Bug 269792 - 13.2-BETA2: sigILL: lzma does not compile to westmere
Summary: 13.2-BETA2: sigILL: lzma does not compile to westmere
Status: Closed Overcome By Events
Alias: None
Product: Base System
Classification: Unclassified
Component: bin (show other bugs)
Version: 13.1-STABLE
Hardware: amd64 Any
: --- Affects Only Me
Assignee: freebsd-bugs (Nobody)
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2023-02-24 00:23 UTC by Peter Much
Modified: 2024-11-17 03:41 UTC (History)
1 user (show)

See Also:


Attachments
pasting above went a bit wrong, here as attachment (4.27 KB, text/plain)
2023-02-24 00:34 UTC, Peter Much
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Peter Much 2023-02-24 00:23:50 UTC
Target is i3-540 (westmere)
Build machine is E5-2660v3 (haswell) - but that's not relevant.

Result is this:

# ./xz < /dev/zero > /dev/null                                            
Illegal instruction (core dumped)                                         
                                                                             
Wikipedia says that the i3-540 is westmere.
The compiler itself, when configured CPUTYPE?=native, says that
the chip is westmere (but it shows a lot of more detailed flags
in addition).
But when then compiling for cpu westmere, the code is defect.
(Compiling on-site with cpu "native", gives working code.)

Bottomline: one cannot currently compile for target westmere on a build machine.

$  /usr/bin/env -i /usr/bin/clang -v -x c -march=3Dnative -fsyntax-only -o =
/dev/null /dev/null 2>&1
FreeBSD clang version 14.0.5 (https://github.com/llvm/llvm-project.git llvm=
org-14.0.5-0-gc12386ae247c)
Target: x86_64-unknown-freebsd13.2
Thread model: posix
InstalledDir: /usr/bin
 (in-process)
 "/usr/bin/clang" -cc1 -triple x86_64-unknown-freebsd13.2 -fsyntax-only -di=
sable-free -clear-ast-before-backend -disable-llvm-verifier -discard-value-=
names -main-file-name null -mrelocation-model static -mframe-pointer=3Dall =
-ffp-contract=3Don -fno-rounding-math -mconstructor-aliases -funwind-tables=
=3D2 -target-cpu westmere -target-feature -avx512pf -target-feature -tsxldt=
rk -target-feature +cx16 -target-feature +sahf -target-feature -tbm -target=
-feature -avx512ifma -target-feature -sha -target-feature +crc32 -target-fe=
ature -fma4 -target-feature -vpclmulqdq -target-feature -prfchw -target-fea=
ture -bmi2 -target-feature -cldemote -target-feature -fsgsbase -target-feat=
ure -ptwrite -target-feature -amx-tile -target-feature -uintr -target-featu=
re -gfni -target-feature +popcnt -target-feature -widekl -target-feature -a=
es -target-feature -avx512bitalg -target-feature -movdiri -target-feature -=
xsaves -target-feature -avx512er -target-feature -avxvnni -target-feature -=
avx512fp16 -target-feature -avx512vnni -target-feature -amx-bf16 -target-fe=
ature -avx512vpopcntdq -target-feature -pconfig -target-feature -clwb -targ=
et-feature -avx512f -target-feature -xsavec -target-feature -clzero -target=
-feature -pku -target-feature +mmx -target-feature -lwp -target-feature -rd=
pid -target-feature -xop -target-feature -rdseed -target-feature -waitpkg -=
target-feature -kl -target-feature -movdir64b -target-feature -sse4a -targe=
t-feature -avx512bw -target-feature -clflushopt -target-feature -xsave -tar=
get-feature -avx512vbmi2 -target-feature +64bit -target-feature -avx512vl -=
target-feature -serialize -target-feature -hreset -target-feature -invpcid =
-target-feature -avx512cd -target-feature -avx -target-feature -vaes -targe=
t-feature -avx512bf16 -target-feature +cx8 -target-feature -fma -target-fea=
ture -rtm -target-feature -bmi -target-feature -enqcmd -target-feature -rdr=
nd -target-feature -mwaitx -target-feature +sse4.1 -target-feature +sse4.2 =
-target-feature -avx2 -target-feature +fxsr -target-feature -wbnoinvd -targ=
et-feature +sse -target-feature -lzcnt -target-feature -pclmul -target-feat=
ure -prefetchwt1 -target-feature -f16c -target-feature +ssse3 -target-featu=
re -sgx -target-feature -shstk -target-feature +cmov -target-feature -avx51=
2vbmi -target-feature -amx-int8 -target-feature -movbe -target-feature -avx=
512vp2intersect -target-feature -xsaveopt -target-feature -avx512dq -target=
-feature +sse2 -target-feature -adx -target-feature +sse3 -mllvm -treat-sca=
lable-fixed-error-as-warning -debugger-tuning=3Dgdb -v -fcoverage-compilati=
on-dir=3D/home/operator -resource-dir /usr/lib/clang/14.0.5 -fdebug-compila=
tion-dir=3D/home/operator -ferror-limit 19 -fgnuc-version=3D4.2.1 -faddrsig=
 -D__GCC_HAVE_DWARF2_CFI_ASM=3D1 -x c /dev/null
clang -cc1 version 14.0.5 based upon LLVM 14.0.5 default target x86_64-unkn=
own-freebsd13.2
#include "..." search starts here:
#include <...> search starts here:
 /usr/lib/clang/14.0.5/include
 /usr/include
End of search list.

$  /usr/bin/env -i /usr/bin/clang -v -x c -march=3Dwestmere -fsyntax-only -=
o /dev/null /dev/null 2>&1
FreeBSD clang version 14.0.5 (https://github.com/llvm/llvm-project.git llvm=
org-14.0.5-0-gc12386ae247c)
Target: x86_64-unknown-freebsd13.2
Thread model: posix
InstalledDir: /usr/bin
 (in-process)
 "/usr/bin/clang" -cc1 -triple x86_64-unknown-freebsd13.2 -fsyntax-only -di=
Comment 1 Peter Much 2023-02-24 00:34:27 UTC
Created attachment 240357 [details]
pasting above went a bit wrong, here as attachment
Comment 2 Dimitry Andric freebsd_committer freebsd_triage 2023-02-24 10:32:36 UTC
Can you get a core dump and/or disassembly? Because it isn't possible to guess what kind of instruction it crashed on (it could be anything).
Comment 3 Peter Much 2023-02-25 01:14:26 UTC
Sure I can make a coredump

(lldb) bt
* thread #1, name = 'xz', stop reason = signal SIGILL
  * frame #0: 0x0000206b53f99114 liblzma.so.5`lzma_stream_encoder_mt_memusage + 308
    frame #1: 0x0000206b53f98d67 liblzma.so.5`___lldb_unnamed_symbol445 + 663
    frame #2: 0x0000206b53f8dc70 liblzma.so.5`lzma_str_to_filters + 1440
    frame #3: 0x0000206b53f906d4 liblzma.so.5`lzma_filters_update + 212
    frame #4: 0x0000206b53f890d8 liblzma.so.5
    frame #5: 0x00002063314013be xz`___lldb_unnamed_symbol238 + 1534
    frame #6: 0x000020633140573c xz`___lldb_unnamed_symbol272 + 828
    frame #7: 0x00002063313ff602 xz
(lldb) disassemble
[...]
    0x206b53f990bd <+221>: callq  0x31360                   ; symbol stub for: lzma_outq_memusage
    0x206b53f990c2 <+226>: cmpq   $-0x1, %rax
    0x206b53f990c6 <+230>: je     0x1f10c                   ; <+300>
    0x206b53f990c8 <+232>: imulq  -0xf8(%rbp), %r15
    0x206b53f990d0 <+240>: movl   0x4(%rbx), %ecx
    0x206b53f990d3 <+243>: imulq  $0x1d8, %rcx, %rcx        ; imm = 0x1D8 
    0x206b53f990da <+250>: movq   $-0x81f1, %rdx            ; imm = 0xFFFF7E0F 
    0x206b53f990e1 <+257>: subq   %rcx, %rdx
    0x206b53f990e4 <+260>: cmpq   %r15, %rdx
    0x206b53f990e7 <+263>: jb     0x1f10c                   ; <+300>
    0x206b53f990e9 <+265>: imulq  -0xf8(%rbp), %r12
    0x206b53f990f1 <+273>: leaq   0x81f0(%r15,%rcx), %rcx
    0x206b53f990f9 <+281>: addq   %r12, %rcx
    0x206b53f990fc <+284>: jb     0x1f10c                   ; <+300>
    0x206b53f990fe <+286>: addq   %rax, %rcx
    0x206b53f99101 <+289>: movq   $-0x1, %r14
    0x206b53f99108 <+296>: cmovaeq %rcx, %r14
    0x206b53f9910c <+300>: movq   (%r13), %rax
    0x206b53f99110 <+304>: cmpq   -0x30(%rbp), %rax
->  0x206b53f99114 <+308>: jne    0x1f12b                   ; <+331>
    0x206b53f99116 <+310>: movq   %r14, %rax
    0x206b53f99119 <+313>: addq   $0xd8, %rsp
    0x206b53f99120 <+320>: popq   %rbx
    0x206b53f99121 <+321>: popq   %r12
    0x206b53f99123 <+323>: popq   %r13
    0x206b53f99125 <+325>: popq   %r14
    0x206b53f99127 <+327>: popq   %r15
    0x206b53f99129 <+329>: popq   %rbp
    0x206b53f9912a <+330>: retq   
    0x206b53f9912b <+331>: callq  0x30540                   ; symbol stub for: __stack_chk_fail

Is this suitable? (I don't get a clue of it)
Comment 4 Peter Much 2023-02-25 01:19:44 UTC
Please disregard prev. message (wrong binary)
Comment 5 Peter Much 2023-02-25 01:44:46 UTC
Now this looks more interesting (I don't do that every day)

* thread #1, name = 'xz', stop reason = signal SIGILL
  * frame #0: 0x0000042592c15114 liblzma.so.5`lzma_crc64 + 276
    frame #1: 0x0000042592c14d67 liblzma.so.5`lzma_check_update + 87
    frame #2: 0x0000042592c09c70 liblzma.so.5`___lldb_unnamed_symbol389 + 256
    frame #3: 0x0000042592c0c6d4 liblzma.so.5`___lldb_unnamed_symbol407 + 484
    frame #4: 0x0000042592c050d8 liblzma.so.5`lzma_code + 424
    frame #5: 0x0000041d7187a3be xz`___lldb_unnamed_symbol238 + 1534
    frame #6: 0x0000041d7187e73c xz`___lldb_unnamed_symbol272 + 828
    frame #7: 0x0000041d71878602 xz

    0x42592c150f7 <+247>: shlq   $0x4, %rsi
    0x42592c150fb <+251>: xorl   %eax, %eax
    0x42592c150fd <+253>: movq   %r8, %xmm0
    0x42592c15102 <+258>: movdqa -0x199ba(%rip), %xmm4
    0x42592c1510a <+266>: nopw   (%rax,%rax)
    0x42592c15110 <+272>: movdqa %xmm1, %xmm5
->  0x42592c15114 <+276>: pclmulqdq $0x0, %xmm0, %xmm5
    0x42592c1511a <+282>: pxor   %xmm3, %xmm5
    0x42592c1511e <+286>: pclmulqdq $0x11, %xmm4, %xmm1
    0x42592c15124 <+292>: pxor   %xmm5, %xmm1
    0x42592c15128 <+296>: movdqa 0x10(%rdi), %xmm3
    0x42592c1512d <+301>: addq   $0x10, %rdi

Googling... hmm...
https://www.intel.com/content/dam/www/public/us/en/documents/white-papers/carry-less-multiplication-instruction-in-gcm-mode-paper.pdf
"The Intel® PCLMULQDQ instruction is a new instruction available
beginning with the all new 2010 Intel® Core™ processor family
based on the 32nm Intel® microarchitecture codename Westmere."

CPU: Intel(R) Core(TM) i3 CPU         540  @ 3.07GHz (3059.12-MHz K8-class CPU)
  Origin="GenuineIntel"  Id=0x20655  Family=0x6  Model=0x25  Stepping=5
Features=0xbfebfbff<FPU,VME,DE,PSE,TSC,MSR,PAE,MCE,CX8,APIC,SEP,MTRR,PGE,MCA,CMOV,PAT,PSE36,CLFLUSH,DTS,ACPI,MMX,FXSR,SSE,SSE2,SS,HTT,TM,PBE>
Features2=0x9ae3bd<SSE3,DTES64,MON,DS_CPL,VMX,EST,TM2,SSSE3,CX16,xTPR,PDCM,PCID,SSE4.1,SSE4.2,POPCNT>
  AMD Features=0x28100800<SYSCALL,NX,RDTSCP,LM>
  AMD Features2=0x1<LAHF>
  Structured Extended Features3=0xc000000<IBPB,STIBP>
  VT-x: PAT,HLT,MTF,PAUSE,EPT,UG,VPID
  TSC: P-state invariant, performance statistics

 Launch Date Q1'10
 Intel® AES New Instructions No

Hm. So AES is only present in i7 series, not in Clarkdale and only part of Arrandale.
I knew the chip doesn't support AES, but didn't know that they mixed it up so horribly...
Comment 6 Mark Linimon freebsd_committer freebsd_triage 2024-11-17 03:41:52 UTC
^Triage: I'm sorry that this PR did not get addressed in a timely fashion.

By now, the version that it was created against is out of suppoprt.
Please re-open if it is still a problem on a supported version.