[PATCH kernel 1/1] powerpc/Documentation/misc-devices/: Fix one compile error

[Feng Gao]

On Wed, Feb 22, 2017 at 5:09 PM, Winkler, Tomas <tomas.winkler at intel.com> wrote:
>
>>
>> On Wed, Feb 22, 2017 at 4:14 PM, Winkler, Tomas <tomas.winkler at intel.com>
>> wrote:
>> > On Wed, 2017-02-22 at 15:22 +0800, fgao at ikuai8.com wrote:
>> >> From: Gao Feng <fgao at ikuai8.com>
>> >>
>> >> When make allyesconfig, there is one compile error on my platform
>> >> "gcc (Ubuntu 4.8.4-2ubuntu1~14.04.3) 4.8.4".
>> >> The following is the output.
>> >>
>> >> Documentation/misc-devices/mei/mei-amt-version.c: In function ‘main’:
>> >> Documentation/misc-devices/mei/mei-amt-version.c:103:5: warning:
>> >> ‘acmd.fd’
>> >> is used uninitialized in this function [-Wuninitialized]
>> >>   if (cl->fd != -1)
>> >>      ^
>> >> Documentation/misc-devices/mei/mei-amt-version.c:443:21: note:
>> >> ‘acmd.fd’
>> >> was declared here
>> >>   struct amt_host_if acmd;
>> >>                      ^
>> >> This commit fixes this compile error.
>> >>
>> >> Signed-off-by: Gao Feng <fgao at ikuai8.com>
>> >
>> > This is false positive, as the variable is assined in mei_init(), in
>> > any case, the code has moved under samples diretory in the current
>> > kernel. Anything need to be fixed there first
>> >
>> > Thanks
>> > Tomas
>>
>> I think it is not false positive.
>> The function stack is main->amt_host_if_init->mei_init->mei_deinit.
>> There is one check "if (cl->fd != -1)".
>
> Isn't  the first line in mei_init me->fd = open("/dev/mei", O_RDWR); ?
> I don't see mei_deinit called w/o mei_init called first.

The codes of powerpc.git(git://git.kernel.org/pub/scm/linux/kernel/git/galak/powerpc.git)
are different with net.git.
The following are the codes from powerpc.git

static bool mei_init(struct mei *me, const uuid_le *guid,
                unsigned char req_protocol_version, bool verbose)
{
        int result;
        struct mei_client *cl;
        struct mei_connect_client_data data;

        mei_deinit(me);

        me->verbose = verbose;

        me->fd = open("/dev/mei", O_RDWR);
        if (me->fd == -1) {
                mei_err(me, "Cannot establish a handle to the Intel
MEI driver\n");
                goto err;
        }

The mei_deinit is invoked before open, so the gcc reports one warning.

Regards
Feng


>
>> Because this fd is not initialized to -1, so it may hit the condition, and execute
>> close.
>>
>> So it should be fixed, although these codes would be moved to sample
>> director.
>
> Please try with the master branch first in case  I'm still wrong.
> Thanks
>
>> Regards
>> Feng
>>
>> >
>> >> ---
>> >>  Documentation/misc-devices/mei/mei-amt-version.c | 1 +
>> >>  1 file changed, 1 insertion(+)
>> >>
>> >> diff --git a/Documentation/misc-devices/mei/mei-amt-version.c
>> >> b/Documentation/misc-devices/mei/mei-amt-version.c
>> >> index 49e4f77..ca035cb 100644
>> >> --- a/Documentation/misc-devices/mei/mei-amt-version.c
>> >> +++ b/Documentation/misc-devices/mei/mei-amt-version.c
>> >> @@ -297,6 +297,7 @@ static bool amt_host_if_init(struct amt_host_if
>> >> *acmd,
>> >>                     unsigned long send_timeout, bool verbose)  {
>> >>       acmd->send_timeout = (send_timeout) ? send_timeout : 20000;
>> >> +     acmd->mei_cl.fd = -1;
>> >>       acmd->initialized = mei_init(&acmd->mei_cl, &MEI_IAMTHIF, 0,
>> >> verbose);
>> >>       return acmd->initialized;
>> >>  }