- 03 Aug, 2014 22 commits
-
-
David S. Miller authored
Thomas Graf says: ==================== Lockless netlink_lookup() with new concurrent hash table Netlink sockets are maintained in a hash table to allow efficient lookup via the port ID for unicast messages. However, lookups currently require a read lock to be taken. This series adds a new generic, resizable, scalable, concurrent hash table based on the paper referenced in the first patch. It then makes use of the new data type to implement lockless netlink_lookup(). Patch 3/3 to convert nft_hash is included for reference but should be merged via the netfilter tree. Inclusion in this series is to provide context for the suggested API. Against net-next since the initial user of the new hash table is in net/ Changes: v4-v5: - use GFP_KERNEL to alloc Netlink buckets as suggested by Nikolay Aleksandrov - free nft hash element on removal as spotted by Nikolay Aleksandrov and Patrick McHardy v3-v4: - fixed wrong shift assignment placement as spotted by Nikolay Aleksandrov - reverted default size of nft_hash to 4 as requested by Patrick McHardy, default size for other hash tables remains at 64 if no hint is given - fixed copyright as requested by Patrick McHardy v2-v3: - fixed typo in nft_hash_destroy() when passing rhashtable handle v1-v2: - fixed traversal off-by-one as spotted by Tobias Klauser - removed unlikely() from BUG_ON() as spotted by Josh Triplett - new 3rd patch to convert nft_hash to rhashtable - make rhashtable_insert() return void - nl_sk_hash_lock must be a mutex - fixed wrong name of rht_shrink_below_30() - exported symbols rht_grow_above_75() and rht_shrink_below_30() - allow table freeing with RCU callback ==================== Signed-off-by:David S. Miller <davem@davemloft.net>
-
Thomas Graf authored
The sizing of the hash table and the practice of requiring a lookup to retrieve the pprev to be stored in the element cookie before the deletion of an entry is left intact. Signed-off-by:
Thomas Graf <tgraf@suug.ch> Acked-by:
Patrick McHardy <kaber@trash.net> Reviewed-by:
Nikolay Aleksandrov <nikolay@redhat.com> Signed-off-by:
-
Thomas Graf authored
Heavy Netlink users such as Open vSwitch spend a considerable amount of time in netlink_lookup() due to the read-lock on nl_table_lock. Use of RCU relieves the lock contention. Makes use of the new resizable hash table to avoid locking on the lookup. The hash table will grow if entries exceeds 75% of table size up to a total table size of 64K. It will automatically shrink if usage falls below 30%. Also splits nl_table_lock into a separate mutex to protect hash table mutations and allow synchronize_rcu() to sleep while waiting for readers during expansion and shrinking. Before: 9.16% kpktgend_0 [openvswitch] [k] masked_flow_lookup 6.42% kpktgend_0 [pktgen] [k] mod_cur_headers 6.26% kpktgend_0 [pktgen] [k] pktgen_thread_worker 6.23% kpktgend_0 [kernel.kallsyms] [k] memset 4.79% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup 4.37% kpktgend_0 [kernel.kallsyms] [k] memcpy 3.60% kpktgend_0 [openvswitch] [k] ovs_flow_extract 2.69% kpktgend_0 [kernel.kallsyms] [k] jhash2 After: 15.26% kpktgend_0 [openvswitch] [k] masked_flow_lookup 8.12% kpktgend_0 [pktgen] [k] pktgen_thread_worker 7.92% kpktgend_0 [pktgen] [k] mod_cur_headers 5.11% kpktgend_0 [kernel.kallsyms] [k] memset 4.11% kpktgend_0 [openvswitch] [k] ovs_flow_extract 4.06% kpktgend_0 [kernel.kallsyms] [k] _raw_spin_lock 3.90% kpktgend_0 [kernel.kallsyms] [k] jhash2 [...] 0.67% kpktgend_0 [kernel.kallsyms] [k] netlink_lookup Signed-off-by:
-
Thomas Graf authored
Generic implementation of a resizable, scalable, concurrent hash table based on [0]. The implementation supports both, fixed size keys specified via an offset and length, or arbitrary keys via own hash and compare functions. Lookups are lockless and protected as RCU read side critical sections. Automatic growing/shrinking based on user configurable watermarks is available while allowing concurrent lookups to take place. Objects to be hashed must include a struct rhash_head. The reason for not using the existing struct hlist_head is that the expansion and shrinking will have two buckets point to a single entry which would lead in obscure reverse chaining behaviour. Code includes a boot selftest if CONFIG_TEST_RHASHTABLE is defined. [0] https://www.usenix.org/legacy/event/atc11/tech/final_files/Triplett.pdfSigned-off-by:
-
David S. Miller authored
Aaron Brown says: ==================== Intel Wired LAN Driver Updates This series contains updates to the i40e and i40evf drivers. Vasu adds FCOE support, build options and a documentation pointer to i40e. Shannon exposes a Firmware API request used to do register writes on the driver's behalf and disables local loopback on VMDQ VSI in order to stop the VEB from echoing the VMDQ packets back at the VSI. Ashish corrects the vf_id offset for virtchnl messages in the case of multiple PFs, removes support for vf unicast promiscuos mode to disallow VFs from receiving traffic intended for another VF, updates the vfr_stat state check to handle the existing and future mechanism and adds an adapter state check to prevent re-arming the watchdog timer after i40evf_remove has been called and the timer has been deleted. Serey fixes an issue where a guest OS would panic when removing the vf driver while the device is being reset due to an attempt to clean a non initialized mac_filter_list. Akeem makes a minor comment change. Jessie changes an instance of sprintf to snprintf that was missed when the driver was converted to use snprintf everywhere. Mitch plugs a few memory leaks. ==================== Signed-off-by: -
Serey Kong authored
Removing VF driver during device still in reset caused guest OS panic. in the i40evf_remove(), we're trying to clean mac_filter_list which has not been initialized since the device is still stuck at the reset. The change is to initialize the filter_list before setting any task. Change-ID: I8b59df7384416c7e6f2d264b598f447e1c2c92b0 Signed-off-by:
Serey Kong <serey.kong@intel.com> Tested-by:
Sibai Li <sibai.li@intel.com> Signed-off-by:
Aaron Brown <aaron.f.brown@intel.com> Signed-off-by:
-
Mitch Williams authored
If the driver is loaded and then unloaded before the interface is brought up, then it will allocate a MAC filter entry and never free it. To fix this, on unload, run through the mac filter list and free all the entries. We also do this during reset recovery when the driver cannot contact the PF and needs to shut down completely. Change-ID: I15fabd67eb4a1bfc57605a7db60d0b5d819839db Signed-off-by:
Mitch Williams <mitch.a.williams@intel.com> Tested-by:
-
Mitch Williams authored
Fix a memory leak. Driver was allocating memory for queue vectors on init but not freeing them on shutdown. These need to be freed at two different times: during module unload, and during reset recovery when the driver cannot contact the PF driver and needs to give up. Change-ID: I7c1d0157a776e960d4da432dfe309035aad7c670 Signed-off-by:
-
Ashish Shah authored
Add in an adapter state check to prevent re-arming watchdog timer after i40evf_remove has been called and timer has been deleted. Change-ID: I636ba7c6322be8cbf053231959f90c0a2d8d803a Signed-off-by:
Ashish Shah <ashish.n.shah@intel.com> Tested-by:
-
Ashish Shah authored
Previously defined state I40E_VFR_VFACTIVE uses bit 1 which is now set to "reserved." Update the state checks to also include I40E_VFR_COMPLETED. This change will allow the VF to work with both existing and future PFs. Change-ID: Ifd1d34f79f3b0ffd6d2550ee4dadc55825ff52f8 Signed-off-by:
-
Ashish Shah authored
Remove the ability of a VF to set unicast promiscuous mode. Considered to be a security risk to allow VFs to receive traffic intended for other VFs so don't allow it, simply ignore the flag. Also fix it to send the correct seid to aq for multicast promiscuous set. Change-ID: Icb9c49a281a8e9d3aeebf991ef1533ac82b84b14 Signed-off-by:
Jim Young <jamesx.m.young@intel.com> Signed-off-by:
-
Akeem G Abodunrin authored
Fixes comment for reset reason Change-ID: I6fda4fa292255e6eb0f874502b4d38d722149b10 Signed-off-by:
Akeem G Abodunrin <akeem.g.abodunrin@intel.com> Tested-by:
-
Jesse Brandeburg authored
The driver was converted to use snprintf everywhere but this one function. Just use snprintf, instead of sprintf. Also a small spelling correction in a comment. Change-ID: I59d45f94a52754c7b4cd6034df9a61d8132b7f77 Signed-off-by:
Jesse Brandeburg <jesse.brandeburg@intel.com> Tested-by:
-
Shannon Nelson authored
The local loopback should only be enabled for VSIs that are supporting cascaded VEBs or VEPA setups. This is not the case here, and we need to stop the VEB from echoing the VMDQ VSI packets back at the VSI. Change-ID: I9dfb6ac79db24d04360d7efde62d81e20abc5090 Signed-off-by:
Shannon Nelson <shannon.nelson@intel.com> Tested-by:
-
Ashish Shah authored
The vf_id needs to be offset by the vf_base_id from hw function capabilities for the case of multiple PFs. Change-ID: I20ca8621f98e9cdf98649380b8eeaa35db52677c Signed-off-by:
-
Shannon Nelson authored
Now that the HW registers are no longer in debug mode and many are locked down for writes, we need to expose the Firmware API request used to do writes on the driver's behalf. Change-ID: I09a05c4dc9ea0b24c00193faac34d7799eaa8496 Signed-off-by:
-
Vasu Dev authored
Adds newly added FCoE files to the build but only if FCoE module is configured. Also, updates i40e document for added FCoE support. Signed-off-by:
Vasu Dev <vasu.dev@intel.com> Tested-by: Jack Morgan<jack.morgan@intel.com> Signed-off-by:
-
Vasu Dev authored
Adds FCoE specific code to existing i40e core driver to:- 1. have separate FCoE VSI with additional FCoE queues pairs. 2. have FCoE related hash defines. 3. have additional FCoE related stats code. 4. export and then re-use existing functions required by FCoE build. Signed-off-by:
-
Vasu Dev authored
This patch adds FCoE ( Fibre Channel Over Ethernet ) code for Intel XL710 adapters. This patch is limited to only new FCoE offloads code in newly added files by this patch and then following patches in the series modifies rest of the existing driver to enable FCoE with i40e driver. Signed-off-by:
-
David S. Miller authored
Tom Lendacky says: ==================== amd-xgbe: AMD XGBE driver update 2014-08-01 The following series of patches includes minor fixes/updates to the driver. - Remove some uses of spinlock around ethtool/phylib areas - Update Rx/Tx ready check logic ==================== Signed-off-by: -
Lendacky, Thomas authored
The current time range waiting for Rx/Tx to become ready can sometimes be too short if a connection is not present. Increase the number of retries and the sleep to give a bit more time. Also, change level of the message issued from _err to _dbg if Rx/Tx do not become ready since the underlying logic will function as if no link is established and retry eventually. Signed-off-by:
Tom Lendacky <thomas.lendacky@amd.com> Signed-off-by:
-
Lendacky, Thomas authored
Remove the spinlocks around the ethtool get and set settings functions and within the link adjustment callback routine. Signed-off-by:
-
- 02 Aug, 2014 18 commits
-
-
Himangi Saraogi authored
This patch introduces the use of managed interfaces like devm_ioremap_resource and does away with the calls to free the allocated memory in the probe and remove functions. Also, some labels and variable are done away with. This fixes a bug as there was a missing release_mem_region in the remove function. Signed-off-by:
Himangi Saraogi <himangi774@gmail.com> Acked-by:
Julia Lawall <julia.lawall@lip6.fr> Signed-off-by:
-
Himangi Saraogi authored
This patch introduces the use of devm_ioremap_resource, devm_kmalloc and does away with the functions to free the allocated memory in the probe and remove functions. Also, some labels are done away with. A bug is fixed as two regions are allocated in the probe function, but only one is freed in the remove function. Signed-off-by:
-
Himangi Saraogi authored
This patch introduces the use of managed interfaces like devm_ioremap_resource and does away with the functions to free the allocated memory in the probe and remove functions. Also, many labels are done away with. The field size in no longer needed and is hence removed from the struct net_local. Signed-off-by:
-
Hisashi Nakamura authored
Signed-off-by:
Hisashi Nakamura <hisashi.nakamura.ak@renesas.com> [uli: added bindings documentation] Signed-off-by:
Ulrich Hecht <ulrich.hecht+renesas@gmail.com> Signed-off-by:
-
David S. Miller authored
Sathya Perla says: ==================== be2net: patch set Patch 1 fixes a regression caused by a previous commit on net-next. Old versions of BE3 FW may not support cmds to re-provision (and hence optimize) resources/queues in SR-IOV config. Do not treat this FW cmd failure as fatal and fail the function initialization. Instead, just enable SR-IOV with the resources provided by the FW. Patch 2 ignores a VF mac address setting if the new mac is already active on the VF. Patch 3 adds support to delete a FW-dump via ethtool on Lancer adapters. ==================== Signed-off-by: -
Kalesh AP authored
This patch adds support to delete an existing FW-dump in Lancer via ethtool. Initiating a new dump is not allowed if a FW dump is already present in the adapter. The existing dump has to be first explicitly deleted. Signed-off-by:
Kalesh AP <kalesh.purayil@emulex.com> Signed-off-by:
Sathya Perla <sathya.perla@emulex.com> Signed-off-by:
-
Vasundhara Volam authored
ndo_set_vf_mac() call may be issued for a mac-addr that is already active on a VF. If so, silently ignore the request. Signed-off-by:
Vasundhara Volam <vasundhara.volam@emulex.com> Signed-off-by:
-
Sathya Perla authored
Old versions of BE3 FW may not support cmds to re-provision (and hence optimize) resources/queues in SR-IOV config. Do not treat this FW cmd failure as fatal and fail the function initialization. Instead, just enable SR-IOV with the resources provided by the FW. Prior to the "create optimal number of queues on SR-IOV config" patch such failures were ignored. Fixes: bec84e6b ("create optimal number of queues on SR-IOV config") Reported-by:
Eduardo Habkost <ehabkost@redhat.com> Signed-off-by:
-
David S. Miller authored
Nikolay Aleksandrov says: ==================== inet: frags: cleanup and kmem_cache use This patchset does a couple of small cleanups in patches 1-5 and then in patch 06 it introduces the use of kmem_cache for allocation/freeing of inet_frag_queue+header objects. v2: Broke up patch 02 into 3 patches as David suggested Here are the results of a couple of netperf runs: netperf options: -l 30 -I95,5 -i 15,10 -m 64k - 10 gig before the patchset MIGRATED UDP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.155.1 () port 0 AF_INET : +/-2.500% @ 95% conf. Socket Message Elapsed Messages Size Size Time Okay Errors Throughput bytes bytes secs # # 10^6bits/sec 212992 64000 30.00 442466 0 7551.39 212992 30.00 439130 7494.45 - 10 gig after the patchset MIGRATED UDP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.155.1 () port 0 AF_INET : +/-2.500% @ 95% conf. Socket Message Elapsed Messages Size Size Time Okay Errors Throughput bytes bytes secs # # 10^6bits/sec 212992 64000 30.00 458846 0 7830.94 212992 30.00 457575 7809.25 - Virtio before the patchset MIGRATED UDP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.144.1 () port 0 AF_INET : +/-2.500% @ 95% conf. Socket Message Elapsed Messages Size Size Time Okay Errors Throughput bytes bytes secs # # 10^6bits/sec 212992 64000 30.00 735000 0 12543.96 212992 30.00 560322 9562.79 - Virtio after the patchset MIGRATED UDP STREAM TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 192.168.144.1 () port 0 AF_INET : +/-2.500% @ 95% conf. Socket Message Elapsed Messages Size Size Time Okay Errors Throughput bytes bytes secs # # 10^6bits/sec 212992 64000 30.00 731729 0 12488.14 212992 30.00 647241 11046.21 ==================== Signed-off-by: -
Nikolay Aleksandrov authored
Use kmem_cache to allocate/free inet_frag_queue objects since they're all the same size per inet_frags user and are alloced/freed in high volumes thus making it a perfect case for kmem_cache. Signed-off-by:
Florian Westphal <fw@strlen.de> Signed-off-by:
-
Nikolay Aleksandrov authored
Now that we have INET_FRAG_EVICTED we might as well use it to stop sending icmp messages in the "frag_expire" functions instead of stripping INET_FRAG_FIRST_IN from their flags when evicting. Also fix the comment style in ip6_expire_frag_queue(). Signed-off-by:
-
Nikolay Aleksandrov authored
Fix a couple of functions' declaration alignments. Signed-off-by:
-
Nikolay Aleksandrov authored
Move the flags to an enum definion, swap FIRST_IN/LAST_IN to be in increasing order and add comments explaining each flag and the inet_frag_queue struct members. Signed-off-by:
-
Nikolay Aleksandrov authored
The last_in field has been used to store various flags different from first/last frag in so give it a more descriptive name: flags. Signed-off-by:
-
Nikolay Aleksandrov authored
Softirqs are already disabled so no need to do it again, thus let's be consistent and use the IP6_INC_STATS_BH variant. Signed-off-by:
Hannes Frederic Sowa <hannes@stressinduktion.org> Signed-off-by:
-
Dan Carpenter authored
The dm9000_release_board() function is called with NULL ->data_req and ->addr_req pointers if dm9000_probe() fails. Signed-off-by:
Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by:
-
Duan Jiong authored
ip_local_deliver_finish() already have a rcu_read_lock/unlock, so the rcu_read_lock/unlock is unnecessary. See the stack below: ip_local_deliver_finish | | ->icmp_rcv | | ->icmp_socket_deliver Suggested-by:
Duan Jiong <duanj.fnst@cn.fujitsu.com> Acked-by:
-
David S. Miller authored
Alexei Starovoitov says: ==================== net: filter: split sk_filter into socket and bpf, cleanup names The main goal of the series is to split 'struct sk_filter' into socket and bpf parts and cleanup names in the following way: - everything that deals with sockets keeps 'sk_*' prefix - everything that is pure BPF is changed to 'bpf_*' prefix split 'struct sk_filter' into struct sk_filter { atomic_t refcnt; struct rcu_head rcu; struct bpf_prog *prog; }; and struct bpf_prog { u32 jited:1, len:31; struct sock_fprog_kern *orig_prog; unsigned int (*bpf_func)(const struct sk_buff *skb, const struct bpf_insn *filter); union { struct sock_filter insns[0]; struct bpf_insn insnsi[0]; struct work_struct work; }; }; so that 'struct bpf_prog' can be used independent of sockets and cleans up 'unattached' bpf use cases: isdn, ppp, team, seccomp, ptp, xt_bpf, cls_bpf, test_bpf which don't need refcnt/rcu fields. It's a follow up to the rcu cleanup started by Pablo in commit 34c5bd66 ("net: filter: don't release unattached filter through call_rcu()") Patch 1 - cleans up socket memory charging and makes it possible for functions sk(bpf)_migrate_filter(), sk(bpf)_prepare_filter() to be socket independent Patches 2-4 - trivial renames Patch 5 - sk_filter split and renames of related sk_*() functions ==================== Signed-off-by:
-
