Hi all, Here are some late fixes for 4.19. The first patch teaches xfs to trigger a sb counter recalculation if the icount is bad; when I added the icount check to the write verifier, I forgot to add the corresponding check to the read side. The second patch fixes the online repair "find AG btree root" function to ignore btree blocks that have siblings and to ignore a btree level if multiple sibling-less blocks are found. The third patch fixes a problem when allocating a transaction for online repairs where if the AGF block or AGI inode counts are totally garbage we can end up ENOSPC'ing out of online repair after an attempt to reserve a ridiculous number of blocks. This is unfortunate, particularly if we were setting up for an AGF/AGI counter repair. The fourth patch fixes some buffer state management bugs so that we don't accidentally clobber b_ops on buffers that were already in-core when we try to find an AG header's btree root blocks. The fifth patch fixes an uninitialized variable usage in the iomap code. The sixth patch avoids a crash in the VFS readlink routines if we have an inline symlink that's corrupted such that if_data is NULL. Comments and questions are, as always, welcome. --D
