linux-kernel - Re: [RFC v2 12/14] arm64/lib: asid: Allow user to update the context under the lock

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]

Date:   Mon, 15 Jul 2019 15:38:42 +0100
From:   Julien Grall <julien.grall@....com>
To:     James Morse <james.morse@....com>
Cc:     linux-kernel@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
        kvmarm@...ts.cs.columbia.edu, marc.zyngier@....com,
        julien.thierry@....com, suzuki.poulose@....com,
        catalin.marinas@....com, will.deacon@....com
Subject: Re: [RFC v2 12/14] arm64/lib: asid: Allow user to update the context
 under the lock



On 03/07/2019 18:35, James Morse wrote:
> Hi Julien,

Hi James,

> On 20/06/2019 14:06, Julien Grall wrote:
>> Some users of the ASID allocator (e.g VMID) will require to update the
>> context when a new ASID is generated. This has to be protected by a lock
>> to prevent concurrent modification.
>>
>> Rather than introducing yet another lock, it is possible to re-use the
>> allocator lock for that purpose. This patch introduces a new callback
>> that will be call when updating the context.
> 
> You're using this later in the series to mask out the generation from the atomic64 to
> leave just the vmid.

You are right.

> 
> Where does this concurrent modification happen? The value is only written if we have a
> rollover, and while its active the only bits that could change are the generation.
> (subsequent vCPUs that take the slow path for the same VM will see the updated generation
> and skip the new_context call)
> 
> If we did the generation filtering in update_vmid() after the call to
> asid_check_context(), what would go wrong?
> It happens more often than is necessary and would need a WRITE_ONCE(), but the vmid can't
> change until we become preemptible and another vCPU gets a chance to make its vmid active.

I think I was over cautious. Pre-filtering after asid_check_context() is equally 
fine as long as update_vttbr() is called from preemptible context.

Cheers,

-- 
Julien Grall