Logical block protection issues with IBM LTO-7 and LTO-8 about mhvtl

[markh794]

Are you able to share the virtual media with me ?
The write did occur (for this use case). I will need all files.
The other option is to create a utility that will report various CRC32C seeding options

[rohr22]

The other option is to create a utility that will report various CRC32C seeding options

What do you have in mind for this?

[markh794]

Hi Peter,
I've just pushed a patch which will log the CRC using the following combinations:

• Seed using values of '0' and '0xffffffff'
• Returned value, and returned value bit inverted.

Can you apply the patch and 're-crash the car' (reproduce the write) and provide the /var/log/messages.
Thanks

[rohr22]

OK, I applied your patch and recreated the problem. Attached is a tar file with vtl entries from when the problem occurred.

lto7_failure2.tar.gz

Thank you,
Peter

[markh794]

thanks for the log..

Jan 10 20:14:03 choctaw /usr/bin/vtltape[7093]: ERROR: get_lbp_crc(): CRC32C mismatch - LBP: 0xcc1e58d7, calculated: 0xd7581ecc Jan 10 20:14:03 choctaw /usr/bin/vtltape[7093]: ERROR: writeBlock_lzo(): LBP mis-compare on write : Returning E_LOGICAL_BLOCK_GUARD_FAILED Jan 10 20:14:03 choctaw /usr/bin/vtltape[7093]: return_sense(): [Key/ASC/ASCQ] [04 10 01] Jan 10 20:14:03 choctaw /usr/bin/vtltape[7093]: ERROR: log_crc_options(): Legacy CRC32C: 0xd7581ecc Jan 10 20:14:03 choctaw /usr/bin/vtltape[7093]: ERROR: log_crc_options(): CRC32C with seed 0: 0xd7581ecc Jan 10 20:14:03 choctaw /usr/bin/vtltape[7093]: ERROR: log_crc_options(): Inverted CRC32C with seed 0: 0x28a7e133 Jan 10 20:14:03 choctaw /usr/bin/vtltape[7093]: ERROR: log_crc_options(): CRC32C with seed 0xffffffff: 0x0892032f Jan 10 20:14:03 choctaw /usr/bin/vtltape[7093]: ERROR: log_crc_options(): Inverted CRC32C with seed 0xffffffff: 0xf76dfcd0

Took me a little while staring at it to realise it's a big-endian/little-endian thing. I didn't need that additional debugging Information after all.

e.g.
ERROR: get_lbp_crc(): CRC32C mismatch - LBP: 0xcc1e58d7, calculated: 0xd7581ecc

I'll hopefully get a fix out ASAP.

[rohr22]

Took me a little while staring at it to realise it's a big-endian/little-endian thing. I didn't need that additional debugging Information after all.

Good. I wonder why it doesn't occur with LTO-6 tapes.

[markh794]

LTO-6 only use RS-CRC - that routine I coped from the IBM SCSI reference manual appendix. It returns the RS-CRC value in big-end order.

[markh794]

uploaded patch 5837bb924f1e9e54d9b83cb2e383777e12ed7f7b - untested, but should fix the CRC32C bug.

[rohr22]

This seems to fix the problem. I can now write to tape with logical block protection with IBM LTO-6, LTO-7, and LTO-8!

Thank you.

[markh794]

Closing this out. Thanks for the verification.