On Sun, 2020-05-31 at 20:28 +0300, Ivan Safonov wrote: > On 5/31/20 7:15 PM, Joe Perches wrote: > > On Sun, 2020-05-31 at 19:08 +0300, Ivan Safonov wrote: > > > Wrappers around skb_clone() do not simplify the driver code. > > [] > > > -inline struct sk_buff *_rtw_skb_clone(struct sk_buff *skb) > > > -{ > > > - return skb_clone(skb, in_interrupt() ? GFP_ATOMIC : GFP_KERNEL); > > > -} > > > - > > [] > > > diff --git a/drivers/staging/rtl8723bs/os_dep/recv_linux.c > > > b/drivers/staging/rtl8723bs/os_dep/recv_linux.c > > [] > > > @@ -110,7 +110,7 @@ void rtw_os_recv_indicate_pkt(struct adapter > > > *padapter, _pkt *pkt, struct rx_pkt > > > if (memcmp(pattrib->dst, > > > myid(&padapter->eeprompriv), ETH_ALEN)) { > > > if (bmcast) { > > > psta = > > > rtw_get_bcmc_stainfo(padapter); > > > - pskb2 = rtw_skb_clone(pkt); > > > + pskb2 = skb_clone(pkt, GFP_ATOMIC); > > > > Why make every clone allocation GFP_ATOMIC ? > > The rtw_os_recv_indicate_pkt() is always called from an interrupt handler.
It'd be better to indicate you know that in the changelog as the subject and changelog just shows removing wrappers and the patch code does not agree with that.