier [1]
MFC after: 1 week
This is a RELENG_6 candidate
validity of ro->ro_rt first. This prevents crashing on any non-normally
routed IP packet.
Coverity CID: 162 (incorrectly, it was re-introduced by previous commit)
support a network w/ split mtu's by assigning each host route the correct
mtu. an aspiring programmer could write a daemon to probe hosts and find
out if they support a larger mtu.
timeouts for TCP and T/TCP connections in the TIME_WAIT
state, and we had two separate timed wait queues for them.
Now that is has gone, the timeout is always 2*MSL again,
and there is no reason to keep two queues (the first was
unused anyway!).
Also, reimplement the remaining queue using a TAILQ (it
was technically impossible before, with two queues).
TSO is only used if we are in a pure bulk sending state. The presence of
TCP-MD5, SACK retransmits, SACK advertizements, IPSEC and IP options prevent
using TSO. With TSO the TCP header is the same (except for the sequence number)
for all generated packets. This makes it impossible to transmit any options
which vary per generated segment or packet.
The length of TSO bursts is limited to TCP_MAXWIN.
The sysctl net.inet.tcp.tso globally controls the use of TSO and is enabled.
TSO enabled sends originating from tcp_output() have the CSUM_TCP and CSUM_TSO
flags set, m_pkthdr.csum_data filled with the header pseudo-checksum and
m_pkthdr.tso_segsz set to the segment size (net payload size, not counting
IP+TCP headers or TCP options).
IPv6 currently lacks a pseudo-header checksum function and thus doesn't support
TSO yet.
Tested by: Jack Vogel <jfvogel-at-gmail.com>
Sponsored by: TCP/IP Optimization Fundraise 2005
o add IFCAP_TSO[46] for drivers to announce this capability for IPv4 and IPv6
o add CSUM_TSO flag to mbuf pkthdr csum_flags field
o add tso_segsz field to mbuf pkthdr
o enhance ip_output() packet length check to allow for large TSO packets
o extend tcp_maxmtu[46]() with a flag pointer to pass interface capabilities
o adjust all callers of tcp_maxmtu[46]() accordingly
Discussed on: -current, -net
Sponsored by: TCP/IP Optimization Fundraise 2005
and skip over the normal IP processing.
Add a supporting function ifa_ifwithbroadaddr() to verify and validate the
supplied subnet broadcast address.
PR: kern/99558
Tested by: Andrey V. Elsukov <bu7cher-at-yandex.ru>
Sponsored by: TCP/IP Optimization Fundraise 2005
MFC after: 3 days
bad under high load. For example with 40k sockets and 25k tcptw
entries, connect() syscall can run for seconds. Debugging showed
that it iterates the cycle millions times and purges thousands of
tcptw entries at a time.
Besides practical unusability this change is architecturally
wrong. First, in_pcblookup_local() is used in connect() and bind()
syscalls. No stale entries purging shouldn't be done here. Second,
it is a layering violation.
o Return back the tcptw purging cycle to tcp_timer_2msl_tw(),
that was removed in rev. 1.78 by rwatson. The commit log of this
revision tells nothing about the reason cycle was removed. Now
we need this cycle, since major cleaner of stale tcptw structures
is removed.
o Disable probably necessary, but now unused
tcp_twrecycleable() function.
Reviewed by: ru
for example:
fwd tablearg ip from any to table(1)
where table 1 has entries of the form:
1.1.1.0/24 10.2.3.4
208.23.2.0/24 router2
This allows trivial implementation of a secondary routing table implemented
in the firewall layer.
I expect more work (under discussion with Glebius) to follow this to clean
up some of the messy parts of ipfw related to tables.
Reviewed by: Glebius
MFC after: 1 month
in older versions of FreeBSD. This option is pointless as it is needed in just
about every interesting usage of forward that I have ever seen. It doesn't make
the system any safer and just wastes huge amounts of develper time
when the system doesn't behave as expected when code is moved from
4.x to 6.x It doesn't make
the system any safer and just wastes huge amounts of develper time
when the system doesn't behave as expected when code is moved from
4.x to 6.x or 7.x
Reviewed by: glebius
MFC after: 1 week
were unused or already in if_var.h so add if_name() to if_var.h and
remove net_osdep.h along with all references to it.
Longer term we may want to kill off if_name() entierly since all modern
BSDs have if_xname variables rendering it unnecessicary.
tcp_twstart(), but not to the other, tcp_detach(), as the socket is
already being torn down and therefore there are no listeners. This avoids
a panic if kqueue state is registered on the socket at close(), and
eliminates to XXX comments. There is one case remaining in which
tcp_discardcb() reaches up to the socket layer as part of the TCP host
cache, which would be good to avoid.
Reported by: Goran Gajic <ggajic at afrodita dot rcub dot bg dot ac dot yu>
function, pru_close, to notify protocols that the file descriptor or
other consumer of a socket is closing the socket. pru_abort is now a
notification of close also, and no longer detaches. pru_detach is no
longer used to notify of close, and will be called during socket
tear-down by sofree() when all references to a socket evaporate after
an earlier call to abort or close the socket. This means detach is now
an unconditional teardown of a socket, whereas previously sockets could
persist after detach of the protocol retained a reference.
This faciliates sharing mutexes between layers of the network stack as
the mutex is required during the checking and removal of references at
the head of sofree(). With this change, pru_detach can now assume that
the mutex will no longer be required by the socket layer after
completion, whereas before this was not necessarily true.
Reviewed by: gnn
( and where appropriate the destruction) of the pcb mutex to the init/finit
functions of the pcb zones.
This allows locking of the pcb entries and race condition free comparison
of the generation count.
Rearrange locking a bit to avoid extra locking operation to update the generation
count in in_pcballoc(). (in_pcballoc now returns the pcb locked)
I am planning to convert pcb list handling from a type safe to a reference count
model soon. ( As this allows really freeing the PCBs)
Reviewed by: rwatson@, mohans@
MFC after: 1 week
parameter that can specify configuration parameters:
o rev cloner api's to add optional parameter block
o add SIOCCREATE2 that accepts parameter data
o rev vlan support to use new api (maintain old code)
Reviewed by: arch@
except in places dealing with ifaddr creation or destruction; and
in such special places incomplete ifaddrs should never be linked
to system-wide data structures. Therefore we can eliminate all the
superfluous checks for "ifa->ifa_addr != NULL" and get ready
to the system crashing honestly instead of masking possible bugs.
Suggested by: glebius, jhb, ru
used since FreeBSD-SA-06:04.ipfw.
Adopt send_reject6 to what had been done for legacy IP: no longer
send or permit sending rejects for any but the first fragment.
Discussed with: oleg, csjp (some weeks ago)
o don't assign remote/local host/port information manually between provided
struct in_conninfo and struct syncache, bcopy() it instead
o rename sc_tsrecent to sc_tsreflect in struct syncache to better capture
the purpose of this field
o rename sc_request_r_scale to sc_requested_r_scale for ditto reasons
o fix IPSEC error case printf's to report correct function name
o in syncache_socket() only transpose enhanced tcp options parameters to
struct tcpcb when the inpcb doesn't has TF_NOOPT set
o in syncache_respond() reorder stack variables
o in syncache_respond() remove bogus KASSERT()
No functional changes.
Sponsored by: TCP/IP Optimization Fundraise 2005
