- 06 Feb, 2021 40 commits
-
-
Alex Elder authored
When stopping a channel, gsi_channel_stop() will ensure NAPI polling is complete when it calls napi_disable(). So there is no need to call napi_synchronize() in that case. Move the call to napi_synchronize() out of __gsi_channel_stop() and into gsi_channel_suspend(), so it's only used where needed. Signed-off-by:
Alex Elder <elder@linaro.org> Signed-off-by:
Jakub Kicinski <kuba@kernel.org>
-
Alex Elder authored
Move the mutex calls out of gsi_channel_stop_retry() and into __gsi_channel_stop(), to make the latter more semantically similar to __gsi_channel_start(). Signed-off-by:
-
Jakub Kicinski authored
Vladimir Oltean says: ==================== LAG offload for Ocelot DSA switches This patch series reworks the ocelot switchdev driver such that it could share the same implementation for LAG offload as the felix DSA driver. Testing has been done in the following topology: +----------------------------------+ | Board 1 br0 | | +---------+ | | / \ | | | | | | | bond0 | | | +-----+ | | | / \ | | eno0 swp0 swp1 swp2 | +---|--------|-------|-------|-----+ | | | | +--------+ | | Cable | | Cable| |Cable Cable | | +--------+ | | | | | | +---|--------|-------|-------|-----+ | eno0 swp0 swp1 swp2 | | | \ / | | | +-----+ | | | bond0 | | | | | | \ / | | +---------+ | | Board 2 br0 | +----------------------------------+ The same script can be run on both Board 1 and Board 2 to set this up: ip link del bond0 ip link add bond0 type bond mode balance-xor miimon 1 OR ip link add bond0 type bond mode 802.3ad ip link set swp1 down && ip link set swp1 master bond0 && ip link set swp1 up ip link set swp2 down && ip link set swp2 master bond0 && ip link set swp2 up ip link del br0 ip link add br0 type bridge ip link set bond0 master br0 ip link set swp0 master br0 Then traffic can be tested between eno0 of Board 1 and eno0 of Board 2. ==================== Link: https://lore.kernel.org/r/20210205220221.255646-1-olteanv@gmail.comSigned-off-by: -
Vladimir Oltean authored
The ocelot switch has been supporting LAG offload since its initial commit, however felix could not make use of that, due to lack of a LAG abstraction in DSA. Now that we have that, let's forward DSA's calls towards the ocelot library, who will deal with setting up the bonding. Signed-off-by:
Vladimir Oltean <vladimir.oltean@nxp.com> Signed-off-by:
-
Vladimir Oltean authored
Given the following topology, and focusing only on Box A: Box A +----------------------------------+ | Board 1 br0 | | +---------+ | | / \ | | | | | | | bond0 | | | +-----+ | |192.168.1.1 | / \ | | eno0 swp0 swp1 swp2 | +---|--------|-------|-------|-----+ | | | | +--------+ | | Cable | | Cable| |Cable Cable | | +--------+ | | | | | | +---|--------|-------|-------|-----+ | eno0 swp0 swp1 swp2 | |192.168.1.2 | \ / | | | +-----+ | | | bond0 | | | | | | \ / | | +---------+ | | Board 2 br0 | +----------------------------------+ Box B The assisted_learning_on_cpu_port logic will see that swp0 is bridged with a "foreign interface" (bond0) and will therefore install all addresses learnt by the software bridge towards bond0 (including the address of eno0 on Box B) as static addresses towards the CPU port. But that's not what we want - bond0 is not really a "foreign interface" but one we can offload including L2 forwarding from/towards it. So we need to refine our logic for assisted learning such that, whenever we see an address learnt on a non-DSA interface, we search through the tree for any port that offloads that non-DSA interface. Some confusion might arise as to why we search through the whole tree instead of just the local switch returned by dsa_slave_dev_lower_find. Or a different angle of the same confusion: why does dsa_slave_dev_lower_find(br_dev) return a single dp that's under br_dev instead of the whole list of bridged DSA ports? To answer the second question, it should be enough to install the static FDB entry on the CPU port of a single switch in the tree, because dsa_port_fdb_add uses DSA_NOTIFIER_FDB_ADD which ensures that all other switches in the tree get notified of that address, and add the entry themselves using dsa_towards_port(). This should help understand the answer to the first question: the port returned by dsa_slave_dev_lower_find may not be on the same switch as the ports that offload the LAG. Nonetheless, if the driver implements .crosschip_lag_join and .crosschip_bridge_join as mv88e6xxx does, there still isn't any reason for trapping addresses learnt on the remote LAG towards the CPU, and we should prevent that. Signed-off-by: -
Vladimir Oltean authored
At present there is an issue when ocelot is offloading a bonding interface, but one of the links of the physical ports goes down. Traffic keeps being hashed towards that destination, and of course gets dropped on egress. Monitor the netdev notifier events emitted by the bonding driver for changes in the physical state of lower interfaces, to determine which ports are active and which ones are no longer. Then extend ocelot_get_bond_mask to return either the configured bonding interfaces, or the active ones, depending on a boolean argument. The code that does rebalancing only needs to do so among the active ports, whereas the bridge forwarding mask and the logical port IDs still need to look at the permanently bonded ports. Signed-off-by:
-
Vladimir Oltean authored
It makes it a bit easier to read and understand the code that deals with balancing the 16 aggregation codes among the ports in a certain LAG. Signed-off-by:
-
Vladimir Oltean authored
We can now simplify the implementation by always using ocelot_get_bond_mask to look up the other ports that are offloading the same bonding interface as us. In ocelot_set_aggr_pgids, the code had a way to uniquely iterate through LAGs. We need to achieve the same behavior by marking each LAG as visited, which we do now by using a temporary 32-bit "visited" bitmask. This is ok and we do not need dynamic memory allocation, because we know that this switch architecture will not have more than 32 ports (the PGID port masks are 32-bit anyway). Signed-off-by:
-
Vladimir Oltean authored
The setup of logical port IDs is done in two places: from the inconclusively named ocelot_setup_lag and from ocelot_port_lag_leave, a function that also calls ocelot_setup_lag (which apparently does an incomplete setup of the LAG). To improve this situation, we can rename ocelot_setup_lag into ocelot_setup_logical_port_ids, and drop the "lag" argument. It will now set up the logical port IDs of all switch ports, which may be just slightly more inefficient but more maintainable. Signed-off-by:
-
Vladimir Oltean authored
The index of the LAG is equal to the logical port ID that all the physical port members have, which is further equal to the index of the first physical port that is a member of the LAG. The code gets a bit carried away with logic like this: if (a == b) c = a; else c = b; which can be simplified, of course, into: c = b; (with a being port, b being lp, c being lag) This further makes the "lp" variable redundant, since we can use "lag" everywhere where "lp" (logical port) was used. So instead of a "c = b" assignment, we can do a complete deletion of b. Only one comment here: if (bond_mask) { lp = __ffs(bond_mask); ocelot->lags[lp] = 0; } lp was clobbered before, because it was used as a temporary variable to hold the new smallest port ID from the bond. Now that we don't have "lp" any longer, we'll just avoid the temporary variable and zeroize the bonding mask directly. Signed-off-by:Alexandre Belloni <alexandre.belloni@bootlin.com> Signed-off-by:
-
Vladimir Oltean authored
Since this code should be called from pure switchdev as well as from DSA, we must find a way to determine the bonding mask not by looking directly at the net_device lowers of the bonding interface, since those could have different private structures. We keep a pointer to the bonding upper interface, if present, in struct ocelot_port. Then the bonding mask becomes the bitwise OR of all ports that have the same bonding upper interface. This adds a duplication of functionality with the current "lags" array, but the duplication will be short-lived, since further patches will remove the latter completely. Signed-off-by:
-
Vladimir Oltean authored
IPv6 header information is not currently part of the entropy source for the 4-bit aggregation code used for LAG offload, even though it could be. The hardware reference manual says about these fields: ANA::AGGR_CFG.AC_IP6_TCPUDP_PORT_ENA Use IPv6 TCP/UDP port when calculating aggregation code. Configure identically for all ports. Recommended value is 1. ANA::AGGR_CFG.AC_IP6_FLOW_LBL_ENA Use IPv6 flow label when calculating AC. Configure identically for all ports. Recommended value is 1. Integration with the xmit_hash_policy of the bonding interface is TBD. Signed-off-by:
-
Vladimir Oltean authored
Since switchdev/DSA exposes network interfaces that fulfill many of the same user space expectations that dedicated NICs do, it makes sense to not deny bonding interfaces with a bonding policy that we cannot offload, but instead allow the bonding driver to select the egress interface in software. Signed-off-by:
-
Vladimir Oltean authored
Make ocelot's net device event handler more streamlined by structuring it in a similar way with others. The inspiration here was dsa_slave_netdevice_event. Signed-off-by:
-
Vladimir Oltean authored
ocelot_netdevice_port_event treats a single event, NETDEV_CHANGEUPPER. So we can remove the check for the type of event, and rename the function to be more suggestive, since there already is a function with a very similar name of ocelot_netdevice_event. Signed-off-by:
-
Jakub Kicinski authored
Vladimir Oltean says: ==================== Automatically manage DSA master interface state This patch series adds code that makes DSA open the master interface automatically whenever one user interface gets opened, either by the user, or by various networking subsystems: netconsole, nfsroot. With that in place, we can remove some of the places in the network stack where DSA-specific code was sprinkled. ==================== Link: https://lore.kernel.org/r/20210205133713.4172846-1-vladimir.oltean@nxp.comSigned-off-by:
-
Vladimir Oltean authored
This reverts commit 728c0208. Since 2015 DSA has gained more integration with the network stack, we can now have the same functionality without explicitly open-coding for it: - It now opens the DSA master netdevice automatically whenever a user netdevice is opened. - The master and switch interfaces are coupled in an upper/lower hierarchy using the netdev adjacency lists. In the nfsroot example below, the interface chosen by autoconfig was swp3, and every interface except that and the DSA master, eth1, was brought down afterwards: [ 8.714215] mscc_felix 0000:00:00.5 swp0 (uninitialized): PHY [0000:00:00.3:10] driver [Microsemi GE VSC8514 SyncE] (irq=POLL) [ 8.978041] mscc_felix 0000:00:00.5 swp1 (uninitialized): PHY [0000:00:00.3:11] driver [Microsemi GE VSC8514 SyncE] (irq=POLL) [ 9.246134] mscc_felix 0000:00:00.5 swp2 (uninitialized): PHY [0000:00:00.3:12] driver [Microsemi GE VSC8514 SyncE] (irq=POLL) [ 9.486203] mscc_felix 0000:00:00.5 swp3 (uninitialized): PHY [0000:00:00.3:13] driver [Microsemi GE VSC8514 SyncE] (irq=POLL) [ 9.512827] mscc_felix 0000:00:00.5: configuring for fixed/internal link mode [ 9.521047] mscc_felix 0000:00:00.5: Link is Up - 2.5Gbps/Full - flow control off [ 9.530382] device eth1 entered promiscuous mode [ 9.535452] DSA: tree 0 setup [ 9.539777] printk: console [netcon0] enabled [ 9.544504] netconsole: network logging started [ 9.555047] fsl_enetc 0000:00:00.2 eth1: configuring for fixed/internal link mode [ 9.562790] fsl_enetc 0000:00:00.2 eth1: Link is Up - 1Gbps/Full - flow control off [ 9.564661] 8021q: adding VLAN 0 to HW filter on device bond0 [ 9.637681] fsl_enetc 0000:00:00.0 eth0: PHY [0000:00:00.0:02] driver [Qualcomm Atheros AR8031/AR8033] (irq=POLL) [ 9.655679] fsl_enetc 0000:00:00.0 eth0: configuring for inband/sgmii link mode [ 9.666611] mscc_felix 0000:00:00.5 swp0: configuring for inband/qsgmii link mode [ 9.676216] 8021q: adding VLAN 0 to HW filter on device swp0 [ 9.682086] mscc_felix 0000:00:00.5 swp1: configuring for inband/qsgmii link mode [ 9.690700] 8021q: adding VLAN 0 to HW filter on device swp1 [ 9.696538] mscc_felix 0000:00:00.5 swp2: configuring for inband/qsgmii link mode [ 9.705131] 8021q: adding VLAN 0 to HW filter on device swp2 [ 9.710964] mscc_felix 0000:00:00.5 swp3: configuring for inband/qsgmii link mode [ 9.719548] 8021q: adding VLAN 0 to HW filter on device swp3 [ 9.747811] Sending DHCP requests .. [ 12.742899] mscc_felix 0000:00:00.5 swp1: Link is Up - 1Gbps/Full - flow control rx/tx [ 12.743828] mscc_felix 0000:00:00.5 swp0: Link is Up - 1Gbps/Full - flow control off [ 12.747062] IPv6: ADDRCONF(NETDEV_CHANGE): swp1: link becomes ready [ 12.755216] fsl_enetc 0000:00:00.0 eth0: Link is Up - 1Gbps/Full - flow control rx/tx [ 12.766603] IPv6: ADDRCONF(NETDEV_CHANGE): swp0: link becomes ready [ 12.783188] mscc_felix 0000:00:00.5 swp2: Link is Up - 1Gbps/Full - flow control rx/tx [ 12.785354] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready [ 12.799535] IPv6: ADDRCONF(NETDEV_CHANGE): swp2: link becomes ready [ 13.803141] mscc_felix 0000:00:00.5 swp3: Link is Up - 1Gbps/Full - flow control rx/tx [ 13.811646] IPv6: ADDRCONF(NETDEV_CHANGE): swp3: link becomes ready [ 15.452018] ., OK [ 15.470336] IP-Config: Got DHCP answer from 10.0.0.1, my address is 10.0.0.39 [ 15.477887] IP-Config: Complete: [ 15.481330] device=swp3, hwaddr=00:04:9f:05:de:0a, ipaddr=10.0.0.39, mask=255.255.255.0, gw=10.0.0.1 [ 15.491846] host=10.0.0.39, domain=(none), nis-domain=(none) [ 15.498429] bootserver=10.0.0.1, rootserver=10.0.0.1, rootpath= [ 15.498481] nameserver0=8.8.8.8 [ 15.627542] fsl_enetc 0000:00:00.0 eth0: Link is Down [ 15.690903] mscc_felix 0000:00:00.5 swp0: Link is Down [ 15.745216] mscc_felix 0000:00:00.5 swp1: Link is Down [ 15.800498] mscc_felix 0000:00:00.5 swp2: Link is Down Signed-off-by:
Andrew Lunn <andrew@lunn.ch> Reviewed-by:
Florian Fainelli <f.fainelli@gmail.com> Signed-off-by:
-
Vladimir Oltean authored
This reverts commit 1532b977. The above commit is good and it works, however it was meant as a bugfix for stable kernels and now we have more self-contained ways in DSA to handle the situation where the DSA master must be brought up. Signed-off-by:
-
Vladimir Oltean authored
This is not fixing any actual bug that I know of, but having a DSA interface that is up even when its lower (master) interface is down is one of those things that just do not sound right. Yes, DSA checks if the master is up before actually bringing the user interface up, but nobody prevents bringing the master interface down immediately afterwards... Then the user ports would attempt dev_queue_xmit on an interface that is down, and wonder what's wrong. This patch prevents that from happening. NETDEV_GOING_DOWN is the notification emitted _before_ the master actually goes down, and we are protected by the rtnl_mutex, so all is well. For those of you reading this because you were doing switch testing such as latency measurements for autonomously forwarded traffic, and you needed a controlled environment with no extra packets sent by the network stack, this patch breaks that, because now the user ports go down too, which may shut down the PHY etc. But please don't do it like that, just do instead: tc qdisc add dev eno2 clsact tc filter add dev eno2 egress flower action drop Tested with two cascaded DSA switches: $ ip link set eno2 down sja1105 spi2.0 sw0p2: Link is Down mscc_felix 0000:00:00.5 swp0: Link is Down fsl_enetc 0000:00:00.2 eno2: Link is Down Signed-off-by:
-
Vladimir Oltean authored
DSA wants the master interface to be open before the user port is due to historical reasons. The promiscuity of interfaces that are down used to have issues, as referenced Lennert Buytenhek in commit df02c6ff ("dsa: fix master interface allmulti/promisc handling"). The bugfix mentioned there, commit b6c40d68 ("net: only invoke dev->change_rx_flags when device is UP"), was basically a "don't do that" approach to working around the promiscuity while down issue. Further work done by Vlad Yasevich in commit d2615bf4 ("net: core: Always propagate flag changes to interfaces") has resolved the underlying issue, and it is strictly up to the DSA and 8021q drivers now, it is no longer mandated by the networking core that the master interface must be up when changing its promiscuity. From DSA's point of view, deciding to error out in dsa_slave_open because the master isn't up is (a) a bad user experience and (b) knocking at an open door. Even if there still was an issue with promiscuity while down, DSA could still just open the master and avoid it. Doing it this way has the additional benefit that user space can now remove DSA-specific workarounds, like systemd-networkd with BindCarrier: https://github.com/systemd/systemd/issues/7478 And we can finally remove one of the 2 bullets in the "Common pitfalls using DSA setups" chapter. Tested with two cascaded DSA switches: $ ip link set sw0p2 up fsl_enetc 0000:00:00.2 eno2: configuring for fixed/internal link mode fsl_enetc 0000:00:00.2 eno2: Link is Up - 1Gbps/Full - flow control rx/tx mscc_felix 0000:00:00.5 swp0: configuring for fixed/sgmii link mode mscc_felix 0000:00:00.5 swp0: Link is Up - 1Gbps/Full - flow control off 8021q: adding VLAN 0 to HW filter on device swp0 sja1105 spi2.0 sw0p2: configuring for phy/rgmii-id link mode IPv6: ADDRCONF(NETDEV_CHANGE): eno2: link becomes ready IPv6: ADDRCONF(NETDEV_CHANGE): swp0: link becomes ready Signed-off-by:
-
Jakub Kicinski authored
Huazhong Tan says: ==================== net: hns3: updates for -next This series adds some code optimizations and compatibility handlings for the HNS3 ethernet driver. change log: V2: refactor #2 as Jukub Kicinski reported and remove the part about RSS size which will not be different in different hw. updates netdev->max_mtu as well in #4 reported by Jakub Kicinski. previous version: V1: https://patchwork.kernel.org/project/netdevbpf/cover/1612269593-18691-1-git-send-email-tanhuazhong@huawei.com/ ==================== Link: https://lore.kernel.org/r/1612513969-9278-1-git-send-email-tanhuazhong@huawei.comSigned-off-by: -
Guangbin Huang authored
The max qset number is a fixed value now and it is defined by a macro. In order to support other value in different kinds of device, it is better to use specification queried from firmware to replace macro. Signed-off-by:
Guangbin Huang <huangguangbin2@huawei.com> Signed-off-by:
Huazhong Tan <tanhuazhong@huawei.com> Signed-off-by:
-
Guangbin Huang authored
In order to add a method to check the specification of max tm rate for debugging, function hns3_dbg_dev_specs() adds this value print. Signed-off-by:
-
Yufeng Mo authored
Since the newer hardware may supports different frame size, so add support to obtain the capability from the firmware instead of the fixed value. Signed-off-by:
Yufeng Mo <moyufeng@huawei.com> Signed-off-by:
-
GuoJia Liao authored
When update the TC info for NIC, there are some differences between PF and VF. Currently, four "vport->vport_id" are used to distinguish PF or VF. So merge them into one to improve readability and maintainability of code. Signed-off-by:
GuoJia Liao <liaoguojia@huawei.com> Signed-off-by:
-
Guangbin Huang authored
As RSS indirection table size may be different in different hardware. Instead of using macro, this value is better to use device specification which querying from firmware. BTW, RSS indirection table should be allocated by the queried size instead the static array. .get_rss_indir_size in struct hnae3_ae_ops is not used now, so remove it as well. Signed-off-by:
-
Jian Shen authored
To improve the compatibility of firmware for driver, help firmware to deal with different api commands, add api capability bits when initialize the command queue. Signed-off-by:
Jian Shen <shenjian15@huawei.com> Signed-off-by:
-
Jakub Kicinski authored
Mat Martineau says: ==================== mptcp: Misc. updates for tests & lock annotation Here are two fixes we've collected in the mptcp tree. Patch 1 refactors a MPTCP selftest script to allow running a subset of the tests. Patch 2 adds some locking & might_sleep assertations. ==================== Link: https://lore.kernel.org/r/20210204232330.202441-1-mathew.j.martineau@linux.intel.comSigned-off-by:
-
Florian Westphal authored
Add a few assertions to make sure functions are called with the needed locks held. Two functions gain might_sleep annotations because they contain conditional calls to functions that sleep. Signed-off-by:
Florian Westphal <fw@strlen.de> Signed-off-by:
Mat Martineau <mathew.j.martineau@linux.intel.com> Signed-off-by:
-
Geliang Tang authored
Since the mptcp_join script is becoming too big, this patch splits it into several smaller chunks, each of them has been defined in a function as a individual test group for several related testcases. Using bash getopts function to parse command line arguments, and invoke each function to do the individual test group. Here are all the arguments: -f subflows_tests -s signal_address_tests -l link_failure_tests -t add_addr_timeout_tests -r remove_tests -a add_tests -6 ipv6_tests -4 v4mapped_tests -b backup_tests -p add_addr_ports_tests -c syncookies_tests -h help Run mptcp_join.sh with no argument will execute all testcases. Signed-off-by:
Geliang Tang <geliangtang@gmail.com> Signed-off-by:
-
Jakub Kicinski authored
Russell King says: ==================== dpaa2: add 1000base-X support This patch series adds 1000base-X support to pcs-lynx and DPAA2, allowing runtime switching between SGMII and 1000base-X. This is a pre-requisit for SFP module support on the SolidRun ComExpress 7. v2: updated with Ioana's r-b's, and comment on backplane support ==================== Link: https://lore.kernel.org/r/20210205103859.GH1463@shell.armlinux.org.ukSigned-off-by:
-
Russell King authored
Add support for backplane link mode, which is, according to discussions with NXP earlier in the year, is a mode where the OS (Linux) is able to manage the PCS and Serdes itself. This commit prepares the ground work for allowing 1G fiber connections to be used with DPAA2 on the SolidRun CEX7 platforms. Signed-off-by:
Russell King <rmk+kernel@armlinux.org.uk> Signed-off-by:
-
Russell King authored
Now that pcs-lynx supports 1000BASE-X, add support for this interface mode to dpaa2-mac. pcs-lynx can be switched at runtime between SGMII and 1000BASE-X mode, so allow dpaa2-mac to switch between these as well. This commit prepares the ground work for allowing 1G fiber connections to be used with DPAA2 on the SolidRun CEX7 platforms. Reviewed-by:
Ioana Ciornei <ioana.ciornei@nxp.com> Signed-off-by:
-
Russell King authored
Add support for 1000BASE-X to pcs-lynx for the LX2160A. This commit prepares the ground work for allowing 1G fiber connections to be used with DPAA2 on the SolidRun CEX7 platforms. Reviewed-by:
-
Emil Renner Berthing authored
This converts the driver to use the new tasklet API introduced in commit 12cc923f ("tasklet: Introduce new initialization API") The new API changes the argument passed to callback functions, but fortunately it is unused so it is straight forward to use DECLARE_TASKLET rather than DECLARE_TASLKLET_OLD. Signed-off-by:
Emil Renner Berthing <kernel@esmil.dk> Link: https://lore.kernel.org/r/20210204173947.92884-1-kernel@esmil.dkSigned-off-by:
-
Jakub Kicinski authored
Kevin Hao says: ==================== net: Avoid the memory waste in some Ethernet drivers In the current implementation of napi_alloc_frag(), it doesn't have any align guarantee for the returned buffer address. We would have to use some ugly workarounds to make sure that we can get a align buffer address for some Ethernet drivers. This patch series tries to introduce some helper functions to make sure that an align buffer is returned. Then we can drop the ugly workarounds and avoid the unnecessary memory waste. ==================== Link: https://lore.kernel.org/r/20210204105638.1584-1-haokexin@gmail.comSigned-off-by:
-
Kevin Hao authored
The napi_alloc_frag_align() will guarantee that a correctly align buffer address is returned. So use this function to simplify the buffer alloc and avoid the unnecessary memory waste. Signed-off-by:
Kevin Hao <haokexin@gmail.com> Reviewed-by:
-
Kevin Hao authored
The napi_alloc_frag_align() will guarantee that a correctly align buffer address is returned. So use this function to simplify the buffer alloc and avoid the unnecessary memory waste. Signed-off-by:
Subbaraya Sundeep <sbhatta@marvell.com> Signed-off-by:
-
Kevin Hao authored
In the current implementation of {netdev,napi}_alloc_frag(), it doesn't have any align guarantee for the returned buffer address, But for some hardwares they do require the DMA buffer to be aligned correctly, so we would have to use some workarounds like below if the buffers allocated by the {netdev,napi}_alloc_frag() are used by these hardwares for DMA. buf = napi_alloc_frag(really_needed_size + align); buf = PTR_ALIGN(buf, align); These codes seems ugly and would waste a lot of memories if the buffers are used in a network driver for the TX/RX. We have added the align support for the page_frag functions, so add the corresponding {netdev,napi}_frag functions. Signed-off-by:Alexander Duyck <alexanderduyck@fb.com> Signed-off-by:
-
Kevin Hao authored
In the current implementation of page_frag_alloc(), it doesn't have any align guarantee for the returned buffer address. But for some hardwares they do require the DMA buffer to be aligned correctly, so we would have to use some workarounds like below if the buffers allocated by the page_frag_alloc() are used by these hardwares for DMA. buf = page_frag_alloc(really_needed_size + align); buf = PTR_ALIGN(buf, align); These codes seems ugly and would waste a lot of memories if the buffers are used in a network driver for the TX/RX. So introduce page_frag_alloc_align() to make sure that an aligned buffer address is returned. Signed-off-by:Vlastimil Babka <vbabka@suse.cz> Reviewed-by:
-
