[PATCH v6 3.2-rc1 18/28] uprobes: slot allocation.

[Srikar Dronamraju]

One page of slots are allocated per mm.
On a probehit one free slot is acquired and released after
singlestep operation completes.

Signed-off-by: Jim Keniston <jkenisto@xxxxxxxxxx>
Signed-off-by: Srikar Dronamraju <srikar@xxxxxxxxxxxxxxxxxx>
---

Changelog (since v5)
- no more spin lock needed for slot allocation.
- use install_special_mapping to add a vma. (previous approach used
  init_creds)
- set uprobes_xol_area while holding map_sem exclusively.

 include/linux/mm_types.h |    2 
 include/linux/uprobes.h  |   24 +++++
 kernel/fork.c            |    2 
 kernel/uprobes.c         |  215 +++++++++++++++++++++++++++++++++++++++++++++-
 4 files changed, 240 insertions(+), 3 deletions(-)

diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 544a0b6..2595c9c 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -12,6 +12,7 @@
 #include <linux/completion.h>
 #include <linux/cpumask.h>
 #include <linux/page-debug-flags.h>
+#include <linux/uprobes.h>
 #include <asm/page.h>
 #include <asm/mmu.h>
 
@@ -391,6 +392,7 @@ struct mm_struct {
 #endif
 #ifdef CONFIG_UPROBES
 	atomic_t mm_uprobes_count;
+	struct uprobes_xol_area *uprobes_xol_area;
 #endif
 };
 
diff --git a/include/linux/uprobes.h b/include/linux/uprobes.h
index c1378a9..add5222 100644
--- a/include/linux/uprobes.h
+++ b/include/linux/uprobes.h
@@ -90,6 +90,26 @@ struct uprobe_task {
 	struct uprobe *active_uprobe;
 };
 
+/*
+ * On a breakpoint hit, thread contests for a slot.  It free the
+ * slot after singlestep.  Only definite number of slots are
+ * allocated.
+ */
+
+struct uprobes_xol_area {
+	wait_queue_head_t wq;	/* if all slots are busy */
+	atomic_t slot_count;	/* currently in use slots */
+	unsigned long *bitmap;	/* 0 = free slot */
+	struct page *page;
+
+	/*
+	 * We keep the vma's vm_start rather than a pointer to the vma
+	 * itself.  The probed process or a naughty kernel module could make
+	 * the vma go away, and we must handle that reasonably gracefully.
+	 */
+	unsigned long vaddr;		/* Page(s) of instruction slots */
+};
+
 #ifdef CONFIG_UPROBES
 extern int __weak set_bkpt(struct mm_struct *mm, struct uprobe *uprobe,
 							unsigned long vaddr);
@@ -101,6 +121,7 @@ extern int register_uprobe(struct inode *inode, loff_t offset,
 extern void unregister_uprobe(struct inode *inode, loff_t offset,
 				struct uprobe_consumer *consumer);
 extern void free_uprobe_utask(struct task_struct *tsk);
+extern void free_uprobes_xol_area(struct mm_struct *mm);
 extern int mmap_uprobe(struct vm_area_struct *vma);
 extern void munmap_uprobe(struct vm_area_struct *vma);
 extern unsigned long __weak get_uprobe_bkpt_addr(struct pt_regs *regs);
@@ -134,5 +155,8 @@ static inline unsigned long get_uprobe_bkpt_addr(struct pt_regs *regs)
 static inline void free_uprobe_utask(struct task_struct *tsk)
 {
 }
+static inline void free_uprobes_xol_area(struct mm_struct *mm)
+{
+}
 #endif /* CONFIG_UPROBES */
 #endif	/* _LINUX_UPROBES_H */
diff --git a/kernel/fork.c b/kernel/fork.c
index a03f436..c605f2a 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -558,6 +558,7 @@ void mmput(struct mm_struct *mm)
 	might_sleep();
 
 	if (atomic_dec_and_test(&mm->mm_users)) {
+		free_uprobes_xol_area(mm);
 		exit_aio(mm);
 		ksm_exit(mm);
 		khugepaged_exit(mm); /* must run before exit_mmap */
@@ -746,6 +747,7 @@ struct mm_struct *dup_mm(struct task_struct *tsk)
 #endif
 #ifdef CONFIG_UPROBES
 	atomic_set(&mm->mm_uprobes_count, 0);
+	mm->uprobes_xol_area = NULL;
 #endif
 
 	if (!mm_init(mm, tsk))
diff --git a/kernel/uprobes.c b/kernel/uprobes.c
index c4ccb89..9e73cef 100644
--- a/kernel/uprobes.c
+++ b/kernel/uprobes.c
@@ -33,6 +33,9 @@
 #include <linux/kdebug.h>	/* notifier mechanism */
 #include <linux/uprobes.h>
 
+#define UINSNS_PER_PAGE	(PAGE_SIZE/UPROBES_XOL_SLOT_BYTES)
+#define MAX_UPROBES_XOL_SLOTS UINSNS_PER_PAGE
+
 static bulkref_t uprobes_srcu;
 static struct rb_root uprobes_tree = RB_ROOT;
 static DEFINE_SPINLOCK(uprobes_treelock);	/* serialize rbtree access */
@@ -1054,6 +1057,201 @@ void munmap_uprobe(struct vm_area_struct *vma)
 	return;
 }
 
+/* Slot allocation for XOL */
+static int xol_add_vma(struct uprobes_xol_area *area)
+{
+	struct mm_struct *mm;
+	int ret;
+
+	area->page = alloc_page(GFP_HIGHUSER);
+	if (!area->page)
+		return -ENOMEM;
+
+	mm = current->mm;
+	down_write(&mm->mmap_sem);
+	ret = -EALREADY;
+	if (mm->uprobes_xol_area)
+		goto fail;
+
+	ret = -ENOMEM;
+
+	/* Try to map as high as possible, this is only a hint. */
+	area->vaddr = get_unmapped_area(NULL, TASK_SIZE - PAGE_SIZE,
+							PAGE_SIZE, 0, 0);
+	if (area->vaddr & ~PAGE_MASK) {
+		ret = area->vaddr;
+		goto fail;
+	}
+
+	ret = install_special_mapping(mm, area->vaddr, PAGE_SIZE,
+				VM_EXEC|VM_MAYEXEC|VM_DONTCOPY|VM_IO,
+				&area->page);
+	if (ret)
+		goto fail;
+
+	smp_wmb();	/* pairs with get_uprobes_xol_area() */
+	mm->uprobes_xol_area = area;
+	ret = 0;
+
+fail:
+	up_write(&mm->mmap_sem);
+	if (ret)
+		__free_page(area->page);
+
+	return ret;
+}
+
+static struct uprobes_xol_area *get_uprobes_xol_area(struct mm_struct *mm)
+{
+	struct uprobes_xol_area *area = mm->uprobes_xol_area;
+	smp_read_barrier_depends();/* pairs with wmb in xol_add_vma() */
+	return area;
+}
+
+/*
+ * xol_alloc_area - Allocate process's uprobes_xol_area.
+ * This area will be used for storing instructions for execution out of
+ * line.
+ *
+ * Returns the allocated area or NULL.
+ */
+static struct uprobes_xol_area *xol_alloc_area(void)
+{
+	struct uprobes_xol_area *area;
+
+	area = kzalloc(sizeof(*area), GFP_KERNEL);
+	if (unlikely(!area))
+		return NULL;
+
+	area->bitmap = kzalloc(BITS_TO_LONGS(UINSNS_PER_PAGE) * sizeof(long),
+								GFP_KERNEL);
+
+	if (!area->bitmap)
+		goto fail;
+
+	init_waitqueue_head(&area->wq);
+	if (!xol_add_vma(area))
+		return area;
+
+fail:
+	kfree(area->bitmap);
+	kfree(area);
+	return get_uprobes_xol_area(current->mm);
+}
+
+/*
+ * free_uprobes_xol_area - Free the area allocated for slots.
+ */
+void free_uprobes_xol_area(struct mm_struct *mm)
+{
+	struct uprobes_xol_area *area = mm->uprobes_xol_area;
+
+	if (!area)
+		return;
+
+	put_page(area->page);
+	kfree(area->bitmap);
+	kfree(area);
+}
+
+/*
+ *  - search for a free slot.
+ */
+static unsigned long xol_take_insn_slot(struct uprobes_xol_area *area)
+{
+	unsigned long slot_addr;
+	int slot_nr;
+
+	do {
+		slot_nr = find_first_zero_bit(area->bitmap, UINSNS_PER_PAGE);
+		if (slot_nr < UINSNS_PER_PAGE) {
+			if (!test_and_set_bit(slot_nr, area->bitmap))
+				break;
+
+			slot_nr = UINSNS_PER_PAGE;
+			continue;
+		}
+		wait_event(area->wq,
+			(atomic_read(&area->slot_count) < UINSNS_PER_PAGE));
+	} while (slot_nr >= UINSNS_PER_PAGE);
+
+	slot_addr = area->vaddr + (slot_nr * UPROBES_XOL_SLOT_BYTES);
+	atomic_inc(&area->slot_count);
+	return slot_addr;
+}
+
+/*
+ * xol_get_insn_slot - If was not allocated a slot, then
+ * allocate a slot.
+ * Returns the allocated slot address or 0.
+ */
+static unsigned long xol_get_insn_slot(struct uprobe *uprobe,
+					unsigned long slot_addr)
+{
+	struct uprobes_xol_area *area;
+	unsigned long offset;
+	void *vaddr;
+
+	area = get_uprobes_xol_area(current->mm);
+	if (!area) {
+		area = xol_alloc_area();
+		if (!area)
+			return 0;
+	}
+	current->utask->xol_vaddr = xol_take_insn_slot(area);
+
+	/*
+	 * Initialize the slot if xol_vaddr points to valid
+	 * instruction slot.
+	 */
+	if (unlikely(!current->utask->xol_vaddr))
+		return 0;
+
+	current->utask->vaddr = slot_addr;
+	offset = current->utask->xol_vaddr & ~PAGE_MASK;
+	vaddr = kmap_atomic(area->page);
+	memcpy(vaddr + offset, uprobe->insn, MAX_UINSN_BYTES);
+	kunmap_atomic(vaddr);
+	return current->utask->xol_vaddr;
+}
+
+/*
+ * xol_free_insn_slot - If slot was earlier allocated by
+ * @xol_get_insn_slot(), make the slot available for
+ * subsequent requests.
+ */
+static void xol_free_insn_slot(struct task_struct *tsk)
+{
+	struct uprobes_xol_area *area;
+	unsigned long vma_end;
+	unsigned long slot_addr;
+
+	if (!tsk->mm || !tsk->mm->uprobes_xol_area || !tsk->utask)
+		return;
+
+	slot_addr = tsk->utask->xol_vaddr;
+
+	if (unlikely(!slot_addr || IS_ERR_VALUE(slot_addr)))
+		return;
+
+	area = tsk->mm->uprobes_xol_area;
+	vma_end = area->vaddr + PAGE_SIZE;
+	if (area->vaddr <= slot_addr && slot_addr < vma_end) {
+		int slot_nr;
+		unsigned long offset = slot_addr - area->vaddr;
+
+		slot_nr = offset / UPROBES_XOL_SLOT_BYTES;
+		if (slot_nr >= UINSNS_PER_PAGE)
+			return;
+
+		clear_bit(slot_nr, area->bitmap);
+		atomic_dec(&area->slot_count);
+		if (waitqueue_active(&area->wq))
+			wake_up(&area->wq);
+		tsk->utask->xol_vaddr = 0;
+	}
+}
+
 /**
  * get_uprobe_bkpt_addr - compute address of bkpt given post-bkpt regs
  * @regs: Reflects the saved state of the task after it has hit a breakpoint
@@ -1082,6 +1280,7 @@ void free_uprobe_utask(struct task_struct *tsk)
 	if (utask->active_uprobe)
 		put_uprobe(utask->active_uprobe);
 
+	xol_free_insn_slot(tsk);
 	kfree(utask);
 	tsk->utask = NULL;
 }
@@ -1111,7 +1310,8 @@ static struct uprobe_task *add_utask(void)
 static int pre_ssout(struct uprobe *uprobe, struct pt_regs *regs,
 				unsigned long vaddr)
 {
-	/* TODO: Yet to be implemented */
+	if (xol_get_insn_slot(uprobe, vaddr) && !pre_xol(uprobe, regs))
+		return 0;
 	return -EFAULT;
 }
 
@@ -1121,8 +1321,16 @@ static int pre_ssout(struct uprobe *uprobe, struct pt_regs *regs,
  */
 static bool sstep_complete(struct uprobe *uprobe, struct pt_regs *regs)
 {
-	/* TODO: Yet to be implemented */
-	return false;
+	unsigned long vaddr = instruction_pointer(regs);
+
+	/*
+	 * If we have executed out of line, Instruction pointer
+	 * cannot be same as virtual address of XOL slot.
+	 */
+	if (vaddr == current->utask->xol_vaddr)
+		return false;
+	post_xol(uprobe, regs);
+	return true;
 }
 
 /*
@@ -1180,6 +1388,7 @@ void uprobe_notify_resume(struct pt_regs *regs)
 			utask->active_uprobe = NULL;
 			utask->state = UTASK_RUNNING;
 			user_disable_single_step(current);
+			xol_free_insn_slot(current);
 
 			/* TODO Stop queueing signals. */
 		}

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/