Stale parser->previous after parse_heredoc causes unsigned location length underflow on declaration line

Aberta
  • Algo errado?
#4,359 0 comentários 0 reações 0 responsáveis Ver no GitHub

Mantenedores costumam responder em até 1 dia

Ninguém assumiu esta issue ainda.

Avaliação

Dificuldade
4/5
Tempo estimado
1-2 dias
Facilidade para iniciantes
55/100
Tipo de issue
Bug
Clareza
Claramente especificada
Status de atividade
Ativa
Stack de tecnologia
c, ruby
Domínio
backend, compilers

Direção de pesquisa

Start in the parser code around parse_heredoc and parse_ternary, following how parser->current and parser->previous change after consuming PM_TOKEN_HEREDOC_END. Reproduce with ./bin/parse -e "//?<<0 a " and inspect the AST location for the resulting ElseNode. Done means the malformed input no longer creates a location whose start is after its end or an underflowed length.

Escrita pelo modelo de indexação a partir do texto da issue.

Descrição

Fuzzing find on bdbd482dc.

When parse_heredoc finishes parsing a heredoc body, it rewinds parser->current back to the declaration line (immediately after the heredoc opening token, e.g. <<0), but leaves parser->previous set to PM_TOKEN_HEREDOC_END at the end of the heredoc body (at a higher byte offset than parser->current).

If the caller then synthesizes a missing token via expect1 (which uses parser->previous.end) before consuming another token on the declaration line, the synthesized token is given a byte offset at the end of the heredoc body while subsequent tokens on the declaration line have smaller byte offsets, causing start > end and underflowing uint32_t location.length in the enclosing AST node.

Reproduction

Simulation here: https://stevenjohnstone.github.io/prism-loupe/?version=v2.1.0&code=Ly8%2FPDwwIGEg.

✗ ./bin/parse -e "//?<<0 a "
Errors:
[#<Prism::ParseError @type=:heredoc_term @message="unterminated heredoc; can't find string \"0\" anywhere before EOF" @location=#<Prism::Location @start_offset=5 @length=1 start_line=1> @level=:syntax>,
 #<Prism::ParseError @type=:heredoc_term @message="unterminated heredoc; can't find string \"0\" anywhere before EOF" @location=#<Prism::Location @start_offset=5 @length=1 start_line=1> @level=:syntax>,
 #<Prism::ParseError @type=:ternary_colon @message="expected a `:` after the true expression of a ternary operator" @location=#<Prism::Location @start_offset=9 @length=0 start_line=1> @level=:syntax>]

AST:
@ ProgramNode (location: (1,0)-(1,8))
├── flags: ∅
├── locals: []
└── statements:
    @ StatementsNode (location: (1,0)-(1,8))
    ├── flags: ∅
    └── body: (length: 1)
        └── @ IfNode (location: (1,0)-(1,8))
            ├── flags: newline
            ├── if_keyword_loc: ∅
            ├── predicate:
            │   @ MatchLastLineNode (location: (1,0)-(1,2))
            │   ├── flags: static_literal, forced_us_ascii_encoding
            │   ├── opening_loc: (1,0)-(1,1) = "/"
            │   ├── content_loc: (1,1)-(1,1) = ""
            │   ├── closing_loc: (1,1)-(1,2) = "/"
            │   └── unescaped: ""
            ├── then_keyword_loc: (1,2)-(1,3) = "?"
            ├── statements:
            │   @ StatementsNode (location: (1,3)-(1,6))
            │   ├── flags: ∅
            │   └── body: (length: 1)
            │       └── @ StringNode (location: (1,3)-(1,6))
            │           ├── flags: newline
            │           ├── opening_loc: (1,3)-(1,6) = "<<0"
            │           ├── content_loc: (1,9)-(1,9) = ""
            │           ├── closing_loc: ∅
            │           └── unescaped: ""
            ├── subsequent:
            │   @ ElseNode (location: (1,9)-(1,4294967304))
            │   ├── flags: ∅
            │   ├── else_keyword_loc: (1,9)-(1,9) = ""
            │   ├── statements:
            │   │   @ StatementsNode (location: (1,7)-(1,8))
            │   │   ├── flags: ∅
            │   │   └── body: (length: 1)
            │   │       └── @ CallNode (location: (1,7)-(1,8))
            │   │           ├── flags: newline, variable_call, ignore_visibility
            │   │           ├── receiver: ∅
            │   │           ├── call_operator_loc: ∅
            │   │           ├── name: :a
            │   │           ├── message_loc: (1,7)-(1,8) = "a"
            │   │           ├── opening_loc: ∅
            │   │           ├── arguments: ∅
            │   │           ├── closing_loc: ∅
            │   │           ├── equal_loc: ∅
            │   │           └── block: ∅
            │   └── end_keyword_loc: ∅
            └── end_keyword_loc: ∅

The ElseNode created in parse_ternary is given start = 9 (EOF) and end = 8 (end of a), wrapping uint32_t location.length (U32(end - start)) to 4294967295 (0xFFFFFFFF). Ruby adds then calculates end_offset=9+4294967295=4294967304(232+8).

Root Cause

  1. In "//?<<0 a ", //? starts a ternary conditional expression (parse_ternary).
  2. The true-branch expression is the heredoc <<0 (offsets 3..6).
  3. parse_heredoc consumes PM_TOKEN_HEREDOC_START (opening, offsets 3..6), lexes the heredoc body up to EOF (offset 9), and consumes PM_TOKEN_HEREDOC_END at 9..9.
  4. Consuming PM_TOKEN_HEREDOC_END sets parser->previous to 9..9, while rewinding parser->current back to offset 6 on line 1 (PM_TOKEN_IDENTIFIER "a" at 7..8).
  5. Back in parse_ternary, the parser expects : (PM_TOKEN_COLON) after the true-branch <<0: 
    expect1(parser, PM_TOKEN_COLON, PM_ERR_TERNARY_COLON);
pm_token_t colon = parser->previous;
    Because parser->current is "a" (7..8) rather than :, expect1 fails and synthesizes the missing : token at parser->previous.end — which is still 9 (PM_TOKEN_HEREDOC_END at EOF) instead of 6 (the end of <<0).
  6. Next, parse_ternary parses the false-branch expression "a" at 7..8 and calls pm_else_node_create spanning from colon.start (9) to the end of "a" (8). Since 9 > 8, end - start underflows uint32_t to 4294967295.

Impact

  • Any syntax error immediately following a heredoc on the same declaration line that synthesizes a missing token via expect1 (such as a missing : in a ternary expression, a missing ) or ], or a missing => in a hash) produces an AST node whose start is at the end of the heredoc body and whose end is on the earlier declaration line (start > end), underflowing uint32_t location.length to ~4 GB and causing out-of-bounds heap reads in C consumers such as pm_prettyprint.
Linguagem predominante
C
Estrelas
1k
Forks
202
Merge médio
5h 55min
PRs com merge (30d)
85

Preparar o ambiente

Primeiros passos

  1. Leia a issue inteira e depois o guia de contribuição do projeto.
  2. Comente na issue dizendo que vai assumir — evita que duas pessoas façam o mesmo trabalho.
  3. Faça um fork do repositório e trabalhe em uma branch.
  4. Abra um pull request que referencie o número da issue.

Mais de ruby/prism

Todas as issues de ruby/prism

Issues semelhantes

Mais issues de C

Receba novas issues na sua caixa de entrada

Um resumo curto de issues do GitHub para quem está começando.