linux-kernel - Re: [PATCH v3 1/1] iommu/sva: Invalidate KVA range on kernel TLB flush

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: <ba372b0d-beba-4564-a8a8-84318e5d1238@intel.com>
Date: Thu, 28 Aug 2025 12:31:40 -0700
From: Dave Hansen <dave.hansen@...el.com>
To: Jason Gunthorpe <jgg@...dia.com>
Cc: "Tian, Kevin" <kevin.tian@...el.com>, Baolu Lu
 <baolu.lu@...ux.intel.com>, Joerg Roedel <joro@...tes.org>,
 Will Deacon <will@...nel.org>, Robin Murphy <robin.murphy@....com>,
 Jann Horn <jannh@...gle.com>, Vasant Hegde <vasant.hegde@....com>,
 Alistair Popple <apopple@...dia.com>, Peter Zijlstra <peterz@...radead.org>,
 Uladzislau Rezki <urezki@...il.com>,
 Jean-Philippe Brucker <jean-philippe@...aro.org>,
 Andy Lutomirski <luto@...nel.org>, "Lai, Yi1" <yi1.lai@...el.com>,
 "iommu@...ts.linux.dev" <iommu@...ts.linux.dev>,
 "security@...nel.org" <security@...nel.org>,
 "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
 "stable@...r.kernel.org" <stable@...r.kernel.org>,
 "vishal.moola@...il.com" <vishal.moola@...il.com>,
 Matthew Wilcox <willy@...radead.org>
Subject: Re: [PATCH v3 1/1] iommu/sva: Invalidate KVA range on kernel TLB
 flush

On 8/28/25 12:10, Jason Gunthorpe wrote:
>> The biggest single chunk of code is defining ptdesc_*_kernel(). The rest
>> of it is painfully simple.
> Seems not great to be casting ptdesc to folio just to use
> folio_set_referenced(), I'm pretty sure that is no the direction
> things are going in..

Ideally, the ptdesc->__page_flags fields would have their own set of
macros just like folios do, right? Alas, I was too lazy to go to the
trouble of conjuring those up for this single bit.

> I'd set the flag yourself or perhaps use the memory of __page_mapping
> to store a flag (but zero it before freeing).

That was my first inclination as well (actually I was going to use
__page_type since the lower 24 bits are evidently open, but same idea).

Willy pointed me in the direction of PG_referenced on IRC, so I took
that path. But I'm open to doing whatever. It's just three lines of code
that probably end up boiling down to

	bts 0x1, 0x8(%rax)
versus
	bts 0x2, 0x16(%rax)

;)