Re: State of the freebsd/crochet project?

From: Mark Millard <marklmi_at_yahoo.com>
Date: Fri, 20 Oct 2023 08:52:30 UTC
[I've dropped Warner from my CC.]

On Oct 20, 2023, at 01:44, Mark Millard <marklmi@yahoo.com> wrote:

> On Oct 20, 2023, at 00:39, Mark Millard <marklmi@yahoo.com> wrote:
> 
>> On Oct 20, 2023, at 00:31, Mark Millard <marklmi@yahoo.com> wrote:
>> 
>>> On Oct 19, 2023, at 22:30, Rahul Rameshbabu <sergeantsagara@protonmail.com> wrote:
>>> 
>>>> On Thu, 19 Oct, 2023 00:45:25 -0700 "Mark Millard" <marklmi@yahoo.com> wrote:
>>>>> On Oct 18, 2023, at 21:41, Rahul Rameshbabu <sergeantsagara@protonmail.com> wrote:
>>>>> 
>>>>>> On Tue, 17 Oct, 2023 09:01:33 -0600 "Warner Losh" <imp@bsdimp.com> wrote:
>>>>>>> On Tue, Oct 17, 2023, 7:44 AM void <void@f-m.fm> wrote:
>>>>>>> 
>>>>>>> On Tue, Oct 17, 2023 at 07:13:28AM -0600, Warner Losh wrote:
>>>>>>> 
>>>>>>>> Crochet has no active maintainers. Most people have moved on to poudriere.
>>>>>>> 
>>>>>>> Does poudriere handle the msdos uboot *and* efi part when
>>>>>>> creating the image?
>>>>>>> 
>>>>>>> Yes. I worked with manu years ago to put all the needed metadata for the different boards into the ports...
>>>>>> 
>>>>>> It does but it seems to have an unfortunate caveat. It assumes that
>>>>>> FAT16 is supported by all embedded targets. The Raspberry Pi 4 and I
>>>>>> assume the Pi 5 as well drop support for FAT16
>>>>> 
>>>>> The snapshot images booted the RPI4B's that I have access to just fine
>>>>> last I tried such. But release/arm64/RPI.conf and release/tools/arm.subr
>>>>> which are used to build such uses (selective axtractions across files):
>>>>> 
>>>>> FAT_SIZE="50m -b 1m"
>>>>> FAT_TYPE="16"
>>>>> . . .
>>>>> gpart add -t efi -l efi -a 512k -s ${FAT_SIZE} ${mddev}
>>>>> newfs_msdos -L efi -F ${FAT_TYPE} /dev/${mddev}s1
>>>>> 
>>>>> FreeBSD release images are also build with such: efi partition
>>>>> type and a FAT16 file system.
>>>>> 
>>>>> Looking at a (my abbreviation) RaspiOS64 boot media used to boot
>>>>> the RPi4B's (official RPi* media content, not FreeBSD materials):
>>>>> 
>>>>> # newfs_msdos -N /dev/da0s1
>>>>> /dev/da0s1: 523984 sectors in 32749 FAT16 clusters (8192 bytes/cluster)
>>>>> BytesPerSec=512 SecPerClust=16 ResSectors=1 FATs=2 RootDirEnts=512 Media=0xf0 FATsecs=128 SecPerTrack=63 Heads=255 HiddenSecs=0 HugeSectors=524288
> 
> Hmm. Linux reports:
> 
> # file  -s /dev/sda1
> /dev/sda1: DOS/MBR boot sector, code offset 0x58+2, OEM-ID "mkfs.fat", sectors/cluster 4, Media descriptor 0xf8, sectors/track 32, heads 64, sectors 524288 (volumes > 32 MB), FAT (32 bit), sectors/FAT 1020, reserved 0x1, serial number 0xf92becc, label: "boot       "
> 
> I must have misinterpreted what "newfs_msdos -N /dev/da0s1" reports
> when /dev/da0s1 has an already exiting file system.
> 
> Sorry for that and the resultant bad example.
> 
> For completeness, FreeBSD reports for that media:
> 
> # file -s /dev/da0s1
> /dev/da0s1: DOS/MBR boot sector, code offset 0x58+2, OEM-ID "mkfs.fat", sectors/cluster 4, Media descriptor 0xf8, sectors/track 32, heads 64, sectors 524288 (volumes > 32 MB), FAT (32 bit), sectors/FAT 1020, serial number 0xf92becc, label: "boot       "
> 
> Generating a valid example using, instead:
> 
> FreeBSD-15.0-CURRENT-arm64-aarch64-RPI-20231019-fb7140b1f928-266042.img.xz
> 
> expanded and dd'd to media:
> 
> # file -s /dev/da0s1
> /dev/da0s1: DOS/MBR boot sector, code offset 0x3c+2, OEM-ID "BSD4.4  ", sectors/cluster 8, root entries 512, sectors/FAT 50, sectors/track 63, heads 255, sectors 102400 (volumes > 32 MB), serial number 0xc90a0d0f, label: "EFI        ", FAT (16 bit)
> 
> I just used that to boot a RPi4B Rev 1.5 "C0T" part that has:
> 
> RPi: BOOTLOADER release VERSION:8ba17717 DATE: 2023/01/11 TIME: 17:40:52
> BOOTMODE: 0x06 partition 63 build-ts BUILD_TIMESTAMP=1673458852 serial c740af3c boardrev d03115 stc 421180
> Halt: wake: 1 power_off: 0
> 
> . . .

The console log for this shows that the RPi* firmware reported:

MBR: 0x00000800,  102400 type: 0x0c
MBR: 0x00019800,468757680 type: 0xa5
MBR: 0x00000000,       0 type: 0x00
MBR: 0x00000000,       0 type: 0x00
Trying partition: 0
type: 16 lba: 2048 oem: 'BSD4.4  ' volume: '      ^     '
rsc 1 fat-sectors 50 c-count 12783 c-size 8
root dir cluster 1 sectors 32 entries 512
FAT16 clusters 12783

> 
> Thu Oct 19 05:57:02 UTC 2023
> 
> FreeBSD/arm64 (generic) (ttyu0)
> 
> login: root
> Password:
> Oct 19 05:59:46 generic login[1474]: ROOT LOGIN (root) ON ttyu0
> FreeBSD 15.0-CURRENT (GENERIC) #0 main-n266042-fb7140b1f928: Thu Oct 19 04:52:33 UTC 2023
> 
>>>>> But it does have a partition type of fat32lba:
>>>>> 
>>>>> # gpart show -p /dev/da0
>>>>> =>       63  468862065    da0  MBR  (224G)
>>>>>      63       8129         - free -  (4.0M)
>>>>>    8192     524288  da0s1  fat32lba  (256M)
>>>>>  532480  468329648  da0s2  linux-data  (223G)
>>>>> 
>>>>> Do you know some specific RPi4B EEPROM content for which a FAT16
>>>>> file syatem is not supported? (The EEPROM has the RPi4B boot
>>>>> loader.) Or are you saying some U-Boot vintage is restricted to
>>>>> FAT32 file systems for loading FreeBSD's EFI/BOOT/bootaa64.efi ?
>>>> 
>>>> Yes, I believe that newer EEPROMs in 2020 and above (don't have the
>>>> exact release version but I can bisect if we need to know) no longer
>>>> support FAT16 unfortunately.
>>> 
>>> I just booted a RPi4B Rev 1.5 "C0T" part that has:
>>> 
>>> RPi: BOOTLOADER release VERSION:8ba17717 DATE: 2023/01/11 TIME: 17:40:52
>>> BOOTMODE: 0x06 partition 63 build-ts BUILD_TIMESTAMP=1673458852 serial c740af3c boardrev d03115 stc 421180
>>> Halt: wake: 1 power_off: 0
>>> 
>>> off the (what I call) RaspiOS64 media that I referenced earlier.
>>> 
>>> That means FAT16 with a partition indicating fat32lba.
> 
> I accidentally had used what was actually a FAT32 context:
> bad example.
> 
> The rest of the types of notes should be okay, including the
> corrected example.
> 
>>> There have been bug fixes, such as the 2022=01-31 EEPROM release that
>>> reported: "FAT/GPT fixes and file-system performance improvements."
>>> 
>>>> Here is a relevant link on Raspberry Pi
>>>> forums but I can experiment with pinning an exact EEPROM version from
>>>> the Raspberry PI repository if need be. When I got my Raspberry Pi 4
>>>> board recently, I did an upgrade to the latest EEPROM version and
>>>> noticed this issue.
>>>> 
>>>> * https://forums.raspberrypi.com/viewtopic.php?t=278295#p1685235
>>> 
>>> At that point (2020-06) there were only 2 tagged EEPROM content
>>> releases:
>>> 
>>> v2020.04.16-137ad
>>> v2019.09.10-137ad
>>> 
>>> There are 11 from after 2020-06.
>>> 
>>>> * https://github.com/raspberrypi/rpi-eeprom/releases
>>>> 
>>>> I am using the BOOT_UART feature of the Raspberry Pi 4 for this
>>>> debugging. I was debugging why the image I created at the had failed and
>>>> noticed the bootloader was failing to actually access/read any content
>>>> from the boot partition of the SD card. Switching to FAT32 resolved the
>>>> issue for me immediately, making me trust the assumption about the state
>>>> of later EEPROM releases from the repository.
>>> 
>>> As I've indicated, the official releases of official RPi*
>>> images have FAT16 files systems for the RPi* firmware --and
>>> they boot just fine when dd'd to the USB3 media that I use.
>>> 
>>> Similarly, the modern official FreeBSD images boot just fine
>>> and also have FAT16 for the msdosfs for the RPi*
>>> firmware+U-Boot+FreeBSED-UEFI-loader.
>>> 
>>> FreeBSD has had problems with a U-Boot vintage that was messed
>>> up for 8 GiByte RPi4B's. But that is now in the past.
>>> 
>>>> I noticed in that first link I added here, there seems to be mixed
>>>> opinions on whether the FAT16 file system is supported or not on latest
>>>> EEPROM releases for the Pi 4. Let me go back and test once again with a
>>>> FAT16 file system for my boot partition. I am currently running Jan 11,
>>>> 2023 release (I see they have a new release for Oct 18, 2023).
>>> 
>>> I've not tested the 2023-10-18 release.
>>> 
>>>> On a side note for myself, might be nice to throw the rpi-eeprom tools
>>>> into a port for others to easily grab.
>>>> 
>>>>> 
>>>>> Or may be you are referencing the partition type (expressed here
>>>>> in gpart terms), instead of the actual file system type that is
>>>>> contained? :
>>>>> 
>>>>>  efi                    The system partition for computers that use the
>>>>>                         Extensible Firmware Interface (EFI).  The scheme-
>>>>>                         specific types are "!239" for MBR, and
>>>>>                         "!c12a7328-f81f-11d2-ba4b-00a0c93ec93b" for GPT.
>>>>> . . .
>>>>>  fat16                  A partition that contains a FAT16 filesystem.  The
>>>>>                         scheme-specific type is "!6" for MBR.
>>>>> 
>>>>>  fat32                  A partition that contains a FAT32 filesystem.  The
>>>>>                         scheme-specific type is "!11" for MBR.
>>>>> 
>>>>>  fat32lba               A partition that contains a FAT32 (LBA)
>>>>>                         filesystem.  The scheme-specific type is "!12" for
>>>>>                         MBR.
>>>>> 
>>>>> (It has been some time since last I tried it, but last I tried
>>>>> partition type fat16, the RPi4B's boot from it just fine if I
>>>>> remember right. But GPT is supported, not just MBR.)
>>>>> 
>>>> 
>>>> I am not referring to the partition type rather than the real filesystem
>>>> type, but thanks for checking. In my boot flow with the image I
>>>> generate, I am using the efi partition type.
>>>> 
>>>>>> , so the boot partition
>>>>>> needs to be FAT32.
>>>>>> 
>>>>> 
>>>>> Not for the actual file system for any fairly modern vintage of
>>>>> RPi4B EEPROM content or U-Boot that I'm aware of. I've less
>>>>> certainty about the range of partition types, not having tested
>>>>> such in recent times.
>>>>> 
>>>>> Is there a chance you are using so large of an msdos file
>>>>> system that a FAT32/FAT32LBA file system is a requirement?
>>>> 
>>>> Great question but I believe that is not the case since for the same
>>>> msdos file system (though with different components from rpi-firmware),
>>>> I am able to boot the Raspberry Pi 3 up correctly. Let me verify once
>>>> more FAT16 (the filesystem) was indeed problematic for me since I was
>>>> debugging other issues like not realizing the Pi 4 needed different
>>>> components from the rpi-firmware project compared to previous boards.
>>>> 
>>> 
>> 
>> One more point: the 1st Capture.JPG image shows:
>> 
>> c-count 0 c-size 0 r-dir 0 r-sec 0
>> 
>> As I understand it, that is showing that the information was corrupt
>> as read: valid FAT16 would not have that combination.
>> 
> 




===
Mark Millard
marklmi at yahoo.com