Manual browser: rndsink(9)

Section:
Page:
RNDSINK(9) Kernel Developer's Manual RNDSINK(9)

NAME

rndsink, rndsink_create, rndsink_destroy, rndsink_request, rndsink_schedule, — functions to asynchronously request entropy from the system entropy pool

SYNOPSIS

#include <sys/rndsink.h>

struct rndsink *
rndsink_create(size_t bytes, void (*callback)(void *, const void *, size_t), void *arg);

void
rndsink_destroy(struct rndsink *rndsink);

bool
rndsink_request(struct rndsink *rndsink, void *buffer, size_t bytes);

void
rndsink_schedule(struct rndsink *rndsink);

DESCRIPTION

The rndsink functions support asynchronous requests for entropy from the system entropy pool. Users must call rndsink_create() to create an rndsink which they may then pass to rndsink_request() to request data from the system entropy pool. If full entropy is not available, the system will call a callback when entropy is next available. Users can schedule a callback without requesting data now using rndsink_schedule(). When users no longer need an rndsink, they must pass it to rndsink_destroy().

This API provides direct access to the system entropy pool. Most users should use the cprng(9) API instead, which interposes a cryptographic pseudorandom number generator between the user and the entropy pool.

FUNCTIONS

rndsink_create(bytes, callback, arg)
Create an rndsink for requests of bytes bytes of entropy, which must be no more than RNDSINK_MAX_BYTES. When requested and enough entropy is available, the system will call callback with three arguments:
  • arg, an arbitrary user-supplied pointer;
  • a pointer to a buffer containing the bytes of entropy; and
  • the number of bytes in the buffer, which will always be bytes.

The callback will be called in soft interrupt context.

rndsink_create() may sleep to allocate memory.

rndsink_destroy(rndsink)
Destroy an rndsink. rndsink_destroy() may sleep to wait for pending callbacks to complete and to deallocate memory.
rndsink_request(rndsink, buffer, bytes)
Store bytes bytes derived from the system entropy pool in buffer. If the bytes have full entropy, return true. Otherwise, schedule a callback as if with rndsink_schedule() and return false. In either case, rndsink_request() will store data in buffer. The argument bytes must be the same as the argument to rndsink_create() that was used to create rndsink. May be called at IPL_VM or lower. The caller should use explicit_memset(3) to clear buffer once it has used the data stored there.
rndsink_schedule(rndsink)
Schedule a callback when the system entropy pool has enough entropy. If a callback is already scheduled, it remains scheduled. May be called at IPL_VM or lower.

CODE REFERENCES

The rndsink API is implemented in sys/kern/kern_rndsink.c and sys/sys/rndsink.h.

HISTORY

The rndsink API first appeared in NetBSD 7.0.
April 10, 2013 NetBSD 7.0