DPDK usage discussions
 help / color / mirror / Atom feed
From: James Huang <jamsphon@gmail.com>
To: users@dpdk.org
Subject: Re: [dpdk-users] DPDK program huge core file size
Date: Tue, 23 Feb 2021 11:22:28 -0800	[thread overview]
Message-ID: <CAFpuyR7uSn4A1AYrZ++9uNx7Oq3VHGs1RqdQVFiZ7JhnPZqdMQ@mail.gmail.com> (raw)
In-Reply-To: <CAFpuyR7Mf-MMOx3Jyzy7rZNfqf6e6WwxBqdAuUAOxGwZHdjrtw@mail.gmail.com>

UPDATE: the 'kill -6' command does not dump the hugepage memory zone into
the core file.

Is there a way to bypass the hugepage memory zone dump into the core file
with running gcore command ?


On Fri, Feb 19, 2021 at 11:18 AM James Huang <jamsphon@gmail.com> wrote:

> On CentOS7, we observed that the program (based on dpdk 19.11) creates a
> huge core file size, i.e. 100+GB, far bigger than the expected <4GB. even
> though the system only installs 16GB memory, and allocates 1GB hugepage
> size at boot time. no matter if the core file is created by program panic
> (segfault), or run with tool gcore.
>
> On CentOS 6, the program (based on dpdk 17.05), the core file is the
> expected size.
>
> On CentOS7, we tried to adjust the process coredump_filter combinations,
> it found only when clean the bit 0 can avoid the huge core size, however, a
> cleared bit 0 generate small core file (200MB) and is meaningless for debug
> purposes, i.e. gdb bt command does not output.
>
> Is there a way to avoid dumping the hugepage memory, while remaining other
> memory in the core file?
>
> The following is the program pmap output comparison.
> on CentOS 6, the hugepage resides on the process user space:
> ...
> 00007f4e80000000 1048576K rw-s-  /mnt/huge_1GB/rtemap_0
> 00007f4ec0000000   2048K rw-s-
> /sys/devices/pci0000:00/0000:00:02.0/0000:04:00.0/resource0
> 00007f4ec0200000     16K rw-s-
> /sys/devices/pci0000:00/0000:00:02.0/0000:04:00.0/resource4
> 00007f4ec0204000   2048K rw-s-
> /sys/devices/pci0000:00/0000:00:02.0/0000:04:00.1/resource0
> 00007f4ec0404000     16K rw-s-
> /sys/devices/pci0000:00/0000:00:02.0/0000:04:00.1/resource4
> ...
>
>
> on CentOS 7, the hugepage resides on the process system space::
> ...
> 0000000100000000     20K rw-s- config
> 0000000100005000    184K rw-s- fbarray_memzone
> 0000000100033000      4K rw-s- fbarray_memseg-1048576k-0-0
> 0000000140000000 1048576K rw-s- rtemap_0
> 0000000180000000 32505856K r----   [ anon ]
> 0000000940000000      4K rw-s- fbarray_memseg-1048576k-0-1
> 0000000980000000 33554432K r----   [ anon ]
> 0000001180000000      4K rw-s- fbarray_memseg-1048576k-0-2
> 00000011c0000000 33554432K r----   [ anon ]
> 00000019c0000000      4K rw-s- fbarray_memseg-1048576k-0-3
> 0000001a00000000 33554432K r----   [ anon ]
> 0000002200000000   1024K rw-s- resource0
> 0000002200100000     16K rw-s- resource3
> 0000002200104000   1024K rw-s- resource0
> 0000002200204000     16K rw-s- resource3
> ...
>
> Thanks,
> -James
>

  reply	other threads:[~2021-02-23 19:22 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-02-19 19:18 James Huang
2021-02-23 19:22 ` James Huang [this message]
2021-02-24  3:59   ` Li Feng
2021-02-25 17:37     ` James Huang
2021-02-25 18:23       ` James Huang
2021-02-26 16:00     ` David Marchand

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=CAFpuyR7uSn4A1AYrZ++9uNx7Oq3VHGs1RqdQVFiZ7JhnPZqdMQ@mail.gmail.com \
    --to=jamsphon@gmail.com \
    --cc=users@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).