From: Jason Gunthorpe To: Jonathan Corbet , Itay Avraham , Jakub Kicinski , Leon Romanovsky , linux-doc@vger.kernel.org, linux-rdma@vger.kernel.org, netdev@vger.kernel.org, Paolo Abeni , Saeed Mahameed , Tariq Toukan Cc: Andy Gospodarek , Aron Silverton , Dan Williams , David Ahern , Christoph Hellwig , Jiri Pirko , Leonid Bloch , Leon Romanovsky , linux-cxl@vger.kernel.org, patches@lists.linux.dev Subject: [PATCH v2 0/8] Introduce fwctl subystem Date: Mon, 24 Jun 2024 19:47:24 -0300 Message-ID: <0-v2-940e479ceba9+3821-fwctl_jgg@nvidia.com> Content-Transfer-Encoding: 8bit Content-Type: text/plain X-Mailing-List: linux-cxl@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Xref: photonic.trudheim.com org.kernel.vger.linux-cxl:29104 org.kernel.vger.netdev:354915 Newsgroups: org.kernel.vger.linux-cxl,dev.linux.lists.patches,org.kernel.vger.linux-doc,org.kernel.vger.linux-rdma,org.kernel.vger.netdev Path: photonic.trudheim.com!nntp.lore.kernel.org!not-for-mail fwctl is a new subsystem intended to bring some common rules and order to the growing pattern of exposing a secure FW interface directly to userspace. Unlike existing places like RDMA/DRM/VFIO/uacce that are exposing a device for datapath operations fwctl is focused on debugging, configuration and provisioning of the device. It will not have the necessary features like interrupt delivery to support a datapath. This concept is similar to the long standing practice in the "HW" RAID space of having a device specific misc device to manager the RAID controller FW. fwctl generalizes this notion of a companion debug and management interface that goes along with a dataplane implemented in an appropriate subsystem. The need for this has reached a critical point as many users are moving to run lockdown enabled kernels. Several existing devices have had long standing tooling for management that relied on /sys/../resource0 or PCI config space access which is not permitted in lockdown. A major point of fwctl is to define and document the rules that a device must follow to expose a lockdown compatible RPC. Based on some discussion fwctl splits the RPCs into four categories FWCTL_RPC_CONFIGURATION FWCTL_RPC_DEBUG_READ_ONLY FWCTL_RPC_DEBUG_WRITE FWCTL_RPC_DEBUG_WRITE_FULL Where the latter two trigger a new TAINT_FWCTL, and the final one requires CAP_SYS_RAWIO - excluding it from lockdown. The device driver and its FW would be responsible to restrict RPCs to the requested security scope, while the core code handles the tainting and CAP checks. For details see the final patch which introduces the documentation. This series incorporates a version of the mlx5ctl interface previously proposed: https://lore.kernel.org/r/20240207072435.14182-1-saeed@kernel.org/ For this series the memory registration mechanism was removed, but I expect it will come back. This series comes with mlx5 as a driver implementation, and I have soft commitments for at least three more drivers. There have been two LWN articles written discussing various aspects of this proposal: https://lwn.net/Articles/955001/ https://lwn.net/Articles/969383/ Several have expressed general support for this concept: Broadcom Networking - https://lore.kernel.org/r/Zf2n02q0GevGdS-Z@C02YVCJELVCG Christoph Hellwig - https://lore.kernel.org/r/Zcx53N8lQjkpEu94@infradead.org/ Enfabrica - https://lore.kernel.org/r/9cc7127f-8674-43bc-b4d7-b1c4c2d96fed@kernel.org/ NVIDIA Networking Oracle Linux - https://lore.kernel.org/r/6lakj6lxlxhdgrewodvj3xh6sxn3d36t5dab6najzyti2navx3@wrge7cyfk6nq Work is ongoing for a robust multi-device open source userspace, currently the mlx5ctl_user that was posted by Saeed has been updated to use fwctl. https://github.com/saeedtx/mlx5ctl.git https://github.com/jgunthorpe/mlx5ctl.git This is on github: https://github.com/jgunthorpe/linux/commits/fwctl v2: - Rebase to v6.10-rc5 - Minor style changes - Follow the style consensus for the guard stuff - Documentation grammer/spelling - Add missed length output for mlx5 get_info - Add two more missed MLX5 CMD's - Collect tags v1: https://lore.kernel.org/r/0-v1-9912f1a11620+2a-fwctl_jgg@nvidia.com Jason Gunthorpe (6): fwctl: Add basic structure for a class subsystem with a cdev fwctl: Basic ioctl dispatch for the character device fwctl: FWCTL_INFO to return basic information about the device taint: Add TAINT_FWCTL fwctl: FWCTL_RPC to execute a Remote Procedure Call to device firmware fwctl: Add documentation Saeed Mahameed (2): fwctl/mlx5: Support for communicating with mlx5 fw mlx5: Create an auxiliary device for fwctl_mlx5 Documentation/admin-guide/tainted-kernels.rst | 5 + Documentation/userspace-api/fwctl.rst | 269 ++++++++++++ Documentation/userspace-api/index.rst | 1 + .../userspace-api/ioctl/ioctl-number.rst | 1 + MAINTAINERS | 16 + drivers/Kconfig | 2 + drivers/Makefile | 1 + drivers/fwctl/Kconfig | 23 + drivers/fwctl/Makefile | 5 + drivers/fwctl/main.c | 412 ++++++++++++++++++ drivers/fwctl/mlx5/Makefile | 4 + drivers/fwctl/mlx5/main.c | 337 ++++++++++++++ drivers/net/ethernet/mellanox/mlx5/core/dev.c | 8 + include/linux/fwctl.h | 112 +++++ include/linux/panic.h | 3 +- include/uapi/fwctl/fwctl.h | 137 ++++++ include/uapi/fwctl/mlx5.h | 36 ++ kernel/panic.c | 1 + 18 files changed, 1372 insertions(+), 1 deletion(-) create mode 100644 Documentation/userspace-api/fwctl.rst create mode 100644 drivers/fwctl/Kconfig create mode 100644 drivers/fwctl/Makefile create mode 100644 drivers/fwctl/main.c create mode 100644 drivers/fwctl/mlx5/Makefile create mode 100644 drivers/fwctl/mlx5/main.c create mode 100644 include/linux/fwctl.h create mode 100644 include/uapi/fwctl/fwctl.h create mode 100644 include/uapi/fwctl/mlx5.h base-commit: f2661062f16b2de5d7b6a5c42a9a5c96326b8454 -- 2.45.2 . Date: Tue, 25 Jun 2024 00:12:26 +0000 X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: Mime-Version: 1.0 Message-ID: <20240625001232.1476315-1-ziweixiao@google.com> Subject: [PATCH net-next v3 0/5] gve: Add flow steering support From: Ziwei Xiao To: netdev@vger.kernel.org Cc: jeroendb@google.com, pkaligineedi@google.com, shailend@google.com, davem@davemloft.net, edumazet@google.com, kuba@kernel.org, pabeni@redhat.com, willemb@google.com, hramamurthy@google.com, ziweixiao@google.com, rushilg@google.com, horms@kernel.org, linux-kernel@vger.kernel.org Content-Type: text/plain; charset="UTF-8" Xref: photonic.trudheim.com org.kernel.vger.linux-kernel:1257534 org.kernel.vger.netdev:354916 Newsgroups: org.kernel.vger.linux-kernel,org.kernel.vger.netdev Path: photonic.trudheim.com!nntp.lore.kernel.org!not-for-mail To support flow steering in GVE driver, there are two adminq changes need to be made in advance. The first one is adding adminq mutex lock, which is to allow the incoming flow steering operations to be able to temporarily drop the rtnl_lock to reduce the latency for registering flow rules among several NICs at the same time. This could be achieved by the future changes to reduce the drivers' dependencies on the rtnl lock for particular ethtool ops. The second one is to add the extended adminq command so that we can support larger adminq command such as configure_flow_rule command. In that patch, there is a new added function called gve_adminq_execute_extended_cmd with the attribute of __maybe_unused. That attribute will be removed in the third patch of this series where it will use the previously unused function. And the other three patches are needed for the actual flow steering feature support in driver. Jeroen de Borst (4): gve: Add adminq extended command gve: Add flow steering device option gve: Add flow steering adminq commands gve: Add flow steering ethtool support Ziwei Xiao (1): gve: Add adminq mutex lock drivers/net/ethernet/google/gve/Makefile | 2 +- drivers/net/ethernet/google/gve/gve.h | 54 +++- drivers/net/ethernet/google/gve/gve_adminq.c | 228 +++++++++++++- drivers/net/ethernet/google/gve/gve_adminq.h | 103 ++++++ drivers/net/ethernet/google/gve/gve_ethtool.c | 72 ++++- .../net/ethernet/google/gve/gve_flow_rule.c | 298 ++++++++++++++++++ drivers/net/ethernet/google/gve/gve_main.c | 83 ++++- 7 files changed, 817 insertions(+), 23 deletions(-) create mode 100644 drivers/net/ethernet/google/gve/gve_flow_rule.c -- 2.45.2.741.gdbec12cfda-goog . From: Kuniyuki Iwashima To: "David S. Miller" , Eric Dumazet , Jakub Kicinski , Paolo Abeni CC: Rao Shoaib , Kuniyuki Iwashima , Kuniyuki Iwashima , Subject: [PATCH v1 net 00/11] af_unix: Fix bunch of MSG_OOB bugs and add new tests. Date: Mon, 24 Jun 2024 18:36:34 -0700 Message-ID: <20240625013645.45034-1-kuniyu@amazon.com> X-Mailing-List: netdev@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Content-Type: text/plain Xref: photonic.trudheim.com org.kernel.vger.netdev:354922 Newsgroups: org.kernel.vger.netdev Path: photonic.trudheim.com!nntp.lore.kernel.org!not-for-mail This series rewrites the selftest for AF_UNIX MSG_OOB and fixes bunch of bugs that AF_UNIX behaves differently compared to TCP. Note that the test discovered few more bugs in TCP side, which will be fixed in another series. Kuniyuki Iwashima (11): selftest: af_unix: Remove test_unix_oob.c. selftest: af_unix: Add msg_oob.c. af_unix: Stop recv(MSG_PEEK) at consumed OOB skb. af_unix: Don't stop recv(MSG_DONTWAIT) if consumed OOB skb is at the head. selftest: af_unix: Add non-TCP-compliant test cases in msg_oob.c. af_unix: Don't stop recv() at consumed ex-OOB skb. selftest: af_unix: Add SO_OOBINLINE test cases in msg_oob.c selftest: af_unix: Check SIGURG after every send() in msg_oob.c selftest: af_unix: Check EPOLLPRI after every send()/recv() in msg_oob.c af_unix: Fix wrong ioctl(SIOCATMARK) when consumed OOB skb is at the head. selftest: af_unix: Check SIOCATMARK after every send()/recv() in msg_oob.c. net/unix/af_unix.c | 37 +- tools/testing/selftests/net/.gitignore | 1 - tools/testing/selftests/net/af_unix/Makefile | 2 +- tools/testing/selftests/net/af_unix/msg_oob.c | 734 ++++++++++++++++++ .../selftests/net/af_unix/test_unix_oob.c | 436 ----------- 5 files changed, 766 insertions(+), 444 deletions(-) create mode 100644 tools/testing/selftests/net/af_unix/msg_oob.c delete mode 100644 tools/testing/selftests/net/af_unix/test_unix_oob.c -- 2.30.2 . From: Li RongQing To: netdev@vger.kernel.org Cc: Li RongQing Subject: [PATCH,net-next,v2][resend] virtio_net: Remove u64_stats_update_begin()/end() for stats fetch Date: Tue, 25 Jun 2024 09:34:07 +0800 Message-Id: <20240625013407.25436-1-lirongqing@baidu.com> X-Mailing-List: netdev@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: Xref: photonic.trudheim.com org.kernel.vger.netdev:354932 Newsgroups: org.kernel.vger.netdev Path: photonic.trudheim.com!nntp.lore.kernel.org!not-for-mail This place is fetching the stats, u64_stats_update_begin()/end() should not be used, and the fetcher of stats is in the same context as the updater of the stats, so don't need any protection Acked-by: Michael S. Tsirkin Suggested-by: Jakub Kicinski Signed-off-by: Li RongQing --- drivers/net/virtio_net.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c index b1f8b72..f212977 100644 --- a/drivers/net/virtio_net.c +++ b/drivers/net/virtio_net.c @@ -2382,12 +2382,13 @@ static void virtnet_rx_dim_update(struct virtnet_info *vi, struct receive_queue if (!rq->packets_in_napi) return; - u64_stats_update_begin(&rq->stats.syncp); + /* Don't need protection when fetching stats, since fetcher and + * updater of the stats are in same context + */ dim_update_sample(rq->calls, u64_stats_read(&rq->stats.packets), u64_stats_read(&rq->stats.bytes), &cur_sample); - u64_stats_update_end(&rq->stats.syncp); net_dim(&rq->dim, cur_sample); rq->packets_in_napi = 0; -- 2.9.4 . From: echken To: pshelar@ovn.org, davem@davemloft.net, edumazet@google.com, kuba@kernel.org, pabeni@redhat.com Cc: netdev@vger.kernel.org, dev@openvswitch.org, linux-kernel@vger.kernel.org, echken Subject: [PATCH 1/2] Add GSO UDP Offloading feature to OVS Internal Port Date: Tue, 25 Jun 2024 08:29:24 +0000 Message-Id: <20240625082924.775877-1-chengcheng.luo@smartx.com> X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Xref: photonic.trudheim.com org.kernel.vger.linux-kernel:1257853 org.kernel.vger.netdev:354948 Newsgroups: org.kernel.vger.linux-kernel,org.kernel.vger.netdev Path: photonic.trudheim.com!nntp.lore.kernel.org!not-for-mail The OVS internal port does not support UDP fragmentation offloading, resulting in large packets sent through the OVS internal port to OVS being prematurely fragmented. This increases the total number of packets processed in the path from the vport to the OVS bridge output port, affecting transmission efficiency. Signed-off-by: echken --- net/openvswitch/vport-internal_dev.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/net/openvswitch/vport-internal_dev.c b/net/openvswitch/vport-internal_dev.c index 74c88a6baa43..c5a72c4dc6fd 100644 --- a/net/openvswitch/vport-internal_dev.c +++ b/net/openvswitch/vport-internal_dev.c @@ -110,7 +110,8 @@ static void do_setup(struct net_device *netdev) netdev->features = NETIF_F_LLTX | NETIF_F_SG | NETIF_F_FRAGLIST | NETIF_F_HIGHDMA | NETIF_F_HW_CSUM | - NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL; + NETIF_F_GSO_SOFTWARE | NETIF_F_GSO_ENCAP_ALL | + NETIF_F_GSO_UDP | NETIF_F_GSO_UDP_L4; netdev->vlan_features = netdev->features; netdev->hw_enc_features = netdev->features; -- 2.34.1 . From: Amit Cohen To: CC: , , , , , , , , Amit Cohen Subject: [PATCH RFC net-next 0/4] Adjust page pool netlink filling to non common case Date: Tue, 25 Jun 2024 15:08:03 +0300 Message-ID: <20240625120807.1165581-1-amcohen@nvidia.com> X-Mailing-List: netdev@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Content-Type: text/plain Xref: photonic.trudheim.com org.kernel.vger.netdev:354967 Newsgroups: org.kernel.vger.netdev Path: photonic.trudheim.com!nntp.lore.kernel.org!not-for-mail Most network drivers has 1:1 mapping between netdevice and event queues, so then each page pool is used by only one netdevice. This is not the case in mlxsw driver. Currently, the netlink message is filled with 'pool->slow.netdev->ifindex', which should be NULL in case that several netdevices use the same pool. Adjust page pool netlink filling to use the netdevice which the pool is stored in its list. See more info in commit messages. Without this set, mlxsw driver cannot dump all page pools: $ ./tools/net/ynl/cli.py --spec Documentation/netlink/specs/netdev.yaml \ --dump page-pool-stats-get --output-json | jq [] With this set, "dump" command prints all the page pools for all the netdevices: $ ./tools/net/ynl/cli.py --spec Documentation/netlink/specs/netdev.yaml \ --dump page-pool-get --output-json | \ jq -e ".[] | select(.ifindex == 64)" | grep "napi-id" | wc -l 56 From driver POV, such queries are supported by associating the pools with an unregistered netdevice (dummy netdevice). The following limitations are caused by such implementation: 1. The get command output specifies the 'ifindex' as 0, which is meaningless. `iproute2` will print this as "*", but there might be other tools which fail in such case. 2. get command does not work when devlink instance is reloaded to namespace which is not the initial one, as the dummy device associated with the pools belongs to the initial namespace. See examples in commit messages. We would like to expose page pool stats and info via the standard interface, but such implementation is not perfect. An additional option is to use debugfs, but we prefer to avoid it, if it is possible. Any suggestions for better implementation in case of pool for several netdevices will be welcomed. Patch set overview: Patch #1 makes netlink filling code more flex Patch #2 changes the 'ifindex' which is used for dump Patch #3 sets netdevice for page pools in mlxsw driver, to allow "do" commands Patch #4 sets page pools list for netdevices in mlxsw driver, to allow "dump" commands Amit Cohen (4): net: core: page_pool_user: Allow flexibility of 'ifindex' value net: core: page_pool_user: Change 'ifindex' for page pool dump mlxsw: pci: Allow get page pool info/stats via netlink mlxsw: Set page pools list for netdevices drivers/net/ethernet/mellanox/mlxsw/core.c | 6 +++++ drivers/net/ethernet/mellanox/mlxsw/core.h | 2 ++ drivers/net/ethernet/mellanox/mlxsw/pci.c | 9 ++++++++ .../net/ethernet/mellanox/mlxsw/spectrum.c | 2 ++ net/core/page_pool_user.c | 22 +++++++++---------- 5 files changed, 29 insertions(+), 12 deletions(-) -- 2.45.1 . From: Kory Maincent Subject: [PATCH net-next v4 0/7] net: pse-pd: Add new PSE c33 features Date: Tue, 25 Jun 2024 14:33:45 +0200 Message-Id: <20240625-feature_poe_power_cap-v4-0-b0813aad57d5@bootlin.com> X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 8bit To: "David S. Miller" , Eric Dumazet , Jakub Kicinski , Paolo Abeni , Donald Hunter , Oleksij Rempel , Jonathan Corbet Cc: Thomas Petazzoni , linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Dent Project , kernel@pengutronix.de, linux-doc@vger.kernel.org, Kory Maincent Xref: photonic.trudheim.com org.kernel.vger.linux-kernel:1258141 org.kernel.vger.netdev:354973 Newsgroups: org.kernel.vger.linux-kernel,org.kernel.vger.linux-doc,org.kernel.vger.netdev Path: photonic.trudheim.com!nntp.lore.kernel.org!not-for-mail From: Kory Maincent (Dent Project) This patch series adds new c33 features to the PSE API. - Expand the PSE PI informations status with power, class and failure reason - Add the possibility to get and set the PSE PIs power limit Changes in v4: - Made few update in PSE extended state an substate. - Add support for c33 pse power limit ranges. - Few changes in the specs and the documentation. - Link to v3: https://lore.kernel.org/r/20240614-feature_poe_power_cap-v3-0-a26784e78311@bootlin.com Changes in v3: - Use u32 instead of u8 size for c33 pse extended state and substate. - Reformat the state and substate enumeration to follow Oleksij proposal which is more IEEE 802.3 standard compliant - Sent the first patch standalone in net. - Link to v2: https://lore.kernel.org/r/20240607-feature_poe_power_cap-v2-0-c03c2deb83ab@bootlin.com Changes in v2: - Use uA and uV instead of mA and mV to have more precision in the power calculation. Need to use 64bit variables for the calculation. - Modify the pd-92x0behavior in case of setting the current out of the available ranges. Report an error now. - Link to v1: https://lore.kernel.org/r/20240529-feature_poe_power_cap-v1-0-0c4b1d5953b8@bootlin.com Signed-off-by: Kory Maincent --- Kory Maincent (7): net: ethtool: pse-pd: Expand C33 PSE status with class, power and extended state netlink: specs: Expand the PSE netlink command with C33 new features net: pse-pd: pd692x0: Expand ethtool status message net: pse-pd: Add new power limit get and set c33 features net: ethtool: Add new power limit get and set features netlink: specs: Expand the PSE netlink command with C33 pw-limit attributes net: pse-pd: pd692x0: Enhance with new current limit and voltage read callbacks Documentation/netlink/specs/ethtool.yaml | 58 +++++ Documentation/networking/ethtool-netlink.rst | 87 +++++++- drivers/net/pse-pd/pd692x0.c | 317 ++++++++++++++++++++++++++- drivers/net/pse-pd/pse_core.c | 172 ++++++++++++++- include/linux/ethtool.h | 20 ++ include/linux/pse-pd/pse.h | 51 +++++ include/uapi/linux/ethtool.h | 191 ++++++++++++++++ include/uapi/linux/ethtool_netlink.h | 12 + net/ethtool/pse-pd.c | 119 +++++++++- 9 files changed, 997 insertions(+), 30 deletions(-) --- base-commit: b98d89947d124c8ca9150b32fdc0c5fb099243cf change-id: 20240425-feature_poe_power_cap-18e90ba7294b Best regards, -- Köry Maincent, Bootlin Embedded Linux and kernel engineering https://bootlin.com . From: Yunsheng Lin To: , , CC: , , Yunsheng Lin , Alexander Duyck , Alexei Starovoitov , Daniel Borkmann , Jesper Dangaard Brouer , John Fastabend , Matthias Brugger , AngeloGioacchino Del Regno , , , Subject: [PATCH net-next v9 00/13] First try to replace page_frag with page_frag_cache Date: Tue, 25 Jun 2024 21:52:03 +0800 Message-ID: <20240625135216.47007-1-linyunsheng@huawei.com> X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Content-Type: text/plain Xref: photonic.trudheim.com org.kernel.vger.linux-kernel:1258244 org.kernel.vger.netdev:354983 Newsgroups: org.kernel.vger.linux-kernel,org.infradead.lists.linux-arm-kernel,org.infradead.lists.linux-mediatek,org.kernel.vger.bpf,org.kernel.vger.netdev Path: photonic.trudheim.com!nntp.lore.kernel.org!not-for-mail After [1], there are still two implementations for page frag: 1. mm/page_alloc.c: net stack seems to be using it in the rx part with 'struct page_frag_cache' and the main API being page_frag_alloc_align(). 2. net/core/sock.c: net stack seems to be using it in the tx part with 'struct page_frag' and the main API being skb_page_frag_refill(). This patchset tries to unfiy the page frag implementation by replacing page_frag with page_frag_cache for sk_page_frag() first. net_high_order_alloc_disable_key for the implementation in net/core/sock.c doesn't seems matter that much now have have pcp support for high-order pages in commit 44042b449872 ("mm/page_alloc: allow high-order pages to be stored on the per-cpu lists"). As the related change is mostly related to networking, so targeting the net-next. And will try to replace the rest of page_frag in the follow patchset. After this patchset: 1. Unify the page frag implementation by taking the best out of two the existing implementations: we are able to save some space for the 'page_frag_cache' API user, and avoid 'get_page()' for the old 'page_frag' API user. 2. Future bugfix and performance can be done in one place, hence improving maintainability of page_frag's implementation. Kernel Image changing: Linux Kernel total | text data bss ------------------------------------------------------ after 45250307 | 27274279 17209996 766032 before 45254134 | 27278118 17209984 766032 delta -3827 | -3839 +12 +0 Performance validation: 1. Using micro-benchmark ko added in patch 1 to test aligned and non-aligned API performance impact for the existing users, there is no notiable performance degradation. Instead we seems to some minor performance boot for both aligned and non-aligned API after this patchset as below. 2. Use the below netcat test case, we also have some minor performance boot for repalcing 'page_frag' with 'page_frag_cache' after this patchset. server: taskset -c 32 nc -l -k 1234 > /dev/null client: perf stat -r 200 -- taskset -c 0 head -c 20G /dev/zero | taskset -c 1 nc 127.0.0.1 1234 In order to avoid performance noise as much as possible, the testing is done in system without any other laod and have enough iterations to prove the data is stable enogh, complete log for testing is below: taskset -c 32 nc -l -k 1234 > /dev/null perf stat -r 200 -- insmod ./page_frag_test.ko test_push_cpu=16 test_pop_cpu=17 perf stat -r 200 -- insmod ./page_frag_test.ko test_push_cpu=16 test_pop_cpu=17 test_align=1 perf stat -r 200 -- taskset -c 0 head -c 20G /dev/zero | taskset -c 1 nc 127.0.0.1 1234 *After* this patchset: Performance counter stats for 'insmod ./page_frag_test.ko test_push_cpu=16 test_pop_cpu=17' (200 runs): 17.829030 task-clock (msec) # 0.001 CPUs utilized ( +- 0.30% ) 7 context-switches # 0.386 K/sec ( +- 0.35% ) 0 cpu-migrations # 0.003 K/sec ( +- 28.06% ) 83 page-faults # 0.005 M/sec ( +- 0.10% ) 46303585 cycles # 2.597 GHz ( +- 0.30% ) 61119216 instructions # 1.32 insn per cycle ( +- 0.01% ) 14811318 branches # 830.742 M/sec ( +- 0.01% ) 21046 branch-misses # 0.14% of all branches ( +- 0.09% ) 23.856064365 seconds time elapsed ( +- 0.08% ) Performance counter stats for 'insmod ./page_frag_test.ko test_push_cpu=16 test_pop_cpu=17 test_align=1' (200 runs): 17.628569 task-clock (msec) # 0.001 CPUs utilized ( +- 0.01% ) 7 context-switches # 0.397 K/sec ( +- 0.12% ) 0 cpu-migrations # 0.000 K/sec 83 page-faults # 0.005 M/sec ( +- 0.10% ) 45785943 cycles # 2.597 GHz ( +- 0.01% ) 60043610 instructions # 1.31 insn per cycle ( +- 0.01% ) 14550182 branches # 825.375 M/sec ( +- 0.01% ) 21492 branch-misses # 0.15% of all branches ( +- 0.08% ) 23.443927103 seconds time elapsed ( +- 0.05% ) Performance counter stats for 'taskset -c 0 head -c 20G /dev/zero' (200 runs): 16626.042731 task-clock (msec) # 0.607 CPUs utilized ( +- 0.03% ) 3291020 context-switches # 0.198 M/sec ( +- 0.05% ) 1 cpu-migrations # 0.000 K/sec ( +- 0.50% ) 85 page-faults # 0.005 K/sec ( +- 0.16% ) 30581044838 cycles # 1.839 GHz ( +- 0.05% ) 34962744631 instructions # 1.14 insn per cycle ( +- 0.01% ) 6483883671 branches # 389.984 M/sec ( +- 0.02% ) 99624551 branch-misses # 1.54% of all branches ( +- 0.17% ) 27.370305077 seconds time elapsed ( +- 0.01% ) *Before* this patchset: Performance counter stats for 'insmod ./page_frag_test.ko test_push_cpu=16 test_pop_cpu=17' (200 runs): 18.143552 task-clock (msec) # 0.001 CPUs utilized ( +- 0.28% ) 7 context-switches # 0.382 K/sec ( +- 0.28% ) 1 cpu-migrations # 0.056 K/sec ( +- 0.97% ) 83 page-faults # 0.005 M/sec ( +- 0.10% ) 47105569 cycles # 2.596 GHz ( +- 0.28% ) 60628757 instructions # 1.29 insn per cycle ( +- 0.04% ) 14686743 branches # 809.475 M/sec ( +- 0.04% ) 21826 branch-misses # 0.15% of all branches ( +- 0.12% ) 23.918006193 seconds time elapsed ( +- 0.10% ) Performance counter stats for 'insmod ./page_frag_test.ko test_push_cpu=16 test_pop_cpu=17 test_align=1' (200 runs): 21.726393 task-clock (msec) # 0.001 CPUs utilized ( +- 0.72% ) 7 context-switches # 0.321 K/sec ( +- 0.24% ) 1 cpu-migrations # 0.047 K/sec ( +- 0.85% ) 83 page-faults # 0.004 M/sec ( +- 0.10% ) 56422898 cycles # 2.597 GHz ( +- 0.72% ) 61271860 instructions # 1.09 insn per cycle ( +- 0.05% ) 14837500 branches # 682.925 M/sec ( +- 0.05% ) 21484 branch-misses # 0.14% of all branches ( +- 0.10% ) 23.876306259 seconds time elapsed ( +- 0.13% ) Performance counter stats for 'taskset -c 0 head -c 20G /dev/zero' (200 runs): 17364.040855 task-clock (msec) # 0.624 CPUs utilized ( +- 0.02% ) 3340375 context-switches # 0.192 M/sec ( +- 0.06% ) 1 cpu-migrations # 0.000 K/sec 85 page-faults # 0.005 K/sec ( +- 0.15% ) 32077623335 cycles # 1.847 GHz ( +- 0.03% ) 35121047596 instructions # 1.09 insn per cycle ( +- 0.01% ) 6519872824 branches # 375.481 M/sec ( +- 0.02% ) 101877022 branch-misses # 1.56% of all branches ( +- 0.14% ) 27.842745343 seconds time elapsed ( +- 0.02% ) Note, ipv4-udp, ipv6-tcp and ipv6-udp is also tested with the below script: nc -u -l -k 1234 > /dev/null perf stat -r 4 -- head -c 51200000000 /dev/zero | nc -N -u 127.0.0.1 1234 nc -l6 -k 1234 > /dev/null perf stat -r 4 -- head -c 51200000000 /dev/zero | nc -N ::1 1234 nc -l6 -k -u 1234 > /dev/null perf stat -r 4 -- head -c 51200000000 /dev/zero | nc -u -N ::1 1234 CC: Alexander Duyck 1. https://lore.kernel.org/all/20240228093013.8263-1-linyunsheng@huawei.com/ Change log: V9: 1. Add check for test_alloc_len and change perm of module_param() to 0 as Wang Wei' comment. 2. Rebased on latest net-next. V8: Remove patch 2 & 3 in V7, as free_unref_page() is changed to call pcp_allowed_order() and used in page_frag API recently in: commit 5b8d75913a0e ("mm: combine free_the_page() and free_unref_page()") V7: Fix doc build warning and error. V6: 1. Fix some typo and compiler error for x86 pointed out by Jakub and Simon. 2. Add two refactoring and optimization patches. V5: 1. Add page_frag_alloc_pg() API for tls_device.c case and refactor some implementation, update kernel bin size changing as bin size is increased after that. 2. Add ack from Mat. RFC v4: 1. Update doc according to Randy and Mat's suggestion. 2. Change probe API to "probe" for a specific amount of available space, rather than "nonzero" space according to Mat's suggestion. 3. Retest and update the test result. v3: 1. Use new layout for 'struct page_frag_cache' as the discussion with Alexander and other sugeestions from Alexander. 2. Add probe API to address Mat' comment about mptcp use case. 3. Some doc updating according to Bagas' suggestion. v2: 1. reorder test module to patch 1. 2. split doc and maintainer updating to two patches. 3. refactor the page_frag before moving. 4. fix a type and 'static' warning in test module. 5. add a patch for xtensa arch to enable using get_order() in BUILD_BUG_ON(). 6. Add test case and performance data for the socket code. Yunsheng Lin (13): mm: page_frag: add a test module for page_frag mm: move the page fragment allocator from page_alloc into its own file mm: page_frag: use initial zero offset for page_frag_alloc_align() mm: page_frag: add '_va' suffix to page_frag API mm: page_frag: avoid caller accessing 'page_frag_cache' directly mm: page_frag: reuse existing space for 'size' and 'pfmemalloc' mm: page_frag: some minor refactoring before adding new API mm: page_frag: use __alloc_pages() to replace alloc_pages_node() net: introduce the skb_copy_to_va_nocache() helper mm: page_frag: introduce prepare/probe/commit API net: replace page_frag with page_frag_cache mm: page_frag: update documentation for page_frag mm: page_frag: add a entry in MAINTAINERS for page_frag Documentation/mm/page_frags.rst | 163 +++++++- MAINTAINERS | 11 + .../chelsio/inline_crypto/chtls/chtls.h | 3 - .../chelsio/inline_crypto/chtls/chtls_io.c | 100 ++--- .../chelsio/inline_crypto/chtls/chtls_main.c | 3 - drivers/net/ethernet/google/gve/gve_rx.c | 4 +- drivers/net/ethernet/intel/ice/ice_txrx.c | 2 +- drivers/net/ethernet/intel/ice/ice_txrx.h | 2 +- drivers/net/ethernet/intel/ice/ice_txrx_lib.c | 2 +- .../net/ethernet/intel/ixgbevf/ixgbevf_main.c | 4 +- .../marvell/octeontx2/nic/otx2_common.c | 2 +- drivers/net/ethernet/mediatek/mtk_wed_wo.c | 4 +- drivers/net/tun.c | 44 +- drivers/nvme/host/tcp.c | 8 +- drivers/nvme/target/tcp.c | 22 +- drivers/vhost/net.c | 8 +- include/linux/gfp.h | 22 - include/linux/mm_types.h | 18 - include/linux/page_frag_cache.h | 299 +++++++++++++ include/linux/sched.h | 3 +- include/linux/skbuff.h | 7 +- include/net/sock.h | 29 +- kernel/bpf/cpumap.c | 2 +- kernel/exit.c | 3 +- kernel/fork.c | 3 +- mm/Kconfig.debug | 8 + mm/Makefile | 2 + mm/page_alloc.c | 136 ------ mm/page_frag_cache.c | 346 +++++++++++++++ mm/page_frag_test.c | 393 ++++++++++++++++++ net/core/skbuff.c | 79 ++-- net/core/skmsg.c | 22 +- net/core/sock.c | 46 +- net/core/xdp.c | 2 +- net/ipv4/ip_output.c | 33 +- net/ipv4/tcp.c | 35 +- net/ipv4/tcp_output.c | 28 +- net/ipv6/ip6_output.c | 33 +- net/kcm/kcmsock.c | 30 +- net/mptcp/protocol.c | 67 +-- net/rxrpc/conn_object.c | 4 +- net/rxrpc/local_object.c | 4 +- net/rxrpc/txbuf.c | 15 +- net/sched/em_meta.c | 2 +- net/sunrpc/svcsock.c | 12 +- net/tls/tls_device.c | 137 +++--- 46 files changed, 1642 insertions(+), 560 deletions(-) create mode 100644 include/linux/page_frag_cache.h create mode 100644 mm/page_frag_cache.c create mode 100644 mm/page_frag_test.c -- 2.33.0 . From: Petr Machata To: "David S. Miller" , Eric Dumazet , Jakub Kicinski , Paolo Abeni , CC: Ido Schimmel , Petr Machata , "Amit Cohen" , Subject: [PATCH net-next 0/2] mlxsw: Reduce memory footprint of mlxsw driver Date: Tue, 25 Jun 2024 15:47:33 +0200 Message-ID: X-Mailing-List: netdev@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Content-Type: text/plain Xref: photonic.trudheim.com org.kernel.vger.netdev:354996 Newsgroups: org.kernel.vger.netdev Path: photonic.trudheim.com!nntp.lore.kernel.org!not-for-mail Amit Cohen writes: A previous patch-set used page pool to allocate buffers, to simplify the change, we first used one continuous buffer, which was allocated with order > 0. This set improves page pool usage to allocate the exact number of pages which are required for packet. This change requires using fragmented SKB, till now all the buffer was in the linear part. Note that 'skb->truesize' is decreased for small packets. This set significantly reduces memory consumption of mlxsw driver. The footprint is reduced by 26%. Patch set overview: Patch #1 calculates number of scatter/gather entries and stores the value Patch #2 converts the driver to use fragmented buffers Amit Cohen (2): mlxsw: pci: Store number of scatter/gather entries for maximum packet size mlxsw: pci: Use fragmented buffers drivers/net/ethernet/mellanox/mlxsw/pci.c | 183 ++++++++++++++++++---- 1 file changed, 149 insertions(+), 34 deletions(-) -- 2.45.0 . From: Bartosz Golaszewski To: Bjorn Andersson , Konrad Dybcio , Rob Herring , Krzysztof Kozlowski , Conor Dooley Cc: linux-arm-msm@vger.kernel.org, devicetree@vger.kernel.org, linux-kernel@vger.kernel.org, netdev@vger.kernel.org, Bartosz Golaszewski Subject: [PATCH v2 0/3] arm64: dts: qcom: sa8775p-ride: support both board variants Date: Tue, 25 Jun 2024 17:14:27 +0200 Message-ID: <20240625151430.34024-1-brgl@bgdev.pl> X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Xref: photonic.trudheim.com org.kernel.vger.linux-kernel:1258371 org.kernel.vger.netdev:355003 Newsgroups: org.kernel.vger.linux-kernel,org.kernel.vger.linux-arm-msm,org.kernel.vger.linux-devicetree,org.kernel.vger.netdev Path: photonic.trudheim.com!nntp.lore.kernel.org!not-for-mail From: Bartosz Golaszewski Split the current .dts into two: the existing one keeps the name and supports revision 2 of the board while patch 2 adds a .dts for revision 3. Changes since v1: - add a new compatible for Rev3 Bartosz Golaszewski (3): dt-bindings: arm: qcom: add sa8775p-ride Rev 3 arm64: dts: qcom: move common parts for sa8775p-ride variants into a .dtsi arm64: dts: qcom: sa8775p-ride-r3: add new board file .../devicetree/bindings/arm/qcom.yaml | 1 + arch/arm64/boot/dts/qcom/Makefile | 1 + arch/arm64/boot/dts/qcom/sa8775p-ride-r3.dts | 47 + arch/arm64/boot/dts/qcom/sa8775p-ride.dts | 836 +----------------- arch/arm64/boot/dts/qcom/sa8775p-ride.dtsi | 814 +++++++++++++++++ 5 files changed, 885 insertions(+), 814 deletions(-) create mode 100644 arch/arm64/boot/dts/qcom/sa8775p-ride-r3.dts create mode 100644 arch/arm64/boot/dts/qcom/sa8775p-ride.dtsi -- 2.43.0 . From: Shannon Nelson To: , , , , CC: , Shannon Nelson Subject: [RFC PATCH net-next] ionic: convert Rx queue buffers to use page_pool Date: Tue, 25 Jun 2024 09:56:58 -0700 Message-ID: <20240625165658.34598-1-shannon.nelson@amd.com> X-Mailing-List: netdev@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Type: text/plain Xref: photonic.trudheim.com org.kernel.vger.netdev:355012 Newsgroups: org.kernel.vger.netdev Path: photonic.trudheim.com!nntp.lore.kernel.org!not-for-mail Our home-grown buffer management needs to go away and we need to be playing nicely with the page_pool infrastructure. This converts the Rx traffic queues to use page_pool. RFC: This is still being tweaked and tested, but has passed some basic testing for both normal and jumbo frames going through normal paths as well as XDP PASS, DROP, ABORTED, TX, and REDIRECT paths. It has not been under any performance test runs, but a quicky iperf3 test shows similar numbers. This patch seems far enough along that a look-over by some page_pool experienced reviewers would be appreciated. Thanks! sln Signed-off-by: Shannon Nelson --- drivers/net/ethernet/pensando/Kconfig | 1 + .../net/ethernet/pensando/ionic/ionic_dev.h | 2 +- .../net/ethernet/pensando/ionic/ionic_lif.c | 43 ++- .../net/ethernet/pensando/ionic/ionic_txrx.c | 318 ++++++++---------- 4 files changed, 189 insertions(+), 175 deletions(-) diff --git a/drivers/net/ethernet/pensando/Kconfig b/drivers/net/ethernet/pensando/Kconfig index 3f7519e435b8..01fe76786f77 100644 --- a/drivers/net/ethernet/pensando/Kconfig +++ b/drivers/net/ethernet/pensando/Kconfig @@ -23,6 +23,7 @@ config IONIC depends on PTP_1588_CLOCK_OPTIONAL select NET_DEVLINK select DIMLIB + select PAGE_POOL help This enables the support for the Pensando family of Ethernet adapters. More specific information on this driver can be diff --git a/drivers/net/ethernet/pensando/ionic/ionic_dev.h b/drivers/net/ethernet/pensando/ionic/ionic_dev.h index 92f16b6c5662..45ad2bf1e1e7 100644 --- a/drivers/net/ethernet/pensando/ionic/ionic_dev.h +++ b/drivers/net/ethernet/pensando/ionic/ionic_dev.h @@ -177,7 +177,6 @@ struct ionic_dev { struct ionic_devinfo dev_info; }; -struct ionic_queue; struct ionic_qcq; #define IONIC_MAX_BUF_LEN ((u16)-1) @@ -262,6 +261,7 @@ struct ionic_queue { }; struct xdp_rxq_info *xdp_rxq_info; struct ionic_queue *partner; + struct page_pool *page_pool; dma_addr_t base_pa; dma_addr_t cmb_base_pa; dma_addr_t sg_base_pa; diff --git a/drivers/net/ethernet/pensando/ionic/ionic_lif.c b/drivers/net/ethernet/pensando/ionic/ionic_lif.c index 38ce35462737..e1cd5982bb2e 100644 --- a/drivers/net/ethernet/pensando/ionic/ionic_lif.c +++ b/drivers/net/ethernet/pensando/ionic/ionic_lif.c @@ -13,6 +13,7 @@ #include #include #include +#include #include "ionic.h" #include "ionic_bus.h" @@ -440,6 +441,8 @@ static void ionic_qcq_free(struct ionic_lif *lif, struct ionic_qcq *qcq) ionic_xdp_unregister_rxq_info(&qcq->q); ionic_qcq_intr_free(lif, qcq); + page_pool_destroy(qcq->q.page_pool); + qcq->q.page_pool = NULL; vfree(qcq->q.info); qcq->q.info = NULL; } @@ -579,6 +582,36 @@ static int ionic_qcq_alloc(struct ionic_lif *lif, unsigned int type, goto err_out_free_qcq; } + if (type == IONIC_QTYPE_RXQ) { + struct page_pool_params pp_params = { + .flags = PP_FLAG_DMA_MAP | PP_FLAG_DMA_SYNC_DEV, + .order = 0, + .pool_size = num_descs, + .nid = NUMA_NO_NODE, + .dev = lif->ionic->dev, + .napi = &new->napi, + .dma_dir = DMA_FROM_DEVICE, + .max_len = PAGE_SIZE, + .offset = 0, + .netdev = lif->netdev, + }; + struct bpf_prog *xdp_prog; + + xdp_prog = READ_ONCE(lif->xdp_prog); + if (xdp_prog) { + pp_params.dma_dir = DMA_BIDIRECTIONAL; + pp_params.offset = XDP_PACKET_HEADROOM; + } + + new->q.page_pool = page_pool_create(&pp_params); + if (IS_ERR(new->q.page_pool)) { + netdev_err(lif->netdev, "Cannot create page_pool\n"); + err = PTR_ERR(new->q.page_pool); + new->q.page_pool = NULL; + goto err_out_free_q_info; + } + } + new->q.type = type; new->q.max_sg_elems = lif->qtype_info[type].max_sg_elems; @@ -586,12 +619,12 @@ static int ionic_qcq_alloc(struct ionic_lif *lif, unsigned int type, desc_size, sg_desc_size, pid); if (err) { netdev_err(lif->netdev, "Cannot initialize queue\n"); - goto err_out_free_q_info; + goto err_out_free_page_pool; } err = ionic_alloc_qcq_interrupt(lif, new); if (err) - goto err_out_free_q_info; + goto err_out_free_page_pool; err = ionic_cq_init(lif, &new->cq, &new->intr, num_descs, cq_desc_size); if (err) { @@ -712,6 +745,8 @@ static int ionic_qcq_alloc(struct ionic_lif *lif, unsigned int type, devm_free_irq(dev, new->intr.vector, &new->napi); ionic_intr_free(lif->ionic, new->intr.index); } +err_out_free_page_pool: + page_pool_destroy(new->q.page_pool); err_out_free_q_info: vfree(new->q.info); err_out_free_qcq: @@ -2681,7 +2716,8 @@ static int ionic_xdp_register_rxq_info(struct ionic_queue *q, unsigned int napi_ goto err_out; } - err = xdp_rxq_info_reg_mem_model(rxq_info, MEM_TYPE_PAGE_ORDER0, NULL); + err = xdp_rxq_info_reg_mem_model(rxq_info, MEM_TYPE_PAGE_POOL, + q->page_pool); if (err) { dev_err(q->dev, "Queue %d xdp_rxq_info_reg_mem_model failed, err %d\n", q->index, err); @@ -2878,6 +2914,7 @@ static void ionic_swap_queues(struct ionic_qcq *a, struct ionic_qcq *b) swap(a->q.base, b->q.base); swap(a->q.base_pa, b->q.base_pa); swap(a->q.info, b->q.info); + swap(a->q.page_pool, b->q.page_pool); swap(a->q.xdp_rxq_info, b->q.xdp_rxq_info); swap(a->q.partner, b->q.partner); swap(a->q_base, b->q_base); diff --git a/drivers/net/ethernet/pensando/ionic/ionic_txrx.c b/drivers/net/ethernet/pensando/ionic/ionic_txrx.c index 5bf13a5d411c..ffef3d66e0df 100644 --- a/drivers/net/ethernet/pensando/ionic/ionic_txrx.c +++ b/drivers/net/ethernet/pensando/ionic/ionic_txrx.c @@ -6,6 +6,7 @@ #include #include #include +#include #include "ionic.h" #include "ionic_lif.h" @@ -117,86 +118,19 @@ static void *ionic_rx_buf_va(struct ionic_buf_info *buf_info) static dma_addr_t ionic_rx_buf_pa(struct ionic_buf_info *buf_info) { - return buf_info->dma_addr + buf_info->page_offset; + return page_pool_get_dma_addr(buf_info->page) + buf_info->page_offset; } -static unsigned int ionic_rx_buf_size(struct ionic_buf_info *buf_info) +static void ionic_rx_put_buf(struct ionic_queue *q, + struct ionic_buf_info *buf_info) { - return min_t(u32, IONIC_MAX_BUF_LEN, IONIC_PAGE_SIZE - buf_info->page_offset); -} - -static int ionic_rx_page_alloc(struct ionic_queue *q, - struct ionic_buf_info *buf_info) -{ - struct device *dev = q->dev; - dma_addr_t dma_addr; - struct page *page; - - page = alloc_pages(IONIC_PAGE_GFP_MASK, 0); - if (unlikely(!page)) { - net_err_ratelimited("%s: %s page alloc failed\n", - dev_name(dev), q->name); - q_to_rx_stats(q)->alloc_err++; - return -ENOMEM; - } - - dma_addr = dma_map_page(dev, page, 0, - IONIC_PAGE_SIZE, DMA_FROM_DEVICE); - if (unlikely(dma_mapping_error(dev, dma_addr))) { - __free_pages(page, 0); - net_err_ratelimited("%s: %s dma map failed\n", - dev_name(dev), q->name); - q_to_rx_stats(q)->dma_map_err++; - return -EIO; - } - - buf_info->dma_addr = dma_addr; - buf_info->page = page; - buf_info->page_offset = 0; - - return 0; -} - -static void ionic_rx_page_free(struct ionic_queue *q, - struct ionic_buf_info *buf_info) -{ - struct device *dev = q->dev; - - if (unlikely(!buf_info)) { - net_err_ratelimited("%s: %s invalid buf_info in free\n", - dev_name(dev), q->name); - return; - } - if (!buf_info->page) return; - dma_unmap_page(dev, buf_info->dma_addr, IONIC_PAGE_SIZE, DMA_FROM_DEVICE); - __free_pages(buf_info->page, 0); + page_pool_put_full_page(q->page_pool, buf_info->page, false); buf_info->page = NULL; -} - -static bool ionic_rx_buf_recycle(struct ionic_queue *q, - struct ionic_buf_info *buf_info, u32 len) -{ - u32 size; - - /* don't re-use pages allocated in low-mem condition */ - if (page_is_pfmemalloc(buf_info->page)) - return false; - - /* don't re-use buffers from non-local numa nodes */ - if (page_to_nid(buf_info->page) != numa_mem_id()) - return false; - - size = ALIGN(len, q->xdp_rxq_info ? IONIC_PAGE_SIZE : IONIC_PAGE_SPLIT_SZ); - buf_info->page_offset += size; - if (buf_info->page_offset >= IONIC_PAGE_SIZE) - return false; - - get_page(buf_info->page); - - return true; + buf_info->len = 0; + buf_info->page_offset = 0; } static void ionic_rx_add_skb_frag(struct ionic_queue *q, @@ -207,18 +141,19 @@ static void ionic_rx_add_skb_frag(struct ionic_queue *q, { if (!synced) dma_sync_single_range_for_cpu(q->dev, ionic_rx_buf_pa(buf_info), - off, len, DMA_FROM_DEVICE); + off, len, + page_pool_get_dma_dir(q->page_pool)); skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags, buf_info->page, buf_info->page_offset + off, - len, - IONIC_PAGE_SIZE); + len, buf_info->len); - if (!ionic_rx_buf_recycle(q, buf_info, len)) { - dma_unmap_page(q->dev, buf_info->dma_addr, - IONIC_PAGE_SIZE, DMA_FROM_DEVICE); - buf_info->page = NULL; - } + /* napi_gro_frags() will release/recycle the + * page_pool buffers from the frags list + */ + buf_info->page = NULL; + buf_info->len = 0; + buf_info->page_offset = 0; } static struct sk_buff *ionic_rx_build_skb(struct ionic_queue *q, @@ -243,12 +178,13 @@ static struct sk_buff *ionic_rx_build_skb(struct ionic_queue *q, q_to_rx_stats(q)->alloc_err++; return NULL; } + skb_mark_for_recycle(skb); if (headroom) frag_len = min_t(u16, len, IONIC_XDP_MAX_LINEAR_MTU + VLAN_ETH_HLEN); else - frag_len = min_t(u16, len, ionic_rx_buf_size(buf_info)); + frag_len = min_t(u16, len, IONIC_PAGE_SIZE); if (unlikely(!buf_info->page)) goto err_bad_buf_page; @@ -259,7 +195,7 @@ static struct sk_buff *ionic_rx_build_skb(struct ionic_queue *q, for (i = 0; i < num_sg_elems; i++, buf_info++) { if (unlikely(!buf_info->page)) goto err_bad_buf_page; - frag_len = min_t(u16, len, ionic_rx_buf_size(buf_info)); + frag_len = min_t(u16, len, buf_info->len); ionic_rx_add_skb_frag(q, skb, buf_info, 0, frag_len, synced); len -= frag_len; } @@ -276,11 +212,14 @@ static struct sk_buff *ionic_rx_copybreak(struct net_device *netdev, struct ionic_rx_desc_info *desc_info, unsigned int headroom, unsigned int len, + unsigned int num_sg_elems, bool synced) { struct ionic_buf_info *buf_info; + enum dma_data_direction dma_dir; struct device *dev = q->dev; struct sk_buff *skb; + int i; buf_info = &desc_info->bufs[0]; @@ -291,54 +230,58 @@ static struct sk_buff *ionic_rx_copybreak(struct net_device *netdev, q_to_rx_stats(q)->alloc_err++; return NULL; } + skb_mark_for_recycle(skb); - if (unlikely(!buf_info->page)) { - dev_kfree_skb(skb); - return NULL; - } - + dma_dir = page_pool_get_dma_dir(q->page_pool); if (!synced) dma_sync_single_range_for_cpu(dev, ionic_rx_buf_pa(buf_info), - headroom, len, DMA_FROM_DEVICE); + headroom, len, dma_dir); skb_copy_to_linear_data(skb, ionic_rx_buf_va(buf_info) + headroom, len); - dma_sync_single_range_for_device(dev, ionic_rx_buf_pa(buf_info), - headroom, len, DMA_FROM_DEVICE); skb_put(skb, len); skb->protocol = eth_type_trans(skb, netdev); + /* recycle the Rx buffer now that we're done with it */ + ionic_rx_put_buf(q, buf_info); + buf_info++; + for (i = 0; i < num_sg_elems; i++, buf_info++) + ionic_rx_put_buf(q, buf_info); + return skb; } +static void ionic_xdp_rx_put_bufs(struct ionic_queue *q, + struct ionic_buf_info *buf_info, + int nbufs) +{ + int i; + + for (i = 0; i < nbufs; i++) { + buf_info->page = NULL; + buf_info++; + } +} + static void ionic_xdp_tx_desc_clean(struct ionic_queue *q, struct ionic_tx_desc_info *desc_info) { - unsigned int nbufs = desc_info->nbufs; - struct ionic_buf_info *buf_info; - struct device *dev = q->dev; - int i; + struct xdp_frame_bulk bq; - if (!nbufs) + if (!desc_info->nbufs) return; - buf_info = desc_info->bufs; - dma_unmap_single(dev, buf_info->dma_addr, - buf_info->len, DMA_TO_DEVICE); - if (desc_info->act == XDP_TX) - __free_pages(buf_info->page, 0); - buf_info->page = NULL; + xdp_frame_bulk_init(&bq); + rcu_read_lock(); /* need for xdp_return_frame_bulk */ - buf_info++; - for (i = 1; i < nbufs + 1 && buf_info->page; i++, buf_info++) { - dma_unmap_page(dev, buf_info->dma_addr, - buf_info->len, DMA_TO_DEVICE); - if (desc_info->act == XDP_TX) - __free_pages(buf_info->page, 0); - buf_info->page = NULL; + if (desc_info->act == XDP_TX) { + xdp_return_frame_rx_napi(desc_info->xdpf); + } else if (desc_info->act == XDP_REDIRECT) { + ionic_tx_desc_unmap_bufs(q, desc_info); + xdp_return_frame_bulk(desc_info->xdpf, &bq); } - if (desc_info->act == XDP_REDIRECT) - xdp_return_frame(desc_info->xdpf); + xdp_flush_frame_bulk(&bq); + rcu_read_unlock(); desc_info->nbufs = 0; desc_info->xdpf = NULL; @@ -362,9 +305,15 @@ static int ionic_xdp_post_frame(struct ionic_queue *q, struct xdp_frame *frame, buf_info = desc_info->bufs; stats = q_to_tx_stats(q); - dma_addr = ionic_tx_map_single(q, frame->data, len); - if (!dma_addr) - return -EIO; + if (act == XDP_TX) { + dma_addr = page_pool_get_dma_addr(page) + off; + dma_sync_single_for_device(q->dev, dma_addr, len, DMA_TO_DEVICE); + } else /* XDP_REDIRECT */ { + dma_addr = ionic_tx_map_single(q, frame->data, len); + if (dma_mapping_error(q->dev, dma_addr)) + return -EIO; + } + buf_info->dma_addr = dma_addr; buf_info->len = len; buf_info->page = page; @@ -386,10 +335,19 @@ static int ionic_xdp_post_frame(struct ionic_queue *q, struct xdp_frame *frame, frag = sinfo->frags; elem = ionic_tx_sg_elems(q); for (i = 0; i < sinfo->nr_frags; i++, frag++, bi++) { - dma_addr = ionic_tx_map_frag(q, frag, 0, skb_frag_size(frag)); - if (!dma_addr) { - ionic_tx_desc_unmap_bufs(q, desc_info); - return -EIO; + if (act == XDP_TX) { + dma_addr = page_pool_get_dma_addr(skb_frag_page(frag)); + dma_addr += skb_frag_off(frag); + dma_sync_single_for_device(q->dev, dma_addr, + skb_frag_size(frag), + DMA_TO_DEVICE); + } else { + dma_addr = ionic_tx_map_frag(q, frag, 0, + skb_frag_size(frag)); + if (dma_mapping_error(q->dev, dma_addr)) { + ionic_tx_desc_unmap_bufs(q, desc_info); + return -EIO; + } } bi->dma_addr = dma_addr; bi->len = skb_frag_size(frag); @@ -493,6 +451,7 @@ static bool ionic_run_xdp(struct ionic_rx_stats *stats, struct netdev_queue *nq; struct xdp_frame *xdpf; int remain_len; + int nbufs = 1; int frag_len; int err = 0; @@ -526,13 +485,13 @@ static bool ionic_run_xdp(struct ionic_rx_stats *stats, do { if (unlikely(sinfo->nr_frags >= MAX_SKB_FRAGS)) { err = -ENOSPC; - goto out_xdp_abort; + break; } frag = &sinfo->frags[sinfo->nr_frags]; sinfo->nr_frags++; bi++; - frag_len = min_t(u16, remain_len, ionic_rx_buf_size(bi)); + frag_len = min_t(u16, remain_len, bi->len); dma_sync_single_range_for_cpu(rxq->dev, ionic_rx_buf_pa(bi), 0, frag_len, DMA_FROM_DEVICE); skb_frag_fill_page_desc(frag, bi->page, 0, frag_len); @@ -542,6 +501,7 @@ static bool ionic_run_xdp(struct ionic_rx_stats *stats, if (page_is_pfmemalloc(bi->page)) xdp_buff_set_frag_pfmemalloc(&xdp_buf); } while (remain_len > 0); + nbufs += sinfo->nr_frags; } xdp_action = bpf_prog_run_xdp(xdp_prog, &xdp_buf); @@ -552,14 +512,15 @@ static bool ionic_run_xdp(struct ionic_rx_stats *stats, return false; /* false = we didn't consume the packet */ case XDP_DROP: - ionic_rx_page_free(rxq, buf_info); stats->xdp_drop++; break; case XDP_TX: xdpf = xdp_convert_buff_to_frame(&xdp_buf); - if (!xdpf) - goto out_xdp_abort; + if (!xdpf) { + err = -ENOSPC; + break; + } txq = rxq->partner; nq = netdev_get_tx_queue(netdev, txq->index); @@ -571,12 +532,10 @@ static bool ionic_run_xdp(struct ionic_rx_stats *stats, ionic_q_space_avail(txq), 1, 1)) { __netif_tx_unlock(nq); - goto out_xdp_abort; + err = -EIO; + break; } - dma_unmap_page(rxq->dev, buf_info->dma_addr, - IONIC_PAGE_SIZE, DMA_FROM_DEVICE); - err = ionic_xdp_post_frame(txq, xdpf, XDP_TX, buf_info->page, buf_info->page_offset, @@ -584,40 +543,35 @@ static bool ionic_run_xdp(struct ionic_rx_stats *stats, __netif_tx_unlock(nq); if (unlikely(err)) { netdev_dbg(netdev, "tx ionic_xdp_post_frame err %d\n", err); - goto out_xdp_abort; + break; } - buf_info->page = NULL; + ionic_xdp_rx_put_bufs(rxq, buf_info, nbufs); stats->xdp_tx++; - /* the Tx completion will free the buffers */ break; case XDP_REDIRECT: - /* unmap the pages before handing them to a different device */ - dma_unmap_page(rxq->dev, buf_info->dma_addr, - IONIC_PAGE_SIZE, DMA_FROM_DEVICE); - err = xdp_do_redirect(netdev, &xdp_buf, xdp_prog); if (unlikely(err)) { netdev_dbg(netdev, "xdp_do_redirect err %d\n", err); - goto out_xdp_abort; + break; } - buf_info->page = NULL; + rxq->xdp_flush = true; + ionic_xdp_rx_put_bufs(rxq, buf_info, nbufs); stats->xdp_redirect++; break; case XDP_ABORTED: default: - goto out_xdp_abort; + err = -EIO; + break; } - return true; - -out_xdp_abort: - trace_xdp_exception(netdev, xdp_prog, xdp_action); - ionic_rx_page_free(rxq, buf_info); - stats->xdp_aborted++; + if (err) { + trace_xdp_exception(netdev, xdp_prog, xdp_action); + stats->xdp_aborted++; + } return true; } @@ -639,6 +593,15 @@ static void ionic_rx_clean(struct ionic_queue *q, stats = q_to_rx_stats(q); if (comp->status) { + struct ionic_rxq_desc *desc = &q->rxq[q->head_idx]; + + /* Most likely status==2 and the pkt received was bigger + * than the buffer available: comp->len will show the + * pkt size received that didn't fit the advertised desc.len + */ + dev_dbg(q->dev, "q%d drop comp->status %d comp->len %d desc.len %d\n", + q->index, comp->status, comp->len, desc->len); + stats->dropped++; return; } @@ -658,7 +621,8 @@ static void ionic_rx_clean(struct ionic_queue *q, use_copybreak = len <= q->lif->rx_copybreak; if (use_copybreak) skb = ionic_rx_copybreak(netdev, q, desc_info, - headroom, len, synced); + headroom, len, + comp->num_sg_elems, synced); else skb = ionic_rx_build_skb(q, desc_info, headroom, len, comp->num_sg_elems, synced); @@ -798,32 +762,38 @@ void ionic_rx_fill(struct ionic_queue *q) for (i = n_fill; i; i--) { unsigned int headroom; - unsigned int buf_len; nfrags = 0; remain_len = len; desc = &q->rxq[q->head_idx]; desc_info = &q->rx_info[q->head_idx]; buf_info = &desc_info->bufs[0]; - - if (!buf_info->page) { /* alloc a new buffer? */ - if (unlikely(ionic_rx_page_alloc(q, buf_info))) { - desc->addr = 0; - desc->len = 0; - return; - } - } + ionic_rx_put_buf(q, buf_info); /* fill main descriptor - buf[0] * XDP uses space in the first buffer, so account for * head room, tail room, and ip header in the first frag size. */ headroom = q->xdp_rxq_info ? XDP_PACKET_HEADROOM : 0; - if (q->xdp_rxq_info) - buf_len = IONIC_XDP_MAX_LINEAR_MTU + VLAN_ETH_HLEN; - else - buf_len = ionic_rx_buf_size(buf_info); - frag_len = min_t(u16, len, buf_len); + if (q->xdp_rxq_info) { + /* Always alloc the full size buffer, but only need + * the actual frag_len in the descriptor + */ + buf_info->len = IONIC_XDP_MAX_LINEAR_MTU + VLAN_ETH_HLEN; + frag_len = min_t(u16, len, buf_info->len); + } else { + /* Start with max buffer size, then use + * the frag size for the actual size to alloc + */ + frag_len = min_t(u16, len, IONIC_PAGE_SIZE); + buf_info->len = frag_len; + } + + buf_info->page = page_pool_alloc(q->page_pool, + &buf_info->page_offset, + &buf_info->len, GFP_ATOMIC); + if (unlikely(!buf_info->page)) + return; desc->addr = cpu_to_le64(ionic_rx_buf_pa(buf_info) + headroom); desc->len = cpu_to_le16(frag_len); @@ -833,20 +803,31 @@ void ionic_rx_fill(struct ionic_queue *q) /* fill sg descriptors - buf[1..n] */ sg_elem = q->rxq_sgl[q->head_idx].elems; - for (j = 0; remain_len > 0 && j < q->max_sg_elems; j++, sg_elem++) { - if (!buf_info->page) { /* alloc a new sg buffer? */ - if (unlikely(ionic_rx_page_alloc(q, buf_info))) { - sg_elem->addr = 0; - sg_elem->len = 0; + for (j = 0; remain_len > 0 && j < q->max_sg_elems; j++) { + frag_len = min_t(u16, remain_len, IONIC_PAGE_SIZE); + + /* Recycle any "wrong" sized buffers */ + if (unlikely(buf_info->page && buf_info->len != frag_len)) + ionic_rx_put_buf(q, buf_info); + + /* Get new buffer if needed */ + if (!buf_info->page) { + buf_info->len = frag_len; + buf_info->page = page_pool_alloc(q->page_pool, + &buf_info->page_offset, + &buf_info->len, GFP_ATOMIC); + if (unlikely(!buf_info->page)) { + buf_info->len = 0; return; } } sg_elem->addr = cpu_to_le64(ionic_rx_buf_pa(buf_info)); - frag_len = min_t(u16, remain_len, ionic_rx_buf_size(buf_info)); sg_elem->len = cpu_to_le16(frag_len); + remain_len -= frag_len; buf_info++; + sg_elem++; nfrags++; } @@ -873,17 +854,12 @@ void ionic_rx_fill(struct ionic_queue *q) void ionic_rx_empty(struct ionic_queue *q) { struct ionic_rx_desc_info *desc_info; - struct ionic_buf_info *buf_info; unsigned int i, j; for (i = 0; i < q->num_descs; i++) { desc_info = &q->rx_info[i]; - for (j = 0; j < ARRAY_SIZE(desc_info->bufs); j++) { - buf_info = &desc_info->bufs[j]; - if (buf_info->page) - ionic_rx_page_free(q, buf_info); - } - + for (j = 0; j < ARRAY_SIZE(desc_info->bufs); j++) + ionic_rx_put_buf(q, &desc_info->bufs[j]); desc_info->nbufs = 0; } -- 2.17.1 . From: Marek Vasut To: netdev@vger.kernel.org Cc: Marek Vasut , Andrew Lunn , Conor Dooley , "David S . Miller" , Eric Dumazet , Florian Fainelli , Heiner Kallweit , Jakub Kicinski , Joakim Zhang , Krzysztof Kozlowski , Paolo Abeni , Rob Herring , devicetree@vger.kernel.org, kernel@dh-electronics.com Subject: [net-next,PATCH v2] dt-bindings: net: realtek,rtl82xx: Document known PHY IDs as compatible strings Date: Tue, 25 Jun 2024 20:42:28 +0200 Message-ID: <20240625184359.153423-1-marex@denx.de> X-Mailing-List: netdev@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Xref: photonic.trudheim.com org.kernel.vger.netdev:355019 Newsgroups: org.kernel.vger.netdev,org.kernel.vger.linux-devicetree Path: photonic.trudheim.com!nntp.lore.kernel.org!not-for-mail Extract known PHY IDs from Linux kernel realtek PHY driver and convert them into supported compatible string list for this DT binding document. Signed-off-by: Marek Vasut --- Cc: Andrew Lunn Cc: Conor Dooley Cc: David S. Miller Cc: Eric Dumazet Cc: Florian Fainelli Cc: Heiner Kallweit Cc: Jakub Kicinski Cc: Joakim Zhang Cc: Krzysztof Kozlowski Cc: Paolo Abeni Cc: Rob Herring Cc: devicetree@vger.kernel.org Cc: kernel@dh-electronics.com Cc: netdev@vger.kernel.org --- V2: Drop ethernet-phy-id0000.8201 entry --- .../bindings/net/realtek,rtl82xx.yaml | 23 +++++++++++++++++++ 1 file changed, 23 insertions(+) diff --git a/Documentation/devicetree/bindings/net/realtek,rtl82xx.yaml b/Documentation/devicetree/bindings/net/realtek,rtl82xx.yaml index bb94a2388520b..18ee72f5c74a8 100644 --- a/Documentation/devicetree/bindings/net/realtek,rtl82xx.yaml +++ b/Documentation/devicetree/bindings/net/realtek,rtl82xx.yaml @@ -18,6 +18,29 @@ allOf: - $ref: ethernet-phy.yaml# properties: + compatible: + enum: + - ethernet-phy-id001c.c800 + - ethernet-phy-id001c.c816 + - ethernet-phy-id001c.c838 + - ethernet-phy-id001c.c840 + - ethernet-phy-id001c.c848 + - ethernet-phy-id001c.c849 + - ethernet-phy-id001c.c84a + - ethernet-phy-id001c.c862 + - ethernet-phy-id001c.c878 + - ethernet-phy-id001c.c880 + - ethernet-phy-id001c.c910 + - ethernet-phy-id001c.c912 + - ethernet-phy-id001c.c913 + - ethernet-phy-id001c.c914 + - ethernet-phy-id001c.c915 + - ethernet-phy-id001c.c916 + - ethernet-phy-id001c.c942 + - ethernet-phy-id001c.c961 + - ethernet-phy-id001c.cad0 + - ethernet-phy-id001c.cb00 + realtek,clkout-disable: type: boolean description: -- 2.43.0 . From: David Wei To: Michael Chan , Andy Gospodarek , Jesper Dangaard Brouer , Ilias Apalodimas , netdev@vger.kernel.org Cc: "David S. Miller" , Eric Dumazet , Jakub Kicinski , Paolo Abeni Subject: [PATCH net-next v1 0/2] page_pool: bnxt_en: unlink old page pool in queue api using helper Date: Tue, 25 Jun 2024 12:55:20 -0700 Message-ID: <20240625195522.2974466-1-dw@davidwei.uk> X-Mailing-List: netdev@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: 8bit Xref: photonic.trudheim.com org.kernel.vger.netdev:355024 Newsgroups: org.kernel.vger.netdev Path: photonic.trudheim.com!nntp.lore.kernel.org!not-for-mail 56ef27e3 unexported page_pool_unlink_napi() and renamed it to page_pool_disable_direct_recycling(). This is because there was no in-tree user of page_pool_unlink_napi(). Since then Rx queue API and an implementation in bnxt got merged. In the bnxt implementation, it broadly follows the following steps: allocate new queue memory + page pool, stop old rx queue, swap, then destroy old queue memory + page pool. The existing NAPI instance is re-used. The page pool to be destroyed is still linked to the re-used NAPI instance. Freeing it as-is will trigger warnings in page_pool_disable_direct_recycling(). In my initial patches I unlinked very directly by setting pp.napi to NULL. Instead, bring back page_pool_unlink_napi() and use that instead of having a driver touch a core struct directly. David Wei (2): page_pool: reintroduce page_pool_unlink_napi() bnxt_en: unlink page pool when stopping Rx queue drivers/net/ethernet/broadcom/bnxt/bnxt.c | 6 +----- include/net/page_pool/types.h | 5 +++++ net/core/page_pool.c | 6 ++++++ 3 files changed, 12 insertions(+), 5 deletions(-) -- 2.43.0 .