From: Stephen Hemminger <stephen@networkplumber.org>
To: "Mattias Rönnblom" <mattias.ronnblom@ericsson.com>
Cc: <dev@dpdk.org>
Subject: Re: [dpdk-dev] [RFC] eal: make rte_rand() MT safe
Date: Fri, 5 Apr 2019 09:57:38 -0700 [thread overview]
Message-ID: <20190405095738.145a622d@shemminger-XPS-13-9360> (raw)
In-Reply-To: <20190405134542.28618-1-mattias.ronnblom@ericsson.com>
On Fri, 5 Apr 2019 15:45:42 +0200
Mattias Rönnblom <mattias.ronnblom@ericsson.com> wrote:
> The rte_rand() documentation left it unspecified if the rte_rand() was
> multi-thread safe or not, and the implementation (based on lrand48())
> was not.
>
> This commit makes rte_rand() safe to use from any lcore thread by
> using lrand48_r() and per-lcore random state structs. Besides the
> obvious improvement in terms of correctness (for concurrent users),
> this also much improves rte_rand() performance, since the threads no
> longer shares state. For the single-threaded case, this patch causes
> ~10% rte_rand() performance degradation.
>
> rte_srand() is left multi-thread unsafe, and external synchronization
> is required to serialize rte_sand() calls from different lcore
> threads, and a rte_srand() call with rte_rand() calls made by other
> lcore threads.
>
> The assumption is that the random number generators will be seeded
> once, during startup.
>
> Bugzilla ID: 114
>
> Signed-off-by: Mattias Rönnblom <mattias.ronnblom@ericsson.com>
> ---
> lib/librte_eal/common/include/rte_random.h | 25 ++++-----
> lib/librte_eal/common/meson.build | 1 +
> lib/librte_eal/common/rte_random.c | 65 ++++++++++++++++++++++
> lib/librte_eal/freebsd/eal/Makefile | 1 +
> lib/librte_eal/freebsd/eal/eal.c | 2 -
> lib/librte_eal/linux/eal/Makefile | 1 +
> lib/librte_eal/linux/eal/eal.c | 2 -
> lib/librte_eal/rte_eal_version.map | 2 +
> 8 files changed, 80 insertions(+), 19 deletions(-)
> create mode 100644 lib/librte_eal/common/rte_random.c
>
> diff --git a/lib/librte_eal/common/include/rte_random.h b/lib/librte_eal/common/include/rte_random.h
> index b2ca1c209..bca85a672 100644
> --- a/lib/librte_eal/common/include/rte_random.h
> +++ b/lib/librte_eal/common/include/rte_random.h
> @@ -16,7 +16,6 @@ extern "C" {
> #endif
>
> #include <stdint.h>
> -#include <stdlib.h>
>
> /**
> * Seed the pseudo-random generator.
> @@ -25,14 +24,15 @@ extern "C" {
> * value. It may need to be re-seeded by the user with a real random
> * value.
> *
> + * This function is not multi-thread safe in regards to other
> + * rte_srand() calls, nor is it in relation to concurrent rte_rand()
> + * calls.
> + *
> * @param seedval
> * The value of the seed.
> */
> -static inline void
> -rte_srand(uint64_t seedval)
> -{
> - srand48((long)seedval);
> -}
> +void
> +rte_srand(uint64_t seedval);
>
> /**
> * Get a pseudo-random value.
> @@ -41,18 +41,13 @@ rte_srand(uint64_t seedval)
> * congruential algorithm and 48-bit integer arithmetic, called twice
> * to generate a 64-bit value.
> *
> + * If called from lcore threads, this function is thread-safe.
> + *
> * @return
> * A pseudo-random value between 0 and (1<<64)-1.
> */
> -static inline uint64_t
> -rte_rand(void)
> -{
> - uint64_t val;
> - val = (uint64_t)lrand48();
> - val <<= 32;
> - val += (uint64_t)lrand48();
> - return val;
> -}
> +uint64_t
> +rte_rand(void);
>
> #ifdef __cplusplus
> }
> diff --git a/lib/librte_eal/common/meson.build b/lib/librte_eal/common/meson.build
> index 0670e4102..bafd23207 100644
> --- a/lib/librte_eal/common/meson.build
> +++ b/lib/librte_eal/common/meson.build
> @@ -35,6 +35,7 @@ common_sources = files(
> 'rte_keepalive.c',
> 'rte_malloc.c',
> 'rte_option.c',
> + 'rte_random.c',
> 'rte_reciprocal.c',
> 'rte_service.c'
> )
> diff --git a/lib/librte_eal/common/rte_random.c b/lib/librte_eal/common/rte_random.c
> new file mode 100644
> index 000000000..9d519d03b
> --- /dev/null
> +++ b/lib/librte_eal/common/rte_random.c
> @@ -0,0 +1,65 @@
> +/* SPDX-License-Identifier: BSD-3-Clause
> + * Copyright(c) 2010-2014 Intel Corporation
> + * Copyright(c) 2019 Ericsson AB
> + */
> +
> +#include <stdlib.h>
> +
> +#include <rte_cycles.h>
> +#include <rte_eal.h>
> +#include <rte_lcore.h>
> +#include <rte_random.h>
> +
> +struct rte_rand_data
> +{
> + struct drand48_data data;
> +} __rte_cache_aligned;
> +
> +static struct rte_rand_data rand_data[RTE_MAX_LCORE];
> +
> +void
> +rte_srand(uint64_t seedval)
> +{
> + unsigned i;
> +
> + /* give the different lcores a different seed, to avoid a
> + situation where they generate the same sequence */
> + for (i = 0; i < RTE_MAX_LCORE; i++)
> + srand48_r((long)seedval + i, &rand_data[i].data);
> +}
> +
> +static inline uint32_t
> +__rte_rand48(struct drand48_data *data)
> +{
> + long res;
> +
> + lrand48_r(data, &res);
> +
> + return (uint32_t)res;
> +}
> +
> +uint64_t
> +rte_rand(void)
> +{
> + unsigned lcore_id;
> + struct drand48_data *data;
> + uint64_t val;
> +
> + lcore_id = rte_lcore_id();
> +
> + if (unlikely(lcore_id == LCORE_ID_ANY))
> + lcore_id = rte_get_master_lcore();
> +
> + data = &rand_data[lcore_id].data;
> +
> + val = __rte_rand48(data);
> + val <<= 32;
> + val += __rte_rand48(data);
> +
> + return val;
> +}
> +
> +RTE_INIT(rte_rand_init)
> +{
> + rte_srand(rte_get_timer_cycles());
> +}
rand48 is a terrible PRNG, why not use something better?
Similar discussion in Linux kernel pointed at:
http://www.pcg-random.org/posts/some-prng-implementations.html
Mail thread here:
https://www.spinics.net/lists/netdev/msg560231.html
next prev parent reply other threads:[~2019-04-05 16:57 UTC|newest]
Thread overview: 137+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-04-05 13:45 Mattias Rönnblom
2019-04-05 13:45 ` Mattias Rönnblom
2019-04-05 13:51 ` Mattias Rönnblom
2019-04-05 13:51 ` Mattias Rönnblom
2019-04-05 14:28 ` Bruce Richardson
2019-04-05 14:28 ` Bruce Richardson
2019-04-05 14:56 ` Mattias Rönnblom
2019-04-05 14:56 ` Mattias Rönnblom
2019-04-05 16:57 ` Stephen Hemminger [this message]
2019-04-05 16:57 ` Stephen Hemminger
2019-04-05 18:04 ` Mattias Rönnblom
2019-04-05 18:04 ` Mattias Rönnblom
2019-04-05 20:50 ` Stephen Hemminger
2019-04-05 20:50 ` Stephen Hemminger
2019-04-06 5:52 ` Mattias Rönnblom
2019-04-06 5:52 ` Mattias Rönnblom
2019-04-08 12:30 ` [dpdk-dev] [RFC 1/3] Replace lrand48-based rte_rand with LFSR generator Mattias Rönnblom
2019-04-08 12:30 ` Mattias Rönnblom
2019-04-08 12:30 ` [dpdk-dev] [RFC 2/3] Add 32-bit version of rte_rand Mattias Rönnblom
2019-04-08 12:30 ` Mattias Rönnblom
2019-04-08 12:30 ` [dpdk-dev] [RFC 3/3] Introduce random generator functions with upper bound Mattias Rönnblom
2019-04-08 12:30 ` Mattias Rönnblom
2019-04-08 12:47 ` [dpdk-dev] [RFC 1/3] Replace lrand48-based rte_rand with LFSR generator Mattias Rönnblom
2019-04-08 12:47 ` Mattias Rönnblom
2019-04-19 21:21 ` [dpdk-dev] [RFC v2 0/2] Pseudo-number generation improvements Mattias Rönnblom
2019-04-19 21:21 ` Mattias Rönnblom
2019-04-19 21:21 ` [dpdk-dev] [RFC v2 1/2] eal: replace libc-based random number generation with LFSR Mattias Rönnblom
2019-04-19 21:21 ` Mattias Rönnblom
2019-04-22 11:34 ` Neil Horman
2019-04-22 11:34 ` Neil Horman
2019-04-22 14:49 ` Stephen Hemminger
2019-04-22 14:49 ` Stephen Hemminger
2019-04-22 15:52 ` Mattias Rönnblom
2019-04-22 15:52 ` Mattias Rönnblom
2019-04-22 17:44 ` Mattias Rönnblom
2019-04-22 17:44 ` Mattias Rönnblom
2019-04-23 11:33 ` Neil Horman
2019-04-23 11:33 ` Neil Horman
2019-04-23 17:13 ` Mattias Rönnblom
2019-04-23 17:13 ` Mattias Rönnblom
2019-04-24 11:37 ` Neil Horman
2019-04-24 11:37 ` Neil Horman
2019-04-23 15:31 ` Stephen Hemminger
2019-04-23 15:31 ` Stephen Hemminger
2019-04-23 17:17 ` Mattias Rönnblom
2019-04-23 17:17 ` Mattias Rönnblom
2019-04-24 7:52 ` Mattias Rönnblom
2019-04-24 7:52 ` Mattias Rönnblom
2019-04-24 12:33 ` [dpdk-dev] [RFC v3 0/2] Pseudo-random number generation improvements Mattias Rönnblom
2019-04-24 12:33 ` Mattias Rönnblom
2019-04-24 12:33 ` [dpdk-dev] [RFC v3 1/2] eal: replace libc-based random number generation with LFSR Mattias Rönnblom
2019-04-24 12:33 ` Mattias Rönnblom
2019-05-08 20:12 ` Stephen Hemminger
2019-05-08 20:12 ` Stephen Hemminger
2019-05-08 20:30 ` Mattias Rönnblom
2019-05-08 20:30 ` Mattias Rönnblom
2019-05-09 1:10 ` Stephen Hemminger
2019-05-09 1:10 ` Stephen Hemminger
2019-05-14 9:20 ` [dpdk-dev] [PATCH 0/6] Pseudo-random number generation improvements Mattias Rönnblom
2019-05-14 9:20 ` Mattias Rönnblom
2019-05-14 9:20 ` [dpdk-dev] [PATCH 1/6] eal: replace libc-based random number generation with LFSR Mattias Rönnblom
2019-05-14 9:20 ` Mattias Rönnblom
2019-05-14 9:32 ` Mattias Rönnblom
2019-05-14 9:32 ` Mattias Rönnblom
2019-05-14 14:16 ` Neil Horman
2019-05-14 14:16 ` Neil Horman
2019-05-14 14:53 ` Mattias Rönnblom
2019-05-14 14:53 ` Mattias Rönnblom
2019-05-17 19:27 ` Neil Horman
2019-05-17 20:57 ` Bruce Richardson
2019-05-17 21:10 ` Mattias Rönnblom
2019-05-19 18:32 ` Neil Horman
2019-05-14 15:27 ` Stephen Hemminger
2019-05-14 15:27 ` Stephen Hemminger
2019-05-14 9:20 ` [dpdk-dev] [PATCH 2/6] eal: add pseudo-random number generation performance test Mattias Rönnblom
2019-05-14 9:20 ` Mattias Rönnblom
2019-05-14 9:20 ` [dpdk-dev] [PATCH 3/6] eal: improve entropy for initial PRNG seed Mattias Rönnblom
2019-05-14 9:20 ` Mattias Rönnblom
2019-05-14 9:36 ` Mattias Rönnblom
2019-05-14 9:36 ` Mattias Rönnblom
2019-05-14 9:39 ` Bruce Richardson
2019-05-14 9:39 ` Bruce Richardson
2019-05-14 11:58 ` Mattias Rönnblom
2019-05-14 11:58 ` Mattias Rönnblom
2019-05-14 9:37 ` Bruce Richardson
2019-05-14 9:37 ` Bruce Richardson
2019-05-14 9:20 ` [dpdk-dev] [PATCH 4/6] eal: introduce random generator function with upper bound Mattias Rönnblom
2019-05-14 9:20 ` Mattias Rönnblom
2019-05-14 9:20 ` [dpdk-dev] [PATCH 5/6] eal: add bounded PRNG performance tests Mattias Rönnblom
2019-05-14 9:20 ` Mattias Rönnblom
2019-05-14 9:20 ` [dpdk-dev] [PATCH 6/6] eal: add pseudo-random number generation to MAINTAINERS Mattias Rönnblom
2019-05-14 9:20 ` Mattias Rönnblom
2019-05-16 17:55 ` [dpdk-dev] [PATCH v2 0/6] Pseudo-random number generation improvements Mattias Rönnblom
2019-05-16 17:55 ` [dpdk-dev] [PATCH v2 1/6] eal: replace libc-based random number generation with LFSR Mattias Rönnblom
2019-05-16 17:55 ` [dpdk-dev] [PATCH v2 2/6] eal: add pseudo-random number generation performance test Mattias Rönnblom
2019-05-16 17:55 ` [dpdk-dev] [PATCH v2 3/6] eal: improve entropy for initial PRNG seed Mattias Rönnblom
2019-05-16 17:55 ` [dpdk-dev] [PATCH v2 4/6] eal: introduce random generator function with upper bound Mattias Rönnblom
2019-05-16 17:55 ` [dpdk-dev] [PATCH v2 5/6] eal: add bounded PRNG performance tests Mattias Rönnblom
2019-05-16 17:55 ` [dpdk-dev] [PATCH v2 6/6] eal: add PRNG to MAINTAINERS and release notes Mattias Rönnblom
2019-05-16 20:35 ` [dpdk-dev] [PATCH v2 0/6] Pseudo-random number generation improvements Bruce Richardson
2019-06-05 10:43 ` [dpdk-dev] [PATCH v3 " Mattias Rönnblom
2019-06-05 10:43 ` [dpdk-dev] [PATCH v3 1/6] eal: replace libc-based random number generation with LFSR Mattias Rönnblom
2019-06-05 10:43 ` [dpdk-dev] [PATCH v3 2/6] eal: add pseudo-random number generation performance test Mattias Rönnblom
2019-06-27 21:23 ` Thomas Monjalon
2019-06-28 8:20 ` Mattias Rönnblom
2019-06-05 10:43 ` [dpdk-dev] [PATCH v3 3/6] eal: improve entropy for initial PRNG seed Mattias Rönnblom
2019-06-05 10:43 ` [dpdk-dev] [PATCH v3 4/6] eal: introduce random generator function with upper bound Mattias Rönnblom
2019-06-05 10:43 ` [dpdk-dev] [PATCH v3 5/6] eal: add bounded PRNG performance tests Mattias Rönnblom
2019-06-05 10:44 ` [dpdk-dev] [PATCH v3 6/6] eal: add PRNG to MAINTAINERS and release notes Mattias Rönnblom
2019-06-27 21:27 ` Thomas Monjalon
2019-06-28 8:17 ` Mattias Rönnblom
2019-06-15 12:23 ` [dpdk-dev] [PATCH v3 0/6] Pseudo-random number generation improvements Mattias Rönnblom
2019-06-28 9:01 ` [dpdk-dev] [PATCH v4 0/5] " Mattias Rönnblom
2019-06-28 9:01 ` [dpdk-dev] [PATCH v4 1/5] eal: replace libc-based random number generation with LFSR Mattias Rönnblom
2019-06-28 9:01 ` [dpdk-dev] [PATCH v4 2/5] eal: add pseudo-random number generation performance test Mattias Rönnblom
2019-06-28 9:01 ` [dpdk-dev] [PATCH v4 3/5] eal: improve entropy for initial PRNG seed Mattias Rönnblom
2019-06-28 19:01 ` Ferruh Yigit
2019-06-28 20:58 ` Mattias Rönnblom
2019-06-28 21:08 ` [dpdk-dev] [PATCH] eal: use 32-bit RDSEED to allow 32-bit x86 usage Mattias Rönnblom
2019-06-29 12:54 ` Thomas Monjalon
2019-06-28 9:01 ` [dpdk-dev] [PATCH v4 4/5] eal: introduce random generator function with upper bound Mattias Rönnblom
2019-06-28 9:01 ` [dpdk-dev] [PATCH v4 5/5] eal: add bounded PRNG performance tests Mattias Rönnblom
2019-06-28 13:24 ` [dpdk-dev] [PATCH v4 0/5] Pseudo-random number generation improvements Thomas Monjalon
2019-04-24 12:33 ` [dpdk-dev] [RFC v3 2/2] eal: introduce random generator function with upper bound Mattias Rönnblom
2019-04-24 12:33 ` Mattias Rönnblom
2019-04-19 21:21 ` [dpdk-dev] [RFC v2 " Mattias Rönnblom
2019-04-19 21:21 ` Mattias Rönnblom
2019-04-20 21:08 ` Wiles, Keith
2019-04-20 21:08 ` Wiles, Keith
2019-04-21 19:05 ` Mattias Rönnblom
2019-04-21 19:05 ` Mattias Rönnblom
2019-04-22 4:33 ` Wiles, Keith
2019-04-22 4:33 ` Wiles, Keith
2019-04-22 7:07 ` Mattias Rönnblom
2019-04-22 7:07 ` Mattias Rönnblom
2019-04-22 13:19 ` Wiles, Keith
2019-04-22 13:19 ` Wiles, Keith
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20190405095738.145a622d@shemminger-XPS-13-9360 \
--to=stephen@networkplumber.org \
--cc=dev@dpdk.org \
--cc=mattias.ronnblom@ericsson.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).