- 05 Sep, 2024 2 commits
-
-
Sean Anderson authored
This NIC doesn't have hardware IRQ coalescing. Under high load, interrupts can adversely affect performance. To mitigate this, enable software IRQ coalescing by default. On my system this increases receive throughput with iperf3 from 853 MBit/sec to 934 MBit/s, decreases interrupts from 69489/sec to 2016/sec, and decreases CPU utilization from 27% (4x Cortex-A53) to 14%. Latency is not affected (as far as I can tell). Signed-off-by:
Sean Anderson <sean.anderson@linux.dev> Reviewed-by:
Simon Horman <horms@kernel.org> Link: https://patch.msgid.link/20240903184912.4151926-1-sean.anderson@linux.devSigned-off-by:
Jakub Kicinski <kuba@kernel.org>
-
Niklas Söderlund authored
When reading registers from the PHY using the SIOCGMIIREG IOCTL any errors returned from either mdiobus_read() or mdiobus_c45_read() are ignored, and parts of the returned error is passed as the register value back to user-space. For example, if mdiobus_c45_read() is used with a bus that do not implement the read_c45() callback -EOPNOTSUPP is returned. This is however directly stored in mii_data->val_out and returned as the registers content. As val_out is a u16 the error code is truncated and returned as a plausible register value. Fix this by first checking the return value for errors before returning it as the register content. Before this patch, # phytool read eth0/0:1/0 0xffa1 After this change, $ phytool read eth0/0:1/0 error: phy_read (-95) Signed-off-by:Niklas Söderlund <niklas.soderlund+renesas@ragnatech.se> Reviewed-by:
Florian Fainelli <florian.fainelli@broadcom.com> Reviewed-by:
Andrew Lunn <andrew@lunn.ch> Reviewed-by:
Yoshihiro Shimoda <yoshihiro.shimoda.uh@renesas.com> Tested-by:
Geert Uytterhoeven <geert+renesas@glider.be> Link: https://patch.msgid.link/20240903171536.628930-1-niklas.soderlund+renesas@ragnatech.seSigned-off-by:
-
- 04 Sep, 2024 28 commits
-
-
Li Zetao authored
Since the debugfs_create_dir() never returns a null pointer, checking the return value for a null pointer is redundant, and using IS_ERR is safe enough. Signed-off-by:
Li Zetao <lizetao1@huawei.com> Link: https://patch.msgid.link/20240903143343.2004652-1-lizetao1@huawei.comSigned-off-by:
-
Li Zetao authored
Since the debugfs_create_dir() never returns a null pointer, checking the return value for a null pointer is redundant, and using IS_ERR is safe enough. Signed-off-by:
-
Jakub Kicinski authored
Ido Schimmel says: ==================== Unmask upper DSCP bits - part 3 tl;dr - This patchset continues to unmask the upper DSCP bits in the IPv4 flow key in preparation for allowing IPv4 FIB rules to match on DSCP. No functional changes are expected. The TOS field in the IPv4 flow key ('flowi4_tos') is used during FIB lookup to match against the TOS selector in FIB rules and routes. It is currently impossible for user space to configure FIB rules that match on the DSCP value as the upper DSCP bits are either masked in the various call sites that initialize the IPv4 flow key or along the path to the FIB core. In preparation for adding a DSCP selector to IPv4 and IPv6 FIB rules, we need to make sure the entire DSCP value is present in the IPv4 flow key. This patchset continues to unmask the upper DSCP bits, but this time in the output route path, specifically in the callers of ip_route_output_ports(). The next patchset (last) will handle the callers of ip_route_output_key(). Split from this patchset to avoid going over the 15 patches limit. No functional changes are expected as commit 1fa3314c ("ipv4: Centralize TOS matching") moved the masking of the upper DSCP bits to the core where 'flowi4_tos' is matched against the TOS selector. ==================== Link: https://patch.msgid.link/20240903135327.2810535-1-idosch@nvidia.comSigned-off-by: -
Ido Schimmel authored
Unmask the upper DSCP bits when calling ip_route_output_ports() so that in the future it could perform the FIB lookup according to the full DSCP value. Signed-off-by:
Ido Schimmel <idosch@nvidia.com> Reviewed-by:
Guillaume Nault <gnault@redhat.com> Reviewed-by:
David Ahern <dsahern@kernel.org> Link: https://patch.msgid.link/20240903135327.2810535-5-idosch@nvidia.comSigned-off-by:
-
Ido Schimmel authored
Unmask the upper DSCP bits when calling ip_route_output_ports() so that in the future it could perform the FIB lookup according to the full DSCP value. Signed-off-by:
-
Ido Schimmel authored
Unmask the upper DSCP bits when calling ip_route_output_ports() so that in the future it could perform the FIB lookup according to the full DSCP value. Signed-off-by:
-
Ido Schimmel authored
The function is passed the full DS field in its 'tos' argument by its two callers. It then masks the upper DSCP bits using RT_TOS() when passing it to ip_route_output_ports(). Unmask the upper DSCP bits when passing 'tos' to ip_route_output_ports() so that in the future it could perform the FIB lookup according to the full DSCP value. Signed-off-by:
-
Chen Ni authored
Replace comma between expressions with semicolons. Using a ',' in place of a ';' can have unintended side effects. Although that is not the case here, it is seems best to use ';' unless ',' is intended. Found by inspection. No functional change intended. Compile tested only. Signed-off-by:
Chen Ni <nichen@iscas.ac.cn> Reviewed-by:
-
Ido Schimmel authored
RT_TOS() from include/uapi/linux/in_route.h is defined using IPTOS_TOS_MASK from include/uapi/linux/ip.h. This is problematic for files such as include/net/ip_fib.h that want to use RT_TOS() as without including both header files kernel compilation fails: In file included from ./include/net/ip_fib.h:25, from ./include/net/route.h:27, from ./include/net/lwtunnel.h:9, from net/core/dst.c:24: ./include/net/ip_fib.h: In function ‘fib_dscp_masked_match’: ./include/uapi/linux/in_route.h:31:32: error: ‘IPTOS_TOS_MASK’ undeclared (first use in this function) 31 | #define RT_TOS(tos) ((tos)&IPTOS_TOS_MASK) | ^~~~~~~~~~~~~~ ./include/net/ip_fib.h:440:45: note: in expansion of macro ‘RT_TOS’ 440 | return dscp == inet_dsfield_to_dscp(RT_TOS(fl4->flowi4_tos)); Therefore, cited commit changed linux/in_route.h to include linux/ip.h. However, as reported by David, this breaks iproute2 compilation due overlapping definitions between linux/ip.h and /usr/include/netinet/ip.h: In file included from ../include/uapi/linux/in_route.h:5, from iproute.c:19: ../include/uapi/linux/ip.h:25:9: warning: "IPTOS_TOS" redefined 25 | #define IPTOS_TOS(tos) ((tos)&IPTOS_TOS_MASK) | ^~~~~~~~~ In file included from iproute.c:17: /usr/include/netinet/ip.h:222:9: note: this is the location of the previous definition 222 | #define IPTOS_TOS(tos) ((tos) & IPTOS_TOS_MASK) Fix by changing include/net/ip_fib.h to include linux/ip.h. Note that usage of RT_TOS() should not spread further in the kernel due to recent work in this area. Fixes: 1fa3314c ("ipv4: Centralize TOS matching") Reported-by: -
James Chapman authored
Commit aa92c1ce ("l2tp: add tunnel/session get_next helpers") uses idr_get_next APIs to iterate over l2tp session IDR lists. Sessions in l2tp_v2_session_idr always have a non-null session->tunnel pointer since l2tp_session_register sets it before inserting the session into the IDR. Therefore the null check on session->tunnel in l2tp_v2_session_get_next is redundant and can be removed. Removing the check avoids a warning from lkp. Reported-by:
kernel test robot <lkp@intel.com> Closes: https://lore.kernel.org/r/202408111407.HtON8jqa-lkp@intel.com/ CC: Dan Carpenter <dan.carpenter@linaro.org> Signed-off-by:
James Chapman <jchapman@katalix.com> Acked-by:
Tom Parkin <tparkin@katalix.com> Reviewed-by:
-
Shradha Gupta authored
The mana_set_channels() function requires detaching the mana driver and reattaching it with changed channel values. During this operation if the system is low on memory, the reattach might fail, causing the network device being down. To avoid this we pre-allocate buffers at the beginning of set operation, to prevent complete network loss Signed-off-by:
Shradha Gupta <shradhagupta@linux.microsoft.com> Reviewed-by:
Gerhard Engleder <gerhard@engleder-embedded.com> Reviewed-by:
Haiyang Zhang <haiyangz@microsoft.com> Link: https://patch.msgid.link/1725248734-21760-1-git-send-email-shradhagupta@linux.microsoft.comSigned-off-by:
-
David S. Miller authored
Mohsin Bashir says: ==================== eth: Add basic ethtool support for fbnic This patch series adds basic ethtool support for fbnic. Specifically, the two patches focus on the following two features respectively: 1: Enable 'ethtool -i <dev>' to provide driver, firmware and bus information. 2: Provide mac group stats. Changes since v2: - Fix v1 reference link - Fix nit --- v2: https://lore.kernel.org/netdev/20240827205904.1944066-2-mohsin.bashr@gmail.com v1: https://lore.kernel.org/netdev/20240822184944.3882360-1-mohsin.bashr@gmail.com ==================== Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Mohsin Bashir authored
Add support for group stats for mac. The fbnic_set_counter help preserve the default values for counters which are not touched by the driver. The 'reset' flag in 'get_eth_mac_stats' allows to choose between resetting the counter to recent most value or fetching the aggregated values of the counter. The 'fbnic_stat_rd64' read 64b stats counters in an atomic fashion using read-read-read approach. This allows to isolate cases where counter is moving too fast making accuracy of the counter questionable. Command: ethtool -S eth0 --groups eth-mac Example Output: eth-mac-FramesTransmittedOK: 421644 eth-mac-FramesReceivedOK: 3849708 eth-mac-FrameCheckSequenceErrors: 0 eth-mac-AlignmentErrors: 0 eth-mac-OctetsTransmittedOK: 64799060 eth-mac-FramesLostDueToIntMACXmitError: 0 eth-mac-OctetsReceivedOK: 5134513531 eth-mac-FramesLostDueToIntMACRcvError: 0 eth-mac-MulticastFramesXmittedOK: 568 eth-mac-BroadcastFramesXmittedOK: 454 eth-mac-MulticastFramesReceivedOK: 276106 eth-mac-BroadcastFramesReceivedOK: 26119 eth-mac-FrameTooLongErrors: 0 Signed-off-by:
Mohsin Bashir <mohsin.bashr@gmail.com> Reviewed-by:
-
Mohsin Bashir authored
Add ethtool ops support and enable 'get_drvinfo' for fbnic. The driver provides firmware version information while the driver name and bus information is provided by ethtool_get_drvinfo(). Signed-off-by:
-
Jason Xing authored
Add the SO_PEEK_OFF selftest for UDP. In this patch, I mainly do three things: 1. rename tcp_so_peek_off.c 2. adjust for UDP protocol 3. add selftests into it Suggested-by:
Jon Maloy <jmaloy@redhat.com> Reviewed-by:
Willem de Bruijn <willemb@google.com> Signed-off-by:
Jason Xing <kernelxing@tencent.com> Signed-off-by:
-
David S. Miller authored
Daniel Machon says: ==================== net: microchip: add FDMA library and use it for Sparx5 This patch series is the first of a 2-part series, that adds a new common FDMA library for Microchip switch chips Sparx5 and lan966x. These chips share the same FDMA engine, and as such will benefit from a common library with a common implementation. This also has the benefit of removing a lot open-coded bookkeeping and duplicate code for the two drivers. Additionally, upstreaming efforts for a third chip, lan969x, will begin in the near future. This chip will use the new library too. In this first series, the FDMA library is introduced and used by the Sparx5 switch driver. ################### # Example of use: # ################### - Initialize the rx and tx fdma structs with values for: number of DCB's, number of DB's, channel ID, DB size (data buffer size), and total size of the requested memory. Also provide two callbacks: nextptr_cb() and dataptr_cb() for getting the nextptr and dataptr. - Allocate memory using fdma_alloc_phys() or fdma_alloc_coherent(). - Initialize the DCB's with fdma_dcb_init(). - Add new DCB's with fdma_dcb_add(). - Free memory with fdma_free_phys() or fdma_free_coherent(). ##################### # Patch breakdown: # ##################### Patch #1: introduces library and selects it for Sparx5. Patch #2: includes the fdma_api.h header and removes old symbols. Patch #3: replaces old rx and tx variables with equivalent ones from the fdma struct. Only the variables that can be changed without breaking traffic is changed in this patch. Patch #4: uses the library for allocation of rx buffers. This requires quite a bit of refactoring in this single patch. Patch #5: uses the library for adding DCB's in the rx path. Patch #6: uses the library for freeing rx buffers. Patch #7: uses the library helpers in the rx path. Patch #8: uses the library for allocation of tx buffers. This requires quite a bit of refactoring in this single patch. Patch #9: uses the library for adding DCB's in the tx path. Patch #10: uses the library helpers in the tx path. Patch #11: ditches the existing linked list for storing buffer addresses, and instead uses offsets into contiguous memory. Patch #12: modifies existing rx and tx functions to be direction independent. ==================== Signed-off-by: -
Daniel Machon authored
These direction specific functions can be ditched in favor of a single function: sparx5_fdma_reload(), which retrieves the channel id from the fdma struct instead. Signed-off-by:
Daniel Machon <daniel.machon@microchip.com> Reviewed-by:
Steen Hegelund <Steen.Hegelund@microchip.com> Reviewed-by:
Jens Emil Schulz Østergaard <jensemil.schulzostergaard@microchip.com> Reviewed-by:
Horatiu Vultur <horatiu.vultur@microchip.com> Signed-off-by:
-
Daniel Machon authored
Currently, the driver uses a linked list for storing the tx buffer addresses. This requires a good amount of extra bookkeeping code. Ditch the linked list in favor of tx buffers being in the same contiguous memory space as the DCB's and the DB's. The FDMA library has a helper for this - so use that. The tx buffer addresses are now retrieved as an offset into the FDMA memory space. Signed-off-by:
-
Daniel Machon authored
The library has the helper fdma_free_phys() for freeing physical FDMA memory. Use it in the exit path. Signed-off-by:
-
Daniel Machon authored
Use the fdma_dcb_add() function to add DCB's in the tx path. This gets rid of the open-coding of nextptr and dataptr handling and leaves it to the library. Also, make sure the fdma indexes are advanced using: fdma_dcb_advance(), so that the correct nextptr and dataptr offsets are retrieved. Signed-off-by:
-
Daniel Machon authored
Use the two functions: fdma_alloc_phys() and fdma_dcb_init() for tx buffer allocation and use the new buffers throughout. In order to replace the old buffers with the new ones, we have to do the following refactoring: - use fdma_alloc_phys() and fdma_dcb_init() - replace the variables: tx->dma, tx->first_entry and tx->curr_entry with the equivalents from the FDMA struct. - replace uses of sparx5_db_hw and sparx5_tx_dcb_hw with fdma_db and fdma_dcb. - add sparx5_fdma_tx_dataptr_cb callback for obtaining the dataptr. - Initialize FDMA struct values. Signed-off-by: -
Daniel Machon authored
The library provides helpers for a number of DCB and DB operations. Use these in the rx path. Signed-off-by:
-
Daniel Machon authored
The library has the helper fdma_free_phys() for freeing physical FDMA memory. Use it in the exit path. Signed-off-by:
-
Daniel Machon authored
Use the fdma_dcb_add() function to add DCB's in the rx path. This gets rid of the open-coding of nextptr and dataptr handling and leaves it to the library. Signed-off-by:
-
Daniel Machon authored
Use the two functions: fdma_alloc_phys() and fdma_dcb_init() for rx buffer allocation and use the new buffers throughout. In order to replace the old buffers with the new ones, we have to do the following refactoring: - use fdma_alloc_phys() and fdma_dcb_init() - replace the variables: rx->dma, rx->dcb_entries and rx->last_entry with the equivalents from the FDMA struct. - replace uses of sparx5_db_hw and sparx5_rx_dcb_hw with fdma_db and fdma_dcb. - add sparx5_fdma_rx_dataptr_cb callback for obtaining the dataptr. - Initialize FDMA struct values. Signed-off-by: -
Daniel Machon authored
Replace the old rx and tx variables: channel_id, FDMA_DCB_MAX, FDMA_RX_DCB_MAX_DBS, FDMA_TX_DCB_MAX_DBS, dcb_index and db_index with the equivalents from the FDMA rx and tx structs. These variables are not entangled in any buffer allocation and can therefore be replaced in advance. Signed-off-by:
-
Daniel Machon authored
Include and use the new FDMA header, which now provides the required masks and bit offsets for operating on the DCB's and DB's. Signed-off-by:
-
Daniel Machon authored
Add new FDMA library for interacting with the FDMA engine on Microchip Sparx5 and lan966x switch chips, in an effort to reduce duplicate code and provide a common set of symbols and functions. Signed-off-by:
-
- 03 Sep, 2024 10 commits
-
-
Yue Haibing authored
Commit 6b7c5b94 ("net: Add be2net driver.") declared be_pci_fnum_get() and be_cmd_reset() but never implemented. And commit 9fa465c0 ("be2net: remove code duplication relating to Lancer reset sequence") removed lancer_test_and_set_rdy_state() but leave declaration. Commit 76a9e08e ("be2net: cleanup wake-on-lan code") left behind be_is_wol_supported() declaration. Commit baaa08d1 ("be2net: do not call be_set/get_fw_log_level() on Skyhawk-R") removed be_get_fw_log_level() but leave declaration. Signed-off-by:
Yue Haibing <yuehaibing@huawei.com> Reviewed-by:
-
Yue Haibing authored
There is no caller and implementation in tree. Signed-off-by:
-
Krzysztof Kozlowski authored
This reverts commit bf4d87f8 because it introduced dev_err_probe() in non-probe path, which is not desired. In general, calling dev_err_probe() after successful probe in case of handling -EPROBE_DEFER error, will set deferred status on the device already probed. This is however not a problem here now, because dev_err_probe() in affected places is used for handling errors from request_firmware(), which does not return -EPROBE_DEFER. Still usage of dev_err_probe() in such case is not correct, because request_firmware() could once return -EPROBE_DEFER. Fixes: bf4d87f8 ("net: alacritech: Switch to use dev_err_probe()") Signed-off-by:
Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> Reviewed-by:
-
Florian Westphal authored
This update allows listing default firewalld ruleset on Fedora 40 via tools/net/ynl/cli.py --spec \ Documentation/netlink/specs/nftables.yaml --dump getrule Default ruleset uses fib, reject and objref expressions which were missing. Other missing expressions can be added later. Improve decoding while at it: - add bitwise, ct and lookup attributes - wire up the quota expression - translate raw verdict codes to a human reable name, e.g. 'code': 4294967293 becomes 'code': 'jump'. v2: forgot fib addrtype in enum list (Donald Hunter) Reviewed-by:Donald Hunter <donald.hunter@gmail.com> Signed-off-by:
Florian Westphal <fw@strlen.de> Link: https://patch.msgid.link/20240902214112.2549-1-fw@strlen.deSigned-off-by:
-
Jakub Kicinski authored
Matthieu Baerts says: ==================== mptcp: MIB counters for MPJ TX + misc improvements Recently, a few issues have been discovered around the creation of additional subflows. Without these counters, it was difficult to point out the reason why some subflows were not created as expected. In patch 3, all error paths from __mptcp_subflow_connect() are covered, except the one related to the 'fully established mode', because it can only happen with the userspace PM, which will propagate the error to the userspace in this case (ENOTCONN). These new counters are also verified in the MPTCP Join selftest in patch 6. While at it, a few other patches are improving the MPTCP path-manager code ... - Patch 1: 'flush' related helpers are renamed to avoid confusions - Patch 2: directly pass known ID and flags to create a new subflow, i/o getting them later by iterating over all endpoints again ... and the MPJoin selftests: - Patch 4: reduce the number of positional parameters - Patch 5: only one line for the 'join' checks, instead of 3 - Patch 7: more explicit check names, instead of sometimes too cryptic ones: rtx, ptx, ftx, ctx, fclzrx, sum - Patch 8: specify client/server instead of 'invert' for some checks not suggesting one specific direction - Patch 9: mute errors of mptcp_connect when ran in the background - Patch 10: simplify checksum_tests by using a for-loop - Patch 11: remove 'define' re-definitions ==================== Link: https://patch.msgid.link/20240902-net-next-mptcp-mib-mpjtx-misc-v1-0-d3e0f3773b90@kernel.orgSigned-off-by: -
Matthieu Baerts (NGI0) authored
'MPTCP_PM_NAME' is defined in 'linux/mptcp_pm.h', included in 'linux/mptcp.h', no need to re-define it. 'MPTCP_PM_EVENTS' is not defined in 'linux/mptcp.h', but 'MPTCP_PM_EV_GRP_NAME' is, with the same value. We can then use the latter, and drop the other one. Reviewed-by:
Geliang Tang <geliang@kernel.org> Signed-off-by:
Matthieu Baerts (NGI0) <matttbe@kernel.org> Link: https://patch.msgid.link/20240902-net-next-mptcp-mib-mpjtx-misc-v1-11-d3e0f3773b90@kernel.orgSigned-off-by:
-
Geliang Tang authored
The four checksum tests are similar, only one line is different. So a for-loop can be used to simplify these tests. Signed-off-by:
Geliang Tang <tanggeliang@kylinos.cn> Reviewed-by:
-
Matthieu Baerts (NGI0) authored
The test is supposed to be killed before the end, which will likely cause "Connection reset by peer" errors. It is confusing, especially because in case of real transfer errors, the test will not be marked as failed. But that's OK, there are many other tests checking that. Reviewed-by:
-
Matthieu Baerts (NGI0) authored
Instead of displaying 'invert' when looking at some events like MP_FAIL, MP_FASTCLOSE, MP_RESET, RM_ADDR, which is a bit vague because they are not traditionnaly sent from one side, the host being checked is now printed. For the ADD_ADDR, only display the host when it is the client sending it, which is more unusual. Also before, the 'invert' message was printed after a few checks, but it was not clear which ones exactly. Reviewed-by:
-
Matthieu Baerts (NGI0) authored
Before, the check names had to be very short. It is no longer the case now that these checks are printed on a dedicated line. Then, it looks better to have more explicit names. Reviewed-by:
-
