snd_soc_hifiberry_dacplus broken in 5.4.21 #3481

fichtennadel · 2020-02-23T21:37:25Z

Describe the bug
Using kernel 5.4.21 module snd_soc_hifiberry_dacplus fails to initialize the hifiberry hardware.

To reproduce
Raspi 3 with hifiberry DAC+ , boot kernel 5.4.21 (not ok) or 5.4.11 (ok)

Expected behaviour
module loads, initializes hifiberry DAC+, aplay -l lists sndrpihifiberry [snd_rpi_hifiberry_dacplus]
dmesg | grep hifi shows

[    6.829712] snd-rpi-hifiberry-dacplus soc: sound: ASoC: failed to init link HiFiBerry DAC+: -517
[    7.112002] snd-rpi-hifiberry-dacplus soc: sound: ASoC: failed to init link HiFiBerry DAC+: -517
[    7.429348] snd-rpi-hifiberry-dacplus soc: sound: ASoC: failed to init link HiFiBerry DAC+: -517
[    7.632537] snd-rpi-hifiberry-dacplus soc: sound: pcm512x-hifi <-> 3f203000.i2s mapping ok

Actual behaviour
module loads but fails to initialize hifiberry DAC+, therefore aplay -l lists aplay: device_list:272: no soundcards found...
dmesg | grep hifi shows

[    6.829712] snd-rpi-hifiberry-dacplus soc: sound: ASoC: failed to init link HiFiBerry DAC+: -517
[    7.112002] snd-rpi-hifiberry-dacplus soc: sound: ASoC: failed to init link HiFiBerry DAC+: -517
[    7.429348] snd-rpi-hifiberry-dacplus soc: sound: ASoC: failed to init link HiFiBerry DAC+: -517

Please note the missing last line, this means the driver failed to load or initialize the hifiberry

Last working kernel version was 5.4.11 , can be reproduced by down/upgrading (see links to forum discusssion below)

Logs
dmesg see https://paste.kodi.tv/sasuxiqure

Additional info
see also discussions on
https://forum.kodi.tv/showthread.php?tid=343068&pid=2926110#pid2926110
and
https://support.hifiberry.com/hc/en-us/community/posts/360009449558-module-snd-soc-hifiberry-dacplus-broken-in-kernel-5-4-18

The text was updated successfully, but these errors were encountered:

popcornmix · 2020-02-24T11:32:48Z

I think we want this: https://patchwork.kernel.org/patch/11394967/ from @HiassofT

HiassofT · 2020-02-24T12:04:17Z

It shouldn't hurt to include this patch until it dribbles down into stable trees, it'll at least avoid the warning and splat in regulator core.

The real issue with the hifiberry dacplus card seems to be though that it's (downstream) clock driver isn't loaded/loading (the warning/splat is just a consequence of that) - see eg lsmod output in the dmesg info linked above.

Most likely the clock driver and/or DT overlay need some changes after the clocks DT changes in 3292779

Commit [1] ([2] upstream) removed the "simple-bus" compatible string, "#address-cells" and "#size-cells" to fix some DT scheme warnings. In doing so it prevented the child clocks from being instantiated - the node might as well be absent. Fix the problem (at least with regards to functionality) by restoring "simple-bus" and the sizes. See: #3481 Fixes: [1] 238506e ("ARM: dts: bcm283x: Remove simple-bus from fixed clocks") Fixes: [2] 4b2d246 ("ARM: dts: bcm283x: Remove simple-bus from fixed clocks") Signed-off-by: Phil Elwell <phil@raspberrypi.com>

pelwell · 2020-02-27T14:01:30Z

By removing "simple-bus", the clock DT change broke the instantiation of its children. I've fixed it with a partial reversion.

Commit [1] ([2] upstream) removed the "simple-bus" compatible string, "#address-cells" and "#size-cells" to fix some DT scheme warnings. In doing so it prevented the child clocks from being instantiated - the node might as well be absent. Fix the problem (at least with regards to functionality) by restoring "simple-bus" and the sizes. See: #3481 Fixes: [1] 238506e ("ARM: dts: bcm283x: Remove simple-bus from fixed clocks") Fixes: [2] 4b2d246 ("ARM: dts: bcm283x: Remove simple-bus from fixed clocks") Signed-off-by: Phil Elwell <phil@raspberrypi.com>

pelwell · 2020-02-28T10:24:52Z

@lategoodbye I now understand what is happening, and I owe you an apology. of_clk_init (called early in the boot sequence) searches the tree for all compatible strings for drivers declared with CLK_OF_DECLARE, including clk_fixed_rate. This is why your change is fine for upstream DTBs.

AFAIK the HifiBerry clock driver can't make use of CLK_OF_DECLARE because it's a module, so it does have to rely on something else causing its driver to be instantiated. If we say that /clocks is only for static system clocks then the overlays that declare custom clocks have to be changed to put the clock node somewhere that will be probed - either in the root or under one of the other simple-bus nodes; as an overlay it shouldn't bother any of the kernel checkers.

commit 4b2d246 upstream. The fixed clocks doesn't form some kind of bus. So let's remove it. This fixes the follow DT schema warnings: clocks: clock@3:reg:0: [3] is too short clocks: clock@4:reg:0: [4] is too short clocks: $nodename:0: 'clocks' does not match '^(bus|soc|axi|ahb|apb)(@[0-9a-f]+)?$' clocks: #size-cells:0:0: 0 is not one of [1, 2] clocks: 'ranges' is a required property clock@3: 'reg' does not match any of the regexes: 'pinctrl-[0-9]+' clock@4: 'reg' does not match any of the regexes: 'pinctrl-[0-9]+' Signed-off-by: Stefan Wahren <wahrenst@gmx.net>

Change [1] removes the simple-bus compatible string from the "/clocks" node, preventing any custom clocks placed there from being initialised. Rather than reinstate the compatible string and trigger DT warnings at kernel build time, change the overlays to instantiate those clocks under the root node ("/"). See: #3481 Signed-off-by: Phil Elwell <phil@raspberrypi.com> [1] 4b2d246 ("ARM: dts: bcm283x: Remove simple-bus from fixed clocks")

fichtennadel · 2020-03-01T11:44:09Z

Builds over at https://forum.kodi.tv/showthread.php?tid=343068&pid=2928218#pid2928218 confirm the fix, hifiberry dacplus card initializes successfully now.