14-STABLE not working on an old CPU?

YuryG · Monday at 7:36 AM

I've tried to upgrade 13-STABLE to 14-STABLE on an old CORE2 CPU. Made a sequence of build (

Code:

git pull ; make buildworld ; make buildkernel ; make installkernel ; reboot ; make installworld

)from Handbook, but when it rebooted with the new world and the new kernel almost nothing works! "Illegal instruction" core files on clang or pkg executables etc.
Is it not supported already? Or what could be wrong? And what steps to solve the problem you can suggest?
STABLE/14-n26804-bb7b0eb7cf in the question

Crivens · Monday at 8:05 AM

What cpu exactly? And do you do run a 32 bit or 64 bit system?

YuryG · Monday at 8:28 AM

Crivens said:
What cpu exactly? And do you do run a 32 bit or 64 bit system?

Core 2 Duo E4600 (Allendale, pre SSE-4, but with SSSE-3).
64bit system.

Code:

CPU microcode: updated from 0xa3 to 0xa4
CPU: Intel(R) Core(TM)2 Duo CPU     E4600  @ 2.40GHz (2467.41-MHz K8-class CPU)
  Origin="GenuineIntel"  Id=0x6fd  Family=0x6  Model=0xf  Stepping=13
  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=0xe39d<SSE3,DTES64,MON,DS_CPL,EST,TM2,SSSE3,CX16,xTPR,PDCM>
  AMD Features=0x20100800<SYSCALL,NX,LM>
  AMD Features2=0x1<LAHF>
  TSC: P-state invariant, performance statistics

Charlie_ · Monday at 10:54 AM

Is this relevant?
SIMD enhancements for amd64
simd(7)

SirDice · Monday at 11:05 AM

YuryG said:
"Illegal instruction" core files on clang or pkg executables etc.

Does pkg bootstrap -f fix pkg(8)?

YuryG · Monday at 11:36 AM

SirDice said:
Does pkg bootstrap -f fix pkg(8)?

pkg does not work at all, crashes at start with "illegal instruction", so I couldn't bootstrap -f
Also I can't rebuild anything, since base clang behaves the same way.

YuryG · Monday at 11:39 AM

Charlie_ said:
Is this relevant?
SIMD enhancements for amd64
simd(7)

How is this relevant? ~~14th version is not mentioned,~~ more than that, it is claimed to be if a choice. This CPU doesn't have AVX or SSE4, but still it has enhancements till SSSE3.

SirDice · Monday at 11:39 AM

From simd(7):

Code:

If ARCHLEVEL is
	       unset, not recognised, or not supported by the  processor,  the
	       highest	level  of SIMD enhancements supported by the processor
	       is used.

That seems to suggest it won't use those instructions if they're not supported by the CPU. So I'm wondering why it apparently still used them.

What's in /etc/make.conf?

YuryG said:
14th version is not mentioned

Yes, it does, 14.1-RELEASE, which means it had to have been added to 14-STABLE before 14.1-RELEASE.

YuryG · Monday at 11:45 AM

SirDice said:
From simd(7):

Code:

If ARCHLEVEL is unset, not recognised, or not supported by the processor, the highest level of SIMD enhancements supported by the processor is used.

That seems to suggest it won't use those instructions if they're not supported by the CPU. So I'm wondering why it apparently still used them.

What's in /etc/make.conf?

Yes, it does, 14.1-RELEASE, which means it had to have been added to 14-STABLE before 14.1-RELEASE.

Yes, I've changed post.

Code:

CPUTYPE=?core2

in make.conf
No ARCHLEVEL is set.

YuryG · Monday at 11:50 AM

SirDice said:
From simd(7):

Code:

If ARCHLEVEL is unset, not recognised, or not supported by the processor, the highest level of SIMD enhancements supported by the processor is used.

That seems to suggest it won't use those instructions if they're not supported by the CPU. So I'm wondering why it apparently still used them.

What's in /etc/make.conf?

Yes, it does, 14.1-RELEASE, which means it had to have been added to 14-STABLE before 14.1-RELEASE.

(According to MAN, this CPU is able of "baseline" ARCHLEVEL.)

By the way, clang from 14.1 live dvd1 release works without illegal instruction.

Erichans · Monday at 12:55 PM

SirDice · Monday at 1:22 PM

This is my oldest CPU:

Code:

CPU: Intel(R) Core(TM) i5-3470 CPU @ 3.20GHz (3192.87-MHz K8-class CPU)
  Origin="GenuineIntel"  Id=0x306a9  Family=0x6  Model=0x3a  Stepping=9
  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=0x7fbae3ff<SSE3,PCLMULQDQ,DTES64,MON,DS_CPL,VMX,SMX,EST,TM2,SSSE3,CX16,xTPR,PDCM,PCID,SSE4.1,SSE4.2,x2APIC,POPCNT,TSCDLT,AESNI,XSAVE,OSXSAVE,AVX,F16C,RDRAND>
  AMD Features=0x28100800<SYSCALL,NX,RDTSCP,LM>
  AMD Features2=0x1<LAHF>
  Structured Extended Features=0x281<FSGSBASE,SMEP,ERMS>
  XSAVE Features=0x1<XSAVEOPT>
  VT-x: PAT,HLT,MTF,PAUSE,EPT,UG,VPID
  TSC: P-state invariant, performance statistics

It's running 14-STABLE. But it does have SSE4.2 and AVX. Don't have anything older to test.

T-Aoki · Monday at 1:26 PM

YuryG said:
pkg does not work at all, crashes at start with "illegal instruction", so I couldn't bootstrap -f

Doesn't pkg-static work, too?

SirDice · Monday at 1:49 PM

Is this ZFS? Did you create a boot environment before the upgrade? If you did I would revert back to 13-STABLE. At least until we figure out what happened.

_martin · Monday at 2:08 PM

YuryG said:
Or what could be wrong?

Can you share the coredump of any failing program? From that we'd see what the actual instruction it hit (and where).

Crivens · Monday at 3:51 PM

For program X failing

 gdb `which X` X.core

disassemble

This should give you the instruction which failed. If gdb does work, or lldb. If those fail you might need to take them from the base distribution.

cracauer@ · Monday at 3:53 PM

YuryG said:
I've tried to upgrade 13-STABLE to 14-STABLE on an old CORE2 CPU. Made a sequence of build (

Code:

git pull ; make buildworld ; make buildkernel ; make installkernel ; reboot ; make installworld

)from Handbook, but when it rebooted with the new world and the new kernel almost nothing works! "Illegal instruction" core files on clang or pkg executables etc.
Is it not supported already? Or what could be wrong? And what steps to solve the problem you can suggest?
STABLE/14-n26804-bb7b0eb7cf in the question

The way you chain your commands with ";" you wouldn't notice when one fails. It is better to do "make buildworld && ..." so that an error aborts the chain and the error message stays visible.

Maybe you kernel and userland went out of sync because of that.

malavon · Monday at 4:04 PM

YuryG said:
in make.conf

Quick correction: should be in /etc/src.conf
For reference dmesg from an E6600 (which is a year older than an E4600) on which I run 14.1-RELEASE built from source (with custom options). CPUTYPE is set to core2, but leaving it empty shouldn't be an issue. I have not built 14-STABLE, so there's always a very small chance something is different there. You can always try out 14.1-RELEASE just to be sure (after restoring clang from either a boot environment/zfs snapshot or an installation image).

Code:

CPU: Intel(R) Core(TM)2 CPU          6600  @ 2.40GHz (2404.14-MHz K8-class CPU)
  Origin="GenuineIntel"  Id=0x6f6  Family=0x6  Model=0xf  Stepping=6
  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=0xe3bd<SSE3,DTES64,MON,DS_CPL,VMX,EST,TM2,SSSE3,CX16,xTPR,PDCM>
  AMD Features=0x20100800<SYSCALL,NX,LM>
  AMD Features2=0x1<LAHF>
  VT-x: HLT,PAUSE
  TSC: P-state invariant, performance statistics

malavon · Monday at 4:07 PM

cracauer@ said:
The way you chain your commands with ";" you wouldn't notice when one fails. It is better to do "make buildworld && ..." so that an error aborts the chain and the error message stays visible.

This ^^. I also do a full clean before building, but if you're using meta mode that's probably not a good idea. I prefer a clean environment.
So basically I first make clean; make cleanworld, check if /usr/obj is indeed empty and only then do I start a build:
make -jxx buildworld && make -jxx kernel && make installworld. Note that I choose to live dangerously and update my entire world before rebooting, rebooting is safer.
Also check freebsd-update -kru after building, although I do not know what that will tell for 14-STABLE.

YuryG · Monday at 4:15 PM

cracauer@ said:
The way you chain your commands with ";" you wouldn't notice when one fails. It is better to do "make buildworld && ..." so that an error aborts the chain and the error message stays visible.

Maybe you kernel and userland went out of sync because of that.

I didn't chain it actually. I did steps one be one. And only after the last reboot I've got the problem.

YuryG · Monday at 4:20 PM

malavon said:
This ^^. I also do a full clean before building, but if you're using meta mode that's probably not a good idea. I prefer a clean environment.
So basically I first make clean; make cleanworld, check if /usr/obj is indeed empty and only then do I start a build:
make -jxx buildworld && make -jxx kernel && make installworld. Note that I choose to live dangerously and update my entire world before rebooting, rebooting is safer.
Also check freebsd-update -kru after building, although I do not know what that will tell for 14-STABLE.

Also it was a clean build. With make cleanworld after a new git clone

YuryG · Monday at 4:25 PM

SirDice said:
Is this ZFS? Did you create a boot environment before the upgrade? If you did I would revert back to 13-STABLE. At least until we figure out what happened.

No, not ZFS. So, somehow I will need to use release DVD1. Not sure how.

cracauer@ · Monday at 4:34 PM

YuryG said:
No, not ZFS. So, somehow I will need to use release DVD1. Not sure how.

Just untar the userland tarfile base.tar.gz.

Crivens · Monday at 5:42 PM

cracauer@ said:
Just untar the userland tarfile base.tar.gz.

Back up /etc before.

cracauer@ · Monday at 5:45 PM

Crivens said:
Back up /etc before.

I usually --exclude etc when untarring.

14-STABLE not working on an old CPU?

YuryG

Crivens

Administrator

YuryG

Charlie_

SirDice

Administrator

YuryG

YuryG

SirDice

Administrator

YuryG

YuryG

Erichans

SirDice

Administrator

T-Aoki

SirDice

Administrator

_martin

Crivens

Administrator

cracauer@

malavon

malavon

YuryG

YuryG

YuryG

cracauer@

Crivens

Administrator

cracauer@