From: "Ananyev, Konstantin" <konstantin.ananyev@intel.com>
To: "Burakov, Anatoly" <anatoly.burakov@intel.com>,
"dev@dpdk.org" <dev@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH 1/9] eal: map shared config into exact same address as primary process
Date: Tue, 17 Jun 2014 16:29:03 +0000 [thread overview]
Message-ID: <2601191342CEEE43887BDE71AB9772580EFB7832@IRSMSX105.ger.corp.intel.com> (raw)
In-Reply-To: <fb3aaffcd381dcd174677d763d59cc80171faa90.1403018971.git.anatoly.burakov@intel.com>
Hi Anatoly
>
> Shared config is shared across primary and secondary processes.
> However,when using rte_malloc, the malloc elements keep references to
> the heap inside themselves. This heap reference might not be referencing
> a local heap because the heap reference points to the heap of whatever
> process has allocated that malloc element. Therefore, there can be
> situations when malloc elements in a given heap actually reference
> different addresses for the same heap - depending on which process has
> allocated the element. This can lead to segmentation faults when dealing
> with malloc elements allocated on the same heap by different processes.
>
> To fix this problem, heaps will now have the same addresses across
> processes. In order to achieve that, a new field in a shared mem_config
> (a structure that holds the heaps, and which is shared across processes)
> was added to keep the address of where this config is mapped in the
> primary process.
>
> Secondary process will now map the config in two stages - first, it'll
> map it into an arbitrary address and read the address the primary
> process has allocated for the shared config. Then, the config is
> unmapped and re-mapped using the address previously read.
>
> Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
> ---
> lib/librte_eal/common/include/rte_eal_memconfig.h | 5 ++++
> lib/librte_eal/linuxapp/eal/eal.c | 31 +++++++++++++++++++----
> 2 files changed, 31 insertions(+), 5 deletions(-)
>
> diff --git a/lib/librte_eal/common/include/rte_eal_memconfig.h b/lib/librte_eal/common/include/rte_eal_memconfig.h
> index 30ce6fc..d6359e5 100644
> --- a/lib/librte_eal/common/include/rte_eal_memconfig.h
> +++ b/lib/librte_eal/common/include/rte_eal_memconfig.h
> @@ -89,6 +89,11 @@ struct rte_mem_config {
>
> /* Heaps of Malloc per socket */
> struct malloc_heap malloc_heaps[RTE_MAX_NUMA_NODES];
> +
> + /* address of mem_config in primary process. used to map shared config into
> + * exact same address the primary process maps it.
> + */
> + uint64_t mem_cfg_addr;
> } __attribute__((__packed__));
>
>
> diff --git a/lib/librte_eal/linuxapp/eal/eal.c b/lib/librte_eal/linuxapp/eal/eal.c
> index 6994303..fedd82f 100644
> --- a/lib/librte_eal/linuxapp/eal/eal.c
> +++ b/lib/librte_eal/linuxapp/eal/eal.c
> @@ -239,6 +239,11 @@ rte_eal_config_create(void)
> }
> memcpy(rte_mem_cfg_addr, &early_mem_config, sizeof(early_mem_config));
> rte_config.mem_config = (struct rte_mem_config *) rte_mem_cfg_addr;
> +
> + /* store address of the config in the config itself so that secondary
> + * processes could later map the config into this exact location */
> + rte_config.mem_config->mem_cfg_addr = (uintptr_t) rte_mem_cfg_addr;
> +
> }
>
> /* attach to an existing shared memory config */
> @@ -246,6 +251,8 @@ static void
> rte_eal_config_attach(void)
> {
> void *rte_mem_cfg_addr;
> + struct rte_mem_config *mem_config;
> +
> const char *pathname = eal_runtime_config_path();
>
> if (internal_config.no_shconf)
> @@ -257,13 +264,27 @@ rte_eal_config_attach(void)
> rte_panic("Cannot open '%s' for rte_mem_config\n", pathname);
> }
>
> - rte_mem_cfg_addr = mmap(NULL, sizeof(*rte_config.mem_config),
> - PROT_READ | PROT_WRITE, MAP_SHARED, mem_cfg_fd, 0);
> - close(mem_cfg_fd);
> - if (rte_mem_cfg_addr == MAP_FAILED)
> + /* map it as read-only first */
> + mem_config = (struct rte_mem_config *) mmap(NULL, sizeof(*mem_config),
> + PROT_READ, MAP_SHARED, mem_cfg_fd, 0);
> + if (mem_config == MAP_FAILED)
> rte_panic("Cannot mmap memory for rte_config\n");
>
> - rte_config.mem_config = (struct rte_mem_config *) rte_mem_cfg_addr;
> + /* store address used by primary process */
> + rte_mem_cfg_addr = (void *) (uintptr_t) mem_config->mem_cfg_addr;
> +
> + /* unmap the config */
> + munmap(mem_config, sizeof(*mem_config));
> +
> + /* map the config again, with the proper virtual address */
> + mem_config = (struct rte_mem_config *) mmap(rte_mem_cfg_addr,
> + sizeof(*mem_config), PROT_READ | PROT_WRITE, MAP_SHARED,
> + mem_cfg_fd, 0);
> + if (mem_config == MAP_FAILED || mem_config != rte_mem_cfg_addr)
> + rte_panic("Cannot mmap memory for rte_config\n");
> + close(mem_cfg_fd);
> +
> + rte_config.mem_config = mem_config;
> }
>
> /* Detect if we are a primary or a secondary process */
> --
I think we introduce a race window here.
If secondary process would do first mmap() before rte_config.mem_config->mem_cfg_addr was properly set by primary process,
then it will try to do second mmap() with wrong address.
I think we need to do second mmap() straight after rte_eal_mcfg_wait_complete(), or even just inside it.
Konstantin
next prev parent reply other threads:[~2014-06-17 16:29 UTC|newest]
Thread overview: 44+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-06-13 15:29 [dpdk-dev] [PATCH 0/7] Make DPDK tailqs fully local Anatoly Burakov
2014-06-13 15:29 ` [dpdk-dev] [PATCH 1/7] rte_tailq: change rte_dummy to rte_tailq_entry, add data pointer Anatoly Burakov
2014-06-13 15:29 ` [dpdk-dev] [PATCH 2/7] rte_ring: make ring tailq completely local Anatoly Burakov
2014-06-13 15:29 ` [dpdk-dev] [PATCH 3/7] rte_hash: make rte_hash tailq fully local Anatoly Burakov
2014-06-13 15:29 ` [dpdk-dev] [PATCH 4/7] rte_fbk_hash: make rte_fbk_hash " Anatoly Burakov
2014-06-13 15:29 ` [dpdk-dev] [PATCH 5/7] rte_mempool: make mempool " Anatoly Burakov
2014-06-13 15:29 ` [dpdk-dev] [PATCH 6/7] rte_lpm: make lpm " Anatoly Burakov
2014-06-13 15:29 ` [dpdk-dev] [PATCH 7/7] rte_lpm6: make lpm6 " Anatoly Burakov
2014-06-17 9:57 ` [dpdk-dev] [PATCH 0/7] Make DPDK tailqs " Burakov, Anatoly
2014-06-17 15:35 ` [dpdk-dev] [PATCH 0/9] " Anatoly Burakov
2014-06-17 15:35 ` [dpdk-dev] [PATCH 1/9] eal: map shared config into exact same address as primary process Anatoly Burakov
2014-06-17 16:29 ` Ananyev, Konstantin [this message]
2014-06-18 9:25 ` Burakov, Anatoly
2014-06-17 15:36 ` [dpdk-dev] [PATCH 2/9] rte_tailq: change rte_dummy to rte_tailq_entry, add data pointer Anatoly Burakov
2014-06-17 15:36 ` [dpdk-dev] [PATCH 3/9] rte_ring: make ring tailq fully local Anatoly Burakov
2014-06-17 15:36 ` [dpdk-dev] [PATCH 4/9] rte_hash: make rte_hash " Anatoly Burakov
2014-06-17 15:36 ` [dpdk-dev] [PATCH 5/9] rte_fbk_hash: make rte_fbk_hash " Anatoly Burakov
2014-06-17 15:36 ` [dpdk-dev] [PATCH 6/9] rte_mempool: make mempool " Anatoly Burakov
2014-06-17 15:36 ` [dpdk-dev] [PATCH 7/9] rte_lpm: make lpm " Anatoly Burakov
2014-06-17 15:36 ` [dpdk-dev] [PATCH 8/9] rte_lpm6: make lpm6 " Anatoly Burakov
2014-06-17 15:36 ` [dpdk-dev] [PATCH 9/9] rte_acl: make acl " Anatoly Burakov
2014-06-18 11:27 ` [dpdk-dev] [PATCH v3 0/9] Make DPDK tailqs " Anatoly Burakov
2014-06-18 11:27 ` [dpdk-dev] [PATCH v3 1/9] eal: map shared config into exact same address as primary process Anatoly Burakov
2014-06-18 11:27 ` [dpdk-dev] [PATCH v3 2/9] rte_tailq: change rte_dummy to rte_tailq_entry, add data pointer Anatoly Burakov
2014-06-18 11:27 ` [dpdk-dev] [PATCH v3 3/9] rte_ring: make ring tailq fully local Anatoly Burakov
2014-06-18 11:27 ` [dpdk-dev] [PATCH v3 4/9] rte_hash: make rte_hash " Anatoly Burakov
2014-06-18 11:27 ` [dpdk-dev] [PATCH v3 5/9] rte_fbk_hash: make rte_fbk_hash " Anatoly Burakov
2014-06-18 11:27 ` [dpdk-dev] [PATCH v3 6/9] rte_mempool: make mempool " Anatoly Burakov
2014-06-18 11:27 ` [dpdk-dev] [PATCH v3 7/9] rte_lpm: make lpm " Anatoly Burakov
2014-06-18 11:27 ` [dpdk-dev] [PATCH v3 8/9] rte_lpm6: make lpm6 " Anatoly Burakov
2014-06-18 11:27 ` [dpdk-dev] [PATCH v3 9/9] rte_acl: make acl " Anatoly Burakov
2014-06-18 14:21 ` [dpdk-dev] [PATCH v3 0/9] Make DPDK tailqs " Ananyev, Konstantin
2014-06-20 15:42 ` [dpdk-dev] [PATCH 00/10] " Anatoly Burakov
2014-06-20 15:42 ` [dpdk-dev] [PATCH 01/10] eal: map shared config into exact same address as primary process Anatoly Burakov
2014-06-20 15:42 ` [dpdk-dev] [PATCH 02/10] eal: use --base-virtaddr for mapping rte_config as well Anatoly Burakov
2014-06-20 15:42 ` [dpdk-dev] [PATCH 03/10] rte_tailq: change rte_dummy to rte_tailq_entry, add data pointer Anatoly Burakov
2014-06-20 15:42 ` [dpdk-dev] [PATCH 04/10] rte_ring: make ring tailq fully local Anatoly Burakov
2014-06-20 15:42 ` [dpdk-dev] [PATCH 05/10] rte_hash: make rte_hash " Anatoly Burakov
2014-06-20 15:42 ` [dpdk-dev] [PATCH 06/10] rte_fbk_hash: make rte_fbk_hash " Anatoly Burakov
2014-06-20 15:42 ` [dpdk-dev] [PATCH 07/10] rte_mempool: make mempool " Anatoly Burakov
2014-06-20 15:42 ` [dpdk-dev] [PATCH 08/10] rte_lpm: make lpm " Anatoly Burakov
2014-06-20 15:42 ` [dpdk-dev] [PATCH 09/10] rte_lpm6: make lpm6 " Anatoly Burakov
2014-06-20 15:42 ` [dpdk-dev] [PATCH 10/10] rte_acl: make acl " Anatoly Burakov
2014-07-22 22:12 ` [dpdk-dev] [PATCH 00/10] Make DPDK tailqs " Thomas Monjalon
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=2601191342CEEE43887BDE71AB9772580EFB7832@IRSMSX105.ger.corp.intel.com \
--to=konstantin.ananyev@intel.com \
--cc=anatoly.burakov@intel.com \
--cc=dev@dpdk.org \
/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).