[dpdk-dev] [PATCH v5] eal: out-of-bounds write

Thomas Monjalon thomas.monjalon at 6wind.com
Thu Jul 21 14:54:59 CEST 2016


2016-07-21 12:01, Mrozowicz, SlawomirX:
> Hi Thomas,
> 
> As I understand Sergio suggested to come back to the solution similar to v1.
> Could you comment or better take decision which solution should be applied, please.
> 
> Best Regards,
> Sławomir 
> 
> 
> >-----Original Message-----
> >From: Gonzalez Monroy, Sergio
> >On 20/06/2016 11:09, Thomas Monjalon wrote:
> >> 2016-06-20 10:38, Sergio Gonzalez Monroy:
> >>> On 20/06/2016 10:14, Thomas Monjalon wrote:
> >>>>> +		RTE_LOG(ERR, EAL,
> >>>>> +			"All memory segments exhausted by IVSHMEM. "
> >>>> There is no evidence that it is related to IVSHMEM.
> >>>> "Not enough memory segments." would be more appropriate.
> >>> Actually we would hit this issue when all memsegs have been used by
> >IVSHMEM.
> >>> So I think the message is accurate.
> >> I think it's saner to avoid mixing "potential root cause of a use
> >> case" and "accurate description of the error".
> >> One day, the root cause could be different and the message will become
> >wrong.
> >> Here there is not enough memory segment.
> >>
> >
> >Right.
> >So the whole point of doing the check before the loop was to display the error
> >message with its specific cause.
> >
> >I would think that if the code changes and the message is not accurate then it
> >should also be updated.
> >
> >So if folks prefer a more generic error message probably we don't need the
> >check before the loop and just change the check condition inside the loop that
> >would end up printing the generic error message (after the loop).
> >
> >Basically v1 would do that.
> >http://dpdk.org/dev/patchwork/patch/12241/

At this point of 16.07 we can apply the v1 if you agree.
The message about IVSHMEM will be totally wrong when the ivshmem specific
code will be removed.
If we need more error messages, feel free to send another patch.


More information about the dev mailing list