Fix statfs(2) for 32-bit user space

When handling a 32-bit statfs() system call the returned fields,
although 64-bit in the kernel, must be limited to 32-bits or an
EOVERFLOW error will be returned.

This is less of an issue for block counts since the default
reported block size in 128KiB. But since it is possible to
set a smaller block size, these values will be scaled as
needed to fit in a 32-bit unsigned long.

Unlike most other filesystems the total possible file counts
are more likely to overflow because they are calculated based
on the available free space in the pool. In order to prevent
this the reported value must be capped at 2^32-1. This is
only for statfs(2) reporting, there are no changes to the
internal ZFS limits.

Reviewed-by: Andreas Dilger <[email protected]>
Reviewed-by: Richard Yao <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]>
Issue #7927
Closes #7122
Closes #7937

5 files changed, 64 insertions, 5 deletions

diff --git a/config/kernel-in-compat-syscall.m4 b/config/kernel-in-compat-syscall.m4
new file mode 100644
index 000000000..9fca9da20
--- /dev/null
+++ b/config/kernel-in-compat-syscall.m4
@@ -0,0 +1,20 @@
+dnl #
+dnl # 4.5 API change
+dnl # Added in_compat_syscall() which can be overridden on a per-
+dnl # architecture basis.  Prior to this is_compat_task() was the
+dnl # provided interface.
+dnl #
+AC_DEFUN([ZFS_AC_KERNEL_IN_COMPAT_SYSCALL], [
+	AC_MSG_CHECKING([whether in_compat_syscall() is available])
+	ZFS_LINUX_TRY_COMPILE([
+		#include <linux/compat.h>
+	],[
+		in_compat_syscall();
+	],[
+		AC_MSG_RESULT(yes)
+		AC_DEFINE(HAVE_IN_COMPAT_SYSCALL, 1,
+		    [in_compat_syscall() is available])
+	],[
+		AC_MSG_RESULT(no)
+	])
+])
diff --git a/config/kernel.m4 b/config/kernel.m4
index c7ca260c5..3777f45c8 100644
--- a/config/kernel.m4
+++ b/config/kernel.m4
@@ -129,6 +129,7 @@ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
 	ZFS_AC_KERNEL_GLOBAL_PAGE_STATE
 	ZFS_AC_KERNEL_ACL_HAS_REFCOUNT
 	ZFS_AC_KERNEL_USERNS_CAPABILITIES
+	ZFS_AC_KERNEL_IN_COMPAT_SYSCALL
 
 	AS_IF([test "$LINUX_OBJ" != "$LINUX"], [
 		KERNELMAKE_PARAMS="$KERNELMAKE_PARAMS O=$LINUX_OBJ"
diff --git a/include/linux/vfs_compat.h b/include/linux/vfs_compat.h
index c8203bd55..90b3cca78 100644
--- a/include/linux/vfs_compat.h
+++ b/include/linux/vfs_compat.h
@@ -30,6 +30,7 @@
 #include <sys/taskq.h>
 #include <sys/cred.h>
 #include <linux/backing-dev.h>
+#include <linux/compat.h>
 
 /*
  * 2.6.28 API change,
@@ -626,4 +627,21 @@ inode_set_iversion(struct inode *ip, u64 val)
 }
 #endif
 
+/*
+ * Returns true when called in the context of a 32-bit system call.
+ */
+static inline int
+zpl_is_32bit_api(void)
+{
+#ifdef CONFIG_COMPAT
+#ifdef HAVE_IN_COMPAT_SYSCALL
+	return (in_compat_syscall());
+#else
+	return (is_compat_task());
+#endif
+#else
+	return (BITS_PER_LONG == 32);
+#endif
+}
+
 #endif /* _ZFS_VFS_H */
diff --git a/module/zfs/zfs_vfsops.c b/module/zfs/zfs_vfsops.c
index 761133936..bcdfa26b9 100644
--- a/module/zfs/zfs_vfsops.c
+++ b/module/zfs/zfs_vfsops.c
@@ -1245,15 +1245,13 @@ zfs_statvfs(struct dentry *dentry, struct kstatfs *statp)
 {
 	zfsvfs_t *zfsvfs = dentry->d_sb->s_fs_info;
 	uint64_t refdbytes, availbytes, usedobjs, availobjs;
-	uint64_t fsid;
-	uint32_t bshift;
 
 	ZFS_ENTER(zfsvfs);
 
 	dmu_objset_space(zfsvfs->z_os,
 	    &refdbytes, &availbytes, &usedobjs, &availobjs);
 
-	fsid = dmu_objset_fsid_guid(zfsvfs->z_os);
+	uint64_t fsid = dmu_objset_fsid_guid(zfsvfs->z_os);
 	/*
 	 * The underlying storage pool actually uses multiple block
 	 * size.  Under Solaris frsize (fragment size) is reported as
@@ -1265,7 +1263,7 @@ zfs_statvfs(struct dentry *dentry, struct kstatfs *statp)
 	 */
 	statp->f_frsize = zfsvfs->z_max_blksz;
 	statp->f_bsize = zfsvfs->z_max_blksz;
-	bshift = fls(statp->f_bsize) - 1;
+	uint32_t bshift = fls(statp->f_bsize) - 1;
 
 	/*
 	 * The following report "total" blocks of various kinds in
@@ -1282,7 +1280,7 @@ zfs_statvfs(struct dentry *dentry, struct kstatfs *statp)
 	 * static metadata.  ZFS doesn't preallocate files, so the best
 	 * we can do is report the max that could possibly fit in f_files,
 	 * and that minus the number actually used in f_ffree.
-	 * For f_ffree, report the smaller of the number of object available
+	 * For f_ffree, report the smaller of the number of objects available
 	 * and the number of blocks (each object will take at least a block).
 	 */
 	statp->f_ffree = MIN(availobjs, availbytes >> DNODE_SHIFT);
diff --git a/module/zfs/zpl_super.c b/module/zfs/zpl_super.c
index 5c426b0a9..216c79401 100644
--- a/module/zfs/zpl_super.c
+++ b/module/zfs/zpl_super.c
@@ -181,6 +181,28 @@ zpl_statfs(struct dentry *dentry, struct kstatfs *statp)
 	spl_fstrans_unmark(cookie);
 	ASSERT3S(error, <=, 0);
 
+	/*
+	 * If required by a 32-bit system call, dynamically scale the
+	 * block size up to 16MiB and decrease the block counts.  This
+	 * allows for a maximum size of 64EiB to be reported.  The file
+	 * counts must be artificially capped at 2^32-1.
+	 */
+	if (unlikely(zpl_is_32bit_api())) {
+		while (statp->f_blocks > UINT32_MAX &&
+		    statp->f_bsize < SPA_MAXBLOCKSIZE) {
+			statp->f_frsize <<= 1;
+			statp->f_bsize <<= 1;
+
+			statp->f_blocks >>= 1;
+			statp->f_bfree >>= 1;
+			statp->f_bavail >>= 1;
+		}
+
+		uint64_t usedobjs = statp->f_files - statp->f_ffree;
+		statp->f_ffree = MIN(statp->f_ffree, UINT32_MAX - usedobjs);
+		statp->f_files = statp->f_ffree + usedobjs;
+	}
+
 	return (error);
 }