Re: [Bug Report] OOB-read BUG in HFS+ filesystem

[Johannes Thumshirn]

On 15.04.25 09:52, Christian Brauner wrote:
> On Mon, Apr 14, 2025 at 06:23:28PM +0200, David Sterba wrote:
>> On Mon, Apr 14, 2025 at 03:21:56PM +0100, Matthew Wilcox wrote:
>>> On Mon, Apr 14, 2025 at 04:18:27PM +0200, Christian Brauner wrote:
>>>> On Mon, Apr 14, 2025 at 09:45:25PM +0800, now4yreal wrote:
>>>>> Dear Linux Security Maintainers,
>>>>> I would like to report a OOB-read vulnerability in the HFS+ file
>>>>> system, which I discovered using our in-house developed kernel fuzzer,
>>>>> Symsyz.
>>>>
>>>> Bug reports from non-official syzbot instances are generally not
>>>> accepted.
>>>>
>>>> hfs and hfsplus are orphaned filesystems since at least 2014. Bug
>>>> reports for such filesystems won't receive much attention from the core
>>>> maintainers.
>>>>
>>>> I'm very very close to putting them on the chopping block as they're
>>>> slowly turning into pointless burdens.
>>>
>>> I've tried asking some people who are long term Apple & Linux people,
>>> but haven't been able to find anyone interested in becoming maintainer.
>>> Let's drop both hfs & hfsplus.  Ten years of being unmaintained is
>>> long enough.
>>
>> Agreed. If needed there are FUSE implementations to access .dmg files
>> with HFS/HFS+ or other standalone tools.
>>
>> https://github.com/0x09/hfsfuse
>> https://github.com/darlinghq/darling-dmg
> 
> Ok, I'm open to trying. I'm adding a deprecation message when initating
> a new hfs{plus} context logged to dmesg and then we can try and remove
> it by the end of the year.
> 
> 

Just a word of caution though, (at least Intel) Macs have their EFI ESP 
partition on HFS+ instead of FAT. I don't own an Apple Silicon Mac so I 
can't check if it's there as well.