2011/11/28 Christoph Hellwig <hch@xxxxxxxxxxxxx>: > Guoquan and hank, > > are you using 32-bit or 64-bit kernels? I just noticed we have a > problem with exporting 64-bit inodes on 32-bit kernel because the > VFS i_ino field is just 32-bits long. The patch below would fix > that issue. > yes, we use inode64 on 32-bit kernel. > --- xfs.orig/fs/xfs/linux-2.6/xfs_export.c 2011-11-28 12:11:08.923630697 +0100 > +++ xfs/fs/xfs/linux-2.6/xfs_export.c 2011-11-28 12:13:21.766244360 +0100 > @@ -61,6 +61,8 @@ xfs_fs_encode_fh( > struct fid *fid = (struct fid *)fh; > struct xfs_fid64 *fid64 = (struct xfs_fid64 *)fh; > struct inode *inode = dentry->d_inode; > + struct inode *parent; > + struct xfs_inode *ip = XFS_I(inode); > int fileid_type; > int len; > > @@ -98,22 +100,24 @@ xfs_fs_encode_fh( > switch (fileid_type) { > case FILEID_INO32_GEN_PARENT: > spin_lock(&dentry->d_lock); > - fid->i32.parent_ino = dentry->d_parent->d_inode->i_ino; > - fid->i32.parent_gen = dentry->d_parent->d_inode->i_generation; > + parent = dentry->d_parent->d_inode; > + fid->i32.parent_ino = XFS_I(parent)->i_ino; > + fid->i32.parent_gen = parent->i_generation; > spin_unlock(&dentry->d_lock); > /*FALLTHRU*/ > case FILEID_INO32_GEN: > - fid->i32.ino = inode->i_ino; > + fid->i32.ino = ip->i_ino; > fid->i32.gen = inode->i_generation; > break; > case FILEID_INO32_GEN_PARENT | XFS_FILEID_TYPE_64FLAG: > spin_lock(&dentry->d_lock); > - fid64->parent_ino = dentry->d_parent->d_inode->i_ino; > - fid64->parent_gen = dentry->d_parent->d_inode->i_generation; > + parent = dentry->d_parent->d_inode; > + fid64->parent_ino = XFS_I(parent)->i_ino; > + fid64->parent_gen = parent->i_generation; > spin_unlock(&dentry->d_lock); > /*FALLTHRU*/ > case FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG: > - fid64->ino = inode->i_ino; > + fid64->ino = ip->i_ino; > fid64->gen = inode->i_generation; > break; > } I haven't tested this patch, but I have a question now: although I use inode64 option when mounting, my filesystem did not exceed 2T limit, so, 32-bit inode would be no problem, right? -- The simplest is not all best but the best is surely the simplest! _______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs