From: vishal@kernel.org
To: linux-nvdimm@lists.01.org
Cc: Vishal Verma <vishal.l.verma@intel.com>,
linux-acpi@vger.kernel.org, linux-fsdevel@vger.kernel.org,
Dan Williams <dan.j.williams@intel.com>,
Ross Zwisler <ross.zwisler@linux.intel.com>,
Jeff Moyer <jmoyer@redhat.com>,
Linda Knippers <linda.knippers@hpe.com>
Subject: [PATCH 2/3] libnvdimm: Add a poison list
Date: Sun, 20 Dec 2015 02:18:41 -0700 [thread overview]
Message-ID: <1450603122-7205-3-git-send-email-vishal@kernel.org> (raw)
In-Reply-To: <1450603122-7205-1-git-send-email-vishal@kernel.org>
From: Vishal Verma <vishal.l.verma@intel.com>
During region creation, perform Address Range Scrubs (ARS) for the SPA
(System Physical Address) ranges to retrieve known poison locations from
firmware. Add a new data structure 'nd_poison' which is used as a list
in nvdimm_bus to store these poison locations for subsequent consumption
by the pmem driver.
Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
---
drivers/acpi/nfit.c | 193 ++++++++++++++++++++++++++++++++++++++++++++++
drivers/nvdimm/core.c | 63 +++++++++++++++
drivers/nvdimm/nd-core.h | 1 +
drivers/nvdimm/nd.h | 6 ++
include/linux/libnvdimm.h | 2 +
5 files changed, 265 insertions(+)
diff --git a/drivers/acpi/nfit.c b/drivers/acpi/nfit.c
index e7ed39b..bd8726d 100644
--- a/drivers/acpi/nfit.c
+++ b/drivers/acpi/nfit.c
@@ -15,6 +15,7 @@
#include <linux/module.h>
#include <linux/mutex.h>
#include <linux/ndctl.h>
+#include <linux/delay.h>
#include <linux/list.h>
#include <linux/acpi.h>
#include <linux/sort.h>
@@ -1473,6 +1474,195 @@ static void acpi_nfit_blk_region_disable(struct nvdimm_bus *nvdimm_bus,
/* devm will free nfit_blk */
}
+static int ars_get_cap(struct nvdimm_bus_descriptor *nd_desc,
+ struct nd_cmd_ars_cap *cmd, u64 addr, u64 length)
+{
+ cmd->address = addr;
+ cmd->length = length;
+
+ return nd_desc->ndctl(nd_desc, NULL, ND_CMD_ARS_CAP, cmd,
+ sizeof(*cmd));
+}
+
+static int ars_do_start(struct nvdimm_bus_descriptor *nd_desc,
+ struct nd_cmd_ars_start *cmd, u64 addr, u64 length)
+{
+ int rc;
+
+ cmd->address = addr;
+ cmd->length = length;
+ cmd->type = ND_ARS_PERSISTENT;
+
+ while (1) {
+ rc = nd_desc->ndctl(nd_desc, NULL, ND_CMD_ARS_START, cmd,
+ sizeof(*cmd));
+ if (rc)
+ return rc;
+ switch (cmd->status) {
+ case 0:
+ return 0;
+ case 1:
+ /* ARS unsupported, but we should never get here */
+ return 0;
+ case 2:
+ return -EINVAL;
+ case 3:
+ /* ARS is in progress */
+ msleep(1000);
+ break;
+ default:
+ return -ENXIO;
+ }
+ }
+}
+
+static int ars_get_status(struct nvdimm_bus_descriptor *nd_desc,
+ struct nd_cmd_ars_status *cmd)
+{
+ int rc;
+
+ while (1) {
+ rc = nd_desc->ndctl(nd_desc, NULL, ND_CMD_ARS_STATUS, cmd,
+ sizeof(*cmd));
+ if (rc || cmd->status & 0xffff)
+ return -ENXIO;
+
+ /* Check extended status (Upper two bytes) */
+ switch (cmd->status & 0xffff0000) {
+ case 0:
+ return 0;
+ case 1:
+ /* ARS is in progress */
+ msleep(1000);
+ break;
+ case 2:
+ /* No ARS performed for the current boot */
+ default:
+ return -ENXIO;
+ }
+ }
+}
+
+static int ars_status_process_records(struct nvdimm_bus *nvdimm_bus,
+ struct nd_cmd_ars_status *ars_status, u64 start)
+{
+ int rc;
+ u32 i;
+
+ /*
+ * The address field returned by ars_status should be either
+ * less than or equal to the address we last started ARS for.
+ * The (start, length) returned by ars_status should also have
+ * non-zero overlap with the range we started ARS for.
+ * If this is not the case, bail.
+ */
+ if (ars_status->address > start ||
+ (ars_status->address + ars_status->length < start))
+ return -ENXIO;
+
+ for (i = 0; i < ars_status->num_records; i++) {
+ rc = nvdimm_bus_add_poison(nvdimm_bus,
+ ars_status->records[i].err_address,
+ ars_status->records[i].length);
+ if (rc)
+ return rc;
+ }
+
+ return 0;
+}
+
+static int acpi_nfit_find_poison(struct acpi_nfit_desc *acpi_desc,
+ struct nd_region_desc *ndr_desc)
+{
+ struct nvdimm_bus_descriptor *nd_desc = &acpi_desc->nd_desc;
+ struct nvdimm_bus *nvdimm_bus = acpi_desc->nvdimm_bus;
+ struct nd_cmd_ars_status *ars_status = NULL;
+ struct nd_cmd_ars_start *ars_start = NULL;
+ struct nd_cmd_ars_cap *ars_cap = NULL;
+ u64 start, len, cur, remaining;
+ int rc;
+
+ ars_cap = kzalloc(sizeof(*ars_cap), GFP_KERNEL);
+ if (!ars_cap)
+ return -ENOMEM;
+
+ start = ndr_desc->res->start;
+ len = ndr_desc->res->end - ndr_desc->res->start + 1;
+
+ rc = ars_get_cap(nd_desc, ars_cap, start, len);
+ if (rc)
+ goto out;
+
+ /* If ARS is unsupported, skip this but continue initialization */
+ if (ars_cap->status & 0xffff) {
+ dev_warn(acpi_desc->dev,
+ "ARS unsupported, won't construct an error list\n");
+ goto out;
+ }
+
+ /*
+ * Check if a full-range ARS has been run. If so, use those results
+ * without having to start a new ARS.
+ */
+ ars_status = kzalloc(ars_cap->max_ars_out + sizeof(*ars_status),
+ GFP_KERNEL);
+ if (!ars_status) {
+ rc = -ENOMEM;
+ goto out;
+ }
+
+ rc = ars_get_status(nd_desc, ars_status);
+ if (rc)
+ goto out;
+
+ if (ars_status->address <= start &&
+ (ars_status->address + ars_status->length >= start + len)) {
+ rc = ars_status_process_records(nvdimm_bus, ars_status, start);
+ goto out;
+ }
+
+ /*
+ * ARS_STATUS can overflow if the number of poison entries found is
+ * greater than the maximum buffer size (ars_cap->max_ars_out)
+ * To detect overflow, check if the length field of ars_status
+ * is less than the length we supplied. If so, process the
+ * error entries we got, adjust the start point, and start again
+ */
+ ars_start = kzalloc(sizeof(*ars_start), GFP_KERNEL);
+ if (!ars_start)
+ return -ENOMEM;
+
+ cur = start;
+ remaining = len;
+ do {
+ u64 done, end;
+
+ rc = ars_do_start(nd_desc, ars_start, cur, remaining);
+ if (rc)
+ goto out;
+
+ rc = ars_get_status(nd_desc, ars_status);
+ if (rc)
+ goto out;
+
+ rc = ars_status_process_records(nvdimm_bus, ars_status, cur);
+ if (rc)
+ goto out;
+
+ end = min(cur + remaining,
+ ars_status->address + ars_status->length);
+ done = end - cur;
+ cur += done;
+ remaining -= done;
+ } while (remaining);
+
+ out:
+ kfree(ars_cap);
+ kfree(ars_start);
+ kfree(ars_status);
+ return rc;
+}
+
static int acpi_nfit_init_mapping(struct acpi_nfit_desc *acpi_desc,
struct nd_mapping *nd_mapping, struct nd_region_desc *ndr_desc,
struct acpi_nfit_memory_map *memdev,
@@ -1585,6 +1775,9 @@ static int acpi_nfit_register_region(struct acpi_nfit_desc *acpi_desc,
nvdimm_bus = acpi_desc->nvdimm_bus;
if (nfit_spa_type(spa) == NFIT_SPA_PM) {
+ rc = acpi_nfit_find_poison(acpi_desc, ndr_desc);
+ if (rc)
+ return rc;
if (!nvdimm_pmem_region_create(nvdimm_bus, ndr_desc))
return -ENOMEM;
} else if (nfit_spa_type(spa) == NFIT_SPA_VOLATILE) {
diff --git a/drivers/nvdimm/core.c b/drivers/nvdimm/core.c
index 82c49bb..8c882c8 100644
--- a/drivers/nvdimm/core.c
+++ b/drivers/nvdimm/core.c
@@ -325,6 +325,7 @@ struct nvdimm_bus *__nvdimm_bus_register(struct device *parent,
if (!nvdimm_bus)
return NULL;
INIT_LIST_HEAD(&nvdimm_bus->list);
+ INIT_LIST_HEAD(&nvdimm_bus->poison_list);
init_waitqueue_head(&nvdimm_bus->probe_wait);
nvdimm_bus->id = ida_simple_get(&nd_ida, 0, 0, GFP_KERNEL);
mutex_init(&nvdimm_bus->reconfig_mutex);
@@ -359,6 +360,67 @@ struct nvdimm_bus *__nvdimm_bus_register(struct device *parent,
}
EXPORT_SYMBOL_GPL(__nvdimm_bus_register);
+struct list_head *nvdimm_bus_get_poison_list(struct nvdimm_bus *nvdimm_bus)
+{
+ return &nvdimm_bus->poison_list;
+}
+EXPORT_SYMBOL_GPL(nvdimm_bus_get_poison_list);
+
+static int __add_poison(struct nvdimm_bus *nvdimm_bus, u64 addr, u64 length)
+{
+ struct nd_poison *pl;
+
+ pl = kzalloc(sizeof(*pl), GFP_KERNEL);
+ if (!pl)
+ return -ENOMEM;
+
+ pl->start = addr;
+ pl->length = length;
+ list_add_tail(&pl->list, &nvdimm_bus->poison_list);
+
+ return 0;
+}
+
+int nvdimm_bus_add_poison(struct nvdimm_bus *nvdimm_bus, u64 addr, u64 length)
+{
+ struct nd_poison *pl;
+
+ if (list_empty(&nvdimm_bus->poison_list))
+ return __add_poison(nvdimm_bus, addr, length);
+
+ /*
+ * There is a chance this is a duplicate, check for those first.
+ * This will be the common case as ARS_STATUS returns all known
+ * errors in the SPA space, and we can't query it per region
+ */
+ list_for_each_entry(pl, &nvdimm_bus->poison_list, list)
+ if (pl->start == addr) {
+ /* If length has changed, update this list entry */
+ if (pl->length != length)
+ pl->length = length;
+ return 0;
+ }
+
+ /*
+ * If not a duplicate or a simple length update, add the entry as is,
+ * as any overlapping ranges will get resolved when the list is consumed
+ * and converted to badblocks
+ */
+ return __add_poison(nvdimm_bus, addr, length);
+}
+EXPORT_SYMBOL_GPL(nvdimm_bus_add_poison);
+
+static void free_poison_list(struct list_head *poison_list)
+{
+ struct nd_poison *pl, *next;
+
+ list_for_each_entry_safe(pl, next, poison_list, list) {
+ list_del(&pl->list);
+ kfree(pl);
+ }
+ list_del_init(poison_list);
+}
+
static int child_unregister(struct device *dev, void *data)
{
/*
@@ -385,6 +447,7 @@ void nvdimm_bus_unregister(struct nvdimm_bus *nvdimm_bus)
nd_synchronize();
device_for_each_child(&nvdimm_bus->dev, NULL, child_unregister);
+ free_poison_list(&nvdimm_bus->poison_list);
nvdimm_bus_destroy_ndctl(nvdimm_bus);
device_unregister(&nvdimm_bus->dev);
diff --git a/drivers/nvdimm/nd-core.h b/drivers/nvdimm/nd-core.h
index 159aed5..29acdaa 100644
--- a/drivers/nvdimm/nd-core.h
+++ b/drivers/nvdimm/nd-core.h
@@ -30,6 +30,7 @@ struct nvdimm_bus {
struct list_head list;
struct device dev;
int id, probe_active;
+ struct list_head poison_list;
struct mutex reconfig_mutex;
};
diff --git a/drivers/nvdimm/nd.h b/drivers/nvdimm/nd.h
index 417e521..ba91fcd 100644
--- a/drivers/nvdimm/nd.h
+++ b/drivers/nvdimm/nd.h
@@ -38,6 +38,12 @@ enum {
#endif
};
+struct nd_poison {
+ u64 start;
+ u64 length;
+ struct list_head list;
+};
+
struct nvdimm_drvdata {
struct device *dev;
int nsindex_size;
diff --git a/include/linux/libnvdimm.h b/include/linux/libnvdimm.h
index 3f021dc..43b82a0 100644
--- a/include/linux/libnvdimm.h
+++ b/include/linux/libnvdimm.h
@@ -116,6 +116,8 @@ static inline struct nd_blk_region_desc *to_blk_region_desc(
}
+struct list_head *nvdimm_bus_get_poison_list(struct nvdimm_bus *nvdimm_bus);
+int nvdimm_bus_add_poison(struct nvdimm_bus *nvdimm_bus, u64 addr, u64 length);
struct nvdimm_bus *__nvdimm_bus_register(struct device *parent,
struct nvdimm_bus_descriptor *nfit_desc, struct module *module);
#define nvdimm_bus_register(parent, desc) \
--
2.5.0
next prev parent reply other threads:[~2015-12-20 9:19 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-12-20 9:18 [PATCH 0/3] Expose known poison in SPA ranges to the block layer vishal
2015-12-20 9:18 ` [PATCH 1/3] nfit_test: Enable DSMs for all test NFITs vishal
2015-12-20 9:18 ` vishal [this message]
2015-12-20 9:18 ` [PATCH 3/3] pmem: Use the poison list to expose badblocks vishal
2015-12-20 9:31 ` kbuild test robot
2015-12-21 1:20 ` Dan Williams
2015-12-21 18:50 ` Verma, Vishal L
2015-12-21 19:10 ` Dan Williams
2015-12-23 20:28 ` Dan Williams
2015-12-23 20:32 ` Verma, Vishal L
2015-12-23 20:38 ` Dan Williams
2015-12-21 18:34 ` [PATCH 0/3] Expose known poison in SPA ranges to the block layer Verma, Vishal L
2015-12-21 21:06 ` Linda Knippers
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1450603122-7205-3-git-send-email-vishal@kernel.org \
--to=vishal@kernel.org \
--cc=dan.j.williams@intel.com \
--cc=jmoyer@redhat.com \
--cc=linda.knippers@hpe.com \
--cc=linux-acpi@vger.kernel.org \
--cc=linux-fsdevel@vger.kernel.org \
--cc=linux-nvdimm@lists.01.org \
--cc=ross.zwisler@linux.intel.com \
--cc=vishal.l.verma@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).