Cleaning up e820_pmem?

Cleaning up e820_pmem?

[Andy Lutomirski]

My laptop has /sys/devices/platform/e820_pmem and autoloads all the
nvdimm infrastructure.  While it would be really cool if my laptop had
pmem, that's a bit of a pipe dream right now.  (Even if it did have
it, this laptop is brand new -- it should use NFIT, not e820_pmem.)

Could we move the iomem_resource loop from drivers/nvdimm/e820.c to
arch/x86/kernel/pmem.c and actually list the iomem resources the
standard way as resources belonging to the platform device?  That
would match accepted practice, and it would keep the grossly
x86-specific part of the driver in arch/x86.  Then we could further
tweak it to skip creating the platform device at all if there are no
resources, and we'd avoid needlessly loading the module.

I'd do this myself, except that my lovely machine that *does* support
e820 pmem has been repurposed, so testing on a machine that actually
supports this turd is awkward for me.

--Andy

Cleaning up e820_pmem?

[Andy Lutomirski]

My laptop has /sys/devices/platform/e820_pmem and autoloads all the
nvdimm infrastructure.  While it would be really cool if my laptop had
pmem, that's a bit of a pipe dream right now.  (Even if it did have
it, this laptop is brand new -- it should use NFIT, not e820_pmem.)

Could we move the iomem_resource loop from drivers/nvdimm/e820.c to
arch/x86/kernel/pmem.c and actually list the iomem resources the
standard way as resources belonging to the platform device?  That
would match accepted practice, and it would keep the grossly
x86-specific part of the driver in arch/x86.  Then we could further
tweak it to skip creating the platform device at all if there are no
resources, and we'd avoid needlessly loading the module.

I'd do this myself, except that my lovely machine that *does* support
e820 pmem has been repurposed, so testing on a machine that actually
supports this turd is awkward for me.

--Andy

Re: Cleaning up e820_pmem?

[Williams, Dan J]

On Sun, 2015-11-29 at 22:26 -0800, Andy Lutomirski wrote:
> My laptop has /sys/devices/platform/e820_pmem and autoloads all the
> nvdimm infrastructure.  While it would be really cool if my laptop
> had
> pmem, that's a bit of a pipe dream right now.  (Even if it did have
> it, this laptop is brand new -- it should use NFIT, not e820_pmem.)
> 
> Could we move the iomem_resource loop from drivers/nvdimm/e820.c to
> arch/x86/kernel/pmem.c and actually list the iomem resources the
> standard way as resources belonging to the platform device?  That
> would match accepted practice, and it would keep the grossly
> x86-specific part of the driver in arch/x86.  Then we could further
> tweak it to skip creating the platform device at all if there are no
> resources, and we'd avoid needlessly loading the module.
> 
> I'd do this myself, except that my lovely machine that *does* support
> e820 pmem has been repurposed, so testing on a machine that actually
> supports this turd is awkward for me.

This works for me...

8<---
Subject: libnvdimm, e820: skip module loading when no type-12

From: Dan Williams <dan.j.williams@intel.com>

If there are no persistent memory ranges present then don't bother
creating the platform device.  Otherwise, it loads the full libnvdimm
sub-system only to discover no resources present.

Reported-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 arch/x86/kernel/pmem.c |   12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/arch/x86/kernel/pmem.c b/arch/x86/kernel/pmem.c
index 4f00b63d7ff3..14415aff1813 100644
--- a/arch/x86/kernel/pmem.c
+++ b/arch/x86/kernel/pmem.c
@@ -4,10 +4,22 @@
  */
 #include <linux/platform_device.h>
 #include <linux/module.h>
+#include <linux/ioport.h>
+
+static int found(u64 start, u64 end, void *data)
+{
+	return 1;
+}
 
 static __init int register_e820_pmem(void)
 {
+	char *pmem = "Persistent Memory (legacy)";
 	struct platform_device *pdev;
+	int rc;
+
+	rc = walk_iomem_res(pmem, IORESOURCE_MEM, 0, -1, NULL, found);
+	if (rc <= 0)
+		return 0;
 
 	/*
 	 * See drivers/nvdimm/e820.c for the implementation, this is

Re: Cleaning up e820_pmem?

[Williams, Dan J]

On Sun, 2015-11-29 at 22:26 -0800, Andy Lutomirski wrote:
> My laptop has /sys/devices/platform/e820_pmem and autoloads all the
> nvdimm infrastructure.  While it would be really cool if my laptop
> had
> pmem, that's a bit of a pipe dream right now.  (Even if it did have
> it, this laptop is brand new -- it should use NFIT, not e820_pmem.)
> 
> Could we move the iomem_resource loop from drivers/nvdimm/e820.c to
> arch/x86/kernel/pmem.c and actually list the iomem resources the
> standard way as resources belonging to the platform device?  That
> would match accepted practice, and it would keep the grossly
> x86-specific part of the driver in arch/x86.  Then we could further
> tweak it to skip creating the platform device at all if there are no
> resources, and we'd avoid needlessly loading the module.
> 
> I'd do this myself, except that my lovely machine that *does* support
> e820 pmem has been repurposed, so testing on a machine that actually
> supports this turd is awkward for me.

This works for me...

8<---
Subject: libnvdimm, e820: skip module loading when no type-12

From: Dan Williams <dan.j.williams@intel.com>

If there are no persistent memory ranges present then don't bother
creating the platform device.  Otherwise, it loads the full libnvdimm
sub-system only to discover no resources present.

Reported-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 arch/x86/kernel/pmem.c |   12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/arch/x86/kernel/pmem.c b/arch/x86/kernel/pmem.c
index 4f00b63d7ff3..14415aff1813 100644
--- a/arch/x86/kernel/pmem.c
+++ b/arch/x86/kernel/pmem.c
@@ -4,10 +4,22 @@
  */
 #include <linux/platform_device.h>
 #include <linux/module.h>
+#include <linux/ioport.h>
+
+static int found(u64 start, u64 end, void *data)
+{
+	return 1;
+}
 
 static __init int register_e820_pmem(void)
 {
+	char *pmem = "Persistent Memory (legacy)";
 	struct platform_device *pdev;
+	int rc;
+
+	rc = walk_iomem_res(pmem, IORESOURCE_MEM, 0, -1, NULL, found);
+	if (rc <= 0)
+		return 0;
 
 	/*
 	 * See drivers/nvdimm/e820.c for the implementation, this is

Re: Cleaning up e820_pmem?

[Andy Lutomirski]

On Mon, Nov 30, 2015 at 10:11 AM, Williams, Dan J
<dan.j.williams@intel.com> wrote:
> On Sun, 2015-11-29 at 22:26 -0800, Andy Lutomirski wrote:
>> My laptop has /sys/devices/platform/e820_pmem and autoloads all the
>> nvdimm infrastructure.  While it would be really cool if my laptop
>> had
>> pmem, that's a bit of a pipe dream right now.  (Even if it did have
>> it, this laptop is brand new -- it should use NFIT, not e820_pmem.)
>>
>> Could we move the iomem_resource loop from drivers/nvdimm/e820.c to
>> arch/x86/kernel/pmem.c and actually list the iomem resources the
>> standard way as resources belonging to the platform device?  That
>> would match accepted practice, and it would keep the grossly
>> x86-specific part of the driver in arch/x86.  Then we could further
>> tweak it to skip creating the platform device at all if there are no
>> resources, and we'd avoid needlessly loading the module.
>>
>> I'd do this myself, except that my lovely machine that *does* support
>> e820 pmem has been repurposed, so testing on a machine that actually
>> supports this turd is awkward for me.
>
> This works for me...
>
> 8<---
> Subject: libnvdimm, e820: skip module loading when no type-12
>
> From: Dan Williams <dan.j.williams@intel.com>
>
> If there are no persistent memory ranges present then don't bother
> creating the platform device.  Otherwise, it loads the full libnvdimm
> sub-system only to discover no resources present.
>
> Reported-by: Andy Lutomirski <luto@amacapital.net>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
> ---
>  arch/x86/kernel/pmem.c |   12 ++++++++++++
>  1 file changed, 12 insertions(+)
>
> diff --git a/arch/x86/kernel/pmem.c b/arch/x86/kernel/pmem.c
> index 4f00b63d7ff3..14415aff1813 100644
> --- a/arch/x86/kernel/pmem.c
> +++ b/arch/x86/kernel/pmem.c
> @@ -4,10 +4,22 @@
>   */
>  #include <linux/platform_device.h>
>  #include <linux/module.h>
> +#include <linux/ioport.h>
> +
> +static int found(u64 start, u64 end, void *data)
> +{
> +       return 1;
> +}
>
>  static __init int register_e820_pmem(void)
>  {
> +       char *pmem = "Persistent Memory (legacy)";
>         struct platform_device *pdev;
> +       int rc;
> +
> +       rc = walk_iomem_res(pmem, IORESOURCE_MEM, 0, -1, NULL, found);
> +       if (rc <= 0)
> +               return 0;
>
>         /*
>          * See drivers/nvdimm/e820.c for the implementation, this is

LGTM.

--Andy

Re: Cleaning up e820_pmem?

[Andy Lutomirski]

On Mon, Nov 30, 2015 at 10:11 AM, Williams, Dan J
<dan.j.williams@intel.com> wrote:
> On Sun, 2015-11-29 at 22:26 -0800, Andy Lutomirski wrote:
>> My laptop has /sys/devices/platform/e820_pmem and autoloads all the
>> nvdimm infrastructure.  While it would be really cool if my laptop
>> had
>> pmem, that's a bit of a pipe dream right now.  (Even if it did have
>> it, this laptop is brand new -- it should use NFIT, not e820_pmem.)
>>
>> Could we move the iomem_resource loop from drivers/nvdimm/e820.c to
>> arch/x86/kernel/pmem.c and actually list the iomem resources the
>> standard way as resources belonging to the platform device?  That
>> would match accepted practice, and it would keep the grossly
>> x86-specific part of the driver in arch/x86.  Then we could further
>> tweak it to skip creating the platform device at all if there are no
>> resources, and we'd avoid needlessly loading the module.
>>
>> I'd do this myself, except that my lovely machine that *does* support
>> e820 pmem has been repurposed, so testing on a machine that actually
>> supports this turd is awkward for me.
>
> This works for me...
>
> 8<---
> Subject: libnvdimm, e820: skip module loading when no type-12
>
> From: Dan Williams <dan.j.williams@intel.com>
>
> If there are no persistent memory ranges present then don't bother
> creating the platform device.  Otherwise, it loads the full libnvdimm
> sub-system only to discover no resources present.
>
> Reported-by: Andy Lutomirski <luto@amacapital.net>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
> ---
>  arch/x86/kernel/pmem.c |   12 ++++++++++++
>  1 file changed, 12 insertions(+)
>
> diff --git a/arch/x86/kernel/pmem.c b/arch/x86/kernel/pmem.c
> index 4f00b63d7ff3..14415aff1813 100644
> --- a/arch/x86/kernel/pmem.c
> +++ b/arch/x86/kernel/pmem.c
> @@ -4,10 +4,22 @@
>   */
>  #include <linux/platform_device.h>
>  #include <linux/module.h>
> +#include <linux/ioport.h>
> +
> +static int found(u64 start, u64 end, void *data)
> +{
> +       return 1;
> +}
>
>  static __init int register_e820_pmem(void)
>  {
> +       char *pmem = "Persistent Memory (legacy)";
>         struct platform_device *pdev;
> +       int rc;
> +
> +       rc = walk_iomem_res(pmem, IORESOURCE_MEM, 0, -1, NULL, found);
> +       if (rc <= 0)
> +               return 0;
>
>         /*
>          * See drivers/nvdimm/e820.c for the implementation, this is

LGTM.

--Andy