在 2021/10/15 2:00, Darrick J. Wong 写道:

On Fri, Sep 24, 2021 at 09:09:53PM +0800, Shiyang Ruan wrote:

quoted

To easily track filesystem from a pmem device, we introduce a holder for
dax_device structure, and also its operation.  This holder is used to
remember who is using this dax_device:
  - When it is the backend of a filesystem, the holder will be the
    superblock of this filesystem.
  - When this pmem device is one of the targets in a mapped device, the
    holder will be this mapped device.  In this case, the mapped device
    has its own dax_device and it will follow the first rule.  So that we
    can finally track to the filesystem we needed.

The holder and holder_ops will be set when filesystem is being mounted,
or an target device is being activated.

Signed-off-by: Shiyang Ruan <redacted>
---
  drivers/dax/super.c | 59 +++++++++++++++++++++++++++++++++++++++++++++
  include/linux/dax.h | 29 ++++++++++++++++++++++
  2 files changed, 88 insertions(+)

diff --git a/drivers/dax/super.c b/drivers/dax/super.c
index 48ce86501d93..7d4a11dcba90 100644
--- a/drivers/dax/super.c
+++ b/drivers/dax/super.c

@@ -23,7 +23,10 @@
   * @cdev: optional character interface for "device dax"
   * @host: optional name for lookups where the device path is not available
   * @private: dax driver private data
+ * @holder_data: holder of a dax_device: could be filesystem or mapped device
   * @flags: state and boolean properties
+ * @ops: operations for dax_device
+ * @holder_ops: operations for the inner holder
   */
  struct dax_device {
  	struct hlist_node list;

@@ -31,8 +34,10 @@ struct dax_device {
  	struct cdev cdev;
  	const char *host;
  	void *private;
+	void *holder_data;
  	unsigned long flags;
  	const struct dax_operations *ops;
+	const struct dax_holder_operations *holder_ops;
  };
  
  static dev_t dax_devt;

@@ -374,6 +379,29 @@ int dax_zero_page_range(struct dax_device *dax_dev, pgoff_t pgoff,
  }
  EXPORT_SYMBOL_GPL(dax_zero_page_range);
  
+int dax_holder_notify_failure(struct dax_device *dax_dev, loff_t offset,
+			      size_t size, int flags)
+{
+	int rc;
+
+	dax_read_lock();
+	if (!dax_alive(dax_dev)) {
+		rc = -ENXIO;
+		goto out;
+	}
+
+	if (!dax_dev->holder_data) {
+		rc = -EOPNOTSUPP;
+		goto out;
+	}
+
+	rc = dax_dev->holder_ops->notify_failure(dax_dev, offset, size, flags);

Shouldn't this check if dax_dev->holder_ops != NULL before dereferencing
it for the function call?  Imagine an implementation that wants to
attach a ->notify_failure function to a dax_device, maintains its own
lookup table, and decides that it doesn't need to set holder_data.

(Or, imagine someone who writes a garbage into holder_data and *boom*)

My mistake. I should check @holder_ops instead of @holder_data.

How does the locking work here?  If there's a media failure, we'll take
dax_rwsem and call ->notify_failure.  If the ->notify_failure function
wants to access the pmem to handle the error by calling back into the
dax code, will that cause nested locking on dax_rwsem?

Won't for now.  I have tested it with my simple testcases.

Jumping ahead a bit, I think the rmap btree accesses that the xfs
implementation performs can cause xfs_buf(fer) cache IO, which would
trigger that if the buffers aren't already in memory, if I'm reading
this correctly?

I didn't think of this case.  But I think this uses read lock too.  It 
won't be blocked.  Only dax_set_holder() takes write lock.

quoted

+out:
+	dax_read_unlock();
+	return rc;
+}
+EXPORT_SYMBOL_GPL(dax_holder_notify_failure);
+
  #ifdef CONFIG_ARCH_HAS_PMEM_API
  void arch_wb_cache_pmem(void *addr, size_t size);
  void dax_flush(struct dax_device *dax_dev, void *addr, size_t size)

@@ -618,6 +646,37 @@ void put_dax(struct dax_device *dax_dev)
  }
  EXPORT_SYMBOL_GPL(put_dax);
  
+void dax_set_holder(struct dax_device *dax_dev, void *holder,
+		const struct dax_holder_operations *ops)
+{
+	dax_write_lock();
+	if (!dax_alive(dax_dev)) {
+		dax_write_unlock();
+		return;
+	}
+
+	dax_dev->holder_data = holder;
+	dax_dev->holder_ops = ops;
+	dax_write_unlock();

I guess this means that the holder has to detach itself before anyone
calls kill_dax, or else a dead dax device ends up with a dangling
reference to the holder?

Yes.

quoted

+}
+EXPORT_SYMBOL_GPL(dax_set_holder);
+
+void *dax_get_holder(struct dax_device *dax_dev)
+{
+	void *holder;
+
+	dax_read_lock();
+	if (!dax_alive(dax_dev)) {
+		dax_read_unlock();
+		return NULL;
+	}
+
+	holder = dax_dev->holder_data;
+	dax_read_unlock();
+	return holder;
+}
+EXPORT_SYMBOL_GPL(dax_get_holder);
+
  /**
   * inode_dax: convert a public inode into its dax_dev
   * @inode: An inode with i_cdev pointing to a dax_dev

diff --git a/include/linux/dax.h b/include/linux/dax.h
index 097b3304f9b9..d273d59723cd 100644
--- a/include/linux/dax.h
+++ b/include/linux/dax.h

@@ -38,9 +38,24 @@ struct dax_operations {
  	int (*zero_page_range)(struct dax_device *, pgoff_t, size_t);
  };
  
+struct dax_holder_operations {
+	/*
+	 * notify_failure - notify memory failure into inner holder device
+	 * @dax_dev: the dax device which contains the holder
+	 * @offset: offset on this dax device where memory failure occurs
+	 * @size: length of this memory failure event
+	 * @flags: action flags for memory failure handler
+	 */
+	int (*notify_failure)(struct dax_device *dax_dev, loff_t offset,
+			size_t size, int flags);

Shouldn't size be u64 or something?  Let's say that 8GB of your pmem go
bad, wouldn't you want a single call?  Though I guess the current
implementation only goes a single page at a time, doesn't it?

Right.

quoted

+};
+
  extern struct attribute_group dax_attribute_group;
  
  #if IS_ENABLED(CONFIG_DAX)
+void dax_set_holder(struct dax_device *dax_dev, void *holder,
+		const struct dax_holder_operations *ops);
+void *dax_get_holder(struct dax_device *dax_dev);
  struct dax_device *alloc_dax(void *private, const char *host,
  		const struct dax_operations *ops, unsigned long flags);
  void put_dax(struct dax_device *dax_dev);

@@ -70,6 +85,18 @@ static inline bool daxdev_mapping_supported(struct vm_area_struct *vma,
  	return dax_synchronous(dax_dev);
  }
  #else
+static inline struct dax_device *dax_get_by_host(const char *host)

Not sure why this is being added here?  AFAICT none of the patches call
this function...?

It's mistake when I rebase my code to the latest.  These lines were 
deleted but I didn't notice.  Will fix it.


--
Thanks,
Ruan.

--D

quoted

+{
+	return NULL;
+}
+static inline void dax_set_holder(struct dax_device *dax_dev, void *holder,
+		const struct dax_holder_operations *ops)
+{
+}
+static inline void *dax_get_holder(struct dax_device *dax_dev)
+{
+	return NULL;
+}
  static inline struct dax_device *alloc_dax(void *private, const char *host,
  		const struct dax_operations *ops, unsigned long flags)
  {

@@ -198,6 +225,8 @@ size_t dax_copy_to_iter(struct dax_device *dax_dev, pgoff_t pgoff, void *addr,
  		size_t bytes, struct iov_iter *i);
  int dax_zero_page_range(struct dax_device *dax_dev, pgoff_t pgoff,
  			size_t nr_pages);
+int dax_holder_notify_failure(struct dax_device *dax_dev, loff_t offset,
+		size_t size, int flags);
  void dax_flush(struct dax_device *dax_dev, void *addr, size_t size);
  
  ssize_t dax_iomap_rw(struct kiocb *iocb, struct iov_iter *iter,

-- 
2.33.0