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

Groups > linux.kernel > #1449060 > unrolled thread

[PATCH v4 0/4] ARS rescanning triggered by latent errors or userspace

Back to article view | Back to linux.kernel

Contents

  [PATCH v4 0/4] ARS rescanning triggered by latent errors or  userspace Dan Williams <dan.j.williams@intel.com> - 2016-07-24 07:30 +0200
    [PATCH v4 4/4] nfit: do an ARS scrub on hitting a latent media error Dan Williams <dan.j.williams@intel.com> - 2016-07-24 07:30 +0200

#1449060 — [PATCH v4 0/4] ARS rescanning triggered by latent errors or userspace

Changes since v3 [1]:

1/ Fixed races of scrub_{store|show} versus driver shutdown.  We need to
   make sure the nvdimm_bus, nvdimm_bus_descriptor, and acpi_nfit_desc data
   structures for a given instance remain valid for the duration of an
   acpi_nfit_ars_rescan() submission.  Patch1 "libnvdimm: register
   nvdimm_bus devices with an nd_bus driver", which is new for v4, enables
   use of device_lock() to pin the nvdimm_bus active for this duration.

2/ Fixed races of scrub_{store|show} versus hotplug.  This was simply
   some missing acquisitions of acpi_desc->init_mutex.

3/ Enforce that scrub_store() only initiates a scrub when writing "1",
   other values are invalid.  This lets us introduce new values down the
   road to, for example, disable scrubs after machine checks or other scrub
   policies.

4/ Require all three ARS DSMs be available before the 'scrub' attribute
   becomes visible.

5/ Fix races of mce notifier chain vs driver shutdown by holding
   acpi_desc_lock over the acpi_nfit_destruct() event.

6/ Given ARM is now re-using the NFIT driver, move the x86 specific
   machine check code to its own conditionally-compiled file.  As a
   precursor to this change the nfit source was moved to its own
   sub-directory in Patch3 "nfit: move to nfit/ sub-directory".

[1]: https://lists.01.org/pipermail/linux-nvdimm/2016-July/006407.html

---
Initial cover letter from Vishal:

This series adds on-demand ARS scanning on both, discovery of
latent media errors, and a sysfs trigger from userspace.

The rescanning part is easy to test using the nfit_test framework
- create a namespace (this will by default have bad sectors in
the middle), clear the bad sectors by writing to them, trigger
the rescan through sysfs, and the bad sectors will reappear in
/sys/block/<pmemX>/badblocks.

For the mce handling, I've tested the notifier chain callback
being called with a mock struct mce (called via another sysfs
trigger - this isn't included in the patch obviously), which
has the address field set to a known address in a SPA range,
and the status field with the MCACOD flag set.

What I haven't easily been able to test is the same callback
path with a 'real world' mce, being called as part of the
x86_mce_decoder_chain notifier. I'd therefore appreciate a
closer look at the initial filtering done in nfit_handle_mce
(patch 3/3) from Tony or anyone more familiar with mce handling.

---

Dan Williams (2):
      libnvdimm: register nvdimm_bus devices with an nd_bus driver
      nfit: move to nfit/ sub-directory

Vishal Verma (2):
      nfit, libnvdimm: allow an ARS scrub to be triggered on demand
      nfit: do an ARS scrub on hitting a latent media error


 drivers/acpi/Kconfig             |   27 -----
 drivers/acpi/Makefile            |    2 
 drivers/acpi/nfit/Kconfig        |   26 +++++
 drivers/acpi/nfit/Makefile       |    3 +
 drivers/acpi/nfit/core.c         |  187 ++++++++++++++++++++++++++++++++++++--
 drivers/acpi/nfit/mce.c          |   89 ++++++++++++++++++
 drivers/acpi/nfit/nfit.h         |   24 +++++
 drivers/nvdimm/bus.c             |  188 +++++++++++++++++++++++++++++++++++++-
 drivers/nvdimm/core.c            |  128 +-------------------------
 include/linux/libnvdimm.h        |    1 
 tools/testing/nvdimm/Kbuild      |    5 +
 tools/testing/nvdimm/test/Kbuild |    2 
 12 files changed, 512 insertions(+), 170 deletions(-)
 create mode 100644 drivers/acpi/nfit/Kconfig
 create mode 100644 drivers/acpi/nfit/Makefile
 rename drivers/acpi/{nfit.c => nfit/core.c}
 create mode 100644 drivers/acpi/nfit/mce.c
 rename drivers/acpi/{nfit.h => nfit/nfit.h}

[toc] | [next] | [standalone]

#1449062 — [PATCH v4 4/4] nfit: do an ARS scrub on hitting a latent media error

From: Vishal Verma <vishal.l.verma@intel.com>

When a latent (unknown to 'badblocks') error is encountered, it will
trigger a machine check exception. On a system with machine check
recovery, this will only SIGBUS the process(es) which had the bad page
mapped (as opposed to a kernel panic on platforms without machine
check recovery features). In the former case, we want to trigger a full
rescan of that nvdimm bus. This will allow any additional, new errors
to be captured in the block devices' badblocks lists, and offending
operations on them can be trapped early, avoiding machine checks.

This is done by registering a callback function with the
x86_mce_decoder_chain and calling the new ars_rescan functionality with
the address in the mce notificatiion.

Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: Tony Luck <tony.luck@intel.com>
Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
---
 drivers/acpi/nfit/Makefile  |    1 
 drivers/acpi/nfit/core.c    |   26 +++++++++++--
 drivers/acpi/nfit/mce.c     |   89 +++++++++++++++++++++++++++++++++++++++++++
 drivers/acpi/nfit/nfit.h    |   20 ++++++++++
 tools/testing/nvdimm/Kbuild |    1 
 5 files changed, 133 insertions(+), 4 deletions(-)
 create mode 100644 drivers/acpi/nfit/mce.c

diff --git a/drivers/acpi/nfit/Makefile b/drivers/acpi/nfit/Makefile
index eb95c5aff83b..a407e769f103 100644
--- a/drivers/acpi/nfit/Makefile
+++ b/drivers/acpi/nfit/Makefile
@@ -1,2 +1,3 @@
 obj-$(CONFIG_ACPI_NFIT) := nfit.o
 nfit-y := core.o
+nfit-$(CONFIG_X86_MCE) += mce.o
diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
index 2eaa03dc185a..686837e8358f 100644
--- a/drivers/acpi/nfit/core.c
+++ b/drivers/acpi/nfit/core.c
@@ -51,6 +51,9 @@ module_param(disable_vendor_specific, bool, S_IRUGO);
 MODULE_PARM_DESC(disable_vendor_specific,
 		"Limit commands to the publicly specified set\n");
 
+LIST_HEAD(acpi_descs);
+DEFINE_MUTEX(acpi_desc_lock);
+
 static struct workqueue_struct *nfit_wq;
 
 struct nfit_table_prev {
@@ -361,7 +364,7 @@ static const char *spa_type_name(u16 type)
 	return to_name[type];
 }
 
-static int nfit_spa_type(struct acpi_nfit_system_address *spa)
+int nfit_spa_type(struct acpi_nfit_system_address *spa)
 {
 	int i;
 
@@ -898,8 +901,6 @@ static ssize_t scrub_show(struct device *dev,
 	return rc;
 }
 
-static int acpi_nfit_ars_rescan(struct acpi_nfit_desc *acpi_desc);
-
 static ssize_t scrub_store(struct device *dev,
 		struct device_attribute *attr, const char *buf, size_t size)
 {
@@ -2400,6 +2401,11 @@ static void acpi_nfit_destruct(void *data)
 	struct acpi_nfit_desc *acpi_desc = data;
 	struct device *bus_dev = to_nvdimm_bus_dev(acpi_desc->nvdimm_bus);
 
+	/*
+	 * Destruct under acpi_desc_lock so that nfit_handle_mce does not
+	 * race teardown
+	 */
+	mutex_lock(&acpi_desc_lock);
 	acpi_desc->cancel = 1;
 	/*
 	 * Bounce the nvdimm bus lock to make sure any in-flight
@@ -2414,6 +2420,8 @@ static void acpi_nfit_destruct(void *data)
 		sysfs_put(acpi_desc->scrub_count_state);
 	nvdimm_bus_unregister(acpi_desc->nvdimm_bus);
 	acpi_desc->nvdimm_bus = NULL;
+	list_del(&acpi_desc->list);
+	mutex_unlock(&acpi_desc_lock);
 }
 
 int acpi_nfit_init(struct acpi_nfit_desc *acpi_desc, void *data, acpi_size sz)
@@ -2441,6 +2449,11 @@ int acpi_nfit_init(struct acpi_nfit_desc *acpi_desc, void *data, acpi_size sz)
 		rc = acpi_nfit_desc_init_scrub_attr(acpi_desc);
 		if (rc)
 			return rc;
+
+		/* register this acpi_desc for mce notifications */
+		mutex_lock(&acpi_desc_lock);
+		list_add_tail(&acpi_desc->list, &acpi_descs);
+		mutex_unlock(&acpi_desc_lock);
 	}
 
 	rc = acpi_nfit_desc_init_scrub_attr(acpi_desc);
@@ -2555,7 +2568,7 @@ static int acpi_nfit_clear_to_send(struct nvdimm_bus_descriptor *nd_desc,
 	return 0;
 }
 
-static int acpi_nfit_ars_rescan(struct acpi_nfit_desc *acpi_desc)
+int acpi_nfit_ars_rescan(struct acpi_nfit_desc *acpi_desc)
 {
 	struct device *dev = acpi_desc->dev;
 	struct nfit_spa *nfit_spa;
@@ -2604,6 +2617,7 @@ void acpi_nfit_desc_init(struct acpi_nfit_desc *acpi_desc, struct device *dev)
 	INIT_LIST_HEAD(&acpi_desc->flushes);
 	INIT_LIST_HEAD(&acpi_desc->memdevs);
 	INIT_LIST_HEAD(&acpi_desc->dimms);
+	INIT_LIST_HEAD(&acpi_desc->list);
 	mutex_init(&acpi_desc->init_mutex);
 	INIT_WORK(&acpi_desc->work, acpi_nfit_scrub);
 }
@@ -2756,13 +2770,17 @@ static __init int nfit_init(void)
 	if (!nfit_wq)
 		return -ENOMEM;
 
+	nfit_mce_register();
+
 	return acpi_bus_register_driver(&acpi_nfit_driver);
 }
 
 static __exit void nfit_exit(void)
 {
+	nfit_mce_unregister();
 	acpi_bus_unregister_driver(&acpi_nfit_driver);
 	destroy_workqueue(nfit_wq);
+	WARN_ON(!list_empty(&acpi_descs));
 }
 
 module_init(nfit_init);
diff --git a/drivers/acpi/nfit/mce.c b/drivers/acpi/nfit/mce.c
new file mode 100644
index 000000000000..4c745bf389fe
--- /dev/null
+++ b/drivers/acpi/nfit/mce.c
@@ -0,0 +1,89 @@
+/*
+ * NFIT - Machine Check Handler
+ *
+ * Copyright(c) 2013-2016 Intel Corporation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ */
+#include <linux/notifier.h>
+#include <linux/acpi.h>
+#include <asm/mce.h>
+#include "nfit.h"
+
+static int nfit_handle_mce(struct notifier_block *nb, unsigned long val,
+			void *data)
+{
+	struct mce *mce = (struct mce *)data;
+	struct acpi_nfit_desc *acpi_desc;
+	struct nfit_spa *nfit_spa;
+
+	/* We only care about memory errors */
+	if (!(mce->status & MCACOD))
+		return NOTIFY_DONE;
+
+	/*
+	 * mce->addr contains the physical addr accessed that caused the
+	 * machine check. We need to walk through the list of NFITs, and see
+	 * if any of them matches that address, and only then start a scrub.
+	 */
+	mutex_lock(&acpi_desc_lock);
+	list_for_each_entry(acpi_desc, &acpi_descs, list) {
+		struct device *dev = acpi_desc->dev;
+		int found_match = 0;
+
+		mutex_lock(&acpi_desc->init_mutex);
+		list_for_each_entry(nfit_spa, &acpi_desc->spas, list) {
+			struct acpi_nfit_system_address *spa = nfit_spa->spa;
+
+			if (nfit_spa_type(spa) == NFIT_SPA_PM)
+				continue;
+			/* find the spa that covers the mce addr */
+			if (spa->address > mce->addr)
+				continue;
+			if ((spa->address + spa->length - 1) < mce->addr)
+				continue;
+			found_match = 1;
+			dev_dbg(dev, "%s: addr in SPA %d (0x%llx, 0x%llx)\n",
+				__func__, spa->range_index, spa->address,
+				spa->length);
+			/*
+			 * We can break at the first match because we're going
+			 * to rescan all the SPA ranges. There shouldn't be any
+			 * aliasing anyway.
+			 */
+			break;
+		}
+		mutex_unlock(&acpi_desc->init_mutex);
+
+		/*
+		 * We can ignore an -EBUSY here because if an ARS is already
+		 * in progress, just let that be the last authoritative one
+		 */
+		if (found_match)
+			acpi_nfit_ars_rescan(acpi_desc);
+	}
+
+	mutex_unlock(&acpi_desc_lock);
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block nfit_mce_dec = {
+	.notifier_call	= nfit_handle_mce,
+};
+
+void nfit_mce_register(void)
+{
+	mce_register_decode_chain(&nfit_mce_dec);
+}
+
+void nfit_mce_unregister(void)
+{
+	mce_unregister_decode_chain(&nfit_mce_dec);
+}
diff --git a/drivers/acpi/nfit/nfit.h b/drivers/acpi/nfit/nfit.h
index 6ecf337c97aa..ba6074a06958 100644
--- a/drivers/acpi/nfit/nfit.h
+++ b/drivers/acpi/nfit/nfit.h
@@ -16,6 +16,7 @@
 #define __NFIT_H__
 #include <linux/workqueue.h>
 #include <linux/libnvdimm.h>
+#include <linux/ndctl.h>
 #include <linux/types.h>
 #include <linux/uuid.h>
 #include <linux/acpi.h>
@@ -148,6 +149,7 @@ struct acpi_nfit_desc {
 	struct nd_cmd_ars_status *ars_status;
 	size_t ars_status_size;
 	struct work_struct work;
+	struct list_head list;
 	struct kernfs_node *scrub_count_state;
 	unsigned int scrub_count;
 	unsigned int cancel:1;
@@ -187,6 +189,24 @@ struct nfit_blk {
 	u32 dimm_flags;
 };
 
+extern struct list_head acpi_descs;
+extern struct mutex acpi_desc_lock;
+int acpi_nfit_ars_rescan(struct acpi_nfit_desc *acpi_desc);
+
+#ifdef CONFIG_X86_MCE
+void nfit_mce_register(void);
+void nfit_mce_unregister(void);
+#else
+static inline void nfit_mce_register(void)
+{
+}
+static inline void nfit_mce_unregister(void)
+{
+}
+#endif
+
+int nfit_spa_type(struct acpi_nfit_system_address *spa);
+
 static inline struct acpi_nfit_memory_map *__to_nfit_memdev(
 		struct nfit_mem *nfit_mem)
 {
diff --git a/tools/testing/nvdimm/Kbuild b/tools/testing/nvdimm/Kbuild
index 0dca8ff7557b..ad6dd0543019 100644
--- a/tools/testing/nvdimm/Kbuild
+++ b/tools/testing/nvdimm/Kbuild
@@ -30,6 +30,7 @@ obj-$(CONFIG_DEV_DAX) += dax.o
 obj-$(CONFIG_DEV_DAX_PMEM) += dax_pmem.o
 
 nfit-y := $(ACPI_SRC)/core.o
+nfit-$(CONFIG_X86_MCE) += $(ACPI_SRC)/mce.o
 nfit-y += config_check.o
 
 nd_pmem-y := $(NVDIMM_SRC)/pmem.o

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web