From 0a04cc0bcc3fe067b92feb40a7467a733df686b1 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Reynir=20Bj=C3=B6rnsson?= <reynir@reynir.dk>
Date: Fri, 24 Jan 2025 20:34:20 +0100
Subject: [PATCH] Drop owee dependency and work on strings

---
 bin/main.ml           |  30 ++++++--
 lib/elf.ml            | 165 ++++++++++++++++++++++++++++++++++++++++++
 lib/solo5_elftool.ml  |  54 ++------------
 lib/solo5_elftool.mli |   4 +-
 4 files changed, 200 insertions(+), 53 deletions(-)
 create mode 100644 lib/elf.ml

diff --git a/bin/main.ml b/bin/main.ml
index c30478a..49adfaa 100644
--- a/bin/main.ml
+++ b/bin/main.ml
@@ -1,13 +1,33 @@
+let read_binary file =
+  let ic = open_in_bin file in
+  let res = Buffer.create 16384 in
+  let buf = Bytes.create 16384 in
+  let rec loop () =
+    let len = input ic buf 0 16384 in
+    if len > 0 then
+      let () = Buffer.add_subbytes res buf 0 len in
+      loop ()
+  in
+  loop ();
+  close_in_noerr ic;
+  Buffer.contents res
+
 let query_manifest file =
-  Owee_buf.map_binary file
+  read_binary file
   |> Solo5_elftool.query_manifest
-  |> Result.iter (fun mft ->
-      Fmt.pr "%a\n" Solo5_elftool.pp_mft mft)
+  |> Result.fold
+    ~ok:(fun mft ->
+        Fmt.pr "%a\n" Solo5_elftool.pp_mft mft)
+    ~error:(fun (`Msg e) ->
+        Fmt.epr "%s\n" e)
 
 let query_abi file =
-  Owee_buf.map_binary file
+  read_binary file
   |> Solo5_elftool.query_abi
-  |> Result.iter (Fmt.pr "%a\n" Solo5_elftool.pp_abi)
+  |> Result.fold
+    ~ok:(fun abi -> Fmt.pr "%a\n" Solo5_elftool.pp_abi abi)
+    ~error:(fun (`Msg e) ->
+        Fmt.epr "%s\n" e)
 
 let file =
   let doc = "Solo5 executable" in
diff --git a/lib/elf.ml b/lib/elf.ml
new file mode 100644
index 0000000..ee89d16
--- /dev/null
+++ b/lib/elf.ml
@@ -0,0 +1,165 @@
+exception Elf_error
+
+(* only the bits we care about *)
+type header = {
+  e_shoff : int;
+  e_shentsize : int;
+  e_shnum : int;
+  e_shstrndx : int;
+}
+
+type section = {
+  sh_offset : int;
+  sh_size : int;
+  sh_name_off : int;
+  sh_name : string;
+}
+
+let section_manifest = ".note.solo5.manifest"
+let section_abi = ".note.solo5.abi"
+let note_name = "Solo5"
+let typ_mft1 = 0x3154464d
+let typ_abi1 = 0x31494241
+
+let get_uint16 = function
+  | `LE -> String.get_uint16_le
+  | `BE -> String.get_uint16_be
+
+let get_uint32 en s off =
+  let get = match en with
+    | `LE -> String.get_int32_le
+    | `BE -> String.get_int32_be
+  in
+  Int32.to_int (get s off) land 0xFFFF_FFFF
+
+let get_uint64 en s off =
+  let get = match en with
+    | `LE -> String.get_int64_le
+    | `BE -> String.get_int64_be
+  in
+  match Int64.unsigned_to_int (get s off) with
+  | None -> raise Elf_error
+  | Some n -> n
+
+let c_string s off maxlen =
+  let rec scan_c_string i =
+    if String.length s < off + i || i = maxlen then
+      raise Elf_error
+    else if s.[i+off] = '\000' then
+      i
+    else
+      scan_c_string (succ i)
+  in
+  String.sub s off (scan_c_string 0)
+
+let read_magic s off =
+  if String.length s < off + 4 then
+    raise Elf_error;
+  let valid =
+    String.get_uint8 s off = 0x7f &&
+    s.[off+1] = 'E' && s.[off+2] = 'L' && s.[off+3] = 'F'
+  in
+  if not valid then
+    raise Elf_error;
+  off+4
+
+let elfclass64 = 2
+
+let read_identification s off =
+  if String.length s < off + 12 then
+    raise Elf_error;
+  let elf_class = String.get_uint8 s off in
+  let elf_data = String.get_uint8 s (off+1) in
+  let _elf_version = String.get_uint8 s (off+2) in
+  let _elf_osabi = String.get_uint8 s (off+3) in
+  let _elf_abiversion = String.get_uint8 s (off+4) in
+  (* Check padding *)
+  for i = off + 5 to off+11 do
+    if s.[i] <> '\000' then
+      raise Elf_error
+  done;
+  (* we only support ELFCLASS64 *)
+  if elf_class <> elfclass64 then
+    raise Elf_error;
+  let endianness =
+    match elf_data with
+    | 1 -> `LE
+    | 2 -> `BE
+    | _ -> raise Elf_error
+  in
+  endianness, off+12
+
+let read_header en s =
+  if String.length s < 16 + 48 then
+    raise Elf_error;
+  let e_shoff = get_uint32 en s 0x28 in
+  let e_shentsize = get_uint16 en s 0x3a in
+  let e_shnum = get_uint16 en s 0x3c in
+  let e_shstrndx = get_uint16 en s 0x3e in
+  if Sys.int_size <= 32 then
+    raise Elf_error;
+  { e_shoff; e_shentsize; e_shnum; e_shstrndx }
+
+let read_section en s hdr i =
+  let off = hdr.e_shoff + i * hdr.e_shentsize in
+  if String.length s < off + 64 then
+    raise Elf_error;
+  let sh_name_off = get_uint32 en s off in
+  let sh_offset = get_uint64 en s (off + 24) in
+  let sh_size = get_uint64 en s (off + 32) in
+  { sh_name_off; sh_offset; sh_size; sh_name = "" }
+
+let read_section_name shstrndx s section =
+  let off = shstrndx.sh_offset + section.sh_name_off in
+  if String.length s < off + 1 then
+    raise Elf_error;
+  c_string s off (shstrndx.sh_size - section.sh_name_off)
+
+let read_sections en s hdr =
+  let sections = Array.init hdr.e_shnum (read_section en s hdr) in
+  let shstrndx = sections.(hdr.e_shstrndx) in
+  Array.map
+    (fun section -> { section with sh_name = read_section_name shstrndx s section })
+    sections
+
+let find_section sections name =
+  Array.find_opt
+    (fun section -> String.equal section.sh_name name)
+    sections
+
+let section_body s section =
+  if section.sh_offset < 0 || String.length s < section.sh_offset + section.sh_size then
+    raise Elf_error;
+  String.sub s section.sh_offset section.sh_size
+
+let desc en section_body ~expected_owner ~expected_type =
+  if String.length section_body < 12 then
+    raise Elf_error;
+  let namesz = get_uint32 en section_body 0 in
+  let descsz = get_uint32 en section_body 4
+  and typ = get_uint32 en section_body 8 in
+  if String.length section_body < 12 + namesz + descsz then
+    raise Elf_error;
+  if typ <> expected_type ||
+     String.length expected_owner + 1 <> namesz ||
+     not (String.equal
+            (expected_owner ^ "\000")
+            (String.sub section_body 12 namesz))
+  then
+    None
+  else
+    let off = 12 + namesz in
+    (* padding *)
+    let off = off + ((4 - (off land 3)) land 3) in
+    Some (String.sub section_body off descsz)
+
+let find s section_name typ =
+  let off = read_magic s 0 in
+  let en, _off = read_identification s off in
+  let hdr = read_header en s in
+  let sections = read_sections en s hdr in
+  match find_section sections section_name with
+  | None -> None
+  | Some section ->
+    let body = section_body s section in
+    desc en body ~expected_owner:note_name ~expected_type:typ
diff --git a/lib/solo5_elftool.ml b/lib/solo5_elftool.ml
index 181579b..8bf26e7 100644
--- a/lib/solo5_elftool.ml
+++ b/lib/solo5_elftool.ml
@@ -160,52 +160,14 @@ let parse_abi buf =
   (* XXX: should we check version = 1l ? *)
   Ok { target; version }
 
-let ( let* ) = Result.bind
-
-let note_name = "Solo5"
-let typ_mft1 = 0x3154464d
-let typ_abi1 = 0x31494241
-
-let query_manifest_exn buf =
-  let _header, sections = Owee_elf.read_elf buf in
-  let* section =
-    Owee_elf.find_section sections ".note.solo5.manifest"
-    |> Option.to_result ~none:(`Msg "section .note.solo5.manifest not found")
-  in
-  let body = Owee_elf.section_body buf section in
-  let cursor = Owee_buf.cursor body in
-  let descsz =
-    Owee_elf_notes.read_desc_size cursor
-      ~expected_owner:note_name
-      ~expected_type:typ_mft1
-  in
-  let desc = Owee_buf.Read.fixed_string cursor descsz in
-  let* () = guard "extra data" (Owee_buf.at_end cursor) in
-  parse_mft desc
-
 let query_manifest buf =
-  try query_manifest_exn buf with
-  | Out_of_memory -> raise Out_of_memory
-  | e -> Error (`Msg ("query manifest failure: " ^ Printexc.to_string e))
-
-let query_abi_exn buf =
-  let _header, sections = Owee_elf.read_elf buf in
-  let* section =
-    Owee_elf.find_section sections ".note.solo5.abi"
-    |> Option.to_result ~none:(`Msg "section .note.solo5.abi not found")
-  in
-  let body = Owee_elf.section_body buf section in
-  let cursor = Owee_buf.cursor body in
-  let descsz =
-    Owee_elf_notes.read_desc_size cursor
-      ~expected_owner:note_name
-      ~expected_type:typ_abi1
-  in
-  let desc = Owee_buf.Read.fixed_string cursor descsz in
-  let* () = guard "extra data" (Owee_buf.at_end cursor) in
-  parse_abi desc
+  match Elf.find buf Elf.section_manifest Elf.typ_mft1 with
+  | None -> Error (`Msg "manifest not found")
+  | Some desc -> parse_mft desc
+  (*| exception Elf.Elf_error -> Error (`Msg "error during ELF parsing")*)
 
 let query_abi buf =
-  try query_abi_exn buf with
-  | Out_of_memory -> raise Out_of_memory
-  | e -> Error (`Msg ("query abi failure: " ^ Printexc.to_string e))
+  match Elf.find buf Elf.section_abi Elf.typ_abi1 with
+  | None -> Error (`Msg "manifest not found")
+  | Some desc -> parse_abi desc
+  (*| exception Elf.Elf_error -> Error (`Msg "error during ELF parsing")*)
diff --git a/lib/solo5_elftool.mli b/lib/solo5_elftool.mli
index 289f1a3..f66c60a 100644
--- a/lib/solo5_elftool.mli
+++ b/lib/solo5_elftool.mli
@@ -36,8 +36,8 @@ val pp_abi : Format.formatter -> abi -> unit
 (** Pretty-prints the manifest as JSON in a similar style as the Solo5 command
  * line tool {[solo5-elftool query-abi]}. *)
 
-val query_manifest : Owee_buf.t -> (mft, [> `Msg of string ]) result
+val query_manifest : string -> (mft, [> `Msg of string ]) result
 (** [query_manifest buf] is the solo5 manifest of [buf], or an error message. *)
 
-val query_abi : Owee_buf.t -> (abi, [> `Msg of string ]) result
+val query_abi : string -> (abi, [> `Msg of string ]) result
 (** [query_abi buf] is the solo5 abi of [buf], or an error message. *)