This paper develops a new approach to verifying a performant file system that isolates crash safety and concurrency reasoning to a transaction system that gives atomic access to the disk, so that the rest of the file system can be verified with sequential reasoning. We demonstrate this approach in DaisyNFS, a Network File System (NFS) server written in Go that runs on top of a disk. DaisyNFS uses GoTxn, a new verified, concurrent transaction system that extends GoJournal  with two-phase locking and an allocator. The transaction system's specification formalizes under what conditions transactions can be verified with only sequential reasoning, and comes with a mechanized proof in Coq  that connects the specification to the implementation. As evidence that proofs enjoy sequential reasoning, DaisyNFS uses Dafny , a sequential verification language, to implement and verify all the NFS operations on top of GoTxn. The sequential proofs helped achieve a number of good properties in DaisyNFS: easy incremental development (for example, adding support for large files), a relatively short proof (only 2× as many lines of proof as code), and a performant implementation (at least 60% the throughput of the Linux NFS server exporting ext4 across a variety of benchmarks).