aboutsummaryrefslogtreecommitdiffstats
path: root/target
Commit message (Collapse)AuthorAgeFilesLines
* kernel: tcindex classifier has been retiredJohn Audia2023-03-181-1/+0
| | | | | | https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/net/sched?h=v5.15.100&id=7c183dc0af472dec33d2c0786a5e356baa8cad19 Signed-off-by: John Audia <therealgraysky@proton.me>
* x86: set deprecated CONFIG_MICROCODE_OLD_INTERACEJohn Audia2023-03-181-1/+1
| | | | | | | | | | | We use late loading[1] so need to set this option despite upstream adding a kernel taint when this option is set. See discussion in PR#12149 for more details. https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/diff/arch/x86/Kconfig?id=v5.15.100&id2=v5.15.98 1. https://github.com/openwrt/openwrt/blob/master/target/linux/x86/base-files/lib/preinit/02_load_x86_ucode Signed-off-by: John Audia <therealgraysky@proton.me>
* uencrypt: add support for mbedtlsEneas U de Queiroz2023-03-172-3/+3
| | | | | | | | | | | | | | This commit includes some additional changes: - better handling of iv and keys in openssl/wolfssl variants - fix compiler warnings and whitespace - build all 3 variants as separate packages - adjust the new package name in targets' DEVICE_PACKAGES - remove PKG_FLAGS:=nonshared [Beeline SmartBox Flash - OK] Tested-by: Mikhail Zhilkin <csharper2005@gmail.com> [after test: replaced a hardcoded IV size of 16 by cipher_info->iv_size] Signed-off-by: Eneas U de Queiroz <cotequeiroz@gmail.com>
* kernel: update v6.3 of patchesRafał Miłecki2023-03-1712-108/+164
| | | | | | | 1. Use final (accepted) patches 2. Add two more required for v6.4 nvmem stuff Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
* ipq40xx: update Crisis Innovation Lab MeshPoint.One BDFMantas Pucka2023-03-171-1/+1
| | | | | | | | | | Make sure it uses updated Jalapeno BDF inherited from Device/8dev_jalapeno-common Fixes: 146eb4925c7f ("ipq40xx: add support for Crisis Innovation Lab MeshPoint.One") Signed-off-by: Mantas Pucka <mantas@8devices.com> [ fix Fixes tag to correct format and fix commit title ] Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
* ipq40xx: set specific BDF file for 8devices Habanero/JalapenoMantas Pucka2023-03-171-0/+2
| | | | | | | | | Set specific BDF file for 8devices Habanero/Jalapeno in ipq40xx generic.mk Signed-off-by: Mantas Pucka <mantas@8devices.com> [ split ipq40xx changes in separate commit ] Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
* layerscape: switch to Kernel 5.15 by defaultPaul Spooren2023-03-171-2/+3
| | | | | | | Getting ready for the next release. Tested-by: Pawel Dembicki <paweldembicki@gmail.com> Signed-off-by: Paul Spooren <mail@aparcar.org>
* bcm4908: add HVC workaround for booting kernel 5.15Rafał Miłecki2023-03-161-0/+30
| | | | Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
* bcm4908: include usbport triggerRafał Miłecki2023-03-161-1/+2
| | | | Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
* bcm4908: backport v6.4 pending DTS changesRafał Miłecki2023-03-1618-2/+1582
| | | | Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
* bmips: tweak Kconfig optionsÁlvaro Fernández Rojas2023-03-163-16/+49
| | | | | | Add patch to reduce and improve selected bmips kernel config options. Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
* bmips: irq-bcm6345-l1: register memory regionÁlvaro Fernández Rojas2023-03-161-0/+38
| | | | | | | Add new patch sent upstream for requesting the memory region in the bcm6345-l1 interrupt controller. Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
* bmips: update patches sent upstreamÁlvaro Fernández Rojas2023-03-162-20/+42
| | | | | | Update bmips patches with the latest version sent upstream Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
* ramips: add support for TP-Link Archer AX23 v1David Bauer2023-03-154-0/+223
| | | | | | | | | | | | | | | | | | | | | | | | | Hardware -------- CPU: MediaTek MT7621 DAT RAM: 128MB DDR3 (integrated) FLASH: 16MB SPI-NOR () WiFi: MediaTek MT7905 + MT7975 (2.4 / 5 DBDC) 802.11ax SERIAL: 115200 8N1 LEDs - (3V3 - GND - RX - TX) - ETH ports Installation ------------ Upload the factory image using the Web-UI. Web-Recovery ------------ The router supports a HTTP recovery mode by holding the reset-button when powering on. The interface is reachable at 192.168.0.1 and supports installation using the factory image. Signed-off-by: David Bauer <mail@david-bauer.net>
* imagebuilder: allow to specific ROOTFS_PARTSIZEPaul Spooren2023-03-141-1/+3
| | | | | | | | | | | Setting this options modifies the rootfs size of created images. When installing a large number of packages it may become necessary to increase the size to have enough storage. This option is only useful for supported devices, i.e. with an attached SD Card or installed on a hard drive. Signed-off-by: Paul Spooren <mail@aparcar.org>
* realtek: switch to Kernel 5.15 by defaultPaul Spooren2023-03-141-2/+1
| | | | | | | | | Getting ready for the next release. Tested-by: Sander Vanheule <sander@svanheule.net> Tested-by: Olliver Schinagl <oliver@schinagl.nl> Tested-by: INAGAKI Hiroshi <musashino.open@gmail.com> Signed-off-by: Paul Spooren <mail@aparcar.org>
* ath79: fix user LED glow on Mikrotik 911 Lite boardsLech Perczak2023-03-121-1/+1
| | | | | | | | | | GPIO3, to which the user LED is connected on RB911-Lite boards seems to still sink current, even when driven high. Enabling open drain for this pin fixes this behaviour and gets rid of the glow when LED is set to off, so enable it. Fixes: 43c7132bf8e7 ("ath79: add support for MikroTik RouterBOARD 911 Lite2/Lite5") Signed-off-by: Lech Perczak <lech.perczak@gmail.com>
* ath79: refactor devolo WiFi pro image definitionsDavid Bauer2023-03-101-18/+11
| | | | | | | | Reuse common parts for the devolo WiFi pro series. The series is discontinued and we support all existing devices, so changes due to new revisions or models are highly unlikely Signed-off-by: David Bauer <mail@david-bauer.net>
* bmips: dts: fix warningsÁlvaro Fernández Rojas2023-03-097-2/+22
| | | | | | | | - Add missing interrupt-parent to ext_intc. - Add missing #address-cells and #size-cells to gpio_cntl. - Use proper gpio handle for gpios (ar-5315u). Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
* mvebu: add support for Fortinet FortiGate 50EINAGAKI Hiroshi2023-03-086-0/+588
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Fortinet FortiGate 50E (FG-50E) is a UTM, based on Armada 385 (88F6820). Specification: - SoC : Marvell Armada 385 88F6820 - RAM : DDR3 2 GiB (4x Micron MT41K512M8DA-107, "D9SGQ") - Flash : SPI-NOR 128 MiB (Macronix MX66L1G45GMI-10G) - Ethernet : 7x 10/100/1000 Mbps - LAN 1-5 : Marvell 88E6176 - WAN 1, 2 : Marvell 88E1512 (2x) - LEDs/Keys : 18x/1x - UART : "CONSOLE" port (RJ-45, RS-232C level) - port : ttyS0 - settings : 9600bps 8n1 - assignment : 1:NC , 2:NC , 3:TXD, 4:GND, 5:GND, 6:RXD, 7:NC , 8:NC - note : compatible with Cisco console cable - HW Monitoring: nuvoTon NCT7802Y - Power : 12 VDC, 2 A - plug : Molex 5557-02R Flash instruction using initramfs image: 1. Power on FG-50E and interrupt to show bootmenu 2. Call "[R]: Review TFTP parameters.", check TFTP parameters and connect computer to "Image download port" in the parameters 3. Prepare TFTP server with the parameters obtained above 4. Rename OpenWrt initramfs image to "image.out" and put to TFTP directory 5. Call "[T]: Initiate TFTP firmware transfer." to download initramfs image from TFTP server 6. Type "r" key when the following message is showed, to boot initramfs image without flashing to spi-nor flash "Save as Default firmware/Backup firmware/Run image without saving:[D/B/R]?" 7. On initramfs image, backup mtd if needed minimum: - "firmware-info" - "kernel" - "rootfs" 7. On initramfs image, upload sysupgrade image to the device and perform sysupgrade 8. Wait ~200 seconds to complete flashing and rebooting. If the device is booted with stock firmware, login to bootmenu and call "[B]: Boot with backup firmware and set as default." to set the first OS image as default and boot it. Notes: - All "SPEED" LEDs(Green/Amber) of LAN and 1000M "SPEED" LEDs(Green) of WAN1/2 are connected to GPIO expander. There is no way to indicate link speed of networking device on Linux Kernel/OpenWrt, so those LEDs cannot be handled like stock firmware. On OpenWrt, use netdev(link) trigger instead. - Both colors of Bi-color LEDs on the front panel cannot be turned on at the same time. - "PWR" and "Logo" LEDs are connected to power source directly. - The following partitions are added for OpenWrt. These partitions are contained in "uboot" partition (0x0-0x1fffff) on stock firmware. - "firmware-info" - "dtb" - "u-boot-env" - "board-info" Image header for bootmenu tftp: 0x0 - 0xf : ? 0x10 - 0x2f : Image Name 0x30 - 0x17f: ? 0x180 - 0x183: Kernel Offset* 0x184 - 0x187: Kernel Length* 0x188 - 0x18b: RootFS Offset (ext2)* 0x18c - 0x18f: RootFS Length (ext2)* 0x190 - 0x193: DTB Offset 0x194 - 0x197: DTB Length 0x198 - 0x19b: Data Offset (jffs2) 0x19c - 0x19f: Data Length (jffs2) 0x1a0 - 0x1ff: ? *: required for initramfs image MAC addresses: (eth0): 70:4C:A5:xx:xx:7C (board-info, 0xd880 (hex)) WAN 1 : 70:4C:A5:xx:xx:7D WAN 2 : 70:4C:A5:xx:xx:7E LAN 1 : 70:4C:A5:xx:xx:7F LAN 2 : 70:4C:A5:xx:xx:80 LAN 3 : 70:4C:A5:xx:xx:81 LAN 4 : 70:4C:A5:xx:xx:82 LAN 5 : 70:4C:A5:xx:xx:83 Signed-off-by: INAGAKI Hiroshi <musashino.open@gmail.com>
* ramips: add support for SNR-CPE-ME2-SFPAleksey Nasibulin2023-03-083-0/+235
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | SNR-CPE-ME2-SFP is a wireless router with SFP cage manufactured by SNR/NAG company. Specification: - SoC: MediaTek MT7621A - CPU: 880MHz - Flash: 16 MB (GD25Q127CSIG) - RAM: 256 MB - WLAN: 2.4 GHz, 5 GHz (MediaTek MT7615DN) - Ethernet: 4x 10/100/1000 Mbps - SFP cage (using RTL8211FS-CG) - USB 3.0 port - Power: 12 VDC, 2 A Flash instruction via TFTP: 1. Boot SNR-CPE-ME2 to recovery mode (press reset button and power on device, hold button for ~10 seconds) 2. Send firmware via TFTP client: TFTP Server address: 192.168.1.1 TFTP Client address: 192.168.1.131 3. Wait ~120 seconds to complete flashing 4. Do sysupgrade using web-interface MAC Addresses(stock) -------------------- +----------+------------------+-------------------+ | use | address | example | +----------+------------------+-------------------+ | Device | label | 6A:C4:DD:xx:xx:28 | | Ethernet | + 1 | 6A:C4:DD:xx:xx:29 | | 2g | + 2 | 6A:C4:DD:xx:xx:2A | | 5g | + 3 | 6A:C4:DD:xx:xx:2B | +----------+------------------+-------------------+ Notes: - Reading sfp eeprom is not supported [1] (driver issue). Stock image has the same situation. References: 1. https://forum.openwrt.org/t/mt7621-and-reading-sfp-eeprom/152249 Signed-off-by: Aleksey Nasibulin <alealexpro100@ya.ru>
* ipq40xx: convert Netgear EX6100/EX6150 v2 to DSADavid Bauer2023-03-083-5/+19
| | | | Signed-off-by: David Bauer <mail@david-bauer.net>
* kernel: Fix perf build with binutils 2.40Hauke Mehrtens2023-03-075-0/+550
| | | | | | | | | | Backport patches from kernel 6.0 which are fixing building of perf with binutils 2.40. perf with kernel 5.10 is also not building but the backporting is more complicated and only a few targets are still using kernel 5.10. Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
* mpc85xx: add support for Watchguard Firebox T10David Bauer2023-03-0711-1/+437
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Hardware -------- SoC: Freescale P1010 RAM: 512MB FLASH: 1 MB SPI-NOR 512 MB NAND ETH: 3x Gigabite Ethernet (Atheros AR8033) SERIAL: Cisco RJ-45 (115200 8N1) RTC: Battery-Backed RTC (I2C) Installation ------------ 1. Patch U-Boot by dumping the content of the SPI-Flash using a SPI programmer. The SHA1 hash for the U-Boot password is currently unknown. A tool for patching U-Boot is available at https://github.com/blocktrron/t10-uboot-patcher/ You can also patch the unknown password yourself. The SHA1 hash is E597301A1D89FF3F6D318DBF4DBA0A5ABC5ECBEA 2. Interrupt the bootmenu by pressing CTRL+C. A password prompt appears. The patched password is '1234' (without quotation marks) 3. Download the OpenWrt initramfs image. Copy it to a TFTP server reachable at 10.0.1.13/24 and rename it to uImage. 4. Connect the TFTP server to ethernet port 0 of the Watchguard T10. 5. Download and boot the initramfs image by entering "tftpboot; bootm;" in U-Boot. 6. After OpenWrt booted, create a UBI volume on the old data partition. The "ubi" mtd partition should be mtd7, check this using $ cat /proc/mtd Create a UBI partition by executing $ ubiformat /dev/mtd7 -y 7. Increase the loadable kernel-size of U-Boot by executing $ fw_setenv SysAKernSize 800000 8. Transfer the OpenWrt sysupgrade image to the Watchguard T10 using scp. Install the image by using sysupgrade: $ sysupgrade -n <path-to-sysupgrade> Note: The LAN ports of the T10 are 1 & 2 while 0 is WAN. You might have to change the ethernet-port. 9. OpenWrt should now boot from the internal NAND. Enjoy. Signed-off-by: David Bauer <mail@david-bauer.net>
* at91: switch to Kernel 5.15 by defaultPaul Spooren2023-03-071-2/+1
| | | | | | | | | | | Getting ready for the next release. Claudiu said: > I tested v5.15 on all targets I have access to previously, when > updating OpenWrt kernel for v5.15 and when preparing this PR. (#11918) Signed-off-by: Paul Spooren <mail@aparcar.org> Reviewed-by: Claudiu Beznea <claudiu.beznea@microchip.com>
* kernel: bump 5.10 to 5.10.172John Audia2023-03-072-99/+1
| | | | | | | | | | | Removed upstreamed: backport-5.10/804-0001-net-Remove-WARN_ON_ONCE-sk-sk_forward_alloc-from-sk_.patch[1] All other patches automatically rebased. 1. https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v5.10.171&id=3e4bbd1f38a8d35bd2d3aaffdb5f6ada546b669a Signed-off-by: John Audia <therealgraysky@proton.me>
* kernel: bump 5.10 to 5.10.169John Audia2023-03-0721-158/+164
| | | | | | | | | | | | | | | | | | | | Manually rebased: backport-5.10/811-v6.1-0001-nvmem-core-Fix-memleak-in-nvmem_register.patch Removed upstreamed: backport-5.10/811-v6.1-0003-nvmem-core-add-error-handling-for-dev_set_name.patch[1] patches-5.10/070-net-bgmac-fix-BCM5358-support-by-setting-correct-fla.patch[2] Add fix: target/linux/generic/backport-5.10/804-0001-net-Remove-WARN_ON_ONCE-sk-sk_forward_alloc-from-sk_.patch[3] All other patches automatically rebased. 1. https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v5.10.169&id=a19a0f67dbb89ad2bfc466f2003841acba645884 2. https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?h=v5.10.169&id=a5c51e0c3202820192db3f3809e072f3ca2b1177 3. https://lore.kernel.org/stable/20230227211548.13923-1-kuniyu@amazon.com Signed-off-by: John Audia <therealgraysky@proton.me>
* bcm47xx: Refresh patchesHauke Mehrtens2023-03-054-10/+10
| | | | | | | | The kernel patches did not apply cleanly any more, refresh them automatically. Fixes: 26bc8f68767e ("generic: MIPS: Add barriers between dcache & icache flushes") Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
* sdk: expose PATENTED an NLS build optionsTomasz Maciej Nowak2023-03-051-0/+17
| | | | | | | | | Some packages offer functionalities guarded by these options and it'll be impossible to reach them without changing Config-build.in. So allow to toggle these in more friendly way, by exposing them in configuration menu. Signed-off-by: Tomasz Maciej Nowak <tmn505@gmail.com>
* mediatek: backport more WED upstream fixesFelix Fietkau2023-03-052-0/+119
| | | | | | Fixes hang if the WO firmware can't be loaded Signed-off-by: Felix Fietkau <nbd@nbd.name>
* ipq806x: Add buttons to OnHubAlan Luck2023-03-051-0/+28
| | | | | | | | | | | These are the factory reset button (external) and "developer mode" button (hidden inside the case (ASUS) or under a screw in the base (TP-Link)) found on the TP-Link and ASUS OnHub devices. Signed-off-by: Alan Luck <luckyhome2008@gmail.com> [Brian: add description; factor out for both ASUS and TP-Link; use existing pinmux definitions; add keycode for dev button] Signed-off-by: Brian Norris <computersforpeace@gmail.com>
* Revert "bmips: add missing switch cpu label"Álvaro Fernández Rojas2023-03-055-5/+0
| | | | | | | | This reverts commit dc0de05e10a9c6e1ef66ca2934e8a0ff0d488943. As pointed out by @BKPepe and @arinc9 this was removed by 9df035b since it isn't needed. Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
* mediatek: add missing DTS entries for WED on MT7986Felix Fietkau2023-03-051-0/+65
| | | | | | Needed to support WLAN offloading Signed-off-by: Felix Fietkau <nbd@nbd.name>
* kernel: backport upstream mtk_eth_soc fixesFelix Fietkau2023-03-058-47/+296
| | | | Signed-off-by: Felix Fietkau <nbd@nbd.name>
* kernel: move accepted mtk_eth_soc patches to generic-backportFelix Fietkau2023-03-0520-52/+52
| | | | Signed-off-by: Felix Fietkau <nbd@nbd.name>
* bmips: add missing switch cpu labelÁlvaro Fernández Rojas2023-03-045-0/+5
| | | | Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
* bcm63xx: kernel: power cycle the bcm6358 USB PLLDaniel González Cabanelas2023-03-044-6/+53
| | | | | | | | | | | | | Some BCM6358 based boards may detect USB2.0 high speed devices as USB1.1 full speed. This is an old well known bug, but nobody cared about it. It is quite random and hard to track. With the latest versions of Openwrt, one user confirmed that the bug is still there (tested router: HG556a). Power cycle the USB PLL to fix it. Signed-off-by: Daniel González Cabanelas <dgcbueu@gmail.com>
* bcm63xx: kernel: fix up bcm63268 roboswitch gpio registersDaniel González Cabanelas2023-03-0412-44/+77
| | | | | | | | | | Some BCM63268 bootloaders may leave gpio registers, related to the roboswitch, disabled before loading the OpenWrt firmware. As result of this the switch won't work. These registers, if not enabled, probably avoid forwarding packets. Signed-off-by: Daniel González Cabanelas <dgcbueu@gmail.com>
* bcm63xx: drop kernel 5.10 supportÁlvaro Fernández Rojas2023-03-04139-21622/+0
| | | | | | Now that the target is on kernel 5.15 this is no longer needed. Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
* bcm63xx: switch to Kernel 5.15 by defaultPaul Spooren2023-03-041-2/+1
| | | | | | Getting ready for the next release. Signed-off-by: Paul Spooren <mail@aparcar.org>
* bmips: check for DMA error when refilling RxSieng-Piaw Liew2023-03-041-8/+20
| | | | | | | dma_mapping_error() should be called as dma_map_single() could fail and return error. Signed-off-by: Sieng-Piaw Liew <liew.s.piaw@gmail.com>
* bmips: fix Tx cleanup when NAPI poll budget is zeroSieng-Piaw Liew2023-03-041-4/+5
| | | | | | | | | | | | NAPI poll() function may be passed a budget value of zero, i.e. during netpoll, which isn't NAPI context. Therefore, napi_consume_skb() must be given budget value instead of !force to truly discern netpoll-like scenarios. https://lore.kernel.org/netdev/20220707141056.2644-1-liew.s.piaw@gmail.com/t/#m470f5c20225e76fb08c44d6cfa2f1b739ffaaea4 Signed-off-by: Sieng-Piaw Liew <liew.s.piaw@gmail.com> [improve code format] Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
* bmips: add BQL supportSieng Piaw Liew2023-03-041-0/+8
| | | | | | Add Byte Queue Limits support to reduce/remove bufferbloat. Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com>
* bmips: batch process eth_type_trans() in rx pathSieng Piaw Liew2023-03-041-2/+3
| | | | | | Improve cache efficiency by running eth_type_trans() in its own loop. Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com>
* bmips: batch process rx pathSieng Piaw Liew2023-03-041-1/+5
| | | | | | | Use netif_receive_skb_list() to batch process rx skb. This improves cache efficiency. Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com>
* bmips: improve rx loopSieng Piaw Liew2023-03-041-2/+3
| | | | | | | | | Use existing rx processed count to track against budget, thereby making budget decrement operation redundant. rx_desc_count can be calculated outside the rx loop, making the loop a bit smaller. Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com>
* bmips: switch to napi_build_skb() to reuse skbuff_headsSieng Piaw Liew2023-03-041-2/+2
| | | | | | | | | | napi_build_skb() reuses NAPI skbuff_head cache in order to save some cycles on freeing/allocating skbuff_heads on every new rx or completed tx. Use napi_consume_skb() to feed the cache with skbuff_heads of completed tx so it's never empty. Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com>
* bmips: check NAPI context when refilling rx SKBSieng Piaw Liew2023-03-041-6/+10
| | | | | | | | | | | Check if we're in NAPI context when refilling rx. Normally we're almost always running in NAPI context. Dispatch to napi_alloc_frag() directly instead of relying on netdev_alloc_frag() which does the same but with the overhead of local_bh_disable/enable. Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com> [improve code format] Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
* bmips: convert to build_skb()Sieng Piaw Liew2023-03-041-41/+61
| | | | | | | | | | | | | 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. Performance is slightly increased but the changes allow more potential optimizations. Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com> [improve code format] Signed-off-by: Álvaro Fernández Rojas <noltari@gmail.com>
* bmips: use netdev_alloc_skb() for copybreakSieng Piaw Liew2023-03-041-1/+1
| | | | | | | | | | | bmips is using Broadcom B53 DSA driver which means the ethernet driver must compensate for 6 bytes tags from the internal switch. This means using NET_IP_ALIGN actually misaligns the skb, lowering performance significantly. Therefore napi_alloc_skb() which uses NET_IP_ALIGN is changed to netdev_alloc_skb(). Performance in iperf3 is increased from ~47Mbps to 52Mbps. Signed-off-by: Sieng Piaw Liew <liew.s.piaw@gmail.com>
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-10 02:26:20 +0000