linux/arch/s390/crypto/arch_random.c
Harald Freudenberger 966f53e750 s390/archrandom: Rework arch random implementation.
The arch_get_random_seed_long() invocation done by the random device
driver is done in interrupt context and may be invoked very very
frequently. The existing s390 arch_get_random_seed*() implementation
uses the PRNO(TRNG) instruction which produces excellent high quality
entropy but is relatively slow and thus expensive.

This fix reworks the arch_get_random_seed* implementation. It
introduces a buffer concept to decouple the delivery of random data
via arch_get_random_seed*() from the generation of new random
bytes. The buffer of random data is filled asynchronously by a
workqueue thread.
If there are enough bytes in the buffer the s390_arch_random_generate()
just delivers these bytes. Otherwise false is returned until the worker
thread refills the buffer.
The worker fills the rng buffer by pulling fresh entropy from the
high quality (but slow) true hardware random generator. This entropy
is then spread over the buffer with an pseudo random generator.
As the arch_get_random_seed_long() fetches 8 bytes and the calling
function add_interrupt_randomness() counts this as 1 bit entropy the
distribution needs to make sure there is in fact 1 bit entropy
contained in 8 bytes of the buffer. The current values pull 32 byte
entropy and scatter this into a 2048 byte buffer. So 8 byte in the
buffer will contain 1 bit of entropy.
The worker thread is rescheduled based on the charge level of the
buffer but at least with 500 ms delay to avoid too much cpu consumption.
So the max. amount of rng data delivered via arch_get_random_seed is
limited to 4Kb per second.

Signed-off-by: Harald Freudenberger <freude@de.ibm.com>
Reviewed-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
2018-05-30 11:18:04 +02:00

124 lines
4.1 KiB
C

// SPDX-License-Identifier: GPL-2.0
/*
* s390 arch random implementation.
*
* Copyright IBM Corp. 2017, 2018
* Author(s): Harald Freudenberger
*
* The s390_arch_random_generate() function may be called from random.c
* in interrupt context. So this implementation does the best to be very
* fast. There is a buffer of random data which is asynchronously checked
* and filled by a workqueue thread.
* If there are enough bytes in the buffer the s390_arch_random_generate()
* just delivers these bytes. Otherwise false is returned until the
* worker thread refills the buffer.
* The worker fills the rng buffer by pulling fresh entropy from the
* high quality (but slow) true hardware random generator. This entropy
* is then spread over the buffer with an pseudo random generator PRNG.
* As the arch_get_random_seed_long() fetches 8 bytes and the calling
* function add_interrupt_randomness() counts this as 1 bit entropy the
* distribution needs to make sure there is in fact 1 bit entropy contained
* in 8 bytes of the buffer. The current values pull 32 byte entropy
* and scatter this into a 2048 byte buffer. So 8 byte in the buffer
* will contain 1 bit of entropy.
* The worker thread is rescheduled based on the charge level of the
* buffer but at least with 500 ms delay to avoid too much CPU consumption.
* So the max. amount of rng data delivered via arch_get_random_seed is
* limited to 4k bytes per second.
*/
#include <linux/kernel.h>
#include <linux/atomic.h>
#include <linux/random.h>
#include <linux/slab.h>
#include <linux/static_key.h>
#include <linux/workqueue.h>
#include <asm/cpacf.h>
DEFINE_STATIC_KEY_FALSE(s390_arch_random_available);
atomic64_t s390_arch_random_counter = ATOMIC64_INIT(0);
EXPORT_SYMBOL(s390_arch_random_counter);
#define ARCH_REFILL_TICKS (HZ/2)
#define ARCH_PRNG_SEED_SIZE 32
#define ARCH_RNG_BUF_SIZE 2048
static DEFINE_SPINLOCK(arch_rng_lock);
static u8 *arch_rng_buf;
static unsigned int arch_rng_buf_idx;
static void arch_rng_refill_buffer(struct work_struct *);
static DECLARE_DELAYED_WORK(arch_rng_work, arch_rng_refill_buffer);
bool s390_arch_random_generate(u8 *buf, unsigned int nbytes)
{
/* lock rng buffer */
if (!spin_trylock(&arch_rng_lock))
return false;
/* try to resolve the requested amount of bytes from the buffer */
arch_rng_buf_idx -= nbytes;
if (arch_rng_buf_idx < ARCH_RNG_BUF_SIZE) {
memcpy(buf, arch_rng_buf + arch_rng_buf_idx, nbytes);
atomic64_add(nbytes, &s390_arch_random_counter);
spin_unlock(&arch_rng_lock);
return true;
}
/* not enough bytes in rng buffer, refill is done asynchronously */
spin_unlock(&arch_rng_lock);
return false;
}
EXPORT_SYMBOL(s390_arch_random_generate);
static void arch_rng_refill_buffer(struct work_struct *unused)
{
unsigned int delay = ARCH_REFILL_TICKS;
spin_lock(&arch_rng_lock);
if (arch_rng_buf_idx > ARCH_RNG_BUF_SIZE) {
/* buffer is exhausted and needs refill */
u8 seed[ARCH_PRNG_SEED_SIZE];
u8 prng_wa[240];
/* fetch ARCH_PRNG_SEED_SIZE bytes of entropy */
cpacf_trng(NULL, 0, seed, sizeof(seed));
/* blow this entropy up to ARCH_RNG_BUF_SIZE with PRNG */
memset(prng_wa, 0, sizeof(prng_wa));
cpacf_prno(CPACF_PRNO_SHA512_DRNG_SEED,
&prng_wa, NULL, 0, seed, sizeof(seed));
cpacf_prno(CPACF_PRNO_SHA512_DRNG_GEN,
&prng_wa, arch_rng_buf, ARCH_RNG_BUF_SIZE, NULL, 0);
arch_rng_buf_idx = ARCH_RNG_BUF_SIZE;
}
delay += (ARCH_REFILL_TICKS * arch_rng_buf_idx) / ARCH_RNG_BUF_SIZE;
spin_unlock(&arch_rng_lock);
/* kick next check */
queue_delayed_work(system_long_wq, &arch_rng_work, delay);
}
static int __init s390_arch_random_init(void)
{
/* all the needed PRNO subfunctions available ? */
if (cpacf_query_func(CPACF_PRNO, CPACF_PRNO_TRNG) &&
cpacf_query_func(CPACF_PRNO, CPACF_PRNO_SHA512_DRNG_GEN)) {
/* alloc arch random working buffer */
arch_rng_buf = kmalloc(ARCH_RNG_BUF_SIZE, GFP_KERNEL);
if (!arch_rng_buf)
return -ENOMEM;
/* kick worker queue job to fill the random buffer */
queue_delayed_work(system_long_wq,
&arch_rng_work, ARCH_REFILL_TICKS);
/* enable arch random to the outside world */
static_branch_enable(&s390_arch_random_available);
}
return 0;
}
arch_initcall(s390_arch_random_init);