On Thu, 6 Oct 2011, Andrew Morton wrote:

On Wed, 14 Sep 2011 14:35:21 -0400 (EDT)
Nicolas Pitre [off-list ref] wrote:

quoted

The existing vm_area_register_early() allows for early vmalloc space
allocation.  However upcoming cleanups in the ARM architecture require
that some fixed locations in the vmalloc area be reserved also very early.

The name "vm_area_register_early" would have been a good name for the
reservation part without the allocation.  Since it is already in use with
different semantics, let's create vm_area_add_early() instead.

Both vm_area_register_early() and vm_area_add_early() can be used together
meaning that the former is now implemented using the later where it is
ensured that no conflicting areas are added, but no attempt is made to
make the allocation scheme in vm_area_register_early() more sophisticated.
After all, you must know what you're doing when using those functions.

Signed-off-by: Nicolas Pitre <redacted>
---

Comments / ACKs appreciated.

Deafening silence?

I interpreted it as "no objections".

quoted

diff --git a/include/linux/vmalloc.h b/include/linux/vmalloc.h
index 9332e52ea8..e7d2cba995 100644
--- a/include/linux/vmalloc.h
+++ b/include/linux/vmalloc.h

@@ -130,6 +130,7 @@ extern long vwrite(char *buf, char *addr, unsigned long count);
  */
 extern rwlock_t vmlist_lock;
 extern struct vm_struct *vmlist;
+extern __init void vm_area_add_early(struct vm_struct *vm);
 extern __init void vm_area_register_early(struct vm_struct *vm, size_t align);
 
 #ifdef CONFIG_SMP

diff --git a/mm/vmalloc.c b/mm/vmalloc.c
index 7ef0903058..bf20a0ff95 100644
--- a/mm/vmalloc.c
+++ b/mm/vmalloc.c

@@ -1118,6 +1118,31 @@ void *vm_map_ram(struct page **pages, unsigned int count, int node, pgprot_t pro
 EXPORT_SYMBOL(vm_map_ram);
 
 /**
+ * vm_area_add_early - add vmap area early during boot
+ * @vm: vm_struct to add
+ *
+ * This function is used to add fixed kernel vm area to vmlist before
+ * vmalloc_init() is called.  @vm->addr, @vm->size, and @vm->flags
+ * should contain proper values and the other fields should be zero.
+ *
+ * DO NOT USE THIS FUNCTION UNLESS YOU KNOW WHAT YOU'RE DOING.
+ */
+void __init vm_area_add_early(struct vm_struct *vm)
+{
+	struct vm_struct *tmp, **p;
+
+	for (p = &vmlist; (tmp = *p) != NULL; p = &tmp->next) {
+		if (tmp->addr >= vm->addr) {
+			BUG_ON(tmp->addr < vm->addr + vm->size);
+			break;
+		} else
+			BUG_ON(tmp->addr + tmp->size > vm->addr);
+	}
+	vm->next = *p;
+	*p = vm;
+}
+
+/**
  * vm_area_register_early - register vmap area early during boot
  * @vm: vm_struct to register
  * @align: requested alignment

@@ -1139,8 +1164,7 @@ void __init vm_area_register_early(struct vm_struct *vm, size_t align)
 
 	vm->addr = (void *)addr;
 
-	vm->next = vmlist;
-	vmlist = vm;
+	vm_area_add_early(vm);
 }
 
 void __init vmalloc_init(void)

I tossed this into my tree for a bit of testing, assuming it's
up-to-date and still desired?

Yes it is, on both counts.

Feel free to add it to some ARM tree.  I'll drop my copy when that
turns up in linux-next.

So I did, however I'm postponing the whole series that depend on this to 
the next cycle because of other prerequisite cleanups that are going in 
this cycle from different paths.

Yes, the naming scheme in there is gruseome.

I didn't want to go as far as renaming it across the tree, in case other 
people are relying on the current name in their own tree.


Nicolas