thesofproject
FW authentication questions regarding Apollo Lake chromebooks
Hi, I have an Apollo lake Chromebook and had a few questions with SOF
According to the docs, they say that Apollo Lake is supported but it requires a BIOS change. What changes are required in the BIOS for this to work? I have a custom coreboot + tianocore firmware loaded on my Apollo lake Chromebook already (and don’t boot chrome os on there), so if the BIOS change could be documented or upstreamed into coreboot that would be great to get audio working with SOF (it seems audio does work on geminilake already with the same coreboot tree).
ditto. I'm not seeing any differences between APL and GLK Chromebooks firmware-wise on the audio side, so clarification on what needs to be changed would be appreciated
According to the docs, they say...
Which docs, this one? https://thesofproject.github.io/latest/platforms/intel-cavs/apollolake/index.html
This section is a bit confusing, ~chromebooks never require any BIOS change~ (only Chromebooks after some date), I mean at least not for authentication reasons. Switching to developer mode is enough, all chromebooks (EDIT: after some date) ship with the community key: https://thesofproject.github.io/latest/getting_started/intel_debug/introduction.html#chromebooks-and-sof
See confirmation in https://github.com/thesofproject/sof-docs/pull/379
Please submit doc fixes to https://github.com/thesofproject/sof-docs , thanks!
I'm not seeing any differences between APL and GLK Chromebooks firmware-wise on the audio side
Indeed, check the symbolic links in https://github.com/thesofproject/sof-bin/tree/main/v2.0.x/sof-v2.0 (cloning may make that easier). They're created by https://github.com/thesofproject/sof/blob/main/installer/GNUmakefile where you can find this:
target_of_glk := apl
@coolstar @MrChromebox the issue is the CSME keys used for the SOF firmware authentication. ApolloLake Chromebooks rely on the Intel production key - which means an SOF firmware that you compile yourself would simply be rejected during the boot time. Starting with GeminiLake the key used changed, and you CAN generate the SOF firmware yourself as long as you have access to the Tensilica tools, put the device in developer mode and remove the write protection.
I did test SOF some 3 years ago on ApolloLake, but we had to find devices where we could update coreBoot with the Servo board. It failed on all devices past a specific production milestone, I have exactly ONE device that I could configure with SOF. It's also impossible to update that key with a regular update.
Bottom line: ApolloLake Chromebooks, despite using the same audio DSP as GeminiLake, will not support SOF. It is what it is, there's nothing we can do about it.
@plbossart just wondering, is it possible to update the key with a custom coreboot (even if an external flasher is required to write over the ME region), or are Apollo Lake chromebooks SOL when it comes to SOF?
@cujomalainey , in the docs you helped review recently we have:
As stated above, starting from 2019/2020, Intel Chromeboooks have been configured with the community key.
Could we add more specific product references there? It would obviously help answer the questions here. Thanks!
(even if an external flasher is required to write over the ME region)
Note merely adding a key does AFAIK not require replacing any firmware code. However I'm not familiar with the flash layout so I unfortunately don't know which region(s) it requires to overwrite (if possible)
This looks relevant: https://www.chromium.org/chromium-os/firmware-porting-guide/2-concepts/
@plbossart just wondering, is it possible to update the key with a custom coreboot (even if an external flasher is required to write over the ME region), or are Apollo Lake chromebooks SOL when it comes to SOF?
I don't fully recall the details but there was something fused or preventing updates at or after the DVT1 hw prototype stage. In other words none of the actual products can be updated.
Starting with GeminiLake [...] as long as you have access to the Tensilica tools,
For non-performance sensitive testing you can use the free, crosstool-ng-based toolchain: https://thesofproject.github.io/latest/introduction/index.html?highlight=toolchain#toolchain-faq Don't release binaries compiled with it.
All binaries released in https://github.com/thesofproject/sof-bin are compiled with XCC.
@plbossart just wondering, is it possible to update the key with a custom coreboot (even if an external flasher is required to write over the ME region), or are Apollo Lake chromebooks SOL when it comes to SOF?
I don't fully recall the details but there was something fused or preventing updates at or after the DVT1 hw prototype stage. In other words none of the actual products can be updated.
I am pretty sure if you disable RO FW write protect you can still get in there and change the key. How to do that is a completely different question as I am not a coreboot guy. Also note the many warnings you will see if you venture down that path.
@cujomalainey , in the docs you helped review recently we have:
As stated above, starting from 2019/2020, Intel Chromeboooks have been configured with the community key.
Could we add more specific product references there? It would obviously help answer the questions here. Thanks!
GLK/CML/TGL/JSL/ADL and all following intel chips are signed with the community key. IIRC BYT/CHT/BDW can also run unsigned FW. There is just that gap of APL/SKL/KBL/AML/HSW that can't be switched over easily. Any AMD/MTK platform supported here will also have the community keys. This is again specific to chromebooks.
I am pretty sure if you disable RO FW write protect you can still get in there and change the key. How to do that is a completely different question as I am not a coreboot guy. Also note the many warnings you will see if you venture down that path.
@cujomalainey you are very optimistic on this one. I tried with the support of the Intel team and the Google team (Dylan, Duncan et. al) and did not succeed except on the 'Elektro' device which was pre-production hardware. The same process failed on all other devices. It would take master hacker skills to make this work without any formal support.
I am pretty sure if you disable RO FW write protect you can still get in there and change the key. How to do that is a completely different question as I am not a coreboot guy. Also note the many warnings you will see if you venture down that path.
@cujomalainey you are very optimistic on this one. I tried with the support of the Intel team and the Google team (Dylan, Duncan et. al) and did not succeed except on the 'Elektro' device which was pre-production hardware. The same process failed on all other devices. It would take master hacker skills to make this work without any formal support.
I don't doubt, I have yet to see anything happen easily in that space. And this would be a high risk activity too, even with servod/ccd
Q: Since APL Chromebooks can load, and SoF provides as part of sof-bin, the intel-signed firmware for Apollolake, why can't that be used instead? Seems others have gotten that to work by disabling route checking on the bxt-da7219-max98357a driver (ref: https://github.com/EMLommers/Apollolake_Audio/)
@MrChromebox I think this should work indeed, I didn't realize we had signed versions for APL. The SOF driver would need to be selected manually with 'options snd-intel-dspconfig dsp_driver=3' since 'officially' Intel does not support this product configuration.
I am not sure why a machine driver change would be needed though, the same topology/machine driver is used for GLK and should work as is. I'd be curious to get logs with and without this .disable_route_check setting.
Here's what I used at the time for the UCM file: https://github.com/plbossart/UCM/tree/master/sof-bxtda7219max. It's loosely based on the Chrome version.
BTW I just realized we're both located in ATX, small world :-)
@plbossart small world indeed :)
I just tested a 5.16.9 mainline kernel with snd-intel-dspconfig dsp_driver=3' and get the following:
[ 3.151290] da7219 i2c-DLGS7219:00: Using default DAI clk names: da7219-dai-wclk, da7219-dai-bclk
[ 3.497627] sof-audio-pci-intel-apl 0000:00:0e.0: DSP detected with PCI class/subclass/prog-if 0x040100
[ 3.865307] sof-audio-pci-intel-apl 0000:00:0e.0: bound 0000:00:02.0 (ops i915_audio_component_bind_ops [i915])
[ 3.913398] sof-audio-pci-intel-apl 0000:00:0e.0: use msi interrupt mode
[ 3.959218] sof-audio-pci-intel-apl 0000:00:0e.0: hda codecs found, mask 4
[ 3.967466] sof-audio-pci-intel-apl 0000:00:0e.0: Firmware info: version 2:0:0-b678a
[ 3.967478] sof-audio-pci-intel-apl 0000:00:0e.0: Firmware: ABI 3:20:0 Kernel ABI 3:18:0
[ 3.967482] sof-audio-pci-intel-apl 0000:00:0e.0: warn: FW ABI is more recent than kernel
[ 3.967495] sof-audio-pci-intel-apl 0000:00:0e.0: unknown sof_ext_man header type 3 size 0x30
[ 4.017054] sof-audio-pci-intel-apl 0000:00:0e.0: Firmware info: version 2:0:0-b678a
[ 4.017065] sof-audio-pci-intel-apl 0000:00:0e.0: Firmware: ABI 3:20:0 Kernel ABI 3:18:0
[ 4.017069] sof-audio-pci-intel-apl 0000:00:0e.0: warn: FW ABI is more recent than kernel
[ 4.203061] sof-audio-pci-intel-apl 0000:00:0e.0: Topology: ABI 3:20:0 Kernel ABI 3:18:0
[ 4.203075] sof-audio-pci-intel-apl 0000:00:0e.0: warn: topology ABI is more recent than kernel
[ 4.203686] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Parent card not yet available, widget card binding deferred
[ 4.211725] da7219 i2c-DLGS7219:00: supply VDD not found, using dummy regulator
[ 4.211824] da7219 i2c-DLGS7219:00: supply VDDMIC not found, using dummy regulator
[ 4.211839] da7219 i2c-DLGS7219:00: supply VDDIO not found, using dummy regulator
[ 4.211863] da7219 i2c-DLGS7219:00: Invalid VDDIO voltage
[ 4.224496] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: sink widget DMic overwritten
[ 4.224516] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no source widget found for iDisp3 Tx
[ 4.224518] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp3 Tx -> direct -> hifi3
[ 4.224522] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no sink widget found for iDisp3 Tx
[ 4.224524] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp3_out -> direct -> iDisp3 Tx
[ 4.224529] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no source widget found for iDisp2 Tx
[ 4.224531] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp2 Tx -> direct -> hifi2
[ 4.224535] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no sink widget found for iDisp2 Tx
[ 4.224537] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp2_out -> direct -> iDisp2 Tx
[ 4.224542] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no source widget found for iDisp1 Tx
[ 4.224544] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp1 Tx -> direct -> hifi1
[ 4.224549] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no sink widget found for iDisp1 Tx
[ 4.224551] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp1_out -> direct -> iDisp1 Tx
[ 4.224559] bxt_da7219_max98357a bxt_da7219_mx98357a: snd_soc_bind_card: snd_soc_dapm_add_routes failed: -19
I get the same output (and non-functional audio) with the .disable_route_check added. I believe I had to copy the signed apl firmware over the community to get it to load still.
[ 4.224496] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: sink widget DMic overwritten
I think this one is harmless, though maybe @ranj063 can help clarify what this means.
[ 4.224516] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no source widget found for iDisp3 Tx [ 4.224518] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp3 Tx -> direct -> hifi3 [ 4.224522] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no sink widget found for iDisp3 Tx [ 4.224524] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp3_out -> direct -> iDisp3 Tx [ 4.224529] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no source widget found for iDisp2 Tx [ 4.224531] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp2 Tx -> direct -> hifi2 [ 4.224535] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no sink widget found for iDisp2 Tx [ 4.224537] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp2_out -> direct -> iDisp2 Tx [ 4.224542] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no source widget found for iDisp1 Tx [ 4.224544] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp1 Tx -> direct -> hifi1 [ 4.224549] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no sink widget found for iDisp1 Tx [ 4.224551] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp1_out -> direct -> iDisp1 Tx
Humm, I think I faced that problem before. I can't recall what I had to do, it could be as bad as commenting out the DAPM routes. I'll look into this.
I get the same output (and non-functional audio) with the
.disable_route_checkadded. I believe I had to copy the signed apl firmware over the community to get it to load still.
Yes, for Google we default to the community key on all platforms. You can work around that by using
options snd-sof-pci fw_path="intel/sof"
Humm, I think we didn't update the topology, there's a difference compared to the GLK version. I think we need to use this to remove the errors on widgets we don't care about.
diff --git a/tools/topology/topology1/sof-apl-da7219.m4 b/tools/topology/topology1/sof-apl-da7219.m4
index 5ce3ea8f0..08a45e894 100644
--- a/tools/topology/topology1/sof-apl-da7219.m4
+++ b/tools/topology/topology1/sof-apl-da7219.m4
@@ -194,6 +194,9 @@ DAI_CONFIG(HDA, 5, 5, iDisp3,
VIRTUAL_WIDGET(ssp5 Tx, out_drv, 0)
VIRTUAL_WIDGET(ssp1 Rx, out_drv, 1)
VIRTUAL_WIDGET(ssp1 Tx, out_drv, 2)
+VIRTUAL_WIDGET(iDisp3 Tx, out_drv, 15)
+VIRTUAL_WIDGET(iDisp2 Tx, out_drv, 16)
+VIRTUAL_WIDGET(iDisp1 Tx, out_drv, 17)
VIRTUAL_WIDGET(DMIC01 Rx, out_drv, 3)
VIRTUAL_WIDGET(DMic, out_drv, 4)
VIRTUAL_WIDGET(dmic01_hifi, out_drv, 5)
[ 4.224496] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: sink widget DMic overwritten
I think this one is harmless, though maybe @ranj063 can help clarify what this means.
[ 4.224516] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no source widget found for iDisp3 Tx [ 4.224518] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp3 Tx -> direct -> hifi3 [ 4.224522] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no sink widget found for iDisp3 Tx [ 4.224524] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp3_out -> direct -> iDisp3 Tx [ 4.224529] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no source widget found for iDisp2 Tx [ 4.224531] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp2 Tx -> direct -> hifi2 [ 4.224535] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no sink widget found for iDisp2 Tx [ 4.224537] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp2_out -> direct -> iDisp2 Tx [ 4.224542] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no source widget found for iDisp1 Tx [ 4.224544] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp1 Tx -> direct -> hifi1 [ 4.224549] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: no sink widget found for iDisp1 Tx [ 4.224551] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Failed to add route iDisp1_out -> direct -> iDisp1 Tx
Humm, I think I faced that problem before. I can't recall what I had to do, it could be as bad as commenting out the DAPM routes. I'll look into this. @plbossart all of the above are because of incorrect virtual widgets in the topology. We need to remove the "DMiC" virtual widget and add virtual widgets for each of the ones missing above.
@ranj063 GLK still have the DMIC widgets? Are they not needed?
VIRTUAL_WIDGET(DMIC01 Rx, out_drv, 3)
VIRTUAL_WIDGET(DMic, out_drv, 4)
VIRTUAL_WIDGET(dmic01_hifi, out_drv, 5)
@MrChromebox Can you try this topology? it needs to replace the existing one in /lib/firmware/sof-tplg (edit: need to unzip it)
@ranj063 GLK still have the DMIC widgets? Are they not needed? VIRTUAL_WIDGET(DMic, out_drv, 4)
I think this one's not needed but the other two are
@MrChromebox Can you try this topology? it needs to replace the existing one in /lib/firmware/sof-tplg (edit: need to unzip it)
that fixes it, although the audio output is listed as Headphones (vs internal speakers). I'll need to find a pair of wired headphones to test JD/HP output
Yes, for Google we default to the community key on all platforms. You can work around that by using
options snd-sof-pci fw_path="intel/sof"
possible to default to that if SMBIOS bios-vendor is 'coreboot' (and system-manufacturer is 'Google')? That would cover APL Chromebooks running my upstream coreboot firmware
one additional note - the volume control seems to be a bit off - it scales from 0-10%, but 11-100 is constant. So 10% is full volume (and quite loud)
possible to default to that if SMBIOS bios-vendor is 'coreboot' (and system-manufacturer is 'Google')? That would cover APL Chromebooks running my upstream coreboot firmware
this is the DMI quirk in sound/soc/sof/sof-pci-dev.c that changes the fw_path and adds the community prefix.
static const struct dmi_system_id community_key_platforms[] = {
{
.ident = "Up boards",
.matches = {
DMI_MATCH(DMI_SYS_VENDOR, "AAEON"),
}
},
{
.ident = "Google Chromebooks",
.matches = {
DMI_MATCH(DMI_SYS_VENDOR, "Google"),
}
},
{},
};
we'd want a quirk that works only for APL but only for Chromebooks. not sure how to go about this.
@MrChromebox you could pass a boot parameter to change the path from your firmware to the driver.
@cujomalainey yup that's easy enough, was just wondering if there were a way to work around via a DMI quirk. Sadly most Chromebook users wanting to run Linux are not terribly experienced and asking them to add a kernel boot param can be challenging at times =D
I mean you could pass straight from your AP build, no need to add alsa-conf options. No user intervention needed. Although why you need it is a bit fuzzy since it appears like this bug might warrant a fix that will be released as part of the basic build and the signing on APL should be the same as Windows so you should be able use the default paths.
Hello everyone :)
Board: reef/snappy (HP Chromebook X360 11 G1 EE)
OS: Fedora 35 w/mainline (5.17-rc5)
Firmware: alsa-sof-firmware from fedora repos (/lib/firmware/intel/sof/)
Topology posted by @plbossart
Additional kernel commandline: snd-intel-dspcfg.dsp_driver=3 loglevel=4 snd_sof_pci.fw_path="intel/sof"
[ 37.663651] sof-audio-pci-intel-apl 0000:00:0e.0: DSP detected with PCI class/subclass/prog-if 0x040100
[ 37.664916] sof-audio-pci-intel-apl 0000:00:0e.0: bound 0000:00:02.0 (ops i915_audio_component_bind_ops [i915])
[ 37.670749] sof-audio-pci-intel-apl 0000:00:0e.0: use msi interrupt mode
[ 37.709619] sof-audio-pci-intel-apl 0000:00:0e.0: hda codecs found, mask 4
[ 37.716195] sof-audio-pci-intel-apl 0000:00:0e.0: Firmware info: version 1:9:3-a9780
[ 37.716206] sof-audio-pci-intel-apl 0000:00:0e.0: Firmware: ABI 3:20:0 Kernel ABI 3:18:0
[ 37.716210] sof-audio-pci-intel-apl 0000:00:0e.0: warn: FW ABI is more recent than kernel
[ 37.716245] sof-audio-pci-intel-apl 0000:00:0e.0: unknown sof_ext_man header type 3 size 0x30
[ 37.765641] sof-audio-pci-intel-apl 0000:00:0e.0: Firmware info: version 1:9:3-a9780
[ 37.765652] sof-audio-pci-intel-apl 0000:00:0e.0: Firmware: ABI 3:20:0 Kernel ABI 3:18:0
[ 37.765655] sof-audio-pci-intel-apl 0000:00:0e.0: warn: FW ABI is more recent than kernel
[ 38.082427] sof-audio-pci-intel-apl 0000:00:0e.0: Topology: ABI 3:20:1 Kernel ABI 3:18:0
[ 38.082440] sof-audio-pci-intel-apl 0000:00:0e.0: warn: topology ABI is more recent than kernel
[ 38.082851] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: Parent card not yet available, widget card binding deferred
[ 38.094804] da7219 i2c-DLGS7219:00: supply VDD not found, using dummy regulator
[ 38.094913] da7219 i2c-DLGS7219:00: supply VDDMIC not found, using dummy regulator
[ 38.096133] da7219 i2c-DLGS7219:00: supply VDDIO not found, using dummy regulator
[ 38.096182] da7219 i2c-DLGS7219:00: Invalid VDDIO voltage
[ 38.127849] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: sink widget DMic overwritten
[ 38.128599] bxt_da7219_max98357a bxt_da7219_mx98357a: ASoC: source widget Capture overwritten
[ 38.145248] input: sof-bxtda7219max Headset Jack as /devices/pci0000:00/0000:00:0e.0/bxt_da7219_mx98357a/sound/card0/input10
[ 38.145659] input: sof-bxtda7219max HDMI/DP,pcm=5 as /devices/pci0000:00/0000:00:0e.0/bxt_da7219_mx98357a/sound/card0/input11
[ 38.146081] input: sof-bxtda7219max HDMI/DP,pcm=6 as /devices/pci0000:00/0000:00:0e.0/bxt_da7219_mx98357a/sound/card0/input12
[ 38.146483] input: sof-bxtda7219max HDMI/DP,pcm=7 as /devices/pci0000:00/0000:00:0e.0/bxt_da7219_mx98357a/sound/card0/input13
Tested 3 and 4-pole jack, detection did not work (posted video in #apollolake-testing on chultrabook server).
