| Message ID | df882eddcf76b5d0ae53c19f368a617713462fd3.1670193080.git.lorenzo@kernel.org (mailing list archive) |
|---|---|
| State | Changes Requested |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net-next] net: ethernet: enetc: set frag flag for non-linear xdp buffers | expand |
Hi Lorenzo, On Sun, Dec 04, 2022 at 11:33:23PM +0100, Lorenzo Bianconi wrote: > Set missing XDP_FLAGS_HAS_FRAGS bit in enetc_add_rx_buff_to_xdp for > non-linear xdp buffers. > > Fixes: d1b15102dd16 ("net: enetc: add support for XDP_DROP and XDP_PASS") This can't be the Fixes: tag, struct xdp_buff didn't even have a "flags" field when that commit was introduced. Also, what does this change aim to achieve? It has a Fixes: tag but it's aimed for net-next. Is it to enable multi-buff XDP support? But we also have this in place, shouldn't that be deleted too? case XDP_REDIRECT: /* xdp_return_frame does not support S/G in the sense * that it leaks the fragments (__xdp_return should not * call page_frag_free only for the initial buffer). * Until XDP_REDIRECT gains support for S/G let's keep * the code structure in place, but dead. We drop the * S/G frames ourselves to avoid memory leaks which * would otherwise leave the kernel OOM. */ if (unlikely(cleaned_cnt - orig_cleaned_cnt != 1)) { enetc_xdp_drop(rx_ring, orig_i, i); rx_ring->stats.xdp_redirect_sg++; break; }
> Hi Lorenzo, Hi Vladimir, > > On Sun, Dec 04, 2022 at 11:33:23PM +0100, Lorenzo Bianconi wrote: > > Set missing XDP_FLAGS_HAS_FRAGS bit in enetc_add_rx_buff_to_xdp for > > non-linear xdp buffers. > > > > Fixes: d1b15102dd16 ("net: enetc: add support for XDP_DROP and XDP_PASS") > > This can't be the Fixes: tag, struct xdp_buff didn't even have a "flags" > field when that commit was introduced. yes, my fault, I did not check dates, I assumed this commit was done after xdp multi-buff support. We should get rid of the Fixes tag. > > Also, what does this change aim to achieve? It has a Fixes: tag but it's > aimed for net-next. Is it to enable multi-buff XDP support? But we also > have this in place, shouldn't that be deleted too? I think we should get rid of the chunk below as well. I am currently working on xdp feature support in order to allow XDP_REDIRECT for S/G xdp frames. I will post a v2. Regards, Lorenzo > > case XDP_REDIRECT: > /* xdp_return_frame does not support S/G in the sense > * that it leaks the fragments (__xdp_return should not > * call page_frag_free only for the initial buffer). > * Until XDP_REDIRECT gains support for S/G let's keep > * the code structure in place, but dead. We drop the > * S/G frames ourselves to avoid memory leaks which > * would otherwise leave the kernel OOM. > */ > if (unlikely(cleaned_cnt - orig_cleaned_cnt != 1)) { > enetc_xdp_drop(rx_ring, orig_i, i); > rx_ring->stats.xdp_redirect_sg++; > break; > }
diff --git a/drivers/net/ethernet/freescale/enetc/enetc.c b/drivers/net/ethernet/freescale/enetc/enetc.c index 8671591cb750..6d08ee2233aa 100644 --- a/drivers/net/ethernet/freescale/enetc/enetc.c +++ b/drivers/net/ethernet/freescale/enetc/enetc.c @@ -1416,6 +1416,8 @@ static void enetc_add_rx_buff_to_xdp(struct enetc_bdr *rx_ring, int i, skb_frag_size_set(frag, size); __skb_frag_set_page(frag, rx_swbd->page); + if (!xdp_buff_has_frags(xdp_buff)) + xdp_buff_set_frags_flag(xdp_buff); shinfo->nr_frags++; }
Set missing XDP_FLAGS_HAS_FRAGS bit in enetc_add_rx_buff_to_xdp for non-linear xdp buffers. Fixes: d1b15102dd16 ("net: enetc: add support for XDP_DROP and XDP_PASS") Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org> --- drivers/net/ethernet/freescale/enetc/enetc.c | 2 ++ 1 file changed, 2 insertions(+)