uprobes: pre-filtering

consumer_rwsem, but I don't think it makes sense to optimize this and

filter_chain() was added into install_breakpoint/remove_breakpoint to simplify the initial changes but this is sub-optimal.

This patch shifts the callsite to the callers, register_for_each_vma() and uprobe_mmap(). This way:

- It will be easier to add the new arguments. This is the main reason, we can do more optimizations later.

- register_for_each_vma(is_register => true) can be optimized, we only need to consult the new consumer. The previous consumers were already asked when they called uprobe_register().

This patch also moves the MMF_HAS_UPROBES check from remove_breakpoint(), this allows to avoid the potentionally costly filter_chain(). Note that register_for_each_vma(is_register => false) doesn't really need to take

Quoted 1 line

consumer_rwsem, but I don't think it makes sense to optimize this and

introduce filter_chain_lockless().

Signed-off-by Oleg Nesterov <[email protected]>

---
 kernel/events/uprobes.c |   44 +++++++++++++++++++++-----------------------
 1 files changed, 21 insertions(+), 23 deletions(-)

kernel/events/uprobes.c +21 -23

index 105ac0d..60b0a90 100644
@@ -579,6 +579,11 @@ static int prepare_uprobe(struct uprobe *uprobe, struct file *file,
 	return ret;
 }
 
+static inline bool consumer_filter(struct uprobe_consumer *uc)
+{
+	return true; /* TODO: !uc->filter || uc->filter(...) */
+}
+
 static bool filter_chain(struct uprobe *uprobe)
 {
 	struct uprobe_consumer *uc;

@@ -586,8 +591,7 @@ static bool filter_chain(struct uprobe *uprobe)
 
 	down_read(&uprobe->consumer_rwsem);
 	for (uc = uprobe->consumers; uc; uc = uc->next) {
-		/* TODO: ret = uc->filter(...) */
-		ret = true;
+		ret = consumer_filter(uc);
 		if (ret)
 			break;
 	}

@@ -603,15 +607,6 @@ install_breakpoint(struct uprobe *uprobe, struct mm_struct *mm,
 	bool first_uprobe;
 	int ret;
 
-	/*
-	 * If probe is being deleted, unregister thread could be done with
-	 * the vma-rmap-walk through. Adding a probe now can be fatal since
-	 * nobody will be able to cleanup. But in this case filter_chain()
-	 * must return false, all consumers have gone away.
-	 */
-	if (!filter_chain(uprobe))
-		return 0;
-
 	ret = prepare_uprobe(uprobe, vma->vm_file, mm, vaddr);
 	if (ret)
 		return ret;

@@ -636,12 +631,6 @@ install_breakpoint(struct uprobe *uprobe, struct mm_struct *mm,
 static int
 remove_breakpoint(struct uprobe *uprobe, struct mm_struct *mm, unsigned long vaddr)
 {
-	if (!test_bit(MMF_HAS_UPROBES, &mm->flags))
-		return 0;
-
-	if (filter_chain(uprobe))
-		return 0;
-
 	set_bit(MMF_RECALC_UPROBES, &mm->flags);
 	return set_orig_insn(&uprobe->arch, mm, vaddr);
 }

@@ -781,10 +770,14 @@ static int register_for_each_vma(struct uprobe *uprobe, bool is_register)
 		    vaddr_to_offset(vma, info->vaddr) != uprobe->offset)
 			goto unlock;
 
-		if (is_register)
-			err = install_breakpoint(uprobe, mm, vma, info->vaddr);
-		else
-			err |= remove_breakpoint(uprobe, mm, info->vaddr);
+		if (is_register) {
+			/* consult only the "caller", new consumer. */
+			if (consumer_filter(uprobe->consumers))
+				err = install_breakpoint(uprobe, mm, vma, info->vaddr);
+		} else if (test_bit(MMF_HAS_UPROBES, &mm->flags)) {
+			if (!filter_chain(uprobe))
+				err |= remove_breakpoint(uprobe, mm, info->vaddr);
+		}
 
  unlock:
 		up_write(&mm->mmap_sem);

@@ -968,9 +961,14 @@ int uprobe_mmap(struct vm_area_struct *vma)
 
 	mutex_lock(uprobes_mmap_hash(inode));
 	build_probe_list(inode, vma, vma->vm_start, vma->vm_end, &tmp_list);
-
+	/*
+	 * We can race with uprobe_unregister(), this uprobe can be already
+	 * removed. But in this case filter_chain() must return false, all
+	 * consumers have gone away.
+	 */
 	list_for_each_entry_safe(uprobe, u, &tmp_list, pending_list) {
-		if (!fatal_signal_pending(current)) {
+		if (!fatal_signal_pending(current) &&
+		    filter_chain(uprobe)) {
 			unsigned long vaddr = offset_to_vaddr(vma, uprobe->offset);
 			install_breakpoint(uprobe, vma->vm_mm, vma, vaddr);
 		}

Signature · 1 line

1.5.5.1

Finally add uprobe_consumer->filter() and change consumer_filter() to actually call this method.

Note that ->filter() accepts mm_struct, not task_struct. Because:

1. We do not have for_each_mm_user(mm, task).

2. Even if we implement for_each_mm_user(), ->filter() can use it itself.

3. It is not clear who will actually need this interface to do the "nontrivial" filtering.

Another argument is "enum uprobe_filter_ctx", consumer->filter() can use it to figure out why/where it was called. For example, perhaps we can add UPROBE_FILTER_PRE_REGISTER used by build_map_info() to quickly "nack" the unwanted mm's. In this case consumer should know that it is called under ->i_mmap_mutex.

See the previous discussion at http://marc.info/?t=135214229700002
Perhaps we should pass more arguments, vma/vaddr?

Note: this patch obviously can't help to filter out the child created by fork(), this will be addressed later.

Signed-off-by Oleg Nesterov <[email protected]>

---
 include/linux/uprobes.h |    9 +++++++++
 kernel/events/uprobes.c |   18 +++++++++++-------
 2 files changed, 20 insertions(+), 7 deletions(-)

include/linux/uprobes.h +9 -0

index 83742b9..c2df693 100644
@@ -35,8 +35,17 @@ struct inode;
 # include <asm/uprobes.h>
 #endif
 
+enum uprobe_filter_ctx {
+	UPROBE_FILTER_REGISTER,
+	UPROBE_FILTER_UNREGISTER,
+	UPROBE_FILTER_MMAP,
+};
+
 struct uprobe_consumer {
 	int (*handler)(struct uprobe_consumer *self, struct pt_regs *regs);
+	bool (*filter)(struct uprobe_consumer *self,
+				enum uprobe_filter_ctx ctx,
+				struct mm_struct *mm);
 
 	struct uprobe_consumer *next;
 };

kernel/events/uprobes.c +11 -7

index 60b0a90..e2ebb6f 100644
@@ -579,19 +579,21 @@ static int prepare_uprobe(struct uprobe *uprobe, struct file *file,
 	return ret;
 }
 
-static inline bool consumer_filter(struct uprobe_consumer *uc)
+static inline bool consumer_filter(struct uprobe_consumer *uc,
+				   enum uprobe_filter_ctx ctx, struct mm_struct *mm)
 {
-	return true; /* TODO: !uc->filter || uc->filter(...) */
+	return !uc->filter || uc->filter(uc, ctx, mm);
 }
 
-static bool filter_chain(struct uprobe *uprobe)
+static bool filter_chain(struct uprobe *uprobe,
+			 enum uprobe_filter_ctx ctx, struct mm_struct *mm)
 {
 	struct uprobe_consumer *uc;
 	bool ret = false;
 
 	down_read(&uprobe->consumer_rwsem);
 	for (uc = uprobe->consumers; uc; uc = uc->next) {
-		ret = consumer_filter(uc);
+		ret = consumer_filter(uc, ctx, mm);
 		if (ret)
 			break;
 	}

@@ -772,10 +774,12 @@ static int register_for_each_vma(struct uprobe *uprobe, bool is_register)
 
 		if (is_register) {
 			/* consult only the "caller", new consumer. */
-			if (consumer_filter(uprobe->consumers))
+			if (consumer_filter(uprobe->consumers,
+					UPROBE_FILTER_REGISTER, mm))
 				err = install_breakpoint(uprobe, mm, vma, info->vaddr);
 		} else if (test_bit(MMF_HAS_UPROBES, &mm->flags)) {
-			if (!filter_chain(uprobe))
+			if (!filter_chain(uprobe,
+					UPROBE_FILTER_UNREGISTER, mm))
 				err |= remove_breakpoint(uprobe, mm, info->vaddr);
 		}
 

@@ -968,7 +972,7 @@ int uprobe_mmap(struct vm_area_struct *vma)
 	 */
 	list_for_each_entry_safe(uprobe, u, &tmp_list, pending_list) {
 		if (!fatal_signal_pending(current) &&
-		    filter_chain(uprobe)) {
+		    filter_chain(uprobe, UPROBE_FILTER_MMAP, vma->vm_mm)) {
 			unsigned long vaddr = offset_to_vaddr(vma, uprobe->offset);
 			install_breakpoint(uprobe, vma->vm_mm, vma, vaddr);
 		}

Signature · 1 line

1.5.5.1