linux-kernel - Re: [PATCH V3] iscsi_ibft: fix crash due to KASLR physical memory remapping

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: <YRN01YySPVucdCF0@infradead.org>
Date:   Wed, 11 Aug 2021 07:57:25 +0100
From:   Christoph Hellwig <hch@...radead.org>
To:     Konrad Rzeszutek Wilk <konrad@...nok.org>
Cc:     Christoph Hellwig <hch@...radead.org>,
        Maurizio Lombardi <mlombard@...hat.com>, rppt@...nel.org,
        bp@...en8.de, tglx@...utronix.de, x86@...nel.org,
        pjones@...hat.com, konrad@...nel.org, george.kennedy@...cle.com,
        rafael@...nel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH V3] iscsi_ibft: fix crash due to KASLR physical memory
 remapping

On Tue, Aug 10, 2021 at 01:55:11PM -0400, Konrad Rzeszutek Wilk wrote:
> On Tue, Aug 10, 2021 at 06:00:24PM +0100, Christoph Hellwig wrote:
> > > Fix this bug by saving the address of the physical location
> > > of the ibft; later the driver will use isa_bus_to_virt() to get
> > > the correct virtual address.
> > 
> > That sound rather broken.  Why not save the physical address in
> > find_ibft_region and then later ioremap that when a virtual address is
> > needed like all other code accessing magic I/O memory?
> 
> That is kind of what he does. The physical address is saved as a global
> static variable and also the physical address is memreserved. Then
> later on the physical address is used to create the virtual address.

Except that it uses isa_bus_to_virt, which is really broken.

> Or are you thinking of making the find_ibft_region reserve the physical
> address, and _cache_ the physical address so there is no global
> variable ?

No.  Just switch to ioremap/early_ioremap insted of this isa_bus_to_virt
mess.