From 1c83469d36a9dd30dbf1fb9fc5ca3be3a0e64ff4 Mon Sep 17 00:00:00 2001 From: Mike Christie Date: Mon, 24 Jul 2006 15:47:26 -0500 Subject: [SCSI] iscsi bugfixes: fix oops when iser is flushing io When we enter recovery and flush the running commands we cannot freee the connection before flushing the commands. Some commands may have a reference to the connection that needs to be released before. iscsi_stop was forcing the term and suspend too early and was causing a oops in iser, so this patch removes those callbacks all together and allows the LLD to handle that detail. Signed-off-by: Mike Christie Signed-off-by: James Bottomley --- drivers/infiniband/ulp/iser/iscsi_iser.c | 22 ++-------------------- 1 file changed, 2 insertions(+), 20 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 34b0da5cfa0a..1437d7ee3b19 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -378,21 +378,6 @@ iscsi_iser_conn_start(struct iscsi_cls_conn *cls_conn) return iser_conn_set_full_featured_mode(conn); } -static void -iscsi_iser_conn_terminate(struct iscsi_conn *conn) -{ - struct iscsi_iser_conn *iser_conn = conn->dd_data; - struct iser_conn *ib_conn = iser_conn->ib_conn; - - BUG_ON(!ib_conn); - /* starts conn teardown process, waits until all previously * - * posted buffers get flushed, deallocates all conn resources */ - iser_conn_terminate(ib_conn); - iser_conn->ib_conn = NULL; - conn->recv_lock = NULL; -} - - static struct iscsi_transport iscsi_iser_transport; static struct iscsi_cls_session * @@ -555,13 +540,13 @@ iscsi_iser_ep_poll(__u64 ep_handle, int timeout_ms) static void iscsi_iser_ep_disconnect(__u64 ep_handle) { - struct iser_conn *ib_conn = iscsi_iser_ib_conn_lookup(ep_handle); + struct iser_conn *ib_conn; + ib_conn = iscsi_iser_ib_conn_lookup(ep_handle); if (!ib_conn) return; iser_err("ib conn %p state %d\n",ib_conn, ib_conn->state); - iser_conn_terminate(ib_conn); } @@ -614,9 +599,6 @@ static struct iscsi_transport iscsi_iser_transport = { .get_session_param = iscsi_session_get_param, .start_conn = iscsi_iser_conn_start, .stop_conn = iscsi_conn_stop, - /* these are called as part of conn recovery */ - .suspend_conn_recv = NULL, /* FIXME is/how this relvant to iser? */ - .terminate_conn = iscsi_iser_conn_terminate, /* IO */ .send_pdu = iscsi_conn_send_pdu, .get_stats = iscsi_iser_conn_get_stats, -- cgit From e795d092507d571d66f2ec98d3efdc7dd284bf80 Mon Sep 17 00:00:00 2001 From: Tom Tucker Date: Sun, 30 Jul 2006 20:44:19 -0700 Subject: [NET] infiniband: Cleanup ib_addr module to use the netevents Signed-off-by: Tom Tucker Signed-off-by: Steve Wise Signed-off-by: David S. Miller --- drivers/infiniband/core/addr.c | 30 ++++++++++++++---------------- 1 file changed, 14 insertions(+), 16 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/addr.c b/drivers/infiniband/core/addr.c index d294bbc42f09..1205e8027829 100644 --- a/drivers/infiniband/core/addr.c +++ b/drivers/infiniband/core/addr.c @@ -35,6 +35,7 @@ #include #include #include +#include #include MODULE_AUTHOR("Sean Hefty"); @@ -326,25 +327,22 @@ void rdma_addr_cancel(struct rdma_dev_addr *addr) } EXPORT_SYMBOL(rdma_addr_cancel); -static int addr_arp_recv(struct sk_buff *skb, struct net_device *dev, - struct packet_type *pkt, struct net_device *orig_dev) +static int netevent_callback(struct notifier_block *self, unsigned long event, + void *ctx) { - struct arphdr *arp_hdr; + if (event == NETEVENT_NEIGH_UPDATE) { + struct neighbour *neigh = ctx; - arp_hdr = (struct arphdr *) skb->nh.raw; - - if (arp_hdr->ar_op == htons(ARPOP_REQUEST) || - arp_hdr->ar_op == htons(ARPOP_REPLY)) - set_timeout(jiffies); - - kfree_skb(skb); + if (neigh->dev->type == ARPHRD_INFINIBAND && + (neigh->nud_state & NUD_VALID)) { + set_timeout(jiffies); + } + } return 0; } -static struct packet_type addr_arp = { - .type = __constant_htons(ETH_P_ARP), - .func = addr_arp_recv, - .af_packet_priv = (void*) 1, +static struct notifier_block nb = { + .notifier_call = netevent_callback }; static int addr_init(void) @@ -353,13 +351,13 @@ static int addr_init(void) if (!addr_wq) return -ENOMEM; - dev_add_pack(&addr_arp); + register_netevent_notifier(&nb); return 0; } static void addr_cleanup(void) { - dev_remove_pack(&addr_arp); + unregister_netevent_notifier(&nb); destroy_workqueue(addr_wq); } -- cgit From bf74c7479ef47652005a2418eeb0d867451690da Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Wed, 26 Jul 2006 16:02:53 +0300 Subject: IB/mthca: Fix mthca_array_clear() thinko mthca_array_clear() does not clear the slot if the used count is positive. This leads to crashes in mthca_qp_event() since that uses mthca_array_get() to check that the qp is valid. Discovered by Ali Ayoub. Signed-off-by: Michael S. Tsirkin Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mthca/mthca_allocator.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mthca/mthca_allocator.c b/drivers/infiniband/hw/mthca/mthca_allocator.c index 9ba3211cef7c..848e583273d4 100644 --- a/drivers/infiniband/hw/mthca/mthca_allocator.c +++ b/drivers/infiniband/hw/mthca/mthca_allocator.c @@ -144,7 +144,9 @@ void mthca_array_clear(struct mthca_array *array, int index) if (--array->page_list[p].used == 0) { free_page((unsigned long) array->page_list[p].page); array->page_list[p].page = NULL; - } + } else + array->page_list[p].page[index & (PAGE_SIZE / + sizeof (void *) - 1)] = NULL; if (array->page_list[p].used < 0) pr_debug("Array %p index %d page %d with ref count %d < 0\n", -- cgit From 69e9fbb460fa8766428960439841ffcf565032c1 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Thu, 3 Aug 2006 09:44:22 -0700 Subject: IB/mthca: Clean up mthca array index mask Define a constant MTHCA_ARRAY_MASK to replace repeated uses of (PAGE_SIZE / sizeof (void *) - 1) in mthca array code. Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mthca/mthca_allocator.c | 15 +++++++-------- 1 file changed, 7 insertions(+), 8 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mthca/mthca_allocator.c b/drivers/infiniband/hw/mthca/mthca_allocator.c index 848e583273d4..25157f57a6d0 100644 --- a/drivers/infiniband/hw/mthca/mthca_allocator.c +++ b/drivers/infiniband/hw/mthca/mthca_allocator.c @@ -108,14 +108,15 @@ void mthca_alloc_cleanup(struct mthca_alloc *alloc) * serialize access to the array. */ +#define MTHCA_ARRAY_MASK (PAGE_SIZE / sizeof (void *) - 1) + void *mthca_array_get(struct mthca_array *array, int index) { int p = (index * sizeof (void *)) >> PAGE_SHIFT; - if (array->page_list[p].page) { - int i = index & (PAGE_SIZE / sizeof (void *) - 1); - return array->page_list[p].page[i]; - } else + if (array->page_list[p].page) + return array->page_list[p].page[index & MTHCA_ARRAY_MASK]; + else return NULL; } @@ -130,8 +131,7 @@ int mthca_array_set(struct mthca_array *array, int index, void *value) if (!array->page_list[p].page) return -ENOMEM; - array->page_list[p].page[index & (PAGE_SIZE / sizeof (void *) - 1)] = - value; + array->page_list[p].page[index & MTHCA_ARRAY_MASK] = value; ++array->page_list[p].used; return 0; @@ -145,8 +145,7 @@ void mthca_array_clear(struct mthca_array *array, int index) free_page((unsigned long) array->page_list[p].page); array->page_list[p].page = NULL; } else - array->page_list[p].page[index & (PAGE_SIZE / - sizeof (void *) - 1)] = NULL; + array->page_list[p].page[index & MTHCA_ARRAY_MASK] = NULL; if (array->page_list[p].used < 0) pr_debug("Array %p index %d page %d with ref count %d < 0\n", -- cgit From 75df23e229acab85b704f4603bdf5efdc7960e6a Mon Sep 17 00:00:00 2001 From: Sean Hefty Date: Tue, 25 Jul 2006 19:52:01 +0300 Subject: IB/cm: Fix error handling in ib_send_cm_req Report error code rather than success (0) on failure allocating timewait_info in ib_send_cm_req(). Signed-off-by: Sean Hefty Signed-off-by: Roland Dreier --- drivers/infiniband/core/cm.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c index f85c97f7500a..0de335b7bfc2 100644 --- a/drivers/infiniband/core/cm.c +++ b/drivers/infiniband/core/cm.c @@ -975,8 +975,10 @@ int ib_send_cm_req(struct ib_cm_id *cm_id, cm_id_priv->timewait_info = cm_create_timewait_info(cm_id_priv-> id.local_id); - if (IS_ERR(cm_id_priv->timewait_info)) + if (IS_ERR(cm_id_priv->timewait_info)) { + ret = PTR_ERR(cm_id_priv->timewait_info); goto out; + } ret = cm_init_av_by_path(param->primary_path, &cm_id_priv->av); if (ret) -- cgit From d916a8f1b43b358685b1672390ead11f2d3b74c6 Mon Sep 17 00:00:00 2001 From: Ishai Rabinovitz Date: Tue, 25 Jul 2006 19:54:09 +0300 Subject: IB/srp: Fix crash in srp_reconnect_target Protect against srp_reset_device() clearing the req_queue while srp_reconnect_target() is in progress (note that state change at the top of srp_reconnect_target() is not sufficient for this since srp_reset_device() ignores the state). Signed-off-by: Ishai Rabinovitz Signed-off-by: Michael S. Tsirkin Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/srp/ib_srp.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c index 8f472e7113b4..ff94e4e32691 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.c +++ b/drivers/infiniband/ulp/srp/ib_srp.c @@ -526,8 +526,10 @@ static int srp_reconnect_target(struct srp_target_port *target) while (ib_poll_cq(target->cq, 1, &wc) > 0) ; /* nothing */ + spin_lock_irq(target->scsi_host->host_lock); list_for_each_entry_safe(req, tmp, &target->req_queue, list) srp_reset_req(target, req); + spin_unlock_irq(target->scsi_host->host_lock); target->rx_head = 0; target->tx_head = 0; -- cgit From 559ce8f150d7d031c79c4d79173860f1bdfe3ce4 Mon Sep 17 00:00:00 2001 From: Ishai Rabinovitz Date: Thu, 3 Aug 2006 10:35:43 -0700 Subject: IB/srp: Work around data corruption bug on Mellanox targets Data corruption has been seen with Mellanox SRP targets when FMRs create a memory region with I/O virtual address != 0. Add a workaround that disables FMR merging for Mellanox targets (OUI 0002c9). Signed-off-by: Ishai Rabinovitz Signed-off-by: Michael S. Tsirkin Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/srp/ib_srp.c | 17 +++++++++++++++-- 1 file changed, 15 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c index ff94e4e32691..8257d5a2c8f8 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.c +++ b/drivers/infiniband/ulp/srp/ib_srp.c @@ -77,6 +77,14 @@ MODULE_PARM_DESC(topspin_workarounds, static const u8 topspin_oui[3] = { 0x00, 0x05, 0xad }; +static int mellanox_workarounds = 1; + +module_param(mellanox_workarounds, int, 0444); +MODULE_PARM_DESC(mellanox_workarounds, + "Enable workarounds for Mellanox SRP target bugs if != 0"); + +static const u8 mellanox_oui[3] = { 0x00, 0x02, 0xc9 }; + static void srp_add_one(struct ib_device *device); static void srp_remove_one(struct ib_device *device); static void srp_completion(struct ib_cq *cq, void *target_ptr); @@ -569,7 +577,7 @@ err: return ret; } -static int srp_map_fmr(struct srp_device *dev, struct scatterlist *scat, +static int srp_map_fmr(struct srp_target_port *target, struct scatterlist *scat, int sg_cnt, struct srp_request *req, struct srp_direct_buf *buf) { @@ -579,10 +587,15 @@ static int srp_map_fmr(struct srp_device *dev, struct scatterlist *scat, int page_cnt; int i, j; int ret; + struct srp_device *dev = target->srp_host->dev; if (!dev->fmr_pool) return -ENODEV; + if ((sg_dma_address(&scat[0]) & ~dev->fmr_page_mask) && + mellanox_workarounds && !memcmp(&target->ioc_guid, mellanox_oui, 3)) + return -EINVAL; + len = page_cnt = 0; for (i = 0; i < sg_cnt; ++i) { if (sg_dma_address(&scat[i]) & ~dev->fmr_page_mask) { @@ -685,7 +698,7 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target, buf->va = cpu_to_be64(sg_dma_address(scat)); buf->key = cpu_to_be32(target->srp_host->dev->mr->rkey); buf->len = cpu_to_be32(sg_dma_len(scat)); - } else if (srp_map_fmr(target->srp_host->dev, scat, count, req, + } else if (srp_map_fmr(target, scat, count, req, (void *) cmd->add_data)) { /* * FMR mapping failed, and the scatterlist has more -- cgit From 8ddc7c5326064434048ec1ecfe57659e08345cc1 Mon Sep 17 00:00:00 2001 From: Or Gerlitz Date: Thu, 13 Jul 2006 11:00:39 +0300 Subject: IB/ipoib: Remove broken link from Kconfig and documentation Remove references to the IPoIB IETF working group as it has been closed. Signed-off-by: Or Gerlitz Signed-off-by: Roland Dreier --- drivers/infiniband/ulp/ipoib/Kconfig | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/ulp/ipoib/Kconfig b/drivers/infiniband/ulp/ipoib/Kconfig index 13d6d01c72c0..d74653d7de1c 100644 --- a/drivers/infiniband/ulp/ipoib/Kconfig +++ b/drivers/infiniband/ulp/ipoib/Kconfig @@ -6,8 +6,7 @@ config INFINIBAND_IPOIB transports IP packets over InfiniBand so you can use your IB device as a fancy NIC. - The IPoIB protocol is defined by the IETF ipoib working - group: . + See Documentation/infiniband/ipoib.txt for more information config INFINIBAND_IPOIB_DEBUG bool "IP-over-InfiniBand debugging" if EMBEDDED -- cgit From fd60ae404f104f12369e654af9cf03b1f1047661 Mon Sep 17 00:00:00 2001 From: Jack Morgenstein Date: Thu, 3 Aug 2006 10:56:42 -0700 Subject: IB/uverbs: Avoid a crash on device hot remove Wait until all users have closed their device context before allowing device unregistration to complete. This prevents a crash caused by referring to stale data structures. A better solution would be to have a way to revoke contexts rather than waiting for userspace to close the context, but that's a much bigger change that will have to wait. For now let's at least avoid the crash. Signed-off-by: Jack Morgenstein Signed-off-by: Michael S. Tsirkin Signed-off-by: Roland Dreier --- drivers/infiniband/core/uverbs.h | 2 ++ drivers/infiniband/core/uverbs_main.c | 8 +++++++- 2 files changed, 9 insertions(+), 1 deletion(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/uverbs.h b/drivers/infiniband/core/uverbs.h index bb9bee56a824..102a59c033ff 100644 --- a/drivers/infiniband/core/uverbs.h +++ b/drivers/infiniband/core/uverbs.h @@ -42,6 +42,7 @@ #include #include #include +#include #include #include @@ -69,6 +70,7 @@ struct ib_uverbs_device { struct kref ref; + struct completion comp; int devnum; struct cdev *dev; struct class_device *class_dev; diff --git a/drivers/infiniband/core/uverbs_main.c b/drivers/infiniband/core/uverbs_main.c index e725cccc7cde..4e16314e8e6d 100644 --- a/drivers/infiniband/core/uverbs_main.c +++ b/drivers/infiniband/core/uverbs_main.c @@ -122,7 +122,7 @@ static void ib_uverbs_release_dev(struct kref *ref) struct ib_uverbs_device *dev = container_of(ref, struct ib_uverbs_device, ref); - kfree(dev); + complete(&dev->comp); } void ib_uverbs_release_ucq(struct ib_uverbs_file *file, @@ -740,6 +740,7 @@ static void ib_uverbs_add_one(struct ib_device *device) return; kref_init(&uverbs_dev->ref); + init_completion(&uverbs_dev->comp); spin_lock(&map_lock); uverbs_dev->devnum = find_first_zero_bit(dev_map, IB_UVERBS_MAX_DEVICES); @@ -793,6 +794,8 @@ err_cdev: err: kref_put(&uverbs_dev->ref, ib_uverbs_release_dev); + wait_for_completion(&uverbs_dev->comp); + kfree(uverbs_dev); return; } @@ -812,7 +815,10 @@ static void ib_uverbs_remove_one(struct ib_device *device) spin_unlock(&map_lock); clear_bit(uverbs_dev->devnum, dev_map); + kref_put(&uverbs_dev->ref, ib_uverbs_release_dev); + wait_for_completion(&uverbs_dev->comp); + kfree(uverbs_dev); } static int uverbs_event_get_sb(struct file_system_type *fs_type, int flags, -- cgit From e54b82d739d4a2ef992976c8c0692cdf89286420 Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Thu, 10 Aug 2006 10:46:56 -0700 Subject: IB/mthca: Make fence flag work for send work requests The fence bit needs to be set in the doorbell too, not just the WQE. Signed-off-by: Michael S. Tsirkin Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mthca/mthca_qp.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mthca/mthca_qp.c b/drivers/infiniband/hw/mthca/mthca_qp.c index cd8b6721ac9c..157b4f8ac407 100644 --- a/drivers/infiniband/hw/mthca/mthca_qp.c +++ b/drivers/infiniband/hw/mthca/mthca_qp.c @@ -99,6 +99,10 @@ enum { MTHCA_QP_BIT_RSC = 1 << 3 }; +enum { + MTHCA_SEND_DOORBELL_FENCE = 1 << 5 +}; + struct mthca_qp_path { __be32 port_pkey; u8 rnr_retry; @@ -1502,7 +1506,7 @@ int mthca_tavor_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, int i; int size; int size0 = 0; - u32 f0 = 0; + u32 f0; int ind; u8 op0 = 0; @@ -1686,6 +1690,8 @@ int mthca_tavor_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, if (!size0) { size0 = size; op0 = mthca_opcode[wr->opcode]; + f0 = wr->send_flags & IB_SEND_FENCE ? + MTHCA_SEND_DOORBELL_FENCE : 0; } ++ind; @@ -1843,7 +1849,7 @@ int mthca_arbel_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, int i; int size; int size0 = 0; - u32 f0 = 0; + u32 f0; int ind; u8 op0 = 0; @@ -2051,6 +2057,8 @@ int mthca_arbel_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr, if (!size0) { size0 = size; op0 = mthca_opcode[wr->opcode]; + f0 = wr->send_flags & IB_SEND_FENCE ? + MTHCA_SEND_DOORBELL_FENCE : 0; } ++ind; -- cgit From a19aa5c5fdda8b556ab238177ee27c5ef7873c94 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Fri, 11 Aug 2006 08:56:57 -0700 Subject: IB/mthca: Fix potential AB-BA deadlock with CQ locks When destroying a QP, mthca locks both the QP's send CQ and receive CQ. However, the following scenario is perfectly valid: QP_a: send_cq == CQ_x, recv_cq == CQ_y QP_b: send_cq == CQ_y, recv_cq == CQ_x The old mthca code simply locked send_cq and then recv_cq, which in this case could lead to an AB-BA deadlock if QP_a and QP_b were destroyed simultaneously. We can fix this by changing the locking code to lock the CQ with the lower CQ number first, which will create a consistent lock ordering. Also, the second CQ is locked with spin_lock_nested() to tell lockdep that we know what we're doing with the lock nesting. This bug was found by lockdep. Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mthca/mthca_provider.h | 4 +-- drivers/infiniband/hw/mthca/mthca_qp.c | 42 ++++++++++++++++++++-------- 2 files changed, 32 insertions(+), 14 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mthca/mthca_provider.h b/drivers/infiniband/hw/mthca/mthca_provider.h index 8de2887ba15c..9a5bece3fa5c 100644 --- a/drivers/infiniband/hw/mthca/mthca_provider.h +++ b/drivers/infiniband/hw/mthca/mthca_provider.h @@ -136,8 +136,8 @@ struct mthca_ah { * We have one global lock that protects dev->cq/qp_table. Each * struct mthca_cq/qp also has its own lock. An individual qp lock * may be taken inside of an individual cq lock. Both cqs attached to - * a qp may be locked, with the send cq locked first. No other - * nesting should be done. + * a qp may be locked, with the cq with the lower cqn locked first. + * No other nesting should be done. * * Each struct mthca_cq/qp also has an ref count, protected by the * corresponding table lock. The pointer from the cq/qp_table to the diff --git a/drivers/infiniband/hw/mthca/mthca_qp.c b/drivers/infiniband/hw/mthca/mthca_qp.c index 157b4f8ac407..2e8f6f36e0a5 100644 --- a/drivers/infiniband/hw/mthca/mthca_qp.c +++ b/drivers/infiniband/hw/mthca/mthca_qp.c @@ -1263,6 +1263,32 @@ int mthca_alloc_qp(struct mthca_dev *dev, return 0; } +static void mthca_lock_cqs(struct mthca_cq *send_cq, struct mthca_cq *recv_cq) +{ + if (send_cq == recv_cq) + spin_lock_irq(&send_cq->lock); + else if (send_cq->cqn < recv_cq->cqn) { + spin_lock_irq(&send_cq->lock); + spin_lock_nested(&recv_cq->lock, SINGLE_DEPTH_NESTING); + } else { + spin_lock_irq(&recv_cq->lock); + spin_lock_nested(&send_cq->lock, SINGLE_DEPTH_NESTING); + } +} + +static void mthca_unlock_cqs(struct mthca_cq *send_cq, struct mthca_cq *recv_cq) +{ + if (send_cq == recv_cq) + spin_unlock_irq(&send_cq->lock); + else if (send_cq->cqn < recv_cq->cqn) { + spin_unlock(&recv_cq->lock); + spin_unlock_irq(&send_cq->lock); + } else { + spin_unlock(&send_cq->lock); + spin_unlock_irq(&recv_cq->lock); + } +} + int mthca_alloc_sqp(struct mthca_dev *dev, struct mthca_pd *pd, struct mthca_cq *send_cq, @@ -1315,17 +1341,13 @@ int mthca_alloc_sqp(struct mthca_dev *dev, * Lock CQs here, so that CQ polling code can do QP lookup * without taking a lock. */ - spin_lock_irq(&send_cq->lock); - if (send_cq != recv_cq) - spin_lock(&recv_cq->lock); + mthca_lock_cqs(send_cq, recv_cq); spin_lock(&dev->qp_table.lock); mthca_array_clear(&dev->qp_table.qp, mqpn); spin_unlock(&dev->qp_table.lock); - if (send_cq != recv_cq) - spin_unlock(&recv_cq->lock); - spin_unlock_irq(&send_cq->lock); + mthca_unlock_cqs(send_cq, recv_cq); err_out: dma_free_coherent(&dev->pdev->dev, sqp->header_buf_size, @@ -1359,9 +1381,7 @@ void mthca_free_qp(struct mthca_dev *dev, * Lock CQs here, so that CQ polling code can do QP lookup * without taking a lock. */ - spin_lock_irq(&send_cq->lock); - if (send_cq != recv_cq) - spin_lock(&recv_cq->lock); + mthca_lock_cqs(send_cq, recv_cq); spin_lock(&dev->qp_table.lock); mthca_array_clear(&dev->qp_table.qp, @@ -1369,9 +1389,7 @@ void mthca_free_qp(struct mthca_dev *dev, --qp->refcount; spin_unlock(&dev->qp_table.lock); - if (send_cq != recv_cq) - spin_unlock(&recv_cq->lock); - spin_unlock_irq(&send_cq->lock); + mthca_unlock_cqs(send_cq, recv_cq); wait_event(qp->wait, !get_qp_refcount(dev, qp)); -- cgit From acaea9ee460d0ba5a14f0066ba26cfa43dd5fdf3 Mon Sep 17 00:00:00 2001 From: Jack Morgenstein Date: Tue, 15 Aug 2006 17:20:50 +0300 Subject: IB/core: Fix SM LID/LID change with client reregister set After commit 12bbb2b7be7f5564952ebe0196623e97464b8ac5, when SM LID change or LID change MAD also has a client reregistration bit set, only CLIENT_REREGISTER event is generated. As a result, the sa_query module and the cache module don't update the port information, and ULPs (e.g. IPoIB) stop working. This is the regression we observe as compared to 2.6.17. Rather than generate multiple events (which would have negative performance impact), let us simply let cache and SA query respond to reregister event in the same way as to LID and SM change events. Signed-off-by: Jack Morgenstein Signed-off-by: Michael S. Tsirkin Signed-off-by: Roland Dreier --- drivers/infiniband/core/cache.c | 3 ++- drivers/infiniband/core/sa_query.c | 3 ++- 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/core/cache.c b/drivers/infiniband/core/cache.c index e05ca2cdc73f..75313ade2e0d 100644 --- a/drivers/infiniband/core/cache.c +++ b/drivers/infiniband/core/cache.c @@ -301,7 +301,8 @@ static void ib_cache_event(struct ib_event_handler *handler, event->event == IB_EVENT_PORT_ACTIVE || event->event == IB_EVENT_LID_CHANGE || event->event == IB_EVENT_PKEY_CHANGE || - event->event == IB_EVENT_SM_CHANGE) { + event->event == IB_EVENT_SM_CHANGE || + event->event == IB_EVENT_CLIENT_REREGISTER) { work = kmalloc(sizeof *work, GFP_ATOMIC); if (work) { INIT_WORK(&work->work, ib_cache_task, work); diff --git a/drivers/infiniband/core/sa_query.c b/drivers/infiniband/core/sa_query.c index aeda484ffd82..d6b84226bba7 100644 --- a/drivers/infiniband/core/sa_query.c +++ b/drivers/infiniband/core/sa_query.c @@ -405,7 +405,8 @@ static void ib_sa_event(struct ib_event_handler *handler, struct ib_event *event event->event == IB_EVENT_PORT_ACTIVE || event->event == IB_EVENT_LID_CHANGE || event->event == IB_EVENT_PKEY_CHANGE || - event->event == IB_EVENT_SM_CHANGE) { + event->event == IB_EVENT_SM_CHANGE || + event->event == IB_EVENT_CLIENT_REREGISTER) { struct ib_sa_device *sa_dev; sa_dev = container_of(handler, typeof(*sa_dev), event_handler); -- cgit From 5beba53230351b2d77c317c22e66c415f2ebaf02 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Fri, 18 Aug 2006 10:41:46 -0700 Subject: IB/mthca: No userspace SRQs if HCA doesn't have SRQ support Leave all SRQ methods out of the device's uverbs_cmd_mask if the device doesn't have SRQ support (because of ancient firmware) so that we don't allow userspace to call the driver's create_srq method. This fixes a userspace-triggerable oops caused by ib_uverbs_create_srq() following the device's ->create_srq function pointer, which will be NULL if the device doesn't support SRQs. Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mthca/mthca_provider.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c index 230ae21db8fd..265b1d1c4a62 100644 --- a/drivers/infiniband/hw/mthca/mthca_provider.c +++ b/drivers/infiniband/hw/mthca/mthca_provider.c @@ -1287,11 +1287,7 @@ int mthca_register_device(struct mthca_dev *dev) (1ull << IB_USER_VERBS_CMD_MODIFY_QP) | (1ull << IB_USER_VERBS_CMD_DESTROY_QP) | (1ull << IB_USER_VERBS_CMD_ATTACH_MCAST) | - (1ull << IB_USER_VERBS_CMD_DETACH_MCAST) | - (1ull << IB_USER_VERBS_CMD_CREATE_SRQ) | - (1ull << IB_USER_VERBS_CMD_MODIFY_SRQ) | - (1ull << IB_USER_VERBS_CMD_QUERY_SRQ) | - (1ull << IB_USER_VERBS_CMD_DESTROY_SRQ); + (1ull << IB_USER_VERBS_CMD_DETACH_MCAST); dev->ib_dev.node_type = IB_NODE_CA; dev->ib_dev.phys_port_cnt = dev->limits.num_ports; dev->ib_dev.dma_device = &dev->pdev->dev; @@ -1316,6 +1312,11 @@ int mthca_register_device(struct mthca_dev *dev) dev->ib_dev.modify_srq = mthca_modify_srq; dev->ib_dev.query_srq = mthca_query_srq; dev->ib_dev.destroy_srq = mthca_destroy_srq; + dev->ib_dev.uverbs_cmd_mask |= + (1ull << IB_USER_VERBS_CMD_CREATE_SRQ) | + (1ull << IB_USER_VERBS_CMD_MODIFY_SRQ) | + (1ull << IB_USER_VERBS_CMD_QUERY_SRQ) | + (1ull << IB_USER_VERBS_CMD_DESTROY_SRQ); if (mthca_is_memfree(dev)) dev->ib_dev.post_srq_recv = mthca_arbel_post_srq_recv; -- cgit From 834ac73d4bc804db8ccb3f2a517e36db5f6bc4bd Mon Sep 17 00:00:00 2001 From: "Michael S. Tsirkin" Date: Tue, 22 Aug 2006 22:45:06 +0300 Subject: IB/mthca: Update HCA firmware revisions Update the driver's list of HCA firmware revisions to make sure people running Sinai firmware older than 1.1.0 get a message suggesting a firmware upgrade. Update the Arbel versions as well while we are at it. Signed-off-by: Michael S. Tsirkin Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mthca/mthca_main.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mthca/mthca_main.c b/drivers/infiniband/hw/mthca/mthca_main.c index 557cde3a4563..7b82c1907f04 100644 --- a/drivers/infiniband/hw/mthca/mthca_main.c +++ b/drivers/infiniband/hw/mthca/mthca_main.c @@ -967,12 +967,12 @@ static struct { } mthca_hca_table[] = { [TAVOR] = { .latest_fw = MTHCA_FW_VER(3, 4, 0), .flags = 0 }, - [ARBEL_COMPAT] = { .latest_fw = MTHCA_FW_VER(4, 7, 400), + [ARBEL_COMPAT] = { .latest_fw = MTHCA_FW_VER(4, 7, 600), .flags = MTHCA_FLAG_PCIE }, - [ARBEL_NATIVE] = { .latest_fw = MTHCA_FW_VER(5, 1, 0), + [ARBEL_NATIVE] = { .latest_fw = MTHCA_FW_VER(5, 1, 400), .flags = MTHCA_FLAG_MEMFREE | MTHCA_FLAG_PCIE }, - [SINAI] = { .latest_fw = MTHCA_FW_VER(1, 0, 800), + [SINAI] = { .latest_fw = MTHCA_FW_VER(1, 1, 0), .flags = MTHCA_FLAG_MEMFREE | MTHCA_FLAG_PCIE | MTHCA_FLAG_SINAI_OPT } -- cgit From 5a4e6dccbc0cd1b726820b782daebf887dcb95e9 Mon Sep 17 00:00:00 2001 From: Roland Dreier Date: Thu, 31 Aug 2006 16:43:06 -0700 Subject: IB/mthca: Use IRQ safe locks to protect allocation bitmaps It is supposed to be OK to call mthca_create_ah() and mthca_destroy_ah() from any context. However, for mem-full HCAs, these functions use the mthca_alloc() and mthca_free() bitmap helpers, and those helpers use non-IRQ-safe spin_lock() internally. Lockdep correctly warns that this could lead to a deadlock. Fix this by changing mthca_alloc() and mthca_free() to use spin_lock_irqsave(). Signed-off-by: Roland Dreier --- drivers/infiniband/hw/mthca/mthca_allocator.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-) (limited to 'drivers/infiniband') diff --git a/drivers/infiniband/hw/mthca/mthca_allocator.c b/drivers/infiniband/hw/mthca/mthca_allocator.c index 25157f57a6d0..f930e55b58fc 100644 --- a/drivers/infiniband/hw/mthca/mthca_allocator.c +++ b/drivers/infiniband/hw/mthca/mthca_allocator.c @@ -41,9 +41,11 @@ /* Trivial bitmap-based allocator */ u32 mthca_alloc(struct mthca_alloc *alloc) { + unsigned long flags; u32 obj; - spin_lock(&alloc->lock); + spin_lock_irqsave(&alloc->lock, flags); + obj = find_next_zero_bit(alloc->table, alloc->max, alloc->last); if (obj >= alloc->max) { alloc->top = (alloc->top + alloc->max) & alloc->mask; @@ -56,19 +58,24 @@ u32 mthca_alloc(struct mthca_alloc *alloc) } else obj = -1; - spin_unlock(&alloc->lock); + spin_unlock_irqrestore(&alloc->lock, flags); return obj; } void mthca_free(struct mthca_alloc *alloc, u32 obj) { + unsigned long flags; + obj &= alloc->max - 1; - spin_lock(&alloc->lock); + + spin_lock_irqsave(&alloc->lock, flags); + clear_bit(obj, alloc->table); alloc->last = min(alloc->last, obj); alloc->top = (alloc->top + alloc->max) & alloc->mask; - spin_unlock(&alloc->lock); + + spin_unlock_irqrestore(&alloc->lock, flags); } int mthca_alloc_init(struct mthca_alloc *alloc, u32 num, u32 mask, -- cgit