From: Cyrill Gorcunov <gorcunov@gmail.com> To: tml <tarantool-patches@freelists.org> Cc: Alexander Turenko <alexander.turenko@tarantool.org>, Kirill Yukhin <kyukhin@tarantool.org> Subject: [tarantool-patches] [PATCH 1/2] bugfix: fixed a segfault when unsinking 64-bit pointers. Date: Thu, 23 May 2019 20:48:24 +0300 [thread overview] Message-ID: <20190523174824.GF11013@uranus> (raw) In-Reply-To: <20190523174634.32314-1-gorcunov@gmail.com> From: Thibault Charbonnier <thibaultcha@me.com> The unsinking code was not using the correct layout for GC64 IR constants (value in adjacent slot) for this case. This patch is a derivative of https://github.com/raptorjit/raptorjit/pull/246 ported for LuaJIT itself. Fixed after an intense debugging session with @lukego. Co-authored-by: Luke Gorrie <lukego@gmail.com> --- src/lj_ir.h | 12 ++++++------ src/lj_snap.c | 2 +- 2 files changed, 7 insertions(+), 7 deletions(-) diff --git a/src/lj_ir.h b/src/lj_ir.h index 34c2785..3059bf6 100644 --- a/src/lj_ir.h +++ b/src/lj_ir.h @@ -560,6 +560,11 @@ typedef union IRIns { TValue tv; /* TValue constant (overlaps entire slot). */ } IRIns; +#define ir_isk64(ir) ((ir)->o == IR_KNUM || (ir)->o == IR_KINT64 || \ + (LJ_GC64 && \ + ((ir)->o == IR_KGC || \ + (ir)->o == IR_KPTR || (ir)->o == IR_KKPTR))) + #define ir_kgc(ir) check_exp((ir)->o == IR_KGC, gcref((ir)[LJ_GC64].gcr)) #define ir_kstr(ir) (gco2str(ir_kgc((ir)))) #define ir_ktab(ir) (gco2tab(ir_kgc((ir)))) @@ -567,12 +572,7 @@ typedef union IRIns { #define ir_kcdata(ir) (gco2cd(ir_kgc((ir)))) #define ir_knum(ir) check_exp((ir)->o == IR_KNUM, &(ir)[1].tv) #define ir_kint64(ir) check_exp((ir)->o == IR_KINT64, &(ir)[1].tv) -#define ir_k64(ir) \ - check_exp((ir)->o == IR_KNUM || (ir)->o == IR_KINT64 || \ - (LJ_GC64 && \ - ((ir)->o == IR_KGC || \ - (ir)->o == IR_KPTR || (ir)->o == IR_KKPTR)), \ - &(ir)[1].tv) +#define ir_k64(ir) check_exp(ir_isk64(ir), &(ir)[1].tv) #define ir_kptr(ir) \ check_exp((ir)->o == IR_KPTR || (ir)->o == IR_KKPTR, \ mref((ir)[LJ_GC64].ptr, void)) diff --git a/src/lj_snap.c b/src/lj_snap.c index 18ce715..7554caf 100644 --- a/src/lj_snap.c +++ b/src/lj_snap.c @@ -685,7 +685,7 @@ static void snap_restoredata(GCtrace *T, ExitState *ex, int32_t *src; uint64_t tmp; if (irref_isk(ref)) { - if (ir->o == IR_KNUM || ir->o == IR_KINT64) { + if (ir_isk64(ir)) { src = (int32_t *)&ir[1]; } else if (sz == 8) { tmp = (uint64_t)(uint32_t)ir->i; -- 2.20.1
next prev parent reply other threads:[~2019-05-23 17:48 UTC|newest] Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top 2019-05-23 17:46 [tarantool-patches] [PATCH 0/2] luajit2: " Cyrill Gorcunov 2019-05-23 17:48 ` Cyrill Gorcunov [this message] 2019-05-23 17:49 ` [tarantool-patches] [PATCH 2/2] test/luajit-tap: Add unsink_64_kptr.test.lua Cyrill Gorcunov 2019-05-23 17:50 ` [tarantool-patches] Re: [PATCH 0/2] luajit2: bugfix: fixed a segfault when unsinking 64-bit pointers Cyrill Gorcunov 2019-05-24 17:38 ` Alexander Turenko 2019-05-24 20:39 ` Cyrill Gorcunov 2019-05-27 7:12 ` Alexander Turenko 2019-05-28 2:02 ` Kirill Yukhin 2019-05-28 12:01 ` Alexander Turenko 2019-05-28 12:11 ` Cyrill Gorcunov 2019-05-28 12:14 ` Alexander Turenko
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=20190523174824.GF11013@uranus \ --to=gorcunov@gmail.com \ --cc=alexander.turenko@tarantool.org \ --cc=kyukhin@tarantool.org \ --cc=tarantool-patches@freelists.org \ --subject='Re: [tarantool-patches] [PATCH 1/2] bugfix: fixed a segfault when unsinking 64-bit pointers.' \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox