[Pcsclite-muscle] Possible generation of duplicate SCARDHANDLE

Ludovic Rousseau ludovic.rousseau
Thu Aug 4 04:12:13 PDT 2016


2016-08-02 19:59 GMT+02:00 Maksim Ivanov <emaxx at google.com>:

> Ludovic,
>
> Thanks for the quick update.
>
> Yes, looks like the suggested patch fixes the problem. Great that the
> resulting patch is so small.
>

Thanks for the feedback.
It is now fixed in
https://github.com/LudovicRousseau/PCSC/commit/bb07a905c896e7a67e3d7b37a24b3637ada0546d

I am not sure to understand your concern about srand (point 3).
>> I agree the code is not thread safe. So srand could be called twice. But
>> that should not be an issue. srand would be called at least once in all
>> cases. Or I am missing something?
>
>
> The effect of calling srand twice is the possible overlaps of the obtained
> pseudo-random numbers. Also, srand is not guaranteed to work correctly at
> all when used from multiple threads simultaneously.
>
> Generally, why not just call srand once somewhere around the daemon's
> entry point (int main)? Then there would be no questions regarding which
> component and how should perform this initialization.
>

Good idea.
Fixed in
https://github.com/LudovicRousseau/PCSC/commit/8b80aa4900cd60ab075802bdcc5a996027d0c74e
and
https://github.com/LudovicRousseau/PCSC/commit/8e820796b338cb1048c51ec462446f61a3979835


I also note that the number of different contexts is limited to RAND_MAX.
On my Debian GNU/Linux system I have RAND_MAX = 2147483647 so that should
be enough but on some other systems may it be more limited (like 32767).

I opened a new feature request so I can remember to work on this (when time
and motivation permits)
https://alioth.debian.org/tracker/index.php?func=detail&aid=315434&group_id=30105&atid=410088

Bye

-- 
 Dr. Ludovic Rousseau
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.alioth.debian.org/pipermail/pcsclite-muscle/attachments/20160804/b03392dc/attachment.html>



More information about the pcsclite-muscle mailing list