On 04/12/2018 17:35, Jeffrey Hugo wrote: > On 12/4/2018 9:17 AM, Marc Gonzalez wrote: > >> I booted a downstream kernel with UFS debug enabled (log provided below) >> >> The one difference that jumps out at me is: >> >> DOWNSTREAM >> [ 10.902119] ufshcd-qcom 1da4000.ufshc: ufshcd_init_clocks: clk: core_clk_unipro, rate: 150000000 >> [ 10.902161] ufshcd-qcom 1da4000.ufshc: ufshcd_init_clocks: clk: core_clk_ice, rate: 300000000 >> [ 10.902198] ufshcd-qcom 1da4000.ufshc: ufshcd_init_clocks: clk: ref_clk, rate: 1000 >> >> UPSTREAM >> [ 2.072820] ufshcd-qcom 1da4000.ufshc: ufshcd_init_clocks: clk: core_clk_unipro, rate: 0 >> [ 2.080304] ufshcd-qcom 1da4000.ufshc: ufshcd_init_clocks: clk: core_clk_ice, rate: 0 >> [ 2.088547] ufshcd-qcom 1da4000.ufshc: ufshcd_init_clocks: clk: ref_clk, rate: 0 >> >> >> Jeffrey, I will check the regulators per your suggestion. >> I'm all ears if you have suggestions for the clocks as well. > > Hmm, this is interesting. > > I know you mentioned before that the clock rates were 0. Even with the > downstream kernel, I've seen clock rates be zero (for other use-cases). > > Since we have a delta between downstream and upstream, that seems > significant. > > When I've seen a clock keep its rate 0 like this, its been because of a > bad parent. You can check this in upstream via debugfs - > <debugfs>/clk/clk_summary > In downstream, I recall having to go into the individual clock sub > directory, and reading the parent file. > > However, maybe a simple solution. Do you have > https://patchwork.codeaurora.org/patch/657871/ ? Yep! $ git show --pretty=fuller 79bf268a13bf commit 79bf268a13bf24e46db54fd836807fdfccaf7d59 Author: Jeffrey Hugo <jhugo@xxxxxxxxxxxxxx> AuthorDate: Thu Nov 15 13:44:53 2018 -0700 Commit: Marc Gonzalez <marc.w.gonzalez@xxxxxxx> CommitDate: Thu Nov 29 10:03:33 2018 +0100 arm64: dts: qcom: msm8998: correct xo clock name Not sure I have the clk debug config knob, /sys/kernel/debug is empty. I'm off to tweak :-) Regards.
