- 02 Jan, 2013 20 commits
-
-
Ariel Elior authored
The FLR indication arrives as an attention from the management processor. Upon VF flr all FLRed function in the indication have already been released by Firmware and now we basically need to free the resources allocated to those VFs, and clean any remainders from the device (FLR final cleanup). Signed-off-by:
Ariel Elior <ariele@broadcom.com> Signed-off-by:
Eilon Greenstein <eilong@broadcom.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
Ariel Elior authored
The 'release' request is the opposite of the 'acquire' request. At release, all the resources allocated to the VF are reclaimed. The release flow applies the close flow if applicable. Note that there are actually two types of release: 1. The VF has been removed, and so issued a 'release' request over the VF <-> PF Channel. 2. The PF is going down and so has to release all of it's VFs. Signed-off-by:
-
Ariel Elior authored
The 'close' command is the opposite of an init request. Here the queues of the VF are closed (if any are opened) and released. This flow applies the 'q_teardown' flow on all the queues. The VF state is changed by this request. Interrupts are disabled for the VF when closed. Signed-off-by:
-
Ariel Elior authored
The 'q_teardown' request is basically the opposite of the 'q_setup'. Here the PF driver removes from the device the queue it opened against the VF fastpath ring at 'setup_q' stage, along with all related rx_mode info. Signed-off-by:
-
Ariel Elior authored
The VF driver uses the 'q_filters' message on the VF <-> PF channel for configuring an open queue, for example when the rxmode changes. Signed-off-by:
-
Ariel Elior authored
Upon receiving a 'setup_q' request from the VF over the VF <-> PF channel the PF driver will open a corresponding queue in the device. The PF driver configures the queue with appropriate mac address, vlan configuration, etc from the VF. Signed-off-by:
-
Ariel Elior authored
Statistics are collected by the PF driver. The collection is performed via a query sent to the device which is basically an array of 3-tuples of the form (statistics client, function, DMAE address). In this patch the PF driver adds to the query, on top of the statistics clients it is maintaining for itself (rss queues, storage, etc), the 3-tuples for the VFs it is maintaining. The addresses used are the GPAs of the statistics buffers supplied by the VF in the init message on the VF <-> PF channel. The function parameter ensures that the iommu will translate the GPA to the correct physical address. Signed-off-by:
-
Ariel Elior authored
The VF driver will send an 'init' request as part of its nic load flow. This message is used by the VF to publish the GPA's of its status blocks, slow path ring and statistics buffer. The PF driver notes all this down in the VF database, and also uses this message to transfer the VF to VF_INIT state internally. Signed-off-by:
-
Ariel Elior authored
When a VF is probed by the VF driver, the VF driver sends an 'acquire' request over the VF <-> PF channel for the resources it needs to operate (interrupts, queues, etc). The PF driver either ratifies the request and allocates the resources, responds with the maximum values it will allow the VF to acquire, or fails the request entirely if there is a problem. Signed-off-by:
-
Ariel Elior authored
Support interrupt from device which indicates VF has placed A request on the VF <-> PF channel. The PF driver issues a DMAE to retrieve the request from the VM memory (the Ghost Physical Address of the request is contained in the interrupt. The PF driver uses the GPA in the DMAE request, which is translated by the IOMMU to the correct physical address). The request which arrives is examined to recognize the sending VF. The PF driver allocates a workitem to handle the VF Operation (vfop). Signed-off-by:
-
Ariel Elior authored
At nic load of the PF, if VFs may be present, prepare the device for the VFs. Initialize the VF database in preparation of VF arrival. Signed-off-by:
-
Ariel Elior authored
When A PF determines that it may have to manage SRIOV VFs it allocates a database for this purpose. The database is intended to keep track of the VF state, the resources allocated for each VF (queues, interrupt vectors, etc), the state of the VF's queues. When the VF loads the database is updated accordingly. When A VF closes the database is consulted to determine which resources need to be released (close queues against device, reclaim interrupt vectors, etc). Signed-off-by:
-
Ariel Elior authored
When VF driver is transmitting it must supply the correct mac address in the parsing BD. This is used for firmware validation and enforcement and also for tx-switching. Refactor interrupt ack flow to allow for different BAR addresses of the hardware in the PF BAR vs the VF BAR. Signed-off-by:
-
Ariel Elior authored
The VF driver uses the 'q_filter' request in the VF <-> PF channel to have the PF configure the requested rxmode to device. ndo_set_rxmode is called under bottom half lock, so sleeping until the response arrives over the VF <-> PF channel is out of the question. For this reason the VF driver returns from the ndo after scheduling a work item, which in turn processes the rx mode request and adds the classification information through the VF <-> PF channel accordingly. Signed-off-by:
-
Ariel Elior authored
When a VF is being closed its queues are released via the 'teardown_q' and the VF itself is closed with 'close'. These are essentially the unload counterparts of 'init' and 'setup_q' from the load flow. Signed-off-by:
-
Ariel Elior authored
'init' - init an acquired VF. Supply allocation GPAs to PF. 'setup_q' - PF to allocate a queue in device on behalf of the VF. 'set_mac' - PF to configure a mac in device on behalf of the VF. VF driver uses these requests in the VF <-> PF channel in nic_load flow. Signed-off-by:
-
Ariel Elior authored
Generally, the VF driver cannot access the chip, except by the narrow window its BAR allows. Care had to be taken so the VF driver will not reach code which accesses the chip elsewhere. Refactor the nic_load flow into parts so it would be easier to separate the VF-only logic from the PF-only logic. Signed-off-by:
-
Ariel Elior authored
VF driver uses this request when removed. The PF driver reclaims all resources allocated for that VF at this time. Signed-off-by:
-
Ariel Elior authored
Add the 'acquire' request to VF <-> PF channel and use it at VF probe. In the acquire request the VF driver lists the resources it would like to have. In the response the PF either ratifies the request, or denies it and supplies the maximum values supported. The VF may then attempt another acquire request. This patch adds the bnx2x_vfpf.c file which contains the implementation of the VF to PF hardware channel. Signed-off-by:
-
Ariel Elior authored
To support probing and removing of a bnx2x virtual function the following were added: 1. add bnx2x_vfpf.h: defines the VF to PF channel 2. add bnx2x_sriov.h: header for bnx2x SR-IOV functionality 3. enumerate VF hw types (identify VFs) 4. if driving a VF, map VF bar 5. if driving a VF, allocate Vf to PF channel 6. refactor interrupt flows to include VF Signed-off-by:
-
- 30 Dec, 2012 6 commits
-
-
Flavio Leitner authored
This patch adds few ethtool operations to team driver. Signed-off-by:
Flavio Leitner <fbl@redhat.com> Acked-by:
Jiri Pirko <jiri@resnulli.us> Signed-off-by:
-
Eric Dumazet authored
veth is lacking most modern facilities, like SG, checksums, TSO. It makes sense to extend dev->features to get them, or GRO aggregation is defeated by a forced segmentation. Reported-by:
Andrew Vagin <avagin@parallels.com> Signed-off-by:
Eric Dumazet <edumazet@google.com> Cc: Michał Mirosław <mirq-linux@rere.qmqm.pl> Signed-off-by:
-
Eric Dumazet authored
veth stats are a bit bloated. There is no need to account transmit and receive stats, since they are absolutely symmetric. Also use a per device atomic64_t for the dropped counter, as it should never be used in fast path. Signed-off-by:
-
Flavio Leitner authored
The user space teamd daemon may need to control the master's carrier state depending on the selected mode. Signed-off-by:
-
stephen hemminger authored
The bridge link detection should follow the operational state of the lower device, rather than the carrier bit. This allows devices like tunnels that are controlled by userspace control plane to work with bridge STP link management. Signed-off-by:
Stephen Hemminger <shemminger@vyatta.com> Reviewed-by:
-
Daniel Borkmann authored
Currently, we return -EINVAL for malformed or wrong BPF filters. However, this is not done for BPF_S_ANC* operations, which makes it more difficult to detect if it's actually supported or not by the BPF machine. Therefore, we should also return -EINVAL if K is within the SKF_AD_OFF universe and the ancillary operation did not match. Why exactly is it needed? If tools such as libpcap/tcpdump want to make use of new ancillary operations (like filtering VLAN in kernel space), there is currently no sane way to test if this feature / BPF_S_ANC* op is present or not, since no error is returned. This patch will make life easier for that and allow for a proper usage for user space applications. There was concern, if this patch will break userland. Short answer: Yes and no. Long answer: It will "break" only for code that calls ... { BPF_LD | BPF_(W|H|B) | BPF_ABS, 0, 0, <K> }, ... where <K> is in [0xfffff000, 0xffffffff] _and_ <K> is *not* an ancillary. And here comes the BUT: assuming some *old* code will have such an instruction where <K> is between [0xfffff000, 0xffffffff] and it doesn't know ancillary operations, then this will give a non-expected / unwanted behavior as well (since we do not return the BPF machine with 0 after a failed load_pointer(), which was the case before introducing ancillary operations, but load sth. into the accumulator instead, and continue with the next instruction, for instance). Thus, user space code would already have been broken by introducing ancillary operations into the BPF machine per se. Code that does such a direct load, e.g. "load word at packet offset 0xffffffff into accumulator" ("ld [0xffffffff]") is quite broken, isn't it? The whole assumption of ancillary operations is that no-one intentionally calls things like "ld [0xffffffff]" and expect this word to be loaded from such a packet offset. Hence, we can also safely make use of this feature testing patch and facilitate application development. Therefore, at least from this patch onwards, we have *for sure* a check whether current or in future implemented BPF_S_ANC* ops are supported in the kernel. Patch was tested on x86_64. (Thanks to Eric for the previous review.) Cc: Eric Dumazet <eric.dumazet@gmail.com> Reported-by:Ani Sinha <ani@aristanetworks.com> Signed-off-by:
Daniel Borkmann <dborkman@redhat.com> Signed-off-by:
-
- 29 Dec, 2012 3 commits
-
-
stephen hemminger authored
Sparse detected case where this local function should be static. It may even allow some compiler optimizations. Signed-off-by:
-
stephen hemminger authored
Detected by sparse. Signed-off-by:
-
stephen hemminger authored
Fix sparse warning about local function that should be static. Signed-off-by:
-
- 28 Dec, 2012 6 commits
-
-
Eric Dumazet authored
Use the new per task frag allocator in skb_append_datato_frags(), to reduce number of frags and page allocator overhead. Tested: ifconfig lo mtu 16436 perf record netperf -t UDP_STREAM ; perf report before : Throughput: 32928 Mbit/s 51.79% netperf [kernel.kallsyms] [k] copy_user_generic_string 5.98% netperf [kernel.kallsyms] [k] __alloc_pages_nodemask 5.58% netperf [kernel.kallsyms] [k] get_page_from_freelist 5.01% netperf [kernel.kallsyms] [k] __rmqueue 3.74% netperf [kernel.kallsyms] [k] skb_append_datato_frags 1.87% netperf [kernel.kallsyms] [k] prep_new_page 1.42% netperf [kernel.kallsyms] [k] next_zones_zonelist 1.28% netperf [kernel.kallsyms] [k] __inc_zone_state 1.26% netperf [kernel.kallsyms] [k] alloc_pages_current 0.78% netperf [kernel.kallsyms] [k] sock_alloc_send_pskb 0.74% netperf [kernel.kallsyms] [k] udp_sendmsg 0.72% netperf [kernel.kallsyms] [k] zone_watermark_ok 0.68% netperf [kernel.kallsyms] [k] __cpuset_node_allowed_softwall 0.67% netperf [kernel.kallsyms] [k] fib_table_lookup 0.60% netperf [kernel.kallsyms] [k] memcpy_fromiovecend 0.55% netperf [kernel.kallsyms] [k] __udp4_lib_lookup after: Throughput: 47185 Mbit/s 61.74% netperf [kernel.kallsyms] [k] copy_user_generic_string 2.07% netperf [kernel.kallsyms] [k] prep_new_page 1.98% netperf [kernel.kallsyms] [k] skb_append_datato_frags 1.02% netperf [kernel.kallsyms] [k] sock_alloc_send_pskb 0.97% netperf [kernel.kallsyms] [k] enqueue_task_fair 0.97% netperf [kernel.kallsyms] [k] udp_sendmsg 0.91% netperf [kernel.kallsyms] [k] __ip_route_output_key 0.88% netperf [kernel.kallsyms] [k] __netif_receive_skb 0.87% netperf [kernel.kallsyms] [k] fib_table_lookup 0.85% netperf [kernel.kallsyms] [k] resched_task 0.78% netperf [kernel.kallsyms] [k] __udp4_lib_lookup 0.77% netperf [kernel.kallsyms] [k] _raw_spin_lock_irqsave Signed-off-by: -
Jiri Pirko authored
Signed-off-by:
-
Jiri Pirko authored
Signed-off-by:
-
Jiri Pirko authored
Make carrier writable Signed-off-by:
-
Jiri Pirko authored
This allows a driver to register change_carrier callback which will be called whenever user will like to change carrier state. This is useful for devices like dummy, gre, team and so on. Signed-off-by:
-
Sasha Levin authored
Signed-off-by:
Sasha Levin <sasha.levin@oracle.com> Signed-off-by:
-
- 27 Dec, 2012 3 commits
-
-
git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-stagingLinus Torvalds authored
Pull hwmon fixes from Guenter Roeck: - Report i2c errors to userspace in lm73 driver - Fix problem with DIV_ROUND_CLOSEST and unsigned divisors in emc6w201 driver * tag 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging: hwmon: (emc6w201) Fix DIV_ROUND_CLOSEST problem with unsigned divisors hwmon: (lm73} Detect and report i2c bus errors
-
git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespaceLinus Torvalds authored
Pull namespace fixes from Eric Biederman: "This tree includes two bug fixes for problems Oleg spotted on his review of the recent pid namespace work. A small fix to not enable bottom halves with irqs disabled, and a trivial build fix for f2fs with user namespaces enabled." * 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace: f2fs: Don't assign e_id in f2fs_acl_from_disk proc: Allow proc_free_inum to be called from any context pidns: Stop pid allocation when init dies pidns: Outlaw thread creation after unshare(CLONE_NEWPID)
-
git://git.kernel.org/pub/scm/linux/kernel/git/davem/netLinus Torvalds authored
Pull networking fixes from David Miller: 1) GRE tunnel drivers don't set the transport header properly, they also blindly deref the inner protocol ipv4 and needs some checks. Fixes from Isaku Yamahata. 2) Fix sleeps while atomic in netdevice rename code, from Eric Dumazet. 3) Fix double-spinlock in solos-pci driver, from Dan Carpenter. 4) More ARP bug fixes. Fix lockdep splat in arp_solicit() and then the bug accidentally added by that fix. From Eric Dumazet and Cong Wang. 5) Remove some __dev* annotations that slipped back in, as well as all HOTPLUG references. From Greg KH 6) RDS protocol uses wrong interfaces to access scatter-gather elements, causing a regression. From Mike Marciniszyn. 7) Fix build error in cpts driver, from Richard Cochran. 8) Fix arithmetic in packet scheduler, from Stefan Hasko. 9) Similarly, fix association during calculation of random backoff in batman-adv. From Akinobu Mita. * git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (21 commits) ipv6/ip6_gre: set transport header correctly ipv4/ip_gre: set transport header correctly to gre header IB/rds: suppress incompatible protocol when version is known IB/rds: Correct ib_api use with gs_dma_address/sg_dma_len net/vxlan: Use the underlying device index when joining/leaving multicast groups tcp: should drop incoming frames without ACK flag set netprio_cgroup: define sk_cgrp_prioidx only if NETPRIO_CGROUP is enabled cpts: fix a run time warn_on. cpts: fix build error by removing useless code. batman-adv: fix random jitter calculation arp: fix a regression in arp_solicit() net: sched: integer overflow fix CONFIG_HOTPLUG removal from networking core Drivers: network: more __dev* removal bridge: call br_netpoll_disable in br_add_if ipv4: arp: fix a lockdep splat in arp_solicit() tuntap: dont use a private kmem_cache net: devnet_rename_seq should be a seqcount ip_gre: fix possible use after free ip_gre: make ipgre_tunnel_xmit() not parse network header as IP unconditionally ...
-
- 26 Dec, 2012 2 commits
-
-
Isaku Yamahata authored
ip6gre_xmit2() incorrectly sets transport header to inner payload instead of GRE header. It seems copy-and-pasted from ipip.c. Set transport header to gre header. (In ipip case the transport header is the inner ip header, so that's correct.) Found by inspection. In practice the incorrect transport header doesn't matter because the skb usually is sent to another net_device or socket, so the transport header isn't referenced. Signed-off-by:
Isaku Yamahata <yamahata@valinux.co.jp> Signed-off-by:
-
Isaku Yamahata authored
ipgre_tunnel_xmit() incorrectly sets transport header to inner payload instead of GRE header. It seems copy-and-pasted from ipip.c. So set transport header to gre header. (In ipip case the transport header is the inner ip header, so that's correct.) Found by inspection. In practice the incorrect transport header doesn't matter because the skb usually is sent to another net_device or socket, so the transport header isn't referenced. Signed-off-by:
-
