- 06 Jan, 2022 15 commits
-
-
Vladimir Oltean authored
As discussed during review here: https://patchwork.kernel.org/project/netdevbpf/patch/20220105132141.2648876-3-vladimir.oltean@nxp.com/ we should inform developers about pitfalls of concurrent access to the boolean properties of dsa_switch and dsa_port, now that they've been converted to bit fields. No other measure than a comment needs to be taken, since the code paths that update these bit fields are not concurrent with each other. Suggested-by:
Florian Fainelli <f.fainelli@gmail.com> Signed-off-by:
Vladimir Oltean <vladimir.oltean@nxp.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Vladimir Oltean authored
This is a cosmetic incremental fixup to commits 7787ff77 ("net: dsa: merge all bools of struct dsa_switch into a single u32") bde82f38 ("net: dsa: merge all bools of struct dsa_port into a single u8") The desire to make this change was enunciated after posting these patches here: https://patchwork.kernel.org/project/netdevbpf/cover/20220105132141.2648876-1-vladimir.oltean@nxp.com/ but due to a slight timing overlap (message posted at 2:28 p.m. UTC, merge commit is at 2:46 p.m. UTC), that comment was missed and the changes were applied as-is. Signed-off-by:
-
David S. Miller authored
Vladimir Oltean says: ==================== DSA initialization cleanups These patches contain miscellaneous work that makes the DSA init code path symmetric with the teardown path, and some additional patches carried by Ansuel Smith for his register access over Ethernet work, but those patches can be applied as-is too. https://patchwork.kernel.org/project/netdevbpf/patch/20211214224409.5770-3-ansuelsmth@gmail.com/ ==================== Signed-off-by:
-
Vladimir Oltean authored
It is said that as soon as a network interface is registered, all its resources should have already been prepared, so that it is available for sending and receiving traffic. One of the resources needed by a DSA slave interface is the master. dsa_tree_setup -> dsa_tree_setup_ports -> dsa_port_setup -> dsa_slave_create -> register_netdevice -> dsa_tree_setup_master -> dsa_master_setup -> sets up master->dsa_ptr, which enables reception Therefore, there is a short period of time after register_netdevice() during which the master isn't prepared to pass traffic to the DSA layer (master->dsa_ptr is checked by eth_type_trans). Same thing during unregistration, there is a time frame in which packets might be missed. Note that this change opens us to another race: dsa_master_find_slave() will get invoked potentially earlier than the slave creation, and later than the slave deletion. Since dp->slave starts off as a NULL pointer, the earlier calls aren't a problem, but the later calls are. To avoid use-after-free, we should zeroize dp->slave before calling dsa_slave_destroy(). In practice I cannot really test real life improvements brought by this change, since in my systems, netdevice creation races with PHY autoneg which takes a few seconds to complete, and that masks quite a few races. Effects might be noticeable in a setup with fixed links all the way to an external system. Signed-off-by: -
Vladimir Oltean authored
After commit a57d8c21 ("net: dsa: flush switchdev workqueue before tearing down CPU/DSA ports"), the port setup and teardown procedure became asymmetric. The fact of the matter is that user ports need the shared ports to be up before they can be used for CPU-initiated termination. And since we register net devices for the user ports, those won't be functional until we also call the setup for the shared (CPU, DSA) ports. But we may do that later, depending on the port numbering scheme of the hardware we are dealing with. It just makes sense that all shared ports are brought up before any user port is. I can't pinpoint any issue due to the current behavior, but let's change it nonetheless, for consistency's sake. Signed-off-by:
-
Vladimir Oltean authored
DSA needs to simulate master tracking events when a binding is first with a DSA master established and torn down, in order to give drivers the simplifying guarantee that ->master_state_change calls are made only when the master's readiness state to pass traffic changes. master_state_change() provide a operational bool that DSA driver can use to understand if DSA master is operational or not. To avoid races, we need to block the reception of NETDEV_UP/NETDEV_CHANGE/NETDEV_GOING_DOWN events in the netdev notifier chain while we are changing the master's dev->dsa_ptr (this changes what netdev_uses_dsa(dev) reports). The dsa_master_setup() and dsa_master_teardown() functions optionally require the rtnl_mutex to be held, if the tagger needs the master to be promiscuous, these functions call dev_set_promiscuity(). Move the rtnl_lock() from that function and make it top-level. Signed-off-by:
-
Vladimir Oltean authored
At present there are two paths for changing the MTU of the DSA master. The first is: dsa_tree_setup -> dsa_tree_setup_ports -> dsa_port_setup -> dsa_slave_create -> dsa_slave_change_mtu -> dev_set_mtu(master) The second is: dsa_tree_setup -> dsa_tree_setup_master -> dsa_master_setup -> dev_set_mtu(dev) So the dev_set_mtu() call from dsa_master_setup() has been effectively superseded by the dsa_slave_change_mtu(slave_dev, ETH_DATA_LEN) that is done from dsa_slave_create() for each user port. The later function also updates the master MTU according to the largest user port MTU from the tree. Therefore, updating the master MTU through a separate code path isn't needed. Signed-off-by: -
Vladimir Oltean authored
Currently dsa_slave_create() has two sequences of rtnl_lock/rtnl_unlock in a row. Remove the rtnl_unlock() and rtnl_lock() in between, such that the operation can execute slighly faster. Signed-off-by:
-
Vladimir Oltean authored
In dsa_slave_create() there are 2 sections that take rtnl_lock(): MTU change and netdev registration. They are separated by PHY initialization. There isn't any strict ordering requirement except for the fact that netdev registration should be last. Therefore, we can perform the MTU change a bit later, after the PHY setup. A future change will then be able to merge the two rtnl_lock sections into one. Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec-nextDavid S. Miller authored
Steffen Klassert says: ==================== pull request (net-next): ipsec-next 2022-01-06 1) Fix some clang_analyzer warnings about never read variables. From luo penghao. 2) Check for pols[0] only once in xfrm_expand_policies(). From Jean Sacren. 3) The SA curlft.use_time was updated only on SA cration time. Update whenever the SA is used. From Antony Antony 4) Add support for SM3 secure hash. From Xu Jia. 5) Add support for SM4 symmetric cipher algorithm. From Xu Jia. 6) Add a rate limit for SA mapping change messages. From Antony Antony. ==================== Signed-off-by: -
Jakub Kicinski authored
Aleksander Jan Bajkowski says: ==================== net: lantiq_xrx200: improve ethernet performance This patchset improves Ethernet performance by 15%. NAT Performance results on BT Home Hub 5A (kernel 5.10.89, mtu 1500): Down Up Before 539 Mbps 599 Mbps After 624 Mbps 695 Mbps ==================== Link: https://lore.kernel.org/r/20220104151144.181736-1-olek2@wp.plSigned-off-by:
Jakub Kicinski <kuba@kernel.org>
-
Aleksander Jan Bajkowski authored
We can increase the efficiency of rx path by using buffers to receive packets then build SKBs around them just before passing into the network stack. In contrast, preallocating SKBs too early reduces CPU cache efficiency. NAT Performance results on BT Home Hub 5A (kernel 5.10.89, mtu 1500): Down Up Before 577 Mbps 648 Mbps After 624 Mbps 695 Mbps Signed-off-by:
Aleksander Jan Bajkowski <olek2@wp.pl> Signed-off-by:
-
Aleksander Jan Bajkowski authored
NAT Performance results on BT Home Hub 5A (kernel 5.10.89, mtu 1500): Down Up Before 545 Mbps 625 Mbps After 577 Mbps 648 Mbps Signed-off-by:
-
Aleksander Jan Bajkowski authored
NAT Performance results on BT Home Hub 5A (kernel 5.10.89, mtu 1500): Down Up Before 539 Mbps 599 Mbps After 545 Mbps 625 Mbps Signed-off-by:
-
Miroslav Lichvar authored
When the -L option of the testptp utility is specified with other options (e.g. -p to enable PPS output), the user probably wants to apply it to the pin configured by the -L option. Reorder the code to set the pin function before other function requests to avoid confusing users. Signed-off-by:
Miroslav Lichvar <mlichvar@redhat.com> Reviewed-by:
Vladimir Oltean <olteanv@gmail.com> Acked-by:
Richard Cochran <richardcochran@gmail.com> Link: https://lore.kernel.org/r/20220105152506.3256026-1-mlichvar@redhat.comSigned-off-by:
-
- 05 Jan, 2022 25 commits
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/netdev/netJakub Kicinski authored
No conflicts. Signed-off-by: -
git://git.kernel.org/pub/scm/linux/kernel/git/netdev/netLinus Torvalds authored
Pull networking fixes from Jakub Kicinski" "Networking fixes, including fixes from bpf, and WiFi. One last pull request, turns out some of the recent fixes did more harm than good. Current release - regressions: - Revert "xsk: Do not sleep in poll() when need_wakeup set", made the problem worse - Revert "net: phy: fixed_phy: Fix NULL vs IS_ERR() checking in __fixed_phy_register", broke EPROBE_DEFER handling - Revert "net: usb: r8152: Add MAC pass-through support for more Lenovo Docks", broke setups without a Lenovo dock Current release - new code bugs: - selftests: set amt.sh executable Previous releases - regressions: - batman-adv: mcast: don't send link-local multicast to mcast routers Previous releases - always broken: - ipv4/ipv6: check attribute length for RTA_FLOW / RTA_GATEWAY - sctp: hold endpoint before calling cb in sctp_transport_lookup_process - mac80211: mesh: embed mesh_paths and mpp_paths into ieee80211_if_mesh to avoid complicated handling of sub-object allocation failures - seg6: fix traceroute in the presence of SRv6 - tipc: fix a kernel-infoleak in __tipc_sendmsg()" * tag 'net-5.16-final' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (36 commits) selftests: set amt.sh executable Revert "net: usb: r8152: Add MAC passthrough support for more Lenovo Docks" sfc: The RX page_ring is optional iavf: Fix limit of total number of queues to active queues of VF i40e: Fix incorrect netdev's real number of RX/TX queues i40e: Fix for displaying message regarding NVM version i40e: fix use-after-free in i40e_sync_filters_subtask() i40e: Fix to not show opcode msg on unsuccessful VF MAC change ieee802154: atusb: fix uninit value in atusb_set_extended_addr mac80211: mesh: embedd mesh_paths and mpp_paths into ieee80211_if_mesh mac80211: initialize variable have_higher_than_11mbit sch_qfq: prevent shift-out-of-bounds in qfq_init_qdisc netrom: fix copying in user data in nr_setsockopt udp6: Use Segment Routing Header for dest address if present icmp: ICMPV6: Examine invoking packet for Segment Route Headers. seg6: export get_srh() for ICMP handling Revert "net: phy: fixed_phy: Fix NULL vs IS_ERR() checking in __fixed_phy_register" ipv6: Do cleanup if attribute validation fails in multipath route ipv6: Continue processing multipath route even if gateway attribute is invalid net/fsl: Remove leftover definition in xgmac_mdio ... -
Russell King (Oracle) authored
The four RGMII interface modes take care of the required RGMII delay configuration at the PHY and should not be limited by the network MAC driver. Sadly, gemini was only permitting RGMII mode with no delays, which would require the required delay to be inserted via PCB tracking or by the MAC. However, there are designs that require the PHY to add the delay, which is impossible without Gemini permitting the other three PHY interface modes. Fix the driver to allow these. Signed-off-by:
Russell King (Oracle) <rmk+kernel@armlinux.org.uk> Reviewed-by:
Linus Walleij <linus.walleij@linaro.org> Tested-by:
Corentin Labbe <clabbe.montjoie@gmail.com> Link: https://lore.kernel.org/r/E1n4mpT-002PLd-Ha@rmk-PC.armlinux.org.ukSigned-off-by:
-
Jakub Kicinski authored
Russell King says: ==================== Fix RGMII delays for 88E1118 This series fixes the RGMII delays for 88E1118 Marvell PHYs, after a report by Corentin Labbe that the Marvell driver fails to work. Patch 1 cleans up the paged register accesses in m88e1118_config_init() and patch 2 adds the RGMII delay configuration. This comes with an element of risk as existing DT may need to be fixed for this in a similar way as we have done in the recent past for other PHY drivers that have misinterpreted the RGMII interface modes. ==================== Link: https://lore.kernel.org/r/YdR3wYFkm4eJApwb@shell.armlinux.org.ukSigned-off-by:
-
Russell King (Oracle) authored
Corentin Labbe reports that the SSI 1328 does not work when allowing the PHY to operate at gigabit speeds, but does work with the generic PHY driver. This appears to be because m88e1118_config_init() writes a fixed value to the MSCR register, claiming that this is to enable 1G speeds. However, this always sets bits 4 and 5, enabling RGMII transmit and receive delays. The suspicion is that the original board this was added for required the delays to make 1G speeds work. Add the necessary configuration for RGMII delays for the 88E1118 to bring this into line with the requirements for RGMII support, and thus make the SSI 1328 work. Corentin Labbe has tested this on gemini-ssi1328 and gemini-ns2502. Reported-by:
Andrew Lunn <andrew@lunn.ch> Signed-off-by:
-
Russell King (Oracle) authored
Use phy_write_paged() in m88e1118_config_init() to set the MSCR value. We leave the other paged write for the LEDs in case the DT register parsing is relying on this page. Signed-off-by:
-
Taehee Yoo authored
amt.sh test script will not work because it doesn't have execution permission. So, it adds execution permission. Reported-by:
Hangbin Liu <liuhangbin@gmail.com> Fixes: c08e8bae ("selftests: add amt interface selftest script") Signed-off-by:
Taehee Yoo <ap420073@gmail.com> Link: https://lore.kernel.org/r/20220105144436.13415-1-ap420073@gmail.comSigned-off-by:
-
Aaron Ma authored
This reverts commit f77b83b5. This change breaks multiple usb to ethernet dongles attached on Lenovo USB hub. Fixes: f77b83b5 ("net: usb: r8152: Add MAC passthrough support for more Lenovo Docks") Signed-off-by:
Aaron Ma <aaron.ma@canonical.com> Link: https://lore.kernel.org/r/20220105155102.8557-1-aaron.ma@canonical.comSigned-off-by:
-
Eric Dumazet authored
As reported by Johannes, the tracker allocated in ethnl_default_notify() is not really needed, as this function is not expected to change a device reference count. Fixes: e4b89540 ("netlink: add net device refcount tracker to struct ethnl_req_info") Signed-off-by:
Eric Dumazet <edumazet@google.com> Reported-by:
Johannes Berg <johannes@sipsolutions.net> Tested-by:
-
Eric Dumazet authored
qdisc_create() error path needs to use dev_put_track() because qdisc_alloc() allocated the tracker. Fixes: 606509f2 ("net/sched: add net device refcount tracker to struct Qdisc") Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/brgl/linuxLinus Torvalds authored
Pull gpio fixes from Bartosz Golaszewski: "Here are two last fixes for this release cycle from the GPIO subsystem: - fix irq offset calculation in gpio-aspeed-sgpio - update the MAINTAINERS entry for gpio-brcmstb" * tag 'gpio-fixes-for-v5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/brgl/linux: MAINTAINERS: update gpio-brcmstb maintainers gpio: gpio-aspeed-sgpio: Fix wrong hwirq base in irq handler
-
Jakub Kicinski authored
Merge tag 'ieee802154-for-net-2022-01-05' of git://git.kernel.org/pub/scm/linux/kernel/git/sschmidt/wpan Stefan Schmidt says: ==================== pull-request: ieee802154 for net 2022-01-05 Below I have a last minute fix for the atusb driver. Pavel fixes a KASAN uninit report for the driver. This version is the minimal impact fix to ease backporting. A bigger rework of the driver to avoid potential similar problems is ongoing and will come through net-next when ready. * tag 'ieee802154-for-net-2022-01-05' of git://git.kernel.org/pub/scm/linux/kernel/git/sschmidt/wpan: ieee802154: atusb: fix uninit value in atusb_set_extended_addr ==================== Link: https://lore.kernel.org/r/20220105153914.512305-1-stefan@datenfreihafen.orgSigned-off-by:
-
David S. Miller authored
Vladimir Oltean says: ==================== DSA cross-chip notifier cleanup This series deletes the no-op cross-chip notifier support for MRP and HSR, features which were introduced relatively recently and did not get full review at the time. The new code is functionally equivalent, but simpler. Cc: Horatiu Vultur <horatiu.vultur@microchip.com> Cc: George McCollister <george.mccollister@gmail.com> ==================== Signed-off-by: -
Vladimir Oltean authored
The cross-chip notifiers for HSR are bypass operations, meaning that even though all switches in a tree are notified, only the switch specified in the info structure is targeted. We can eliminate the unnecessary complexity by deleting the cross-chip notifier logic and calling the ds->ops straight from port.c. Cc: George McCollister <george.mccollister@gmail.com> Signed-off-by:
George McCollister <george.mccollister@gmail.com> Signed-off-by:
-
Vladimir Oltean authored
The cross-chip notifiers for MRP are bypass operations, meaning that even though all switches in a tree are notified, only the switch specified in the info structure is targeted. We can eliminate the unnecessary complexity by deleting the cross-chip notifier logic and calling the ds->ops straight from port.c. Cc: Horatiu Vultur <horatiu.vultur@microchip.com> Signed-off-by:
-
Vladimir Oltean authored
The cross-chip notifier boilerplate code meant to check the presence of ds->ops->port_mrp_add_ring_role before calling it, but checked ds->ops->port_mrp_add instead, before calling ds->ops->port_mrp_add_ring_role. Therefore, a driver which implements one operation but not the other would trigger a NULL pointer dereference. There isn't any such driver in DSA yet, so there is no reason to backport the change. Issue found through code inspection. Cc: Horatiu Vultur <horatiu.vultur@microchip.com> Fixes: c595c433 ("net: dsa: add MRP support") Signed-off-by:
-
Danielle Ratson authored
Locally generated packets ingress the device through its CPU port. When the CPU port is congested and there are not enough credits in its headroom buffer, packets can be dropped. While this might be acceptable for data packets that traverse the network, configuration packets exchanged between the host and the device (EMADs) should not be subjected to this flow control. The "sdq_lp" bit in the SDQ (Send Descriptor Queue) context allows the host to instruct the device to treat packets sent on this queue as "local processing" and always process them, regardless of the state of the CPU port's headroom. Add the definition of this bit and set it for the dedicated SDQ reserved for the transmission of EMAD packets. This makes the "local processing" bit in the WQE (Work Queue Element) redundant, so clear it. Signed-off-by:
Danielle Ratson <danieller@nvidia.com> Signed-off-by:
Ido Schimmel <idosch@nvidia.com> Signed-off-by:
-
David S. Miller authored
Merge tag 'linux-can-next-for-5.17-20220105' of git://git.kernel.org/pub/scm/linux/kernel/git/mkl/linux-can-next Marc Kleine-Budde says: ==================== pull-request: can-next 2022-01-05 this is a pull request of 15 patches for net-next/master. The first patch is by me and removed an unused variable from the usb_8dev driver. Andy Shevchenko contributes a patch for the mcp251x driver, which removes an unneeded assignment. Jimmy Assarsson's patch for the kvaser_usb makes use of units.h in the assignment of frequencies. Lad Prabhakar provides 2 patches, converting the ti_hecc and the sja1000 driver to make use of platform_get_irq(). The 10 remaining patches are by Vincent Mailhol. First the etas_es58x driver populates the net_device::dev_port. The next 5 patches cleanup the handling of CAN error and CAN RTR messages of all drivers. The remaining 4 patches enhance the CAN controller mode flag handling and export it via netlink to user space. ==================== Signed-off-by:
-
David S. Miller authored
Vladimir Oltean says: ==================== Cleanup to main DSA structures This series contains changes that do the following: - struct dsa_port reduced from 576 to 544 bytes, and first cache line a bit better organized - struct dsa_switch from 160 to 136 bytes, and first cache line a bit better organized - struct dsa_switch_tree from 112 to 104 bytes, and first cache line a bit better organized ==================== Signed-off-by: -
Vladimir Oltean authored
There is a 7 byte hole after dst->setup and a 4 byte hole after dst->default_proto. Combining them, we have a single hole of just 3 bytes on 64 bit machines. Before: pahole -C dsa_switch_tree net/dsa/slave.o struct dsa_switch_tree { struct list_head list; /* 0 16 */ struct list_head ports; /* 16 16 */ struct raw_notifier_head nh; /* 32 8 */ unsigned int index; /* 40 4 */ struct kref refcount; /* 44 4 */ struct net_device * * lags; /* 48 8 */ bool setup; /* 56 1 */ /* XXX 7 bytes hole, try to pack */ /* --- cacheline 1 boundary (64 bytes) --- */ const struct dsa_device_ops * tag_ops; /* 64 8 */ enum dsa_tag_protocol default_proto; /* 72 4 */ /* XXX 4 bytes hole, try to pack */ struct dsa_platform_data * pd; /* 80 8 */ struct list_head rtable; /* 88 16 */ unsigned int lags_len; /* 104 4 */ unsigned int last_switch; /* 108 4 */ /* size: 112, cachelines: 2, members: 13 */ /* sum members: 101, holes: 2, sum holes: 11 */ /* last cacheline: 48 bytes */ }; After: pahole -C dsa_switch_tree net/dsa/slave.o struct dsa_switch_tree { struct list_head list; /* 0 16 */ struct list_head ports; /* 16 16 */ struct raw_notifier_head nh; /* 32 8 */ unsigned int index; /* 40 4 */ struct kref refcount; /* 44 4 */ struct net_device * * lags; /* 48 8 */ const struct dsa_device_ops * tag_ops; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ enum dsa_tag_protocol default_proto; /* 64 4 */ bool setup; /* 68 1 */ /* XXX 3 bytes hole, try to pack */ struct dsa_platform_data * pd; /* 72 8 */ struct list_head rtable; /* 80 16 */ unsigned int lags_len; /* 96 4 */ unsigned int last_switch; /* 100 4 */ /* size: 104, cachelines: 2, members: 13 */ /* sum members: 101, holes: 1, sum holes: 3 */ /* last cacheline: 40 bytes */ }; Signed-off-by: -
Vladimir Oltean authored
dst->ports is accessed most notably by dsa_master_find_slave(), which is invoked in the RX path. dst->lags is accessed by dsa_lag_dev(), which is invoked in the RX path of tag_dsa.c. dst->tag_ops, dst->default_proto and dst->pd don't need to be in the first cache line, so they are moved out by this change. Before: pahole -C dsa_switch_tree net/dsa/slave.o struct dsa_switch_tree { struct list_head list; /* 0 16 */ struct raw_notifier_head nh; /* 16 8 */ unsigned int index; /* 24 4 */ struct kref refcount; /* 28 4 */ bool setup; /* 32 1 */ /* XXX 7 bytes hole, try to pack */ const struct dsa_device_ops * tag_ops; /* 40 8 */ enum dsa_tag_protocol default_proto; /* 48 4 */ /* XXX 4 bytes hole, try to pack */ struct dsa_platform_data * pd; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct list_head ports; /* 64 16 */ struct list_head rtable; /* 80 16 */ struct net_device * * lags; /* 96 8 */ unsigned int lags_len; /* 104 4 */ unsigned int last_switch; /* 108 4 */ /* size: 112, cachelines: 2, members: 13 */ /* sum members: 101, holes: 2, sum holes: 11 */ /* last cacheline: 48 bytes */ }; After: pahole -C dsa_switch_tree net/dsa/slave.o struct dsa_switch_tree { struct list_head list; /* 0 16 */ struct list_head ports; /* 16 16 */ struct raw_notifier_head nh; /* 32 8 */ unsigned int index; /* 40 4 */ struct kref refcount; /* 44 4 */ struct net_device * * lags; /* 48 8 */ bool setup; /* 56 1 */ /* XXX 7 bytes hole, try to pack */ /* --- cacheline 1 boundary (64 bytes) --- */ const struct dsa_device_ops * tag_ops; /* 64 8 */ enum dsa_tag_protocol default_proto; /* 72 4 */ /* XXX 4 bytes hole, try to pack */ struct dsa_platform_data * pd; /* 80 8 */ struct list_head rtable; /* 88 16 */ unsigned int lags_len; /* 104 4 */ unsigned int last_switch; /* 108 4 */ /* size: 112, cachelines: 2, members: 13 */ /* sum members: 101, holes: 2, sum holes: 11 */ /* last cacheline: 48 bytes */ }; Signed-off-by: -
Vladimir Oltean authored
Currently, num_ports is declared as size_t, which is defined as __kernel_ulong_t, therefore it occupies 8 bytes of memory. Even switches with port numbers in the range of tens are exotic, so there is no need for this amount of storage. Additionally, because the max_num_bridges member right above it is also 4 bytes, it means the compiler needs to add padding between the last 2 fields. By reducing the size, we don't need that padding and can reduce the struct size. Before: pahole -C dsa_switch net/dsa/slave.o struct dsa_switch { struct device * dev; /* 0 8 */ struct dsa_switch_tree * dst; /* 8 8 */ unsigned int index; /* 16 4 */ u32 setup:1; /* 20: 0 4 */ u32 vlan_filtering_is_global:1; /* 20: 1 4 */ u32 needs_standalone_vlan_filtering:1; /* 20: 2 4 */ u32 configure_vlan_while_not_filtering:1; /* 20: 3 4 */ u32 untag_bridge_pvid:1; /* 20: 4 4 */ u32 assisted_learning_on_cpu_port:1; /* 20: 5 4 */ u32 vlan_filtering:1; /* 20: 6 4 */ u32 pcs_poll:1; /* 20: 7 4 */ u32 mtu_enforcement_ingress:1; /* 20: 8 4 */ /* XXX 23 bits hole, try to pack */ struct notifier_block nb; /* 24 24 */ /* XXX last struct has 4 bytes of padding */ void * priv; /* 48 8 */ void * tagger_data; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct dsa_chip_data * cd; /* 64 8 */ const struct dsa_switch_ops * ops; /* 72 8 */ u32 phys_mii_mask; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct mii_bus * slave_mii_bus; /* 88 8 */ unsigned int ageing_time_min; /* 96 4 */ unsigned int ageing_time_max; /* 100 4 */ struct dsa_8021q_context * tag_8021q_ctx; /* 104 8 */ struct devlink * devlink; /* 112 8 */ unsigned int num_tx_queues; /* 120 4 */ unsigned int num_lag_ids; /* 124 4 */ /* --- cacheline 2 boundary (128 bytes) --- */ unsigned int max_num_bridges; /* 128 4 */ /* XXX 4 bytes hole, try to pack */ size_t num_ports; /* 136 8 */ /* size: 144, cachelines: 3, members: 27 */ /* sum members: 132, holes: 2, sum holes: 8 */ /* sum bitfield members: 9 bits, bit holes: 1, sum bit holes: 23 bits */ /* paddings: 1, sum paddings: 4 */ /* last cacheline: 16 bytes */ }; After: pahole -C dsa_switch net/dsa/slave.o struct dsa_switch { struct device * dev; /* 0 8 */ struct dsa_switch_tree * dst; /* 8 8 */ unsigned int index; /* 16 4 */ u32 setup:1; /* 20: 0 4 */ u32 vlan_filtering_is_global:1; /* 20: 1 4 */ u32 needs_standalone_vlan_filtering:1; /* 20: 2 4 */ u32 configure_vlan_while_not_filtering:1; /* 20: 3 4 */ u32 untag_bridge_pvid:1; /* 20: 4 4 */ u32 assisted_learning_on_cpu_port:1; /* 20: 5 4 */ u32 vlan_filtering:1; /* 20: 6 4 */ u32 pcs_poll:1; /* 20: 7 4 */ u32 mtu_enforcement_ingress:1; /* 20: 8 4 */ /* XXX 23 bits hole, try to pack */ struct notifier_block nb; /* 24 24 */ /* XXX last struct has 4 bytes of padding */ void * priv; /* 48 8 */ void * tagger_data; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct dsa_chip_data * cd; /* 64 8 */ const struct dsa_switch_ops * ops; /* 72 8 */ u32 phys_mii_mask; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct mii_bus * slave_mii_bus; /* 88 8 */ unsigned int ageing_time_min; /* 96 4 */ unsigned int ageing_time_max; /* 100 4 */ struct dsa_8021q_context * tag_8021q_ctx; /* 104 8 */ struct devlink * devlink; /* 112 8 */ unsigned int num_tx_queues; /* 120 4 */ unsigned int num_lag_ids; /* 124 4 */ /* --- cacheline 2 boundary (128 bytes) --- */ unsigned int max_num_bridges; /* 128 4 */ unsigned int num_ports; /* 132 4 */ /* size: 136, cachelines: 3, members: 27 */ /* sum members: 128, holes: 1, sum holes: 4 */ /* sum bitfield members: 9 bits, bit holes: 1, sum bit holes: 23 bits */ /* paddings: 1, sum paddings: 4 */ /* last cacheline: 8 bytes */ }; Signed-off-by: -
Vladimir Oltean authored
struct dsa_switch has 9 boolean properties, many of which are in fact set by drivers for custom behavior (vlan_filtering_is_global, needs_standalone_vlan_filtering, etc etc). The binary layout of the structure could be improved. For example, the "bool setup" at the beginning introduces a gratuitous 7 byte hole in the first cache line. The change merges all boolean properties into bitfields of an u32, and places that u32 in the first cache line of the structure, since many bools are accessed from the data path (untag_bridge_pvid, vlan_filtering, vlan_filtering_is_global). We place this u32 after the existing ds->index, which is also 4 bytes in size. As a positive side effect, ds->tagger_data now fits into the first cache line too, because 4 bytes are saved. Before: pahole -C dsa_switch net/dsa/slave.o struct dsa_switch { bool setup; /* 0 1 */ /* XXX 7 bytes hole, try to pack */ struct device * dev; /* 8 8 */ struct dsa_switch_tree * dst; /* 16 8 */ unsigned int index; /* 24 4 */ /* XXX 4 bytes hole, try to pack */ struct notifier_block nb; /* 32 24 */ /* XXX last struct has 4 bytes of padding */ void * priv; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ void * tagger_data; /* 64 8 */ struct dsa_chip_data * cd; /* 72 8 */ const struct dsa_switch_ops * ops; /* 80 8 */ u32 phys_mii_mask; /* 88 4 */ /* XXX 4 bytes hole, try to pack */ struct mii_bus * slave_mii_bus; /* 96 8 */ unsigned int ageing_time_min; /* 104 4 */ unsigned int ageing_time_max; /* 108 4 */ struct dsa_8021q_context * tag_8021q_ctx; /* 112 8 */ struct devlink * devlink; /* 120 8 */ /* --- cacheline 2 boundary (128 bytes) --- */ unsigned int num_tx_queues; /* 128 4 */ bool vlan_filtering_is_global; /* 132 1 */ bool needs_standalone_vlan_filtering; /* 133 1 */ bool configure_vlan_while_not_filtering; /* 134 1 */ bool untag_bridge_pvid; /* 135 1 */ bool assisted_learning_on_cpu_port; /* 136 1 */ bool vlan_filtering; /* 137 1 */ bool pcs_poll; /* 138 1 */ bool mtu_enforcement_ingress; /* 139 1 */ unsigned int num_lag_ids; /* 140 4 */ unsigned int max_num_bridges; /* 144 4 */ /* XXX 4 bytes hole, try to pack */ size_t num_ports; /* 152 8 */ /* size: 160, cachelines: 3, members: 27 */ /* sum members: 141, holes: 4, sum holes: 19 */ /* paddings: 1, sum paddings: 4 */ /* last cacheline: 32 bytes */ }; After: pahole -C dsa_switch net/dsa/slave.o struct dsa_switch { struct device * dev; /* 0 8 */ struct dsa_switch_tree * dst; /* 8 8 */ unsigned int index; /* 16 4 */ u32 setup:1; /* 20: 0 4 */ u32 vlan_filtering_is_global:1; /* 20: 1 4 */ u32 needs_standalone_vlan_filtering:1; /* 20: 2 4 */ u32 configure_vlan_while_not_filtering:1; /* 20: 3 4 */ u32 untag_bridge_pvid:1; /* 20: 4 4 */ u32 assisted_learning_on_cpu_port:1; /* 20: 5 4 */ u32 vlan_filtering:1; /* 20: 6 4 */ u32 pcs_poll:1; /* 20: 7 4 */ u32 mtu_enforcement_ingress:1; /* 20: 8 4 */ /* XXX 23 bits hole, try to pack */ struct notifier_block nb; /* 24 24 */ /* XXX last struct has 4 bytes of padding */ void * priv; /* 48 8 */ void * tagger_data; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct dsa_chip_data * cd; /* 64 8 */ const struct dsa_switch_ops * ops; /* 72 8 */ u32 phys_mii_mask; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct mii_bus * slave_mii_bus; /* 88 8 */ unsigned int ageing_time_min; /* 96 4 */ unsigned int ageing_time_max; /* 100 4 */ struct dsa_8021q_context * tag_8021q_ctx; /* 104 8 */ struct devlink * devlink; /* 112 8 */ unsigned int num_tx_queues; /* 120 4 */ unsigned int num_lag_ids; /* 124 4 */ /* --- cacheline 2 boundary (128 bytes) --- */ unsigned int max_num_bridges; /* 128 4 */ /* XXX 4 bytes hole, try to pack */ size_t num_ports; /* 136 8 */ /* size: 144, cachelines: 3, members: 27 */ /* sum members: 132, holes: 2, sum holes: 8 */ /* sum bitfield members: 9 bits, bit holes: 1, sum bit holes: 23 bits */ /* paddings: 1, sum paddings: 4 */ /* last cacheline: 16 bytes */ }; Signed-off-by: -
Vladimir Oltean authored
Both dsa_port :: type and dsa_port :: index introduce a 4 octet hole after them, so we can group them together and the holes would be eliminated, turning 16 octets of storage into just 8. This makes the cpu_dp pointer fit in the first cache line, which is good, because dsa_slave_to_master(), called by dsa_enqueue_skb(), uses it. Before: pahole -C dsa_port net/dsa/slave.o struct dsa_port { union { struct net_device * master; /* 0 8 */ struct net_device * slave; /* 0 8 */ }; /* 0 8 */ const struct dsa_device_ops * tag_ops; /* 8 8 */ struct dsa_switch_tree * dst; /* 16 8 */ struct sk_buff * (*rcv)(struct sk_buff *, struct net_device *); /* 24 8 */ enum { DSA_PORT_TYPE_UNUSED = 0, DSA_PORT_TYPE_CPU = 1, DSA_PORT_TYPE_DSA = 2, DSA_PORT_TYPE_USER = 3, } type; /* 32 4 */ /* XXX 4 bytes hole, try to pack */ struct dsa_switch * ds; /* 40 8 */ unsigned int index; /* 48 4 */ /* XXX 4 bytes hole, try to pack */ const char * name; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct dsa_port * cpu_dp; /* 64 8 */ u8 mac[6]; /* 72 6 */ u8 stp_state; /* 78 1 */ u8 vlan_filtering:1; /* 79: 0 1 */ u8 learning:1; /* 79: 1 1 */ u8 lag_tx_enabled:1; /* 79: 2 1 */ u8 devlink_port_setup:1; /* 79: 3 1 */ u8 setup:1; /* 79: 4 1 */ /* XXX 3 bits hole, try to pack */ struct device_node * dn; /* 80 8 */ unsigned int ageing_time; /* 88 4 */ /* XXX 4 bytes hole, try to pack */ struct dsa_bridge * bridge; /* 96 8 */ struct devlink_port devlink_port; /* 104 288 */ /* --- cacheline 6 boundary (384 bytes) was 8 bytes ago --- */ struct phylink * pl; /* 392 8 */ struct phylink_config pl_config; /* 400 40 */ struct net_device * lag_dev; /* 440 8 */ /* --- cacheline 7 boundary (448 bytes) --- */ struct net_device * hsr_dev; /* 448 8 */ struct list_head list; /* 456 16 */ const struct ethtool_ops * orig_ethtool_ops; /* 472 8 */ const struct dsa_netdevice_ops * netdev_ops; /* 480 8 */ struct mutex addr_lists_lock; /* 488 32 */ /* --- cacheline 8 boundary (512 bytes) was 8 bytes ago --- */ struct list_head fdbs; /* 520 16 */ struct list_head mdbs; /* 536 16 */ /* size: 552, cachelines: 9, members: 30 */ /* sum members: 539, holes: 3, sum holes: 12 */ /* sum bitfield members: 5 bits, bit holes: 1, sum bit holes: 3 bits */ /* last cacheline: 40 bytes */ }; After: pahole -C dsa_port net/dsa/slave.o struct dsa_port { union { struct net_device * master; /* 0 8 */ struct net_device * slave; /* 0 8 */ }; /* 0 8 */ const struct dsa_device_ops * tag_ops; /* 8 8 */ struct dsa_switch_tree * dst; /* 16 8 */ struct sk_buff * (*rcv)(struct sk_buff *, struct net_device *); /* 24 8 */ struct dsa_switch * ds; /* 32 8 */ unsigned int index; /* 40 4 */ enum { DSA_PORT_TYPE_UNUSED = 0, DSA_PORT_TYPE_CPU = 1, DSA_PORT_TYPE_DSA = 2, DSA_PORT_TYPE_USER = 3, } type; /* 44 4 */ const char * name; /* 48 8 */ struct dsa_port * cpu_dp; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ u8 mac[6]; /* 64 6 */ u8 stp_state; /* 70 1 */ u8 vlan_filtering:1; /* 71: 0 1 */ u8 learning:1; /* 71: 1 1 */ u8 lag_tx_enabled:1; /* 71: 2 1 */ u8 devlink_port_setup:1; /* 71: 3 1 */ u8 setup:1; /* 71: 4 1 */ /* XXX 3 bits hole, try to pack */ struct device_node * dn; /* 72 8 */ unsigned int ageing_time; /* 80 4 */ /* XXX 4 bytes hole, try to pack */ struct dsa_bridge * bridge; /* 88 8 */ struct devlink_port devlink_port; /* 96 288 */ /* --- cacheline 6 boundary (384 bytes) --- */ struct phylink * pl; /* 384 8 */ struct phylink_config pl_config; /* 392 40 */ struct net_device * lag_dev; /* 432 8 */ struct net_device * hsr_dev; /* 440 8 */ /* --- cacheline 7 boundary (448 bytes) --- */ struct list_head list; /* 448 16 */ const struct ethtool_ops * orig_ethtool_ops; /* 464 8 */ const struct dsa_netdevice_ops * netdev_ops; /* 472 8 */ struct mutex addr_lists_lock; /* 480 32 */ /* --- cacheline 8 boundary (512 bytes) --- */ struct list_head fdbs; /* 512 16 */ struct list_head mdbs; /* 528 16 */ /* size: 544, cachelines: 9, members: 30 */ /* sum members: 539, holes: 1, sum holes: 4 */ /* sum bitfield members: 5 bits, bit holes: 1, sum bit holes: 3 bits */ /* last cacheline: 32 bytes */ }; Signed-off-by: -
Vladimir Oltean authored
struct dsa_port has 5 bool members which create quite a number of 7 byte holes in the structure layout. By merging them all into bitfields of an u8, and placing that u8 in the 1-byte hole after dp->mac and dp->stp_state, we can reduce the structure size from 576 bytes to 552 bytes on arm64. Before: pahole -C dsa_port net/dsa/slave.o struct dsa_port { union { struct net_device * master; /* 0 8 */ struct net_device * slave; /* 0 8 */ }; /* 0 8 */ const struct dsa_device_ops * tag_ops; /* 8 8 */ struct dsa_switch_tree * dst; /* 16 8 */ struct sk_buff * (*rcv)(struct sk_buff *, struct net_device *); /* 24 8 */ enum { DSA_PORT_TYPE_UNUSED = 0, DSA_PORT_TYPE_CPU = 1, DSA_PORT_TYPE_DSA = 2, DSA_PORT_TYPE_USER = 3, } type; /* 32 4 */ /* XXX 4 bytes hole, try to pack */ struct dsa_switch * ds; /* 40 8 */ unsigned int index; /* 48 4 */ /* XXX 4 bytes hole, try to pack */ const char * name; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct dsa_port * cpu_dp; /* 64 8 */ u8 mac[6]; /* 72 6 */ u8 stp_state; /* 78 1 */ /* XXX 1 byte hole, try to pack */ struct device_node * dn; /* 80 8 */ unsigned int ageing_time; /* 88 4 */ bool vlan_filtering; /* 92 1 */ bool learning; /* 93 1 */ /* XXX 2 bytes hole, try to pack */ struct dsa_bridge * bridge; /* 96 8 */ struct devlink_port devlink_port; /* 104 288 */ /* --- cacheline 6 boundary (384 bytes) was 8 bytes ago --- */ bool devlink_port_setup; /* 392 1 */ /* XXX 7 bytes hole, try to pack */ struct phylink * pl; /* 400 8 */ struct phylink_config pl_config; /* 408 40 */ /* --- cacheline 7 boundary (448 bytes) --- */ struct net_device * lag_dev; /* 448 8 */ bool lag_tx_enabled; /* 456 1 */ /* XXX 7 bytes hole, try to pack */ struct net_device * hsr_dev; /* 464 8 */ struct list_head list; /* 472 16 */ const struct ethtool_ops * orig_ethtool_ops; /* 488 8 */ const struct dsa_netdevice_ops * netdev_ops; /* 496 8 */ struct mutex addr_lists_lock; /* 504 32 */ /* --- cacheline 8 boundary (512 bytes) was 24 bytes ago --- */ struct list_head fdbs; /* 536 16 */ struct list_head mdbs; /* 552 16 */ bool setup; /* 568 1 */ /* size: 576, cachelines: 9, members: 30 */ /* sum members: 544, holes: 6, sum holes: 25 */ /* padding: 7 */ }; After: pahole -C dsa_port net/dsa/slave.o struct dsa_port { union { struct net_device * master; /* 0 8 */ struct net_device * slave; /* 0 8 */ }; /* 0 8 */ const struct dsa_device_ops * tag_ops; /* 8 8 */ struct dsa_switch_tree * dst; /* 16 8 */ struct sk_buff * (*rcv)(struct sk_buff *, struct net_device *); /* 24 8 */ enum { DSA_PORT_TYPE_UNUSED = 0, DSA_PORT_TYPE_CPU = 1, DSA_PORT_TYPE_DSA = 2, DSA_PORT_TYPE_USER = 3, } type; /* 32 4 */ /* XXX 4 bytes hole, try to pack */ struct dsa_switch * ds; /* 40 8 */ unsigned int index; /* 48 4 */ /* XXX 4 bytes hole, try to pack */ const char * name; /* 56 8 */ /* --- cacheline 1 boundary (64 bytes) --- */ struct dsa_port * cpu_dp; /* 64 8 */ u8 mac[6]; /* 72 6 */ u8 stp_state; /* 78 1 */ u8 vlan_filtering:1; /* 79: 0 1 */ u8 learning:1; /* 79: 1 1 */ u8 lag_tx_enabled:1; /* 79: 2 1 */ u8 devlink_port_setup:1; /* 79: 3 1 */ u8 setup:1; /* 79: 4 1 */ /* XXX 3 bits hole, try to pack */ struct device_node * dn; /* 80 8 */ unsigned int ageing_time; /* 88 4 */ /* XXX 4 bytes hole, try to pack */ struct dsa_bridge * bridge; /* 96 8 */ struct devlink_port devlink_port; /* 104 288 */ /* --- cacheline 6 boundary (384 bytes) was 8 bytes ago --- */ struct phylink * pl; /* 392 8 */ struct phylink_config pl_config; /* 400 40 */ struct net_device * lag_dev; /* 440 8 */ /* --- cacheline 7 boundary (448 bytes) --- */ struct net_device * hsr_dev; /* 448 8 */ struct list_head list; /* 456 16 */ const struct ethtool_ops * orig_ethtool_ops; /* 472 8 */ const struct dsa_netdevice_ops * netdev_ops; /* 480 8 */ struct mutex addr_lists_lock; /* 488 32 */ /* --- cacheline 8 boundary (512 bytes) was 8 bytes ago --- */ struct list_head fdbs; /* 520 16 */ struct list_head mdbs; /* 536 16 */ /* size: 552, cachelines: 9, members: 30 */ /* sum members: 539, holes: 3, sum holes: 12 */ /* sum bitfield members: 5 bits, bit holes: 1, sum bit holes: 3 bits */ /* last cacheline: 40 bytes */ }; Signed-off-by:
-
