Re: [PATCH] scripts/ksize: Add kernel build size report

[Darren Hart]

On 11/8/14 19:26, Josh Triplett wrote:
> On Sat, Nov 08, 2014 at 03:18:18PM -0800, Darren Hart wrote:
>> ksize generates hierarchical build size reports from vmlinux, *.o, and
>> built-in.o files.
>>
>> ksize is useful in preparing minimal configurations and comparing
>> similar configurations across kernel versions.
>>
>> Signed-off-by: Darren Hart <dvhart@xxxxxxxxxxxxxxx>
>> Cc: Josh Triplett <josh@xxxxxxxxxxxxxxxx>
> 
> One comment below; with that addressed:
> Reviewed-by: Josh Triplett <josh@xxxxxxxxxxxxxxxx>
> 
>> +def main(argv):
>> +    try:
>> +        opts, args = getopt.getopt(argv[1:], "dh", ["help"])
>> +    except getopt.GetoptError, err:
>> +        print '%s' % str(err)
>> +        usage()
>> +        return 2
>> +
>> +    driver_detail = False
>> +    for o, a in opts:
>> +        if o == '-d':
>> +            driver_detail = True
>> +        elif o in ('-h', '--help'):
>> +            usage()
>> +            return 0
>> +        else:
>> +            assert False, "unhandled option"
>> +
>> +    cols = term_width()
>> +
>> +    # Determine kernel version
>> +    p = Popen("strings vmlinux | grep 'Linux version' | cut -d ' ' -f 3",
>> +              shell=True, stdout=PIPE, stderr=PIPE)
>> +    version = p.communicate()[0].strip()
> 
> This seems like a very fragile, Perl-y way to obtain the kernel version.
> 
> I'd suggest either not including the version (just as bloat-o-meter
> doesn't), or parsing it out using objdump -h -t and file offsetting
> (looking for the offset and size of linux_banner).  Personally I'd go
> with the former.

I agree it's a hack, but it isn't a performance sensitive area and I
couldn't see spending the time to calculate offsets and such to save a
couple seconds for something like this. There are more advanced
approaches to extracting bits of information from vmlinux, such as
scripts/extract-ikconfig. I suppose something along those lines could be
created.

As to fragile, I have not tested every version, but this technique has
been used at least as far back as 2.6.24 [1], which seem fairly stable
to me.

Personally, I think it's nice to have. But, if nobody chimes in asking
to keep it, or with a better way to retrieve it, I can resend with it
removed.

Thanks Josh :-)

1.
http://plastilinux.blogspot.com/2009/11/how-to-know-version-of-kernel-without.html
(A slightly faster hack)

-- 
Darren Hart
Intel Open Source Technology Center
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/