linux-kernel - Re: [RFC PATCH v10 0/5] fix vt-d hard lockup when hotplug ATS capable device

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] [day] [month] [year] [list]

Message-ID: <f6e57309-be5a-4a84-a280-6ed00a550548@linux.intel.com>
Date: Thu, 18 Jan 2024 10:32:20 +0800
From: Ethan Zhao <haifeng.zhao@...ux.intel.com>
To: Baolu Lu <baolu.lu@...ux.intel.com>, kevin.tian@...el.com,
 bhelgaas@...gle.com, dwmw2@...radead.org, will@...nel.org,
 robin.murphy@....com, lukas@...ner.de
Cc: linux-pci@...r.kernel.org, iommu@...ts.linux.dev,
 linux-kernel@...r.kernel.org
Subject: Re: [RFC PATCH v10 0/5] fix vt-d hard lockup when hotplug ATS capable
 device


On 1/18/2024 10:26 AM, Ethan Zhao wrote:
>
> On 1/18/2024 8:46 AM, Baolu Lu wrote:
>> On 1/17/24 5:00 PM, Ethan Zhao wrote:
>>> +       /*
>>> +        * If the ATS invalidation target device is gone this moment 
>>> (surprise
>>> +        * removed, died, no response) don't try this request again. 
>>> this
>>> +        * request will not get valid result anymore. but the 
>>> request was
>>> +        * already submitted to hardware and we predict to get a ITE in
>>> +        * followed batch of request, if so, it will get handled then.
>>> +        */
>>> +       if (target_pdev && !pci_device_is_present(target_pdev))
>>> +               return -EINVAL;
>>
>> Again, we should not ignore the error triggered by the current request.
>> Do not leave it to the next one. The WAIT descriptor is a fence. Handle
>> everything within its boundary.
>
>  We didn't set fence bit to every ATS invalidation wait descriptor,
>
> only the intel_drain_pasid_prq() queue a drain page requests with FN
>
> sit, but that is not called in hotplug removal path.

So, in fact so far, it doesn't act as a fence except the 
intel_drain_pasid_prq() ,

and it never handle everthing within its border.


Thanks,

Ethan

>
>
> Thanks,
>
> Ethan
>
>
>
>>
>> Best regards,
>> baolu
>