02:05 <bamse> fun stuff

10:15 <Mis012[m]> bamse: would it be possible to somewhat easily support Linux applying the static configuration, assuming that it needs to be described in the device tree anyway?

10:15 <Mis012[m]> could make running Linux in EL2 quite a bit easier

10:17 <Mis012[m]> though I guess for keeping the security properties, one would have to put the static configuration in TZ

10:18 <Mis012[m]> not like it makes more sense to have it in the hypervisor when it's concerning completely different cores as well

10:23 <Mis012[m]> could also do it in u-boot while it's in EL3, but not exactly sure how that would work

13:05 <bamse> Mis012[m]: i don't know, but i think the "stream mapping registers are read only" is a rather concrete problem to discuss/solve

13:09 <Mis012[m]> well, would definitely be nice to have it as quirk for the generic smmu driver

13:10 <Mis012[m]> bamse: would you happen to have any pointers as to what the `snps_ctrl_usb2_phy` and `usb2_utmi_clk_en` bits do?

13:10 <Mis012[m]> they don't seem to be femtophy signals

13:10 <Mis012[m]> but they're treated similarly by the magic init seq

15:00 <bamse> Mis012[m]: i'm not familiar with those and i'm not sure where to look

15:02 <Mis012[m]> apparently not femtophy databook :/

15:03 <Mis012[m]> msm8976 register description probably

15:05 <Mis012[m]> or qcs405

15:05 <Mis012[m]> for qcs405, I still wonder if the reset values are different from msm8976

15:06 <Mis012[m]> because if not, then half of the overrides it's setting are not overrides

19:08 <Mis012[m]> uh hm...

19:08 <Mis012[m]> bamse: does this also apply for qcs405?

19:09 <Mis012[m]> it seems that `UTMI_PHY_CMN_CTRL_OVERRIDE_EN` should be used instead, but the qcs405 init seq uses `PICOPHY_UTMI_CONTROL_OVERWRITE_EN`...

19:10 <Mis012[m]> hm, actually nvm

19:11 <Mis012[m]> `PICOPHY_UTMI_CONTROL_OVERWRITE_EN` is enabled on reset

19:12 <Mis012[m]> and the PICOPHY part does hint at it not being too relevant

19:14 <Mis012[m]> but then on 8976 and 8937, the reset value for `UTMI_PHY_CMN_CTRL_OVERRIDE_EN` is 0...

19:14 <Mis012[m]> so is it different on qcs405?

19:15 <Mis012[m]> the init seq doesn't make any attempt at setting it to 1...

20:16 <Mis012[m]> managed to get a FLAT fro qcs40x

20:16 <Mis012[m]> seems that the reset value is the same

20:17 <Mis012[m]> bamse: how the HELL does the init sequence do anything when `UTMI_PHY_CMN_CTRL_OVERRIDE_EN` stays at `0` the whole time????

20:19 <bamse> :)

20:19 <Mis012[m]> bamse: don't :) me

20:20 <Mis012[m]> is there some magic making `UTMI_PHY_CMN_CTRL_OVERRIDE_EN` redundant or have I just wasted a bunch of time again

20:20 <bamse> as i said, i don't know the usb phys well enough to know

20:22 <Mis012[m]> bamse: oh, stupid me, turns out that the overrides which are not NOOPs don't have that disclaimer?

20:24 <Mis012[m]> except one

20:25 <Mis012[m]> well, that's not an override, but it's used as part of the init seq

20:28 <Mis012[m]> looks like I will need more testing than I though, because now I wonder of if the USB2_SUSPEND_N_SEL toggling is necessary or not

20:28 <Mis012[m]> *thought

20:31 <Mis012[m]> and if it is necessary, then whether it's only necessary because it works around the POR changes being NOOPs

20:38 <Mis012[m]> actually, there's another fun possibility...

20:39 <Mis012[m]> that the overrides are not in fact applied, but since the things that are technically not overrides are what's really important for usb to work, noone ever noticed

20:40 <Mis012[m]> bamse: considering how busy the qcs40x owner probably is, I'm afraid getting my theories tested will be a nightmare...