[PATCH v3] cxl/test: Enforce PMD alignment for volatile mock regions

[PATCH v3] cxl/test: Enforce PMD alignment for volatile mock regions

[Richard Cheng]

cxl_test allocates synthetic CFMWS HPA windows from a gen_pool with
SZ_256M alignment. On arm64 with CONFIG_ARM64_64K_PAGES=y and
CONFIG_PGTABLE_LEVELS=3, PMD_SIZE is 512M, so every CXL region carved
from a volatile window inherits a non-PMD-aligned start, and
cxl_dax_region_probe() -> alloc_dax_region() fails:

"""
    cxl_dax_region dax_region1: probe with driver cxl_dax_region failed
                                with error -12
"""

Enforce that every volatile mock CFMWS is PMD-aligned in both start and
size

Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Acked-by: Kai-Heng Feng <kaihengf@nvidia.com>
Signed-off-by: Richard Cheng <icheng@nvidia.com>
---
Changelog:

v2->v3:
    - Correct hunk headers
    - Add base commit SHA

v1->v2:
    - Reframe from "fix start alignment" to "enforce PMD invariant for
      volatile CFMWS"
    - Bump cfmws5.window_size to max(SZ_256M, PMD_SIZE)
    - Gate populate_cedt() alignment on
      ACPI_CEDT_CFMWS_RESTRICT_VOLATILE so PMEM windows stay at SZ_256M
    - Add BUILD_BUG_ON for MOCK_AUTO_REGION_SIZE_DEFAULT v.s. PMD_SIZE
      in cxl_test_init()

Best regards,
Richard Cheng.
---
 tools/testing/cxl/test/cxl.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/tools/testing/cxl/test/cxl.c b/tools/testing/cxl/test/cxl.c
index 418669927fb0..7a1105f9a0bb 100644
--- a/tools/testing/cxl/test/cxl.c
+++ b/tools/testing/cxl/test/cxl.c
@@ -318,7 +318,7 @@ static struct {
 			.restrictions = ACPI_CEDT_CFMWS_RESTRICT_HOSTONLYMEM |
 					ACPI_CEDT_CFMWS_RESTRICT_VOLATILE,
 			.qtg_id = FAKE_QTG_ID,
-			.window_size = SZ_256M,
+			.window_size = SZ_256M > PMD_SIZE ? SZ_256M : PMD_SIZE,
 		},
 		.target = { 3 },
 	},
@@ -495,9 +495,12 @@ static int populate_cedt(void)
 
 	for (i = cfmws_start; i <= cfmws_end; i++) {
 		struct acpi_cedt_cfmws *window = mock_cfmws[i];
+		int align = SZ_256M;
 
 		cfmws_elc_update(window, i);
-		res = alloc_mock_res(window->window_size, SZ_256M);
+		if (window->restrictions & ACPI_CEDT_CFMWS_RESTRICT_VOLATILE)
+			align = max_t(int, SZ_256M, PMD_SIZE);
+		res = alloc_mock_res(window->window_size, align);
 		if (!res)
 			return -ENOMEM;
 		window->base_hpa = res->range.start;
@@ -1826,6 +1829,8 @@ static __init int cxl_test_init(void)
 	int rc, i;
 	struct range mappable;
 
+	BUILD_BUG_ON(!IS_ALIGNED(MOCK_AUTO_REGION_SIZE_DEFAULT, PMD_SIZE));
+
 	cxl_acpi_test();
 	cxl_core_test();
 	cxl_mem_test();

base-commit: 6779b50faa562e6cca1aa6a4649a4d764c6c7e28
-- 
2.43.0

Re: [PATCH v3] cxl/test: Enforce PMD alignment for volatile mock regions

[Alison Schofield]

On Fri, May 22, 2026 at 01:44:57PM +0800, Richard Cheng wrote:
> cxl_test allocates synthetic CFMWS HPA windows from a gen_pool with
> SZ_256M alignment. On arm64 with CONFIG_ARM64_64K_PAGES=y and
> CONFIG_PGTABLE_LEVELS=3, PMD_SIZE is 512M, so every CXL region carved
> from a volatile window inherits a non-PMD-aligned start, and
> cxl_dax_region_probe() -> alloc_dax_region() fails:
> 
> """
>     cxl_dax_region dax_region1: probe with driver cxl_dax_region failed
>                                 with error -12
> """
> 
> Enforce that every volatile mock CFMWS is PMD-aligned in both start and
> size
> 
> Reviewed-by: Dave Jiang <dave.jiang@intel.com>
> Acked-by: Kai-Heng Feng <kaihengf@nvidia.com>
> Signed-off-by: Richard Cheng <icheng@nvidia.com>
> ---
> Changelog:
> 
> v2->v3:
>     - Correct hunk headers
>     - Add base commit SHA
> 
> v1->v2:
>     - Reframe from "fix start alignment" to "enforce PMD invariant for
>       volatile CFMWS"
>     - Bump cfmws5.window_size to max(SZ_256M, PMD_SIZE)
>     - Gate populate_cedt() alignment on
>       ACPI_CEDT_CFMWS_RESTRICT_VOLATILE so PMEM windows stay at SZ_256M
>     - Add BUILD_BUG_ON for MOCK_AUTO_REGION_SIZE_DEFAULT v.s. PMD_SIZE
>       in cxl_test_init()
> 
> Best regards,
> Richard Cheng.
> ---
>  tools/testing/cxl/test/cxl.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/tools/testing/cxl/test/cxl.c b/tools/testing/cxl/test/cxl.c
> index 418669927fb0..7a1105f9a0bb 100644
> --- a/tools/testing/cxl/test/cxl.c
> +++ b/tools/testing/cxl/test/cxl.c
> @@ -318,7 +318,7 @@ static struct {
>  			.restrictions = ACPI_CEDT_CFMWS_RESTRICT_HOSTONLYMEM |
>  					ACPI_CEDT_CFMWS_RESTRICT_VOLATILE,
>  			.qtg_id = FAKE_QTG_ID,
> -			.window_size = SZ_256M,
> +			.window_size = SZ_256M > PMD_SIZE ? SZ_256M : PMD_SIZE,

Why not this: max_t(resource_size_t, SZ_256M, PMD_SIZE)


>  		},
>  		.target = { 3 },
>  	},
> @@ -495,9 +495,12 @@ static int populate_cedt(void)
>  
>  	for (i = cfmws_start; i <= cfmws_end; i++) {
>  		struct acpi_cedt_cfmws *window = mock_cfmws[i];
> +		int align = SZ_256M;
>  
>  		cfmws_elc_update(window, i);
> -		res = alloc_mock_res(window->window_size, SZ_256M);
> +		if (window->restrictions & ACPI_CEDT_CFMWS_RESTRICT_VOLATILE)
> +			align = max_t(int, SZ_256M, PMD_SIZE);
> +		res = alloc_mock_res(window->window_size, align);
>  		if (!res)
>  			return -ENOMEM;
>  		window->base_hpa = res->range.start;
> @@ -1826,6 +1829,8 @@ static __init int cxl_test_init(void)
>  	int rc, i;
>  	struct range mappable;
>  
> +	BUILD_BUG_ON(!IS_ALIGNED(MOCK_AUTO_REGION_SIZE_DEFAULT, PMD_SIZE));

The rule we need to enforce is tha the actual auto region size is
PMD aligned, not the default definition. So I think this needs to
be a runtime check of IS_ALIGNED(mock_auto_region_size, PMD_SIZE).

-- Alison

> +
>  	cxl_acpi_test();
>  	cxl_core_test();
>  	cxl_mem_test();
> 
> base-commit: 6779b50faa562e6cca1aa6a4649a4d764c6c7e28
> -- 
> 2.43.0
> 

Re: [PATCH v3] cxl/test: Enforce PMD alignment for volatile mock regions

[Richard Cheng]

On Tue, May 26, 2026 at 04:21:30PM +0800, Alison Schofield wrote:
> On Fri, May 22, 2026 at 01:44:57PM +0800, Richard Cheng wrote:
> > cxl_test allocates synthetic CFMWS HPA windows from a gen_pool with
> > SZ_256M alignment. On arm64 with CONFIG_ARM64_64K_PAGES=y and
> > CONFIG_PGTABLE_LEVELS=3, PMD_SIZE is 512M, so every CXL region carved
> > from a volatile window inherits a non-PMD-aligned start, and
> > cxl_dax_region_probe() -> alloc_dax_region() fails:
> > 
> > """
> >     cxl_dax_region dax_region1: probe with driver cxl_dax_region failed
> >                                 with error -12
> > """
> > 
> > Enforce that every volatile mock CFMWS is PMD-aligned in both start and
> > size
> > 
> > Reviewed-by: Dave Jiang <dave.jiang@intel.com>
> > Acked-by: Kai-Heng Feng <kaihengf@nvidia.com>
> > Signed-off-by: Richard Cheng <icheng@nvidia.com>
> > ---
> > Changelog:
> > 
> > v2->v3:
> >     - Correct hunk headers
> >     - Add base commit SHA
> > 
> > v1->v2:
> >     - Reframe from "fix start alignment" to "enforce PMD invariant for
> >       volatile CFMWS"
> >     - Bump cfmws5.window_size to max(SZ_256M, PMD_SIZE)
> >     - Gate populate_cedt() alignment on
> >       ACPI_CEDT_CFMWS_RESTRICT_VOLATILE so PMEM windows stay at SZ_256M
> >     - Add BUILD_BUG_ON for MOCK_AUTO_REGION_SIZE_DEFAULT v.s. PMD_SIZE
> >       in cxl_test_init()
> > 
> > Best regards,
> > Richard Cheng.
> > ---
> >  tools/testing/cxl/test/cxl.c | 9 +++++++--
> >  1 file changed, 7 insertions(+), 2 deletions(-)
> > 
> > diff --git a/tools/testing/cxl/test/cxl.c b/tools/testing/cxl/test/cxl.c
> > index 418669927fb0..7a1105f9a0bb 100644
> > --- a/tools/testing/cxl/test/cxl.c
> > +++ b/tools/testing/cxl/test/cxl.c
> > @@ -318,7 +318,7 @@ static struct {
> >  			.restrictions = ACPI_CEDT_CFMWS_RESTRICT_HOSTONLYMEM |
> >  					ACPI_CEDT_CFMWS_RESTRICT_VOLATILE,
> >  			.qtg_id = FAKE_QTG_ID,
> > -			.window_size = SZ_256M,
> > +			.window_size = SZ_256M > PMD_SIZE ? SZ_256M : PMD_SIZE,
> 
> Why not this: max_t(resource_size_t, SZ_256M, PMD_SIZE)
>

Hi Alison,

Thanks for the feedback, but I doubt this will work.
"max_t()" expands to a statement expression, which isn't a constant expression so it can't be used in this file-scope static initializer.

I did the change and compile, the error is
"""
  include/linux/minmax.h:86:9: error: braced-group within expression
                                allowed only inside a function
     86 |     ({ type ux = (x); type uy = (y); __cmp(op, ux, uy); })
        |     ^
  test/cxl.c:295:40: note: in expansion of macro 'max_t'
    295 |     .window_size = max_t(resource_size_t, SZ_256M, PMD_SIZE),
"""

So I don't think this approach works, I would like to stick to ternary form if that's ok for you. 
> 
> >  		},
> >  		.target = { 3 },
> >  	},
> > @@ -495,9 +495,12 @@ static int populate_cedt(void)
> >  
> >  	for (i = cfmws_start; i <= cfmws_end; i++) {
> >  		struct acpi_cedt_cfmws *window = mock_cfmws[i];
> > +		int align = SZ_256M;
> >  
> >  		cfmws_elc_update(window, i);
> > -		res = alloc_mock_res(window->window_size, SZ_256M);
> > +		if (window->restrictions & ACPI_CEDT_CFMWS_RESTRICT_VOLATILE)
> > +			align = max_t(int, SZ_256M, PMD_SIZE);
> > +		res = alloc_mock_res(window->window_size, align);
> >  		if (!res)
> >  			return -ENOMEM;
> >  		window->base_hpa = res->range.start;
> > @@ -1826,6 +1829,8 @@ static __init int cxl_test_init(void)
> >  	int rc, i;
> >  	struct range mappable;
> >  
> > +	BUILD_BUG_ON(!IS_ALIGNED(MOCK_AUTO_REGION_SIZE_DEFAULT, PMD_SIZE));
> 
> The rule we need to enforce is tha the actual auto region size is
> PMD aligned, not the default definition. So I think this needs to
> be a runtime check of IS_ALIGNED(mock_auto_region_size, PMD_SIZE).
> 
> -- Alison
> 

Thanks for the catch, I'll fix this and send a v4 to make it runtime check.

Do you prefer WARN_ON_ONCE() to be ammend or no? I mean

"""
if (!IS_ALIGNED(mock_auto_region_size, PMD_SIZE))
	return -EINVAL;
"""
or
"""
if (WARN_ON_ONCE(!IS_ALIGNED(mock_auto_region_size, PMD_SIZE)))
	return -EINVAL;
"""

Which one do you prefer ? or any other idea ? please let me know and I'll make the change.

Best regards,
Richard Cheng.

> > +
> >  	cxl_acpi_test();
> >  	cxl_core_test();
> >  	cxl_mem_test();
> > 
> > base-commit: 6779b50faa562e6cca1aa6a4649a4d764c6c7e28
> > -- 
> > 2.43.0
> > 

Re: [PATCH v3] cxl/test: Enforce PMD alignment for volatile mock regions

[Alison Schofield]

On Wed, May 27, 2026 at 12:08:07PM +0800, Richard Cheng wrote:
> On Tue, May 26, 2026 at 04:21:30PM +0800, Alison Schofield wrote:
> > On Fri, May 22, 2026 at 01:44:57PM +0800, Richard Cheng wrote:

snip

> > > diff --git a/tools/testing/cxl/test/cxl.c b/tools/testing/cxl/test/cxl.c
> > > index 418669927fb0..7a1105f9a0bb 100644
> > > --- a/tools/testing/cxl/test/cxl.c
> > > +++ b/tools/testing/cxl/test/cxl.c
> > > @@ -318,7 +318,7 @@ static struct {
> > >  			.restrictions = ACPI_CEDT_CFMWS_RESTRICT_HOSTONLYMEM |
> > >  					ACPI_CEDT_CFMWS_RESTRICT_VOLATILE,
> > >  			.qtg_id = FAKE_QTG_ID,
> > > -			.window_size = SZ_256M,
> > > +			.window_size = SZ_256M > PMD_SIZE ? SZ_256M : PMD_SIZE,
> > 
> > Why not this: max_t(resource_size_t, SZ_256M, PMD_SIZE)
> >
> 
> Hi Alison,
> 
> Thanks for the feedback, but I doubt this will work.
> "max_t()" expands to a statement expression, which isn't a constant expression so it can't be used in this file-scope static initializer.
> 
> I did the change and compile, the error is
> """
>   include/linux/minmax.h:86:9: error: braced-group within expression
>                                 allowed only inside a function
>      86 |     ({ type ux = (x); type uy = (y); __cmp(op, ux, uy); })
>         |     ^
>   test/cxl.c:295:40: note: in expansion of macro 'max_t'
>     295 |     .window_size = max_t(resource_size_t, SZ_256M, PMD_SIZE),
> """
> 
> So I don't think this approach works, I would like to stick to ternary form if that's ok for you. 

Sounds good, and right.  Thanks for the explanation.


> > > +	BUILD_BUG_ON(!IS_ALIGNED(MOCK_AUTO_REGION_SIZE_DEFAULT, PMD_SIZE));
> > 
> > The rule we need to enforce is tha the actual auto region size is
> > PMD aligned, not the default definition. So I think this needs to
> > be a runtime check of IS_ALIGNED(mock_auto_region_size, PMD_SIZE).
> > 
> > -- Alison
> > 
> 
> Thanks for the catch, I'll fix this and send a v4 to make it runtime check.
> 
> Do you prefer WARN_ON_ONCE() to be ammend or no? I mean
> 
> """
> if (!IS_ALIGNED(mock_auto_region_size, PMD_SIZE))
> 	return -EINVAL;
> """
> or
> """
> if (WARN_ON_ONCE(!IS_ALIGNED(mock_auto_region_size, PMD_SIZE)))
> 	return -EINVAL;
> """
> 
> Which one do you prefer ? or any other idea ? please let me know and I'll make the change.

I don't think we need the backtrace of warn, how about this:

if (!IS_ALIGNED(mock_auto_region_size, PMD_SIZE)) {
    pr_err_once("mock_auto_region_size %d must be PMD-aligned\n",
                mock_auto_region_size);
    return -EINVAL;
}

> 
> Best regards,
> Richard Cheng.
> 
> > > +
> > >  	cxl_acpi_test();
> > >  	cxl_core_test();
> > >  	cxl_mem_test();
> > > 
> > > base-commit: 6779b50faa562e6cca1aa6a4649a4d764c6c7e28
> > > -- 
> > > 2.43.0
> > >