[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [bluetooth-dev] patch: more big endian changes
Le Lundi 12 Mars 2001 15:46, Matthias Fuchs a écrit :
> > Hi Everybody,
> here are the rest of the big endian changes for openbt stack. The
> current cvs version really does not support big endian CPUs. The only BE
> support that is available was indented for one of the older releases
> before the cvs time.
> In the last days gordon put big endian support for the RFCOMM layer into
> the cvs and though that the lower layers were already corrected. They
> were not, but now they are :-)
> I took Gordons old patches and applied them by hand to the current cvs
> version. I also moved some conversion macros from rfcomm.c to local.h,
> where I think they belong to.
> I tested the big endian stuff with real (but old) hardware between a
> Linux PC (kernel 2.4.0) and our embedded Linux custom board with an
> IBM405CR CPU (PowerPC). I could do inquiries, rfcomm connections and ppp
> over rfcomm with some pingin' over it ! No problems ... nearly no
> problems ... After some ftp over the link I got error from
> send_acl_packet saying 'negativ cur_len -135'. But I think this is
> caused by my faulty ROK101007 firmware, because after the message the
> HCI fsm gets out of sync ... (those 'discarding data ... waiting for
> ever' messages). Are there some other possible reasons ?
It could also be a race condition beetween the "receive" task and the "send"
task allocating/deallocating buffers when an ncp packet is received (it
happened to us in user mode, and we temporarily fixed it with a mutex). But
that shouldn't happen in kernel mode (btw, was your test in kernel mode ?). I
have not been able to see this problem yet on my (little endian) system...
> I also did some changes on apps/Rules.elinux and
> linux/drivers/char/bluetooth/Makefile for cross compiling for my
> powerpc-linux system (I am using hardhat linux). You can now do a
> make INCLUDEDIR=/usr/src/hardhatlinux/include
> (if your kernel includes for your target are under that path and if your
> crossdev tools are named like 'powerpc-linux-gcc' etc.
> I hope that the endianess problem are fixed by now. SDP still has to be
> checked !!!
> Can anybody check this in ?
Content-Type: text/plain; charset="us-ascii";
To unsubscribe from this list: send the line "unsubscribe bluetooth-dev" in
the body of a message to firstname.lastname@example.org