Re: [f2fs-dev] [PATCH] f2fs: avoid f2fs_map_blocks() for consecutive holes in readpages

[Chao Yu <chao@kernel.org>]

Nanzhe,

On 1/20/2026 8:24 AM, Nanzhe Zhao wrote:
> Hi Chao:
> At 2026-01-19 21:44:48, "Chao Yu" <chao@kernel.org> wrote:
>>
>> I guess f2fs_map_no_dnode() will update map->m_next_pgofs to pgofs of next
>> potential valid dnode.
>>
>> Thanks,
>>
> 
> I guess we were discussing the cases that f2fs_get_dnode_of_data won't return
> -ENOENT in f2fs_map_blocks but dn.blkaddr is still NULL_ADDR or NEW_ADDR ?

I may misunderstand your question at first place, sorry about that.

> 
> I think I might understand the intention behind your repeated emphasis on the
> f2fs_map_no_dnode case?  Are you saying that, on F2FS, the vast majority of sparse
> files fall into holes where the dnode hasn't been allocated at all, and that within the
> dnode the blkaddr values NULL_ADDR and NEW_ADDR—especially the latter on the read path
> —are relatively uncommon?

Actually, no, we'd better investigate into details of hole data pattern in device
to see whether there are 1) more small holes in dnode (NULL_ADDR/NEW_ADDR case) or
2) there are more large hole in {di,i}dnode, do you have free slot to check that? :)

Thanks,

> 
> Thanks,
> Nanzhe Zhao