Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <3f27c84f-f8b0-bddf-cdd5-952c9576c6c3@quicinc.com>
Date:   Tue, 18 Apr 2023 15:51:23 -0600
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:91.0) Gecko/20100101
 Thunderbird/91.6.0
Subject: Re: [PATCH] accel/qaic: initialize ret variable to 0
Content-Language: en-US
To:     Nick Desaulniers <ndesaulniers@google.com>
CC:     Tom Rix <trix@redhat.com>, <ogabbay@kernel.org>,
        <nathan@kernel.org>, <jacek.lawrynowicz@linux.intel.com>,
        <quic_carlv@quicinc.com>, <stanislaw.gruszka@linux.intel.com>,
        <quic_pkanojiy@quicinc.com>, <linux-arm-msm@vger.kernel.org>,
        <dri-devel@lists.freedesktop.org>, <linux-kernel@vger.kernel.org>,
        <llvm@lists.linux.dev>
References: <20230418192046.3235870-1-trix@redhat.com>
 <fe7da2c1-904e-ee4f-9a6c-443049c214b7@quicinc.com>
 <CAKwvOdmFGVPddi98yt22i+U+3aow+dwhxKgdJ45h3n1i-n3bzw@mail.gmail.com>
From:   Jeffrey Hugo <quic_jhugo@quicinc.com>
In-Reply-To: <CAKwvOdmFGVPddi98yt22i+U+3aow+dwhxKgdJ45h3n1i-n3bzw@mail.gmail.com>
Content-Type: text/plain; charset="UTF-8"; format=flowed
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 4/18/2023 2:48 PM, Nick Desaulniers wrote:
> On Tue, Apr 18, 2023 at 1:46 PM Jeffrey Hugo <quic_jhugo@quicinc.com> wrote:
>>
>> On 4/18/2023 1:20 PM, Tom Rix wrote:
>>> clang static analysis reports
>>> drivers/accel/qaic/qaic_data.c:610:2: warning: Undefined or garbage
>>>     value returned to caller [core.uninitialized.UndefReturn]
>>>           return ret;
>>>           ^~~~~~~~~~
>>>
>>> The ret variable is only set some of the time but is always returned.
>>> So initialize ret to 0.
>>
>> This does not appear to be entirely accurate to me.
>>
>> Do you know what analysis Clang is doing?  Is it limited to the function
>> itself?
>>
>> remap_pfn_range, which initializes ret, will always run at-least once.
> 
> What happens if the loop body is never executed, say if `bo->sgt->sgl` is NULL?

qaic_gem_object_mmap() doesn't get called unless a valid GEM object is 
provided by userspace.  For userspace to get a valid GEM object, it has 
to go through qaic_create_bo_ioctl().  qaic_create_bo_ioctl() will not 
return a valid GEM object unless sgt is allocated and initialized.

The loop body will execute at-least once.  The if body will execute 
at-least once.  remap_pfn_range() will run at-least once.  Therefore, 
ret is always initialized.

This is how the code is intended to operate, and how it appears to be 
implemented from what I see.  Unless I'm missing something.

I can see how Clang might not be able to infer these things, but I don't 
believe the code is broken.  I feel that the commit text is unclear and 
suggests that the code is infact, broken.

Userspace should not call mmap() in a critical path thus I don't see a 
true performance concern here.  So while my understanding of the coding 
style is that redundant initialization of variables are to be avoided, I 
think we can say that this is not redundant because it silences a 
warning (because Clang is limited).

>> Feels more accurate to say that Clang cannot detect that ret will be
>> initialized, and we want to quiet the warning from the false positive.
>>
>>> Fixes: ff13be830333 ("accel/qaic: Add datapath")
>>> Signed-off-by: Tom Rix <trix@redhat.com>
>>> ---
>>>    drivers/accel/qaic/qaic_data.c | 2 +-
>>>    1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/accel/qaic/qaic_data.c b/drivers/accel/qaic/qaic_data.c
>>> index c0a574cd1b35..b46a16fb3080 100644
>>> --- a/drivers/accel/qaic/qaic_data.c
>>> +++ b/drivers/accel/qaic/qaic_data.c
>>> @@ -591,7 +591,7 @@ static int qaic_gem_object_mmap(struct drm_gem_object *obj, struct vm_area_struc
>>>        struct qaic_bo *bo = to_qaic_bo(obj);
>>>        unsigned long offset = 0;
>>>        struct scatterlist *sg;
>>> -     int ret;
>>> +     int ret = 0;
>>>
>>>        if (obj->import_attach)
>>>                return -EINVAL;
>>
> 
>