All of lore.kernel.org
 help / color / mirror / Atom feed
From: <ian_bruce@mail.ru>
To: NeilBrown <neilb@suse.com>
Cc: linux-raid@vger.kernel.org
Subject: Re: [BUG] non-metadata arrays cannot use more than 27 component devices
Date: Wed, 1 Mar 2017 05:05:47 -0800	[thread overview]
Message-ID: <20170301050529.5978de82.ian_bruce@mail.ru> (raw)
In-Reply-To: <87mvd6oy8n.fsf@notabene.neil.brown.name>

On Wed, 01 Mar 2017 07:29:28 +1100
NeilBrown <neilb@suse.com> wrote:

> Thanks.  That makes it easy.
> Test works with my patch applied.

Thanks for fixing that.

If anybody is curious, the application for this capability is as
follows. For live systems running from a USB flashdrive, we need to
loop-mount an ext4 filesystem image from the fat32-formatted flashdrive.
Unfortunately, the maximum file size on fat32 is 4GB, which is a severe
limitation, when 128GB flashdrives are commonly available.

The solution is to split the ext4 image into multiple sub-4GB chunks,
associate a /dev/loop device with each of those files, have mdadm turn
those into a single RAID device, and mount that as the ext4 filesystem.
It is preferable to use non-metadata, linear-mode RAID for this, because
we can then convert back and forth between the single filesystem image
and its constituent chunks using the non-privileged utilities "cat" and
"split". With a maximum of 27 RAID component devices, the maximum
filesystem size would be 108GB, which is not quite a complete solution.


On Fri, 24 Feb 2017 15:46:19 -0500
Phil Turmel <philip@turmel.org> wrote:

> Note that build mode doesn't support a bunch of other MD raid features
> either, like all of the parity raid levels. That it doesn't support
> v1+ metadata isn't a surprise, and isn't the only legacy feature that
> only uses legacy metadata (built-in kernel auto-assembly gets the most
> whining, actually).

> If you think its trivial to implement --build with v1.x metadata, go
> right ahead. Post your patches for review.

I haven't tested "mdadm --build" with parity RAID myself (although the
/dev/loop trick would probably suffice for that too), but if this is so,
would the change to provide that be as simple as the patch to remove the
27-component limitation? (Although I suppose that unlike linear mode,
the component devices for parity mode would have to be initialized with
consistent data, first.)

Somebody might find a use for non-metadata, parity-mode RAID, if it were
available.


-- Ian Bruce

      reply	other threads:[~2017-03-01 13:05 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-02-24 12:08 [BUG] non-metadata arrays cannot use more than 27 component devices ian_bruce
2017-02-24 15:20 ` Phil Turmel
2017-02-24 16:40   ` ian_bruce
2017-02-24 20:46     ` Phil Turmel
2017-02-25 20:05       ` Anthony Youngman
2017-02-25 22:00         ` Phil Turmel
2017-02-25 23:30           ` Wols Lists
2017-02-25 23:41             ` Phil Turmel
2017-02-25 23:55               ` Wols Lists
2017-02-26  0:07                 ` Phil Turmel
2017-03-01 15:02                   ` Wols Lists
2017-03-01 17:23                     ` Phil Turmel
2017-03-01 18:13                       ` Phil Turmel
2017-03-01 19:50                         ` Anthony Youngman
2017-03-01 22:20                           ` Phil Turmel
2017-02-27  5:55 ` NeilBrown
2017-02-28 10:25   ` ian_bruce
2017-02-28 20:29     ` NeilBrown
2017-03-01 13:05       ` ian_bruce [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=20170301050529.5978de82.ian_bruce@mail.ru \
    --to=ian_bruce@mail.ru \
    --cc=linux-raid@vger.kernel.org \
    --cc=neilb@suse.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.