You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I really appreciate your work. I was wondering whether there are plans for upstreaming this driver? Although as far as I understand, since the code of lzfse is not GPL, I guess it won't.
Anyways, I compile my own kernels for the purpose of running Linux on my Mac, and the read support of your driver is quite useful. I now add your driver in my kernel itself. Although, it compiles well, I wanted to know the exact kconfig of this driver that I should use.
My current config is below:
# SPDX-License-Identifier: GPL-2.0-only
config APFS_FS
tristate "Apple file system support"
select CRYPTO
select CRYPTO_CRC32C
select LIBCRC32C
select CRYPTO_XXHASH
select CRYPTO_SHA256
select CRYPTO_BLAKE2B
select BUFFER_HEAD
select FS_IOMAP
select LEGACY_DIRECT_IO
help
If you say Y here, you will be able to mount APFS partitions
with read-only access. Write access is experimental and will
corrupt your container.
Reacted by yoshimo, neko_code, James Cuzella, Integral and Chung-Chiang Cheng
I was wondering whether there are plans for upstreaming this driver?
This question has come up before. The situation has not changed that much. At the very least I would need to fix the locking and handle ENOSPC better. But I don't expect upstream to be all that interested.
the code of lzfse is not GPL
It's under the 3-clause bsd license, which is compatible with the gpl. The real problem with the lzfse code is that it triggers an objtool warning, probably because of all the crazy gotos.
I wanted to know the exact kconfig
Frankly I have no idea. I only build out of tree these days, and always for standard distro kernels. I'm sure you at least need to select ZLIB_INFLATE as well.
You can leave this open if you prefer, I do need to think about Kconfig eventually. As for your question, I don't think we use any of the crypto hashes you selected right now, only crc32c.
You can leave this open if you prefer, I do need to think about Kconfig eventually. As for your question, I don't think we use any of the crypto hashes you selected right now, only crc32c.
# SPDX-License-Identifier: GPL-2.0-only
config APFS_FS
tristate "Apple file system support"
select LIBCRC32C
select ZLIB_INFLATE
select NLS
select BUFFER_HEAD
select FS_IOMAP
select LEGACY_DIRECT_IO
help
If you say Y here, you will be able to mount APFS partitions
with read-only access. Write access is experimental and will
corrupt your container.
@eafer . I would be more than happy to collaborate with you on upstreaming this. I wouldn't mind handling all of the grunt work/bouncing it off the mailing list if you would help me test/help review it once I have a working setup.
@ethancedwards8 I don't see it happening at this point, the upstream filesystem people are very demanding (which is reasonable because a bad filesystem can do a lot of damage). At the very least I would need to get rid of buffer heads first, and I haven't even looked into that.
I'm seeing some traffic to this issue, so for the record: upstream is no longer interested in picking up new filesystems without a good reason, so I don't plan to ever submit it myself. The code has all the proper signoffs and everything though, so I can't/won't stop others from trying if they want to, like Ethan did last year.
Reacted by Ethan Carter Edwards and Daniel Engbert
Activity
eafer commentedon Jun 9, 2024
This question has come up before. The situation has not changed that much. At the very least I would need to fix the locking and handle ENOSPC better. But I don't expect upstream to be all that interested.
It's under the 3-clause bsd license, which is compatible with the gpl. The real problem with the lzfse code is that it triggers an objtool warning, probably because of all the crazy gotos.
Frankly I have no idea. I only build out of tree these days, and always for standard distro kernels. I'm sure you at least need to select
ZLIB_INFLATEas well.eafer commentedon Jun 9, 2024
I found this old commit from the original repo:
linux-apfs/linux-apfs@885a04c
I guess you also need to select
NLS.AdityaGarg8 commentedon Jun 9, 2024
Thanks for the reply!
AdityaGarg8 commentedon Jun 10, 2024
As a last question @eafer , anything in the kconfig above which you feel is absolutely unnecessary
eafer commentedon Jun 12, 2024
You can leave this open if you prefer, I do need to think about Kconfig eventually. As for your question, I don't think we use any of the crypto hashes you selected right now, only crc32c.
AdityaGarg8 commentedon Jun 12, 2024
Alright, I'll reopened
AdityaGarg8 commentedon Jun 12, 2024
I settled with this
ethancedwards8 commentedon Jan 22, 2025
@eafer . I would be more than happy to collaborate with you on upstreaming this. I wouldn't mind handling all of the grunt work/bouncing it off the mailing list if you would help me test/help review it once I have a working setup.
eafer commentedon Jan 23, 2025
@ethancedwards8 I don't see it happening at this point, the upstream filesystem people are very demanding (which is reasonable because a bad filesystem can do a lot of damage). At the very least I would need to get rid of buffer heads first, and I haven't even looked into that.
eafer commentedon Jun 26, 2026
I'm seeing some traffic to this issue, so for the record: upstream is no longer interested in picking up new filesystems without a good reason, so I don't plan to ever submit it myself. The code has all the proper signoffs and everything though, so I can't/won't stop others from trying if they want to, like Ethan did last year.