public inbox for initramfs@vger.kernel.org
 help / color / mirror / Atom feed
From: Xunlei Pang <xpang-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
To: Xunlei Pang <xlpang-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>,
	initramfs-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Cc: Dave Young <dyoung-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>,
	Harald Hoyer <harald-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
Subject: Re: [PATCH] kernel-modules: handle nfs host fstype properly
Date: Wed, 15 Jun 2016 18:01:41 +0800	[thread overview]
Message-ID: <57612785.1030200@redhat.com> (raw)
In-Reply-To: <1465983624-12964-1-git-send-email-xlpang-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>

Hi Harald,

On 2016/06/15 at 17:40, Xunlei Pang wrote:
> There are scenarios that support fstype target not mounted
> beforehand, kdump can utilize this feature to avoid imposing
> severe overload on the shared nfs server when thousands of
> diskless clients with nfs kdumping boot and trigger kdump
> initramfs rebuild synchronously.
>
> Because target is not mounted beforehand, some of the kernel
> modules will fail to be installed in hostonly mode. Dracut
> actually can support this by instmods all "host_fs_types" with
> hostonly unset in 90kernel-modules/module-setup.sh installkernel().
>
> This works properly for most host fstypes(ext[234]/xfs, etc), but
> that's not the case for nfs.
>
> From nfs manpage, we can clearly see:
> "The fstype field contains "nfs". Use of the "nfs4" fstype in
> /etc/fstab is deprecated."
>
> It means "nfs" specified can be actually mounted as "nfs[3-4]", so
> for these cases, the current 90kernel-modules implementation only
> installs ko modules for "nfs"(nfs.ko, etc), lacking nfsv[3-4].ko,
> so it cannot work properly. In order to address the issue, we should
> install all the possbile kernel modules for "nfs" fstype:
> "nfs sunrpc ipv6 nfsv2 nfsv3 nfsv4 nfs_acl nfs_layout_nfsv41_files".

The list is directly copied from "95nfs/module_setup.sh installkernel()",
I am not sure if it is enough for all the nfs cases.

I also tried to use insmods "=fs/nfs" instead, it installed more kernel
modules compared with the list, for example installed extra:

  fs/nfs/blocklayout/blocklayoutdriver.ko
  fs/nfs/flexfilelayout/nfs_layout_flexfiles.ko
  fs/nfs/objlayout/objlayoutdriver.ko


I don't know if these kernel modules are necessary, if so I can change
to use "=fs/nfs" instead.

Regards,
Xunlei

>
> Additionally, we need a special remapping for "nfs[3-4]" deprecated
> cases, for example, we should map "nfs4" to "nfsv4.ko" not "nfs4.ko".
>
> Signed-off-by: Xunlei Pang <xlpang-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
> ---
>  modules.d/90kernel-modules/module-setup.sh | 20 +++++++++++++++++++-
>  1 file changed, 19 insertions(+), 1 deletion(-)
>
> diff --git a/modules.d/90kernel-modules/module-setup.sh b/modules.d/90kernel-modules/module-setup.sh
> index 7904c02..1f57ec7 100755
> --- a/modules.d/90kernel-modules/module-setup.sh
> +++ b/modules.d/90kernel-modules/module-setup.sh
> @@ -44,7 +44,25 @@ installkernel() {
>                  dracut_instmods -o -P ".*/(kernel/fs/nfs|kernel/fs/nfsd|kernel/fs/lockd)/.*" '=fs'
>              fi
>          else
> -            hostonly='' instmods "${host_fs_types[@]}"
> +            for i in "${host_fs_types[@]}"; do
> +                if [[ $i = nfs ]]; then
> +                    # nfs manpage says:
> +                    # "The fstype field contains "nfs". Use of the "nfs4" fstype in
> +                    # /etc/fstab is deprecated."
> +                    #
> +                    # It means "nfs" can be actually mounted as "nfs4", so for "nfs"
> +                    # host fstype we better install all the possible kernel modules,
> +                    # this is useful for kdump "nfs" dumping not mounted beforehand.
> +                    #
> +                    # The list is copied from "95nfs/module_setup.sh installkernel()".
> +                    i="nfs sunrpc ipv6 nfsv2 nfsv3 nfsv4 nfs_acl nfs_layout_nfsv41_files"
> +                elif [[ $i = nfs[3-4] ]]; then
> +                    # Deprecated cases, we provide support, but need a name mapping:
> +                    # For example, should map "nfs4" to use "nfsv4.ko" not "nfs4.ko".
> +                    i=${i/nfs/nfsv}
> +                fi
> +                hostonly='' instmods $i
> +            done
>          fi
>      fi
>      :

  parent reply	other threads:[~2016-06-15 10:01 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-06-15  9:40 [PATCH] kernel-modules: handle nfs host fstype properly Xunlei Pang
     [not found] ` <1465983624-12964-1-git-send-email-xlpang-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2016-06-15 10:01   ` Xunlei Pang [this message]
     [not found]     ` <57612785.1030200-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2016-10-26  7:43       ` Harald Hoyer
     [not found]         ` <462e1435-0412-904c-e8df-05b1e71b1d39-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
2016-10-26  7:49           ` Xunlei Pang
2016-06-15 10:06   ` Dracut GitHub Import Bot

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=57612785.1030200@redhat.com \
    --to=xpang-h+wxahxf7alqt0dzr+alfa@public.gmane.org \
    --cc=dyoung-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org \
    --cc=harald-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org \
    --cc=initramfs-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
    --cc=xlpang-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.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