GNATS-1110: in_cksum_hdr error in PC386 BSP
ralf.corsepius at rtems.org
Wed Jun 28 15:01:49 UTC 2006
On Wed, 2006-06-28 at 09:54 -0500, Eric Norum wrote:
> On Jun 28, 2006, at 9:39 AM, Ralf Corsepius wrote:
> > On Wed, 2006-06-28 at 08:47 -0500, Eric Norum wrote:
> > > The checksum routines are very heavily used so it's really
> > > important
> > > that they be as fast as possible. Do you have an estimate of the
> > >
> > > relative times of the old and new versions of this code?
> > heavily used and broken don't align smoothly.
> Yes, I think that they might in this case.
> 1) This problem may have just shown up as a result of the change from
> gcc-3.x to gcc-4.x.
> 2) Joel has reported for some time that qemu has been broken.
Somebody might want to test on real hardware.
> > Either these routines have been heavily used, then they must have
> > been
> > working, or not - Then I don't understand why nobody has tripped
> > over
> > this issue in the years this code is in place.
> Because in previous years they were using a different tool chain.
It's always a comfortable excuse to accuse the toolchain :-)
Did you check the asm being generated?
> > IMO, the PR is way to vague to give any trust in it.
> Others have reported problems with networking on the pc386.
This doesn't mean much, esp. because most networking on little endian
targets always is a magnitude difficult than on big endian ones.
> This is the first time I've seen a patch though.
I presume, explicitly testing this checksum code on real hardware with
somebody being familiar to i386s isn't much of a problem.
More information about the users