xref: /freebsd/share/man/man9/epoch.9 (revision 70398c2f86e2fbc2120b4a11d9ec9284a6a395f5)
1.\"
2.\" Copyright (C) 2018 Matthew Macy <mmacy@FreeBSD.org>.
3.\"
4.\" Redistribution and use in source and binary forms, with or without
5.\" modification, are permitted provided that the following conditions
6.\" are met:
7.\" 1. Redistributions of source code must retain the above copyright
8.\"    notice(s), this list of conditions and the following disclaimer as
9.\"    the first lines of this file unmodified other than the possible
10.\"    addition of one or more copyright notices.
11.\" 2. Redistributions in binary form must reproduce the above copyright
12.\"    notice(s), this list of conditions and the following disclaimer in the
13.\"    documentation and/or other materials provided with the distribution.
14.\"
15.\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDER(S) ``AS IS'' AND ANY
16.\" EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
17.\" WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
18.\" DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY
19.\" DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
20.\" (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
21.\" SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
22.\" CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH
25.\" DAMAGE.
26.\"
27.\" $FreeBSD$
28.\"
29.Dd May 13, 2018
30.Dt EPOCH 9
31.Os
32.Sh NAME
33.Nm epoch ,
34.Nm epoch_context ,
35.Nm epoch_alloc ,
36.Nm epoch_free ,
37.Nm epoch_enter ,
38.Nm epoch_exit ,
39.Nm epoch_wait ,
40.Nm epoch_call ,
41.Nm in_epoch ,
42.Nd kernel epoch based reclaimation
43.Sh SYNOPSIS
44.In sys/param.h
45.In sys/proc.h
46.In sys/epoch.h
47.Ft epoch_t
48.Fn epoch_alloc "int flags"
49.Ft void
50.Fn epoch_enter "epoch_t epoch"
51.Ft void
52.Fn epoch_enter_preempt "epoch_t epoch"
53.Ft void
54.Fn epoch_exit "epoch_t epoch"
55.Ft void
56.Fn epoch_exit_preempt "epoch_t epoch"
57.Ft void
58.Fn epoch_wait "epoch_t epoch"
59.Ft void
60.Fn epoch_wait_preempt "epoch_t epoch"
61.Ft void
62.Fn epoch_call "epoch_t epoch" "epoch_context_t ctx" "void (*callback) (epoch_context_t)"
63.Ft int
64.Fn in_epoch "void"
65.Sh DESCRIPTION
66Epochs are used to guarantee liveness and immutability of data by
67deferring reclamation and mutation until a grace period has elapsed.
68Epochs do not have any lock ordering issues. Entering and leaving
69an epoch section will never block.
70.Pp
71Epochs are allocated with
72.Fn epoch_alloc
73and freed with
74.Fn epoch_free .
75The flags passed to epoch_alloc determine whether preemption is
76allowed during a section or not (the dafult), as specified by
77EPOCH_PREEMPT.
78Threads indicate the start of an epoch critical section by calling
79.Fn epoch_enter .
80The end of a critical section is indicated by calling
81.Fn epoch_exit .
82The _preempt variants can be used around code which requires preemption.
83A thread can wait until a grace period has elapsed
84since any threads have entered
85the epoch by calling
86.Fn epoch_wait
87or
88.Fn epoch_wait_preempt ,
89depending on the epoch_type.
90The use of a default epoch type allows one to use
91.Fn epoch_wait
92which is guaranteed to have much shorter completion times since
93we know that none of the threads in an epoch section will be preempted
94before completing its section.
95If the thread can't sleep or is otherwise in a performance sensitive
96path it can ensure that a grace period has elapsed by calling
97.Fn epoch_call
98with a callback with any work that needs to wait for an epoch to elapse.
99Only non-sleepable locks can be acquired during a section protected by
100.Fn epoch_enter_preempt
101and
102.Fn epoch_exit_preempt .
103INVARIANTS can assert that a thread is in an epoch by using
104.Fn in_epoch .
105.Pp
106The epoch API currently does not support sleeping in epoch_preempt sections.
107A caller cannot do epoch_enter recursively on different preemptible epochs. A
108caller should never call
109.Fn epoch_wait
110in the middle of an epoch section as this will lead to a deadlock.
111.Pp
112Note that epochs are not a straight replacement for read locks. Callers
113must use safe list and tailq traversal routines in an epoch (see ck_queue).
114When modifying a list referenced from an epoch section safe removal
115routines must be used and the caller can no longer modify a list entry
116in place. An item to be modified must be handled with copy on write
117and frees must be deferred until after a grace period has elapsed.
118.Sh RETURN VALUES
119.Fn in_epoch
120will return 1 if curthread is in an epoch, 0 otherwise.
121.Sh CAVEATS
122One must be cautious when using
123.Fn epoch_wait_preempt
124threads are pinned during epoch sections so if a thread in a section is then
125preempted by a higher priority compute bound thread on that CPU it can be
126prevented from leaving the section. Thus the wait time for the waiter is
127potentially unbounded.
128.Sh EXAMPLES
129Async free example:
130
131Thread 1:
132.Bd -literal
133int
134in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_laddr *laddr,
135    struct ucred *cred)
136{
137   /* ... */
138   epoch_enter(net_epoch);
139    CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
140        sa = ifa->ifa_addr;
141	if (sa->sa_family != AF_INET)
142	    continue;
143	sin = (struct sockaddr_in *)sa;
144	if (prison_check_ip4(cred, &sin->sin_addr) == 0) {
145	     ia = (struct in_ifaddr *)ifa;
146	     break;
147	}
148    }
149    epoch_exit(net_epoch);
150   /* ... */
151}
152.Ed
153Thread 2:
154.Bd -literal
155void
156ifa_free(struct ifaddr *ifa)
157{
158
159    if (refcount_release(&ifa->ifa_refcnt))
160        epoch_call(net_epoch, &ifa->ifa_epoch_ctx, ifa_destroy);
161}
162
163void
164if_purgeaddrs(struct ifnet *ifp)
165{
166
167    /* .... */
168    IF_ADDR_WLOCK(ifp);
169    CK_STAILQ_REMOVE(&ifp->if_addrhead, ifa, ifaddr, ifa_link);
170    IF_ADDR_WUNLOCK(ifp);
171    ifa_free(ifa);
172}
173.Ed
174.Pp
175Thread 1 traverses the ifaddr list in an epoch. Thread 2 unlinks
176with the corresponding epoch safe macro, marks as logically free,
177and then defers deletion. More general mutation or a synchronous
178free would have to follow a a call to
179.Fn epoch_wait .
180.Sh ERRORS
181None.
182.El
183.Sh SEE ALSO
184.Xr locking 9 ,
185.Xr mtx_pool 9 ,
186.Xr mutex 9 ,
187.Xr rwlock 9 ,
188.Xr sema 9 ,
189.Xr sleep 9 ,
190.Xr sx 9 ,
191.Xr timeout 9
192