Re: [PATCH] mm: shmem: implement POSIX_FADV_[WILL|DONT]NEED for shmem

Thanks Suren for the reply!!

On 11/2/2021 11:42 PM, Suren Baghdasaryan wrote:

quoted

quoted

Also not sure if you need to isolate the page before this writeback.

Isolation may not be required but this is to avoid trying of the reclaim
on the same page in some parallel path as both any way places the page
in the swap. Or you want me to remove the isolation here?

Maybe you could isolate them in bulk then? Seems like a waste
isolating them one-by-one.

Okay. Should the bulk here be SWAP_CLUSTER_MAX fine?

quoted

quoted

Won't deleting the page from swap cache interfere with ongoing
writeback if it has not yet completed?

I just followed the path of how to reclaim the page:
shrink_page_list()
     pageout()
     ..........
     __remove_mapping()
        __delete_from_swap_cache()

You see some issue here which I can't understand.

I think shrink_page_list() would not call __remove_mapping() if after
pageout() PageWriteback() is still true. Maybe I'm missing some path?

I agree here. PG_writeback might have still set on the pages under
background writeback thus should be excluded.

I just considered the case for sync writes, eg:zram,  where PG_writeback
is cleared before returns.

Will update the change accordingly.

-- 
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora
Forum, a Linux Foundation Collaborative Project