Skip to contentSkip to conversation

mm/vma: do not try to unmap a VMA if mmap_prepare() invoked from mmap()

PATCH Applied Lorenzo Stoakes 2 participants 3 messages Apr 21

Subsystems (3)
lore
Share & tools
lore

Lorenzo Stoakes

author maintainer
#0 lore raw

PATCH 1/1 mm/vma: do not try to unmap a VMA if mmap_prepare() invoked from mmap()

The mmap_prepare hook functionality includes the ability to invoke mmap_prepare() from the mmap() hook of existing 'stacked' drivers, that is ones which are capable of calling the mmap hooks of other drivers/file systems (e.g. overlayfs, shm).

As part of the mmap_prepare action functionality, we deal with errors by unmapping the VMA should one arise. This works in the usual mmap_prepare case, as we invoke this action at the last moment, when the VMA is established in the maple tree.

However, the mmap() hook passes a not-fully-established VMA pointer to the caller (which is the motivation behind the mmap_prepare() work), which is detached.

So attempting to unmap a VMA in this state will be problematic, with the most obvious symptom being a warning in vma_mark_detached(), because the VMA is already detached.

It's also unncessary - the mmap() handler will clean up the VMA on error.

So to fix this issue, this patch propagates whether or not an mmap action is being completed via the compatibility layer or directly.

If the former, then we do not attempt VMA cleanup, if the latter, then we do.

This patch also updates the userland VMA tests to reflect the change.

Fixes ac0a3fc9c07d ("mm: add ability to take further action in vm_area_desc")
Reported-by [email protected]
---
 include/linux/mm.h                |  2 +-
 mm/util.c                         | 26 +++++++++++++++++---------
 mm/vma.c                          |  3 ++-
 tools/testing/vma/include/dup.h   |  2 +-
 tools/testing/vma/include/stubs.h |  3 ++-
 5 files changed, 23 insertions(+), 13 deletions(-)
include/linux/mm.h +1 -1
index 0b776907152e..af23453e9dbd 100644
@@ -4391,7 +4391,7 @@ static inline void mmap_action_map_kernel_pages_full(struct vm_area_desc *desc,
 
 int mmap_action_prepare(struct vm_area_desc *desc);
 int mmap_action_complete(struct vm_area_struct *vma,
-			 struct mmap_action *action);
+			 struct mmap_action *action, bool is_compat);
 
 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
mm/util.c +17 -9
index 232c3930a662..3cc949a0b7ed 100644
@@ -1232,7 +1232,7 @@ int __compat_vma_mmap(struct vm_area_desc *desc,
 	/* Update the VMA from the descriptor. */
 	compat_set_vma_from_desc(vma, desc);
 	/* Complete any specified mmap actions. */
-	return mmap_action_complete(vma, &desc->action);
+	return mmap_action_complete(vma, &desc->action, /*is_compat=*/true);
 }
 EXPORT_SYMBOL(__compat_vma_mmap);
 
@@ -1389,7 +1389,8 @@ static int call_vma_mapped(struct vm_area_struct *vma)
 }
 
 static int mmap_action_finish(struct vm_area_struct *vma,
-			      struct mmap_action *action, int err)
+			      struct mmap_action *action, int err,
+			      bool is_compat)
 {
 	size_t len;
 
@@ -1400,8 +1401,12 @@ static int mmap_action_finish(struct vm_area_struct *vma,
 
 	/* do_munmap() might take rmap lock, so release if held. */
 	maybe_rmap_unlock_action(vma, action);
-	if (!err)
-		return 0;
+	/*
+	 * If this is invoked from the compatibility layer, post-mmap() hook
+	 * logic will handle cleanup for us.
+	 */
+	if (!err || is_compat)
+		return err;
 
 	/*
 	 * If an error occurs, unmap the VMA altogether and return an error. We
@@ -1451,13 +1456,15 @@ EXPORT_SYMBOL(mmap_action_prepare);
  * mmap_action_complete - Execute VMA descriptor action.
  * @vma: The VMA to perform the action upon.
  * @action: The action to perform.
+ * @is_compat: Is this being invoked from the compatibility layer?
  *
  * Similar to mmap_action_prepare().
  *
- * Return: 0 on success, or error, at which point the VMA will be unmapped.
+ * Return: 0 on success, or error, at which point the VMA will be unmapped if
+ * !@is_compat.
  */
 int mmap_action_complete(struct vm_area_struct *vma,
-			 struct mmap_action *action)
+			 struct mmap_action *action, bool is_compat)
 {
 	int err = 0;
 
@@ -1478,7 +1485,7 @@ int mmap_action_complete(struct vm_area_struct *vma,
 		break;
 	}
 
-	return mmap_action_finish(vma, action, err);
+	return mmap_action_finish(vma, action, err, is_compat);
 }
 EXPORT_SYMBOL(mmap_action_complete);
 #else
@@ -1500,7 +1507,8 @@ int mmap_action_prepare(struct vm_area_desc *desc)
 EXPORT_SYMBOL(mmap_action_prepare);
 
 int mmap_action_complete(struct vm_area_struct *vma,
-			 struct mmap_action *action)
+			 struct mmap_action *action,
+			 bool is_compat)
 {
 	int err = 0;
 
@@ -1517,7 +1525,7 @@ int mmap_action_complete(struct vm_area_struct *vma,
 		break;
 	}
 
-	return mmap_action_finish(vma, action, err);
+	return mmap_action_finish(vma, action, err, is_compat);
 }
 EXPORT_SYMBOL(mmap_action_complete);
 #endif
mm/vma.c +2 -1
index 377321b48734..d90791b00a7b 100644
@@ -2780,7 +2780,8 @@ static unsigned long __mmap_region(struct file *file, unsigned long addr,
 	__mmap_complete(&map, vma);
 
 	if (have_mmap_prepare && allocated_new) {
-		error = mmap_action_complete(vma, &desc.action);
+		error = mmap_action_complete(vma, &desc.action,
+					     /*is_compat=*/false);
 		if (error)
 			return error;
 	}
tools/testing/vma/include/dup.h +1 -1
index b4864aad2db0..9e0dfd3a85b0 100644
@@ -1330,7 +1330,7 @@ static inline int __compat_vma_mmap(struct vm_area_desc *desc,
 	/* Update the VMA from the descriptor. */
 	compat_set_vma_from_desc(vma, desc);
 	/* Complete any specified mmap actions. */
-	return mmap_action_complete(vma, &desc->action);
+	return mmap_action_complete(vma, &desc->action, /*is_compat=*/true);
 }
 
 static inline int compat_vma_mmap(struct file *file, struct vm_area_struct *vma)
tools/testing/vma/include/stubs.h +2 -1
index a30b8bc84955..64164e25658f 100644
@@ -87,7 +87,8 @@ static inline int mmap_action_prepare(struct vm_area_desc *desc)
 }
 
 static inline int mmap_action_complete(struct vm_area_struct *vma,
-				       struct mmap_action *action)
+				       struct mmap_action *action,
+				       bool is_compat)
 {
 	return 0;
 }

--
2.53.0

reviewer
#1 lore raw
Quoted 26 lines · Lorenzo Stoakes

The mmap_prepare hook functionality includes the ability to invoke mmap_prepare() from the mmap() hook of existing 'stacked' drivers, that is ones which are capable of calling the mmap hooks of other drivers/file systems (e.g. overlayfs, shm).

As part of the mmap_prepare action functionality, we deal with errors by unmapping the VMA should one arise. This works in the usual mmap_prepare case, as we invoke this action at the last moment, when the VMA is established in the maple tree.

However, the mmap() hook passes a not-fully-established VMA pointer to the caller (which is the motivation behind the mmap_prepare() work), which is detached.

So attempting to unmap a VMA in this state will be problematic, with the most obvious symptom being a warning in vma_mark_detached(), because the VMA is already detached.

It's also unncessary - the mmap() handler will clean up the VMA on error.

So to fix this issue, this patch propagates whether or not an mmap action is being completed via the compatibility layer or directly.

If the former, then we do not attempt VMA cleanup, if the latter, then we do.

Quoted 7 lines

This patch also updates the userland VMA tests to reflect the change.

Fixes ac0a3fc9c07d ("mm: add ability to take further action in vm_area_desc")
Reported-by [email protected]

How about something like the following:

include/linux/mm_types.h +6 -0
index a308e2c23b82..c29165de6d5c 100644
@@ -868,6 +868,12 @@ struct mmap_action {
         * completely set up.
         */
        bool hide_from_rmap_until_complete :1;
+
+       /*
+        * Set if this mmap_action is part of compatibility with ->mmap().
+        * Internal flag.
+        */
+       bool compat_mmap :1;
 };
 
 /*
mm/util.c +6 -1
index 232c3930a662..5134f879566d 100644
@@ -1229,6 +1229,7 @@ int __compat_vma_mmap(struct vm_area_desc *desc,
        err = mmap_action_prepare(desc);
        if (err)
                return err;
+       desc->action.compat_mmap = 1;
        /* Update the VMA from the descriptor. */
        compat_set_vma_from_desc(vma, desc);
        /* Complete any specified mmap actions. */
@@ -1400,7 +1401,11 @@ static int mmap_action_finish(struct vm_area_struct *vma,
 
        /* do_munmap() might take rmap lock, so release if held. */
        maybe_rmap_unlock_action(vma, action);
-       if (!err)
+       /*
+        * If this is invoked from the compatibility layer, post-mmap() hook
+        * logic will handle cleanup for us.
+        */
+       if (!err || action->compat_mmap)
                return 0;
 
        /*

We have plenty of free bits in mmap_action and this is a little nicer than passing is_compat bools down the callchain.

(that comment over compat_mmap is really... vague and bad, but I couldn't think of something else)

Signature · 1 line
Pedro
author maintainer
#2 lore raw
Quoted 29 lines · Pedro Falcato
Quoted 28 lines · Lorenzo Stoakes

The mmap_prepare hook functionality includes the ability to invoke mmap_prepare() from the mmap() hook of existing 'stacked' drivers, that is ones which are capable of calling the mmap hooks of other drivers/file systems (e.g. overlayfs, shm).

As part of the mmap_prepare action functionality, we deal with errors by unmapping the VMA should one arise. This works in the usual mmap_prepare case, as we invoke this action at the last moment, when the VMA is established in the maple tree.

However, the mmap() hook passes a not-fully-established VMA pointer to the caller (which is the motivation behind the mmap_prepare() work), which is detached.

So attempting to unmap a VMA in this state will be problematic, with the most obvious symptom being a warning in vma_mark_detached(), because the VMA is already detached.

It's also unncessary - the mmap() handler will clean up the VMA on error.

So to fix this issue, this patch propagates whether or not an mmap action is being completed via the compatibility layer or directly.

If the former, then we do not attempt VMA cleanup, if the latter, then we do.

This patch also updates the userland VMA tests to reflect the change.

Quoted 7 lines
Quoted 5 lines
Fixes ac0a3fc9c07d ("mm: add ability to take further action in vm_area_desc")
Reported-by [email protected]

How about something like the following:

(Normally I'd love helper struct stuf but...)

I was going to say in the commit message as to why I'm not doing that :) but thought maybe I shouldn't spell it out.

But to spell it out - I don't want to do that, because it could be abused by drivers and I don't want to add any possibility of doing that, as it defeats the purpose of the change.

And adding checks to make sure a driver didn't mess with it complicates everything.

I'd rather live with the added param, it's temporary and can be removed once the mmap() hook is removed...

Quoted 45 lines
include/linux/mm_types.h +6 -0
index a308e2c23b82..c29165de6d5c 100644
@@ -868,6 +868,12 @@ struct mmap_action {
         * completely set up.
         */
        bool hide_from_rmap_until_complete :1;
+
+       /*
+        * Set if this mmap_action is part of compatibility with ->mmap().
+        * Internal flag.
+        */
+       bool compat_mmap :1;
 };
 
 /*
mm/util.c +6 -1
index 232c3930a662..5134f879566d 100644
@@ -1229,6 +1229,7 @@ int __compat_vma_mmap(struct vm_area_desc *desc,
        err = mmap_action_prepare(desc);
        if (err)
                return err;
+       desc->action.compat_mmap = 1;
        /* Update the VMA from the descriptor. */
        compat_set_vma_from_desc(vma, desc);
        /* Complete any specified mmap actions. */
@@ -1400,7 +1401,11 @@ static int mmap_action_finish(struct vm_area_struct *vma,
 
        /* do_munmap() might take rmap lock, so release if held. */
        maybe_rmap_unlock_action(vma, action);
-       if (!err)
+       /*
+        * If this is invoked from the compatibility layer, post-mmap() hook
+        * logic will handle cleanup for us.
+        */
+       if (!err || action->compat_mmap)
                return 0;
 
        /*
Quoted 8 lines

We have plenty of free bits in mmap_action and this is a little nicer than passing is_compat bools down the callchain.

(that comment over compat_mmap is really... vague and bad, but I couldn't think of something else)

--
Pedro

Cheers, Lorenzo