Hi, On Fri, Aug 24, 2018 at 9:16 PM, Eric Biggers <ebiggers@xxxxxxxxxx> wrote: > Hi Gao, > > On Sat, Aug 25, 2018 at 10:29:26AM +0800, Gao Xiang wrote: >> Hi, >> >> At last, I hope filesystems could select the on-disk position of hash tree and 'struct fsverity_descriptor' >> rather than fixed in the end of verity files...I think if fs-verity preparing such support and interfaces could be better.....hmmm... :( > > In theory it would be a much cleaner design to store verity metadata separately > from the data. But the Merkle tree can be very large. For example, a 1 GB file > using SHA-512 would have a 16.6 MB Merkle tree. So the Merkle tree can't be an > extended attribute, since the xattrs API requires xattrs to be small (<= 64 KB), > and most filesystems further limit xattr sizes in their on-disk format to as > little as 4 KB. Furthermore, even if both of these limits were to be increased, > the xattrs functions (both the syscalls, and the internal functions that > filesystems have) are all based around getting/setting the entire xattr value. > > Also when used with fscrypt, we want the Merkle tree and fsverity_descriptor to > be encrypted, so they doesn't leak plaintext hashes. And we want the Merkle > tree to be paged into memory, just like the file contents, to take advantage of > the usual Linux memory management. > > What we really need is *streams*, like NTFS has. But the filesystems we're > targetting don't support streams, nor does the Linux syscall interface have any > API for accessing streams, nor does the VFS support them. > > Adding streams support to all those things would be a huge multi-year effort, > controversial, and almost certainly not worth it just for fs-verity. > > So simply storing the verity metadata past i_size seems like the best solution > for now. > > That being said, in the future we could pretty easily swap out the calls to > read_mapping_page() with something else if a particular filesystem wanted to > store the metadata somewhere else. We actually even originally had a function > ->read_metadata_page() in the filesystem's fsverity_operations, but it turned > out to be unnecessary and I replaced it with directly calling > read_mapping_page(), but it could be changed back at any time. What about an xattr not to hold the Merkle tree, but to contain a suitable reference to a file/inode+offset that contains it (+ toplevel hash for said tree/file or the descriptor/struct)? If you also expose said file in the directory structure, things such as backups might be easier to handle. For where the tree is appended to the file, you could self-reference. -Olof
