3 files changed, 40 insertions, 10 deletions

diff --git a/mm/filemap.c b/mm/filemap.c
index e989fb1..7905fe7 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2546,7 +2546,6 @@ ssize_t generic_file_aio_write(struct kiocb *iocb, const struct iovec *iov,
 
 	BUG_ON(iocb->ki_pos != pos);
 
-	sb_start_write(inode->i_sb);
 	mutex_lock(&inode->i_mutex);
 	ret = __generic_file_aio_write(iocb, iov, nr_segs, &iocb->ki_pos);
 	mutex_unlock(&inode->i_mutex);
@@ -2558,7 +2557,6 @@ ssize_t generic_file_aio_write(struct kiocb *iocb, const struct iovec *iov,
 		if (err < 0 && ret > 0)
 			ret = err;
 	}
-	sb_end_write(inode->i_sb);
 	return ret;
 }
 EXPORT_SYMBOL(generic_file_aio_write);
diff --git a/mm/filemap_xip.c b/mm/filemap_xip.c
index a912da6..28fe26b 100644
--- a/mm/filemap_xip.c
+++ b/mm/filemap_xip.c
@@ -404,8 +404,6 @@ xip_file_write(struct file *filp, const char __user *buf, size_t len,
 	loff_t pos;
 	ssize_t ret;
 
-	sb_start_write(inode->i_sb);
-
 	mutex_lock(&inode->i_mutex);
 
 	if (!access_ok(VERIFY_READ, buf, len)) {
@@ -439,7 +437,6 @@ xip_file_write(struct file *filp, const char __user *buf, size_t len,
 	current->backing_dev_info = NULL;
  out_up:
 	mutex_unlock(&inode->i_mutex);
-	sb_end_write(inode->i_sb);
 	return ret;
 }
 EXPORT_SYMBOL_GPL(xip_file_write);
diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 72043d6..b12fd86 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c
@@ -27,10 +27,30 @@
 #include <linux/pfn.h>
 #include <linux/kmemleak.h>
 #include <linux/atomic.h>
+#include <linux/llist.h>
 #include <asm/uaccess.h>
 #include <asm/tlbflush.h>
 #include <asm/shmparam.h>
 
+struct vfree_deferred {
+	struct llist_head list;
+	struct work_struct wq;
+};
+static DEFINE_PER_CPU(struct vfree_deferred, vfree_deferred);
+
+static void __vunmap(const void *, int);
+
+static void free_work(struct work_struct *w)
+{
+	struct vfree_deferred *p = container_of(w, struct vfree_deferred, wq);
+	struct llist_node *llnode = llist_del_all(&p->list);
+	while (llnode) {
+		void *p = llnode;
+		llnode = llist_next(llnode);
+		__vunmap(p, 1);
+	}
+}
+
 /*** Page table manipulation functions ***/
 
 static void vunmap_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end)
@@ -1175,10 +1195,14 @@ void __init vmalloc_init(void)
 
 	for_each_possible_cpu(i) {
 		struct vmap_block_queue *vbq;
+		struct vfree_deferred *p;
 
 		vbq = &per_cpu(vmap_block_queue, i);
 		spin_lock_init(&vbq->lock);
 		INIT_LIST_HEAD(&vbq->free);
+		p = &per_cpu(vfree_deferred, i);
+		init_llist_head(&p->list);
+		INIT_WORK(&p->wq, free_work);
 	}
 
 	/* Import existing vmlist entries. */
@@ -1486,7 +1510,7 @@ static void __vunmap(const void *addr, int deallocate_pages)
 	kfree(area);
 	return;
 }
-
+ 
 /**
  *	vfree  -  release memory allocated by vmalloc()
  *	@addr:		memory base address
@@ -1495,15 +1519,25 @@ static void __vunmap(const void *addr, int deallocate_pages)
  *	obtained from vmalloc(), vmalloc_32() or __vmalloc(). If @addr is
  *	NULL, no operation is performed.
  *
- *	Must not be called in interrupt context.
+ *	Must not be called in NMI context (strictly speaking, only if we don't
+ *	have CONFIG_ARCH_HAVE_NMI_SAFE_CMPXCHG, but making the calling
+ *	conventions for vfree() arch-depenedent would be a really bad idea)
+ *	
  */
 void vfree(const void *addr)
 {
-	BUG_ON(in_interrupt());
+	BUG_ON(in_nmi());
 
 	kmemleak_free(addr);
 
-	__vunmap(addr, 1);
+	if (!addr)
+		return;
+	if (unlikely(in_interrupt())) {
+		struct vfree_deferred *p = &__get_cpu_var(vfree_deferred);
+		llist_add((struct llist_node *)addr, &p->list);
+		schedule_work(&p->wq);
+	} else
+		__vunmap(addr, 1);
 }
 EXPORT_SYMBOL(vfree);
 
@@ -1520,7 +1554,8 @@ void vunmap(const void *addr)
 {
 	BUG_ON(in_interrupt());
 	might_sleep();
-	__vunmap(addr, 0);
+	if (addr)
+		__vunmap(addr, 0);
 }
 EXPORT_SYMBOL(vunmap);