Re: [PATCH v2 08/29] ktap: add bytecode reader(kernel/trace/ktap/kp_bcread.[c|h])

[Jovi Zhangwei]

On Sun, Mar 30, 2014 at 10:47 AM, Andi Kleen <andi@xxxxxxxxxxxxxx> wrote:
>> +/* Read debug info of a prototype. */
>> +static void bcread_dbg(BCReadCtx *ctx, ktap_proto_t *pt, int sizedbg)
>> +{
>> +     void *lineinfo = (void *)proto_lineinfo(pt);
>> +
>> +     bcread_block(ctx, lineinfo, sizedbg);
>> +     /* Swap lineinfo if the endianess differs. */
>
>
> Why does this care about endianness? Can't that be handled in the user
> space? And why would the user space create different endianness than
> the host is?
>
That's designed for portability initially, it means we can just run bytecode
without compile script file everywhere, especially when compilation  is
a heavily task for some embedded platform, even though ktap compilation
is extremely fast.

I doubt maybe there will have this bytecode portability requirement in future?

>> +     for (i = 0; i < sizekgc; i++, kr++) {
>> +             int tp = bcread_uint32(ctx);
>> +             if (tp >= BCDUMP_KGC_STR) {
>
> The signedness handling all over this file is a scary.
> What happens if the user puts in negative values or near overflow
> values.
>
> Most likely a lot of these checks should be unsigned
> and need to be audited again (and ideally fuzzed too)
>
>> +
>> +     /* Allocate prototype object and initialize its fields. */
>> +     pt = (ktap_proto_t *)kp_obj_new(ctx->ks, (int)sizept);
>
> Error check?
>
> Lots of other similar cases.
>
I will take more check in kp_bcread.c file, will fix in next version.

Thanks.

Jovi
--
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/