[BUG] bind command leads to invalid state where plaform data is NULL

[Heinrich Schuchardt]

On 6/19/23 14:37, Simon Glass wrote:
> Hi Heinrich,
>
> On Sun, 18 Jun 2023 at 11:12, Heinrich Schuchardt <xypron.glpk at gmx.de> wrote:
>>
>> Hello Simon,
>>
>> from origin/next I build qemu_arm64_defconfig with CONFIG_CMD_BIND=yes.
>>
>> I ran the image with:
>>
>> qemu-system-aarch64 -semihosting \
>>       -machine virt,gic-version=max -accel $(ACCEL) -m 1G -smp cores=2 \
>>       -bios u-boot.bin -cpu $(CPU) -nographic -gdb tcp::1234 \
>>       -netdev user,id=eth0,tftp=tftp -device e1000,netdev=eth0,romfile= \
>>       -drive if=none,file=arm64.img,format=raw,id=mydisk \
>>       -drive if=pflash,format=raw,index=1,file=envstore.img \
>>       -device virtio-rng-pci \
>>       -device ich9-ahci,id=ahci -device ide-hd,drive=mydisk,bus=ahci.0
>>
>> => scsi bind
>> => dm tree
>>    Class     Index  Probed  Driver                Name
>> -----------------------------------------------------------
>>    root          0  [ + ]   root_driver           root_driver
>>    pci           0  [ + ]   pci_generic_ecam      |-- pcie at 10000000
>>    pci_generi    0  [   ]   pci_generic_drv       |   |-- pci_0:0.0
>>    ahci          0  [ + ]   ahci_pci              |   `-- ahci_pci
>>    scsi          0  [ + ]   ahci_scsi             |       `-- ahci_scsi
>>    blk           0  [ + ]   scsi_blk              |           |--
>> ahci_scsi.id0lun0
>>    partition     0  [ + ]   blk_partition         |           |   |--
>> ahci_scsi.id0lun0:1
>>    partition     1  [ + ]   blk_partition         |           |   `--
>> ahci_scsi.id0lun0:15
>>    bootdev       2  [   ]   scsi_bootdev          |           `--
>> ahci_scsi.id0lun0.bootdev
>> => unbind blk 0
>> => dm tree
>>    Class     Index  Probed  Driver                Name
>> -----------------------------------------------------------
>>    root          0  [ + ]   root_driver           root_driver
>>    pci           0  [ + ]   pci_generic_ecam      |-- pcie at 10000000
>>    pci_generi    0  [   ]   pci_generic_drv       |   |-- pci_0:0.0
>>    ahci          0  [ + ]   ahci_pci              |   `-- ahci_pci
>>    scsi          0  [ + ]   ahci_scsi             |       `-- ahci_scsi
>>    bootdev       2  [   ]   scsi_bootdev          |           `--
>> ahci_scsi.id0lun0.bootdev
>> => bind scsi 0 scsi_blk
>> => dm tree
>>    Class     Index  Probed  Driver                Name
>> -----------------------------------------------------------
>>    root          0  [ + ]   root_driver           root_driver
>>    pci           0  [ + ]   pci_generic_ecam      |-- pcie at 10000000
>>    pci_generi    0  [   ]   pci_generic_drv       |   |-- pci_0:0.0
>>    ahci          0  [ + ]   ahci_pci              |   `-- ahci_pci
>>    scsi          0  [ + ]   ahci_scsi             |       `-- ahci_scsi
>>    bootdev       2  [   ]   scsi_bootdev          |           |--
>> ahci_scsi.id0lun0.bootdev
>>    blk           0  [   ]   scsi_blk              |           `-- scsi_blk
>> => part list scsi 0
>> dev_get_uclass_plat: null device
>> "Synchronous Abort" handler, esr 0x96000004, far 0xd503201f14000072
>> elr: 000000000003392c lr : 0000000000033924 (reloc)
>> elr: 000000007ff1492c lr : 000000007ff14924
>> x0 : d503201f1400000a x1 : 0000000000000000
>> x2 : 0000000000000090 x3 : 0000000000000013
>> x4 : 000000007eda0419 x5 : 0000000000000020
>> x6 : 00000000ffffffd0 x7 : 000000007eda06b0
>> x8 : 0000000000000010 x9 : 0000000000000000
>> x10: 0000000000000006 x11: 000000000001869f
>> x12: 0000000000000016 x13: 0000000000000004
>> x14: 0000000000000000 x15: 000000007eda073c
>> x16: 000000007ff2547c x17: 0000000000000000
>> x18: 000000007eea0db0 x19: 0000000000000000
>> x20: 0000000000000000 x21: 0000000000000001
>> x22: 0000000000000000 x23: 000000007eda0740
>> x24: 0000000000000000 x25: 0000000000000000
>> x26: 0000000000000000 x27: 0000000000000000
>> x28: 0000000000000000 x29: 000000007eda06d0
>>
>> Code: aa0303f7 94000287 aa0003f4 f9400260 (f9403418)
>> Resetting CPU ...
>>
>> The crash occurs in blk_read(). blk_read() does not check if desc == NULL.
>>
>> I would have expected "bind scsi 0 scsi_blk" to create a valid state
>> where the platform data is set up.
>>
>> Best regards
>>
>> Heinrich
>
> I am not sure what 'scsi bind' does.
>
> But binding a blk device needs to be done by its parent device. I
> suspect the 'bind' command does not know how to set up things
> correctly. See for example blk_create_device() which does quite a lot
> of setup.
>
> What are you trying to fix / do?

I just wanted to have an example for creating a man-page for the bind
command. Unfortunately new commands were created without man-page.

I have an example for "<node path> <driver>". Do you have a better
example for "bind <class> <index> <driver>"?

Best regards

Heinrich