linux-kernel - Re: [PATCH 2/3] arm64: pageattr: Use walk_page_range

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: Windows password security audit tool. GUI, reports in PDF.

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

Message-ID: <eb0b075a-e6f8-4561-ba3c-3019710847d7@lucifer.local>
Date: Mon, 9 Jun 2025 12:00:00 +0100
From: Lorenzo Stoakes <lorenzo.stoakes@...cle.com>
To: Dev Jain <dev.jain@....com>
Cc: akpm@...ux-foundation.org, david@...hat.com, catalin.marinas@....com,
        will@...nel.org, Liam.Howlett@...cle.com, vbabka@...e.cz,
        rppt@...nel.org, surenb@...gle.com, mhocko@...e.com,
        linux-mm@...ck.org, linux-kernel@...r.kernel.org,
        suzuki.poulose@....com, steven.price@....com, gshan@...hat.com,
        linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH 2/3] arm64: pageattr: Use walk_page_range_novma() to
 change memory permissions

On Mon, Jun 09, 2025 at 03:11:59PM +0530, Dev Jain wrote:
>
> On 06/06/25 3:19 pm, Lorenzo Stoakes wrote:
> > On Fri, May 30, 2025 at 02:34:06PM +0530, Dev Jain wrote:
> > > Move away from apply_to_page_range(), which does not honour leaf mappings,
> > > to walk_page_range_novma(). The callbacks emit a warning and return EINVAL
> > > if a partial range is detected.
> > Hm a follow up question here - why not just improve apply_to_page_range() to
> > honour leaf mappings?
> >
> > What does honouring leaf mappings actually mean? You mean handling huge pages?
> >
> > Would it be all that difficult to implement?
> >
> > It seems like you're pushing a bunch of the 'applying' logic over from there to
> > a walker that isn't maybe best suited to it and having to introduce an iffy new
> > form of locking...
> >
> > Can we go vice-versa? :)
> >
> > Also obviously walk_page_range_novma() doesn't exist any more :P
> > walk_kernel_page_table_range() is the preferred solution.
>
> I cannot see this function in mm-unstable. Also, mm-unstable is broken - I get
> some RCU message in dmesg, and after some 20-30 second delay the kernel boots.
> So on which branch do I base my work on...

mm-unstable shouldn't be broken as that's what should be in -next, concerning!
Worth investigating... But rebase! :)

Sorry maybe isn't clear as we changed this recently - you should base all
changes intended for the next release on mm-new.

As I understand it:

- mm-new is the _rawest_ form of the state of mm, Andrew described it as
  the 'crazy' branch that basiclaly has everything.

- mm-unstable is when things have been percolating for a while and are
  considered reasonably stable-ish kinda, but most importantly - ready for
  -next testing. And is what goes to -next.

- mm-stable is gathered shortly before the merge window starts and is all
  the stuff Andrew will send to Linus.

To pick up on the most recent changes therefore use mm-new. Using anything
else risks issues like this where your patch will conflict, etc.

Another point to note is, during the merge window, mm-new is where changes
due for the release after the one being currently merged are kept
(e.g. over the past 2 weeks of 6.16 merge window, this would be changes for
6.17).

Not all subsystems even take patches at all during the merge window, but mm
does.

So especially during merge window it's important to base you changes on
mm-new.

>
> >

Cheers, Lorenzo