This change intends to run the whole network stack in softint context
(or normal LWP), not hardware interrupt context. Note that the work is
still incomplete by this change; to that end, we also have to softint-ify
if_link_state_change (and bpf) which can still run in hardware interrupt.
This change softint-ifies at ifp->if_input that is called from
each device driver (and ieee80211_input) to ensure Layer 2 runs
in softint (e.g., ether_input and bridge_input). To this end,
we provide a framework (called percpuq) that utlizes softint(9)
and percpu ifqueues. With this patch, rxintr of most drivers just
queues received packets and schedules a softint, and the softint
dequeues packets and does rest packet processing.
To minimize changes to each driver, percpuq is allocated in struct
ifnet for now and that is initialized by default (in if_attach).
We probably have to move percpuq to softc of each driver, but it's
future work. At this point, only wm(4) has percpuq in its softc
as a reference implementation.
Additional information including performance numbers can be found
in the thread at tech-kern@ and tech-net@:
http://mail-index.netbsd.org/tech-kern/2016/01/14/msg019997.html
Acknowledgment: riastradh@ greatly helped this work.
Thank you very much!
This change was intended, but Nakahara-san had already made a better
one locally! So I'll let him commit that one, and I'll try not to
step on anyone's toes again.
Mostly mechanical change to replace it, culling some now-needless
boilerplate around all the users.
This does not substantively change the ip_encap API or eliminate
abuse of sketchy pointer casts -- that will come later, and will be
easier now that it is not tangled up with struct protosw.
You can't use this unless you know what it is a priori: the formal
prototype is variadic, and the different instances (e.g., ip_output,
route_output) have different real prototypes.
Convert the only user of it, raw_send in net/raw_cb.c, to take an
explicit callback argument. Convert the only instances of it,
route_output and key_output, to such explicit callbacks for raw_send.
Use assertions to make sure the conversion to explicit callbacks is
warranted.
Discussed on tech-net with no objections:
https://mail-index.netbsd.org/tech-net/2016/01/16/msg005484.html
lltable and llentry were introduced to replace ARP cache data structure
for further restructuring of the routing table: L2 nexthop cache
separation. This change replaces the NDP cache data structure
(llinfo_nd6) with them as well as ARP.
One noticeable change is for neighbor cache GC mechanism that was
introduced to prevent IPv6 DoS attacks. net.inet6.ip6.neighborgcthresh
was the max number of caches that we store in the system. After
introducing lltable/llentry, the value is changed to be per-interface
basis because lltable/llentry stores neighbor caches in each interface
separately. And the change brings one degradation; the old GC mechanism
dropped exceeded packets based on LRU while the new implementation drops
packets in order from the beginning of lltable (a hash table + linked
lists). It would be improved in the future.
Added functions in in6.c come from FreeBSD (as of r286629) and are
tweaked for NetBSD.
Proposed on tech-kern and tech-net.
This is a restructuring for coming changes to nd6 (replacing
llinfo_nd6 with llentry). Once we have a lock of llinfo_nd6,
we need to pass it to nd6_ns_output with holding the lock.
However, in a function subsequent to nd6_ns_output, the llinfo_nd6
may be looked up, i.e., its lock would be acquired again.
To avoid such a situation, pass only required data (in6_addr) to
nd6_ns_output instead of passing whole llinfo_nd6.
Inspired by FreeBSD
We have to update icmp6_redirect_timeout_q as well as icmp6_redirtimeout
when changing net.inet6.icmp6.redirtimeout via sysctl. The updating logic
is copied from sysctl_net_inet_icmp_redirtimeout.
This change is from s-yamaguchi@IIJ (with KNF by ozaki-r) and fixes
PR kern/50240.
We have to check and avoid to rtfree the original rtentry passed to
nd6_output even when manipulating gateway routes.
This fixes panic on assertion "ro->_ro_rt ==NULL || ro->_ro_rt->rt_refcnt > 0"
failure and probably PR kern/50161.
Some codes in sys/net* use time_second to manage time periods such as
cache expirations. However, time_second doesn't increase monotonically
and can leap by say settimeofday(2) according to time_second(9). We
should use time_uptime instead of it to avoid such time leaps.
This change replaces time_second with time_uptime. Additionally it
converts a time based on time_uptime to a time based on time_second
when the kernel passes the time to userland programs that expect
the latter, and vice versa.
Note that we shouldn't leak time_uptime to other hosts over the
netowrk. My investigation shows there is no such leak:
http://mail-index.netbsd.org/tech-net/2015/08/06/msg005332.html
Discussed on tech-kern and tech-net.
rt_refcnt of rtentry was used in bad manners, for example, direct rt_refcnt++
and rt_refcnt-- outside route.c, "rt->rt_refcnt++; rtfree(rt);" idiom, and
touching rt after rt->rt_refcnt--.
These abuses seem to be needed because rt_refcnt manages only references
between rtentry and doesn't take care of references during packet processing
(IOW references from local variables). In order to reduce the above abuses,
the latter cases should be counted by rt_refcnt as well as the former cases.
This change improves consistency of use of rt_refcnt:
- rtentry is always accessed with rt_refcnt incremented
- rtentry's rt_refcnt is decremented after use (rtfree is always used instead
of rt_refcnt--)
- functions returning rtentry increment its rt_refcnt (and caller rtfree it)
Note that rt_refcnt prevents rtentry from being freed but doesn't prevent
rtentry from being updated. Toward MP-safe, we need to provide another
protection for rtentry, e.g., locks. (Or introduce a better data structure
allowing concurrent readers during updates.)
nd6_numroutes is intended to be incremented when a route is added via RA
and decremented when a RA route is deleted. However, a decrement of a RA
route was skipped when there remained references to the RA route.
converting protocol user requests to accept sockaddr instead of mbufs.
remove tcp_input copy in to mbuf from sockaddr and just copy to sockaddr
to make it possible for the transitional functions to go away.
no version bump since these functions only existed for a short time and
were commented as adapters (they appeared in 7.99.15).
nam parameter type from buf * to sockaddr *.
final commit for parameter type changes to protocol user requests
* bump kernel version to 7.99.15 for parameter type changes to pr_{send,connect}
pr_{accept,sockname,peername} nam parameter type from mbuf * to sockaddr *.
* retained use of mbuftypes[MT_SONAME] for now.
* bump to netbsd version 7.99.12 for parameter type change.
patch posted to tech-net@ 2015/04/19
* update protocol bind implementations to use/expect sockaddr *
instead of mbuf *
* introduce sockaddr_big struct for storage of addr data passed via
sys_bind; sockaddr_big is of sufficient size and alignment to
accommodate all addr data sizes received.
* modify sys_bind to allocate sockaddr_big instead of using an mbuf.
* bump kernel version to 7.99.9 for change to pr_bind() parameter type.
Patch posted to tech-net@
http://mail-index.netbsd.org/tech-net/2015/03/15/msg005004.html
The choice to use a new structure sockaddr_big has been retained since
changing sockaddr_storage size would lead to unnecessary ABI change. The
use of the new structure does not preclude future work that increases
the size of sockaddr_storage and at that time sockaddr_big may be
trivially replaced.
Tested by mrg@ and myself, discussed with rmind@, posted to tech-net@
they are created on the fly. This makes it clear what the route is for
and allows an optimisation in ip_output() by avoiding a call to
in_broadcast() because most of the time we do talk to a host.
It also avoids a needless allocation for the storage of llinfo_arp and
thus vanishes from arp(8) - it showed as incomplete anyway so this
is a nice side effect.
Guard against this and routes marked with RTF_BLACKHOLE in
ip_fastforward().
While here, guard against routes marked with RTF_BLACKHOLE in
ip6_fastforward().
RTF_BROADCAST is IPv4 only, so don't bother checking that here.
Add functions rt_ifa_addlocal() and rt_ifa_remlocal() to add and remove
local routes for the address and announce the new address and route
to the routing socket.
Add in_ifaddlocal() and in_ifremlocal() to use these functions.
Rename in6_if{add,rem}loop() to in6_if{add,rem}local() and use these
functions.
rtinit() no longer announces the address, just the network route for the
address. As such, calls to rt_newaddrmsg() have been removed from
in_addprefix() and in_scrubprefix().
This solves the problem of potentially more than one announcement, or no
announcement at all for the address in certain situations.
per-interface data, make sure to call nd6_purge() with it to remove
routing entries pointing to the going interface.
When we should happen to call this function again later, with the data
already gone, just return.
Fixes PR kern/49682, ok: christos.
(v4 multicast options off v4 mapped v6 socket) on interface destruction. The
code to clean this up in a true v4 socket was moved to its own function
which is now also called in the corresponding place for v6 sockets on
interface destruction.
Call callout_halt before callout_destroy. And also let callout (mld_timeo)
not call callout_schedule when we already called callout_halt.
This fixes PR 47881.
- allow binding to mapped v4 multicast addresses
- define v4moptions, allow setting it via ioctl, pass it to ip_output,
free it when killing the pcb.
Ideally we would allow the IPV6 multicast setsockopts work on mapped addresses
too, but this is a lot more work and linux does not do it either.
switches and put into separate functions
- always KASSERT(solocked(so)) even if not implemented
(for PRU_CONNECT2 only)
- replace calls to pr_generic() with req = PRU_CONNECT2 with calls to
pr_connect2()
- replace calls to pr_generic() with req = PRU_PURGEIF with calls to
pr_purgeif()
put common code from unp_connect2() (used by unp_connect() into
unp_connect1() and call out to it when needed
patch only briefly reviewed by rmind@
separate functions
xxx_send(struct socket *, struct mbuf *, struct mbuf *,
struct mbuf *, struct lwp *)
- always KASSERT(solocked(so)) even if not implemented
- replace calls to pr_generic() with req = PRU_SEND with calls to
pr_send()
rename existing functions that operate on PCB for consistency (and to
free up their names for xxx_send() PRUs
- l2cap_send() -> l2cap_send_pcb()
- sco_send() -> sco_send_pcb()
- rfcomm_send() -> rfcomm_send_pcb()
patch reviewed by rmind
pr_generic() usrreq switches and put into separate functions
xxx_disconnect(struct socket *)
xxx_shutdown(struct socket *)
xxx_abort(struct socket *)
- always KASSERT(solocked(so)) even if not implemented
- replace calls to pr_generic() with req =
PRU_{DISCONNECT,SHUTDOWN,ABORT}
with calls to pr_{disconnect,shutdown,abort}() respectively
rename existing internal functions used to implement above functionality
to permit use of the names for xxx_{disconnect,shutdown,abort}().
- {l2cap,sco,rfcomm}_disconnect() ->
{l2cap,sco,rfcomm}_disconnect_pcb()
- {unp,rip,tcp}_disconnect() -> {unp,rip,tcp}_disconnect1()
- unp_shutdown() -> unp_shutdown1()
patch reviewed by rmind
into seaparate functions
xxx_listen(struct socket *, struct mbuf *)
- always KASSERT(solocked(so)) and KASSERT(nam != NULL)
- replace calls to pr_generic() with req = PRU_CONNECT with
pr_connect()
- rename existin {l2cap,sco,rfcomm}_connect() to
{l2cap,sco,rfcomm}_connect_pcb() respectively to permit
naming consistency with other protocols functions.
- drop struct lwp * parameter from unp_connect() and at_pcbconnect()
and use curlwp instead where appropriate.
patch reviewed by rmind
switches and put into separate functions
xxx_bind(struct socket *, struct mbuf *)
xxx_listen(struct socket *)
- always KASSERT(solocked(so)) even if not implemented
- replace calls to pr_generic() with req = PRU_BIND with call to
pr_bind()
- replace calls to pr_generic() with req = PRU_LISTEN with call to
pr_listen()
- drop struct lwp * parameter from at_pcbsetaddr(), in_pcbbind() and
unp_bind() and always use curlwp.
rename existing functions that operate on PCB for consistency (and to
free up their names for xxx_{bind,listen}() PRUs
- l2cap_{bind,listen}() -> l2cap_{bind,listen}_pcb()
- sco_{bind,listen}() -> sco_{bind,listen}_pcb()
- rfcomm_{bind,listen}() -> rfcomm_{bind,listen}_pcb()
patch reviewed by rmind
welcome to netbsd 6.99.48
switches and put into separate functions
xxx_sendoob(struct socket *, struct mbuf *, struct mbuf *)
xxx_recvoob(struct socket *, struct mbuf *, int)
- always KASSERT(solocked(so)) even if request is not implemented
- replace calls to pr_generic() with req = PRU_{SEND,RCV}OOB with
calls to pr_{send,recv}oob() respectively.
there is still some tweaking of m_freem(m) and m_freem(control) to come
for consistency. not performed with this commit for clarity.
reviewed by rmind
into a separate function xxx_accept(struct socket *, struct mbuf *)
note: future cleanup will take place to remove struct mbuf parameter
type and replace it with a more appropriate type.
patch reviewed by rmind
usrreq switches and put into separate functions
xxx_{peer,sock}addr(struct socket *, struct mbuf *).
- KASSERT(solocked(so)) always in new functions even if request
is not implemented
- KASSERT(pcb != NULL) and KASSERT(nam) if the request is
implemented and not for tcp.
* for tcp roll #ifdef KPROF and #ifdef DEBUG code from tcp_usrreq() into
easier to cut & paste functions tcp_debug_capture() and
tcp_debug_trace()
- functions provided by rmind
- remaining use of PRU_{PEER,SOCK}ADDR #define to be removed in a
future commit.
* rename netbt functions to permit consistency of pru function names
(as has been done with other requests already split out).
- l2cap_{peer,sock}addr() -> l2cap_{peer,sock}_addr_pcb()
- rfcomm_{peer,sock}addr() -> rfcomm_{peer,sock}_addr_pcb()
- sco_{peer,sock}addr() -> sco_{peer,sock}_addr_pcb()
* split/refactor do_sys_getsockname(lwp, fd, which, nam) into
two functions do_sys_get{peer,sock}name(fd, nam).
- move PRU_PEERADDR handling into do_sys_getpeername() from
do_sys_getsockname()
- have svr4_stream directly call do_sys_get{sock,peer}name()
respectively instead of providing `which' & fix a DPRINTF string
that incorrectly wrote "getpeername" when it meant "getsockname"
- fix sys_getpeername() and sys_getsockname() to call
do_sys_get{sock,peer}name() without `which' and `lwp' & adjust
comments
- bump kernel version for removal of lwp & which parameters from
do_sys_getsockname()
note: future cleanup to remove struct mbuf * abuse in
xxx_{peer,sock}name()
still to come, not done in this commit since it is easier to do post
split.
patch reviewed by rmind
welcome to 6.99.47
were not filling in struct stat.
decision made after further discussion with rmind and investigation of
how other operating systems behave. soo_stat() is doing just enough to
be able to call what gets returned valid and thus justifys a return of
success.
additional review will be done to determine of the pr_stat functions
that were already returning EOPNOTSUPP can be considered successful with
what soo_stat() is doing.
not fill in struct stat instead of returning success.
* in pr_stat remove all checks for non-NULL so->so_pcb except where the
pcb is actually used (i.e. cases where we don't return EOPNOTSUPP).
proposed on tech-net@
separate xxx_stat(struct socket *, struct stat *) functions.
* replace calls using pr_generic with req == PRU_SENSE with pr_stat().
further change will follow that cleans up the pattern used to extract the
pcb and test for its presence.
reviewed by rmind
nd6_dad_start uses callout when xtick > 0 while doesn't when
xtick == 0. So if we pass a random value ranging from 0 to N,
nd6_dad_start uses callout randomly. This behavior makes
debugging difficult.
Discussed in http://mail-index.netbsd.org/tech-kern/2014/06/25/msg017278.html
abuse of pointer to struct mbuf type.
param2 changed to u_long type and uses parameter name 'cmd' (ioctl command)
param3 changed to void * type and uses parameter name 'data'
param4 changed to struct ifnet * and uses parameter name 'ifp'
param5 has been removed (formerly struct lwp *) and uses of 'l' have been
replaced with curlwp from curproc(9).
callers have had (now unnecessary) casts to struct mbuf * removed, called
code has had (now unnecessary) casts to u_long, void * and struct ifnet *
respectively removed.
reviewed by rmind@
`control' to `ifp' after split from xxx_usrreq().
sys_socket.c
fix wrapping of arguments to be consistent with other function calls
in the file after replacing pr_usrreq() call with pr_ioctl() which
required one less argument.
link_proto.c
fix indentation of parameters in link_ioctl() prototype to be
consistent with the rest of the file.
discussed with rmind@
into separate xxx_ioctl() functions.
* place KASSERT(req != PRU_CONTROL) inside xxx_userreq() as it is now
inappropriate for req = PRU_CONTROL in xxx_userreq().
* replace calls to pr_generic() with req = PRU_CONTROL with pr_ioctl().
* remove & fixup references to PRU_CONTROL xxx_userreq() function comments.
* fix various comments references for xxx_userreq() that mentioned
PRU_CONTROL as xxx_userreq() no longer handles the request.
a further change will follow to fix parameter and naming inconsistencies
retained from original code.
Reviewed by rmind@
- net.sdl for the active link-layer adddress (the MAC)
- net.ether.multicast for the Ethernet multicast addresses
- net.inet6.multicast for the IPv6 multicast groups
- net.inet6.multicast_kludge for temporarily removed multicast groups
Use this sysctls for replacing the kmem grovelling in ifmcstat(8).
- Replace ipintrq and ip6intrq with the pktqueue mechanism.
- Eliminate kernel-lock from ipintr() and ip6intr().
- Some preparation work to push softnet_lock out of ipintr().
Discussed on tech-net.
dismantling of pr_usrreq in the protocols; no functional change intended.
PRU_ATTACH/PRU_DETACH changes will follow soon.
Bump for struct protosw. Welcome to 6.99.62!
If enabled, check for a duplicated link-local address and abort enabling
as per RFC 4862, section 5.4.5. If allowed to enable, perform DAD
on the tentative addresses.
Taken from FreeBSD.
friendly (there are only few hooks in the system). Make the structures
opaque and the interface more strict.
- Remove PFIL_HOOKS option by making pfil(9) mandatory.
and into in6_if_link_up.
This fixes a possible panic where link is up but not the interface.
Note that a better solution would be to listen to the routing socket
in the kernel, but I don't know how to do that.
Reachable Router tests for IFF_UP as well.
addresses as detached.
Likewise, when the link state changes to up, mark all detached IPv6
as tentative and start DAD on them.
Advertised router reachability now checks that link state is not down.
This means that when an interface link state changes, the default IPv6
router may change as well.
ip6_insertfraghdr either sets a result parameter or returns an error.
While the caller only uses the result parameter in the non-error case,
knowing that requires cross-module static analysis, and that's not
robust against distant code changes. Therfore, set ip6f to NULL
before the function call that maybe sets it, avoiding a spuruious
warning and changing the future possible bug from an unitialized
dereference to a NULL deferrence.
From "http://tools.ietf.org/html/draft-ietf-6man-ipv6-atomic-fragments-00":
A host that receives an IPv6 packet which includes a Fragment
Header with the "Fragment Offset" equal to 0 and the "M" bit equal
to 0 MUST process such packet in isolation from any other packets/
fragments, even if such packets/fragments contain the same set
{IPV6 Source Address, IPv6 Destination Address, Fragment
Identification}. That is, the Fragment Header of "atomic
fragments" should be removed by the receiving host, and the
resulting packet should be processed as a non-fragmented IPv6
datagram. Additionally, any fragments already queued with the
same set {IPV6 Source Address, IPv6 Destination Address, Fragment
Identification} should not be discarded upon receipt of the
"colliding" IPv6 atomic fragment, since IPv6 atomic fragments do
not really interfere with "normal" fragmented traffic.
Depending on compiler options, this code can be involved in an
(apparently) spurious compiler warning. However, it was not
immediately obvious the the compiler was wrong.
