Revision 6e474083f3daf3a3546737f5d7d502ad12eb257c authored by Wei Xu on 01 December 2017, 10:10:36 UTC, committed by David S. Miller on 03 December 2017, 02:31:03 UTC
Matthew found a roughly 40% tcp throughput regression with commit
c67df11f(vhost_net: try batch dequing from skb array) as discussed
in the following thread:
https://www.mail-archive.com/netdev@vger.kernel.org/msg187936.html

Eventually we figured out that it was a skb leak in handle_rx()
when sending packets to the VM. This usually happens when a guest
can not drain out vq as fast as vhost fills in, afterwards it sets
off the traffic jam and leaks skb(s) which occurs as no headcount
to send on the vq from vhost side.

This can be avoided by making sure we have got enough headcount
before actually consuming a skb from the batched rx array while
transmitting, which is simply done by moving checking the zero
headcount a bit ahead.

Signed-off-by: Wei Xu <wexu@redhat.com>
Reported-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
1 parent fa935ca
Raw File
rng.c
/*
 * Cryptographic API.
 *
 * RNG operations.
 *
 * Copyright (c) 2008 Neil Horman <nhorman@tuxdriver.com>
 * Copyright (c) 2015 Herbert Xu <herbert@gondor.apana.org.au>
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License as published by the Free
 * Software Foundation; either version 2 of the License, or (at your option)
 * any later version.
 *
 */

#include <linux/atomic.h>
#include <crypto/internal/rng.h>
#include <linux/err.h>
#include <linux/module.h>
#include <linux/mutex.h>
#include <linux/random.h>
#include <linux/seq_file.h>
#include <linux/slab.h>
#include <linux/string.h>
#include <linux/cryptouser.h>
#include <linux/compiler.h>
#include <net/netlink.h>

#include "internal.h"

static DEFINE_MUTEX(crypto_default_rng_lock);
struct crypto_rng *crypto_default_rng;
EXPORT_SYMBOL_GPL(crypto_default_rng);
static int crypto_default_rng_refcnt;

int crypto_rng_reset(struct crypto_rng *tfm, const u8 *seed, unsigned int slen)
{
	u8 *buf = NULL;
	int err;

	if (!seed && slen) {
		buf = kmalloc(slen, GFP_KERNEL);
		if (!buf)
			return -ENOMEM;

		err = get_random_bytes_wait(buf, slen);
		if (err)
			goto out;
		seed = buf;
	}

	err = crypto_rng_alg(tfm)->seed(tfm, seed, slen);
out:
	kzfree(buf);
	return err;
}
EXPORT_SYMBOL_GPL(crypto_rng_reset);

static int crypto_rng_init_tfm(struct crypto_tfm *tfm)
{
	return 0;
}

static unsigned int seedsize(struct crypto_alg *alg)
{
	struct rng_alg *ralg = container_of(alg, struct rng_alg, base);

	return ralg->seedsize;
}

#ifdef CONFIG_NET
static int crypto_rng_report(struct sk_buff *skb, struct crypto_alg *alg)
{
	struct crypto_report_rng rrng;

	strncpy(rrng.type, "rng", sizeof(rrng.type));

	rrng.seedsize = seedsize(alg);

	if (nla_put(skb, CRYPTOCFGA_REPORT_RNG,
		    sizeof(struct crypto_report_rng), &rrng))
		goto nla_put_failure;
	return 0;

nla_put_failure:
	return -EMSGSIZE;
}
#else
static int crypto_rng_report(struct sk_buff *skb, struct crypto_alg *alg)
{
	return -ENOSYS;
}
#endif

static void crypto_rng_show(struct seq_file *m, struct crypto_alg *alg)
	__maybe_unused;
static void crypto_rng_show(struct seq_file *m, struct crypto_alg *alg)
{
	seq_printf(m, "type         : rng\n");
	seq_printf(m, "seedsize     : %u\n", seedsize(alg));
}

static const struct crypto_type crypto_rng_type = {
	.extsize = crypto_alg_extsize,
	.init_tfm = crypto_rng_init_tfm,
#ifdef CONFIG_PROC_FS
	.show = crypto_rng_show,
#endif
	.report = crypto_rng_report,
	.maskclear = ~CRYPTO_ALG_TYPE_MASK,
	.maskset = CRYPTO_ALG_TYPE_MASK,
	.type = CRYPTO_ALG_TYPE_RNG,
	.tfmsize = offsetof(struct crypto_rng, base),
};

struct crypto_rng *crypto_alloc_rng(const char *alg_name, u32 type, u32 mask)
{
	return crypto_alloc_tfm(alg_name, &crypto_rng_type, type, mask);
}
EXPORT_SYMBOL_GPL(crypto_alloc_rng);

int crypto_get_default_rng(void)
{
	struct crypto_rng *rng;
	int err;

	mutex_lock(&crypto_default_rng_lock);
	if (!crypto_default_rng) {
		rng = crypto_alloc_rng("stdrng", 0, 0);
		err = PTR_ERR(rng);
		if (IS_ERR(rng))
			goto unlock;

		err = crypto_rng_reset(rng, NULL, crypto_rng_seedsize(rng));
		if (err) {
			crypto_free_rng(rng);
			goto unlock;
		}

		crypto_default_rng = rng;
	}

	crypto_default_rng_refcnt++;
	err = 0;

unlock:
	mutex_unlock(&crypto_default_rng_lock);

	return err;
}
EXPORT_SYMBOL_GPL(crypto_get_default_rng);

void crypto_put_default_rng(void)
{
	mutex_lock(&crypto_default_rng_lock);
	crypto_default_rng_refcnt--;
	mutex_unlock(&crypto_default_rng_lock);
}
EXPORT_SYMBOL_GPL(crypto_put_default_rng);

#if defined(CONFIG_CRYPTO_RNG) || defined(CONFIG_CRYPTO_RNG_MODULE)
int crypto_del_default_rng(void)
{
	int err = -EBUSY;

	mutex_lock(&crypto_default_rng_lock);
	if (crypto_default_rng_refcnt)
		goto out;

	crypto_free_rng(crypto_default_rng);
	crypto_default_rng = NULL;

	err = 0;

out:
	mutex_unlock(&crypto_default_rng_lock);

	return err;
}
EXPORT_SYMBOL_GPL(crypto_del_default_rng);
#endif

int crypto_register_rng(struct rng_alg *alg)
{
	struct crypto_alg *base = &alg->base;

	if (alg->seedsize > PAGE_SIZE / 8)
		return -EINVAL;

	base->cra_type = &crypto_rng_type;
	base->cra_flags &= ~CRYPTO_ALG_TYPE_MASK;
	base->cra_flags |= CRYPTO_ALG_TYPE_RNG;

	return crypto_register_alg(base);
}
EXPORT_SYMBOL_GPL(crypto_register_rng);

void crypto_unregister_rng(struct rng_alg *alg)
{
	crypto_unregister_alg(&alg->base);
}
EXPORT_SYMBOL_GPL(crypto_unregister_rng);

int crypto_register_rngs(struct rng_alg *algs, int count)
{
	int i, ret;

	for (i = 0; i < count; i++) {
		ret = crypto_register_rng(algs + i);
		if (ret)
			goto err;
	}

	return 0;

err:
	for (--i; i >= 0; --i)
		crypto_unregister_rng(algs + i);

	return ret;
}
EXPORT_SYMBOL_GPL(crypto_register_rngs);

void crypto_unregister_rngs(struct rng_alg *algs, int count)
{
	int i;

	for (i = count - 1; i >= 0; --i)
		crypto_unregister_rng(algs + i);
}
EXPORT_SYMBOL_GPL(crypto_unregister_rngs);

MODULE_LICENSE("GPL");
MODULE_DESCRIPTION("Random Number Generator");
back to top