man/man9/zone.9

.\"-
.\" Copyright (c) 2001 Dag-Erling Smørgrav
.\" All rights reserved.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, are permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\"    notice, this list of conditions and the following disclaimer.
.\" 2. Redistributions in binary form must reproduce the above copyright
.\"    notice, this list of conditions and the following disclaimer in the
.\"    documentation and/or other materials provided with the distribution.
.\"
.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\" ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\" SUCH DAMAGE.
.\"
.Dd January 16, 2023
.Dt UMA 9
.Os
.Sh NAME
.Nm UMA
.Nd general-purpose kernel object allocator
.Sh SYNOPSIS
.In sys/param.h
.In sys/queue.h
.In vm/uma.h
.Bd -literal
typedef int (*uma_ctor)(void *mem, int size, void *arg, int flags);
typedef void (*uma_dtor)(void *mem, int size, void *arg);
typedef int (*uma_init)(void *mem, int size, int flags);
typedef void (*uma_fini)(void *mem, int size);
typedef int (*uma_import)(void *arg, void **store, int count, int domain,
    int flags);
typedef void (*uma_release)(void *arg, void **store, int count);
typedef void *(*uma_alloc)(uma_zone_t zone, vm_size_t size, int domain,
    uint8_t *pflag, int wait);
typedef void (*uma_free)(void *item, vm_size_t size, uint8_t pflag);

.Ed
.Ft uma_zone_t
.Fo uma_zcreate
.Fa "char *name" "size_t size"
.Fa "uma_ctor ctor" "uma_dtor dtor" "uma_init zinit" "uma_fini zfini"
.Fa "int align" "uint16_t flags"
.Fc
.Ft uma_zone_t
.Fo uma_zcache_create
.Fa "char *name" "int size"
.Fa "uma_ctor ctor" "uma_dtor dtor" "uma_init zinit" "uma_fini zfini"
.Fa "uma_import zimport" "uma_release zrelease"
.Fa "void *arg" "int flags"
.Fc
.Ft uma_zone_t
.Fo uma_zsecond_create
.Fa "char *name"
.Fa "uma_ctor ctor" "uma_dtor dtor" "uma_init zinit" "uma_fini zfini"
.Fa "uma_zone_t master"
.Fc
.Ft void
.Fn uma_zdestroy "uma_zone_t zone"
.Ft "void *"
.Fn uma_zalloc "uma_zone_t zone" "int flags"
.Ft "void *"
.Fn uma_zalloc_arg "uma_zone_t zone" "void *arg" "int flags"
.Ft "void *"
.Fn uma_zalloc_domain "uma_zone_t zone" "void *arg" "int domain" "int flags"
.Ft "void *"
.Fn uma_zalloc_pcpu "uma_zone_t zone" "int flags"
.Ft "void *"
.Fn uma_zalloc_pcpu_arg "uma_zone_t zone" "void *arg" "int flags"
.Ft "void *"
.Fn uma_zalloc_smr "uma_zone_t zone" "int flags"
.Ft void
.Fn uma_zfree "uma_zone_t zone" "void *item"
.Ft void
.Fn uma_zfree_arg "uma_zone_t zone" "void *item" "void *arg"
.Ft void
.Fn uma_zfree_pcpu "uma_zone_t zone" "void *item"
.Ft void
.Fn uma_zfree_pcpu_arg "uma_zone_t zone" "void *item" "void *arg"
.Ft void
.Fn uma_zfree_smr "uma_zone_t zone" "void *item"
.Ft void
.Fn uma_prealloc "uma_zone_t zone" "int nitems"
.Ft void
.Fn uma_zone_reserve "uma_zone_t zone" "int nitems"
.Ft void
.Fn uma_zone_reserve_kva "uma_zone_t zone" "int nitems"
.Ft void
.Fn uma_reclaim "int req"
.Ft void
.Fn uma_reclaim_domain "int req" "int domain"
.Ft void
.Fn uma_zone_reclaim "uma_zone_t zone" "int req"
.Ft void
.Fn uma_zone_reclaim_domain "uma_zone_t zone" "int req" "int domain"
.Ft void
.Fn uma_zone_set_allocf "uma_zone_t zone" "uma_alloc allocf"
.Ft void
.Fn uma_zone_set_freef "uma_zone_t zone" "uma_free freef"
.Ft int
.Fn uma_zone_set_max "uma_zone_t zone" "int nitems"
.Ft void
.Fn uma_zone_set_maxcache "uma_zone_t zone" "int nitems"
.Ft int
.Fn uma_zone_get_max "uma_zone_t zone"
.Ft int
.Fn uma_zone_get_cur "uma_zone_t zone"
.Ft void
.Fn uma_zone_set_warning "uma_zone_t zone" "const char *warning"
.Ft void
.Fn uma_zone_set_maxaction "uma_zone_t zone" "void (*maxaction)(uma_zone_t)"
.Ft smr_t
.Fn uma_zone_get_smr "uma_zone_t zone"
.Ft void
.Fn uma_zone_set_smr "uma_zone_t zone" "smr_t smr"
.In sys/sysctl.h
.Fn SYSCTL_UMA_MAX parent nbr name access zone descr
.Fn SYSCTL_ADD_UMA_MAX ctx parent nbr name access zone descr
.Fn SYSCTL_UMA_CUR parent nbr name access zone descr
.Fn SYSCTL_ADD_UMA_CUR ctx parent nbr name access zone descr
.Sh DESCRIPTION
UMA (Universal Memory Allocator) provides an efficient interface for managing
dynamically-sized collections of items of identical size, referred to as zones.
Zones keep track of which items are in use and which
are not, and UMA provides functions for allocating items from a zone and
for releasing them back, making them available for subsequent allocation requests.
Zones maintain per-CPU caches with linear scalability on SMP
systems as well as round-robin and first-touch policies for NUMA
systems.
The number of items cached per CPU is bounded, and each zone additionally
maintains an unbounded cache of items that is used to quickly satisfy
per-CPU cache allocation misses.
.Pp
Two types of zones exist: regular zones and cache zones.
In a regular zone, items are allocated from a slab, which is one or more
virtually contiguous memory pages that have been allocated from the kernel's
page allocator.
Internally, slabs are managed by a UMA keg, which is responsible for allocating
slabs and keeping track of their usage by one or more zones.
In typical usage, there is one keg per zone, so slabs are not shared among
multiple zones.
.Pp
Normal zones import items from a keg, and release items back to that keg if
requested.
Cache zones do not have a keg, and instead use custom import and release
methods.
For example, some collections of kernel objects are statically allocated
at boot-time, and the size of the collection does not change.
A cache zone can be used to implement an efficient allocator for the objects in
such a collection.
.Pp
The
.Fn uma_zcreate
and
.Fn uma_zcache_create
functions create a new regular zone and cache zone, respectively.
The
.Fn uma_zsecond_create
function creates a regular zone which shares the keg of the zone
specified by the
.Fa master
argument.
The
.Fa name
argument is a text name of the zone for debugging and stats; this memory
should not be freed until the zone has been deallocated.
.Pp
The
.Fa ctor
and
.Fa dtor
arguments are callback functions that are called by
the UMA subsystem at the time of the call to
.Fn uma_zalloc
and
.Fn uma_zfree
respectively.
Their purpose is to provide hooks for initializing or
destroying things that need to be done at the time of the allocation
or release of a resource.
A good usage for the
.Fa ctor
and
.Fa dtor
callbacks might be to initialize a data structure embedded in the item,
such as a
.Xr queue 3
head.
.Pp
The
.Fa zinit
and
.Fa zfini
arguments are used to optimize the allocation of items from the zone.
They are called by the UMA subsystem whenever
it needs to allocate or free items to satisfy requests or memory pressure.
A good use for the
.Fa zinit
and
.Fa zfini
callbacks might be to
initialize and destroy a mutex contained within an item.
This would allow one to avoid destroying and re-initializing the mutex
each time the item is freed and re-allocated.
They are not called on each call to
.Fn uma_zalloc
and
.Fn uma_zfree
but rather when an item is imported into a zone's cache, and when a zone
releases an item to the slab allocator, typically as a response to memory
pressure.
.Pp
For
.Fn uma_zcache_create ,
the
.Fa zimport
and
.Fa zrelease
functions are called to import items into the zone and to release items
from the zone, respectively.
The
.Fa zimport
function should store pointers to items in the
.Fa store
array, which contains a maximum of
.Fa count
entries.
The function must return the number of imported items, which may be less than
the maximum.
Similarly, the
.Fa store
parameter to the
.Fa zrelease
function contains an array of
.Fa count
pointers to items.
The
.Fa arg
parameter passed to
.Fn uma_zcache_create
is provided to the import and release functions.
The
.Fa domain
parameter to
.Fa zimport
specifies the requested
.Xr numa 4
domain for the allocation.
It is either a NUMA domain number or the special value
.Dv UMA_ANYDOMAIN .
.Pp
The
.Fa flags
argument of
.Fn uma_zcreate
and
.Fn uma_zcache_create
is a subset of the following flags:
.Bl -tag -width "foo"
.It Dv UMA_ZONE_NOFREE
Slabs allocated to the zone's keg are never freed.
.It Dv UMA_ZONE_NODUMP
Pages belonging to the zone will not be included in minidumps.
.It Dv UMA_ZONE_PCPU
An allocation from zone would have
.Va mp_ncpu
shadow copies, that are privately assigned to CPUs.
A CPU can address its private copy using base the allocation address plus
a multiple of the current CPU ID and
.Fn sizeof "struct pcpu" :
.Bd -literal -offset indent
foo_zone = uma_zcreate(..., UMA_ZONE_PCPU);
 ...
foo_base = uma_zalloc(foo_zone, ...);
 ...
critical_enter();
foo_pcpu = (foo_t *)zpcpu_get(foo_base);
/* do something with foo_pcpu */
critical_exit();

.Ed
Note that
.Dv M_ZERO
cannot be used when allocating items from a PCPU zone.
To obtain zeroed memory from a PCPU zone, use the
.Fn uma_zalloc_pcpu
function and its variants instead, and pass
.Dv M_ZERO .
.It Dv UMA_ZONE_NOTOUCH
The UMA subsystem may not directly touch (i.e. read or write) the slab memory.
Otherwise, by default, book-keeping of items within a slab may be done in the
slab page itself, and
.Dv INVARIANTS
kernels may also do use-after-free checking by accessing the slab memory.
.It Dv UMA_ZONE_ZINIT
The zone will have its
.Ft uma_init
method set to internal method that initializes a new allocated slab
to all zeros.
Do not mistake
.Ft uma_init
method with
.Ft uma_ctor .
A zone with
.Dv UMA_ZONE_ZINIT
flag would not return zeroed memory on every
.Fn uma_zalloc .
.It Dv UMA_ZONE_NOTPAGE
An allocator function will be supplied with
.Fn uma_zone_set_allocf
and the memory that it returns may not be kernel virtual memory backed by VM
pages in the page array.
.It Dv UMA_ZONE_MALLOC
The zone is for the
.Xr malloc 9
subsystem.
.It Dv UMA_ZONE_VM
The zone is for the VM subsystem.
.It Dv UMA_ZONE_CONTIG
Items in this zone must be contiguous in physical address space.
Items will follow normal alignment constraints and may span page boundaries
between pages with contiguous physical addresses.
.It Dv UMA_ZONE_UNMANAGED
By default, UMA zone caches are shrunk to help resolve free page shortages.
Cached items that have not been used for a long period may also be freed from
zone.
When this flag is set, the system will not reclaim memory from the zone's
caches.
.It Dv UMA_ZONE_SMR
Create a zone whose items will be synchronized using the
.Xr smr 9
mechanism.
Upon creation the zone will have an associated
.Dt smr_t
structure which can be fetched using
.Fn uma_zone_get_smr .
.El
.Pp
Zones can be destroyed using
.Fn uma_zdestroy ,
freeing all memory that is cached in the zone.
All items allocated from the zone must be freed to the zone before the zone
may be safely destroyed.
.Pp
To allocate an item from a zone, simply call
.Fn uma_zalloc
with a pointer to that zone and set the
.Fa flags
argument to selected flags as documented in
.Xr malloc 9 .
It will return a pointer to an item if successful, or
.Dv NULL
in the rare case where all items in the zone are in use and the
allocator is unable to grow the zone and
.Dv M_NOWAIT
is specified.
.Pp
Items are released back to the zone from which they were allocated by
calling
.Fn uma_zfree
with a pointer to the zone and a pointer to the item.
If
.Fa item
is
.Dv NULL ,
then
.Fn uma_zfree
does nothing.
.Pp
The variants
.Fn uma_zalloc_arg
and
.Fn uma_zfree_arg
allow callers to
specify an argument for the
.Dv ctor
and
.Dv dtor
functions of the zone, respectively.
The variants
.Fn uma_zalloc_pcpu
and
.Fn uma_zfree_pcpu
allocate and free
.Va mp_ncpu
shadow copies as described for
.Dv UMA_ZONE_PCPU .
If
.Fa item
is
.Dv NULL ,
then
.Fn uma_zfree_pcpu
does nothing.
.Pp
The
.Fn uma_zalloc_smr
and
.Fn uma_zfree_smr
functions allocate and free items from an SMR-enabled zone, that is,
a zone created with
.Dv UMA_ZONE_SMR
or a zone that has had
.Fn uma_zone_set_smr
called.
.Pp
The
.Fn uma_zalloc_domain
function allows callers to specify a fixed
.Xr numa 4
domain to allocate from.
This uses a guaranteed but slow path in the allocator which reduces
concurrency.
.Pp
The
.Fn uma_prealloc
function allocates slabs for the requested number of items, typically following
the initial creation of a zone.
Subsequent allocations from the zone will be satisfied using the pre-allocated
slabs.
Note that slab allocation is performed with the
.Dv M_WAITOK
flag, so
.Fn uma_prealloc
may sleep.
.Pp
The
.Fn uma_zone_reserve
function sets the number of reserved items for the zone.
.Fn uma_zalloc
and variants will ensure that the zone contains at least the reserved number
of free items.
Reserved items may be allocated by specifying
.Dv M_USE_RESERVE
in the allocation request flags.
.Fn uma_zone_reserve
does not perform any pre-allocation by itself.
.Pp
The
.Fn uma_zone_reserve_kva
function pre-allocates kernel virtual address space for the requested
number of items.
Subsequent allocations from the zone will be satisfied using the pre-allocated
address space.
Note that unlike
.Fn uma_zone_reserve ,
.Fn uma_zone_reserve_kva
does not restrict the use of the pre-allocation to
.Dv M_USE_RESERVE
requests.
.Pp
The
.Fn uma_reclaim
and
.Fn uma_zone_reclaim
functions reclaim cached items from UMA zones, releasing unused memory.
The
.Fn uma_reclaim
function reclaims items from all regular zones, while
.Fn uma_zone_reclaim
reclaims items only from the specified zone.
The
.Fa req
parameter must be one of three values which specify how aggressively
items are to be reclaimed:
.Bl -tag -width indent
.It Dv UMA_RECLAIM_TRIM
Reclaim items only in excess of the zone's estimated working set size.
The working set size is periodically updated and tracks the recent history
of the zone's usage.
.It Dv UMA_RECLAIM_DRAIN
Reclaim all items from the unbounded cache.
Free items in the per-CPU caches are left alone.
.It Dv UMA_RECLAIM_DRAIN_CPU
Reclaim all cached items.
.El
The
.Fn uma_reclaim_domain
and
.Fn uma_zone_reclaim_domain
functions apply only to items allocated from the specified domain.
In the case of domains using a round-robin NUMA policy, cached items from all
domains are freed to the keg, but only slabs from the specific domain will
be freed.
.Pp
The
.Fn uma_zone_set_allocf
and
.Fn uma_zone_set_freef
functions allow a zone's default slab allocation and free functions to be
overridden.
This is useful if memory with special constraints such as attributes,
alignment, or address ranges must be used.
.Pp
The
.Fn uma_zone_set_max
function limits the number of items
.Pq and therefore memory
that can be allocated to
.Fa zone .
The
.Fa nitems
argument specifies the requested upper limit number of items.
The effective limit is returned to the caller, as it may end up being higher
than requested due to the implementation rounding up to ensure all memory pages
allocated to the zone are utilised to capacity.
The limit applies to the total number of items in the zone, which includes
allocated items, free items and free items in the per-cpu caches.
On systems with more than one CPU it may not be possible to allocate
the specified number of items even when there is no shortage of memory,
because all of the remaining free items may be in the caches of the
other CPUs when the limit is hit.
.Pp
The
.Fn uma_zone_set_maxcache
function limits the number of free items which may be cached in the zone.
This limit applies to both the per-CPU caches and the cache of free buckets.
.Pp
The
.Fn uma_zone_get_max
function returns the effective upper limit number of items for a zone.
.Pp
The
.Fn uma_zone_get_cur
function returns an approximation of the number of items currently allocated
from the zone.
The returned value is approximate because appropriate synchronisation to
determine an exact value is not performed by the implementation.
This ensures low overhead at the expense of potentially stale data being used
in the calculation.
.Pp
The
.Fn uma_zone_set_warning
function sets a warning that will be printed on the system console when the
given zone becomes full and fails to allocate an item.
The warning will be printed no more often than every five minutes.
Warnings can be turned off globally by setting the
.Va vm.zone_warnings
sysctl tunable to
.Va 0 .
.Pp
The
.Fn uma_zone_set_maxaction
function sets a function that will be called when the given zone becomes full
and fails to allocate an item.
The function will be called with the zone locked.
Also, the function
that called the allocation function may have held additional locks.
Therefore,
this function should do very little work (similar to a signal handler).
.Pp
The
.Fn uma_zone_set_smr
function associates an existing
.Xr smr 9
structure with a UMA zone.
The effect is similar to creating a zone with the
.Dv UMA_ZONE_SMR
flag, except that a new SMR structure is not created.
This function must be called before any allocations from the zone are performed.
.Pp
The
.Fn SYSCTL_UMA_MAX parent nbr name access zone descr
macro declares a static
.Xr sysctl 9
oid that exports the effective upper limit number of items for a zone.
The
.Fa zone
argument should be a pointer to
.Vt uma_zone_t .
A read of the oid returns value obtained through
.Fn uma_zone_get_max .
A write to the oid sets new value via
.Fn uma_zone_set_max .
The
.Fn SYSCTL_ADD_UMA_MAX ctx parent nbr name access zone descr
macro is provided to create this type of oid dynamically.
.Pp
The
.Fn SYSCTL_UMA_CUR parent nbr name access zone descr
macro declares a static read-only
.Xr sysctl 9
oid that exports the approximate current occupancy of the zone.
The
.Fa zone
argument should be a pointer to
.Vt uma_zone_t .
A read of the oid returns value obtained through
.Fn uma_zone_get_cur .
The
.Fn SYSCTL_ADD_UMA_CUR ctx parent nbr name zone descr
macro is provided to create this type of oid dynamically.
.Sh IMPLEMENTATION NOTES
The memory that these allocation calls return is not executable.
The
.Fn uma_zalloc
function does not support the
.Dv M_EXEC
flag to allocate executable memory.
Not all platforms enforce a distinction between executable and
non-executable memory.
.Sh SEE ALSO
.Xr numa 4 ,
.Xr vmstat 8 ,
.Xr malloc 9 ,
.Xr smr 9
.Rs
.%A Jeff Bonwick
.%T "The Slab Allocator: An Object-Caching Kernel Memory Allocator"
.%D 1994
.Re
.Sh HISTORY
The zone allocator first appeared in
.Fx 3.0 .
It was radically changed in
.Fx 5.0
to function as a slab allocator.
.Sh AUTHORS
.An -nosplit
The zone allocator was written by
.An John S. Dyson .
The zone allocator was rewritten in large parts by
.An Jeff Roberson Aq Mt jeff@FreeBSD.org
to function as a slab allocator.
.Pp
This manual page was written by
.An Dag-Erling Sm\(/orgrav Aq Mt des@FreeBSD.org .
Changes for UMA by
.An Jeroen Ruigrok van der Werven Aq Mt asmodai@FreeBSD.org .