The SYSCTL_NODE macro defines a list that stores all child-elements of
that node. If there's no SYSCTL_DECL macro anywhere else, there's no
reason why it shouldn't be static.
effectively negative. Often seen as upstream fastcgi connection timeouts
in nginx when using sendfile over unix domain sockets for communication.
Sendfile(2) may send more bytes then currently allowed by the
hiwatermark of the socket, e.g. because the so_snd sockbuf lock is
dropped after sbspace() call in the kern_sendfile() loop. In this case,
recalculated hiwatermark will overflow. Since lowatermark is renewed
as half of the hiwatermark by sendfile code, and both are unsigned,
the send buffer never reaches the free space requested by lowatermark,
causing indefinite wait in sendfile.
Reviewed by: rwatson
Approved by: re (bz)
MFC after: 2 weeks
VNET socket push back:
try to minimize the number of places where we have to switch vnets
and narrow down the time we stay switched. Add assertions to the
socket code to catch possibly unset vnets as seen in r204147.
While this reduces the number of vnet recursion in some places like
NFS, POSIX local sockets and some netgraph, .. recursions are
impossible to fix.
The current expectations are documented at the beginning of
uipc_socket.c along with the other information there.
Sponsored by: The FreeBSD Foundation
Sponsored by: CK Software GmbH
Reviewed by: jhb
Tested by: zec
Tested by: Mikolaj Golub (to.my.trociny gmail.com)
MFC after: 2 weeks
collect phases. The unp_discard() function executes
unp_externalize_fp(), which might make the socket eligible for gc-ing,
and then, later, taskqueue will close the socket. Since unp_gc()
dropped the list lock to do the malloc, close might happen after the
mark step but before the collection step, causing collection to not
find the socket and miss one array element.
I believe that the race was there before r216158, but the stated
revision made the window much wider by postponing the close to
taskqueue sometimes.
Only process as much array elements as we find the sockets during
second phase of gc [1]. Take linkage lock and recheck the eligibility
of the socket for gc, as well as call fhold() under the linkage lock.
Reported and tested by: jmallett
Submitted by: jmallett [1]
Reviewed by: rwatson, jeff (possibly)
MFC after: 1 week
proper log message for r216150.
MFC after: 1 week
If unix socket has a unix socket attached as the rights that has a
unix socket attached as the rights that has a unix socket attached as
the rights ... Kernel may overflow the stack on attempt to close such
socket.
Only close the rights file in the context of the current close if the
file is not unix domain socket. Otherwise, postpone the work to
taskqueue, preventing unlimited recursion.
The pass of the unix domain sockets over the SCM_RIGHTS message
control is not widely used, and more, the close of the socket with
still attached rights is mostly an application failure. The change
should not affect the performance of typical users of SCM_RIGHTS.
Reviewed by: jeff, rwatson
sockets. This allows for reliable bi-directional datagram communication
over UNIX domain sockets, in contrast to SOCK_DGRAM (M:N, unreliable) or
SOCK_STERAM (bi-directional bytestream). Largely, this reuses existing
UNIX domain socket code. This allows applications requiring record-
oriented semantics to do so reliably via local IPC.
Some implementation notes (also present in XXX comments):
- Currently we lack an sbappend variant able to do datagrams and control
data without doing addresses, so we mark SOCK_SEQPACKET as PR_ADDR.
Adding a new variant will solve this problem.
- UNIX domain sockets on FreeBSD provide back-pressure/flow control
notification for stream sockets by manipulating the send socket
buffer's size during pru_send and pru_rcvd. This trick works less well
for SOCK_SEQPACKET as sosend_generic() uses sb_hiwat not just to
manage blocking, but also to determine maximum datagram size. Fixing
this requires rethinking how back-pressure is done for SOCK_SEQPACKET;
in the mean time, it's possible to get EMSGSIZE when buffers fill,
instead of blocking.
Discussed with: benl
Reviewed by: bz, rpaulo
MFC after: 3 months
Sponsored by: Google
vnet.h, we now use jails (rather than vimages) as the abstraction
for virtualization management, and what remained was specific to
virtual network stacks. Minor cleanups are done in the process,
and comments updated to reflect these changes.
Reviewed by: bz
Approved by: re (vimage blanket)
Instead of locking the local unp followed by the remote unp, use the same
locking model as accept() and read lock the global link lock followed by
the remote unp while fetching the remote sockaddr.
Reported by: Mel Flynn mel.flynn of mailing.thruhere.net
Reviewed by: rwatson
MFC after: 1 week
and used in a large number of files, but also because an increasing number
of incorrect uses of MAC calls were sneaking in due to copy-and-paste of
MAC-aware code without the associated opt_mac.h include.
Discussed with: pjd
count of the number of registered policies.
Rather than unconditionally locking sockets before passing them into MAC,
lock them in the MAC entry points only if mac_policy_count is non-zero.
This avoids locking overhead for a number of socket system calls when no
policies are registered, eliminating measurable overhead for the MAC
Framework for the socket subsystem when there are no active policies.
Possibly socket locks should be acquired by policies if they are required
for socket labels, which would further avoid locking overhead when there
are policies but they don't require labeling of sockets, or possibly
don't even implement socket controls.
Obtained from: TrustedBSD Project
previously always pointing to the default vnet context, to a
dynamically changing thread-local one. The currvnet context
should be set on entry to networking code via CURVNET_SET() macros,
and reverted to previous state via CURVNET_RESTORE(). Recursions
on curvnet are permitted, though strongly discuouraged.
This change should have no functional impact on nooptions VIMAGE
kernel builds, where CURVNET_* macros expand to whitespace.
The curthread->td_vnet (aka curvnet) variable's purpose is to be an
indicator of the vnet context in which the current network-related
operation takes place, in case we cannot deduce the current vnet
context from any other source, such as by looking at mbuf's
m->m_pkthdr.rcvif->if_vnet, sockets's so->so_vnet etc. Moreover, so
far curvnet has turned out to be an invaluable consistency checking
aid: it helps to catch cases when sockets, ifnets or any other
vnet-aware structures may have leaked from one vnet to another.
The exact placement of the CURVNET_SET() / CURVNET_RESTORE() macros
was a result of an empirical iterative process, whith an aim to
reduce recursions on CURVNET_SET() to a minimum, while still reducing
the scope of CURVNET_SET() to networking only operations - the
alternative would be calling CURVNET_SET() on each system call entry.
In general, curvnet has to be set in three typicall cases: when
processing socket-related requests from userspace or from within the
kernel; when processing inbound traffic flowing from device drivers
to upper layers of the networking stack, and when executing
timer-driven networking functions.
This change also introduces a DDB subcommand to show the list of all
vnet instances.
Approved by: julian (mentor)
the removal of NQNFS, but was left in in case it was required for NFSv4.
Since our new NFSv4 client and server can't use it for their
requirements, GC the old mechanism, as well as other unused lease-
related code and interfaces.
Due to its impact on kernel programming and binary interfaces, this
change should not be MFC'd.
Proposed by: jeff
Reviewed by: jeff
Discussed with: rmacklem, zach loafman @ isilon
locks: a global list/counter/generation counter protected by a new
mutex unp_list_lock, and a global linkage rwlock, unp_global_rwlock,
which protects the connections between UNIX domain sockets.
This eliminates conditional lock acquisition that was previously a
property of the global lock being held over sonewconn() leading to a
call to uipc_attach(), which also required the global lock, but
couldn't rely on it as other paths existed to uipc_attach() that
didn't hold it: now uipc_attach() uses only the list lock, which
follows the linkage lock in the lock order. It may also reduce
contention on the global lock for some workloads.
Add global UNIX domain socket locks to hard-coded witness lock
order.
MFC after: 1 week
Discussed with: kris
descriptor pointer in unp_freerights: we can no longer recurse into
unp_gc due to unp_gc being invoked in a deferred way, but it's still
a good idea.
MFC after: 3 days
- Staticize and locally prototype functions uipc_ctloutput(), unp_dispose(),
unp_init(), and unp_externalize(), none of which have been required
outside of uipc_usrreq.c since uipc_proto.c was removed.
- Remove stale prototype for uipc_usrreq(), which has not existed in the
code since 1997
- Forward declare and staticize uipc_usrreqs structure in uipc_usrreq.c and
not un.h.
- Comment on why uipc_connect2() is still non-static -- it is used directly
by fifofs.
- Remove stale comments, tidy up whitespace.
MFC after: 3 days (where applicable)
soun->sun_path isn't a null-terminated string. As UNIX(4) states, "the
terminating NUL is not part of the address." Since strlcpy has to return
"the total length of the string [it] tried to create," it walks off the end
of soun->sun_path looking for a \0.
This reverts r105332.
Reported by: Ryan Stone
unp_connect(): it is expected to return with the lock held, and two
possible error paths otherwise returned with it unlocked.
The fix committed here is slightly different from the patch in the
PR, but along an alternative line suggested in the PR.
PR: 119778
MFC after: 3 days
Submitted by: James Juran <james dot juran at baesystems dot com>
conjuction with 'thread' argument passing which is always curthread.
Remove the unuseful extra-argument and pass explicitly curthread to lower
layer functions, when necessary.
KPI results broken by this change, which should affect several ports, so
version bumping and manpage update will be further committed.
Tested by: kris, pho, Diego Sardina <siarodx at gmail dot com>
- Clear all of the gc flags before doing a run. Stale flags were causing
us to skip some descriptors.
- If a unp socket has been marked REF in a gc pass it can't be dead.
Found by: rwatson's test tool.
- Introduce a finit() which is used to initailize the fields of struct file
in such a way that the ops vector is only valid after the data, type,
and flags are valid.
- Protect f_flag and f_count with atomic operations.
- Remove the global list of all files and associated accounting.
- Rewrite the unp garbage collection such that it no longer requires
the global list of all files and instead uses a list of all unp sockets.
- Mark sockets in the accept queue so we don't incorrectly gc them.
Tested by: kris, pho
from Mac OS X Leopard--rationalize naming for entry points to
the following general forms:
mac_<object>_<method/action>
mac_<object>_check_<method/action>
The previous naming scheme was inconsistent and mostly
reversed from the new scheme. Also, make object types more
consistent and remove spaces from object types that contain
multiple parts ("posix_sem" -> "posixsem") to make mechanical
parsing easier. Introduce a new "netinet" object type for
certain IPv4/IPv6-related methods. Also simplify, slightly,
some entry point names.
All MAC policy modules will need to be recompiled, and modules
not updates as part of this commit will need to be modified to
conform to the new KPI.
Sponsored by: SPARTA (original patches against Mac OS X)
Obtained from: TrustedBSD Project, Apple Computer
and flags with an sxlock. This leads to a significant and measurable
performance improvement as a result of access to shared locking for
frequent lookup operations, reduced general overhead, and reduced overhead
in the event of contention. All of these are imported for threaded
applications where simultaneous access to a shared file descriptor array
occurs frequently. Kris has reported 2x-4x transaction rate improvements
on 8-core MySQL benchmarks; smaller improvements can be expected for many
workloads as a result of reduced overhead.
- Generally eliminate the distinction between "fast" and regular
acquisisition of the filedesc lock; the plan is that they will now all
be fast. Change all locking instances to either shared or exclusive
locks.
- Correct a bug (pointed out by kib) in fdfree() where previously msleep()
was called without the mutex held; sx_sleep() is now always called with
the sxlock held exclusively.
- Universally hold the struct file lock over changes to struct file,
rather than the filedesc lock or no lock. Always update the f_ops
field last. A further memory barrier is required here in the future
(discussed with jhb).
- Improve locking and reference management in linux_at(), which fails to
properly acquire vnode references before using vnode pointers. Annotate
improper use of vn_fullpath(), which will be replaced at a future date.
In fcntl(), we conservatively acquire an exclusive lock, even though in
some cases a shared lock may be sufficient, which should be revisited.
The dropping of the filedesc lock in fdgrowtable() is no longer required
as the sxlock can be held over the sleep operation; we should consider
removing that (pointed out by attilio).
Tested by: kris
Discussed with: jhb, kris, attilio, jeff
uipc_send in cases where only a global read lock is held by breaking
them out and avoiding the unpcb lock acquire in the common case. This
avoids deadlocks which manifested with X11, and should also marginally
further improve performance.
Reported by: sepotvin, brooks
concurrency:
- Add per-unpcb mutexes protecting unpcb connection state, fields, etc.
- Replace global UNP mutex with a global UNP rwlock, which will protect the
UNIX domain socket connection topology, v_socket, and be acquired
exclusively before acquiring more than per-unpcb at a time in order to
avoid lock order issues.
In performance measurements involving MySQL, this change has little or no
overhead on UP (+/- 1%), but leads to a significant (5%-30%) improvement in
multi-processor measurements using the sysbench and supersmack benchmarks.
Much testing by: kris
Approved by: re (kensmith)
