linux-kernel - Re: [PATCH] media: ov7670: check status of ov7670

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: <db41a471-b5cd-6ffe-c24f-ba14d9125920@redhat.com>
Date:   Fri, 28 Aug 2020 08:36:43 -0700
From:   Tom Rix <trix@...hat.com>
To:     Nick Desaulniers <ndesaulniers@...gle.com>
Cc:     Jonathan Corbet <corbet@....net>,
        Mauro Carvalho Chehab <mchehab@...nel.org>,
        Nathan Chancellor <natechancellor@...il.com>,
        linux-media@...r.kernel.org, LKML <linux-kernel@...r.kernel.org>,
        Masahiro Yamada <masahiroy@...nel.org>,
        Nathan Huckleberry <nhuck15@...il.com>
Subject: Re: [PATCH] media: ov7670: check status of ov7670_read


On 8/28/20 8:13 AM, Nick Desaulniers wrote:
> On Fri, Aug 28, 2020 at 7:55 AM <trix@...hat.com> wrote:
>> From: Tom Rix <trix@...hat.com>
>>
>> clang static analysis flags this representative problem
>>
>> drivers/media/i2c/ov7670.c:1463:9: warning: Assigned
>>   value is garbage or undefined
>>         *value = gain;
>>                ^ ~~~~
>>
>> gain is set by a successful call to ov7670_read()
> Indeed, it looks like gain is only valid if the return value from
> ov7670_read() is >= 0.  Would it be simpler to just initialize gain to
> 0 in ov7670_g_gain?

It looks like ov7670_read() subroutines both convert > 0 returns to 0.

Setting gain, and similar cases, to 0 means it is ok the setting side effect.  maybe it would be ok if this was a void return or there was some known poison value.  I did not look into that.  There are other routines that are more than a simple get() and i wanted to use the same fix for all of them.

yeah on clang static analysis making it into kbuild!  I have been doing these fixes for a couple of months.  with a just-built clang, i run 'scan-build make randomconfig', 'scan-build make', then run scan-view on the html.  the randomconfig accounts for the very scattered fixes i have posted.  generally i would say kernel is doing pretty well, maybe 20 to 1 false to real problems.

Tom

>
> Side question; I'm super happy to see someone sending patches for
> things identified by clang's static analyzer.  I'm curious, Tom, how
> did you run it?  Did you use the recently landed in kbuild-next
> support for the make target `make clang-analyzer`?
>
>> So check that ov7670_read() is successful.
>>
>> The remaining static analysis problems are false positives.
>> There appears to be a limitation with checking the
>> aggregated returns.
>>
>> Signed-off-by: Tom Rix <trix@...hat.com>
>> ---
>>  drivers/media/i2c/ov7670.c | 17 +++++++++++++----
>>  1 file changed, 13 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/media/i2c/ov7670.c b/drivers/media/i2c/ov7670.c
>> index b42b289faaef..001d4b09db72 100644
>> --- a/drivers/media/i2c/ov7670.c
>> +++ b/drivers/media/i2c/ov7670.c
>> @@ -929,6 +929,8 @@ static int ov7670_set_hw(struct v4l2_subdev *sd, int hstart, int hstop,
>>         ret =  ov7670_write(sd, REG_HSTART, (hstart >> 3) & 0xff);
>>         ret += ov7670_write(sd, REG_HSTOP, (hstop >> 3) & 0xff);
>>         ret += ov7670_read(sd, REG_HREF, &v);
>> +       if (ret)
>> +               return ret;
>>         v = (v & 0xc0) | ((hstop & 0x7) << 3) | (hstart & 0x7);
>>         msleep(10);
>>         ret += ov7670_write(sd, REG_HREF, v);
>> @@ -938,6 +940,8 @@ static int ov7670_set_hw(struct v4l2_subdev *sd, int hstart, int hstop,
>>         ret += ov7670_write(sd, REG_VSTART, (vstart >> 2) & 0xff);
>>         ret += ov7670_write(sd, REG_VSTOP, (vstop >> 2) & 0xff);
>>         ret += ov7670_read(sd, REG_VREF, &v);
>> +       if (ret)
>> +               return ret;
>>         v = (v & 0xf0) | ((vstop & 0x3) << 2) | (vstart & 0x3);
>>         msleep(10);
>>         ret += ov7670_write(sd, REG_VREF, v);
>> @@ -1460,6 +1464,8 @@ static int ov7670_g_gain(struct v4l2_subdev *sd, __s32 *value)
>>         unsigned char gain;
>>
>>         ret = ov7670_read(sd, REG_GAIN, &gain);
>> +       if (ret)
>> +               return ret;
>>         *value = gain;
>>         return ret;
>>  }
>> @@ -1470,11 +1476,14 @@ static int ov7670_s_gain(struct v4l2_subdev *sd, int value)
>>         unsigned char com8;
>>
>>         ret = ov7670_write(sd, REG_GAIN, value & 0xff);
>> +       if (ret)
>> +               return ret;
>>         /* Have to turn off AGC as well */
>> -       if (ret == 0) {
>> -               ret = ov7670_read(sd, REG_COM8, &com8);
>> -               ret = ov7670_write(sd, REG_COM8, com8 & ~COM8_AGC);
>> -       }
>> +       ret = ov7670_read(sd, REG_COM8, &com8);
>> +       if (ret)
>> +               return ret;
>> +       ret = ov7670_write(sd, REG_COM8, com8 & ~COM8_AGC);
>> +
>>         return ret;
>>  }
>>
>> --
>> 2.18.1
>>
>