Re: reboot broken on RPi4 on main [breaks at git: e6cf1a0826c9 - main - physmem: add ram0 pseudo-driver]

On Feb 24, 2023, at 22:16, Mark Millard <marklmi@yahoo.com> wrote:

> [The first main version to not reboot RPi4B's is:
> git: e6cf1a0826c9 - main - physmem: add ram0 pseudo-driver Mitchell Horne
> ]
> 
> On Feb 24, 2023, at 17:45, Mark Millard <marklmi@yahoo.com> wrote:
> 
>> On Feb 24, 2023, at 15:18, Mark Millard <marklmi@yahoo.com> wrote:
>> 
>>> On Feb 23, 2023, at 14:01, Mike Karels <mike@karels.net> wrote:
>>> 
>>>> Reboot (shutdown -r) is hanging on RPi4 on main as of today’s snapshot.
>>>> It hangs after printing the Uptime.  The initial bootstrap and boot from
>>>> power-up work.  I haven’t tested a snapshot since Jan 1, so I’m not sure
>>>> when it stopped working.  Any ideas what might have broken it?  I can bisect
>>>> the recent snapshots if nothing else.
>>>> 
>>>> 13.2-BETA2 works, however.
>>> 
>>> For reference . . .
>>> 
>>> While it is a personal build instead of a snapshot,
>>> I see the "shutdown -r now" hang problem for my build
>>> based on:
>>> 
>>> # ~/fbsd-based-on-what-commit.sh -C /usr/main-src/
>>> d04c86717c8c (HEAD -> main, freebsd/main, freebsd/HEAD) bsd.sys.mk: Add NO_WSTRICT_PROTOTYPES like in kernel
>>> branch: main
>>> merge-base: d04c86717c8ca3aa1bd9d8927a37a1f5443925b5
>>> merge-base: CommitDate: 2023-02-19 07:02:12 +0000
>>> n261026 (--first-parent --count for merge-base)
>>> 
>>> The context has a "B0T" 8 GiByte RPi4B.
>> 
>> For the artifact builds available for arm64, testing
>> the kernels:
>> 
>> The last that has "shutdown -r now" working on the RPi4B I'm using:
>> 
>>   • git: 8b418c83d175 - main - cp: Adjust the sparse file tests. Dag-Erling Smørgrav
>> ( https://artifact.ci.freebsd.org/snapshot/main/8b418c83d175fde3b1f65210509ddcf2ac248d9f/arm64/aarch64/kernel.txz )
>> 
>> No arm64 artifacts after that until . . .
>> 
>> The first that has "shutdown -r now" failing on the same RPi4B:
>> 
>>   • git: ded5f2954e1a - main - nfsd: Fix handling of the error case for nfsvno_open Rick Macklem
>> ( https://artifact.ci.freebsd.org/snapshot/main/ded5f2954e1a1bb7748646888938af767ee6257a/arm64/aarch64/kernel.txz )
>> 
>> So the range is limited to (including those end points):
>> (All are 2023-Feb-08 artifact builds.)
>> 
>>   • git: 8b418c83d175 - main - cp: Adjust the sparse file tests. Dag-Erling Smørgrav 
>>   • git: 87d405eab911 - main - iommu_gas: initialize start_gap as first node Doug Moore 
>>   • git: 8c784bb8cf36 - main - lua: Update to 5.4.4 Warner Losh 
>>   • git: 5fff09660e06 - main - geli: split the initalization of HMAC Gordon Tetlow 
>>   • git: 81ad626541db - main - Merge llvm-project main llvmorg-15-init-15358-g53dc0f10787 Dimitry Andric 
>>   • git: 753f127f3ace - main - Merge llvm-project main llvmorg-15-init-16436-g18a6ab5b8d1f Dimitry Andric 
>>   • git: fcaf7f8644a9 - main - Merge llvm-project main llvmorg-15-init-17485-ga3e38b4a206b Dimitry Andric 
>>   • git: 972a253a57b6 - main - Merge llvm-project main llvmorg-15-init-17826-g1f8ae9d7e7e4 Dimitry Andric 
>>   • git: 61cfbce3347e - main - Merge llvm-project release/15.x llvmorg-15.0.0-rc2-40-gfbd2950d8d0d Dimitry Andric 
>>   • git: a4a491e2238b - main - Merge llvm-project release/15.x llvmorg-15.0.0-9-g1c73596d3454 Dimitry Andric 
>>   • git: 6246ae0b85d8 - main - Merge llvm-project release/15.x llvmorg-15.0.2-10-gf3c5289e7846 Dimitry Andric 
>>   • git: f3fd488f1e19 - main - Merge llvm-project release/15.x llvmorg-15.0.6-0-g088f33605d8a Dimitry Andric 
>>   • git: 50d7464c3fe6 - main - Merge llvm-project release/15.x llvmorg-15.0.7-0-g8dfdcc7b7bf6 Dimitry Andric 
>>   • git: 3264f6b88fce - main - Bump __FreeBSD_version for llvm 15.0.7 merge Dimitry Andric 
>>   • git: 89a072d11cd2 - main - Makefile.amd64: remove construct that serves no purpose Warner Losh 
>>   • git: ae1dca798e0f - main - e1000: fix I219 hang on reset Kevin Bowling 
>>   • git: 647f2d2bc0cb - main - e1000: bump driver version Kevin Bowling 
>>   • git: 48bfd3597654 - main - Add nproc(1) Mateusz Guzik 
>>   • git: 1d03c3578d05 - main - arm: add an interrupt rman to nexus Mitchell Horne 
> 
> My build of the above version works for "shutdown -r now".
> 
>>   • git: f9bdaab95ec4 - main - ofwbus: remove handling of resources from ofwbus Mitchell Horne 
> 
> My build of the above version works for "shutdown -r now".
> It is the last version to do so.
> 
>>   • git: e6cf1a0826c9 - main - physmem: add ram0 pseudo-driver Mitchell Horne  
> 
> My build of the above version fails for "shutdown -r now":
> hang-up just after Uptime: . . . message.
> 
>>   • git: fa3f6655421f - main - netmap: drop redundant if_mtu assignment Vincenzo Maffione 
>>   • git: ded5f2954e1a - main - nfsd: Fix handling of the error case for nfsvno_open Rick Macklem
>> 
>> . . .
> 
> 

The following from boot -v looks somewhat odd:

ram0: reserving memory region:   2000-7ef0000
ram0: reserving memory region:   7f10000-31c00000
ram0: reserving memory region:   331d7000-39c2a000
ram0: reserving memory region:   39c2b000-39c2e000
ram0: reserving memory region:   39c2f000-39c30000
ram0: reserving memory region:   39c32000-39c33000
ram0: reserving memory region:   39c37000-3b050000
ram0: reserving memory region:   3b060000-3b300000
ram0: reserving memory region:   40000000-fc000000
ram0: reserving excluded region: 0-1fff
ram0: reserving excluded region: 7ef0000-7f0ffff
ram0: reserving excluded region: 31c00000-331d6fff
ram0: reserving excluded region: 39c2a000-39c2afff
ram0: reserving excluded region: 39c2e000-39c2efff
ram0: reserving excluded region: 39c30000-39c31fff
ram0: reserving excluded region: 39c33000-39c36fff
ram0: reserving excluded region: 3b050000-3b05ffff
ram0: reserving excluded region: 3ee5c000-3ee5cfff
ram0: reserving excluded region: 3ee5c000-3ee5cfff
ram0: failed to reserve region
ram0: reserving excluded region: fe100000-fe100fff

Possible oddities (4 GiByte RPi4B example):

Nothing covers any part of [3b300000,3ee5c000) .
Exclusion [3ee5c000-3ee5cfff] is repeated. (Related to the above?)
The "failed to reserve region" notice. (Related to repetition?)
Nothing covers any part of [fc000000,fe100000) .
Nothing covers any part of [fe101000,ffffffff] .

(The 2 different styles of specifying high bounds
reads oddly.)

===
Mark Millard
marklmi at yahoo.com