- 13 Feb, 2014 40 commits
-
-
Sergei Shtylyov authored
The driver #define's and uses ETHERSMALL macro for the minimum Ethernet frame size for which we have a standard macro ETH_ZLEN. Use the latter instead of the home-grown one. Signed-off-by:
Sergei Shtylyov <sergei.shtylyov@cogentembedded.com> Signed-off-by:
David S. Miller <davem@davemloft.net>
-
David S. Miller authored
Jon Maloy says: ==================== tipc: clean up media and bearer layer This commit series aims at facilitating future changes to the locking policy around nodes, links and bearers. Currently, we have a big read/write lock (net_lock) that is used for serializing all changes to the node, link and bearer lists, as well as to their mutual pointers and references. But, in order to allow for concurrent access to the contents of these structures, net_lock is only used in read mode by the data path code, and hence a finer granular locking policy must be applied inside the scope of net_lock: a spinlock (node_lock) for each node structure, and another one (bearer_lock) for protection of bearer structures. This locking policy has proved hard to maintain. We have several times encountered contention problems between node_lock and bearer_lock, and with the advent of the RCU locking mechanism we feel it is anyway obsolete and ripe for improvements. We now plan to replace net_lock with an RCU lock, as well as getting rid of bearer_lock altogether. This will both reduce data path overhead and make the code more manageable, while reducing the risk of future lock contention problems. Prior to these changes, we need to do some necessary cleanup and code consolidation. This is what we do with this commit series, before we finally remove bearer_lock. In a later series we will replace net_lock with an RCU lock. v2: - Re-inserted a removed kerneldoc entry in commit#5, based on feedback from D. Miller. ==================== Signed-off-by: -
Jon Paul Maloy authored
In an earlier commit, ("tipc: remove links list from bearer struct") we described three issues that need to be pre-emptively resolved before we can remove tipc_net_lock. Here we resolve issue a) described in that commit: "a) In access method #2, we access the link before taking the protecting node_lock. This will not work once net_lock is gone, so we will have to change the access order. We will deal with this in a later commit in this series." Here, we change that access order, by ensuring that the function link_find_link() returns only a safe reference for finding the link, i.e., a node pointer and an index into its 'links' array, not the link pointer itself. We also change all callers of this function to first take the node lock before they can check if there still is a valid link pointer at the returned index. Since the function now returns a node pointer rather than a link pointer, we rename it to the more appropriate 'tipc_link_find_owner(). Signed-off-by:Jon Maloy <jon.maloy@ericsson.com> Reviewed-by:
Ying Xue <ying.xue@windriver.com> Signed-off-by:
-
Ying Xue authored
After the earlier commits ("tipc: remove 'links' list from tipc_bearer struct") and ("tipc: introduce new spinlock to protect struct link_req"), there is no longer any need to protect struct link_req or or any link list by use of bearer_lock. Furthermore, we have eliminated the need for using bearer_lock during downcalls (send) from the link to the bearer, since we have ensured that bearers always have a longer life cycle that their associated links, and always contain valid data. So, the only need now for a lock protecting bearers is for guaranteeing consistency of the bearer list itself. For this, it is sufficient, at least for the time being, to continue applying 'net_lock´ in write mode. By removing bearer_lock we also pre-empt introduction of issue b) descibed in the previous commit "tipc: remove 'links' list from tipc_bearer struct": "b) When the outer protection from net_lock is gone, taking bearer_lock and node_lock in opposite order of method 1) and 2) will become an obvious deadlock hazard". Therefore, we now eliminate the bearer_lock spinlock. Signed-off-by:Paul Gortmaker <paul.gortmaker@windriver.com> Signed-off-by:
-
Jon Paul Maloy authored
When a bearer is disabled, all its attached links are deleted. Ideally, we should do link failover to redundant links on other bearers, if there are any, in such cases. This would be consistent with current behavior when a link is reset, but not deleted. However, due to the complexity involved, and the (wrongly) perceived low demand for this feature, it was never implemented until now. We mark the doomed link for deletion with a new flag, but wait until the failover process is finished before we actually delete it. With the improved link tunnelling/failover code introduced earlier in this commit series, it is now easy to identify a spot in the code where the failover is finished and it is safe to delete the marked link. Moreover, the test for the flag and the deletion can be done synchronously, and outside the most time critical data path. Signed-off-by:
-
Jon Paul Maloy authored
We change the order of checking for destination users when processing incoming packets. By placing the checks for users that may potentially replace the processed buffer, i.e., CHANGEOVER_PROTOCOL and MSG_FRAGMENTER, in a separate step before we check for the true end users, we get rid of a label and a 'goto', at the same time making the code more comprehensible and easy to follow. This commit does not change any functionality, it is just a cosmetic code reshuffle. Signed-off-by:
-
Jon Paul Maloy authored
After the previous redesign of the tunnel reception algorithm and functions, we finalize it by renaming a couple of stack variables in tipc_tunnel_rcv(). This makes it more consistent with the naming scheme elsewhere in this part of the code. This change is purely cosmetic, with no functional changes. Signed-off-by:
-
Jon Paul Maloy authored
We simplify and slim down the code in function tipc_tunnel_rcv() No impact on the users of this function. Signed-off-by:
-
Jon Paul Maloy authored
After the earlier commits in this series related to the function tipc_link_tunnel_rcv(), we can now go further and simplify its signature. The function now consumes all DUPLICATE packets, and only returns such ORIGINAL packets that are ready for immediate delivery, i.e., no more link level protocol processing needs to be done by the caller. As a consequence, the the caller, tipc_rcv(), does not access the link pointer after call return, and it becomes unnecessary to pass a link pointer reference in the call. Instead, we now only pass it the tunnel link's owner node, which is sufficient to find the destination link for the tunnelled packet. Signed-off-by:
-
Jon Paul Maloy authored
When a link is reset, and there is a redundant link available, all sender sockets will steer their subsequent traffic through the remaining link. In order to guarantee preserved packet order and cardinality during the transition, we tunnel the failing link's send queue through the remaining link before we allow any sockets to use it. In this commit, we change the algorithm for receiving failover ("ORIGINAL_MSG") packets in tipc_link_tunnel_rcv(), at the same time delegating it to a new subfuncton, tipc_link_failover_rcv(). Instead of directly returning an extracted inner packet to the packet reception loop in tipc_rcv(), we first check if it is a message fragment, in which case we append it to the reset link's fragment chain. If the fragment chain is complete, we return the whole chain instead of the individual buffer, eliminating any need for the tipc_rcv() loop to do reassembly of tunneled packets. This change makes it possible to further simplify tipc_link_tunnel_rcv(), as well as the calling tipc_rcv() loop. We will do that in later commits. It also makes it possible to identify a single spot in the code where we can tell that a failover procedure is finished, something that is useful when we are deleting links after a failover. This will also be done in a later commit. Signed-off-by: -
Jon Paul Maloy authored
When a second link to a destination comes up, some sender sockets will steer their subsequent traffic through the new link. In order to guarantee preserved packet order and cardinality for those sockets, we tunnel a duplicate of the old link's send queue through the new link before we open it for regular traffic. The last arriving packet copy, on whichever link, will be dropped at the receiving end based on the original sequence number, to ensure that only one copy is delivered to the end receiver. In this commit, we change the algorithm for receiving DUPLICATE_MSG packets, at the same time delegating it to a new subfunction, tipc_link_dup_rcv(). Instead of returning an extracted inner packet to the packet reception loop in tipc_rcv(), we just add it to the receiving (new) link's deferred packet queue. The packet will then be processed by that link when it receives its first non-tunneled packet, i.e., at latest when the changeover procedure is finished. Because tipc_link_tunnel_rcv()/tipc_link_dup_rcv() now is consuming all packets of type DUPLICATE_MSG, the calling tipc_rcv() function can omit testing for this. This in turn means that the current conditional jump to the label 'protocol_check' becomes redundant, and we can remove that label. Signed-off-by:
-
Ying Xue authored
In our ongoing effort to simplify the TIPC locking structure, we see a need to remove the linked list for tipc_links in the bearer. This can be explained as follows. Currently, we have three different ways to access a link, via three different lists/tables: 1: Via a node hash table: Used by the time-critical outgoing/incoming data paths. (e.g. link_send_sections_fast() and tipc_recv_msg() ): grab net_lock(read) find node from node hash table grab node_lock select link grab bearer_lock send_msg() release bearer_lock release node lock release net_lock 2: Via a global linked list for nodes: Used by configuration commands (link_cmd_set_value()) grab net_lock(read) find node and link from global node list (using link name) grab node_lock update link release node lock release net_lock (Same locking order as above. No problem.) 3: Via the bearer's linked link list: Used by notifications from interface (e.g. tipc_disable_bearer() ) grab net_lock(write) grab bearer_lock get link ptr from bearer's link list get node from link grab node_lock delete link release node lock release bearer_lock release net_lock (Different order from above, but works because we grab the outer net_lock in write mode first, excluding all other access.) The first major goal in our simplification effort is to get rid of the "big" net_lock, replacing it with rcu-locks when accessing the node list and node hash array. This will come in a later patch series. But to get there we first need to rewrite access methods ##2 and 3, since removal of net_lock would introduce three major problems: a) In access method #2, we access the link before taking the protecting node_lock. This will not work once net_lock is gone, so we will have to change the access order. We will deal with this in a later commit in this series, "tipc: add node lock protection to link found by link_find_link()". b) When the outer protection from net_lock is gone, taking bearer_lock and node_lock in opposite order of method 1) and 2) will become an obvious deadlock hazard. This is fixed in the commit ("tipc: remove bearer_lock from tipc_bearer struct") later in this series. c) Similar to what is described in problem a), access method #3 starts with using a link pointer that is unprotected by node_lock, in order to via that pointer find the correct node struct and lock it. Before we remove net_lock, this access order must be altered. This is what we do with this commit. We can avoid introducing problem problem c) by even here using the global node list to find the node, before accessing its links. When we loop though the node list we use the own bearer identity as search criteria, thus easily finding the links that are associated to the resetting/disabling bearer. It should be noted that although this method is somewhat slower than the current list traversal, it is in no way time critical. This is only about resetting or deleting links, something that must be considered relatively infrequent events. As a bonus, we can get rid of the mutual pointers between links and bearers. After this commit, pointer dependency go in one direction only: from the link to the bearer. This commit pre-empts introduction of problem c) as described above. Signed-off-by: -
Ying Xue authored
Currently, the 'started' field in struct tipc_link represents only a binary state, 'started' or 'not started'. We need it to represent more link execution states in the coming commits in this series. Hence, we rename the field to 'flags', and define the current started/non-started state to be represented by the LSB bit of that field. Signed-off-by:
-
Ying Xue authored
We break out the code for deleting attached links in the function bearer_disable(), and define a new function named tipc_link_delete_list() to do this job. This commit incurs no functional changes, but makes the code of function bearer_disable() cleaner. It is also a preparation for a more important change to the bearer code, in a subsequent commit in this series. Signed-off-by:
-
Ying Xue authored
We break out the code for resetting attached links in the function tipc_reset_bearer(), and define a new function named tipc_link_reset_list() to do this job. This commit incurs no functional changes, but makes the code of function tipc_reset_bearer() cleaner. It is also a preparation for a more important change to the bearer code, in a subsequent commit in this series. Signed-off-by:
-
Jon Paul Maloy authored
The function tipc_link_recv_fragment(struct sk_buff **buf) currently leaves the value of the input buffer pointer undefined when it returns, except when the return code indicates that the reassembly is complete. This despite the fact that it always consumes the input buffer. Here, we enforce a stricter behavior by this function, ensuring that the returned buffer pointer is non-NULL if and only if the reassembly is complete. This makes it possible to test for the buffer pointer as criteria for successful reassembly. We also rename the function to tipc_link_frag_rcv(), which is both shorter and more in line with common naming practice in the network subsystem. Apart from the new name, these changes have no impact on current users of the function, but makes it more practical for use in some planned future commits. Signed-off-by:
-
Andreas Bofjäll authored
The inline functions in addr.h uses tipc_own_addr which is exported by core.h, but addr.h never actually includes it. It works because it is explicitly included where this is used, but it looks a bit strange. Include core.h in addr.h explicitly to make the dependency clearer. Signed-off-by:
Andreas Bofjäll <andreas.bofjall@ericsson.com> Reviewed-by:
-
dingtianhong authored
The dev_valid_name() will check the buffer length for input name, no need to check it twice. Cc: Jay Vosburgh <fubar@us.ibm.com> Cc: Veaceslav Falico <vfalico@redhat.com> Cc: Andy Gospodarek <andy@greyhouse.net> Signed-off-by:
Ding Tianhong <dingtianhong@huawei.com> Reviewed-by:
Jiri Pirko <jiri@resnulli.us> Signed-off-by:
-
dingtianhong authored
The dev_set_mac_address() will check the dev->netdev_ops->ndo_set_mac_address, so no need to check it in bond_set_mac_address(). Cc: Jay Vosburgh <fubar@us.ibm.com> Cc: Veaceslav Falico <vfalico@redhat.com> Cc: Andy Gospodarek <andy@greyhouse.net> Signed-off-by:
-
WANG Cong authored
We are trying to mirror the local traffic from lo to eth0, allowing setting mac address of lo to eth0 would make the ether addresses in these packets correct, so that we don't have to modify the ether header again. Since usually no one cares about its mac address (all-zero), it is safe to allow those who care to set its mac address. Cc: Hannes Frederic Sowa <hannes@stressinduktion.org> Cc: Neil Horman <nhorman@tuxdriver.com> Cc: Stephen Hemminger <stephen@networkplumber.org> Cc: Eric Dumazet <edumazet@google.com> Cc: David S. Miller <davem@davemloft.net> Signed-off-by:
Cong Wang <xiyou.wangcong@gmail.com> Acked-by:
Neil Horman <nhorman@tuxdriver.com> Signed-off-by:
-
David S. Miller authored
Aaron Brown says: ==================== Intel Wired LAN Driver Updates This series contains updates to i40e and i40evf, primarily reset handling / refactoring along with a fair amount of minor cleanup. Jesse fixes some spelling, bumps the version and other trivial fixes. Akeem sets a bit that is needed before shutdown in the case of tx_timeout recovery failure. Mitch refactors reset handling along with a whole bunch of clean up. ==================== Signed-off-by: -
Mitch Williams authored
Respond better to a VF reset event. When a reset is signaled by the PF, or detected by the watchdog task, prevent the watchdog from processing admin queue requests, and schedule the reset task. In the reset task, wait first for the reset to start, then for it to complete, then reinit the driver. If the reset never appears to complete after a long, long time (>10 seconds is possible depending on what's going on with the PF driver), then set a flag to indicate that PF communications have failed. If this flag is set, check for the reset to complete in the watchdog, and attempt to do a full reinitialization of the driver from scratch. With these changes the VF driver correctly handles a PF reset event while running on bare metal, or in a VM. Also update copyrights. Change-ID: I93513efd0b50523a8345e7f6a33a5e4f8a2a5996 Signed-off-by:
Mitch Williams <mitch.a.williams@intel.com> Signed-off-by:
Jesse Brandeburg <jesse.brandeburg@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
As pointed out by Dan Carpenter (from Oracle), the flags variable is declared as a 64-bit long but all of the flags are defined as u32, which may lead to unintended consequences. Fix this by declaring flags as u32 (since we don't need any more than about a dozen flags right now), and remove the volatile qualifier, since it's unnecessary and just makes checkpatch cry. Change-ID: I137d3bb1842bf7e9456b5929ca54e3b0ed45dcab Signed-off-by:
-
Mitch Williams authored
Since we store the traffic vector names in the queue vector struct, we don't need to maintain an array of strings for these names in the adapter structure. Replace this array with a single string and use it when allocating the misc irq vector. Also update copyrights. Change-ID: I664f096c3c008210d6a04a487163e8aa934fee5b Signed-off-by:
-
Mitch Williams authored
Locate the structure in the correct header file. Change-ID: Ic7853131728812093a44a75d6b70953311a48dab Signed-off-by:
-
Mitch Williams authored
Remove a bunch of unused structure members that are just wasting space. Remove a completely unused info structure definition as well. Also update copyrights. Change-ID: I028ab92d9b7bd13a832cf3363bd1dc6610d8a535 Signed-off-by:
-
Mitch Williams authored
Set the DOWN flag before attempting to disable VFs when unloading the driver. Also, don't attempt to reset the VFs when the driver is unloading, because the switch configuration will fail. This fixes a panic on unload when VFs are enabled. Change-ID: I25a6567e89c9687145f510ff4f630932412c5c5d Signed-off-by:
-
Mitch Williams authored
If VFs are present when the driver loads, then set up some resources so they can function. Change-ID: I485916a811609a9990ce663d06dc645f625b07ff Signed-off-by:
-
Mitch Williams authored
Reset all of the VFs after a PF reset, so that they are in a known state, and the VF driver can detect the reset and reinit itself. Change-ID: I93c5b3a0f8b1371d0da078f92de948b9d3a6413f Signed-off-by:
-
Mitch Williams authored
Without this, the VF can never communicate with the PF after a VF reset. Change-ID: I8d10f1d0d0638d50d39f0aff263422e05d83ad83 Signed-off-by:
-
Mitch Williams authored
This looks like a cut and paste error. The code makes no sense where it is, and accomplishes nothing. Since we've removed the goto, we can also get rid of the extraneous brackets. Change-ID: I9315e3eafeee0a5713c94b0dc57b58b60a849124 Signed-off-by:
-
Akeem G Abodunrin authored
If tx_timeout recovery failed, then it becomes necessary to set i40e_down bit before actually shutdown the connection. Change-ID: Iaac81df0e302116571827aa0cff450697fbb7fa3 Signed-off-by:
Akeem G Abodunrin <akeem.g.abodunrin@intel.com> Signed-off-by:
Kavindya Deegala <kavindya.s.deegala@intel.com> Signed-off-by:
-
Mitch Williams authored
As politely pointed out by Dave Miller, calls to memset do not need a void pointer cast. Additionally, it is preferred to use sizeof(*the actual object) instead of sizeof(type). Change-ID: Id6a02429b7040111531f3865ea03fbe619167cb3 Signed-off-by:
Kevin Scott <kevin.c.scott@intel.com> Signed-off-by:
-
Jesse Brandeburg authored
This change moves one operator up to the previous line and deletes the duplicate declaration of ETH_ALEN. Also update copyrights. Change-ID: I88de73093b584e0f3b29d481ccd83fc4b1a1afa5 Signed-off-by:
-
Jesse Brandeburg authored
Update the driver version to 0.3.31-k. Signed-off-by:
-
Jesse Brandeburg authored
Fix a spelling error, s/extention/extension/. Signed-off-by:
-
Sabrina Dubroca authored
IPX doesn't implement shutdown, which poses a problem to some users: https://bugzilla.kernel.org/show_bug.cgi?id=67841 This patch is heavily based on the shutdown implementation for unix sockets. Reported-by:
Bruno Jesus <00cpxxx@gmail.com> Signed-off-by:
Sabrina Dubroca <sd@queasysnail.net> Signed-off-by:
-
David S. Miller authored
Cong Wang says: ==================== net_sched: act: more cleanup and improvement v2 -> v3: * fix a mis-splitted patch * keep hinfo as a pointer in ops v1 -> v2: * Fix a bug noticed by Jamal * Drop patches already merged into net-next * Add patch 5/5 Patches are cleanup's for the structures of tc actions, except patch 4 which is an improvement. See each patch for details. ==================== Cc: Stephen Hemminger <stephen@networkplumber.org> Cc: Jamal Hadi Salim <jhs@mojatatu.com> Cc: David S. Miller <davem@davemloft.net> Signed-off-by:
-
WANG Cong authored
We could allocate tc_action on stack in tca_action_flush(), since it is not large. Also, we could use create_a() in tcf_action_get_1(). Cc: Jamal Hadi Salim <jhs@mojatatu.com> Cc: David S. Miller <davem@davemloft.net> Signed-off-by:
Jamal Hadi Salim <jhs@mojatatu.com> Signed-off-by:
-
WANG Cong authored
When an action is bonnd to a filter, there is no point to remove it outside. Currently we just silently decrease the refcnt, we should reject this explicitly with EPERM. Cc: Jamal Hadi Salim <jhs@mojatatu.com> Cc: David S. Miller <davem@davemloft.net> Signed-off-by:
-
