Re: [syzbot] [hfs?] WARNING in hfs_write_inode

[Michael Schmitz <schmitzmic@gmail.com>]

Hi Arnd,

Am 05.01.23 um 11:33 schrieb Arnd Bergmann:
> On Wed, Jan 4, 2023, at 20:06, Linus Torvalds wrote:
>> I suspect this code is basically all dead. From what I can tell, hfs
>> only gets updates for
>>
>>  (a) syzbot reports
>>
>>  (b) vfs interface changes
> There is clearly no new work going into it, and most data exchange
> with MacOS would use HFS+, but I think there are still some users.
PowerPC yaboot boot partitions spring to mind here. Plain HFS is still
used in places where it can't be replaced AFAIK.
>
>> and the last real changes seem to have been by Ernesto A. Fernández
>> back in 2018.
>>
>> Hmm. Looking at that code, we have another bug in there, introduced by
>> an earlier fix for a similar issue: commit 8d824e69d9f3 ("hfs: fix OOB
>> Read in __hfs_brec_find") added
>>
>> +       if (HFS_I(main_inode)->cat_key.CName.len > HFS_NAMELEN)
>> +               return -EIO;
>>
>> but it's after hfs_find_init(), so it should actually have done a
>> hfs_find_exit() to not leak memory.
>>
>> So we should probably fix that too.
>>
>> Something like this ENTIRELY UNTESTED patch?

Looking at Linus' patch, I wonder whether the missing fd.entrylength
size test in the HFS_IS_RSRC(inode) case was due to the fact that a
file's resource fork may be empty?

Adding Brad Boyer (bfind.c author) to Cc. Brad might know what
fd.entrylength should be set to in such a case.

Cheers,

    Michael


>>
>> Do we have anybody who looks at hfs?
> Adding Viacheslav Dubeyko to Cc, he's at least been reviewing
> patches for HFS and HFS+ somewhat recently. The linux-m68k
> list may have some users dual-booting old MacOS.
>
> Viacheslav, see the start of the thread at
> https://lore.kernel.org/lkml/000000000000dbce4e05f170f289@google.com/
>
>      Arnd