On 03/05/2024 03:35, Ping-Ke Shih wrote: > Bitterblue Smith <rtl8821cerfe2@xxxxxxxxx> wrote: >> The skb created in this function always has the same headroom, >> the chip's TX descriptor size. Use chip->tx_pkt_desc_sz directly. >> >> Signed-off-by: Bitterblue Smith <rtl8821cerfe2@xxxxxxxxx> >> --- >> This is the patch I promised earlier: >> https://lore.kernel.org/linux-wireless/cae2d330-a4fb-4570-9dde-09684af23ffd@xxxxxxxxx/ >> --- >> drivers/net/wireless/realtek/rtw88/usb.c | 14 +++++--------- >> 1 file changed, 5 insertions(+), 9 deletions(-) >> >> diff --git a/usb.c b/usb.c >> index 1dfe7c6ae4ba..ff57976b9d3b 100644 >> --- a/usb.c >> +++ b/usb.c >> @@ -440,23 +440,21 @@ static int rtw_usb_write_data(struct rtw_dev *rtwdev, >> { >> const struct rtw_chip_info *chip = rtwdev->chip; >> struct sk_buff *skb; >> - unsigned int desclen, headsize, size; >> + unsigned int size; >> u8 qsel; >> int ret = 0; >> >> size = pkt_info->tx_pkt_size; >> qsel = pkt_info->qsel; >> - desclen = chip->tx_pkt_desc_sz; >> - headsize = pkt_info->offset ? pkt_info->offset : desclen; >> >> - skb = dev_alloc_skb(headsize + size); >> + skb = dev_alloc_skb(chip->tx_pkt_desc_sz + size); >> if (unlikely(!skb)) >> return -ENOMEM; >> >> - skb_reserve(skb, headsize); >> + skb_reserve(skb, chip->tx_pkt_desc_sz); >> skb_put_data(skb, buf, size); >> - skb_push(skb, headsize); >> - memset(skb->data, 0, headsize); >> + skb_push(skb, chip->tx_pkt_desc_sz); >> + memset(skb->data, 0, chip->tx_pkt_desc_sz); >> rtw_tx_fill_tx_desc(pkt_info, skb); >> rtw_tx_fill_txdesc_checksum(rtwdev, pkt_info, skb->data); >> >> @@ -471,12 +469,10 @@ static int rtw_usb_write_data(struct rtw_dev *rtwdev, >> static int rtw_usb_write_data_rsvd_page(struct rtw_dev *rtwdev, u8 *buf, >> u32 size) >> { >> - const struct rtw_chip_info *chip = rtwdev->chip; >> struct rtw_tx_pkt_info pkt_info = {0}; >> >> pkt_info.tx_pkt_size = size; >> pkt_info.qsel = TX_DESC_QSEL_BEACON; >> - pkt_info.offset = chip->tx_pkt_desc_sz; > > pkt_info.offset is still used by rtw_tx_fill_tx_desc(), so you can't remove > this line. I think this is because Larry NAKed this patch. > Yes, that's the problem. I will send v2 without this hunk.
