On Fri, Nov 14, 2025 at 06:54:05PM -0800, Alison Schofield wrote: > On Fri, Oct 31, 2025 at 10:39:59AM -0700, Dave Jiang wrote: > > Add a unit test that verifies the extended linear cache setup paths > > in the kernel driver. cxl_test provides a mock'd version. The test > > verifies the sysfs attribute that indicates extended linear cache support > > is correctly reported. It also verifies the sizing and offset of the > > regions and decoders. > > > > The expecation is that CFMWS covers the entire extended linear cache > > region. The first part is DRAM and second part is CXL memory in a 1:1 > > setup. The start base for hardware decoders should be offsetted by the > > DRAM size. > > > > Signed-off-by: Dave Jiang <[email protected]> > > --- > > test/cxl-elc.sh | 89 ++++++++++++++++++++++++++++++++++++++++++++++++ > > test/meson.build | 2 ++ > > 2 files changed, 91 insertions(+) > > create mode 100755 test/cxl-elc.sh > > snip > > > > The call to this: > > > +find_region() > > +{ > > + json="$($CXL list -b cxl_test -R)" > > + region=$(echo "$json" | jq -r '.[] | > > select(has("extended_linear_cache_size") and .extended_linear_cache_size != > > null) | .region') > > + [[ -n "$region" && "$region" != "null" ]] || do_skip "no test extended > > linear cache region found" > > +} > > + > > is missing from here. Just found while testing w patches supporting the > parameter. So you can fix this and ignore my review comment about > looking up the parameter.
I got that wrong, we do need both checks: Skip if the param is missing Fail if the ELC region is not found since that signals cxl-test breakage > > > +retrieve_info > > +compare_sizes > > +compare_bases > > >
