mlock: make mlock error return Posixly Correct
Lee Schermerhorn [Sun, 19 Oct 2008 03:26:56 +0000 (20:26 -0700)]
Rework Posix error return for mlock().

Posix requires error code for mlock*() system calls for some conditions
that differ from what kernel low level functions, such as
get_user_pages(), return for those conditions.  For more info, see:

http://marc.info/?l=linux-kernel&m=121750892930775&w=2

This patch provides the same translation of get_user_pages()
error codes to posix specified error codes in the context
of the mlock rework for unevictable lru.

[akpm@linux-foundation.org: fix build]
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory.c
mm/mlock.c

index b8487f8..ba86b43 100644 (file)
@@ -2821,7 +2821,7 @@ int make_pages_present(unsigned long addr, unsigned long end)
                        len, write, 0, NULL, NULL);
        if (ret < 0)
                return ret;
-       return ret == len ? 0 : -1;
+       return ret == len ? 0 : -EFAULT;
 }
 
 #if !defined(__HAVE_ARCH_GATE_AREA)
index bce1b22..008ea70 100644 (file)
@@ -248,11 +248,24 @@ static long __mlock_vma_pages_range(struct vm_area_struct *vma,
                        addr += PAGE_SIZE;      /* for next get_user_pages() */
                        nr_pages--;
                }
+               ret = 0;
        }
 
        lru_add_drain_all();    /* to update stats */
 
-       return 0;       /* count entire vma as locked_vm */
+       return ret;     /* count entire vma as locked_vm */
+}
+
+/*
+ * convert get_user_pages() return value to posix mlock() error
+ */
+static int __mlock_posix_error_return(long retval)
+{
+       if (retval == -EFAULT)
+               retval = -ENOMEM;
+       else if (retval == -ENOMEM)
+               retval = -EAGAIN;
+       return retval;
 }
 
 #else /* CONFIG_UNEVICTABLE_LRU */
@@ -265,9 +278,15 @@ static long __mlock_vma_pages_range(struct vm_area_struct *vma,
                                   int mlock)
 {
        if (mlock && (vma->vm_flags & VM_LOCKED))
-               make_pages_present(start, end);
+               return make_pages_present(start, end);
+       return 0;
+}
+
+static inline int __mlock_posix_error_return(long retval)
+{
        return 0;
 }
+
 #endif /* CONFIG_UNEVICTABLE_LRU */
 
 /**
@@ -434,10 +453,7 @@ success:
                downgrade_write(&mm->mmap_sem);
 
                ret = __mlock_vma_pages_range(vma, start, end, 1);
-               if (ret > 0) {
-                       mm->locked_vm -= ret;
-                       ret = 0;
-               }
+
                /*
                 * Need to reacquire mmap sem in write mode, as our callers
                 * expect this.  We have no support for atomically upgrading
@@ -451,6 +467,11 @@ success:
                /* non-NULL *prev must contain @start, but need to check @end */
                if (!(*prev) || end > (*prev)->vm_end)
                        ret = -ENOMEM;
+               else if (ret > 0) {
+                       mm->locked_vm -= ret;
+                       ret = 0;
+               } else
+                       ret = __mlock_posix_error_return(ret); /* translate if needed */
        } else {
                /*
                 * TODO:  for unlocking, pages will already be resident, so