linux-kernel - Re: [PATCH 1/1] nvdimm: allow exposing RAM carveouts as NVDIMM DIMM devices

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <6965857b34958_f1ef61008b@iweiny-mobl.notmuch>
Date: Mon, 12 Jan 2026 17:36:27 -0600
From: Ira Weiny <ira.weiny@...el.com>
To: Michał Cłapiński <mclapinski@...gle.com>,
	<dan.j.williams@...el.com>
CC: Mike Rapoport <rppt@...nel.org>, Ira Weiny <ira.weiny@...el.com>, "Dave
 Jiang" <dave.jiang@...el.com>, Vishal Verma <vishal.l.verma@...el.com>,
	<jane.chu@...cle.com>, Pasha Tatashin <pasha.tatashin@...een.com>, "Tyler
 Hicks" <code@...icks.com>, <linux-kernel@...r.kernel.org>,
	<nvdimm@...ts.linux.dev>
Subject: Re: [PATCH 1/1] nvdimm: allow exposing RAM carveouts as NVDIMM DIMM
 devices

Michał Cłapiński wrote:
> On Wed, Dec 17, 2025 at 4:14 AM <dan.j.williams@...el.com> wrote:
> >
> > Michał Cłapiński wrote:
> > [..]
> > > > Sure, then make it 1280K of label space. There's no practical limit in
> > > > the implementation.
> > >
> > > Hi Dan,
> > > I just had the time to try this out. So I modified the code to
> > > increase the label space to 2M and I was able to create the
> > > namespaces. It put the metadata in volatile memory.
> > >
> > > But the infoblocks are still within the namespaces, right? If I try to
> > > create a 3G namespace with alignment set to 1G, its actual usable size
> > > is 2G. So I can't divide the whole pmem into 1G devices with 1G
> > > alignment.
> >
> > Ugh, yes, I failed to predict that outcome.
> >
> > > If I modify the code to remove the infoblocks, the namespace mode
> > > won't be persistent, right? In my solution I get that information from
> > > the kernel command line, so I don't need the infoblocks.
> >
> > So, I dislike the command line option ABI expansion proposal enough to
> > invest some time to find an alternative. One observation is that the
> > label is able to indicate the namespace mode independent of an
> > info-block. The info-block is only really needed when deciding whether
> > and how much space to reserve to allocate 'struct page' metadata.
> >
> > -- 8< --
> > From 4f44cbb6e3bd4cac9481bdd4caf28975a4f1e471 Mon Sep 17 00:00:00 2001
> > From: Dan Williams <dan.j.williams@...el.com>
> > Date: Mon, 15 Dec 2025 17:10:04 -0800
> > Subject: [PATCH] nvdimm: Allow fsdax and devdax namespace modes without
> >  info-blocks
> > MIME-Version: 1.0
> > Content-Type: text/plain; charset=UTF-8
> > Content-Transfer-Encoding: 8bit
> >
> > Michał reports that the new ramdax facility does not meet his needs which
> > is to carve large reservations of memory into multiple 1GB aligned
> > namespaces/volumes. While ramdax solves the problem of in-memory
> > description of the volume layout, the nvdimm "infoblocks" eat capacity and
> > destroy alignment properties.
> >
> > The infoblock serves 2 purposes, it indicates whether the namespace should
> > operate in fsdax or devdax mode, Michał needs this, and it optionally
> > reserves namespace capacity for storing 'struct page' metadata, Michał does
> > not need this. It turns out the mode information is already recorded in the
> > namespace label, and if no reservation is needed for 'struct page' metadata
> > then infoblock settings can just be hard coded.
> >
> > Introduce a new ND_REGION_VIRT_INFOBLOCK flag for ramdax to indicate that
> > all infoblocks be synthesized and not consume any capacity from the
> > namespace.
> >
> > With that ramdax can create a full sized namespace:
> >
> > $ ndctl create-namespace -r region0 -s 1G -a 1G -M mem
> > {
> >   "dev":"namespace0.0",
> >   "mode":"fsdax",
> >   "map":"mem",
> >   "size":"1024.00 MiB (1073.74 MB)",
> >   "uuid":"c48c4991-86af-4de1-8c7c-8919358df1f9",
> >   "sector_size":512,
> >   "align":1073741824,
> >   "blockdev":"pmem0"
> > }
> 
> Thank you for working on this.
> 
> I tried it an indeed it works with fsdax. It doesn't seem to work with
> devdax though.
> 
> $ ndctl create-namespace -v -r region1 -m devdax -a 1G -M mem -s 1G
> libndctl: ndctl_dax_enable: dax1.0: failed to enable
>   Error: namespace1.1: failed to enable
> 
> failed to create namespace: No such device or address
> 
> $ dmesg | grep dax
> [...]
> [   29.504763] dax_pmem dax1.0: could not reserve metadata
> [   29.504766] dax_pmem dax1.0: probe with driver dax_pmem failed with error -16
> [   29.506553] probe of dax1.0 returned 16 after 1815 usecs
> [...]
> 
> I think the dax_pmem driver needs to be modified too.

Michał

Did yall have a suggestion on how?  I lost track of this thread over the
holidays and so I'm not sure where this stands.

Ira