Scroll to navigation

COUNTER(9) Kernel Developer's Manual COUNTER(9)

NAME

counterSMP-friendly kernel counter implementation

SYNOPSIS

#include <sys/types.h>
#include <sys/systm.h>
#include <sys/counter.h>

counter_u64_t
counter_u64_alloc(int wait);

void
counter_u64_free(counter_u64_t c);

void
counter_u64_add(counter_u64_t c, int64_t v);

void
counter_enter();

void
counter_exit();

void
counter_u64_add_protected(counter_u64_t c, int64_t v);

uint64_t
counter_u64_fetch(counter_u64_t c);

void
counter_u64_zero(counter_u64_t c);

int64_t
counter_ratecheck(struct counter_rate *cr, int64_t limit);

COUNTER_U64_SYSINIT(counter_u64_t c);

COUNTER_U64_DEFINE_EARLY(counter_u64_t c);

#include <sys/sysctl.h>

SYSCTL_COUNTER_U64(parent, nbr, name, access, ptr, descr);

SYSCTL_ADD_COUNTER_U64(ctx, parent, nbr, name, access, ptr, descr);

SYSCTL_COUNTER_U64_ARRAY(parent, nbr, name, access, ptr, len, descr);

SYSCTL_ADD_COUNTER_U64_ARRAY(ctx, parent, nbr, name, access, ptr, len, descr);

DESCRIPTION

counter is a generic facility to create counters that can be utilized for any purpose (such as collecting statistical data). A counter is guaranteed to be lossless when several kernel threads do simultaneous updates. However, counter does not block the calling thread, also no atomic(9) operations are used for the update, therefore the counters can be used in any non-interrupt context. Moreover, counter has special optimisations for SMP environments, making counter update faster than simple arithmetic on the global variable. Thus counter is considered suitable for accounting in the performance-critical code paths.

(wait)
Allocate a new 64-bit unsigned counter. The wait argument is the malloc(9) wait flag, should be either M_NOWAIT or M_WAITOK. If M_NOWAIT is specified the operation may fail.
(c)
Free the previously allocated counter c.
(c, v)
Add v to c. The KPI does not guarantee any protection from wraparound.
()
Enter mode that would allow the safe update of several counters via (). On some machines this expands to critical(9) section, while on other is a nop. See IMPLEMENTATION DETAILS.
()
Exit mode for updating several counters.
counter_u64_add_protected(c, v)
Same as counter_u64_add(), but should be preceded by counter_enter().
(c)
Take a snapshot of counter c. The data obtained is not guaranteed to reflect the real cumulative value for any moment.
(c)
Clear the counter c and set it to zero.
(cr, limit)
The function is a multiprocessor-friendly version of () which uses counter internally. Returns non-negative value if the rate is not yet reached during the current second, and a negative value otherwise. If the limit was reached on previous second, but was just reset back to zero, then counter_ratecheck() returns number of events since previous reset.
(c)
Define a SYSINIT(9) initializer for the global counter c.
(c)
Define and initialize a global counter c. It is always safe to increment c, though updates prior to the SI_SUB_COUNTER SYSINIT(9) event are lost.
(parent, nbr, name, access, ptr, descr)
Declare a static sysctl(9) oid that would represent a counter. The ptr argument should be a pointer to allocated counter_u64_t. A read of the oid returns value obtained through counter_u64_fetch(). Any write to the oid zeroes it.
(ctx, parent, nbr, name, access, ptr, descr)
Create a sysctl(9) oid that would represent a counter. The ptr argument should be a pointer to allocated counter_u64_t. A read of the oid returns value obtained through counter_u64_fetch(). Any write to the oid zeroes it.
(parent, nbr, name, access, ptr, len, descr)
Declare a static sysctl(9) oid that would represent an array of counter. The ptr argument should be a pointer to allocated array of counter_u64_t's. The len argument should specify number of elements in the array. A read of the oid returns len-sized array of uint64_t values obtained through counter_u64_fetch(). Any write to the oid zeroes all array elements.
(ctx, parent, nbr, name, access, ptr, len, descr)
Create a sysctl(9) oid that would represent an array of counter. The ptr argument should be a pointer to allocated array of counter_u64_t's. The len argument should specify number of elements in the array. A read of the oid returns len-sized array of uint64_t values obtained through counter_u64_fetch(). Any write to the oid zeroes all array elements.

IMPLEMENTATION DETAILS

On all architectures counter is implemented using per-CPU data fields that are specially aligned in memory, to avoid inter-CPU bus traffic due to shared use of the variables between CPUs. These are allocated using UMA_ZONE_PCPU uma(9) zone. The update operation only touches the field that is private to current CPU. Fetch operation loops through all per-CPU fields and obtains a snapshot sum of all fields.

On amd64 a counter update is implemented as a single instruction without lock semantics, operating on the private data for the current CPU, which is safe against preemption and interrupts.

On i386 architecture, when machine supports the cmpxchg8 instruction, this instruction is used. The multi-instruction sequence provides the same guarantees as the amd64 single-instruction implementation.

On some architectures updating a counter require a critical(9) section.

EXAMPLES

The following example creates a static counter array exported to userspace through a sysctl:

#define MY_SIZE 8
static counter_u64_t array[MY_SIZE];
SYSCTL_COUNTER_U64_ARRAY(_debug, OID_AUTO, counter_array, CTLFLAG_RW,
    &array[0], MY_SIZE, "Test counter array");

SEE ALSO

atomic(9), critical(9), locking(9), malloc(9), ratecheck(9), sysctl(9), SYSINIT(9), uma(9)

HISTORY

The counter facility first appeared in FreeBSD 10.0.

AUTHORS

The counter facility was written by Gleb Smirnoff and Konstantin Belousov.

March 6, 2020 Debian