linux-kernel - Re: [PATCH 4/4] remoteproc: qcom_q6v5_mss: Use a carveout to authenticate modem headers

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:   Thu, 15 Dec 2022 09:51:28 +0100
From:   Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>
To:     Sibi Sankar <quic_sibis@...cinc.com>, andersson@...nel.org,
        krzysztof.kozlowski+dt@...aro.org, robh+dt@...nel.org,
        manivannan.sadhasivam@...aro.org
Cc:     agross@...nel.org, linux-arm-msm@...r.kernel.org,
        devicetree@...r.kernel.org, linux-kernel@...r.kernel.org,
        konrad.dybcio@...ainline.org, amit.pundir@...aro.org,
        regressions@...mhuis.info, sumit.semwal@...aro.org,
        will@...nel.org, catalin.marinas@....com, robin.murphy@....com
Subject: Re: [PATCH 4/4] remoteproc: qcom_q6v5_mss: Use a carveout to
 authenticate modem headers

On 14/12/2022 12:51, Sibi Sankar wrote:
> 
> 
> On 12/14/22 16:58, Krzysztof Kozlowski wrote:
>> On 14/12/2022 11:33, Sibi Sankar wrote:
>>>
>>>
>>> On 12/14/22 01:17, Krzysztof Kozlowski wrote:
>>>> On 13/12/2022 15:07, Sibi Sankar wrote:
>>>>> The memory region allocated using dma_alloc_attr with no kernel mapping
>>>>> attribute set would still be a part of the linear kernel map. Any access
>>>>> to this region by the application processor after assigning it to the
>>>>> remote Q6 will result in a XPU violation. Fix this by replacing the
>>>>> dynamically allocated memory region with a no-map carveout and unmap the
>>>>> modem metadata memory region before passing control to the remote Q6.
>>>>>
>>>>> Reported-by: Amit Pundir <amit.pundir@...aro.org>
>>>>> Fixes: 6c5a9dc2481b ("remoteproc: qcom: Make secure world call for mem ownership switch")
>>>>> Signed-off-by: Sibi Sankar <quic_sibis@...cinc.com>
>>>>> ---
>>>>
>>>> Thank you for your patch. There is something to discuss/improve.
>>>>>    
>>>>>    	return ret < 0 ? ret : 0;
>>>>> @@ -1882,6 +1899,26 @@ static int q6v5_alloc_memory_region(struct q6v5 *qproc)
>>>>>    	qproc->mpss_phys = qproc->mpss_reloc = r.start;
>>>>>    	qproc->mpss_size = resource_size(&r);
>>>>>    
>>>>> +	if (!child) {
>>>>> +		node = of_parse_phandle(qproc->dev->of_node, "memory-region", 2);
>>>>> +	} else {
>>>>> +		child = of_get_child_by_name(qproc->dev->of_node, "metadata");
>>>>
>>>> Bindings do not allow to have child "metadata", do they?
>>>
>>> memory-region property was used to specify mba/mpss region in a phandle
>>> array only from SC7180 SoC. All the older dtbs in the wild/upstream
>>> still had sub-nodes to achieve the same. Patch 3 allows for a sub-set
>>> of the SoCs (MSM8996/MSM8998/SDM845) to use metadata as a sub-node so
>>> as to not break bindings when newer kernel uses a older dtb.
>>
>> This does not explain why you extend the driver without extending the
>> bindings. You do not do it for legacy stuff but for SC7180. But even for
>> legacy devices you cannot add new properties without having it in some
>> legacy bindings.
> 
> https://patchwork.kernel.org/project/linux-arm-msm/patch/20221213140724.8612-4-quic_sibis@quicinc.com/
> 
> The legacy bindings are a part of patch 3 ^^.

Ah, ok.

Best regards,
Krzysztof