i2c issues in 6.18 on R4, but not r4pro (both mt7988 with i2c-mux on i2c2)
Frank Wunderlich (linux)
linux at fw-web.de
Sun Nov 2 03:39:19 PST 2025
Hi,
it seems that is caused by these 2 commits:
94c296776403 2025-06-03 i2c: muxes: pca954x: Reset if (de)select fails
Wojciech Siudy
690de2902dca 2025-06-03 i2c: muxes: pca954x: Use reset controller only
Wojciech Siudy
I'm sure that i had tested with them already reverted, but maybe
powercycle was not complete.
Git bisect pointed also to "i2c: muxes: pca954x: Use reset controller
only".
Added Author in To. Maybe he can tell where the gpio-reset should happen
in i2c core (have not found any recent commit which takes the
reset-gpios) like stated in the commit
commit 690de2902dca98aec96de004428c020ca902f047
Author: Wojciech Siudy <wojciech.siudy at nokia.com>
Date: Tue Jun 3 14:47:20 2025 +0200
i2c: muxes: pca954x: Use reset controller only
Fallback to legacy reset-gpios is no more needed, because the
framework
can use reset-gpios properity now as well.
my dts-config:
pca9545: i2c-mux at 70 {
compatible = "nxp,pca9545";
reg = <0x70>;
reset-gpios = <&pio 5 GPIO_ACTIVE_LOW>;
i'm unsure what devices are shown in the i2cdetect output, as there
should not be any on same bus, only behind the mux (rtc, eeprom and
sfp).
[ 1.637174] pca954x 2-0070: supply vdd not found, using dummy
regulator
[ 1.675701] i2c i2c-2: Added multiplexed i2c bus 3
[ 1.680597] i2c i2c-2: Added multiplexed i2c bus 4
[ 1.685472] i2c i2c-2: Added multiplexed i2c bus 5
[ 1.690344] i2c i2c-2: Added multiplexed i2c bus 6
[ 1.695137] pca954x 2-0070: registered 4 multiplexed busses for I2C
switch pca9545
looks like devices from channel 0 (rtc at 51,eeprom at 57) are shown on the
main bus too.
root at bpi-r4-v11:~
# i2cdetect -y 2
0 1 2 3 4 5 6 7 8 9 a b c d e f
00: -- -- -- -- -- -- -- --
10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
50: -- UU -- -- -- -- -- UU -- -- -- -- -- -- -- --
60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
70: UU -- -- -- -- -- -- --
root at bpi-r4-v11:~
# i2cdetect -y 3
0 1 2 3 4 5 6 7 8 9 a b c d e f
00: -- -- -- -- -- -- -- --
10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
50: -- UU -- -- -- -- -- UU -- -- -- -- -- -- -- --
60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
70: UU -- -- -- -- -- -- --
root at bpi-r4-v11:~
# i2cdetect -y 4
0 1 2 3 4 5 6 7 8 9 a b c d e f
00: -- -- -- -- -- -- -- --
10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
50: 50 51 -- -- -- -- -- -- -- -- -- -- -- -- -- --
60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
70: UU -- -- -- -- -- -- --
root at bpi-r4-v11:~
# i2cdetect -y 5
0 1 2 3 4 5 6 7 8 9 a b c d e f
00: -- -- -- -- -- -- -- --
10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
50: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
70: UU -- -- -- -- -- -- --
root at bpi-r4-v11:~
# i2cdetect -y 6
0 1 2 3 4 5 6 7 8 9 a b c d e f
00: -- -- -- -- -- -- -- --
10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
50: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
70: UU -- -- -- -- -- -- --
root at bpi-r4-v11:~
regards Frank
Am 2025-10-30 11:38, schrieb Frank Wunderlich (linux):
> Hi
>
> sorry for html-mail, my main email-provider (GMX) sent it as html (have
> enabled text-mail as default in settings) when using web-mailer.
>
> i've noticed i2c-issues (i2c2) on mt7988 BPI-R4 with 6.18. On most
> bootups the i2c-mux is not detected.
>
> root at bpi-r4-v11:~
> # i2cdetect -y 2
> 0 1 2 3 4 5 6 7 8 9 a b c d e f
> 00: -- -- -- -- -- -- -- --
> 10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 50: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 70: -- -- -- -- -- -- -- --
>
> i2c-probe is completed successful (no errors and debugs shown before
> final "return 0;")
>
> sometimes the mux came up with same kernel-binary, and i see also other
> devices on same i2c bus, but mostly all devices are not shown when i
> use i2cdetect on the bus.
>
> root at bpi-r4-v11:~
> # i2cdetect -y 2
> 0 1 2 3 4 5 6 7 8 9 a b c d e f
> 00: -- -- -- -- -- -- -- --
> 10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 50: -- UU -- -- -- -- -- UU -- -- -- -- -- -- -- --
> 60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
> 70: UU -- -- -- -- -- -- --
> root at bpi-r4-v11:~
> # [ 58.594846] sfp sfp1: failed to read EEPROM: -ENXIO
>
> 6.17 seems ok (have not seen the issue there yet).
>
> i looked through many commits, reverted some for clock (my own, lauras
> clk_gate,...), pinctrl and the i2c-mux (some reset-related) itself but
> nothing helped.
>
> One strange thing is that i have noticed no issues on bpi-r4pro which
> uses same SoC, same config and should use same drivers and
> configuration via dt.
> Only diference was the reset for the mux itself in dts as far as i
> see....so tried to disable it to have it nearly same as on r4pro
> without any effect.
>
> Maybe you have any idea how i can nail it down?
>
> root at bpi-r4-v11:~
> # dmesg | grep -i 'sfp\|i2c'
> [ 0.000000] Machine model: Banana Pi BPI-R4 (2x SFP+)
> [ 1.600387] i2c_dev: i2c /dev entries driver
> [ 1.605291] /soc/i2c at 11003000/rt5190a at 64: Fixed dependency cycle(s)
> with /soc/i2c at 11003000/rt5190a at 64/regulators/buck1 #i2c0 seems not
> affected
> [ 12.685157] platform sfp1: deferred probe pending: (reason unknown)
> [ 12.691440] platform sfp2: deferred probe pending: (reason unknown)
>
> # dmesg | grep -i 'err\|fail\|clk'
> ... #nothing related to i2c or clk
> [ 1.623805] pca954x 2-0070: probe failed
> ....
>
> wondered why i2c-clocks are always disabled when i look into the
> clk_summary (also when i2c was working)
>
> root at bpi-r4-v11:~
> # cat /sys/kernel/debug/clk/clk_summary | grep -i -B2 INFRA_I2C_BCK
> infra_usb_sys 0 0 0
> 125000000 0 0 50000 N deviceless
>
> i2c_sel 0 1 0
> 125000000 0 0 50000 N deviceless
>
> infra_i2c_bck 0 3 0
> 125000000 0 0 50000 N 11005000.i2c
>
> root at bpi-r4-v11:~
>
> but looking at the driver the clocks are disabled on probe which looks
> strange to me
>
> https://elixir.bootlin.com/linux/v6.18-rc1/source/drivers/i2c/busses/i2c-mt65xx.c#L1477
>
> i guess it is some kind of timing issue where clocks are still disabled
> but not reenabled in mtk_i2c_transfer, but
> my debug in this function shows that is called and ret=0 after the
> bulk_enable (and of course flooding console :p ).
> maybe some speed-calculation-issue?
>
> what i had tried to revert (also my own non-mainline like "convert
> invalid GATE to MUX"), so they seem not the rootcause:
>
> d51e7cfca3fe 2025-09-25 i2c: mt65xx: convert set_speed function to void
> Wolfram Sang
> b49218365280 2025-09-06 i2c: mediatek: fix potential incorrect use of
> I2C_MASTER_WRRD Leilk.Liu
> 614b1c3cbfb0 2025-06-12 i2c: use inclusive callbacks in struct
> i2c_algorithm Wolfram Sang
>
> c90fa5493f7a 2025-07-31 i2c: mux: pca9541: Use I2C adapter timeout
> value for arbitration timeout Manikanta Guntupalli
> 94c296776403 2025-06-03 i2c: muxes: pca954x: Reset if (de)select fails
> Wojciech Siudy
> 690de2902dca 2025-06-03 i2c: muxes: pca954x: Use reset controller only
> Wojciech Siudy
>
> e504d3bdb3d0 2025-09-15 clk: mediatek: clk-gate: Add ops for gates with
> HW voter Laura Nao
> 8ceff24a754a 2025-09-15 clk: mediatek: clk-gate: Refactor
> mtk_clk_register_gate to use mtk_gate struct Laura Nao
>
> bd6f4a91401f 2025-09-02 pinctrl: mediatek: moore: replace struct
> function_desc with struct pinfunction Bartosz Golaszewski
> 7a24f1f5b214 2025-09-02 pinctrl: mediatek: mt7988: use
> PINCTRL_PIN_FUNCTION() Bartosz Golaszewski
>
> regards Frank
More information about the Linux-mediatek
mailing list