[RFC,V2,1/9] include/linux/pagemap.h: introduce attach/clear_page_private

Commit Message

Guoqing Jiang April 30, 2020, 9:44 p.m. UTC

The logic in attach_page_buffers and  __clear_page_buffers are quite
paired, but

1. they are located in different files.

2. attach_page_buffers is implemented in buffer_head.h, so it could be
   used by other files. But __clear_page_buffers is static function in
   buffer.c and other potential users can't call the function, md-bitmap
   even copied the function.

So, introduce the new attach/clear_page_private to replace them. With
the new pair of function, we will remove the usage of attach_page_buffers
and  __clear_page_buffers in next patches. Thanks for the new names from
Christoph Hellwig.

Suggested-by: Matthew Wilcox <[e-mail]>
Cc: Andrew Morton <[e-mail]>
Cc: "Darrick J. Wong" <[e-mail]>
Cc: William Kucharski <[e-mail]>
Cc: "Kirill A. Shutemov" <[e-mail]>
Cc: Andreas Gruenbacher <[e-mail]>
Cc: Yang Shi <[e-mail]>
Cc: Yafang Shao <[e-mail]>
Cc: Song Liu <[e-mail]>
Cc: [e-mail]
Cc: Chris Mason <[e-mail]>
Cc: Josef Bacik <[e-mail]>
Cc: David Sterba <[e-mail]>
Cc: [e-mail]
Cc: Alexander Viro <[e-mail]>
Cc: Jaegeuk Kim <[e-mail]>
Cc: Chao Yu <[e-mail]>
Cc: [e-mail]
Cc: Christoph Hellwig <[e-mail]>
Cc: [e-mail]
Cc: Anton Altaparmakov <[e-mail]>
Cc: [e-mail]
Cc: Mike Marshall <[e-mail]>
Cc: Martin Brandenburg <[e-mail]>
Cc: [e-mail]
Cc: Thomas Gleixner <[e-mail]>
Cc: Sebastian Andrzej Siewior <[e-mail]>
Cc: Roman Gushchin <[e-mail]>
Cc: Andreas Dilger <[e-mail]>
Signed-off-by: Guoqing Jiang <[e-mail]>
---
RFC -> RFC V2:  Address the comments from Christoph Hellwig
1. change function names to attach/clear_page_private and add comments.
2. change the return type of attach_page_private.

 include/linux/pagemap.h | 35 +++++++++++++++++++++++++++++++++++
 1 file changed, 35 insertions(+)

Comments

Andreas Grünbacher April 30, 2020, 10:10 p.m. UTC | #1

Hi,

Am Do., 30. Apr. 2020 um 23:56 Uhr schrieb Guoqing Jiang
<[e-mail]>:
> The logic in attach_page_buffers and  __clear_page_buffers are quite
> paired, but
>
> 1. they are located in different files.
>
> 2. attach_page_buffers is implemented in buffer_head.h, so it could be
>    used by other files. But __clear_page_buffers is static function in
>    buffer.c and other potential users can't call the function, md-bitmap
>    even copied the function.
>
> So, introduce the new attach/clear_page_private to replace them. With
> the new pair of function, we will remove the usage of attach_page_buffers
> and  __clear_page_buffers in next patches. Thanks for the new names from
> Christoph Hellwig.
>
> Suggested-by: Matthew Wilcox <[e-mail]>
> Cc: Andrew Morton <[e-mail]>
> Cc: "Darrick J. Wong" <[e-mail]>
> Cc: William Kucharski <[e-mail]>
> Cc: "Kirill A. Shutemov" <[e-mail]>
> Cc: Andreas Gruenbacher <[e-mail]>
> Cc: Yang Shi <[e-mail]>
> Cc: Yafang Shao <[e-mail]>
> Cc: Song Liu <[e-mail]>
> Cc: [e-mail]
> Cc: Chris Mason <[e-mail]>
> Cc: Josef Bacik <[e-mail]>
> Cc: David Sterba <[e-mail]>
> Cc: [e-mail]
> Cc: Alexander Viro <[e-mail]>
> Cc: Jaegeuk Kim <[e-mail]>
> Cc: Chao Yu <[e-mail]>
> Cc: [e-mail]
> Cc: Christoph Hellwig <[e-mail]>
> Cc: [e-mail]
> Cc: Anton Altaparmakov <[e-mail]>
> Cc: [e-mail]
> Cc: Mike Marshall <[e-mail]>
> Cc: Martin Brandenburg <[e-mail]>
> Cc: [e-mail]
> Cc: Thomas Gleixner <[e-mail]>
> Cc: Sebastian Andrzej Siewior <[e-mail]>
> Cc: Roman Gushchin <[e-mail]>
> Cc: Andreas Dilger <[e-mail]>
> Signed-off-by: Guoqing Jiang <[e-mail]>
> ---
> RFC -> RFC V2:  Address the comments from Christoph Hellwig
> 1. change function names to attach/clear_page_private and add comments.
> 2. change the return type of attach_page_private.
>
>  include/linux/pagemap.h | 35 +++++++++++++++++++++++++++++++++++
>  1 file changed, 35 insertions(+)
>
> diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
> index a8f7bd8ea1c6..2e515f210b18 100644
> --- a/include/linux/pagemap.h
> +++ b/include/linux/pagemap.h
> @@ -205,6 +205,41 @@ static inline int page_cache_add_speculative(struct page *page, int count)
>         return __page_cache_add_speculative(page, count);
>  }
>
> +/**
> + * attach_page_private - attach data to page's private field and set PG_private.
> + * @page: page to be attached and set flag.
> + * @data: data to attach to page's private field.
> + *
> + * Need to take reference as mm.h said "Setting PG_private should also increment
> + * the refcount".
> + */
> +static inline void attach_page_private(struct page *page, void *data)
> +{
> +       get_page(page);
> +       set_page_private(page, (unsigned long)data);
> +       SetPagePrivate(page);
> +}
> +
> +/**
> + * clear_page_private - clear page's private field and PG_private.
> + * @page: page to be cleared.
> + *
> + * The counterpart function of attach_page_private.
> + * Return: private data of page or NULL if page doesn't have private data.
> + */
> +static inline void *clear_page_private(struct page *page)
> +{
> +       void *data = (void *)page_private(page);
> +
> +       if (!PagePrivate(page))
> +               return NULL;
> +       ClearPagePrivate(page);
> +       set_page_private(page, 0);
> +       put_page(page);
> +
> +       return data;
> +}
> +

I like this in general, but the name clear_page_private suggests that
this might be the inverse operation of set_page_private, which it is
not. So maybe this can be renamed to detach_page_private to more
clearly indicate that it pairs with attach_page_private?

>  #ifdef CONFIG_NUMA
>  extern struct page *__page_cache_alloc(gfp_t gfp);
>  #else
> --
> 2.17.1
>

Thanks,
Andreas

Matthew Wilcox April 30, 2020, 10:13 p.m. UTC | #2

On Thu, Apr 30, 2020 at 11:44:42PM +0200, Guoqing Jiang wrote:
> +/**
> + * attach_page_private - attach data to page's private field and set PG_private.
> + * @page: page to be attached and set flag.
> + * @data: data to attach to page's private field.
> + *
> + * Need to take reference as mm.h said "Setting PG_private should also increment
> + * the refcount".
> + */

I don't think this will read well when added to the API documentation.
Try this:

/**
 * attach_page_private - Attach private data to a page.
 * @page: Page to attach data to.
 * @data: Data to attach to page.
 *
 * Attaching private data to a page increments the page's reference count.
 * The data must be detached before the page will be freed.
 */

> +/**
> + * clear_page_private - clear page's private field and PG_private.
> + * @page: page to be cleared.
> + *
> + * The counterpart function of attach_page_private.
> + * Return: private data of page or NULL if page doesn't have private data.
> + */

Seems to me that the opposite of "attach" is "detach", not "clear".

/**
 * detach_page_private - Detach private data from a page.
 * @page: Page to detach data from.
 *
 * Removes the data that was previously attached to the page and decrements
 * the refcount on the page.
 *
 * Return: Data that was attached to the page.
 */

Al Viro May 1, 2020, 1:42 a.m. UTC | #3

On Thu, Apr 30, 2020 at 03:13:38PM -0700, Matthew Wilcox wrote:

> > +/**
> > + * clear_page_private - clear page's private field and PG_private.
> > + * @page: page to be cleared.
> > + *
> > + * The counterpart function of attach_page_private.
> > + * Return: private data of page or NULL if page doesn't have private data.
> > + */
> 
> Seems to me that the opposite of "attach" is "detach", not "clear".

Or "remove", perhaps...

Al Viro May 1, 2020, 1:49 a.m. UTC | #4

On Fri, May 01, 2020 at 02:42:29AM +0100, Al Viro wrote:
> On Thu, Apr 30, 2020 at 03:13:38PM -0700, Matthew Wilcox wrote:
> 
> > > +/**
> > > + * clear_page_private - clear page's private field and PG_private.
> > > + * @page: page to be cleared.
> > > + *
> > > + * The counterpart function of attach_page_private.
> > > + * Return: private data of page or NULL if page doesn't have private data.
> > > + */
> > 
> > Seems to me that the opposite of "attach" is "detach", not "clear".
> 
> Or "remove", perhaps...

Actually, "detach" is better - neither "clear" nor "remove" imply "... and give
me what used to be attached there", as this thing is doing.

Guoqing Jiang May 1, 2020, 6:38 a.m. UTC | #5

On 5/1/20 12:10 AM, Andreas Grünbacher wrote:
> Hi,
>
> Am Do., 30. Apr. 2020 um 23:56 Uhr schrieb Guoqing Jiang
> <[e-mail]>:
>> The logic in attach_page_buffers and  __clear_page_buffers are quite
>> paired, but
>>
>> 1. they are located in different files.
>>
>> 2. attach_page_buffers is implemented in buffer_head.h, so it could be
>>     used by other files. But __clear_page_buffers is static function in
>>     buffer.c and other potential users can't call the function, md-bitmap
>>     even copied the function.
>>
>> So, introduce the new attach/clear_page_private to replace them. With
>> the new pair of function, we will remove the usage of attach_page_buffers
>> and  __clear_page_buffers in next patches. Thanks for the new names from
>> Christoph Hellwig.
>>
>> Suggested-by: Matthew Wilcox <[e-mail]>
>> Cc: Andrew Morton <[e-mail]>
>> Cc: "Darrick J. Wong" <[e-mail]>
>> Cc: William Kucharski <[e-mail]>
>> Cc: "Kirill A. Shutemov" <[e-mail]>
>> Cc: Andreas Gruenbacher <[e-mail]>
>> Cc: Yang Shi <[e-mail]>
>> Cc: Yafang Shao <[e-mail]>
>> Cc: Song Liu <[e-mail]>
>> Cc: [e-mail]
>> Cc: Chris Mason <[e-mail]>
>> Cc: Josef Bacik <[e-mail]>
>> Cc: David Sterba <[e-mail]>
>> Cc: [e-mail]
>> Cc: Alexander Viro <[e-mail]>
>> Cc: Jaegeuk Kim <[e-mail]>
>> Cc: Chao Yu <[e-mail]>
>> Cc: [e-mail]
>> Cc: Christoph Hellwig <[e-mail]>
>> Cc: [e-mail]
>> Cc: Anton Altaparmakov <[e-mail]>
>> Cc: [e-mail]
>> Cc: Mike Marshall <[e-mail]>
>> Cc: Martin Brandenburg <[e-mail]>
>> Cc: [e-mail]
>> Cc: Thomas Gleixner <[e-mail]>
>> Cc: Sebastian Andrzej Siewior <[e-mail]>
>> Cc: Roman Gushchin <[e-mail]>
>> Cc: Andreas Dilger <[e-mail]>
>> Signed-off-by: Guoqing Jiang <[e-mail]>
>> ---
>> RFC -> RFC V2:  Address the comments from Christoph Hellwig
>> 1. change function names to attach/clear_page_private and add comments.
>> 2. change the return type of attach_page_private.
>>
>>   include/linux/pagemap.h | 35 +++++++++++++++++++++++++++++++++++
>>   1 file changed, 35 insertions(+)
>>
>> diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
>> index a8f7bd8ea1c6..2e515f210b18 100644
>> --- a/include/linux/pagemap.h
>> +++ b/include/linux/pagemap.h
>> @@ -205,6 +205,41 @@ static inline int page_cache_add_speculative(struct page *page, int count)
>>          return __page_cache_add_speculative(page, count);
>>   }
>>
>> +/**
>> + * attach_page_private - attach data to page's private field and set PG_private.
>> + * @page: page to be attached and set flag.
>> + * @data: data to attach to page's private field.
>> + *
>> + * Need to take reference as mm.h said "Setting PG_private should also increment
>> + * the refcount".
>> + */
>> +static inline void attach_page_private(struct page *page, void *data)
>> +{
>> +       get_page(page);
>> +       set_page_private(page, (unsigned long)data);
>> +       SetPagePrivate(page);
>> +}
>> +
>> +/**
>> + * clear_page_private - clear page's private field and PG_private.
>> + * @page: page to be cleared.
>> + *
>> + * The counterpart function of attach_page_private.
>> + * Return: private data of page or NULL if page doesn't have private data.
>> + */
>> +static inline void *clear_page_private(struct page *page)
>> +{
>> +       void *data = (void *)page_private(page);
>> +
>> +       if (!PagePrivate(page))
>> +               return NULL;
>> +       ClearPagePrivate(page);
>> +       set_page_private(page, 0);
>> +       put_page(page);
>> +
>> +       return data;
>> +}
>> +
> I like this in general, but the name clear_page_private suggests that
> this might be the inverse operation of set_page_private, which it is
> not. So maybe this can be renamed to detach_page_private to more
> clearly indicate that it pairs with attach_page_private?

Yes, the new name is better, thank you!

Cheers,
Guoqing

Guoqing Jiang May 1, 2020, 6:39 a.m. UTC | #6

On 5/1/20 12:13 AM, Matthew Wilcox wrote:
> On Thu, Apr 30, 2020 at 11:44:42PM +0200, Guoqing Jiang wrote:
>> +/**
>> + * attach_page_private - attach data to page's private field and set PG_private.
>> + * @page: page to be attached and set flag.
>> + * @data: data to attach to page's private field.
>> + *
>> + * Need to take reference as mm.h said "Setting PG_private should also increment
>> + * the refcount".
>> + */
> I don't think this will read well when added to the API documentation.
> Try this:
>
> /**
>   * attach_page_private - Attach private data to a page.
>   * @page: Page to attach data to.
>   * @data: Data to attach to page.
>   *
>   * Attaching private data to a page increments the page's reference count.
>   * The data must be detached before the page will be freed.
>   */
>
>> +/**
>> + * clear_page_private - clear page's private field and PG_private.
>> + * @page: page to be cleared.
>> + *
>> + * The counterpart function of attach_page_private.
>> + * Return: private data of page or NULL if page doesn't have private data.
>> + */
> Seems to me that the opposite of "attach" is "detach", not "clear".
>
> /**
>   * detach_page_private - Detach private data from a page.
>   * @page: Page to detach data from.
>   *
>   * Removes the data that was previously attached to the page and decrements
>   * the refcount on the page.
>   *
>   * Return: Data that was attached to the page.
>   */

Thanks you very much, Mattew! Will change them in next version.

Best Regards,
Guoqing

Guoqing Jiang May 1, 2020, 6:41 a.m. UTC | #7

On 5/1/20 3:49 AM, Al Viro wrote:
> On Fri, May 01, 2020 at 02:42:29AM +0100, Al Viro wrote:
>> On Thu, Apr 30, 2020 at 03:13:38PM -0700, Matthew Wilcox wrote:
>>
>>>> +/**
>>>> + * clear_page_private - clear page's private field and PG_private.
>>>> + * @page: page to be cleared.
>>>> + *
>>>> + * The counterpart function of attach_page_private.
>>>> + * Return: private data of page or NULL if page doesn't have private data.
>>>> + */
>>> Seems to me that the opposite of "attach" is "detach", not "clear".
>> Or "remove", perhaps...
> Actually, "detach" is better - neither "clear" nor "remove" imply "... and give
> me what used to be attached there", as this thing is doing.

Ok, seems we have reached the agreement about the new name ;-), will 
follow the instruction.

Thanks & Regards,
Guoqing

Patch

diff --git a/include/linux/pagemap.h b/include/linux/pagemap.h
index a8f7bd8ea1c6..2e515f210b18 100644
--- a/include/linux/pagemap.h
+++ b/include/linux/pagemap.h
@@ -205,6 +205,41 @@  static inline int page_cache_add_speculative(struct page *page, int count)
 	return __page_cache_add_speculative(page, count);
 }
 
+/**
+ * attach_page_private - attach data to page's private field and set PG_private.
+ * @page: page to be attached and set flag.
+ * @data: data to attach to page's private field.
+ *
+ * Need to take reference as mm.h said "Setting PG_private should also increment
+ * the refcount".
+ */
+static inline void attach_page_private(struct page *page, void *data)
+{
+	get_page(page);
+	set_page_private(page, (unsigned long)data);
+	SetPagePrivate(page);
+}
+
+/**
+ * clear_page_private - clear page's private field and PG_private.
+ * @page: page to be cleared.
+ *
+ * The counterpart function of attach_page_private.
+ * Return: private data of page or NULL if page doesn't have private data.
+ */
+static inline void *clear_page_private(struct page *page)
+{
+	void *data = (void *)page_private(page);
+
+	if (!PagePrivate(page))
+		return NULL;
+	ClearPagePrivate(page);
+	set_page_private(page, 0);
+	put_page(page);
+
+	return data;
+}
+
 #ifdef CONFIG_NUMA
 extern struct page *__page_cache_alloc(gfp_t gfp);
 #else