Repository navigation
Config: need auto detection and fallback between SOF/SST/Legacy drivers #86
Description
The current code requires users to compile out two out of the three drivers for APL/GLK registered for the same PCI ID - or use mod probe blacklists. This needs to eventually move to an autodetect when the probe is handled in sequence, e.g. SOF first, then fallback to SST if firmware authentication fails, and last legacy HDaudio.
Filing it as a bug since it's really required for distributions such as ClearLinux and Ubuntu
Activity
mengdonglin commentedon Sep 13, 2018
@keyonjie could you help to take care of this?
keyonjie commentedon Sep 13, 2018
This may need refine/unified probing(especially for sof and sst) part, furthermore, sometimes using sof or sst is totally an user end choice(e.g. on those no FW authentication skl- platforms), we can't make selection for end-user in hard-coded logic.
I thought of using exclusive kernel config item for SOF and SST, which means if you select one(the other one will be unselected automatically), but the cons of doing that is we can't configure e.g byt with SST + apl with SOF, I am afraid someone make want to try something like that, even though this is not important for me.
@lgirdwood any comments?
mengdonglin commentedon Sep 18, 2018
This is an enhancement actually, so set label "enhancement"
lgirdwood commentedon Sep 19, 2018
@keyonjie We will have a single PCI and single ACPI entry point for all Intel SOF and legacy drivers. Drivers will default to legacy unless they are mentioned in the SOF lookup table (table of DMI names, DT names).
net: Fix rtl8192cu build errors on other platforms
plbossart commentedon Dec 21, 2018
Mechanism implemented upstream (will be in 4.21), SOF needs to implement the PCI class/subclass detection as described in #468
io_uring: fix KASAN use after free in io_sq_wq_submit_work
cpufreq: Cancel policy update work scheduled before freeing
cpuidle: psci: Fix suspicious RCU usage
Merge pull request thesofproject#86 from Rust-for-Linux/rust-extern-c…
reset: zynqmp: replace spaces with tabs
qed: fix possible unpaired spin_{un}lock_bh in _qed_mcp_cmd_and_union()
net: Fix zero-copy head len calculation.
arch/x86/mm/hugetlbpage.c: pud_huge() returns 0 when using 2-level pa…
arm64: cmpxchg_double*: hazard against entire exchange variable
arm64: cmpxchg_double*: hazard against entire exchange variable
powerpc/bpf/32: Fix failing test_bpf tests
crypto: qat - initialize user_input.lock for rate_limiting
btrfs: do proper folio cleanup when cow_file_range() failed
riscv, bpf: Add support for BPF_MOVSX in RV32 JIT