Linux semantics for fallocate()

config/kernel-truncate-pagecache-range.m4

@@ -0,0 +1,18 @@
+dnl #
+dnl # 3.4 API Change
+dnl # Added truncate_pagecache_range() function.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_TRUNCATE_PAGECACHE_RANGE],
+	[AC_MSG_CHECKING([whether truncate_pagecache_range() is available])
+	ZFS_LINUX_TRY_COMPILE_SYMBOL([
+		#include <linux/mm.h>
+	], [
+		truncate_pagecache_range(NULL, 0, 0);
+	], [truncate_pagecache_range], [mm/truncate.c], [
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_TRUNCATE_PAGECACHE_RANGE, 1,
+		          [truncate_pagecache_range() is available])
+	], [
+		AC_MSG_RESULT(no)
+	])
+])

config/kernel.m4

@@ -69,6 +69,7 @@ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
 	ZFS_AC_KERNEL_MKDIR_UMODE_T
 	ZFS_AC_KERNEL_LOOKUP_NAMEIDATA
 	ZFS_AC_KERNEL_CREATE_NAMEIDATA
+	ZFS_AC_KERNEL_TRUNCATE_PAGECACHE_RANGE
 	ZFS_AC_KERNEL_TRUNCATE_RANGE
 	ZFS_AC_KERNEL_AUTOMOUNT
 	ZFS_AC_KERNEL_ENCODE_FH_WITH_INODE

include/linux/vfs_compat.h

@@ -62,6 +62,37 @@ truncate_setsize(struct inode *ip, loff_t new)
 }
 #endif /* HAVE_TRUNCATE_SETSIZE */
 
+/*
+ * 3.4 API change,
+ * Add truncate_pagecache_range if it is not exported by the Linux kernel.
+ *
+ * Comment from the Linux kernel sources:
+ *
+ * This function should typically be called before the filesystem
+ * releases resources associated with the freed range (eg. deallocates
+ * blocks). This way, pagecache will always stay logically coherent
+ * with on-disk format, and the filesystem would not have to deal with
+ * situations such as writepage being called for a page that has already
+ * had its underlying blocks deallocated.
+ */
+#ifndef HAVE_TRUNCATE_PAGECACHE_RANGE
+static inline void
+truncate_pagecache_range(struct inode *ip, loff_t off, loff_t endoff)
+{
+	struct address_space *mapping = ip->i_mapping;
+	loff_t unmap_start, unmap_end;
+
+	ASSERT(ISP2(PAGE_SIZE));
+	unmap_start = roundup(off, PAGE_SIZE);
+	unmap_end = ((1 + endoff) & (PAGE_SIZE - 1)) - 1;
+
+	if ((uint64_t)unmap_end > (uint64_t)unmap_start)
+		unmap_mapping_range(mapping, unmap_start,
+		    1 + unmap_end - unmap_start, 0);
+	truncate_inode_pages_range(mapping, off, endoff);
+}
+#endif /*  HAVE_TRUNCATE_PAGECACHE_RANGE */
+
 #if defined(HAVE_BDI) && !defined(HAVE_BDI_SETUP_AND_REGISTER)
 /*
  * 2.6.34 API change,

module/zfs/zpl_file.c

@@ -486,22 +486,43 @@ zpl_fallocate_common(struct inode *ip, int mode, loff_t offset, loff_t len)
 	cred_t *cr = CRED();
 	int error = -EOPNOTSUPP;
 
+#ifdef FALLOC_FL_PUNCH_HOLE
+	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
+#else
 	if (mode & FALLOC_FL_KEEP_SIZE)
+#endif
 		return (-EOPNOTSUPP);
 
 	crhold(cr);
 
 #ifdef FALLOC_FL_PUNCH_HOLE
 	if (mode & FALLOC_FL_PUNCH_HOLE) {
 		flock64_t bf;
+		loff_t olen;
 
+		if (len <= 0)
+			return (-EINVAL);
+
+		spl_inode_lock(ip);
+		olen = i_size_read(ip);
+
+		if (offset > olen) {
+			spl_inode_unlock(ip);
+			return (0);
+		}
+		if (offset + len > olen)
+			len = olen - offset;
 		bf.l_type = F_WRLCK;
 		bf.l_whence = 0;
 		bf.l_start = offset;
 		bf.l_len = len;
 		bf.l_pid = 0;
 
 		error = -zfs_space(ip, F_FREESP, &bf, FWRITE, offset, cr);
+		if (!error)
+			truncate_pagecache_range(ip, offset,
+			    offset + len - 1);
+		spl_inode_unlock(ip);
 	}
 #endif /* FALLOC_FL_PUNCH_HOLE */