On Tue, 17 Jun 2025 18:09:37 +0530 Neeraj Kumar <[email protected]> wrote:
> In 84ec985944ef3, For cxl pmem region auto-assembly after endpoint port > probing, cxl_nvd presence was required. And for cxl region persistency, > region creation happens during nvdimm_probe which need the completion > of endpoint probe. > > It is therefore refactored cxl pmem region auto-assembly after endpoint > probing to cxl mem probing > > Signed-off-by: Neeraj Kumar <[email protected]> For ordering requirements this needs more eyes. I've never cared that much about he persistency and auto assembly code so not something I have a good mental model of! > --- > drivers/cxl/core/port.c | 38 ++++++++++++++++++++++++++++++++++++++ > drivers/cxl/cxl.h | 1 + > drivers/cxl/mem.c | 27 ++++++++++++++++++--------- > drivers/cxl/port.c | 38 -------------------------------------- > 4 files changed, 57 insertions(+), 47 deletions(-) > diff --git a/drivers/cxl/mem.c b/drivers/cxl/mem.c > index 2f03a4d5606e..aaea4eb178ef 100644 > --- a/drivers/cxl/mem.c > +++ b/drivers/cxl/mem.c > @@ -180,6 +171,24 @@ static int cxl_mem_probe(struct device *dev) > return rc; > } > > + if (resource_size(&cxlds->pmem_res) && IS_ENABLED(CONFIG_CXL_PMEM)) { > + rc = devm_cxl_add_nvdimm(parent_port, cxlmd); > + if (rc) { > + if (rc == -ENODEV) > + dev_info(dev, "PMEM disabled by platform\n"); > + return rc; > + } > + } > + > + /* > + * Now that all endpoint decoders are successfully enumerated, try to > + * assemble region autodiscovery from committed decoders. > + * Earlier it was part of cxl_endpoint_port_probe, So moved it here I would drop this history statement. Good to have in the patch description but no point in keeping it in the code. Just state what the requirements are now. > + * as cxl_nvd of the memdev needs to be available during the pmem > + * region auto-assembling > + */ > + cxl_region_discovery(cxlmd->endpoint); > + > /* > * The kernel may be operating out of CXL memory on this device,
