Re: ALPHA1 on Raspberry Pi 3B+ [added: and RPi4B]

On 14 Aug 2023, at 10:07, Emmanuel Vadot wrote:

> On Mon, 14 Aug 2023 09:51:27 -0500
> Mike Karels <mike@karels.net> wrote:
>
>> On 14 Aug 2023, at 0:30, Emmanuel Vadot wrote:
>>
>>> On Mon, 14 Aug 2023 07:03:19 +0200
>>> Emmanuel Vadot <manu@bidouilliste.com> wrote:
>>>
>>>> On Sun, 13 Aug 2023 12:35:31 -0500
>>>> Mike Karels <mike@karels.net> wrote:
>>>>
>>>>> On 13 Aug 2023, at 12:19, Emmanuel Vadot wrote:
>>>>>
>>>>>> On Sun, 13 Aug 2023 11:25:25 -0500
>>>>>> Mike Karels <mike@karels.net> wrote:
>>>>>>
>>>>>>> On 13 Aug 2023, at 11:10, Mark Millard wrote:
>>>>>>>
>>>>>>>> On Aug 13, 2023, at 08:17, Warner Losh <imp@bsdimp.com> wrote:
>>>>>>>>
>>>>>>>>> Manu just updated Linux DTS in the tree. Maybe see if you revert that if the problem persists.
>>>>>>>>
>>>>>>>> git: 69f8cc60aa1e - main - ofw_firmware: Only match if there is no compatible
>>>>>>>>
>>>>>>>> is the fix that Manu has committed:
>>>>>>>>
>>>>>>>> QUOTE
>>>>>>>>     ofw_firmware: Only match if there is no compatible
>>>>>>>>
>>>>>>>>     If there is a compatible string it likely means that the firmware needs
>>>>>>>>     a dedicated driver (like on RPI*).
>>>>>>>>
>>>>>>>>     PR:     273087
>>>>>>>>     Tested-by:      Mark Millard <marklmi26-fbsd@yahoo.com>
>>>>>>>>     Sponsored by:   Beckhoff Automation GmbH & Co. KG
>>>>>>>>     Fixes:          fdfd3a90b6ce ("ofw: Add a ofw_firmware driver")
>>>>>>>> END QUOTE
>>>>>>>
>>>>>>> Just for completeness: that change fixes the bcm2835_cpufreq0/powerd
>>>>>>> problem and the gpioled0 problem, but not the clk_fixed2 problem
>>>>>>> (clk_fixed4 on rpi4).  Installing an msdos boot partition from the
>>>>>>> 3 Aug image makes that problem disappear.
>>>>>>>
>>>>>>> 		Mike
>>>>>>
>>>>>>  There is two fixed-clock in the DTB without clock-frequency property
>>>>>> and with a status set to "disabled", this isn't conforming to the
>>>>>> bindings
>>>>>> (https://cgit.freebsd.org/src/tree/sys/contrib/device-tree/Bindings/clock/fixed-clock.yaml)
>>>>>> so we complain on this, this is normal.
>>>>>
>>>>> Would it be possible to detect the disabled status to prevent the errors
>>>>> (I'm guessing not) or to suppress the repeats?  150 lines of errors seems
>>>>> like a lot for an out-of-spec DTB entry, and makes it hard to ignore.
>>>>>
>>>>> 		Mike
>>>>
>>>>  Detecting the disabled status makes no sense, a fixed clock cannot be
>>>> disable, it's always present and running.
>>>>  But I think that if we check that clock-frenquency isn't present in
>>>> the probe function, print a message and bail we will not attempt to
>>>> attach the driver at each pass.
>>>>  That's the only clean solution that I can see without making dirty
>>>> hacks for some non-conforming DTB.
>>>
>>>  Something like this :
>>> https://people.freebsd.org/~manu/0001-clk-fixed-Bail-early-if-there-is-no-clock-frequency-.patch
>>>
>>>  Please let me know if that works.
>>
>> Not exactly.  This results in:
>>
>> clk_fixed4: clock-fixed have no clock-frequency
>>
>> but this appears 1562 times.  btw, "have" should be "has".
>
>  Then I don't know how to fix this properly.

Too bad.  The new message is much more obvious, but that's a lot
of spam.

>> I hadn't noticed them before, but there are a lock order reversal
>> and a malloc while holding a non-sleepable lock, associated with
>> the gpio fix:
>
>  Are you sure that they are new ?

They were not in ALPHA1, where gpioled0 did not configure.  They happen
once that problem was fixed, and before this patch.  I don't remember
if they happened in the 3 Aug snapshot; I can test that if it helps.
I would have noticed if I was watching, but I don't always connect to
the serial console or check dmesg.boot.

		Mike

>> gpioled0: <GPIO LEDs> on ofwbus0
>> lock order reversal: (sleepable after non-sleepable)
>>  1st 0xffff000000d0f6f8 LED mtx (LED mtx, sleep mutex) @ /usr/src/sys/dev/led/led.c:298
>>  2nd 0xffffa00001857c10 Raspberry Pi firmware gpio (Raspberry Pi firmware gpio, sx) @ /usr/src/sys/arm/broadcom/bcm2835/raspberrypi_gpio.c:252
>> lock order LED mtx -> Raspberry Pi firmware gpio attempted at:
>> #0 0xffff0000004d3984 at witness_checkorder+0xa98
>> #1 0xffff00000046ecb4 at _sx_xlock+0x7c
>> #2 0xffff0000008b1700 at rpi_fw_gpio_pin_set+0xe0
>> #3 0xffff0000001c7400 at led_create_state+0x158
>> #4 0xffff0000001910d4 at gpioled_attach+0x290
>> #5 0xffff00000049efa4 at device_attach+0x3f8
>> #6 0xffff00000049eb14 at device_probe_and_attach+0x7c
>> #7 0xffff0000004a0e54 at bus_generic_new_pass+0xfc
>> #8 0xffff0000004a0e04 at bus_generic_new_pass+0xac
>> #9 0xffff0000004a0e04 at bus_generic_new_pass+0xac
>> #10 0xffff00000049bd30 at bus_set_pass+0x4c
>> #11 0xffff0000003ea3bc at mi_startup+0x1fc
>> #12 0xffff0000000008ac at virtdone+0x70
>> uma_zalloc_debug: zone "malloc-64" with the following non-sleepable locks held:
>> exclusive sleep mutex LED mtx (LED mtx) r = 0 (0xffff000000d0f6f8) locked @ /usr/src/sys/dev/led/led.c:298
>> stack backtrace:
>> #0 0xffff0000004d3dc8 at witness_debugger+0x5c
>> #1 0xffff0000004d4fcc at witness_warn+0x400
>> #2 0xffff0000007830a0 at uma_zalloc_debug+0x30
>> #3 0xffff000000782c20 at uma_zalloc_arg+0x2c
>> #4 0xffff000000437abc at malloc+0x8c
>> #5 0xffff0000008a69b4 at bcm2835_firmware_property+0x44
>> #6 0xffff0000008b1718 at rpi_fw_gpio_pin_set+0xf8
>> #7 0xffff0000001c7400 at led_create_state+0x158
>> #8 0xffff0000001910d4 at gpioled_attach+0x290
>> #9 0xffff00000049efa4 at device_attach+0x3f8
>> #10 0xffff00000049eb14 at device_probe_and_attach+0x7c
>> #11 0xffff0000004a0e54 at bus_generic_new_pass+0xfc
>> #12 0xffff0000004a0e04 at bus_generic_new_pass+0xac
>> #13 0xffff0000004a0e04 at bus_generic_new_pass+0xac
>> #14 0xffff00000049bd30 at bus_set_pass+0x4c
>> #15 0xffff0000003ea3bc at mi_startup+0x1fc
>> #16 0xffff0000000008ac at virtdone+0x70
>> uma_zalloc_debug: zone "malloc-16" with the following non-sleepable locks held:
>> exclusive sleep mutex LED mtx (LED mtx) r = 0 (0xffff000000d0f6f8) locked @ /usr/src/sys/dev/led/led.c:298
>> stack backtrace:
>> #0 0xffff0000004d3dc8 at witness_debugger+0x5c
>> #1 0xffff0000004d4fcc at witness_warn+0x400
>> #2 0xffff0000007830a0 at uma_zalloc_debug+0x30
>> #3 0xffff000000782c20 at uma_zalloc_arg+0x2c
>> #4 0xffff000000437abc at malloc+0x8c
>> #5 0xffff0000007ce15c at bounce_bus_dmamem_alloc+0x50
>> #6 0xffff0000008a9404 at bcm2835_mbox_property+0xdc
>> #7 0xffff0000008a69e8 at bcm2835_firmware_property+0x78
>> #8 0xffff0000008b1718 at rpi_fw_gpio_pin_set+0xf8
>> #9 0xffff0000001c7400 at led_create_state+0x158
>> #10 0xffff0000001910d4 at gpioled_attach+0x290
>> #11 0xffff00000049efa4 at device_attach+0x3f8
>> #12 0xffff00000049eb14 at device_probe_and_attach+0x7c
>> #13 0xffff0000004a0e54 at bus_generic_new_pass+0xfc
>> #14 0xffff0000004a0e04 at bus_generic_new_pass+0xac
>> #15 0xffff0000004a0e04 at bus_generic_new_pass+0xac
>> #16 0xffff00000049bd30 at bus_set_pass+0x4c
>> #17 0xffff0000003ea3bc at mi_startup+0x1fc
>> uma_zalloc_debug: zone "malloc-128" with the following non-sleepable locks held:
>> exclusive sleep mutex LED mtx (LED mtx) r = 0 (0xffff000000d0f6f8) locked @ /usr/src/sys/dev/led/led.c:298
>> stack backtrace:
>> #0 0xffff0000004d3dc8 at witness_debugger+0x5c
>> #1 0xffff0000004d4fcc at witness_warn+0x400
>> #2 0xffff0000007830a0 at uma_zalloc_debug+0x30
>> #3 0xffff000000782c20 at uma_zalloc_arg+0x2c
>> #4 0xffff000000437abc at malloc+0x8c
>> #5 0xffff0000007ce1ac at bounce_bus_dmamem_alloc+0xa0
>> #6 0xffff0000008a9404 at bcm2835_mbox_property+0xdc
>> #7 0xffff0000008a69e8 at bcm2835_firmware_property+0x78
>> #8 0xffff0000008b1718 at rpi_fw_gpio_pin_set+0xf8
>> #9 0xffff0000001c7400 at led_create_state+0x158
>> #10 0xffff0000001910d4 at gpioled_attach+0x290
>> #11 0xffff00000049efa4 at device_attach+0x3f8
>> #12 0xffff00000049eb14 at device_probe_and_attach+0x7c
>> #13 0xffff0000004a0e54 at bus_generic_new_pass+0xfc
>> #14 0xffff0000004a0e04 at bus_generic_new_pass+0xac
>> #15 0xffff0000004a0e04 at bus_generic_new_pass+0xac
>> #16 0xffff00000049bd30 at bus_set_pass+0x4c
>> #17 0xffff0000003ea3bc at mi_startup+0x1fc
>>
>> 		Mike
>>
>
>
> -- 
> Emmanuel Vadot <manu@bidouilliste.com> <manu@freebsd.org>