xref: /freebsd/share/man/man9/zone.9 (revision a831d7d1c538b93ffec095657d9a6e6e778db195)
1.\"-
2.\" Copyright (c) 2001 Dag-Erling Coïdan Smørgrav
3.\" All rights reserved.
4.\"
5.\" Redistribution and use in source and binary forms, with or without
6.\" modification, are permitted provided that the following conditions
7.\" are met:
8.\" 1. Redistributions of source code must retain the above copyright
9.\"    notice, this list of conditions and the following disclaimer.
10.\" 2. Redistributions in binary form must reproduce the above copyright
11.\"    notice, this list of conditions and the following disclaimer in the
12.\"    documentation and/or other materials provided with the distribution.
13.\"
14.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17.\" ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24.\" SUCH DAMAGE.
25.\"
26.\" $FreeBSD$
27.\"
28.Dd February 7, 2014
29.Dt ZONE 9
30.Os
31.Sh NAME
32.Nm uma_zcreate ,
33.Nm uma_zalloc ,
34.Nm uma_zalloc_arg ,
35.Nm uma_zfree ,
36.Nm uma_zfree_arg ,
37.Nm uma_find_refcnt ,
38.Nm uma_zdestroy ,
39.Nm uma_zone_set_max,
40.Nm uma_zone_get_max,
41.Nm uma_zone_get_cur,
42.Nm uma_zone_set_warning
43.Nd zone allocator
44.Sh SYNOPSIS
45.In sys/param.h
46.In sys/queue.h
47.In vm/uma.h
48.Ft uma_zone_t
49.Fo uma_zcreate
50.Fa "char *name" "int size"
51.Fa "uma_ctor ctor" "uma_dtor dtor" "uma_init uminit" "uma_fini fini"
52.Fa "int align" "uint16_t flags"
53.Fc
54.Ft "void *"
55.Fn uma_zalloc "uma_zone_t zone" "int flags"
56.Ft "void *"
57.Fn uma_zalloc_arg "uma_zone_t zone" "void *arg" "int flags"
58.Ft void
59.Fn uma_zfree "uma_zone_t zone" "void *item"
60.Ft void
61.Fn uma_zfree_arg "uma_zone_t zone" "void *item" "void *arg"
62.Ft "uint32_t *"
63.Fn uma_find_refcnt "uma_zone_t zone" "void *item"
64.Ft void
65.Fn uma_zdestroy "uma_zone_t zone"
66.Ft int
67.Fn uma_zone_set_max "uma_zone_t zone" "int nitems"
68.Ft int
69.Fn uma_zone_get_max "uma_zone_t zone"
70.Ft int
71.Fn uma_zone_get_cur "uma_zone_t zone"
72.Ft void
73.Fn uma_zone_set_warning "uma_zone_t zone" "const char *warning"
74.In sys/sysctl.h
75.Fn SYSCTL_UMA_MAX parent nbr name access zone descr
76.Fn SYSCTL_ADD_UMA_MAX ctx parent nbr name access zone descr
77.Fn SYSCTL_UMA_CUR parent nbr name access zone descr
78.Fn SYSCTL_ADD_UMA_CUR ctx parent nbr name access zone descr
79.Sh DESCRIPTION
80The zone allocator provides an efficient interface for managing
81dynamically-sized collections of items of similar size.
82The zone allocator can work with preallocated zones as well as with
83runtime-allocated ones, and is therefore available much earlier in the
84boot process than other memory management routines.
85.Pp
86A zone is an extensible collection of items of identical size.
87The zone allocator keeps track of which items are in use and which
88are not, and provides functions for allocating items from the zone and
89for releasing them back (which makes them available for later use).
90.Pp
91After the first allocation of an item,
92it will have been cleared to zeroes, however subsequent allocations
93will retain the contents as of the last free.
94.Pp
95The
96.Fn uma_zcreate
97function creates a new zone from which items may then be allocated from.
98The
99.Fa name
100argument is a text name of the zone for debugging and stats; this memory
101should not be freed until the zone has been deallocated.
102.Pp
103The
104.Fa ctor
105and
106.Fa dtor
107arguments are callback functions that are called by
108the uma subsystem at the time of the call to
109.Fn uma_zalloc
110and
111.Fn uma_zfree
112respectively.
113Their purpose is to provide hooks for initializing or
114destroying things that need to be done at the time of the allocation
115or release of a resource.
116A good usage for the
117.Fa ctor
118and
119.Fa dtor
120callbacks
121might be to adjust a global count of the number of objects allocated.
122.Pp
123The
124.Fa uminit
125and
126.Fa fini
127arguments are used to optimize the allocation of
128objects from the zone.
129They are called by the uma subsystem whenever
130it needs to allocate or free several items to satisfy requests or memory
131pressure.
132A good use for the
133.Fa uminit
134and
135.Fa fini
136callbacks might be to
137initialize and destroy mutexes contained within the object.
138This would
139allow one to re-use already initialized mutexes when an object is returned
140from the uma subsystem's object cache.
141They are not called on each call to
142.Fn uma_zalloc
143and
144.Fn uma_zfree
145but rather in a batch mode on several objects.
146.Pp
147The
148.Fa flags
149argument of the
150.Fn uma_zcreate
151is a subset of the following flags:
152.Bl -tag -width "foo"
153.It Dv UMA_ZONE_NOFREE
154Slabs of the zone are never returned back to VM.
155.It Dv UMA_ZONE_REFCNT
156Each item in the zone would have internal reference counter associated with it.
157See
158.Fn uma_find_refcnt .
159.It Dv UMA_ZONE_NODUMP
160Pages belonging to the zone will not be included into mini-dumps.
161.It Dv UMA_ZONE_PCPU
162An allocation from zone would have
163.Va mp_ncpu
164shadow copies, that are privately assigned to CPUs.
165A CPU can address its private copy using base allocation address plus
166multiple of current CPU id and
167.Fn sizeof "struct pcpu" :
168.Bd -literal -offset indent
169foo_zone = uma_zcreate(..., UMA_ZONE_PCPU);
170 ...
171foo_base = uma_zalloc(foo_zone, ...);
172 ...
173critical_enter();
174foo_pcpu = (foo_t *)zpcpu_get(foo_base);
175/* do something with foo_pcpu */
176critical_exit();
177.Ed
178.It Dv UMA_ZONE_OFFPAGE
179By default book-keeping of items within a slab is done in the slab page itself.
180This flag explicitly tells subsystem that book-keeping structure should be
181allocated separately from special internal zone.
182This flag requires either
183.Dv UMA_ZONE_VTOSLAB
184or
185.Dv UMA_ZONE_HASH ,
186since subsystem requires a mechanism to find a book-keeping structure
187to an item beeing freed.
188The subsystem may choose to prefer offpage book-keeping for certain zones
189implicitly.
190.It Dv UMA_ZONE_ZINIT
191The zone will have its
192.Ft uma_init
193method set to internal method that initializes a new allocated slab
194to all zeros.
195Do not mistake
196.Ft uma_init
197method with
198.Ft uma_ctor .
199A zone with
200.Dv UMA_ZONE_ZINIT
201flag would not return zeroed memory on every
202.Fn uma_zalloc .
203.It Dv UMA_ZONE_HASH
204The zone should use an internal hash table to find slab book-keeping
205structure where an allocation being freed belongs to.
206.It Dv UMA_ZONE_VTOSLAB
207The zone should use special field of
208.Vt vm_page_t
209to find slab book-keeping structure where an allocation being freed belongs to.
210.It Dv UMA_ZONE_MALLOC
211The zone is for the
212.Xr malloc 9
213subsystem.
214.It Dv UMA_ZONE_VM
215The zone is for the VM subsystem.
216.El
217.Pp
218To allocate an item from a zone, simply call
219.Fn uma_zalloc
220with a pointer to that zone
221and set the
222.Fa flags
223argument to selected flags as documented in
224.Xr malloc 9 .
225It will return a pointer to an item if successful,
226or
227.Dv NULL
228in the rare case where all items in the zone are in use and the
229allocator is unable to grow the zone
230and
231.Dv M_NOWAIT
232is specified.
233.Pp
234Items are released back to the zone from which they were allocated by
235calling
236.Fn uma_zfree
237with a pointer to the zone and a pointer to the item.
238If
239.Fa item
240is
241.Dv NULL ,
242then
243.Fn uma_zfree
244does nothing.
245.Pp
246The variations
247.Fn uma_zalloc_arg
248and
249.Fn uma_zfree_arg
250allow to
251specify an argument for the
252.Dv ctor
253and
254.Dv dtor
255functions, respectively.
256.Pp
257If zone was created with
258.Dv UMA_ZONE_REFCNT
259flag, then pointer to reference counter for an item can be retrieved with
260help of the
261.Fn uma_find_refcnt
262function.
263.Pp
264Created zones,
265which are empty,
266can be destroyed using
267.Fn uma_zdestroy ,
268freeing all memory that was allocated for the zone.
269All items allocated from the zone with
270.Fn uma_zalloc
271must have been freed with
272.Fn uma_zfree
273before.
274.Pp
275The
276.Fn uma_zone_set_max
277function limits the number of items
278.Pq and therefore memory
279that can be allocated to
280.Fa zone .
281The
282.Fa nitems
283argument specifies the requested upper limit number of items.
284The effective limit is returned to the caller, as it may end up being higher
285than requested due to the implementation rounding up to ensure all memory pages
286allocated to the zone are utilised to capacity.
287The limit applies to the total number of items in the zone, which includes
288allocated items, free items and free items in the per-cpu caches.
289On systems with more than one CPU it may not be possible to allocate
290the specified number of items even when there is no shortage of memory,
291because all of the remaining free items may be in the caches of the
292other CPUs when the limit is hit.
293.Pp
294The
295.Fn uma_zone_get_max
296function returns the effective upper limit number of items for a zone.
297.Pp
298The
299.Fn uma_zone_get_cur
300function returns the approximate current occupancy of the zone.
301The returned value is approximate because appropriate synchronisation to
302determine an exact value is not performed by the implementation.
303This ensures low overhead at the expense of potentially stale data being used
304in the calculation.
305.Pp
306The
307.Fn uma_zone_set_warning
308function sets a warning that will be printed on the system console when the
309given zone becomes full and fails to allocate an item.
310The warning will be printed not often than every five minutes.
311Warnings can be turned off globally by setting the
312.Va vm.zone_warnings
313sysctl tunable to
314.Va 0 .
315.Pp
316The
317.Fn SYSCTL_UMA_MAX parent nbr name access zone descr
318macro declares a static
319.Xr sysctl
320oid that exports the effective upper limit number of items for a zone.
321The
322.Fa zone
323argument should be a pointer to
324.Vt uma_zone_t .
325A read of the oid returns value obtained through
326.Fn uma_zone_get_max .
327A write to the oid sets new value via
328.Fn uma_zone_set_max .
329The
330.Fn SYSCTL_ADD_UMA_MAX ctx parent nbr name access zone descr
331macro is provided to create this type of oid dynamically.
332.Pp
333The
334.Fn SYSCTL_UMA_CUR parent nbr name access zone descr
335macro declares a static read only
336.Xr sysctl
337oid that exports the approximate current occupancy of the zone.
338The
339.Fa zone
340argument should be a pointer to
341.Vt uma_zone_t .
342A read of the oid returns value obtained through
343.Fn uma_zone_get_cur .
344The
345.Fn SYSCTL_ADD_UMA_CUR ctx parent nbr name zone descr
346macro is provided to create this type of oid dynamically.
347.Sh RETURN VALUES
348The
349.Fn uma_zalloc
350function returns a pointer to an item, or
351.Dv NULL
352if the zone ran out of unused items
353and
354.Dv M_NOWAIT
355was specified.
356.Sh SEE ALSO
357.Xr malloc 9
358.Sh HISTORY
359The zone allocator first appeared in
360.Fx 3.0 .
361It was radically changed in
362.Fx 5.0
363to function as a slab allocator.
364.Sh AUTHORS
365.An -nosplit
366The zone allocator was written by
367.An John S. Dyson .
368The zone allocator was rewritten in large parts by
369.An Jeff Roberson Aq Mt jeff@FreeBSD.org
370to function as a slab allocator.
371.Pp
372This manual page was written by
373.An Dag-Erling Sm\(/orgrav Aq Mt des@FreeBSD.org .
374Changes for UMA by
375.An Jeroen Ruigrok van der Werven Aq Mt asmodai@FreeBSD.org .
376