linux-kernel - Re: [PATCH 03/11] firmware: qcom-scm: atomically assign and read the global _

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: <36dc8000-6baf-5af0-1921-8ed509796ed0@linaro.org>
Date:   Tue, 29 Aug 2023 14:48:41 +0200
From:   Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>
To:     Bartosz Golaszewski <bartosz.golaszewski@...aro.org>
Cc:     Andy Gross <agross@...nel.org>,
        Bjorn Andersson <andersson@...nel.org>,
        Konrad Dybcio <konrad.dybcio@...aro.org>,
        Rob Herring <robh+dt@...nel.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
        Conor Dooley <conor+dt@...nel.org>,
        Catalin Marinas <catalin.marinas@....com>,
        Will Deacon <will@...nel.org>, Arnd Bergmann <arnd@...db.de>,
        Alex Elder <elder@...aro.org>,
        Srini Kandagatla <srinivas.kandagatla@...aro.org>,
        kernel@...cinc.com, linux-arm-msm@...r.kernel.org,
        devicetree@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org
Subject: Re: [PATCH 03/11] firmware: qcom-scm: atomically assign and read the
 global __scm pointer

On 29/08/2023 14:31, Bartosz Golaszewski wrote:
> On Tue, 29 Aug 2023 at 09:59, Krzysztof Kozlowski
> <krzysztof.kozlowski@...aro.org> wrote:
>>
>> On 28/08/2023 21:24, Bartosz Golaszewski wrote:
>>> Checking for the availability of SCM bridge can happen from any context.
>>> It's only by chance that we haven't run into concurrency issues but with
>>> the upcoming SHM Bridge driver that will be initiated at the same
>>> initcall level, we need to assure the assignment and readback of the
>>> __scm pointer is atomic.
>>>
>>> Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@...aro.org>
>>> ---
>>>  drivers/firmware/qcom_scm.c | 6 +++---
>>>  1 file changed, 3 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/drivers/firmware/qcom_scm.c b/drivers/firmware/qcom_scm.c
>>> index 980fcfa20b9f..422de70faff8 100644
>>> --- a/drivers/firmware/qcom_scm.c
>>> +++ b/drivers/firmware/qcom_scm.c
>>> @@ -1331,7 +1331,7 @@ static int qcom_scm_find_dload_address(struct device *dev, u64 *addr)
>>>   */
>>>  bool qcom_scm_is_available(void)
>>>  {
>>> -     return !!__scm;
>>> +     return !!READ_ONCE(__scm);
>>>  }
>>>  EXPORT_SYMBOL(qcom_scm_is_available);
>>>
>>> @@ -1477,8 +1477,8 @@ static int qcom_scm_probe(struct platform_device *pdev)
>>>       if (ret)
>>>               return ret;
>>>
>>> -     __scm = scm;
>>> -     __scm->dev = &pdev->dev;
>>> +     scm->dev = &pdev->dev;
>>> +     WRITE_ONCE(__scm, scm);
>>
>> Your re-ordering is not effective here, I think. I don't understand it's
>> purpose exactly, but scm->dev assignment is not WRITE_ONCE(), thus it
>> can be reordered:
>>
>> "compiler is also forbidden from reordering successive instances of
>> READ_ONCE and WRITE_ONCE" <- so compiler is not forbidden to reorder
>> other stuff.
>>
>> "Ensuring that the compiler does not fold, spindle, or otherwise
>> mutilate accesses that either do not require ordering or that interact"
>> <- which means you do not require ordering here.
>>
> 
> Hmm, I had the list_add() implementation in mind as well as examples
> from https://www.kernel.org/doc/Documentation/memory-barriers.txt and
> was under the impression that WRITE_ONCE() here is enough. I need to
> double check it.

It all depends what do you want to achieve. If strict ordering of both
writes, then all the examples and descriptions from memory-barriers.txt
say that you need WRITE_ONCE in both places.

If you want to achieve something else, like scm->dev visible for other
CPUs before __scm=scm then you would need full barriers, IMHO.


Best regards,
Krzysztof