more adequate TCP performance with IPv6.
Changes for IPv4, r166403 and r172795, both ignored the
IPv6 counterpart and left it in the state of art of year 2000.
The same logic in syncache already shares code between v4 and v6 so
things do not need to be adapted there.
Reported by: Steinar Haug (sthaug nethelp.no)
Tested by: Steinar Haug (sthaug nethelp.no)
MFC after: 3 days
from existing functions for initializing global state.
At this stage, the new per-vnet initializer functions are
directly called from the existing global initialization code,
which should in most cases result in compiler inlining those
new functions, hence yielding a near-zero functional change.
Modify the existing initializer functions which are invoked via
protosw, like ip_init() et. al., to allow them to be invoked
multiple times, i.e. per each vnet. Global state, if any,
is initialized only if such functions are called within the
context of vnet0, which will be determined via the
IS_DEFAULT_VNET(curvnet) check (currently always true).
While here, V_irtualize a few remaining global UMA zones
used by net/netinet/netipsec networking code. While it is
not yet clear to me or anybody else whether this is the right
thing to do, at this stage this makes the code more readable,
and makes it easier to track uncollected UMA-zone-backed
objects on vnet removal. In the long run, it's quite possible
that some form of shared use of UMA zone pools among multiple
vnets should be considered.
Bump __FreeBSD_version due to changes in layout of structs
vnet_ipfw, vnet_inet and vnet_net.
Approved by: julian (mentor)
in the case where a single mbuf is allocated due to
m_getcl() returning NULL, we already call MH_ALIGN,
so do not increment m->m_data in this case.
Found during MLDv2 port.
- PR-SCTP had major issues when skipping through a multi-part message.
o Did not look at socket buffer.
o Did not properly handle the reassmebly queue.
o The MARKED segments could interfere and un-skip a chunk causing
a problem with the proper FWD-TSN.
o No FR of FWD-TSN's was being done.
- NR-Sack code was basically disabled. It needed fixes that
never got into the real code.
- CMT code had issues when the two paths were NOT the same b/w. We
found a few small bugs, but also the critcal one here was not
dividing the rwnd amongst the paths.
Obtained from: Michael Tuexen and myself at the IETF hack-fest ;-)
they were passed uninitialized to in6_pcblookup_hash. Instead, do as is done
for IPv4 and use the addresses within the sockaddr structure, which are
correctly populated.
This fixes tcpdrop(8) for IPv6 address pairs.
Reviewed by: bz
This is purely a forwarding plane cleanup; no control plane
code is involved.
Summary:
* Split IPv4 and IPv6 MROUTING support. The static compile-time
kernel option remains the same, however, the modules may now
be built for IPv4 and IPv6 separately as ip_mroute_mod and
ip6_mroute_mod.
* Clean up the IPv4 multicast forwarding code to use BSD queue
and hash table constructs. Don't build our own timer abstractions
when ratecheck() and timevalclear() etc will do.
* Expose the multicast forwarding cache (MFC) and virtual interface
table (VIF) as sysctls, to reduce netstat's dependence on libkvm
for this information for running kernels.
* bandwidth meters however still require libkvm.
* Make the MFC hash table size a boot/load-time tunable ULONG,
net.inet.ip.mfchashsize (defaults to 256).
* Remove unused members from struct vif and struct mfc.
* Kill RSVP support, as no current RSVP implementation uses it.
These stubs could be moved to raw_ip.c.
* Don't share locks or initialization between IPv4 and IPv6.
* Don't use a static struct route_in6 in ip6_mroute.c.
The v6 code is still using a cached struct route_in6, this is
moved to mif6 for the time being.
* More cleanup remains to be merged from ip_mroute.c to ip6_mroute.c.
v4 path tested using ports/net/mcast-tools.
v6 changes are mostly mechanical locking and *have not* been tested.
As these changes partially break some kernel ABIs, they will not
be MFCed. There is a lot more work to be done here.
Reviewed by: Pavlin Radoslavov
any IPv4 multicast operations which reference it.
There is a potential race because ifma_protospec is set to NULL
when we discover the underlying ifnet has gone away. This write
is not covered by the IF_ADDR_LOCK, and it's difficult to widen
its scope without making it a recursive lock. It isn't clear why
this manifests more quickly with 802.11 interfaces, but does not
seem to manifest at all with wired interfaces.
With this change, the 802.11 related panics reported by sam@
and cokane@ should go away. It is not the right fix, that requires
more thought before 8.0.
Idea from: sam
Tested by: cokane
in FreeBSD 5.x to allow network device drivers to run with Giant
despite the network stack being Giant-free. This significantly
simplifies calls into ioctl() on network interfaces, especially
in the multicast code, as well as eliminates deferred invocation
of interface if_start routines.
Disable the build on device drivers still depending on
IFF_NEEDSGIANT as they no longer compile. They will be removed
in a few weeks if they haven't been made MPSAFE in that time.
Disabled drivers:
if_ar
if_axe
if_aue
if_cdce
if_cue
if_kue
if_ray
if_rue
if_rum
if_sr
if_udav
if_ural
if_zyd
Drivers that were already disabled because of tty changes:
if_ppp
if_sl
Discussed on: arch@
certain flags that should have been in inp_flags ended up in inp_vflag,
meaning that they were inconsistently locked, and in one case,
interpreted. Move the following flags from inp_vflag to gaps in the
inp_flags space (and clean up the inp_flags constants to make gaps
more obvious to future takers):
INP_TIMEWAIT
INP_SOCKREF
INP_ONESBCAST
INP_DROPPED
Some aspects of this change have no effect on kernel ABI at all, as these
are UDP/TCP/IP-internal uses; however, netstat and sockstat detect
INP_TIMEWAIT when listing TCP sockets, so any MFC will need to take this
into account.
MFC after: 1 week (or after dependencies are MFC'd)
Reviewed by: bz
- When sending large PR-SCTP messages over a
lossy link we would incorrectly calculate the fwd-tsn
- When receiving large multipart pr-sctp packets we would
incorrectly send back a SACK that would renege improperly
on already received packets thus causing unneeded retransmissions.
or not the inpcb is currenty on various hash lookup lists, rather
than using (lport != 0) to detect this. This means that the full
4-tuple of a connection can be retained after close, which should
lead to more sensible netstat output in the window between TCP
close and socket close.
MFC after: 2 weeks
in6p_ip6_nxt
in6p_vflag
in6p_flags
in6p_socket
in6p_lport
in6p_fport
in6p_ppcb
Remove unused v6 macro aliases for inpcb flags:
IN6P_HIGHPORT
IN6P_LOWPORT
IN6P_ANONPORT
IN6P_RECVIF
IN6P_MTUDISC
IN6P_FAITH
IN6P_CONTROLOPTS
References to in6p_lport and in6_fport in sockstat are also replaced with
normal inp_lport and inp_fport references.
MFC after: 3 days
Reviewed by: bz
IPv4 stack.
Diffs are minimized against p4.
PCS has been used for some protocol verification, more widespread
testing of recorded sources in Group-and-Source queries is needed.
sizeof(struct igmpstat) has changed.
__FreeBSD_version is bumped to 800070.
1) WP should never be marked unless flight size is 0
2) When recovering from wp if the peer ack's it we don't mark for retran
3) When recovering, we must assure a timer is still running.
into the advance_peer_ack point so we would incorrectly
send a wrong value in the FWD-TSN
- PR-SCTP bug, where an PR packet is used for a window
probe which could incorrectly get the packet moved
back into the send_queue, which will cause major issues and
should not happen.
- Fix a trace to use the proper macro.
and do not attempt to perform a group lookup.
This is a socket layer lock, and the bottom half of IP
really has no business taking it.
Use the value of the in_mcast_loop sysctl to determine
if we should loop back by default, in the absence of
any multicast socket options. Because the check on
group membership is now deferred to the input path,
an m_copym() is now required.
This should increase multicast send performance where the
source has not requested loopback, although this has not been
benchmarked or measured.
It is also a necessary change for IN_MULTI_LOCK to become
non-recursive, which is required in order to implement IGMPv3
in a thread-safe way.
IPv4 multicast sends are looped back to senders by default
on a stack-wide basis, rather than relying on the socket option.
Note that the sysctl only applies to newly created multicast sockets.
RH0 was deprecated by RFC 5095.
While most of the code had been disabled by #if 0 already, leave a
bit of infrastructure for possible RH2 code and a log message under
BURN_BRIDGES in case a user still tries to send RH0 packets.
Reviewed by: gnn (a bit back, earlier version)
as a handler.
The variable was exported only for debugging, but there is little reason
to do it now that the timekeeping is supported by various other variables.
For the time being just comment out the sysctl, but I think this
should go away.
of sysctl_variables.
I would also remove it from the VNET record but I am unsure if
there is any ABI issue -- so for the time being just mark it as
unused in ip_fw.h, and then we will collect the garbage at some
appropriate time in the future.
MFC after: 3 days
which are not in a module of their own like gif.
Single kernel compiles and universe will fail if the size of the struct
changes. Th expected values are given in sys/vimage.h.
See the comments where how to handle this.
Requested by: peter
- Fix the copy, we can't do a blind copy but must transfer
the data from the old to the new.
- Fix the ACK processing so we properly stop retransmitting
the thing.
- Fix it so if we get a retran we will properly reply with
the saved response without doing anything.
MFC after: 1 month
net/route.h.
Remove the hidden include of opt_route.h and net/route.h from net/vnet.h.
We need to make sure that both opt_route.h and net/route.h are included
before net/vnet.h because of the way MRT figures out the number of FIBs
from the kernel option. If we do not, we end up with the default number
of 1 when including net/vnet.h and array sizes are wrong.
This does not change the list of files which depend on opt_route.h
but we can identify them now more easily.
checks for the tcpcb, previously used to detect complete disconnection,
with INP_DROPPED checks. Correct that, preventing shutdown() from
improperly generating a TCP segment with destination IP and port of
0.0.0.0:0.
PR: kern/132050
Reported by: david gueluy <david.gueluy at netasq.com>
MFC after: 3 weeks
to in_rtrequest(); the radix head lock is already acquired before
rnh_walktree is called in in_rtqtimo_one(). This avoids a recursive
acquisition that is no longer permitted in 8.x due to use of an rwlock
for the radix head lock.
Reported by: dikshie <dikshie at gmail.com>
MFC after: 3 days
longer do we require SCTP to be in the kernel for the
lib to be able to handle SCTP. We do this by moving
the CRC32c checksum into libkern/crc32.c and then adjusting
all routines to use the common methods. Note that this
will improve the performance of iSCSI since they were
using the old single 256 bit table lookup versus the
slicing 8 algorithm (which gives a 4x speed up in
CRC32c calculation :-D)
Reviewed by:rwatson, gnn, scottl, paolo
MFC after: 4 week? (assuming we MFC the alias_sctp changes)
Add a note next to fields in network format.
The n_* types are not enough for compiler checks on endianness, and their
use often requires an otherwise unnecessary #include <netinet/in_systm.h>
The typedef in in_systm.h are still there.
(duplicate) code in sys/netipsec/ipsec.c and fold it into
common, INET/6 independent functions.
The file local functions ipsec4_setspidx_inpcb() and
ipsec6_setspidx_inpcb() were 1:1 identical after the change
in r186528. Rename to ipsec_setspidx_inpcb() and remove the
duplicate.
Public functions ipsec[46]_get_policy() were 1:1 identical.
Remove one copy and merge in the factored out code from
ipsec_get_policy() into the other. The public function left
is now called ipsec_get_policy() and callers were adapted.
Public functions ipsec[46]_set_policy() were 1:1 identical.
Rename file local ipsec_set_policy() function to
ipsec_set_policy_internal().
Remove one copy of the public functions, rename the other
to ipsec_set_policy() and adapt callers.
Public functions ipsec[46]_hdrsiz() were logically identical
(ignoring one questionable assert in the v6 version).
Rename the file local ipsec_hdrsiz() to ipsec_hdrsiz_internal(),
the public function to ipsec_hdrsiz(), remove the duplicate
copy and adapt the callers.
The v6 version had been unused anyway. Cleanup comments.
Public functions ipsec[46]_in_reject() were logically identical
apart from statistics. Move the common code into a file local
ipsec46_in_reject() leaving vimage+statistics in small AF specific
wrapper functions. Note: unfortunately we already have a public
ipsec_in_reject().
Reviewed by: sam
Discussed with: rwatson (renaming to *_internal)
MFC after: 26 days
X-MFC: keep wrapper functions for public symbols?
return zero on success and an error code otherwise. The possible errors
are EADDRNOTAVAIL if an address being checked for doesn't match the
prison, and EAFNOSUPPORT if the prison doesn't have any addresses in
that address family. For most callers of these functions, use the
returned error code instead of e.g. a hard-coded EADDRNOTAVAIL or
EINVAL.
Always include a jailed() check in these functions, where a non-jailed
cred always returns success (and makes no changes). Remove the explicit
jailed() checks that preceded many of the function calls.
Approved by: bz (mentor)
is messing with the UDP tunnel. This means
that if two users actually tried to change the
tunnel port at the same time interesting things COULD
result, but its probably very unlikely to happen :-)
- Prepare for CRC offloading, add MIB counters (RS/MT).
- Bugfix: Disable CRC computation for IPv6 addresses with local scope (MT).
- Bugfix: Handle close() with SO_LINGER correctly when notifications
are generated during the close() call(MT).
- Bugfix: Generate DRY event when sender is dry during subscription.
Only for 1-to-1 style sockets (RS/MT)
- Bugfix: Put vtags for the correct amount of time into time-wait (MT).
- Bugfix: Clear vtag entries correctly on expiration (MT).
- Bugfix: shutdown() indicates ENOTCONN when called for unconnected
1-to-1 style sockets (MT).
- Bugfix: In sctp Auth code (PL).
- Add support for devices that support SCTP csum offload (igb).
- Add missing sctp_associd to mib sysctl xsctp_tcb structure (RS)
Obtained from: With help from Peter Lei and Michael Tuexen
we, like TCP and UDP, move the checksum calculation
into the IP routines when there is no hardware support
we call into the normal SCTP checksum routine.
The next round of SCTP updates will use
this functionality. Of course the IGB driver needs
a few updates to support the new intel controller set
that actually does SCTP csum offload too.
Reviewed by: gnn, rwatson, kmacy
section included a lot of stuff that did not belong there.
So split the block in multiple components each around the relevant stuff.
This said, I wonder if building a kernel where SYSCTL_NODE is not
defined is supported at all.
Submitted by: Marta Carbone
The new behaviour is on by default, and can be disabled by setting the
net.inet.tcp.rfc3465 sysctl to 0 to obtain previous behaviour.
The patch changes struct tcpcb in sys/netinet/tcp_var.h which breaks
the ABI. Bump __FreeBSD_version to 800061 accordingly. User space tools
that rely on the size of struct tcpcb (e.g. sockstat) need to be recompiled.
Reviewed by: rpaulo, gnn
Approved by: gnn, kmacy (mentors)
Sponsored by: FreeBSD Foundation
read with libkvm) to the addresses of a prison, when inside a
jail. [1]
As the patch from the PR was pre-'new-arp', add checks to the
llt_dump handlers as well.
While touching RTM_GET in route_output(), consistently use
curthread credentials rather than the creds from the socket
there. [2]
PR: kern/68189
Submitted by: Mark Delany <sxcg2-fuwxj@qmda.emu.st> [1]
Discussed with: rwatson [2]
Reviewed by: rwatson
MFC after: 4 weeks
applications to specify a non-local IP address when bind()'ing a socket
to a local endpoint.
This allows applications to spoof the client IP address of connections
if (obviously!) they somehow are able to receive the traffic normally
destined to said clients.
This patch doesn't include any changes to ipfw or the bridging code to
redirect the client traffic through the PCB checks so TCP gets a shot
at it. The normal behaviour is that packets with a non-local destination
IP address are not handled locally. This can be dealth with some IPFW hackery;
modifications to IPFW to make this less hacky will occur in subsequent
commmits.
Thanks to Julian Elischer and others at Ironport. This work was approved
and donated before Cisco acquired them.
Obtained from: Julian Elischer and others
MFC after: 2 weeks
jail-aware. Up to now we returned the first address of the interface
for SIOCGIFADDR w/o an ifr_addr in the query. This caused problems for
programs querying for an address but running inside a jail, as the
address returned usually did not belong to the jail.
Like for v6, if there was an ifr_addr given on v4, you could probe
for more addresses on the interfaces that you were not allowed to see
from inside a jail. Return an error (EADDRNOTAVAIL) in that case
now unless the address is on the given interface and valid for the
jail.
PR: kern/114325
Reviewed by: rwatson
MFC after: 4 weeks
the KASSERT and checks suggested.
Reviewed by: The udp tunneling was discussed on net@ under the
thread entitled "Heads up -- Thinking about UDP and tunneling"
disabled entirely, which is its default state before set to a
non-zero value.
PR: 128790
Submitted by: Nick Hilliard <nick at foobar dot org>
MFC after: 3 weeks
to ip_output(). The destionation is represented in a sockaddr{} object
that may contain other pieces of information, e.g., port number. This
same destination sockaddr{} object may be passed into L2 code, which
could be used to create a L2 entry. Since there exists a L2 table per
address family, the L2 lookup function can make address family specific
comparison instead of the generic bcmp() operation over the entire
sockaddr{} structure.
Note in the IPv6 case the sin6_scope_id is not compared because the
address is currently stored in the embedded form inside the kernel.
The in6_lltable_lookup() has to account for the scope-id if this
storage format were to change in the future.
1. The "route" command allows route insertion through the interface-direct
option "-iface". During if_attach(), an sockaddr_dl{} entry is created
for the interface and is part of the interface address list. This
sockaddr_dl{} entry describes the interface in detail. The "route"
command selects this entry as the "gateway" object when the "-iface"
option is present. The "arp" and "ndp" commands also interact with the
kernel through the routing socket when adding and removing static L2
entries. The static L2 information is also provided through the
"gateway" object with an AF_LINK family type, similar to what is
provided by the "route" command. In order to differentiate between
these two types of operations, a RTF_LLDATA flag is introduced. This
flag is set by the "arp" and "ndp" commands when issuing the add and
delete commands. This flag is also set in each L2 entry returned by the
kernel. The "arp" and "ndp" command follows a convention where a RTM_GET
is issued first followed by a RTM_ADD/DELETE. This RTM_GET request fills
in the fields for a "rtm" object, which is reinjected into the kernel by
a subsequent RTM_ADD/DELETE command. The entry returend from RTM_GET
is a prefix route, so the RTF_LLDATA flag must be specified when issuing
the RTM_ADD/DELETE messages.
2. Enforce the convention that NET_RT_FLAGS with a 0 w_arg is the
specification for retrieving L2 information. Also optimized the
code logic.
Reviewed by: julian
had been the only flag with random usage patterns.
Switch inc_flags to be used as a real bit field by using
INC_ISIPV6 with bitops to check for the 'isipv6' condition.
While here fix a place or two where in case of v4 inc_flags
were not properly initialized before.[1]
Found by: rwatson during review [1]
Discussed with: rwatson
Reviewed by: rwatson
MFC after: 4 weeks
the inpcb names rather than the following IPv6 compat macros:
in6pcb,in6p_sp, in6p_ip6_nxt,in6p_flowinfo,in6p_vflag,
in6p_flags,in6p_socket,in6p_lport,in6p_fport,in6p_ppcb and
sotoin6pcb().
Apart from removing duplicate code in netipsec, this is a pure
whitespace, not a functional change.
Discussed with: rwatson
Reviewed by: rwatson (version before review requested changes)
MFC after: 4 weeks (set the timer and see then)
1. separating L2 tables (ARP, NDP) from the L3 routing tables
2. removing as much locking dependencies among these layers as
possible to allow for some parallelism in the search operations
3. simplify the logic in the routing code,
The most notable end result is the obsolescent of the route
cloning (RTF_CLONING) concept, which translated into code reduction
in both IPv4 ARP and IPv6 NDP related modules, and size reduction in
struct rtentry{}. The change in design obsoletes the semantics of
RTF_CLONING, RTF_WASCLONE and RTF_LLINFO routing flags. The userland
applications such as "arp" and "ndp" have been modified to reflect
those changes. The output from "netstat -r" shows only the routing
entries.
Quite a few developers have contributed to this project in the
past: Glebius Smirnoff, Luigi Rizzo, Alessandro Cerri, and
Andre Oppermann. And most recently:
- Kip Macy revised the locking code completely, thus completing
the last piece of the puzzle, Kip has also been conducting
active functional testing
- Sam Leffler has helped me improving/refactoring the code, and
provided valuable reviews
- Julian Elischer setup the perforce tree for me and has helped
me maintaining that branch before the svn conversion
to keep for 7-STABLE when MFCing in_pcbladdr() to not change the
behaviour there.
With this a destination route via a loopback interface is treated as
a valid and reachable thing for IPv4 source address selection, even
though nothing of that network is ever directly reachable, but it is
more like a blackhole route.
With this the source address will be selected and IPsec can grab the
packets before we would discard them at a later point, encapsulate them
and send them out from a different tunnel endpoint IP.
Discussed on: net
Reported by: Frank Behrens <frank@harz.behrens.de>
Tested by: Frank Behrens <frank@harz.behrens.de>
MFC after: 4 weeks (just so that I get the mail)
but formerly missed under VIMAGE_GLOBAL.
Put the extern declarations of the virtualized globals
under VIMAGE_GLOBAL as the globals themsevles are already.
This will help by the time when we are going to remove the globals
entirely.
Sponsored by: The FreeBSD Foundation
missed under VIMAGE_GLOBAL.
Start putting the extern declarations of the virtualized globals
under VIMAGE_GLOBAL as the globals themsevles are already.
This will help by the time when we are going to remove the globals
entirely.
While there garbage collect a few dead externs from ip6_var.h.
Sponsored by: The FreeBSD Foundation
container structures, depending on VIMAGE_GLOBALS compile time option.
Make VIMAGE_GLOBALS a new compile-time option, which by default will not
be defined, resulting in instatiations of global variables selected for
V_irtualization (enclosed in #ifdef VIMAGE_GLOBALS blocks) to be
effectively compiled out. Instantiate new global container structures
to hold V_irtualized variables: vnet_net_0, vnet_inet_0, vnet_inet6_0,
vnet_ipsec_0, vnet_netgraph_0, and vnet_gif_0.
Update the VSYM() macro so that depending on VIMAGE_GLOBALS the V_
macros resolve either to the original globals, or to fields inside
container structures, i.e. effectively
#ifdef VIMAGE_GLOBALS
#define V_rt_tables rt_tables
#else
#define V_rt_tables vnet_net_0._rt_tables
#endif
Update SYSCTL_V_*() macros to operate either on globals or on fields
inside container structs.
Extend the internal kldsym() lookups with the ability to resolve
selected fields inside the virtualization container structs. This
applies only to the fields which are explicitly registered for kldsym()
visibility via VNET_MOD_DECLARE() and vnet_mod_register(), currently
this is done only in sys/net/if.c.
Fix a few broken instances of MODULE_GLOBAL() macro use in SCTP code,
and modify the MODULE_GLOBAL() macro to resolve to V_ macros, which in
turn result in proper code being generated depending on VIMAGE_GLOBALS.
De-virtualize local static variables in sys/contrib/pf/net/pf_subr.c
which were prematurely V_irtualized by automated V_ prepending scripts
during earlier merging steps. PF virtualization will be done
separately, most probably after next PF import.
Convert a few variable initializations at instantiation to
initialization in init functions, most notably in ipfw. Also convert
TUNABLE_INT() initializers for V_ variables to TUNABLE_FETCH_INT() in
initializer functions.
Discussed at: devsummit Strassburg
Reviewed by: bz, julian
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
to read-locking in the TCP input path, allowing greater TCP
input parallelism where multiple ithreads or ithread and netisr
are able to run in parallel. Previously, most TCP input paths
held a write lock on the global tcbinfo lock, effectively
serializing TCP input.
Before looking up the connection, acquire a write lock if a
potentially state-changing flag is set on the TCP segment header
(FIN, RST, SYN), and otherwise a read lock. We may later have
to upgrade to a write lock in certain cases (ACKs received by the
syncache or during TIMEWAIT) in order to support global state
transitions, but this is never required for steady-state packets.
Upgrading from a write lock to a read lock must be done as a
trylock operation to avoid deadlocks, and actually violates the
lock order as the tcbinfo lock preceeds the inpcb lock held at
the time of upgrade. If the trylock fails, we bump the refcount
on the inpcb, drop both locks, and re-acquire in-order. If
another thread has freed the connection while the locks are
dropped, we free the inpcb and repeat the lookup (this should
hardly ever or never happen in practice).
For now, maintain a number of new counters measuring how many
times various cases execute, and in particular whether various
optimistic assumptions about when read locks can be used, whether
upgrades are done using the fast path, and whether connections
close in practice in the above-described race, actually occur.
MFC after: 6 weeks
Discussed with: kmacy
Reviewed by: bz, gnn, kmacy
Tested by: kmacy
incremented using in_pcbref(), and decremented using in_pcbfree()
or inpcbrele(). Protocols using only current in_pcballoc() and
in_pcbfree() calls will see the same semantics, but it is now
possible for TCP to call in_pcbref() and in_pcbrele() to prevent
an inpcb from being freed when both tcbinfo and per-inpcb locks
are released. This makes it possible to safely transition from
holding only the inpcb lock to both tcbinfo and inpcb lock
without re-looking up a connection in the input path, timer
path, etc.
Notice that in_pcbrele() does not unlock the connection after
decrementing the refcount, if the connection remains, so that
the caller can continue to use it; in_pcbrele() returns a flag
indicating whether or not the inpcb pointer is still valid, and
in_pcbfee() is now a simple wrapper around in_pcbrele().
MFC after: 1 month
Discussed with: bz, kmacy
Reviewed by: bz, gnn, kmacy
Tested by: kmacy
RTFREE_LOCKED() here. This macro makes sure the reference count
on the route is being managed properly. This elimates another
case which results in the following message being printed to the
console:
rtfree: 0xc841ee88 has 1 refs
Reviewed by: bz
MFC after: 2 weeks
bit of debugging afterwards):
- Fix protection code for notification generation.
- Decouple associd from vtag
- Allow vtags to have less strigent requirements in non-uniqueness.
o don't pre-hash them when you issue one in a cookie.
o Allow duplicates and use addresses and ports to
discriminate amongst the duplicates during lookup.
- Add support for the NAT draft draft-ietf-behave-sctpnat-00, this
is still experimental and needs more extensive testing with the
Jason Butt ipfw changes.
- Support for the SENDER_DRY event to get DTLS in OpenSSL working
with a set of patches from Michael Tuexen (hopefully heading to OpenSSL soon).
- Update the support of SCTP-AUTH by Peter Lei.
- Use macros for refcounting.
- Fix MTU for UDP encapsulation.
- Fix reporting back of unsent data.
- Update assoc send counter handling to be consistent with endpoint sent counter.
- Fix a bug in PR-SCTP.
- Fix so we only send another FWD-TSN when a SACK arrives IF and only
if the adv-peer-ack point progressed. However we still make sure
a timer is running if we do have an adv_peer_ack point.
- Fix PR-SCTP bug where chunks were retransmitted if they are sent
unreliable but not abandoned yet.
With the help of: Michael Teuxen and Peter Lei :-)
MFC after: 4 weeks
directly include only the header files needed. This reduces the
unneeded spamming of various headers into lots of files.
For now, this leaves us with very few modules including vnet.h
and thus needing to depend on opt_route.h.
Reviewed by: brooks, gnn, des, zec, imp
Sponsored by: The FreeBSD Foundation
Bring in updated jail support from bz_jail branch.
This enhances the current jail implementation to permit multiple
addresses per jail. In addtion to IPv4, IPv6 is supported as well.
Due to updated checks it is even possible to have jails without
an IP address at all, which basically gives one a chroot with
restricted process view, no networking,..
SCTP support was updated and supports IPv6 in jails as well.
Cpuset support permits jails to be bound to specific processor
sets after creation.
Jails can have an unrestricted (no duplicate protection, etc.) name
in addition to the hostname. The jail name cannot be changed from
within a jail and is considered to be used for management purposes
or as audit-token in the future.
DDB 'show jails' command was added to aid debugging.
Proper compat support permits 32bit jail binaries to be used on 64bit
systems to manage jails. Also backward compatibility was preserved where
possible: for jail v1 syscalls, as well as with user space management
utilities.
Both jail as well as prison version were updated for the new features.
A gap was intentionally left as the intermediate versions had been
used by various patches floating around the last years.
Bump __FreeBSD_version for the afore mentioned and in kernel changes.
Special thanks to:
- Pawel Jakub Dawidek (pjd) for his multi-IPv4 patches
and Olivier Houchard (cognet) for initial single-IPv6 patches.
- Jeff Roberson (jeff) and Randall Stewart (rrs) for their
help, ideas and review on cpuset and SCTP support.
- Robert Watson (rwatson) for lots and lots of help, discussions,
suggestions and review of most of the patch at various stages.
- John Baldwin (jhb) for his help.
- Simon L. Nielsen (simon) as early adopter testing changes
on cluster machines as well as all the testers and people
who provided feedback the last months on freebsd-jail and
other channels.
- My employer, CK Software GmbH, for the support so I could work on this.
Reviewed by: (see above)
MFC after: 3 months (this is just so that I get the mail)
X-MFC Before: 7.2-RELEASE if possible
underneath #ifdef VIMAGE blocks.
This change introduces some churn in #include ordering and nesting
throughout the network stack and drivers but is not expected to cause
any additional issues.
In the next step this will allow us to instantiate the virtualization
container structures and switch from using global variables to their
"containerized" counterparts.
Reviewed by: bz, julian
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
IPv6 socket by comparing a constant inp vflag.
This is expected to help to reduce extra locking.
Suggested by: rwatson
Reviewed by: rwatson
MFC after: 6 weeks
IPsec change in r185366 only differed in two additonal IPv6 lines.
Rather than splattering conditional code everywhere add the v6
check centrally at this single place.
Reviewed by: rwatson (as part of a larger changset)
MFC after: 6 weeks (*)
(*) possibly need to leave a stub wrapper in 7 to keep the symbol.
Ignoring different names because of macros (in6pcb, in6p_sp) and
inp vs. in6p variable name both functions were entirely identical.
Reviewed by: rwatson (as part of a larger changeset)
MFC after: 6 weeks (*)
(*) possibly need to leave a stub wrappers in 7 to keep the symbols.
whitespace) macros from p4/vimage branch.
Do a better job at enclosing all instantiations of globals
scheduled for virtualization in #ifdef VIMAGE_GLOBALS blocks.
De-virtualize and mark as const saorder_state_alive and
saorder_state_any arrays from ipsec code, given that they are never
updated at runtime, so virtualizing them would be pointless.
Reviewed by: bz, julian
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
as in_pcbdetach() and we don't need the code twice.
Reviewed by: rwatson
MFC after: 6 weeks (*)
(*) possibly need to leave a stub wrapper in 7 to keep the symbol.
for virtualization.
Instead of initializing the affected global variables at instatiation,
assign initial values to them in initializer functions. As a rule,
initialization at instatiation for such variables should never be
introduced again from now on. Furthermore, enclose all instantiations
of such global variables in #ifdef VIMAGE_GLOBALS blocks.
Essentialy, this change should have zero functional impact. In the next
phase of merging network stack virtualization infrastructure from
p4/vimage branch, the new initialization methology will allow us to
switch between using global variables and their counterparts residing in
virtualization containers with minimum code churn, and in the long run
allow us to intialize multiple instances of such container structures.
Discussed at: devsummit Strassburg
Reviewed by: bz, julian
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
-Improvement: panic() on INVARIANTS kernels if memory allocation
fails for a tagblock in sctp_add_vtag_to_timewait().
-Bugfix: Protect code in sctp_is_in_timewait() by
SCTP_INP_INFO_WLOCK/SCTP_INP_INFO_WUNLOCK.
-Cleanup: Get rid of unused variable now in sctp_init_asoc().
-Bugfix: Reuse the correct vtag in sctp_add_vtag_to_timewait().
-Cleanup: Get rid of unused constant SCTP_TIME_WAIT_SHORT
in sctp_constants.h.
-Improvement: Use all hash buckets of the vtag hash table.
-Cleanup: Get rid of then unused constant SCTP_STACK_VTAG_HASH_SIZE_A.
-Bugfix: Handle SHUTDOWN;SACK packet correctly.
-Bugfix: Last TSN in a gap ack block was not being "ack'd"
in the internal scoreboard.
Obtained from: (with help from Michael Tuexen)
tcp_mss() and tcp_mss_update() so that tcp_mtudisc() could
re-use the same code.
Move the TSO logic back to tcp_mss() and out of tcp_mss_update().
We tried to avoid that initially but if were are called from
tcp_output() with EMSGSIZE, we cleared the TSO flag on the tcpcb
there, called into tcp_mtudisc() and tcp_mss_update() which
then would reenable TSO on the tcpcb based on TSO capabilities
of the interface as learnt in tcp_maxmtu/6().
So if TSO was enabled on the (possibly new) outgoing interface
it was turned back on, which lead to an endless loop between
tcp_output() and tcp_mtudisc() until we overflew the stack.
Reported by: kmacy
MFC after: 2 months (along with r182851)
tcp_mss() and tcp_mss_update() so that tcp_mtudisc() could
re-use the same code.
In case we return early and got a metricptr to pass the hostcache
info back to the caller we need to initialize the data to a defined
state (zero it) as tcp_hc_get() would do if there was no hit.
Without that the caller would check on random stack garbage which
could lead to undefined results.
This only affected tcp_mss() if there was no routing entry for the peer,
tcp_mtudisc() was not affected.
MFC after: 2 months (along with r182851)
This should fix q_time overflow, which happens after 2^32/(86400*hz) days of
uptime (~50days for hz = 1000).
q_time overflow cause following:
- traffic shaping may not work in 'fast' mode (not enabled by default).
- incorrect average queue length calculation in RED/GRED algorithm.
NB: due to ABI change this change is not applicable to stable.
PR: kern/128401
a) Need for EEOR mode to take the min of the socket buffer size and the
add more threshold, otherwise if you are so silly as to set a send
buf size less than the add-more you could block forever in eeor mode.
b) We were incorrectly using the sysctl vs the calculated value. This
causes us to block forever if the addmore theshold is larger than
then the socket buffer size.
- If we send EXACTLY the size left in the send buffer
and then send again, we end up with exactly 0 bytes and
don't hit the pre-block code to wait for more space.
- If we fall into the loop with our max_len == 0 (the bug
above) we then call in to copy out the data, setup the length
of the waiting to transmit data to 0 and call the mbuf copy routine
which 0 indicates copy all the data to the mbuf chain.. which it
does. This then leaves a "stuck" message on the stream queue with
its size exactly 0 bytes but all the data there and thus nothing
left in the uio structure. We then reach a stuck forever state
never being able to send data.
sooner to decomplicate locking and eliminate the need for a rather
chatty comment about why we have to handle the global lock in a
special way for the benefit of ipfw and pf cred rules.
MFC after: 3 days
- Consistently add parentheses to return statements.
- Use NULL instead of 0 when comparing pointers, also avoiding
unnecessary casts.
- Do not use pointers as booleans.
Reviewed by: rwatson (earlier version)
MFC after: 2 months
already (but probably had been way above as the code was there twice)
and describe what was last changed in rev. 1.199 there (which now is
in sync with in6_src.c r184096).
Pointed at by: mlaier
MFC after: 2 mmonths
ephemeral port allocation as implemented in netinet/in_pcb.c rev. 1.143
(initially from OpenBSD) and follow-up commits during the last four and
a half years including rev. 1.157, 1.162 and 1.199.
This now is relying on the same infrastructure as has been implemented
in in_pcb.c since rev. 1.199.
Reviewed by: silby, rpaulo, mlaier
MFC after: 2 months
be given when the user has enabled it). (Michael Tuexen)
- Sack Immediately was not being set properly on the actual chunk, it
was only put in the rcvd_flags which is incorrect. (Michael Tuexen)
- added an ifndef userspace to one of the already present macro's for
inet (Brad Penoff)
Obtained from: Michael Tuexen and Brad Penoff
MFC after: 4 weeks
credentials from inp_cred which is also available after the
socket is gone.
Switch cr_canseesocket consumers to cr_canseeinpcb.
This removes an extra acquisition of the socket lock.
Reviewed by: rwatson
MFC after: 3 months (set timer; decide then)
netisr or ithread's socket buffer size limit is not the right limit to
use. Instead, pass NULL as the other two calls to sbreserve_locked()
in the TCP input path (tcp_mss()) do.
In practice, this is a no-op, as ithreads and the netisr run without a
process limit on socket buffer use, and a NULL thread pointer leads to
not using the process's limit, if any. However, if tcp_input() is
called in other contexts that do have limits, this may prevent the
incorrect limit from being used.
MFC after: 3 days
This means that inp_cred is always there, even after the socket
has gone away. It also means that it is constant for the lifetime
of the inp.
Both facts lead to simpler code and possibly less locking.
Suggested by: rwatson
Reviewed by: rwatson
MFC after: 6 weeks
X-MFC Note: use a inp_pspare for inp_cred
For the jail case we are already looping over the interface addresses
before falling back to the only IP address of a jail in case of no
match. This is in preparation for the upcoming multi-IPv4/v6/no-IP
jail patch this change was developed with initially.
This also changes the semantics of selecting the IP for processes within
a jail as it now uses the same logic as outside the jail (with additional
checks) but no longer is on a mutually exclusive code path.
Benchmarks had shown no difference at 95.0% confidence for neither the
plain nor the jail case (even with the additional overhead). See:
http://lists.freebsd.org/pipermail/freebsd-net/2008-September/019531.html
Inpsired by a patch from: Yahoo! (partially)
Tested by: latest multi-IP jail patch users (implictly)
Discussed with: rwatson (general things around this)
Reviewed by: mostly silence (feedback from bms)
Help with benchmarking from: kris
MFC after: 2 months
from the vimage project, as per plan established at devsummit 08/08:
http://wiki.freebsd.org/Image/Notes200808DevSummit
Introduce INIT_VNET_*() initializer macros, VNET_FOREACH() iterator
macros, and CURVNET_SET() context setting macros, all currently
resolving to NOPs.
Prepare for virtualization of selected SYSCTL objects by introducing a
family of SYSCTL_V_*() macros, currently resolving to their global
counterparts, i.e. SYSCTL_V_INT() == SYSCTL_INT().
Move selected #defines from sys/sys/vimage.h to newly introduced header
files specific to virtualized subsystems (sys/net/vnet.h,
sys/netinet/vinet.h etc.).
All the changes are verified to have zero functional impact at this
point in time by doing MD5 comparision between pre- and post-change
object files(*).
(*) netipsec/keysock.c did not validate depending on compile time options.
Implemented by: julian, bz, brooks, zec
Reviewed by: julian, bz, brooks, kris, rwatson, ...
Approved by: julian (mentor)
Obtained from: //depot/projects/vimage-commit2/...
X-MFC after: never
Sponsored by: NLnet Foundation, The FreeBSD Foundation
in the transmit path, such as TCPS_TIMEWAIT, fail the credential
extraction immediately rather than acquiring locks and looking up
the inpcb on the global lists in order to reach the conclusion that
the credential extraction has failed.
This is more efficient, but more importantly, it avoids lock
recursion on the inpcbinfo, which is no longer allowed with rwlocks.
This appears to have been responsible for at least two reported
panics.
MFC after: 3 days
Reported by: ganbold
called without an inpcb pointer despite holding the tcbinfo global
lock, which lead to a deadlock or panic when ipfw tried to further
acquire it recursively.
Reported by: Stefan Ehmann <shoesoft at gmx dot net>
MFC after: 3 days
unconditionally drop the tcbinfo lock (after all, we assert it lines
before), but call tcp_dropwithreset() under both inpcb and inpcbinfo
locks only if we pass in an tcpcb. Otherwise, if the pointer is NULL,
firewall code may later recurse the global tcbinfo lock trying to look
up an inpcb.
This is an instance where a layering violation leads not only
potentially to code reentrace and recursion, but also to lock
recursion, and was revealed by the conversion to rwlocks because
acquiring a read lock on an rwlock already held with a write lock is
forbidden. When these locks were mutexes, they simply recursed.
Reported by: Stefan Ehmann <shoesoft at gmx dot net>
MFC after: 3 days
rt_check() in its original form proved to be sufficient and
rt_check_fib() can go away (as can its evil twin in_rt_check()).
I believe this does NOT address the crashes people have been seeing
in rt_check.
MFC after: 1 week
the same way it has been implemented for IPv4.
Reviewed by: bms (skimmed)
Tested by: Nick Hilliard (nick netability.ie) (with more changes)
MFC after: 2 months
congestion window not being incremented, if cwnd > maxseg^2.
As suggested in RFC2581 increment the cwnd by 1 in this case.
See http://caia.swin.edu.au/reports/080829A/CAIA-TR-080829A.pdf
for more details.
Submitted by: Alana Huebner, Lawrence Stewart,
Grenville Armitage (caia.swin.edu.au)
Reviewed by: dwmalone, gnn, rpaulo
MFC After: 3 days
Payload Length) as set in tcpip_fillheaders().
ip6_output() will calculate it based of the length from the
mbuf packet header itself.
So initialize the value in tcpip_fillheaders() in correct
(network) byte order.
With the above change, to my reading, all places calling tcp_trace()
pass in the ip6 header via ipgen as serialized in the mbuf and with
ip6_plen in network byte order.
Thus convert the IPv6 payload length to host byte order before printing.
MFC after: 2 months
calls the latter.
Merge tcp_mss_update() with code from tcp_mtudisc() basically
doing the same thing.
This gives us one central place where we calcuate and check mss values
to update t_maxopd (maximum mss + options length) instead of two slightly
different but almost equal implementations to maintain.
PR: kern/118455
Reviewed by: silby (back in March)
MFC after: 2 months
SYSCTL_PROCs and check that the default mss for neither v4 nor
v6 goes below the minimum MSS constant (216).
This prevents people from shooting themselves in the foot.
PR: kern/118455 (remotely related)
Reviewed by: silby (as part of a larger patch in March)
MFC after: 2 months
This is different to the first one (as len gets updated between those
two) and would have caught various edge cases (read bugs) at a well
defined place I had been debugging the last months instead of
triggering (random) panics further down the call graph.
MFC after: 2 months
the default rule number but also the maximum rule number. User space
software such as ipfw and natd should be aware of its value. The
software that already includes ip_fw.h should use the defined value. All
other a expected to use sysctl (as discussed on net@).
MFC after: 5 days.
Discussed on: net@
translation. It turns out this is useful for applications which require
source port randomization for security (i.e. dns servers).
Discussed with: secteam
Requested by: mlaier
MFC after: 2 weeks
wind up with the incorrect checksum on the wire when transmitted via
devices that do checksum offloading.
PR: kern/119635
Reviewed by: rwatson
MFC after: 5 days
- Change it so that without INVARIANTs there are
no panics in SCTP.
- sctp_timer changes so that we have a recovery mechanism
when the sent list is out of order.
storage. We can safely remove the label copying operations since
M_MOVE_PKTHDR will move the mbuf tags (which contain MAC labels) to
the destination mbuf.
MFC after: 1 week
Discussed with: rwatson
we can be sure that it's valid.
In case we abort early free it again else put it into the syncache.
We need the cred in the syncache to be able to restrict what will be
exportet by the sysctl helper function syncache_pcblist() (to netstat)
within jails.
PR: kern/126493
Reviewed by: rwatson (earlier versions)
MFC after: 3 days
the IP multicast input code from the output path; we don't allow
reentrance of the input path from the IP output path, it must use the
netisr due to potential lock recursion.
MFC after: 3 days
virtualization work done by Marko Zec (zec@).
This is the first in a series of commits over the course
of the next few weeks.
Mark all uses of global variables to be virtualized
with a V_ prefix.
Use macros to map them back to their global names for
now, so this is a NOP change only.
We hope to have caught at least 85-90% of what is needed
so we do not invalidate a lot of outstanding patches again.
Obtained from: //depot/projects/vimage-commit2/...
Reviewed by: brooks, des, ed, mav, julian,
jamie, kris, rwatson, zec, ...
(various people I forgot, different versions)
md5 (with a bit of help)
Sponsored by: NLnet Foundation, The FreeBSD Foundation
X-MFC after: never
V_Commit_Message_Reviewed_By: more people than the patch
into v4-only vs. v6-only inp_flags processing.
When ip6_savecontrol_v4() is called from ip6_savecontrol() we
were not passing back the **mp thus the information will be missing
in userland.
Istead of going with a *** as suggested in the PR we are returning
**mp now and passing in the v4only flag as a pointer argument.
PR: kern/126349
Reviewed by: rwatson, dwmalone
keyword. But it doesn't work. Two options.. make it no longer accept it,
or actually make it work.. I chose the 2nd..
Allow the tablearg to be used to specify a skipto destination.
This is actually a very powerful construct if used correctly, or a sink
of cpu cycles if used badly.
changes t teh man page will follow.
This gives significant performance improvements when many raw sockets used.
Benchmarks of mpd handeling 1000 simultaneous PPTP connections show up to 50%
performance boost. With higher number of connections benefit becomes even
bigger. PopTop snd others should also get some benefits.
- removing 'const' qualifier from an input parameter to conform to the type
required by rw_assert();
- using in_addr->s_addr to retrive 32 bits address value.
Observed by: tinderbox
information from rip_input() to rip_append(). Instead, pass the source
address for an IP datagram to rip_append() using a stack-allocated
sockaddr_in, similar to udp_input() and udp_append().
Prior to the move to rwlocks for inpcbinfo, this was not a problem, as
use of the global was synchronized using the ripcbinfo mutex, but with
read-locking there is the potential for a race during concurrent
receive.
This problem is not present in the IPv6 raw IP socket code, which
already used a stack variable for the address.
Spotted by: mav
MFC after: 1 week (before inpcbinfo rwlock changes)
completes the move to a fully parallel UDP transmit path by using
global read, rather than write, locking of inpcbinfo in further
semi-connected cases:
- Add macros to allow try-locking of inpcb and inpcbinfo.
- Always acquire an incpcb read lock in udp_output(), which stablizes the
local inpcb address and port bindings in order to determine what further
locking is required:
- If the inpcb is currently not bound (at all) and are implicitly
connecting, we require inpcbinfo and inpcb write locks, so drop the
read lock and re-acquire.
- If the inpcb is bound for at least one of the port or address, but an
explicit source or destination is requested, trylock the inpcbinfo
lock, and if that fails, drop the inpcb lock, lock the global lock,
and relock the inpcb lock.
- Otherwise, no further locking is required (common case).
- Update comments.
In practice, this means that the vast majority of consumers of UDP sockets
will not acquire any exclusive locks at the socket or UDP levels of the
network stack. This leads to a marked performance improvement in several
important workloads, including BIND, nsd, and memcached over UDP, as well
as significant improvements in pps microbenchmarks.
The plan is to MFC all of the rwlock changes to RELENG_7 once they have
settled for a weeks in the tree.
Tested by: ps, kris (older revision), bde
MFC after: 3 weeks
