diff options
Diffstat (limited to 'drivers/infiniband')
41 files changed, 653 insertions, 539 deletions
diff --git a/drivers/infiniband/core/cm.c b/drivers/infiniband/core/cm.c index 7cfedb8d9bc..86fee43502c 100644 --- a/drivers/infiniband/core/cm.c +++ b/drivers/infiniband/core/cm.c @@ -34,6 +34,8 @@ * * $Id: cm.c 2821 2005-07-08 17:07:28Z sean.hefty $ */ + +#include <linux/completion.h> #include <linux/dma-mapping.h> #include <linux/err.h> #include <linux/idr.h> @@ -122,7 +124,7 @@ struct cm_id_private { struct rb_node service_node; struct rb_node sidr_id_node; spinlock_t lock; /* Do not acquire inside cm.lock */ - wait_queue_head_t wait; + struct completion comp; atomic_t refcount; struct ib_mad_send_buf *msg; @@ -159,7 +161,7 @@ static void cm_work_handler(void *data); static inline void cm_deref_id(struct cm_id_private *cm_id_priv) { if (atomic_dec_and_test(&cm_id_priv->refcount)) - wake_up(&cm_id_priv->wait); + complete(&cm_id_priv->comp); } static int cm_alloc_msg(struct cm_id_private *cm_id_priv, @@ -559,7 +561,7 @@ struct ib_cm_id *ib_create_cm_id(struct ib_device *device, goto error; spin_lock_init(&cm_id_priv->lock); - init_waitqueue_head(&cm_id_priv->wait); + init_completion(&cm_id_priv->comp); INIT_LIST_HEAD(&cm_id_priv->work_list); atomic_set(&cm_id_priv->work_count, -1); atomic_set(&cm_id_priv->refcount, 1); @@ -724,8 +726,8 @@ retest: } cm_free_id(cm_id->local_id); - atomic_dec(&cm_id_priv->refcount); - wait_event(cm_id_priv->wait, !atomic_read(&cm_id_priv->refcount)); + cm_deref_id(cm_id_priv); + wait_for_completion(&cm_id_priv->comp); while ((work = cm_dequeue_work(cm_id_priv)) != NULL) cm_free_work(work); if (cm_id_priv->private_data && cm_id_priv->private_data_len) diff --git a/drivers/infiniband/core/mad.c b/drivers/infiniband/core/mad.c index 3a702da83e4..5ad41a64314 100644 --- a/drivers/infiniband/core/mad.c +++ b/drivers/infiniband/core/mad.c @@ -228,10 +228,7 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, goto error1; } /* Make sure class supplied is consistent with RMPP */ - if (ib_is_mad_class_rmpp(mad_reg_req->mgmt_class)) { - if (!rmpp_version) - goto error1; - } else { + if (!ib_is_mad_class_rmpp(mad_reg_req->mgmt_class)) { if (rmpp_version) goto error1; } @@ -355,7 +352,7 @@ struct ib_mad_agent *ib_register_mad_agent(struct ib_device *device, INIT_WORK(&mad_agent_priv->local_work, local_completions, mad_agent_priv); atomic_set(&mad_agent_priv->refcount, 1); - init_waitqueue_head(&mad_agent_priv->wait); + init_completion(&mad_agent_priv->comp); return &mad_agent_priv->agent; @@ -470,7 +467,7 @@ struct ib_mad_agent *ib_register_mad_snoop(struct ib_device *device, mad_snoop_priv->agent.qp = port_priv->qp_info[qpn].qp; mad_snoop_priv->agent.port_num = port_num; mad_snoop_priv->mad_snoop_flags = mad_snoop_flags; - init_waitqueue_head(&mad_snoop_priv->wait); + init_completion(&mad_snoop_priv->comp); mad_snoop_priv->snoop_index = register_snoop_agent( &port_priv->qp_info[qpn], mad_snoop_priv); @@ -489,6 +486,18 @@ error1: } EXPORT_SYMBOL(ib_register_mad_snoop); +static inline void deref_mad_agent(struct ib_mad_agent_private *mad_agent_priv) +{ + if (atomic_dec_and_test(&mad_agent_priv->refcount)) + complete(&mad_agent_priv->comp); +} + +static inline void deref_snoop_agent(struct ib_mad_snoop_private *mad_snoop_priv) +{ + if (atomic_dec_and_test(&mad_snoop_priv->refcount)) + complete(&mad_snoop_priv->comp); +} + static void unregister_mad_agent(struct ib_mad_agent_private *mad_agent_priv) { struct ib_mad_port_private *port_priv; @@ -512,9 +521,8 @@ static void unregister_mad_agent(struct ib_mad_agent_private *mad_agent_priv) flush_workqueue(port_priv->wq); ib_cancel_rmpp_recvs(mad_agent_priv); - atomic_dec(&mad_agent_priv->refcount); - wait_event(mad_agent_priv->wait, - !atomic_read(&mad_agent_priv->refcount)); + deref_mad_agent(mad_agent_priv); + wait_for_completion(&mad_agent_priv->comp); kfree(mad_agent_priv->reg_req); ib_dereg_mr(mad_agent_priv->agent.mr); @@ -532,9 +540,8 @@ static void unregister_mad_snoop(struct ib_mad_snoop_private *mad_snoop_priv) atomic_dec(&qp_info->snoop_count); spin_unlock_irqrestore(&qp_info->snoop_lock, flags); - atomic_dec(&mad_snoop_priv->refcount); - wait_event(mad_snoop_priv->wait, - !atomic_read(&mad_snoop_priv->refcount)); + deref_snoop_agent(mad_snoop_priv); + wait_for_completion(&mad_snoop_priv->comp); kfree(mad_snoop_priv); } @@ -603,8 +610,7 @@ static void snoop_send(struct ib_mad_qp_info *qp_info, spin_unlock_irqrestore(&qp_info->snoop_lock, flags); mad_snoop_priv->agent.snoop_handler(&mad_snoop_priv->agent, send_buf, mad_send_wc); - if (atomic_dec_and_test(&mad_snoop_priv->refcount)) - wake_up(&mad_snoop_priv->wait); + deref_snoop_agent(mad_snoop_priv); spin_lock_irqsave(&qp_info->snoop_lock, flags); } spin_unlock_irqrestore(&qp_info->snoop_lock, flags); @@ -629,8 +635,7 @@ static void snoop_recv(struct ib_mad_qp_info *qp_info, spin_unlock_irqrestore(&qp_info->snoop_lock, flags); mad_snoop_priv->agent.recv_handler(&mad_snoop_priv->agent, mad_recv_wc); - if (atomic_dec_and_test(&mad_snoop_priv->refcount)) - wake_up(&mad_snoop_priv->wait); + deref_snoop_agent(mad_snoop_priv); spin_lock_irqsave(&qp_info->snoop_lock, flags); } spin_unlock_irqrestore(&qp_info->snoop_lock, flags); @@ -971,8 +976,7 @@ void ib_free_send_mad(struct ib_mad_send_buf *send_buf) free_send_rmpp_list(mad_send_wr); kfree(send_buf->mad); - if (atomic_dec_and_test(&mad_agent_priv->refcount)) - wake_up(&mad_agent_priv->wait); + deref_mad_agent(mad_agent_priv); } EXPORT_SYMBOL(ib_free_send_mad); @@ -1760,8 +1764,7 @@ static void ib_mad_complete_recv(struct ib_mad_agent_private *mad_agent_priv, mad_recv_wc = ib_process_rmpp_recv_wc(mad_agent_priv, mad_recv_wc); if (!mad_recv_wc) { - if (atomic_dec_and_test(&mad_agent_priv->refcount)) - wake_up(&mad_agent_priv->wait); + deref_mad_agent(mad_agent_priv); return; } } @@ -1773,8 +1776,7 @@ static void ib_mad_complete_recv(struct ib_mad_agent_private *mad_agent_priv, if (!mad_send_wr) { spin_unlock_irqrestore(&mad_agent_priv->lock, flags); ib_free_recv_mad(mad_recv_wc); - if (atomic_dec_and_test(&mad_agent_priv->refcount)) - wake_up(&mad_agent_priv->wait); + deref_mad_agent(mad_agent_priv); return; } ib_mark_mad_done(mad_send_wr); @@ -1793,8 +1795,7 @@ static void ib_mad_complete_recv(struct ib_mad_agent_private *mad_agent_priv, } else { mad_agent_priv->agent.recv_handler(&mad_agent_priv->agent, mad_recv_wc); - if (atomic_dec_and_test(&mad_agent_priv->refcount)) - wake_up(&mad_agent_priv->wait); + deref_mad_agent(mad_agent_priv); } } @@ -2024,8 +2025,7 @@ void ib_mad_complete_send_wr(struct ib_mad_send_wr_private *mad_send_wr, mad_send_wc); /* Release reference on agent taken when sending */ - if (atomic_dec_and_test(&mad_agent_priv->refcount)) - wake_up(&mad_agent_priv->wait); + deref_mad_agent(mad_agent_priv); return; done: spin_unlock_irqrestore(&mad_agent_priv->lock, flags); diff --git a/drivers/infiniband/core/mad_priv.h b/drivers/infiniband/core/mad_priv.h index 6c9c133d71e..b4fa28d3160 100644 --- a/drivers/infiniband/core/mad_priv.h +++ b/drivers/infiniband/core/mad_priv.h @@ -37,6 +37,7 @@ #ifndef __IB_MAD_PRIV_H__ #define __IB_MAD_PRIV_H__ +#include <linux/completion.h> #include <linux/pci.h> #include <linux/kthread.h> #include <linux/workqueue.h> @@ -108,7 +109,7 @@ struct ib_mad_agent_private { struct list_head rmpp_list; atomic_t refcount; - wait_queue_head_t wait; + struct completion comp; }; struct ib_mad_snoop_private { @@ -117,7 +118,7 @@ struct ib_mad_snoop_private { int snoop_index; int mad_snoop_flags; atomic_t refcount; - wait_queue_head_t wait; + struct completion comp; }; struct ib_mad_send_wr_private { diff --git a/drivers/infiniband/core/mad_rmpp.c b/drivers/infiniband/core/mad_rmpp.c index dfd4e588ce0..d4704e054e3 100644 --- a/drivers/infiniband/core/mad_rmpp.c +++ b/drivers/infiniband/core/mad_rmpp.c @@ -49,7 +49,7 @@ struct mad_rmpp_recv { struct list_head list; struct work_struct timeout_work; struct work_struct cleanup_work; - wait_queue_head_t wait; + struct completion comp; enum rmpp_state state; spinlock_t lock; atomic_t refcount; @@ -69,10 +69,16 @@ struct mad_rmpp_recv { u8 method; }; +static inline void deref_rmpp_recv(struct mad_rmpp_recv *rmpp_recv) +{ + if (atomic_dec_and_test(&rmpp_recv->refcount)) + complete(&rmpp_recv->comp); +} + static void destroy_rmpp_recv(struct mad_rmpp_recv *rmpp_recv) { - atomic_dec(&rmpp_recv->refcount); - wait_event(rmpp_recv->wait, !atomic_read(&rmpp_recv->refcount)); + deref_rmpp_recv(rmpp_recv); + wait_for_completion(&rmpp_recv->comp); ib_destroy_ah(rmpp_recv->ah); kfree(rmpp_recv); } @@ -253,7 +259,7 @@ create_rmpp_recv(struct ib_mad_agent_private *agent, goto error; rmpp_recv->agent = agent; - init_waitqueue_head(&rmpp_recv->wait); + init_completion(&rmpp_recv->comp); INIT_WORK(&rmpp_recv->timeout_work, recv_timeout_handler, rmpp_recv); INIT_WORK(&rmpp_recv->cleanup_work, recv_cleanup_handler, rmpp_recv); spin_lock_init(&rmpp_recv->lock); @@ -279,12 +285,6 @@ error: kfree(rmpp_recv); return NULL; } -static inline void deref_rmpp_recv(struct mad_rmpp_recv *rmpp_recv) -{ - if (atomic_dec_and_test(&rmpp_recv->refcount)) - wake_up(&rmpp_recv->wait); -} - static struct mad_rmpp_recv * find_rmpp_recv(struct ib_mad_agent_private *agent, struct ib_mad_recv_wc *mad_recv_wc) diff --git a/drivers/infiniband/core/sysfs.c b/drivers/infiniband/core/sysfs.c index 15121cb5a1f..21f9282c1b2 100644 --- a/drivers/infiniband/core/sysfs.c +++ b/drivers/infiniband/core/sysfs.c @@ -336,7 +336,7 @@ static ssize_t show_pma_counter(struct ib_port *p, struct port_attribute *attr, switch (width) { case 4: ret = sprintf(buf, "%u\n", (out_mad->data[40 + offset / 8] >> - (offset % 4)) & 0xf); + (4 - (offset % 8))) & 0xf); break; case 8: ret = sprintf(buf, "%u\n", out_mad->data[40 + offset / 8]); diff --git a/drivers/infiniband/core/ucm.c b/drivers/infiniband/core/ucm.c index f6a05965a4e..9164a09b6cc 100644 --- a/drivers/infiniband/core/ucm.c +++ b/drivers/infiniband/core/ucm.c @@ -32,6 +32,8 @@ * * $Id: ucm.c 2594 2005-06-13 19:46:02Z libor $ */ + +#include <linux/completion.h> #include <linux/init.h> #include <linux/fs.h> #include <linux/module.h> @@ -72,7 +74,7 @@ struct ib_ucm_file { struct ib_ucm_context { int id; - wait_queue_head_t wait; + struct completion comp; atomic_t ref; int events_reported; @@ -138,7 +140,7 @@ static struct ib_ucm_context *ib_ucm_ctx_get(struct ib_ucm_file *file, int id) static void ib_ucm_ctx_put(struct ib_ucm_context *ctx) { if (atomic_dec_and_test(&ctx->ref)) - wake_up(&ctx->wait); + complete(&ctx->comp); } static inline int ib_ucm_new_cm_id(int event) @@ -178,7 +180,7 @@ static struct ib_ucm_context *ib_ucm_ctx_alloc(struct ib_ucm_file *file) return NULL; atomic_set(&ctx->ref, 1); - init_waitqueue_head(&ctx->wait); + init_completion(&ctx->comp); ctx->file = file; INIT_LIST_HEAD(&ctx->events); @@ -586,8 +588,8 @@ static ssize_t ib_ucm_destroy_id(struct ib_ucm_file *file, if (IS_ERR(ctx)) return PTR_ERR(ctx); - atomic_dec(&ctx->ref); - wait_event(ctx->wait, !atomic_read(&ctx->ref)); + ib_ucm_ctx_put(ctx); + wait_for_completion(&ctx->comp); /* No new events will be generated after destroying the cm_id. */ ib_destroy_cm_id(ctx->cm_id); diff --git a/drivers/infiniband/core/uverbs_mem.c b/drivers/infiniband/core/uverbs_mem.c index 36a32c31566..efe147dbeb4 100644 --- a/drivers/infiniband/core/uverbs_mem.c +++ b/drivers/infiniband/core/uverbs_mem.c @@ -211,8 +211,10 @@ void ib_umem_release_on_close(struct ib_device *dev, struct ib_umem *umem) */ work = kmalloc(sizeof *work, GFP_KERNEL); - if (!work) + if (!work) { + mmput(mm); return; + } INIT_WORK(&work->work, ib_umem_account, work); work->mm = mm; diff --git a/drivers/infiniband/hw/ipath/ipath_debug.h b/drivers/infiniband/hw/ipath/ipath_debug.h index 593e28969c6..46762387f5f 100644 --- a/drivers/infiniband/hw/ipath/ipath_debug.h +++ b/drivers/infiniband/hw/ipath/ipath_debug.h @@ -60,11 +60,11 @@ #define __IPATH_KERNEL_SEND 0x2000 /* use kernel mode send */ #define __IPATH_EPKTDBG 0x4000 /* print ethernet packet data */ #define __IPATH_SMADBG 0x8000 /* sma packet debug */ -#define __IPATH_IPATHDBG 0x10000 /* Ethernet (IPATH) general debug on */ -#define __IPATH_IPATHWARN 0x20000 /* Ethernet (IPATH) warnings on */ -#define __IPATH_IPATHERR 0x40000 /* Ethernet (IPATH) errors on */ -#define __IPATH_IPATHPD 0x80000 /* Ethernet (IPATH) packet dump on */ -#define __IPATH_IPATHTABLE 0x100000 /* Ethernet (IPATH) table dump on */ +#define __IPATH_IPATHDBG 0x10000 /* Ethernet (IPATH) gen debug */ +#define __IPATH_IPATHWARN 0x20000 /* Ethernet (IPATH) warnings */ +#define __IPATH_IPATHERR 0x40000 /* Ethernet (IPATH) errors */ +#define __IPATH_IPATHPD 0x80000 /* Ethernet (IPATH) packet dump */ +#define __IPATH_IPATHTABLE 0x100000 /* Ethernet (IPATH) table dump */ #else /* _IPATH_DEBUGGING */ @@ -79,11 +79,12 @@ #define __IPATH_TRSAMPLE 0x0 /* generate trace buffer sample entries */ #define __IPATH_VERBDBG 0x0 /* very verbose debug */ #define __IPATH_PKTDBG 0x0 /* print packet data */ -#define __IPATH_PROCDBG 0x0 /* print process startup (init)/exit messages */ +#define __IPATH_PROCDBG 0x0 /* process startup (init)/exit messages */ /* print mmap/nopage stuff, not using VDBG any more */ #define __IPATH_MMDBG 0x0 #define __IPATH_EPKTDBG 0x0 /* print ethernet packet data */ -#define __IPATH_SMADBG 0x0 /* print process startup (init)/exit messages */#define __IPATH_IPATHDBG 0x0 /* Ethernet (IPATH) table dump on */ +#define __IPATH_SMADBG 0x0 /* process startup (init)/exit messages */ +#define __IPATH_IPATHDBG 0x0 /* Ethernet (IPATH) table dump on */ #define __IPATH_IPATHWARN 0x0 /* Ethernet (IPATH) warnings on */ #define __IPATH_IPATHERR 0x0 /* Ethernet (IPATH) errors on */ #define __IPATH_IPATHPD 0x0 /* Ethernet (IPATH) packet dump on */ diff --git a/drivers/infiniband/hw/ipath/ipath_diag.c b/drivers/infiniband/hw/ipath/ipath_diag.c index cd533cf951c..28ddceb260e 100644 --- a/drivers/infiniband/hw/ipath/ipath_diag.c +++ b/drivers/infiniband/hw/ipath/ipath_diag.c @@ -277,13 +277,14 @@ static int ipath_diag_open(struct inode *in, struct file *fp) bail: spin_unlock_irqrestore(&ipath_devs_lock, flags); - mutex_unlock(&ipath_mutex); /* Only expose a way to reset the device if we make it into diag mode. */ if (ret == 0) ipath_expose_reset(&dd->pcidev->dev); + mutex_unlock(&ipath_mutex); + return ret; } @@ -365,15 +366,3 @@ static ssize_t ipath_diag_write(struct file *fp, const char __user *data, bail: return ret; } - -void ipath_diag_bringup_link(struct ipath_devdata *dd) -{ - if (diag_set_link || (dd->ipath_flags & IPATH_LINKACTIVE)) - return; - - diag_set_link = 1; - ipath_cdbg(VERBOSE, "Trying to set to set link active for " - "diag pkt\n"); - ipath_layer_set_linkstate(dd, IPATH_IB_LINKARM); - ipath_layer_set_linkstate(dd, IPATH_IB_LINKACTIVE); -} diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c index 58a94efb007..dddcdae736a 100644 --- a/drivers/infiniband/hw/ipath/ipath_driver.c +++ b/drivers/infiniband/hw/ipath/ipath_driver.c @@ -116,10 +116,9 @@ static int __devinit ipath_init_one(struct pci_dev *, #define PCI_DEVICE_ID_INFINIPATH_PE800 0x10 static const struct pci_device_id ipath_pci_tbl[] = { - {PCI_DEVICE(PCI_VENDOR_ID_PATHSCALE, - PCI_DEVICE_ID_INFINIPATH_HT)}, - {PCI_DEVICE(PCI_VENDOR_ID_PATHSCALE, - PCI_DEVICE_ID_INFINIPATH_PE800)}, + { PCI_DEVICE(PCI_VENDOR_ID_PATHSCALE, PCI_DEVICE_ID_INFINIPATH_HT) }, + { PCI_DEVICE(PCI_VENDOR_ID_PATHSCALE, PCI_DEVICE_ID_INFINIPATH_PE800) }, + { 0, } }; MODULE_DEVICE_TABLE(pci, ipath_pci_tbl); @@ -418,9 +417,19 @@ static int __devinit ipath_init_one(struct pci_dev *pdev, ret = pci_set_dma_mask(pdev, DMA_64BIT_MASK); if (ret) { - dev_info(&pdev->dev, "pci_set_dma_mask unit %u " - "fails: %d\n", dd->ipath_unit, ret); - goto bail_regions; + /* + * if the 64 bit setup fails, try 32 bit. Some systems + * do not setup 64 bit maps on systems with 2GB or less + * memory installed. + */ + ret = pci_set_dma_mask(pdev, DMA_32BIT_MASK); + if (ret) { + dev_info(&pdev->dev, "pci_set_dma_mask unit %u " + "fails: %d\n", dd->ipath_unit, ret); + goto bail_regions; + } + else + ipath_dbg("No 64bit DMA mask, used 32 bit mask\n"); } pci_set_master(pdev); @@ -1729,7 +1738,7 @@ void ipath_free_pddata(struct ipath_devdata *dd, u32 port, int freehdrq) } } -int __init infinipath_init(void) +static int __init infinipath_init(void) { int ret; @@ -1896,19 +1905,19 @@ static void __exit infinipath_cleanup(void) } else ipath_dbg("irq is 0, not doing free_irq " "for unit %u\n", dd->ipath_unit); - dd->pcidev = NULL; - } - /* - * we check for NULL here, because it's outside the kregbase - * check, and we need to call it after the free_irq. Thus - * it's possible that the function pointers were never - * initialized. - */ - if (dd->ipath_f_cleanup) - /* clean up chip-specific stuff */ - dd->ipath_f_cleanup(dd); + /* + * we check for NULL here, because it's outside + * the kregbase check, and we need to call it + * after the free_irq. Thus it's possible that + * the function pointers were never initialized. + */ + if (dd->ipath_f_cleanup) + /* clean up chip-specific stuff */ + dd->ipath_f_cleanup(dd); + dd->pcidev = NULL; + } spin_lock_irqsave(&ipath_devs_lock, flags); } @@ -1949,7 +1958,7 @@ int ipath_reset_device(int unit) } if (dd->ipath_pd) - for (i = 1; i < dd->ipath_portcnt; i++) { + for (i = 1; i < dd->ipath_cfgports; i++) { if (dd->ipath_pd[i] && dd->ipath_pd[i]->port_cnt) { ipath_dbg("unit %u port %d is in use " "(PID %u cmd %s), can't reset\n", diff --git a/drivers/infiniband/hw/ipath/ipath_eeprom.c b/drivers/infiniband/hw/ipath/ipath_eeprom.c index f11a900e8cd..a2f1ceafcca 100644 --- a/drivers/infiniband/hw/ipath/ipath_eeprom.c +++ b/drivers/infiniband/hw/ipath/ipath_eeprom.c @@ -505,11 +505,10 @@ static u8 flash_csum(struct ipath_flash *ifp, int adjust) * ipath_get_guid - get the GUID from the i2c device * @dd: the infinipath device * - * When we add the multi-chip support, we will probably have to add - * the ability to use the number of guids field, and get the guid from - * the first chip's flash, to use for all of them. + * We have the capability to use the ipath_nguid field, and get + * the guid from the first chip's flash, to use for all of them. */ -void ipath_get_guid(struct ipath_devdata *dd) +void ipath_get_eeprom_info(struct ipath_devdata *dd) { void *buf; struct ipath_flash *ifp; diff --git a/drivers/infiniband/hw/ipath/ipath_file_ops.c b/drivers/infiniband/hw/ipath/ipath_file_ops.c index c347191f02b..ada267e41f6 100644 --- a/drivers/infiniband/hw/ipath/ipath_file_ops.c +++ b/drivers/infiniband/hw/ipath/ipath_file_ops.c @@ -139,7 +139,7 @@ static int ipath_get_base_info(struct ipath_portdata *pd, kinfo->spi_piosize = dd->ipath_ibmaxlen; kinfo->spi_mtu = dd->ipath_ibmaxlen; /* maxlen, not ibmtu */ kinfo->spi_port = pd->port_port; - kinfo->spi_sw_version = IPATH_USER_SWVERSION; + kinfo->spi_sw_version = IPATH_KERN_SWVERSION; kinfo->spi_hw_version = dd->ipath_revision; if (copy_to_user(ubase, kinfo, sizeof(*kinfo))) @@ -1224,6 +1224,10 @@ static unsigned int ipath_poll(struct file *fp, if (tail == head) { set_bit(IPATH_PORT_WAITING_RCV, &pd->port_flag); + if(dd->ipath_rhdrhead_intr_off) /* arm rcv interrupt */ + (void)ipath_write_ureg(dd, ur_rcvhdrhead, + dd->ipath_rhdrhead_intr_off + | head, pd->port_port); poll_wait(fp, &pd->port_wait, pt); if (test_bit(IPATH_PORT_WAITING_RCV, &pd->port_flag)) { diff --git a/drivers/infiniband/hw/ipath/ipath_ht400.c b/drivers/infiniband/hw/ipath/ipath_ht400.c index 4652435998f..fac0a2b74de 100644 --- a/drivers/infiniband/hw/ipath/ipath_ht400.c +++ b/drivers/infiniband/hw/ipath/ipath_ht400.c @@ -607,7 +607,12 @@ static int ipath_ht_boardname(struct ipath_devdata *dd, char *name, case 4: /* Ponderosa is one of the bringup boards */ n = "Ponderosa"; break; - case 5: /* HT-460 original production board */ + case 5: + /* + * HT-460 original production board; two production levels, with + * different serial number ranges. See ipath_ht_early_init() for + * case where we enable IPATH_GPIO_INTR for later serial # range. + */ n = "InfiniPath_HT-460"; break; case 6: @@ -642,7 +647,7 @@ static int ipath_ht_boardname(struct ipath_devdata *dd, char *name, if (n) snprintf(name, namelen, "%s", n); - if (dd->ipath_majrev != 3 || dd->ipath_minrev != 2) { + if (dd->ipath_majrev != 3 || (dd->ipath_minrev < 2 || dd->ipath_minrev > 3)) { /* * This version of the driver only supports the HT-400 * Rev 3.2 @@ -1520,6 +1525,18 @@ static int ipath_ht_early_init(struct ipath_devdata *dd) */ ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl, INFINIPATH_S_ABORT); + + ipath_get_eeprom_info(dd); + if(dd->ipath_boardrev == 5 && dd->ipath_serial[0] == '1' && + dd->ipath_serial[1] == '2' && dd->ipath_serial[2] == '8') { + /* + * Later production HT-460 has same changes as HT-465, so + * can use GPIO interrupts. They have serial #'s starting + * with 128, rather than 112. + */ + dd->ipath_flags |= IPATH_GPIO_INTR; + dd->ipath_flags &= ~IPATH_POLL_RX_INTR; + } return 0; } diff --git a/drivers/infiniband/hw/ipath/ipath_init_chip.c b/drivers/infiniband/hw/ipath/ipath_init_chip.c index 2823ff9c0c6..dc83250d26a 100644 --- a/drivers/infiniband/hw/ipath/ipath_init_chip.c +++ b/drivers/infiniband/hw/ipath/ipath_init_chip.c @@ -53,13 +53,19 @@ MODULE_PARM_DESC(cfgports, "Set max number of ports to use"); /* * Number of buffers reserved for driver (layered drivers and SMA - * send). Reserved at end of buffer list. + * send). Reserved at end of buffer list. Initialized based on + * number of PIO buffers if not set via module interface. + * The problem with this is that it's global, but we'll use different + * numbers for different chip types. So the default value is not + * very useful. I've redefined it for the 1.3 release so that it's + * zero unless set by the user to something else, in which case we + * try to respect it. */ -static ushort ipath_kpiobufs = 32; +static ushort ipath_kpiobufs; static int ipath_set_kpiobufs(const char *val, struct kernel_param *kp); -module_param_call(kpiobufs, ipath_set_kpiobufs, param_get_uint, +module_param_call(kpiobufs, ipath_set_kpiobufs, param_get_ushort, &ipath_kpiobufs, S_IWUSR | S_IRUGO); MODULE_PARM_DESC(kpiobufs, "Set number of PIO buffers for driver"); @@ -531,8 +537,11 @@ static int init_housekeeping(struct ipath_devdata *dd, * Don't clear ipath_flags as 8bit mode was set before * entering this func. However, we do set the linkstate to * unknown, so we can watch for a transition. + * PRESENT is set because we want register reads to work, + * and the kernel infrastructure saw it in config space; + * We clear it if we have failures. */ - dd->ipath_flags |= IPATH_LINKUNK; + dd->ipath_flags |= IPATH_LINKUNK | IPATH_PRESENT; dd->ipath_flags &= ~(IPATH_LINKACTIVE | IPATH_LINKARMED | IPATH_LINKDOWN | IPATH_LINKINIT); @@ -560,6 +569,7 @@ static int init_housekeeping(struct ipath_devdata *dd, || (dd->ipath_uregbase & 0xffffffff) == 0xffffffff) { ipath_dev_err(dd, "Register read failures from chip, " "giving up initialization\n"); + dd->ipath_flags &= ~IPATH_PRESENT; ret = -ENODEV; goto done; } @@ -682,16 +692,14 @@ int ipath_init_chip(struct ipath_devdata *dd, int reinit) */ dd->ipath_pioavregs = ALIGN(val, sizeof(u64) * BITS_PER_BYTE / 2) / (sizeof(u64) * BITS_PER_BYTE / 2); - if (!ipath_kpiobufs) /* have to have at least 1, for SMA */ - kpiobufs = ipath_kpiobufs = 1; - else if ((dd->ipath_piobcnt2k + dd->ipath_piobcnt4k) < - (dd->ipath_cfgports * IPATH_MIN_USER_PORT_BUFCNT)) { - dev_info(&dd->pcidev->dev, "Too few PIO buffers (%u) " - "for %u ports to have %u each!\n", - dd->ipath_piobcnt2k + dd->ipath_piobcnt4k, - dd->ipath_cfgports, IPATH_MIN_USER_PORT_BUFCNT); - kpiobufs = 1; /* reserve just the minimum for SMA/ether */ - } else + if (ipath_kpiobufs == 0) { + /* not set by user, or set explictly to default */ + if ((dd->ipath_piobcnt2k + dd->ipath_piobcnt4k) > 128) + kpiobufs = 32; + else + kpiobufs = 16; + } + else kpiobufs = ipath_kpiobufs; if (kpiobufs > @@ -871,7 +879,6 @@ int ipath_init_chip(struct ipath_devdata *dd, int reinit) done: if (!ret) { - ipath_get_guid(dd); *dd->ipath_statusp |= IPATH_STATUS_CHIP_PRESENT; if (!dd->ipath_f_intrsetup(dd)) { /* now we can enable all interrupts from the chip */ diff --git a/drivers/infiniband/hw/ipath/ipath_intr.c b/drivers/infiniband/hw/ipath/ipath_intr.c index 60f5f410806..3e72a1fe3d7 100644 --- a/drivers/infiniband/hw/ipath/ipath_intr.c +++ b/drivers/infiniband/hw/ipath/ipath_intr.c @@ -172,8 +172,8 @@ static void handle_e_ibstatuschanged(struct ipath_devdata *dd, "was %s\n", dd->ipath_unit, ib_linkstate(lstate), ib_linkstate((unsigned) - dd->ipath_lastibcstat - & IPATH_IBSTATE_MASK)); + dd->ipath_lastibcstat + & IPATH_IBSTATE_MASK)); } else { lstate = dd->ipath_lastibcstat & IPATH_IBSTATE_MASK; @@ -665,14 +665,14 @@ static void handle_layer_pioavail(struct ipath_devdata *dd) ret = __ipath_layer_intr(dd, IPATH_LAYER_INT_SEND_CONTINUE); if (ret > 0) - goto clear; + goto set; ret = __ipath_verbs_piobufavail(dd); if (ret > 0) - goto clear; + goto set; return; -clear: +set: set_bit(IPATH_S_PIOINTBUFAVAIL, &dd->ipath_sendctrl); ipath_write_kreg(dd, dd->ipath_kregs->kr_sendctrl, dd->ipath_sendctrl); @@ -719,11 +719,24 @@ static void handle_rcv(struct ipath_devdata *dd, u32 istat) irqreturn_t ipath_intr(int irq, void *data, struct pt_regs *regs) { struct ipath_devdata *dd = data; - u32 istat = ipath_read_kreg32(dd, dd->ipath_kregs->kr_intstatus); + u32 istat; ipath_err_t estat = 0; static unsigned unexpected = 0; irqreturn_t ret; + if(!(dd->ipath_flags & IPATH_PRESENT)) { + /* this is mostly so we don't try to touch the chip while + * it is being reset */ + /* + * This return value is perhaps odd, but we do not want the + * interrupt core code to remove our interrupt handler + * because we don't appear to be handling an interrupt + * during a chip reset. + */ + return IRQ_HANDLED; + } + + istat = ipath_read_kreg32(dd, dd->ipath_kregs->kr_intstatus); if (unlikely(!istat)) { ipath_stats.sps_nullintr++; ret = IRQ_NONE; /* not our interrupt, or already handled */ diff --git a/drivers/infiniband/hw/ipath/ipath_kernel.h b/drivers/infiniband/hw/ipath/ipath_kernel.h index 159d0aed31a..5d92d57b6f5 100644 --- a/drivers/infiniband/hw/ipath/ipath_kernel.h +++ b/drivers/infiniband/hw/ipath/ipath_kernel.h @@ -528,7 +528,6 @@ extern spinlock_t ipath_devs_lock; extern struct ipath_devdata *ipath_lookup(int unit); extern u16 ipath_layer_rcv_opcode; -extern int ipath_verbs_registered; extern int __ipath_layer_intr(struct ipath_devdata *, u32); extern int ipath_layer_intr(struct ipath_devdata *, u32); extern int __ipath_layer_rcv(struct ipath_devdata *, void *, @@ -651,7 +650,7 @@ u32 __iomem *ipath_getpiobuf(struct ipath_devdata *, u32 *); void ipath_init_pe800_funcs(struct ipath_devdata *); /* init HT-400-specific func */ void ipath_init_ht400_funcs(struct ipath_devdata *); -void ipath_get_guid(struct ipath_devdata *); +void ipath_get_eeprom_info(struct ipath_devdata *); u64 ipath_snap_cntr(struct ipath_devdata *, ipath_creg); /* @@ -732,7 +731,7 @@ u64 ipath_read_kreg64_port(const struct ipath_devdata *, ipath_kreg, static inline u32 ipath_read_ureg32(const struct ipath_devdata *dd, ipath_ureg regno, int port) { - if (!dd->ipath_kregbase) + if (!dd->ipath_kregbase || !(dd->ipath_flags & IPATH_PRESENT)) return 0; return readl(regno + (u64 __iomem *) @@ -763,7 +762,7 @@ static inline void ipath_write_ureg(const struct ipath_devdata *dd, static inline u32 ipath_read_kreg32(const struct ipath_devdata *dd, ipath_kreg regno) { - if (!dd->ipath_kregbase) + if (!dd->ipath_kregbase || !(dd->ipath_flags & IPATH_PRESENT)) return -1; return readl((u32 __iomem *) & dd->ipath_kregbase[regno]); } @@ -771,7 +770,7 @@ static inline u32 ipath_read_kreg32(const struct ipath_devdata *dd, static inline u64 ipath_read_kreg64(const struct ipath_devdata *dd, ipath_kreg regno) { - if (!dd->ipath_kregbase) + if (!dd->ipath_kregbase || !(dd->ipath_flags & IPATH_PRESENT)) return -1; return readq(&dd->ipath_kregbase[regno]); @@ -787,7 +786,7 @@ static inline void ipath_write_kreg(const struct ipath_devdata *dd, static inline u64 ipath_read_creg(const struct ipath_devdata *dd, ipath_sreg regno) { - if (!dd->ipath_kregbase) + if (!dd->ipath_kregbase || !(dd->ipath_flags & IPATH_PRESENT)) return 0; return readq(regno + (u64 __iomem *) @@ -798,7 +797,7 @@ static inline u64 ipath_read_creg(const struct ipath_devdata *dd, static inline u32 ipath_read_creg32(const struct ipath_devdata *dd, ipath_sreg regno) { - if (!dd->ipath_kregbase) + if (!dd->ipath_kregbase || !(dd->ipath_flags & IPATH_PRESENT)) return 0; return readl(regno + (u64 __iomem *) (dd->ipath_cregbase + diff --git a/drivers/infiniband/hw/ipath/ipath_keys.c b/drivers/infiniband/hw/ipath/ipath_keys.c index aa33b0e9f2f..5ae8761f9dd 100644 --- a/drivers/infiniband/hw/ipath/ipath_keys.c +++ b/drivers/infiniband/hw/ipath/ipath_keys.c @@ -136,9 +136,7 @@ int ipath_lkey_ok(struct ipath_lkey_table *rkt, struct ipath_sge *isge, ret = 1; goto bail; } - spin_lock(&rkt->lock); mr = rkt->table[(sge->lkey >> (32 - ib_ipath_lkey_table_size))]; - spin_unlock(&rkt->lock); if (unlikely(mr == NULL || mr->lkey != sge->lkey)) { ret = 0; goto bail; @@ -184,8 +182,6 @@ bail: * @acc: access flags * * Return 1 if successful, otherwise 0. - * - * The QP r_rq.lock should be held. */ int ipath_rkey_ok(struct ipath_ibdev *dev, struct ipath_sge_state *ss, u32 len, u64 vaddr, u32 rkey, int acc) @@ -196,9 +192,7 @@ int ipath_rkey_ok(struct ipath_ibdev *dev, struct ipath_sge_state *ss, size_t off; int ret; - spin_lock(&rkt->lock); mr = rkt->table[(rkey >> (32 - ib_ipath_lkey_table_size))]; - spin_unlock(&rkt->lock); if (unlikely(mr == NULL || mr->lkey != rkey)) { ret = 0; goto bail; diff --git a/drivers/infiniband/hw/ipath/ipath_layer.c b/drivers/infiniband/hw/ipath/ipath_layer.c index 2cabf634057..9ec4ac77b87 100644 --- a/drivers/infiniband/hw/ipath/ipath_layer.c +++ b/drivers/infiniband/hw/ipath/ipath_layer.c @@ -46,13 +46,15 @@ /* Acquire before ipath_devs_lock. */ static DEFINE_MUTEX(ipath_layer_mutex); +static int ipath_verbs_registered; + u16 ipath_layer_rcv_opcode; + static int (*layer_intr)(void *, u32); static int (*layer_rcv)(void *, void *, struct sk_buff *); static int (*layer_rcv_lid)(void *, void *); static int (*verbs_piobufavail)(void *); static void (*verbs_rcv)(void *, void *, void *, u32); -int ipath_verbs_registered; static void *(*layer_add_one)(int, struct ipath_devdata *); static void (*layer_remove_one)(void *); @@ -586,6 +588,8 @@ void ipath_verbs_unregister(void) verbs_rcv = NULL; verbs_timer_cb = NULL; + ipath_verbs_registered = 0; + mutex_unlock(&ipath_layer_mutex); } @@ -868,12 +872,13 @@ static void copy_io(u32 __iomem *piobuf, struct ipath_sge_state *ss, update_sge(ss, len); length -= len; } + /* Update address before sending packet. */ + update_sge(ss, length); /* must flush early everything before trigger word */ ipath_flush_wc(); __raw_writel(last, piobuf); /* be sure trigger word is written */ ipath_flush_wc(); - update_sge(ss, length); } /** @@ -939,17 +944,18 @@ int ipath_verbs_send(struct ipath_devdata *dd, u32 hdrwords, if (likely(ss->num_sge == 1 && len <= ss->sge.length && !((unsigned long)ss->sge.vaddr & (sizeof(u32) - 1)))) { u32 w; + u32 *addr = (u32 *) ss->sge.vaddr; + /* Update address before sending packet. */ + update_sge(ss, len); /* Need to round up for the last dword in the packet. */ w = (len + 3) >> 2; - __iowrite32_copy(piobuf, ss->sge.vaddr, w - 1); + __iowrite32_copy(piobuf, addr, w - 1); /* must flush early everything before trigger word */ ipath_flush_wc(); - __raw_writel(((u32 *) ss->sge.vaddr)[w - 1], - piobuf + w - 1); + __raw_writel(addr[w - 1], piobuf + w - 1); /* be sure trigger word is written */ ipath_flush_wc(); - update_sge(ss, len); ret = 0; goto bail; } diff --git a/drivers/infiniband/hw/ipath/ipath_pe800.c b/drivers/infiniband/hw/ipath/ipath_pe800.c index e693a7a8266..02e8c75b24f 100644 --- a/drivers/infiniband/hw/ipath/ipath_pe800.c +++ b/drivers/infiniband/hw/ipath/ipath_pe800.c @@ -305,8 +305,8 @@ static const struct ipath_cregs ipath_pe_cregs = { * we'll print them and continue. We reuse the same message buffer as * ipath_handle_errors() to avoid excessive stack usage. */ -void ipath_pe_handle_hwerrors(struct ipath_devdata *dd, char *msg, - size_t msgl) +static void ipath_pe_handle_hwerrors(struct ipath_devdata *dd, char *msg, + size_t msgl) { ipath_err_t hwerrs; u32 bits, ctrl; @@ -552,7 +552,7 @@ static int ipath_pe_boardname(struct ipath_devdata *dd, char *name, * freeze mode), and enable hardware errors as errors (along with * everything else) in errormask */ -void ipath_pe_init_hwerrors(struct ipath_devdata *dd) +static void ipath_pe_init_hwerrors(struct ipath_devdata *dd) { ipath_err_t val; u64 extsval; @@ -577,7 +577,7 @@ void ipath_pe_init_hwerrors(struct ipath_devdata *dd) * ipath_pe_bringup_serdes - bring up the serdes * @dd: the infinipath device */ -int ipath_pe_bringup_serdes(struct ipath_devdata *dd) +static int ipath_pe_bringup_serdes(struct ipath_devdata *dd) { u64 val, tmp, config1; int ret = 0, change = 0; @@ -694,7 +694,7 @@ int ipath_pe_bringup_serdes(struct ipath_devdata *dd) * @dd: the infinipath device * Called when driver is being unloaded */ -void ipath_pe_quiet_serdes(struct ipath_devdata *dd) +static void ipath_pe_quiet_serdes(struct ipath_devdata *dd) { u64 val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_serdesconfig0); @@ -972,6 +972,8 @@ static int ipath_setup_pe_reset(struct ipath_devdata *dd) /* Use ERROR so it shows up in logs, etc. */ ipath_dev_err(dd, "Resetting PE-800 unit %u\n", dd->ipath_unit); + /* keep chip from being accessed in a few places */ + dd->ipath_flags &= ~(IPATH_INITTED|IPATH_PRESENT); val = dd->ipath_control | INFINIPATH_C_RESET; ipath_write_kreg(dd, dd->ipath_kregs->kr_control, val); mb(); @@ -997,6 +999,8 @@ static int ipath_setup_pe_reset(struct ipath_devdata *dd) if ((r = pci_enable_device(dd->pcidev))) ipath_dev_err(dd, "pci_enable_device failed after " "reset: %d\n", r); + /* whether it worked or not, mark as present, again */ + dd->ipath_flags |= IPATH_PRESENT; val = ipath_read_kreg64(dd, dd->ipath_kregs->kr_revision); if (val == dd->ipath_revision) { ipath_cdbg(VERBOSE, "Got matching revision " @@ -1176,6 +1180,8 @@ static int ipath_pe_early_init(struct ipath_devdata *dd) */ dd->ipath_rhdrhead_intr_off = 1ULL<<32; + ipath_get_eeprom_info(dd); + return 0; } diff --git a/drivers/infiniband/hw/ipath/ipath_qp.c b/drivers/infiniband/hw/ipath/ipath_qp.c index 6058d70d757..9f8855d970c 100644 --- a/drivers/infiniband/hw/ipath/ipath_qp.c +++ b/drivers/infiniband/hw/ipath/ipath_qp.c @@ -188,8 +188,8 @@ static void free_qpn(struct ipath_qp_table *qpt, u32 qpn) * Allocate the next available QPN and put the QP into the hash table. * The hash table holds a reference to the QP. */ -int ipath_alloc_qpn(struct ipath_qp_table *qpt, struct ipath_qp *qp, - enum ib_qp_type type) +static int ipath_alloc_qpn(struct ipath_qp_table *qpt, struct ipath_qp *qp, + enum ib_qp_type type) { unsigned long flags; u32 qpn; @@ -232,7 +232,7 @@ bail: * Remove the QP from the table so it can't be found asynchronously by * the receive interrupt routine. */ -void ipath_free_qp(struct ipath_qp_table *qpt, struct ipath_qp *qp) +static void ipath_free_qp(struct ipath_qp_table *qpt, struct ipath_qp *qp) { struct ipath_qp *q, **qpp; unsigned long flags; @@ -358,6 +358,65 @@ static void ipath_reset_qp(struct ipath_qp *qp) } /** + * ipath_error_qp - put a QP into an error state + * @qp: the QP to put into an error state + * + * Flushes both send and receive work queues. + * QP r_rq.lock and s_lock should be held. + */ + +static void ipath_error_qp(struct ipath_qp *qp) +{ + struct ipath_ibdev *dev = to_idev(qp->ibqp.device); + struct ib_wc wc; + + _VERBS_INFO("QP%d/%d in error state\n", + qp->ibqp.qp_num, qp->remote_qpn); + + spin_lock(&dev->pending_lock); + /* XXX What if its already removed by the timeout code? */ + if (!list_empty(&qp->timerwait)) + list_del_init(&qp->timerwait); + if (!list_empty(&qp->piowait)) + list_del_init(&qp->piowait); + spin_unlock(&dev->pending_lock); + + wc.status = IB_WC_WR_FLUSH_ERR; + wc.vendor_err = 0; + wc.byte_len = 0; + wc.imm_data = 0; + wc.qp_num = qp->ibqp.qp_num; + wc.src_qp = 0; + wc.wc_flags = 0; + wc.pkey_index = 0; + wc.slid = 0; + wc.sl = 0; + wc.dlid_path_bits = 0; + wc.port_num = 0; + + while (qp->s_last != qp->s_head) { + struct ipath_swqe *wqe = get_swqe_ptr(qp, qp->s_last); + + wc.wr_id = wqe->wr.wr_id; + wc.opcode = ib_ipath_wc_opcode[wqe->wr.opcode]; + if (++qp->s_last >= qp->s_size) + qp->s_last = 0; + ipath_cq_enter(to_icq(qp->ibqp.send_cq), &wc, 1); + } + qp->s_cur = qp->s_tail = qp->s_head; + qp->s_hdrwords = 0; + qp->s_ack_state = IB_OPCODE_RC_ACKNOWLEDGE; + + wc.opcode = IB_WC_RECV; + while (qp->r_rq.tail != qp->r_rq.head) { + wc.wr_id = get_rwqe_ptr(&qp->r_rq, qp->r_rq.tail)->wr_id; + if (++qp->r_rq.tail >= qp->r_rq.size) + qp->r_rq.tail = 0; + ipath_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1); + } +} + +/** * ipath_modify_qp - modify the attributes of a queue pair * @ibqp: the queue pair who's attributes we're modifying * @attr: the new attributes @@ -368,6 +427,7 @@ static void ipath_reset_qp(struct ipath_qp *qp) int ipath_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask) { + struct ipath_ibdev *dev = to_idev(ibqp->device); struct ipath_qp *qp = to_iqp(ibqp); enum ib_qp_state cur_state, new_state; unsigned long flags; @@ -384,6 +444,19 @@ int ipath_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, attr_mask)) goto inval; + if (attr_mask & IB_QP_AV) + if (attr->ah_attr.dlid == 0 || + attr->ah_attr.dlid >= IPS_MULTICAST_LID_BASE) + goto inval; + + if (attr_mask & IB_QP_PKEY_INDEX) + if (attr->pkey_index >= ipath_layer_get_npkeys(dev->dd)) + goto inval; + + if (attr_mask & IB_QP_MIN_RNR_TIMER) + if (attr->min_rnr_timer > 31) + goto inval; + switch (new_state) { case IB_QPS_RESET: ipath_reset_qp(qp); @@ -398,13 +471,8 @@ int ipath_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, } - if (attr_mask & IB_QP_PKEY_INDEX) { - struct ipath_ibdev *dev = to_idev(ibqp->device); - - if (attr->pkey_index >= ipath_layer_get_npkeys(dev->dd)) - goto inval; + if (attr_mask & IB_QP_PKEY_INDEX) qp->s_pkey_index = attr->pkey_index; - } if (attr_mask & IB_QP_DEST_QPN) qp->remote_qpn = attr->dest_qp_num; @@ -420,12 +488,8 @@ int ipath_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, if (attr_mask & IB_QP_ACCESS_FLAGS) qp->qp_access_flags = attr->qp_access_flags; - if (attr_mask & IB_QP_AV) { - if (attr->ah_attr.dlid == 0 || - attr->ah_attr.dlid >= IPS_MULTICAST_LID_BASE) - goto inval; + if (attr_mask & IB_QP_AV) qp->remote_ah_attr = attr->ah_attr; - } if (attr_mask & IB_QP_PATH_MTU) qp->path_mtu = attr->path_mtu; @@ -440,11 +504,8 @@ int ipath_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, qp->s_rnr_retry_cnt = qp->s_rnr_retry; } - if (attr_mask & IB_QP_MIN_RNR_TIMER) { - if (attr->min_rnr_timer > 31) - goto inval; + if (attr_mask & IB_QP_MIN_RNR_TIMER) qp->s_min_rnr_timer = attr->min_rnr_timer; - } if (attr_mask & IB_QP_QKEY) qp->qkey = attr->qkey; @@ -651,10 +712,8 @@ struct ib_qp *ipath_create_qp(struct ib_pd *ibpd, init_attr->qp_type == IB_QPT_RC ? ipath_do_rc_send : ipath_do_uc_send, (unsigned long)qp); - qp->piowait.next = LIST_POISON1; - qp->piowait.prev = LIST_POISON2; - qp->timerwait.next = LIST_POISON1; - qp->timerwait.prev = LIST_POISON2; + INIT_LIST_HEAD(&qp->piowait); + INIT_LIST_HEAD(&qp->timerwait); qp->state = IB_QPS_RESET; qp->s_wq = swq; qp->s_size = init_attr->cap.max_send_wr + 1; @@ -675,7 +734,7 @@ struct ib_qp *ipath_create_qp(struct ib_pd *ibpd, ipath_reset_qp(qp); /* Tell the core driver that the kernel SMA is present. */ - if (qp->ibqp.qp_type == IB_QPT_SMI) + if (init_attr->qp_type == IB_QPT_SMI) ipath_layer_set_verbs_flags(dev->dd, IPATH_VERBS_KERNEL_SMA); break; @@ -724,10 +783,10 @@ int ipath_destroy_qp(struct ib_qp *ibqp) /* Make sure the QP isn't on the timeout list. */ spin_lock_irqsave(&dev->pending_lock, flags); - if (qp->timerwait.next != LIST_POISON1) - list_del(&qp->timerwait); - if (qp->piowait.next != LIST_POISON1) - list_del(&qp->piowait); + if (!list_empty(&qp->timerwait)) + list_del_init(&qp->timerwait); + if (!list_empty(&qp->piowait)) + list_del_init(&qp->piowait); spin_unlock_irqrestore(&dev->pending_lock, flags); /* @@ -796,10 +855,10 @@ void ipath_sqerror_qp(struct ipath_qp *qp, struct ib_wc *wc) spin_lock(&dev->pending_lock); /* XXX What if its already removed by the timeout code? */ - if (qp->timerwait.next != LIST_POISON1) - list_del(&qp->timerwait); - if (qp->piowait.next != LIST_POISON1) - list_del(&qp->piowait); + if (!list_empty(&qp->timerwait)) + list_del_init(&qp->timerwait); + if (!list_empty(&qp->piowait)) + list_del_init(&qp->piowait); spin_unlock(&dev->pending_lock); ipath_cq_enter(to_icq(qp->ibqp.send_cq), wc, 1); @@ -821,65 +880,6 @@ void ipath_sqerror_qp(struct ipath_qp *qp, struct ib_wc *wc) } /** - * ipath_error_qp - put a QP into an error state - * @qp: the QP to put into an error state - * - * Flushes both send and receive work queues. - * QP r_rq.lock and s_lock should be held. - */ - -void ipath_error_qp(struct ipath_qp *qp) -{ - struct ipath_ibdev *dev = to_idev(qp->ibqp.device); - struct ib_wc wc; - - _VERBS_INFO("QP%d/%d in error state\n", - qp->ibqp.qp_num, qp->remote_qpn); - - spin_lock(&dev->pending_lock); - /* XXX What if its already removed by the timeout code? */ - if (qp->timerwait.next != LIST_POISON1) - list_del(&qp->timerwait); - if (qp->piowait.next != LIST_POISON1) - list_del(&qp->piowait); - spin_unlock(&dev->pending_lock); - - wc.status = IB_WC_WR_FLUSH_ERR; - wc.vendor_err = 0; - wc.byte_len = 0; - wc.imm_data = 0; - wc.qp_num = qp->ibqp.qp_num; - wc.src_qp = 0; - wc.wc_flags = 0; - wc.pkey_index = 0; - wc.slid = 0; - wc.sl = 0; - wc.dlid_path_bits = 0; - wc.port_num = 0; - - while (qp->s_last != qp->s_head) { - struct ipath_swqe *wqe = get_swqe_ptr(qp, qp->s_last); - - wc.wr_id = wqe->wr.wr_id; - wc.opcode = ib_ipath_wc_opcode[wqe->wr.opcode]; - if (++qp->s_last >= qp->s_size) - qp->s_last = 0; - ipath_cq_enter(to_icq(qp->ibqp.send_cq), &wc, 1); - } - qp->s_cur = qp->s_tail = qp->s_head; - qp->s_hdrwords = 0; - qp->s_ack_state = IB_OPCODE_RC_ACKNOWLEDGE; - - wc.opcode = IB_WC_RECV; - while (qp->r_rq.tail != qp->r_rq.head) { - wc.wr_id = get_rwqe_ptr(&qp->r_rq, qp->r_rq.tail)->wr_id; - if (++qp->r_rq.tail >= qp->r_rq.size) - qp->r_rq.tail = 0; - ipath_cq_enter(to_icq(qp->ibqp.recv_cq), &wc, 1); - } -} - -/** * ipath_get_credit - flush the send work queue of a QP * @qp: the qp who's send work queue to flush * @aeth: the Acknowledge Extended Transport Header diff --git a/drivers/infiniband/hw/ipath/ipath_rc.c b/drivers/infiniband/hw/ipath/ipath_rc.c index a4055ca0061..493b1821a93 100644 --- a/drivers/infiniband/hw/ipath/ipath_rc.c +++ b/drivers/infiniband/hw/ipath/ipath_rc.c @@ -57,7 +57,7 @@ static void ipath_init_restart(struct ipath_qp *qp, struct ipath_swqe *wqe) qp->s_len = wqe->length - len; dev = to_idev(qp->ibqp.device); spin_lock(&dev->pending_lock); - if (qp->timerwait.next == LIST_POISON1) + if (list_empty(&qp->timerwait)) list_add_tail(&qp->timerwait, &dev->pending[dev->pending_index]); spin_unlock(&dev->pending_lock); @@ -356,7 +356,7 @@ static inline int ipath_make_rc_req(struct ipath_qp *qp, if ((int)(qp->s_psn - qp->s_next_psn) > 0) qp->s_next_psn = qp->s_psn; spin_lock(&dev->pending_lock); - if (qp->timerwait.next == LIST_POISON1) + if (list_empty(&qp->timerwait)) list_add_tail(&qp->timerwait, &dev->pending[dev->pending_index]); spin_unlock(&dev->pending_lock); @@ -726,8 +726,8 @@ void ipath_restart_rc(struct ipath_qp *qp, u32 psn, struct ib_wc *wc) */ dev = to_idev(qp->ibqp.device); spin_lock(&dev->pending_lock); - if (qp->timerwait.next != LIST_POISON1) - list_del(&qp->timerwait); + if (!list_empty(&qp->timerwait)) + list_del_init(&qp->timerwait); spin_unlock(&dev->pending_lock); if (wqe->wr.opcode == IB_WR_RDMA_READ) @@ -886,8 +886,8 @@ static int do_rc_ack(struct ipath_qp *qp, u32 aeth, u32 psn, int opcode) * just won't find anything to restart if we ACK everything. */ spin_lock(&dev->pending_lock); - if (qp->timerwait.next != LIST_POISON1) - list_del(&qp->timerwait); + if (!list_empty(&qp->timerwait)) + list_del_init(&qp->timerwait); spin_unlock(&dev->pending_lock); /* @@ -1194,8 +1194,7 @@ static inline void ipath_rc_rcv_resp(struct ipath_ibdev *dev, IB_WR_RDMA_READ)) goto ack_done; spin_lock(&dev->pending_lock); - if (qp->s_rnr_timeout == 0 && - qp->timerwait.next != LIST_POISON1) + if (qp->s_rnr_timeout == 0 && !list_empty(&qp->timerwait)) list_move_tail(&qp->timerwait, &dev->pending[dev->pending_index]); spin_unlock(&dev->pending_lock); diff --git a/drivers/infiniband/hw/ipath/ipath_registers.h b/drivers/infiniband/hw/ipath/ipath_registers.h index 1e59750c5f6..402126eb79c 100644 --- a/drivers/infiniband/hw/ipath/ipath_registers.h +++ b/drivers/infiniband/hw/ipath/ipath_registers.h @@ -34,8 +34,9 @@ #define _IPATH_REGISTERS_H /* - * This file should only be included by kernel source, and by the diags. - * It defines the registers, and their contents, for the InfiniPath HT-400 chip + * This file should only be included by kernel source, and by the diags. It + * defines the registers, and their contents, for the InfiniPath HT-400 + * chip. */ /* @@ -156,8 +157,10 @@ #define INFINIPATH_IBCC_FLOWCTRLWATERMARK_SHIFT 8 #define INFINIPATH_IBCC_LINKINITCMD_MASK 0x3ULL #define INFINIPATH_IBCC_LINKINITCMD_DISABLE 1 -#define INFINIPATH_IBCC_LINKINITCMD_POLL 2 /* cycle through TS1/TS2 till OK */ -#define INFINIPATH_IBCC_LINKINITCMD_SLEEP 3 /* wait for TS1, then go on */ +/* cycle through TS1/TS2 till OK */ +#define INFINIPATH_IBCC_LINKINITCMD_POLL 2 +/* wait for TS1, then go on */ +#define INFINIPATH_IBCC_LINKINITCMD_SLEEP 3 #define INFINIPATH_IBCC_LINKINITCMD_SHIFT 16 #define INFINIPATH_IBCC_LINKCMD_MASK 0x3ULL #define INFINIPATH_IBCC_LINKCMD_INIT 1 /* move to 0x11 */ @@ -182,7 +185,8 @@ #define INFINIPATH_IBCS_LINKSTATE_SHIFT 4 #define INFINIPATH_IBCS_TXREADY 0x40000000 #define INFINIPATH_IBCS_TXCREDITOK 0x80000000 -/* link training states (shift by INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) */ +/* link training states (shift by + INFINIPATH_IBCS_LINKTRAININGSTATE_SHIFT) */ #define INFINIPATH_IBCS_LT_STATE_DISABLED 0x00 #define INFINIPATH_IBCS_LT_STATE_LINKUP 0x01 #define INFINIPATH_IBCS_LT_STATE_POLLACTIVE 0x02 @@ -267,10 +271,12 @@ /* kr_serdesconfig0 bits */ #define INFINIPATH_SERDC0_RESET_MASK 0xfULL /* overal reset bits */ #define INFINIPATH_SERDC0_RESET_PLL 0x10000000ULL /* pll reset */ -#define INFINIPATH_SERDC0_TXIDLE 0xF000ULL /* tx idle enables (per lane) */ -#define INFINIPATH_SERDC0_RXDETECT_EN 0xF0000ULL /* rx detect enables (per lane) */ -#define INFINIPATH_SERDC0_L1PWR_DN 0xF0ULL /* L1 Power down; use with RXDETECT, - Otherwise not used on IB side */ +/* tx idle enables (per lane) */ +#define INFINIPATH_SERDC0_TXIDLE 0xF000ULL +/* rx detect enables (per lane) */ +#define INFINIPATH_SERDC0_RXDETECT_EN 0xF0000ULL +/* L1 Power down; use with RXDETECT, Otherwise not used on IB side */ +#define INFINIPATH_SERDC0_L1PWR_DN 0xF0ULL /* kr_xgxsconfig bits */ #define INFINIPATH_XGXS_RESET 0x7ULL @@ -390,12 +396,13 @@ struct ipath_kregs { ipath_kreg kr_txintmemsize; ipath_kreg kr_xgxsconfig; ipath_kreg kr_ibpllcfg; - /* use these two (and the following N ports) only with ipath_k*_kreg64_port(); - * not *kreg64() */ + /* use these two (and the following N ports) only with + * ipath_k*_kreg64_port(); not *kreg64() */ ipath_kreg kr_rcvhdraddr; ipath_kreg kr_rcvhdrtailaddr; - /* remaining registers are not present on all types of infinipath chips */ + /* remaining registers are not present on all types of infinipath + chips */ ipath_kreg kr_rcvpktledcnt; ipath_kreg kr_pcierbuftestreg0; ipath_kreg kr_pcierbuftestreg1; diff --git a/drivers/infiniband/hw/ipath/ipath_ruc.c b/drivers/infiniband/hw/ipath/ipath_ruc.c index f232e77b78e..d38f4f3cfd1 100644 --- a/drivers/infiniband/hw/ipath/ipath_ruc.c +++ b/drivers/infiniband/hw/ipath/ipath_ruc.c @@ -435,7 +435,7 @@ void ipath_no_bufs_available(struct ipath_qp *qp, struct ipath_ibdev *dev) unsigned long flags; spin_lock_irqsave(&dev->pending_lock, flags); - if (qp->piowait.next == LIST_POISON1) + if (list_empty(&qp->piowait)) list_add_tail(&qp->piowait, &dev->piowait); spin_unlock_irqrestore(&dev->pending_lock, flags); /* @@ -531,19 +531,12 @@ int ipath_post_rc_send(struct ipath_qp *qp, struct ib_send_wr *wr) } wqe->wr.num_sge = j; qp->s_head = next; - /* - * Wake up the send tasklet if the QP is not waiting - * for an RNR timeout. - */ - next = qp->s_rnr_timeout; spin_unlock_irqrestore(&qp->s_lock, flags); - if (next == 0) { - if (qp->ibqp.qp_type == IB_QPT_UC) - ipath_do_uc_send((unsigned long) qp); - else - ipath_do_rc_send((unsigned long) qp); - } + if (qp->ibqp.qp_type == IB_QPT_UC) + ipath_do_uc_send((unsigned long) qp); + else + ipath_do_rc_send((unsigned long) qp); ret = 0; diff --git a/drivers/infiniband/hw/ipath/ipath_sysfs.c b/drivers/infiniband/hw/ipath/ipath_sysfs.c index 32acd8048b4..f323791cc49 100644 --- a/drivers/infiniband/hw/ipath/ipath_sysfs.c +++ b/drivers/infiniband/hw/ipath/ipath_sysfs.c @@ -711,10 +711,22 @@ static struct attribute_group dev_attr_group = { * enters diag mode. A device reset is quite likely to crash the * machine entirely, so we don't want to normally make it * available. + * + * Called with ipath_mutex held. */ int ipath_expose_reset(struct device *dev) { - return device_create_file(dev, &dev_attr_reset); + static int exposed; + int ret; + + if (!exposed) { + ret = device_create_file(dev, &dev_attr_reset); + exposed = 1; + } + else + ret = 0; + + return ret; } int ipath_driver_create_group(struct device_driver *drv) diff --git a/drivers/infiniband/hw/ipath/ipath_ud.c b/drivers/infiniband/hw/ipath/ipath_ud.c index 5ff3de6128b..e606daf8321 100644 --- a/drivers/infiniband/hw/ipath/ipath_ud.c +++ b/drivers/infiniband/hw/ipath/ipath_ud.c @@ -46,8 +46,10 @@ * This is called from ipath_post_ud_send() to forward a WQE addressed * to the same HCA. */ -void ipath_ud_loopback(struct ipath_qp *sqp, struct ipath_sge_state *ss, - u32 length, struct ib_send_wr *wr, struct ib_wc *wc) +static void ipath_ud_loopback(struct ipath_qp *sqp, + struct ipath_sge_state *ss, + u32 length, struct ib_send_wr *wr, + struct ib_wc *wc) { struct ipath_ibdev *dev = to_idev(sqp->ibqp.device); struct ipath_qp *qp; diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.c b/drivers/infiniband/hw/ipath/ipath_verbs.c index 9f27fd35cdb..28fdbdaa789 100644 --- a/drivers/infiniband/hw/ipath/ipath_verbs.c +++ b/drivers/infiniband/hw/ipath/ipath_verbs.c @@ -41,7 +41,7 @@ /* Not static, because we don't want the compiler removing it */ const char ipath_verbs_version[] = "ipath_verbs " IPATH_IDSTR; -unsigned int ib_ipath_qp_table_size = 251; +static unsigned int ib_ipath_qp_table_size = 251; module_param_named(qp_table_size, ib_ipath_qp_table_size, uint, S_IRUGO); MODULE_PARM_DESC(qp_table_size, "QP table size"); @@ -87,7 +87,7 @@ const enum ib_wc_opcode ib_ipath_wc_opcode[] = { /* * System image GUID. */ -__be64 sys_image_guid; +static __be64 sys_image_guid; /** * ipath_copy_sge - copy data to SGE memory @@ -449,7 +449,6 @@ static void ipath_ib_timer(void *arg) { struct ipath_ibdev *dev = (struct ipath_ibdev *) arg; struct ipath_qp *resend = NULL; - struct ipath_qp *rnr = NULL; struct list_head *last; struct ipath_qp *qp; unsigned long flags; @@ -465,32 +464,18 @@ static void ipath_ib_timer(void *arg) last = &dev->pending[dev->pending_index]; while (!list_empty(last)) { qp = list_entry(last->next, struct ipath_qp, timerwait); - if (last->next == LIST_POISON1 || - last->next != &qp->timerwait || - qp->timerwait.prev != last) { - INIT_LIST_HEAD(last); - } else { - list_del(&qp->timerwait); - qp->timerwait.prev = (struct list_head *) resend; - resend = qp; - atomic_inc(&qp->refcount); - } + list_del_init(&qp->timerwait); + qp->timer_next = resend; + resend = qp; + atomic_inc(&qp->refcount); } last = &dev->rnrwait; if (!list_empty(last)) { qp = list_entry(last->next, struct ipath_qp, timerwait); if (--qp->s_rnr_timeout == 0) { do { - if (last->next == LIST_POISON1 || - last->next != &qp->timerwait || - qp->timerwait.prev != last) { - INIT_LIST_HEAD(last); - break; - } - list_del(&qp->timerwait); - qp->timerwait.prev = - (struct list_head *) rnr; - rnr = qp; + list_del_init(&qp->timerwait); + tasklet_hi_schedule(&qp->s_task); if (list_empty(last)) break; qp = list_entry(last->next, struct ipath_qp, @@ -530,8 +515,7 @@ static void ipath_ib_timer(void *arg) spin_unlock_irqrestore(&dev->pending_lock, flags); /* XXX What if timer fires again while this is running? */ - for (qp = resend; qp != NULL; - qp = (struct ipath_qp *) qp->timerwait.prev) { + for (qp = resend; qp != NULL; qp = qp->timer_next) { struct ib_wc wc; spin_lock_irqsave(&qp->s_lock, flags); @@ -545,9 +529,6 @@ static void ipath_ib_timer(void *arg) if (atomic_dec_and_test(&qp->refcount)) wake_up(&qp->wait); } - for (qp = rnr; qp != NULL; - qp = (struct ipath_qp *) qp->timerwait.prev) - tasklet_hi_schedule(&qp->s_task); } /** @@ -556,9 +537,9 @@ static void ipath_ib_timer(void *arg) * * This is called from ipath_intr() at interrupt level when a PIO buffer is * available after ipath_verbs_send() returned an error that no buffers were - * available. Return 0 if we consumed all the PIO buffers and we still have + * available. Return 1 if we consumed all the PIO buffers and we still have * QPs waiting for buffers (for now, just do a tasklet_hi_schedule and - * return one). + * return zero). */ static int ipath_ib_piobufavail(void *arg) { @@ -573,13 +554,13 @@ static int ipath_ib_piobufavail(void *arg) while (!list_empty(&dev->piowait)) { qp = list_entry(dev->piowait.next, struct ipath_qp, piowait); - list_del(&qp->piowait); + list_del_init(&qp->piowait); tasklet_hi_schedule(&qp->s_task); } spin_unlock_irqrestore(&dev->pending_lock, flags); bail: - return 1; + return 0; } static int ipath_query_device(struct ib_device *ibdev, @@ -970,6 +951,7 @@ static void *ipath_register_ib_device(int unit, struct ipath_devdata *dd) idev->dd = dd; strlcpy(dev->name, "ipath%d", IB_DEVICE_NAME_MAX); + dev->owner = THIS_MODULE; dev->node_guid = ipath_layer_get_guid(dd); dev->uverbs_abi_ver = IPATH_UVERBS_ABI_VERSION; dev->uverbs_cmd_mask = @@ -1110,7 +1092,7 @@ static void ipath_unregister_ib_device(void *arg) ib_dealloc_device(ibdev); } -int __init ipath_verbs_init(void) +static int __init ipath_verbs_init(void) { return ipath_verbs_register(ipath_register_ib_device, ipath_unregister_ib_device, @@ -1118,33 +1100,33 @@ int __init ipath_verbs_init(void) ipath_ib_timer); } -void __exit ipath_verbs_cleanup(void) +static void __exit ipath_verbs_cleanup(void) { ipath_verbs_unregister(); } static ssize_t show_rev(struct class_device *cdev, char *buf) { - struct ipath_ibdev *dev = - container_of(cdev, struct ipath_ibdev, ibdev.class_dev); - int vendor, boardrev, majrev, minrev; + struct ipath_ibdev *dev = + container_of(cdev, struct ipath_ibdev, ibdev.class_dev); + int vendor, boardrev, majrev, minrev; - ipath_layer_query_device(dev->dd, &vendor, &boardrev, - &majrev, &minrev); - return sprintf(buf, "%d.%d\n", majrev, minrev); + ipath_layer_query_device(dev->dd, &vendor, &boardrev, + &majrev, &minrev); + return sprintf(buf, "%d.%d\n", majrev, minrev); } static ssize_t show_hca(struct class_device *cdev, char *buf) { - struct ipath_ibdev *dev = - container_of(cdev, struct ipath_ibdev, ibdev.class_dev); - int ret; + struct ipath_ibdev *dev = + container_of(cdev, struct ipath_ibdev, ibdev.class_dev); + int ret; - ret = ipath_layer_get_boardname(dev->dd, buf, 128); - if (ret < 0) - goto bail; - strcat(buf, "\n"); - ret = strlen(buf); + ret = ipath_layer_get_boardname(dev->dd, buf, 128); + if (ret < 0) + goto bail; + strcat(buf, "\n"); + ret = strlen(buf); bail: return ret; @@ -1152,40 +1134,40 @@ bail: static ssize_t show_stats(struct class_device *cdev, char *buf) { - struct ipath_ibdev *dev = - container_of(cdev, struct ipath_ibdev, ibdev.class_dev); - int i; - int len; - - len = sprintf(buf, - "RC resends %d\n" - "RC QACKs %d\n" - "RC ACKs %d\n" - "RC SEQ NAKs %d\n" - "RC RDMA seq %d\n" - "RC RNR NAKs %d\n" - "RC OTH NAKs %d\n" - "RC timeouts %d\n" - "RC RDMA dup %d\n" - "piobuf wait %d\n" - "no piobuf %d\n" - "PKT drops %d\n" - "WQE errs %d\n", - dev->n_rc_resends, dev->n_rc_qacks, dev->n_rc_acks, - dev->n_seq_naks, dev->n_rdma_seq, dev->n_rnr_naks, - dev->n_other_naks, dev->n_timeouts, - dev->n_rdma_dup_busy, dev->n_piowait, - dev->n_no_piobuf, dev->n_pkt_drops, dev->n_wqe_errs); - for (i = 0; i < ARRAY_SIZE(dev->opstats); i++) { + struct ipath_ibdev *dev = + container_of(cdev, struct ipath_ibdev, ibdev.class_dev); + int i; + int len; + + len = sprintf(buf, + "RC resends %d\n" + "RC no QACK %d\n" + "RC ACKs %d\n" + "RC SEQ NAKs %d\n" + "RC RDMA seq %d\n" + "RC RNR NAKs %d\n" + "RC OTH NAKs %d\n" + "RC timeouts %d\n" + "RC RDMA dup %d\n" + "piobuf wait %d\n" + "no piobuf %d\n" + "PKT drops %d\n" + "WQE errs %d\n", + dev->n_rc_resends, dev->n_rc_qacks, dev->n_rc_acks, + dev->n_seq_naks, dev->n_rdma_seq, dev->n_rnr_naks, + dev->n_other_naks, dev->n_timeouts, + dev->n_rdma_dup_busy, dev->n_piowait, + dev->n_no_piobuf, dev->n_pkt_drops, dev->n_wqe_errs); + for (i = 0; i < ARRAY_SIZE(dev->opstats); i++) { const struct ipath_opcode_stats *si = &dev->opstats[i]; - if (!si->n_packets && !si->n_bytes) - continue; - len += sprintf(buf + len, "%02x %llu/%llu\n", i, + if (!si->n_packets && !si->n_bytes) + continue; + len += sprintf(buf + len, "%02x %llu/%llu\n", i, (unsigned long long) si->n_packets, - (unsigned long long) si->n_bytes); - } - return len; + (unsigned long long) si->n_bytes); + } + return len; } static CLASS_DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL); @@ -1194,25 +1176,25 @@ static CLASS_DEVICE_ATTR(board_id, S_IRUGO, show_hca, NULL); static CLASS_DEVICE_ATTR(stats, S_IRUGO, show_stats, NULL); static struct class_device_attribute *ipath_class_attributes[] = { - &class_device_attr_hw_rev, - &class_device_attr_hca_type, - &class_device_attr_board_id, - &class_device_attr_stats + &class_device_attr_hw_rev, + &class_device_attr_hca_type, + &class_device_attr_board_id, + &class_device_attr_stats }; static int ipath_verbs_register_sysfs(struct ib_device *dev) { - int i; + int i; int ret; - for (i = 0; i < ARRAY_SIZE(ipath_class_attributes); ++i) - if (class_device_create_file(&dev->class_dev, - ipath_class_attributes[i])) { - ret = 1; + for (i = 0; i < ARRAY_SIZE(ipath_class_attributes); ++i) + if (class_device_create_file(&dev->class_dev, + ipath_class_attributes[i])) { + ret = 1; goto bail; } - ret = 0; + ret = 0; bail: return ret; diff --git a/drivers/infiniband/hw/ipath/ipath_verbs.h b/drivers/infiniband/hw/ipath/ipath_verbs.h index b824632b2a8..4f8d59300e9 100644 --- a/drivers/infiniband/hw/ipath/ipath_verbs.h +++ b/drivers/infiniband/hw/ipath/ipath_verbs.h @@ -282,7 +282,8 @@ struct ipath_srq { */ struct ipath_qp { struct ib_qp ibqp; - struct ipath_qp *next; /* link list for QPN hash table */ + struct ipath_qp *next; /* link list for QPN hash table */ + struct ipath_qp *timer_next; /* link list for ipath_ib_timer() */ struct list_head piowait; /* link for wait PIO buf */ struct list_head timerwait; /* link for waiting for timeouts */ struct ib_ah_attr remote_ah_attr; @@ -577,8 +578,6 @@ int ipath_init_qp_table(struct ipath_ibdev *idev, int size); void ipath_sqerror_qp(struct ipath_qp *qp, struct ib_wc *wc); -void ipath_error_qp(struct ipath_qp *qp); - void ipath_get_credit(struct ipath_qp *qp, u32 aeth); void ipath_do_rc_send(unsigned long data); @@ -607,9 +606,6 @@ void ipath_rc_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr, void ipath_restart_rc(struct ipath_qp *qp, u32 psn, struct ib_wc *wc); -void ipath_ud_loopback(struct ipath_qp *sqp, struct ipath_sge_state *ss, - u32 length, struct ib_send_wr *wr, struct ib_wc *wc); - int ipath_post_ud_send(struct ipath_qp *qp, struct ib_send_wr *wr); void ipath_ud_rcv(struct ipath_ibdev *dev, struct ipath_ib_header *hdr, diff --git a/drivers/infiniband/hw/ipath/ips_common.h b/drivers/infiniband/hw/ipath/ips_common.h index 410a764dfce..ab7cbbbfd03 100644 --- a/drivers/infiniband/hw/ipath/ips_common.h +++ b/drivers/infiniband/hw/ipath/ips_common.h @@ -95,7 +95,7 @@ struct ether_header { __u8 seq_num; __le32 len; /* MUST be of word size due to PIO write requirements */ - __u32 csum; + __le32 csum; __le16 csum_offset; __le16 flags; __u16 first_2_bytes; diff --git a/drivers/infiniband/hw/mthca/mthca_cmd.c b/drivers/infiniband/hw/mthca/mthca_cmd.c index 1985b5dfa48..798e13e14fa 100644 --- a/drivers/infiniband/hw/mthca/mthca_cmd.c +++ b/drivers/infiniband/hw/mthca/mthca_cmd.c @@ -182,7 +182,7 @@ struct mthca_cmd_context { u8 status; }; -static int fw_cmd_doorbell = 1; +static int fw_cmd_doorbell = 0; module_param(fw_cmd_doorbell, int, 0644); MODULE_PARM_DESC(fw_cmd_doorbell, "post FW commands through doorbell page if nonzero " "(and supported by FW)"); diff --git a/drivers/infiniband/hw/mthca/mthca_cq.c b/drivers/infiniband/hw/mthca/mthca_cq.c index 312cf90731e..205854e9c66 100644 --- a/drivers/infiniband/hw/mthca/mthca_cq.c +++ b/drivers/infiniband/hw/mthca/mthca_cq.c @@ -238,9 +238,9 @@ void mthca_cq_event(struct mthca_dev *dev, u32 cqn, spin_lock(&dev->cq_table.lock); cq = mthca_array_get(&dev->cq_table.cq, cqn & (dev->limits.num_cqs - 1)); - if (cq) - atomic_inc(&cq->refcount); + ++cq->refcount; + spin_unlock(&dev->cq_table.lock); if (!cq) { @@ -254,8 +254,10 @@ void mthca_cq_event(struct mthca_dev *dev, u32 cqn, if (cq->ibcq.event_handler) cq->ibcq.event_handler(&event, cq->ibcq.cq_context); - if (atomic_dec_and_test(&cq->refcount)) + spin_lock(&dev->cq_table.lock); + if (!--cq->refcount) wake_up(&cq->wait); + spin_unlock(&dev->cq_table.lock); } static inline int is_recv_cqe(struct mthca_cqe *cqe) @@ -267,23 +269,13 @@ static inline int is_recv_cqe(struct mthca_cqe *cqe) return !(cqe->is_send & 0x80); } -void mthca_cq_clean(struct mthca_dev *dev, u32 cqn, u32 qpn, +void mthca_cq_clean(struct mthca_dev *dev, struct mthca_cq *cq, u32 qpn, struct mthca_srq *srq) { - struct mthca_cq *cq; struct mthca_cqe *cqe; u32 prod_index; int nfreed = 0; - spin_lock_irq(&dev->cq_table.lock); - cq = mthca_array_get(&dev->cq_table.cq, cqn & (dev->limits.num_cqs - 1)); - if (cq) - atomic_inc(&cq->refcount); - spin_unlock_irq(&dev->cq_table.lock); - - if (!cq) - return; - spin_lock_irq(&cq->lock); /* @@ -301,7 +293,7 @@ void mthca_cq_clean(struct mthca_dev *dev, u32 cqn, u32 qpn, if (0) mthca_dbg(dev, "Cleaning QPN %06x from CQN %06x; ci %d, pi %d\n", - qpn, cqn, cq->cons_index, prod_index); + qpn, cq->cqn, cq->cons_index, prod_index); /* * Now sweep backwards through the CQ, removing CQ entries @@ -325,8 +317,6 @@ void mthca_cq_clean(struct mthca_dev *dev, u32 cqn, u32 qpn, } spin_unlock_irq(&cq->lock); - if (atomic_dec_and_test(&cq->refcount)) - wake_up(&cq->wait); } void mthca_cq_resize_copy_cqes(struct mthca_cq *cq) @@ -821,7 +811,7 @@ int mthca_init_cq(struct mthca_dev *dev, int nent, } spin_lock_init(&cq->lock); - atomic_set(&cq->refcount, 1); + cq->refcount = 1; init_waitqueue_head(&cq->wait); memset(cq_context, 0, sizeof *cq_context); @@ -896,6 +886,17 @@ err_out: return err; } +static inline int get_cq_refcount(struct mthca_dev *dev, struct mthca_cq *cq) +{ + int c; + + spin_lock_irq(&dev->cq_table.lock); + c = cq->refcount; + spin_unlock_irq(&dev->cq_table.lock); + + return c; +} + void mthca_free_cq(struct mthca_dev *dev, struct mthca_cq *cq) { @@ -929,6 +930,7 @@ void mthca_free_cq(struct mthca_dev *dev, spin_lock_irq(&dev->cq_table.lock); mthca_array_clear(&dev->cq_table.cq, cq->cqn & (dev->limits.num_cqs - 1)); + --cq->refcount; spin_unlock_irq(&dev->cq_table.lock); if (dev->mthca_flags & MTHCA_FLAG_MSI_X) @@ -936,8 +938,7 @@ void mthca_free_cq(struct mthca_dev *dev, else synchronize_irq(dev->pdev->irq); - atomic_dec(&cq->refcount); - wait_event(cq->wait, !atomic_read(&cq->refcount)); + wait_event(cq->wait, !get_cq_refcount(dev, cq)); if (cq->is_kernel) { mthca_free_cq_buf(dev, &cq->buf, cq->ibcq.cqe); diff --git a/drivers/infiniband/hw/mthca/mthca_dev.h b/drivers/infiniband/hw/mthca/mthca_dev.h index 4c1dcb4c182..f8160b8de09 100644 --- a/drivers/infiniband/hw/mthca/mthca_dev.h +++ b/drivers/infiniband/hw/mthca/mthca_dev.h @@ -496,7 +496,7 @@ void mthca_free_cq(struct mthca_dev *dev, void mthca_cq_completion(struct mthca_dev *dev, u32 cqn); void mthca_cq_event(struct mthca_dev *dev, u32 cqn, enum ib_event_type event_type); -void mthca_cq_clean(struct mthca_dev *dev, u32 cqn, u32 qpn, +void mthca_cq_clean(struct mthca_dev *dev, struct mthca_cq *cq, u32 qpn, struct mthca_srq *srq); void mthca_cq_resize_copy_cqes(struct mthca_cq *cq); int mthca_alloc_cq_buf(struct mthca_dev *dev, struct mthca_cq_buf *buf, int nent); diff --git a/drivers/infiniband/hw/mthca/mthca_mad.c b/drivers/infiniband/hw/mthca/mthca_mad.c index f235c7ea42f..4730863ece9 100644 --- a/drivers/infiniband/hw/mthca/mthca_mad.c +++ b/drivers/infiniband/hw/mthca/mthca_mad.c @@ -49,7 +49,7 @@ enum { MTHCA_VENDOR_CLASS2 = 0xa }; -int mthca_update_rate(struct mthca_dev *dev, u8 port_num) +static int mthca_update_rate(struct mthca_dev *dev, u8 port_num) { struct ib_port_attr *tprops = NULL; int ret; diff --git a/drivers/infiniband/hw/mthca/mthca_mr.c b/drivers/infiniband/hw/mthca/mthca_mr.c index 25e1c1db9a4..a486dec1707 100644 --- a/drivers/infiniband/hw/mthca/mthca_mr.c +++ b/drivers/infiniband/hw/mthca/mthca_mr.c @@ -761,6 +761,7 @@ void mthca_arbel_fmr_unmap(struct mthca_dev *dev, struct mthca_fmr *fmr) int __devinit mthca_init_mr_table(struct mthca_dev *dev) { + unsigned long addr; int err, i; err = mthca_alloc_init(&dev->mr_table.mpt_alloc, @@ -796,9 +797,12 @@ int __devinit mthca_init_mr_table(struct mthca_dev *dev) goto err_fmr_mpt; } + addr = pci_resource_start(dev->pdev, 4) + + ((pci_resource_len(dev->pdev, 4) - 1) & + dev->mr_table.mpt_base); + dev->mr_table.tavor_fmr.mpt_base = - ioremap(dev->mr_table.mpt_base, - (1 << i) * sizeof (struct mthca_mpt_entry)); + ioremap(addr, (1 << i) * sizeof(struct mthca_mpt_entry)); if (!dev->mr_table.tavor_fmr.mpt_base) { mthca_warn(dev, "MPT ioremap for FMR failed.\n"); @@ -806,9 +810,12 @@ int __devinit mthca_init_mr_table(struct mthca_dev *dev) goto err_fmr_mpt; } + addr = pci_resource_start(dev->pdev, 4) + + ((pci_resource_len(dev->pdev, 4) - 1) & + dev->mr_table.mtt_base); + dev->mr_table.tavor_fmr.mtt_base = - ioremap(dev->mr_table.mtt_base, - (1 << i) * MTHCA_MTT_SEG_SIZE); + ioremap(addr, (1 << i) * MTHCA_MTT_SEG_SIZE); if (!dev->mr_table.tavor_fmr.mtt_base) { mthca_warn(dev, "MTT ioremap for FMR failed.\n"); err = -ENOMEM; diff --git a/drivers/infiniband/hw/mthca/mthca_provider.c b/drivers/infiniband/hw/mthca/mthca_provider.c index 565a24b1756..a2eae8a3016 100644 --- a/drivers/infiniband/hw/mthca/mthca_provider.c +++ b/drivers/infiniband/hw/mthca/mthca_provider.c @@ -306,7 +306,7 @@ static int mthca_query_gid(struct ib_device *ibdev, u8 port, goto out; } - memcpy(gid->raw + 8, out_mad->data + (index % 8) * 16, 8); + memcpy(gid->raw + 8, out_mad->data + (index % 8) * 8, 8); out: kfree(in_mad); diff --git a/drivers/infiniband/hw/mthca/mthca_provider.h b/drivers/infiniband/hw/mthca/mthca_provider.h index 6676a786d69..179a8f610d0 100644 --- a/drivers/infiniband/hw/mthca/mthca_provider.h +++ b/drivers/infiniband/hw/mthca/mthca_provider.h @@ -139,11 +139,12 @@ struct mthca_ah { * a qp may be locked, with the send cq locked first. No other * nesting should be done. * - * Each struct mthca_cq/qp also has an atomic_t ref count. The - * pointer from the cq/qp_table to the struct counts as one reference. - * This reference also is good for access through the consumer API, so - * modifying the CQ/QP etc doesn't need to take another reference. - * Access because of a completion being polled does need a reference. + * 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 + * struct counts as one reference. This reference also is good for + * access through the consumer API, so modifying the CQ/QP etc doesn't + * need to take another reference. Access to a QP because of a + * completion being polled does not need a reference either. * * Finally, each struct mthca_cq/qp has a wait_queue_head_t for the * destroy function to sleep on. @@ -159,8 +160,9 @@ struct mthca_ah { * - decrement ref count; if zero, wake up waiters * * To destroy a CQ/QP, we can do the following: - * - lock cq/qp_table, remove pointer, unlock cq/qp_table lock - * - decrement ref count + * - lock cq/qp_table + * - remove pointer and decrement ref count + * - unlock cq/qp_table lock * - wait_event until ref count is zero * * It is the consumer's responsibilty to make sure that no QP @@ -197,7 +199,7 @@ struct mthca_cq_resize { struct mthca_cq { struct ib_cq ibcq; spinlock_t lock; - atomic_t refcount; + int refcount; int cqn; u32 cons_index; struct mthca_cq_buf buf; @@ -217,7 +219,7 @@ struct mthca_cq { struct mthca_srq { struct ib_srq ibsrq; spinlock_t lock; - atomic_t refcount; + int refcount; int srqn; int max; int max_gs; @@ -254,7 +256,7 @@ struct mthca_wq { struct mthca_qp { struct ib_qp ibqp; - atomic_t refcount; + int refcount; u32 qpn; int is_direct; u8 port; /* for SQP and memfree use only */ diff --git a/drivers/infiniband/hw/mthca/mthca_qp.c b/drivers/infiniband/hw/mthca/mthca_qp.c index f37b0e36732..07c13be07a4 100644 --- a/drivers/infiniband/hw/mthca/mthca_qp.c +++ b/drivers/infiniband/hw/mthca/mthca_qp.c @@ -240,7 +240,7 @@ void mthca_qp_event(struct mthca_dev *dev, u32 qpn, spin_lock(&dev->qp_table.lock); qp = mthca_array_get(&dev->qp_table.qp, qpn & (dev->limits.num_qps - 1)); if (qp) - atomic_inc(&qp->refcount); + ++qp->refcount; spin_unlock(&dev->qp_table.lock); if (!qp) { @@ -257,8 +257,10 @@ void mthca_qp_event(struct mthca_dev *dev, u32 qpn, if (qp->ibqp.event_handler) qp->ibqp.event_handler(&event, qp->ibqp.qp_context); - if (atomic_dec_and_test(&qp->refcount)) + spin_lock(&dev->qp_table.lock); + if (!--qp->refcount) wake_up(&qp->wait); + spin_unlock(&dev->qp_table.lock); } static int to_mthca_state(enum ib_qp_state ib_state) @@ -833,10 +835,10 @@ int mthca_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, int attr_mask) * entries and reinitialize the QP. */ if (new_state == IB_QPS_RESET && !qp->ibqp.uobject) { - mthca_cq_clean(dev, to_mcq(qp->ibqp.send_cq)->cqn, qp->qpn, + mthca_cq_clean(dev, to_mcq(qp->ibqp.send_cq), qp->qpn, qp->ibqp.srq ? to_msrq(qp->ibqp.srq) : NULL); if (qp->ibqp.send_cq != qp->ibqp.recv_cq) - mthca_cq_clean(dev, to_mcq(qp->ibqp.recv_cq)->cqn, qp->qpn, + mthca_cq_clean(dev, to_mcq(qp->ibqp.recv_cq), qp->qpn, qp->ibqp.srq ? to_msrq(qp->ibqp.srq) : NULL); mthca_wq_init(&qp->sq); @@ -1096,7 +1098,7 @@ static int mthca_alloc_qp_common(struct mthca_dev *dev, int ret; int i; - atomic_set(&qp->refcount, 1); + qp->refcount = 1; init_waitqueue_head(&qp->wait); qp->state = IB_QPS_RESET; qp->atomic_rd_en = 0; @@ -1318,6 +1320,17 @@ int mthca_alloc_sqp(struct mthca_dev *dev, return err; } +static inline int get_qp_refcount(struct mthca_dev *dev, struct mthca_qp *qp) +{ + int c; + + spin_lock_irq(&dev->qp_table.lock); + c = qp->refcount; + spin_unlock_irq(&dev->qp_table.lock); + + return c; +} + void mthca_free_qp(struct mthca_dev *dev, struct mthca_qp *qp) { @@ -1339,14 +1352,14 @@ void mthca_free_qp(struct mthca_dev *dev, spin_lock(&dev->qp_table.lock); mthca_array_clear(&dev->qp_table.qp, qp->qpn & (dev->limits.num_qps - 1)); + --qp->refcount; spin_unlock(&dev->qp_table.lock); if (send_cq != recv_cq) spin_unlock(&recv_cq->lock); spin_unlock_irq(&send_cq->lock); - atomic_dec(&qp->refcount); - wait_event(qp->wait, !atomic_read(&qp->refcount)); + wait_event(qp->wait, !get_qp_refcount(dev, qp)); if (qp->state != IB_QPS_RESET) mthca_MODIFY_QP(dev, qp->state, IB_QPS_RESET, qp->qpn, 0, @@ -1358,10 +1371,10 @@ void mthca_free_qp(struct mthca_dev *dev, * unref the mem-free tables and free the QPN in our table. */ if (!qp->ibqp.uobject) { - mthca_cq_clean(dev, to_mcq(qp->ibqp.send_cq)->cqn, qp->qpn, + mthca_cq_clean(dev, to_mcq(qp->ibqp.send_cq), qp->qpn, qp->ibqp.srq ? to_msrq(qp->ibqp.srq) : NULL); if (qp->ibqp.send_cq != qp->ibqp.recv_cq) - mthca_cq_clean(dev, to_mcq(qp->ibqp.recv_cq)->cqn, qp->qpn, + mthca_cq_clean(dev, to_mcq(qp->ibqp.recv_cq), qp->qpn, qp->ibqp.srq ? to_msrq(qp->ibqp.srq) : NULL); mthca_free_memfree(dev, qp); @@ -1714,23 +1727,7 @@ int mthca_tavor_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr, ind = qp->rq.next_ind; - for (nreq = 0; wr; ++nreq, wr = wr->next) { - if (unlikely(nreq == MTHCA_TAVOR_MAX_WQES_PER_RECV_DB)) { - nreq = 0; - - doorbell[0] = cpu_to_be32((qp->rq.next_ind << qp->rq.wqe_shift) | size0); - doorbell[1] = cpu_to_be32(qp->qpn << 8); - - wmb(); - - mthca_write64(doorbell, - dev->kar + MTHCA_RECEIVE_DOORBELL, - MTHCA_GET_DOORBELL_LOCK(&dev->doorbell_lock)); - - qp->rq.head += MTHCA_TAVOR_MAX_WQES_PER_RECV_DB; - size0 = 0; - } - + for (nreq = 0; wr; wr = wr->next) { if (mthca_wq_overflow(&qp->rq, nreq, qp->ibqp.recv_cq)) { mthca_err(dev, "RQ %06x full (%u head, %u tail," " %d max, %d nreq)\n", qp->qpn, @@ -1784,6 +1781,23 @@ int mthca_tavor_post_receive(struct ib_qp *ibqp, struct ib_recv_wr *wr, ++ind; if (unlikely(ind >= qp->rq.max)) ind -= qp->rq.max; + + ++nreq; + if (unlikely(nreq == MTHCA_TAVOR_MAX_WQES_PER_RECV_DB)) { + nreq = 0; + + doorbell[0] = cpu_to_be32((qp->rq.next_ind << qp->rq.wqe_shift) | size0); + doorbell[1] = cpu_to_be32(qp->qpn << 8); + + wmb(); + + mthca_write64(doorbell, + dev->kar + MTHCA_RECEIVE_DOORBELL, + MTHCA_GET_DOORBELL_LOCK(&dev->doorbell_lock)); + + qp->rq.head += MTHCA_TAVOR_MAX_WQES_PER_RECV_DB; + size0 = 0; + } } out: diff --git a/drivers/infiniband/hw/mthca/mthca_srq.c b/drivers/infiniband/hw/mthca/mthca_srq.c index adcaf85355a..b292fefa3b4 100644 --- a/drivers/infiniband/hw/mthca/mthca_srq.c +++ b/drivers/infiniband/hw/mthca/mthca_srq.c @@ -241,7 +241,7 @@ int mthca_alloc_srq(struct mthca_dev *dev, struct mthca_pd *pd, goto err_out_mailbox; spin_lock_init(&srq->lock); - atomic_set(&srq->refcount, 1); + srq->refcount = 1; init_waitqueue_head(&srq->wait); if (mthca_is_memfree(dev)) @@ -308,6 +308,17 @@ err_out: return err; } +static inline int get_srq_refcount(struct mthca_dev *dev, struct mthca_srq *srq) +{ + int c; + + spin_lock_irq(&dev->srq_table.lock); + c = srq->refcount; + spin_unlock_irq(&dev->srq_table.lock); + + return c; +} + void mthca_free_srq(struct mthca_dev *dev, struct mthca_srq *srq) { struct mthca_mailbox *mailbox; @@ -329,10 +340,10 @@ void mthca_free_srq(struct mthca_dev *dev, struct mthca_srq *srq) spin_lock_irq(&dev->srq_table.lock); mthca_array_clear(&dev->srq_table.srq, srq->srqn & (dev->limits.num_srqs - 1)); + --srq->refcount; spin_unlock_irq(&dev->srq_table.lock); - atomic_dec(&srq->refcount); - wait_event(srq->wait, !atomic_read(&srq->refcount)); + wait_event(srq->wait, !get_srq_refcount(dev, srq)); if (!srq->ibsrq.uobject) { mthca_free_srq_buf(dev, srq); @@ -414,7 +425,7 @@ void mthca_srq_event(struct mthca_dev *dev, u32 srqn, spin_lock(&dev->srq_table.lock); srq = mthca_array_get(&dev->srq_table.srq, srqn & (dev->limits.num_srqs - 1)); if (srq) - atomic_inc(&srq->refcount); + ++srq->refcount; spin_unlock(&dev->srq_table.lock); if (!srq) { @@ -431,8 +442,10 @@ void mthca_srq_event(struct mthca_dev *dev, u32 srqn, srq->ibsrq.event_handler(&event, srq->ibsrq.srq_context); out: - if (atomic_dec_and_test(&srq->refcount)) + spin_lock(&dev->srq_table.lock); + if (!--srq->refcount) wake_up(&srq->wait); + spin_unlock(&dev->srq_table.lock); } /* @@ -477,26 +490,7 @@ int mthca_tavor_post_srq_recv(struct ib_srq *ibsrq, struct ib_recv_wr *wr, first_ind = srq->first_free; - for (nreq = 0; wr; ++nreq, wr = wr->next) { - if (unlikely(nreq == MTHCA_TAVOR_MAX_WQES_PER_RECV_DB)) { - nreq = 0; - - doorbell[0] = cpu_to_be32(first_ind << srq->wqe_shift); - doorbell[1] = cpu_to_be32(srq->srqn << 8); - - /* - * Make sure that descriptors are written - * before doorbell is rung. - */ - wmb(); - - mthca_write64(doorbell, - dev->kar + MTHCA_RECEIVE_DOORBELL, - MTHCA_GET_DOORBELL_LOCK(&dev->doorbell_lock)); - - first_ind = srq->first_free; - } - + for (nreq = 0; wr; wr = wr->next) { ind = srq->first_free; if (ind < 0) { @@ -556,6 +550,26 @@ int mthca_tavor_post_srq_recv(struct ib_srq *ibsrq, struct ib_recv_wr *wr, srq->wrid[ind] = wr->wr_id; srq->first_free = next_ind; + + ++nreq; + if (unlikely(nreq == MTHCA_TAVOR_MAX_WQES_PER_RECV_DB)) { + nreq = 0; + + doorbell[0] = cpu_to_be32(first_ind << srq->wqe_shift); + doorbell[1] = cpu_to_be32(srq->srqn << 8); + + /* + * Make sure that descriptors are written + * before doorbell is rung. + */ + wmb(); + + mthca_write64(doorbell, + dev->kar + MTHCA_RECEIVE_DOORBELL, + MTHCA_GET_DOORBELL_LOCK(&dev->doorbell_lock)); + + first_ind = srq->first_free; + } } if (likely(nreq)) { diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c index a54da42849a..8406839b91c 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c @@ -275,6 +275,7 @@ static void ipoib_ib_handle_wc(struct net_device *dev, spin_lock_irqsave(&priv->tx_lock, flags); ++priv->tx_tail; if (netif_queue_stopped(dev) && + test_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags) && priv->tx_head - priv->tx_tail <= ipoib_sendq_size >> 1) netif_wake_queue(dev); spin_unlock_irqrestore(&priv->tx_lock, flags); diff --git a/drivers/infiniband/ulp/ipoib/ipoib_vlan.c b/drivers/infiniband/ulp/ipoib/ipoib_vlan.c index 4ca175553f9..f887780e809 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_vlan.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_vlan.c @@ -158,10 +158,8 @@ int ipoib_vlan_delete(struct net_device *pdev, unsigned short pkey) if (priv->pkey == pkey) { unregister_netdev(priv->dev); ipoib_dev_cleanup(priv->dev); - list_del(&priv->list); - - kfree(priv); + free_netdev(priv->dev); ret = 0; break; diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c index 5f2b3f6e4c4..9cbdffa08dc 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.c +++ b/drivers/infiniband/ulp/srp/ib_srp.c @@ -340,7 +340,10 @@ static void srp_disconnect_target(struct srp_target_port *target) /* XXX should send SRP_I_LOGOUT request */ init_completion(&target->done); - ib_send_cm_dreq(target->cm_id, NULL, 0); + if (ib_send_cm_dreq(target->cm_id, NULL, 0)) { + printk(KERN_DEBUG PFX "Sending CM DREQ failed\n"); + return; + } wait_for_completion(&target->done); } @@ -351,7 +354,6 @@ static void srp_remove_work(void *target_ptr) spin_lock_irq(target->scsi_host->host_lock); if (target->state != SRP_TARGET_DEAD) { spin_unlock_irq(target->scsi_host->host_lock); - scsi_host_put(target->scsi_host); return; } target->state = SRP_TARGET_REMOVED; @@ -365,8 +367,6 @@ static void srp_remove_work(void *target_ptr) ib_destroy_cm_id(target->cm_id); srp_free_target_ib(target); scsi_host_put(target->scsi_host); - /* And another put to really free the target port... */ - scsi_host_put(target->scsi_host); } static int srp_connect_target(struct srp_target_port *target) @@ -409,6 +409,34 @@ static int srp_connect_target(struct srp_target_port *target) } } +static void srp_unmap_data(struct scsi_cmnd *scmnd, + struct srp_target_port *target, + struct srp_request *req) +{ + struct scatterlist *scat; + int nents; + + if (!scmnd->request_buffer || + (scmnd->sc_data_direction != DMA_TO_DEVICE && + scmnd->sc_data_direction != DMA_FROM_DEVICE)) + return; + + /* + * This handling of non-SG commands can be killed when the + * SCSI midlayer no longer generates non-SG commands. + */ + if (likely(scmnd->use_sg)) { + nents = scmnd->use_sg; + scat = scmnd->request_buffer; + } else { + nents = 1; + scat = &req->fake_sg; + } + + dma_unmap_sg(target->srp_host->dev->dma_device, scat, nents, + scmnd->sc_data_direction); +} + static int srp_reconnect_target(struct srp_target_port *target) { struct ib_cm_id *new_cm_id; @@ -455,16 +483,16 @@ static int srp_reconnect_target(struct srp_target_port *target) list_for_each_entry(req, &target->req_queue, list) { req->scmnd->result = DID_RESET << 16; req->scmnd->scsi_done(req->scmnd); + srp_unmap_data(req->scmnd, target, req); } target->rx_head = 0; target->tx_head = 0; target->tx_tail = 0; - target->req_head = 0; - for (i = 0; i < SRP_SQ_SIZE - 1; ++i) - target->req_ring[i].next = i + 1; - target->req_ring[SRP_SQ_SIZE - 1].next = -1; + INIT_LIST_HEAD(&target->free_reqs); INIT_LIST_HEAD(&target->req_queue); + for (i = 0; i < SRP_SQ_SIZE; ++i) + list_add_tail(&target->req_ring[i].list, &target->free_reqs); ret = srp_connect_target(target); if (ret) @@ -589,32 +617,10 @@ static int srp_map_data(struct scsi_cmnd *scmnd, struct srp_target_port *target, return len; } -static void srp_unmap_data(struct scsi_cmnd *scmnd, - struct srp_target_port *target, - struct srp_request *req) +static void srp_remove_req(struct srp_target_port *target, struct srp_request *req) { - struct scatterlist *scat; - int nents; - - if (!scmnd->request_buffer || - (scmnd->sc_data_direction != DMA_TO_DEVICE && - scmnd->sc_data_direction != DMA_FROM_DEVICE)) - return; - - /* - * This handling of non-SG commands can be killed when the - * SCSI midlayer no longer generates non-SG commands. - */ - if (likely(scmnd->use_sg)) { - nents = scmnd->use_sg; - scat = scmnd->request_buffer; - } else { - nents = 1; - scat = &req->fake_sg; - } - - dma_unmap_sg(target->srp_host->dev->dma_device, scat, nents, - scmnd->sc_data_direction); + srp_unmap_data(req->scmnd, target, req); + list_move_tail(&req->list, &target->free_reqs); } static void srp_process_rsp(struct srp_target_port *target, struct srp_rsp *rsp) @@ -639,7 +645,7 @@ static void srp_process_rsp(struct srp_target_port *target, struct srp_rsp *rsp) req->tsk_status = rsp->data[3]; complete(&req->done); } else { - scmnd = req->scmnd; + scmnd = req->scmnd; if (!scmnd) printk(KERN_ERR "Null scmnd for RSP w/tag %016llx\n", (unsigned long long) rsp->tag); @@ -657,16 +663,11 @@ static void srp_process_rsp(struct srp_target_port *target, struct srp_rsp *rsp) else if (rsp->flags & (SRP_RSP_FLAG_DIOVER | SRP_RSP_FLAG_DIUNDER)) scmnd->resid = be32_to_cpu(rsp->data_in_res_cnt); - srp_unmap_data(scmnd, target, req); - if (!req->tsk_mgmt) { - req->scmnd = NULL; scmnd->host_scribble = (void *) -1L; scmnd->scsi_done(scmnd); - list_del(&req->list); - req->next = target->req_head; - target->req_head = rsp->tag & ~SRP_TAG_TSK_MGMT; + srp_remove_req(target, req); } else req->cmd_done = 1; } @@ -853,7 +854,6 @@ static int srp_queuecommand(struct scsi_cmnd *scmnd, struct srp_request *req; struct srp_iu *iu; struct srp_cmd *cmd; - long req_index; int len; if (target->state == SRP_TARGET_CONNECTING) @@ -873,22 +873,20 @@ static int srp_queuecommand(struct scsi_cmnd *scmnd, dma_sync_single_for_cpu(target->srp_host->dev->dma_device, iu->dma, SRP_MAX_IU_LEN, DMA_TO_DEVICE); - req_index = target->req_head; + req = list_entry(target->free_reqs.next, struct srp_request, list); scmnd->scsi_done = done; scmnd->result = 0; - scmnd->host_scribble = (void *) req_index; + scmnd->host_scribble = (void *) (long) req->index; cmd = iu->buf; memset(cmd, 0, sizeof *cmd); cmd->opcode = SRP_CMD; cmd->lun = cpu_to_be64((u64) scmnd->device->lun << 48); - cmd->tag = req_index; + cmd->tag = req->index; memcpy(cmd->cdb, scmnd->cmnd, scmnd->cmd_len); - req = &target->req_ring[req_index]; - req->scmnd = scmnd; req->cmd = iu; req->cmd_done = 0; @@ -913,8 +911,7 @@ static int srp_queuecommand(struct scsi_cmnd *scmnd, goto err_unmap; } - target->req_head = req->next; - list_add_tail(&req->list, &target->req_queue); + list_move_tail(&req->list, &target->req_queue); return 0; @@ -1137,30 +1134,20 @@ static int srp_cm_handler(struct ib_cm_id *cm_id, struct ib_cm_event *event) return 0; } -static int srp_send_tsk_mgmt(struct scsi_cmnd *scmnd, u8 func) +static int srp_send_tsk_mgmt(struct srp_target_port *target, + struct srp_request *req, u8 func) { - struct srp_target_port *target = host_to_target(scmnd->device->host); - struct srp_request *req; struct srp_iu *iu; struct srp_tsk_mgmt *tsk_mgmt; - int req_index; - int ret = FAILED; spin_lock_irq(target->scsi_host->host_lock); if (target->state == SRP_TARGET_DEAD || target->state == SRP_TARGET_REMOVED) { - scmnd->result = DID_BAD_TARGET << 16; + req->scmnd->result = DID_BAD_TARGET << 16; goto out; } - if (scmnd->host_scribble == (void *) -1L) - goto out; - - req_index = (long) scmnd->host_scribble; - printk(KERN_ERR "Abort for req_index %d\n", req_index); - - req = &target->req_ring[req_index]; init_completion(&req->done); iu = __srp_get_tx_iu(target); @@ -1171,10 +1158,10 @@ static int srp_send_tsk_mgmt(struct scsi_cmnd *scmnd, u8 func) memset(tsk_mgmt, 0, sizeof *tsk_mgmt); tsk_mgmt->opcode = SRP_TSK_MGMT; - tsk_mgmt->lun = cpu_to_be64((u64) scmnd->device->lun << 48); - tsk_mgmt->tag = req_index | SRP_TAG_TSK_MGMT; + tsk_mgmt->lun = cpu_to_be64((u64) req->scmnd->device->lun << 48); + tsk_mgmt->tag = req->index | SRP_TAG_TSK_MGMT; tsk_mgmt->tsk_mgmt_func = func; - tsk_mgmt->task_tag = req_index; + tsk_mgmt->task_tag = req->index; if (__srp_post_send(target, iu, sizeof *tsk_mgmt)) goto out; @@ -1182,39 +1169,85 @@ static int srp_send_tsk_mgmt(struct scsi_cmnd *scmnd, u8 func) req->tsk_mgmt = iu; spin_unlock_irq(target->scsi_host->host_lock); + if (!wait_for_completion_timeout(&req->done, msecs_to_jiffies(SRP_ABORT_TIMEOUT_MS))) - return FAILED; - spin_lock_irq(target->scsi_host->host_lock); + return -1; - if (req->cmd_done) { - list_del(&req->list); - req->next = target->req_head; - target->req_head = req_index; - - scmnd->scsi_done(scmnd); - } else if (!req->tsk_status) { - scmnd->result = DID_ABORT << 16; - ret = SUCCESS; - } + return 0; out: spin_unlock_irq(target->scsi_host->host_lock); - return ret; + return -1; +} + +static int srp_find_req(struct srp_target_port *target, + struct scsi_cmnd *scmnd, + struct srp_request **req) +{ + if (scmnd->host_scribble == (void *) -1L) + return -1; + + *req = &target->req_ring[(long) scmnd->host_scribble]; + + return 0; } static int srp_abort(struct scsi_cmnd *scmnd) { + struct srp_target_port *target = host_to_target(scmnd->device->host); + struct srp_request *req; + int ret = SUCCESS; + printk(KERN_ERR "SRP abort called\n"); - return srp_send_tsk_mgmt(scmnd, SRP_TSK_ABORT_TASK); + if (srp_find_req(target, scmnd, &req)) + return FAILED; + if (srp_send_tsk_mgmt(target, req, SRP_TSK_ABORT_TASK)) + return FAILED; + + spin_lock_irq(target->scsi_host->host_lock); + + if (req->cmd_done) { + srp_remove_req(target, req); + scmnd->scsi_done(scmnd); + } else if (!req->tsk_status) { + srp_remove_req(target, req); + scmnd->result = DID_ABORT << 16; + } else + ret = FAILED; + + spin_unlock_irq(target->scsi_host->host_lock); + + return ret; } static int srp_reset_device(struct scsi_cmnd *scmnd) { + struct srp_target_port *target = host_to_target(scmnd->device->host); + struct srp_request *req, *tmp; + printk(KERN_ERR "SRP reset_device called\n"); - return srp_send_tsk_mgmt(scmnd, SRP_TSK_LUN_RESET); + if (srp_find_req(target, scmnd, &req)) + return FAILED; + if (srp_send_tsk_mgmt(target, req, SRP_TSK_LUN_RESET)) + return FAILED; + if (req->tsk_status) + return FAILED; + + spin_lock_irq(target->scsi_host->host_lock); + + list_for_each_entry_safe(req, tmp, &target->req_queue, list) + if (req->scmnd->device == scmnd->device) { + req->scmnd->result = DID_RESET << 16; + req->scmnd->scsi_done(req->scmnd); + srp_remove_req(target, req); + } + + spin_unlock_irq(target->scsi_host->host_lock); + + return SUCCESS; } static int srp_reset_host(struct scsi_cmnd *scmnd) @@ -1514,10 +1547,12 @@ static ssize_t srp_create_target(struct class_device *class_dev, INIT_WORK(&target->work, srp_reconnect_work, target); - for (i = 0; i < SRP_SQ_SIZE - 1; ++i) - target->req_ring[i].next = i + 1; - target->req_ring[SRP_SQ_SIZE - 1].next = -1; + INIT_LIST_HEAD(&target->free_reqs); INIT_LIST_HEAD(&target->req_queue); + for (i = 0; i < SRP_SQ_SIZE; ++i) { + target->req_ring[i].index = i; + list_add_tail(&target->req_ring[i].list, &target->free_reqs); + } ret = srp_parse_options(buf, target); if (ret) diff --git a/drivers/infiniband/ulp/srp/ib_srp.h b/drivers/infiniband/ulp/srp/ib_srp.h index bd7f7c3115d..c5cd43aae86 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.h +++ b/drivers/infiniband/ulp/srp/ib_srp.h @@ -101,7 +101,7 @@ struct srp_request { */ struct scatterlist fake_sg; struct completion done; - short next; + short index; u8 cmd_done; u8 tsk_status; }; @@ -133,7 +133,7 @@ struct srp_target_port { unsigned tx_tail; struct srp_iu *tx_ring[SRP_SQ_SIZE + 1]; - int req_head; + struct list_head free_reqs; struct list_head req_queue; struct srp_request req_ring[SRP_SQ_SIZE]; |