]> git.baikalelectronics.ru Git - kernel.git/commitdiff
iov_iter_get_pages_alloc(): lift freeing pages array on failure exits into wrapper
authorAl Viro <viro@zeniv.linux.org.uk>
Sat, 11 Jun 2022 00:38:20 +0000 (20:38 -0400)
committerAl Viro <viro@zeniv.linux.org.uk>
Tue, 9 Aug 2022 02:37:19 +0000 (22:37 -0400)
Incidentally, ITER_XARRAY did *not* free the sucker in case when
iter_xarray_populate_pages() returned 0...

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
lib/iov_iter.c

index 12dda1013bea033fa70d79f0b658dc55898dde59..e14749711e342f0286f160c7f914de0a800bf717 100644 (file)
@@ -1506,15 +1506,10 @@ static ssize_t pipe_get_pages_alloc(struct iov_iter *i,
                maxsize = n;
        else
                npages = DIV_ROUND_UP(maxsize + off, PAGE_SIZE);
-       p = get_pages_array(npages);
+       *pages = p = get_pages_array(npages);
        if (!p)
                return -ENOMEM;
-       n = __pipe_get_pages(i, maxsize, p, off);
-       if (n > 0)
-               *pages = p;
-       else
-               kvfree(p);
-       return n;
+       return __pipe_get_pages(i, maxsize, p, off);
 }
 
 static ssize_t iter_xarray_get_pages_alloc(struct iov_iter *i,
@@ -1544,10 +1539,9 @@ static ssize_t iter_xarray_get_pages_alloc(struct iov_iter *i,
                        count++;
        }
 
-       p = get_pages_array(count);
+       *pages = p = get_pages_array(count);
        if (!p)
                return -ENOMEM;
-       *pages = p;
 
        nr = iter_xarray_populate_pages(p, i->xarray, index, count);
        if (nr == 0)
@@ -1556,7 +1550,7 @@ static ssize_t iter_xarray_get_pages_alloc(struct iov_iter *i,
        return min_t(size_t, nr * PAGE_SIZE - offset, maxsize);
 }
 
-ssize_t iov_iter_get_pages_alloc(struct iov_iter *i,
+static ssize_t __iov_iter_get_pages_alloc(struct iov_iter *i,
                   struct page ***pages, size_t maxsize,
                   size_t *start)
 {
@@ -1583,16 +1577,12 @@ ssize_t iov_iter_get_pages_alloc(struct iov_iter *i,
                *start = addr % PAGE_SIZE;
                addr &= PAGE_MASK;
                n = DIV_ROUND_UP(maxsize + *start, PAGE_SIZE);
-               p = get_pages_array(n);
+               *pages = p = get_pages_array(n);
                if (!p)
                        return -ENOMEM;
                res = get_user_pages_fast(addr, n, gup_flags, p);
-               if (unlikely(res <= 0)) {
-                       kvfree(p);
-                       *pages = NULL;
+               if (unlikely(res <= 0))
                        return res;
-               }
-               *pages = p;
                return min_t(size_t, maxsize, res * PAGE_SIZE - *start);
        }
        if (iov_iter_is_bvec(i)) {
@@ -1613,6 +1603,22 @@ ssize_t iov_iter_get_pages_alloc(struct iov_iter *i,
                return iter_xarray_get_pages_alloc(i, pages, maxsize, start);
        return -EFAULT;
 }
+
+ssize_t iov_iter_get_pages_alloc(struct iov_iter *i,
+                  struct page ***pages, size_t maxsize,
+                  size_t *start)
+{
+       ssize_t len;
+
+       *pages = NULL;
+
+       len = __iov_iter_get_pages_alloc(i, pages, maxsize, start);
+       if (len <= 0) {
+               kvfree(*pages);
+               *pages = NULL;
+       }
+       return len;
+}
 EXPORT_SYMBOL(iov_iter_get_pages_alloc);
 
 size_t csum_and_copy_from_iter(void *addr, size_t bytes, __wsum *csum,