[PATCH v2 0/2] papr: Implement initial support for injecting smart errors

[PATCH v2 0/2] papr: Implement initial support for injecting smart errors

[Shivaprasad G Bhat]

From: Vaibhav Jain <vaibhav@linux.ibm.com>

Changes since v1:
Link: https://patchwork.kernel.org/project/linux-nvdimm/cover/20210712173132.1205192-1-vaibhav@linux.ibm.com/
* Minor update to patch description
* The changes are based on the new kernel patch [1]

The patch series implements limited support for injecting smart errors for PAPR
NVDIMMs via ndctl-inject-smart(1) command. SMART errors are emulating in
papr_scm module as presently PAPR doesn't support injecting smart errors on an
NVDIMM. Currently support for injecting 'fatal' health state and 'dirty'
shutdown state is implemented. With the proposed ndctl patched and with
corresponding kernel patch [1] following command flow is expected:

$ sudo ndctl list -DH -d nmem0
...
      "health_state":"ok",
      "shutdown_state":"clean",
...
 # inject unsafe shutdown and fatal health error
$ sudo ndctl inject-smart nmem0 -Uf
...
      "health_state":"fatal",
      "shutdown_state":"dirty",
...
 # uninject all errors
$ sudo ndctl inject-smart nmem0 -N
...
      "health_state":"ok",
      "shutdown_state":"clean",
...

Structure of the patch series
=============================

* First patch updates 'inject-smart' code to not always assume support for
  injecting all smart-errors. It also updates 'intel.c' to explicitly indicate
  the type of smart-inject errors supported.

* Update 'papr.c' to add support for injecting smart 'fatal' health and
  'dirty-shutdown' errors.

[1] : https://patchwork.kernel.org/project/linux-nvdimm/patch/163091917031.334.16212158243308361834.stgit@82313cf9f602/
---

Vaibhav Jain (2):
      libndctl, intel: Indicate supported smart-inject types
      libndctl/papr: Add limited support for inject-smart


 ndctl/inject-smart.c  | 33 ++++++++++++++++++-----
 ndctl/lib/intel.c     |  7 ++++-
 ndctl/lib/papr.c      | 61 +++++++++++++++++++++++++++++++++++++++++++
 ndctl/lib/papr_pdsm.h | 17 ++++++++++++
 ndctl/libndctl.h      |  8 ++++++
 5 files changed, 118 insertions(+), 8 deletions(-)

--
Signature

[PATCH v2 1/2] libndctl, intel: Indicate supported smart-inject types

[Shivaprasad G Bhat]

From: Vaibhav Jain <vaibhav@linux.ibm.com>

Presently the inject-smart code assumes support for injecting all
smart-errors namely media-temperature, controller-temperature,
spares-remaining, fatal-health and unsafe-shutdown. This assumption
may break in case of other non-Intel NVDIMM types namely PAPR NVDIMMs
which presently only have support for injecting unsafe-shutdown and
fatal health events.

Trying to inject-smart errors on PAPR NVDIMMs causes problems as
smart_inject() prematurely exits when trying to inject
media-temperature smart-error errors out.

To fix this issue the patch proposes extending the definition of
dimm_op 'smart_inject_supported' to return bitmap of flags indicating
the type of smart-error injections supported by an NVDIMM. These types
are indicated by the newly introduced defines ND_SMART_INJECT_* . A
dimm-ops provide can return an bitmap composed of these flags back
from its implementation of 'smart_inject_supported' to indicate to
dimm_inject_smart() which type of smart-error injection it
supports. In case of an error the dimm-op is still expected to return
a negative error code back to the caller.

The patch updates intel_dimm_smart_inject_supported() to return a
bitmap composed of all ND_SMART_INJECT_* flags to indicate support for
all smart-error types.

Finally the patch also updates smart_inject() to test for specific
ND_START_INJECT_* flags before sending a smart-inject command via
dimm-provider.

Signed-off-by: Vaibhav Jain <vaibhav@linux.ibm.com>
---
Changelog:

Since v1:
Link: https://lore.kernel.org/nvdimm/20210712173132.1205192-2-vaibhav@linux.ibm.com
* Minor update to patch description

 ndctl/inject-smart.c |   33 ++++++++++++++++++++++++++-------
 ndctl/lib/intel.c    |    7 ++++++-
 ndctl/libndctl.h     |    8 ++++++++
 3 files changed, 40 insertions(+), 8 deletions(-)

diff --git a/ndctl/inject-smart.c b/ndctl/inject-smart.c
index 9077bca2..ef0620f5 100644
--- a/ndctl/inject-smart.c
+++ b/ndctl/inject-smart.c
@@ -393,18 +393,26 @@ out:
 	} \
 }
 
-static int smart_inject(struct ndctl_dimm *dimm)
+static int smart_inject(struct ndctl_dimm *dimm, unsigned int inject_types)
 {
 	const char *name = ndctl_dimm_get_devname(dimm);
 	struct ndctl_cmd *si_cmd = NULL;
 	int rc = -EOPNOTSUPP;
 
-	send_inject_val(media_temperature)
-	send_inject_val(ctrl_temperature)
-	send_inject_val(spares)
-	send_inject_bool(fatal)
-	send_inject_bool(unsafe_shutdown)
+	if (inject_types & ND_SMART_INJECT_MEDIA_TEMPERATURE)
+		send_inject_val(media_temperature);
 
+	if (inject_types & ND_SMART_INJECT_CTRL_TEMPERATURE)
+		send_inject_val(ctrl_temperature);
+
+	if (inject_types & ND_SMART_INJECT_SPARES_REMAINING)
+		send_inject_val(spares);
+
+	if (inject_types & ND_SMART_INJECT_HEALTH_STATE)
+		send_inject_bool(fatal);
+
+	if (inject_types & ND_SMART_INJECT_UNCLEAN_SHUTDOWN)
+		send_inject_bool(unsafe_shutdown);
 out:
 	ndctl_cmd_unref(si_cmd);
 	return rc;
@@ -415,8 +423,10 @@ static int dimm_inject_smart(struct ndctl_dimm *dimm)
 	struct json_object *jhealth;
 	struct json_object *jdimms;
 	struct json_object *jdimm;
+	unsigned int supported_types;
 	int rc;
 
+	/* Get supported smart injection types */
 	rc = ndctl_dimm_smart_inject_supported(dimm);
 	switch (rc) {
 	case -ENOTTY:
@@ -431,6 +441,15 @@ static int dimm_inject_smart(struct ndctl_dimm *dimm)
 		error("%s: smart injection not supported by either platform firmware or the kernel.",
 			ndctl_dimm_get_devname(dimm));
 		return rc;
+	default:
+		if (rc < 0) {
+			error("%s: Unknown error %d while checking for smart injection support",
+			      ndctl_dimm_get_devname(dimm), rc);
+			return rc;
+		}
+		/* Assigning to an unsigned type since rc < 0 */
+		supported_types = rc;
+		break;
 	}
 
 	if (sctx.op_mask & (1 << OP_SET)) {
@@ -439,7 +458,7 @@ static int dimm_inject_smart(struct ndctl_dimm *dimm)
 			goto out;
 	}
 	if (sctx.op_mask & (1 << OP_INJECT)) {
-		rc = smart_inject(dimm);
+		rc = smart_inject(dimm, supported_types);
 		if (rc)
 			goto out;
 	}
diff --git a/ndctl/lib/intel.c b/ndctl/lib/intel.c
index a3df26e6..13148545 100644
--- a/ndctl/lib/intel.c
+++ b/ndctl/lib/intel.c
@@ -455,7 +455,12 @@ static int intel_dimm_smart_inject_supported(struct ndctl_dimm *dimm)
 		return -EIO;
 	}
 
-	return 0;
+	/* Indicate all smart injection types are supported */
+	return ND_SMART_INJECT_SPARES_REMAINING |
+		ND_SMART_INJECT_MEDIA_TEMPERATURE |
+		ND_SMART_INJECT_CTRL_TEMPERATURE |
+		ND_SMART_INJECT_HEALTH_STATE |
+		ND_SMART_INJECT_UNCLEAN_SHUTDOWN;
 }
 
 static const char *intel_cmd_desc(int fn)
diff --git a/ndctl/libndctl.h b/ndctl/libndctl.h
index cdadd5fd..54539ac7 100644
--- a/ndctl/libndctl.h
+++ b/ndctl/libndctl.h
@@ -69,6 +69,13 @@ extern "C" {
 #define ND_EVENT_HEALTH_STATE		(1 << 3)
 #define ND_EVENT_UNCLEAN_SHUTDOWN	(1 << 4)
 
+/* Flags indicating support for various smart injection types */
+#define ND_SMART_INJECT_SPARES_REMAINING	(1 << 0)
+#define ND_SMART_INJECT_MEDIA_TEMPERATURE	(1 << 1)
+#define ND_SMART_INJECT_CTRL_TEMPERATURE	(1 << 2)
+#define ND_SMART_INJECT_HEALTH_STATE		(1 << 3)
+#define ND_SMART_INJECT_UNCLEAN_SHUTDOWN	(1 << 4)
+
 size_t ndctl_min_namespace_size(void);
 size_t ndctl_sizeof_namespace_index(void);
 size_t ndctl_sizeof_namespace_label(void);
@@ -310,6 +317,7 @@ int ndctl_cmd_smart_inject_spares(struct ndctl_cmd *cmd, bool enable,
 		unsigned int spares);
 int ndctl_cmd_smart_inject_fatal(struct ndctl_cmd *cmd, bool enable);
 int ndctl_cmd_smart_inject_unsafe_shutdown(struct ndctl_cmd *cmd, bool enable);
+/* Returns a bitmap of ND_SMART_INJECT_* supported */
 int ndctl_dimm_smart_inject_supported(struct ndctl_dimm *dimm);
 
 struct ndctl_cmd *ndctl_dimm_cmd_new_vendor_specific(struct ndctl_dimm *dimm,

[PATCH v2 2/2] libndctl/papr: Add limited support for inject-smart

[Shivaprasad G Bhat]

From: Vaibhav Jain <vaibhav@linux.ibm.com>

Implements support for ndctl inject-smart command by providing an
implementation of 'smart_inject*' dimm-ops callbacks. Presently only
support for injecting unsafe-shutdown and fatal-health states is
available.

The patch also introduce various PAPR PDSM structures that are used to
communicate the inject-smart errors to the papr_scm kernel
module. This is done via SMART_INJECT PDSM which sends a payload of
type 'struct nd_papr_pdsm_smart_inject'.

With the patch following output from ndctl inject-smart command is
expected for PAPR NVDIMMs:

$ sudo ndctl inject-smart -fU nmem0
[
  {
    "dev":"nmem0",
    "flag_failed_flush":true,
    "flag_smart_event":true,
    "health":{
      "health_state":"fatal",
      "shutdown_state":"dirty",
      "shutdown_count":0
    }
  }
]

$ sudo ndctl inject-smart -N nmem0
[
  {
    "dev":"nmem0",
    "health":{
      "health_state":"ok",
      "shutdown_state":"clean",
      "shutdown_count":0
    }
  }
]

The patch depends on the kernel PAPR PDSM implementation for
PDSM_SMART_INJECT posted at [1].

[1] : https://patchwork.kernel.org/project/linux-nvdimm/patch/163091917031.334.16212158243308361834.stgit@82313cf9f602/
Signed-off-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Signed-off-by: Vaibhav Jain <vaibhav@linux.ibm.com>
---
Changelog:

Since v1:
Link: https://lore.kernel.org/nvdimm/20210712173132.1205192-3-vaibhav@linux.ibm.com/
* Updates to patch description.

 ndctl/lib/papr.c      |   61 +++++++++++++++++++++++++++++++++++++++++++++++++
 ndctl/lib/papr_pdsm.h |   17 ++++++++++++++
 2 files changed, 78 insertions(+)

diff --git a/ndctl/lib/papr.c b/ndctl/lib/papr.c
index 42ff200d..b797e1e5 100644
--- a/ndctl/lib/papr.c
+++ b/ndctl/lib/papr.c
@@ -221,6 +221,41 @@ static unsigned int papr_smart_get_shutdown_state(struct ndctl_cmd *cmd)
 	return health.dimm_bad_shutdown;
 }
 
+static int papr_smart_inject_supported(struct ndctl_dimm *dimm)
+{
+	if (!ndctl_dimm_is_cmd_supported(dimm, ND_CMD_CALL))
+		return -EOPNOTSUPP;
+
+	if (!test_dimm_dsm(dimm, PAPR_PDSM_SMART_INJECT))
+		return -EIO;
+
+	return ND_SMART_INJECT_HEALTH_STATE | ND_SMART_INJECT_UNCLEAN_SHUTDOWN;
+}
+
+static int papr_smart_inject_valid(struct ndctl_cmd *cmd)
+{
+	if (cmd->type != ND_CMD_CALL ||
+	    to_pdsm(cmd)->cmd_status != 0 ||
+	    to_pdsm_cmd(cmd) != PAPR_PDSM_SMART_INJECT)
+		return -EINVAL;
+
+	return 0;
+}
+
+static struct ndctl_cmd *papr_new_smart_inject(struct ndctl_dimm *dimm)
+{
+	struct ndctl_cmd *cmd;
+
+	cmd = allocate_cmd(dimm, PAPR_PDSM_SMART_INJECT,
+			sizeof(struct nd_papr_pdsm_smart_inject));
+	if (!cmd)
+		return NULL;
+	/* Set the input payload size */
+	to_ndcmd(cmd)->nd_size_in = ND_PDSM_HDR_SIZE +
+		sizeof(struct nd_papr_pdsm_smart_inject);
+	return cmd;
+}
+
 static unsigned int papr_smart_get_life_used(struct ndctl_cmd *cmd)
 {
 	struct nd_papr_pdsm_health health;
@@ -255,11 +290,37 @@ static unsigned int papr_smart_get_shutdown_count(struct ndctl_cmd *cmd)
 
 	return (health.extension_flags & PDSM_DIMM_DSC_VALID) ?
 		(health.dimm_dsc) : 0;
+}
+
+static int papr_cmd_smart_inject_fatal(struct ndctl_cmd *cmd, bool enable)
+{
+	if (papr_smart_inject_valid(cmd) < 0)
+		return -EINVAL;
+
+	to_payload(cmd)->inject.flags |= PDSM_SMART_INJECT_HEALTH_FATAL;
+	to_payload(cmd)->inject.fatal_enable = enable;
 
+	return 0;
+}
+
+static int papr_cmd_smart_inject_unsafe_shutdown(struct ndctl_cmd *cmd,
+						 bool enable)
+{
+	if (papr_smart_inject_valid(cmd) < 0)
+		return -EINVAL;
+
+	to_payload(cmd)->inject.flags |= PDSM_SMART_INJECT_BAD_SHUTDOWN;
+	to_payload(cmd)->inject.unsafe_shutdown_enable = enable;
+
+	return 0;
 }
 
 struct ndctl_dimm_ops * const papr_dimm_ops = &(struct ndctl_dimm_ops) {
 	.cmd_is_supported = papr_cmd_is_supported,
+	.new_smart_inject = papr_new_smart_inject,
+	.smart_inject_supported = papr_smart_inject_supported,
+	.smart_inject_fatal = papr_cmd_smart_inject_fatal,
+	.smart_inject_unsafe_shutdown = papr_cmd_smart_inject_unsafe_shutdown,
 	.smart_get_flags = papr_smart_get_flags,
 	.get_firmware_status =  papr_get_firmware_status,
 	.xlat_firmware_status = papr_xlat_firmware_status,
diff --git a/ndctl/lib/papr_pdsm.h b/ndctl/lib/papr_pdsm.h
index f45b1e40..20ac20f8 100644
--- a/ndctl/lib/papr_pdsm.h
+++ b/ndctl/lib/papr_pdsm.h
@@ -121,12 +121,29 @@ struct nd_papr_pdsm_health {
 enum papr_pdsm {
 	PAPR_PDSM_MIN = 0x0,
 	PAPR_PDSM_HEALTH,
+	PAPR_PDSM_SMART_INJECT,
 	PAPR_PDSM_MAX,
 };
+/* Flags for injecting specific smart errors */
+#define PDSM_SMART_INJECT_HEALTH_FATAL		(1 << 0)
+#define PDSM_SMART_INJECT_BAD_SHUTDOWN		(1 << 1)
+
+struct nd_papr_pdsm_smart_inject {
+	union {
+		struct {
+			/* One or more of PDSM_SMART_INJECT_ */
+			__u32 flags;
+			__u8 fatal_enable;
+			__u8 unsafe_shutdown_enable;
+		};
+		__u8 buf[ND_PDSM_PAYLOAD_MAX_SIZE];
+	};
+};
 
 /* Maximal union that can hold all possible payload types */
 union nd_pdsm_payload {
 	struct nd_papr_pdsm_health health;
+	struct nd_papr_pdsm_smart_inject inject;
 	__u8 buf[ND_PDSM_PAYLOAD_MAX_SIZE];
 } __attribute__((packed));