linux-kernel - Re: [PATCH 1/3] x86/intel_rdt: Fix issue when mkdir uses a freed CLOSid

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:   Wed, 5 Apr 2017 11:03:05 -0700 (PDT)
From:   Shivappa Vikas <vikas.shivappa@...el.com>
To:     Thomas Gleixner <tglx@...utronix.de>
cc:     Vikas Shivappa <vikas.shivappa@...ux.intel.com>,
        vikas.shivappa@...el.com, x86@...nel.org,
        linux-kernel@...r.kernel.org, hpa@...or.com, mingo@...nel.org,
        peterz@...radead.org, ravi.v.shankar@...el.com,
        tony.luck@...el.com, fenghua.yu@...el.com, h.peter.anvin@...el.com
Subject: Re: [PATCH 1/3] x86/intel_rdt: Fix issue when mkdir uses a freed
 CLOSid



On Wed, 5 Apr 2017, Thomas Gleixner wrote:

> On Mon, 3 Apr 2017, Vikas Shivappa wrote:
>
>> Subject: x86/intel_rdt: Fix issue when mkdir uses a freed CLOSid
>
> This subject line is useless again. It want's to be descriptive.
>
> "Fix issue" Which issue?
>
>> Each resctrl directory has one CLOSid allocated which is mapped to a
>> control register/QOS_MSR. During an rmdir this CLOSid is freed and can
>> be reused later when a new directory is created.  Currently we do not
>> reset the QOS_MSR to a default when the CLOSid is freed. So when the
>> next mkdir uses a freed CLOSid, the new directory is associated with a
>> stale QOS_MSR.
>
> That's not an issue. That's maybe something people are not expecting.
>
>> Fix this issue by writing a default value to the QOS_MSR when the
>> associated CLOSid is freed. The default value is all 1s for CBM which
>> means no control is enforced when a mkdir reuses this CLOSid.
>
> That's just wrong.
>
> The proper behaviour for a new control group is, that at the time when it
> is created it copies the CBM values of the default group and not claiming
> access to ALL of the cache by default.

The behaviour of the default/root group is having access to all cache. Because 
we set the value of all CBMs to all 1s and then root group takes CLOS 0. This is 
trying to set the same values for all newly created groups.

Thanks,
Vikas

>
> Thanks,
>
> 	tglx
>
>
>