DPDK patches and discussions
 help / color / mirror / Atom feed
From: Olivier Matz <olivier.matz@6wind.com>
To: Bruce Richardson <bruce.richardson@intel.com>
Cc: David Marchand <david.marchand@redhat.com>,
	Shanmugasundaram M <shansirtt@gmail.com>, dev <dev@dpdk.org>,
	Guillaume Gaudonville <guillaume.gaudonville@6wind.com>,
	Jean-Mickael Guerin <jmg@6wind.com>
Subject: Re: [dpdk-dev] Reg 6Wind fastpath DPDK 18.11.1 virtual memory allocation ( 54.5GB of anonymous mapping)
Date: Thu, 10 Jun 2021 09:17:10 +0200	[thread overview]
Message-ID: <YMG8dmzzJYtoMmzr@platinum> (raw)
In-Reply-To: <YMC4K7WUlmLUha9K@bricha3-MOBL.ger.corp.intel.com>

Hi Shanmugasundaram,

On Wed, Jun 09, 2021 at 01:46:35PM +0100, Bruce Richardson wrote:
> On Wed, Jun 09, 2021 at 02:28:04PM +0200, David Marchand wrote:
> > On Wed, Jun 9, 2021 at 2:02 PM Shanmugasundaram M <shansirtt@gmail.com> wrote:
> > > We are using  6Wind fastpath(fp-rte) which is using DPDK version 18.11.1.
> > > We are seeing  65.5G(preallocated during startup) of virtual memory for
> > > fp-rte on a 32GB machine, 10G of that was huge pages (5000 mappings of 2M.)
> > >
> > > We found 4 strange anon (4K size) segments totaling 54.5G (16+16+16+6.5).
> > > Note: we changed / customized memory CONFIG_RTE_MAX_MEMSEG_PER_LIST=24576
> > > and default value of  CONFIG_RTE_MAX_MEMSEG_LISTS=64 used.
> > > Kindly let us know
> > > 1. Why are allocating  54.5 GB anonymous mapping
> > > 2. When will we use it ?
> > 
> > This is likely pre-reserved virtual memory from the DPDK allocator.
> > This virtual memory won't probably never be used.
> > 
> 
> There may not be actually any memory backing this, since DPDK will reserve
> huge amounts of virtual address space without actually using it. So those
> GB's you see in the virtual memory may be completely unused, the addresses
> are just reserved to ensure that if memory does need to be mapped into the
> process, we have an appropriate contiguous address block to put it in.
> 
> > But you'd better ask 6WIND guys as they maintain their own DPDK downstream.
> > 
> > Copying a few guys.

I agree with David and Bruce. This is probably related to virtual memory
areas reserved by eal_get_virtual_area() (mmap of /dev/zero). Enabling EAL
debug logs should show traces about it.

If you need support (even on old 6WINDGate versions based on
dpdk-18.11), please open a ticket at 6WIND support, we'll be happy to
help.

Regards,
Olivier

      reply	other threads:[~2021-06-10  7:17 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-03-09 19:00 [dpdk-dev] Reg DPDK EAL log-level=8 options is not throwing any DEBUG messages in /var/log/messages Shanmugasundaram M
2021-03-09 20:29 ` Stephen Hemminger
2021-03-10 18:19   ` Shanmugasundaram M
2021-03-10 18:25     ` Stephen Hemminger
2021-03-10 18:32       ` Shanmugasundaram M
2021-06-09 12:02 ` [dpdk-dev] Reg 6Wind fastpath DPDK 18.11.1 virtual memory allocation ( 54.5GB of anonymous mapping) Shanmugasundaram M
2021-06-09 12:28   ` David Marchand
2021-06-09 12:46     ` Bruce Richardson
2021-06-10  7:17       ` Olivier Matz [this message]

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=YMG8dmzzJYtoMmzr@platinum \
    --to=olivier.matz@6wind.com \
    --cc=bruce.richardson@intel.com \
    --cc=david.marchand@redhat.com \
    --cc=dev@dpdk.org \
    --cc=guillaume.gaudonville@6wind.com \
    --cc=jmg@6wind.com \
    --cc=shansirtt@gmail.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).