Skip to contentSkip to conversation

libxfs: check all defer ops types

PATCH In review Darrick J. Wong 3 participants 7 messages Oct 27, 2018

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 libxfs: check all defer ops types

From: Darrick J. Wong <[email protected]>

Make sure we have all the defer ops types loaded before proceeding.

---
 libxfs/init.c      |    5 +++++
 libxfs/xfs_defer.c |   12 ++++++++++++
 libxfs/xfs_defer.h |    1 +
 3 files changed, 18 insertions(+)
libxfs/init.c +5 -0
index 750b8f20..d1fdc3a8 100644
@@ -253,6 +253,11 @@ libxfs_init(libxfs_init_t *a)
 	xfs_rmap_update_init_defer_op();
 	xfs_refcount_update_init_defer_op();
 	xfs_bmap_update_init_defer_op();
+	if (!xfs_defer_check_all_present()) {
+		fprintf(stderr, _("%s: defer ops not all there\n"),
+			progname);
+		goto done;
+	}
 
 	radix_tree_init();
 
libxfs/xfs_defer.c +12 -0
index 5a6da0f3..cd3a8ef3 100644
@@ -545,3 +545,15 @@ xfs_defer_move(
 
 	xfs_defer_reset(stp);
 }
+
+/* Make sure we filled out all the defer_ops types. */
+bool
+xfs_defer_check_all_present(void)
+{
+	int		x;
+
+	for(x = 0; x < XFS_DEFER_OPS_TYPE_MAX; x++)
+		if (defer_op_types[x] == NULL)
+			return false;
+	return true;
+}
libxfs/xfs_defer.h +1 -0
index 2584a5b9..bc7ae7ff 100644
@@ -57,5 +57,6 @@ struct xfs_defer_op_type {
 };
 
 void xfs_defer_init_op_type(const struct xfs_defer_op_type *type);
+bool xfs_defer_check_all_present(void);
 
 #endif /* __XFS_DEFER_H__ */
#1 lore raw
Quoted 3 lines · Darrick J. Wong

From: Darrick J. Wong <[email protected]>

Make sure we have all the defer ops types loaded before proceeding.

So, the only reason we have this problem is that userspace doesn't exactly share the code which handles all the defer ops.

I think this really needs to be a BUILD_BUG_ON if anything, by the time we get to runtime it's really too late, and a weird thing to run at every invocation.

The only thing I can think of is to make an array containing each of the xfs_defer_op_types, and iterate over it in some sort of xfs_defer_ops_init(), then we can compare the size of that array to XFS_DEFER_OPS_TYPE_MAX with a BUILD_BUG_ON. Worth it?

-Eric

Quoted 56 lines
---
 libxfs/init.c      |    5 +++++
 libxfs/xfs_defer.c |   12 ++++++++++++
 libxfs/xfs_defer.h |    1 +
 3 files changed, 18 insertions(+)
libxfs/init.c +5 -0
index 750b8f20..d1fdc3a8 100644
@@ -253,6 +253,11 @@ libxfs_init(libxfs_init_t *a)
 	xfs_rmap_update_init_defer_op();
 	xfs_refcount_update_init_defer_op();
 	xfs_bmap_update_init_defer_op();
+	if (!xfs_defer_check_all_present()) {
+		fprintf(stderr, _("%s: defer ops not all there\n"),
+			progname);
+		goto done;
+	}
 
 	radix_tree_init();
 
libxfs/xfs_defer.c +12 -0
index 5a6da0f3..cd3a8ef3 100644
@@ -545,3 +545,15 @@ xfs_defer_move(
 
 	xfs_defer_reset(stp);
 }
+
+/* Make sure we filled out all the defer_ops types. */
+bool
+xfs_defer_check_all_present(void)
+{
+	int		x;
+
+	for(x = 0; x < XFS_DEFER_OPS_TYPE_MAX; x++)
+		if (defer_op_types[x] == NULL)
+			return false;
+	return true;
+}
libxfs/xfs_defer.h +1 -0
index 2584a5b9..bc7ae7ff 100644
@@ -57,5 +57,6 @@ struct xfs_defer_op_type {
 };
 
 void xfs_defer_init_op_type(const struct xfs_defer_op_type *type);
+bool xfs_defer_check_all_present(void);
 
 #endif /* __XFS_DEFER_H__ */
#2 lore raw
Quoted 8 lines · Eric Sandeen
Quoted 3 lines · Darrick J. Wong

From: Darrick J. Wong <[email protected]>

Make sure we have all the defer ops types loaded before proceeding.

So, the only reason we have this problem is that userspace doesn't exactly share the code which handles all the defer ops.

Quoted 8 lines

I think this really needs to be a BUILD_BUG_ON if anything, by the time we get to runtime it's really too late, and a weird thing to run at every invocation.

The only thing I can think of is to make an array containing each of the xfs_defer_op_types, and iterate over it in some sort of xfs_defer_ops_init(), then we can compare the size of that array to XFS_DEFER_OPS_TYPE_MAX with a BUILD_BUG_ON. Worth it?

Uh, or just make a statically initialized array?
I guess nothing about this directly enforces that the ->type number is in the right slot of the array.

static const struct
xfs_defer_op_type *defer_op_types[XFS_DEFER_OPS_TYPE_MAX] = {
xfs_bmap_update_defer_type, xfs_refcount_update_defer_type, xfs_rmap_update_defer_type, xfs_extent_free_defer_type, xfs_agfl_free_defer_type,
};

#3 lore raw
Quoted 24 lines · Eric Sandeen
Quoted 8 lines · Eric Sandeen
Quoted 3 lines · Darrick J. Wong

From: Darrick J. Wong <[email protected]>

Make sure we have all the defer ops types loaded before proceeding.

So, the only reason we have this problem is that userspace doesn't exactly share the code which handles all the defer ops.

Quoted 8 lines

I think this really needs to be a BUILD_BUG_ON if anything, by the time we get to runtime it's really too late, and a weird thing to run at every invocation.

The only thing I can think of is to make an array containing each of the xfs_defer_op_types, and iterate over it in some sort of xfs_defer_ops_init(), then we can compare the size of that array to XFS_DEFER_OPS_TYPE_MAX with a BUILD_BUG_ON. Worth it?

Uh, or just make a statically initialized array?
I guess nothing about this directly enforces that the ->type number is in the right slot of the array.

Quoted 8 lines

static const struct
xfs_defer_op_type *defer_op_types[XFS_DEFER_OPS_TYPE_MAX] = {
xfs_bmap_update_defer_type, xfs_refcount_update_defer_type, xfs_rmap_update_defer_type, xfs_extent_free_defer_type, xfs_agfl_free_defer_type,
};

I was thinking more:

static const struct
xfs_defer_op_type defer_op_types[XFS_DEFER_OPS_TYPE_MAX] = {
[XFS_DEFER_OPS_TYPE_AGFL_FREE] = {
.type = XFS_DEFER_OPS_TYPE_AGFL_FREE,
.diff_items = xfs_extent_free_diff_items,
.create_intent = xfs_extent_free_create_intent,
.abort_intent = xfs_extent_free_abort_intent,
.log_item = xfs_extent_free_log_item,
.create_done = xfs_extent_free_create_done,
.finish_item = xfs_agfl_free_finish_item,
.cancel_item = xfs_extent_free_cancel_item,
},
...
};

The downside of this is that now we have to declare all those functions somewhere so that the fs/xfs/ and libxfs/defer_item.c can match the prototypes. But that's probably better than what we have now, which is a big gaping hole in xfsprogs 4.18.

--D

Quoted 4 lines
#4 lore raw
Quoted 3 lines · Darrick J. Wong

The downside of this is that now we have to declare all those functions somewhere so that the fs/xfs/ and libxfs/defer_item.c can match the prototypes.

I don't think that is worth the pain vs the version I posted. I don't think removing a single indirection really buys us so much as to expose all the internals.

#5 lore raw
In reply to Eric Sandeen
Quoted 8 lines · Eric Sandeen

static const struct
xfs_defer_op_type *defer_op_types[XFS_DEFER_OPS_TYPE_MAX] = {
xfs_bmap_update_defer_type, xfs_refcount_update_defer_type, xfs_rmap_update_defer_type, xfs_extent_free_defer_type, xfs_agfl_free_defer_type,
};

I think this should be:

static const struct xfs_defer_op_type *defer_op_types[] = {
[XFS_DEFER_OPS_TYPE_BMAP] = &xfs_bmap_update_defer_type,
[XFS_DEFER_OPS_TYPE_REFCOUNT] = &xfs_refcount_update_defer_type,
[XFS_DEFER_OPS_TYPE_RMAP] = &xfs_rmap_update_defer_type,
[XFS_DEFER_OPS_TYPE_FREE] = &xfs_extent_free_defer_type,
[XFS_DEFER_OPS_TYPE_AGFL_FREE] = &xfs_agfl_free_defer_type,
};

somefunc()

{
	BUILD_BUG_ON(ARRAY_SIZE(defer_op_types) != XFS_DEFER_OPS_TYPE_MAX);
}
#6 lore raw
Quoted 20 lines · Christoph Hellwig
Quoted 8 lines · Eric Sandeen

static const struct
xfs_defer_op_type *defer_op_types[XFS_DEFER_OPS_TYPE_MAX] = {
xfs_bmap_update_defer_type, xfs_refcount_update_defer_type, xfs_rmap_update_defer_type, xfs_extent_free_defer_type, xfs_agfl_free_defer_type,
};

I think this should be:

static const struct xfs_defer_op_type *defer_op_types[] = {
[XFS_DEFER_OPS_TYPE_BMAP] = &xfs_bmap_update_defer_type,
[XFS_DEFER_OPS_TYPE_REFCOUNT] = &xfs_refcount_update_defer_type,
[XFS_DEFER_OPS_TYPE_RMAP] = &xfs_rmap_update_defer_type,
[XFS_DEFER_OPS_TYPE_FREE] = &xfs_extent_free_defer_type,
[XFS_DEFER_OPS_TYPE_AGFL_FREE] = &xfs_agfl_free_defer_type,
};

Quoted 5 lines

somefunc()

{
	BUILD_BUG_ON(ARRAY_SIZE(defer_op_types) != XFS_DEFER_OPS_TYPE_MAX);
}

Uh, yeah, that's what I was trying but pitifully failing to come up with, thanks. :)

-Eric