[PATCH v2] DAX: warn when kmem regions are truncated for memory block alignment.

Gregory Price posted 1 patch 10 months, 1 week ago
There is a newer version of this series
drivers/dax/kmem.c | 10 +++++++++-
1 file changed, 9 insertions(+), 1 deletion(-)
[PATCH v2] DAX: warn when kmem regions are truncated for memory block alignment.
Posted by Gregory Price 10 months, 1 week ago
Device capacity intended for use as system ram should be aligned to the
archite-defined memory block size or that capacity will be silently
truncated and capacity stranded.

As hotplug dax memory becomes more prevelant, the memory block size
alignment becomes more important for platform and device vendors to
pay attention to - so this truncation should not be silent.

This issue is particularly relevant for CXL Dynamic Capacity devices,
whose capacity may arrive in spec-aligned but block-misaligned chunks.

Suggested-by: David Hildenbrand <david@redhat.com>
Suggested-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Gregory Price <gourry@gourry.net>
---
 drivers/dax/kmem.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

diff --git a/drivers/dax/kmem.c b/drivers/dax/kmem.c
index e97d47f42ee2..32fe3215e11e 100644
--- a/drivers/dax/kmem.c
+++ b/drivers/dax/kmem.c
@@ -13,6 +13,7 @@
 #include <linux/mman.h>
 #include <linux/memory-tiers.h>
 #include <linux/memory_hotplug.h>
+#include <linux/string_helpers.h>
 #include "dax-private.h"
 #include "bus.h"
 
@@ -68,7 +69,7 @@ static void kmem_put_memory_types(void)
 static int dev_dax_kmem_probe(struct dev_dax *dev_dax)
 {
 	struct device *dev = &dev_dax->dev;
-	unsigned long total_len = 0;
+	unsigned long total_len = 0, orig_len = 0;
 	struct dax_kmem_data *data;
 	struct memory_dev_type *mtype;
 	int i, rc, mapped = 0;
@@ -97,6 +98,7 @@ static int dev_dax_kmem_probe(struct dev_dax *dev_dax)
 	for (i = 0; i < dev_dax->nr_range; i++) {
 		struct range range;
 
+		orig_len += range_len(&dev_dax->ranges[i].range);
 		rc = dax_kmem_range(dev_dax, i, &range);
 		if (rc) {
 			dev_info(dev, "mapping%d: %#llx-%#llx too small after alignment\n",
@@ -109,6 +111,12 @@ static int dev_dax_kmem_probe(struct dev_dax *dev_dax)
 	if (!total_len) {
 		dev_warn(dev, "rejecting DAX region without any memory after alignment\n");
 		return -EINVAL;
+	} else if (total_len != orig_len) {
+		char buf[16];
+
+		string_get_size((orig_len - total_len), 1, STRING_UNITS_2,
+				buf, sizeof(buf));
+		dev_warn(dev, "DAX region truncated by %s due to alignment\n", buf);
 	}
 
 	init_node_memory_type(numa_node, mtype);
-- 
2.47.1
Re: [PATCH v2] DAX: warn when kmem regions are truncated for memory block alignment.
Posted by Dan Williams 10 months ago
Gregory Price wrote:
> Device capacity intended for use as system ram should be aligned to the
> archite-defined memory block size or that capacity will be silently
> truncated and capacity stranded.
> 
> As hotplug dax memory becomes more prevelant, the memory block size
> alignment becomes more important for platform and device vendors to
> pay attention to - so this truncation should not be silent.
> 
> This issue is particularly relevant for CXL Dynamic Capacity devices,
> whose capacity may arrive in spec-aligned but block-misaligned chunks.
> 
> Suggested-by: David Hildenbrand <david@redhat.com>
> Suggested-by: Dan Williams <dan.j.williams@intel.com>
> Signed-off-by: Gregory Price <gourry@gourry.net>

The typo in the changelog was already noted, and this addresses all my
feedback so feel free to add:

Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Re: [PATCH v2] DAX: warn when kmem regions are truncated for memory block alignment.
Posted by Alison Schofield 10 months ago
On Tue, Apr 01, 2025 at 09:59:20PM -0400, Gregory Price wrote:
> Device capacity intended for use as system ram should be aligned to the
> archite-defined memory block size or that capacity will be silently
> truncated and capacity stranded.
> 
> As hotplug dax memory becomes more prevelant, the memory block size
> alignment becomes more important for platform and device vendors to
> pay attention to - so this truncation should not be silent.
> 
> This issue is particularly relevant for CXL Dynamic Capacity devices,
> whose capacity may arrive in spec-aligned but block-misaligned chunks.
> 
> Suggested-by: David Hildenbrand <david@redhat.com>
> Suggested-by: Dan Williams <dan.j.williams@intel.com>
> Signed-off-by: Gregory Price <gourry@gourry.net>

Existing unit test 'daxctl-devices.sh' hits this case:
[   52.547521] kmem dax3.0: DAX region truncated by 62.0 MiB due to alignment

Tested-by: Alison Schofield <alison.schofield@intel.com>

snip

>
Re: [PATCH v2] DAX: warn when kmem regions are truncated for memory block alignment.
Posted by Gregory Price 10 months ago
On Wed, Apr 09, 2025 at 03:21:53PM -0700, Alison Schofield wrote:
> Existing unit test 'daxctl-devices.sh' hits this case:
> [   52.547521] kmem dax3.0: DAX region truncated by 62.0 MiB due to alignment
> 
> Tested-by: Alison Schofield <alison.schofield@intel.com>
>

Thanks for testing, good to know there's an existing test for this

will respin w/ Jonathan's recommendations and tags.

~Gregory
Re: [PATCH v2] DAX: warn when kmem regions are truncated for memory block alignment.
Posted by Jonathan Cameron 10 months, 1 week ago
On Tue,  1 Apr 2025 21:59:20 -0400
Gregory Price <gourry@gourry.net> wrote:

> Device capacity intended for use as system ram should be aligned to the
> archite-defined memory block size or that capacity will be silently

archite?

> truncated and capacity stranded.
> 
> As hotplug dax memory becomes more prevelant, the memory block size
> alignment becomes more important for platform and device vendors to
> pay attention to - so this truncation should not be silent.
> 
> This issue is particularly relevant for CXL Dynamic Capacity devices,
> whose capacity may arrive in spec-aligned but block-misaligned chunks.
> 
> Suggested-by: David Hildenbrand <david@redhat.com>
> Suggested-by: Dan Williams <dan.j.williams@intel.com>
> Signed-off-by: Gregory Price <gourry@gourry.net>

One trivial comment inline otherwise seems reasonable to me.
Reviewed-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>


> ---
>  drivers/dax/kmem.c | 10 +++++++++-
>  1 file changed, 9 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/dax/kmem.c b/drivers/dax/kmem.c
> index e97d47f42ee2..32fe3215e11e 100644
> --- a/drivers/dax/kmem.c
> +++ b/drivers/dax/kmem.c
> @@ -13,6 +13,7 @@
>  #include <linux/mman.h>
>  #include <linux/memory-tiers.h>
>  #include <linux/memory_hotplug.h>
> +#include <linux/string_helpers.h>
>  #include "dax-private.h"
>  #include "bus.h"
>  
> @@ -68,7 +69,7 @@ static void kmem_put_memory_types(void)
>  static int dev_dax_kmem_probe(struct dev_dax *dev_dax)
>  {
>  	struct device *dev = &dev_dax->dev;
> -	unsigned long total_len = 0;
> +	unsigned long total_len = 0, orig_len = 0;
>  	struct dax_kmem_data *data;
>  	struct memory_dev_type *mtype;
>  	int i, rc, mapped = 0;
> @@ -97,6 +98,7 @@ static int dev_dax_kmem_probe(struct dev_dax *dev_dax)
>  	for (i = 0; i < dev_dax->nr_range; i++) {
>  		struct range range;
>  
> +		orig_len += range_len(&dev_dax->ranges[i].range);
>  		rc = dax_kmem_range(dev_dax, i, &range);
>  		if (rc) {
>  			dev_info(dev, "mapping%d: %#llx-%#llx too small after alignment\n",
> @@ -109,6 +111,12 @@ static int dev_dax_kmem_probe(struct dev_dax *dev_dax)
>  	if (!total_len) {
>  		dev_warn(dev, "rejecting DAX region without any memory after alignment\n");
>  		return -EINVAL;
> +	} else if (total_len != orig_len) {
> +		char buf[16];
> +
> +		string_get_size((orig_len - total_len), 1, STRING_UNITS_2,

Trivial but do those inner brackets really add anything?

> +				buf, sizeof(buf));
> +		dev_warn(dev, "DAX region truncated by %s due to alignment\n", buf);
>  	}
>  
>  	init_node_memory_type(numa_node, mtype);
Re: [PATCH v2] DAX: warn when kmem regions are truncated for memory block alignment.
Posted by David Hildenbrand 10 months, 1 week ago
On 02.04.25 03:59, Gregory Price wrote:
> Device capacity intended for use as system ram should be aligned to the
> archite-defined memory block size or that capacity will be silently
> truncated and capacity stranded.
> 
> As hotplug dax memory becomes more prevelant, the memory block size
> alignment becomes more important for platform and device vendors to
> pay attention to - so this truncation should not be silent.
> 
> This issue is particularly relevant for CXL Dynamic Capacity devices,
> whose capacity may arrive in spec-aligned but block-misaligned chunks.
> 
> Suggested-by: David Hildenbrand <david@redhat.com>
> Suggested-by: Dan Williams <dan.j.williams@intel.com>
> Signed-off-by: Gregory Price <gourry@gourry.net>
> ---
>   drivers/dax/kmem.c | 10 +++++++++-
>   1 file changed, 9 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/dax/kmem.c b/drivers/dax/kmem.c
> index e97d47f42ee2..32fe3215e11e 100644
> --- a/drivers/dax/kmem.c
> +++ b/drivers/dax/kmem.c
> @@ -13,6 +13,7 @@
>   #include <linux/mman.h>
>   #include <linux/memory-tiers.h>
>   #include <linux/memory_hotplug.h>
> +#include <linux/string_helpers.h>
>   #include "dax-private.h"
>   #include "bus.h"
>   
> @@ -68,7 +69,7 @@ static void kmem_put_memory_types(void)
>   static int dev_dax_kmem_probe(struct dev_dax *dev_dax)
>   {
>   	struct device *dev = &dev_dax->dev;
> -	unsigned long total_len = 0;
> +	unsigned long total_len = 0, orig_len = 0;
>   	struct dax_kmem_data *data;
>   	struct memory_dev_type *mtype;
>   	int i, rc, mapped = 0;
> @@ -97,6 +98,7 @@ static int dev_dax_kmem_probe(struct dev_dax *dev_dax)
>   	for (i = 0; i < dev_dax->nr_range; i++) {
>   		struct range range;
>   
> +		orig_len += range_len(&dev_dax->ranges[i].range);
>   		rc = dax_kmem_range(dev_dax, i, &range);
>   		if (rc) {
>   			dev_info(dev, "mapping%d: %#llx-%#llx too small after alignment\n",
> @@ -109,6 +111,12 @@ static int dev_dax_kmem_probe(struct dev_dax *dev_dax)
>   	if (!total_len) {
>   		dev_warn(dev, "rejecting DAX region without any memory after alignment\n");
>   		return -EINVAL;
> +	} else if (total_len != orig_len) {
> +		char buf[16];
> +
> +		string_get_size((orig_len - total_len), 1, STRING_UNITS_2,
> +				buf, sizeof(buf));
> +		dev_warn(dev, "DAX region truncated by %s due to alignment\n", buf);

Acked-by: David Hildenbrand <david@redhat.com>

-- 
Cheers,

David / dhildenb