Re: [gentoo-user] AMD microcode problem - Fam15h ( FYI )

[taii...@gmx.com]

On 05/26/2018 07:51 AM, Corbin Bird wrote:

> On 05/25/2018 08:50 PM, Adam Carter wrote:
>> > For me dmesg says;
>> > [    1.538275] microcode: CPU0: patch_level=0x06000852
>> >
>> > but i still have lwp in /proc/cpuinfo. Are you at 0x06000852 ?
>> .
>> This is my dmesg output :
>> .
>> [    1.111448] microcode: microcode updated early to new
>> patch_level=0x06000852
>>
>>
>> Ok then it looks like the mno-lwp is responsible for lwp's absence in
>> your /proc/cpuinfo.
>>
>> FWIW, no stability problems for me so far. (FX-8350 + 4.16.11).
>>
> .
> I should have clarified ... the '-mno-lwp' was added as a result of the
> comparison of the two /proc/cpuinfo files.I was very curious about WHAT
> exactly the microcode update did.
>
> The CPU I am using is a FX-9590.
>
> Question : Is there a PSP in your CPU?

Bulldozer/piledriver CPUs like the FX series and their corresponding
G34/C32 opterons don't have PSP - they are the last and best owner
controlled x86_64 CPU's and for now can still play the latest games at
max settings (even supports playing in a VM via IOMMU :D)
The 8350 and 93xx CPU's are pretty much the same the niners are just
better binned to support the uber OC's

You can have a 100% blob free owner controlled libre firmware
workstation/server with the KGPE-D16/KCMA-D8 which is the best option
for those who need to run x86 and thus can't get a talos or some other
power workstation.


0xDF372A17.asc
Description: application/pgp-keys

Re: [gentoo-user] AMD microcode problem - Fam15h ( FYI )

[Corbin Bird]

On 05/27/2018 06:12 AM, Adam Carter wrote:
> Also, are you saying that the sequence of events for you was;
> - booted with new microcode, and lwp was missing from /proc/cpuinfo
> - you had instability
> - you rebuilt with -mno-lwp
> - stability returned
.
- quickpackage'd 'sys-kernel/linux-firmware' v20180507
- cat'd the /proc/cpuinfo file to oldcpuinfo.text ( in home directory )
- updated 'sys-kernel/linux-firmware' from v20180507 to v20180518
- compilied the updated / new kernel released, and modules, then
installed them
- booted with new microcode
- cat'd the new /proc/cpuinfo to newcpuinfo.text ( in home directory )
and lwp was missing from /proc/cpuinfo
- began experiencing 'kernel hard locks'
- next day, updated kernel version released, added '-mno-lwp' to
'make.conf' CFLAGS, built it && installed it
- rebooted, stability returned

Corbin

Re: [gentoo-user] AMD microcode problem - Fam15h ( FYI )

[Adam Carter]

>
> I should have clarified ... the '-mno-lwp' was added as a result of the
> comparison of the two /proc/cpuinfo files. I was very curious about WHAT
> exactly the microcode update did.
>
> The CPU I am using is a FX-9590.
>
> Question : Is there a PSP in your CPU?
>

According to libreboot, only fam16 onwards have PSP, so if that's true, no.
AFAIK the 9590 is identical to the 8350 other than being clocked higher, so
i'm expecting our experience to be the same.

Also, are you saying that the sequence of events for you was;
- booted with new microcode, and lwp was missing from /proc/cpuinfo
- you had instability
- you rebuilt with -mno-lwp
- stability returned

Re: [gentoo-user] AMD microcode problem - Fam15h ( FYI )

[Corbin Bird]

On 05/25/2018 08:50 PM, Adam Carter wrote:
>
> > For me dmesg says;
> > [    1.538275] microcode: CPU0: patch_level=0x06000852
> >
> > but i still have lwp in /proc/cpuinfo. Are you at 0x06000852 ?
> .
> This is my dmesg output :
> .
> [    1.111448] microcode: microcode updated early to new
> patch_level=0x06000852
>
>
> Ok then it looks like the mno-lwp is responsible for lwp's absence in
> your /proc/cpuinfo.
>
> FWIW, no stability problems for me so far. (FX-8350 + 4.16.11).
>
.
I should have clarified ... the '-mno-lwp' was added as a result of the
comparison of the two /proc/cpuinfo files.I was very curious about WHAT
exactly the microcode update did.

The CPU I am using is a FX-9590.

Question : Is there a PSP in your CPU?

Corbin

Re: [gentoo-user] AMD microcode problem - Fam15h ( FYI )

[Adam Carter]

>
> > For me dmesg says;
> > [1.538275] microcode: CPU0: patch_level=0x06000852
> >
> > but i still have lwp in /proc/cpuinfo. Are you at 0x06000852 ?
> .
> This is my dmesg output :
> .
> [1.111448] microcode: microcode updated early to new
> patch_level=0x06000852
>

Ok then it looks like the mno-lwp is responsible for lwp's absence in your
/proc/cpuinfo.

FWIW, no stability problems for me so far. (FX-8350 + 4.16.11).

Re: [gentoo-user] AMD microcode problem - Fam15h ( FYI )

[Corbin Bird]

On 05/24/2018 05:32 PM, Adam Carter wrote:
>
> For me dmesg says;
> [    1.538275] microcode: CPU0: patch_level=0x06000852
>
> but i still have lwp in /proc/cpuinfo. Are you at 0x06000852 ?
.
This is my dmesg output :
.
[    1.111448] microcode: microcode updated early to new
patch_level=0x06000852
[    1.111574] microcode: CPU0: patch_level=0x06000852
[    1.111691] microcode: CPU1: patch_level=0x06000852
[    1.111809] microcode: CPU2: patch_level=0x06000852
[    1.111925] microcode: CPU3: patch_level=0x06000852
[    1.112117] microcode: CPU4: patch_level=0x06000852
[    1.112236] microcode: CPU5: patch_level=0x06000852
[    1.112354] microcode: CPU6: patch_level=0x06000852
[    1.112469] microcode: CPU7: patch_level=0x06000852
[    1.112609] microcode: Microcode Update Driver: v2.01
, Peter Oruba
.
current /proc/cpuinfo ( kernel v4.9.102 ):
--
processor    : 0
vendor_id    : AuthenticAMD
cpu family    : 21
model        : 2
model name    : AMD FX(tm)-9590 Eight-Core Processor
stepping    : 0
microcode    : 0x6000852
.
flags        : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt
pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc extd_apicid
aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2
popcnt aes xsave avx f16c lahf_lm cmp_legacy svm extapic cr8_legacy abm
sse4a misalignsse 3dnowprefetch osvw ibs xop skinit wdt fma4 tce
nodeid_msr tbm topoext perfctr_core perfctr_nb cpb hw_pstate ssbd ibpb
vmmcall bmi1 amd_ibpb arat npt lbrv svm_lock nrip_save tsc_scale
vmcb_clean flushbyasid decodeassists pausefilter pfthreshold
.
bugs        : fxsave_leak sysret_ss_attrs null_seg spectre_v1 spectre_v2
spec_store_bypass
.
old /proc/cpuinfo : ( pre-microcode update / kernel v4.9.99 )
--
processor    : 0
vendor_id    : AuthenticAMD
cpu family    : 21
model        : 2
model name    : AMD FX(tm)-9590 Eight-Core Processor
stepping    : 0
microcode    : 0x600084f
.
flags        : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt
pdpe1gb rdtscp lm constant_tsc rep_good nopl nonstop_tsc extd_apicid
aperfmperf eagerfpu pni pclmulqdq monitor ssse3 fma cx16 sse4_1 sse4_2
popcnt aes xsave avx f16c lahf_lm cmp_legacy svm extapic cr8_legacy abm
sse4a misalignsse 3dnowprefetch osvw ibs xop skinit wdt lwp fma4 tce
nodeid_msr tbm topoext perfctr_core perfctr_nb cpb hw_pstate vmmcall
bmi1 arat npt lbrv svm_lock nrip_save tsc_scale vmcb_clean flushbyasid
decodeassists pausefilter pfthreshold
.
bugs        : fxsave_leak sysret_ss_attrs null_seg spectre_v1 spectre_v2
.
Just noticed the "ssbd" flag ADDED in the new /proc/cpuinfo.
.
Corbin

Re: [gentoo-user] AMD microcode problem - Fam15h ( FYI )

[Adam Carter]

On Thu, May 24, 2018 at 12:31 PM, Corbin Bird 
wrote:

> FYI :
>
> Did the microcode update, compiled and installed the 4.9.101 kernel with
> it.
> After rebooting, I started having kernel hard lock problems.
>
> Before compiling / installing kernel 4.9.102, I made an addition to the
> CFLAGS var in make.conf : -mno-lwp
> ( 'lwp' is 'Light Weight Profiling' )
>
> No kernel hard locks after the reboot.
>
> Lesson Learned :
> The kernel needs to be compiled, installed, system rebooted,
> re-compiled, re-installed and re-booted again.
> Otherwise the kernel compile process doesn't adjust to the removal of
> instruction sets.
> The 'make.conf' entry is to catch any other uses, system wide.
>
>
> Config Info :
>
> Using 'sys-kernel/gentoo-sources' USE flags : experimental modules symlink
>
> -march=bdver2
>
> kernel .config settings :
> # CONFIG_PROFILING is not set
> CONFIG_HAVE_OPROFILE=y
> CONFIG_OPROFILE_NMI_TIMER=y
> CONFIG_KPROBES=y
>
>
For me dmesg says;
[1.538275] microcode: CPU0: patch_level=0x06000852

but i still have lwp in /proc/cpuinfo. Are you at 0x06000852 ?

[gentoo-user] AMD microcode problem - Fam15h ( FYI )

[Corbin Bird]

FYI :

Did the microcode update, compiled and installed the 4.9.101 kernel with it.
After rebooting, I started having kernel hard lock problems.

Before compiling / installing kernel 4.9.102, I made an addition to the
CFLAGS var in make.conf : -mno-lwp
( 'lwp' is 'Light Weight Profiling' )

No kernel hard locks after the reboot.

Lesson Learned :
The kernel needs to be compiled, installed, system rebooted,
re-compiled, re-installed and re-booted again.
Otherwise the kernel compile process doesn't adjust to the removal of
instruction sets.
The 'make.conf' entry is to catch any other uses, system wide.


Config Info :

Using 'sys-kernel/gentoo-sources' USE flags : experimental modules symlink

-march=bdver2

kernel .config settings :
# CONFIG_PROFILING is not set
CONFIG_HAVE_OPROFILE=y
CONFIG_OPROFILE_NMI_TIMER=y
CONFIG_KPROBES=y


Corbin