From 3eb0ee59130f5c9bea6f4346b7b929c46c65bacb Mon Sep 17 00:00:00 2001 From: Mark Wielaard Date: Thu, 13 Jan 2022 13:31:27 +0100 Subject: [PATCH] Add valgrind-3.18.1-ppc-hwcaps.patch Resolves: #2037450 Valgrind segfaults during initialization on Power10 --- valgrind-3.18.1-ppc-hwcaps.patch | 136 +++++++++++++++++++++++++++++++ valgrind.spec | 7 ++ 2 files changed, 143 insertions(+) create mode 100644 valgrind-3.18.1-ppc-hwcaps.patch diff --git a/valgrind-3.18.1-ppc-hwcaps.patch b/valgrind-3.18.1-ppc-hwcaps.patch new file mode 100644 index 0000000..09b2714 --- /dev/null +++ b/valgrind-3.18.1-ppc-hwcaps.patch @@ -0,0 +1,136 @@ +commit 9d1d6cd6acc612cd94261956a8a94a6403a5d528 +Author: Will Schmidt +Date: Tue Jan 4 16:41:00 2022 -0600 + + Subject: Assorted changes to protect from side affects from the feature checking code. + + This problem was initially reported by Tulio, he assisted me in + identifying the underlying issue here. + + This was discovered on a Power10, and occurs since the ISA 3.1 support + check uses the brh instruction via a hardcoded ".long 0x7f1401b6" asm stanza. + That encoding writes to r20, and since the stanza does not contain a clobber + the compiler did not know to save or restore that register upon entry or exit. + The junk value remaining in r20 subsequently caused a segfault. + + This patch adds clobber masks to the instruction stanzas, as well as + updates the associated comments to clarify which registers are being + used. + As part of this change I've also + - updated the .long for the cnttzw instruction to write to r20, and + zeroed the reserved bits from that instruction so it is properly + decoded by the disassembler. + - updated the .long for the dadd instruction to write to f0. + + I've inspected the current codegen with these changes in place, and + confirm that r20 is now saved and restored on entry and exit from the + machine_get_hwcaps() function. + +diff --git a/coregrind/m_machine.c b/coregrind/m_machine.c +index 0b60ecc0fd44..a860ed67a334 100644 +--- a/coregrind/m_machine.c ++++ b/coregrind/m_machine.c +@@ -1244,11 +1244,11 @@ Bool VG_(machine_get_hwcaps)( void ) + /* Check for ISA 3.0 support. */ + have_isa_3_0 = True; + if (VG_MINIMAL_SETJMP(env_unsup_insn)) { + have_isa_3_0 = False; + } else { +- __asm__ __volatile__(".long 0x7d205434"); /* cnttzw RT, RB */ ++ __asm__ __volatile__(".long 00x7f140434"::"r20"); /* cnttzw r20,r24 */ + } + + // ISA 3.1 not supported on 32-bit systems + + /* determine dcbz/dcbzl sizes while we still have the signal +@@ -1356,79 +1356,79 @@ Bool VG_(machine_get_hwcaps)( void ) + /* Altivec insns */ + have_V = True; + if (VG_MINIMAL_SETJMP(env_unsup_insn)) { + have_V = False; + } else { +- __asm__ __volatile__(".long 0x10000484"); /*vor 0,0,0*/ ++ __asm__ __volatile__(".long 0x10000484"); /* vor v0,v0,v0 */ + } + + /* General-Purpose optional (fsqrt, fsqrts) */ + have_FX = True; + if (VG_MINIMAL_SETJMP(env_unsup_insn)) { + have_FX = False; + } else { +- __asm__ __volatile__(".long 0xFC00002C"); /*fsqrt 0,0*/ ++ __asm__ __volatile__(".long 0xFC00002C"); /* fsqrt f0,f0 */ + } + + /* Graphics optional (stfiwx, fres, frsqrte, fsel) */ + have_GX = True; + if (VG_MINIMAL_SETJMP(env_unsup_insn)) { + have_GX = False; + } else { +- __asm__ __volatile__(".long 0xFC000034"); /*frsqrte 0,0*/ ++ __asm__ __volatile__(".long 0xFC000034"); /* frsqrte f0,f0 */ + } + + /* VSX support implies Power ISA 2.06 */ + have_VX = True; + if (VG_MINIMAL_SETJMP(env_unsup_insn)) { + have_VX = False; + } else { +- __asm__ __volatile__(".long 0xf0000564"); /* xsabsdp XT,XB */ ++ __asm__ __volatile__(".long 0xf0000564"); /* xsabsdp vs0,vs0 */ + } + + /* Check for Decimal Floating Point (DFP) support. */ + have_DFP = True; + if (VG_MINIMAL_SETJMP(env_unsup_insn)) { + have_DFP = False; + } else { +- __asm__ __volatile__(".long 0xee4e8005"); /* dadd FRT,FRA, FRB */ ++ __asm__ __volatile__(".long 0xec0e8005"); /* dadd f0,f14,f16 */ + } + + /* Check for ISA 2.07 support. */ + have_isa_2_07 = True; + if (VG_MINIMAL_SETJMP(env_unsup_insn)) { + have_isa_2_07 = False; + } else { +- __asm__ __volatile__(".long 0x7c000166"); /* mtvsrd XT,RA */ ++ __asm__ __volatile__(".long 0x7c000166"); /* mtvsrd f0,r0 */ + } + + /* Check for ISA 3.0 support. */ + have_isa_3_0 = True; + if (VG_MINIMAL_SETJMP(env_unsup_insn)) { + have_isa_3_0 = False; + } else { +- __asm__ __volatile__(".long 0x7d205434"); /* cnttzw RT, RB */ ++ __asm__ __volatile__(".long 0x7f140434":::"r20"); /* cnttzw r20,r24 */ + } + + /* Check for ISA 3.1 support. */ + have_isa_3_1 = True; + if (VG_MINIMAL_SETJMP(env_unsup_insn)) { + have_isa_3_1 = False; + } else { +- __asm__ __volatile__(".long 0x7f1401b6"); /* brh RA, RS */ ++ __asm__ __volatile__(".long 0x7f1401b6":::"r20"); /* brh r20,r24 */ + } + + /* Check if Host supports scv instruction */ + have_scv_support = True; + if (VG_MINIMAL_SETJMP(env_unsup_insn)) { + have_scv_support = False; + } else { + /* Set r0 to 13 for the system time call. Don't want to make a random + system call. */ +- __asm__ __volatile__(".long 0x7c000278"); /* clear r0 */ +- __asm__ __volatile__(".long 0x6009000d"); /* set r0 to 13 */ +- __asm__ __volatile__(".long 0x44000001"); /* scv */ ++ __asm__ __volatile__(".long 0x7c000278"); /* clear r0 with xor r0,r0,r0 */ ++ __asm__ __volatile__(".long 0x6009000d"); /* set r0 to 13 with ori r9,r0,13 */ ++ __asm__ __volatile__(".long 0x44000001"); /* scv 0 */ + } + + /* determine dcbz/dcbzl sizes while we still have the signal + * handlers registered */ + find_ppc_dcbz_sz(&vai); diff --git a/valgrind.spec b/valgrind.spec index 4838ada..465ff8a 100644 --- a/valgrind.spec +++ b/valgrind.spec @@ -134,6 +134,9 @@ Patch19: valgrind-3.18.1-rseq-enosys.patch # KDE#444481 gdb_server test failures on s390x Patch20: valgrind-3.18.1-s390x-vdso.patch +# KDE#447995 Valgrind segfault on power10 due to hwcap checking code +Patch21: valgrind-3.18.1-ppc-hwcaps.patch + BuildRequires: make BuildRequires: glibc-devel @@ -286,6 +289,7 @@ Valgrind User Manual for details. %patch18 -p1 %patch19 -p1 %patch20 -p1 +%patch21 -p1 %build # LTO triggers undefined symbols in valgrind. Valgrind has a --enable-lto @@ -515,6 +519,9 @@ fi %endif %changelog +* Thu Jan 13 2022 Mark Wielaard +- Add valgrind-3.18.1-ppc-hwcaps.patch + * Tue Dec 14 2021 Mark Wielaard - 3.18.1-6 - Add valgrind-3.18.1-rseq-enosys.patch - Add valgrind-3.18.1-s390x-vdso.patch