Amazon EFS provides an NFSv4.1 filesystem which appears to use (close to) full length (128 byte) file handles. This causes the handle reported by name_to_handle_at() to exceed MAX_HANDLE_SZ, resulting in EOVERFLOW if 128 bytes were requested, or EINVAL if the size reported by the previous call was requested. To fix this, increase MAX_HANDLE_SIZE a little, and add a BUILD_BUG so that this sort of inconsistent error reporting won't happen again. Link: https://github.com/systemd/systemd/issues/7082#issuecomment-347380436 Signed-off-by: NeilBrown <neilb@xxxxxxxx> --- fs/nfs/export.c | 2 ++ include/linux/exportfs.h | 10 ++++++++-- 2 files changed, 10 insertions(+), 2 deletions(-) diff --git a/fs/nfs/export.c b/fs/nfs/export.c index 83fd09fc8f77..23b2fc3ab2bb 100644 --- a/fs/nfs/export.c +++ b/fs/nfs/export.c @@ -39,6 +39,8 @@ nfs_encode_fh(struct inode *inode, __u32 *p, int *max_len, struct inode *parent) size_t fh_size = offsetof(struct nfs_fh, data) + server_fh->size; int len = EMBED_FH_OFF + XDR_QUADLEN(fh_size); + BUILD_BUG_ON_MSG(EMBED_FH_OFF + NFS_MAXFHSIZE > MAX_HANDLE_SZ, + "MAX_HANDLE_SZ is too small"); dprintk("%s: max fh len %d inode %p parent %p", __func__, *max_len, inode, parent); diff --git a/include/linux/exportfs.h b/include/linux/exportfs.h index 0d3037419bc7..71eb9c2cc2fb 100644 --- a/include/linux/exportfs.h +++ b/include/linux/exportfs.h @@ -11,8 +11,14 @@ struct iomap; struct super_block; struct vfsmount; -/* limit the handle size to NFSv4 handle size now */ -#define MAX_HANDLE_SZ 128 +/* Must be larger than NFSv4 file handle. + * overlayfs doesn't want this too close to 255. + * NOTE: This value MUST NOT be exported to user-space. + * Applications must only ever see MAX_HANDLE_SZ == 128. + * If they try a larger number on older kernels, they + * will get -EINVAL which will be confusing. + */ +#define MAX_HANDLE_SZ 200 /* * The fileid_type identifies how the file within the filesystem is encoded.
