fuse: add no_create fallback and offer ATOMIC_O_TRUNC/BIG_WRITES in INIT

[copybara-service]

fuse: add no_create fallback and offer ATOMIC_O_TRUNC/BIG_WRITES in INIT

Three independent fixes to the FUSE client, found while getting mountpoint-s3 to work inside a sandbox. Each applies to any FUSE server.

1. Fall back to FUSE_MKNOD when the server does not implement FUSE_CREATE.
FUSE_CREATE is optional; a server may answer it with ENOSYS. Linux then latches fc->no_create and creates the file with FUSE_MKNOD followed by FUSE_OPEN (fs/fuse/dir.c:fuse_atomic_open()). The sentry returns the ENOSYS to the application, so open(O_CREAT) fails on any such server while mknod(2) then open(2) succeeds. mountpoint-s3 is one such server: it implements mknod and open but not create, so nothing that creates files the ordinary way — open(path, "w"), shell >, cp — works on it under gVisor. This adds conn.noCreate, mirroring the existing conn.noOpen.

2. Offer FUSE_ATOMIC_O_TRUNC in FUSE_INIT.
The sentry already implements the flag on the reply side but never offers it, so no server can enable it and O_TRUNC is always emulated with a separate SETATTR. Servers that require atomic truncate refuse to start: mountpoint-s3 with --allow-overwrite panics at INIT, which leaves the mount unusable. Two latent bugs become reachable once the flag is offered and are fixed first: a server that has ENOSYSed FUSE_OPEN and negotiated the flag had O_TRUNC silently dropped, and the first FUSE_OPEN carrying a delegated O_TRUNC could itself return ENOSYS and leave the file untruncated.

3. Offer FUSE_BIG_WRITES in FUSE_INIT.
Also already implemented on the reply side but never offered, so every FUSE_WRITE is clamped to one page. With the flag, writes go out at up to min(max_pages * 4096, max_write). Two prerequisites are fixed first: the write loop bounded a byte count by an absolute file offset (one spurious zero-length FUSE_WRITE per write(), and over-sized allocations once the clamp is gone), and the daemon read-buffer minimum used the reply header instead of FUSEWriteIn, 24 bytes short of what Linux requires. Note that each in-flight write is held twice in sentry memory until the daemon reads it, so the worst-case memory pinned by writes to a stalled daemon grows with the larger request size (up to maxActiveRequests × 2 MiB).

Linux has offered both flags since FUSE 7.9.

Measured with mountpoint-s3 1.24.0 inside a sandbox, before → after: open(O_CREAT) ENOSYS → succeeds; --allow-overwrite panics at INIT → mounts and overwrites; INIT flags offered 0x400000 → 0x400028 (mount-s3 echoes 0x400028 with --allow-overwrite, 0x400020 without); dd bs=1M write throughput 138 → 854 MB/s (median of 3; this measures FUSE_WRITE round trips, the S3 upload happens on close). bazel test //pkg/sentry/fsimpl/fuse:fuse_test passes.

Updates #3199

Assisted-by: Claude Code
FUTURE_COPYBARA_INTEGRATE_REVIEW=#14447 from nicolaslara:upstream/fuse-init-flags-and-no-create 392210b