On Fri, 24 Jan 2025 17:28:02 +0100 Philippe Mathieu-Daudé <[email protected]> wrote:
> On 24/1/25 17:20, Jonathan Cameron wrote: > > On Thu, 23 Jan 2025 09:51:51 +0000 > > Jonathan Cameron via <[email protected]> wrote: > > > >> On Wed, 22 Jan 2025 07:56:24 +0100 > >> Philippe Mathieu-Daudé <[email protected]> wrote: > >> > >>> Map the component I/O region as UnimplementedDevice > >>> to be able to trace guest I/O accesses with '-d unimp'. > >>> > >>> Signed-off-by: Philippe Mathieu-Daudé <[email protected]> > >> I'm not that familiar with this infrastructure but seems > >> fine to me. > >> > >> I'd definitely be curious if anything is touching this space so > >> tracing may be helpful for that! > > Hi Philippe > > > >> > >> Reviewed-by: Jonathan Cameron <[email protected]> > > Retract that. > > > > Can't instantiate a CXL device with this patch because: > > > > Device unimplemented-device can not be dynamically instantiated. > > Hmm the qtests using CXL devices pass, how do you trigger that? Seems oddly enough to be arm64 only (well not x86 - there is no support for other arches than those two) which isn't upstream yet hence no tests upstream - I'll get back to that sometime soonish as the blocker of dt-bindings for PXBs may be less of an issue than it was. source of that print seems to be hw/core/sysbus-fdt.c add_fdt_node(). Seems it is looking for sysbus devices and gets confused when it finds this one. I'm not sure if it is in /peripheral or /peripheral-anon. For now you could try my gitlab tree gitlab.com/jic23/qemu.git cxl-* whatever has latest date. Or shout if you want me to try anything. Jonathan > > > > > Reverting this patch on my tree fixes that. > > > >>> --- > >>> include/hw/cxl/cxl_component.h | 3 ++- > >>> hw/cxl/cxl-component-utils.c | 14 +++++++++++--- > >>> hw/cxl/Kconfig | 1 + > >>> 3 files changed, 14 insertions(+), 4 deletions(-) >
