El 22/08/18 a las 06:24, Christian Hesse escribió:
> Gerardo Exequiel Pozzi via arch-releng  on Sat,
> 2018/08/18 23:29:
>> +# Prepare amd_ucode
>> +make_amd_ucode_img() {
>> +    mkdir -p ${work_dir}/amd-ucode/kernel/x86/microcode
>> +    cat
>> ${work_dir}/x86_64/airootfs/lib/firmware/amd-ucode/microcode_amd*.bin >
>> ${work_dir}/amd-ucode/kernel/x86/microcode/AuthenticAMD.bin
>> +    pushd ${work_dir}/amd-ucode > /dev/null
>> +    echo kernel/x86/microcode/AuthenticAMD.bin | bsdcpio -o -H newc -R 0:0
>>> amd-ucode.img
>> +    popd > /dev/null
>> +    cp ${work_dir}/amd-ucode/amd-ucode.img
>> ${work_dir}/x86_64/airootfs/boot/amd-ucode.img +}
> 
> Wondering if we should put this into a package... Could be a split package of
> linux-firmware. Any opinion on this?
> 

Yes, split package from linux-firmware should be the best way.

Christian Hesse  on Wed, 2018/08/22 11:24:
> Gerardo Exequiel Pozzi via arch-releng  on Sat,
> 2018/08/18 23:29:
> > +# Prepare amd_ucode
> > +make_amd_ucode_img() {
> > +    mkdir -p ${work_dir}/amd-ucode/kernel/x86/microcode
> > +    cat
> > ${work_dir}/x86_64/airootfs/lib/firmware/amd-ucode/microcode_amd*.bin >
> > ${work_dir}/amd-ucode/kernel/x86/microcode/AuthenticAMD.bin
> > +    pushd ${work_dir}/amd-ucode > /dev/null
> > +    echo kernel/x86/microcode/AuthenticAMD.bin | bsdcpio -o -H newc -R
> > 0:0  
> > > amd-ucode.img  
> > +    popd > /dev/null
> > +    cp ${work_dir}/amd-ucode/amd-ucode.img
> > ${work_dir}/x86_64/airootfs/boot/amd-ucode.img +}  
> 
> Wondering if we should put this into a package... Could be a split package
> of linux-firmware. Any opinion on this?

[ Added Laurent to CC as he maintains the linux-firmware package. ]

I've attached a proposal patch to add a split package amd-ucode-early to
linux-firmware. Any objections? Is the name ok or do we prefer just
'amd-ucode'?

And another note... Did anybody test if microcode is updated properly if both,
intel-ucode.img and amd-ucode, are loaded?
-- 
main(a){char*c=/*    Schoene Gruesse                         */"B?IJj;MEH"
"CX:;",b;for(a/*    Best regards             my address:    */=0;b=c[a++];)
putchar(b-1/(/*    Chris            cc -ox -xc - && ./x    */b/42*2-3)*42);}

El 23/08/18 a las 06:20, Christian Hesse escribió:
> Christian Hesse  on Wed, 2018/08/22 11:24:
>> Gerardo Exequiel Pozzi via arch-releng  on Sat,
>> 2018/08/18 23:29:
>>> +# Prepare amd_ucode
>>> +make_amd_ucode_img() {
>>> +    mkdir -p ${work_dir}/amd-ucode/kernel/x86/microcode
>>> +    cat
>>> ${work_dir}/x86_64/airootfs/lib/firmware/amd-ucode/microcode_amd*.bin >
>>> ${work_dir}/amd-ucode/kernel/x86/microcode/AuthenticAMD.bin
>>> +    pushd ${work_dir}/amd-ucode > /dev/null
>>> +    echo kernel/x86/microcode/AuthenticAMD.bin | bsdcpio -o -H newc -R
>>> 0:0  
>>>> amd-ucode.img  
>>> +    popd > /dev/null
>>> +    cp ${work_dir}/amd-ucode/amd-ucode.img
>>> ${work_dir}/x86_64/airootfs/boot/amd-ucode.img +}  
>>
>> Wondering if we should put this into a package... Could be a split package
>> of linux-firmware. Any opinion on this?
> 
> [ Added Laurent to CC as he maintains the linux-firmware package. ]
> 
> I've attached a proposal patch to add a split package amd-ucode-early to
> linux-firmware. Any objections? Is the name ok or do we prefer just
> 'amd-ucode'?

amd-ucode is right (like intel-ucode)
missing LICENSE file [#1]

> 
> And another note... Did anybody test if microcode is updated properly if both,
> intel-ucode.img and amd-ucode, are loaded?
> 

At least one feedback from intel user [#2], fw is loaded fine.

Thanks.


[#1] https://bugs.archlinux.org/task/59796
[#2] https://bugs.archlinux.org/task/59694#comment172278

I'd like to think I could do some verification since I'm on AMD... I'm new here, so help me along if there's something I've missed. I enabled the [testing] repo and installed: pacman -S testing/linux testing/amd-ucode testing/linux-firmware pacman -Q linux linux-firmware amd-ucode linux 4.18.5.arch1-1 linux-firmware 20180821.1d17c18-2 amd-ucode 20180821.1d17c18-2 I did a mkinitcpio -p linux just in case, then rebooted. output of `dmesg | grep -ie code -ie amd` (with obvious fluff removed): [ 0.000000] AMD AuthenticAMD [ 0.000000] random: get_random_u32 called from bsp_init_amd+0x1f9/0x290 with crng_init=0 [ 0.000000] ACPI: SSDT 0x00000000BFDDFB10 001714 (v01 AMD POWERNOW 00000001 AMD 00000001) [ 0.000000] ACPI: IVRS 0x00000000BFDE12A0 0000E8 (v01 AMD RD890S 00202031 AMD 00000000) [ 0.025991] Spectre V2 : Mitigation: Full AMD retpoline [ 0.036666] smpboot: CPU0: AMD FX(tm)-8120 Eight-Core Processor (family: 0x15, model: 0x1, stepping: 0x2) [ 0.053345] Performance Events: Fam15h core perfctr, AMD PMU driver. [ 1.950177] amd_uncore: AMD NB counters detected [ 1.950462] perf: AMD IBS detected (0x000000ff) [ 2.018364] microcode: CPU0: patch_level=0x0600063d [ 2.018371] microcode: CPU1: patch_level=0x0600063d [ 2.018379] microcode: CPU2: patch_level=0x0600063d [ 2.018388] microcode: CPU3: patch_level=0x0600063d [ 2.018396] microcode: CPU4: patch_level=0x0600063d [ 2.018403] microcode: CPU5: patch_level=0x0600063d [ 2.018410] microcode: CPU6: patch_level=0x0600063d [ 2.018418] microcode: CPU7: patch_level=0x0600063d [ 2.018457] microcode: Microcode Update Driver: v2.2. [ 2.375803] QUIRK: Enable AMD PLL fix [ 2.375865] ehci-pci 0000:00:12.2: applying AMD SB700/SB800/Hudson-2/3 EHCI dummy qh workaround [ 2.391345] ehci-pci 0000:00:13.2: applying AMD SB700/SB800/Hudson-2/3 EHCI dummy qh workaround [ 2.404862] ehci-pci 0000:00:16.2: applying AMD SB700/SB800/Hudson-2/3 EHCI dummy qh workaround [ 5.980964] microcode: CPU1: new patch_level=0x0600063e [ 5.992588] microcode: CPU4: new patch_level=0x0600063e [ 6.013367] microcode: CPU2: new patch_level=0x0600063e [ 6.034148] microcode: CPU6: new patch_level=0x0600063e [ 6.034221] x86/CPU: CPU features have changed after loading microcode, but might not take effect. Those microcode: lines at the end total 4, while I have an 8-core AMD FX-8120 - and the CPU numbers are usually different at each boot. Is there anything else I can do to assist? On Fri, Aug 24, 2018 at 5:10 PM Christian Hesse wrote:

Gerardo Exequiel Pozzi via arch-releng on Fri, 2018/08/24 17:34:

...

...

And another note... Did anybody test if microcode is updated properly if both, intel-ucode.img and amd-ucode, are loaded?

At least one feedback from intel user [#2], fw is loaded fine.

Well, intel ucode is loaded first, so that result is kind of expected. Feedback from an amd user would be welcome. -- main(a){char*c=/* Schoene Gruesse */"B?IJj;MEH" "CX:;",b;for(a/* Best regards my address: */=0;b=c[a++];) putchar(b-1/(/* Chris cc -ox -xc - && ./x */b/42*2-3)*42);}

"Bryan L. Gay" on Fri, 2018/08/24 22:24:

Well, fudge.

I see now that I needed to update GRUB so that it would put the microcode in the kernel boot config:

grub-mkconfig -o /boot/grub/grub.cfg

To now see: initrd /intel-ucode.img /amd-ucode.img /initramfs-linux.img

in the grub.cfg

Without this manual step, simply installing the amd-ucode doesn't take effect. Duh.

Interesting part of dmesg output is now:

[ 2.055382] microcode: microcode updated early to new patch_level=0x0600063e [ 2.055412] microcode: CPU0: patch_level=0x0600063e [ 2.055420] microcode: CPU1: patch_level=0x0600063e [ 2.055428] microcode: CPU2: patch_level=0x0600063e [ 2.055435] microcode: CPU3: patch_level=0x0600063e [ 2.055443] microcode: CPU4: patch_level=0x0600063e [ 2.055451] microcode: CPU5: patch_level=0x0600063e [ 2.055456] microcode: CPU6: patch_level=0x0600063e [ 2.055462] microcode: CPU7: patch_level=0x0600063e [ 2.055503] microcode: Microcode Update Driver: v2.2. [ 2.422079] QUIRK: Enable AMD PLL fix

So, I guess that answers the question sought after.

That looks good. Thanks a lot for testing! -- main(a){char*c=/* Schoene Gruesse */"B?IJj;MEH" "CX:;",b;for(a/* Best regards my address: */=0;b=c[a++];) putchar(b-1/(/* Chris cc -ox -xc - && ./x */b/42*2-3)*42);}