diff mbox series

[RFC,v11.1,2/2] mm, pmem, xfs: Introduce MF_MEM_REMOVE for unbind

Message ID 1681296735-2-1-git-send-email-ruansy.fnst@fujitsu.com (mailing list archive)
State New
Headers show
Series None | expand

Commit Message

Shiyang Ruan April 12, 2023, 10:52 a.m. UTC
This is a RFC HOTFIX.

This hotfix adds a exclusive forzen state to make sure any others won't
thaw the fs during xfs_dax_notify_failure():

  #define SB_FREEZE_EXCLUSIVE	(SB_FREEZE_COMPLETE + 2)
Using +2 here is because Darrick's patch[0] is using +1.  So, should we
make these definitions global?

Another thing I can't make up my mind is: when another freezer has freeze
the fs, should we wait unitl it finish, or print a warning in dmesg and
return -EBUSY?

Since there are at least 2 places needs exclusive forzen state, I think
we can refactor helper functions of freeze/thaw for them.  e.g.
  int freeze_super_exclusive(struct super_block *sb, int frozen);
  int thaw_super_exclusive(struct super_block *sb, int frozen);

[0] https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/commit/?h=repair-fscounters&id=c3a0d1de4d54ffb565dbc7092dfe1fb851940669


--- Original commit message ---
This patch is inspired by Dan's "mm, dax, pmem: Introduce
dev_pagemap_failure()"[1].  With the help of dax_holder and
->notify_failure() mechanism, the pmem driver is able to ask filesystem
(or mapped device) on it to unmap all files in use and notify processes
who are using those files.

Call trace:
trigger unbind
 -> unbind_store()
  -> ... (skip)
   -> devres_release_all()
    -> kill_dax()
     -> dax_holder_notify_failure(dax_dev, 0, U64_MAX, MF_MEM_PRE_REMOVE)
      -> xfs_dax_notify_failure()
      `-> freeze_super()
      `-> do xfs rmap
      ` -> mf_dax_kill_procs()
      `  -> collect_procs_fsdax()    // all associated
      `  -> unmap_and_kill()
      ` -> invalidate_inode_pages2() // drop file's cache
      `-> thaw_super()

Introduce MF_MEM_PRE_REMOVE to let filesystem know this is a remove
event.  Also introduce a exclusive freeze/thaw to lock the filesystem to
prevent new dax mapping from being created.  And do not shutdown
filesystem directly if something not supported, or if failure range
includes metadata area.  Make sure all files and processes are handled
correctly.  Also drop the cache of associated files before pmem is
removed.

[1]: https://lore.kernel.org/linux-mm/161604050314.1463742.14151665140035795571.stgit@dwillia2-desk3.amr.corp.intel.com/

Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
---
 drivers/dax/super.c         |   3 +-
 fs/xfs/xfs_notify_failure.c | 151 ++++++++++++++++++++++++++++++++++--
 include/linux/mm.h          |   1 +
 mm/memory-failure.c         |  17 +++-
 4 files changed, 162 insertions(+), 10 deletions(-)

Comments

Shiyang Ruan April 20, 2023, 2:07 a.m. UTC | #1
Ping~

在 2023/4/12 18:52, Shiyang Ruan 写道:
> This is a RFC HOTFIX.
> 
> This hotfix adds a exclusive forzen state to make sure any others won't
> thaw the fs during xfs_dax_notify_failure():
> 
>    #define SB_FREEZE_EXCLUSIVE	(SB_FREEZE_COMPLETE + 2)
> Using +2 here is because Darrick's patch[0] is using +1.  So, should we
> make these definitions global?
> 
> Another thing I can't make up my mind is: when another freezer has freeze
> the fs, should we wait unitl it finish, or print a warning in dmesg and
> return -EBUSY?
> 
> Since there are at least 2 places needs exclusive forzen state, I think
> we can refactor helper functions of freeze/thaw for them.  e.g.
>    int freeze_super_exclusive(struct super_block *sb, int frozen);
>    int thaw_super_exclusive(struct super_block *sb, int frozen);
> 
> [0] https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/commit/?h=repair-fscounters&id=c3a0d1de4d54ffb565dbc7092dfe1fb851940669
> 
> 
> --- Original commit message ---
> This patch is inspired by Dan's "mm, dax, pmem: Introduce
> dev_pagemap_failure()"[1].  With the help of dax_holder and
> ->notify_failure() mechanism, the pmem driver is able to ask filesystem
> (or mapped device) on it to unmap all files in use and notify processes
> who are using those files.
> 
> Call trace:
> trigger unbind
>   -> unbind_store()
>    -> ... (skip)
>     -> devres_release_all()
>      -> kill_dax()
>       -> dax_holder_notify_failure(dax_dev, 0, U64_MAX, MF_MEM_PRE_REMOVE)
>        -> xfs_dax_notify_failure()
>        `-> freeze_super()
>        `-> do xfs rmap
>        ` -> mf_dax_kill_procs()
>        `  -> collect_procs_fsdax()    // all associated
>        `  -> unmap_and_kill()
>        ` -> invalidate_inode_pages2() // drop file's cache
>        `-> thaw_super()
> 
> Introduce MF_MEM_PRE_REMOVE to let filesystem know this is a remove
> event.  Also introduce a exclusive freeze/thaw to lock the filesystem to
> prevent new dax mapping from being created.  And do not shutdown
> filesystem directly if something not supported, or if failure range
> includes metadata area.  Make sure all files and processes are handled
> correctly.  Also drop the cache of associated files before pmem is
> removed.
> 
> [1]: https://lore.kernel.org/linux-mm/161604050314.1463742.14151665140035795571.stgit@dwillia2-desk3.amr.corp.intel.com/
> 
> Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
> ---
>   drivers/dax/super.c         |   3 +-
>   fs/xfs/xfs_notify_failure.c | 151 ++++++++++++++++++++++++++++++++++--
>   include/linux/mm.h          |   1 +
>   mm/memory-failure.c         |  17 +++-
>   4 files changed, 162 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/dax/super.c b/drivers/dax/super.c
> index c4c4728a36e4..2e1a35e82fce 100644
> --- a/drivers/dax/super.c
> +++ b/drivers/dax/super.c
> @@ -323,7 +323,8 @@ void kill_dax(struct dax_device *dax_dev)
>   		return;
>   
>   	if (dax_dev->holder_data != NULL)
> -		dax_holder_notify_failure(dax_dev, 0, U64_MAX, 0);
> +		dax_holder_notify_failure(dax_dev, 0, U64_MAX,
> +				MF_MEM_PRE_REMOVE);
>   
>   	clear_bit(DAXDEV_ALIVE, &dax_dev->flags);
>   	synchronize_srcu(&dax_srcu);
> diff --git a/fs/xfs/xfs_notify_failure.c b/fs/xfs/xfs_notify_failure.c
> index 1e2eddb8f90f..796dd954d33a 100644
> --- a/fs/xfs/xfs_notify_failure.c
> +++ b/fs/xfs/xfs_notify_failure.c
> @@ -22,6 +22,7 @@
>   
>   #include <linux/mm.h>
>   #include <linux/dax.h>
> +#include <linux/fs.h>
>   
>   struct xfs_failure_info {
>   	xfs_agblock_t		startblock;
> @@ -73,10 +74,16 @@ xfs_dax_failure_fn(
>   	struct xfs_mount		*mp = cur->bc_mp;
>   	struct xfs_inode		*ip;
>   	struct xfs_failure_info		*notify = data;
> +	struct address_space		*mapping;
> +	pgoff_t				pgoff;
> +	unsigned long			pgcnt;
>   	int				error = 0;
>   
>   	if (XFS_RMAP_NON_INODE_OWNER(rec->rm_owner) ||
>   	    (rec->rm_flags & (XFS_RMAP_ATTR_FORK | XFS_RMAP_BMBT_BLOCK))) {
> +		/* The device is about to be removed.  Not a really failure. */
> +		if (notify->mf_flags & MF_MEM_PRE_REMOVE)
> +			return 0;
>   		notify->want_shutdown = true;
>   		return 0;
>   	}
> @@ -92,14 +99,120 @@ xfs_dax_failure_fn(
>   		return 0;
>   	}
>   
> -	error = mf_dax_kill_procs(VFS_I(ip)->i_mapping,
> -				  xfs_failure_pgoff(mp, rec, notify),
> -				  xfs_failure_pgcnt(mp, rec, notify),
> -				  notify->mf_flags);
> +	mapping = VFS_I(ip)->i_mapping;
> +	pgoff = xfs_failure_pgoff(mp, rec, notify);
> +	pgcnt = xfs_failure_pgcnt(mp, rec, notify);
> +
> +	/* Continue the rmap query if the inode isn't a dax file. */
> +	if (dax_mapping(mapping))
> +		error = mf_dax_kill_procs(mapping, pgoff, pgcnt,
> +				notify->mf_flags);
> +
> +	/* Invalidate the cache anyway. */
> +	invalidate_inode_pages2_range(mapping, pgoff, pgoff + pgcnt - 1);
> +
>   	xfs_irele(ip);
>   	return error;
>   }
>   
> +#define SB_FREEZE_EXCLUSIVE	(SB_FREEZE_COMPLETE + 2)
> +
> +static int
> +xfs_dax_notify_failure_freeze(
> +	struct xfs_mount	*mp)
> +{
> +	struct super_block	*sb = mp->m_super;
> +	int			error = 0;
> +	int			level;
> +
> +	/* Wait until we're ready to freeze. */
> +	down_write(&sb->s_umount);
> +	while (sb->s_writers.frozen != SB_UNFROZEN) {
> +		up_write(&sb->s_umount);
> +
> +		// just wait, or print warning in dmesg then return -EBUSY?
> +
> +		delay(HZ / 10);
> +		down_write(&sb->s_umount);
> +	}
> +
> +	if (sb_rdonly(sb)) {
> +		sb->s_writers.frozen = SB_FREEZE_EXCLUSIVE;
> +		goto out;
> +	}
> +
> +	sb->s_writers.frozen = SB_FREEZE_WRITE;
> +	/* Release s_umount to preserve sb_start_write -> s_umount ordering */
> +	up_write(&sb->s_umount);
> +	percpu_down_write(sb->s_writers.rw_sem + SB_FREEZE_WRITE - 1);
> +	down_write(&sb->s_umount);
> +
> +	/* Now we go and block page faults... */
> +	sb->s_writers.frozen = SB_FREEZE_PAGEFAULT;
> +	percpu_down_write(sb->s_writers.rw_sem + SB_FREEZE_PAGEFAULT - 1);
> +
> +	/* All writers are done so after syncing there won't be dirty data */
> +	error = sync_filesystem(sb);
> +	if (error) {
> +		sb->s_writers.frozen = SB_UNFROZEN;
> +		for (level = SB_FREEZE_PAGEFAULT - 1; level >= 0; level--)
> +			percpu_up_write(sb->s_writers.rw_sem + level);
> +		wake_up(&sb->s_writers.wait_unfrozen);
> +		goto out;
> +	}
> +
> +	/* Now wait for internal filesystem counter */
> +	sb->s_writers.frozen = SB_FREEZE_FS;
> +	percpu_down_write(sb->s_writers.rw_sem + SB_FREEZE_FS - 1);
> +
> +	/*
> +	 * To prevent anyone else from unfreezing us, set the VFS freeze level
> +	 * to one higher than SB_FREEZE_COMPLETE.
> +	 */
> +	sb->s_writers.frozen = SB_FREEZE_EXCLUSIVE;
> +	for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--)
> +		percpu_rwsem_release(sb->s_writers.rw_sem + level, 0,
> +				_THIS_IP_);
> +
> +out:
> +	up_write(&sb->s_umount);
> +	return error;
> +}
> +
> +static void
> +xfs_dax_notify_failure_thaw(
> +	struct xfs_mount	*mp)
> +{
> +	struct super_block	*sb = mp->m_super;
> +	int			level;
> +
> +	down_write(&sb->s_umount);
> +	if (sb->s_writers.frozen != SB_FREEZE_EXCLUSIVE) {
> +		/* somebody snuck in and unfroze us? */
> +		ASSERT(0);
> +		up_write(&sb->s_umount);
> +		return;
> +	}
> +
> +	if (sb_rdonly(sb)) {
> +		sb->s_writers.frozen = SB_UNFROZEN;
> +		goto out;
> +	}
> +
> +	for (level = 0; level < SB_FREEZE_LEVELS; ++level)
> +		percpu_rwsem_acquire(sb->s_writers.rw_sem + level, 0,
> +				_THIS_IP_);
> +
> +	sb->s_writers.frozen = SB_UNFROZEN;
> +	for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--)
> +		percpu_up_write(sb->s_writers.rw_sem + level);
> +
> +out:
> +	wake_up(&sb->s_writers.wait_unfrozen);
> +	up_write(&sb->s_umount);
> +}
> +
>   static int
>   xfs_dax_notify_ddev_failure(
>   	struct xfs_mount	*mp,
> @@ -164,11 +277,22 @@ xfs_dax_notify_ddev_failure(
>   	}
>   
>   	xfs_trans_cancel(tp);
> +
> +	/* Thaw the fs if it is freezed before. */
> +	if (mf_flags & MF_MEM_PRE_REMOVE)
> +		xfs_dax_notify_failure_thaw(mp);
> +
> +	/*
> +	 * Determine how to shutdown the filesystem according to the
> +	 * error code and flags.
> +	 */
>   	if (error || notify.want_shutdown) {
>   		xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_ONDISK);
>   		if (!error)
>   			error = -EFSCORRUPTED;
> -	}
> +	} else if (mf_flags & MF_MEM_PRE_REMOVE)
> +		xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
> +
>   	return error;
>   }
>   
> @@ -182,6 +306,7 @@ xfs_dax_notify_failure(
>   	struct xfs_mount	*mp = dax_holder(dax_dev);
>   	u64			ddev_start;
>   	u64			ddev_end;
> +	int			error;
>   
>   	if (!(mp->m_super->s_flags & SB_BORN)) {
>   		xfs_warn(mp, "filesystem is not ready for notify_failure()!");
> @@ -196,6 +321,8 @@ xfs_dax_notify_failure(
>   
>   	if (mp->m_logdev_targp && mp->m_logdev_targp->bt_daxdev == dax_dev &&
>   	    mp->m_logdev_targp != mp->m_ddev_targp) {
> +		if (mf_flags & MF_MEM_PRE_REMOVE)
> +			return 0;
>   		xfs_err(mp, "ondisk log corrupt, shutting down fs!");
>   		xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_ONDISK);
>   		return -EFSCORRUPTED;
> @@ -209,6 +336,12 @@ xfs_dax_notify_failure(
>   	ddev_start = mp->m_ddev_targp->bt_dax_part_off;
>   	ddev_end = ddev_start + bdev_nr_bytes(mp->m_ddev_targp->bt_bdev) - 1;
>   
> +	/* Notify failure on the whole device. */
> +	if (offset == 0 && len == U64_MAX) {
> +		offset = ddev_start;
> +		len = bdev_nr_bytes(mp->m_ddev_targp->bt_bdev);
> +	}
> +
>   	/* Ignore the range out of filesystem area */
>   	if (offset + len - 1 < ddev_start)
>   		return -ENXIO;
> @@ -225,6 +358,14 @@ xfs_dax_notify_failure(
>   	if (offset + len - 1 > ddev_end)
>   		len = ddev_end - offset + 1;
>   
> +	if (mf_flags & MF_MEM_PRE_REMOVE) {
> +		xfs_info(mp, "device is about to be removed!");
> +		/* Freeze fs to prevent new mappings from being created. */
> +		error = xfs_dax_notify_failure_freeze(mp);
> +		if (error)
> +			return error;
> +	}
> +
>   	return xfs_dax_notify_ddev_failure(mp, BTOBB(offset), BTOBB(len),
>   			mf_flags);
>   }
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 1f79667824eb..ac3f22c20e1d 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -3436,6 +3436,7 @@ enum mf_flags {
>   	MF_UNPOISON = 1 << 4,
>   	MF_SW_SIMULATED = 1 << 5,
>   	MF_NO_RETRY = 1 << 6,
> +	MF_MEM_PRE_REMOVE = 1 << 7,
>   };
>   int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
>   		      unsigned long count, int mf_flags);
> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> index fae9baf3be16..6e6acec45568 100644
> --- a/mm/memory-failure.c
> +++ b/mm/memory-failure.c
> @@ -623,7 +623,7 @@ static void collect_procs_file(struct page *page, struct list_head *to_kill,
>    */
>   static void collect_procs_fsdax(struct page *page,
>   		struct address_space *mapping, pgoff_t pgoff,
> -		struct list_head *to_kill)
> +		struct list_head *to_kill, bool pre_remove)
>   {
>   	struct vm_area_struct *vma;
>   	struct task_struct *tsk;
> @@ -631,8 +631,15 @@ static void collect_procs_fsdax(struct page *page,
>   	i_mmap_lock_read(mapping);
>   	read_lock(&tasklist_lock);
>   	for_each_process(tsk) {
> -		struct task_struct *t = task_early_kill(tsk, true);
> +		struct task_struct *t = tsk;
>   
> +		/*
> +		 * Search for all tasks while MF_MEM_PRE_REMOVE, because the
> +		 * current may not be the one accessing the fsdax page.
> +		 * Otherwise, search for the current task.
> +		 */
> +		if (!pre_remove)
> +			t = task_early_kill(tsk, true);
>   		if (!t)
>   			continue;
>   		vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
> @@ -1732,6 +1739,7 @@ int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
>   	dax_entry_t cookie;
>   	struct page *page;
>   	size_t end = index + count;
> +	bool pre_remove = mf_flags & MF_MEM_PRE_REMOVE;
>   
>   	mf_flags |= MF_ACTION_REQUIRED | MF_MUST_KILL;
>   
> @@ -1743,9 +1751,10 @@ int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
>   		if (!page)
>   			goto unlock;
>   
> -		SetPageHWPoison(page);
> +		if (!pre_remove)
> +			SetPageHWPoison(page);
>   
> -		collect_procs_fsdax(page, mapping, index, &to_kill);
> +		collect_procs_fsdax(page, mapping, index, &to_kill, pre_remove);
>   		unmap_and_kill(&to_kill, page_to_pfn(page), mapping,
>   				index, mf_flags);
>   unlock:
Jan Kara April 20, 2023, 12:09 p.m. UTC | #2
On Thu 20-04-23 10:07:39, Shiyang Ruan wrote:
> 在 2023/4/12 18:52, Shiyang Ruan 写道:
> > This is a RFC HOTFIX.
> > 
> > This hotfix adds a exclusive forzen state to make sure any others won't
> > thaw the fs during xfs_dax_notify_failure():
> > 
> >    #define SB_FREEZE_EXCLUSIVE	(SB_FREEZE_COMPLETE + 2)
> > Using +2 here is because Darrick's patch[0] is using +1.  So, should we
> > make these definitions global?
> > 
> > Another thing I can't make up my mind is: when another freezer has freeze
> > the fs, should we wait unitl it finish, or print a warning in dmesg and
> > return -EBUSY?
> > 
> > Since there are at least 2 places needs exclusive forzen state, I think
> > we can refactor helper functions of freeze/thaw for them.  e.g.
> >    int freeze_super_exclusive(struct super_block *sb, int frozen);
> >    int thaw_super_exclusive(struct super_block *sb, int frozen);
> > 
> > [0] https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/commit/?h=repair-fscounters&id=c3a0d1de4d54ffb565dbc7092dfe1fb851940669

I'm OK with the idea of new freeze state that does not allow userspace to
thaw the filesystem. But I don't really like the guts of filesystem
freezing being replicated inside XFS. It is bad enough that they are
replicated in [0], replicating them *once more* in another XFS file shows
we are definitely doing something wrong. And Luis will need yet another
incantation of the exlusive freeze for suspend-to-disk. So please guys get
together and reorganize the generic freezing code so that it supports
exclusive freeze (for in-kernel users) and works for your usecases instead
of replicating it inside XFS...

								Honza

> > --- Original commit message ---
> > This patch is inspired by Dan's "mm, dax, pmem: Introduce
> > dev_pagemap_failure()"[1].  With the help of dax_holder and
> > ->notify_failure() mechanism, the pmem driver is able to ask filesystem
> > (or mapped device) on it to unmap all files in use and notify processes
> > who are using those files.
> > 
> > Call trace:
> > trigger unbind
> >   -> unbind_store()
> >    -> ... (skip)
> >     -> devres_release_all()
> >      -> kill_dax()
> >       -> dax_holder_notify_failure(dax_dev, 0, U64_MAX, MF_MEM_PRE_REMOVE)
> >        -> xfs_dax_notify_failure()
> >        `-> freeze_super()
> >        `-> do xfs rmap
> >        ` -> mf_dax_kill_procs()
> >        `  -> collect_procs_fsdax()    // all associated
> >        `  -> unmap_and_kill()
> >        ` -> invalidate_inode_pages2() // drop file's cache
> >        `-> thaw_super()
> > 
> > Introduce MF_MEM_PRE_REMOVE to let filesystem know this is a remove
> > event.  Also introduce a exclusive freeze/thaw to lock the filesystem to
> > prevent new dax mapping from being created.  And do not shutdown
> > filesystem directly if something not supported, or if failure range
> > includes metadata area.  Make sure all files and processes are handled
> > correctly.  Also drop the cache of associated files before pmem is
> > removed.
> > 
> > [1]: https://lore.kernel.org/linux-mm/161604050314.1463742.14151665140035795571.stgit@dwillia2-desk3.amr.corp.intel.com/
> > 
> > Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
> > ---
> >   drivers/dax/super.c         |   3 +-
> >   fs/xfs/xfs_notify_failure.c | 151 ++++++++++++++++++++++++++++++++++--
> >   include/linux/mm.h          |   1 +
> >   mm/memory-failure.c         |  17 +++-
> >   4 files changed, 162 insertions(+), 10 deletions(-)
> > 
> > diff --git a/drivers/dax/super.c b/drivers/dax/super.c
> > index c4c4728a36e4..2e1a35e82fce 100644
> > --- a/drivers/dax/super.c
> > +++ b/drivers/dax/super.c
> > @@ -323,7 +323,8 @@ void kill_dax(struct dax_device *dax_dev)
> >   		return;
> >   	if (dax_dev->holder_data != NULL)
> > -		dax_holder_notify_failure(dax_dev, 0, U64_MAX, 0);
> > +		dax_holder_notify_failure(dax_dev, 0, U64_MAX,
> > +				MF_MEM_PRE_REMOVE);
> >   	clear_bit(DAXDEV_ALIVE, &dax_dev->flags);
> >   	synchronize_srcu(&dax_srcu);
> > diff --git a/fs/xfs/xfs_notify_failure.c b/fs/xfs/xfs_notify_failure.c
> > index 1e2eddb8f90f..796dd954d33a 100644
> > --- a/fs/xfs/xfs_notify_failure.c
> > +++ b/fs/xfs/xfs_notify_failure.c
> > @@ -22,6 +22,7 @@
> >   #include <linux/mm.h>
> >   #include <linux/dax.h>
> > +#include <linux/fs.h>
> >   struct xfs_failure_info {
> >   	xfs_agblock_t		startblock;
> > @@ -73,10 +74,16 @@ xfs_dax_failure_fn(
> >   	struct xfs_mount		*mp = cur->bc_mp;
> >   	struct xfs_inode		*ip;
> >   	struct xfs_failure_info		*notify = data;
> > +	struct address_space		*mapping;
> > +	pgoff_t				pgoff;
> > +	unsigned long			pgcnt;
> >   	int				error = 0;
> >   	if (XFS_RMAP_NON_INODE_OWNER(rec->rm_owner) ||
> >   	    (rec->rm_flags & (XFS_RMAP_ATTR_FORK | XFS_RMAP_BMBT_BLOCK))) {
> > +		/* The device is about to be removed.  Not a really failure. */
> > +		if (notify->mf_flags & MF_MEM_PRE_REMOVE)
> > +			return 0;
> >   		notify->want_shutdown = true;
> >   		return 0;
> >   	}
> > @@ -92,14 +99,120 @@ xfs_dax_failure_fn(
> >   		return 0;
> >   	}
> > -	error = mf_dax_kill_procs(VFS_I(ip)->i_mapping,
> > -				  xfs_failure_pgoff(mp, rec, notify),
> > -				  xfs_failure_pgcnt(mp, rec, notify),
> > -				  notify->mf_flags);
> > +	mapping = VFS_I(ip)->i_mapping;
> > +	pgoff = xfs_failure_pgoff(mp, rec, notify);
> > +	pgcnt = xfs_failure_pgcnt(mp, rec, notify);
> > +
> > +	/* Continue the rmap query if the inode isn't a dax file. */
> > +	if (dax_mapping(mapping))
> > +		error = mf_dax_kill_procs(mapping, pgoff, pgcnt,
> > +				notify->mf_flags);
> > +
> > +	/* Invalidate the cache anyway. */
> > +	invalidate_inode_pages2_range(mapping, pgoff, pgoff + pgcnt - 1);
> > +
> >   	xfs_irele(ip);
> >   	return error;
> >   }
> > +#define SB_FREEZE_EXCLUSIVE	(SB_FREEZE_COMPLETE + 2)
> > +
> > +static int
> > +xfs_dax_notify_failure_freeze(
> > +	struct xfs_mount	*mp)
> > +{
> > +	struct super_block	*sb = mp->m_super;
> > +	int			error = 0;
> > +	int			level;
> > +
> > +	/* Wait until we're ready to freeze. */
> > +	down_write(&sb->s_umount);
> > +	while (sb->s_writers.frozen != SB_UNFROZEN) {
> > +		up_write(&sb->s_umount);
> > +
> > +		// just wait, or print warning in dmesg then return -EBUSY?
> > +
> > +		delay(HZ / 10);
> > +		down_write(&sb->s_umount);
> > +	}
> > +
> > +	if (sb_rdonly(sb)) {
> > +		sb->s_writers.frozen = SB_FREEZE_EXCLUSIVE;
> > +		goto out;
> > +	}
> > +
> > +	sb->s_writers.frozen = SB_FREEZE_WRITE;
> > +	/* Release s_umount to preserve sb_start_write -> s_umount ordering */
> > +	up_write(&sb->s_umount);
> > +	percpu_down_write(sb->s_writers.rw_sem + SB_FREEZE_WRITE - 1);
> > +	down_write(&sb->s_umount);
> > +
> > +	/* Now we go and block page faults... */
> > +	sb->s_writers.frozen = SB_FREEZE_PAGEFAULT;
> > +	percpu_down_write(sb->s_writers.rw_sem + SB_FREEZE_PAGEFAULT - 1);
> > +
> > +	/* All writers are done so after syncing there won't be dirty data */
> > +	error = sync_filesystem(sb);
> > +	if (error) {
> > +		sb->s_writers.frozen = SB_UNFROZEN;
> > +		for (level = SB_FREEZE_PAGEFAULT - 1; level >= 0; level--)
> > +			percpu_up_write(sb->s_writers.rw_sem + level);
> > +		wake_up(&sb->s_writers.wait_unfrozen);
> > +		goto out;
> > +	}
> > +
> > +	/* Now wait for internal filesystem counter */
> > +	sb->s_writers.frozen = SB_FREEZE_FS;
> > +	percpu_down_write(sb->s_writers.rw_sem + SB_FREEZE_FS - 1);
> > +
> > +	/*
> > +	 * To prevent anyone else from unfreezing us, set the VFS freeze level
> > +	 * to one higher than SB_FREEZE_COMPLETE.
> > +	 */
> > +	sb->s_writers.frozen = SB_FREEZE_EXCLUSIVE;
> > +	for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--)
> > +		percpu_rwsem_release(sb->s_writers.rw_sem + level, 0,
> > +				_THIS_IP_);
> > +
> > +out:
> > +	up_write(&sb->s_umount);
> > +	return error;
> > +}
> > +
> > +static void
> > +xfs_dax_notify_failure_thaw(
> > +	struct xfs_mount	*mp)
> > +{
> > +	struct super_block	*sb = mp->m_super;
> > +	int			level;
> > +
> > +	down_write(&sb->s_umount);
> > +	if (sb->s_writers.frozen != SB_FREEZE_EXCLUSIVE) {
> > +		/* somebody snuck in and unfroze us? */
> > +		ASSERT(0);
> > +		up_write(&sb->s_umount);
> > +		return;
> > +	}
> > +
> > +	if (sb_rdonly(sb)) {
> > +		sb->s_writers.frozen = SB_UNFROZEN;
> > +		goto out;
> > +	}
> > +
> > +	for (level = 0; level < SB_FREEZE_LEVELS; ++level)
> > +		percpu_rwsem_acquire(sb->s_writers.rw_sem + level, 0,
> > +				_THIS_IP_);
> > +
> > +	sb->s_writers.frozen = SB_UNFROZEN;
> > +	for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--)
> > +		percpu_up_write(sb->s_writers.rw_sem + level);
> > +
> > +out:
> > +	wake_up(&sb->s_writers.wait_unfrozen);
> > +	up_write(&sb->s_umount);
> > +}
> > +
> >   static int
> >   xfs_dax_notify_ddev_failure(
> >   	struct xfs_mount	*mp,
> > @@ -164,11 +277,22 @@ xfs_dax_notify_ddev_failure(
> >   	}
> >   	xfs_trans_cancel(tp);
> > +
> > +	/* Thaw the fs if it is freezed before. */
> > +	if (mf_flags & MF_MEM_PRE_REMOVE)
> > +		xfs_dax_notify_failure_thaw(mp);
> > +
> > +	/*
> > +	 * Determine how to shutdown the filesystem according to the
> > +	 * error code and flags.
> > +	 */
> >   	if (error || notify.want_shutdown) {
> >   		xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_ONDISK);
> >   		if (!error)
> >   			error = -EFSCORRUPTED;
> > -	}
> > +	} else if (mf_flags & MF_MEM_PRE_REMOVE)
> > +		xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
> > +
> >   	return error;
> >   }
> > @@ -182,6 +306,7 @@ xfs_dax_notify_failure(
> >   	struct xfs_mount	*mp = dax_holder(dax_dev);
> >   	u64			ddev_start;
> >   	u64			ddev_end;
> > +	int			error;
> >   	if (!(mp->m_super->s_flags & SB_BORN)) {
> >   		xfs_warn(mp, "filesystem is not ready for notify_failure()!");
> > @@ -196,6 +321,8 @@ xfs_dax_notify_failure(
> >   	if (mp->m_logdev_targp && mp->m_logdev_targp->bt_daxdev == dax_dev &&
> >   	    mp->m_logdev_targp != mp->m_ddev_targp) {
> > +		if (mf_flags & MF_MEM_PRE_REMOVE)
> > +			return 0;
> >   		xfs_err(mp, "ondisk log corrupt, shutting down fs!");
> >   		xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_ONDISK);
> >   		return -EFSCORRUPTED;
> > @@ -209,6 +336,12 @@ xfs_dax_notify_failure(
> >   	ddev_start = mp->m_ddev_targp->bt_dax_part_off;
> >   	ddev_end = ddev_start + bdev_nr_bytes(mp->m_ddev_targp->bt_bdev) - 1;
> > +	/* Notify failure on the whole device. */
> > +	if (offset == 0 && len == U64_MAX) {
> > +		offset = ddev_start;
> > +		len = bdev_nr_bytes(mp->m_ddev_targp->bt_bdev);
> > +	}
> > +
> >   	/* Ignore the range out of filesystem area */
> >   	if (offset + len - 1 < ddev_start)
> >   		return -ENXIO;
> > @@ -225,6 +358,14 @@ xfs_dax_notify_failure(
> >   	if (offset + len - 1 > ddev_end)
> >   		len = ddev_end - offset + 1;
> > +	if (mf_flags & MF_MEM_PRE_REMOVE) {
> > +		xfs_info(mp, "device is about to be removed!");
> > +		/* Freeze fs to prevent new mappings from being created. */
> > +		error = xfs_dax_notify_failure_freeze(mp);
> > +		if (error)
> > +			return error;
> > +	}
> > +
> >   	return xfs_dax_notify_ddev_failure(mp, BTOBB(offset), BTOBB(len),
> >   			mf_flags);
> >   }
> > diff --git a/include/linux/mm.h b/include/linux/mm.h
> > index 1f79667824eb..ac3f22c20e1d 100644
> > --- a/include/linux/mm.h
> > +++ b/include/linux/mm.h
> > @@ -3436,6 +3436,7 @@ enum mf_flags {
> >   	MF_UNPOISON = 1 << 4,
> >   	MF_SW_SIMULATED = 1 << 5,
> >   	MF_NO_RETRY = 1 << 6,
> > +	MF_MEM_PRE_REMOVE = 1 << 7,
> >   };
> >   int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
> >   		      unsigned long count, int mf_flags);
> > diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> > index fae9baf3be16..6e6acec45568 100644
> > --- a/mm/memory-failure.c
> > +++ b/mm/memory-failure.c
> > @@ -623,7 +623,7 @@ static void collect_procs_file(struct page *page, struct list_head *to_kill,
> >    */
> >   static void collect_procs_fsdax(struct page *page,
> >   		struct address_space *mapping, pgoff_t pgoff,
> > -		struct list_head *to_kill)
> > +		struct list_head *to_kill, bool pre_remove)
> >   {
> >   	struct vm_area_struct *vma;
> >   	struct task_struct *tsk;
> > @@ -631,8 +631,15 @@ static void collect_procs_fsdax(struct page *page,
> >   	i_mmap_lock_read(mapping);
> >   	read_lock(&tasklist_lock);
> >   	for_each_process(tsk) {
> > -		struct task_struct *t = task_early_kill(tsk, true);
> > +		struct task_struct *t = tsk;
> > +		/*
> > +		 * Search for all tasks while MF_MEM_PRE_REMOVE, because the
> > +		 * current may not be the one accessing the fsdax page.
> > +		 * Otherwise, search for the current task.
> > +		 */
> > +		if (!pre_remove)
> > +			t = task_early_kill(tsk, true);
> >   		if (!t)
> >   			continue;
> >   		vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
> > @@ -1732,6 +1739,7 @@ int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
> >   	dax_entry_t cookie;
> >   	struct page *page;
> >   	size_t end = index + count;
> > +	bool pre_remove = mf_flags & MF_MEM_PRE_REMOVE;
> >   	mf_flags |= MF_ACTION_REQUIRED | MF_MUST_KILL;
> > @@ -1743,9 +1751,10 @@ int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
> >   		if (!page)
> >   			goto unlock;
> > -		SetPageHWPoison(page);
> > +		if (!pre_remove)
> > +			SetPageHWPoison(page);
> > -		collect_procs_fsdax(page, mapping, index, &to_kill);
> > +		collect_procs_fsdax(page, mapping, index, &to_kill, pre_remove);
> >   		unmap_and_kill(&to_kill, page_to_pfn(page), mapping,
> >   				index, mf_flags);
> >   unlock:
Shiyang Ruan April 25, 2023, 12:47 p.m. UTC | #3
在 2023/4/20 20:09, Jan Kara 写道:
> On Thu 20-04-23 10:07:39, Shiyang Ruan wrote:
>> 在 2023/4/12 18:52, Shiyang Ruan 写道:
>>> This is a RFC HOTFIX.
>>>
>>> This hotfix adds a exclusive forzen state to make sure any others won't
>>> thaw the fs during xfs_dax_notify_failure():
>>>
>>>     #define SB_FREEZE_EXCLUSIVE	(SB_FREEZE_COMPLETE + 2)
>>> Using +2 here is because Darrick's patch[0] is using +1.  So, should we
>>> make these definitions global?
>>>
>>> Another thing I can't make up my mind is: when another freezer has freeze
>>> the fs, should we wait unitl it finish, or print a warning in dmesg and
>>> return -EBUSY?
>>>
>>> Since there are at least 2 places needs exclusive forzen state, I think
>>> we can refactor helper functions of freeze/thaw for them.  e.g.
>>>     int freeze_super_exclusive(struct super_block *sb, int frozen);
>>>     int thaw_super_exclusive(struct super_block *sb, int frozen);
>>>
>>> [0] https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/commit/?h=repair-fscounters&id=c3a0d1de4d54ffb565dbc7092dfe1fb851940669
> 
> I'm OK with the idea of new freeze state that does not allow userspace to
> thaw the filesystem. But I don't really like the guts of filesystem
> freezing being replicated inside XFS. It is bad enough that they are
> replicated in [0], replicating them *once more* in another XFS file shows
> we are definitely doing something wrong. And Luis will need yet another
> incantation of the exlusive freeze for suspend-to-disk. So please guys get
> together and reorganize the generic freezing code so that it supports
> exclusive freeze (for in-kernel users) and works for your usecases instead
> of replicating it inside XFS...

I agree that too much replicating code is not good.  It's necessary to 
create a generic exclusive freeze/thaw for all users.  But for me, I 
don't have the confidence to do it well, because it requires good design 
and code changes will involve other filesystems.  It's diffcult.

However, I hope to be able to make progress on this unbind feature. 
Thus, I tend to refactor a common helper function for xfs first, and 
update the code later when the generic freeze is done.


--
Thanks,
Ruan.

> 
> 								Honza
> 
>>> --- Original commit message ---
>>> This patch is inspired by Dan's "mm, dax, pmem: Introduce
>>> dev_pagemap_failure()"[1].  With the help of dax_holder and
>>> ->notify_failure() mechanism, the pmem driver is able to ask filesystem
>>> (or mapped device) on it to unmap all files in use and notify processes
>>> who are using those files.
>>>
>>> Call trace:
>>> trigger unbind
>>>    -> unbind_store()
>>>     -> ... (skip)
>>>      -> devres_release_all()
>>>       -> kill_dax()
>>>        -> dax_holder_notify_failure(dax_dev, 0, U64_MAX, MF_MEM_PRE_REMOVE)
>>>         -> xfs_dax_notify_failure()
>>>         `-> freeze_super()
>>>         `-> do xfs rmap
>>>         ` -> mf_dax_kill_procs()
>>>         `  -> collect_procs_fsdax()    // all associated
>>>         `  -> unmap_and_kill()
>>>         ` -> invalidate_inode_pages2() // drop file's cache
>>>         `-> thaw_super()
>>>
>>> Introduce MF_MEM_PRE_REMOVE to let filesystem know this is a remove
>>> event.  Also introduce a exclusive freeze/thaw to lock the filesystem to
>>> prevent new dax mapping from being created.  And do not shutdown
>>> filesystem directly if something not supported, or if failure range
>>> includes metadata area.  Make sure all files and processes are handled
>>> correctly.  Also drop the cache of associated files before pmem is
>>> removed.
>>>
>>> [1]: https://lore.kernel.org/linux-mm/161604050314.1463742.14151665140035795571.stgit@dwillia2-desk3.amr.corp.intel.com/
>>>
>>> Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
>>> ---
>>>    drivers/dax/super.c         |   3 +-
>>>    fs/xfs/xfs_notify_failure.c | 151 ++++++++++++++++++++++++++++++++++--
>>>    include/linux/mm.h          |   1 +
>>>    mm/memory-failure.c         |  17 +++-
>>>    4 files changed, 162 insertions(+), 10 deletions(-)
>>>
>>> diff --git a/drivers/dax/super.c b/drivers/dax/super.c
>>> index c4c4728a36e4..2e1a35e82fce 100644
>>> --- a/drivers/dax/super.c
>>> +++ b/drivers/dax/super.c
>>> @@ -323,7 +323,8 @@ void kill_dax(struct dax_device *dax_dev)
>>>    		return;
>>>    	if (dax_dev->holder_data != NULL)
>>> -		dax_holder_notify_failure(dax_dev, 0, U64_MAX, 0);
>>> +		dax_holder_notify_failure(dax_dev, 0, U64_MAX,
>>> +				MF_MEM_PRE_REMOVE);
>>>    	clear_bit(DAXDEV_ALIVE, &dax_dev->flags);
>>>    	synchronize_srcu(&dax_srcu);
>>> diff --git a/fs/xfs/xfs_notify_failure.c b/fs/xfs/xfs_notify_failure.c
>>> index 1e2eddb8f90f..796dd954d33a 100644
>>> --- a/fs/xfs/xfs_notify_failure.c
>>> +++ b/fs/xfs/xfs_notify_failure.c
>>> @@ -22,6 +22,7 @@
>>>    #include <linux/mm.h>
>>>    #include <linux/dax.h>
>>> +#include <linux/fs.h>
>>>    struct xfs_failure_info {
>>>    	xfs_agblock_t		startblock;
>>> @@ -73,10 +74,16 @@ xfs_dax_failure_fn(
>>>    	struct xfs_mount		*mp = cur->bc_mp;
>>>    	struct xfs_inode		*ip;
>>>    	struct xfs_failure_info		*notify = data;
>>> +	struct address_space		*mapping;
>>> +	pgoff_t				pgoff;
>>> +	unsigned long			pgcnt;
>>>    	int				error = 0;
>>>    	if (XFS_RMAP_NON_INODE_OWNER(rec->rm_owner) ||
>>>    	    (rec->rm_flags & (XFS_RMAP_ATTR_FORK | XFS_RMAP_BMBT_BLOCK))) {
>>> +		/* The device is about to be removed.  Not a really failure. */
>>> +		if (notify->mf_flags & MF_MEM_PRE_REMOVE)
>>> +			return 0;
>>>    		notify->want_shutdown = true;
>>>    		return 0;
>>>    	}
>>> @@ -92,14 +99,120 @@ xfs_dax_failure_fn(
>>>    		return 0;
>>>    	}
>>> -	error = mf_dax_kill_procs(VFS_I(ip)->i_mapping,
>>> -				  xfs_failure_pgoff(mp, rec, notify),
>>> -				  xfs_failure_pgcnt(mp, rec, notify),
>>> -				  notify->mf_flags);
>>> +	mapping = VFS_I(ip)->i_mapping;
>>> +	pgoff = xfs_failure_pgoff(mp, rec, notify);
>>> +	pgcnt = xfs_failure_pgcnt(mp, rec, notify);
>>> +
>>> +	/* Continue the rmap query if the inode isn't a dax file. */
>>> +	if (dax_mapping(mapping))
>>> +		error = mf_dax_kill_procs(mapping, pgoff, pgcnt,
>>> +				notify->mf_flags);
>>> +
>>> +	/* Invalidate the cache anyway. */
>>> +	invalidate_inode_pages2_range(mapping, pgoff, pgoff + pgcnt - 1);
>>> +
>>>    	xfs_irele(ip);
>>>    	return error;
>>>    }
>>> +#define SB_FREEZE_EXCLUSIVE	(SB_FREEZE_COMPLETE + 2)
>>> +
>>> +static int
>>> +xfs_dax_notify_failure_freeze(
>>> +	struct xfs_mount	*mp)
>>> +{
>>> +	struct super_block	*sb = mp->m_super;
>>> +	int			error = 0;
>>> +	int			level;
>>> +
>>> +	/* Wait until we're ready to freeze. */
>>> +	down_write(&sb->s_umount);
>>> +	while (sb->s_writers.frozen != SB_UNFROZEN) {
>>> +		up_write(&sb->s_umount);
>>> +
>>> +		// just wait, or print warning in dmesg then return -EBUSY?
>>> +
>>> +		delay(HZ / 10);
>>> +		down_write(&sb->s_umount);
>>> +	}
>>> +
>>> +	if (sb_rdonly(sb)) {
>>> +		sb->s_writers.frozen = SB_FREEZE_EXCLUSIVE;
>>> +		goto out;
>>> +	}
>>> +
>>> +	sb->s_writers.frozen = SB_FREEZE_WRITE;
>>> +	/* Release s_umount to preserve sb_start_write -> s_umount ordering */
>>> +	up_write(&sb->s_umount);
>>> +	percpu_down_write(sb->s_writers.rw_sem + SB_FREEZE_WRITE - 1);
>>> +	down_write(&sb->s_umount);
>>> +
>>> +	/* Now we go and block page faults... */
>>> +	sb->s_writers.frozen = SB_FREEZE_PAGEFAULT;
>>> +	percpu_down_write(sb->s_writers.rw_sem + SB_FREEZE_PAGEFAULT - 1);
>>> +
>>> +	/* All writers are done so after syncing there won't be dirty data */
>>> +	error = sync_filesystem(sb);
>>> +	if (error) {
>>> +		sb->s_writers.frozen = SB_UNFROZEN;
>>> +		for (level = SB_FREEZE_PAGEFAULT - 1; level >= 0; level--)
>>> +			percpu_up_write(sb->s_writers.rw_sem + level);
>>> +		wake_up(&sb->s_writers.wait_unfrozen);
>>> +		goto out;
>>> +	}
>>> +
>>> +	/* Now wait for internal filesystem counter */
>>> +	sb->s_writers.frozen = SB_FREEZE_FS;
>>> +	percpu_down_write(sb->s_writers.rw_sem + SB_FREEZE_FS - 1);
>>> +
>>> +	/*
>>> +	 * To prevent anyone else from unfreezing us, set the VFS freeze level
>>> +	 * to one higher than SB_FREEZE_COMPLETE.
>>> +	 */
>>> +	sb->s_writers.frozen = SB_FREEZE_EXCLUSIVE;
>>> +	for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--)
>>> +		percpu_rwsem_release(sb->s_writers.rw_sem + level, 0,
>>> +				_THIS_IP_);
>>> +
>>> +out:
>>> +	up_write(&sb->s_umount);
>>> +	return error;
>>> +}
>>> +
>>> +static void
>>> +xfs_dax_notify_failure_thaw(
>>> +	struct xfs_mount	*mp)
>>> +{
>>> +	struct super_block	*sb = mp->m_super;
>>> +	int			level;
>>> +
>>> +	down_write(&sb->s_umount);
>>> +	if (sb->s_writers.frozen != SB_FREEZE_EXCLUSIVE) {
>>> +		/* somebody snuck in and unfroze us? */
>>> +		ASSERT(0);
>>> +		up_write(&sb->s_umount);
>>> +		return;
>>> +	}
>>> +
>>> +	if (sb_rdonly(sb)) {
>>> +		sb->s_writers.frozen = SB_UNFROZEN;
>>> +		goto out;
>>> +	}
>>> +
>>> +	for (level = 0; level < SB_FREEZE_LEVELS; ++level)
>>> +		percpu_rwsem_acquire(sb->s_writers.rw_sem + level, 0,
>>> +				_THIS_IP_);
>>> +
>>> +	sb->s_writers.frozen = SB_UNFROZEN;
>>> +	for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--)
>>> +		percpu_up_write(sb->s_writers.rw_sem + level);
>>> +
>>> +out:
>>> +	wake_up(&sb->s_writers.wait_unfrozen);
>>> +	up_write(&sb->s_umount);
>>> +}
>>> +
>>>    static int
>>>    xfs_dax_notify_ddev_failure(
>>>    	struct xfs_mount	*mp,
>>> @@ -164,11 +277,22 @@ xfs_dax_notify_ddev_failure(
>>>    	}
>>>    	xfs_trans_cancel(tp);
>>> +
>>> +	/* Thaw the fs if it is freezed before. */
>>> +	if (mf_flags & MF_MEM_PRE_REMOVE)
>>> +		xfs_dax_notify_failure_thaw(mp);
>>> +
>>> +	/*
>>> +	 * Determine how to shutdown the filesystem according to the
>>> +	 * error code and flags.
>>> +	 */
>>>    	if (error || notify.want_shutdown) {
>>>    		xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_ONDISK);
>>>    		if (!error)
>>>    			error = -EFSCORRUPTED;
>>> -	}
>>> +	} else if (mf_flags & MF_MEM_PRE_REMOVE)
>>> +		xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
>>> +
>>>    	return error;
>>>    }
>>> @@ -182,6 +306,7 @@ xfs_dax_notify_failure(
>>>    	struct xfs_mount	*mp = dax_holder(dax_dev);
>>>    	u64			ddev_start;
>>>    	u64			ddev_end;
>>> +	int			error;
>>>    	if (!(mp->m_super->s_flags & SB_BORN)) {
>>>    		xfs_warn(mp, "filesystem is not ready for notify_failure()!");
>>> @@ -196,6 +321,8 @@ xfs_dax_notify_failure(
>>>    	if (mp->m_logdev_targp && mp->m_logdev_targp->bt_daxdev == dax_dev &&
>>>    	    mp->m_logdev_targp != mp->m_ddev_targp) {
>>> +		if (mf_flags & MF_MEM_PRE_REMOVE)
>>> +			return 0;
>>>    		xfs_err(mp, "ondisk log corrupt, shutting down fs!");
>>>    		xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_ONDISK);
>>>    		return -EFSCORRUPTED;
>>> @@ -209,6 +336,12 @@ xfs_dax_notify_failure(
>>>    	ddev_start = mp->m_ddev_targp->bt_dax_part_off;
>>>    	ddev_end = ddev_start + bdev_nr_bytes(mp->m_ddev_targp->bt_bdev) - 1;
>>> +	/* Notify failure on the whole device. */
>>> +	if (offset == 0 && len == U64_MAX) {
>>> +		offset = ddev_start;
>>> +		len = bdev_nr_bytes(mp->m_ddev_targp->bt_bdev);
>>> +	}
>>> +
>>>    	/* Ignore the range out of filesystem area */
>>>    	if (offset + len - 1 < ddev_start)
>>>    		return -ENXIO;
>>> @@ -225,6 +358,14 @@ xfs_dax_notify_failure(
>>>    	if (offset + len - 1 > ddev_end)
>>>    		len = ddev_end - offset + 1;
>>> +	if (mf_flags & MF_MEM_PRE_REMOVE) {
>>> +		xfs_info(mp, "device is about to be removed!");
>>> +		/* Freeze fs to prevent new mappings from being created. */
>>> +		error = xfs_dax_notify_failure_freeze(mp);
>>> +		if (error)
>>> +			return error;
>>> +	}
>>> +
>>>    	return xfs_dax_notify_ddev_failure(mp, BTOBB(offset), BTOBB(len),
>>>    			mf_flags);
>>>    }
>>> diff --git a/include/linux/mm.h b/include/linux/mm.h
>>> index 1f79667824eb..ac3f22c20e1d 100644
>>> --- a/include/linux/mm.h
>>> +++ b/include/linux/mm.h
>>> @@ -3436,6 +3436,7 @@ enum mf_flags {
>>>    	MF_UNPOISON = 1 << 4,
>>>    	MF_SW_SIMULATED = 1 << 5,
>>>    	MF_NO_RETRY = 1 << 6,
>>> +	MF_MEM_PRE_REMOVE = 1 << 7,
>>>    };
>>>    int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
>>>    		      unsigned long count, int mf_flags);
>>> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
>>> index fae9baf3be16..6e6acec45568 100644
>>> --- a/mm/memory-failure.c
>>> +++ b/mm/memory-failure.c
>>> @@ -623,7 +623,7 @@ static void collect_procs_file(struct page *page, struct list_head *to_kill,
>>>     */
>>>    static void collect_procs_fsdax(struct page *page,
>>>    		struct address_space *mapping, pgoff_t pgoff,
>>> -		struct list_head *to_kill)
>>> +		struct list_head *to_kill, bool pre_remove)
>>>    {
>>>    	struct vm_area_struct *vma;
>>>    	struct task_struct *tsk;
>>> @@ -631,8 +631,15 @@ static void collect_procs_fsdax(struct page *page,
>>>    	i_mmap_lock_read(mapping);
>>>    	read_lock(&tasklist_lock);
>>>    	for_each_process(tsk) {
>>> -		struct task_struct *t = task_early_kill(tsk, true);
>>> +		struct task_struct *t = tsk;
>>> +		/*
>>> +		 * Search for all tasks while MF_MEM_PRE_REMOVE, because the
>>> +		 * current may not be the one accessing the fsdax page.
>>> +		 * Otherwise, search for the current task.
>>> +		 */
>>> +		if (!pre_remove)
>>> +			t = task_early_kill(tsk, true);
>>>    		if (!t)
>>>    			continue;
>>>    		vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
>>> @@ -1732,6 +1739,7 @@ int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
>>>    	dax_entry_t cookie;
>>>    	struct page *page;
>>>    	size_t end = index + count;
>>> +	bool pre_remove = mf_flags & MF_MEM_PRE_REMOVE;
>>>    	mf_flags |= MF_ACTION_REQUIRED | MF_MUST_KILL;
>>> @@ -1743,9 +1751,10 @@ int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
>>>    		if (!page)
>>>    			goto unlock;
>>> -		SetPageHWPoison(page);
>>> +		if (!pre_remove)
>>> +			SetPageHWPoison(page);
>>> -		collect_procs_fsdax(page, mapping, index, &to_kill);
>>> +		collect_procs_fsdax(page, mapping, index, &to_kill, pre_remove);
>>>    		unmap_and_kill(&to_kill, page_to_pfn(page), mapping,
>>>    				index, mf_flags);
>>>    unlock:
Jan Kara April 25, 2023, 1:23 p.m. UTC | #4
On Tue 25-04-23 20:47:35, Shiyang Ruan wrote:
> 
> 
> 在 2023/4/20 20:09, Jan Kara 写道:
> > On Thu 20-04-23 10:07:39, Shiyang Ruan wrote:
> > > 在 2023/4/12 18:52, Shiyang Ruan 写道:
> > > > This is a RFC HOTFIX.
> > > > 
> > > > This hotfix adds a exclusive forzen state to make sure any others won't
> > > > thaw the fs during xfs_dax_notify_failure():
> > > > 
> > > >     #define SB_FREEZE_EXCLUSIVE	(SB_FREEZE_COMPLETE + 2)
> > > > Using +2 here is because Darrick's patch[0] is using +1.  So, should we
> > > > make these definitions global?
> > > > 
> > > > Another thing I can't make up my mind is: when another freezer has freeze
> > > > the fs, should we wait unitl it finish, or print a warning in dmesg and
> > > > return -EBUSY?
> > > > 
> > > > Since there are at least 2 places needs exclusive forzen state, I think
> > > > we can refactor helper functions of freeze/thaw for them.  e.g.
> > > >     int freeze_super_exclusive(struct super_block *sb, int frozen);
> > > >     int thaw_super_exclusive(struct super_block *sb, int frozen);
> > > > 
> > > > [0] https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/commit/?h=repair-fscounters&id=c3a0d1de4d54ffb565dbc7092dfe1fb851940669
> > 
> > I'm OK with the idea of new freeze state that does not allow userspace to
> > thaw the filesystem. But I don't really like the guts of filesystem
> > freezing being replicated inside XFS. It is bad enough that they are
> > replicated in [0], replicating them *once more* in another XFS file shows
> > we are definitely doing something wrong. And Luis will need yet another
> > incantation of the exlusive freeze for suspend-to-disk. So please guys get
> > together and reorganize the generic freezing code so that it supports
> > exclusive freeze (for in-kernel users) and works for your usecases instead
> > of replicating it inside XFS...
> 
> I agree that too much replicating code is not good.  It's necessary to
> create a generic exclusive freeze/thaw for all users.  But for me, I don't
> have the confidence to do it well, because it requires good design and code
> changes will involve other filesystems.  It's diffcult.
> 
> However, I hope to be able to make progress on this unbind feature. Thus, I
> tend to refactor a common helper function for xfs first, and update the code
> later when the generic freeze is done.

I think Darrick was thinking about working on a proper generic interface.
So please coordinate with him.

								Honza
Darrick J. Wong April 25, 2023, 3:18 p.m. UTC | #5
On Tue, Apr 25, 2023 at 03:23:15PM +0200, Jan Kara wrote:
> On Tue 25-04-23 20:47:35, Shiyang Ruan wrote:
> > 
> > 
> > 在 2023/4/20 20:09, Jan Kara 写道:
> > > On Thu 20-04-23 10:07:39, Shiyang Ruan wrote:
> > > > 在 2023/4/12 18:52, Shiyang Ruan 写道:
> > > > > This is a RFC HOTFIX.
> > > > > 
> > > > > This hotfix adds a exclusive forzen state to make sure any others won't
> > > > > thaw the fs during xfs_dax_notify_failure():
> > > > > 
> > > > >     #define SB_FREEZE_EXCLUSIVE	(SB_FREEZE_COMPLETE + 2)
> > > > > Using +2 here is because Darrick's patch[0] is using +1.  So, should we
> > > > > make these definitions global?
> > > > > 
> > > > > Another thing I can't make up my mind is: when another freezer has freeze
> > > > > the fs, should we wait unitl it finish, or print a warning in dmesg and
> > > > > return -EBUSY?
> > > > > 
> > > > > Since there are at least 2 places needs exclusive forzen state, I think
> > > > > we can refactor helper functions of freeze/thaw for them.  e.g.
> > > > >     int freeze_super_exclusive(struct super_block *sb, int frozen);
> > > > >     int thaw_super_exclusive(struct super_block *sb, int frozen);
> > > > > 
> > > > > [0] https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/commit/?h=repair-fscounters&id=c3a0d1de4d54ffb565dbc7092dfe1fb851940669
> > > 
> > > I'm OK with the idea of new freeze state that does not allow userspace to
> > > thaw the filesystem. But I don't really like the guts of filesystem
> > > freezing being replicated inside XFS. It is bad enough that they are
> > > replicated in [0], replicating them *once more* in another XFS file shows
> > > we are definitely doing something wrong. And Luis will need yet another
> > > incantation of the exlusive freeze for suspend-to-disk. So please guys get
> > > together and reorganize the generic freezing code so that it supports
> > > exclusive freeze (for in-kernel users) and works for your usecases instead
> > > of replicating it inside XFS...
> > 
> > I agree that too much replicating code is not good.  It's necessary to
> > create a generic exclusive freeze/thaw for all users.  But for me, I don't
> > have the confidence to do it well, because it requires good design and code
> > changes will involve other filesystems.  It's diffcult.
> > 
> > However, I hope to be able to make progress on this unbind feature. Thus, I
> > tend to refactor a common helper function for xfs first, and update the code
> > later when the generic freeze is done.
> 
> I think Darrick was thinking about working on a proper generic interface.
> So please coordinate with him.

I'll post a vfs generic kernelfreeze series later today.

One thing I haven't figured out yet is what's supposed to happen when
PREREMOVE is called on a frozen filesystem.  We don't want userspace to
be able to thaw the fs while PREREMOVE is running, so I /guess/ that
means we need some method for the kernel to take over a userspace
freeze and then put it back when we're done?

--D

> 								Honza
> 
> -- 
> Jan Kara <jack@suse.com>
> SUSE Labs, CR
Shiyang Ruan April 26, 2023, 2:27 a.m. UTC | #6
在 2023/4/25 23:18, Darrick J. Wong 写道:
> On Tue, Apr 25, 2023 at 03:23:15PM +0200, Jan Kara wrote:
>> On Tue 25-04-23 20:47:35, Shiyang Ruan wrote:
>>>
>>>
>>> 在 2023/4/20 20:09, Jan Kara 写道:
>>>> On Thu 20-04-23 10:07:39, Shiyang Ruan wrote:
>>>>> 在 2023/4/12 18:52, Shiyang Ruan 写道:
>>>>>> This is a RFC HOTFIX.
>>>>>>
>>>>>> This hotfix adds a exclusive forzen state to make sure any others won't
>>>>>> thaw the fs during xfs_dax_notify_failure():
>>>>>>
>>>>>>      #define SB_FREEZE_EXCLUSIVE	(SB_FREEZE_COMPLETE + 2)
>>>>>> Using +2 here is because Darrick's patch[0] is using +1.  So, should we
>>>>>> make these definitions global?
>>>>>>
>>>>>> Another thing I can't make up my mind is: when another freezer has freeze
>>>>>> the fs, should we wait unitl it finish, or print a warning in dmesg and
>>>>>> return -EBUSY?
>>>>>>
>>>>>> Since there are at least 2 places needs exclusive forzen state, I think
>>>>>> we can refactor helper functions of freeze/thaw for them.  e.g.
>>>>>>      int freeze_super_exclusive(struct super_block *sb, int frozen);
>>>>>>      int thaw_super_exclusive(struct super_block *sb, int frozen);
>>>>>>
>>>>>> [0] https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/commit/?h=repair-fscounters&id=c3a0d1de4d54ffb565dbc7092dfe1fb851940669
>>>>
>>>> I'm OK with the idea of new freeze state that does not allow userspace to
>>>> thaw the filesystem. But I don't really like the guts of filesystem
>>>> freezing being replicated inside XFS. It is bad enough that they are
>>>> replicated in [0], replicating them *once more* in another XFS file shows
>>>> we are definitely doing something wrong. And Luis will need yet another
>>>> incantation of the exlusive freeze for suspend-to-disk. So please guys get
>>>> together and reorganize the generic freezing code so that it supports
>>>> exclusive freeze (for in-kernel users) and works for your usecases instead
>>>> of replicating it inside XFS...
>>>
>>> I agree that too much replicating code is not good.  It's necessary to
>>> create a generic exclusive freeze/thaw for all users.  But for me, I don't
>>> have the confidence to do it well, because it requires good design and code
>>> changes will involve other filesystems.  It's diffcult.
>>>
>>> However, I hope to be able to make progress on this unbind feature. Thus, I
>>> tend to refactor a common helper function for xfs first, and update the code
>>> later when the generic freeze is done.
>>
>> I think Darrick was thinking about working on a proper generic interface.
>> So please coordinate with him.
> 
> I'll post a vfs generic kernelfreeze series later today.
> 
> One thing I haven't figured out yet is what's supposed to happen when
> PREREMOVE is called on a frozen filesystem.

call PREREMOVE when:
1. freezed by kernel:    we wait unitl kernel thaws -> not sure
2. freezed by userspace: we take over the control of freeze state:
      a. userspace can't thaw before PREREMOVE is done
      b. kernel keeps freeze state after PREREMOVE is done and before 
userspace thaws

Since the unbind interface doesn't return any other errcode except 
-ENODEV, the only thing I can think of to do is wait for the other one 
done?  If another one doesn't thaw after a long time waitting, we print 
a "waitting too long" warning in dmesg.  But I'm not sure if this is good.

> We don't want userspace to
> be able to thaw the fs while PREREMOVE is running, so I /guess/ that
> means we need some method for the kernel to take over a userspace
> freeze and then put it back when we're done?

As is designed by Luis, we can add sb->s_writers.frozen_by_user flag to 
distinguish whether current freeze state is initiated by kernel or 
userspace.  In his patch, userspace can take over kernel's freeze.  We 
just need to switch the order.


--
Thanks,
Ruan.

> 
> --D
> 
>> 								Honza
>>
>> -- 
>> Jan Kara <jack@suse.com>
>> SUSE Labs, CR
Darrick J. Wong April 26, 2023, 2:37 a.m. UTC | #7
On Wed, Apr 26, 2023 at 10:27:43AM +0800, Shiyang Ruan wrote:
> 
> 
> 在 2023/4/25 23:18, Darrick J. Wong 写道:
> > On Tue, Apr 25, 2023 at 03:23:15PM +0200, Jan Kara wrote:
> > > On Tue 25-04-23 20:47:35, Shiyang Ruan wrote:
> > > > 
> > > > 
> > > > 在 2023/4/20 20:09, Jan Kara 写道:
> > > > > On Thu 20-04-23 10:07:39, Shiyang Ruan wrote:
> > > > > > 在 2023/4/12 18:52, Shiyang Ruan 写道:
> > > > > > > This is a RFC HOTFIX.
> > > > > > > 
> > > > > > > This hotfix adds a exclusive forzen state to make sure any others won't
> > > > > > > thaw the fs during xfs_dax_notify_failure():
> > > > > > > 
> > > > > > >      #define SB_FREEZE_EXCLUSIVE	(SB_FREEZE_COMPLETE + 2)
> > > > > > > Using +2 here is because Darrick's patch[0] is using +1.  So, should we
> > > > > > > make these definitions global?
> > > > > > > 
> > > > > > > Another thing I can't make up my mind is: when another freezer has freeze
> > > > > > > the fs, should we wait unitl it finish, or print a warning in dmesg and
> > > > > > > return -EBUSY?
> > > > > > > 
> > > > > > > Since there are at least 2 places needs exclusive forzen state, I think
> > > > > > > we can refactor helper functions of freeze/thaw for them.  e.g.
> > > > > > >      int freeze_super_exclusive(struct super_block *sb, int frozen);
> > > > > > >      int thaw_super_exclusive(struct super_block *sb, int frozen);
> > > > > > > 
> > > > > > > [0] https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/commit/?h=repair-fscounters&id=c3a0d1de4d54ffb565dbc7092dfe1fb851940669
> > > > > 
> > > > > I'm OK with the idea of new freeze state that does not allow userspace to
> > > > > thaw the filesystem. But I don't really like the guts of filesystem
> > > > > freezing being replicated inside XFS. It is bad enough that they are
> > > > > replicated in [0], replicating them *once more* in another XFS file shows
> > > > > we are definitely doing something wrong. And Luis will need yet another
> > > > > incantation of the exlusive freeze for suspend-to-disk. So please guys get
> > > > > together and reorganize the generic freezing code so that it supports
> > > > > exclusive freeze (for in-kernel users) and works for your usecases instead
> > > > > of replicating it inside XFS...
> > > > 
> > > > I agree that too much replicating code is not good.  It's necessary to
> > > > create a generic exclusive freeze/thaw for all users.  But for me, I don't
> > > > have the confidence to do it well, because it requires good design and code
> > > > changes will involve other filesystems.  It's diffcult.
> > > > 
> > > > However, I hope to be able to make progress on this unbind feature. Thus, I
> > > > tend to refactor a common helper function for xfs first, and update the code
> > > > later when the generic freeze is done.
> > > 
> > > I think Darrick was thinking about working on a proper generic interface.
> > > So please coordinate with him.
> > 
> > I'll post a vfs generic kernelfreeze series later today.
> > 
> > One thing I haven't figured out yet is what's supposed to happen when
> > PREREMOVE is called on a frozen filesystem.
> 
> call PREREMOVE when:
> 1. freezed by kernel:    we wait unitl kernel thaws -> not sure
> 2. freezed by userspace: we take over the control of freeze state:
>      a. userspace can't thaw before PREREMOVE is done
>      b. kernel keeps freeze state after PREREMOVE is done and before
> userspace thaws
> 
> Since the unbind interface doesn't return any other errcode except -ENODEV,
> the only thing I can think of to do is wait for the other one done?  If
> another one doesn't thaw after a long time waitting, we print a "waitting
> too long" warning in dmesg.  But I'm not sure if this is good.
> 
> > We don't want userspace to
> > be able to thaw the fs while PREREMOVE is running, so I /guess/ that
> > means we need some method for the kernel to take over a userspace
> > freeze and then put it back when we're done?
> 
> As is designed by Luis, we can add sb->s_writers.frozen_by_user flag to
> distinguish whether current freeze state is initiated by kernel or
> userspace.  In his patch, userspace can take over kernel's freeze.  We just
> need to switch the order.

<nod> How does this patchset
https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/commit/?h=djwong-wtf&id=a97da76ed5256d692a02ece01b4032dbf68cbf89
https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/commit/?h=djwong-wtf&id=93310faf77480265b3bc784f6883f5af9ccfce3b
https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux.git/commit/?h=djwong-wtf&id=a68cea1aa317775046372840ee4f0ba5bdb75d9f

strike you?

I think for #2 above I could write a freeze_super_excl variant that
turns a userspace freeze into a kernel freeze, and a thaw_super_excl
variant that changes it back.

--D

> 
> 
> --
> Thanks,
> Ruan.
> 
> > 
> > --D
> > 
> > > 								Honza
> > > 
> > > -- 
> > > Jan Kara <jack@suse.com>
> > > SUSE Labs, CR
diff mbox series

Patch

diff --git a/drivers/dax/super.c b/drivers/dax/super.c
index c4c4728a36e4..2e1a35e82fce 100644
--- a/drivers/dax/super.c
+++ b/drivers/dax/super.c
@@ -323,7 +323,8 @@  void kill_dax(struct dax_device *dax_dev)
 		return;
 
 	if (dax_dev->holder_data != NULL)
-		dax_holder_notify_failure(dax_dev, 0, U64_MAX, 0);
+		dax_holder_notify_failure(dax_dev, 0, U64_MAX,
+				MF_MEM_PRE_REMOVE);
 
 	clear_bit(DAXDEV_ALIVE, &dax_dev->flags);
 	synchronize_srcu(&dax_srcu);
diff --git a/fs/xfs/xfs_notify_failure.c b/fs/xfs/xfs_notify_failure.c
index 1e2eddb8f90f..796dd954d33a 100644
--- a/fs/xfs/xfs_notify_failure.c
+++ b/fs/xfs/xfs_notify_failure.c
@@ -22,6 +22,7 @@ 
 
 #include <linux/mm.h>
 #include <linux/dax.h>
+#include <linux/fs.h>
 
 struct xfs_failure_info {
 	xfs_agblock_t		startblock;
@@ -73,10 +74,16 @@  xfs_dax_failure_fn(
 	struct xfs_mount		*mp = cur->bc_mp;
 	struct xfs_inode		*ip;
 	struct xfs_failure_info		*notify = data;
+	struct address_space		*mapping;
+	pgoff_t				pgoff;
+	unsigned long			pgcnt;
 	int				error = 0;
 
 	if (XFS_RMAP_NON_INODE_OWNER(rec->rm_owner) ||
 	    (rec->rm_flags & (XFS_RMAP_ATTR_FORK | XFS_RMAP_BMBT_BLOCK))) {
+		/* The device is about to be removed.  Not a really failure. */
+		if (notify->mf_flags & MF_MEM_PRE_REMOVE)
+			return 0;
 		notify->want_shutdown = true;
 		return 0;
 	}
@@ -92,14 +99,120 @@  xfs_dax_failure_fn(
 		return 0;
 	}
 
-	error = mf_dax_kill_procs(VFS_I(ip)->i_mapping,
-				  xfs_failure_pgoff(mp, rec, notify),
-				  xfs_failure_pgcnt(mp, rec, notify),
-				  notify->mf_flags);
+	mapping = VFS_I(ip)->i_mapping;
+	pgoff = xfs_failure_pgoff(mp, rec, notify);
+	pgcnt = xfs_failure_pgcnt(mp, rec, notify);
+
+	/* Continue the rmap query if the inode isn't a dax file. */
+	if (dax_mapping(mapping))
+		error = mf_dax_kill_procs(mapping, pgoff, pgcnt,
+				notify->mf_flags);
+
+	/* Invalidate the cache anyway. */
+	invalidate_inode_pages2_range(mapping, pgoff, pgoff + pgcnt - 1);
+
 	xfs_irele(ip);
 	return error;
 }
 
+#define SB_FREEZE_EXCLUSIVE	(SB_FREEZE_COMPLETE + 2)
+
+static int
+xfs_dax_notify_failure_freeze(
+	struct xfs_mount	*mp)
+{
+	struct super_block	*sb = mp->m_super;
+	int			error = 0;
+	int			level;
+
+	/* Wait until we're ready to freeze. */
+	down_write(&sb->s_umount);
+	while (sb->s_writers.frozen != SB_UNFROZEN) {
+		up_write(&sb->s_umount);
+
+		// just wait, or print warning in dmesg then return -EBUSY?
+
+		delay(HZ / 10);
+		down_write(&sb->s_umount);
+	}
+
+	if (sb_rdonly(sb)) {
+		sb->s_writers.frozen = SB_FREEZE_EXCLUSIVE;
+		goto out;
+	}
+
+	sb->s_writers.frozen = SB_FREEZE_WRITE;
+	/* Release s_umount to preserve sb_start_write -> s_umount ordering */
+	up_write(&sb->s_umount);
+	percpu_down_write(sb->s_writers.rw_sem + SB_FREEZE_WRITE - 1);
+	down_write(&sb->s_umount);
+
+	/* Now we go and block page faults... */
+	sb->s_writers.frozen = SB_FREEZE_PAGEFAULT;
+	percpu_down_write(sb->s_writers.rw_sem + SB_FREEZE_PAGEFAULT - 1);
+
+	/* All writers are done so after syncing there won't be dirty data */
+	error = sync_filesystem(sb);
+	if (error) {
+		sb->s_writers.frozen = SB_UNFROZEN;
+		for (level = SB_FREEZE_PAGEFAULT - 1; level >= 0; level--)
+			percpu_up_write(sb->s_writers.rw_sem + level);
+		wake_up(&sb->s_writers.wait_unfrozen);
+		goto out;
+	}
+
+	/* Now wait for internal filesystem counter */
+	sb->s_writers.frozen = SB_FREEZE_FS;
+	percpu_down_write(sb->s_writers.rw_sem + SB_FREEZE_FS - 1);
+
+	/*
+	 * To prevent anyone else from unfreezing us, set the VFS freeze level
+	 * to one higher than SB_FREEZE_COMPLETE.
+	 */
+	sb->s_writers.frozen = SB_FREEZE_EXCLUSIVE;
+	for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--)
+		percpu_rwsem_release(sb->s_writers.rw_sem + level, 0,
+				_THIS_IP_);
+
+out:
+	up_write(&sb->s_umount);
+	return error;
+}
+
+static void
+xfs_dax_notify_failure_thaw(
+	struct xfs_mount	*mp)
+{
+	struct super_block	*sb = mp->m_super;
+	int			level;
+
+	down_write(&sb->s_umount);
+	if (sb->s_writers.frozen != SB_FREEZE_EXCLUSIVE) {
+		/* somebody snuck in and unfroze us? */
+		ASSERT(0);
+		up_write(&sb->s_umount);
+		return;
+	}
+
+	if (sb_rdonly(sb)) {
+		sb->s_writers.frozen = SB_UNFROZEN;
+		goto out;
+	}
+
+	for (level = 0; level < SB_FREEZE_LEVELS; ++level)
+		percpu_rwsem_acquire(sb->s_writers.rw_sem + level, 0,
+				_THIS_IP_);
+
+	sb->s_writers.frozen = SB_UNFROZEN;
+	for (level = SB_FREEZE_LEVELS - 1; level >= 0; level--)
+		percpu_up_write(sb->s_writers.rw_sem + level);
+
+out:
+	wake_up(&sb->s_writers.wait_unfrozen);
+	up_write(&sb->s_umount);
+}
+
 static int
 xfs_dax_notify_ddev_failure(
 	struct xfs_mount	*mp,
@@ -164,11 +277,22 @@  xfs_dax_notify_ddev_failure(
 	}
 
 	xfs_trans_cancel(tp);
+
+	/* Thaw the fs if it is freezed before. */
+	if (mf_flags & MF_MEM_PRE_REMOVE)
+		xfs_dax_notify_failure_thaw(mp);
+
+	/*
+	 * Determine how to shutdown the filesystem according to the
+	 * error code and flags.
+	 */
 	if (error || notify.want_shutdown) {
 		xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_ONDISK);
 		if (!error)
 			error = -EFSCORRUPTED;
-	}
+	} else if (mf_flags & MF_MEM_PRE_REMOVE)
+		xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
+
 	return error;
 }
 
@@ -182,6 +306,7 @@  xfs_dax_notify_failure(
 	struct xfs_mount	*mp = dax_holder(dax_dev);
 	u64			ddev_start;
 	u64			ddev_end;
+	int			error;
 
 	if (!(mp->m_super->s_flags & SB_BORN)) {
 		xfs_warn(mp, "filesystem is not ready for notify_failure()!");
@@ -196,6 +321,8 @@  xfs_dax_notify_failure(
 
 	if (mp->m_logdev_targp && mp->m_logdev_targp->bt_daxdev == dax_dev &&
 	    mp->m_logdev_targp != mp->m_ddev_targp) {
+		if (mf_flags & MF_MEM_PRE_REMOVE)
+			return 0;
 		xfs_err(mp, "ondisk log corrupt, shutting down fs!");
 		xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_ONDISK);
 		return -EFSCORRUPTED;
@@ -209,6 +336,12 @@  xfs_dax_notify_failure(
 	ddev_start = mp->m_ddev_targp->bt_dax_part_off;
 	ddev_end = ddev_start + bdev_nr_bytes(mp->m_ddev_targp->bt_bdev) - 1;
 
+	/* Notify failure on the whole device. */
+	if (offset == 0 && len == U64_MAX) {
+		offset = ddev_start;
+		len = bdev_nr_bytes(mp->m_ddev_targp->bt_bdev);
+	}
+
 	/* Ignore the range out of filesystem area */
 	if (offset + len - 1 < ddev_start)
 		return -ENXIO;
@@ -225,6 +358,14 @@  xfs_dax_notify_failure(
 	if (offset + len - 1 > ddev_end)
 		len = ddev_end - offset + 1;
 
+	if (mf_flags & MF_MEM_PRE_REMOVE) {
+		xfs_info(mp, "device is about to be removed!");
+		/* Freeze fs to prevent new mappings from being created. */
+		error = xfs_dax_notify_failure_freeze(mp);
+		if (error)
+			return error;
+	}
+
 	return xfs_dax_notify_ddev_failure(mp, BTOBB(offset), BTOBB(len),
 			mf_flags);
 }
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 1f79667824eb..ac3f22c20e1d 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -3436,6 +3436,7 @@  enum mf_flags {
 	MF_UNPOISON = 1 << 4,
 	MF_SW_SIMULATED = 1 << 5,
 	MF_NO_RETRY = 1 << 6,
+	MF_MEM_PRE_REMOVE = 1 << 7,
 };
 int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
 		      unsigned long count, int mf_flags);
diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index fae9baf3be16..6e6acec45568 100644
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -623,7 +623,7 @@  static void collect_procs_file(struct page *page, struct list_head *to_kill,
  */
 static void collect_procs_fsdax(struct page *page,
 		struct address_space *mapping, pgoff_t pgoff,
-		struct list_head *to_kill)
+		struct list_head *to_kill, bool pre_remove)
 {
 	struct vm_area_struct *vma;
 	struct task_struct *tsk;
@@ -631,8 +631,15 @@  static void collect_procs_fsdax(struct page *page,
 	i_mmap_lock_read(mapping);
 	read_lock(&tasklist_lock);
 	for_each_process(tsk) {
-		struct task_struct *t = task_early_kill(tsk, true);
+		struct task_struct *t = tsk;
 
+		/*
+		 * Search for all tasks while MF_MEM_PRE_REMOVE, because the
+		 * current may not be the one accessing the fsdax page.
+		 * Otherwise, search for the current task.
+		 */
+		if (!pre_remove)
+			t = task_early_kill(tsk, true);
 		if (!t)
 			continue;
 		vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
@@ -1732,6 +1739,7 @@  int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
 	dax_entry_t cookie;
 	struct page *page;
 	size_t end = index + count;
+	bool pre_remove = mf_flags & MF_MEM_PRE_REMOVE;
 
 	mf_flags |= MF_ACTION_REQUIRED | MF_MUST_KILL;
 
@@ -1743,9 +1751,10 @@  int mf_dax_kill_procs(struct address_space *mapping, pgoff_t index,
 		if (!page)
 			goto unlock;
 
-		SetPageHWPoison(page);
+		if (!pre_remove)
+			SetPageHWPoison(page);
 
-		collect_procs_fsdax(page, mapping, index, &to_kill);
+		collect_procs_fsdax(page, mapping, index, &to_kill, pre_remove);
 		unmap_and_kill(&to_kill, page_to_pfn(page), mapping,
 				index, mf_flags);
 unlock: