- 04 Mar, 2020 4 commits
-
-
David S. Miller authored
Mat Martineau says: ==================== mptcp: Improve DATA_FIN transmission MPTCP's DATA_FIN flag is sent in a DSS option when closing the MPTCP-level connection. This patch series prepares for correct DATA_FIN handling across multiple subflows (where individual subflows may disconnect without closing the entire MPTCP connection) by changing the way the MPTCP-level socket requests a DATA_FIN on a subflow and propagates the necessary data for the TCP option. ==================== Signed-off-by:David S. Miller <davem@davemloft.net>
-
Mat Martineau authored
When a DATA_FIN is sent in a MPTCP DSS option that contains a data mapping, the DATA_FIN consumes one byte of space in the mapping. In this case, the DATA_FIN should only be included in the DSS option if its sequence number aligns with the end of the mapped data. Otherwise the subflow can send an incorrect implicit sequence number for the DATA_FIN, and the DATA_ACK for that sequence number would not close the MPTCP-level connection correctly. Signed-off-by:
Mat Martineau <mathew.j.martineau@linux.intel.com> Signed-off-by:
-
Mat Martineau authored
Instead of reading the MPTCP-level sequence number when sending DATA_FIN, store the data in the subflow so it can be safely accessed when the subflow TCP headers are written to the packet without the MPTCP-level lock held. This also allows the MPTCP-level socket to close individual subflows without closing the MPTCP connection. Signed-off-by:
-
Mat Martineau authored
MPTCP should wait for an active connection or skip sending depending on the connection state, as TCP does. This happens before the possible passthrough to a regular TCP sendmsg because the subflow's socket type (MPTCP or TCP fallback) is not known until the connection is complete. This is also relevent at disconnect time, where data should not be sent in certain MPTCP-level connection states. Signed-off-by:
-
- 03 Mar, 2020 11 commits
-
-
David S. Miller authored
Parav Pandit says: ==================== devlink: Introduce devlink port flavour virtual Currently PCI PF and VF devlink devices register their ports as physical port in non-representors mode. Introduce a new port flavour as virtual so that virtual devices can register 'virtual' flavour to make it more clear to users. An example of one PCI PF and 2 PCI virtual functions, each having one devlink port. $ devlink port show pci/0000:06:00.0/1: type eth netdev ens2f0 flavour physical port 0 pci/0000:06:00.2/1: type eth netdev ens2f2 flavour virtual port 0 pci/0000:06:00.3/1: type eth netdev ens2f3 flavour virtual port 0 Patch summary: Patch-1 Introduces new devlink port flavour 'virtual'. Patch-2 Uses new flavour to register PCI VF virtual ports. ==================== Signed-off-by: -
Parav Pandit authored
Use newly introduce 'virtual' port flavour for devlink port of PCI VF devlink device in non-representors mode. While at it, remove recently introduced empty lines at end of the file. Reviewed-by:
Jiri Pirko <jiri@mellanox.com> Signed-off-by:
Parav Pandit <parav@mellanox.com> Signed-off-by:
-
Parav Pandit authored
Currently mlx5 PCI PF and VF devlink devices register their ports as physical port in non-representors mode. Introduce a new port flavour as virtual so that virtual devices can register 'virtual' flavour to make it more clear to users. An example of one PCI PF and 2 PCI virtual functions, each having one devlink port. $ devlink port show pci/0000:06:00.0/1: type eth netdev ens2f0 flavour physical port 0 pci/0000:06:00.2/1: type eth netdev ens2f2 flavour virtual port 0 pci/0000:06:00.3/1: type eth netdev ens2f3 flavour virtual port 0 Reviewed-by:
Jakub Kicinski <kuba@kernel.org> Signed-off-by:
-
Russell King authored
Using vim to edit the phylink documentation reveals some mistakes due to the "invisible" pythonesque white space indentation that can't be seen with other editors. Fix it. Signed-off-by:
Russell King <rmk+kernel@armlinux.org.uk> Signed-off-by:
-
David S. Miller authored
Paul Blakey says: ==================== act_ct: Software offload of conntrack_in This series adds software offload of connections with an established ct state using the NF flow table offload infrastructure, so once such flows are offloaded, they will not pass through conntrack again, and instead act_ct will restore the conntrack info metadata on the skb to the state it had on the offload event - established. Act_ct maintains an FT instance per ct zone. Flow table entries are created, per ct connection, when connections enter an established state and deleted otherwise. Once an entry is created, the FT assumes ownership of the entry, and manages it's aging. On the datapath, first lookup the skb in the zone's FT before going into conntrack, and if a matching flow is found, restore the conntrack info metadata on the skb, and skip calling conntrack. Note that this patchset is part of the connection tracking offload feature. Hardware offload of connections with an established ct state series will follow this one. Changelog: v1->v2: Removed now unused netfilter patches ==================== Signed-off-by: -
Paul Blakey authored
Offload nf conntrack processing by looking up the 5-tuple in the zone's flow table. The nf conntrack module will process the packets until a connection is in established state. Once in established state, the ct state pointer (nf_conn) will be restored on the skb from a successful ft lookup. Signed-off-by:
Paul Blakey <paulb@mellanox.com> Acked-by:
-
Paul Blakey authored
Add a ft entry when connections enter an established state and delete the connections when they leave the established state. The flow table assumes ownership of the connection. In the following patch act_ct will lookup the ct state from the FT. In future patches, drivers will register for callbacks for ft add/del events and will be able to use the information to offload the connections. Note that connection aging is managed by the FT. Signed-off-by:
-
Paul Blakey authored
Use the NF flow tables infrastructure for CT offload. Create a nf flow table per zone. Next patches will add FT entries to this table, and do the software offload. Signed-off-by:
-
Colin Ian King authored
There is a spelling mistake in a dev_warn message. Fix it. Signed-off-by:
Colin Ian King <colin.king@canonical.com> Signed-off-by:
-
Oleksij Rempel authored
Validate 100baseT1_Full to make this driver work with TJA1102 PHY. Signed-off-by:
Oleksij Rempel <o.rempel@pengutronix.de> Acked-by:
Vladimir Oltean <olteanv@gmail.com> Signed-off-by:
-
Cambda Zhu authored
The data pointers of ipv6 sysctl are set one by one which is hard to maintain, especially with kconfig. This patch simplifies it by using math to point the per net sysctls into the appropriate struct net, just like what we did for ipv4. Signed-off-by:
Cambda Zhu <cambda@linux.alibaba.com> Reviewed-by:
Eric Dumazet <edumazet@google.com> Signed-off-by:
-
- 02 Mar, 2020 25 commits
-
-
Jesper Dangaard Brouer authored
Adding ethtool stats for when XDP transmitted packets overrun the TX queue. This is recorded separately for XDP_TX and ndo_xdp_xmit. This is an important aid for troubleshooting XDP based setups. It is currently a known weakness and property of XDP that there isn't any push-back or congestion feedback when transmitting frames via XDP. It's easy to realise when redirecting from a higher speed link into a slower speed link, or simply two ingress links into a single egress. The situation can also happen when Ethernet flow control is active. For testing the patch and provoking the situation to occur on my Espressobin board, I configured the TX-queue to be smaller (434) than RX-queue (512) and overload network with large MTU size frames (as a larger frame takes longer to transmit). Hopefully the upcoming XDP TX hook can be extended to provide insight into these TX queue overflows, to allow programmable adaptation strategies. Signed-off-by:
Jesper Dangaard Brouer <brouer@redhat.com> Acked-by:
Lorenzo Bianconi <lorenzo@kernel.org> Signed-off-by:
-
David S. Miller authored
More zero-length array transformations from Gustavo A. R. Silva. Signed-off-by: -
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by:Gustavo A. R. Silva <gustavo@embeddedor.com> Signed-off-by:
-
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: -
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: -
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: -
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: -
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: -
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: -
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: -
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: -
Gustavo A. R. Silva authored
The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732 ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: -
David S. Miller authored
Sunil Goutham says: ==================== net: thunderx: Miscellaneous changes This patchset has changes wrt driver performance optimization, load time optimization. And a change to PCI device regiatration table for timestamp device. ==================== Signed-off-by: -
Prakash Brahmajyosyula authored
Across Cavium's ThunderX and Marvell's OcteonTx2 silicons the PTP timestamping block's PCI device ID and vendor ID have remained same but the HW architecture has changed. Hence added PCI subsystem IDs to the device table to avoid this driver from being probed on OcteonTx2 silicons. Signed-off-by:
Prakash Brahmajyosyula <bprakash@marvell.com> Signed-off-by:
Sunil Goutham <sgoutham@marvell.com> Signed-off-by:
-
Geetha sowjanya authored
Replace msleep() with usleep_range() as internally it uses hrtimers. This will put a cap on maximum wait time. Signed-off-by:
Geetha sowjanya <gakula@marvell.com> Signed-off-by:
-
Sunil Goutham authored
With the current RX RED/DROP levels of 192/184 for CQE_RX, when packet incoming rate is high, LLC is getting polluted resulting in more cache misses and higher latency in packet processing. This slows down the whole process and performance loss. Hence reduced the levels to 224/216 (ie for a CQ size of 1024, Rx pkts will be red dropped or dropped when unused CQE are less than 128/160 respectively) Signed-off-by:
-
David S. Miller authored
Sunil Goutham says: ==================== octeontx2: Flow control support and other misc changes This patch series adds flow control support (802.3 pause frames) and has other changes wrt generic admin function (AF) driver functionality. ==================== Signed-off-by: -
Sunil Goutham authored
Currently on the first check if the operation is still not finished, the poll goes to sleep for 2-5 usecs. But if for some reason (due to other priority stuff like interrupts etc) by the time the poll wakes up the 10ms time is expired then we don't check if operation is finished or not and return failure. This patch modifies poll logic to check HW operation after sleep so that the status is checked atleast twice. Signed-off-by:
-
Sunil Goutham authored
Bus mastering is enabled by firmware, but when this driver is unbinded bus mastering gets disabled by the PCI subsystem which results interrupts not working when driver is reloaded. Hence set bus mastering everytime in probe(). Also - Converted pci_set_dma_mask() and pci_set_consistent_dma_mask() to dma_set_mask_and_coherent(). - Cleared transaction pending bit which gets set during driver unbind due to clearing of bus mastering (ME bit). Signed-off-by:
-
Sunil Goutham authored
Currently there is no way for AF dependent drivers in any domain to check if the AF driver is loaded. This patch sets an ID for RVUM block which will automatically reflects in PF/VFs discovery register which they can check and defer their probe until AF is up. Signed-off-by:
-
Linu Cherian authored
For retrieving info like interface MAC addresses, packet parser key extraction config etc currently a command is sent to firmware and firmware which periodically polls for commands, processes these and returns the info. This is resulting in interface initialization taking lot of time. To optimize this a memory region is shared between firmware and this driver, firmware while booting puts static info like these into that region for driver to read directly without using commands. With this - Logic for retrieving packet parser extraction config via commands is removed and repalced with using the shared 'fwdata' structure. - Now RVU MSIX vector address is also retrieved from this fwdata struct instead of from CSR. Otherwise when kexec/kdump crash kernel loads CSR will have a IOVA setup by primary kernel which impacts RVU PF/VF's interrupts. - Also added a mbox handler for PF/VF interfaces to retrieve their MAC addresses from AF. Signed-off-by:
Linu Cherian <lcherian@marvell.com> Signed-off-by:
Christina Jacob <cjacob@marvell.com> Signed-off-by:
Rakesh Babu <rsaladi2@marvell.com> Signed-off-by:
-
Geetha sowjanya authored
Added mailbox requests to retrieve backpressure IDs from AF and Aura, CQ contexts are configured with these BPIDs. So that when resource levels reach configured thresholds they assert backpressure on the interface which is also mapped to same BPID. Also added support to enable/disable pause frames generation via ethtool. Signed-off-by:
Andrew Lunn <andrew@lunn.ch> Signed-off-by:
-
Geetha sowjanya authored
CGX LMAC, the physical interface can generate pause frames when internal resources asserts backpressure due to exhaustion. This patch configures CGX to generate 802.3 pause frames. Also enabled processing of received pause frames on the line which will assert backpressure on the internal transmit path. Also added mailbox handlers for PF drivers to enable or disable pause frames anytime. Signed-off-by:
-
Geetha sowjanya authored
Each of the interface receive channels can be backpressured by resources upon exhaustion or reaching configured threshold levels. Resources here are receive buffer queues (Auras) and pkt notification descriptor queues (CQs). Resources and interface channels are mapped using backpressure IDs (BPIDs). HW supports upto 512 BPIDs, this patch divides these BPIDs statically across CGX/LBK/SDP interfaces as follows. BPIDs 0 - 191 are mapped to LMAC channels, 16 per LMAC. BPIDs 192 - 255 are mapped to LBK channels. BPIDs 256 - 511 are mapped to SDP channels. Also did the needed basic configuration of BPIDs. Added mbox handlers with which a PF device can request for a BPID which it will use to configure Auras and CQs. Signed-off-by:
-
Heiner Kallweit authored
Add constants for the used interrupts bits. This avoids the magic number for MII_VSC85XX_INT_MASK_MASK. Signed-off-by:
Heiner Kallweit <hkallweit1@gmail.com> Reviewed-by:
-
