diff options
Diffstat (limited to 'drivers/infiniband/hw/ipath')
34 files changed, 2508 insertions, 2319 deletions
diff --git a/drivers/infiniband/hw/ipath/Kconfig b/drivers/infiniband/hw/ipath/Kconfig index 1db9489f1e8..574a678e7fd 100644 --- a/drivers/infiniband/hw/ipath/Kconfig +++ b/drivers/infiniband/hw/ipath/Kconfig @@ -1,16 +1,9 @@ -config IPATH_CORE - tristate "QLogic InfiniPath Driver" - depends on 64BIT && PCI_MSI && NET - ---help--- - This is a low-level driver for QLogic InfiniPath host channel - adapters (HCAs) based on the HT-400 and PE-800 chips. - config INFINIBAND_IPATH - tristate "QLogic InfiniPath Verbs Driver" - depends on IPATH_CORE && INFINIBAND + tristate "QLogic InfiniPath Driver" + depends on PCI_MSI && 64BIT && INFINIBAND ---help--- - This is a driver that provides InfiniBand verbs support for - QLogic InfiniPath host channel adapters (HCAs). This - allows these devices to be used with both kernel upper level - protocols such as IP-over-InfiniBand as well as with userspace - applications (in conjunction with InfiniBand userspace access). + This is a driver for QLogic InfiniPath host channel adapters, + including InfiniBand verbs support. This driver allows these + devices to be used with both kernel upper level protocols such + as IP-over-InfiniBand as well as with userspace applications + (in conjunction with InfiniBand userspace access). diff --git a/drivers/infiniband/hw/ipath/Makefile b/drivers/infiniband/hw/ipath/Makefile index b0bf7286413..5e29cb0095e 100644 --- a/drivers/infiniband/hw/ipath/Makefile +++ b/drivers/infiniband/hw/ipath/Makefile @@ -1,36 +1,35 @@ EXTRA_CFLAGS += -DIPATH_IDSTR='"QLogic kernel.org driver"' \ -DIPATH_KERN_TYPE=0 -obj-$(CONFIG_IPATH_CORE) += ipath_core.o obj-$(CONFIG_INFINIBAND_IPATH) += ib_ipath.o -ipath_core-y := \ +ib_ipath-y := \ + ipath_cq.o \ ipath_diag.o \ ipath_driver.o \ ipath_eeprom.o \ ipath_file_ops.o \ ipath_fs.o \ - ipath_ht400.o \ + ipath_iba6110.o \ + ipath_iba6120.o \ ipath_init_chip.o \ ipath_intr.o \ - ipath_layer.o \ - ipath_pe800.o \ - ipath_stats.o \ - ipath_sysfs.o \ - ipath_user_pages.o - -ipath_core-$(CONFIG_X86_64) += ipath_wc_x86_64.o - -ib_ipath-y := \ - ipath_cq.o \ ipath_keys.o \ + ipath_layer.o \ ipath_mad.o \ + ipath_mmap.o \ ipath_mr.o \ ipath_qp.o \ ipath_rc.o \ ipath_ruc.o \ ipath_srq.o \ + ipath_stats.o \ + ipath_sysfs.o \ ipath_uc.o \ ipath_ud.o \ - ipath_verbs.o \ - ipath_verbs_mcast.o + ipath_user_pages.o \ + ipath_verbs_mcast.o \ + ipath_verbs.o + +ib_ipath-$(CONFIG_X86_64) += ipath_wc_x86_64.o +ib_ipath-$(CONFIG_PPC64) += ipath_wc_ppc64.o diff --git a/drivers/infiniband/hw/ipath/ipath_common.h b/drivers/infiniband/hw/ipath/ipath_common.h index 062bd392e7e..f577905e3ac 100644 --- a/drivers/infiniband/hw/ipath/ipath_common.h +++ b/drivers/infiniband/hw/ipath/ipath_common.h @@ -106,9 +106,9 @@ struct infinipath_stats { __u64 sps_ether_spkts; /* number of "ethernet" packets received by driver */ __u64 sps_ether_rpkts; - /* number of SMA packets sent by driver */ + /* number of SMA packets sent by driver. Obsolete. */ __u64 sps_sma_spkts; - /* number of SMA packets received by driver */ + /* number of SMA packets received by driver. Obsolete. */ __u64 sps_sma_rpkts; /* number of times all ports rcvhdrq was full and packet dropped */ __u64 sps_hdrqfull; @@ -138,7 +138,7 @@ struct infinipath_stats { __u64 sps_pageunlocks; /* * Number of packets dropped in kernel other than errors (ether - * packets if ipath not configured, sma/mad, etc.) + * packets if ipath not configured, etc.) */ __u64 sps_krdrops; /* pad for future growth */ @@ -153,8 +153,6 @@ struct infinipath_stats { #define IPATH_STATUS_DISABLED 0x2 /* hardware disabled */ /* Device has been disabled via admin request */ #define IPATH_STATUS_ADMIN_DISABLED 0x4 -#define IPATH_STATUS_OIB_SMA 0x8 /* ipath_mad kernel SMA running */ -#define IPATH_STATUS_SMA 0x10 /* user SMA running */ /* Chip has been found and initted */ #define IPATH_STATUS_CHIP_PRESENT 0x20 /* IB link is at ACTIVE, usable for data traffic */ @@ -465,12 +463,11 @@ struct __ipath_sendpkt { struct ipath_iovec sps_iov[4]; }; -/* Passed into SMA special file's ->read and ->write methods. */ -struct ipath_sma_pkt -{ - __u32 unit; /* unit on which to send packet */ - __u64 data; /* address of payload in userspace */ - __u32 len; /* length of payload */ +/* Passed into diag data special file's ->write method. */ +struct ipath_diag_pkt { + __u32 unit; + __u64 data; + __u32 len; }; /* diff --git a/drivers/infiniband/hw/ipath/ipath_cq.c b/drivers/infiniband/hw/ipath/ipath_cq.c index 3efee341c9b..049221bc590 100644 --- a/drivers/infiniband/hw/ipath/ipath_cq.c +++ b/drivers/infiniband/hw/ipath/ipath_cq.c @@ -42,20 +42,28 @@ * @entry: work completion entry to add * @sig: true if @entry is a solicitated entry * - * This may be called with one of the qp->s_lock or qp->r_rq.lock held. + * This may be called with qp->s_lock held. */ void ipath_cq_enter(struct ipath_cq *cq, struct ib_wc *entry, int solicited) { + struct ipath_cq_wc *wc = cq->queue; unsigned long flags; + u32 head; u32 next; spin_lock_irqsave(&cq->lock, flags); - if (cq->head == cq->ibcq.cqe) + /* + * Note that the head pointer might be writable by user processes. + * Take care to verify it is a sane value. + */ + head = wc->head; + if (head >= (unsigned) cq->ibcq.cqe) { + head = cq->ibcq.cqe; next = 0; - else - next = cq->head + 1; - if (unlikely(next == cq->tail)) { + } else + next = head + 1; + if (unlikely(next == wc->tail)) { spin_unlock_irqrestore(&cq->lock, flags); if (cq->ibcq.event_handler) { struct ib_event ev; @@ -67,8 +75,8 @@ void ipath_cq_enter(struct ipath_cq *cq, struct ib_wc *entry, int solicited) } return; } - cq->queue[cq->head] = *entry; - cq->head = next; + wc->queue[head] = *entry; + wc->head = next; if (cq->notify == IB_CQ_NEXT_COMP || (cq->notify == IB_CQ_SOLICITED && solicited)) { @@ -101,19 +109,20 @@ void ipath_cq_enter(struct ipath_cq *cq, struct ib_wc *entry, int solicited) int ipath_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *entry) { struct ipath_cq *cq = to_icq(ibcq); + struct ipath_cq_wc *wc = cq->queue; unsigned long flags; int npolled; spin_lock_irqsave(&cq->lock, flags); for (npolled = 0; npolled < num_entries; ++npolled, ++entry) { - if (cq->tail == cq->head) + if (wc->tail == wc->head) break; - *entry = cq->queue[cq->tail]; - if (cq->tail == cq->ibcq.cqe) - cq->tail = 0; + *entry = wc->queue[wc->tail]; + if (wc->tail >= cq->ibcq.cqe) + wc->tail = 0; else - cq->tail++; + wc->tail++; } spin_unlock_irqrestore(&cq->lock, flags); @@ -160,38 +169,74 @@ struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries, { struct ipath_ibdev *dev = to_idev(ibdev); struct ipath_cq *cq; - struct ib_wc *wc; + struct ipath_cq_wc *wc; struct ib_cq *ret; - if (entries > ib_ipath_max_cqes) { + if (entries < 1 || entries > ib_ipath_max_cqes) { ret = ERR_PTR(-EINVAL); - goto bail; + goto done; } if (dev->n_cqs_allocated == ib_ipath_max_cqs) { ret = ERR_PTR(-ENOMEM); - goto bail; + goto done; } - /* - * Need to use vmalloc() if we want to support large #s of - * entries. - */ + /* Allocate the completion queue structure. */ cq = kmalloc(sizeof(*cq), GFP_KERNEL); if (!cq) { ret = ERR_PTR(-ENOMEM); - goto bail; + goto done; } /* - * Need to use vmalloc() if we want to support large #s of entries. + * Allocate the completion queue entries and head/tail pointers. + * This is allocated separately so that it can be resized and + * also mapped into user space. + * We need to use vmalloc() in order to support mmap and large + * numbers of entries. */ - wc = vmalloc(sizeof(*wc) * (entries + 1)); + wc = vmalloc_user(sizeof(*wc) + sizeof(struct ib_wc) * entries); if (!wc) { - kfree(cq); ret = ERR_PTR(-ENOMEM); - goto bail; + goto bail_cq; } + + /* + * Return the address of the WC as the offset to mmap. + * See ipath_mmap() for details. + */ + if (udata && udata->outlen >= sizeof(__u64)) { + struct ipath_mmap_info *ip; + __u64 offset = (__u64) wc; + int err; + + err = ib_copy_to_udata(udata, &offset, sizeof(offset)); + if (err) { + ret = ERR_PTR(err); + goto bail_wc; + } + + /* Allocate info for ipath_mmap(). */ + ip = kmalloc(sizeof(*ip), GFP_KERNEL); + if (!ip) { + ret = ERR_PTR(-ENOMEM); + goto bail_wc; + } + cq->ip = ip; + ip->context = context; + ip->obj = wc; + kref_init(&ip->ref); + ip->mmap_cnt = 0; + ip->size = PAGE_ALIGN(sizeof(*wc) + + sizeof(struct ib_wc) * entries); + spin_lock_irq(&dev->pending_lock); + ip->next = dev->pending_mmaps; + dev->pending_mmaps = ip; + spin_unlock_irq(&dev->pending_lock); + } else + cq->ip = NULL; + /* * ib_create_cq() will initialize cq->ibcq except for cq->ibcq.cqe. * The number of entries should be >= the number requested or return @@ -202,15 +247,22 @@ struct ib_cq *ipath_create_cq(struct ib_device *ibdev, int entries, cq->triggered = 0; spin_lock_init(&cq->lock); tasklet_init(&cq->comptask, send_complete, (unsigned long)cq); - cq->head = 0; - cq->tail = 0; + wc->head = 0; + wc->tail = 0; cq->queue = wc; ret = &cq->ibcq; dev->n_cqs_allocated++; + goto done; -bail: +bail_wc: + vfree(wc); + +bail_cq: + kfree(cq); + +done: return ret; } @@ -229,7 +281,10 @@ int ipath_destroy_cq(struct ib_cq *ibcq) tasklet_kill(&cq->comptask); dev->n_cqs_allocated--; - vfree(cq->queue); + if (cq->ip) + kref_put(&cq->ip->ref, ipath_release_mmap_info); + else + vfree(cq->queue); kfree(cq); return 0; @@ -253,7 +308,7 @@ int ipath_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify notify) spin_lock_irqsave(&cq->lock, flags); /* * Don't change IB_CQ_NEXT_COMP to IB_CQ_SOLICITED but allow - * any other transitions. + * any other transitions (see C11-31 and C11-32 in ch. 11.4.2.2). */ if (cq->notify != IB_CQ_NEXT_COMP) cq->notify = notify; @@ -264,46 +319,86 @@ int ipath_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify notify) int ipath_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata) { struct ipath_cq *cq = to_icq(ibcq); - struct ib_wc *wc, *old_wc; - u32 n; + struct ipath_cq_wc *old_wc = cq->queue; + struct ipath_cq_wc *wc; + u32 head, tail, n; int ret; + if (cqe < 1 || cqe > ib_ipath_max_cqes) { + ret = -EINVAL; + goto bail; + } + /* * Need to use vmalloc() if we want to support large #s of entries. */ - wc = vmalloc(sizeof(*wc) * (cqe + 1)); + wc = vmalloc_user(sizeof(*wc) + sizeof(struct ib_wc) * cqe); if (!wc) { ret = -ENOMEM; goto bail; } + /* + * Return the address of the WC as the offset to mmap. + * See ipath_mmap() for details. + */ + if (udata && udata->outlen >= sizeof(__u64)) { + __u64 offset = (__u64) wc; + + ret = ib_copy_to_udata(udata, &offset, sizeof(offset)); + if (ret) + goto bail; + } + spin_lock_irq(&cq->lock); - if (cq->head < cq->tail) - n = cq->ibcq.cqe + 1 + cq->head - cq->tail; + /* + * Make sure head and tail are sane since they + * might be user writable. + */ + head = old_wc->head; + if (head > (u32) cq->ibcq.cqe) + head = (u32) cq->ibcq.cqe; + tail = old_wc->tail; + if (tail > (u32) cq->ibcq.cqe) + tail = (u32) cq->ibcq.cqe; + if (head < tail) + n = cq->ibcq.cqe + 1 + head - tail; else - n = cq->head - cq->tail; + n = head - tail; if (unlikely((u32)cqe < n)) { spin_unlock_irq(&cq->lock); vfree(wc); ret = -EOVERFLOW; goto bail; } - for (n = 0; cq->tail != cq->head; n++) { - wc[n] = cq->queue[cq->tail]; - if (cq->tail == cq->ibcq.cqe) - cq->tail = 0; + for (n = 0; tail != head; n++) { + wc->queue[n] = old_wc->queue[tail]; + if (tail == (u32) cq->ibcq.cqe) + tail = 0; else - cq->tail++; + tail++; } cq->ibcq.cqe = cqe; - cq->head = n; - cq->tail = 0; - old_wc = cq->queue; + wc->head = n; + wc->tail = 0; cq->queue = wc; spin_unlock_irq(&cq->lock); vfree(old_wc); + if (cq->ip) { + struct ipath_ibdev *dev = to_idev(ibcq->device); + struct ipath_mmap_info *ip = cq->ip; + + ip->obj = wc; + ip->size = PAGE_ALIGN(sizeof(*wc) + + sizeof(struct ib_wc) * cqe); + spin_lock_irq(&dev->pending_lock); + ip->next = dev->pending_mmaps; + dev->pending_mmaps = ip; + spin_unlock_irq(&dev->pending_lock); + } + ret = 0; bail: diff --git a/drivers/infiniband/hw/ipath/ipath_debug.h b/drivers/infiniband/hw/ipath/ipath_debug.h index f415beda0d3..df69f0d80b8 100644 --- a/drivers/infiniband/hw/ipath/ipath_debug.h +++ b/drivers/infiniband/hw/ipath/ipath_debug.h @@ -60,7 +60,6 @@ #define __IPATH_USER_SEND 0x1000 /* use user mode send */ #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) gen debug */ #define __IPATH_IPATHWARN 0x20000 /* Ethernet (IPATH) warnings */ #define __IPATH_IPATHERR 0x40000 /* Ethernet (IPATH) errors */ @@ -84,7 +83,6 @@ /* 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 /* 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 */ diff --git a/drivers/infiniband/hw/ipath/ipath_diag.c b/drivers/infiniband/hw/ipath/ipath_diag.c index 147dd89e21c..29958b6e021 100644 --- a/drivers/infiniband/hw/ipath/ipath_diag.c +++ b/drivers/infiniband/hw/ipath/ipath_diag.c @@ -41,11 +41,12 @@ * through the /sys/bus/pci resource mmap interface. */ +#include <linux/io.h> #include <linux/pci.h> +#include <linux/vmalloc.h> #include <asm/uaccess.h> #include "ipath_kernel.h" -#include "ipath_layer.h" #include "ipath_common.h" int ipath_diag_inuse; @@ -274,6 +275,158 @@ bail: return ret; } +static ssize_t ipath_diagpkt_write(struct file *fp, + const char __user *data, + size_t count, loff_t *off); + +static struct file_operations diagpkt_file_ops = { + .owner = THIS_MODULE, + .write = ipath_diagpkt_write, +}; + +static struct cdev *diagpkt_cdev; +static struct class_device *diagpkt_class_dev; + +int __init ipath_diagpkt_add(void) +{ + return ipath_cdev_init(IPATH_DIAGPKT_MINOR, + "ipath_diagpkt", &diagpkt_file_ops, + &diagpkt_cdev, &diagpkt_class_dev); +} + +void __exit ipath_diagpkt_remove(void) +{ + ipath_cdev_cleanup(&diagpkt_cdev, &diagpkt_class_dev); +} + +/** + * ipath_diagpkt_write - write an IB packet + * @fp: the diag data device file pointer + * @data: ipath_diag_pkt structure saying where to get the packet + * @count: size of data to write + * @off: unused by this code + */ +static ssize_t ipath_diagpkt_write(struct file *fp, + const char __user *data, + size_t count, loff_t *off) +{ + u32 __iomem *piobuf; + u32 plen, clen, pbufn; + struct ipath_diag_pkt dp; + u32 *tmpbuf = NULL; + struct ipath_devdata *dd; + ssize_t ret = 0; + u64 val; + + if (count < sizeof(dp)) { + ret = -EINVAL; + goto bail; + } + + if (copy_from_user(&dp, data, sizeof(dp))) { + ret = -EFAULT; + goto bail; + } + + /* send count must be an exact number of dwords */ + if (dp.len & 3) { + ret = -EINVAL; + goto bail; + } + + clen = dp.len >> 2; + + dd = ipath_lookup(dp.unit); + if (!dd || !(dd->ipath_flags & IPATH_PRESENT) || + !dd->ipath_kregbase) { + ipath_cdbg(VERBOSE, "illegal unit %u for diag data send\n", + dp.unit); + ret = -ENODEV; + goto bail; + } + + if (ipath_diag_inuse && !diag_set_link && + !(dd->ipath_flags & IPATH_LINKACTIVE)) { + diag_set_link = 1; + ipath_cdbg(VERBOSE, "Trying to set to set link active for " + "diag pkt\n"); + ipath_set_linkstate(dd, IPATH_IB_LINKARM); + ipath_set_linkstate(dd, IPATH_IB_LINKACTIVE); + } + + if (!(dd->ipath_flags & IPATH_INITTED)) { + /* no hardware, freeze, etc. */ + ipath_cdbg(VERBOSE, "unit %u not usable\n", dd->ipath_unit); + ret = -ENODEV; + goto bail; + } + val = dd->ipath_lastibcstat & IPATH_IBSTATE_MASK; + if (val != IPATH_IBSTATE_INIT && val != IPATH_IBSTATE_ARM && + val != IPATH_IBSTATE_ACTIVE) { + ipath_cdbg(VERBOSE, "unit %u not ready (state %llx)\n", + dd->ipath_unit, (unsigned long long) val); + ret = -EINVAL; + goto bail; + } + + /* need total length before first word written */ + /* +1 word is for the qword padding */ + plen = sizeof(u32) + dp.len; + + if ((plen + 4) > dd->ipath_ibmaxlen) { + ipath_dbg("Pkt len 0x%x > ibmaxlen %x\n", + plen - 4, dd->ipath_ibmaxlen); + ret = -EINVAL; + goto bail; /* before writing pbc */ + } + tmpbuf = vmalloc(plen); + if (!tmpbuf) { + dev_info(&dd->pcidev->dev, "Unable to allocate tmp buffer, " + "failing\n"); + ret = -ENOMEM; + goto bail; + } + + if (copy_from_user(tmpbuf, + (const void __user *) (unsigned long) dp.data, + dp.len)) { + ret = -EFAULT; + goto bail; + } + + piobuf = ipath_getpiobuf(dd, &pbufn); + if (!piobuf) { + ipath_cdbg(VERBOSE, "No PIO buffers avail unit for %u\n", + dd->ipath_unit); + ret = -EBUSY; + goto bail; + } + + plen >>= 2; /* in dwords */ + + if (ipath_debug & __IPATH_PKTDBG) + ipath_cdbg(VERBOSE, "unit %u 0x%x+1w pio%d\n", + dd->ipath_unit, plen - 1, pbufn); + + /* we have to flush after the PBC for correctness on some cpus + * or WC buffer can be written out of order */ + writeq(plen, piobuf); + ipath_flush_wc(); + /* copy all by the trigger word, then flush, so it's written + * to chip before trigger word, then write trigger word, then + * flush again, so packet is sent. */ + __iowrite32_copy(piobuf + 2, tmpbuf, clen - 1); + ipath_flush_wc(); + __raw_writel(tmpbuf[clen - 1], piobuf + clen + 1); + ipath_flush_wc(); + + ret = sizeof(dp); + +bail: + vfree(tmpbuf); + return ret; +} + static int ipath_diag_release(struct inode *in, struct file *fp) { mutex_lock(&ipath_mutex); diff --git a/drivers/infiniband/hw/ipath/ipath_driver.c b/drivers/infiniband/hw/ipath/ipath_driver.c index f98518d912b..2108466c7e3 100644 --- a/drivers/infiniband/hw/ipath/ipath_driver.c +++ b/drivers/infiniband/hw/ipath/ipath_driver.c @@ -39,7 +39,7 @@ #include <linux/vmalloc.h> #include "ipath_kernel.h" -#include "ipath_layer.h" +#include "ipath_verbs.h" #include "ipath_common.h" static void ipath_update_pio_bufs(struct ipath_devdata *); @@ -51,8 +51,6 @@ const char *ipath_get_unit_name(int unit) return iname; } -EXPORT_SYMBOL_GPL(ipath_get_unit_name); - #define DRIVER_LOAD_MSG "QLogic " IPATH_DRV_NAME " loaded: " #define PFX IPATH_DRV_NAME ": " @@ -60,13 +58,13 @@ EXPORT_SYMBOL_GPL(ipath_get_unit_name); * The size has to be longer than this string, so we can append * board/chip information to it in the init code. */ -const char ipath_core_version[] = IPATH_IDSTR "\n"; +const char ib_ipath_version[] = IPATH_IDSTR "\n"; static struct idr unit_table; DEFINE_SPINLOCK(ipath_devs_lock); LIST_HEAD(ipath_dev_list); -wait_queue_head_t ipath_sma_state_wait; +wait_queue_head_t ipath_state_wait; unsigned ipath_debug = __IPATH_INFO; @@ -403,10 +401,10 @@ static int __devinit ipath_init_one(struct pci_dev *pdev, /* setup the chip-specific functions, as early as possible. */ switch (ent->device) { case PCI_DEVICE_ID_INFINIPATH_HT: - ipath_init_ht400_funcs(dd); + ipath_init_iba6110_funcs(dd); break; case PCI_DEVICE_ID_INFINIPATH_PE800: - ipath_init_pe800_funcs(dd); + ipath_init_iba6120_funcs(dd); break; default: ipath_dev_err(dd, "Found unknown QLogic deviceid 0x%x, " @@ -440,7 +438,13 @@ static int __devinit ipath_init_one(struct pci_dev *pdev, } dd->ipath_pcirev = rev; +#if defined(__powerpc__) + /* There isn't a generic way to specify writethrough mappings */ + dd->ipath_kregbase = __ioremap(addr, len, + (_PAGE_NO_CACHE|_PAGE_WRITETHRU)); +#else dd->ipath_kregbase = ioremap_nocache(addr, len); +#endif if (!dd->ipath_kregbase) { ipath_dbg("Unable to map io addr %llx to kvirt, failing\n", @@ -503,7 +507,7 @@ static int __devinit ipath_init_one(struct pci_dev *pdev, ipathfs_add_device(dd); ipath_user_add(dd); ipath_diag_add(dd); - ipath_layer_add(dd); + ipath_register_ib_device(dd); goto bail; @@ -532,7 +536,7 @@ static void __devexit ipath_remove_one(struct pci_dev *pdev) return; dd = pci_get_drvdata(pdev); - ipath_layer_remove(dd); + ipath_unregister_ib_device(dd->verbs_dev); ipath_diag_remove(dd); ipath_user_remove(dd); ipathfs_remove_device(dd); @@ -607,21 +611,23 @@ void ipath_disarm_piobufs(struct ipath_devdata *dd, unsigned first, * * wait up to msecs milliseconds for IB link state change to occur for * now, take the easy polling route. Currently used only by - * ipath_layer_set_linkstate. Returns 0 if state reached, otherwise + * ipath_set_linkstate. Returns 0 if state reached, otherwise * -ETIMEDOUT state can have multiple states set, for any of several * transitions. */ -int ipath_wait_linkstate(struct ipath_devdata *dd, u32 state, int msecs) +static int ipath_wait_linkstate(struct ipath_devdata *dd, u32 state, + int msecs) { - dd->ipath_sma_state_wanted = state; - wait_event_interruptible_timeout(ipath_sma_state_wait, + dd->ipath_state_wanted = state; + wait_event_interruptible_timeout(ipath_state_wait, (dd->ipath_flags & state), msecs_to_jiffies(msecs)); - dd->ipath_sma_state_wanted = 0; + dd->ipath_state_wanted = 0; if (!(dd->ipath_flags & state)) { u64 val; - ipath_cdbg(SMA, "Didn't reach linkstate %s within %u ms\n", + ipath_cdbg(VERBOSE, "Didn't reach linkstate %s within %u" + " ms\n", /* test INIT ahead of DOWN, both can be set */ (state & IPATH_LINKINIT) ? "INIT" : ((state & IPATH_LINKDOWN) ? "DOWN" : @@ -807,58 +813,6 @@ bail: return skb; } -/** - * ipath_rcv_layer - receive a packet for the layered (ethernet) driver - * @dd: the infinipath device - * @etail: the sk_buff number - * @tlen: the total packet length - * @hdr: the ethernet header - * - * Separate routine for better overall optimization - */ -static void ipath_rcv_layer(struct ipath_devdata *dd, u32 etail, - u32 tlen, struct ether_header *hdr) -{ - u32 elen; - u8 pad, *bthbytes; - struct sk_buff *skb, *nskb; - - if (dd->ipath_port0_skbs && - hdr->sub_opcode == IPATH_ITH4X_OPCODE_ENCAP) { - /* - * Allocate a new sk_buff to replace the one we give - * to the network stack. - */ - nskb = ipath_alloc_skb(dd, GFP_ATOMIC); - if (!nskb) { - /* count OK packets that we drop */ - ipath_stats.sps_krdrops++; - return; - } - - bthbytes = (u8 *) hdr->bth; - pad = (bthbytes[1] >> 4) & 3; - /* +CRC32 */ - elen = tlen - (sizeof(*hdr) + pad + sizeof(u32)); - - skb = dd->ipath_port0_skbs[etail]; - dd->ipath_port0_skbs[etail] = nskb; - skb_put(skb, elen); - - dd->ipath_f_put_tid(dd, etail + (u64 __iomem *) - ((char __iomem *) dd->ipath_kregbase - + dd->ipath_rcvegrbase), 0, - virt_to_phys(nskb->data)); - - __ipath_layer_rcv(dd, hdr, skb); - - /* another ether packet received */ - ipath_stats.sps_ether_rpkts++; - } - else if (hdr->sub_opcode == IPATH_ITH4X_OPCODE_LID_ARP) - __ipath_layer_rcv_lid(dd, hdr); -} - static void ipath_rcv_hdrerr(struct ipath_devdata *dd, u32 eflags, u32 l, @@ -972,26 +926,17 @@ reloop: if (unlikely(eflags)) ipath_rcv_hdrerr(dd, eflags, l, etail, rc); else if (etype == RCVHQ_RCV_TYPE_NON_KD) { - int ret = __ipath_verbs_rcv(dd, rc + 1, - ebuf, tlen); - if (ret == -ENODEV) - ipath_cdbg(VERBOSE, - "received IB packet, " - "not SMA (QP=%x)\n", qp); - if (dd->ipath_lli_counter) - dd->ipath_lli_counter--; - - } else if (etype == RCVHQ_RCV_TYPE_EAGER) { - if (qp == IPATH_KD_QP && - bthbytes[0] == ipath_layer_rcv_opcode && - ebuf) - ipath_rcv_layer(dd, etail, tlen, - (struct ether_header *)hdr); - else - ipath_cdbg(PKT, "typ %x, opcode %x (eager, " - "qp=%x), len %x; ignored\n", - etype, bthbytes[0], qp, tlen); + ipath_ib_rcv(dd->verbs_dev, rc + 1, ebuf, tlen); + if (dd->ipath_lli_counter) + dd->ipath_lli_counter--; + ipath_cdbg(PKT, "typ %x, opcode %x (eager, " + "qp=%x), len %x; ignored\n", + etype, bthbytes[0], qp, tlen); } + else if (etype == RCVHQ_RCV_TYPE_EAGER) + ipath_cdbg(PKT, "typ %x, opcode %x (eager, " + "qp=%x), len %x; ignored\n", + etype, bthbytes[0], qp, tlen); else if (etype == RCVHQ_RCV_TYPE_EXPECTED) ipath_dbg("Bug: Expected TID, opcode %x; ignored\n", be32_to_cpu(hdr->bth[0]) & 0xff); @@ -1024,7 +969,8 @@ reloop: */ if (l == hdrqtail || (i && !(i&0xf))) { |