RFC: FPA support removal from gdb and its impact on kgdb

[Russell King (Oracle)]

On Thu, Oct 06, 2022 at 02:42:09PM +0100, Luis Machado wrote:
> Hi Russel,
> 
> On 10/5/22 17:54, Russell King (Oracle) wrote:
> > Given this, and this is a fundamental interface between two different
> > pieces of software, I fail to see how you can even consider removing
> > support for this - if you remove support from gdb, then later gdb
> > will be unable to debug existing kernels.
> 
> It is being considered because not even the Linux kernel uses these
> registers anymore. The kgdb implementation was based on what GDB exposed
> long ago.

The kernel has never used the FPA not even an emulation of the FPA. The
kernel has provided a number of FPA emulation options over the years for
userspace to use, but never for its own use.

So kgdb's inclusion if these is merely to make kgdb's stub work with
gdb and nothing else.

> > In kernel-land, we have a rule: don't break userspace. I think there
> > should also be a rule for userspace: don't break the kernel!
> 
> That's fair, and I'm not seeking to proactively break the kernel. That's
> why I listed some possible solutions.
> 
> The best one is to update kgdb to send back XML data and drop the fp
> registers. On GDB's side, we can carry compatibility code that will
> guarantee things will work until it makes sense to remove the
> compatibility layer altogether.
> 
> Does that sound like a good compromise?

Yes, thanks.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!