- 30 Aug, 2019 26 commits
-
-
Thomas Bogendoerfer authored
Move common code for rx buffer setup into ioc3_alloc_skb and deal with allocation failures. Also clean up allocation size calculation. Signed-off-by:
Thomas Bogendoerfer <tbogendoerfer@suse.de> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Thomas Bogendoerfer authored
Do tx ring cleaning and freeing of rx buffers, when chip is shutdown and allocate buffers before bringing chip up. Signed-off-by:
-
Thomas Bogendoerfer authored
ioc3_init did everything from reset to init rings to starting the chip. This change move out chip start into a new function as preparation for easier handling of receive buffer allocation failures. Signed-off-by:
-
Thomas Bogendoerfer authored
After allocation of descriptor memory is now done once in probe handling of tx ring is completely done by ioc3_clean_tx_ring. So we remove the remaining tx ring actions out of ioc3_alloc_rings and ioc3_free_rings and rename it to ioc3_[alloc|free]_rx_bufs to better describe what they are doing. Signed-off-by:
-
Thomas Bogendoerfer authored
Move clearing of the descriptor valid bit into ioc3_alloc_rings. This makes ioc3_clean_rx_ring obsolete. Signed-off-by:
-
Thomas Bogendoerfer authored
Memory for descriptor rings are allocated/freed, when interface is brought up/down. Since the size of the rings is not changeable by hardware, we now allocate rings now during probe and free it, when device is removed. Signed-off-by:
-
Thomas Bogendoerfer authored
Descriptor ring sizes of the IOC3 are more or less fixed size. To make clearer where there is a relation to ring sizes use defines. Reviewed-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Signed-off-by:
-
Thomas Bogendoerfer authored
Before massaging the driver further fix oddities found by checkpatch like - wrong indention - comment formatting - use of printk instead or netdev_xxx/pr_xxx Signed-off-by:
-
Thomas Bogendoerfer authored
Break up the big ioc3 register struct into functional pieces to make use in sub-function drivers more straightforward. And while doing that get rid of all volatile access by using readX/writeX. Signed-off-by:
-
Thomas Bogendoerfer authored
Removed not needed disabling of ethernet interrupts in IP27 platform code. Acked-by:
Paul Burton <paul.burton@mips.com> Signed-off-by:
-
Colin Ian King authored
The subtraction of the two pointers is automatically scaled by the size of the size of the object the pointers point to, so the division by sizeof(*i2400m->barker) is incorrect. This has been broken since day one of the driver and is only debug, so remove the debug completely. Also move && in condition to clean up a checkpatch warning. Addresses-Coverity: ("Extra sizeof expression") Signed-off-by:Colin Ian King <colin.king@canonical.com> Reviewed-by:
Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by:
-
Gustavo A. R. Silva authored
One of the more common cases of allocation size calculations is finding the size of a structure that has a zero-sized array at the end, along with memory for some number of elements for that array. For example: struct spider_net_card { ... struct spider_net_descr darray[0]; }; Make use of the struct_size() helper instead of an open-coded version in order to avoid any potential type mistakes. So, replace the following form: sizeof(struct spider_net_card) + (tx_descriptors + rx_descriptors) * sizeof(struct spider_net_descr) with: struct_size(card, darray, tx_descriptors + rx_descriptors) Notice that, in this case, variable alloc_size is not necessary, hence it is removed. Building: allmodconfig powerpc. This code was detected with the help of Coccinelle. Signed-off-by:Gustavo A. R. Silva <gustavo@embeddedor.com> Signed-off-by:
-
David S. Miller authored
Heiner Kallweit says: ==================== r8169: add support for RTL8125 This series adds support for the 2.5Gbps chip RTl8125. It can be found on PCIe network cards, and on an increasing number of consumer gaming mainboards. Series is partially based on the r8125 vendor driver. Tested with a Delock 89531 PCIe card against a Netgear GS110MX Multi-Gig switch. Firmware isn't strictly needed, but on some systems there may be compatibility issues w/o firmware. Firmware has been submitted to linux-firmware. v2: - split first patch into 6 smaller ones to facilitate bisecting ==================== Signed-off-by: -
Heiner Kallweit authored
This adds EEE support for RTL8125 based on the vendor driver. Supported is EEE for 100Mbps and 1Gbps. Realtek recommended to not yet enable EEE for 2.5Gbps due to potential compatibility issues. Also ethtool doesn't support yet controlling EEE for 2.5Gbps and 5Gbps. Signed-off-by:
Heiner Kallweit <hkallweit1@gmail.com> Signed-off-by:
-
Heiner Kallweit authored
This patch adds PHY initialization magic copied from the r8125 vendor driver. In addition it supports loading the firmware for chip version RTL_GIGA_MAC_VER_61. Signed-off-by:
-
Heiner Kallweit authored
This adds support for 2.5Gbps chip RTL8125, it's partially based on the r8125 vendor driver. Tested with a Delock 89531 PCIe card against a Netgear GS110MX Multi-Gig switch. Firmware isn't strictly needed, but on some systems there may be compatibility issues w/o firmware. Firmware has been submitted to linux-firmware. Signed-off-by:
-
Heiner Kallweit authored
On RTL8125 this bit is always cleared after send. Therefore check for tx_skb->skb being set what is functionally equivalent. Signed-off-by:
-
Heiner Kallweit authored
RTL8125 uses a different register number for IntrMask. To net have side effects by reading a random register let's use a register that is the same on all supported chip families. Signed-off-by:
-
Heiner Kallweit authored
RTL8125 doesn't support the same coalescing registers, therefore move this initialization to the 8168/6169-specific init. Signed-off-by:
-
Heiner Kallweit authored
For RTL8125 we will have to read the MAC address also from another register range, therefore create a small helper. Signed-off-by:
-
Heiner Kallweit authored
Extend helper rtl_is_8168evl_up to properly work once we add mac version numbers >51 for RTL8125. Signed-off-by:
-
Heiner Kallweit authored
RTL8125 uses a 32 bit interrupt mask even though only bits in the lower 16 bits are used. Change interrupt mask size to u32 to be prepared and reintroduce helper rtl_get_events. Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linuxDavid S. Miller authored
Saeed Mahameed says: ==================== mlx5-updates-2019-08-22 Misc updates for mlx5e net device driver 1) Maxim and Tariq add the support for LAG TX port affinity distribution When VF LAG is enabled, VFs netdevs will round-robin the TX affinity of their tx queues among the different LAG ports. 2) Aya adds the support for ip-in-ip RSS. 3) Marina adds the support for ip-in-ip TX TSO and checksum offloads. 4) Moshe adds a device internal drop counter to mlx5 ethtool stats. ==================== Signed-off-by: -
Vivien Didelot authored
Now mv88e6xxx does not enable its ports at setup itself and let the DSA core handle this, unused ports are disabled without being powered on first. While that is expected, the SERDES powering code was assuming that a port was already set up before powering it down, resulting in freeing an unused IRQ. The patch fixes this assumption. Fixes: b759f528 ("net: dsa: mv88e6xxx: enable SERDES after setup") Signed-off-by:
Vivien Didelot <vivien.didelot@gmail.com> Reviewed-by:
Marek Behún <marek.behun@nic.cz> Tested-by:
Andrew Lunn <andrew@lunn.ch> Signed-off-by:
-
Vivien Didelot authored
This is a follow-up patch for commit 7a3007d2 ("net: dsa: mv88e6xxx: fully support SERDES on Topaz family"). Since .port_set_cmode is only called from mv88e6xxx_port_setup_mac and mv88e6xxx_phylink_mac_config, it is fine to keep this "make writable" code private to the mv88e6341_port_set_cmode implementation, instead of adding yet another operation to the switch info structure. Signed-off-by:
-
Vivien Didelot authored
This is a follow-up patch for commit 17deaf5c ("net: dsa: mv88e6xxx: create serdes_get_lane chip operation"). The .serdes_get_lane implementations access the CMODE of a port, even though it is cached at the moment, it is safer to call them after the mutex is locked, not before. At the same time, check for an eventual error and return IRQ_DONE, instead of blindly ignoring it. Signed-off-by:
-
- 29 Aug, 2019 14 commits
-
-
David S. Miller authored
Huazhong Tan says: ==================== net: hns3: add some cleanups and optimizations This patch-set includes cleanups, optimizations and bugfix for the HNS3 ethernet controller driver. [patch 01/12] adds code optimization for debugfs command "dump reg". [patch 02/12] fixes magic number issues. [patch 03/12] modifies some parameters about hclge_dbg_dump_tm_map(). [patch 04/12] removes some unused parameters. [patch 05/12] refactors some logs to make them more readable. [patch 06/12] makes some resusable codes into functions. [patch 07/12] fixes some type errors. [patch 08/12] reduces the waiting time for per TQP reset. [patch 09/12] implements .process_hw_error for hns3 client. [patch 10/12] adds phy selftest for HNS3 driver. [patch 11/12] adds checking for reset interrupt status when reset fails. [patch 12/12] prevents SSU loopback when running ethtool -t. ==================== Signed-off-by: -
Yufeng Mo authored
The current loopback mode is to add 0x1F to the SMAC address as the DMAC address and enable the promiscuous mode. However, if the VF address is the same as the DMAC address, the loopback test fails. Loopback can be enabled in three places: SSU, MAC, and serdes. By default, SSU loopback is enabled, so if the SMAC and the DMAC are the same, the packets are looped back in the SSU. If SSU loopback is disabled, packets can reach MAC even if SMAC is the same as DMAC. Therefore, this patch disables the SSU loopback before the loopback test. In this way, the SMAC and DMAC can be the same, and the promiscuous mode does not need to be enabled. And this is not valid in version 0x20. This patch also uses a macro to replace 0x1F. Fixes: c39c4d98 ("net: hns3: Add mac loopback selftest support in hns3 driver") Signed-off-by:
Yufeng Mo <moyufeng@huawei.com> Reviewed-by:
Peng Li <lipeng321@huawei.com> Signed-off-by:
Huazhong Tan <tanhuazhong@huawei.com> Signed-off-by:
-
Huazhong Tan authored
Currently, the reset interrupt will be cleared firstly, so when reset fails, if interrupt status register has reset interrupt, it means there is a new coming reset. Fixes: 72e2fb07 ("net: hns3: clear reset interrupt status in hclge_irq_handle()") Signed-off-by:
-
Yufeng Mo authored
Currently, the loopback test supports only mac selftest and serdes selftest. This patch adds phy selftest. Signed-off-by:
-
Weihang Li authored
When hardware or IMP get specified error it may need the client to take some special operations. This patch implements the hns3 client's process_hw_errorx. Signed-off-by:
Weihang Li <liweihang@hisilicon.com> Signed-off-by:
-
Zhongzhu Liu authored
This patch optimizes the waiting time for TQP reset. Signed-off-by:
Zhongzhu Liu <liuzhongzhu@huawei.com> Reviewed-by:
Yunsheng Lin <linyunsheng@huawei.com> Reviewed-by:
-
Guojia Liao authored
This patch fixes some incorrect type in assignment reported by sparse. Those sparse warning as below: - warning : restricted __le16 degrades to integer - warning : cast from restricted __le32 - warning : expected restricted __le32 - warning : cast from restricted __be32 - warning : cast from restricted __be16 - warning : cast to restricted __le16 Signed-off-by:
Guojia Liao <liaoguojia@huawei.com> Reviewed-by:
-
Yonglong Liu authored
In hclge_dcb.c, these pair of codes: hclge_notify_client(hdev, HNAE3_DOWN_CLIENT); hclge_notify_client(hdev, HNAE3_UNINIT_CLIENT); and hclge_notify_client(hdev, HNAE3_INIT_CLIENT); hclge_notify_client(hdev, HNAE3_UP_CLIENT); are called many times, so make them into a function. Signed-off-by:
Yonglong Liu <liuyonglong@huawei.com> Reviewed-by:
-
Yufeng Mo authored
To better identify abnormal conditions, this patch modifies or adds some logs to show driver status more accurately. Signed-off-by:
Guangbin Huang <huangguangbin2@huawei.com> Signed-off-by:
-
Guojia Liao authored
This patch simplifies parameters of some functions by deleting unused parameter. Signed-off-by:
-
Huazhong Tan authored
This patch replaces kstrtouint()'s patameter base with 0 in the hclge_dbg_dump_tm_mac(), which makes it more flexible. Also uses a macro to replace string "dump tm map", since it has been used multiple times. Signed-off-by:
-
Huazhong Tan authored
This patch uses macro to replace some magic number. Signed-off-by:
-
Zhongzhu Liu authored
For making the code more readable, this patch uses a array to keep the information about the dumping register, and then uses it to parse the parameter cmd_buf which passing into hclge_dbg_dump_reg_cmd(). Also replaces parameter "base" of kstrtouint with 0 in the hclge_dbg_dump_reg_common(), which makes it more flexible. Signed-off-by:
-
Ioana Radulescu authored
Starting with firmware version MC10.18.0, we have support for L2 flow control. Asymmetrical configuration (Rx or Tx only) is supported, but not pause frame autonegotioation. Pause frame configuration is done via ethtool. By default, we start with flow control enabled on both Rx and Tx. Changes are propagated to hardware through firmware commands, using two flags (PAUSE, ASYM_PAUSE) to specify Rx and Tx pause configuration, as follows: PAUSE | ASYM_PAUSE | Rx pause | Tx pause ---------------------------------------- 0 | 0 | disabled | disabled 0 | 1 | disabled | enabled 1 | 0 | enabled | enabled 1 | 1 | enabled | disabled The hardware can automatically send pause frames when the number of buffers in the pool goes below a predefined threshold. Due to this, flow control is incompatible with Rx frame queue taildrop (both mechanisms target the case when processing of ingress frames can't keep up with the Rx rate; for large frames, the number of buffers in the pool may never get low enough to trigger pause frames as long as taildrop is enabled). So we set pause frame generation and Rx FQ taildrop as mutually exclusive. Signed-off-by:
Ioana Radulescu <ruxandra.radulescu@nxp.com> Reviewed-by:
-
