Re: [PATCH v7] cxl/region: check interleave capability

[Dan Williams <dan.j.williams@intel.com>]

Alison Schofield wrote:
[..]
> > > diff --git a/drivers/cxl/cxlmem.h b/drivers/cxl/cxlmem.h
> > > index 36cee9c30ceb..7fe617122d33 100644
> > > --- a/drivers/cxl/cxlmem.h
> > > +++ b/drivers/cxl/cxlmem.h
> > > @@ -848,11 +848,21 @@ static inline void cxl_mem_active_dec(void)
> > >  
> > >  int cxl_mem_sanitize(struct cxl_memdev *cxlmd, u16 cmd);
> > >  
> > > +/*
> > 
> > Minor detail that can come in a follow-on patch is that this needs to
> > be:
> > 
> > /**
> > 
> > ...in order for the kernel-doc system to autoformat it like it does
> > other 'struct' documentation:
> > 
> > https://docs.kernel.org/driver-api/cxl/memory-devices.html
> > 
> > However, the reason it needs to be a follow-on patch is that this file
> > is not currently included for parsing and needs something like this:
> 
> We'll need another revision of this patch to address the cxl-test module

Thanks for running that by the way!

Now, I tend to not want to leave people hanging figuring out how
cxl_test works, so in this case I think the incremental fix on top of
the crash fix is this:

diff --git a/tools/testing/cxl/test/cxl.c b/tools/testing/cxl/test/cxl.c
index 908e0d083936..4b2416d9f66f 100644
--- a/tools/testing/cxl/test/cxl.c
+++ b/tools/testing/cxl/test/cxl.c
@@ -635,6 +635,9 @@ static struct cxl_hdm *mock_cxl_setup_hdm(struct cxl_port *port,
                return ERR_PTR(-ENOMEM);
 
        cxlhdm->port = port;
+       cxlhdm->interleave_mask = ~0U;
+       cxlhdm->iw_cap_mask = ~0UL;
+
        return cxlhdm;
 }
 
I.e. allow everything for now.

The good news is the patch does work to reject decoders without the
proper capabilities.

> dependencies so how about adding the "/**" in the next revision.
> 
> I'm suggesting this because cxl_mem.h contains other kernel doc
> comments that are not being picked up because cxl_mem.h is missing
> in Documentation/driver-api/cxl/memory-devices.rst. There also seem
> to be other ommissions when compared with the kernel doc notations
> in drivers/cxl/ :  core/cdat.c, core/hdm.c. I say 'seems' because
> I guess it could be intentional.
> 
> Can Yao add the kernel doc notation in the next rev of this patch
> and then come back soon and sync memory-devices.rst for all of
> drivers/cxl/ ?

Sure, but any review issues on the second patch should not hold up
merging the first since they're unrelated changes.