OpenE2K
/
linux
6
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity
linux/drivers/infiniband/ulp/ipoib/ipoib.h

764 lines
20 KiB
C
Raw Normal View History

Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/*
* Copyright (c) 2004, 2005 Topspin Communications. All rights reserved.
[PATCH] IB: Add copyright notices Make some lawyers happy and add copyright notices for people who forgot to include them when they actually touched the code. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-08-11 08:03:10 +02:00
* Copyright (c) 2005 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2004 Voltaire, Inc. All rights reserved.
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
*
* This software is available to you under a choice of one of two
* licenses. You may choose to be licensed under the terms of the GNU
* General Public License (GPL) Version 2, available from the file
* COPYING in the main directory of this source tree, or the
* OpenIB.org BSD license below:
*
* Redistribution and use in source and binary forms, with or
* without modification, are permitted provided that the following
* conditions are met:
*
* - Redistributions of source code must retain the above
* copyright notice, this list of conditions and the following
* disclaimer.
*
* - Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following
* disclaimer in the documentation and/or other materials
* provided with the distribution.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#ifndef _IPOIB_H
#define _IPOIB_H
#include <linux/list.h>
#include <linux/skbuff.h>
#include <linux/netdevice.h>
#include <linux/workqueue.h>
#include <linux/kref.h>
#include <linux/if_infiniband.h>
IB: convert from semaphores to mutexes semaphore to mutex conversion by Ingo and Arjan's script. Signed-off-by: Ingo Molnar <mingo@elte.hu> [ Sanity-checked on real IB hardware ] Signed-off-by: Roland Dreier <rolandd@cisco.com>
2006-01-13 23:51:39 +01:00
#include <linux/mutex.h>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
#include <net/neighbour.h>
IPoIB: Stop lying about hard_header_len and use skb->cb to stash LL addresses Commit a0417fa3a18a ("net: Make qdisc_skb_cb upper size bound explicit.") made it possible for a netdev driver to use skb->cb between its header_ops.create method and its .ndo_start_xmit method. Use this in ipoib_hard_header() to stash away the LL address (GID + QPN), instead of the "ipoib_pseudoheader" hack. This allows IPoIB to stop lying about its hard_header_len, which will let us fix the L2 check for GRO. Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-02-07 15:51:21 +01:00
#include <net/sch_generic.h>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
atomic: use <linux/atomic.h> This allows us to move duplicated code in <asm/atomic.h> (atomic_inc_not_zero() for now) to <linux/atomic.h> Signed-off-by: Arun Sharma <asharma@fb.com> Reviewed-by: Eric Dumazet <eric.dumazet@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: David Miller <davem@davemloft.net> Cc: Eric Dumazet <eric.dumazet@gmail.com> Acked-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-07-27 01:09:06 +02:00
#include <linux/atomic.h>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
[PATCH] IB: move include files to include/rdma Move the InfiniBand headers from drivers/infiniband/include to include/rdma. This allows InfiniBand-using code to live elsewhere, and lets us remove the ugly EXTRA_CFLAGS include path from the InfiniBand Makefiles. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-08-25 22:40:04 +02:00
#include <rdma/ib_verbs.h>
#include <rdma/ib_pack.h>
#include <rdma/ib_sa.h>
IPoIB: Remove LRO support As a first step in moving from LRO to GRO, revert commit af40da894e9 ("IPoIB: add LRO support"). Also eliminate the ethtool set_flags callback which isn't needed anymore. Finally, we need to include <linux/sched.h> directly to get the declaration of restart_syscall() (which used to be included implicitly through <linux/inet_lro.h>). Cc: Ben Hutchings <bhutchings@solarflare.com> Cc: Eric W. Biederman <ebiederm@xmission.com> Cc: Vladimir Sokolovsky <vlad@mellanox.co.il> Signed-off-by: Or Gerlitz <ogerlitz@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2011-01-11 02:41:54 +01:00
#include <linux/sched.h>
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/* constants */
IPoIB: Refresh paths instead of flushing them on SM change events The patch tries to solve the problem of device going down and paths being flushed on an SM change event. The method is to mark the paths as candidates for refresh (by setting the new valid flag to 0), and wait for an ARP probe a new path record query. The solution requires a different and less intrusive handling of SM change event. For that, the second argument of the flush function changes its meaning from a boolean flag to a level. In most cases, SM failover doesn't cause LID change so traffic won't stop. In the rare cases of LID change, the remote host (the one that hadn't changed its LID) will lose connectivity until paths are refreshed. This is no worse than the current state. In fact, preventing the device from going down saves packets that otherwise would be lost. Signed-off-by: Moni Levy <monil@voltaire.com> Signed-off-by: Moni Shoua <monis@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-07-15 08:48:49 +02:00
enum ipoib_flush_level {
IPOIB_FLUSH_LIGHT,
IPOIB_FLUSH_NORMAL,
IPOIB_FLUSH_HEAVY
};
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
enum {
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
IPOIB_ENCAP_LEN = 4,
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
IPoIB: Handle 4K IB MTU for UD (datagram) mode This patch enables IPoIB to use 4K UD messages (when the underlying device and fabrics support a 4K MTU) by using two scatter buffers when PAGE_SIZE is less than or equal to thhe HCA IB MTU size. The first buffer is for IPoIB header + GRH header, and the second buffer is the IPoIB payload, which is 4K-4. Signed-off-by: Shirley Ma <xma@us.ibm.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-04-23 20:55:45 +02:00
IPOIB_UD_HEAD_SIZE = IB_GRH_BYTES + IPOIB_ENCAP_LEN,
IPOIB_UD_RX_SG = 2, /* max buffer needed for 4K mtu */
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
IPOIB_CM_MTU = 0x10000 - 0x10, /* padding to align header to 16 */
IPOIB_CM_BUF_SIZE = IPOIB_CM_MTU + IPOIB_ENCAP_LEN,
IPOIB_CM_HEAD_SIZE = IPOIB_CM_BUF_SIZE % PAGE_SIZE,
IPOIB_CM_RX_SG = ALIGN(IPOIB_CM_BUF_SIZE, PAGE_SIZE) / PAGE_SIZE,
IPoIB: Double default RX/TX ring sizes Increase IPoIB ring sizes to twice their original sizes (RX: 128->256, TX: 64->128) to act as a shock absorber for high traffic peaks. With the current settings, we have seen cases that there are many calls to netif_stop_queue(), which causes degradation in throughput. Also, larger receive buffer sizes help IPoIB in CM mode to avoid experiencing RNR NAK conditions due to insufficient receive buffers at the SRQ. Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-07-15 08:48:52 +02:00
IPOIB_RX_RING_SIZE = 256,
IPOIB_TX_RING_SIZE = 128,
IPoIB: Make send and receive queue sizes tunable Make IPoIB's send and receive queue sizes tunable via module parameters ("send_queue_size" and "recv_queue_size"). This allows the queue sizes to be enlarged to fix disastrously bad performance on some platforms and workloads, without bloating memory usage when large queues aren't needed. Signed-off-by: Shirley Ma <xma@us.ibm.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2006-04-10 18:43:58 +02:00
IPOIB_MAX_QUEUE_SIZE = 8192,
IPOIB_MIN_QUEUE_SIZE = 2,
IPoIB/cm: Add connected mode support for devices without SRQs Some IB adapters (notably IBM's eHCA) do not implement SRQs (shared receive queues). The current IPoIB connected mode support only works on devices that support SRQs. Fix this by adding support for using the receive queue of each connected mode receive QP. The disadvantage of this compared to using an SRQ is that it means a full queue of receives must be posted for each remote connected mode peer, which means that total memory usage is potentially much higher than when using SRQs. To manage this, add a new module parameter "max_nonsrq_conn_qp" that limits the number of connections allowed per interface. The rest of the changes are fairly straightforward: we use a table of struct ipoib_cm_rx to hold all the active connections, and put the table index of the connection in the high bits of receive WR IDs. This is needed because we cannot rely on the struct ib_wc.qp field for non-SRQ receive completions. Most of the rest of the changes just test whether or not an SRQ is available, and post receives or find received packets in the right place depending on the answer. Cleaning up dead connections actually becomes simpler, because we do not have to do the "last WQE reached" dance that is required to destroy QPs attached to an SRQ. We just move the QP to the error state and wait for all pending receives to be flushed. Signed-off-by: Pradeep Satyanarayana <pradeeps@linux.vnet.ibm.com> [ Completely rewritten and split up, based on Pradeep's work. Several bugs fixed and no doubt several bugs introduced. - Roland ] Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-01-25 23:15:24 +01:00
IPOIB_CM_MAX_CONN_QP = 4096,
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
IPOIB_NUM_WC = 4,
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
IPOIB_MAX_PATH_REC_QUEUE = 3,
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
IPOIB_MAX_MCAST_QUEUE = 3,
IPOIB_FLAG_OPER_UP = 0,
IPOIB_FLAG_INITIALIZED = 1,
IPOIB_FLAG_ADMIN_UP = 2,
IPOIB_PKEY_ASSIGNED = 3,
IPOIB_PKEY_STOP = 4,
IPOIB_FLAG_SUBINTERFACE = 5,
IPOIB_MCAST_RUN = 6,
IPOIB_STOP_REAPER = 7,
IPOIB_FLAG_ADMIN_CM = 9,
IPoIB/cm: Use common CQ for CM send completions Use the same CQ for CM send completions as for all other IPoIB completions. This means all completions are processed via the same NAPI polling routine. This should help reduce the number of interrupts for bi-directional traffic (such as TCP) and fixes "driver is hogging interrupts" errors reported for IPoIB send side, e.g. <https://bugs.openfabrics.org/show_bug.cgi?id=508> To do this, keep a per-interface counter of outstanding send WRs, and stop the interface when this counter reaches the send queue size to avoid CQ overruns. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-08-16 14:36:16 +02:00
IPOIB_FLAG_UMCAST = 10,
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
IPOIB_STOP_NEIGH_GC = 11,
IPOIB_NEIGH_TBL_FLUSH = 12,
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
IPOIB_MAX_BACKOFF_SECONDS = 16,
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
IPOIB_MCAST_FLAG_FOUND = 0, /* used in set_multicast_list */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
IPOIB_MCAST_FLAG_SENDONLY = 1,
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
IPOIB_MCAST_FLAG_BUSY = 2, /* joining or already joined */
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
IPOIB_MCAST_FLAG_ATTACHED = 3,
IPoIB: Use separate CQ for UD send completions Use a dedicated CQ for UD send completions. Also, do not arm the UD send CQ, which reduces the number of interrupts generated. This patch farther reduces overhead by not calling poll CQ for every posted send WR -- it does polls only when there 16 or more outstanding work requests. Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-04-29 22:46:53 +02:00
MAX_SEND_CQE = 16,
IPoIB: Copy small received SKBs in connected mode The connected mode implementation in the IPoIB driver has a large overhead in the way SKBs are handled in the receive flow. It usually allocates an SKB with as big as was used in the currently received SKB and moves unused fragments from the old SKB to the new one. This involves a loop on all the remaining fragments and incurs overhead on the CPU. This patch, for small SKBs, allocates an SKB just large enough to contain the received data and copies to it the data from the received SKB. The newly allocated SKB is passed to the stack and the old SKB is reposted. When running netperf, UDP small messages, without this pach I get: UDP UNIDIRECTIONAL SEND TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to 14.4.3.178 (14.4.3.178) port 0 AF_INET Socket Message Elapsed Messages Size Size Time Okay Errors Throughput bytes bytes secs # # 10^6bits/sec 114688 128 10.00 5142034 0 526.31 114688 10.00 1130489 115.71 With this patch I get both send and receive at ~315 mbps. The reason that send performance actually slows down is as follows: When using this patch, the overhead of the CPU for handling RX packets is dramatically reduced. As a result, we do not experience RNR NAK messages from the receiver which cause the connection to be closed and reopened again; when the patch is not used, the receiver cannot handle the packets fast enough so there is less time to post new buffers and hence the mentioned RNR NACKs. So what happens is that the application *thinks* it posted a certain number of packets for transmission but these packets are flushed and do not really get transmitted. Since the connection gets opened and closed many times, each time netperf gets the CPU time that otherwise would have been given to IPoIB to actually transmit the packets. This can be verified when looking at the port counters -- the output of ifconfig and the oputput of netperf (this is for the case without the patch): tx packets ========== port counter: 1,543,996 ifconfig: 1,581,426 netperf: 5,142,034 rx packets ========== netperf 1,1304,089 Signed-off-by: Eli Cohen <eli@mellanox.co.il>
2008-07-15 08:48:44 +02:00
IPOIB_CM_COPYBREAK = 256,
IB/ipoib: Add rtnl_link_ops support Add rtnl_link_ops to IPoIB, with the first usage being child device create/delete through them. Childs devices are now either legacy ones, created/deleted through the ipoib sysfs entries, or RTNL ones. Adding support for RTNL childs involved refactoring of ipoib_vlan_add which is now used by both the sysfs and the link_ops code. Also, added ndo_uninit entry to support calling unregister_netdevice_queue from the rtnl dellink entry. This required removal of calls to ipoib_dev_cleanup from the driver in flows which use unregister_netdevice, since the networking core will invoke ipoib_uninit which does exactly that. Signed-off-by: Erez Shitrit <erezsh@mellanox.co.il> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-13 07:56:36 +02:00
IPOIB_NON_CHILD = 0,
IPOIB_LEGACY_CHILD = 1,
IPOIB_RTNL_CHILD = 2,
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
};
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
#define IPOIB_OP_RECV (1ul << 31)
#ifdef CONFIG_INFINIBAND_IPOIB_CM
IPoIB/cm: Use common CQ for CM send completions Use the same CQ for CM send completions as for all other IPoIB completions. This means all completions are processed via the same NAPI polling routine. This should help reduce the number of interrupts for bi-directional traffic (such as TCP) and fixes "driver is hogging interrupts" errors reported for IPoIB send side, e.g. <https://bugs.openfabrics.org/show_bug.cgi?id=508> To do this, keep a per-interface counter of outstanding send WRs, and stop the interface when this counter reaches the send queue size to avoid CQ overruns. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-08-16 14:36:16 +02:00
#define IPOIB_OP_CM (1ul << 30)
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
#else
IPoIB/cm: Use common CQ for CM send completions Use the same CQ for CM send completions as for all other IPoIB completions. This means all completions are processed via the same NAPI polling routine. This should help reduce the number of interrupts for bi-directional traffic (such as TCP) and fixes "driver is hogging interrupts" errors reported for IPoIB send side, e.g. <https://bugs.openfabrics.org/show_bug.cgi?id=508> To do this, keep a per-interface counter of outstanding send WRs, and stop the interface when this counter reaches the send queue size to avoid CQ overruns. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-08-16 14:36:16 +02:00
#define IPOIB_OP_CM (0)
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
#endif
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/* structs */
struct ipoib_header {
[PATCH] IB: sparse endianness cleanup Fix sparse warnings. Use __be* where appropriate. Signed-off-by: Sean Hefty <sean.hefty@intel.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-08-14 06:05:57 +02:00
__be16 proto;
u16 reserved;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
};
IPoIB: Stop lying about hard_header_len and use skb->cb to stash LL addresses Commit a0417fa3a18a ("net: Make qdisc_skb_cb upper size bound explicit.") made it possible for a netdev driver to use skb->cb between its header_ops.create method and its .ndo_start_xmit method. Use this in ipoib_hard_header() to stash away the LL address (GID + QPN), instead of the "ipoib_pseudoheader" hack. This allows IPoIB to stop lying about its hard_header_len, which will let us fix the L2 check for GRO. Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-02-07 15:51:21 +01:00
struct ipoib_cb {
struct qdisc_skb_cb qdisc_cb;
u8 hwaddr[INFINIBAND_ALEN];
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
};
IPoIB: Specify Traffic Class with path record queries for QoS support To support QoS within and between subnets, modify IPoIB to request specific Traffic Class values with path record queries, using the value associated with the IPoIB broadcast group. Signed-off-by: Sean Hefty <sean.hefty@intel.com> [ See some comments I made on this at v1 and v2 of the posts <http://lists.openfabrics.org/pipermail/general/2007-August/039275.html> <http://lists.openfabrics.org/pipermail/general/2007-September/040312.html> ] Reviewed-by: Or Gerlitz <ogerlitz@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-08-02 21:21:31 +02:00
/* Used for all multicast joins (broadcast, IPv4 mcast and IPv6 mcast) */
struct ipoib_mcast {
struct ib_sa_mcmember_rec mcmember;
struct ib_sa_multicast *mc;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
struct ipoib_ah *ah;
IPoIB: Specify Traffic Class with path record queries for QoS support To support QoS within and between subnets, modify IPoIB to request specific Traffic Class values with path record queries, using the value associated with the IPoIB broadcast group. Signed-off-by: Sean Hefty <sean.hefty@intel.com> [ See some comments I made on this at v1 and v2 of the posts <http://lists.openfabrics.org/pipermail/general/2007-August/039275.html> <http://lists.openfabrics.org/pipermail/general/2007-September/040312.html> ] Reviewed-by: Or Gerlitz <ogerlitz@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-08-02 21:21:31 +02:00
struct rb_node rb_node;
struct list_head list;
unsigned long created;
unsigned long backoff;
unsigned long flags;
unsigned char logcount;
struct list_head neigh_list;
struct sk_buff_head pkt_queue;
struct net_device *dev;
};
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
[IPoIB] Drop RX packets when out of memory Change the way IPoIB handles RX packets when it can't allocate a new receive skbuff. If the allocation of a new receive skb fails, we now drop the packet we just received and repost the original receive skb. This means that the receive ring always stays full and we don't have to monkey around with trying to schedule a refill task for later. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-10-29 00:30:34 +02:00
struct ipoib_rx_buf {
struct sk_buff *skb;
IPoIB: Handle 4K IB MTU for UD (datagram) mode This patch enables IPoIB to use 4K UD messages (when the underlying device and fabrics support a 4K MTU) by using two scatter buffers when PAGE_SIZE is less than or equal to thhe HCA IB MTU size. The first buffer is for IPoIB header + GRH header, and the second buffer is the IPoIB payload, which is 4K-4. Signed-off-by: Shirley Ma <xma@us.ibm.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-04-23 20:55:45 +02:00
u64 mapping[IPOIB_UD_RX_SG];
[IPoIB] Drop RX packets when out of memory Change the way IPoIB handles RX packets when it can't allocate a new receive skbuff. If the allocation of a new receive skb fails, we now drop the packet we just received and repost the original receive skb. This means that the receive ring always stays full and we don't have to monkey around with trying to schedule a refill task for later. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-10-29 00:30:34 +02:00
};
struct ipoib_tx_buf {
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
struct sk_buff *skb;
IPoIB: Add send gather support This patch acts as a preparation for using checksum offload for IB devices capable of inserting/verifying checksum in IP packets. The patch does not actaully turn on NETIF_F_SG - we defer that to the patches adding checksum offload capabilities. We only add support for send gathers for datagram mode, since existing HW does not support checksum offload on connected QPs. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-01-30 17:30:53 +01:00
u64 mapping[MAX_SKB_FRAGS + 1];
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
};
IPoIB/cm: Reduce connected mode TX object size Since IPoIB connected mode does not NETIF_F_SG, we only have one DMA mapping per send, so we don't need a mapping[] array. Define a new struct with a single u64 mapping member and use it for the CM tx_ring. Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-07-15 08:48:52 +02:00
struct ipoib_cm_tx_buf {
struct sk_buff *skb;
u64 mapping;
};
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
struct ib_cm_id;
struct ipoib_cm_data {
__be32 qpn; /* High byte MUST be ignored on receive */
__be32 mtu;
};
IPoIB/cm: Fix SRQ WR leak SRQ WR leakage has been observed with IPoIB/CM: e.g. flipping ports on and off will, with time, leak out all WRs and then all connections will start getting RNR NAKs. Fix this in the way suggested by spec: move the QP being destroyed to the error state, wait for "Last WQE Reached" event and then post WR on a "drain QP" connected to the same CQ. Once we observe a completion on the drain QP, it's safe to call ib_destroy_qp. Signed-off-by: Michael S. Tsirkin <mst@dev.mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-05-21 14:04:59 +02:00
/*
* Quoting 10.3.1 Queue Pair and EE Context States:
*
* Note, for QPs that are associated with an SRQ, the Consumer should take the
* QP through the Error State before invoking a Destroy QP or a Modify QP to the
* Reset State. The Consumer may invoke the Destroy QP without first performing
* a Modify QP to the Error State and waiting for the Affiliated Asynchronous
* Last WQE Reached Event. However, if the Consumer does not wait for the
* Affiliated Asynchronous Last WQE Reached Event, then WQE and Data Segment
* leakage may occur. Therefore, it is good programming practice to tear down a
* QP that is associated with an SRQ by using the following process:
*
* - Put the QP in the Error State
* - Wait for the Affiliated Asynchronous Last WQE Reached Event;
* - either:
* drain the CQ by invoking the Poll CQ verb and either wait for CQ
* to be empty or the number of Poll CQ operations has exceeded
* CQ capacity size;
* - or
* post another WR that completes on the same CQ and wait for this
* WR to return as a WC;
* - and then invoke a Destroy QP or Reset QP.
*
* We use the second option and wait for a completion on the
IPoIB/cm: Fix performance regression on Mellanox commit 518b1646 ("IPoIB/cm: Fix SRQ WR leak") introduced a severe performance regression on Mellanox cards, because keeping a QP in the error state for extended periods of time moves hardware to the slow path (until the QP is destroyed). For example, MPI latency goes from ~3 usecs to ~7 usecs. Fix this by posting a send WR on one of the QPs that are being flushed, instead of using a separate drain QP that is kept in the error state. This fixes bug <https://bugs.openfabrics.org/show_bug.cgi?id=636>, reported and bisected by Scott Weitzenkamp at Cisco and debugged by Sasha Mikheev at Voltaire. Signed-off-by: Michael S. Tsirkin <mst@dev.mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-05-28 13:37:27 +02:00
* same CQ before destroying QPs attached to our SRQ.
IPoIB/cm: Fix SRQ WR leak SRQ WR leakage has been observed with IPoIB/CM: e.g. flipping ports on and off will, with time, leak out all WRs and then all connections will start getting RNR NAKs. Fix this in the way suggested by spec: move the QP being destroyed to the error state, wait for "Last WQE Reached" event and then post WR on a "drain QP" connected to the same CQ. Once we observe a completion on the drain QP, it's safe to call ib_destroy_qp. Signed-off-by: Michael S. Tsirkin <mst@dev.mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-05-21 14:04:59 +02:00
*/
enum ipoib_cm_state {
IPOIB_CM_RX_LIVE,
IPOIB_CM_RX_ERROR, /* Ignored by stale task */
IPOIB_CM_RX_FLUSH /* Last WQE Reached event observed */
};
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
struct ipoib_cm_rx {
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
struct ib_cm_id *id;
struct ib_qp *qp;
IPoIB/cm: Add connected mode support for devices without SRQs Some IB adapters (notably IBM's eHCA) do not implement SRQs (shared receive queues). The current IPoIB connected mode support only works on devices that support SRQs. Fix this by adding support for using the receive queue of each connected mode receive QP. The disadvantage of this compared to using an SRQ is that it means a full queue of receives must be posted for each remote connected mode peer, which means that total memory usage is potentially much higher than when using SRQs. To manage this, add a new module parameter "max_nonsrq_conn_qp" that limits the number of connections allowed per interface. The rest of the changes are fairly straightforward: we use a table of struct ipoib_cm_rx to hold all the active connections, and put the table index of the connection in the high bits of receive WR IDs. This is needed because we cannot rely on the struct ib_wc.qp field for non-SRQ receive completions. Most of the rest of the changes just test whether or not an SRQ is available, and post receives or find received packets in the right place depending on the answer. Cleaning up dead connections actually becomes simpler, because we do not have to do the "last WQE reached" dance that is required to destroy QPs attached to an SRQ. We just move the QP to the error state and wait for all pending receives to be flushed. Signed-off-by: Pradeep Satyanarayana <pradeeps@linux.vnet.ibm.com> [ Completely rewritten and split up, based on Pradeep's work. Several bugs fixed and no doubt several bugs introduced. - Roland ] Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-01-25 23:15:24 +01:00
struct ipoib_cm_rx_buf *rx_ring;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
struct list_head list;
struct net_device *dev;
unsigned long jiffies;
enum ipoib_cm_state state;
IPoIB/cm: Add connected mode support for devices without SRQs Some IB adapters (notably IBM's eHCA) do not implement SRQs (shared receive queues). The current IPoIB connected mode support only works on devices that support SRQs. Fix this by adding support for using the receive queue of each connected mode receive QP. The disadvantage of this compared to using an SRQ is that it means a full queue of receives must be posted for each remote connected mode peer, which means that total memory usage is potentially much higher than when using SRQs. To manage this, add a new module parameter "max_nonsrq_conn_qp" that limits the number of connections allowed per interface. The rest of the changes are fairly straightforward: we use a table of struct ipoib_cm_rx to hold all the active connections, and put the table index of the connection in the high bits of receive WR IDs. This is needed because we cannot rely on the struct ib_wc.qp field for non-SRQ receive completions. Most of the rest of the changes just test whether or not an SRQ is available, and post receives or find received packets in the right place depending on the answer. Cleaning up dead connections actually becomes simpler, because we do not have to do the "last WQE reached" dance that is required to destroy QPs attached to an SRQ. We just move the QP to the error state and wait for all pending receives to be flushed. Signed-off-by: Pradeep Satyanarayana <pradeeps@linux.vnet.ibm.com> [ Completely rewritten and split up, based on Pradeep's work. Several bugs fixed and no doubt several bugs introduced. - Roland ] Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-01-25 23:15:24 +01:00
int recv_count;
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
};
struct ipoib_cm_tx {
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
struct ib_cm_id *id;
struct ib_qp *qp;
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
struct list_head list;
struct net_device *dev;
struct ipoib_neigh *neigh;
struct ipoib_path *path;
IPoIB/cm: Reduce connected mode TX object size Since IPoIB connected mode does not NETIF_F_SG, we only have one DMA mapping per send, so we don't need a mapping[] array. Define a new struct with a single u64 mapping member and use it for the CM tx_ring. Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-07-15 08:48:52 +02:00
struct ipoib_cm_tx_buf *tx_ring;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
unsigned tx_head;
unsigned tx_tail;
unsigned long flags;
u32 mtu;
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
};
struct ipoib_cm_rx_buf {
struct sk_buff *skb;
u64 mapping[IPOIB_CM_RX_SG];
};
struct ipoib_cm_dev_priv {
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
struct ib_srq *srq;
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
struct ipoib_cm_rx_buf *srq_ring;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
struct ib_cm_id *id;
struct list_head passive_ids; /* state: LIVE */
struct list_head rx_error_list; /* state: ERROR */
struct list_head rx_flush_list; /* state: FLUSH, drain not started */
struct list_head rx_drain_list; /* state: FLUSH, drain started */
struct list_head rx_reap_list; /* state: FLUSH, drain done */
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
struct work_struct start_task;
struct work_struct reap_task;
struct work_struct skb_task;
IPoIB/cm: Fix SRQ WR leak SRQ WR leakage has been observed with IPoIB/CM: e.g. flipping ports on and off will, with time, leak out all WRs and then all connections will start getting RNR NAKs. Fix this in the way suggested by spec: move the QP being destroyed to the error state, wait for "Last WQE Reached" event and then post WR on a "drain QP" connected to the same CQ. Once we observe a completion on the drain QP, it's safe to call ib_destroy_qp. Signed-off-by: Michael S. Tsirkin <mst@dev.mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-05-21 14:04:59 +02:00
struct work_struct rx_reap_task;
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
struct delayed_work stale_task;
struct sk_buff_head skb_queue;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
struct list_head start_list;
struct list_head reap_list;
struct ib_wc ibwc[IPOIB_NUM_WC];
struct ib_sge rx_sge[IPOIB_CM_RX_SG];
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
struct ib_recv_wr rx_wr;
IPoIB/cm: Add connected mode support for devices without SRQs Some IB adapters (notably IBM's eHCA) do not implement SRQs (shared receive queues). The current IPoIB connected mode support only works on devices that support SRQs. Fix this by adding support for using the receive queue of each connected mode receive QP. The disadvantage of this compared to using an SRQ is that it means a full queue of receives must be posted for each remote connected mode peer, which means that total memory usage is potentially much higher than when using SRQs. To manage this, add a new module parameter "max_nonsrq_conn_qp" that limits the number of connections allowed per interface. The rest of the changes are fairly straightforward: we use a table of struct ipoib_cm_rx to hold all the active connections, and put the table index of the connection in the high bits of receive WR IDs. This is needed because we cannot rely on the struct ib_wc.qp field for non-SRQ receive completions. Most of the rest of the changes just test whether or not an SRQ is available, and post receives or find received packets in the right place depending on the answer. Cleaning up dead connections actually becomes simpler, because we do not have to do the "last WQE reached" dance that is required to destroy QPs attached to an SRQ. We just move the QP to the error state and wait for all pending receives to be flushed. Signed-off-by: Pradeep Satyanarayana <pradeeps@linux.vnet.ibm.com> [ Completely rewritten and split up, based on Pradeep's work. Several bugs fixed and no doubt several bugs introduced. - Roland ] Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-01-25 23:15:24 +01:00
int nonsrq_conn_qp;
IPoIB/CM: Enable SRQ support on HCAs that support fewer than 16 SG entries Some HCAs (such as ehca2) support SRQ, but only support fewer than 16 SG entries for SRQs. Currently IPoIB/CM implicitly assumes all HCAs will support 16 SG entries for SRQs (to handle a 64K MTU with 4K pages). This patch removes that restriction by limiting the maximum MTU in connected mode to what the maximum number of SRQ SG entries allows. This patch addresses <https://bugs.openfabrics.org/show_bug.cgi?id=728> Signed-off-by: Pradeep Satyanarayana <pradeeps@linux.vnet.ibm.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-12-21 22:08:23 +01:00
int max_cm_mtu;
int num_frags;
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
};
IPoIB: Support modifying IPoIB CQ event moderation This can be used to tune at run time the parameters controlling the event (interrupt) generation rate and thus reduce the overhead incurred by handling interrupts resulting in better throughput. Since IPoIB uses a single CQ for both RX and TX, RX is chosen to dictate configuration for both RX and TX. Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-04-17 06:09:33 +02:00
struct ipoib_ethtool_st {
u16 coalesce_usecs;
u16 max_coalesced_frames;
};
IPoIB: Fix memory leak in the neigh table deletion flow If the neighbours hash table is empty when unloading the module, then ipoib_flush_neighs(), the cleanup routine, isn't called and the memory used for the hash table itself leaked. To fix this, ipoib_flush_neighs() is allways called, and another completion object is added to signal when the table is freed. Once invoked, ipoib_flush_neighs() flushes all the neighbours (if there are any), calls the the hash table RCU free routine, which now signals completion of the deletion process, and waits for the last neighbour to be freed. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-29 17:14:33 +02:00
struct ipoib_neigh_table;
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
struct ipoib_neigh_hash {
IPoIB: Fix memory leak in the neigh table deletion flow If the neighbours hash table is empty when unloading the module, then ipoib_flush_neighs(), the cleanup routine, isn't called and the memory used for the hash table itself leaked. To fix this, ipoib_flush_neighs() is allways called, and another completion object is added to signal when the table is freed. Once invoked, ipoib_flush_neighs() flushes all the neighbours (if there are any), calls the the hash table RCU free routine, which now signals completion of the deletion process, and waits for the last neighbour to be freed. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-29 17:14:33 +02:00
struct ipoib_neigh_table *ntbl;
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
struct ipoib_neigh __rcu **buckets;
struct rcu_head rcu;
u32 mask;
u32 size;
};
struct ipoib_neigh_table {
struct ipoib_neigh_hash __rcu *htbl;
atomic_t entries;
struct completion flushed;
IPoIB: Fix memory leak in the neigh table deletion flow If the neighbours hash table is empty when unloading the module, then ipoib_flush_neighs(), the cleanup routine, isn't called and the memory used for the hash table itself leaked. To fix this, ipoib_flush_neighs() is allways called, and another completion object is added to signal when the table is freed. Once invoked, ipoib_flush_neighs() flushes all the neighbours (if there are any), calls the the hash table RCU free routine, which now signals completion of the deletion process, and waits for the last neighbour to be freed. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-08-29 17:14:33 +02:00
struct completion deleted;
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
};
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
/*
IPoIB: Use netif_tx_lock() and get rid of private tx_lock, LLTX Currently, IPoIB is an LLTX driver that uses its own IRQ-disabling tx_lock. Not only do we want to get rid of LLTX, this actually causes problems because of the skb_orphan() done with this tx_lock held: some skb destructors expect to be run with interrupts enabled. The simplest fix for this is to get rid of the driver-private tx_lock and stop using LLTX. We kill off priv->tx_lock and use netif_tx_lock[_bh]() instead; the patch to do this is a tiny bit tricky because we need to update places that take priv->lock inside the tx_lock to disable IRQs, rather than relying on tx_lock having already disabled IRQs. Also, there are a couple of places where we need to disable BHs to make sure we have a consistent context to call netif_tx_lock() (since we no longer can use _irqsave() variants), and we also have to change ipoib_send_comp_handler() to call drain_tx_cq() through a timer rather than directly, because ipoib_send_comp_handler() runs in interrupt context and drain_tx_cq() must run in BH context so it can call netif_tx_lock(). Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-09-30 19:36:21 +02:00
* Device private locking: network stack tx_lock protects members used
* in TX fast path, lock protects everything else. lock nests inside
* of tx_lock (ie tx_lock must be acquired first if needed).
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
*/
struct ipoib_dev_priv {
spinlock_t lock;
struct net_device *dev;
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 01:41:36 +02:00
struct napi_struct napi;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
unsigned long flags;
IB: convert from semaphores to mutexes semaphore to mutex conversion by Ingo and Arjan's script. Signed-off-by: Ingo Molnar <mingo@elte.hu> [ Sanity-checked on real IB hardware ] Signed-off-by: Roland Dreier <rolandd@cisco.com>
2006-01-13 23:51:39 +01:00
struct mutex vlan_mutex;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
struct rb_root path_tree;
struct list_head path_list;
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
struct ipoib_neigh_table ntbl;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
struct ipoib_mcast *broadcast;
struct list_head multicast_list;
struct rb_root multicast_tree;
IPoIB: Handle P_Key table reordering SM reconfiguration or failover possibly causes a shuffling of the values in the P_Key table. Right now, IPoIB only queries for the P_Key index once when it creates the device QP, and hence there are problems if the index of a P_Key value changes. Fix this by using the PKEY_CHANGE event to trigger a recheck of the P_Key index. Signed-off-by: Yosef Etigin <yosefe@voltaire.com> Acked-by: Michael S. Tsirkin <mst@dev.mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-05-19 17:51:54 +02:00
struct delayed_work pkey_poll_task;
WorkStruct: make allyesconfig Fix up for make allyesconfig. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-11-22 15:57:56 +01:00
struct delayed_work mcast_task;
IPoIB: Fix deadlock on RTNL between bcast join comp and ipoib_stop() Taking rtnl_lock in ipoib_mcast_join_complete() causes a deadlock with ipoib_stop(). We avoid it by scheduling the piece of code that takes the lock on ipoib_workqueue instead of executing it directly. This works because we only flush the ipoib_workqueue with the RTNL not held. The deadlock happens because ipoib_stop() calls ipoib_ib_dev_down() which calls ipoib_mcast_dev_flush(), which calls ipoib_mcast_free(), which calls ipoib_mcast_leave(). The latter calls ib_sa_free_multicast(), and this waits until the multicast completion handler finishes. This handler is ipoib_mcast_join_complete(), which waits for the rtnl_lock(), which was already taken by ipoib_stop(). This bug was introduced in commit a77a57a1 ("IPoIB: Fix deadlock on RTNL in ipoib_stop()"). Signed-off-by: Yossi Etigin <yosefe@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-09-16 20:57:45 +02:00
struct work_struct carrier_on_task;
IPoIB: Refresh paths instead of flushing them on SM change events The patch tries to solve the problem of device going down and paths being flushed on an SM change event. The method is to mark the paths as candidates for refresh (by setting the new valid flag to 0), and wait for an ARP probe a new path record query. The solution requires a different and less intrusive handling of SM change event. For that, the second argument of the flush function changes its meaning from a boolean flag to a level. In most cases, SM failover doesn't cause LID change so traffic won't stop. In the rare cases of LID change, the remote host (the one that hadn't changed its LID) will lose connectivity until paths are refreshed. This is no worse than the current state. In fact, preventing the device from going down saves packets that otherwise would be lost. Signed-off-by: Moni Levy <monil@voltaire.com> Signed-off-by: Moni Shoua <monis@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-07-15 08:48:49 +02:00
struct work_struct flush_light;
struct work_struct flush_normal;
struct work_struct flush_heavy;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
struct work_struct restart_task;
WorkStruct: make allyesconfig Fix up for make allyesconfig. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-11-22 15:57:56 +01:00
struct delayed_work ah_reap_task;
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
struct delayed_work neigh_reap_task;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
struct ib_device *ca;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
u8 port;
u16 pkey;
u16 pkey_index;
struct ib_pd *pd;
struct ib_mr *mr;
IPoIB: Use separate CQ for UD send completions Use a dedicated CQ for UD send completions. Also, do not arm the UD send CQ, which reduces the number of interrupts generated. This patch farther reduces overhead by not calling poll CQ for every posted send WR -- it does polls only when there 16 or more outstanding work requests. Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-04-29 22:46:53 +02:00
struct ib_cq *recv_cq;
struct ib_cq *send_cq;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
struct ib_qp *qp;
u32 qkey;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
union ib_gid local_gid;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
u16 local_lid;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
unsigned int admin_mtu;
unsigned int mcast_mtu;
IPoIB: Handle 4K IB MTU for UD (datagram) mode This patch enables IPoIB to use 4K UD messages (when the underlying device and fabrics support a 4K MTU) by using two scatter buffers when PAGE_SIZE is less than or equal to thhe HCA IB MTU size. The first buffer is for IPoIB header + GRH header, and the second buffer is the IPoIB payload, which is 4K-4. Signed-off-by: Shirley Ma <xma@us.ibm.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-04-23 20:55:45 +02:00
unsigned int max_ib_mtu;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
[IPoIB] Drop RX packets when out of memory Change the way IPoIB handles RX packets when it can't allocate a new receive skbuff. If the allocation of a new receive skb fails, we now drop the packet we just received and repost the original receive skb. This means that the receive ring always stays full and we don't have to monkey around with trying to schedule a refill task for later. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-10-29 00:30:34 +02:00
struct ipoib_rx_buf *rx_ring;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
[IPoIB] Drop RX packets when out of memory Change the way IPoIB handles RX packets when it can't allocate a new receive skbuff. If the allocation of a new receive skb fails, we now drop the packet we just received and repost the original receive skb. This means that the receive ring always stays full and we don't have to monkey around with trying to schedule a refill task for later. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-10-29 00:30:34 +02:00
struct ipoib_tx_buf *tx_ring;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
unsigned tx_head;
unsigned tx_tail;
IPoIB: Add send gather support This patch acts as a preparation for using checksum offload for IB devices capable of inserting/verifying checksum in IP packets. The patch does not actaully turn on NETIF_F_SG - we defer that to the patches adding checksum offload capabilities. We only add support for send gathers for datagram mode, since existing HW does not support checksum offload on connected QPs. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-01-30 17:30:53 +01:00
struct ib_sge tx_sge[MAX_SKB_FRAGS + 1];
[IPoIB] Drop RX packets when out of memory Change the way IPoIB handles RX packets when it can't allocate a new receive skbuff. If the allocation of a new receive skb fails, we now drop the packet we just received and repost the original receive skb. This means that the receive ring always stays full and we don't have to monkey around with trying to schedule a refill task for later. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-10-29 00:30:34 +02:00
struct ib_send_wr tx_wr;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
unsigned tx_outstanding;
IPoIB: Use separate CQ for UD send completions Use a dedicated CQ for UD send completions. Also, do not arm the UD send CQ, which reduces the number of interrupts generated. This patch farther reduces overhead by not calling poll CQ for every posted send WR -- it does polls only when there 16 or more outstanding work requests. Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-04-29 22:46:53 +02:00
struct ib_wc send_wc[MAX_SEND_CQE];
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
IPoIB: Handle 4K IB MTU for UD (datagram) mode This patch enables IPoIB to use 4K UD messages (when the underlying device and fabrics support a 4K MTU) by using two scatter buffers when PAGE_SIZE is less than or equal to thhe HCA IB MTU size. The first buffer is for IPoIB header + GRH header, and the second buffer is the IPoIB payload, which is 4K-4. Signed-off-by: Shirley Ma <xma@us.ibm.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-04-23 20:55:45 +02:00
struct ib_recv_wr rx_wr;
struct ib_sge rx_sge[IPOIB_UD_RX_SG];
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
struct ib_wc ibwc[IPOIB_NUM_WC];
struct list_head dead_ahs;
struct ib_event_handler event_handler;
struct net_device *parent;
struct list_head child_intfs;
struct list_head list;
IB/ipoib: Add rtnl_link_ops support Add rtnl_link_ops to IPoIB, with the first usage being child device create/delete through them. Childs devices are now either legacy ones, created/deleted through the ipoib sysfs entries, or RTNL ones. Adding support for RTNL childs involved refactoring of ipoib_vlan_add which is now used by both the sysfs and the link_ops code. Also, added ndo_uninit entry to support calling unregister_netdevice_queue from the rtnl dellink entry. This required removal of calls to ipoib_dev_cleanup from the driver in flows which use unregister_netdevice, since the networking core will invoke ipoib_uninit which does exactly that. Signed-off-by: Erez Shitrit <erezsh@mellanox.co.il> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-13 07:56:36 +02:00
int child_type;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
#ifdef CONFIG_INFINIBAND_IPOIB_CM
struct ipoib_cm_dev_priv cm;
#endif
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
#ifdef CONFIG_INFINIBAND_IPOIB_DEBUG
struct list_head fs_list;
struct dentry *mcg_dentry;
[IPoIB] add path record information in debugfs Add ibX_path files to debugfs that contain information about the IPoIB path cache. IPoIB ARP only gives GIDs, which the IPoIB driver must resolve to real IB paths through the ib_sa module. For debugging, when the ARP table looks OK but traffic isn't flowing, it's useful to be able to see if the resolution from GID to path worked. Also clean up the formatting of the existing _mcg debugfs files. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-11-07 19:33:11 +01:00
struct dentry *path_dentry;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
#endif
IPoIB: Add LSO support For HCAs that support TCP segmentation offload (IB_DEVICE_UD_TSO), set NETIF_F_TSO and use HW LSO to offload TCP segmentation. Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-04-17 06:09:27 +02:00
int hca_caps;
IPoIB: Support modifying IPoIB CQ event moderation This can be used to tune at run time the parameters controlling the event (interrupt) generation rate and thus reduce the overhead incurred by handling interrupts resulting in better throughput. Since IPoIB uses a single CQ for both RX and TX, RX is chosen to dictate configuration for both RX and TX. Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-04-17 06:09:33 +02:00
struct ipoib_ethtool_st ethtool;
IB/ipoib: Fix transmit queue stalling forever Commit f56bcd80 ("IPoIB: Use separate CQ for UD send completions") introduced a bug where the transmit queue could get stopped and never woken up. The problem is that send completions are only polled at the end of the xmit function, so if the send queue fills up and the xmit path stops the queue, then there is no way for send completions to ever get polled, and so the transmit queue stays stopped forever. Fix this by arming the send CQ just before posting the last send request that fills the send queue. Then, when the completion event handler is called, drain the send CQ. Since it is possible that not enough send completions are in the CQ, verify that the the net queue has been woken up after draining the send CQ, and if not arm a timer and drain again at the timer function. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-05-01 05:02:45 +02:00
struct timer_list poll_timer;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
};
struct ipoib_ah {
struct net_device *dev;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
struct ib_ah *ah;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
struct list_head list;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
struct kref ref;
unsigned last_send;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
};
struct ipoib_path {
struct net_device *dev;
struct ib_sa_path_rec pathrec;
struct ipoib_ah *ah;
struct sk_buff_head queue;
struct list_head neigh_list;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
int query_id;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
struct ib_sa_query *query;
struct completion done;
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
struct rb_node rb_node;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
struct list_head list;
IPoIB: Refresh paths instead of flushing them on SM change events The patch tries to solve the problem of device going down and paths being flushed on an SM change event. The method is to mark the paths as candidates for refresh (by setting the new valid flag to 0), and wait for an ARP probe a new path record query. The solution requires a different and less intrusive handling of SM change event. For that, the second argument of the flush function changes its meaning from a boolean flag to a level. In most cases, SM failover doesn't cause LID change so traffic won't stop. In the rare cases of LID change, the remote host (the one that hadn't changed its LID) will lose connectivity until paths are refreshed. This is no worse than the current state. In fact, preventing the device from going down saves packets that otherwise would be lost. Signed-off-by: Moni Levy <monil@voltaire.com> Signed-off-by: Moni Shoua <monis@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-07-15 08:48:49 +02:00
int valid;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
};
struct ipoib_neigh {
struct ipoib_ah *ah;
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
#ifdef CONFIG_INFINIBAND_IPOIB_CM
struct ipoib_cm_tx *cm;
#endif
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
u8 daddr[INFINIBAND_ALEN];
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
struct sk_buff_head queue;
IB/ipoib: Bound the net device to the ipoib_neigh structue IPoIB uses a two layer neighboring scheme, such that for each struct neighbour whose device is an ipoib one, there is a struct ipoib_neigh buddy which is created on demand at the tx flow by an ipoib_neigh_alloc(skb->dst->neighbour) call. When using the bonding driver, neighbours are created by the net stack on behalf of the bonding (master) device. On the tx flow the bonding code gets an skb such that skb->dev points to the master device, it changes this skb to point on the slave device and calls the slave hard_start_xmit function. Under this scheme, ipoib_neigh_destructor assumption that for each struct neighbour it gets, n->dev is an ipoib device and hence netdev_priv(n->dev) can be casted to struct ipoib_dev_priv is buggy. To fix it, this patch adds a dev field to struct ipoib_neigh which is used instead of the struct neighbour dev one, when n->dev->flags has the IFF_MASTER bit set. Signed-off-by: Moni Shoua <monis at voltaire.com> Signed-off-by: Or Gerlitz <ogerlitz at voltaire.com> Acked-by: Roland Dreier <rdreier@cisco.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-10-10 04:43:36 +02:00
struct net_device *dev;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
struct list_head list;
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
struct ipoib_neigh __rcu *hnext;
struct rcu_head rcu;
atomic_t refcnt;
unsigned long alive;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
};
IPoIB: Handle 4K IB MTU for UD (datagram) mode This patch enables IPoIB to use 4K UD messages (when the underlying device and fabrics support a 4K MTU) by using two scatter buffers when PAGE_SIZE is less than or equal to thhe HCA IB MTU size. The first buffer is for IPoIB header + GRH header, and the second buffer is the IPoIB payload, which is 4K-4. Signed-off-by: Shirley Ma <xma@us.ibm.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-04-23 20:55:45 +02:00
#define IPOIB_UD_MTU(ib_mtu) (ib_mtu - IPOIB_ENCAP_LEN)
#define IPOIB_UD_BUF_SIZE(ib_mtu) (ib_mtu + IB_GRH_BYTES)
static inline int ipoib_ud_need_sg(unsigned int ib_mtu)
{
return IPOIB_UD_BUF_SIZE(ib_mtu) > PAGE_SIZE;
}
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
void ipoib_neigh_dtor(struct ipoib_neigh *neigh);
static inline void ipoib_neigh_put(struct ipoib_neigh *neigh)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
{
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
if (atomic_dec_and_test(&neigh->refcnt))
ipoib_neigh_dtor(neigh);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
}
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
struct ipoib_neigh *ipoib_neigh_get(struct net_device *dev, u8 *daddr);
struct ipoib_neigh *ipoib_neigh_alloc(u8 *daddr,
IB/ipoib: Bound the net device to the ipoib_neigh structue IPoIB uses a two layer neighboring scheme, such that for each struct neighbour whose device is an ipoib one, there is a struct ipoib_neigh buddy which is created on demand at the tx flow by an ipoib_neigh_alloc(skb->dst->neighbour) call. When using the bonding driver, neighbours are created by the net stack on behalf of the bonding (master) device. On the tx flow the bonding code gets an skb such that skb->dev points to the master device, it changes this skb to point on the slave device and calls the slave hard_start_xmit function. Under this scheme, ipoib_neigh_destructor assumption that for each struct neighbour it gets, n->dev is an ipoib device and hence netdev_priv(n->dev) can be casted to struct ipoib_dev_priv is buggy. To fix it, this patch adds a dev field to struct ipoib_neigh which is used instead of the struct neighbour dev one, when n->dev->flags has the IFF_MASTER bit set. Signed-off-by: Moni Shoua <monis at voltaire.com> Signed-off-by: Or Gerlitz <ogerlitz at voltaire.com> Acked-by: Roland Dreier <rdreier@cisco.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-10-10 04:43:36 +02:00
struct net_device *dev);
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
void ipoib_neigh_free(struct ipoib_neigh *neigh);
void ipoib_del_neighs_by_gid(struct net_device *dev, u8 *gid);
IPoIB: Consolidate private neighbour data handling Consolidate IPoIB's private neighbour data handling into ipoib_neigh_alloc() and ipoib_neigh_free(). This will make it easier to keep track of the neighbour structures that IPoIB is handling, and is a nice cleanup of the code: add/remove: 2/1 grow/shrink: 1/8 up/down: 100/-178 (-78) function old new delta ipoib_neigh_alloc - 61 +61 ipoib_neigh_free - 36 +36 ipoib_mcast_join_finish 1288 1291 +3 path_rec_completion 575 573 -2 ipoib_mcast_join_task 664 660 -4 ipoib_neigh_destructor 101 92 -9 ipoib_neigh_setup_dev 14 3 -11 ipoib_neigh_setup 17 - -17 path_free 238 215 -23 ipoib_mcast_free 329 306 -23 ipoib_mcast_send 718 684 -34 neigh_add_path 705 650 -55 Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2006-04-04 18:59:40 +02:00
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
extern struct workqueue_struct *ipoib_workqueue;
/* functions */
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 01:41:36 +02:00
int ipoib_poll(struct napi_struct *napi, int budget);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
void ipoib_ib_completion(struct ib_cq *cq, void *dev_ptr);
IB/ipoib: Fix transmit queue stalling forever Commit f56bcd80 ("IPoIB: Use separate CQ for UD send completions") introduced a bug where the transmit queue could get stopped and never woken up. The problem is that send completions are only polled at the end of the xmit function, so if the send queue fills up and the xmit path stops the queue, then there is no way for send completions to ever get polled, and so the transmit queue stays stopped forever. Fix this by arming the send CQ just before posting the last send request that fills the send queue. Then, when the completion event handler is called, drain the send CQ. Since it is possible that not enough send completions are in the CQ, verify that the the net queue has been woken up after draining the send CQ, and if not arm a timer and drain again at the timer function. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-05-01 05:02:45 +02:00
void ipoib_send_comp_handler(struct ib_cq *cq, void *dev_ptr);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
struct ipoib_ah *ipoib_create_ah(struct net_device *dev,
struct ib_pd *pd, struct ib_ah_attr *attr);
void ipoib_free_ah(struct kref *kref);
static inline void ipoib_put_ah(struct ipoib_ah *ah)
{
kref_put(&ah->ref, ipoib_free_ah);
}
[IPoIB] don't compile debug code if debugging isn't enabled Don't build ipoib_mcast_iter_ functions if CONFIG_INFINIBAND_IPOIB_DEBUG is not enabled -- their only callers will not be built either. Also move the prototype for ipoib_open() to ipoib.h to fix a sparse warning. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-11-03 05:51:01 +01:00
int ipoib_open(struct net_device *dev);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
int ipoib_add_pkey_attr(struct net_device *dev);
IPoIB: Allow setting policy to ignore multicast groups The kernel IB stack allows (through the RDMA CM) userspace applications to join and use multicast groups from the IPoIB MGID range. This allows multicast traffic to be handled directly from userspace QPs, without going through the kernel stack, which gives better performance for some applications. However, to fully interoperate with IP multicast, such userspace applications need to participate in IGMP reports and queries, or else routers may not forward the multicast traffic to the system where the application is running. The simplest way to do this is to share the kernel IGMP implementation by using the IP_ADD_MEMBERSHIP option to join multicast groups that are being handled directly in userspace. However, in such cases, the actual multicast traffic should not also be handled by the IPoIB interface, because that would burn resources handling multicast packets that will just be discarded in the kernel. To handle this, this patch adds lookup on the database used for IB multicast group reference counting when IPoIB is joining multicast groups, and if a multicast group is already handled by user space, then the IPoIB kernel driver ignores the group. This is controlled by a per-interface policy flag. When the flag is set, IPoIB will not join and attach its QP to a multicast group which already has an entry in the database; when the flag is cleared, IPoIB will behave as before this change. For each IPoIB interface, the /sys/class/net/$intf/umcast attribute controls the policy flag. The default value is off/0. Signed-off-by: Or Gerlitz <ogerlitz@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-08 10:13:00 +02:00
int ipoib_add_umcast_attr(struct net_device *dev);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
void ipoib_send(struct net_device *dev, struct sk_buff *skb,
struct ipoib_ah *address, u32 qpn);
WorkStruct: make allyesconfig Fix up for make allyesconfig. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-11-22 15:57:56 +01:00
void ipoib_reap_ah(struct work_struct *work);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
IPoIB: Refresh paths instead of flushing them on SM change events The patch tries to solve the problem of device going down and paths being flushed on an SM change event. The method is to mark the paths as candidates for refresh (by setting the new valid flag to 0), and wait for an ARP probe a new path record query. The solution requires a different and less intrusive handling of SM change event. For that, the second argument of the flush function changes its meaning from a boolean flag to a level. In most cases, SM failover doesn't cause LID change so traffic won't stop. In the rare cases of LID change, the remote host (the one that hadn't changed its LID) will lose connectivity until paths are refreshed. This is no worse than the current state. In fact, preventing the device from going down saves packets that otherwise would be lost. Signed-off-by: Moni Levy <monil@voltaire.com> Signed-off-by: Moni Shoua <monis@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-07-15 08:48:49 +02:00
void ipoib_mark_paths_invalid(struct net_device *dev);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
void ipoib_flush_paths(struct net_device *dev);
struct ipoib_dev_priv *ipoib_intf_alloc(const char *format);
int ipoib_ib_dev_init(struct net_device *dev, struct ib_device *ca, int port);
IPoIB: Refresh paths instead of flushing them on SM change events The patch tries to solve the problem of device going down and paths being flushed on an SM change event. The method is to mark the paths as candidates for refresh (by setting the new valid flag to 0), and wait for an ARP probe a new path record query. The solution requires a different and less intrusive handling of SM change event. For that, the second argument of the flush function changes its meaning from a boolean flag to a level. In most cases, SM failover doesn't cause LID change so traffic won't stop. In the rare cases of LID change, the remote host (the one that hadn't changed its LID) will lose connectivity until paths are refreshed. This is no worse than the current state. In fact, preventing the device from going down saves packets that otherwise would be lost. Signed-off-by: Moni Levy <monil@voltaire.com> Signed-off-by: Moni Shoua <monis@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-07-15 08:48:49 +02:00
void ipoib_ib_dev_flush_light(struct work_struct *work);
void ipoib_ib_dev_flush_normal(struct work_struct *work);
void ipoib_ib_dev_flush_heavy(struct work_struct *work);
IPoIB: Handle P_Key table reordering SM reconfiguration or failover possibly causes a shuffling of the values in the P_Key table. Right now, IPoIB only queries for the P_Key index once when it creates the device QP, and hence there are problems if the index of a P_Key value changes. Fix this by using the PKEY_CHANGE event to trigger a recheck of the P_Key index. Signed-off-by: Yosef Etigin <yosefe@voltaire.com> Acked-by: Michael S. Tsirkin <mst@dev.mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-05-19 17:51:54 +02:00
void ipoib_pkey_event(struct work_struct *work);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
void ipoib_ib_dev_cleanup(struct net_device *dev);
int ipoib_ib_dev_open(struct net_device *dev);
int ipoib_ib_dev_up(struct net_device *dev);
IPoIB: Move ipoib_ib_dev_flush() to ipoib workqueue Move ipoib_ib_dev_flush() to ipoib's workqueue. This keeps it ordered with respect to other work scheduled by the ipoib driver. This fixes problems with races, for example: - ipoib_ib_dev_flush() has started running because of an IB event - user does ifconfig ib0 down - ipoib_mcast_stop_thread() gets called twice and waits for the same completion twice Signed-off-by: Jack Morgenstein <jackm@mellanox.co.il> Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2006-03-20 19:08:24 +01:00
int ipoib_ib_dev_down(struct net_device *dev, int flush);
IPoIB: Handle P_Key table reordering SM reconfiguration or failover possibly causes a shuffling of the values in the P_Key table. Right now, IPoIB only queries for the P_Key index once when it creates the device QP, and hence there are problems if the index of a P_Key value changes. Fix this by using the PKEY_CHANGE event to trigger a recheck of the P_Key index. Signed-off-by: Yosef Etigin <yosefe@voltaire.com> Acked-by: Michael S. Tsirkin <mst@dev.mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-05-19 17:51:54 +02:00
int ipoib_ib_dev_stop(struct net_device *dev, int flush);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
int ipoib_dev_init(struct net_device *dev, struct ib_device *ca, int port);
void ipoib_dev_cleanup(struct net_device *dev);
WorkStruct: make allyesconfig Fix up for make allyesconfig. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-11-22 15:57:56 +01:00
void ipoib_mcast_join_task(struct work_struct *work);
IPoIB: Fix deadlock on RTNL between bcast join comp and ipoib_stop() Taking rtnl_lock in ipoib_mcast_join_complete() causes a deadlock with ipoib_stop(). We avoid it by scheduling the piece of code that takes the lock on ipoib_workqueue instead of executing it directly. This works because we only flush the ipoib_workqueue with the RTNL not held. The deadlock happens because ipoib_stop() calls ipoib_ib_dev_down() which calls ipoib_mcast_dev_flush(), which calls ipoib_mcast_free(), which calls ipoib_mcast_leave(). The latter calls ib_sa_free_multicast(), and this waits until the multicast completion handler finishes. This handler is ipoib_mcast_join_complete(), which waits for the rtnl_lock(), which was already taken by ipoib_stop(). This bug was introduced in commit a77a57a1 ("IPoIB: Fix deadlock on RTNL in ipoib_stop()"). Signed-off-by: Yossi Etigin <yosefe@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-09-16 20:57:45 +02:00
void ipoib_mcast_carrier_on_task(struct work_struct *work);
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
void ipoib_mcast_send(struct net_device *dev, u8 *daddr, struct sk_buff *skb);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
WorkStruct: make allyesconfig Fix up for make allyesconfig. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-11-22 15:57:56 +01:00
void ipoib_mcast_restart_task(struct work_struct *work);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
int ipoib_mcast_start_thread(struct net_device *dev);
[PATCH] IPoIB: Don't flush workqueue from within workqueue ipoib_mcast_restart_task() is always called from within the single-threaded IPoIB workqueue, so flushing the workqueue from within the function can lead to a recursion overflow. But since we're running in a single-threaded workqueue, we're already synchronized against other items in the workqueue, so just get rid of the flush in ipoib_mcast_restart_task(). Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-09-20 19:52:04 +02:00
int ipoib_mcast_stop_thread(struct net_device *dev, int flush);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
void ipoib_mcast_dev_down(struct net_device *dev);
void ipoib_mcast_dev_flush(struct net_device *dev);
[IPoIB] don't compile debug code if debugging isn't enabled Don't build ipoib_mcast_iter_ functions if CONFIG_INFINIBAND_IPOIB_DEBUG is not enabled -- their only callers will not be built either. Also move the prototype for ipoib_open() to ipoib.h to fix a sparse warning. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-11-03 05:51:01 +01:00
#ifdef CONFIG_INFINIBAND_IPOIB_DEBUG
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
struct ipoib_mcast_iter *ipoib_mcast_iter_init(struct net_device *dev);
int ipoib_mcast_iter_next(struct ipoib_mcast_iter *iter);
void ipoib_mcast_iter_read(struct ipoib_mcast_iter *iter,
union ib_gid *gid,
unsigned long *created,
unsigned int *queuelen,
unsigned int *complete,
unsigned int *send_only);
[IPoIB] add path record information in debugfs Add ibX_path files to debugfs that contain information about the IPoIB path cache. IPoIB ARP only gives GIDs, which the IPoIB driver must resolve to real IB paths through the ib_sa module. For debugging, when the ARP table looks OK but traffic isn't flowing, it's useful to be able to see if the resolution from GID to path worked. Also clean up the formatting of the existing _mcg debugfs files. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-11-07 19:33:11 +01:00
struct ipoib_path_iter *ipoib_path_iter_init(struct net_device *dev);
int ipoib_path_iter_next(struct ipoib_path_iter *iter);
void ipoib_path_iter_read(struct ipoib_path_iter *iter,
struct ipoib_path *path);
[IPoIB] don't compile debug code if debugging isn't enabled Don't build ipoib_mcast_iter_ functions if CONFIG_INFINIBAND_IPOIB_DEBUG is not enabled -- their only callers will not be built either. Also move the prototype for ipoib_open() to ipoib.h to fix a sparse warning. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-11-03 05:51:01 +01:00
#endif
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
int ipoib_mcast_attach(struct net_device *dev, u16 mlid,
IPoIB: Only set Q_Key once: after joining broadcast group The current code will set the Q_Key for any join of a non-sendonly multicast group. The operation involves a modify QP operation, which is fairly heavyweight, and is only really required after the join of the broadcast group. Fix this by adding a parameter to ipoib_mcast_attach() to control when the Q_Key is set. Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-07-15 08:48:50 +02:00
union ib_gid *mgid, int set_qkey);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
[IPoIB] Rename ipoib_create_qp() -> ipoib_init_qp() and fix error cleanup ipoib_create_qp() no longer creates IPoIB's QP, so it shouldn't destroy the QP on failure -- that unwinding happens elsewhere, so the current code can cause a double free. While we're at it, the function's name should match what it actually does, so rename it to ipoib_init_qp(). Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-10-11 20:08:24 +02:00
int ipoib_init_qp(struct net_device *dev);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
int ipoib_transport_dev_init(struct net_device *dev, struct ib_device *ca);
void ipoib_transport_dev_cleanup(struct net_device *dev);
void ipoib_event(struct ib_event_handler *handler,
struct ib_event *record);
int ipoib_vlan_add(struct net_device *pdev, unsigned short pkey);
int ipoib_vlan_delete(struct net_device *pdev, unsigned short pkey);
IB/ipoib: Add rtnl_link_ops support Add rtnl_link_ops to IPoIB, with the first usage being child device create/delete through them. Childs devices are now either legacy ones, created/deleted through the ipoib sysfs entries, or RTNL ones. Adding support for RTNL childs involved refactoring of ipoib_vlan_add which is now used by both the sysfs and the link_ops code. Also, added ndo_uninit entry to support calling unregister_netdevice_queue from the rtnl dellink entry. This required removal of calls to ipoib_dev_cleanup from the driver in flows which use unregister_netdevice, since the networking core will invoke ipoib_uninit which does exactly that. Signed-off-by: Erez Shitrit <erezsh@mellanox.co.il> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-13 07:56:36 +02:00
int __ipoib_vlan_add(struct ipoib_dev_priv *ppriv, struct ipoib_dev_priv *priv,
u16 pkey, int child_type);
int __init ipoib_netlink_init(void);
void __exit ipoib_netlink_fini(void);
IB/ipoib: Add more rtnl_link_ops callbacks Add the rtnl_link_ops changelink and fill_info callbacks, through which the admin can now set/get the driver mode, etc policies. Maintain the proprietary sysfs entries only for legacy childs. For child devices, set dev->iflink to point to the parent device ifindex, such that user space tools can now correctly show the uplink relation as done for vlan, macvlan, etc devices. Pointed out by Patrick McHardy <kaber@trash.net> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-27 14:06:02 +02:00
void ipoib_set_umcast(struct net_device *ndev, int umcast_val);
int ipoib_set_mode(struct net_device *dev, const char *buf);
IB/ipoib: Add rtnl_link_ops support Add rtnl_link_ops to IPoIB, with the first usage being child device create/delete through them. Childs devices are now either legacy ones, created/deleted through the ipoib sysfs entries, or RTNL ones. Adding support for RTNL childs involved refactoring of ipoib_vlan_add which is now used by both the sysfs and the link_ops code. Also, added ndo_uninit entry to support calling unregister_netdevice_queue from the rtnl dellink entry. This required removal of calls to ipoib_dev_cleanup from the driver in flows which use unregister_netdevice, since the networking core will invoke ipoib_uninit which does exactly that. Signed-off-by: Erez Shitrit <erezsh@mellanox.co.il> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2012-09-13 07:56:36 +02:00
void ipoib_setup(struct net_device *dev);
WorkStruct: make allyesconfig Fix up for make allyesconfig. Signed-Off-By: David Howells <dhowells@redhat.com>
2006-11-22 15:57:56 +01:00
void ipoib_pkey_poll(struct work_struct *work);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
int ipoib_pkey_dev_delay_open(struct net_device *dev);
IPoIB/cm: Drain cq in ipoib_cm_dev_stop() Since NAPI polling is disabled while ipoib_cm_dev_stop() is running, ipoib_cm_dev_stop() must poll the CQ itself in order to see the packets draining. Signed-off-by: Michael S. Tsirkin <mst@dev.mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-05-24 17:32:46 +02:00
void ipoib_drain_cq(struct net_device *dev);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
IPoIB: Add basic ethtool support Just add the infrastructure so we can add functionality later. Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-04-17 06:09:32 +02:00
void ipoib_set_ethtool_ops(struct net_device *dev);
IPoIB: Set netdev offload features properly for child (VLAN) interfaces Child devices were created without any offload features set, fix this by moving the code that computes the features into generic function which is now called through non-child and child device creation. Signed-off-by: Or Gerlitz <ogerlitz@voltaire.com> -- v1 has a bug where the 'result' flag in ipoib_vlan_add may be used uninitialized Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-10-23 00:49:49 +02:00
int ipoib_set_dev_features(struct ipoib_dev_priv *priv, struct ib_device *hca);
IPoIB: Add basic ethtool support Just add the infrastructure so we can add functionality later. Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-04-17 06:09:32 +02:00
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
#define IPOIB_FLAGS_RC 0x80
#define IPOIB_FLAGS_UC 0x40
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
/* We don't support UC connections at the moment */
#define IPOIB_CM_SUPPORTED(ha) (ha[0] & (IPOIB_FLAGS_RC))
IPoIB: Fix build with CONFIG_INFINIBAND_IPOIB_CM=n With the new netlink support in commit 862096a8bbf8 ("IB/ipoib: Add more rtnl_link_ops callbacks") we need ipoib_set_mode() to be available even if connected mode isn't built. Move the function from ipoib_cm.c to ipoib_main.c (and make a few CM-related macros available unconditonally). This fixes the build error drivers/built-in.o: In function 'ipoib_changelink': ipoib_netlink.c:(.text+0x6a5fc9): undefined reference to 'ipoib_set_mode' ipoib_netlink.c:(.text+0x6a5fe3): undefined reference to 'ipoib_set_mode' when CONFIG_INFINIBAND_IPOIB_CM isn't set. Reported-by: Randy Dunlap <rdunlap@xenotime.net> Reported-by: Michael Neuling <mikey@neuling.org> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-10-03 06:23:43 +02:00
#ifdef CONFIG_INFINIBAND_IPOIB_CM
IPoIB/cm: Add connected mode support for devices without SRQs Some IB adapters (notably IBM's eHCA) do not implement SRQs (shared receive queues). The current IPoIB connected mode support only works on devices that support SRQs. Fix this by adding support for using the receive queue of each connected mode receive QP. The disadvantage of this compared to using an SRQ is that it means a full queue of receives must be posted for each remote connected mode peer, which means that total memory usage is potentially much higher than when using SRQs. To manage this, add a new module parameter "max_nonsrq_conn_qp" that limits the number of connections allowed per interface. The rest of the changes are fairly straightforward: we use a table of struct ipoib_cm_rx to hold all the active connections, and put the table index of the connection in the high bits of receive WR IDs. This is needed because we cannot rely on the struct ib_wc.qp field for non-SRQ receive completions. Most of the rest of the changes just test whether or not an SRQ is available, and post receives or find received packets in the right place depending on the answer. Cleaning up dead connections actually becomes simpler, because we do not have to do the "last WQE reached" dance that is required to destroy QPs attached to an SRQ. We just move the QP to the error state and wait for all pending receives to be flushed. Signed-off-by: Pradeep Satyanarayana <pradeeps@linux.vnet.ibm.com> [ Completely rewritten and split up, based on Pradeep's work. Several bugs fixed and no doubt several bugs introduced. - Roland ] Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-01-25 23:15:24 +01:00
extern int ipoib_max_conn_qp;
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
static inline int ipoib_cm_admin_enabled(struct net_device *dev)
{
struct ipoib_dev_priv *priv = netdev_priv(dev);
return IPOIB_CM_SUPPORTED(dev->dev_addr) &&
test_bit(IPOIB_FLAG_ADMIN_CM, &priv->flags);
}
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
static inline int ipoib_cm_enabled(struct net_device *dev, u8 *hwaddr)
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
{
struct ipoib_dev_priv *priv = netdev_priv(dev);
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
return IPOIB_CM_SUPPORTED(hwaddr) &&
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
test_bit(IPOIB_FLAG_ADMIN_CM, &priv->flags);
}
static inline int ipoib_cm_up(struct ipoib_neigh *neigh)
{
return test_bit(IPOIB_FLAG_OPER_UP, &neigh->cm->flags);
}
static inline struct ipoib_cm_tx *ipoib_cm_get(struct ipoib_neigh *neigh)
{
return neigh->cm;
}
static inline void ipoib_cm_set(struct ipoib_neigh *neigh, struct ipoib_cm_tx *tx)
{
neigh->cm = tx;
}
IPoIB/cm: Add connected mode support for devices without SRQs Some IB adapters (notably IBM's eHCA) do not implement SRQs (shared receive queues). The current IPoIB connected mode support only works on devices that support SRQs. Fix this by adding support for using the receive queue of each connected mode receive QP. The disadvantage of this compared to using an SRQ is that it means a full queue of receives must be posted for each remote connected mode peer, which means that total memory usage is potentially much higher than when using SRQs. To manage this, add a new module parameter "max_nonsrq_conn_qp" that limits the number of connections allowed per interface. The rest of the changes are fairly straightforward: we use a table of struct ipoib_cm_rx to hold all the active connections, and put the table index of the connection in the high bits of receive WR IDs. This is needed because we cannot rely on the struct ib_wc.qp field for non-SRQ receive completions. Most of the rest of the changes just test whether or not an SRQ is available, and post receives or find received packets in the right place depending on the answer. Cleaning up dead connections actually becomes simpler, because we do not have to do the "last WQE reached" dance that is required to destroy QPs attached to an SRQ. We just move the QP to the error state and wait for all pending receives to be flushed. Signed-off-by: Pradeep Satyanarayana <pradeeps@linux.vnet.ibm.com> [ Completely rewritten and split up, based on Pradeep's work. Several bugs fixed and no doubt several bugs introduced. - Roland ] Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-01-25 23:15:24 +01:00
static inline int ipoib_cm_has_srq(struct net_device *dev)
{
struct ipoib_dev_priv *priv = netdev_priv(dev);
return !!priv->cm.srq;
}
IPoIB/CM: Enable SRQ support on HCAs that support fewer than 16 SG entries Some HCAs (such as ehca2) support SRQ, but only support fewer than 16 SG entries for SRQs. Currently IPoIB/CM implicitly assumes all HCAs will support 16 SG entries for SRQs (to handle a 64K MTU with 4K pages). This patch removes that restriction by limiting the maximum MTU in connected mode to what the maximum number of SRQ SG entries allows. This patch addresses <https://bugs.openfabrics.org/show_bug.cgi?id=728> Signed-off-by: Pradeep Satyanarayana <pradeeps@linux.vnet.ibm.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-12-21 22:08:23 +01:00
static inline unsigned int ipoib_cm_max_mtu(struct net_device *dev)
{
struct ipoib_dev_priv *priv = netdev_priv(dev);
return priv->cm.max_cm_mtu;
}
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_tx *tx);
int ipoib_cm_dev_open(struct net_device *dev);
void ipoib_cm_dev_stop(struct net_device *dev);
int ipoib_cm_dev_init(struct net_device *dev);
int ipoib_cm_add_mode_attr(struct net_device *dev);
void ipoib_cm_dev_cleanup(struct net_device *dev);
struct ipoib_cm_tx *ipoib_cm_create_tx(struct net_device *dev, struct ipoib_path *path,
struct ipoib_neigh *neigh);
void ipoib_cm_destroy_tx(struct ipoib_cm_tx *tx);
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
void ipoib_cm_skb_too_long(struct net_device *dev, struct sk_buff *skb,
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
unsigned int mtu);
void ipoib_cm_handle_rx_wc(struct net_device *dev, struct ib_wc *wc);
IPoIB/cm: Use common CQ for CM send completions Use the same CQ for CM send completions as for all other IPoIB completions. This means all completions are processed via the same NAPI polling routine. This should help reduce the number of interrupts for bi-directional traffic (such as TCP) and fixes "driver is hogging interrupts" errors reported for IPoIB send side, e.g. <https://bugs.openfabrics.org/show_bug.cgi?id=508> To do this, keep a per-interface counter of outstanding send WRs, and stop the interface when this counter reaches the send queue size to avoid CQ overruns. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-08-16 14:36:16 +02:00
void ipoib_cm_handle_tx_wc(struct net_device *dev, struct ib_wc *wc);
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
#else
struct ipoib_cm_tx;
IPoIB/cm: Add connected mode support for devices without SRQs Some IB adapters (notably IBM's eHCA) do not implement SRQs (shared receive queues). The current IPoIB connected mode support only works on devices that support SRQs. Fix this by adding support for using the receive queue of each connected mode receive QP. The disadvantage of this compared to using an SRQ is that it means a full queue of receives must be posted for each remote connected mode peer, which means that total memory usage is potentially much higher than when using SRQs. To manage this, add a new module parameter "max_nonsrq_conn_qp" that limits the number of connections allowed per interface. The rest of the changes are fairly straightforward: we use a table of struct ipoib_cm_rx to hold all the active connections, and put the table index of the connection in the high bits of receive WR IDs. This is needed because we cannot rely on the struct ib_wc.qp field for non-SRQ receive completions. Most of the rest of the changes just test whether or not an SRQ is available, and post receives or find received packets in the right place depending on the answer. Cleaning up dead connections actually becomes simpler, because we do not have to do the "last WQE reached" dance that is required to destroy QPs attached to an SRQ. We just move the QP to the error state and wait for all pending receives to be flushed. Signed-off-by: Pradeep Satyanarayana <pradeeps@linux.vnet.ibm.com> [ Completely rewritten and split up, based on Pradeep's work. Several bugs fixed and no doubt several bugs introduced. - Roland ] Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-01-25 23:15:24 +01:00
#define ipoib_max_conn_qp 0
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
static inline int ipoib_cm_admin_enabled(struct net_device *dev)
{
return 0;
}
IPoIB: Use a private hash table for path lookup in xmit path Dave Miller <davem@davemloft.net> provided a detailed description of why the way IPoIB is using neighbours for its own ipoib_neigh struct is buggy: Any time an ipoib_neigh is changed, a sequence like the following is made: spin_lock_irqsave(&priv->lock, flags); /* * It's safe to call ipoib_put_ah() inside * priv->lock here, because we know that * path->ah will always hold one more reference, * so ipoib_put_ah() will never do more than * decrement the ref count. */ if (neigh->ah) ipoib_put_ah(neigh->ah); list_del(&neigh->list); ipoib_neigh_free(dev, neigh); spin_unlock_irqrestore(&priv->lock, flags); ipoib_path_lookup(skb, n, dev); This doesn't work, because you're leaving a stale pointer to the freed up ipoib_neigh in the special neigh->ha pointer cookie. Yes, it even fails with all the locking done to protect _changes_ to *ipoib_neigh(n), and with the code in ipoib_neigh_free() that NULLs out the pointer. The core issue is that read side calls to *to_ipoib_neigh(n) are not being synchronized at all, they are performed without any locking. So whether we hold the lock or not when making changes to *ipoib_neigh(n) you still can have threads see references to freed up ipoib_neigh objects. cpu 1 cpu 2 n = *ipoib_neigh() *ipoib_neigh() = NULL kfree(n) n->foo == OOPS [..] Perhaps the ipoib code can have a private path database it manages entirely itself, which holds all the necessary information and is looked up by some generic key which is available easily at transmit time and does not involve generic neighbour entries. See <http://marc.info/?l=linux-rdma&m=132812793105624&w=2> and <http://marc.info/?l=linux-rdma&w=2&r=1&s=allows+references+to+freed+memory&q=b> for the full discussion. This patch aims to solve the race conditions found in the IPoIB driver. The patch removes the connection between the core networking neighbour structure and the ipoib_neigh structure. In addition to avoiding the race described above, it allows us to handle SKBs carrying IP packets that don't have any associated neighbour. We add an ipoib_neigh hash table with N buckets where the key is the destination hardware address. The ipoib_neigh is fetched from the hash table and instead of the stashed location in the neighbour structure. The hash table uses both RCU and reference counting to guarantee that no ipoib_neigh instance is ever deleted while in use. Fetching the ipoib_neigh structure instance from the hash also makes the special code in ipoib_start_xmit that handles remote and local bonding failover redundant. Aged ipoib_neigh instances are deleted by a garbage collection task that runs every M seconds and deletes every ipoib_neigh instance that was idle for at least 2*M seconds. The deletion is safe since the ipoib_neigh instances are protected using RCU and reference count mechanisms. The number of buckets (N) and frequency of running the GC thread (M), are taken from the exported arb_tbl. Signed-off-by: Shlomo Pongratz <shlomop@mellanox.com> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: Roland Dreier <roland@purestorage.com>
2012-07-24 19:05:22 +02:00
static inline int ipoib_cm_enabled(struct net_device *dev, u8 *hwaddr)
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
{
return 0;
}
static inline int ipoib_cm_up(struct ipoib_neigh *neigh)
{
return 0;
}
static inline struct ipoib_cm_tx *ipoib_cm_get(struct ipoib_neigh *neigh)
{
return NULL;
}
static inline void ipoib_cm_set(struct ipoib_neigh *neigh, struct ipoib_cm_tx *tx)
{
}
IPoIB/cm: Add connected mode support for devices without SRQs Some IB adapters (notably IBM's eHCA) do not implement SRQs (shared receive queues). The current IPoIB connected mode support only works on devices that support SRQs. Fix this by adding support for using the receive queue of each connected mode receive QP. The disadvantage of this compared to using an SRQ is that it means a full queue of receives must be posted for each remote connected mode peer, which means that total memory usage is potentially much higher than when using SRQs. To manage this, add a new module parameter "max_nonsrq_conn_qp" that limits the number of connections allowed per interface. The rest of the changes are fairly straightforward: we use a table of struct ipoib_cm_rx to hold all the active connections, and put the table index of the connection in the high bits of receive WR IDs. This is needed because we cannot rely on the struct ib_wc.qp field for non-SRQ receive completions. Most of the rest of the changes just test whether or not an SRQ is available, and post receives or find received packets in the right place depending on the answer. Cleaning up dead connections actually becomes simpler, because we do not have to do the "last WQE reached" dance that is required to destroy QPs attached to an SRQ. We just move the QP to the error state and wait for all pending receives to be flushed. Signed-off-by: Pradeep Satyanarayana <pradeeps@linux.vnet.ibm.com> [ Completely rewritten and split up, based on Pradeep's work. Several bugs fixed and no doubt several bugs introduced. - Roland ] Signed-off-by: Roland Dreier <rolandd@cisco.com>
2008-01-25 23:15:24 +01:00
static inline int ipoib_cm_has_srq(struct net_device *dev)
{
return 0;
}
IPoIB/CM: Enable SRQ support on HCAs that support fewer than 16 SG entries Some HCAs (such as ehca2) support SRQ, but only support fewer than 16 SG entries for SRQs. Currently IPoIB/CM implicitly assumes all HCAs will support 16 SG entries for SRQs (to handle a 64K MTU with 4K pages). This patch removes that restriction by limiting the maximum MTU in connected mode to what the maximum number of SRQ SG entries allows. This patch addresses <https://bugs.openfabrics.org/show_bug.cgi?id=728> Signed-off-by: Pradeep Satyanarayana <pradeeps@linux.vnet.ibm.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-12-21 22:08:23 +01:00
static inline unsigned int ipoib_cm_max_mtu(struct net_device *dev)
{
return 0;
}
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
static inline
void ipoib_cm_send(struct net_device *dev, struct sk_buff *skb, struct ipoib_cm_tx *tx)
{
return;
}
static inline
int ipoib_cm_dev_open(struct net_device *dev)
{
return 0;
}
static inline
void ipoib_cm_dev_stop(struct net_device *dev)
{
return;
}
static inline
int ipoib_cm_dev_init(struct net_device *dev)
{
return -ENOSYS;
}
static inline
void ipoib_cm_dev_cleanup(struct net_device *dev)
{
return;
}
static inline
struct ipoib_cm_tx *ipoib_cm_create_tx(struct net_device *dev, struct ipoib_path *path,
struct ipoib_neigh *neigh)
{
return NULL;
}
static inline
void ipoib_cm_destroy_tx(struct ipoib_cm_tx *tx)
{
return;
}
static inline
int ipoib_cm_add_mode_attr(struct net_device *dev)
{
return 0;
}
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
static inline void ipoib_cm_skb_too_long(struct net_device *dev, struct sk_buff *skb,
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
unsigned int mtu)
{
dev_kfree_skb_any(skb);
}
static inline void ipoib_cm_handle_rx_wc(struct net_device *dev, struct ib_wc *wc)
{
}
IPoIB/cm: Use common CQ for CM send completions Use the same CQ for CM send completions as for all other IPoIB completions. This means all completions are processed via the same NAPI polling routine. This should help reduce the number of interrupts for bi-directional traffic (such as TCP) and fixes "driver is hogging interrupts" errors reported for IPoIB send side, e.g. <https://bugs.openfabrics.org/show_bug.cgi?id=508> To do this, keep a per-interface counter of outstanding send WRs, and stop the interface when this counter reaches the send queue size to avoid CQ overruns. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-08-16 14:36:16 +02:00
static inline void ipoib_cm_handle_tx_wc(struct net_device *dev, struct ib_wc *wc)
{
}
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
#endif
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
#ifdef CONFIG_INFINIBAND_IPOIB_DEBUG
[IPoIB] add path record information in debugfs Add ibX_path files to debugfs that contain information about the IPoIB path cache. IPoIB ARP only gives GIDs, which the IPoIB driver must resolve to real IB paths through the ib_sa module. For debugging, when the ARP table looks OK but traffic isn't flowing, it's useful to be able to see if the resolution from GID to path worked. Also clean up the formatting of the existing _mcg debugfs files. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-11-07 19:33:11 +01:00
void ipoib_create_debug_files(struct net_device *dev);
void ipoib_delete_debug_files(struct net_device *dev);
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
int ipoib_register_debugfs(void);
void ipoib_unregister_debugfs(void);
#else
[IPoIB] add path record information in debugfs Add ibX_path files to debugfs that contain information about the IPoIB path cache. IPoIB ARP only gives GIDs, which the IPoIB driver must resolve to real IB paths through the ib_sa module. For debugging, when the ARP table looks OK but traffic isn't flowing, it's useful to be able to see if the resolution from GID to path worked. Also clean up the formatting of the existing _mcg debugfs files. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2005-11-07 19:33:11 +01:00
static inline void ipoib_create_debug_files(struct net_device *dev) { }
static inline void ipoib_delete_debug_files(struct net_device *dev) { }
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
static inline int ipoib_register_debugfs(void) { return 0; }
static inline void ipoib_unregister_debugfs(void) { }
#endif
#define ipoib_printk(level, priv, format, arg...) \
printk(level "%s: " format, ((struct ipoib_dev_priv *) priv)->dev->name , ## arg)
#define ipoib_warn(priv, format, arg...) \
ipoib_printk(KERN_WARNING, priv, format , ## arg)
IPoIB: Make send and receive queue sizes tunable Make IPoIB's send and receive queue sizes tunable via module parameters ("send_queue_size" and "recv_queue_size"). This allows the queue sizes to be enlarged to fix disastrously bad performance on some platforms and workloads, without bloating memory usage when large queues aren't needed. Signed-off-by: Shirley Ma <xma@us.ibm.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2006-04-10 18:43:58 +02:00
extern int ipoib_sendq_size;
extern int ipoib_recvq_size;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
IB/sa: Require SA registration Require users to register with SA module, to prevent the sa_query module text from going away while an SA query callback is still running. Update all in-tree users for the new interface. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Sean Hefty <sean.hefty@intel.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2006-08-22 01:40:12 +02:00
extern struct ib_sa_client ipoib_sa_client;
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
#ifdef CONFIG_INFINIBAND_IPOIB_DEBUG
extern int ipoib_debug_level;
#define ipoib_dbg(priv, format, arg...) \
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
do { \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
if (ipoib_debug_level > 0) \
ipoib_printk(KERN_DEBUG, priv, format , ## arg); \
} while (0)
#define ipoib_dbg_mcast(priv, format, arg...) \
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
do { \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
if (mcast_debug_level > 0) \
ipoib_printk(KERN_DEBUG, priv, format , ## arg); \
} while (0)
#else /* CONFIG_INFINIBAND_IPOIB_DEBUG */
#define ipoib_dbg(priv, format, arg...) \
do { (void) (priv); } while (0)
#define ipoib_dbg_mcast(priv, format, arg...) \
do { (void) (priv); } while (0)
#endif /* CONFIG_INFINIBAND_IPOIB_DEBUG */
#ifdef CONFIG_INFINIBAND_IPOIB_DEBUG_DATA
#define ipoib_dbg_data(priv, format, arg...) \
IPoIB: Trivial formatting cleanups Fix whitespace blunders, convert "foo* bar" to "foo *bar", etc. Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-24 04:57:54 +02:00
do { \
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
if (data_debug_level > 0) \
ipoib_printk(KERN_DEBUG, priv, format , ## arg); \
} while (0)
#else /* CONFIG_INFINIBAND_IPOIB_DEBUG_DATA */
#define ipoib_dbg_data(priv, format, arg...) \
do { (void) (priv); } while (0)
#endif /* CONFIG_INFINIBAND_IPOIB_DEBUG_DATA */
IPoIB: Connected mode experimental support The following patch adds experimental support for IPoIB connected mode, as defined by the draft from the IETF ipoib working group. The idea is to increase performance by increasing the MTU from the maximum of 2K (theoretically 4K) supported by IPoIB on top of UD. With this code, I'm able to get 800MByte/sec or more with netperf without options on a Mellanox 4x back-to-back DDR system. Some notes on code: 1. SRQ is used for scalability to large cluster sizes 2. Only RC connections are used (UC does not support SRQ now) 3. Retry count is set to 0 since spec draft warns against retries 4. Each connection is used for data transfers in only 1 direction, so each connection is either active(TX) or passive (RX). 2 sides that want to communicate create 2 connections. 5. Each active (TX) connection has a separate CQ for send completions - this keeps the code simple without CQ resize and other tricks 6. To detect stale passive side connections (where the remote side is down), we keep an LRU list of passive connections (updated once per second per connection) and destroy a connection after it has been unused for several seconds. The LRU rule makes it possible to avoid scanning connections that have recently been active. Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-02-05 21:12:23 +01:00
#define IPOIB_QPN(ha) (be32_to_cpup((__be32 *) ha) & 0xffffff)
Linux-2.6.12-rc2 Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
2005-04-17 00:20:36 +02:00
#endif /* _IPOIB_H */