Re: [PATCH net-next 1/5] ch_ktls: Use kmap_local_page() instead of kmap_atomic()

[Ira Weiny <ira.weiny@intel.com>]

On Thu, Nov 17, 2022 at 02:25:53PM -0800, Venkataramanan, Anirudh wrote:
> kmap_atomic() is being deprecated in favor of kmap_local_page().
> Replace kmap_atomic() and kunmap_atomic() with kmap_local_page()
> and kunmap_local() respectively.
> 
> Note that kmap_atomic() disables preemption and page-fault processing,
> but kmap_local_page() doesn't. Converting the former to the latter is safe
> only if there isn't an implicit dependency on preemption and page-fault
> handling being disabled, which does appear to be the case here.
                                 does not?

Also, say 'is not the case here'.  Appearances are not enough.  But we know
that this code is safe doing only memcpy's.

> 
> Also note that the page being mapped is not allocated by the driver,
> and so the driver doesn't know if the page is in normal memory. This is the
> reason kmap_local_page() is used as opposed to page_address().
> 
> I don't have hardware, so this change has only been compile tested.
> 
> Cc: Ayush Sawal <ayush.sawal@chelsio.com>
> Cc: Vinay Kumar Yadav <vinay.yadav@chelsio.com>
> Cc: Rohit Maheshwari <rohitm@chelsio.com>
> Cc: Ira Weiny <ira.weiny@intel.com>
> Cc: Fabio M. De Francesco <fmdefrancesco@gmail.com>
> Signed-off-by: Anirudh Venkataramanan <anirudh.venkataramanan@intel.com>
> ---
>  .../ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c | 10 +++++-----
>  1 file changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c b/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c
> index da9973b..d95f230 100644
> --- a/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c
> +++ b/drivers/net/ethernet/chelsio/inline_crypto/ch_ktls/chcr_ktls.c
> @@ -1853,24 +1853,24 @@ static int chcr_short_record_handler(struct chcr_ktls_info *tx_info,
>  				i++;
>  			}
>  			f = &record->frags[i];
> -			vaddr = kmap_atomic(skb_frag_page(f));
> +			vaddr = kmap_local_page(skb_frag_page(f));
>  
>  			data = vaddr + skb_frag_off(f)  + remaining;
>  			frag_delta = skb_frag_size(f) - remaining;
>  
>  			if (frag_delta >= prior_data_len) {
>  				memcpy(prior_data, data, prior_data_len);
> -				kunmap_atomic(vaddr);
> +				kunmap_local(vaddr);
>  			} else {
>  				memcpy(prior_data, data, frag_delta);
> -				kunmap_atomic(vaddr);
> +				kunmap_local(vaddr);
>  				/* get the next page */
>  				f = &record->frags[i + 1];
> -				vaddr = kmap_atomic(skb_frag_page(f));
> +				vaddr = kmap_local_page(skb_frag_page(f));
>  				data = vaddr + skb_frag_off(f);
>  				memcpy(prior_data + frag_delta,
>  				       data, (prior_data_len - frag_delta));
> -				kunmap_atomic(vaddr);
> +				kunmap_local(vaddr);

Agree with Fabio that this needs to be memcpy_from_page().  We should be
consistent in using it.

Ira

>  			}
>  			/* reset tcp_seq as per the prior_data_required len */
>  			tcp_seq -= prior_data_len;
> -- 
> 2.37.2
>