Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1623785 > unrolled thread

[PATCH 0/5] libnvdimm: acpi updates and a revert

Back to article view | Back to linux.kernel

Contents

  [PATCH 0/5] libnvdimm: acpi updates and a revert Dan Williams <dan.j.williams@intel.com> - 2017-04-14 19:10 +0200
    [PATCH 5/5] acpi, nfit: limit ->flush_probe() to initialization work Dan Williams <dan.j.williams@intel.com> - 2017-04-14 19:10 +0200
      Re: [PATCH 5/5] acpi, nfit: limit ->flush_probe() to initialization work Dan Williams <dan.j.williams@intel.com> - 2017-04-14 19:40 +0200
    [PATCH 2/5] acpi, nfit: add support for acpi 6.1 dimm state flags Dan Williams <dan.j.williams@intel.com> - 2017-04-14 19:10 +0200
      Re: [PATCH 2/5] acpi, nfit: add support for acpi 6.1 dimm state flags "Kani, Toshimitsu" <toshi.kani@hpe.com> - 2017-04-17 19:30 +0200

#1623785 — [PATCH 0/5] libnvdimm: acpi updates and a revert

With Dave's recent fix [1], we can restore error clearing for btt i/o in
4.12.

ACPI 6.1 introduced new health state flags. Beyond reflecting them in
the dimmX/flags sysfs attribute we also need to handle the deeper
implications of the ACPI_NFIT_MEM_MAP_FAILED flag which changes
assumptions on how the driver discovers dimms. In the "map failed" case
there may missing or no SPA entries associated with a dimm. Those dimms
should still be registered with libnvdimm so that the error state can be
communicated and recovery attempted.

[1]: https://patchwork.kernel.org/patch/9680035/

---

Dan Williams (5):
      Revert "libnvdimm: band aid btt vs clear poison locking"
      acpi, nfit: add support for acpi 6.1 dimm state flags
      tools/testing/nvdimm: test acpi 6.1 health state flags
      acpi, nfit: support "map failed" dimms
      acpi, nfit: limit ->flush_probe() to initialization work


 drivers/acpi/nfit/core.c         |   61 +++++++++++++++++++++++++++++++-------
 drivers/acpi/nfit/nfit.h         |    1 +
 drivers/nvdimm/claim.c           |   10 +-----
 tools/testing/nvdimm/test/nfit.c |   40 +++++++++++++++++++++++--
 4 files changed, 88 insertions(+), 24 deletions(-)

[toc] | [next] | [standalone]

#1623786 — [PATCH 5/5] acpi, nfit: limit ->flush_probe() to initialization work

The nvdimm probe flushing mechanism gives userspace a sync point where
it knows all asynchronous driver probe sequences have completed.
However, it need not wait for other asynchronous actions, like
on-demand address-range-scrub. Track the init work separately from other
work in the workqueue, and only flush the former.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/acpi/nfit/core.c |   13 ++++++++++---
 drivers/acpi/nfit/nfit.h |    1 +
 2 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
index 06738df477db..17cac9d369e0 100644
--- a/drivers/acpi/nfit/core.c
+++ b/drivers/acpi/nfit/core.c
@@ -738,7 +738,7 @@ static void nfit_mem_init_bdw(struct acpi_nfit_desc *acpi_desc,
 	}
 }
 
-static int nfit_mem_dcr_init(struct acpi_nfit_desc *acpi_desc,
+static int __nfit_mem_init(struct acpi_nfit_desc *acpi_desc,
 		struct acpi_nfit_system_address *spa)
 {
 	struct nfit_mem *nfit_mem, *found;
@@ -898,7 +898,7 @@ static int nfit_mem_init(struct acpi_nfit_desc *acpi_desc)
 	 * BDWs are optional.
 	 */
 	list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
-		rc = nfit_mem_dcr_init(acpi_desc, nfit_spa->spa);
+		rc = __nfit_mem_init(acpi_desc, nfit_spa->spa);
 		if (rc)
 			return rc;
 	}
@@ -908,7 +908,7 @@ static int nfit_mem_init(struct acpi_nfit_desc *acpi_desc)
 	 * SPA entries above. Find and register all the unmapped DIMMs
 	 * for reporting and recovery purposes.
 	 */
-	rc = nfit_mem_dcr_init(acpi_desc, NULL);
+	rc = __nfit_mem_init(acpi_desc, NULL);
 	if (rc)
 		return rc;
 
@@ -2568,6 +2568,7 @@ static void acpi_nfit_scrub(struct work_struct *work)
 			acpi_nfit_register_region(acpi_desc, nfit_spa);
 		}
 	}
+	acpi_desc->init_complete = 1;
 
 	list_for_each_entry(nfit_spa, &acpi_desc->spas, list)
 		acpi_nfit_async_scrub(acpi_desc, nfit_spa);
@@ -2771,6 +2772,12 @@ static int acpi_nfit_flush_probe(struct nvdimm_bus_descriptor *nd_desc)
 	device_lock(dev);
 	device_unlock(dev);
 
+	/* bounce the init_mutex to make init_complete valid */
+	mutex_lock(&acpi_desc->init_mutex);
+	mutex_unlock(&acpi_desc->init_mutex);
+	if (acpi_desc->init_complete)
+		return 0;
+
 	/*
 	 * Scrub work could take 10s of seconds, userspace may give up so we
 	 * need to be interruptible while waiting.
diff --git a/drivers/acpi/nfit/nfit.h b/drivers/acpi/nfit/nfit.h
index fc29c2e9832e..256829597585 100644
--- a/drivers/acpi/nfit/nfit.h
+++ b/drivers/acpi/nfit/nfit.h
@@ -163,6 +163,7 @@ struct acpi_nfit_desc {
 	unsigned int scrub_count;
 	unsigned int scrub_mode;
 	unsigned int cancel:1;
+	unsigned int init_complete:1;
 	unsigned long dimm_cmd_force_en;
 	unsigned long bus_cmd_force_en;
 	int (*blk_do_io)(struct nd_blk_region *ndbr, resource_size_t dpa,

[toc] | [prev] | [next] | [standalone]

#1623807 — Re: [PATCH 5/5] acpi, nfit: limit ->flush_probe() to initialization work

On Fri, Apr 14, 2017 at 9:58 AM, Dan Williams <dan.j.williams@intel.com> wrote:
> The nvdimm probe flushing mechanism gives userspace a sync point where
> it knows all asynchronous driver probe sequences have completed.
> However, it need not wait for other asynchronous actions, like
> on-demand address-range-scrub. Track the init work separately from other
> work in the workqueue, and only flush the former.
>
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
> ---
>  drivers/acpi/nfit/core.c |   13 ++++++++++---
>  drivers/acpi/nfit/nfit.h |    1 +
>  2 files changed, 11 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
> index 06738df477db..17cac9d369e0 100644
> --- a/drivers/acpi/nfit/core.c
> +++ b/drivers/acpi/nfit/core.c
> @@ -738,7 +738,7 @@ static void nfit_mem_init_bdw(struct acpi_nfit_desc *acpi_desc,
>         }
>  }
>
> -static int nfit_mem_dcr_init(struct acpi_nfit_desc *acpi_desc,
> +static int __nfit_mem_init(struct acpi_nfit_desc *acpi_desc,
>                 struct acpi_nfit_system_address *spa)
>  {
>         struct nfit_mem *nfit_mem, *found;
> @@ -898,7 +898,7 @@ static int nfit_mem_init(struct acpi_nfit_desc *acpi_desc)
>          * BDWs are optional.
>          */
>         list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
> -               rc = nfit_mem_dcr_init(acpi_desc, nfit_spa->spa);
> +               rc = __nfit_mem_init(acpi_desc, nfit_spa->spa);

Whoops, this function rename should have been folded into the previous patch.

[toc] | [prev] | [next] | [standalone]

#1623787 — [PATCH 2/5] acpi, nfit: add support for acpi 6.1 dimm state flags

Add support for the ACPI_NFIT_MEM_MAP_FAILED ("map_fail") and
ACPI_NFIT_MEM_HEALTH_ENABLED ("smart_notify") health state flags. The
"map_fail" flag identifies DIMMs that were not mapped into one or more
physical address ranges. The "health_notify" flag indicates whether
platform firmware will send notifications when there is new SMART health
data to consume.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/acpi/nfit/core.c |    6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
index 53943d6f4214..05829de43b1d 100644
--- a/drivers/acpi/nfit/core.c
+++ b/drivers/acpi/nfit/core.c
@@ -1246,12 +1246,14 @@ static ssize_t flags_show(struct device *dev,
 {
 	u16 flags = to_nfit_memdev(dev)->flags;
 
-	return sprintf(buf, "%s%s%s%s%s\n",
+	return sprintf(buf, "%s%s%s%s%s%s%s\n",
 		flags & ACPI_NFIT_MEM_SAVE_FAILED ? "save_fail " : "",
 		flags & ACPI_NFIT_MEM_RESTORE_FAILED ? "restore_fail " : "",
 		flags & ACPI_NFIT_MEM_FLUSH_FAILED ? "flush_fail " : "",
 		flags & ACPI_NFIT_MEM_NOT_ARMED ? "not_armed " : "",
-		flags & ACPI_NFIT_MEM_HEALTH_OBSERVED ? "smart_event " : "");
+		flags & ACPI_NFIT_MEM_HEALTH_OBSERVED ? "smart_event " : "",
+		flags & ACPI_NFIT_MEM_MAP_FAILED ? "map_fail " : "",
+		flags & ACPI_NFIT_MEM_HEALTH_ENABLED ? "smart_notify " : "");
 }
 static DEVICE_ATTR_RO(flags);
 

[toc] | [prev] | [next] | [standalone]

#1624765 — Re: [PATCH 2/5] acpi, nfit: add support for acpi 6.1 dimm state flags

On Fri, 2017-04-14 at 09:58 -0700, Dan Williams wrote:
> Add support for the ACPI_NFIT_MEM_MAP_FAILED ("map_fail") and
> ACPI_NFIT_MEM_HEALTH_ENABLED ("smart_notify") health state flags. The
> "map_fail" flag identifies DIMMs that were not mapped into one or
> more physical address ranges. The "health_notify" flag indicates
> whether platform firmware will send notifications when there is new
> SMART health data to consume.
> 
> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
> ---
>  drivers/acpi/nfit/core.c |    6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
> index 53943d6f4214..05829de43b1d 100644
> --- a/drivers/acpi/nfit/core.c
> +++ b/drivers/acpi/nfit/core.c
> @@ -1246,12 +1246,14 @@ static ssize_t flags_show(struct device *dev,
>  {
>  	u16 flags = to_nfit_memdev(dev)->flags;
>  
> -	return sprintf(buf, "%s%s%s%s%s\n",
> +	return sprintf(buf, "%s%s%s%s%s%s%s\n",
>  		flags & ACPI_NFIT_MEM_SAVE_FAILED ? "save_fail " :
> "",
>  		flags & ACPI_NFIT_MEM_RESTORE_FAILED ? "restore_fail
> " : "",
>  		flags & ACPI_NFIT_MEM_FLUSH_FAILED ? "flush_fail " :
> "",
>  		flags & ACPI_NFIT_MEM_NOT_ARMED ? "not_armed " : "",
> -		flags & ACPI_NFIT_MEM_HEALTH_OBSERVED ? "smart_event
> " : "");
> +		flags & ACPI_NFIT_MEM_HEALTH_OBSERVED ? "smart_event
> " : "",
> +		flags & ACPI_NFIT_MEM_MAP_FAILED ? "map_fail " : "",
> +		flags & ACPI_NFIT_MEM_HEALTH_ENABLED ? "smart_notify
> " : "");

Thanks for the update!  The above change looks good, but we will also
need to make the same changes to the dev_info() in
acpi_nfit_register_dimms().

-Toshi

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web