Re: [Tarantool-patches] [PATCH luajit] Prevent loop in snap_usedef().

[Maxim Kokryashkin via Tarantool-patches <tarantool-patches@dev.tarantool.org>]

Hi, Sergey!
Thanks for the patch!
Please consider my comments below.
On Tue, Mar 26, 2024 at 06:29:11PM +0300, Sergey Bronnikov wrote:
> From: Sergey Bronnikov <sergeyb@tarantool.org>
>
> Reported by XmiliaH.
>
> (cherry picked from commit 0e66fc96377853d898390f1a02723c54ec3a42f7)
>
> It is possible to get an infinite loop in a function `snap_usedef`
> when a `UCLO` makes a tight loop.
The description should include explanation for the cause of the issue
and should explain how it was resolved.
>
> Sergey Bronnikov:
> * added the description and the test for the problem
>
> Part of tarantool/tarantool#9595
> ---
> Branch: https://github.com/tarantool/luajit/tree/ligurio/lj-736-prevent-loop-in-snap_usedef
> Issues:
> - https://github.com/LuaJIT/LuaJIT/issues/736
> - https://github.com/tarantool/tarantool/issues/9595
>
>  src/lj_snap.c                                 |  7 ++-
>  .../lj-736-BC_UCLO-triggers-infinite-loop.lua | 59 +++++++++++++++++++
>  2 files changed, 65 insertions(+), 1 deletion(-)
>  create mode 100644 test/tarantool-tests/lj-736-BC_UCLO-triggers-infinite-loop.lua
>
> diff --git a/src/lj_snap.c b/src/lj_snap.c
> index 5a00b5cd..0710e1f0 100644
> --- a/src/lj_snap.c
> +++ b/src/lj_snap.c
> @@ -252,7 +252,12 @@ static BCReg snap_usedef(jit_State *J, uint8_t *udf,
>        BCReg minslot = bc_a(ins);
>        if (op >= BC_FORI && op <= BC_JFORL) minslot += FORL_EXT;
>        else if (op >= BC_ITERL && op <= BC_JITERL) minslot += bc_b(pc[-2])-1;
> -      else if (op == BC_UCLO) { pc += bc_j(ins); break; }
> +      else if (op == BC_UCLO) {
> +	ptrdiff_t delta = bc_j(ins);
> +	if (delta < 0) return maxslot;  /* Prevent loop. */
> +	pc += delta;
> +	break;
> +      }
>        for (s = minslot; s < maxslot; s++) DEF_SLOT(s);
>        return minslot < maxslot ? minslot : maxslot;
>        }
> diff --git a/test/tarantool-tests/lj-736-BC_UCLO-triggers-infinite-loop.lua b/test/tarantool-tests/lj-736-BC_UCLO-triggers-infinite-loop.lua
> new file mode 100644
> index 00000000..28a2b61b
> --- /dev/null
> +++ b/test/tarantool-tests/lj-736-BC_UCLO-triggers-infinite-loop.lua
> @@ -0,0 +1,59 @@
> +local tap = require('tap')
> +local test = tap.test('lj-736-BC_UCLO-triggers-infinite-loop'):skipcond({
> +  ['Test requires JIT enabled'] = not jit.status(),
> +})
> +
> +test:plan(1)
> +
> +-- Test reproduces an issue when BC_UCLO triggers an infinite loop.
> +-- See details in https://github.com/LuaJIT/LuaJIT/issues/736.
> +--
> +-- Listing below demonstrates a problem -
> +-- the bytecode UCLO on the line 13 makes a loop at 0013-0014:
> +--
> +-- - BYTECODE -- bc_uclo.lua:0-20
> +-- 0001    KPRI     0   0
> +-- 0002    FNEW     1   0      ; bc_uclo.lua:5
> +-- 0003    KSHORT   2   1
> +-- 0004    KSHORT   3   4
> +-- 0005    KSHORT   4   1
> +-- 0006    FORI     2 => 0011
> +-- 0007 => ISNEN    5   0      ; 2
> +-- 0008    JMP      6 => 0010
> +-- 0009    UCLO     0 => 0012
> +-- 0010 => FORL     2 => 0007
> +-- 0011 => UCLO     0 => 0012
> +-- 0012 => KPRI     0   0
> +-- 0013    UCLO     0 => 0012
> +-- 0014    FNEW     1   1      ; bc_uclo.lua:18
> +-- 0015    UCLO     0 => 0016
> +-- 0016 => RET0     0   1
> +
> +jit.opt.start('hotloop=1')
> +
> +do
> +  local uv = 0
> +  local w = function() return uv end -- luacheck: no unused
> +  for i = 1, 2 do
Add a comment that we have two iterations only because we only
need to record the trace.
> +    -- Infinite loop is here.
> +    if i == 2 then
> +      if i == 2 then
> +        goto pass
> +      end
> +      goto unreachable
> +    end
> +  end
> +end
> +
> +::unreachable::
> +-- Lua chunk below is required for reproducing a bug.
> +do
> +  local uv = 0 -- luacheck: no unused
> +  goto unreachable
> +  local w = function() return uv end -- luacheck: ignore
> +end
> +
> +::pass::
Please add a comment explaining why do we need a goto statement
and an unreachable code segment.
> +
> +test:ok(true, 'BC_UCLO does not trigger an infinite loop')
> +os.exit(test:check() and 0 or 1)
The test executes without any failures for the x86 non-GC64
build before the patch. The exact command:
$ cmake .. -DCMAKE_BUILD_TYPE=Debug -DLUA_USE_ASSERT=ON -DLUA_USE_APICHECK=ON && make -j && make test

> --
> 2.34.1
>