feat(tvix/nar-bridge): init r/6600

This provides a Nix HTTP Binary Cache interface in front of a tvix-store
that's reachable via gRPC.

TODOs:

 - remove import command, move serve up to toplevel. We have nix-copy-
   closure and tvix-store commands.
 - loop into CI. We should be able to fetch the protos as a third-party
   dependency.
 - Check if we can test nar-bridge slightly easier in an integration
   test.
 - Ensure we support connecting to unix sockets and grpc+http at least,
   using the same syntax as tvix-store.
 - Don't buffer the entire blob when rendering NAR

Co-Authored-By: Connor Brewster <cbrewster@hey.com>
Co-Authored-By: Márton Boros <martonboros@gmail.com>
Co-Authored-By: Vo Minh Thu <noteed@gmail.com>

Change-Id: I6064474e49dfe78cea67676957462d9f28658d4a
Reviewed-on: https://cl.tvl.fyi/c/depot/+/9339
Tested-by: BuildkiteCI
Reviewed-by: tazjin <tazjin@tvl.su>

1 files changed, 66 insertions, 0 deletions

diff --git a/tvix/nar-bridge/pkg/reader/hashers.go b/tvix/nar-bridge/pkg/reader/hashers.go
new file mode 100644
index 000000000000..7e98d2062d21
--- /dev/null
+++ b/tvix/nar-bridge/pkg/reader/hashers.go
@@ -0,0 +1,66 @@
+package reader
+
+import (
+	"errors"
+	"fmt"
+	"hash"
+	"io"
+)
+
+var _ io.Reader = &Hasher{}
+
+// Hasher wraps io.Reader.
+// You can ask it for the digest of the hash function used internally, and the
+// number of bytes written.
+type Hasher struct {
+	r         io.Reader
+	h         hash.Hash
+	bytesRead uint32
+}
+
+func NewHasher(r io.Reader, h hash.Hash) *Hasher {
+	return &Hasher{
+		r:         r,
+		h:         h,
+		bytesRead: 0,
+	}
+}
+
+func (h *Hasher) Read(p []byte) (int, error) {
+	nRead, rdErr := h.r.Read(p)
+
+	// write the number of bytes read from the reader to the hash.
+	// We need to do this independently on whether there's been error.
+	// n always describes the number of successfully written bytes.
+	nHash, hashErr := h.h.Write(p[0:nRead])
+	if hashErr != nil {
+		return nRead, fmt.Errorf("unable to write to hash: %w", hashErr)
+	}
+
+	// We assume here the hash function accepts the whole p in one Go,
+	// and doesn't early-return on the Write.
+	// We compare it with nRead and bail out if that was not the case.
+	if nHash != nRead {
+		return nRead, fmt.Errorf("hash didn't accept the full write")
+	}
+
+	// update bytesWritten
+	h.bytesRead += uint32(nRead)
+
+	if rdErr != nil {
+		if errors.Is(rdErr, io.EOF) {
+			return nRead, rdErr
+		}
+		return nRead, fmt.Errorf("error from underlying reader: %w", rdErr)
+	}
+
+	return nRead, hashErr
+}
+
+func (h *Hasher) BytesWritten() uint32 {
+	return h.bytesRead
+}
+
+func (h *Hasher) Sum(b []byte) []byte {
+	return h.h.Sum(b)
+}