Re: [Intel-wired-lan] [Question] ixgbe:Mechanism of RSS

2025-01-02 Thread Haifeng Xu
On 2025/1/3 00:01, Edward Cree wrote: > On 02/01/2025 11:23, Haifeng Xu wrote: >> We want to make full use of cpu resources to receive packets. So >> we enable 63 rx queues. But we found the rate of interrupt growth >> on cpu 0~15 is faster than other cpus(almost twice). > ... >> I am confused t

Re: [Intel-wired-lan] [Question] ixgbe:Mechanism of RSS

2025-01-02 Thread Haifeng Xu
On 2025/1/3 00:39, Jakub Kicinski wrote: > On Thu, 2 Jan 2025 16:01:18 + Edward Cree wrote: >> On 02/01/2025 11:23, Haifeng Xu wrote: >>> We want to make full use of cpu resources to receive packets. So >>> we enable 63 rx queues. But we found the rate of interrupt growth >>> on cpu 0~15 is

Re: [Intel-wired-lan] [Question] ixgbe:Mechanism of RSS

2025-01-02 Thread Haifeng Xu
On 2025/1/2 19:46, Eric Dumazet wrote: > On Thu, Jan 2, 2025 at 12:23 PM Haifeng Xu wrote: >> >> >> >> On 2025/1/2 18:34, Eric Dumazet wrote: >>> On Thu, Jan 2, 2025 at 9:43 AM Haifeng Xu wrote: On 2025/1/2 16:13, Eric Dumazet wrote: > On Thu, Jan 2, 2025 at 4:53 AM Hai

Re: [Intel-wired-lan] [PATCH net-next v2 0/8] net: napi: add CPU affinity to napi->config

2025-01-02 Thread Ahmed Zaki
On 2024-12-22 2:23 a.m., Shay Drori wrote: On 18/12/2024 18:58, Ahmed Zaki wrote: External email: Use caution opening links or attachments Move the IRQ affinity management to the napi struct. All drivers that are already using netif_napi_set_irq() are modified to the new API. Except mlx5

[Intel-wired-lan] [PATCH iwl-net] ice: Fix switchdev slow-path in LAG

2025-01-02 Thread Marcin Szycik
Ever since removing switchdev control VSI and using PF for port representor Tx/Rx, switchdev slow-path has been working improperly after failover in SR-IOV LAG. LAG assumes that the first uplink to be added to the aggregate will own VFs and have switchdev configured. After failing-over to the other

[Intel-wired-lan] [PATCH net-next 2/3] igc: Remove unused igc_read/write_pci_cfg wrappers

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" igc_read_pci_cfg() and igc_write_pci_cfg were added in 2018 as part of commit 146740f9abc4 ("igc: Add support for PF") but have remained unused. Remove them. Signed-off-by: Dr. David Alan Gilbert --- drivers/net/ethernet/intel/igc/igc_hw.h | 2 -- drivers/net

[Intel-wired-lan] [PATCH net-next 3/3] igc: Remove unused igc_read/write_pcie_cap_reg

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" The last uses of igc_read_pcie_cap_reg() and igc_write_pcie_cap_reg() were removed in 2019 by commit 16ecd8d9af26 ("igc: Remove the obsolete workaround") Remove them. Signed-off-by: Dr. David Alan Gilbert --- drivers/net/ethernet/intel/igc/igc_hw.h | 3 --- d

[Intel-wired-lan] [PATCH net-next 1/9] i40e: Deadcode i40e_aq_*

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" i40e_aq_add_mirrorrule(), i40e_aq_delete_mirrorrule() and i40e_aq_set_vsi_vlan_promisc() were added in 2016 by commit 7bd6875bef70 ("i40e: APIs to Add/remove port mirroring rules") but haven't been used. They were the last user of i40e_mirrorrule_op(). i40e_aq_rea

[Intel-wired-lan] [PATCH net-next 0/3] igc deadcoding

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" Hi, This set removes some functions that are entirely unused and have been since ~2018. Build tested. Signed-off-by: Dr. David Alan Gilbert (Repost now netdev is open) Dr. David Alan Gilbert (3): igc: Remove unused igc_acquire/release_nvm igc: Remove unuse

[Intel-wired-lan] [PATCH net-next 1/3] igc: Remove unused igc_acquire/release_nvm

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" igc_acquire_nvm() and igc_release_nvm() were added in 2018 as part of commit ab4056126813 ("igc: Add NVM support") but never used. Remove them. The igc_1225.c has it's own specific implementations. Signed-off-by: Dr. David Alan Gilbert --- drivers/net/ethernet

[Intel-wired-lan] [PATCH net-next] ixgbevf: Remove unused ixgbevf_hv_mbx_ops

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" The const struct ixgbevf_hv_mbx_ops was added in 2016 as part of commit c6d45171d706 ("ixgbevf: Support Windows hosts (Hyper-V)") but has remained unused. The functions it references are still referenced elsewhere. Remove it. Signed-off-by: Dr. David Alan Gilber

[Intel-wired-lan] [PATCH net-next 2/9] i40e: Remove unused i40e_blink_phy_link_led

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" i40e_blink_phy_link_led() was added in 2016 by commit fd077cd3399b ("i40e: Add functions to blink led on 10GBaseT PHY") but hasn't been used. Remove it. Signed-off-by: Dr. David Alan Gilbert Reviewed-by: Kalesh AP --- drivers/net/ethernet/intel/i40e/i40e_commo

[Intel-wired-lan] [PATCH net-next 0/9] i40e deadcoding

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" Hi, This is a bunch of deadcoding of functions that are entirely uncalled in the i40e driver. Build tested only. Signed-off-by: Dr. David Alan Gilbert (Resend now that netdev has woken up) Dr. David Alan Gilbert (9): i40e: Deadcode i40e_aq_* i40e: Remove

[Intel-wired-lan] [PATCH net-next 4/9] i40e: Deadcode profile code

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" i40e_add_pinfo_to_list() was added in 2017 by commit 1d5c960c5ef5 ("i40e: new AQ commands") i40e_find_section_in_profile() was added in 2019 by commit cdc594e00370 ("i40e: Implement DDP support in i40e driver") Neither have been used. Remove them. Signed-off-by:

[Intel-wired-lan] [PATCH net-next 3/9] i40e: Remove unused i40e_(read|write)_phy_register

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" i40e_read_phy_register() and i40e_write_phy_register() were added in 2016 by commit f62ba91458b5 ("i40e: Add functions which apply correct PHY access method for read and write operation") but haven't been used. Remove them. (There are more specific _clause* varia

[Intel-wired-lan] [PATCH net-next 9/9] i40e: Remove unused i40e_dcb_hw_get_num_tc

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" The last useof i40e_dcb_hw_get_num_tc() was removed in 2022 by commit fe20371578ef ("Revert "i40e: Fix reset bw limit when DCB enabled with 1 TC"") Remove it. Signed-off-by: Dr. David Alan Gilbert Reviewed-by: Kalesh AP --- drivers/net/ethernet/intel/i40e/i40e_

[Intel-wired-lan] [PATCH net-next 6/9] i40e: Remove unused i40e_del_filter

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" The last use of i40e_del_filter() was removed in 2016 by commit 9569a9a4547d ("i40e: when adding or removing MAC filters, correctly handle VLANs") Remove it. Fix up a comment that referenced it. Note: The __ version of this function is still used. Signed-off-by:

[Intel-wired-lan] [PATCH net-next 8/9] i40e: Remove unused i40e_asq_send_command_v2

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" i40e_asq_send_command_v2() was added in 2022 by commit 74073848b0d7 ("i40e: Add new versions of send ASQ command functions") but hasn't been used. Remove it. (The _atomic_v2 version of the function is used, so leave it). Signed-off-by: Dr. David Alan Gilbert ---

[Intel-wired-lan] [PATCH net-next 5/9] i40e: Remove unused i40e_get_cur_guaranteed_fd_count

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" The last use of i40e_get_cur_guaranteed_fd_count() was removed in 2015 by commit 04294e38a451 ("i40e: FD filters flush policy changes") Remove it. Signed-off-by: Dr. David Alan Gilbert Reviewed-by: Kalesh AP --- drivers/net/ethernet/intel/i40e/i40e.h | 1

[Intel-wired-lan] [PATCH net-next 7/9] i40e: Remove unused i40e_commit_partition_bw_setting

2025-01-02 Thread linux
From: "Dr. David Alan Gilbert" i40e_commit_partition_bw_setting() was added in 2017 by commit 4fc8c6763957 ("i40e: genericize the partition bandwidth control") but hasn't been used. Remove it. Signed-off-by: Dr. David Alan Gilbert Reviewed-by: Kalesh AP --- drivers/net/ethernet/intel/i40e/i4

Re: [Intel-wired-lan] [Question] ixgbe:Mechanism of RSS

2025-01-02 Thread Jakub Kicinski
On Thu, 2 Jan 2025 16:01:18 + Edward Cree wrote: > On 02/01/2025 11:23, Haifeng Xu wrote: > > We want to make full use of cpu resources to receive packets. So > > we enable 63 rx queues. But we found the rate of interrupt growth > > on cpu 0~15 is faster than other cpus(almost twice). > ... >

Re: [Intel-wired-lan] [Question] ixgbe:Mechanism of RSS

2025-01-02 Thread Edward Cree
On 02/01/2025 11:23, Haifeng Xu wrote: > We want to make full use of cpu resources to receive packets. So > we enable 63 rx queues. But we found the rate of interrupt growth > on cpu 0~15 is faster than other cpus(almost twice). ... > I am confused that why ixgbe NIC can dispatch the packets > to t

Re: [Intel-wired-lan] [Question] ixgbe:Mechanism of RSS

2025-01-02 Thread Eric Dumazet
On Thu, Jan 2, 2025 at 12:23 PM Haifeng Xu wrote: > > > > On 2025/1/2 18:34, Eric Dumazet wrote: > > On Thu, Jan 2, 2025 at 9:43 AM Haifeng Xu wrote: > >> > >> > >> > >> On 2025/1/2 16:13, Eric Dumazet wrote: > >>> On Thu, Jan 2, 2025 at 4:53 AM Haifeng Xu wrote: > > Hi masters, >

Re: [Intel-wired-lan] [Question] ixgbe:Mechanism of RSS

2025-01-02 Thread Haifeng Xu
On 2025/1/2 18:34, Eric Dumazet wrote: > On Thu, Jan 2, 2025 at 9:43 AM Haifeng Xu wrote: >> >> >> >> On 2025/1/2 16:13, Eric Dumazet wrote: >>> On Thu, Jan 2, 2025 at 4:53 AM Haifeng Xu wrote: Hi masters, We use the Intel Corporation 82599ES NIC in our production

Re: [Intel-wired-lan] [Question] ixgbe:Mechanism of RSS

2025-01-02 Thread Eric Dumazet
On Thu, Jan 2, 2025 at 9:43 AM Haifeng Xu wrote: > > > > On 2025/1/2 16:13, Eric Dumazet wrote: > > On Thu, Jan 2, 2025 at 4:53 AM Haifeng Xu wrote: > >> > >> Hi masters, > >> > >> We use the Intel Corporation 82599ES NIC in our production > >> environment. And it has 63 rx queues, every

Re: [Intel-wired-lan] [Question] ixgbe:Mechanism of RSS

2025-01-02 Thread Haifeng Xu
On 2025/1/2 16:13, Eric Dumazet wrote: > On Thu, Jan 2, 2025 at 4:53 AM Haifeng Xu wrote: >> >> Hi masters, >> >> We use the Intel Corporation 82599ES NIC in our production >> environment. And it has 63 rx queues, every rx queue interrupt is processed >> by a single cpu. >> Th

Re: [Intel-wired-lan] [Question] ixgbe:Mechanism of RSS

2025-01-02 Thread Eric Dumazet
On Thu, Jan 2, 2025 at 4:53 AM Haifeng Xu wrote: > > Hi masters, > > We use the Intel Corporation 82599ES NIC in our production > environment. And it has 63 rx queues, every rx queue interrupt is processed > by a single cpu. > The RSS configuration can be seen as follow: > >