Hi Chao,
On 2017/9/23 20:50, Chao Yu wrote:
Hi Xiang Gao,
On 2017/9/22 9:54, gaoxiang (P) wrote:
> Hi Chao,
>
> Thanks for your reply.
> This patch originates from the our sdcardfs compatible
case-insensitive lookup acceleration.
>
> For the customized case-insensitive create operation since f2fs has
no hash and it needs to decrypt all the dirent name,
> it is recommended to read ahead the whole dir data in advance.
>
> However I found f2fs cannot use page_cache_sync_readahead with null
@file pointer as the other common Linux file system,
> this feature is something like a bonus and I know that Linux has no
written words that we should support null @file pointer, yet
> page_cache_sync_readahead is an EXPORT_SYMBOL_GPL and could be used
by other non-standard 3rd kernel modules and
> of course I think it is really no problem.
Looks like some filesystems store private data in file->private_data
for some
purpose during ->open, and will use the data in late ->readpages. So
passing
@file with NULL value will simply cause kernel bug for these filesystems.
As I checked, most generic local filesystems just use mapping->host
instead of
file->f_mapping->host, so look this patch again, following generic
filesystem
looks not bad. ;)
I'm glad to hear that the community could apply it. :)
How about just changing file->f_mapping->host to mapping->host in out
patch?
To my knowledge of f2fs, I am not quite sure whether f_mapping in f2fs is
really equal to mapping or not. Therefore I add a extra BUG_ON and do some
stress tests to re-confirm that.
Yes, it's fine with me that we only change file->f_mapping->host to
mapping->host.
Thanks,
Thanks,
>
> Anyway, if it has no use to the community, I will apply it only on
the internal branch.
> And I will also find some common examples of data page readahead. :)
>
> Thanks,
>
>> -----Original Message-----
>> From: Chao Yu [mailto:[email protected]]
>> Sent: Thursday, September 21, 2017 10:33 PM
>> To: gaoxiang (P) <[email protected]>; [email protected];
Yuchao (T)
>> <[email protected]>
>> Cc: [email protected]
>> Subject: Re: [f2fs-dev] [PATCH] f2fs: allow readpages with NULL
file pointer
>>
>> On 2017/9/21 13:00, gaoxiang (P) wrote:
>>> Keep in line with the other Linux file system implementations since
>>> page_cache_sync_readahead supports NULL file pointer, and thus we can
>>> readahead data by f2fs itself without file opening (something like the
>>> btrfs behavior).
>>
>> Let's keep what it is until there is an example doing readahead
passing @file
>> with NULL.
>>
>> Thanks,
>>
>>>
>>> Signed-off-by: Gao Xiang <[email protected]>
>>> ---
>>> fs/f2fs/data.c | 4 +++-
>>> 1 file changed, 3 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index 95f30f0..afa12f1
>>> 100644
>>> --- a/fs/f2fs/data.c
>>> +++ b/fs/f2fs/data.c
>>> @@ -1333,9 +1333,11 @@ static int f2fs_read_data_pages(struct file *file,
>>> struct address_space *mapping,
>>> struct list_head *pages, unsigned nr_pages) {
>>> - struct inode *inode = file->f_mapping->host;
>>> + struct inode *inode = mapping->host;
>>> struct page *page = list_last_entry(pages, struct page, lru);
>>>
>>> + if (likely(file != NULL))
>>> + BUG_ON(file->f_mapping != mapping);
>>> trace_f2fs_readpages(inode, page, nr_pages);
>>>
>>> /* If the file has inline data, skip readpages */
>>>
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
