This is the mail archive of the ecos-patches@sourceware.org mailing list for the eCos project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [ECOS] Re: Improving TFTP performance


On Thu, Dec 27, 2007 at 03:02:03PM +0100, ?yvind Harboe wrote:
> On Dec 21, 2007 5:10 PM, Jonathan Larmour <jifl@ecoscentric.com> wrote:
> > ?yvind Harboe wrote:
> > > Switched to using memcpy() for tftp get, which matters w/larger packet sizes.
> >
> > Thanks.
> >
> > It would be good if you could tweak a few things: firstly, the use of
> > sprintf means the CDL will need to require CYGINT_ISO_STDIO_FORMATTED_IO.
> 
> Fixed.
> 
> >
> > Secondly, and I realise the problem was there before as well, but I am
> > concerned that there are no boundary checks on overrunning the end of
> > 'data', despite the variable length of filename (which could well originate
> > from user input, so it's up to us to enforce our own size limits). I think
> > this should be checked. Would you be able to do that?
> 
> That will take some work + invalidate all the testing I've done.
> That's for another patch.
> 
> > Finally, Since tftp_client_put is nearly identical in this respect, can you
> > make a similar change to it? Seems strange to do one and not the other.
> 
> I think that makes sense after any followup patches to tftp get are in place.

I've committed this patch. I also tested it against a tftpd which does
not support this option and it did not break.

    Andrew


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]