nvdimm.lists.linux.dev archive mirror
 help / color / mirror / Atom feed
* [ndctl PATCH 0/3] daxctl-device fixes
@ 2021-03-31  3:12 Vishal Verma
  2021-03-31  3:12 ` [ndctl PATCH 1/3] daxctl: fail reconfigure-device based on kernel onlining policy Vishal Verma
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: Vishal Verma @ 2021-03-31  3:12 UTC (permalink / raw)
  To: linux-nvdimm; +Cc: Dave Hansen, Chunye Xu

This series contains a couple of fixes:

1/ Check the kernel policy for onlining blocks. If that is set to
'online', and if daxctl is passed --no-online or is onlining to
ZONE_MOVABLE (default), then fail the reconfigure command as we know
we can't satisfy that ask. (Prior to these patches, the reconfigure
would succeed, and we'd print a warning after the fact, which may
result in a surprising experience).

2/ For ndctl disable-namespace (and destroy-namespace), fail the
operation if the dax device is active as system-ram.

Additionally, augment the daxctl-devices.sh unit test to test for both
of these conditions and the expected (fixed) behavior in both cases.

Vishal Verma (3):
  daxctl: fail reconfigure-device based on kernel onlining policy
  libdaxctl: add an API to check if a device is active
  libndctl: check for active system-ram before disabling daxctl devices

 .../daxctl/daxctl-reconfigure-device.txt      | 12 ++++-
 daxctl/lib/libdaxctl-private.h                |  1 +
 daxctl/lib/libdaxctl.c                        | 31 +++++++++++
 ndctl/lib/libndctl.c                          | 25 ++++++++-
 daxctl/libdaxctl.h                            |  2 +
 daxctl/device.c                               | 10 ++++
 daxctl/lib/libdaxctl.sym                      |  6 +++
 test/daxctl-devices.sh                        | 52 +++++++++++++++++++
 8 files changed, 137 insertions(+), 2 deletions(-)

-- 
2.30.2
_______________________________________________
Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org
To unsubscribe send an email to linux-nvdimm-leave@lists.01.org

^ permalink raw reply	[flat|nested] 4+ messages in thread

* [ndctl PATCH 1/3] daxctl: fail reconfigure-device based on kernel onlining policy
  2021-03-31  3:12 [ndctl PATCH 0/3] daxctl-device fixes Vishal Verma
@ 2021-03-31  3:12 ` Vishal Verma
  2021-03-31  3:12 ` [ndctl PATCH 2/3] libdaxctl: add an API to check if a device is active Vishal Verma
  2021-03-31  3:12 ` [ndctl PATCH 3/3] libndctl: check for active system-ram before disabling daxctl devices Vishal Verma
  2 siblings, 0 replies; 4+ messages in thread
From: Vishal Verma @ 2021-03-31  3:12 UTC (permalink / raw)
  To: linux-nvdimm; +Cc: Dave Hansen, Chunye Xu, Dave Hansen

If the kernel has a policy set to auto-online any new memory blocks, we
know that an attempt to reconfigure a device either in ZONE_MOVABLE, or
with the --no-online is going to fail. While we detect this race after
the fact, and print a warning, that is often insufficient as the user
may be forced to reboot to get out of the situation, resulting in an
unpleasant experience.

Detect whether the kernel policy is set to auto-online. If so, fail
device reconfigure operations that we know can't be satisfied. Allow
for overriding this safety check via the -f (--force) option. Update the
man page to talk about this, and the unit test to test for an expected
failure by enabling auto-onlining.

Cc: Dave Hansen <dave.hansen@intel.com>
Reported-by: Chunye Xu <chunye.xu@intel.com>
Reported-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
---
 .../daxctl/daxctl-reconfigure-device.txt      | 12 ++++++-
 daxctl/lib/libdaxctl-private.h                |  1 +
 daxctl/lib/libdaxctl.c                        | 21 +++++++++++
 daxctl/libdaxctl.h                            |  1 +
 daxctl/device.c                               | 10 ++++++
 daxctl/lib/libdaxctl.sym                      |  5 +++
 test/daxctl-devices.sh                        | 36 +++++++++++++++++++
 7 files changed, 85 insertions(+), 1 deletion(-)

diff --git a/Documentation/daxctl/daxctl-reconfigure-device.txt b/Documentation/daxctl/daxctl-reconfigure-device.txt
index ad33eda..f112b3c 100644
--- a/Documentation/daxctl/daxctl-reconfigure-device.txt
+++ b/Documentation/daxctl/daxctl-reconfigure-device.txt
@@ -119,6 +119,10 @@ recommended to use the --no-online option described below. This will abridge
 the device reconfiguration operation to just hotplugging the memory, and
 refrain from then onlining it.
 
+In case daxctl detects that there is a kernel policy to auto-online blocks
+(via /sys/devices/system/memory/auto_online_blocks), then reconfiguring to
+system-ram will result in a failure. This can be overridden with '--force'.
+
 OPTIONS
 -------
 include::region-option.txt[]
@@ -162,12 +166,18 @@ include::movable-options.txt[]
 
 -f::
 --force::
-	When converting from "system-ram" mode to "devdax", it is expected
+	- When converting from "system-ram" mode to "devdax", it is expected
 	that all the memory sections are first made offline. By default,
 	daxctl won't touch online memory. However with this option, attempt
 	to offline the memory on the NUMA node associated with the dax device
 	before converting it back to "devdax" mode.
 
+	- Additionally, if a kernel policy to auto-online blocks is detected,
+	reconfiguration to system-ram fails. With this option, the failure can
+	be overridden to allow reconfiguration regardless of kernel policy.
+	Doing this may result in a successful reconfiguration, but it may
+	not be possible to subsequently offline the memory without a reboot.
+
 
 include::human-option.txt[]
 
diff --git a/daxctl/lib/libdaxctl-private.h b/daxctl/lib/libdaxctl-private.h
index af257fd..ae45311 100644
--- a/daxctl/lib/libdaxctl-private.h
+++ b/daxctl/lib/libdaxctl-private.h
@@ -111,6 +111,7 @@ struct daxctl_memory {
 	char *node_path;
 	unsigned long block_size;
 	enum memory_zones zone;
+	bool auto_online;
 };
 
 
diff --git a/daxctl/lib/libdaxctl.c b/daxctl/lib/libdaxctl.c
index 479e8f6..879f7e6 100644
--- a/daxctl/lib/libdaxctl.c
+++ b/daxctl/lib/libdaxctl.c
@@ -1644,3 +1644,24 @@ DAXCTL_EXPORT int daxctl_memory_is_movable(struct daxctl_memory *mem)
 		return rc;
 	return (mem->zone == MEM_ZONE_MOVABLE) ? 1 : 0;
 }
+
+DAXCTL_EXPORT int daxctl_dev_will_auto_online_memory(struct daxctl_dev *dev)
+{
+	const char *auto_path = "/sys/devices/system/memory/auto_online_blocks";
+	const char *devname = daxctl_dev_get_devname(dev);
+	struct daxctl_ctx *ctx = daxctl_dev_get_ctx(dev);
+	char buf[SYSFS_ATTR_SIZE];
+
+	/*
+	 * If we can't read the policy for some reason, don't fail yet. Assume
+	 * the auto-onlining policy is absent, and carry on. If onlining blocks
+	 * does result in the memory being in an inconsistent state, we have a
+	 * check and warning for it after the fact
+	 */
+	if (sysfs_read_attr(ctx, auto_path, buf) != 0)
+		err(ctx, "%s: Unable to determine auto-online policy: %s\n",
+				devname, strerror(errno));
+
+	/* match both "online" and "online_movable" */
+	return !strncmp(buf, "online", 6);
+}
diff --git a/daxctl/libdaxctl.h b/daxctl/libdaxctl.h
index e82b274..30ab51a 100644
--- a/daxctl/libdaxctl.h
+++ b/daxctl/libdaxctl.h
@@ -71,6 +71,7 @@ int daxctl_dev_disable(struct daxctl_dev *dev);
 int daxctl_dev_enable_devdax(struct daxctl_dev *dev);
 int daxctl_dev_enable_ram(struct daxctl_dev *dev);
 int daxctl_dev_get_target_node(struct daxctl_dev *dev);
+int daxctl_dev_will_auto_online_memory(struct daxctl_dev *dev);
 
 struct daxctl_memory;
 struct daxctl_memory *daxctl_dev_get_memory(struct daxctl_dev *dev);
diff --git a/daxctl/device.c b/daxctl/device.c
index 0721a57..a427b7d 100644
--- a/daxctl/device.c
+++ b/daxctl/device.c
@@ -541,8 +541,18 @@ static int disable_devdax_device(struct daxctl_dev *dev)
 
 static int reconfig_mode_system_ram(struct daxctl_dev *dev)
 {
+	const char *devname = daxctl_dev_get_devname(dev);
 	int rc, skip_enable = 0;
 
+	if (param.no_online || !param.no_movable) {
+		if (!param.force && daxctl_dev_will_auto_online_memory(dev)) {
+			fprintf(stderr,
+				"%s: error: kernel policy will auto-online memory, aborting\n",
+				devname);
+			return -EBUSY;
+		}
+	}
+
 	if (daxctl_dev_is_enabled(dev)) {
 		rc = disable_devdax_device(dev);
 		if (rc < 0)
diff --git a/daxctl/lib/libdaxctl.sym b/daxctl/lib/libdaxctl.sym
index a4e1684..892e393 100644
--- a/daxctl/lib/libdaxctl.sym
+++ b/daxctl/lib/libdaxctl.sym
@@ -91,3 +91,8 @@ global:
 	daxctl_mapping_get_size;
 	daxctl_dev_set_mapping;
 } LIBDAXCTL_7;
+
+LIBDAXCTL_9 {
+global:
+	daxctl_dev_will_auto_online_memory;
+} LIBDAXCTL_8;
diff --git a/test/daxctl-devices.sh b/test/daxctl-devices.sh
index 496e4f2..eed5906 100755
--- a/test/daxctl-devices.sh
+++ b/test/daxctl-devices.sh
@@ -64,6 +64,26 @@ daxctl_get_mode()
 	"$DAXCTL" list -d "$1" | jq -er '.[].mode'
 }
 
+set_online_policy()
+{
+	echo "online" > /sys/devices/system/memory/auto_online_blocks
+}
+
+unset_online_policy()
+{
+	echo "offline" > /sys/devices/system/memory/auto_online_blocks
+}
+
+save_online_policy()
+{
+	saved_policy="$(cat /sys/devices/system/memory/auto_online_blocks)"
+}
+
+restore_online_policy()
+{
+	echo "$saved_policy" > /sys/devices/system/memory/auto_online_blocks
+}
+
 daxctl_test()
 {
 	local daxdev
@@ -71,6 +91,9 @@ daxctl_test()
 	daxdev=$(daxctl_get_dev "$testdev")
 	test -n "$daxdev"
 
+	# these tests need to run with kernel onlining policy turned off
+	save_online_policy
+	unset_online_policy
 	"$DAXCTL" reconfigure-device -N -m system-ram "$daxdev"
 	[[ $(daxctl_get_mode "$daxdev") == "system-ram" ]]
 	"$DAXCTL" online-memory "$daxdev"
@@ -81,6 +104,19 @@ daxctl_test()
 	[[ $(daxctl_get_mode "$daxdev") == "system-ram" ]]
 	"$DAXCTL" reconfigure-device -f -m devdax "$daxdev"
 	[[ $(daxctl_get_mode "$daxdev") == "devdax" ]]
+
+	# this tests for reconfiguration failure if an online-policy is set
+	set_online_policy
+	: "This command is expected to fail:"
+	if ! "$DAXCTL" reconfigure-device -N -m system-ram "$daxdev"; then
+		echo "reconfigure failed as expected"
+	else
+		echo "reconfigure succeded, expected failure"
+		restore_online_policy
+		return 1
+	fi
+
+	restore_online_policy
 }
 
 find_testdev
-- 
2.30.2
_______________________________________________
Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org
To unsubscribe send an email to linux-nvdimm-leave@lists.01.org

^ permalink raw reply	[flat|nested] 4+ messages in thread

* [ndctl PATCH 2/3] libdaxctl: add an API to check if a device is active
  2021-03-31  3:12 [ndctl PATCH 0/3] daxctl-device fixes Vishal Verma
  2021-03-31  3:12 ` [ndctl PATCH 1/3] daxctl: fail reconfigure-device based on kernel onlining policy Vishal Verma
@ 2021-03-31  3:12 ` Vishal Verma
  2021-03-31  3:12 ` [ndctl PATCH 3/3] libndctl: check for active system-ram before disabling daxctl devices Vishal Verma
  2 siblings, 0 replies; 4+ messages in thread
From: Vishal Verma @ 2021-03-31  3:12 UTC (permalink / raw)
  To: linux-nvdimm; +Cc: Dave Hansen, Chunye Xu

Add an API to check whether a daxctl device is active in system-ram
mode. This would be used from libndctl during
ndctl_namespace_disable_safe(), so that we don't disable/destroy an
underlying namespace while the memory is active and online.

Reported-by: Chunye Xu <chunye.xu@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
---
 daxctl/lib/libdaxctl.c   | 10 ++++++++++
 daxctl/libdaxctl.h       |  1 +
 daxctl/lib/libdaxctl.sym |  1 +
 3 files changed, 12 insertions(+)

diff --git a/daxctl/lib/libdaxctl.c b/daxctl/lib/libdaxctl.c
index 879f7e6..860bd9c 100644
--- a/daxctl/lib/libdaxctl.c
+++ b/daxctl/lib/libdaxctl.c
@@ -1665,3 +1665,13 @@ DAXCTL_EXPORT int daxctl_dev_will_auto_online_memory(struct daxctl_dev *dev)
 	/* match both "online" and "online_movable" */
 	return !strncmp(buf, "online", 6);
 }
+
+DAXCTL_EXPORT int daxctl_dev_has_online_memory(struct daxctl_dev *dev)
+{
+	struct daxctl_memory *mem = daxctl_dev_get_memory(dev);
+
+	if (mem)
+		return daxctl_memory_is_online(mem);
+	else
+		return 0;
+}
diff --git a/daxctl/libdaxctl.h b/daxctl/libdaxctl.h
index 30ab51a..683ae9c 100644
--- a/daxctl/libdaxctl.h
+++ b/daxctl/libdaxctl.h
@@ -72,6 +72,7 @@ int daxctl_dev_enable_devdax(struct daxctl_dev *dev);
 int daxctl_dev_enable_ram(struct daxctl_dev *dev);
 int daxctl_dev_get_target_node(struct daxctl_dev *dev);
 int daxctl_dev_will_auto_online_memory(struct daxctl_dev *dev);
+int daxctl_dev_has_online_memory(struct daxctl_dev *dev);
 
 struct daxctl_memory;
 struct daxctl_memory *daxctl_dev_get_memory(struct daxctl_dev *dev);
diff --git a/daxctl/lib/libdaxctl.sym b/daxctl/lib/libdaxctl.sym
index 892e393..a13e93d 100644
--- a/daxctl/lib/libdaxctl.sym
+++ b/daxctl/lib/libdaxctl.sym
@@ -95,4 +95,5 @@ global:
 LIBDAXCTL_9 {
 global:
 	daxctl_dev_will_auto_online_memory;
+	daxctl_dev_has_online_memory;
 } LIBDAXCTL_8;
-- 
2.30.2
_______________________________________________
Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org
To unsubscribe send an email to linux-nvdimm-leave@lists.01.org

^ permalink raw reply	[flat|nested] 4+ messages in thread

* [ndctl PATCH 3/3] libndctl: check for active system-ram before disabling daxctl devices
  2021-03-31  3:12 [ndctl PATCH 0/3] daxctl-device fixes Vishal Verma
  2021-03-31  3:12 ` [ndctl PATCH 1/3] daxctl: fail reconfigure-device based on kernel onlining policy Vishal Verma
  2021-03-31  3:12 ` [ndctl PATCH 2/3] libdaxctl: add an API to check if a device is active Vishal Verma
@ 2021-03-31  3:12 ` Vishal Verma
  2 siblings, 0 replies; 4+ messages in thread
From: Vishal Verma @ 2021-03-31  3:12 UTC (permalink / raw)
  To: linux-nvdimm; +Cc: Dave Hansen, Chunye Xu

Teach ndctl_namespace_disable_safe() to look at the state of a
daxctl_dev with respect to whether it is active in 'system-ram' mode
before disabling it. This is similar to checking whether a filesystem is
actively mounted on a namespace before disabling it.

Without this, libndctl would happily disable a devdax namespace while the
device was active in system-ram mode. If the namespace was subsequently
also destroyed, this would leave the memory without any sort of a
'handle' to perform any subsequent operation on it, and the system would
have to be rebooted to get out of this situation.

Reported-by: Chunye Xu <chunye.xu@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Signed-off-by: Vishal Verma <vishal.l.verma@intel.com>
---
 ndctl/lib/libndctl.c   | 25 ++++++++++++++++++++++++-
 test/daxctl-devices.sh | 16 ++++++++++++++++
 2 files changed, 40 insertions(+), 1 deletion(-)

diff --git a/ndctl/lib/libndctl.c b/ndctl/lib/libndctl.c
index 2f6d806..2eda56c 100644
--- a/ndctl/lib/libndctl.c
+++ b/ndctl/lib/libndctl.c
@@ -4593,21 +4593,40 @@ NDCTL_EXPORT int ndctl_namespace_disable_invalidate(struct ndctl_namespace *ndns
 	return ndctl_namespace_disable(ndns);
 }
 
+static int ndctl_dax_has_active_memory(struct ndctl_dax *dax)
+{
+	struct daxctl_region *dax_region;
+	struct daxctl_dev *dax_dev;
+
+	dax_region = ndctl_dax_get_daxctl_region(dax);
+	if (!dax_region)
+		return 0;
+
+	daxctl_dev_foreach(dax_region, dax_dev)
+		if (daxctl_dev_has_online_memory(dax_dev))
+			return 1;
+
+	return 0;
+}
+
 NDCTL_EXPORT int ndctl_namespace_disable_safe(struct ndctl_namespace *ndns)
 {
 	const char *devname = ndctl_namespace_get_devname(ndns);
 	struct ndctl_ctx *ctx = ndctl_namespace_get_ctx(ndns);
 	struct ndctl_pfn *pfn = ndctl_namespace_get_pfn(ndns);
 	struct ndctl_btt *btt = ndctl_namespace_get_btt(ndns);
+	struct ndctl_dax *dax = ndctl_namespace_get_dax(ndns);
 	const char *bdev = NULL;
+	int fd, active = 0;
 	char path[50];
-	int fd;
 	unsigned long long size = ndctl_namespace_get_size(ndns);
 
 	if (pfn && ndctl_pfn_is_enabled(pfn))
 		bdev = ndctl_pfn_get_block_device(pfn);
 	else if (btt && ndctl_btt_is_enabled(btt))
 		bdev = ndctl_btt_get_block_device(btt);
+	else if (dax && ndctl_dax_is_enabled(dax))
+		active = ndctl_dax_has_active_memory(dax);
 	else if (ndctl_namespace_is_enabled(ndns))
 		bdev = ndctl_namespace_get_block_device(ndns);
 
@@ -4632,6 +4651,10 @@ NDCTL_EXPORT int ndctl_namespace_disable_safe(struct ndctl_namespace *ndns)
 					devname, bdev, strerror(errno));
 			return -errno;
 		}
+	} else if (active) {
+		dbg(ctx, "%s: active as system-ram, refusing to disable\n",
+				devname);
+		return -EBUSY;
 	} else {
 		if (size == 0)
 			/* No disable necessary due to no capacity allocated */
diff --git a/test/daxctl-devices.sh b/test/daxctl-devices.sh
index eed5906..56c9691 100755
--- a/test/daxctl-devices.sh
+++ b/test/daxctl-devices.sh
@@ -105,6 +105,22 @@ daxctl_test()
 	"$DAXCTL" reconfigure-device -f -m devdax "$daxdev"
 	[[ $(daxctl_get_mode "$daxdev") == "devdax" ]]
 
+	# fail 'ndctl-disable-namespace' while the devdax namespace is active
+	# as system-ram. If this test fails, a reboot will be required to
+	# recover from the resulting state.
+	test -n "$testdev"
+	"$DAXCTL" reconfigure-device -m system-ram "$daxdev"
+	[[ $(daxctl_get_mode "$daxdev") == "system-ram" ]]
+	if ! "$NDCTL" disable-namespace "$testdev"; then
+		echo "disable-namespace failed as expected"
+	else
+		echo "disable-namespace succeded, expected failure"
+		echo "reboot required to recover from this state"
+		return 1
+	fi
+	"$DAXCTL" reconfigure-device -f -m devdax "$daxdev"
+	[[ $(daxctl_get_mode "$daxdev") == "devdax" ]]
+
 	# this tests for reconfiguration failure if an online-policy is set
 	set_online_policy
 	: "This command is expected to fail:"
-- 
2.30.2
_______________________________________________
Linux-nvdimm mailing list -- linux-nvdimm@lists.01.org
To unsubscribe send an email to linux-nvdimm-leave@lists.01.org

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2021-03-31  3:12 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-03-31  3:12 [ndctl PATCH 0/3] daxctl-device fixes Vishal Verma
2021-03-31  3:12 ` [ndctl PATCH 1/3] daxctl: fail reconfigure-device based on kernel onlining policy Vishal Verma
2021-03-31  3:12 ` [ndctl PATCH 2/3] libdaxctl: add an API to check if a device is active Vishal Verma
2021-03-31  3:12 ` [ndctl PATCH 3/3] libndctl: check for active system-ram before disabling daxctl devices Vishal Verma

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).