[Cluster-devel] [PATCH v3 2/6] gfs2: Replace kmap_atomic()+memcpy by memcpy_from_page()

[Fabio M. De Francesco <fmdefrancesco@gmail.com>]

On gioved? 29 giugno 2023 23:50:07 CEST Deepak R Varma wrote:
> kmap_atomic() is deprecated in favor of kmap_local_{folio,page}().
> 
> kmap_atomic() disables page-faults and preemption (the latter only for
> !PREEMPT_RT kernels), However, the code within the mapping/un-mapping in
> gfs2_internal_read() does not depend on the above-mentioned side effects.
> 
> Further, memcpy_{from,to}_page() wrappers combine the
> {kmap, unmap}_local_page() blocks when they are intended exclusively to
> copy contents from/to the temporary mapped page. So, replace the
> kmap_atomic()/kunmap_automic()

s/kunmap_automic/kunmap_atomic/

> block by the memcpy_from_page() API call.
> This change allows to tidy-up code and also eliminate unused variable p.
> 
> Suggested-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>

I don't like to read the list of the functions replaced by the 
memcpy_from_page() helper in the subject of the patch. These details are 
better suited in commit messages.

Furthermore, you have not been consistent: "kmap_atomic()+memcpy" contains 
parentheses only in one of the two functions. Both or none.

The thing that really matters here is that You are replacing the deprecated 
kmap_atomic() with kmap_local_page(). This is what the subject should show.

Please cite the use of the suited helper _only_ within the commit message and 
explain why it is preferred to open code kmap_local_page() + memcpy() + 
kunmap_local(). 

> Signed-off-by: Deepak R Varma <drv@mailo.com>
> ---
> Changes in v3:
>    - Split as a separate patch for conversion in gfs2_internal_read()


>    - Use memcpy_from_page() as suggested by Fabio and  Andreas G
>    - Included split version in patch set
> 
> Changes in v2:
>    - Update patch description to correct the replacement function name
>      from kmap_local_folio to kmap_local _page
> 
> 
> 
>  fs/gfs2/aops.c | 9 ++++-----
>  1 file changed, 4 insertions(+), 5 deletions(-)
> 
> diff --git a/fs/gfs2/aops.c b/fs/gfs2/aops.c
> index 3eac4f2f5c27..f47fed657763 100644
> --- a/fs/gfs2/aops.c
> +++ b/fs/gfs2/aops.c
> @@ -489,7 +489,6 @@ int gfs2_internal_read(struct gfs2_inode *ip, char *buf,
> loff_t *pos, unsigned copied = 0;
>  	unsigned amt;
>  	struct page *page;
> -	void *p;
> 
>  	do {
>  		page = read_cache_page(mapping, index, gfs2_read_folio, 
NULL);
> @@ -498,12 +497,12 @@ int gfs2_internal_read(struct gfs2_inode *ip, char 
*buf,
> loff_t *pos, continue;
>  			return PTR_ERR(page);
>  		}
> -		p = kmap_atomic(page);
> -		amt = size - copied;
>  		if (offset + size > PAGE_SIZE)
>  			amt = PAGE_SIZE - offset;
> -		memcpy(buf + copied, p + offset, amt);
> -		kunmap_atomic(p);
> +		else
> +			amt = size - copied;

This is a different logical change. Please put it in a different patch with 
proper subject and commit message.

Thanks,

Fabio 

> +
> +		memcpy_from_page(buf, page, offset, amt);
>  		put_page(page);
>  		copied += amt;
>  		index++;
> --
> 2.34.1