- 18 Nov, 2018 9 commits
-
-
Samuel Mendoza-Jonas authored
When the NCSI driver is stopped with ncsi_stop_dev() the channel monitors are stopped and the state set to "inactive". However the channels are still configured and active from the perspective of the network controller. We should suspend each active channel but in the context of ncsi_stop_dev() the transmit queue has been or is about to be stopped so we won't have time to do so. Instead when ncsi_start_dev() is called if the NCSI topology has already been probed then call ncsi_reset_dev() to suspend any channels that were previously active. This resets the network controller to a known state, provides an up to date view of channel link state, and makes sure that mode flags such as NCSI_MODE_TX_ENABLE are properly reset. In addition to ncsi_start_dev() use ncsi_reset_dev() in ncsi-netlink.c to update the channel configuration more cleanly. Signed-off-by:
Samuel Mendoza-Jonas <sam@mendozajonas.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Samuel Mendoza-Jonas authored
The concepts of a channel being 'active' and it having link are slightly muddled in the NCSI driver. Tweak this slightly so that NCSI_CHANNEL_ACTIVE represents a channel that has been configured and enabled, and NCSI_CHANNEL_INACTIVE represents a de-configured channel. This distinction is important because a channel can be 'active' but have its link down; in this case the channel may still need to be configured so that it may receive AEN link-state-change packets. Signed-off-by:
Samuel Mendoza-Jonas <sam@mendozajonas.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Samuel Mendoza-Jonas authored
When a package is deselected all channels of that package cease communication. If there are other channels active on the package of the suspended channel this will disable them as well, so only send a deselect-package command if no other channels are active. Signed-off-by:
Samuel Mendoza-Jonas <sam@mendozajonas.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Samuel Mendoza-Jonas authored
Currently the NCSI driver sends a select-package command to all possible packages simultaneously to discover what packages are available. However at this stage in the probe process the driver does not know if hardware arbitration is available: if it isn't then this process could cause collisions on the RMII bus when packages try to respond. Update the probe loop to probe each package one by one, and once complete check if HWA is universally supported. Signed-off-by:
Samuel Mendoza-Jonas <sam@mendozajonas.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Samuel Mendoza-Jonas authored
NCSI hardware arbitration allows multiple packages to be enabled at once and share the same wiring. If the NCSI driver recognises that HWA is available it unconditionally enables all packages and channels; but that is a configuration decision rather than something required by HWA. Additionally the current implementation will not failover on link events which can cause connectivity to be lost unless the interface is manually bounced. Retain basic HWA support but remove the separate configuration path to enable all channels, leaving this to be handled by a later implementation. Signed-off-by:
Samuel Mendoza-Jonas <sam@mendozajonas.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Arjun Vynipadath authored
The SGE Host Page Size has nothing to do with the actual Host Page Size. It's the SGE's BAR2 Doorbell/GTS Page Size for interpreting the SGE Ingress/Egress Queue per Page values. Firmware reads all of these things and makes all the subsequent changes necessary. The Host Driver uses the SGE Host Page Size in order to properly calculate BAR2 Offsets. Signed-off-by:
Casey Leedom <leedom@chelsio.com> Signed-off-by:
Arjun Vynipadath <arjun@chelsio.com> Signed-off-by:
Ganesh Goudar <ganeshgr@chelsio.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Yousuk Seung authored
Add TCP_NLA_SRTT to SCM_TIMESTAMPING_OPT_STATS that reports the smoothed round trip time in microseconds (tcp_sock.srtt_us >> 3). Signed-off-by:
Yousuk Seung <ysseung@google.com> Signed-off-by:
Eric Dumazet <edumazet@google.com> Acked-by:
Soheil Hassas Yeganeh <soheil@google.com> Acked-by:
Neal Cardwell <ncardwell@google.com> Acked-by:
Yuchung Cheng <ycheng@google.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Stephen Hemminger authored
Trivial spelling errors found by codespell. Signed-off-by:
Stephen Hemminger <stephen@networkplumber.org> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
David S. Miller authored
Instead of constantly playing with the struct initializer syntax trying to make gcc and CLang both happy, just clear it out using memset(). >> drivers/net/tun.c:2503:42: warning: Using plain integer as NULL pointer Reported-by:
kbuild test robot <lkp@intel.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
- 17 Nov, 2018 31 commits
-
-
Jason Wang authored
Thanks to the batched XDP buffs through msg_control. Instead of calling put_page() for each page which involves a atomic operation, let's batch them by record the last page that needs to be freed and its refcnt count and free them in a batch. Testpmd(virtio-user + vhost_net) + XDP_DROP shows 3.8% improvement. Before: 4.71Mpps After : 4.89Mpps Signed-off-by:
Jason Wang <jasowang@redhat.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Jason Wang authored
We do a get_page() which involves a atomic operation. This patch tries to mitigate a per packet atomic operation by maintaining a reference bias which is initially USHRT_MAX. Each time a page is got, instead of calling get_page() we decrease the bias and when we find it's time to use a new page we will decrease the bias at one time through __page_cache_drain_cache(). Testpmd(virtio_user + vhost_net) + XDP_DROP on TAP shows about 1.6% improvement. Before: 4.63Mpps After: 4.71Mpps Signed-off-by:
Jason Wang <jasowang@redhat.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
David S. Miller authored
Jakub Kicinski says: ==================== net: sched: gred: introduce per-virtual queue attributes This series updates the GRED Qdisc. The Qdisc matches nfp offload very well, but before we can offload it there are a number of improvements to make. First few patches add extack messages to the Qdisc and pass extack to netlink validation. Next a new netlink attribute group is added, to allow GRED to be extended more easily. Currently GRED passes C structures as attributes, and even an array of C structs for virtual queue configuration. User space has hard coded the expected length of that array, so adding new fields is not possible. New two-level attribute group is added: [TCA_GRED_VQ_LIST] [TCA_GRED_VQ_ENTRY] [TCA_GRED_VQ_DP] [TCA_GRED_VQ_FLAGS] [TCA_GRED_VQ_STAT_*] [TCA_GRED_VQ_ENTRY] [TCA_GRED_VQ_DP] [TCA_GRED_VQ_FLAGS] [TCA_GRED_VQ_STAT_*] [TCA_GRED_VQ_ENTRY] ... Statistics are dump only. Patch 4 switches the byte counts to be 64 bit, and patch 5 introduces the new stats attributes for dump. Patch 6 switches RED flags to be per-virtual queue, and patch 7 allows them to be dumped and set at virtual queue granularity. ==================== Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Jakub Kicinski authored
Allow users to set and dump RED flags (ECN enabled and harddrop) on per-virtual queue basis. Validation of attributes is split from changes to make sure we won't have to undo previous operations when we find out configuration is invalid. The objective is to allow changing per-Qdisc parameters without overwriting the per-vq configured flags. Old user space will not pass the TCA_GRED_VQ_FLAGS attribute and per-Qdisc flags will always get propagated to the virtual queues. New user space which wants to make use of per-vq flags should set per-Qdisc flags to 0 and then configure per-vq flags as it sees fit. Once per-vq flags are set per-Qdisc flags can't be changed to non-zero. Vice versa - if the per-Qdisc flags are non-zero the TCA_GRED_VQ_FLAGS attribute has to either be omitted or set to the same value as per-Qdisc flags. Update per-Qdisc parameters: per-Qdisc | per-VQ | result 0 | 0 | all vq flags updated 0 | non-0 | error (vq flags in use) non-0 | 0 | -- impossible -- non-0 | non-0 | all vq flags updated Update per-VQ state (flags parameter not specified): no change to flags Update per-VQ state (flags parameter set): per-Qdisc | per-VQ | result 0 | any | per-vq flags updated non-0 | 0 | -- impossible -- non-0 | non-0 | error (per-Qdisc flags in use) Signed-off-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Reviewed-by:
John Hurley <john.hurley@netronome.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Jakub Kicinski authored
Right now ECN marking and HARD drop (the common RED flags) can only be configured for the entire Qdisc. In preparation for per-vq flags store the values in the virtual queue structure. Setting per-vq flags will only be allowed when no flags are set for the entire Qdisc. For the new flags we will also make sure undefined bits are 0. Signed-off-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Reviewed-by:
John Hurley <john.hurley@netronome.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Jakub Kicinski authored
Currently all GRED's virtual queue data is dumped in a single array in a single attribute. This makes it pretty much impossible to add new fields. In order to expose more detailed stats add a new set of attributes. We can now expose the 64 bit value of bytesin and all the mark stats which were not part of the original design. Signed-off-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Reviewed-by:
John Hurley <john.hurley@netronome.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Jakub Kicinski authored
32 bit counters for bytes are not really going to last long in modern world. Make sch_gred count bytes on a 64 bit counter. It will still get truncated during dump but follow up patch will add set of new stat dump attributes. Signed-off-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Reviewed-by:
John Hurley <john.hurley@netronome.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Jakub Kicinski authored
Add extack messages to -EINVAL errors, to help users identify their mistakes. Signed-off-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Reviewed-by:
John Hurley <john.hurley@netronome.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Jakub Kicinski authored
In case netlink wants to provide parsing error pass extack to nla_parse_nested(). Signed-off-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Reviewed-by:
John Hurley <john.hurley@netronome.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Jakub Kicinski authored
We will soon want to add more code to the non-error path, separate it from the error handling flow. Signed-off-by:
Jakub Kicinski <jakub.kicinski@netronome.com> Reviewed-by:
John Hurley <john.hurley@netronome.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Paolo Abeni authored
This covers for proper accounting of encap needed static keys Signed-off-by:
Paolo Abeni <pabeni@redhat.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
YueHaibing authored
Fixes gcc '-Wunused-but-set-variable' warning: drivers/isdn/hisax/hfc_pci.c:277:6: warning: variable ‘total’ set but not used [-Wunused-but-set-variable] It never used since git history start. Signed-off-by:
YueHaibing <yuehaibing@huawei.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Paolo Abeni authored
The commit 60fb9567 ("udp: implement complete book-keeping for encap_needed") introduced a severe misuse of jump label APIs, which syzbot, as reported by Eric, was able to exploit. When multiple sockets/process can concurrently request (and than disable) the udp encap, we need to track the activation counter with *_inc()/*_dec() jump label variants, or we can experience bad things at disable time. Fixes: 60fb9567 ("udp: implement complete book-keeping for encap_needed") Reported-by:
Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by:
Paolo Abeni <pabeni@redhat.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Jesus Sanchez-Palencia authored
Currently on dequeue() ETF only drops the first expired packet, which causes a problem if the next packet is already expired. When this happens, the watchdog will be configured with a time in the past, fire straight way and the packet will finally be dropped once the dequeue() function of the qdisc is called again. We can save quite a few cycles and improve the overall behavior of the qdisc if we drop all expired packets if the next packet is expired. This should allow ETF to recover faster from bad situations. But packet drops are still a very serious warning that the requirements imposed on the system aren't reasonable. This was inspired by how the implementation of hrtimers use the rb_tree inside the kernel. Signed-off-by:
Jesus Sanchez-Palencia <jesus.s.palencia@gmail.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Jesus Sanchez-Palencia authored
This is just a refactor that will simplify the implementation of the next patch in this series which will drop all expired packets on the dequeue flow. Signed-off-by:
Jesus Sanchez-Palencia <jesus.s.palencia@gmail.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Jesus Sanchez-Palencia authored
ETF's peek() operation is heavily used so use an rb_root_cached instead and leverage rb_first_cached() which will run in O(1) instead of O(log n). Even if on 'timesortedlist_clear()' we could be using rb_erase(), we choose to use rb_erase_cached(), because if in the future we allow runtime changes to ETF parameters, and need to do a '_clear()', this might cause some hard to debug issues. Signed-off-by:
Jesus Sanchez-Palencia <jesus.s.palencia@gmail.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Jesus Sanchez-Palencia authored
There is no point in firing the qdisc watchdog if there are no future skbs pending in the queue and the watchdog had been set previously. Signed-off-by:
Jesus Sanchez-Palencia <jesus.s.palencia@gmail.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Yafang Shao authored
Currently we can use bpf or tcp tracepoint to conveniently trace the tcp state transition at the run time. So we don't need to do this stuff at the compile time anymore. Signed-off-by:
Yafang Shao <laoar.shao@gmail.com> Signed-off-by:
Eric Dumazet <edumazet@google.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
David S. Miller authored
Ben Dooks says: ==================== SMSC95xx driver updates (round 1) This is a series of a few driver cleanups and some fixups of the code for the SMSC95XX driver. There have been a few reviews, and the issues have been fixed so this should be ready for merging. I will work on the tx-alignment and the other bits of usbnet changes and produce at least two more patch series for this later. ==================== Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Ben Dooks authored
The manual states that the checksum cannot lie in the last DWORD of the transmission, so add a basic check for this and fall back to software checksumming the packet. This only seems to trigger for ACK packets with no options or data to return to the other end, and the use of the tx-alignment option makes it more likely to happen. Signed-off-by:
Ben Dooks <ben.dooks@codethink.co.uk> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Ben Dooks authored
Change the RX code to use get_unaligned_le32() instead of the combo of memcpy and cpu_to_le32s(&var). Signed-off-by:
Ben Dooks <ben.dooks@codethink.co.uk> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Ben Dooks authored
The smsc95xx_tx_fixup is doing multiple calls to skb_push() to put an 8-byte command header onto the packet. It would be easier to do one skb_push() and then copy the data in once the push is done. We also make the code smaller by using proper unaligned puts for the header. This merges in the CPU to LE32 conversion as well and makes the whole sequence easier to understand hopefully. Signed-off-by:
Ben Dooks <ben.dooks@codethink.co.uk> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Ben Dooks authored
The smsc95xx driver already takes into account the NET_IP_ALIGN parameter when setting up the receive packet data, which means we do not need to worry about aligning the packets in the usbnet driver. Adding the EVENT_NO_IP_ALIGN means that the IPv4 header is now passed to the ip_rcv() routine with the start on an aligned address. Tested on Raspberry Pi B3. Signed-off-by:
Ben Dooks <ben.dooks@codethink.co.uk> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
David S. Miller authored
Ioana Ciocoi Radulescu says: ==================== dpaa2-eth: add bql support The first two patches make minor tweaks to the driver to simplify bql implementation. The third patch adds the actual bql support. ==================== Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Ioana Ciocoi Radulescu authored
Add support for byte queue limit. On NAPI poll, we save the total number of Tx confirmed frames/bytes and register them with bql at the end of the poll function. Signed-off-by:
Ioana Radulescu <ruxandra.radulescu@nxp.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Ioana Ciocoi Radulescu authored
Change the frame consume callback signature: * the entire FQ structure is passed to the callback instead of just the queue index * the NAPI structure can be easily obtained from the channel it is associated to, so we don't need to pass it explicitly Signed-off-by:
Ioana Radulescu <ruxandra.radulescu@nxp.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Ioana Ciocoi Radulescu authored
The DPNI object on which we build a network interface has a certain number of {Rx, Tx, Tx confirmation} frame queues as resources. The default hardware setup offers one queue of each type, as well as one DPCON channel, for each core available in the system. There are however cases where the number of queues is greater than the number of cores or channels. Until now, we configured and used all the frame queues associated with a DPNI, even if it meant assigning multiple queues of one type to the same channel. Update the driver to only use a number of queues equal to the number of channels, ensuring each channel will contain exactly one Rx and one Tx confirmation queue. >From the user viewpoint, this change is completely transparent. Performance wise there is no impact in most scenarios. In case the number of queues is larger than and not a multiple of the number of channels, Rx hash distribution offers now better load balancing between cores, which can have a positive impact on overall system performance. Signed-off-by:
Ioana Radulescu <ruxandra.radulescu@nxp.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Jiri Pirko authored
Currently, the vlan packet offloads are registered only upon 8021q module load. However, even without this module loaded, the offloads could be utilized, for example by openvswitch datapath. As reported by Michael, that causes 2x to 5x performance improvement, depending on a testcase. So move the vlan offload registrations into vlan_core and make this available even without 8021q module loaded. Reported-by:
Michael Shteinbok <michaelsh86@gmail.com> Signed-off-by:
Jiri Pirko <jiri@mellanox.com> Tested-by:
Michael Shteinbok <michaelsh86@gmail.com> Reviewed-by:
David Ahern <dsahern@gmail.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Colin Ian King authored
There is a missing indentation before the declaration of port. Add it. Signed-off-by:
Colin Ian King <colin.king@canonical.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Colin Ian King authored
Trivial fix, the spelling of "failded" is incorrect in dev_err and dev_warn messages. Fix this. Signed-off-by:
Colin Ian King <colin.king@canonical.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Cong Wang authored
Signed-off-by:
Cong Wang <xiyou.wangcong@gmail.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-