* [PATCH 01/11] docs: discuss autonomous self healing in the xfs online repair design doc
2026-01-13 0:32 [PATCHSET v5] xfs: autonomous self healing of filesystems Darrick J. Wong
@ 2026-01-13 0:32 ` Darrick J. Wong
2026-01-13 16:00 ` Christoph Hellwig
2026-01-13 0:33 ` [PATCH 02/11] xfs: start creating infrastructure for health monitoring Darrick J. Wong
` (9 subsequent siblings)
10 siblings, 1 reply; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 0:32 UTC (permalink / raw)
To: djwong, cem; +Cc: hch, linux-fsdevel, linux-xfs
From: Darrick J. Wong <djwong@kernel.org>
Update the XFS online repair document to describe the motivation and
design of the autonomous filesystem healing agent known as xfs_healer.
Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
---
.../filesystems/xfs/xfs-online-fsck-design.rst | 153 ++++++++++++++++++++
1 file changed, 151 insertions(+), 2 deletions(-)
diff --git a/Documentation/filesystems/xfs/xfs-online-fsck-design.rst b/Documentation/filesystems/xfs/xfs-online-fsck-design.rst
index 3d9233f403dbb1..fd936d1b7a32a2 100644
--- a/Documentation/filesystems/xfs/xfs-online-fsck-design.rst
+++ b/Documentation/filesystems/xfs/xfs-online-fsck-design.rst
@@ -166,9 +166,12 @@ The current XFS tools leave several problems unsolved:
malicious actors **exploit quirks of Unicode** to place misleading names
in directories.
+8. **Site Reliability and Support Engineers** would like to reduce the
+ frequency of incidents requiring **manual intervention**.
+
Given this definition of the problems to be solved and the actors who would
benefit, the proposed solution is a third fsck tool that acts on a running
-filesystem.
+filesystem, and an autononmous agent that fixes problems as they arise.
This new third program has three components: an in-kernel facility to check
metadata, an in-kernel facility to repair metadata, and a userspace driver
@@ -203,6 +206,13 @@ Even if a piece of filesystem metadata can only be regenerated by scanning the
entire system, the scan can still be done in the background while other file
operations continue.
+The autonomous self healing agent should listen for metadata health impact
+reports coming from the kernel and automatically schedule repairs for the
+damaged metadata.
+If the required repairs are larger in scope than a single metadata structure,
+``xfs_scrub`` should be invoked to perform a full analysis.
+``xfs_healer`` is the name of this program.
+
In summary, online fsck takes advantage of resource sharding and redundant
metadata to enable targeted checking and repair operations while the system
is running.
@@ -850,11 +860,16 @@ variable in the following service files:
* ``xfs_scrub_all_fail.service``
The decision to enable the background scan is left to the system administrator.
-This can be done by enabling either of the following services:
+This can be done system-wide by enabling either of the following services:
* ``xfs_scrub_all.timer`` on systemd systems
* ``xfs_scrub_all.cron`` on non-systemd systems
+To enable online repair for specific filesystems, the ``autofsck``
+filesystem property should be set to ``repair``.
+To enable only scanning, the property should be set to ``check``.
+To disable online fsck entirely, the property should be set to ``none``.
+
This automatic weekly scan is configured out of the box to perform an
additional media scan of all file data once per month.
This is less foolproof than, say, storing file data block checksums, but much
@@ -897,6 +912,36 @@ notifications and initiate a repair?
*Answer*: These questions remain unanswered, but should be a part of the
conversation with early adopters and potential downstream users of XFS.
+Autonomous Self Healing
+-----------------------
+
+The autonomous self healing agent is a background system service that starts
+when the filesystem is mounted and runs until unmount.
+When starting up, the agent opens a special pseudofile under the specific
+mount.
+When the filesystem generates new adverse health events, the events will be
+made available for reading via the special pseudofile.
+The events need not be limited to metadata concerns; they can also reflect
+events outside of the filesystem's direct control such as file I/O errors.
+
+The agent reads these events in a loop and responds to the events
+appropriately.
+For a single trouble report about metadata, the agent initiates a targeted
+repair of the specific structure.
+If that repair fails or the agent observes too many metadata trouble reports
+over a short interval, it should then initiate a full scan of the filesystem
+via the ``xfs_scrub`` service.
+
+The decision to enable the background scan is left to the system administrator.
+This can be done system-wide by enabling the following services:
+
+* ``xfs_healer@.service`` on systemd systems
+
+To enable autonomous healing for specific filesystems, the ``autofsck``
+filesystem property should be set to ``repair``.
+To disable self healing, the property should be set to ``check``,
+``optimize``, or ``none``.
+
5. Kernel Algorithms and Data Structures
========================================
@@ -4780,6 +4825,70 @@ Orphaned files are adopted by the orphanage as follows:
7. If a runtime error happens, call ``xrep_adoption_cancel`` to release all
resources.
+Health Monitoring
+-----------------
+
+A self-correcting filesystem responds to observations of problems by scheduling
+repairs of the affected areas.
+The filesystem must therefore create event objects in response to stimuli
+(metadata corruption, file I/O errors, etc.) and dispatch these events to
+downstream consumers.
+
+However, the decision to translate an adverse metadata health report into a
+repair should be made by userspace, and the actual scheduling done by userspace.
+Some users (e.g. containers) would prefer to fast-fail the container and restart
+it on another node at a previous checkpoint.
+For workloads running in isolation, repairs may be preferable; either way this
+is something the system administrator knows, and not the kernel.
+A userspace agent (``xfs_healer``, described later) will collect events from the
+kernel and dispatch them appropriately.
+
+Exporting health events to userspace requires the creation of a new component,
+known as the health monitor.
+Because the monitor exposes itself to userspace to deliver information, a file
+descriptor is the natural abstraction to use here.
+The health monitor hooks all the relevant sources of metadata health events.
+Upon activation of the hook, a new event object is created and added to a queue.
+When the agent reads from the fd, event objects are pulled from the start of the
+queue and formatted into the user's buffer.
+The events are freed, and the read call returns to userspace to allow the agent
+to perform some work.
+Memory usage is constrained on a per-fd basis to prevent memory exhaustion; if
+an event must be discarded, a special "lost event" event is delivered to the
+agent.
+
+In short, health events are captured, queued, and eventually copied out to
+userspace for dispatching.
+
++----------------------------------------------------------------------+
+| **Sidebar**: |
++----------------------------------------------------------------------+
+| **Question**: Why use a pseudofile and not use existing notification |
+| methods such as fanotify? |
+| |
+| *Answer*: The pseudofile is a private filesystem interface only |
+| available to processes with the CAP_SYS_ADMIN priviledge and the |
+| ability to open the root directory of an XFS filesystem. |
+| Using a pseudofile gives the kernel and ``xfs_healer`` the |
+| flexibility to expose XFS-specific filesystem details to a special |
+| userspace daemon without cluttering up fanotify's userspace ABI. |
+| Normal userpace programs are not expected to subscribe to the XFS |
+| events in this manner. |
+| Instead, they should subscribe to the generic events provided by |
+| fanotify. |
+| |
+| The pseudofile can also accept ioctls, which gives the userspace |
+| ``xfs_healer`` program a programmatic means to validate that prior |
+| to a repair, its reopened mountpoint is actually the same filesystem |
+| that is being monitored. |
+| |
+| Finally, on an implementation level, fsnotify provides rather little |
+| in the way of an actual event queue implementation; it's really more |
+| of an event dispatcher. |
+| This means there's little advantage in terms of the quantity of new |
+| code added since we still have to write our own queuing discipline! |
++----------------------------------------------------------------------+
+
6. Userspace Algorithms and Data Structures
===========================================
@@ -5071,6 +5180,46 @@ and report what has been lost.
For media errors in blocks owned by files, parent pointers can be used to
construct file paths from inode numbers for user-friendly reporting.
+Autonomous Self Healing
+-----------------------
+
+When a filesystem mounts, the Linux kernel initiates a fsnotify event
+describing the mount point and the path to the data device.
+A separate ``xfs_healer_start`` systemd service listens for these mount
+events via fanotify, and starts a mountpoint-specific ``xfs_healer``
+service instance.
+The ``xfs_healer`` service opens the mountpoint and issues the
+XFS_IOC_HEALTH_MONITOR ioctl to open a special health monitoring file.
+After that is set up, the mountpoint is closed to avoid pinning the mount.
+
+The health monitoring file hooks certain points of the filesystem so that it
+may receive events about metadata health, filesystem shutdowns, media errors,
+file I/O errors, and unmounting of the filesystem.
+Events are queued up for each health monitor file and encoded into a
+``struct xfs_health_monitor_event`` object when the agent calls ``read()`` on
+the file.
+All health events are dispatched to a background threadpool to reduce stalls
+in the main event loop.
+Events can be logged into the system log for further analysis.
+
+For metadata health events, the specific details are used to construct a call
+to the scrub ioctl.
+The filesystem mountpoint is reopened, and the kernel is called.
+If events are lost or the repairs fail, a full scan will be initiated by
+starting up an ``xfs_scrub@.service`` for the given mountpoint.
+
+A filesystem shutdown causes all future repair work to cease, and an unmount
+causes the agent to exit.
+
+**Future Work Question**: Should the healer daemon also register a dbus
+listener and publish events there?
+
+*Answer*: This is unclear -- if there's a demand for system monitoring daemons
+to consume this information and make decisions, then yes, this could be wired
+up in ``xfs_healer``.
+On the other hand, systemd is in the middle of a transition to varlink, so
+it makes more sense to wait and see what happens.
+
7. Conclusion and Future Work
=============================
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH 02/11] xfs: start creating infrastructure for health monitoring
2026-01-13 0:32 [PATCHSET v5] xfs: autonomous self healing of filesystems Darrick J. Wong
2026-01-13 0:32 ` [PATCH 01/11] docs: discuss autonomous self healing in the xfs online repair design doc Darrick J. Wong
@ 2026-01-13 0:33 ` Darrick J. Wong
2026-01-13 16:03 ` Christoph Hellwig
2026-01-13 0:33 ` [PATCH 03/11] xfs: create event queuing, formatting, and discovery infrastructure Darrick J. Wong
` (8 subsequent siblings)
10 siblings, 1 reply; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 0:33 UTC (permalink / raw)
To: djwong, cem; +Cc: hch, linux-fsdevel, linux-xfs
From: Darrick J. Wong <djwong@kernel.org>
Start creating helper functions and infrastructure to pass filesystem
health events to a health monitoring file. Since this is an
administrative interface, we only support a single health monitor
process per filesystem, so we don't need to use anything fancy such as
notifier chains (== tons of indirect calls).
Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
---
fs/xfs/libxfs/xfs_fs.h | 7 +
fs/xfs/xfs_healthmon.h | 36 +++++++
fs/xfs/xfs_mount.h | 4 +
fs/xfs/Makefile | 1
fs/xfs/xfs_health.c | 1
fs/xfs/xfs_healthmon.c | 262 ++++++++++++++++++++++++++++++++++++++++++++++++
fs/xfs/xfs_ioctl.c | 4 +
fs/xfs/xfs_mount.c | 2
8 files changed, 317 insertions(+)
create mode 100644 fs/xfs/xfs_healthmon.h
create mode 100644 fs/xfs/xfs_healthmon.c
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 12463ba766da05..c58e55b3df4099 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -1003,6 +1003,12 @@ struct xfs_rtgroup_geometry {
#define XFS_RTGROUP_GEOM_SICK_RMAPBT (1U << 3) /* reverse mappings */
#define XFS_RTGROUP_GEOM_SICK_REFCNTBT (1U << 4) /* reference counts */
+struct xfs_health_monitor {
+ __u64 flags; /* flags */
+ __u8 format; /* output format */
+ __u8 pad[23]; /* zeroes */
+};
+
/*
* ioctl commands that are used by Linux filesystems
*/
@@ -1042,6 +1048,7 @@ struct xfs_rtgroup_geometry {
#define XFS_IOC_GETPARENTS_BY_HANDLE _IOWR('X', 63, struct xfs_getparents_by_handle)
#define XFS_IOC_SCRUBV_METADATA _IOWR('X', 64, struct xfs_scrub_vec_head)
#define XFS_IOC_RTGROUP_GEOMETRY _IOWR('X', 65, struct xfs_rtgroup_geometry)
+#define XFS_IOC_HEALTH_MONITOR _IOW ('X', 68, struct xfs_health_monitor)
/*
* ioctl commands that replace IRIX syssgi()'s
diff --git a/fs/xfs/xfs_healthmon.h b/fs/xfs/xfs_healthmon.h
new file mode 100644
index 00000000000000..218d5aac87b012
--- /dev/null
+++ b/fs/xfs/xfs_healthmon.h
@@ -0,0 +1,36 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Copyright (c) 2024-2026 Oracle. All Rights Reserved.
+ * Author: Darrick J. Wong <djwong@kernel.org>
+ */
+#ifndef __XFS_HEALTHMON_H__
+#define __XFS_HEALTHMON_H__
+
+struct xfs_healthmon {
+ /*
+ * Weak reference to the xfs filesystem that is being monitored. It
+ * will be set to zero when the filesystem detaches from the monitor.
+ * Do not dereference this pointer.
+ */
+ uintptr_t mount_cookie;
+
+ /*
+ * Device number of the filesystem being monitored. This is for
+ * consistent tracing even after unmount.
+ */
+ dev_t dev;
+
+ /*
+ * Reference count of this structure. The open healthmon fd holds one
+ * ref, the xfs_mount holds another ref if it points to this object,
+ * and running event handlers hold their own refs.
+ */
+ refcount_t ref;
+};
+
+void xfs_healthmon_unmount(struct xfs_mount *mp);
+
+long xfs_ioc_health_monitor(struct file *file,
+ struct xfs_health_monitor __user *arg);
+
+#endif /* __XFS_HEALTHMON_H__ */
diff --git a/fs/xfs/xfs_mount.h b/fs/xfs/xfs_mount.h
index b871dfde372b52..61c71128d171cb 100644
--- a/fs/xfs/xfs_mount.h
+++ b/fs/xfs/xfs_mount.h
@@ -13,6 +13,7 @@ struct xfs_ail;
struct xfs_quotainfo;
struct xfs_da_geometry;
struct xfs_perag;
+struct xfs_healthmon;
/* dynamic preallocation free space thresholds, 5% down to 1% */
enum {
@@ -342,6 +343,9 @@ typedef struct xfs_mount {
/* Hook to feed dirent updates to an active online repair. */
struct xfs_hooks m_dir_update_hooks;
+
+ /* Private data referring to a health monitor object. */
+ struct xfs_healthmon *m_healthmon;
} xfs_mount_t;
#define M_IGEO(mp) (&(mp)->m_ino_geo)
diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index 5bf501cf827172..1b7385e23b3463 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -88,6 +88,7 @@ xfs-y += xfs_aops.o \
xfs_globals.o \
xfs_handle.o \
xfs_health.o \
+ xfs_healthmon.o \
xfs_icache.o \
xfs_ioctl.o \
xfs_iomap.o \
diff --git a/fs/xfs/xfs_health.c b/fs/xfs/xfs_health.c
index fbb8886c72fe5e..3d50397f8f7c00 100644
--- a/fs/xfs/xfs_health.c
+++ b/fs/xfs/xfs_health.c
@@ -19,6 +19,7 @@
#include "xfs_da_btree.h"
#include "xfs_quota_defs.h"
#include "xfs_rtgroup.h"
+#include "xfs_healthmon.h"
#include <linux/fserror.h>
diff --git a/fs/xfs/xfs_healthmon.c b/fs/xfs/xfs_healthmon.c
new file mode 100644
index 00000000000000..b7095ea55897c5
--- /dev/null
+++ b/fs/xfs/xfs_healthmon.c
@@ -0,0 +1,262 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Copyright (c) 2024-2026 Oracle. All Rights Reserved.
+ * Author: Darrick J. Wong <djwong@kernel.org>
+ */
+#include "xfs.h"
+#include "xfs_fs.h"
+#include "xfs_shared.h"
+#include "xfs_format.h"
+#include "xfs_log_format.h"
+#include "xfs_trans_resv.h"
+#include "xfs_mount.h"
+#include "xfs_inode.h"
+#include "xfs_trace.h"
+#include "xfs_ag.h"
+#include "xfs_btree.h"
+#include "xfs_da_format.h"
+#include "xfs_da_btree.h"
+#include "xfs_quota_defs.h"
+#include "xfs_rtgroup.h"
+#include "xfs_healthmon.h"
+
+#include <linux/anon_inodes.h>
+#include <linux/eventpoll.h>
+#include <linux/poll.h>
+
+/*
+ * Live Health Monitoring
+ * ======================
+ *
+ * Autonomous self-healing of XFS filesystems requires a means for the kernel
+ * to send filesystem health events to a monitoring daemon in userspace. To
+ * accomplish this, we establish a thread_with_file kthread object to handle
+ * translating internal events about filesystem health into a format that can
+ * be parsed easily by userspace. When those internal events occur, the core
+ * filesystem code calls this health monitor to convey the events to userspace.
+ * Userspace reads events from the file descriptor returned by the ioctl.
+ *
+ * The healthmon abstraction has a weak reference to the host filesystem mount
+ * so that the queueing and processing of the events do not pin the mount and
+ * cannot slow down the main filesystem. The healthmon object can exist past
+ * the end of the filesystem mount.
+ */
+
+/* sign of a detached health monitor */
+#define DETACHED_MOUNT_COOKIE ((uintptr_t)0)
+
+/* spinlock for atomically updating xfs_mount <-> xfs_healthmon pointers */
+static DEFINE_SPINLOCK(xfs_healthmon_lock);
+
+/* Grab a reference to the healthmon object for a given mount, if any. */
+static struct xfs_healthmon *
+xfs_healthmon_get(
+ struct xfs_mount *mp)
+{
+ struct xfs_healthmon *hm;
+
+ rcu_read_lock();
+ hm = mp->m_healthmon;
+ if (hm && !refcount_inc_not_zero(&hm->ref))
+ hm = NULL;
+ rcu_read_unlock();
+
+ return hm;
+}
+
+/*
+ * Release the reference to a healthmon object. If there are no more holders,
+ * free the health monitor after an RCU grace period to eliminate possibility
+ * of races with xfs_healthmon_get.
+ */
+static void
+xfs_healthmon_put(
+ struct xfs_healthmon *hm)
+{
+ if (refcount_dec_and_test(&hm->ref))
+ kfree_rcu_mightsleep(hm);
+}
+
+/* Attach a health monitor to an xfs_mount. Only one allowed at a time. */
+STATIC int
+xfs_healthmon_attach(
+ struct xfs_mount *mp,
+ struct xfs_healthmon *hm)
+{
+ spin_lock(&xfs_healthmon_lock);
+ if (mp->m_healthmon != NULL) {
+ spin_unlock(&xfs_healthmon_lock);
+ return -EEXIST;
+ }
+
+ refcount_inc(&hm->ref);
+ mp->m_healthmon = hm;
+ hm->mount_cookie = (uintptr_t)mp->m_super;
+ spin_unlock(&xfs_healthmon_lock);
+
+ return 0;
+}
+
+/* Detach a xfs mount from a specific healthmon instance. */
+STATIC void
+xfs_healthmon_detach(
+ struct xfs_healthmon *hm)
+{
+ spin_lock(&xfs_healthmon_lock);
+ if (hm->mount_cookie == DETACHED_MOUNT_COOKIE) {
+ spin_unlock(&xfs_healthmon_lock);
+ return;
+ }
+
+ XFS_M((struct super_block *)hm->mount_cookie)->m_healthmon = NULL;
+ hm->mount_cookie = DETACHED_MOUNT_COOKIE;
+ spin_unlock(&xfs_healthmon_lock);
+
+ xfs_healthmon_put(hm);
+}
+
+/* Detach the xfs mount from this healthmon instance. */
+void
+xfs_healthmon_unmount(
+ struct xfs_mount *mp)
+{
+ struct xfs_healthmon *hm = xfs_healthmon_get(mp);
+
+ if (!hm)
+ return;
+
+ xfs_healthmon_detach(hm);
+ xfs_healthmon_put(hm);
+}
+
+STATIC ssize_t
+xfs_healthmon_read_iter(
+ struct kiocb *iocb,
+ struct iov_iter *to)
+{
+ return -EIO;
+}
+
+/* Free the health monitoring information. */
+STATIC int
+xfs_healthmon_release(
+ struct inode *inode,
+ struct file *file)
+{
+ struct xfs_healthmon *hm = file->private_data;
+
+ /*
+ * We might be closing the healthmon file before the filesystem
+ * unmounts, because userspace processes can terminate at any time and
+ * for any reason. Null out xfs_mount::m_healthmon so that another
+ * process can create another health monitor file.
+ */
+ xfs_healthmon_detach(hm);
+
+ xfs_healthmon_put(hm);
+ return 0;
+}
+
+/* Validate ioctl parameters. */
+static inline bool
+xfs_healthmon_validate(
+ const struct xfs_health_monitor *hmo)
+{
+ if (hmo->flags)
+ return false;
+ if (hmo->format)
+ return false;
+ if (memchr_inv(&hmo->pad, 0, sizeof(hmo->pad)))
+ return false;
+ return true;
+}
+
+/* Emit some data about the health monitoring fd. */
+static void
+xfs_healthmon_show_fdinfo(
+ struct seq_file *m,
+ struct file *file)
+{
+ struct xfs_healthmon *hm = file->private_data;
+
+ seq_printf(m, "state:\t%s\ndev:\t%d:%d\n",
+ hm->mount_cookie == DETACHED_MOUNT_COOKIE ?
+ "dead" : "alive",
+ MAJOR(hm->dev), MINOR(hm->dev));
+}
+
+static const struct file_operations xfs_healthmon_fops = {
+ .owner = THIS_MODULE,
+ .show_fdinfo = xfs_healthmon_show_fdinfo,
+ .read_iter = xfs_healthmon_read_iter,
+ .release = xfs_healthmon_release,
+};
+
+/*
+ * Create a health monitoring file. Returns an index to the fd table or a
+ * negative errno.
+ */
+long
+xfs_ioc_health_monitor(
+ struct file *file,
+ struct xfs_health_monitor __user *arg)
+{
+ struct xfs_health_monitor hmo;
+ struct xfs_healthmon *hm;
+ struct xfs_inode *ip = XFS_I(file_inode(file));
+ struct xfs_mount *mp = ip->i_mount;
+ int ret;
+
+ /*
+ * The only intended user of the health monitoring system should be the
+ * xfs_healer daemon running on behalf of the whole filesystem in the
+ * initial user namespace. IOWs, we don't allow unprivileged userspace
+ * (they can use fsnotify) nor do we allow containers.
+ */
+ if (!capable(CAP_SYS_ADMIN))
+ return -EPERM;
+ if (ip->i_ino != mp->m_sb.sb_rootino)
+ return -EPERM;
+ if (current_user_ns() != &init_user_ns)
+ return -EPERM;
+
+ if (copy_from_user(&hmo, arg, sizeof(hmo)))
+ return -EFAULT;
+
+ if (!xfs_healthmon_validate(&hmo))
+ return -EINVAL;
+
+ hm = kzalloc(sizeof(*hm), GFP_KERNEL);
+ if (!hm)
+ return -ENOMEM;
+ hm->dev = mp->m_super->s_dev;
+ refcount_set(&hm->ref, 1);
+
+ /*
+ * Try to attach this health monitor to the xfs_mount. The monitor is
+ * considered live and will receive events if this succeeds.
+ */
+ ret = xfs_healthmon_attach(mp, hm);
+ if (ret)
+ goto out_hm;
+
+ /*
+ * Create the anonymous file and install a fd for it. If it succeeds,
+ * the file owns hm and can go away at any time, so we must not access
+ * it again. This must go last because we can't undo a fd table
+ * installation.
+ */
+ ret = anon_inode_getfd("xfs_healthmon", &xfs_healthmon_fops, hm,
+ O_CLOEXEC | O_RDONLY);
+ if (ret < 0)
+ goto out_mp;
+
+ return ret;
+
+out_mp:
+ xfs_healthmon_detach(hm);
+out_hm:
+ ASSERT(refcount_read(&hm->ref) == 1);
+ xfs_healthmon_put(hm);
+ return ret;
+}
diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index 59eaad77437181..c04c41ca924e37 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -41,6 +41,7 @@
#include "xfs_exchrange.h"
#include "xfs_handle.h"
#include "xfs_rtgroup.h"
+#include "xfs_healthmon.h"
#include <linux/mount.h>
#include <linux/fileattr.h>
@@ -1419,6 +1420,9 @@ xfs_file_ioctl(
case XFS_IOC_COMMIT_RANGE:
return xfs_ioc_commit_range(filp, arg);
+ case XFS_IOC_HEALTH_MONITOR:
+ return xfs_ioc_health_monitor(filp, arg);
+
default:
return -ENOTTY;
}
diff --git a/fs/xfs/xfs_mount.c b/fs/xfs/xfs_mount.c
index 0953f6ae94abc8..ab67c91915384c 100644
--- a/fs/xfs/xfs_mount.c
+++ b/fs/xfs/xfs_mount.c
@@ -41,6 +41,7 @@
#include "xfs_rtrefcount_btree.h"
#include "scrub/stats.h"
#include "xfs_zone_alloc.h"
+#include "xfs_healthmon.h"
static DEFINE_MUTEX(xfs_uuid_table_mutex);
static int xfs_uuid_table_size;
@@ -625,6 +626,7 @@ xfs_unmount_flush_inodes(
cancel_delayed_work_sync(&mp->m_reclaim_work);
xfs_reclaim_inodes(mp);
xfs_health_unmount(mp);
+ xfs_healthmon_unmount(mp);
}
static void
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH 03/11] xfs: create event queuing, formatting, and discovery infrastructure
2026-01-13 0:32 [PATCHSET v5] xfs: autonomous self healing of filesystems Darrick J. Wong
2026-01-13 0:32 ` [PATCH 01/11] docs: discuss autonomous self healing in the xfs online repair design doc Darrick J. Wong
2026-01-13 0:33 ` [PATCH 02/11] xfs: start creating infrastructure for health monitoring Darrick J. Wong
@ 2026-01-13 0:33 ` Darrick J. Wong
2026-01-13 16:05 ` Christoph Hellwig
2026-01-13 0:33 ` [PATCH 04/11] xfs: convey filesystem unmount events to the health monitor Darrick J. Wong
` (7 subsequent siblings)
10 siblings, 1 reply; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 0:33 UTC (permalink / raw)
To: djwong, cem; +Cc: hch, linux-fsdevel, linux-xfs
From: Darrick J. Wong <djwong@kernel.org>
Create the basic infrastructure that we need to report health events to
userspace. We need a compact form for recording critical information
about an event and queueing them; a means to notice that we've lost some
events; and a means to format the events into something that userspace
can handle. Make the kernel export C structures via read().
In a previous iteration of this new subsystem, I wanted to explore data
exchange formats that are more flexible and easier for humans to read
than C structures. The thought being that when we want to rev (or
worse, enlarge) the event format, it ought to be trivially easy to do
that in a way that doesn't break old userspace.
I looked at formats such as protobufs and capnproto. These look really
nice in that extending the wire format is fairly easy, you can give it a
data schema and it generates the serialization code for you, handles
endianness problems, etc. The huge downside is that neither support C
all that well.
Too hard, and didn't want to port either of those huge sprawling
libraries first to the kernel and then again to xfsprogs. Then I
thought, how about JSON? Javascript objects are human readable, the
kernel can emit json without much fuss (it's all just strings!) and
there are plenty of interpreters for python/rust/c/etc.
There's a proposed schema format for json, which means that xfs can
publish a description of the events that kernel will emit. Userspace
consumers (e.g. xfsprogs/xfs_healer) can embed the same schema document
and use it to validate the incoming events from the kernel, which means
it can discard events that it doesn't understand, or garbage being
emitted due to bugs.
However, json has a huge crutch -- javascript is well known for its
vague definitions of what are numbers. This makes expressing a large
number rather fraught, because the runtime is free to represent a number
in nearly any way it wants. Stupider ones will truncate values to word
size, others will roll out doubles for uint52_t (yes, fifty-two) with
the resulting loss of precision. Not good when you're dealing with
discrete units.
It just so happens that python's json library is smart enough to see a
sequence of digits and put them in a u64 (at least on x86_64/aarch64)
but an actual javascript interpreter (pasting into Firefox) isn't
necessarily so clever.
It turns out that none of the proposed json schemas were ever ratified
even in an open-consensus way, so json blobs are still just loosely
structured blobs. The parsing in userspace was also noticeably slow and
memory-consumptive.
Hence only the C interface survives.
Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
---
fs/xfs/libxfs/xfs_fs.h | 47 +++++
fs/xfs/xfs_healthmon.h | 63 ++++++
fs/xfs/xfs_trace.h | 171 +++++++++++++++++
fs/xfs/xfs_healthmon.c | 495 +++++++++++++++++++++++++++++++++++++++++++++++-
fs/xfs/xfs_trace.c | 2
5 files changed, 772 insertions(+), 6 deletions(-)
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index c58e55b3df4099..22b86bc888de5a 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -1003,12 +1003,59 @@ struct xfs_rtgroup_geometry {
#define XFS_RTGROUP_GEOM_SICK_RMAPBT (1U << 3) /* reverse mappings */
#define XFS_RTGROUP_GEOM_SICK_REFCNTBT (1U << 4) /* reference counts */
+/* Health monitor event domains */
+
+/* affects the whole fs */
+#define XFS_HEALTH_MONITOR_DOMAIN_MOUNT (0)
+
+/* Health monitor event types */
+
+/* status of the monitor itself */
+#define XFS_HEALTH_MONITOR_TYPE_RUNNING (0)
+#define XFS_HEALTH_MONITOR_TYPE_LOST (1)
+
+/* lost events */
+struct xfs_health_monitor_lost {
+ __u64 count;
+};
+
+struct xfs_health_monitor_event {
+ /* XFS_HEALTH_MONITOR_DOMAIN_* */
+ __u32 domain;
+
+ /* XFS_HEALTH_MONITOR_TYPE_* */
+ __u32 type;
+
+ /* Timestamp of the event, in nanoseconds since the Unix epoch */
+ __u64 time_ns;
+
+ /*
+ * Details of the event. The primary clients are written in python
+ * and rust, so break this up because bindgen hates anonymous structs
+ * and unions.
+ */
+ union {
+ struct xfs_health_monitor_lost lost;
+ } e;
+
+ /* zeroes */
+ __u64 pad[2];
+};
+
struct xfs_health_monitor {
__u64 flags; /* flags */
__u8 format; /* output format */
__u8 pad[23]; /* zeroes */
};
+/* Return all health status events, not just deltas */
+#define XFS_HEALTH_MONITOR_VERBOSE (1ULL << 0)
+
+#define XFS_HEALTH_MONITOR_ALL (XFS_HEALTH_MONITOR_VERBOSE)
+
+/* Initial return format version */
+#define XFS_HEALTH_MONITOR_FMT_V0 (0)
+
/*
* ioctl commands that are used by Linux filesystems
*/
diff --git a/fs/xfs/xfs_healthmon.h b/fs/xfs/xfs_healthmon.h
index 218d5aac87b012..600f8a3e52196d 100644
--- a/fs/xfs/xfs_healthmon.h
+++ b/fs/xfs/xfs_healthmon.h
@@ -26,10 +26,73 @@ struct xfs_healthmon {
* and running event handlers hold their own refs.
*/
refcount_t ref;
+
+ /* lock for event list and event counters */
+ struct mutex lock;
+
+ /* list of event objects */
+ struct xfs_healthmon_event *first_event;
+ struct xfs_healthmon_event *last_event;
+
+ /* number of events in the list */
+ unsigned int events;
+
+ /* do we want all events? */
+ bool verbose:1;
+
+ /* waiter so read/poll can sleep until the arrival of events */
+ struct wait_queue_head wait;
+
+ /*
+ * Buffer for formatting events for a read_iter call. Events are
+ * formatted into the buffer at bufhead, and buftail determines where
+ * to start a copy_iter to get those events to userspace. All buffer
+ * fields are protected by inode_lock.
+ */
+ char *buffer;
+ size_t bufsize;
+ size_t bufhead;
+ size_t buftail;
+
+ /* did we lose previous events? */
+ unsigned long long lost_prev_event;
+
+ /* total counts of events observed and lost events */
+ unsigned long long total_events;
+ unsigned long long total_lost;
};
void xfs_healthmon_unmount(struct xfs_mount *mp);
+enum xfs_healthmon_type {
+ XFS_HEALTHMON_RUNNING, /* monitor running */
+ XFS_HEALTHMON_LOST, /* message lost */
+};
+
+enum xfs_healthmon_domain {
+ XFS_HEALTHMON_MOUNT, /* affects the whole fs */
+};
+
+struct xfs_healthmon_event {
+ struct xfs_healthmon_event *next;
+
+ enum xfs_healthmon_type type;
+ enum xfs_healthmon_domain domain;
+
+ uint64_t time_ns;
+
+ union {
+ /* lost events */
+ struct {
+ uint64_t lostcount;
+ };
+ /* mount */
+ struct {
+ unsigned int flags;
+ };
+ };
+};
+
long xfs_ioc_health_monitor(struct file *file,
struct xfs_health_monitor __user *arg);
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index f70afbf3cb196b..04727470b3b410 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -103,6 +103,8 @@ struct xfs_refcount_intent;
struct xfs_metadir_update;
struct xfs_rtgroup;
struct xfs_open_zone;
+struct xfs_healthmon_event;
+struct xfs_healthmon;
#define XFS_ATTR_FILTER_FLAGS \
{ XFS_ATTR_ROOT, "ROOT" }, \
@@ -5906,6 +5908,175 @@ DEFINE_EVENT(xfs_freeblocks_resv_class, name, \
DEFINE_FREEBLOCKS_RESV_EVENT(xfs_freecounter_reserved);
DEFINE_FREEBLOCKS_RESV_EVENT(xfs_freecounter_enospc);
+TRACE_EVENT(xfs_healthmon_lost_event,
+ TP_PROTO(const struct xfs_healthmon *hm),
+ TP_ARGS(hm),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(unsigned long long, lost_prev)
+ ),
+ TP_fast_assign(
+ __entry->dev = hm->dev;
+ __entry->lost_prev = hm->lost_prev_event;
+ ),
+ TP_printk("dev %d:%d lost_prev %llu",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ __entry->lost_prev)
+);
+
+#define XFS_HEALTHMON_FLAGS_STRINGS \
+ { XFS_HEALTH_MONITOR_VERBOSE, "verbose" }
+#define XFS_HEALTHMON_FMT_STRINGS \
+ { XFS_HEALTH_MONITOR_FMT_V0, "v0" }
+
+TRACE_EVENT(xfs_healthmon_create,
+ TP_PROTO(dev_t dev, u64 flags, u8 format),
+ TP_ARGS(dev, flags, format),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(u64, flags)
+ __field(u8, format)
+ ),
+ TP_fast_assign(
+ __entry->dev = dev;
+ __entry->flags = flags;
+ __entry->format = format;
+ ),
+ TP_printk("dev %d:%d flags %s format %s",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ __print_flags(__entry->flags, "|", XFS_HEALTHMON_FLAGS_STRINGS),
+ __print_symbolic(__entry->format, XFS_HEALTHMON_FMT_STRINGS))
+);
+
+TRACE_EVENT(xfs_healthmon_copybuf,
+ TP_PROTO(const struct xfs_healthmon *hm, const struct iov_iter *iov),
+ TP_ARGS(hm, iov),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(size_t, bufsize)
+ __field(size_t, inpos)
+ __field(size_t, outpos)
+ __field(size_t, to_copy)
+ __field(size_t, iter_count)
+ ),
+ TP_fast_assign(
+ __entry->dev = hm->dev;
+ __entry->bufsize = hm->bufsize;
+ __entry->inpos = hm->bufhead;
+ __entry->outpos = hm->buftail;
+ if (hm->bufhead > hm->buftail)
+ __entry->to_copy = hm->bufhead - hm->buftail;
+ else
+ __entry->to_copy = 0;
+ __entry->iter_count = iov_iter_count(iov);
+ ),
+ TP_printk("dev %d:%d bufsize %zu in_pos %zu out_pos %zu to_copy %zu iter_count %zu",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ __entry->bufsize,
+ __entry->inpos,
+ __entry->outpos,
+ __entry->to_copy,
+ __entry->iter_count)
+);
+
+DECLARE_EVENT_CLASS(xfs_healthmon_class,
+ TP_PROTO(const struct xfs_healthmon *hm),
+ TP_ARGS(hm),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(unsigned int, events)
+ __field(unsigned long long, lost_prev)
+ ),
+ TP_fast_assign(
+ __entry->dev = hm->dev;
+ __entry->events = hm->events;
+ __entry->lost_prev = hm->lost_prev_event;
+ ),
+ TP_printk("dev %d:%d events %u lost_prev? %llu",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ __entry->events,
+ __entry->lost_prev)
+);
+#define DEFINE_HEALTHMON_EVENT(name) \
+DEFINE_EVENT(xfs_healthmon_class, name, \
+ TP_PROTO(const struct xfs_healthmon *hm), \
+ TP_ARGS(hm))
+DEFINE_HEALTHMON_EVENT(xfs_healthmon_read_start);
+DEFINE_HEALTHMON_EVENT(xfs_healthmon_read_finish);
+DEFINE_HEALTHMON_EVENT(xfs_healthmon_release);
+DEFINE_HEALTHMON_EVENT(xfs_healthmon_detach);
+
+#define XFS_HEALTHMON_TYPE_STRINGS \
+ { XFS_HEALTHMON_LOST, "lost" }
+
+#define XFS_HEALTHMON_DOMAIN_STRINGS \
+ { XFS_HEALTHMON_MOUNT, "mount" }
+
+TRACE_DEFINE_ENUM(XFS_HEALTHMON_LOST);
+TRACE_DEFINE_ENUM(XFS_HEALTHMON_MOUNT);
+
+DECLARE_EVENT_CLASS(xfs_healthmon_event_class,
+ TP_PROTO(const struct xfs_healthmon *hm,
+ const struct xfs_healthmon_event *event),
+ TP_ARGS(hm, event),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(unsigned int, type)
+ __field(unsigned int, domain)
+ __field(unsigned int, mask)
+ __field(unsigned long long, ino)
+ __field(unsigned int, gen)
+ __field(unsigned int, group)
+ __field(unsigned long long, offset)
+ __field(unsigned long long, length)
+ __field(unsigned long long, lostcount)
+ ),
+ TP_fast_assign(
+ __entry->dev = hm->dev;
+ __entry->type = event->type;
+ __entry->domain = event->domain;
+ __entry->mask = 0;
+ __entry->group = 0;
+ __entry->ino = 0;
+ __entry->gen = 0;
+ __entry->offset = 0;
+ __entry->length = 0;
+ __entry->lostcount = 0;
+ switch (__entry->domain) {
+ case XFS_HEALTHMON_MOUNT:
+ switch (__entry->type) {
+ case XFS_HEALTHMON_LOST:
+ __entry->lostcount = event->lostcount;
+ break;
+ }
+ break;
+ }
+ ),
+ TP_printk("dev %d:%d type %s domain %s mask 0x%x ino 0x%llx gen 0x%x offset 0x%llx len 0x%llx group 0x%x lost %llu",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ __print_symbolic(__entry->type, XFS_HEALTHMON_TYPE_STRINGS),
+ __print_symbolic(__entry->domain, XFS_HEALTHMON_DOMAIN_STRINGS),
+ __entry->mask,
+ __entry->ino,
+ __entry->gen,
+ __entry->offset,
+ __entry->length,
+ __entry->group,
+ __entry->lostcount)
+);
+#define DEFINE_HEALTHMONEVENT_EVENT(name) \
+DEFINE_EVENT(xfs_healthmon_event_class, name, \
+ TP_PROTO(const struct xfs_healthmon *hm, \
+ const struct xfs_healthmon_event *event), \
+ TP_ARGS(hm, event))
+DEFINE_HEALTHMONEVENT_EVENT(xfs_healthmon_insert);
+DEFINE_HEALTHMONEVENT_EVENT(xfs_healthmon_push);
+DEFINE_HEALTHMONEVENT_EVENT(xfs_healthmon_pop);
+DEFINE_HEALTHMONEVENT_EVENT(xfs_healthmon_format);
+DEFINE_HEALTHMONEVENT_EVENT(xfs_healthmon_format_overflow);
+DEFINE_HEALTHMONEVENT_EVENT(xfs_healthmon_drop);
+DEFINE_HEALTHMONEVENT_EVENT(xfs_healthmon_merge);
+
#endif /* _TRACE_XFS_H */
#undef TRACE_INCLUDE_PATH
diff --git a/fs/xfs/xfs_healthmon.c b/fs/xfs/xfs_healthmon.c
index b7095ea55897c5..f1c6782f5e3915 100644
--- a/fs/xfs/xfs_healthmon.c
+++ b/fs/xfs/xfs_healthmon.c
@@ -45,6 +45,13 @@
/* sign of a detached health monitor */
#define DETACHED_MOUNT_COOKIE ((uintptr_t)0)
+/* Constrain the number of event objects that can build up in memory. */
+#define XFS_HEALTHMON_MAX_EVENTS (SZ_32K / \
+ sizeof(struct xfs_healthmon_event))
+
+/* Constrain the size of the output buffer for read_iter. */
+#define XFS_HEALTHMON_MAX_OUTBUF SZ_64K
+
/* spinlock for atomically updating xfs_mount <-> xfs_healthmon pointers */
static DEFINE_SPINLOCK(xfs_healthmon_lock);
@@ -73,8 +80,20 @@ static void
xfs_healthmon_put(
struct xfs_healthmon *hm)
{
- if (refcount_dec_and_test(&hm->ref))
+ if (refcount_dec_and_test(&hm->ref)) {
+ struct xfs_healthmon_event *event;
+ struct xfs_healthmon_event *next = hm->first_event;
+
+ while ((event = next) != NULL) {
+ trace_xfs_healthmon_drop(hm, event);
+ next = event->next;
+ kfree(event);
+ }
+
+ kfree(hm->buffer);
+ mutex_destroy(&hm->lock);
kfree_rcu_mightsleep(hm);
+ }
}
/* Attach a health monitor to an xfs_mount. Only one allowed at a time. */
@@ -112,9 +131,182 @@ xfs_healthmon_detach(
hm->mount_cookie = DETACHED_MOUNT_COOKIE;
spin_unlock(&xfs_healthmon_lock);
+ trace_xfs_healthmon_detach(hm);
xfs_healthmon_put(hm);
}
+static inline void xfs_healthmon_bump_events(struct xfs_healthmon *hm)
+{
+ hm->events++;
+ hm->total_events++;
+}
+
+static inline void xfs_healthmon_bump_lost(struct xfs_healthmon *hm)
+{
+ hm->lost_prev_event++;
+ hm->total_lost++;
+}
+
+/*
+ * If possible, merge a new event into an existing event. Returns whether or
+ * not it merged anything.
+ */
+static bool
+xfs_healthmon_merge_events(
+ struct xfs_healthmon_event *existing,
+ const struct xfs_healthmon_event *new)
+{
+ if (!existing)
+ return false;
+
+ /* type and domain must match to merge events */
+ if (existing->type != new->type ||
+ existing->domain != new->domain)
+ return false;
+
+ switch (existing->type) {
+ case XFS_HEALTHMON_RUNNING:
+ /* should only ever be one of these events anyway */
+ return false;
+
+ case XFS_HEALTHMON_LOST:
+ existing->lostcount += new->lostcount;
+ return true;
+ }
+
+ return false;
+}
+
+/* Insert an event onto the start of the queue. */
+static inline void
+__xfs_healthmon_insert(
+ struct xfs_healthmon *hm,
+ struct xfs_healthmon_event *event)
+{
+ struct timespec64 now;
+
+ ktime_get_coarse_real_ts64(&now);
+ event->time_ns = (now.tv_sec * NSEC_PER_SEC) + now.tv_nsec;
+
+ event->next = hm->first_event;
+ if (!hm->first_event)
+ hm->first_event = event;
+ if (!hm->last_event)
+ hm->last_event = event;
+ xfs_healthmon_bump_events(hm);
+ wake_up(&hm->wait);
+
+ trace_xfs_healthmon_insert(hm, event);
+}
+
+/* Push an event onto the end of the queue. */
+static inline void
+__xfs_healthmon_push(
+ struct xfs_healthmon *hm,
+ struct xfs_healthmon_event *event)
+{
+ struct timespec64 now;
+
+ ktime_get_coarse_real_ts64(&now);
+ event->time_ns = (now.tv_sec * NSEC_PER_SEC) + now.tv_nsec;
+
+ if (!hm->first_event)
+ hm->first_event = event;
+ if (hm->last_event)
+ hm->last_event->next = event;
+ hm->last_event = event;
+ event->next = NULL;
+ xfs_healthmon_bump_events(hm);
+ wake_up(&hm->wait);
+
+ trace_xfs_healthmon_push(hm, event);
+}
+
+/* Deal with any previously lost events */
+static int
+xfs_healthmon_clear_lost_prev(
+ struct xfs_healthmon *hm)
+{
+ struct xfs_healthmon_event lost_event = {
+ .type = XFS_HEALTHMON_LOST,
+ .domain = XFS_HEALTHMON_MOUNT,
+ .lostcount = hm->lost_prev_event,
+ };
+ struct xfs_healthmon_event *event = NULL;
+
+ if (xfs_healthmon_merge_events(hm->last_event, &lost_event)) {
+ trace_xfs_healthmon_merge(hm, hm->last_event);
+ wake_up(&hm->wait);
+ goto cleared;
+ }
+
+ if (hm->events < XFS_HEALTHMON_MAX_EVENTS)
+ event = kmemdup(&lost_event, sizeof(struct xfs_healthmon_event),
+ GFP_NOFS);
+ if (!event)
+ return -ENOMEM;
+
+ __xfs_healthmon_push(hm, event);
+cleared:
+ hm->lost_prev_event = 0;
+ return 0;
+}
+
+/*
+ * Push an event onto the end of the list after dealing with lost events and
+ * possibly full queues.
+ */
+STATIC int
+xfs_healthmon_push(
+ struct xfs_healthmon *hm,
+ const struct xfs_healthmon_event *template)
+{
+ struct xfs_healthmon_event *event = NULL;
+ int error = 0;
+
+ /*
+ * Locklessly check if the health monitor has already detached from the
+ * mount. If so, ignore the event. If we race with deactivation,
+ * we'll queue the event but never send it.
+ */
+ if (hm->mount_cookie == DETACHED_MOUNT_COOKIE)
+ return -ESHUTDOWN;
+
+ mutex_lock(&hm->lock);
+
+ /* Report previously lost events before we do anything else */
+ if (hm->lost_prev_event) {
+ error = xfs_healthmon_clear_lost_prev(hm);
+ if (error)
+ goto out_unlock;
+ }
+
+ /* Try to merge with the newest event */
+ if (xfs_healthmon_merge_events(hm->last_event, template)) {
+ trace_xfs_healthmon_merge(hm, hm->last_event);
+ wake_up(&hm->wait);
+ goto out_unlock;
+ }
+
+ /* Only create a heap event object if we're not already at capacity. */
+ if (hm->events < XFS_HEALTHMON_MAX_EVENTS)
+ event = kmemdup(template, sizeof(struct xfs_healthmon_event),
+ GFP_NOFS);
+ if (!event) {
+ /* No memory means we lose the event */
+ trace_xfs_healthmon_lost_event(hm);
+ xfs_healthmon_bump_lost(hm);
+ error = -ENOMEM;
+ goto out_unlock;
+ }
+
+ __xfs_healthmon_push(hm, event);
+
+out_unlock:
+ mutex_unlock(&hm->lock);
+ return error;
+}
+
/* Detach the xfs mount from this healthmon instance. */
void
xfs_healthmon_unmount(
@@ -129,12 +321,271 @@ xfs_healthmon_unmount(
xfs_healthmon_put(hm);
}
+static inline void
+xfs_healthmon_reset_outbuf(
+ struct xfs_healthmon *hm)
+{
+ hm->buftail = 0;
+ hm->bufhead = 0;
+}
+
+static const unsigned int domain_map[] = {
+ [XFS_HEALTHMON_MOUNT] = XFS_HEALTH_MONITOR_DOMAIN_MOUNT,
+};
+
+static const unsigned int type_map[] = {
+ [XFS_HEALTHMON_RUNNING] = XFS_HEALTH_MONITOR_TYPE_RUNNING,
+ [XFS_HEALTHMON_LOST] = XFS_HEALTH_MONITOR_TYPE_LOST,
+};
+
+/* Render event as a V0 structure */
+STATIC int
+xfs_healthmon_format_v0(
+ struct xfs_healthmon *hm,
+ const struct xfs_healthmon_event *event)
+{
+ struct xfs_health_monitor_event hme = {
+ .time_ns = event->time_ns,
+ };
+
+ trace_xfs_healthmon_format(hm, event);
+
+ if (event->domain < 0 || event->domain >= ARRAY_SIZE(domain_map) ||
+ event->type < 0 || event->type >= ARRAY_SIZE(type_map))
+ return -EFSCORRUPTED;
+
+ hme.domain = domain_map[event->domain];
+ hme.type = type_map[event->type];
+
+ /* fill in the event-specific details */
+ switch (event->domain) {
+ case XFS_HEALTHMON_MOUNT:
+ switch (event->type) {
+ case XFS_HEALTHMON_LOST:
+ hme.e.lost.count = event->lostcount;
+ break;
+ default:
+ break;
+ }
+ break;
+ default:
+ break;
+ }
+
+ ASSERT(hm->bufhead + sizeof(hme) <= hm->bufsize);
+
+ /* copy formatted object to the outbuf */
+ if (hm->bufhead + sizeof(hme) <= hm->bufsize) {
+ memcpy(hm->buffer + hm->bufhead, &hme, sizeof(hme));
+ hm->bufhead += sizeof(hme);
+ }
+
+ return 0;
+}
+
+/* How many bytes are waiting in the outbuf to be copied? */
+static inline size_t
+xfs_healthmon_outbuf_bytes(
+ struct xfs_healthmon *hm)
+{
+ if (hm->bufhead > hm->buftail)
+ return hm->bufhead - hm->buftail;
+ return 0;
+}
+
+/*
+ * Do we have something for userspace to read? This can mean unmount events,
+ * events pending in the queue, or pending bytes in the outbuf.
+ */
+static inline bool
+xfs_healthmon_has_eventdata(
+ struct xfs_healthmon *hm)
+{
+ /*
+ * If the health monitor is already detached from the xfs_mount, we
+ * want reads to return 0 bytes even if there are no events, because
+ * userspace interprets that as EOF. If we race with deactivation,
+ * read_iter will take the necessary locks to discover that there are
+ * no events to send.
+ */
+ if (hm->mount_cookie == DETACHED_MOUNT_COOKIE)
+ return true;
+
+ /*
+ * Either there are events waiting to be formatted into the buffer, or
+ * there's unread bytes in the buffer.
+ */
+ return hm->events > 0 || xfs_healthmon_outbuf_bytes(hm) > 0;
+}
+
+/* Try to copy the rest of the outbuf to the iov iter. */
+STATIC ssize_t
+xfs_healthmon_copybuf(
+ struct xfs_healthmon *hm,
+ struct iov_iter *to)
+{
+ size_t to_copy;
+ size_t w = 0;
+
+ trace_xfs_healthmon_copybuf(hm, to);
+
+ to_copy = xfs_healthmon_outbuf_bytes(hm);
+ if (to_copy) {
+ w = copy_to_iter(hm->buffer + hm->buftail, to_copy, to);
+ if (!w)
+ return -EFAULT;
+
+ hm->buftail += w;
+ }
+
+ /*
+ * Nothing left to copy? Reset the output buffer cursors to the start
+ * since there's no live data in the buffer.
+ */
+ if (xfs_healthmon_outbuf_bytes(hm) == 0)
+ xfs_healthmon_reset_outbuf(hm);
+ return w;
+}
+
+/*
+ * Return a health monitoring event for formatting into the output buffer if
+ * there's enough space in the outbuf and an event waiting for us. Caller
+ * must hold i_rwsem on the healthmon file.
+ */
+static inline struct xfs_healthmon_event *
+xfs_healthmon_format_pop(
+ struct xfs_healthmon *hm)
+{
+ struct xfs_healthmon_event *event;
+
+ if (hm->bufhead + sizeof(*event) > hm->bufsize)
+ return NULL;
+
+ mutex_lock(&hm->lock);
+ event = hm->first_event;
+ if (event) {
+ if (hm->last_event == event)
+ hm->last_event = NULL;
+ hm->first_event = event->next;
+ hm->events--;
+
+ trace_xfs_healthmon_pop(hm, event);
+ }
+ mutex_unlock(&hm->lock);
+ return event;
+}
+
+/* Allocate formatting buffer */
+STATIC int
+xfs_healthmon_alloc_outbuf(
+ struct xfs_healthmon *hm,
+ size_t user_bufsize)
+{
+ void *outbuf;
+ size_t bufsize =
+ min(XFS_HEALTHMON_MAX_OUTBUF, max(PAGE_SIZE, user_bufsize));
+
+ outbuf = kzalloc(bufsize, GFP_KERNEL);
+ if (!outbuf) {
+ if (bufsize == PAGE_SIZE)
+ return -ENOMEM;
+
+ bufsize = PAGE_SIZE;
+ outbuf = kzalloc(bufsize, GFP_KERNEL);
+ if (!outbuf)
+ return -ENOMEM;
+ }
+
+ hm->buffer = outbuf;
+ hm->bufsize = bufsize;
+ hm->bufhead = 0;
+ hm->buftail = 0;
+
+ return 0;
+}
+
+/*
+ * Convey queued event data to userspace. First copy any remaining bytes in
+ * the outbuf, then format the oldest event into the outbuf and copy that too.
+ */
STATIC ssize_t
xfs_healthmon_read_iter(
struct kiocb *iocb,
struct iov_iter *to)
{
- return -EIO;
+ struct file *file = iocb->ki_filp;
+ struct inode *inode = file_inode(file);
+ struct xfs_healthmon *hm = file->private_data;
+ struct xfs_healthmon_event *event;
+ size_t copied = 0;
+ ssize_t ret = 0;
+
+ if (file->f_flags & O_NONBLOCK) {
+ if (!xfs_healthmon_has_eventdata(hm) || !inode_trylock(inode))
+ return -EAGAIN;
+ } else {
+ ret = wait_event_interruptible(hm->wait,
+ xfs_healthmon_has_eventdata(hm));
+ if (ret)
+ return ret;
+
+ inode_lock(inode);
+ }
+
+ if (hm->bufsize == 0) {
+ ret = xfs_healthmon_alloc_outbuf(hm, iov_iter_count(to));
+ if (ret)
+ goto out_unlock;
+ }
+
+ trace_xfs_healthmon_read_start(hm);
+
+ /*
+ * If there's anything left in the output buffer, copy that before
+ * formatting more events.
+ */
+ ret = xfs_healthmon_copybuf(hm, to);
+ if (ret < 0)
+ goto out_unlock;
+ copied += ret;
+
+ while (iov_iter_count(to) > 0) {
+ /* Format the next events into the outbuf until it's full. */
+ while ((event = xfs_healthmon_format_pop(hm)) != NULL) {
+ ret = xfs_healthmon_format_v0(hm, event);
+ kfree(event);
+ if (ret)
+ goto out_unlock;
+ }
+
+ /* Copy anything formatted into outbuf to userspace */
+ ret = xfs_healthmon_copybuf(hm, to);
+ if (ret <= 0)
+ break;
+
+ copied += ret;
+ }
+
+out_unlock:
+ trace_xfs_healthmon_read_finish(hm);
+ inode_unlock(inode);
+ return copied ?: ret;
+}
+
+/* Poll for available events. */
+STATIC __poll_t
+xfs_healthmon_poll(
+ struct file *file,
+ struct poll_table_struct *wait)
+{
+ struct xfs_healthmon *hm = file->private_data;
+ __poll_t mask = 0;
+
+ poll_wait(file, &hm->wait, wait);
+
+ if (xfs_healthmon_has_eventdata(hm))
+ mask |= EPOLLIN;
+ return mask;
}
/* Free the health monitoring information. */
@@ -145,6 +596,8 @@ xfs_healthmon_release(
{
struct xfs_healthmon *hm = file->private_data;
+ trace_xfs_healthmon_release(hm);
+
/*
* We might be closing the healthmon file before the filesystem
* unmounts, because userspace processes can terminate at any time and
@@ -153,6 +606,12 @@ xfs_healthmon_release(
*/
xfs_healthmon_detach(hm);
+ /*
+ * Wake up any readers that might be left. There shouldn't be any
+ * because the only users of the waiter are read and poll.
+ */
+ wake_up_all(&hm->wait);
+
xfs_healthmon_put(hm);
return 0;
}
@@ -162,9 +621,9 @@ static inline bool
xfs_healthmon_validate(
const struct xfs_health_monitor *hmo)
{
- if (hmo->flags)
+ if (hmo->flags & ~XFS_HEALTH_MONITOR_ALL)
return false;
- if (hmo->format)
+ if (hmo->format != XFS_HEALTH_MONITOR_FMT_V0)
return false;
if (memchr_inv(&hmo->pad, 0, sizeof(hmo->pad)))
return false;
@@ -179,16 +638,21 @@ xfs_healthmon_show_fdinfo(
{
struct xfs_healthmon *hm = file->private_data;
- seq_printf(m, "state:\t%s\ndev:\t%d:%d\n",
+ mutex_lock(&hm->lock);
+ seq_printf(m, "state:\t%s\ndev:\t%d:%d\nformat:\tv0\nevents:\t%llu\nlost:\t%llu\n",
hm->mount_cookie == DETACHED_MOUNT_COOKIE ?
"dead" : "alive",
- MAJOR(hm->dev), MINOR(hm->dev));
+ MAJOR(hm->dev), MINOR(hm->dev),
+ hm->total_events,
+ hm->total_lost);
+ mutex_unlock(&hm->lock);
}
static const struct file_operations xfs_healthmon_fops = {
.owner = THIS_MODULE,
.show_fdinfo = xfs_healthmon_show_fdinfo,
.read_iter = xfs_healthmon_read_iter,
+ .poll = xfs_healthmon_poll,
.release = xfs_healthmon_release,
};
@@ -202,6 +666,7 @@ xfs_ioc_health_monitor(
struct xfs_health_monitor __user *arg)
{
struct xfs_health_monitor hmo;
+ struct xfs_healthmon_event *running_event;
struct xfs_healthmon *hm;
struct xfs_inode *ip = XFS_I(file_inode(file));
struct xfs_mount *mp = ip->i_mount;
@@ -232,6 +697,22 @@ xfs_ioc_health_monitor(
hm->dev = mp->m_super->s_dev;
refcount_set(&hm->ref, 1);
+ mutex_init(&hm->lock);
+ init_waitqueue_head(&hm->wait);
+
+ if (hmo.flags & XFS_HEALTH_MONITOR_VERBOSE)
+ hm->verbose = true;
+
+ /* Queue up the first event that lets the client know we're running. */
+ running_event = kzalloc(sizeof(struct xfs_healthmon_event), GFP_NOFS);
+ if (!running_event) {
+ ret = -ENOMEM;
+ goto out_hm;
+ }
+ running_event->type = XFS_HEALTHMON_RUNNING;
+ running_event->domain = XFS_HEALTHMON_MOUNT;
+ __xfs_healthmon_insert(hm, running_event);
+
/*
* Try to attach this health monitor to the xfs_mount. The monitor is
* considered live and will receive events if this succeeds.
@@ -251,6 +732,8 @@ xfs_ioc_health_monitor(
if (ret < 0)
goto out_mp;
+ trace_xfs_healthmon_create(mp->m_super->s_dev, hmo.flags, hmo.format);
+
return ret;
out_mp:
diff --git a/fs/xfs/xfs_trace.c b/fs/xfs/xfs_trace.c
index a60556dbd172ee..d42b864a3837a2 100644
--- a/fs/xfs/xfs_trace.c
+++ b/fs/xfs/xfs_trace.c
@@ -51,6 +51,8 @@
#include "xfs_rtgroup.h"
#include "xfs_zone_alloc.h"
#include "xfs_zone_priv.h"
+#include "xfs_health.h"
+#include "xfs_healthmon.h"
/*
* We include this last to have the helpers above available for the trace
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH 04/11] xfs: convey filesystem unmount events to the health monitor
2026-01-13 0:32 [PATCHSET v5] xfs: autonomous self healing of filesystems Darrick J. Wong
` (2 preceding siblings ...)
2026-01-13 0:33 ` [PATCH 03/11] xfs: create event queuing, formatting, and discovery infrastructure Darrick J. Wong
@ 2026-01-13 0:33 ` Darrick J. Wong
2026-01-13 16:11 ` Christoph Hellwig
2026-01-13 0:33 ` [PATCH 05/11] xfs: convey metadata health " Darrick J. Wong
` (6 subsequent siblings)
10 siblings, 1 reply; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 0:33 UTC (permalink / raw)
To: djwong, cem; +Cc: hch, linux-fsdevel, linux-xfs
From: Darrick J. Wong <djwong@kernel.org>
In xfs_healthmon_unmount, send events to xfs_healer so that it knows
that nothing further can be done for the filesystem.
Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
---
fs/xfs/libxfs/xfs_fs.h | 3 +++
fs/xfs/xfs_healthmon.h | 1 +
fs/xfs/xfs_trace.h | 6 +++++-
fs/xfs/xfs_healthmon.c | 34 +++++++++++++++++++++++++++++++++-
4 files changed, 42 insertions(+), 2 deletions(-)
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 22b86bc888de5a..59de6ab69fb319 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -1014,6 +1014,9 @@ struct xfs_rtgroup_geometry {
#define XFS_HEALTH_MONITOR_TYPE_RUNNING (0)
#define XFS_HEALTH_MONITOR_TYPE_LOST (1)
+/* filesystem was unmounted */
+#define XFS_HEALTH_MONITOR_TYPE_UNMOUNT (2)
+
/* lost events */
struct xfs_health_monitor_lost {
__u64 count;
diff --git a/fs/xfs/xfs_healthmon.h b/fs/xfs/xfs_healthmon.h
index 600f8a3e52196d..63e6c110bd378a 100644
--- a/fs/xfs/xfs_healthmon.h
+++ b/fs/xfs/xfs_healthmon.h
@@ -67,6 +67,7 @@ void xfs_healthmon_unmount(struct xfs_mount *mp);
enum xfs_healthmon_type {
XFS_HEALTHMON_RUNNING, /* monitor running */
XFS_HEALTHMON_LOST, /* message lost */
+ XFS_HEALTHMON_UNMOUNT, /* filesystem is unmounting */
};
enum xfs_healthmon_domain {
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 04727470b3b410..305cae8f497b43 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -6005,14 +6005,18 @@ DEFINE_HEALTHMON_EVENT(xfs_healthmon_read_start);
DEFINE_HEALTHMON_EVENT(xfs_healthmon_read_finish);
DEFINE_HEALTHMON_EVENT(xfs_healthmon_release);
DEFINE_HEALTHMON_EVENT(xfs_healthmon_detach);
+DEFINE_HEALTHMON_EVENT(xfs_healthmon_report_unmount);
#define XFS_HEALTHMON_TYPE_STRINGS \
- { XFS_HEALTHMON_LOST, "lost" }
+ { XFS_HEALTHMON_LOST, "lost" }, \
+ { XFS_HEALTHMON_UNMOUNT, "unmount" }
#define XFS_HEALTHMON_DOMAIN_STRINGS \
{ XFS_HEALTHMON_MOUNT, "mount" }
TRACE_DEFINE_ENUM(XFS_HEALTHMON_LOST);
+TRACE_DEFINE_ENUM(XFS_HEALTHMON_UNMOUNT);
+
TRACE_DEFINE_ENUM(XFS_HEALTHMON_MOUNT);
DECLARE_EVENT_CLASS(xfs_healthmon_event_class,
diff --git a/fs/xfs/xfs_healthmon.c b/fs/xfs/xfs_healthmon.c
index f1c6782f5e3915..122e5a4d8e24cb 100644
--- a/fs/xfs/xfs_healthmon.c
+++ b/fs/xfs/xfs_healthmon.c
@@ -166,6 +166,7 @@ xfs_healthmon_merge_events(
switch (existing->type) {
case XFS_HEALTHMON_RUNNING:
+ case XFS_HEALTHMON_UNMOUNT:
/* should only ever be one of these events anyway */
return false;
@@ -307,16 +308,47 @@ xfs_healthmon_push(
return error;
}
-/* Detach the xfs mount from this healthmon instance. */
+/*
+ * Report that the filesystem is being unmounted, then detach the xfs mount
+ * from this healthmon instance.
+ */
void
xfs_healthmon_unmount(
struct xfs_mount *mp)
{
struct xfs_healthmon *hm = xfs_healthmon_get(mp);
+ struct xfs_healthmon_event *event;
if (!hm)
return;
+ event = kzalloc(sizeof(struct xfs_healthmon_event), GFP_NOFS);
+ mutex_lock(&hm->lock);
+
+ trace_xfs_healthmon_report_unmount(hm);
+
+ if (event) {
+ /*
+ * Insert the unmount notification at the start of the event
+ * queue so that userspace knows the filesystem went away as
+ * soon as possible. There's nothing actionable for userspace
+ * after an unmount.
+ */
+ event->type = XFS_HEALTHMON_UNMOUNT;
+ event->domain = XFS_HEALTHMON_MOUNT;
+
+ __xfs_healthmon_insert(hm, event);
+ } else {
+ /*
+ * Wake up the reader directly in case we didn't have enough
+ * memory to queue the unmount event. The filesystem is about
+ * to go away so we don't care about reporting previously lost
+ * events.
+ */
+ wake_up(&hm->wait);
+ }
+ mutex_unlock(&hm->lock);
+
xfs_healthmon_detach(hm);
xfs_healthmon_put(hm);
}
^ permalink raw reply related [flat|nested] 36+ messages in thread* Re: [PATCH 04/11] xfs: convey filesystem unmount events to the health monitor
2026-01-13 0:33 ` [PATCH 04/11] xfs: convey filesystem unmount events to the health monitor Darrick J. Wong
@ 2026-01-13 16:11 ` Christoph Hellwig
2026-01-13 18:48 ` Darrick J. Wong
0 siblings, 1 reply; 36+ messages in thread
From: Christoph Hellwig @ 2026-01-13 16:11 UTC (permalink / raw)
To: Darrick J. Wong; +Cc: cem, hch, linux-fsdevel, linux-xfs
> + event = kzalloc(sizeof(struct xfs_healthmon_event), GFP_NOFS);
> + mutex_lock(&hm->lock);
> +
> + trace_xfs_healthmon_report_unmount(hm);
> +
> + if (event) {
> + /*
> + * Insert the unmount notification at the start of the event
> + * queue so that userspace knows the filesystem went away as
> + * soon as possible. There's nothing actionable for userspace
> + * after an unmount.
> + */
> + event->type = XFS_HEALTHMON_UNMOUNT;
> + event->domain = XFS_HEALTHMON_MOUNT;
> +
> + __xfs_healthmon_insert(hm, event);
The allocation (and locking) could move into __xfs_healthmon_insert
sharing this code with xfs_ioc_health_monitor(). That means the wake_up
would not be covered by the lock, but I can't see how that would need
it.
> + } else {
> + /*
> + * Wake up the reader directly in case we didn't have enough
> + * memory to queue the unmount event. The filesystem is about
> + * to go away so we don't care about reporting previously lost
> + * events.
> + */
> + wake_up(&hm->wait);
> + }
And then even after reading this many times I'm still not understanding
it. Yes, this wakes up the reader direct, but why is waking it up
without an even ok? And if it is ok, why do we even bother with the
unmount even? If OTOH the unmount event is actually important, should
we preallocate the event? (and render my above suggestion impractical).
^ permalink raw reply [flat|nested] 36+ messages in thread* Re: [PATCH 04/11] xfs: convey filesystem unmount events to the health monitor
2026-01-13 16:11 ` Christoph Hellwig
@ 2026-01-13 18:48 ` Darrick J. Wong
0 siblings, 0 replies; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 18:48 UTC (permalink / raw)
To: Christoph Hellwig; +Cc: cem, linux-fsdevel, linux-xfs
On Tue, Jan 13, 2026 at 05:11:05PM +0100, Christoph Hellwig wrote:
> > + event = kzalloc(sizeof(struct xfs_healthmon_event), GFP_NOFS);
> > + mutex_lock(&hm->lock);
> > +
> > + trace_xfs_healthmon_report_unmount(hm);
> > +
> > + if (event) {
> > + /*
> > + * Insert the unmount notification at the start of the event
> > + * queue so that userspace knows the filesystem went away as
> > + * soon as possible. There's nothing actionable for userspace
> > + * after an unmount.
> > + */
> > + event->type = XFS_HEALTHMON_UNMOUNT;
> > + event->domain = XFS_HEALTHMON_MOUNT;
> > +
> > + __xfs_healthmon_insert(hm, event);
>
> The allocation (and locking) could move into __xfs_healthmon_insert
> sharing this code with xfs_ioc_health_monitor(). That means the wake_up
> would not be covered by the lock, but I can't see how that would need
> it.
>
> > + } else {
> > + /*
> > + * Wake up the reader directly in case we didn't have enough
> > + * memory to queue the unmount event. The filesystem is about
> > + * to go away so we don't care about reporting previously lost
> > + * events.
> > + */
> > + wake_up(&hm->wait);
> > + }
>
> And then even after reading this many times I'm still not understanding
> it. Yes, this wakes up the reader direct, but why is waking it up
> without an even ok?
I think that was an artifact from an earlier stage in development where
it was necessary to wake up the reader to get it to drain the event
queue no matter what. At this point I'd much rather preallocate the
unmount event because...
> And if it is ok, why do we even bother with the
> unmount even? If OTOH the unmount event is actually important, should
> we preallocate the event? (and render my above suggestion impractical).
...the unmount is important, because the userspace daemon uses that as
the signal to quit immediately. If userspace misses the event then
it'll just keep running forever with a dead healthmon fd.
xfs_healer_start will restart the whole service if someone mounts
another xfs filesystem at the same mountpoint, but if nobody does that
then we'll be stuck with a defunct daemon forever.
Let's preallocate the unmount event, and then this whole thing becomes:
/*
* Insert the unmount notification at the start of the event
* queue so that userspace knows the filesystem went away as
* soon as possible. There's nothing actionable for userspace
* after an unmount. Once we've inserted unmount_event, hm no
* longer owns that event.
*/
mutex_lock(&hm->lock);
__xfs_healthmon_insert(hm, hm->unmount_event);
hm->unmount_event = NULL;
mutex_unlock(&hm->lock);
--D
^ permalink raw reply [flat|nested] 36+ messages in thread
* [PATCH 05/11] xfs: convey metadata health events to the health monitor
2026-01-13 0:32 [PATCHSET v5] xfs: autonomous self healing of filesystems Darrick J. Wong
` (3 preceding siblings ...)
2026-01-13 0:33 ` [PATCH 04/11] xfs: convey filesystem unmount events to the health monitor Darrick J. Wong
@ 2026-01-13 0:33 ` Darrick J. Wong
2026-01-13 16:11 ` Christoph Hellwig
2026-01-13 0:34 ` [PATCH 06/11] xfs: convey filesystem shutdown " Darrick J. Wong
` (5 subsequent siblings)
10 siblings, 1 reply; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 0:33 UTC (permalink / raw)
To: djwong, cem; +Cc: hch, linux-fsdevel, linux-xfs
From: Darrick J. Wong <djwong@kernel.org>
Connect the filesystem metadata health event collection system to the
health monitor so that xfs can send events to xfs_healer as it collects
information.
Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
---
fs/xfs/libxfs/xfs_fs.h | 35 +++++++++
fs/xfs/libxfs/xfs_health.h | 5 +
fs/xfs/xfs_healthmon.h | 39 +++++++++
fs/xfs/xfs_trace.h | 130 +++++++++++++++++++++++++++++++-
fs/xfs/xfs_health.c | 123 ++++++++++++++++++++++++++++++
fs/xfs/xfs_healthmon.c | 181 ++++++++++++++++++++++++++++++++++++++++++++
6 files changed, 511 insertions(+), 2 deletions(-)
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 59de6ab69fb319..04e1dcf61257d0 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -1008,6 +1008,12 @@ struct xfs_rtgroup_geometry {
/* affects the whole fs */
#define XFS_HEALTH_MONITOR_DOMAIN_MOUNT (0)
+/* metadata health events */
+#define XFS_HEALTH_MONITOR_DOMAIN_FS (1)
+#define XFS_HEALTH_MONITOR_DOMAIN_AG (2)
+#define XFS_HEALTH_MONITOR_DOMAIN_INODE (3)
+#define XFS_HEALTH_MONITOR_DOMAIN_RTGROUP (4)
+
/* Health monitor event types */
/* status of the monitor itself */
@@ -1017,11 +1023,37 @@ struct xfs_rtgroup_geometry {
/* filesystem was unmounted */
#define XFS_HEALTH_MONITOR_TYPE_UNMOUNT (2)
+/* metadata health events */
+#define XFS_HEALTH_MONITOR_TYPE_SICK (3)
+#define XFS_HEALTH_MONITOR_TYPE_CORRUPT (4)
+#define XFS_HEALTH_MONITOR_TYPE_HEALTHY (5)
+
/* lost events */
struct xfs_health_monitor_lost {
__u64 count;
};
+/* fs/rt metadata */
+struct xfs_health_monitor_fs {
+ /* XFS_FSOP_GEOM_SICK_* flags */
+ __u32 mask;
+};
+
+/* ag/rtgroup metadata */
+struct xfs_health_monitor_group {
+ /* XFS_{AG,RTGROUP}_SICK_* flags */
+ __u32 mask;
+ __u32 gno;
+};
+
+/* inode metadata */
+struct xfs_health_monitor_inode {
+ /* XFS_BS_SICK_* flags */
+ __u32 mask;
+ __u32 gen;
+ __u64 ino;
+};
+
struct xfs_health_monitor_event {
/* XFS_HEALTH_MONITOR_DOMAIN_* */
__u32 domain;
@@ -1039,6 +1071,9 @@ struct xfs_health_monitor_event {
*/
union {
struct xfs_health_monitor_lost lost;
+ struct xfs_health_monitor_fs fs;
+ struct xfs_health_monitor_group group;
+ struct xfs_health_monitor_inode inode;
} e;
/* zeroes */
diff --git a/fs/xfs/libxfs/xfs_health.h b/fs/xfs/libxfs/xfs_health.h
index b31000f7190ce5..1d45cf5789e864 100644
--- a/fs/xfs/libxfs/xfs_health.h
+++ b/fs/xfs/libxfs/xfs_health.h
@@ -289,4 +289,9 @@ void xfs_bulkstat_health(struct xfs_inode *ip, struct xfs_bulkstat *bs);
#define xfs_metadata_is_sick(error) \
(unlikely((error) == -EFSCORRUPTED || (error) == -EFSBADCRC))
+unsigned int xfs_healthmon_inode_mask(unsigned int sick_mask);
+unsigned int xfs_healthmon_rtgroup_mask(unsigned int sick_mask);
+unsigned int xfs_healthmon_perag_mask(unsigned int sick_mask);
+unsigned int xfs_healthmon_fs_mask(unsigned int sick_mask);
+
#endif /* __XFS_HEALTH_H__ */
diff --git a/fs/xfs/xfs_healthmon.h b/fs/xfs/xfs_healthmon.h
index 63e6c110bd378a..4632fa873ec087 100644
--- a/fs/xfs/xfs_healthmon.h
+++ b/fs/xfs/xfs_healthmon.h
@@ -68,10 +68,21 @@ enum xfs_healthmon_type {
XFS_HEALTHMON_RUNNING, /* monitor running */
XFS_HEALTHMON_LOST, /* message lost */
XFS_HEALTHMON_UNMOUNT, /* filesystem is unmounting */
+
+ /* metadata health events */
+ XFS_HEALTHMON_SICK, /* runtime corruption observed */
+ XFS_HEALTHMON_CORRUPT, /* fsck reported corruption */
+ XFS_HEALTHMON_HEALTHY, /* fsck reported healthy structure */
};
enum xfs_healthmon_domain {
XFS_HEALTHMON_MOUNT, /* affects the whole fs */
+
+ /* metadata health events */
+ XFS_HEALTHMON_FS, /* main filesystem metadata */
+ XFS_HEALTHMON_AG, /* allocation group metadata */
+ XFS_HEALTHMON_INODE, /* inode metadata */
+ XFS_HEALTHMON_RTGROUP, /* realtime group metadata */
};
struct xfs_healthmon_event {
@@ -91,9 +102,37 @@ struct xfs_healthmon_event {
struct {
unsigned int flags;
};
+ /* fs/rt metadata */
+ struct {
+ /* XFS_SICK_* flags */
+ unsigned int fsmask;
+ };
+ /* ag/rtgroup metadata */
+ struct {
+ /* XFS_SICK_(AG|RG)* flags */
+ unsigned int grpmask;
+ unsigned int group;
+ };
+ /* inode metadata */
+ struct {
+ /* XFS_SICK_INO_* flags */
+ unsigned int imask;
+ uint32_t gen;
+ xfs_ino_t ino;
+ };
};
};
+void xfs_healthmon_report_fs(struct xfs_mount *mp,
+ enum xfs_healthmon_type type, unsigned int old_mask,
+ unsigned int new_mask);
+void xfs_healthmon_report_group(struct xfs_group *xg,
+ enum xfs_healthmon_type type, unsigned int old_mask,
+ unsigned int new_mask);
+void xfs_healthmon_report_inode(struct xfs_inode *ip,
+ enum xfs_healthmon_type type, unsigned int old_mask,
+ unsigned int new_mask);
+
long xfs_ioc_health_monitor(struct file *file,
struct xfs_health_monitor __user *arg);
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 305cae8f497b43..debe9846418a04 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -6009,15 +6009,29 @@ DEFINE_HEALTHMON_EVENT(xfs_healthmon_report_unmount);
#define XFS_HEALTHMON_TYPE_STRINGS \
{ XFS_HEALTHMON_LOST, "lost" }, \
- { XFS_HEALTHMON_UNMOUNT, "unmount" }
+ { XFS_HEALTHMON_UNMOUNT, "unmount" }, \
+ { XFS_HEALTHMON_SICK, "sick" }, \
+ { XFS_HEALTHMON_CORRUPT, "corrupt" }, \
+ { XFS_HEALTHMON_HEALTHY, "healthy" }
#define XFS_HEALTHMON_DOMAIN_STRINGS \
- { XFS_HEALTHMON_MOUNT, "mount" }
+ { XFS_HEALTHMON_MOUNT, "mount" }, \
+ { XFS_HEALTHMON_FS, "fs" }, \
+ { XFS_HEALTHMON_AG, "ag" }, \
+ { XFS_HEALTHMON_INODE, "inode" }, \
+ { XFS_HEALTHMON_RTGROUP, "rtgroup" }
TRACE_DEFINE_ENUM(XFS_HEALTHMON_LOST);
TRACE_DEFINE_ENUM(XFS_HEALTHMON_UNMOUNT);
+TRACE_DEFINE_ENUM(XFS_HEALTHMON_SICK);
+TRACE_DEFINE_ENUM(XFS_HEALTHMON_CORRUPT);
+TRACE_DEFINE_ENUM(XFS_HEALTHMON_HEALTHY);
TRACE_DEFINE_ENUM(XFS_HEALTHMON_MOUNT);
+TRACE_DEFINE_ENUM(XFS_HEALTHMON_FS);
+TRACE_DEFINE_ENUM(XFS_HEALTHMON_AG);
+TRACE_DEFINE_ENUM(XFS_HEALTHMON_INODE);
+TRACE_DEFINE_ENUM(XFS_HEALTHMON_RTGROUP);
DECLARE_EVENT_CLASS(xfs_healthmon_event_class,
TP_PROTO(const struct xfs_healthmon *hm,
@@ -6054,6 +6068,19 @@ DECLARE_EVENT_CLASS(xfs_healthmon_event_class,
break;
}
break;
+ case XFS_HEALTHMON_FS:
+ __entry->mask = event->fsmask;
+ break;
+ case XFS_HEALTHMON_AG:
+ case XFS_HEALTHMON_RTGROUP:
+ __entry->mask = event->grpmask;
+ __entry->group = event->group;
+ break;
+ case XFS_HEALTHMON_INODE:
+ __entry->mask = event->imask;
+ __entry->ino = event->ino;
+ __entry->gen = event->gen;
+ break;
}
),
TP_printk("dev %d:%d type %s domain %s mask 0x%x ino 0x%llx gen 0x%x offset 0x%llx len 0x%llx group 0x%x lost %llu",
@@ -6081,6 +6108,105 @@ DEFINE_HEALTHMONEVENT_EVENT(xfs_healthmon_format_overflow);
DEFINE_HEALTHMONEVENT_EVENT(xfs_healthmon_drop);
DEFINE_HEALTHMONEVENT_EVENT(xfs_healthmon_merge);
+TRACE_EVENT(xfs_healthmon_report_fs,
+ TP_PROTO(const struct xfs_healthmon *hm,
+ unsigned int old_mask, unsigned int new_mask,
+ const struct xfs_healthmon_event *event),
+ TP_ARGS(hm, old_mask, new_mask, event),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(unsigned int, type)
+ __field(unsigned int, domain)
+ __field(unsigned int, old_mask)
+ __field(unsigned int, new_mask)
+ __field(unsigned int, fsmask)
+ ),
+ TP_fast_assign(
+ __entry->dev = hm->dev;
+ __entry->type = event->type;
+ __entry->domain = event->domain;
+ __entry->old_mask = old_mask;
+ __entry->new_mask = new_mask;
+ __entry->fsmask = event->fsmask;
+ ),
+ TP_printk("dev %d:%d type %s domain %s oldmask 0x%x newmask 0x%x fsmask 0x%x",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ __print_symbolic(__entry->type, XFS_HEALTHMON_TYPE_STRINGS),
+ __print_symbolic(__entry->domain, XFS_HEALTHMON_DOMAIN_STRINGS),
+ __entry->old_mask,
+ __entry->new_mask,
+ __entry->fsmask)
+);
+
+TRACE_EVENT(xfs_healthmon_report_group,
+ TP_PROTO(const struct xfs_healthmon *hm,
+ unsigned int old_mask, unsigned int new_mask,
+ const struct xfs_healthmon_event *event),
+ TP_ARGS(hm, old_mask, new_mask, event),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(unsigned int, type)
+ __field(unsigned int, domain)
+ __field(unsigned int, old_mask)
+ __field(unsigned int, new_mask)
+ __field(unsigned int, grpmask)
+ __field(unsigned int, group)
+ ),
+ TP_fast_assign(
+ __entry->dev = hm->dev;
+ __entry->type = event->type;
+ __entry->domain = event->domain;
+ __entry->old_mask = old_mask;
+ __entry->new_mask = new_mask;
+ __entry->grpmask = event->grpmask;
+ __entry->group = event->group;
+ ),
+ TP_printk("dev %d:%d type %s domain %s oldmask 0x%x newmask 0x%x grpmask 0x%x group 0x%x",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ __print_symbolic(__entry->type, XFS_HEALTHMON_TYPE_STRINGS),
+ __print_symbolic(__entry->domain, XFS_HEALTHMON_DOMAIN_STRINGS),
+ __entry->old_mask,
+ __entry->new_mask,
+ __entry->grpmask,
+ __entry->group)
+);
+
+TRACE_EVENT(xfs_healthmon_report_inode,
+ TP_PROTO(const struct xfs_healthmon *hm,
+ unsigned int old_mask, unsigned int new_mask,
+ const struct xfs_healthmon_event *event),
+ TP_ARGS(hm, old_mask, new_mask, event),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(unsigned int, type)
+ __field(unsigned int, domain)
+ __field(unsigned int, old_mask)
+ __field(unsigned int, new_mask)
+ __field(unsigned int, imask)
+ __field(unsigned long long, ino)
+ __field(unsigned int, gen)
+ ),
+ TP_fast_assign(
+ __entry->dev = hm->dev;
+ __entry->type = event->type;
+ __entry->domain = event->domain;
+ __entry->old_mask = old_mask;
+ __entry->new_mask = new_mask;
+ __entry->imask = event->imask;
+ __entry->ino = event->ino;
+ __entry->gen = event->gen;
+ ),
+ TP_printk("dev %d:%d type %s domain %s oldmask 0x%x newmask 0x%x imask 0x%x ino 0x%llx gen 0x%x",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ __print_symbolic(__entry->type, XFS_HEALTHMON_TYPE_STRINGS),
+ __print_symbolic(__entry->domain, XFS_HEALTHMON_DOMAIN_STRINGS),
+ __entry->old_mask,
+ __entry->new_mask,
+ __entry->imask,
+ __entry->ino,
+ __entry->gen)
+);
+
#endif /* _TRACE_XFS_H */
#undef TRACE_INCLUDE_PATH
diff --git a/fs/xfs/xfs_health.c b/fs/xfs/xfs_health.c
index 3d50397f8f7c00..f243c06fd44762 100644
--- a/fs/xfs/xfs_health.c
+++ b/fs/xfs/xfs_health.c
@@ -108,14 +108,19 @@ xfs_fs_mark_sick(
struct xfs_mount *mp,
unsigned int mask)
{
+ unsigned int old_mask;
+
ASSERT(!(mask & ~XFS_SICK_FS_ALL));
trace_xfs_fs_mark_sick(mp, mask);
spin_lock(&mp->m_sb_lock);
+ old_mask = mp->m_fs_sick;
mp->m_fs_sick |= mask;
spin_unlock(&mp->m_sb_lock);
fserror_report_metadata(mp->m_super, -EFSCORRUPTED, GFP_NOFS);
+ if (mask)
+ xfs_healthmon_report_fs(mp, XFS_HEALTHMON_SICK, old_mask, mask);
}
/* Mark per-fs metadata as having been checked and found unhealthy by fsck. */
@@ -124,15 +129,21 @@ xfs_fs_mark_corrupt(
struct xfs_mount *mp,
unsigned int mask)
{
+ unsigned int old_mask;
+
ASSERT(!(mask & ~XFS_SICK_FS_ALL));
trace_xfs_fs_mark_corrupt(mp, mask);
spin_lock(&mp->m_sb_lock);
+ old_mask = mp->m_fs_sick;
mp->m_fs_sick |= mask;
mp->m_fs_checked |= mask;
spin_unlock(&mp->m_sb_lock);
fserror_report_metadata(mp->m_super, -EFSCORRUPTED, GFP_NOFS);
+ if (mask)
+ xfs_healthmon_report_fs(mp, XFS_HEALTHMON_CORRUPT, old_mask,
+ mask);
}
/* Mark a per-fs metadata healed. */
@@ -141,15 +152,22 @@ xfs_fs_mark_healthy(
struct xfs_mount *mp,
unsigned int mask)
{
+ unsigned int old_mask;
+
ASSERT(!(mask & ~XFS_SICK_FS_ALL));
trace_xfs_fs_mark_healthy(mp, mask);
spin_lock(&mp->m_sb_lock);
+ old_mask = mp->m_fs_sick;
mp->m_fs_sick &= ~mask;
if (!(mp->m_fs_sick & XFS_SICK_FS_PRIMARY))
mp->m_fs_sick &= ~XFS_SICK_FS_SECONDARY;
mp->m_fs_checked |= mask;
spin_unlock(&mp->m_sb_lock);
+
+ if (mask)
+ xfs_healthmon_report_fs(mp, XFS_HEALTHMON_HEALTHY, old_mask,
+ mask);
}
/* Sample which per-fs metadata are unhealthy. */
@@ -199,14 +217,20 @@ xfs_group_mark_sick(
struct xfs_group *xg,
unsigned int mask)
{
+ unsigned int old_mask;
+
xfs_group_check_mask(xg, mask);
trace_xfs_group_mark_sick(xg, mask);
spin_lock(&xg->xg_state_lock);
+ old_mask = xg->xg_sick;
xg->xg_sick |= mask;
spin_unlock(&xg->xg_state_lock);
fserror_report_metadata(xg->xg_mount->m_super, -EFSCORRUPTED, GFP_NOFS);
+ if (mask)
+ xfs_healthmon_report_group(xg, XFS_HEALTHMON_SICK, old_mask,
+ mask);
}
/*
@@ -217,15 +241,21 @@ xfs_group_mark_corrupt(
struct xfs_group *xg,
unsigned int mask)
{
+ unsigned int old_mask;
+
xfs_group_check_mask(xg, mask);
trace_xfs_group_mark_corrupt(xg, mask);
spin_lock(&xg->xg_state_lock);
+ old_mask = xg->xg_sick;
xg->xg_sick |= mask;
xg->xg_checked |= mask;
spin_unlock(&xg->xg_state_lock);
fserror_report_metadata(xg->xg_mount->m_super, -EFSCORRUPTED, GFP_NOFS);
+ if (mask)
+ xfs_healthmon_report_group(xg, XFS_HEALTHMON_CORRUPT, old_mask,
+ mask);
}
/*
@@ -236,15 +266,22 @@ xfs_group_mark_healthy(
struct xfs_group *xg,
unsigned int mask)
{
+ unsigned int old_mask;
+
xfs_group_check_mask(xg, mask);
trace_xfs_group_mark_healthy(xg, mask);
spin_lock(&xg->xg_state_lock);
+ old_mask = xg->xg_sick;
xg->xg_sick &= ~mask;
if (!(xg->xg_sick & XFS_SICK_AG_PRIMARY))
xg->xg_sick &= ~XFS_SICK_AG_SECONDARY;
xg->xg_checked |= mask;
spin_unlock(&xg->xg_state_lock);
+
+ if (mask)
+ xfs_healthmon_report_group(xg, XFS_HEALTHMON_HEALTHY, old_mask,
+ mask);
}
/* Sample which per-ag metadata are unhealthy. */
@@ -283,10 +320,13 @@ xfs_inode_mark_sick(
struct xfs_inode *ip,
unsigned int mask)
{
+ unsigned int old_mask;
+
ASSERT(!(mask & ~XFS_SICK_INO_ALL));
trace_xfs_inode_mark_sick(ip, mask);
spin_lock(&ip->i_flags_lock);
+ old_mask = ip->i_sick;
ip->i_sick |= mask;
spin_unlock(&ip->i_flags_lock);
@@ -300,6 +340,9 @@ xfs_inode_mark_sick(
spin_unlock(&VFS_I(ip)->i_lock);
fserror_report_file_metadata(VFS_I(ip), -EFSCORRUPTED, GFP_NOFS);
+ if (mask)
+ xfs_healthmon_report_inode(ip, XFS_HEALTHMON_SICK, old_mask,
+ mask);
}
/* Mark inode metadata as having been checked and found unhealthy by fsck. */
@@ -308,10 +351,13 @@ xfs_inode_mark_corrupt(
struct xfs_inode *ip,
unsigned int mask)
{
+ unsigned int old_mask;
+
ASSERT(!(mask & ~XFS_SICK_INO_ALL));
trace_xfs_inode_mark_corrupt(ip, mask);
spin_lock(&ip->i_flags_lock);
+ old_mask = ip->i_sick;
ip->i_sick |= mask;
ip->i_checked |= mask;
spin_unlock(&ip->i_flags_lock);
@@ -326,6 +372,9 @@ xfs_inode_mark_corrupt(
spin_unlock(&VFS_I(ip)->i_lock);
fserror_report_file_metadata(VFS_I(ip), -EFSCORRUPTED, GFP_NOFS);
+ if (mask)
+ xfs_healthmon_report_inode(ip, XFS_HEALTHMON_CORRUPT, old_mask,
+ mask);
}
/* Mark parts of an inode healed. */
@@ -334,15 +383,22 @@ xfs_inode_mark_healthy(
struct xfs_inode *ip,
unsigned int mask)
{
+ unsigned int old_mask;
+
ASSERT(!(mask & ~XFS_SICK_INO_ALL));
trace_xfs_inode_mark_healthy(ip, mask);
spin_lock(&ip->i_flags_lock);
+ old_mask = ip->i_sick;
ip->i_sick &= ~mask;
if (!(ip->i_sick & XFS_SICK_INO_PRIMARY))
ip->i_sick &= ~XFS_SICK_INO_SECONDARY;
ip->i_checked |= mask;
spin_unlock(&ip->i_flags_lock);
+
+ if (mask)
+ xfs_healthmon_report_inode(ip, XFS_HEALTHMON_HEALTHY, old_mask,
+ mask);
}
/* Sample which parts of an inode are unhealthy. */
@@ -422,6 +478,25 @@ xfs_fsop_geom_health(
}
}
+/*
+ * Translate XFS_SICK_FS_* into XFS_FSOP_GEOM_SICK_* except for the rt free
+ * space codes, which are sent via the rtgroup events.
+ */
+unsigned int
+xfs_healthmon_fs_mask(
+ unsigned int sick_mask)
+{
+ const struct ioctl_sick_map *m;
+ unsigned int ioctl_mask = 0;
+
+ for_each_sick_map(fs_map, m) {
+ if (sick_mask & m->sick_mask)
+ ioctl_mask |= m->ioctl_mask;
+ }
+
+ return ioctl_mask;
+}
+
static const struct ioctl_sick_map ag_map[] = {
{ XFS_SICK_AG_SB, XFS_AG_GEOM_SICK_SB },
{ XFS_SICK_AG_AGF, XFS_AG_GEOM_SICK_AGF },
@@ -458,6 +533,22 @@ xfs_ag_geom_health(
}
}
+/* Translate XFS_SICK_AG_* into XFS_AG_GEOM_SICK_*. */
+unsigned int
+xfs_healthmon_perag_mask(
+ unsigned int sick_mask)
+{
+ const struct ioctl_sick_map *m;
+ unsigned int ioctl_mask = 0;
+
+ for_each_sick_map(ag_map, m) {
+ if (sick_mask & m->sick_mask)
+ ioctl_mask |= m->ioctl_mask;
+ }
+
+ return ioctl_mask;
+}
+
static const struct ioctl_sick_map rtgroup_map[] = {
{ XFS_SICK_RG_SUPER, XFS_RTGROUP_GEOM_SICK_SUPER },
{ XFS_SICK_RG_BITMAP, XFS_RTGROUP_GEOM_SICK_BITMAP },
@@ -488,6 +579,22 @@ xfs_rtgroup_geom_health(
}
}
+/* Translate XFS_SICK_RG_* into XFS_RTGROUP_GEOM_SICK_*. */
+unsigned int
+xfs_healthmon_rtgroup_mask(
+ unsigned int sick_mask)
+{
+ const struct ioctl_sick_map *m;
+ unsigned int ioctl_mask = 0;
+
+ for_each_sick_map(rtgroup_map, m) {
+ if (sick_mask & m->sick_mask)
+ ioctl_mask |= m->ioctl_mask;
+ }
+
+ return ioctl_mask;
+}
+
static const struct ioctl_sick_map ino_map[] = {
{ XFS_SICK_INO_CORE, XFS_BS_SICK_INODE },
{ XFS_SICK_INO_BMBTD, XFS_BS_SICK_BMBTD },
@@ -526,6 +633,22 @@ xfs_bulkstat_health(
}
}
+/* Translate XFS_SICK_INO_* into XFS_BS_SICK_*. */
+unsigned int
+xfs_healthmon_inode_mask(
+ unsigned int sick_mask)
+{
+ const struct ioctl_sick_map *m;
+ unsigned int ioctl_mask = 0;
+
+ for_each_sick_map(ino_map, m) {
+ if (sick_mask & m->sick_mask)
+ ioctl_mask |= m->ioctl_mask;
+ }
+
+ return ioctl_mask;
+}
+
/* Mark a block mapping sick. */
void
xfs_bmap_mark_sick(
diff --git a/fs/xfs/xfs_healthmon.c b/fs/xfs/xfs_healthmon.c
index 122e5a4d8e24cb..20020022b0611c 100644
--- a/fs/xfs/xfs_healthmon.c
+++ b/fs/xfs/xfs_healthmon.c
@@ -18,6 +18,7 @@
#include "xfs_da_btree.h"
#include "xfs_quota_defs.h"
#include "xfs_rtgroup.h"
+#include "xfs_health.h"
#include "xfs_healthmon.h"
#include <linux/anon_inodes.h>
@@ -173,6 +174,33 @@ xfs_healthmon_merge_events(
case XFS_HEALTHMON_LOST:
existing->lostcount += new->lostcount;
return true;
+
+ case XFS_HEALTHMON_SICK:
+ case XFS_HEALTHMON_CORRUPT:
+ case XFS_HEALTHMON_HEALTHY:
+ switch (existing->domain) {
+ case XFS_HEALTHMON_FS:
+ existing->fsmask |= new->fsmask;
+ return true;
+ case XFS_HEALTHMON_AG:
+ case XFS_HEALTHMON_RTGROUP:
+ if (existing->group == new->group){
+ existing->grpmask |= new->grpmask;
+ return true;
+ }
+ return false;
+ case XFS_HEALTHMON_INODE:
+ if (existing->ino == new->ino &&
+ existing->gen == new->gen) {
+ existing->imask |= new->imask;
+ return true;
+ }
+ return false;
+ default:
+ ASSERT(0);
+ return false;
+ }
+ return false;
}
return false;
@@ -353,6 +381,135 @@ xfs_healthmon_unmount(
xfs_healthmon_put(hm);
}
+/* Compute the reporting mask for non-unmount metadata health events. */
+static inline unsigned int
+metadata_event_mask(
+ struct xfs_healthmon *hm,
+ enum xfs_healthmon_type type,
+ unsigned int old_mask,
+ unsigned int new_mask)
+{
+ /* If we want all events, return all events. */
+ if (hm->verbose)
+ return new_mask;
+
+ switch (type) {
+ case XFS_HEALTHMON_SICK:
+ /* Always report runtime corruptions */
+ return new_mask;
+ case XFS_HEALTHMON_CORRUPT:
+ /* Only report new fsck errors */
+ return new_mask & ~old_mask;
+ case XFS_HEALTHMON_HEALTHY:
+ /* Only report healthy metadata that got fixed */
+ return new_mask & old_mask;
+ default:
+ ASSERT(0);
+ break;
+ }
+
+ return 0;
+}
+
+/* Report XFS_FS_SICK_* events to healthmon */
+void
+xfs_healthmon_report_fs(
+ struct xfs_mount *mp,
+ enum xfs_healthmon_type type,
+ unsigned int old_mask,
+ unsigned int new_mask)
+{
+ struct xfs_healthmon_event event = {
+ .type = type,
+ .domain = XFS_HEALTHMON_FS,
+ };
+ struct xfs_healthmon *hm = xfs_healthmon_get(mp);
+
+ if (!hm)
+ return;
+
+ event.fsmask = metadata_event_mask(hm, type, old_mask, new_mask) &
+ ~XFS_SICK_FS_SECONDARY;
+ trace_xfs_healthmon_report_fs(hm, old_mask, new_mask, &event);
+
+ if (event.fsmask)
+ xfs_healthmon_push(hm, &event);
+
+ xfs_healthmon_put(hm);
+}
+
+/* Report XFS_SICK_(AG|RG)* flags to healthmon */
+void
+xfs_healthmon_report_group(
+ struct xfs_group *xg,
+ enum xfs_healthmon_type type,
+ unsigned int old_mask,
+ unsigned int new_mask)
+{
+ struct xfs_healthmon_event event = {
+ .type = type,
+ .group = xg->xg_gno,
+ };
+ struct xfs_healthmon *hm = xfs_healthmon_get(xg->xg_mount);
+
+ if (!hm)
+ return;
+
+ switch (xg->xg_type) {
+ case XG_TYPE_RTG:
+ event.domain = XFS_HEALTHMON_RTGROUP;
+ event.grpmask = metadata_event_mask(hm, type, old_mask,
+ new_mask) &
+ ~XFS_SICK_RG_SECONDARY;
+ break;
+ case XG_TYPE_AG:
+ event.domain = XFS_HEALTHMON_AG;
+ event.grpmask = metadata_event_mask(hm, type, old_mask,
+ new_mask) &
+ ~XFS_SICK_AG_SECONDARY;
+ break;
+ default:
+ ASSERT(0);
+ break;
+ }
+
+ trace_xfs_healthmon_report_group(hm, old_mask, new_mask, &event);
+
+ if (event.grpmask)
+ xfs_healthmon_push(hm, &event);
+
+ xfs_healthmon_put(hm);
+}
+
+/* Report XFS_SICK_INO_* flags to healthmon */
+void
+xfs_healthmon_report_inode(
+ struct xfs_inode *ip,
+ enum xfs_healthmon_type type,
+ unsigned int old_mask,
+ unsigned int new_mask)
+{
+ struct xfs_healthmon_event event = {
+ .type = type,
+ .domain = XFS_HEALTHMON_INODE,
+ .ino = ip->i_ino,
+ .gen = VFS_I(ip)->i_generation,
+ };
+ struct xfs_healthmon *hm = xfs_healthmon_get(ip->i_mount);
+
+ if (!hm)
+ return;
+
+ event.imask = metadata_event_mask(hm, type, old_mask, new_mask) &
+ ~XFS_SICK_INO_SECONDARY;
+ trace_xfs_healthmon_report_inode(hm, old_mask, event.imask, &event);
+
+ if (event.imask)
+ xfs_healthmon_push(hm, &event);
+
+ xfs_healthmon_put(hm);
+}
+
static inline void
xfs_healthmon_reset_outbuf(
struct xfs_healthmon *hm)
@@ -363,11 +520,19 @@ xfs_healthmon_reset_outbuf(
static const unsigned int domain_map[] = {
[XFS_HEALTHMON_MOUNT] = XFS_HEALTH_MONITOR_DOMAIN_MOUNT,
+ [XFS_HEALTHMON_FS] = XFS_HEALTH_MONITOR_DOMAIN_FS,
+ [XFS_HEALTHMON_AG] = XFS_HEALTH_MONITOR_DOMAIN_AG,
+ [XFS_HEALTHMON_INODE] = XFS_HEALTH_MONITOR_DOMAIN_INODE,
+ [XFS_HEALTHMON_RTGROUP] = XFS_HEALTH_MONITOR_DOMAIN_RTGROUP,
};
static const unsigned int type_map[] = {
[XFS_HEALTHMON_RUNNING] = XFS_HEALTH_MONITOR_TYPE_RUNNING,
[XFS_HEALTHMON_LOST] = XFS_HEALTH_MONITOR_TYPE_LOST,
+ [XFS_HEALTHMON_SICK] = XFS_HEALTH_MONITOR_TYPE_SICK,
+ [XFS_HEALTHMON_CORRUPT] = XFS_HEALTH_MONITOR_TYPE_CORRUPT,
+ [XFS_HEALTHMON_HEALTHY] = XFS_HEALTH_MONITOR_TYPE_HEALTHY,
+ [XFS_HEALTHMON_UNMOUNT] = XFS_HEALTH_MONITOR_TYPE_UNMOUNT,
};
/* Render event as a V0 structure */
@@ -400,6 +565,22 @@ xfs_healthmon_format_v0(
break;
}
break;
+ case XFS_HEALTHMON_FS:
+ hme.e.fs.mask = xfs_healthmon_fs_mask(event->fsmask);
+ break;
+ case XFS_HEALTHMON_RTGROUP:
+ hme.e.group.mask = xfs_healthmon_rtgroup_mask(event->grpmask);
+ hme.e.group.gno = event->group;
+ break;
+ case XFS_HEALTHMON_AG:
+ hme.e.group.mask = xfs_healthmon_perag_mask(event->grpmask);
+ hme.e.group.gno = event->group;
+ break;
+ case XFS_HEALTHMON_INODE:
+ hme.e.inode.mask = xfs_healthmon_inode_mask(event->imask);
+ hme.e.inode.ino = event->ino;
+ hme.e.inode.gen = event->gen;
+ break;
default:
break;
}
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH 06/11] xfs: convey filesystem shutdown events to the health monitor
2026-01-13 0:32 [PATCHSET v5] xfs: autonomous self healing of filesystems Darrick J. Wong
` (4 preceding siblings ...)
2026-01-13 0:33 ` [PATCH 05/11] xfs: convey metadata health " Darrick J. Wong
@ 2026-01-13 0:34 ` Darrick J. Wong
2026-01-13 16:14 ` Christoph Hellwig
2026-01-13 0:34 ` [PATCH 07/11] xfs: convey externally discovered fsdax media errors " Darrick J. Wong
` (4 subsequent siblings)
10 siblings, 1 reply; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 0:34 UTC (permalink / raw)
To: djwong, cem; +Cc: hch, linux-fsdevel, linux-xfs
From: Darrick J. Wong <djwong@kernel.org>
Connect the filesystem shutdown code to the health monitor so that xfs
can send events about that to the xfs_healer daemon.
Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
---
fs/xfs/libxfs/xfs_fs.h | 18 ++++++++++++
fs/xfs/xfs_healthmon.h | 7 ++++-
fs/xfs/xfs_trace.h | 23 +++++++++++++++-
fs/xfs/xfs_fsops.c | 15 +++++-----
fs/xfs/xfs_healthmon.c | 70 ++++++++++++++++++++++++++++++++++++++++++++++++
5 files changed, 124 insertions(+), 9 deletions(-)
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 04e1dcf61257d0..c8f7011a7ef8ef 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -1028,6 +1028,9 @@ struct xfs_rtgroup_geometry {
#define XFS_HEALTH_MONITOR_TYPE_CORRUPT (4)
#define XFS_HEALTH_MONITOR_TYPE_HEALTHY (5)
+/* filesystem shutdown */
+#define XFS_HEALTH_MONITOR_TYPE_SHUTDOWN (6)
+
/* lost events */
struct xfs_health_monitor_lost {
__u64 count;
@@ -1054,6 +1057,20 @@ struct xfs_health_monitor_inode {
__u64 ino;
};
+/* shutdown reasons */
+#define XFS_HEALTH_SHUTDOWN_META_IO_ERROR (1u << 0)
+#define XFS_HEALTH_SHUTDOWN_LOG_IO_ERROR (1u << 1)
+#define XFS_HEALTH_SHUTDOWN_FORCE_UMOUNT (1u << 2)
+#define XFS_HEALTH_SHUTDOWN_CORRUPT_INCORE (1u << 3)
+#define XFS_HEALTH_SHUTDOWN_CORRUPT_ONDISK (1u << 4)
+#define XFS_HEALTH_SHUTDOWN_DEVICE_REMOVED (1u << 5)
+
+/* shutdown */
+struct xfs_health_monitor_shutdown {
+ /* XFS_HEALTH_SHUTDOWN_* flags */
+ __u32 reasons;
+};
+
struct xfs_health_monitor_event {
/* XFS_HEALTH_MONITOR_DOMAIN_* */
__u32 domain;
@@ -1074,6 +1091,7 @@ struct xfs_health_monitor_event {
struct xfs_health_monitor_fs fs;
struct xfs_health_monitor_group group;
struct xfs_health_monitor_inode inode;
+ struct xfs_health_monitor_shutdown shutdown;
} e;
/* zeroes */
diff --git a/fs/xfs/xfs_healthmon.h b/fs/xfs/xfs_healthmon.h
index 4632fa873ec087..542e37069c6f56 100644
--- a/fs/xfs/xfs_healthmon.h
+++ b/fs/xfs/xfs_healthmon.h
@@ -69,6 +69,9 @@ enum xfs_healthmon_type {
XFS_HEALTHMON_LOST, /* message lost */
XFS_HEALTHMON_UNMOUNT, /* filesystem is unmounting */
+ /* filesystem shutdown */
+ XFS_HEALTHMON_SHUTDOWN,
+
/* metadata health events */
XFS_HEALTHMON_SICK, /* runtime corruption observed */
XFS_HEALTHMON_CORRUPT, /* fsck reported corruption */
@@ -98,7 +101,7 @@ struct xfs_healthmon_event {
struct {
uint64_t lostcount;
};
- /* mount */
+ /* shutdown */
struct {
unsigned int flags;
};
@@ -133,6 +136,8 @@ void xfs_healthmon_report_inode(struct xfs_inode *ip,
enum xfs_healthmon_type type, unsigned int old_mask,
unsigned int new_mask);
+void xfs_healthmon_report_shutdown(struct xfs_mount *mp, uint32_t flags);
+
long xfs_ioc_health_monitor(struct file *file,
struct xfs_health_monitor __user *arg);
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index debe9846418a04..ec99a6d3dd318c 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -6012,7 +6012,8 @@ DEFINE_HEALTHMON_EVENT(xfs_healthmon_report_unmount);
{ XFS_HEALTHMON_UNMOUNT, "unmount" }, \
{ XFS_HEALTHMON_SICK, "sick" }, \
{ XFS_HEALTHMON_CORRUPT, "corrupt" }, \
- { XFS_HEALTHMON_HEALTHY, "healthy" }
+ { XFS_HEALTHMON_HEALTHY, "healthy" }, \
+ { XFS_HEALTHMON_SHUTDOWN, "shutdown" }
#define XFS_HEALTHMON_DOMAIN_STRINGS \
{ XFS_HEALTHMON_MOUNT, "mount" }, \
@@ -6022,6 +6023,7 @@ DEFINE_HEALTHMON_EVENT(xfs_healthmon_report_unmount);
{ XFS_HEALTHMON_RTGROUP, "rtgroup" }
TRACE_DEFINE_ENUM(XFS_HEALTHMON_LOST);
+TRACE_DEFINE_ENUM(XFS_HEALTHMON_SHUTDOWN);
TRACE_DEFINE_ENUM(XFS_HEALTHMON_UNMOUNT);
TRACE_DEFINE_ENUM(XFS_HEALTHMON_SICK);
TRACE_DEFINE_ENUM(XFS_HEALTHMON_CORRUPT);
@@ -6063,6 +6065,9 @@ DECLARE_EVENT_CLASS(xfs_healthmon_event_class,
switch (__entry->domain) {
case XFS_HEALTHMON_MOUNT:
switch (__entry->type) {
+ case XFS_HEALTHMON_SHUTDOWN:
+ __entry->mask = event->flags;
+ break;
case XFS_HEALTHMON_LOST:
__entry->lostcount = event->lostcount;
break;
@@ -6207,6 +6212,22 @@ TRACE_EVENT(xfs_healthmon_report_inode,
__entry->gen)
);
+TRACE_EVENT(xfs_healthmon_report_shutdown,
+ TP_PROTO(const struct xfs_healthmon *hm, uint32_t shutdown_flags),
+ TP_ARGS(hm, shutdown_flags),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(uint32_t, shutdown_flags)
+ ),
+ TP_fast_assign(
+ __entry->dev = hm->dev;
+ __entry->shutdown_flags = shutdown_flags;
+ ),
+ TP_printk("dev %d:%d shutdown_flags %s",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ __print_flags(__entry->shutdown_flags, "|", XFS_SHUTDOWN_STRINGS))
+);
+
#endif /* _TRACE_XFS_H */
#undef TRACE_INCLUDE_PATH
diff --git a/fs/xfs/xfs_fsops.c b/fs/xfs/xfs_fsops.c
index b7c21f68edc78d..63be01a9443cb0 100644
--- a/fs/xfs/xfs_fsops.c
+++ b/fs/xfs/xfs_fsops.c
@@ -25,6 +25,7 @@
#include "xfs_rtrmap_btree.h"
#include "xfs_rtrefcount_btree.h"
#include "xfs_metafile.h"
+#include "xfs_healthmon.h"
#include <linux/fserror.h>
@@ -497,14 +498,13 @@ xfs_fs_goingdown(
*/
void
xfs_do_force_shutdown(
- struct xfs_mount *mp,
- uint32_t flags,
- char *fname,
- int lnnum)
+ struct xfs_mount *mp,
+ uint32_t flags,
+ char *fname,
+ int lnnum)
{
- int tag;
- const char *why;
-
+ int tag;
+ const char *why;
if (xfs_set_shutdown(mp)) {
xlog_shutdown_wait(mp->m_log);
@@ -544,6 +544,7 @@ xfs_do_force_shutdown(
xfs_stack_trace();
fserror_report_shutdown(mp->m_super, GFP_KERNEL);
+ xfs_healthmon_report_shutdown(mp, flags);
}
/*
diff --git a/fs/xfs/xfs_healthmon.c b/fs/xfs/xfs_healthmon.c
index 20020022b0611c..c9c8d9f10fc49c 100644
--- a/fs/xfs/xfs_healthmon.c
+++ b/fs/xfs/xfs_healthmon.c
@@ -20,6 +20,7 @@
#include "xfs_rtgroup.h"
#include "xfs_health.h"
#include "xfs_healthmon.h"
+#include "xfs_fsops.h"
#include <linux/anon_inodes.h>
#include <linux/eventpoll.h>
@@ -201,6 +202,11 @@ xfs_healthmon_merge_events(
return false;
}
return false;
+
+ case XFS_HEALTHMON_SHUTDOWN:
+ /* yes, we can race to shutdown */
+ existing->flags |= new->flags;
+ return true;
}
return false;
@@ -510,6 +516,28 @@ xfs_healthmon_report_inode(
xfs_healthmon_put(hm);
}
+/* Add a shutdown event to the reporting queue. */
+void
+xfs_healthmon_report_shutdown(
+ struct xfs_mount *mp,
+ uint32_t flags)
+{
+ struct xfs_healthmon_event event = {
+ .type = XFS_HEALTHMON_SHUTDOWN,
+ .domain = XFS_HEALTHMON_MOUNT,
+ .flags = flags,
+ };
+ struct xfs_healthmon *hm = xfs_healthmon_get(mp);
+
+ if (!hm)
+ return;
+
+ trace_xfs_healthmon_report_shutdown(hm, flags);
+
+ xfs_healthmon_push(hm, &event);
+ xfs_healthmon_put(hm);
+}
+
static inline void
xfs_healthmon_reset_outbuf(
struct xfs_healthmon *hm)
@@ -518,6 +546,44 @@ xfs_healthmon_reset_outbuf(
hm->bufhead = 0;
}
+struct flags_map {
+ unsigned int in_mask;
+ unsigned int out_mask;
+};
+
+static const struct flags_map shutdown_map[] = {
+ { SHUTDOWN_META_IO_ERROR, XFS_HEALTH_SHUTDOWN_META_IO_ERROR },
+ { SHUTDOWN_LOG_IO_ERROR, XFS_HEALTH_SHUTDOWN_LOG_IO_ERROR },
+ { SHUTDOWN_FORCE_UMOUNT, XFS_HEALTH_SHUTDOWN_FORCE_UMOUNT },
+ { SHUTDOWN_CORRUPT_INCORE, XFS_HEALTH_SHUTDOWN_CORRUPT_INCORE },
+ { SHUTDOWN_CORRUPT_ONDISK, XFS_HEALTH_SHUTDOWN_CORRUPT_ONDISK },
+ { SHUTDOWN_DEVICE_REMOVED, XFS_HEALTH_SHUTDOWN_DEVICE_REMOVED },
+};
+
+static inline unsigned int
+__map_flags(
+ const struct flags_map *map,
+ size_t array_len,
+ unsigned int flags)
+{
+ const struct flags_map *m;
+ unsigned int ret = 0;
+
+ for (m = map; m < map + array_len; m++) {
+ if (flags & m->in_mask)
+ ret |= m->out_mask;
+ }
+
+ return ret;
+}
+
+#define map_flags(map, flags) __map_flags((map), ARRAY_SIZE(map), (flags))
+
+static inline unsigned int shutdown_mask(unsigned int in)
+{
+ return map_flags(shutdown_map, in);
+}
+
static const unsigned int domain_map[] = {
[XFS_HEALTHMON_MOUNT] = XFS_HEALTH_MONITOR_DOMAIN_MOUNT,
[XFS_HEALTHMON_FS] = XFS_HEALTH_MONITOR_DOMAIN_FS,
@@ -533,6 +599,7 @@ static const unsigned int type_map[] = {
[XFS_HEALTHMON_CORRUPT] = XFS_HEALTH_MONITOR_TYPE_CORRUPT,
[XFS_HEALTHMON_HEALTHY] = XFS_HEALTH_MONITOR_TYPE_HEALTHY,
[XFS_HEALTHMON_UNMOUNT] = XFS_HEALTH_MONITOR_TYPE_UNMOUNT,
+ [XFS_HEALTHMON_SHUTDOWN] = XFS_HEALTH_MONITOR_TYPE_SHUTDOWN,
};
/* Render event as a V0 structure */
@@ -561,6 +628,9 @@ xfs_healthmon_format_v0(
case XFS_HEALTHMON_LOST:
hme.e.lost.count = event->lostcount;
break;
+ case XFS_HEALTHMON_SHUTDOWN:
+ hme.e.shutdown.reasons = shutdown_mask(event->flags);
+ break;
default:
break;
}
^ permalink raw reply related [flat|nested] 36+ messages in thread* Re: [PATCH 06/11] xfs: convey filesystem shutdown events to the health monitor
2026-01-13 0:34 ` [PATCH 06/11] xfs: convey filesystem shutdown " Darrick J. Wong
@ 2026-01-13 16:14 ` Christoph Hellwig
2026-01-13 19:01 ` Darrick J. Wong
0 siblings, 1 reply; 36+ messages in thread
From: Christoph Hellwig @ 2026-01-13 16:14 UTC (permalink / raw)
To: Darrick J. Wong; +Cc: cem, hch, linux-fsdevel, linux-xfs
On Mon, Jan 12, 2026 at 04:34:07PM -0800, Darrick J. Wong wrote:
> - /* mount */
> + /* shutdown */
huh?
> @@ -497,14 +498,13 @@ xfs_fs_goingdown(
> */
> void
> xfs_do_force_shutdown(
> - struct xfs_mount *mp,
> - uint32_t flags,
> - char *fname,
> - int lnnum)
> + struct xfs_mount *mp,
> + uint32_t flags,
> + char *fname,
> + int lnnum)
> {
> - int tag;
> - const char *why;
> -
> + int tag;
> + const char *why;
I like this cleanup, but it seems to be entirely unrelated to
the rest of the patch?
^ permalink raw reply [flat|nested] 36+ messages in thread* Re: [PATCH 06/11] xfs: convey filesystem shutdown events to the health monitor
2026-01-13 16:14 ` Christoph Hellwig
@ 2026-01-13 19:01 ` Darrick J. Wong
0 siblings, 0 replies; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 19:01 UTC (permalink / raw)
To: Christoph Hellwig; +Cc: cem, linux-fsdevel, linux-xfs
On Tue, Jan 13, 2026 at 05:14:42PM +0100, Christoph Hellwig wrote:
> On Mon, Jan 12, 2026 at 04:34:07PM -0800, Darrick J. Wong wrote:
> > - /* mount */
> > + /* shutdown */
>
> huh?
Oh, heh. That part of the union isn't needed until this patch, so I'll
move its definition to this patch.
> > @@ -497,14 +498,13 @@ xfs_fs_goingdown(
> > */
> > void
> > xfs_do_force_shutdown(
> > - struct xfs_mount *mp,
> > - uint32_t flags,
> > - char *fname,
> > - int lnnum)
> > + struct xfs_mount *mp,
> > + uint32_t flags,
> > + char *fname,
> > + int lnnum)
> > {
> > - int tag;
> > - const char *why;
> > -
> > + int tag;
> > + const char *why;
>
> I like this cleanup, but it seems to be entirely unrelated to
> the rest of the patch?
Yeah. I think it was needed back when xfs_do_force_shutdown actually
had to call xfs_healthmon_get to make the new variables line up, but
now that the change is a oneliner I'll revert these whitspace cleanups.
--D
^ permalink raw reply [flat|nested] 36+ messages in thread
* [PATCH 07/11] xfs: convey externally discovered fsdax media errors to the health monitor
2026-01-13 0:32 [PATCHSET v5] xfs: autonomous self healing of filesystems Darrick J. Wong
` (5 preceding siblings ...)
2026-01-13 0:34 ` [PATCH 06/11] xfs: convey filesystem shutdown " Darrick J. Wong
@ 2026-01-13 0:34 ` Darrick J. Wong
2026-01-13 16:15 ` Christoph Hellwig
2026-01-13 0:34 ` [PATCH 08/11] xfs: convey file I/O " Darrick J. Wong
` (3 subsequent siblings)
10 siblings, 1 reply; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 0:34 UTC (permalink / raw)
To: djwong, cem; +Cc: hch, linux-fsdevel, linux-xfs
From: Darrick J. Wong <djwong@kernel.org>
Connect the fsdax media failure notification code to the health monitor
so that xfs can send events about that to the xfs_healer daemon.
Later on we'll add the ability for the xfs_scrub media scan (phase 6) to
report the errors that it finds to the kernel so that those are also
logged by xfs_healer.
Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
---
fs/xfs/libxfs/xfs_fs.h | 15 ++++++++++
fs/xfs/xfs_healthmon.h | 16 ++++++++++
fs/xfs/xfs_trace.h | 38 +++++++++++++++++++++++++
fs/xfs/xfs_healthmon.c | 66 +++++++++++++++++++++++++++++++++++++++++++
fs/xfs/xfs_notify_failure.c | 17 ++++++++---
fs/xfs/xfs_trace.c | 1 +
6 files changed, 148 insertions(+), 5 deletions(-)
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index c8f7011a7ef8ef..38aeb1b0d87b5e 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -1014,6 +1014,11 @@ struct xfs_rtgroup_geometry {
#define XFS_HEALTH_MONITOR_DOMAIN_INODE (3)
#define XFS_HEALTH_MONITOR_DOMAIN_RTGROUP (4)
+/* disk events */
+#define XFS_HEALTH_MONITOR_DOMAIN_DATADEV (5)
+#define XFS_HEALTH_MONITOR_DOMAIN_RTDEV (6)
+#define XFS_HEALTH_MONITOR_DOMAIN_LOGDEV (7)
+
/* Health monitor event types */
/* status of the monitor itself */
@@ -1031,6 +1036,9 @@ struct xfs_rtgroup_geometry {
/* filesystem shutdown */
#define XFS_HEALTH_MONITOR_TYPE_SHUTDOWN (6)
+/* media errors */
+#define XFS_HEALTH_MONITOR_TYPE_MEDIA_ERROR (7)
+
/* lost events */
struct xfs_health_monitor_lost {
__u64 count;
@@ -1071,6 +1079,12 @@ struct xfs_health_monitor_shutdown {
__u32 reasons;
};
+/* disk media errors */
+struct xfs_health_monitor_media {
+ __u64 daddr;
+ __u64 bbcount;
+};
+
struct xfs_health_monitor_event {
/* XFS_HEALTH_MONITOR_DOMAIN_* */
__u32 domain;
@@ -1092,6 +1106,7 @@ struct xfs_health_monitor_event {
struct xfs_health_monitor_group group;
struct xfs_health_monitor_inode inode;
struct xfs_health_monitor_shutdown shutdown;
+ struct xfs_health_monitor_media media;
} e;
/* zeroes */
diff --git a/fs/xfs/xfs_healthmon.h b/fs/xfs/xfs_healthmon.h
index 542e37069c6f56..cbf1e98bde3d00 100644
--- a/fs/xfs/xfs_healthmon.h
+++ b/fs/xfs/xfs_healthmon.h
@@ -76,6 +76,9 @@ enum xfs_healthmon_type {
XFS_HEALTHMON_SICK, /* runtime corruption observed */
XFS_HEALTHMON_CORRUPT, /* fsck reported corruption */
XFS_HEALTHMON_HEALTHY, /* fsck reported healthy structure */
+
+ /* media errors */
+ XFS_HEALTHMON_MEDIA_ERROR,
};
enum xfs_healthmon_domain {
@@ -86,6 +89,11 @@ enum xfs_healthmon_domain {
XFS_HEALTHMON_AG, /* allocation group metadata */
XFS_HEALTHMON_INODE, /* inode metadata */
XFS_HEALTHMON_RTGROUP, /* realtime group metadata */
+
+ /* media errors */
+ XFS_HEALTHMON_DATADEV,
+ XFS_HEALTHMON_RTDEV,
+ XFS_HEALTHMON_LOGDEV,
};
struct xfs_healthmon_event {
@@ -123,6 +131,11 @@ struct xfs_healthmon_event {
uint32_t gen;
xfs_ino_t ino;
};
+ /* media errors */
+ struct {
+ xfs_daddr_t daddr;
+ uint64_t bbcount;
+ };
};
};
@@ -138,6 +151,9 @@ void xfs_healthmon_report_inode(struct xfs_inode *ip,
void xfs_healthmon_report_shutdown(struct xfs_mount *mp, uint32_t flags);
+void xfs_healthmon_report_media(struct xfs_mount *mp, enum xfs_device fdev,
+ xfs_daddr_t daddr, uint64_t bbcount);
+
long xfs_ioc_health_monitor(struct file *file,
struct xfs_health_monitor __user *arg);
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index ec99a6d3dd318c..fe7295a4e917ee 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -6086,6 +6086,12 @@ DECLARE_EVENT_CLASS(xfs_healthmon_event_class,
__entry->ino = event->ino;
__entry->gen = event->gen;
break;
+ case XFS_HEALTHMON_DATADEV:
+ case XFS_HEALTHMON_LOGDEV:
+ case XFS_HEALTHMON_RTDEV:
+ __entry->offset = event->daddr;
+ __entry->length = event->bbcount;
+ break;
}
),
TP_printk("dev %d:%d type %s domain %s mask 0x%x ino 0x%llx gen 0x%x offset 0x%llx len 0x%llx group 0x%x lost %llu",
@@ -6228,6 +6234,38 @@ TRACE_EVENT(xfs_healthmon_report_shutdown,
__print_flags(__entry->shutdown_flags, "|", XFS_SHUTDOWN_STRINGS))
);
+#define XFS_DEVICE_STRINGS \
+ { XFS_DEV_DATA, "datadev" }, \
+ { XFS_DEV_RT, "rtdev" }, \
+ { XFS_DEV_LOG, "logdev" }
+
+TRACE_DEFINE_ENUM(XFS_DEV_DATA);
+TRACE_DEFINE_ENUM(XFS_DEV_RT);
+TRACE_DEFINE_ENUM(XFS_DEV_LOG);
+
+TRACE_EVENT(xfs_healthmon_report_media,
+ TP_PROTO(const struct xfs_healthmon *hm, enum xfs_device fdev,
+ const struct xfs_healthmon_event *event),
+ TP_ARGS(hm, fdev, event),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(unsigned int, error_dev)
+ __field(uint64_t, daddr)
+ __field(uint64_t, bbcount)
+ ),
+ TP_fast_assign(
+ __entry->dev = hm->dev;
+ __entry->error_dev = fdev;
+ __entry->daddr = event->daddr;
+ __entry->bbcount = event->bbcount;
+ ),
+ TP_printk("dev %d:%d %s daddr 0x%llx bbcount 0x%llx",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ __print_symbolic(__entry->error_dev, XFS_DEVICE_STRINGS),
+ __entry->daddr,
+ __entry->bbcount)
+);
+
#endif /* _TRACE_XFS_H */
#undef TRACE_INCLUDE_PATH
diff --git a/fs/xfs/xfs_healthmon.c b/fs/xfs/xfs_healthmon.c
index c9c8d9f10fc49c..88ffbe2e995bd6 100644
--- a/fs/xfs/xfs_healthmon.c
+++ b/fs/xfs/xfs_healthmon.c
@@ -21,6 +21,7 @@
#include "xfs_health.h"
#include "xfs_healthmon.h"
#include "xfs_fsops.h"
+#include "xfs_notify_failure.h"
#include <linux/anon_inodes.h>
#include <linux/eventpoll.h>
@@ -207,6 +208,19 @@ xfs_healthmon_merge_events(
/* yes, we can race to shutdown */
existing->flags |= new->flags;
return true;
+
+ case XFS_HEALTHMON_MEDIA_ERROR:
+ /* physically adjacent errors can merge */
+ if (existing->daddr + existing->bbcount == new->daddr) {
+ existing->bbcount += new->bbcount;
+ return true;
+ }
+ if (new->daddr + new->bbcount == existing->daddr) {
+ existing->daddr = new->daddr;
+ existing->bbcount += new->bbcount;
+ return true;
+ }
+ return false;
}
return false;
@@ -538,6 +552,48 @@ xfs_healthmon_report_shutdown(
xfs_healthmon_put(hm);
}
+static inline enum xfs_healthmon_domain
+media_error_domain(
+ enum xfs_device fdev)
+{
+ switch (fdev) {
+ case XFS_DEV_DATA:
+ return XFS_HEALTHMON_DATADEV;
+ case XFS_DEV_LOG:
+ return XFS_HEALTHMON_LOGDEV;
+ case XFS_DEV_RT:
+ return XFS_HEALTHMON_RTDEV;
+ }
+
+ ASSERT(0);
+ return 0;
+}
+
+/* Add a media error event to the reporting queue. */
+void
+xfs_healthmon_report_media(
+ struct xfs_mount *mp,
+ enum xfs_device fdev,
+ xfs_daddr_t daddr,
+ uint64_t bbcount)
+{
+ struct xfs_healthmon_event event = {
+ .type = XFS_HEALTHMON_MEDIA_ERROR,
+ .domain = media_error_domain(fdev),
+ .daddr = daddr,
+ .bbcount = bbcount,
+ };
+ struct xfs_healthmon *hm = xfs_healthmon_get(mp);
+
+ if (!hm)
+ return;
+
+ trace_xfs_healthmon_report_media(hm, fdev, &event);
+
+ xfs_healthmon_push(hm, &event);
+ xfs_healthmon_put(hm);
+}
+
static inline void
xfs_healthmon_reset_outbuf(
struct xfs_healthmon *hm)
@@ -590,6 +646,9 @@ static const unsigned int domain_map[] = {
[XFS_HEALTHMON_AG] = XFS_HEALTH_MONITOR_DOMAIN_AG,
[XFS_HEALTHMON_INODE] = XFS_HEALTH_MONITOR_DOMAIN_INODE,
[XFS_HEALTHMON_RTGROUP] = XFS_HEALTH_MONITOR_DOMAIN_RTGROUP,
+ [XFS_HEALTHMON_DATADEV] = XFS_HEALTH_MONITOR_DOMAIN_DATADEV,
+ [XFS_HEALTHMON_RTDEV] = XFS_HEALTH_MONITOR_DOMAIN_RTDEV,
+ [XFS_HEALTHMON_LOGDEV] = XFS_HEALTH_MONITOR_DOMAIN_LOGDEV,
};
static const unsigned int type_map[] = {
@@ -600,6 +659,7 @@ static const unsigned int type_map[] = {
[XFS_HEALTHMON_HEALTHY] = XFS_HEALTH_MONITOR_TYPE_HEALTHY,
[XFS_HEALTHMON_UNMOUNT] = XFS_HEALTH_MONITOR_TYPE_UNMOUNT,
[XFS_HEALTHMON_SHUTDOWN] = XFS_HEALTH_MONITOR_TYPE_SHUTDOWN,
+ [XFS_HEALTHMON_MEDIA_ERROR] = XFS_HEALTH_MONITOR_TYPE_MEDIA_ERROR,
};
/* Render event as a V0 structure */
@@ -651,6 +711,12 @@ xfs_healthmon_format_v0(
hme.e.inode.ino = event->ino;
hme.e.inode.gen = event->gen;
break;
+ case XFS_HEALTHMON_DATADEV:
+ case XFS_HEALTHMON_LOGDEV:
+ case XFS_HEALTHMON_RTDEV:
+ hme.e.media.daddr = event->daddr;
+ hme.e.media.bbcount = event->bbcount;
+ break;
default:
break;
}
diff --git a/fs/xfs/xfs_notify_failure.c b/fs/xfs/xfs_notify_failure.c
index 6d5002413c2cb4..1edc4ddd10cdb2 100644
--- a/fs/xfs/xfs_notify_failure.c
+++ b/fs/xfs/xfs_notify_failure.c
@@ -22,6 +22,7 @@
#include "xfs_notify_failure.h"
#include "xfs_rtgroup.h"
#include "xfs_rtrmap_btree.h"
+#include "xfs_healthmon.h"
#include <linux/mm.h>
#include <linux/dax.h>
@@ -219,6 +220,8 @@ xfs_dax_notify_logdev_failure(
if (error)
return error;
+ xfs_healthmon_report_media(mp, XFS_DEV_LOG, daddr, bblen);
+
/*
* In the pre-remove case the failure notification is attempting to
* trigger a force unmount. The expectation is that the device is
@@ -252,16 +255,20 @@ xfs_dax_notify_dev_failure(
uint64_t bblen;
struct xfs_group *xg = NULL;
+ error = xfs_dax_translate_range(xfs_group_type_buftarg(mp, type),
+ offset, len, &daddr, &bblen);
+ if (error)
+ return error;
+
+ xfs_healthmon_report_media(mp,
+ type == XG_TYPE_RTG ? XFS_DEV_RT : XFS_DEV_DATA,
+ daddr, bblen);
+
if (!xfs_has_rmapbt(mp)) {
xfs_debug(mp, "notify_failure() needs rmapbt enabled!");
return -EOPNOTSUPP;
}
- error = xfs_dax_translate_range(xfs_group_type_buftarg(mp, type),
- offset, len, &daddr, &bblen);
- if (error)
- return error;
-
if (type == XG_TYPE_RTG) {
start_bno = xfs_daddr_to_rtb(mp, daddr);
end_bno = xfs_daddr_to_rtb(mp, daddr + bblen - 1);
diff --git a/fs/xfs/xfs_trace.c b/fs/xfs/xfs_trace.c
index d42b864a3837a2..08ddab700a6cd3 100644
--- a/fs/xfs/xfs_trace.c
+++ b/fs/xfs/xfs_trace.c
@@ -53,6 +53,7 @@
#include "xfs_zone_priv.h"
#include "xfs_health.h"
#include "xfs_healthmon.h"
+#include "xfs_notify_failure.h"
/*
* We include this last to have the helpers above available for the trace
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH 08/11] xfs: convey file I/O errors to the health monitor
2026-01-13 0:32 [PATCHSET v5] xfs: autonomous self healing of filesystems Darrick J. Wong
` (6 preceding siblings ...)
2026-01-13 0:34 ` [PATCH 07/11] xfs: convey externally discovered fsdax media errors " Darrick J. Wong
@ 2026-01-13 0:34 ` Darrick J. Wong
2026-01-13 16:15 ` Christoph Hellwig
2026-01-13 0:34 ` [PATCH 09/11] xfs: allow reconfiguration of the health monitoring device Darrick J. Wong
` (2 subsequent siblings)
10 siblings, 1 reply; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 0:34 UTC (permalink / raw)
To: djwong, cem; +Cc: hch, linux-fsdevel, linux-xfs
From: Darrick J. Wong <djwong@kernel.org>
Connect the fserror reporting to the health monitor so that xfs can send
events about file I/O errors to the xfs_healer daemon. These events are
entirely informational because xfs cannot regenerate user data, so
hopefully the fsnotify I/O error event gets noticed by the relevant
management systems.
Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
---
fs/xfs/libxfs/xfs_fs.h | 24 ++++++++++++++
fs/xfs/xfs_healthmon.h | 21 ++++++++++++
fs/xfs/xfs_trace.h | 54 ++++++++++++++++++++++++++++++
fs/xfs/xfs_healthmon.c | 85 ++++++++++++++++++++++++++++++++++++++++++++++++
fs/xfs/xfs_super.c | 12 +++++++
fs/xfs/xfs_trace.c | 2 +
6 files changed, 198 insertions(+)
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 38aeb1b0d87b5e..4ec1b2aede976f 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -1019,6 +1019,9 @@ struct xfs_rtgroup_geometry {
#define XFS_HEALTH_MONITOR_DOMAIN_RTDEV (6)
#define XFS_HEALTH_MONITOR_DOMAIN_LOGDEV (7)
+/* file range events */
+#define XFS_HEALTH_MONITOR_DOMAIN_FILERANGE (8)
+
/* Health monitor event types */
/* status of the monitor itself */
@@ -1039,6 +1042,17 @@ struct xfs_rtgroup_geometry {
/* media errors */
#define XFS_HEALTH_MONITOR_TYPE_MEDIA_ERROR (7)
+/* pagecache I/O to a file range failed */
+#define XFS_HEALTH_MONITOR_TYPE_BUFREAD (8)
+#define XFS_HEALTH_MONITOR_TYPE_BUFWRITE (9)
+
+/* direct I/O to a file range failed */
+#define XFS_HEALTH_MONITOR_TYPE_DIOREAD (10)
+#define XFS_HEALTH_MONITOR_TYPE_DIOWRITE (11)
+
+/* out of band media error reported for a file range */
+#define XFS_HEALTH_MONITOR_TYPE_DATALOST (12)
+
/* lost events */
struct xfs_health_monitor_lost {
__u64 count;
@@ -1079,6 +1093,15 @@ struct xfs_health_monitor_shutdown {
__u32 reasons;
};
+/* file range events */
+struct xfs_health_monitor_filerange {
+ __u64 pos;
+ __u64 len;
+ __u64 ino;
+ __u32 gen;
+ __u32 error;
+};
+
/* disk media errors */
struct xfs_health_monitor_media {
__u64 daddr;
@@ -1107,6 +1130,7 @@ struct xfs_health_monitor_event {
struct xfs_health_monitor_inode inode;
struct xfs_health_monitor_shutdown shutdown;
struct xfs_health_monitor_media media;
+ struct xfs_health_monitor_filerange filerange;
} e;
/* zeroes */
diff --git a/fs/xfs/xfs_healthmon.h b/fs/xfs/xfs_healthmon.h
index cbf1e98bde3d00..215290ade011d5 100644
--- a/fs/xfs/xfs_healthmon.h
+++ b/fs/xfs/xfs_healthmon.h
@@ -79,6 +79,13 @@ enum xfs_healthmon_type {
/* media errors */
XFS_HEALTHMON_MEDIA_ERROR,
+
+ /* file range events */
+ XFS_HEALTHMON_BUFREAD,
+ XFS_HEALTHMON_BUFWRITE,
+ XFS_HEALTHMON_DIOREAD,
+ XFS_HEALTHMON_DIOWRITE,
+ XFS_HEALTHMON_DATALOST,
};
enum xfs_healthmon_domain {
@@ -94,6 +101,9 @@ enum xfs_healthmon_domain {
XFS_HEALTHMON_DATADEV,
XFS_HEALTHMON_RTDEV,
XFS_HEALTHMON_LOGDEV,
+
+ /* file range events */
+ XFS_HEALTHMON_FILERANGE,
};
struct xfs_healthmon_event {
@@ -136,6 +146,14 @@ struct xfs_healthmon_event {
xfs_daddr_t daddr;
uint64_t bbcount;
};
+ /* file range events */
+ struct {
+ xfs_ino_t fino;
+ loff_t fpos;
+ uint64_t flen;
+ uint32_t fgen;
+ int error;
+ };
};
};
@@ -154,6 +172,9 @@ void xfs_healthmon_report_shutdown(struct xfs_mount *mp, uint32_t flags);
void xfs_healthmon_report_media(struct xfs_mount *mp, enum xfs_device fdev,
xfs_daddr_t daddr, uint64_t bbcount);
+void xfs_healthmon_report_file_ioerror(struct xfs_inode *ip,
+ const struct fserror_event *p);
+
long xfs_ioc_health_monitor(struct file *file,
struct xfs_health_monitor __user *arg);
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index fe7295a4e917ee..0cf4877753584f 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -105,6 +105,7 @@ struct xfs_rtgroup;
struct xfs_open_zone;
struct xfs_healthmon_event;
struct xfs_healthmon;
+struct fserror_event;
#define XFS_ATTR_FILTER_FLAGS \
{ XFS_ATTR_ROOT, "ROOT" }, \
@@ -6092,6 +6093,12 @@ DECLARE_EVENT_CLASS(xfs_healthmon_event_class,
__entry->offset = event->daddr;
__entry->length = event->bbcount;
break;
+ case XFS_HEALTHMON_FILERANGE:
+ __entry->ino = event->fino;
+ __entry->gen = event->fgen;
+ __entry->offset = event->fpos;
+ __entry->length = event->flen;
+ break;
}
),
TP_printk("dev %d:%d type %s domain %s mask 0x%x ino 0x%llx gen 0x%x offset 0x%llx len 0x%llx group 0x%x lost %llu",
@@ -6266,6 +6273,53 @@ TRACE_EVENT(xfs_healthmon_report_media,
__entry->bbcount)
);
+#define FS_ERROR_STRINGS \
+ { FSERR_BUFFERED_READ, "buffered_read" }, \
+ { FSERR_BUFFERED_WRITE, "buffered_write" }, \
+ { FSERR_DIRECTIO_READ, "directio_read" }, \
+ { FSERR_DIRECTIO_WRITE, "directio_write" }, \
+ { FSERR_DATA_LOST, "data_lost" }, \
+ { FSERR_METADATA, "metadata" }
+
+TRACE_DEFINE_ENUM(FSERR_BUFFERED_READ);
+TRACE_DEFINE_ENUM(FSERR_BUFFERED_WRITE);
+TRACE_DEFINE_ENUM(FSERR_DIRECTIO_READ);
+TRACE_DEFINE_ENUM(FSERR_DIRECTIO_WRITE);
+TRACE_DEFINE_ENUM(FSERR_DATA_LOST);
+TRACE_DEFINE_ENUM(FSERR_METADATA);
+
+TRACE_EVENT(xfs_healthmon_report_file_ioerror,
+ TP_PROTO(const struct xfs_healthmon *hm,
+ const struct fserror_event *p),
+ TP_ARGS(hm, p),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(unsigned int, type)
+ __field(unsigned long long, ino)
+ __field(unsigned int, gen)
+ __field(long long, pos)
+ __field(unsigned long long, len)
+ __field(int, error)
+ ),
+ TP_fast_assign(
+ __entry->dev = hm->dev;
+ __entry->type = p->type;
+ __entry->ino = XFS_I(p->inode)->i_ino;
+ __entry->gen = p->inode->i_generation;
+ __entry->pos = p->pos;
+ __entry->len = p->len;
+ __entry->error = p->error;
+ ),
+ TP_printk("dev %d:%d ino 0x%llx gen 0x%x op %s pos 0x%llx bytecount 0x%llx error %d",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ __entry->ino,
+ __entry->gen,
+ __print_symbolic(__entry->type, FS_ERROR_STRINGS),
+ __entry->pos,
+ __entry->len,
+ __entry->error)
+);
+
#endif /* _TRACE_XFS_H */
#undef TRACE_INCLUDE_PATH
diff --git a/fs/xfs/xfs_healthmon.c b/fs/xfs/xfs_healthmon.c
index 88ffbe2e995bd6..a4668bb6141530 100644
--- a/fs/xfs/xfs_healthmon.c
+++ b/fs/xfs/xfs_healthmon.c
@@ -22,10 +22,12 @@
#include "xfs_healthmon.h"
#include "xfs_fsops.h"
#include "xfs_notify_failure.h"
+#include "xfs_file.h"
#include <linux/anon_inodes.h>
#include <linux/eventpoll.h>
#include <linux/poll.h>
+#include <linux/fserror.h>
/*
* Live Health Monitoring
@@ -221,6 +223,27 @@ xfs_healthmon_merge_events(
return true;
}
return false;
+
+ case XFS_HEALTHMON_BUFREAD:
+ case XFS_HEALTHMON_BUFWRITE:
+ case XFS_HEALTHMON_DIOREAD:
+ case XFS_HEALTHMON_DIOWRITE:
+ case XFS_HEALTHMON_DATALOST:
+ /* logically adjacent file ranges can merge */
+ if (existing->fino != new->fino || existing->fgen != new->fgen)
+ return false;
+
+ if (existing->fpos + existing->flen == new->fpos) {
+ existing->flen += new->flen;
+ return true;
+ }
+
+ if (new->fpos + new->flen == existing->fpos) {
+ existing->fpos = new->fpos;
+ existing->flen += new->flen;
+ return true;
+ }
+ return false;
}
return false;
@@ -594,6 +617,55 @@ xfs_healthmon_report_media(
xfs_healthmon_put(hm);
}
+static inline enum xfs_healthmon_type file_ioerr_type(enum fserror_type action)
+{
+ switch (action) {
+ case FSERR_BUFFERED_READ:
+ return XFS_HEALTHMON_BUFREAD;
+ case FSERR_BUFFERED_WRITE:
+ return XFS_HEALTHMON_BUFWRITE;
+ case FSERR_DIRECTIO_READ:
+ return XFS_HEALTHMON_DIOREAD;
+ case FSERR_DIRECTIO_WRITE:
+ return XFS_HEALTHMON_DIOWRITE;
+ case FSERR_DATA_LOST:
+ return XFS_HEALTHMON_DATALOST;
+ case FSERR_METADATA:
+ /* filtered out by xfs_fs_report_error */
+ break;
+ }
+
+ ASSERT(0);
+ return -1;
+}
+
+/* Add a file io error event to the reporting queue. */
+void
+xfs_healthmon_report_file_ioerror(
+ struct xfs_inode *ip,
+ const struct fserror_event *p)
+{
+ struct xfs_healthmon_event event = {
+ .type = file_ioerr_type(p->type),
+ .domain = XFS_HEALTHMON_FILERANGE,
+ .fino = ip->i_ino,
+ .fgen = VFS_I(ip)->i_generation,
+ .fpos = p->pos,
+ .flen = p->len,
+ /* send positive error number to userspace */
+ .error = -p->error,
+ };
+ struct xfs_healthmon *hm = xfs_healthmon_get(ip->i_mount);
+
+ if (!hm)
+ return;
+
+ trace_xfs_healthmon_report_file_ioerror(hm, p);
+
+ xfs_healthmon_push(hm, &event);
+ xfs_healthmon_put(hm);
+}
+
static inline void
xfs_healthmon_reset_outbuf(
struct xfs_healthmon *hm)
@@ -649,6 +721,7 @@ static const unsigned int domain_map[] = {
[XFS_HEALTHMON_DATADEV] = XFS_HEALTH_MONITOR_DOMAIN_DATADEV,
[XFS_HEALTHMON_RTDEV] = XFS_HEALTH_MONITOR_DOMAIN_RTDEV,
[XFS_HEALTHMON_LOGDEV] = XFS_HEALTH_MONITOR_DOMAIN_LOGDEV,
+ [XFS_HEALTHMON_FILERANGE] = XFS_HEALTH_MONITOR_DOMAIN_FILERANGE,
};
static const unsigned int type_map[] = {
@@ -660,6 +733,11 @@ static const unsigned int type_map[] = {
[XFS_HEALTHMON_UNMOUNT] = XFS_HEALTH_MONITOR_TYPE_UNMOUNT,
[XFS_HEALTHMON_SHUTDOWN] = XFS_HEALTH_MONITOR_TYPE_SHUTDOWN,
[XFS_HEALTHMON_MEDIA_ERROR] = XFS_HEALTH_MONITOR_TYPE_MEDIA_ERROR,
+ [XFS_HEALTHMON_BUFREAD] = XFS_HEALTH_MONITOR_TYPE_BUFREAD,
+ [XFS_HEALTHMON_BUFWRITE] = XFS_HEALTH_MONITOR_TYPE_BUFWRITE,
+ [XFS_HEALTHMON_DIOREAD] = XFS_HEALTH_MONITOR_TYPE_DIOREAD,
+ [XFS_HEALTHMON_DIOWRITE] = XFS_HEALTH_MONITOR_TYPE_DIOWRITE,
+ [XFS_HEALTHMON_DATALOST] = XFS_HEALTH_MONITOR_TYPE_DATALOST,
};
/* Render event as a V0 structure */
@@ -717,6 +795,13 @@ xfs_healthmon_format_v0(
hme.e.media.daddr = event->daddr;
hme.e.media.bbcount = event->bbcount;
break;
+ case XFS_HEALTHMON_FILERANGE:
+ hme.e.filerange.ino = event->fino;
+ hme.e.filerange.gen = event->fgen;
+ hme.e.filerange.pos = event->fpos;
+ hme.e.filerange.len = event->flen;
+ hme.e.filerange.error = abs(event->error);
+ break;
default:
break;
}
diff --git a/fs/xfs/xfs_super.c b/fs/xfs/xfs_super.c
index bc71aa9dcee8d6..d0cef9ce6b8992 100644
--- a/fs/xfs/xfs_super.c
+++ b/fs/xfs/xfs_super.c
@@ -47,12 +47,14 @@
#include "xfs_parent.h"
#include "xfs_rtalloc.h"
#include "xfs_zone_alloc.h"
+#include "xfs_healthmon.h"
#include "scrub/stats.h"
#include "scrub/rcbag_btree.h"
#include <linux/magic.h>
#include <linux/fs_context.h>
#include <linux/fs_parser.h>
+#include <linux/fserror.h>
static const struct super_operations xfs_super_operations;
@@ -1301,6 +1303,15 @@ xfs_fs_show_stats(
return 0;
}
+static void
+xfs_fs_report_error(
+ const struct fserror_event *event)
+{
+ /* healthmon already knows about non-inode and metadata errors */
+ if (event->inode && event->type != FSERR_METADATA)
+ xfs_healthmon_report_file_ioerror(XFS_I(event->inode), event);
+}
+
static const struct super_operations xfs_super_operations = {
.alloc_inode = xfs_fs_alloc_inode,
.destroy_inode = xfs_fs_destroy_inode,
@@ -1317,6 +1328,7 @@ static const struct super_operations xfs_super_operations = {
.free_cached_objects = xfs_fs_free_cached_objects,
.shutdown = xfs_fs_shutdown,
.show_stats = xfs_fs_show_stats,
+ .report_error = xfs_fs_report_error,
};
static int
diff --git a/fs/xfs/xfs_trace.c b/fs/xfs/xfs_trace.c
index 08ddab700a6cd3..3ae449646eb9b2 100644
--- a/fs/xfs/xfs_trace.c
+++ b/fs/xfs/xfs_trace.c
@@ -54,6 +54,8 @@
#include "xfs_health.h"
#include "xfs_healthmon.h"
#include "xfs_notify_failure.h"
+#include "xfs_file.h"
+#include <linux/fserror.h>
/*
* We include this last to have the helpers above available for the trace
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH 09/11] xfs: allow reconfiguration of the health monitoring device
2026-01-13 0:32 [PATCHSET v5] xfs: autonomous self healing of filesystems Darrick J. Wong
` (7 preceding siblings ...)
2026-01-13 0:34 ` [PATCH 08/11] xfs: convey file I/O " Darrick J. Wong
@ 2026-01-13 0:34 ` Darrick J. Wong
2026-01-13 16:17 ` Christoph Hellwig
2026-01-13 0:35 ` [PATCH 10/11] xfs: check if an open file is on the health monitored fs Darrick J. Wong
2026-01-13 0:35 ` [PATCH 11/11] xfs: add media verification ioctl Darrick J. Wong
10 siblings, 1 reply; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 0:34 UTC (permalink / raw)
To: djwong, cem; +Cc: hch, linux-fsdevel, linux-xfs
From: Darrick J. Wong <djwong@kernel.org>
Make it so that we can reconfigure the health monitoring device by
calling the XFS_IOC_HEALTH_MONITOR ioctl on it. As of right now we can
only toggle the verbose flag, but this is less annoying than having to
closing the monitor fd and reopen it.
Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
---
fs/xfs/xfs_healthmon.c | 44 ++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 44 insertions(+)
diff --git a/fs/xfs/xfs_healthmon.c b/fs/xfs/xfs_healthmon.c
index a4668bb6141530..548c5777d4c596 100644
--- a/fs/xfs/xfs_healthmon.c
+++ b/fs/xfs/xfs_healthmon.c
@@ -23,6 +23,7 @@
#include "xfs_fsops.h"
#include "xfs_notify_failure.h"
#include "xfs_file.h"
+#include "xfs_ioctl.h"
#include <linux/anon_inodes.h>
#include <linux/eventpoll.h>
@@ -1082,12 +1083,55 @@ xfs_healthmon_show_fdinfo(
mutex_unlock(&hm->lock);
}
+/* Reconfigure the health monitor. */
+STATIC long
+xfs_healthmon_reconfigure(
+ struct file *file,
+ unsigned int cmd,
+ void __user *arg)
+{
+ struct xfs_health_monitor hmo;
+ struct xfs_healthmon *hm = file->private_data;
+
+ if (copy_from_user(&hmo, arg, sizeof(hmo)))
+ return -EFAULT;
+
+ if (!xfs_healthmon_validate(&hmo))
+ return -EINVAL;
+
+ mutex_lock(&hm->lock);
+ hm->verbose = !!(hmo.flags & XFS_HEALTH_MONITOR_VERBOSE);
+ mutex_unlock(&hm->lock);
+
+ return 0;
+}
+
+/* Handle ioctls for the health monitoring thread. */
+STATIC long
+xfs_healthmon_ioctl(
+ struct file *file,
+ unsigned int cmd,
+ unsigned long p)
+{
+ void __user *arg = (void __user *)p;
+
+ switch (cmd) {
+ case XFS_IOC_HEALTH_MONITOR:
+ return xfs_healthmon_reconfigure(file, cmd, arg);
+ default:
+ break;
+ }
+
+ return -ENOTTY;
+}
+
static const struct file_operations xfs_healthmon_fops = {
.owner = THIS_MODULE,
.show_fdinfo = xfs_healthmon_show_fdinfo,
.read_iter = xfs_healthmon_read_iter,
.poll = xfs_healthmon_poll,
.release = xfs_healthmon_release,
+ .unlocked_ioctl = xfs_healthmon_ioctl,
};
/*
^ permalink raw reply related [flat|nested] 36+ messages in thread* Re: [PATCH 09/11] xfs: allow reconfiguration of the health monitoring device
2026-01-13 0:34 ` [PATCH 09/11] xfs: allow reconfiguration of the health monitoring device Darrick J. Wong
@ 2026-01-13 16:17 ` Christoph Hellwig
2026-01-13 18:28 ` Darrick J. Wong
0 siblings, 1 reply; 36+ messages in thread
From: Christoph Hellwig @ 2026-01-13 16:17 UTC (permalink / raw)
To: Darrick J. Wong; +Cc: cem, hch, linux-fsdevel, linux-xfs
On Mon, Jan 12, 2026 at 04:34:54PM -0800, Darrick J. Wong wrote:
> From: Darrick J. Wong <djwong@kernel.org>
>
> Make it so that we can reconfigure the health monitoring device by
> calling the XFS_IOC_HEALTH_MONITOR ioctl on it. As of right now we can
> only toggle the verbose flag, but this is less annoying than having to
> closing the monitor fd and reopen it.
"reconfiguration" stills sounds rather pretentious for toggling the
verbose flag, but the code looks good:
Reviewed-by: Christoph Hellwig <hch@lst.de>
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH 09/11] xfs: allow reconfiguration of the health monitoring device
2026-01-13 16:17 ` Christoph Hellwig
@ 2026-01-13 18:28 ` Darrick J. Wong
0 siblings, 0 replies; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 18:28 UTC (permalink / raw)
To: Christoph Hellwig; +Cc: cem, linux-fsdevel, linux-xfs
On Tue, Jan 13, 2026 at 05:17:15PM +0100, Christoph Hellwig wrote:
> On Mon, Jan 12, 2026 at 04:34:54PM -0800, Darrick J. Wong wrote:
> > From: Darrick J. Wong <djwong@kernel.org>
> >
> > Make it so that we can reconfigure the health monitoring device by
> > calling the XFS_IOC_HEALTH_MONITOR ioctl on it. As of right now we can
> > only toggle the verbose flag, but this is less annoying than having to
> > closing the monitor fd and reopen it.
>
> "reconfiguration" stills sounds rather pretentious for toggling the
> verbose flag, but the code looks good:
I'll change the title to:
"xfs: allow toggling verbose logging on the health monitoring file"
> Reviewed-by: Christoph Hellwig <hch@lst.de>
Thanks!
--D
^ permalink raw reply [flat|nested] 36+ messages in thread
* [PATCH 10/11] xfs: check if an open file is on the health monitored fs
2026-01-13 0:32 [PATCHSET v5] xfs: autonomous self healing of filesystems Darrick J. Wong
` (8 preceding siblings ...)
2026-01-13 0:34 ` [PATCH 09/11] xfs: allow reconfiguration of the health monitoring device Darrick J. Wong
@ 2026-01-13 0:35 ` Darrick J. Wong
2026-01-13 16:17 ` Christoph Hellwig
2026-01-13 0:35 ` [PATCH 11/11] xfs: add media verification ioctl Darrick J. Wong
10 siblings, 1 reply; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 0:35 UTC (permalink / raw)
To: djwong, cem; +Cc: hch, linux-fsdevel, linux-xfs
From: Darrick J. Wong <djwong@kernel.org>
Create a new ioctl for the healthmon file that checks that a given fd
points to the same filesystem that the healthmon file is monitoring.
This allows xfs_healer to check that when it reopens a mountpoint to
perform repairs, the file that it gets matches the filesystem that
generated the corruption report.
(Note that xfs_healer doesn't maintain an open fd to a filesystem that
it's monitoring so that it doesn't pin the mount.)
Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
---
fs/xfs/libxfs/xfs_fs.h | 12 +++++++++++-
fs/xfs/xfs_healthmon.c | 34 ++++++++++++++++++++++++++++++++++
2 files changed, 45 insertions(+), 1 deletion(-)
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index 4ec1b2aede976f..a01303c5de6ce6 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -1151,6 +1151,15 @@ struct xfs_health_monitor {
/* Initial return format version */
#define XFS_HEALTH_MONITOR_FMT_V0 (0)
+/*
+ * Check that a given fd points to the same filesystem that the health monitor
+ * is monitoring.
+ */
+struct xfs_health_file_on_monitored_fs {
+ __s32 fd;
+ __u32 flags; /* zero for now */
+};
+
/*
* ioctl commands that are used by Linux filesystems
*/
@@ -1191,7 +1200,8 @@ struct xfs_health_monitor {
#define XFS_IOC_SCRUBV_METADATA _IOWR('X', 64, struct xfs_scrub_vec_head)
#define XFS_IOC_RTGROUP_GEOMETRY _IOWR('X', 65, struct xfs_rtgroup_geometry)
#define XFS_IOC_HEALTH_MONITOR _IOW ('X', 68, struct xfs_health_monitor)
-
+#define XFS_IOC_HEALTH_FD_ON_MONITORED_FS \
+ _IOW ('X', 69, struct xfs_health_file_on_monitored_fs)
/*
* ioctl commands that replace IRIX syssgi()'s
*/
diff --git a/fs/xfs/xfs_healthmon.c b/fs/xfs/xfs_healthmon.c
index 548c5777d4c596..e8e8dc5b489104 100644
--- a/fs/xfs/xfs_healthmon.c
+++ b/fs/xfs/xfs_healthmon.c
@@ -1106,6 +1106,38 @@ xfs_healthmon_reconfigure(
return 0;
}
+/* Does the fd point to the same filesystem as the one we're monitoring? */
+STATIC long
+xfs_healthmon_file_on_monitored_fs(
+ struct file *file,
+ unsigned int cmd,
+ void __user *arg)
+{
+ struct xfs_health_file_on_monitored_fs hms;
+ struct xfs_healthmon *hm = file->private_data;
+ struct inode *hms_inode;
+
+ if (copy_from_user(&hms, arg, sizeof(hms)))
+ return -EFAULT;
+
+ if (hms.flags)
+ return -EINVAL;
+
+ CLASS(fd, hms_fd)(hms.fd);
+ if (fd_empty(hms_fd))
+ return -EBADF;
+
+ hms_inode = file_inode(fd_file(hms_fd));
+ mutex_lock(&hm->lock);
+ if (hm->mount_cookie != (uintptr_t)hms_inode->i_sb) {
+ mutex_unlock(&hm->lock);
+ return -ESTALE;
+ }
+
+ mutex_unlock(&hm->lock);
+ return 0;
+}
+
/* Handle ioctls for the health monitoring thread. */
STATIC long
xfs_healthmon_ioctl(
@@ -1118,6 +1150,8 @@ xfs_healthmon_ioctl(
switch (cmd) {
case XFS_IOC_HEALTH_MONITOR:
return xfs_healthmon_reconfigure(file, cmd, arg);
+ case XFS_IOC_HEALTH_FD_ON_MONITORED_FS:
+ return xfs_healthmon_file_on_monitored_fs(file, cmd, arg);
default:
break;
}
^ permalink raw reply related [flat|nested] 36+ messages in thread* [PATCH 11/11] xfs: add media verification ioctl
2026-01-13 0:32 [PATCHSET v5] xfs: autonomous self healing of filesystems Darrick J. Wong
` (9 preceding siblings ...)
2026-01-13 0:35 ` [PATCH 10/11] xfs: check if an open file is on the health monitored fs Darrick J. Wong
@ 2026-01-13 0:35 ` Darrick J. Wong
2026-01-13 15:57 ` Christoph Hellwig
10 siblings, 1 reply; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 0:35 UTC (permalink / raw)
To: djwong, cem; +Cc: hch, linux-fsdevel, linux-xfs
From: Darrick J. Wong <djwong@kernel.org>
Add a new privileged ioctl so that xfs_scrub can ask the kernel to
verify the media of the devices backing an xfs filesystem, and have any
resulting media errors reported to fsnotify and xfs_healer.
Signed-off-by: "Darrick J. Wong" <djwong@kernel.org>
---
fs/xfs/libxfs/xfs_fs.h | 27 +++
fs/xfs/xfs_notify_failure.h | 4
fs/xfs/xfs_trace.h | 97 +++++++++++
fs/xfs/Makefile | 6 -
fs/xfs/xfs_ioctl.c | 3
fs/xfs/xfs_notify_failure.c | 375 +++++++++++++++++++++++++++++++++++++++++++
6 files changed, 507 insertions(+), 5 deletions(-)
diff --git a/fs/xfs/libxfs/xfs_fs.h b/fs/xfs/libxfs/xfs_fs.h
index a01303c5de6ce6..10fd94e3b9f5b4 100644
--- a/fs/xfs/libxfs/xfs_fs.h
+++ b/fs/xfs/libxfs/xfs_fs.h
@@ -1160,6 +1160,31 @@ struct xfs_health_file_on_monitored_fs {
__u32 flags; /* zero for now */
};
+/* Verify the media of the underlying devices */
+struct xfs_verify_media {
+ __u32 dev; /* I: XFS_VERIFY_*DEV */
+ __u32 flags; /* I: all other XFS_VERIFY_* */
+
+ /* IO: inclusive start of disk range to verify, in 512b blocks */
+ __u64 start_daddr;
+ /*
+ * I: exclusive end of the disk range to verify, in 512b blocks
+ * or XFS_VERIFY_TO_EOD to verify to the end of the device
+ */
+ __u64 end_daddr;
+
+ __u32 ioerror; /* O: I/O error (positive) */
+ __u32 pad; /* zero */
+};
+
+#define XFS_VERIFY_DATADEV (1) /* data device */
+#define XFS_VERIFY_LOGDEV (2) /* external log device */
+#define XFS_VERIFY_RTDEV (3) /* realtime device */
+
+#define XFS_VERIFY_TO_EOD (~0ULL) /* end of disk */
+
+#define XFS_VERIFY_REPORT_ERRORS (1 << 0) /* report to fsnotify */
+
/*
* ioctl commands that are used by Linux filesystems
*/
@@ -1202,6 +1227,8 @@ struct xfs_health_file_on_monitored_fs {
#define XFS_IOC_HEALTH_MONITOR _IOW ('X', 68, struct xfs_health_monitor)
#define XFS_IOC_HEALTH_FD_ON_MONITORED_FS \
_IOW ('X', 69, struct xfs_health_file_on_monitored_fs)
+#define XFS_IOC_VERIFY_MEDIA _IOWR('X', 70, struct xfs_verify_media)
+
/*
* ioctl commands that replace IRIX syssgi()'s
*/
diff --git a/fs/xfs/xfs_notify_failure.h b/fs/xfs/xfs_notify_failure.h
index 8d08ec29dd2949..042d78a2979c60 100644
--- a/fs/xfs/xfs_notify_failure.h
+++ b/fs/xfs/xfs_notify_failure.h
@@ -8,4 +8,8 @@
extern const struct dax_holder_operations xfs_dax_holder_operations;
+struct xfs_verify_media;
+int xfs_ioc_verify_media(struct file *file,
+ struct xfs_verify_media __user *arg);
+
#endif /* __XFS_NOTIFY_FAILURE_H__ */
diff --git a/fs/xfs/xfs_trace.h b/fs/xfs/xfs_trace.h
index 0cf4877753584f..9ab97f93119198 100644
--- a/fs/xfs/xfs_trace.h
+++ b/fs/xfs/xfs_trace.h
@@ -6320,6 +6320,103 @@ TRACE_EVENT(xfs_healthmon_report_file_ioerror,
__entry->error)
);
+TRACE_EVENT(xfs_verify_media,
+ TP_PROTO(const struct xfs_mount *mp, const struct xfs_verify_media *me,
+ dev_t fdev, xfs_daddr_t bio_daddr, uint64_t bio_bbcount,
+ unsigned int bufsize),
+ TP_ARGS(mp, me, fdev, bio_daddr, bio_bbcount, bufsize),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(dev_t, fdev)
+ __field(xfs_daddr_t, start_daddr)
+ __field(xfs_daddr_t, end_daddr)
+ __field(unsigned int, flags)
+ __field(xfs_daddr_t, bio_daddr)
+ __field(uint64_t, bio_bbcount)
+ __field(unsigned int, bufsize)
+ ),
+ TP_fast_assign(
+ __entry->dev = mp->m_ddev_targp->bt_dev;
+ __entry->fdev = fdev;
+ __entry->start_daddr = me->start_daddr;
+ __entry->end_daddr = me->end_daddr;
+ __entry->flags = me->flags;
+ __entry->bio_daddr = bio_daddr;
+ __entry->bio_bbcount = bio_bbcount;
+ __entry->bufsize = bufsize;
+ ),
+ TP_printk("dev %d:%d fdev %d:%d start_daddr 0x%llx end_daddr 0x%llx flags 0x%x bio_daddr 0x%llx bio_bbcount 0x%llx bufsize 0x%x",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ MAJOR(__entry->fdev), MINOR(__entry->fdev),
+ __entry->start_daddr,
+ __entry->end_daddr,
+ __entry->flags,
+ __entry->bio_daddr,
+ __entry->bio_bbcount,
+ __entry->bufsize)
+);
+
+TRACE_EVENT(xfs_verify_media_end,
+ TP_PROTO(const struct xfs_mount *mp, const struct xfs_verify_media *me,
+ dev_t fdev),
+ TP_ARGS(mp, me, fdev),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(dev_t, fdev)
+ __field(xfs_daddr_t, start_daddr)
+ __field(xfs_daddr_t, end_daddr)
+ __field(int, ioerror)
+ ),
+ TP_fast_assign(
+ __entry->dev = mp->m_ddev_targp->bt_dev;
+ __entry->fdev = fdev;
+ __entry->start_daddr = me->start_daddr;
+ __entry->end_daddr = me->end_daddr;
+ __entry->ioerror = me->ioerror;
+ ),
+ TP_printk("dev %d:%d fdev %d:%d start_daddr 0x%llx end_daddr 0x%llx ioerror %d",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ MAJOR(__entry->fdev), MINOR(__entry->fdev),
+ __entry->start_daddr,
+ __entry->end_daddr,
+ __entry->ioerror)
+);
+
+TRACE_EVENT(xfs_verify_media_error,
+ TP_PROTO(const struct xfs_mount *mp, const struct xfs_verify_media *me,
+ dev_t fdev, xfs_daddr_t daddr, uint64_t bbcount, int error),
+ TP_ARGS(mp, me, fdev, daddr, bbcount, error),
+ TP_STRUCT__entry(
+ __field(dev_t, dev)
+ __field(dev_t, fdev)
+ __field(xfs_daddr_t, start_daddr)
+ __field(xfs_daddr_t, end_daddr)
+ __field(unsigned int, flags)
+ __field(xfs_daddr_t, daddr)
+ __field(uint64_t, bbcount)
+ __field(int, error)
+ ),
+ TP_fast_assign(
+ __entry->dev = mp->m_ddev_targp->bt_dev;
+ __entry->fdev = fdev;
+ __entry->start_daddr = me->start_daddr;
+ __entry->end_daddr = me->end_daddr;
+ __entry->flags = me->flags;
+ __entry->daddr = daddr;
+ __entry->bbcount = bbcount;
+ __entry->error = error;
+ ),
+ TP_printk("dev %d:%d fdev %d:%d start_daddr 0x%llx end_daddr 0x%llx flags 0x%x daddr 0x%llx bbcount 0x%llx error %d",
+ MAJOR(__entry->dev), MINOR(__entry->dev),
+ MAJOR(__entry->fdev), MINOR(__entry->fdev),
+ __entry->start_daddr,
+ __entry->end_daddr,
+ __entry->flags,
+ __entry->daddr,
+ __entry->bbcount,
+ __entry->error)
+);
+
#endif /* _TRACE_XFS_H */
#undef TRACE_INCLUDE_PATH
diff --git a/fs/xfs/Makefile b/fs/xfs/Makefile
index 1b7385e23b3463..a879d1d2155ed7 100644
--- a/fs/xfs/Makefile
+++ b/fs/xfs/Makefile
@@ -99,6 +99,7 @@ xfs-y += xfs_aops.o \
xfs_message.o \
xfs_mount.o \
xfs_mru_cache.o \
+ xfs_notify_failure.o \
xfs_pwork.o \
xfs_reflink.o \
xfs_stats.o \
@@ -149,11 +150,6 @@ xfs-$(CONFIG_SYSCTL) += xfs_sysctl.o
xfs-$(CONFIG_COMPAT) += xfs_ioctl32.o
xfs-$(CONFIG_EXPORTFS_BLOCK_OPS) += xfs_pnfs.o
-# notify failure
-ifeq ($(CONFIG_MEMORY_FAILURE),y)
-xfs-$(CONFIG_FS_DAX) += xfs_notify_failure.o
-endif
-
xfs-$(CONFIG_XFS_DRAIN_INTENTS) += xfs_drain.o
xfs-$(CONFIG_XFS_LIVE_HOOKS) += xfs_hooks.o
xfs-$(CONFIG_XFS_MEMORY_BUFS) += xfs_buf_mem.o
diff --git a/fs/xfs/xfs_ioctl.c b/fs/xfs/xfs_ioctl.c
index c04c41ca924e37..b74b9e0f023bf8 100644
--- a/fs/xfs/xfs_ioctl.c
+++ b/fs/xfs/xfs_ioctl.c
@@ -42,6 +42,7 @@
#include "xfs_handle.h"
#include "xfs_rtgroup.h"
#include "xfs_healthmon.h"
+#include "xfs_notify_failure.h"
#include <linux/mount.h>
#include <linux/fileattr.h>
@@ -1422,6 +1423,8 @@ xfs_file_ioctl(
case XFS_IOC_HEALTH_MONITOR:
return xfs_ioc_health_monitor(filp, arg);
+ case XFS_IOC_VERIFY_MEDIA:
+ return xfs_ioc_verify_media(filp, arg);
default:
return -ENOTTY;
diff --git a/fs/xfs/xfs_notify_failure.c b/fs/xfs/xfs_notify_failure.c
index 1edc4ddd10cdb2..5ef4109cc062d2 100644
--- a/fs/xfs/xfs_notify_failure.c
+++ b/fs/xfs/xfs_notify_failure.c
@@ -23,12 +23,14 @@
#include "xfs_rtgroup.h"
#include "xfs_rtrmap_btree.h"
#include "xfs_healthmon.h"
+#include "xfs_trace.h"
#include <linux/mm.h>
#include <linux/dax.h>
#include <linux/fs.h>
#include <linux/fserror.h>
+#if defined(CONFIG_MEMORY_FAILURE) && defined(CONFIG_FS_DAX)
struct xfs_failure_info {
xfs_agblock_t startblock;
xfs_extlen_t blockcount;
@@ -395,3 +397,376 @@ xfs_dax_notify_failure(
const struct dax_holder_operations xfs_dax_holder_operations = {
.notify_failure = xfs_dax_notify_failure,
};
+#endif /* CONFIG_MEMORY_FAILURE && CONFIG_FS_DAX */
+
+struct xfs_group_data_lost {
+ xfs_agblock_t startblock;
+ xfs_extlen_t blockcount;
+};
+
+static int
+xfs_report_one_data_lost(
+ struct xfs_btree_cur *cur,
+ const struct xfs_rmap_irec *rec,
+ void *data)
+{
+ struct xfs_mount *mp = cur->bc_mp;
+ struct xfs_inode *ip;
+ struct xfs_group_data_lost *lost = data;
+ xfs_fileoff_t fileoff = rec->rm_offset;
+ xfs_extlen_t blocks = rec->rm_blockcount;
+ const xfs_agblock_t lost_end =
+ lost->startblock + lost->blockcount;
+ const xfs_agblock_t rmap_end =
+ rec->rm_startblock + rec->rm_blockcount;
+ int error = 0;
+
+ if (XFS_RMAP_NON_INODE_OWNER(rec->rm_owner) ||
+ (rec->rm_flags & (XFS_RMAP_ATTR_FORK | XFS_RMAP_BMBT_BLOCK)))
+ return 0;
+
+ error = xfs_iget(mp, cur->bc_tp, rec->rm_owner, 0, 0, &ip);
+ if (error)
+ return 0;
+
+ if (lost->startblock > rec->rm_startblock) {
+ fileoff += lost->startblock - rec->rm_startblock;
+ blocks -= lost->startblock - rec->rm_startblock;
+ }
+ if (rmap_end > lost_end)
+ blocks -= rmap_end - lost_end;
+
+ fserror_report_data_lost(VFS_I(ip), XFS_FSB_TO_B(mp, fileoff),
+ XFS_FSB_TO_B(mp, blocks), GFP_NOFS);
+
+ xfs_irele(ip);
+ return 0;
+}
+
+static int
+xfs_report_data_lost(
+ struct xfs_mount *mp,
+ enum xfs_group_type type,
+ xfs_daddr_t daddr,
+ u64 bblen)
+{
+ struct xfs_group *xg = NULL;
+ struct xfs_trans *tp;
+ xfs_fsblock_t start_bno, end_bno;
+ uint32_t start_gno, end_gno;
+ int error;
+
+ if (type == XG_TYPE_RTG) {
+ start_bno = xfs_daddr_to_rtb(mp, daddr);
+ end_bno = xfs_daddr_to_rtb(mp, daddr + bblen - 1);
+ } else {
+ start_bno = XFS_DADDR_TO_FSB(mp, daddr);
+ end_bno = XFS_DADDR_TO_FSB(mp, daddr + bblen - 1);
+ }
+
+ tp = xfs_trans_alloc_empty(mp);
+ start_gno = xfs_fsb_to_gno(mp, start_bno, type);
+ end_gno = xfs_fsb_to_gno(mp, end_bno, type);
+ while ((xg = xfs_group_next_range(mp, xg, start_gno, end_gno, type))) {
+ struct xfs_buf *agf_bp = NULL;
+ struct xfs_rtgroup *rtg = NULL;
+ struct xfs_btree_cur *cur;
+ struct xfs_rmap_irec ri_low = { };
+ struct xfs_rmap_irec ri_high;
+ struct xfs_group_data_lost lost;
+
+ if (type == XG_TYPE_AG) {
+ struct xfs_perag *pag = to_perag(xg);
+
+ error = xfs_alloc_read_agf(pag, tp, 0, &agf_bp);
+ if (error) {
+ xfs_perag_put(pag);
+ break;
+ }
+
+ cur = xfs_rmapbt_init_cursor(mp, tp, agf_bp, pag);
+ } else {
+ rtg = to_rtg(xg);
+ xfs_rtgroup_lock(rtg, XFS_RTGLOCK_RMAP);
+ cur = xfs_rtrmapbt_init_cursor(tp, rtg);
+ }
+
+ /*
+ * Set the rmap range from ri_low to ri_high, which represents
+ * a [start, end] where we looking for the files or metadata.
+ */
+ memset(&ri_high, 0xFF, sizeof(ri_high));
+ if (xg->xg_gno == start_gno)
+ ri_low.rm_startblock =
+ xfs_fsb_to_gbno(mp, start_bno, type);
+ if (xg->xg_gno == end_gno)
+ ri_high.rm_startblock =
+ xfs_fsb_to_gbno(mp, end_bno, type);
+
+ lost.startblock = ri_low.rm_startblock;
+ lost.blockcount = min(xg->xg_block_count,
+ ri_high.rm_startblock + 1) -
+ ri_low.rm_startblock;
+
+ error = xfs_rmap_query_range(cur, &ri_low, &ri_high,
+ xfs_report_one_data_lost, &lost);
+ xfs_btree_del_cursor(cur, error);
+ if (agf_bp)
+ xfs_trans_brelse(tp, agf_bp);
+ if (rtg)
+ xfs_rtgroup_unlock(rtg, XFS_RTGLOCK_RMAP);
+ if (error) {
+ xfs_group_put(xg);
+ break;
+ }
+ }
+
+ xfs_trans_cancel(tp);
+ return 0;
+}
+
+/* Verify the media of an xfs device by submitting read requests to the disk. */
+static int
+xfs_verify_media(
+ struct xfs_mount *mp,
+ enum xfs_device fdev,
+ struct xfs_verify_media *me)
+{
+ struct blk_plug plug;
+ struct xfs_buftarg *btp = NULL;
+ struct bio *bio = NULL;
+ struct folio *folio;
+ xfs_daddr_t new_start_daddr = me->start_daddr;
+ xfs_daddr_t bio_daddr;
+ uint64_t bio_bbcount;
+ const unsigned int iosize = BIO_MAX_VECS << PAGE_SHIFT;
+ unsigned int bufsize = iosize;
+ unsigned int bio_submitted = 0;
+ enum xfs_group_type group;
+ int error = 0;
+
+ me->ioerror = 0;
+
+ switch (fdev) {
+ case XFS_DEV_DATA:
+ btp = mp->m_ddev_targp;
+ break;
+ case XFS_DEV_LOG:
+ if (mp->m_logdev_targp->bt_bdev != mp->m_ddev_targp->bt_bdev)
+ btp = mp->m_logdev_targp;
+ break;
+ case XFS_DEV_RT:
+ btp = mp->m_rtdev_targp;
+ break;
+ }
+ if (!btp)
+ return -ENODEV;
+
+ /*
+ * If the caller told us to verify to EOD, tell the user exactly where
+ * that was.
+ */
+ if (me->end_daddr == XFS_VERIFY_TO_EOD)
+ me->end_daddr = btp->bt_nr_sectors;
+
+ if (me->start_daddr > me->end_daddr)
+ return 0;
+
+ /* We've already hit EOD, so advance to the end */
+ if (me->start_daddr >= btp->bt_nr_sectors) {
+ me->start_daddr = me->end_daddr;
+ return 0;
+ }
+
+ bio_daddr = me->start_daddr;
+ bio_bbcount = min_t(sector_t, me->end_daddr, btp->bt_nr_sectors) -
+ me->start_daddr;
+
+ /*
+ * There are three ranges involved here:
+ *
+ * - [me->start_daddr, me->end_daddr) is the range that the user wants
+ * to verify. If me->end_daddr is XFS_VERIFY_TO_EOD then that means
+ * to verify to the end of the disk.
+ *
+ * - [new_start_daddr, me->end_daddr) is the range that we have not
+ * yet verified. We update new_start_daddr after each successful
+ * read. me->start_daddr is set to new_start_daddr before
+ * returning.
+ *
+ * - [bio_daddr, bio_daddr + bio_bbcount) is the range of the next
+ * read bio(s) that we'll submit.
+ *
+ * Try to create bios of maximal size, whether we allocate one large
+ * folio of that size, or a single page.
+ */
+ if ((bufsize >> SECTOR_SHIFT) > bio_bbcount)
+ bufsize = bio_bbcount << SECTOR_SHIFT;
+
+ folio = folio_alloc(GFP_KERNEL, get_order(bufsize));
+ if (!folio)
+ folio = folio_alloc(GFP_KERNEL, 0);
+ if (!folio)
+ return -ENOMEM;
+ bufsize = folio_size(folio);
+
+ trace_xfs_verify_media(mp, me, btp->bt_bdev->bd_dev, bio_daddr,
+ bio_bbcount, bufsize);
+
+ blk_start_plug(&plug);
+ while (bio_bbcount > 0) {
+ unsigned int nr_sects =
+ min_t(sector_t, bio_bbcount, iosize >> SECTOR_SHIFT);
+ const unsigned int nr_vecs =
+ howmany(nr_sects << SECTOR_SHIFT, bufsize);
+ unsigned int i;
+
+ bio = blk_next_bio(bio, btp->bt_bdev, nr_vecs, REQ_OP_READ,
+ GFP_KERNEL);
+ if (!bio) {
+ error = -ENOMEM;
+ goto out_folio;
+ }
+ bio->bi_iter.bi_sector = bio_daddr;
+
+ for (i = 0; i < nr_vecs; i++) {
+ unsigned int vec_sects =
+ min(nr_sects, bufsize >> SECTOR_SHIFT);
+
+ bio_add_folio_nofail(bio, folio,
+ vec_sects << SECTOR_SHIFT, 0);
+
+ bio_daddr += vec_sects;
+ bio_bbcount -= vec_sects;
+ bio_submitted += vec_sects;
+ }
+
+ /* Don't let too many IOs accumulate */
+ if (bio_submitted > SZ_256M >> SECTOR_SHIFT) {
+ blk_finish_plug(&plug);
+ error = submit_bio_wait(bio);
+ if (error)
+ goto media_error;
+ bio_put(bio);
+ bio = NULL;
+
+ if (fatal_signal_pending(current)) {
+ error = -EINTR;
+ goto out_folio;
+ }
+
+ cond_resched();
+ new_start_daddr += bio_submitted;
+ bio_submitted = 0;
+ blk_start_plug(&plug);
+ }
+
+ }
+ blk_finish_plug(&plug);
+
+ /* Finish up a partially constructed bio if there is one */
+ if (!bio)
+ goto out_folio;
+
+ error = submit_bio_wait(bio);
+ if (error)
+ goto media_error;
+
+ new_start_daddr += bio_submitted;
+
+out_bio:
+ bio_put(bio);
+out_folio:
+ folio_put(folio);
+ if (!error) {
+ /*
+ * Advance start_daddr to the end of what we verified if there
+ * wasn't an operational error; or to end_daddr if we reached
+ * the end of the disk.
+ */
+ me->start_daddr = new_start_daddr;
+ if (me->start_daddr >= btp->bt_nr_sectors)
+ me->start_daddr = me->end_daddr;
+ }
+
+ trace_xfs_verify_media_end(mp, me, btp->bt_bdev->bd_dev);
+ return error;
+
+media_error:
+ trace_xfs_verify_media_error(mp, me, btp->bt_bdev->bd_dev,
+ new_start_daddr, bio_submitted, error);
+
+ /* Only report the I/O error if we didn't verify any bytes at all. */
+ if (me->start_daddr == new_start_daddr)
+ me->ioerror = -error;
+ error = 0;
+
+ if (!(me->flags & XFS_VERIFY_REPORT_ERRORS))
+ goto out_bio;
+
+ xfs_healthmon_report_media(mp, fdev, new_start_daddr, bio_submitted);
+
+ if (!xfs_has_rmapbt(mp))
+ goto out_bio;
+
+ switch (fdev) {
+ case XFS_DEV_DATA:
+ group = XG_TYPE_AG;
+ break;
+ case XFS_DEV_RT:
+ group = XG_TYPE_RTG;
+ break;
+ default:
+ goto out_bio;
+ }
+
+ xfs_report_data_lost(mp, group, new_start_daddr, bio_submitted);
+ goto out_bio;
+}
+
+#define XFS_VALID_VERIFY_MEDIA_FLAGS (XFS_VERIFY_REPORT_ERRORS)
+int
+xfs_ioc_verify_media(
+ struct file *file,
+ struct xfs_verify_media __user *arg)
+{
+ struct xfs_verify_media me;
+ struct xfs_inode *ip = XFS_I(file_inode(file));
+ struct xfs_mount *mp = ip->i_mount;
+ enum xfs_device fdev;
+ int error;
+
+ if (!capable(CAP_SYS_ADMIN))
+ return -EPERM;
+
+ if (copy_from_user(&me, arg, sizeof(me)))
+ return -EFAULT;
+
+ if (me.pad)
+ return -EINVAL;
+ if (me.flags & ~XFS_VALID_VERIFY_MEDIA_FLAGS)
+ return -EINVAL;
+
+ switch (me.dev) {
+ case XFS_VERIFY_DATADEV:
+ fdev = XFS_DEV_DATA;
+ break;
+ case XFS_VERIFY_RTDEV:
+ fdev = XFS_DEV_RT;
+ break;
+ case XFS_VERIFY_LOGDEV:
+ fdev = XFS_DEV_LOG;
+ break;
+ default:
+ return -EINVAL;
+ }
+
+ error = xfs_verify_media(mp, fdev, &me);
+ if (error)
+ return error;
+
+ if (copy_to_user(arg, &me, sizeof(me)))
+ return -EFAULT;
+
+ return 0;
+}
^ permalink raw reply related [flat|nested] 36+ messages in thread* Re: [PATCH 11/11] xfs: add media verification ioctl
2026-01-13 0:35 ` [PATCH 11/11] xfs: add media verification ioctl Darrick J. Wong
@ 2026-01-13 15:57 ` Christoph Hellwig
2026-01-13 23:21 ` Darrick J. Wong
0 siblings, 1 reply; 36+ messages in thread
From: Christoph Hellwig @ 2026-01-13 15:57 UTC (permalink / raw)
To: Darrick J. Wong; +Cc: cem, hch, linux-fsdevel, linux-xfs
On Mon, Jan 12, 2026 at 04:35:25PM -0800, Darrick J. Wong wrote:
> From: Darrick J. Wong <djwong@kernel.org>
>
> Add a new privileged ioctl so that xfs_scrub can ask the kernel to
> verify the media of the devices backing an xfs filesystem, and have any
> resulting media errors reported to fsnotify and xfs_healer.
Hmm, the description is a bit sparse?
> +/* Verify the media of the underlying devices */
> +struct xfs_verify_media {
> + __u32 dev; /* I: XFS_VERIFY_*DEV */
This should probably use the enum xfs_device values?
> +#define XFS_VERIFY_TO_EOD (~0ULL) /* end of disk */
Is there much of a point in this flag? scrub/healer really should
know the device size, shouldn't they?
> diff --git a/fs/xfs/xfs_notify_failure.c b/fs/xfs/xfs_notify_failure.c
> index 1edc4ddd10cdb2..5ef4109cc062d2 100644
> --- a/fs/xfs/xfs_notify_failure.c
> +++ b/fs/xfs/xfs_notify_failure.c
There's basically no overlap with the existing code in this file,
why not add a new one?
> + const unsigned int iosize = BIO_MAX_VECS << PAGE_SHIFT;
> + unsigned int bufsize = iosize;
That's a pretty gigantic buffer size. In general a low number of
MB should max out most current devices, and for a background scrub
you generally do not want to actually max out the device..
The in the background is also a good point here - we probably want
a way to tune the size as it might put too much of a load onto the
system pretty easily, and we need a way to dial it back.
> + folio = folio_alloc(GFP_KERNEL, get_order(bufsize));
> + if (!folio)
That first folio_alloc will cause nasty stack traces when it fails.
> + folio = folio_alloc(GFP_KERNEL, 0);
.. and then we fall back to just a single page. This is what I ended
up writing for an about to submitted series elsewhere:
static struct folio *folio_alloc_greedy(gfp_t gfp, size_t *size)
{
struct folio *folio;
while (*size > PAGE_SIZE) {
folio = folio_alloc(gfp | __GFP_NORETRY, get_order(*size));
if (folio)
return folio;
*size = rounddown_pow_of_two(*size - 1);
}
return folio_alloc(gfp, get_order(*size));
}
although that is a bit more complicated as we never want to round
up the actual size.
> + for (i = 0; i < nr_vecs; i++) {
> + unsigned int vec_sects =
> + min(nr_sects, bufsize >> SECTOR_SHIFT);
> +
> + bio_add_folio_nofail(bio, folio,
> + vec_sects << SECTOR_SHIFT, 0);
> +
> + bio_daddr += vec_sects;
> + bio_bbcount -= vec_sects;
> + bio_submitted += vec_sects;
> + }
A single folio is always just a single vetor in the bio. No need
for any of the looping here.
> + /* Don't let too many IOs accumulate */
> + if (bio_submitted > SZ_256M >> SECTOR_SHIFT) {
> + blk_finish_plug(&plug);
> + error = submit_bio_wait(bio);
Also the building up and chaining here seems harmful. If you're
on SSDs you want to fire things off ASAP if you have large I/O.
On a HDD we'll take care of it below, but the bios will usually
actually be split, not merged anyway as they are beyond the
supported I/O size of the HBAs.
^ permalink raw reply [flat|nested] 36+ messages in thread* Re: [PATCH 11/11] xfs: add media verification ioctl
2026-01-13 15:57 ` Christoph Hellwig
@ 2026-01-13 23:21 ` Darrick J. Wong
2026-01-14 5:40 ` Darrick J. Wong
2026-01-14 6:02 ` Christoph Hellwig
0 siblings, 2 replies; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-13 23:21 UTC (permalink / raw)
To: Christoph Hellwig; +Cc: cem, linux-fsdevel, linux-xfs
On Tue, Jan 13, 2026 at 04:57:01PM +0100, Christoph Hellwig wrote:
> On Mon, Jan 12, 2026 at 04:35:25PM -0800, Darrick J. Wong wrote:
> > From: Darrick J. Wong <djwong@kernel.org>
> >
> > Add a new privileged ioctl so that xfs_scrub can ask the kernel to
> > verify the media of the devices backing an xfs filesystem, and have any
> > resulting media errors reported to fsnotify and xfs_healer.
>
> Hmm, the description is a bit sparse?
>
> > +/* Verify the media of the underlying devices */
> > +struct xfs_verify_media {
> > + __u32 dev; /* I: XFS_VERIFY_*DEV */
>
> This should probably use the enum xfs_device values?
Yes, that's a good point.
> > +#define XFS_VERIFY_TO_EOD (~0ULL) /* end of disk */
>
> Is there much of a point in this flag? scrub/healer really should
> know the device size, shouldn't they?
Yes, scrub and healer both know the size they want to verify. I put
that in for the sake of xfs_io so that it wouldn't have to figure out
the device size, but as the ioctl always decreases @end_daddr to the
actual EOD, I think it'd be ok if xfs_io blindly wrote in ~0ULL.
> > diff --git a/fs/xfs/xfs_notify_failure.c b/fs/xfs/xfs_notify_failure.c
> > index 1edc4ddd10cdb2..5ef4109cc062d2 100644
> > --- a/fs/xfs/xfs_notify_failure.c
> > +++ b/fs/xfs/xfs_notify_failure.c
>
> There's basically no overlap with the existing code in this file,
> why not add a new one?
Good idea! I didn't even realize that it shares nothing now.
> > + const unsigned int iosize = BIO_MAX_VECS << PAGE_SHIFT;
> > + unsigned int bufsize = iosize;
>
> That's a pretty gigantic buffer size. In general a low number of
> MB should max out most current devices, and for a background scrub
> you generally do not want to actually max out the device..
256 * 4k (= 1MB) is too large a buffer?
I guess that /is/ 16M on a 64k-page system.
Maybe I need to reset my expectations. :)
> The in the background is also a good point here - we probably want
> a way to tune the size as it might put too much of a load onto the
> system pretty easily, and we need a way to dial it back.
I guess I could add a u32 max_io_size to constrain iosize; and a u16
rest_us that would enforce a sleep in between each submit_bio.
For xfs_scrub@.service I rely on systemd to set the io and cpu priority
and the kernel not to schedule xfs_scrub for resource usage control.
> > + folio = folio_alloc(GFP_KERNEL, get_order(bufsize));
> > + if (!folio)
>
> That first folio_alloc will cause nasty stack traces when it fails.
>
> > + folio = folio_alloc(GFP_KERNEL, 0);
>
> .. and then we fall back to just a single page. This is what I ended
> up writing for an about to submitted series elsewhere:
>
> static struct folio *folio_alloc_greedy(gfp_t gfp, size_t *size)
> {
> struct folio *folio;
>
> while (*size > PAGE_SIZE) {
> folio = folio_alloc(gfp | __GFP_NORETRY, get_order(*size));
> if (folio)
> return folio;
> *size = rounddown_pow_of_two(*size - 1);
> }
>
> return folio_alloc(gfp, get_order(*size));
> }
<nod>
> although that is a bit more complicated as we never want to round
> up the actual size.
I made a simpler version that does
for (order = get_order(); order > 0; order--)
folio = folio_alloc(...);
> > + for (i = 0; i < nr_vecs; i++) {
> > + unsigned int vec_sects =
> > + min(nr_sects, bufsize >> SECTOR_SHIFT);
> > +
> > + bio_add_folio_nofail(bio, folio,
> > + vec_sects << SECTOR_SHIFT, 0);
> > +
> > + bio_daddr += vec_sects;
> > + bio_bbcount -= vec_sects;
> > + bio_submitted += vec_sects;
> > + }
>
> A single folio is always just a single vetor in the bio. No need
> for any of the looping here.
If we have to fall back to a single base page, shouldn't we still try to
create a larger bio? A subtle assumption here is that it's ok to have
all the bvecs pointing to the same memory, and that the device won't
screw up if someone asks it to DMA to the same page simultaneously.
Obviously that all goes away with REQ_OP_VERIFY.
> > + /* Don't let too many IOs accumulate */
> > + if (bio_submitted > SZ_256M >> SECTOR_SHIFT) {
> > + blk_finish_plug(&plug);
> > + error = submit_bio_wait(bio);
>
> Also the building up and chaining here seems harmful. If you're
> on SSDs you want to fire things off ASAP if you have large I/O.
> On a HDD we'll take care of it below, but the bios will usually
> actually be split, not merged anyway as they are beyond the
> supported I/O size of the HBAs.
Hrm, maybe I should query the block device for max_sectors_kb then?
Though curiously the nvme devices all report 128K for that, and the
spinning rust reports 1M.
Anyway I've changed the loop body to allocate a bio, add up to iosize
worth of meory to it, submit it, and wait. This avoids building up huge
chains of bios, and now I ask the device what's the biggest IO that it
supports:
iosize = min(queue_max_bytes(bdev_get_queue(btp->bt_bdev)), SZ_1M);
if (me->me_maxio && iosize > me->me_maxio)
iosize = me->me_maxio;
if (iosize < SECTOR_SIZE)
iosize = SECTOR_SIZE;
...
while (bio_bbcount > 0) {
struct bio *bio;
unsigned int nr_sects =
min_t(sector_t, bio_bbcount, iosize >> SECTOR_SHIFT);
const unsigned int nr_vecs =
howmany(nr_sects << SECTOR_SHIFT, bufsize);
unsigned int bio_submitted = 0;
unsigned int i;
bio = bio_alloc(btp->bt_bdev, nr_vecs, REQ_OP_READ, GFP_KERNEL);
if (!bio) {
error = -ENOMEM;
break;
}
bio->bi_iter.bi_sector = bio_daddr;
for (i = 0; i < nr_vecs; i++) {
unsigned int vec_sects =
min(nr_sects, bufsize >> SECTOR_SHIFT);
bio_add_folio_nofail(bio, folio,
vec_sects << SECTOR_SHIFT, 0);
bio_daddr += vec_sects;
bio_bbcount -= vec_sects;
bio_submitted += vec_sects;
}
error = submit_bio_wait(bio);
bio_put(bio);
bio = NULL;
if (error) {
xfs_verify_media_error(mp, me, btp, fdev,
new_start_daddr, bio_submitted, error);
error = 0;
break;
}
cond_resched();
new_start_daddr += bio_submitted;
}
The only downside of this simple loop is that we can't pipeline read
requests, but maybe we don't want to flood the device with IO requests
anyway.
However, in the case where memory is fragmented and we can only get
(say) a single base page, it'll still try to load up the bio with as
many vecs as it can to try to keep the io size large, because issuing
256x 4k IOs is a lot slower than issuing 1x 1M IO with the same page
added 256 times.
I wonder if nr_vecs ought to be capped by queue_max_segments?
--D
^ permalink raw reply [flat|nested] 36+ messages in thread* Re: [PATCH 11/11] xfs: add media verification ioctl
2026-01-13 23:21 ` Darrick J. Wong
@ 2026-01-14 5:40 ` Darrick J. Wong
2026-01-14 6:02 ` Christoph Hellwig
1 sibling, 0 replies; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-14 5:40 UTC (permalink / raw)
To: Christoph Hellwig; +Cc: cem, linux-fsdevel, linux-xfs
On Tue, Jan 13, 2026 at 03:21:13PM -0800, Darrick J. Wong wrote:
> On Tue, Jan 13, 2026 at 04:57:01PM +0100, Christoph Hellwig wrote:
> > On Mon, Jan 12, 2026 at 04:35:25PM -0800, Darrick J. Wong wrote:
> > > From: Darrick J. Wong <djwong@kernel.org>
> > >
> > > Add a new privileged ioctl so that xfs_scrub can ask the kernel to
> > > verify the media of the devices backing an xfs filesystem, and have any
> > > resulting media errors reported to fsnotify and xfs_healer.
> >
> > Hmm, the description is a bit sparse?
> >
> > > +/* Verify the media of the underlying devices */
> > > +struct xfs_verify_media {
> > > + __u32 dev; /* I: XFS_VERIFY_*DEV */
> >
> > This should probably use the enum xfs_device values?
>
> Yes, that's a good point.
FYI, today's draft of this ioctl can be read here:
https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/commit/?h=health-monitoring_2026-01-13&id=e3fee7d7ead8b3e630845304b9030b5c7c5f27da
It contains all the alterations I talked about earlier today.
(I might be coming down with a cold, so I thought it best to git push
now and find out if I'm at all coherent tomorrow.)
--D
^ permalink raw reply [flat|nested] 36+ messages in thread* Re: [PATCH 11/11] xfs: add media verification ioctl
2026-01-13 23:21 ` Darrick J. Wong
2026-01-14 5:40 ` Darrick J. Wong
@ 2026-01-14 6:02 ` Christoph Hellwig
2026-01-14 6:07 ` Darrick J. Wong
1 sibling, 1 reply; 36+ messages in thread
From: Christoph Hellwig @ 2026-01-14 6:02 UTC (permalink / raw)
To: Darrick J. Wong; +Cc: Christoph Hellwig, cem, linux-fsdevel, linux-xfs
On Tue, Jan 13, 2026 at 03:21:13PM -0800, Darrick J. Wong wrote:
> > > +#define XFS_VERIFY_TO_EOD (~0ULL) /* end of disk */
> >
> > Is there much of a point in this flag? scrub/healer really should
> > know the device size, shouldn't they?
>
> Yes, scrub and healer both know the size they want to verify. I put
> that in for the sake of xfs_io so that it wouldn't have to figure out
> the device size, but as the ioctl always decreases @end_daddr to the
> actual EOD, I think it'd be ok if xfs_io blindly wrote in ~0ULL.
That's the best of both worlds.
> > > + const unsigned int iosize = BIO_MAX_VECS << PAGE_SHIFT;
> > > + unsigned int bufsize = iosize;
> >
> > That's a pretty gigantic buffer size. In general a low number of
> > MB should max out most current devices, and for a background scrub
> > you generally do not want to actually max out the device..
>
> 256 * 4k (= 1MB) is too large a buffer?
No, my reading comprehension just sucks :) And of course the way
it's written isn't very helpful either.
> I guess that /is/ 16M on a 64k-page system.
Yeah, just stick to SZ_1M.
> > > + min(nr_sects, bufsize >> SECTOR_SHIFT);
> > > +
> > > + bio_add_folio_nofail(bio, folio,
> > > + vec_sects << SECTOR_SHIFT, 0);
> > > +
> > > + bio_daddr += vec_sects;
> > > + bio_bbcount -= vec_sects;
> > > + bio_submitted += vec_sects;
> > > + }
> >
> > A single folio is always just a single vetor in the bio. No need
> > for any of the looping here.
>
> If we have to fall back to a single base page, shouldn't we still try to
> create a larger bio?
How do you create a larger bio if you only have a single bio available?
> A subtle assumption here is that it's ok to have
> all the bvecs pointing to the same memory, and that the device won't
> screw up if someone asks it to DMA to the same page simultaneously.
Ooooh. Yes, that will screw up badly when using PI.
> > > + /* Don't let too many IOs accumulate */
> > > + if (bio_submitted > SZ_256M >> SECTOR_SHIFT) {
> > > + blk_finish_plug(&plug);
> > > + error = submit_bio_wait(bio);
> >
> > Also the building up and chaining here seems harmful. If you're
> > on SSDs you want to fire things off ASAP if you have large I/O.
> > On a HDD we'll take care of it below, but the bios will usually
> > actually be split, not merged anyway as they are beyond the
> > supported I/O size of the HBAs.
>
> Hrm, maybe I should query the block device for max_sectors_kb then?
No. max_sectors_kb is kida stupid. I think a sensible default and
a tunable is a better choice here at least for now.
> However, in the case where memory is fragmented and we can only get
> (say) a single base page, it'll still try to load up the bio with as
> many vecs as it can to try to keep the io size large, because issuing
> 256x 4k IOs is a lot slower than issuing 1x 1M IO with the same page
> added 256 times.
Yeah. But seriously, if the MM is pretty good and is getting better
at finding large allocations. We need to start relying on that.
> I wonder if nr_vecs ought to be capped by queue_max_segments?
No, leave all that splitting to the block layer. max_segments is
an implementation detail.
^ permalink raw reply [flat|nested] 36+ messages in thread* Re: [PATCH 11/11] xfs: add media verification ioctl
2026-01-14 6:02 ` Christoph Hellwig
@ 2026-01-14 6:07 ` Darrick J. Wong
2026-01-14 6:15 ` Christoph Hellwig
0 siblings, 1 reply; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-14 6:07 UTC (permalink / raw)
To: Christoph Hellwig; +Cc: cem, linux-fsdevel, linux-xfs
On Wed, Jan 14, 2026 at 07:02:14AM +0100, Christoph Hellwig wrote:
> On Tue, Jan 13, 2026 at 03:21:13PM -0800, Darrick J. Wong wrote:
> > > > +#define XFS_VERIFY_TO_EOD (~0ULL) /* end of disk */
> > >
> > > Is there much of a point in this flag? scrub/healer really should
> > > know the device size, shouldn't they?
> >
> > Yes, scrub and healer both know the size they want to verify. I put
> > that in for the sake of xfs_io so that it wouldn't have to figure out
> > the device size, but as the ioctl always decreases @end_daddr to the
> > actual EOD, I think it'd be ok if xfs_io blindly wrote in ~0ULL.
>
> That's the best of both worlds.
>
> > > > + const unsigned int iosize = BIO_MAX_VECS << PAGE_SHIFT;
> > > > + unsigned int bufsize = iosize;
> > >
> > > That's a pretty gigantic buffer size. In general a low number of
> > > MB should max out most current devices, and for a background scrub
> > > you generally do not want to actually max out the device..
> >
> > 256 * 4k (= 1MB) is too large a buffer?
>
> No, my reading comprehension just sucks :) And of course the way
> it's written isn't very helpful either.
>
> > I guess that /is/ 16M on a 64k-page system.
>
> Yeah, just stick to SZ_1M.
Ok, will do.
> > > > + min(nr_sects, bufsize >> SECTOR_SHIFT);
> > > > +
> > > > + bio_add_folio_nofail(bio, folio,
> > > > + vec_sects << SECTOR_SHIFT, 0);
> > > > +
> > > > + bio_daddr += vec_sects;
> > > > + bio_bbcount -= vec_sects;
> > > > + bio_submitted += vec_sects;
> > > > + }
> > >
> > > A single folio is always just a single vetor in the bio. No need
> > > for any of the looping here.
> >
> > If we have to fall back to a single base page, shouldn't we still try to
> > create a larger bio?
>
> How do you create a larger bio if you only have a single bio available?
>
> > A subtle assumption here is that it's ok to have
> > all the bvecs pointing to the same memory, and that the device won't
> > screw up if someone asks it to DMA to the same page simultaneously.
>
> Ooooh. Yes, that will screw up badly when using PI.
DOH! I forgot about that.
Yep, that's correct. I'll change the code to one folio, one bio.
> > > > + /* Don't let too many IOs accumulate */
> > > > + if (bio_submitted > SZ_256M >> SECTOR_SHIFT) {
> > > > + blk_finish_plug(&plug);
> > > > + error = submit_bio_wait(bio);
> > >
> > > Also the building up and chaining here seems harmful. If you're
> > > on SSDs you want to fire things off ASAP if you have large I/O.
> > > On a HDD we'll take care of it below, but the bios will usually
> > > actually be split, not merged anyway as they are beyond the
> > > supported I/O size of the HBAs.
> >
> > Hrm, maybe I should query the block device for max_sectors_kb then?
>
> No. max_sectors_kb is kida stupid. I think a sensible default and
> a tunable is a better choice here at least for now.
<nod> I'll set iosize to 1MB by default and userspace can decrease it if
it so desires.
Also it occurs to me that max_hw_sectors_kb seems to be 128K for all of
my consumer nvme devices, and the fancy Intel ones too. Funny that the
sata ssds set it to 4MB, but I guess capping at 128k is one way to
reduce the command latency...? (Or the kernel's broken? I can't figure
out how to get mpsmin with nvme-cli...)
> > However, in the case where memory is fragmented and we can only get
> > (say) a single base page, it'll still try to load up the bio with as
> > many vecs as it can to try to keep the io size large, because issuing
> > 256x 4k IOs is a lot slower than issuing 1x 1M IO with the same page
> > added 256 times.
>
> Yeah. But seriously, if the MM is pretty good and is getting better
> at finding large allocations. We need to start relying on that.
>
> > I wonder if nr_vecs ought to be capped by queue_max_segments?
>
> No, leave all that splitting to the block layer. max_segments is
> an implementation detail.
Will do. Only using the folio once per bio will make the control loop
easier to understand too.
--D
^ permalink raw reply [flat|nested] 36+ messages in thread* Re: [PATCH 11/11] xfs: add media verification ioctl
2026-01-14 6:07 ` Darrick J. Wong
@ 2026-01-14 6:15 ` Christoph Hellwig
2026-01-14 6:19 ` Darrick J. Wong
0 siblings, 1 reply; 36+ messages in thread
From: Christoph Hellwig @ 2026-01-14 6:15 UTC (permalink / raw)
To: Darrick J. Wong; +Cc: Christoph Hellwig, cem, linux-fsdevel, linux-xfs
On Tue, Jan 13, 2026 at 10:07:05PM -0800, Darrick J. Wong wrote:
> > a tunable is a better choice here at least for now.
>
> <nod> I'll set iosize to 1MB by default and userspace can decrease it if
> it so desires.
>
> Also it occurs to me that max_hw_sectors_kb seems to be 128K for all of
> my consumer nvme devices, and the fancy Intel ones too. Funny that the
> sata ssds set it to 4MB, but I guess capping at 128k is one way to
> reduce the command latency...? (Or the kernel's broken? I can't figure
> out how to get mpsmin with nvme-cli...)
mpsmin is basically always 4k.
On something unrelated: SSDs remap all the time by definition, and
HDDs are really good at remapping bad sectors these days as well.
So verifying blocks that do not actually contain file system (meta)data
is pretty pointless. Can we come up with a way to verify only blocks
that have valid data in them, while still being resonably sequential?
I.e. walk the rmap?
^ permalink raw reply [flat|nested] 36+ messages in thread
* Re: [PATCH 11/11] xfs: add media verification ioctl
2026-01-14 6:15 ` Christoph Hellwig
@ 2026-01-14 6:19 ` Darrick J. Wong
0 siblings, 0 replies; 36+ messages in thread
From: Darrick J. Wong @ 2026-01-14 6:19 UTC (permalink / raw)
To: Christoph Hellwig; +Cc: cem, linux-fsdevel, linux-xfs
On Wed, Jan 14, 2026 at 07:15:59AM +0100, Christoph Hellwig wrote:
> On Tue, Jan 13, 2026 at 10:07:05PM -0800, Darrick J. Wong wrote:
> > > a tunable is a better choice here at least for now.
> >
> > <nod> I'll set iosize to 1MB by default and userspace can decrease it if
> > it so desires.
> >
> > Also it occurs to me that max_hw_sectors_kb seems to be 128K for all of
> > my consumer nvme devices, and the fancy Intel ones too. Funny that the
> > sata ssds set it to 4MB, but I guess capping at 128k is one way to
> > reduce the command latency...? (Or the kernel's broken? I can't figure
> > out how to get mpsmin with nvme-cli...)
>
> mpsmin is basically always 4k.
>
>
> On something unrelated: SSDs remap all the time by definition, and
> HDDs are really good at remapping bad sectors these days as well.
> So verifying blocks that do not actually contain file system (meta)data
> is pretty pointless. Can we come up with a way to verify only blocks
> that have valid data in them, while still being resonably sequential?
> I.e. walk the rmap?
xfs_scrub phase6 already calls getfsmap to figure out which parts of the
disk actually contain written data of any kind and are worth verifying.
--D
^ permalink raw reply [flat|nested] 36+ messages in thread