linux-kernel - Re: [RESEND PATCH v1 1/5] mm: vmalloc must set pte via arch code

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: <db1b8b8f-c3ac-4dbb-8b15-125c0929856a@lucifer.local>
Date:   Sat, 13 May 2023 14:14:18 +0100
From:   Lorenzo Stoakes <lstoakes@...il.com>
To:     Ryan Roberts <ryan.roberts@....com>
Cc:     Andrew Morton <akpm@...ux-foundation.org>,
        "Matthew Wilcox (Oracle)" <willy@...radead.org>,
        "Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com>,
        SeongJae Park <sj@...nel.org>, linux-kernel@...r.kernel.org,
        linux-mm@...ck.org, damon@...ts.linux.dev,
        Christoph Hellwig <hch@...radead.org>,
        Uladzislau Rezki <urezki@...il.com>
Subject: Re: [RESEND PATCH v1 1/5] mm: vmalloc must set pte via arch code

You've not cc'd the vmalloc reviewers, including the author of 3e9a9e256b1e
whose patch you purport to fix. Please remember to run get_maintainers.pl
on all files you patch and cc them at least on relevant patches.

Have added Christoph + Uladzislau as cc.

You'll definitely want an ack from Christoph on this!

On Thu, May 11, 2023 at 02:21:09PM +0100, Ryan Roberts wrote:
> It is bad practice to directly set pte entries within a pte table.
> Instead all modifications must go through arch-provided helpers such as
> set_pte_at() to give the arch code visibility and allow it to validate
> (and potentially modify) the operation.

This does make sense, and I see for example in xtensa that an arch-specific
instruction is issued under certain circumstances so I do suspect we should
do this.

As for validation, the function never indicates an error, so only in the
sense that a WARN_ON() could _in theory_ trigger is it being
validated. This might be quite a nitty point :) as set_pte_at() has no
means of indicating an error. But maybe to be pedantic 'check' rather than
'validate'?

>
> Fixes: 3e9a9e256b1e ("mm: add a vmap_pfn function")

Not sure if this is really 'fixing' anything, I mean ostensibly, but not
sure if the tag is relevant here, that is more so for a bug being
introduced, and unless an issue has arisen not sure if it's
appropriate. But this might be a nit, again!

> Signed-off-by: Ryan Roberts <ryan.roberts@....com>
> ---
>  mm/vmalloc.c | 5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/mm/vmalloc.c b/mm/vmalloc.c
> index 9683573f1225..d8d2fe797c55 100644
> --- a/mm/vmalloc.c
> +++ b/mm/vmalloc.c
> @@ -2899,10 +2899,13 @@ struct vmap_pfn_data {
>  static int vmap_pfn_apply(pte_t *pte, unsigned long addr, void *private)
>  {
>  	struct vmap_pfn_data *data = private;
> +	pte_t ptent;
>
>  	if (WARN_ON_ONCE(pfn_valid(data->pfns[data->idx])))
>  		return -EINVAL;
> -	*pte = pte_mkspecial(pfn_pte(data->pfns[data->idx++], data->prot));
> +
> +	ptent = pte_mkspecial(pfn_pte(data->pfns[data->idx++], data->prot));
> +	set_pte_at(&init_mm, addr, pte, ptent);

While we're refactoring, it'd be nice to stash data->pfns[data->idx] into a
local pfn variable.

>  	return 0;
>  }
>
> --
> 2.25.1
>