| Message ID | 20240729200716.681496-1-anthony.l.nguyen@intel.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | ice: fix AF_XDP ZC timeout and concurrency issues | expand |
Hello: This series was applied to netdev/net.git (main) by Tony Nguyen <anthony.l.nguyen@intel.com>: On Mon, 29 Jul 2024 13:07:06 -0700 you wrote: > Maciej Fijalkowski says: > > Changes included in this patchset address an issue that customer has > been facing when AF_XDP ZC Tx sockets were used in combination with flow > control and regular Tx traffic. > > After executing: > ethtool --set-priv-flags $dev link-down-on-close on > ethtool -A $dev rx on tx on > > [...] Here is the summary with links: - [net,v2,1/8] ice: respect netif readiness in AF_XDP ZC related ndo's https://git.kernel.org/netdev/net/c/ec145a18687f - [net,v2,2/8] ice: don't busy wait for Rx queue disable in ice_qp_dis() https://git.kernel.org/netdev/net/c/1ff72a2f6779 - [net,v2,3/8] ice: replace synchronize_rcu with synchronize_net https://git.kernel.org/netdev/net/c/405d9999aa0b - [net,v2,4/8] ice: modify error handling when setting XSK pool in ndo_bpf https://git.kernel.org/netdev/net/c/d59227179949 - [net,v2,5/8] ice: toggle netif_carrier when setting up XSK pool https://git.kernel.org/netdev/net/c/9da75a511c55 - [net,v2,6/8] ice: improve updating ice_{t,r}x_ring::xsk_pool https://git.kernel.org/netdev/net/c/ebc33a3f8d0a - [net,v2,7/8] ice: add missing WRITE_ONCE when clearing ice_rx_ring::xdp_prog https://git.kernel.org/netdev/net/c/6044ca26210b - [net,v2,8/8] ice: xsk: fix txq interrupt mapping https://git.kernel.org/netdev/net/c/963fb4612295 You are awesome, thank you!
Maciej Fijalkowski says: Changes included in this patchset address an issue that customer has been facing when AF_XDP ZC Tx sockets were used in combination with flow control and regular Tx traffic. After executing: ethtool --set-priv-flags $dev link-down-on-close on ethtool -A $dev rx on tx on launching multiple ZC Tx sockets on $dev + pinging remote interface (so that regular Tx traffic is present) and then going through down/up of $dev, Tx timeout occurred and then most of the time ice driver was unable to recover from that state. These patches combined together solve the described above issue on customer side. Main focus here is to forbid producing Tx descriptors when either carrier is not yet initialized or process of bringing interface down has already started. v2: * in patch 6, use a single READ_ONCE against xsk_pool within napi [Jakub] v1: https://lore.kernel.org/netdev/20240708221416.625850-1-anthony.l.nguyen@intel.com/ --- Olek, we decided not to check IFF_UP as you initially suggested. Reason is that when link goes down netif_running() has broader scope than IFF_UP being set as the former (the __LINK_STATE_START bit) is cleared earlier in the core. The following are changes since commit 039564d2fd37b122ec0d268e2ee6334e7169e225: Merge branch 'mptcp-endpoint-readd-fixes' into main and are available in the git repository at: git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue 100GbE Maciej Fijalkowski (7): ice: don't busy wait for Rx queue disable in ice_qp_dis() ice: replace synchronize_rcu with synchronize_net ice: modify error handling when setting XSK pool in ndo_bpf ice: toggle netif_carrier when setting up XSK pool ice: improve updating ice_{t,r}x_ring::xsk_pool ice: add missing WRITE_ONCE when clearing ice_rx_ring::xdp_prog ice: xsk: fix txq interrupt mapping Michal Kubiak (1): ice: respect netif readiness in AF_XDP ZC related ndo's drivers/net/ethernet/intel/ice/ice.h | 11 +- drivers/net/ethernet/intel/ice/ice_base.c | 4 +- drivers/net/ethernet/intel/ice/ice_main.c | 2 +- drivers/net/ethernet/intel/ice/ice_txrx.c | 10 +- drivers/net/ethernet/intel/ice/ice_xsk.c | 184 +++++++++++++--------- drivers/net/ethernet/intel/ice/ice_xsk.h | 14 +- 6 files changed, 135 insertions(+), 90 deletions(-)