- 16 Jun, 2016 40 commits
-
-
Xin Long authored
Commit d46e416c ("sctp: sctp should change socket state when shutdown is received") may set sk_state CLOSING in sctp_sock_migrate, but inet_accept doesn't allow the sk_state other than ESTABLISHED/ CLOSED for sctp. So we will change sk_state to CLOSED, instead of CLOSING, as actually sk is closed already there. Fixes: d46e416c ("sctp: sctp should change socket state when shutdown is received") Reported-by:
Ye Xiaolong <xiaolong.ye@intel.com> Signed-off-by:
Xin Long <lucien.xin@gmail.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Fabien Siron authored
Signed-off-by:
Fabien Siron <fabien.siron@epita.fr> Signed-off-by:
-
David S. Miller authored
Daniel Borkmann says: ==================== bpf: improve fd array release This set improves BPF perf fd array map release wrt to purging entries, first two extend the API as needed. Please see individual patches for more details. ==================== Signed-off-by: -
Daniel Borkmann authored
The behavior of perf event arrays are quite different from all others as they are tightly coupled to perf event fds, f.e. shown recently by commit e03e7ee3 ("perf/bpf: Convert perf_event_array to use struct file") to make refcounting on perf event more robust. A remaining issue that the current code still has is that since additions to the perf event array take a reference on the struct file via perf_event_get() and are only released via fput() (that cleans up the perf event eventually via perf_event_release_kernel()) when the element is either manually removed from the map from user space or automatically when the last reference on the perf event map is dropped. However, this leads us to dangling struct file's when the map gets pinned after the application owning the perf event descriptor exits, and since the struct file reference will in such case only be manually dropped or via pinned file removal, it leads to the perf event living longer than necessary, consuming needlessly resources for that time. Relations between perf event fds and bpf perf event map fds can be rather complex. F.e. maps can act as demuxers among different perf event fds that can possibly be owned by different threads and based on the index selection from the program, events get dispatched to one of the per-cpu fd endpoints. One perf event fd (or, rather a per-cpu set of them) can also live in multiple perf event maps at the same time, listening for events. Also, another requirement is that perf event fds can get closed from application side after they have been attached to the perf event map, so that on exit perf event map will take care of dropping their references eventually. Likewise, when such maps are pinned, the intended behavior is that a user application does bpf_obj_get(), puts its fds in there and on exit when fd is released, they are dropped from the map again, so the map acts rather as connector endpoint. This also makes perf event maps inherently different from program arrays as described in more detail in commit c9da161c ("bpf: fix clearing on persistent program array maps"). To tackle this, map entries are marked by the map struct file that added the element to the map. And when the last reference to that map struct file is released from user space, then the tracked entries are purged from the map. This is okay, because new map struct files instances resp. frontends to the anon inode are provided via bpf_map_new_fd() that is called when we invoke bpf_obj_get_user() for retrieving a pinned map, but also when an initial instance is created via map_create(). The rest is resolved by the vfs layer automatically for us by keeping reference count on the map's struct file. Any concurrent updates on the map slot are fine as well, it just means that perf_event_fd_array_release() needs to delete less of its own entires. Signed-off-by:
Daniel Borkmann <daniel@iogearbox.net> Acked-by:
Alexei Starovoitov <ast@kernel.org> Signed-off-by:
-
Daniel Borkmann authored
This patch extends map_fd_get_ptr() callback that is used by fd array maps, so that struct file pointer from the related map can be passed in. It's safe to remove map_update_elem() callback for the two maps since this is only allowed from syscall side, but not from eBPF programs for these two map types. Like in per-cpu map case, bpf_fd_array_map_update_elem() needs to be called directly here due to the extra argument. Signed-off-by:
-
Daniel Borkmann authored
Add a release callback for maps that is invoked when the last reference to its struct file is gone and the struct file about to be released by vfs. The handler will be used by fd array maps. Signed-off-by:
-
David S. Miller authored
Edward Cree says: ==================== sfc: RX VLAN filtering Adds support for VLAN-qualified receive filters on EF10 hardware. This is needed when running as a guest if the hypervisor has enabled vfs-vlan-restrict, in which case the firmware rejects filters not qualified with VLAN 0. ==================== Signed-off-by: -
Andrew Rybchenko authored
If vPort has VLAN_RESTRICT flag, VLAN tagged traffic will not be delivered without corresponding Rx filters which may be proxied to and moderated by hypervisor. Signed-off-by:
Edward Cree <ecree@solarflare.com> Signed-off-by:
-
Edward Cree authored
Signed-off-by:
-
Andrew Rybchenko authored
If should be done after net_dev->hw_features initialization, to keep the feature there to be able to enable it later using ethtool. VLAN filtering is enforced and fixed if vPort requires usage of VLAN filters to receive tagged traffic. Signed-off-by:
-
Martin Habets authored
If it is not supported we simply disable the feature. For the feature to work we need firmware filter support for OUTER_VID + LOC_MAC and for OUTER_VID + LOC_MAC_IG. The low-latency firmware can match on OUTER_VID + LOC_MAC but not on OUTER_VID + LOC_MAC_IG. For the capture packet firmware it is the other way around. Only the full-feature variant can match on both combinations. Incorporates a fix by Andrew Rybchenko <Andrew.Rybchenko@oktetlabs.ru> in the net_dev->[hw_]features handling. Signed-off-by:
-
Andrew Rybchenko authored
Filter match flags are not unique criteria to be mapped to priority because of both unknown unicast and unknown multicast are mapped to LOC_MAC_IG. So, local MAC is required to map filter to priority. MCDI filter flags is unique criteria to find filter priority. Signed-off-by:
-
Edward Cree authored
Nearly every time we call efx_ef10_filter_remove_unsafe, we first check for EFX_EF10_FILTER_ID_INVALID, in which case we do nothing. So move that check into the function, simplifying all the call sites. Also, change the return type to void, since none of the callers check it. Signed-off-by:
-
Martin Habets authored
When trying to enslave an SFC interface to a bond the following BUG_ON was hit: kernel BUG [in ef10.c]! CPU: 0 PID: 4383 Comm: ifenslave Tainted: G ... Call Trace: efx_ef10_filter_add_vlan+0x121/0x180 [sfc] efx_ef10_filter_table_probe+0x2a2/0x4f0 [sfc] efx_ef10_set_mac_address+0x370/0x6d0 [sfc] efx_set_mac_address+0x7d/0x120 [sfc] dev_set_mac_address+0x43/0xa0 bond_enslave+0x337/0xea0 [bonding] This comes from function efx_ef10_filter_vlan_sync_rx_mode. To solve the bug we ensure the mac_lock is taken before calling efx_ef10_filter_add_vlan. But to avoid a priority inversion mac_lock must be taken before filter_sem. To satisfy these requirements we end up taking mac_lock in efx_ef10_vport_set_mac_address, efx_ef10_set_mac_address, efx_ef10_sriov_set_vf_vlan and efx_probe_filters. Signed-off-by:
-
Andrew Rybchenko authored
Supports HW VLAN filtering, en/disabled using ethtool. Signed-off-by:
-
Andrew Rybchenko authored
Right now it contains dummy VLAN entry with unspecified VID only. The entry is used for the case when HW VLAN filtering is not used. Signed-off-by:
-
Andrew Rybchenko authored
It is a step to support VLAN filtering in HW. Signed-off-by:
-
Andrew Rybchenko authored
These flags are built when address cache is updated. The information will be required when VLAN filtering is added and address cache is used without re-sync. Signed-off-by:
-
Andrew Rybchenko authored
It is a step to support VLAN filtering in HW. Until then, there is only one struct efx_ef10_filter_vlan per struct efx_ef10_filter_table, with no VLAN information yet. Signed-off-by:
-
Andrew Rybchenko authored
It is required to remove setting of filter IDs to invalid from multicast and unicast addresses caching functions. Add initialization to invalid when filter table is created. Add paranoid checks to track consistency. Signed-off-by:
-
Edward Cree authored
Based on a patch by Andrew Rybchenko <Andrew.Rybchenko@oktetlabs.ru> Signed-off-by:
-
Andrew Rybchenko authored
It allows to change set of fixed features on datapath reset. Signed-off-by:
-
Andrew Rybchenko authored
It is used for EF10 only and logically belongs to EF10 filter table state. It is OK that it is reset to false on filter table recreation since all filters are removed on destruction. Signed-off-by:
-
Andrew Rybchenko authored
It is useful to simplify features addition. Signed-off-by:
-
David S. Miller authored
Merge tag 'rxrpc-rewrite-20160615' of git://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs David Howells says: ==================== rxrpc: Rework endpoint record handling Here's the next part of the AF_RXRPC rewrite. In this set I rework endpoint record handling. There are two types of endpoint record, local and peer. The local endpoint record is used as an anchor for the transport socket that AF_RXRPC uses (at the moment a UDP socket). Local endpoints can be shared between AF_RXRPC sockets under certain restricted circumstances. The peer endpoint is a record of the remote end. It is (or will be) used to keep track MTU and RTT values and, with these changes, is used to find the call(s) to abort when a network error occurs. The following significant changes are made: (1) The local endpoint event handling code is split out into its own file. (2) The local endpoint list bottom half-excluding spinlock is removed as things are arranged such that sk_user_data will not change whilst the transport socket callbacks are in progress. (3) Local endpoints can now only be shared if they have the same transport address (as before) and have a local service ID of 0 (ie. they're not listening for incoming calls). This prevents callbacks from a server to one process being picked up by another process. (4) Local endpoint destruction is now accomplished by the same work item as processes events, meaning that the destructor doesn't need to wait for the event processor. (5) Peer endpoints are now held in a hash table rather than a flat list. (6) Peer endpoints are now destroyed by RCU rather than by work item. (7) Peer endpoints are now differentiated by local endpoint and remote transport port in addition to remote transport address and transport type and family. This means that a firewall that excludes access between a particular local port and remote port won't cause calls to be aborted that use a different port pair. (8) Error report handling now no longer assumes that the source is always an IPv4 ICMP message from a UDP port and has assumptions that an ICMP message comes from an IPv4 socket removed. At some point IPv6 support will be added. (9) Peer endpoints rather than local endpoints are now the anchor point for distributing network error reports. (10) Both types of endpoint records are now disposed of as soon as all references to them are gone. There is less hanging around and once their usage counts hit zero, records can no longer be resurrected. ==================== Signed-off-by:
-
David S. Miller authored
Raghu Vatsavayi says: ==================== liquidio: Updates and Bug fixes Following are updates for liquidio bug fixes and driver support for new firmware interface. These updates are divided into smaller logical patches as mentioned by you. These set of nine patches should be applied in the following order as some of them depend on earlier patches in the list. ==================== Signed-off-by: -
Raghu Vatsavayi authored
Added support for new instruction header for octeon2/octeon3(ih) and corresponding changes. Signed-off-by:
Derek Chickles <derek.chickles@caviumnetworks.com> Signed-off-by:
Satanand Burla <satananda.burla@caviumnetworks.com> Signed-off-by:
Felix Manlunas <felix.manlunas@caviumnetworks.com> Signed-off-by:
Raghu Vatsavayi <raghu.vatsavayi@caviumnetworks.com> Signed-off-by:
-
Raghu Vatsavayi authored
This patch decoupled the firmware side ifidx and host side interface number. It also has some minor name change for linkinfo sturct field. Signed-off-by:
-
Raghu Vatsavayi authored
This patch is for new driver/firmware control command structure (octnic_packet_params and octnic_cmd_setup ) and resultant code changes. Signed-off-by:
-
Raghu Vatsavayi authored
This patch is to refactor packet size calculations to support PTP enabled for 66xx and 68xx cards and also other cards that do not support PTP. Signed-off-by:
-
Raghu Vatsavayi authored
This patch is to add page based buffers for receive side descriptors of the driver and separate free routines for rx and tx buffers. Signed-off-by:
-
Raghu Vatsavayi authored
This patch is to allocate rx queue's memory based on numa node and also use page based buffers for rx traffic improvements. Signed-off-by:
-
Raghu Vatsavayi authored
This patch is to allocate and manage scatter gather lists per input queue(iq's) and remove queue's interdependence. Signed-off-by:
-
Raghu Vatsavayi authored
This patch is to allocate the input queues based on Numa node in tx path and queue mapping changes based on the mapping info provided by firmware. Signed-off-by:
-
Raghu Vatsavayi authored
This patch is to resolve the double free issue by checking proper return values from soft command. Signed-off-by:
-
Tom Herbert authored
The algorithm for checksum neutral mapping is incorrect. This problem was being hidden since we were previously always performing checksum offload on the translated addresses and only with IPv6 HW csum. Enabling an ILA router shows the issue. Corrected algorithm: old_loc is the original locator in the packet, new_loc is the value to overwrite with and is found in the lookup table. old_flag is the old flag value (zero of CSUM_NEUTRAL_FLAG) and new_flag is then (old_flag ^ CSUM_NEUTRAL_FLAG) & CSUM_NEUTRAL_FLAG. Need SUM(new_id + new_flag + diff) == SUM(old_id + old_flag) for checksum neutral translation. Solving for diff gives: diff = (old_id - new_id) + (old_flag - new_flag) compute_csum_diff8(new_id, old_id) gives old_id - new_id If old_flag is set old_flag - new_flag = old_flag = CSUM_NEUTRAL_FLAG Else old_flag - new_flag = -new_flag = ~CSUM_NEUTRAL_FLAG Tested: - Implemented a user space program that creates random addresses and random locators to overwrite. Compares the checksum over the address before and after translation (must always be equal) - Enabled ILA router and showed proper operation. Signed-off-by:Tom Herbert <tom@herbertland.com> Signed-off-by:
-
Philip Prindeville authored
In the presence of firewalls which improperly block ICMP Unreachable (including Fragmentation Required) messages, Path MTU Discovery is prevented from working. A workaround is to handle IPv4 payloads opaquely, ignoring the DF bit--as is done for other payloads like AppleTalk--and doing transparent fragmentation and reassembly. Redux includes the enforcement of mutual exclusion between this feature and Path MTU Discovery as suggested by Alexander Duyck. Cc: Alexander Duyck <alexander.duyck@gmail.com> Reviewed-by:Stephen Hemminger <stephen@networkplumber.org> Signed-off-by:
Philip Prindeville <philipp@redfish-solutions.com> Signed-off-by:
-
David Ahern authored
Attempting to delete a VRF device with a socket bound to it can stall: unregister_netdevice: waiting for red to become free. Usage count = 1 The unregister is waiting for the dst to be released and with it references to the vrf device. Similar to dst_ifdown switch the dst dev to loopback on delete for all of the dst's for the vrf device and release the references to the vrf device. Fixes: 193125db ("net: Introduce VRF device driver") Fixes: 35402e31 ("net: Add IPv6 support to VRF device") Signed-off-by:
David Ahern <dsa@cumulusnetworks.com> Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/leon/linux-rdmaDavid S. Miller authored
Mellanox shared code between RDMA and net-next trees This is Mellanox mlx5_core shared code for both net-next and RDMA trees for 4.8 kernel cycle.
-
Arnd Bergmann authored
The latest changes to the MDIO code introduced a false-positive warning with gcc-6 (possibly others): drivers/net/phy/mdio-mux.c: In function 'mdio_mux_init': drivers/net/phy/mdio-mux.c:188:3: error: 'parent_bus_node' may be used uninitialized in this function [-Werror=maybe-uninitialized] It's easy to avoid the warning by making sure the parent_bus_node is initialized in both cases at the start of the function, since the later 'of_node_put()' call is also valid for a NULL pointer argument. Signed-off-by:
Arnd Bergmann <arnd@arndb.de> Fixes: f20e6657 ("mdio: mux: Enhanced MDIO mux framework for integrated multiplexers") Reviewed-by:
Andrew Lunn <andrew@lunn.ch> Signed-off-by:
-
