diff options
author | Steve Wise <swise@opengridcomputing.com> | 2007-02-12 16:16:18 -0800 |
---|---|---|
committer | Roland Dreier <rolandd@cisco.com> | 2007-02-12 16:16:18 -0800 |
commit | b038ced7b3705bf0ac9b30e118af0f56ab48b847 (patch) | |
tree | 0e293376fd164c187dbe7f0a320b6f14b62f5958 /drivers/infiniband/hw/cxgb3/iwch_provider.c | |
parent | c7f743a669c27f9c392e78fda8829db9d6d50f43 (diff) | |
download | blackbird-op-linux-b038ced7b3705bf0ac9b30e118af0f56ab48b847.tar.gz blackbird-op-linux-b038ced7b3705bf0ac9b30e118af0f56ab48b847.zip |
RDMA/cxgb3: Add driver for Chelsio T3 RNIC
Add an RDMA/iWARP driver for the Chelsio T3 1GbE and 10GbE adapters.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Diffstat (limited to 'drivers/infiniband/hw/cxgb3/iwch_provider.c')
-rw-r--r-- | drivers/infiniband/hw/cxgb3/iwch_provider.c | 1203 |
1 files changed, 1203 insertions, 0 deletions
diff --git a/drivers/infiniband/hw/cxgb3/iwch_provider.c b/drivers/infiniband/hw/cxgb3/iwch_provider.c new file mode 100644 index 000000000000..6861087d776c --- /dev/null +++ b/drivers/infiniband/hw/cxgb3/iwch_provider.c @@ -0,0 +1,1203 @@ +/* + * Copyright (c) 2006 Chelsio, Inc. All rights reserved. + * Copyright (c) 2006 Open Grid Computing, Inc. All rights reserved. + * + * This software is available to you under a choice of one of two + * licenses. You may choose to be licensed under the terms of the GNU + * General Public License (GPL) Version 2, available from the file + * COPYING in the main directory of this source tree, or the + * OpenIB.org BSD license below: + * + * Redistribution and use in source and binary forms, with or + * without modification, are permitted provided that the following + * conditions are met: + * + * - Redistributions of source code must retain the above + * copyright notice, this list of conditions and the following + * disclaimer. + * + * - Redistributions in binary form must reproduce the above + * copyright notice, this list of conditions and the following + * disclaimer in the documentation and/or other materials + * provided with the distribution. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND + * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS + * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN + * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ +#include <linux/module.h> +#include <linux/moduleparam.h> +#include <linux/device.h> +#include <linux/netdevice.h> +#include <linux/etherdevice.h> +#include <linux/delay.h> +#include <linux/errno.h> +#include <linux/list.h> +#include <linux/spinlock.h> +#include <linux/ethtool.h> + +#include <asm/io.h> +#include <asm/irq.h> +#include <asm/byteorder.h> + +#include <rdma/iw_cm.h> +#include <rdma/ib_verbs.h> +#include <rdma/ib_smi.h> +#include <rdma/ib_user_verbs.h> + +#include "cxio_hal.h" +#include "iwch.h" +#include "iwch_provider.h" +#include "iwch_cm.h" +#include "iwch_user.h" + +static int iwch_modify_port(struct ib_device *ibdev, + u8 port, int port_modify_mask, + struct ib_port_modify *props) +{ + return -ENOSYS; +} + +static struct ib_ah *iwch_ah_create(struct ib_pd *pd, + struct ib_ah_attr *ah_attr) +{ + return ERR_PTR(-ENOSYS); +} + +static int iwch_ah_destroy(struct ib_ah *ah) +{ + return -ENOSYS; +} + +static int iwch_multicast_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid) +{ + return -ENOSYS; +} + +static int iwch_multicast_detach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid) +{ + return -ENOSYS; +} + +static int iwch_process_mad(struct ib_device *ibdev, + int mad_flags, + u8 port_num, + struct ib_wc *in_wc, + struct ib_grh *in_grh, + struct ib_mad *in_mad, struct ib_mad *out_mad) +{ + return -ENOSYS; +} + +static int iwch_dealloc_ucontext(struct ib_ucontext *context) +{ + struct iwch_dev *rhp = to_iwch_dev(context->device); + struct iwch_ucontext *ucontext = to_iwch_ucontext(context); + struct iwch_mm_entry *mm, *tmp; + + PDBG("%s context %p\n", __FUNCTION__, context); + list_for_each_entry_safe(mm, tmp, &ucontext->mmaps, entry) + kfree(mm); + cxio_release_ucontext(&rhp->rdev, &ucontext->uctx); + kfree(ucontext); + return 0; +} + +static struct ib_ucontext *iwch_alloc_ucontext(struct ib_device *ibdev, + struct ib_udata *udata) +{ + struct iwch_ucontext *context; + struct iwch_dev *rhp = to_iwch_dev(ibdev); + + PDBG("%s ibdev %p\n", __FUNCTION__, ibdev); + context = kzalloc(sizeof(*context), GFP_KERNEL); + if (!context) + return ERR_PTR(-ENOMEM); + cxio_init_ucontext(&rhp->rdev, &context->uctx); + INIT_LIST_HEAD(&context->mmaps); + spin_lock_init(&context->mmap_lock); + return &context->ibucontext; +} + +static int iwch_destroy_cq(struct ib_cq *ib_cq) +{ + struct iwch_cq *chp; + + PDBG("%s ib_cq %p\n", __FUNCTION__, ib_cq); + chp = to_iwch_cq(ib_cq); + + remove_handle(chp->rhp, &chp->rhp->cqidr, chp->cq.cqid); + atomic_dec(&chp->refcnt); + wait_event(chp->wait, !atomic_read(&chp->refcnt)); + + cxio_destroy_cq(&chp->rhp->rdev, &chp->cq); + kfree(chp); + return 0; +} + +static struct ib_cq *iwch_create_cq(struct ib_device *ibdev, int entries, + struct ib_ucontext *ib_context, + struct ib_udata *udata) +{ + struct iwch_dev *rhp; + struct iwch_cq *chp; + struct iwch_create_cq_resp uresp; + struct iwch_create_cq_req ureq; + struct iwch_ucontext *ucontext = NULL; + + PDBG("%s ib_dev %p entries %d\n", __FUNCTION__, ibdev, entries); + rhp = to_iwch_dev(ibdev); + chp = kzalloc(sizeof(*chp), GFP_KERNEL); + if (!chp) + return ERR_PTR(-ENOMEM); + + if (ib_context) { + ucontext = to_iwch_ucontext(ib_context); + if (!t3a_device(rhp)) { + if (ib_copy_from_udata(&ureq, udata, sizeof (ureq))) { + kfree(chp); + return ERR_PTR(-EFAULT); + } + chp->user_rptr_addr = (u32 __user *)(unsigned long)ureq.user_rptr_addr; + } + } + + if (t3a_device(rhp)) { + + /* + * T3A: Add some fluff to handle extra CQEs inserted + * for various errors. + * Additional CQE possibilities: + * TERMINATE, + * incoming RDMA WRITE Failures + * incoming RDMA READ REQUEST FAILUREs + * NOTE: We cannot ensure the CQ won't overflow. + */ + entries += 16; + } + entries = roundup_pow_of_two(entries); + chp->cq.size_log2 = ilog2(entries); + + if (cxio_create_cq(&rhp->rdev, &chp->cq)) { + kfree(chp); + return ERR_PTR(-ENOMEM); + } + chp->rhp = rhp; + chp->ibcq.cqe = (1 << chp->cq.size_log2) - 1; + spin_lock_init(&chp->lock); + atomic_set(&chp->refcnt, 1); + init_waitqueue_head(&chp->wait); + insert_handle(rhp, &rhp->cqidr, chp, chp->cq.cqid); + + if (ucontext) { + struct iwch_mm_entry *mm; + + mm = kmalloc(sizeof *mm, GFP_KERNEL); + if (!mm) { + iwch_destroy_cq(&chp->ibcq); + return ERR_PTR(-ENOMEM); + } + uresp.cqid = chp->cq.cqid; + uresp.size_log2 = chp->cq.size_log2; + spin_lock(&ucontext->mmap_lock); + uresp.key = ucontext->key; + ucontext->key += PAGE_SIZE; + spin_unlock(&ucontext->mmap_lock); + if (ib_copy_to_udata(udata, &uresp, sizeof (uresp))) { + kfree(mm); + iwch_destroy_cq(&chp->ibcq); + return ERR_PTR(-EFAULT); + } + mm->key = uresp.key; + mm->addr = virt_to_phys(chp->cq.queue); + mm->len = PAGE_ALIGN((1UL << uresp.size_log2) * + sizeof (struct t3_cqe)); + insert_mmap(ucontext, mm); + } + PDBG("created cqid 0x%0x chp %p size 0x%0x, dma_addr 0x%0llx\n", + chp->cq.cqid, chp, (1 << chp->cq.size_log2), + (unsigned long long) chp->cq.dma_addr); + return &chp->ibcq; +} + +static int iwch_resize_cq(struct ib_cq *cq, int cqe, struct ib_udata *udata) +{ +#ifdef notyet + struct iwch_cq *chp = to_iwch_cq(cq); + struct t3_cq oldcq, newcq; + int ret; + + PDBG("%s ib_cq %p cqe %d\n", __FUNCTION__, cq, cqe); + + /* We don't downsize... */ + if (cqe <= cq->cqe) + return 0; + + /* create new t3_cq with new size */ + cqe = roundup_pow_of_two(cqe+1); + newcq.size_log2 = ilog2(cqe); + + /* Dont allow resize to less than the current wce count */ + if (cqe < Q_COUNT(chp->cq.rptr, chp->cq.wptr)) { + return -ENOMEM; + } + + /* Quiesce all QPs using this CQ */ + ret = iwch_quiesce_qps(chp); + if (ret) { + return ret; + } + + ret = cxio_create_cq(&chp->rhp->rdev, &newcq); + if (ret) { + return ret; + } + + /* copy CQEs */ + memcpy(newcq.queue, chp->cq.queue, (1 << chp->cq.size_log2) * + sizeof(struct t3_cqe)); + + /* old iwch_qp gets new t3_cq but keeps old cqid */ + oldcq = chp->cq; + chp->cq = newcq; + chp->cq.cqid = oldcq.cqid; + + /* resize new t3_cq to update the HW context */ + ret = cxio_resize_cq(&chp->rhp->rdev, &chp->cq); + if (ret) { + chp->cq = oldcq; + return ret; + } + chp->ibcq.cqe = (1<<chp->cq.size_log2) - 1; + + /* destroy old t3_cq */ + oldcq.cqid = newcq.cqid; + ret = cxio_destroy_cq(&chp->rhp->rdev, &oldcq); + if (ret) { + printk(KERN_ERR MOD "%s - cxio_destroy_cq failed %d\n", + __FUNCTION__, ret); + } + + /* add user hooks here */ + + /* resume qps */ + ret = iwch_resume_qps(chp); + return ret; +#else + return -ENOSYS; +#endif +} + +static int iwch_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify notify) +{ + struct iwch_dev *rhp; + struct iwch_cq *chp; + enum t3_cq_opcode cq_op; + int err; + unsigned long flag; + u32 rptr; + + chp = to_iwch_cq(ibcq); + rhp = chp->rhp; + if (notify == IB_CQ_SOLICITED) + cq_op = CQ_ARM_SE; + else + cq_op = CQ_ARM_AN; + if (chp->user_rptr_addr) { + if (get_user(rptr, chp->user_rptr_addr)) + return -EFAULT; + spin_lock_irqsave(&chp->lock, flag); + chp->cq.rptr = rptr; + } else + spin_lock_irqsave(&chp->lock, flag); + PDBG("%s rptr 0x%x\n", __FUNCTION__, chp->cq.rptr); + err = cxio_hal_cq_op(&rhp->rdev, &chp->cq, cq_op, 0); + spin_unlock_irqrestore(&chp->lock, flag); + if (err) + printk(KERN_ERR MOD "Error %d rearming CQID 0x%x\n", err, + chp->cq.cqid); + return err; +} + +static int iwch_mmap(struct ib_ucontext *context, struct vm_area_struct *vma) +{ + int len = vma->vm_end - vma->vm_start; + u32 key = vma->vm_pgoff << PAGE_SHIFT; + struct cxio_rdev *rdev_p; + int ret = 0; + struct iwch_mm_entry *mm; + struct iwch_ucontext *ucontext; + + PDBG("%s pgoff 0x%lx key 0x%x len %d\n", __FUNCTION__, vma->vm_pgoff, + key, len); + + if (vma->vm_start & (PAGE_SIZE-1)) { + return -EINVAL; + } + + rdev_p = &(to_iwch_dev(context->device)->rdev); + ucontext = to_iwch_ucontext(context); + + mm = remove_mmap(ucontext, key, len); + if (!mm) + return -EINVAL; + kfree(mm); + + if ((mm->addr >= rdev_p->rnic_info.udbell_physbase) && + (mm->addr < (rdev_p->rnic_info.udbell_physbase + + rdev_p->rnic_info.udbell_len))) { + + /* + * Map T3 DB register. + */ + if (vma->vm_flags & VM_READ) { + return -EPERM; + } + + vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot); + vma->vm_flags |= VM_DONTCOPY | VM_DONTEXPAND; + vma->vm_flags &= ~VM_MAYREAD; + ret = io_remap_pfn_range(vma, vma->vm_start, + mm->addr >> PAGE_SHIFT, + len, vma->vm_page_prot); + } else { + + /* + * Map WQ or CQ contig dma memory... + */ + ret = remap_pfn_range(vma, vma->vm_start, + mm->addr >> PAGE_SHIFT, + len, vma->vm_page_prot); + } + + return ret; +} + +static int iwch_deallocate_pd(struct ib_pd *pd) +{ + struct iwch_dev *rhp; + struct iwch_pd *php; + + php = to_iwch_pd(pd); + rhp = php->rhp; + PDBG("%s ibpd %p pdid 0x%x\n", __FUNCTION__, pd, php->pdid); + cxio_hal_put_pdid(rhp->rdev.rscp, php->pdid); + kfree(php); + return 0; +} + +static struct ib_pd *iwch_allocate_pd(struct ib_device *ibdev, + struct ib_ucontext *context, + struct ib_udata *udata) +{ + struct iwch_pd *php; + u32 pdid; + struct iwch_dev *rhp; + + PDBG("%s ibdev %p\n", __FUNCTION__, ibdev); + rhp = (struct iwch_dev *) ibdev; + pdid = cxio_hal_get_pdid(rhp->rdev.rscp); + if (!pdid) + return ERR_PTR(-EINVAL); + php = kzalloc(sizeof(*php), GFP_KERNEL); + if (!php) { + cxio_hal_put_pdid(rhp->rdev.rscp, pdid); + return ERR_PTR(-ENOMEM); + } + php->pdid = pdid; + php->rhp = rhp; + if (context) { + if (ib_copy_to_udata(udata, &php->pdid, sizeof (__u32))) { + iwch_deallocate_pd(&php->ibpd); + return ERR_PTR(-EFAULT); + } + } + PDBG("%s pdid 0x%0x ptr 0x%p\n", __FUNCTION__, pdid, php); + return &php->ibpd; +} + +static int iwch_dereg_mr(struct ib_mr *ib_mr) +{ + struct iwch_dev *rhp; + struct iwch_mr *mhp; + u32 mmid; + + PDBG("%s ib_mr %p\n", __FUNCTION__, ib_mr); + /* There can be no memory windows */ + if (atomic_read(&ib_mr->usecnt)) + return -EINVAL; + + mhp = to_iwch_mr(ib_mr); + rhp = mhp->rhp; + mmid = mhp->attr.stag >> 8; + cxio_dereg_mem(&rhp->rdev, mhp->attr.stag, mhp->attr.pbl_size, + mhp->attr.pbl_addr); + remove_handle(rhp, &rhp->mmidr, mmid); + if (mhp->kva) + kfree((void *) (unsigned long) mhp->kva); + PDBG("%s mmid 0x%x ptr %p\n", __FUNCTION__, mmid, mhp); + kfree(mhp); + return 0; +} + +static struct ib_mr *iwch_register_phys_mem(struct ib_pd *pd, + struct ib_phys_buf *buffer_list, + int num_phys_buf, + int acc, + u64 *iova_start) +{ + __be64 *page_list; + int shift; + u64 total_size; + int npages; + struct iwch_dev *rhp; + struct iwch_pd *php; + struct iwch_mr *mhp; + int ret; + + PDBG("%s ib_pd %p\n", __FUNCTION__, pd); + php = to_iwch_pd(pd); + rhp = php->rhp; + + acc = iwch_convert_access(acc); + + + mhp = kzalloc(sizeof(*mhp), GFP_KERNEL); + if (!mhp) + return ERR_PTR(-ENOMEM); + + /* First check that we have enough alignment */ + if ((*iova_start & ~PAGE_MASK) != (buffer_list[0].addr & ~PAGE_MASK)) { + ret = -EINVAL; + goto err; + } + + if (num_phys_buf > 1 && + ((buffer_list[0].addr + buffer_list[0].size) & ~PAGE_MASK)) { + ret = -EINVAL; + goto err; + } + + ret = build_phys_page_list(buffer_list, num_phys_buf, iova_start, + &total_size, &npages, &shift, &page_list); + if (ret) + goto err; + + mhp->rhp = rhp; + mhp->attr.pdid = php->pdid; + mhp->attr.zbva = 0; + + /* NOTE: TPT perms are backwards from BIND WR perms! */ + mhp->attr.perms = (acc & 0x1) << 3; + mhp->attr.perms |= (acc & 0x2) << 1; + mhp->attr.perms |= (acc & 0x4) >> 1; + mhp->attr.perms |= (acc & 0x8) >> 3; + + mhp->attr.va_fbo = *iova_start; + mhp->attr.page_size = shift - 12; + + mhp->attr.len = (u32) total_size; + mhp->attr.pbl_size = npages; + ret = iwch_register_mem(rhp, php, mhp, shift, page_list); + kfree(page_list); + if (ret) { + goto err; + } + return &mhp->ibmr; +err: + kfree(mhp); + return ERR_PTR(ret); + +} + +static int iwch_reregister_phys_mem(struct ib_mr *mr, + int mr_rereg_mask, + struct ib_pd *pd, + struct ib_phys_buf *buffer_list, + int num_phys_buf, + int acc, u64 * iova_start) +{ + + struct iwch_mr mh, *mhp; + struct iwch_pd *php; + struct iwch_dev *rhp; + int new_acc; + __be64 *page_list = NULL; + int shift = 0; + u64 total_size; + int npages; + int ret; + + PDBG("%s ib_mr %p ib_pd %p\n", __FUNCTION__, mr, pd); + + /* There can be no memory windows */ + if (atomic_read(&mr->usecnt)) + return -EINVAL; + + mhp = to_iwch_mr(mr); + rhp = mhp->rhp; + php = to_iwch_pd(mr->pd); + + /* make sure we are on the same adapter */ + if (rhp != php->rhp) + return -EINVAL; + + new_acc = mhp->attr.perms; + + memcpy(&mh, mhp, sizeof *mhp); + + if (mr_rereg_mask & IB_MR_REREG_PD) + php = to_iwch_pd(pd); + if (mr_rereg_mask & IB_MR_REREG_ACCESS) + mh.attr.perms = iwch_convert_access(acc); + if (mr_rereg_mask & IB_MR_REREG_TRANS) + ret = build_phys_page_list(buffer_list, num_phys_buf, + iova_start, + &total_size, &npages, + &shift, &page_list); + + ret = iwch_reregister_mem(rhp, php, &mh, shift, page_list, npages); + kfree(page_list); + if (ret) { + return ret; + } + if (mr_rereg_mask & IB_MR_REREG_PD) + mhp->attr.pdid = php->pdid; + if (mr_rereg_mask & IB_MR_REREG_ACCESS) + mhp->attr.perms = acc; + if (mr_rereg_mask & IB_MR_REREG_TRANS) { + mhp->attr.zbva = 0; + mhp->attr.va_fbo = *iova_start; + mhp->attr.page_size = shift - 12; + mhp->attr.len = (u32) total_size; + mhp->attr.pbl_size = npages; + } + + return 0; +} + + +static struct ib_mr *iwch_reg_user_mr(struct ib_pd *pd, struct ib_umem *region, + int acc, struct ib_udata *udata) +{ + __be64 *pages; + int shift, n, len; + int i, j, k; + int err = 0; + struct ib_umem_chunk *chunk; + struct iwch_dev *rhp; + struct iwch_pd *php; + struct iwch_mr *mhp; + struct iwch_reg_user_mr_resp uresp; + + PDBG("%s ib_pd %p\n", __FUNCTION__, pd); + shift = ffs(region->page_size) - 1; + + php = to_iwch_pd(pd); + rhp = php->rhp; + mhp = kzalloc(sizeof(*mhp), GFP_KERNEL); + if (!mhp) + return ERR_PTR(-ENOMEM); + + n = 0; + list_for_each_entry(chunk, ®ion->chunk_list, list) + n += chunk->nents; + + pages = kmalloc(n * sizeof(u64), GFP_KERNEL); + if (!pages) { + err = -ENOMEM; + goto err; + } + + acc = iwch_convert_access(acc); + + i = n = 0; + + list_for_each_entry(chunk, ®ion->chunk_list, list) + for (j = 0; j < chunk->nmap; ++j) { + len = sg_dma_len(&chunk->page_list[j]) >> shift; + for (k = 0; k < len; ++k) { + pages[i++] = cpu_to_be64(sg_dma_address( + &chunk->page_list[j]) + + region->page_size * k); + } + } + + mhp->rhp = rhp; + mhp->attr.pdid = php->pdid; + mhp->attr.zbva = 0; + mhp->attr.perms = (acc & 0x1) << 3; + mhp->attr.perms |= (acc & 0x2) << 1; + mhp->attr.perms |= (acc & 0x4) >> 1; + mhp->attr.perms |= (acc & 0x8) >> 3; + mhp->attr.va_fbo = region->virt_base; + mhp->attr.page_size = shift - 12; + mhp->attr.len = (u32) region->length; + mhp->attr.pbl_size = i; + err = iwch_register_mem(rhp, php, mhp, shift, pages); + kfree(pages); + if (err) + goto err; + + if (udata && t3b_device(rhp)) { + uresp.pbl_addr = (mhp->attr.pbl_addr - + rhp->rdev.rnic_info.pbl_base) >> 3; + PDBG("%s user resp pbl_addr 0x%x\n", __FUNCTION__, + uresp.pbl_addr); + + if (ib_copy_to_udata(udata, &uresp, sizeof (uresp))) { + iwch_dereg_mr(&mhp->ibmr); + err = -EFAULT; + goto err; + } + } + + return &mhp->ibmr; + +err: + kfree(mhp); + return ERR_PTR(err); +} + +static struct ib_mr *iwch_get_dma_mr(struct ib_pd *pd, int acc) +{ + struct ib_phys_buf bl; + u64 kva; + struct ib_mr *ibmr; + + PDBG("%s ib_pd %p\n", __FUNCTION__, pd); + + /* + * T3 only supports 32 bits of size. + */ + bl.size = 0xffffffff; + bl.addr = 0; + kva = 0; + ibmr = iwch_register_phys_mem(pd, &bl, 1, acc, &kva); + return ibmr; +} + +static struct ib_mw *iwch_alloc_mw(struct ib_pd *pd) +{ + struct iwch_dev *rhp; + struct iwch_pd *php; + struct iwch_mw *mhp; + u32 mmid; + u32 stag = 0; + int ret; + + php = to_iwch_pd(pd); + rhp = php->rhp; + mhp = kzalloc(sizeof(*mhp), GFP_KERNEL); + if (!mhp) + return ERR_PTR(-ENOMEM); + ret = cxio_allocate_window(&rhp->rdev, &stag, php->pdid); + if (ret) { + kfree(mhp); + return ERR_PTR(ret); + } + mhp->rhp = rhp; + mhp->attr.pdid = php->pdid; + mhp->attr.type = TPT_MW; + mhp->attr.stag = stag; + mmid = (stag) >> 8; + insert_handle(rhp, &rhp->mmidr, mhp, mmid); + PDBG("%s mmid 0x%x mhp %p stag 0x%x\n", __FUNCTION__, mmid, mhp, stag); + return &(mhp->ibmw); +} + +static int iwch_dealloc_mw(struct ib_mw *mw) +{ + struct iwch_dev *rhp; + struct iwch_mw *mhp; + u32 mmid; + + mhp = to_iwch_mw(mw); + rhp = mhp->rhp; + mmid = (mw->rkey) >> 8; + cxio_deallocate_window(&rhp->rdev, mhp->attr.stag); + remove_handle(rhp, &rhp->mmidr, mmid); + kfree(mhp); + PDBG("%s ib_mw %p mmid 0x%x ptr %p\n", __FUNCTION__, mw, mmid, mhp); + return 0; +} + +static int iwch_destroy_qp(struct ib_qp *ib_qp) +{ + struct iwch_dev *rhp; + struct iwch_qp *qhp; + struct iwch_qp_attributes attrs; + struct iwch_ucontext *ucontext; + + qhp = to_iwch_qp(ib_qp); + rhp = qhp->rhp; + + if (qhp->attr.state == IWCH_QP_STATE_RTS) { + attrs.next_state = IWCH_QP_STATE_ERROR; + iwch_modify_qp(rhp, qhp, IWCH_QP_ATTR_NEXT_STATE, &attrs, 0); + } + wait_event(qhp->wait, !qhp->ep); + + remove_handle(rhp, &rhp->qpidr, qhp->wq.qpid); + + atomic_dec(&qhp->refcnt); + wait_event(qhp->wait, !atomic_read(&qhp->refcnt)); + + ucontext = ib_qp->uobject ? to_iwch_ucontext(ib_qp->uobject->context) + : NULL; + cxio_destroy_qp(&rhp->rdev, &qhp->wq, + ucontext ? &ucontext->uctx : &rhp->rdev.uctx); + + PDBG("%s ib_qp %p qpid 0x%0x qhp %p\n", __FUNCTION__, + ib_qp, qhp->wq.qpid, qhp); + kfree(qhp); + return 0; +} + +static struct ib_qp *iwch_create_qp(struct ib_pd *pd, + struct ib_qp_init_attr *attrs, + struct ib_udata *udata) +{ + struct iwch_dev *rhp; + struct iwch_qp *qhp; + struct iwch_pd *php; + struct iwch_cq *schp; + struct iwch_cq *rchp; + struct iwch_create_qp_resp uresp; + int wqsize, sqsize, rqsize; + struct iwch_ucontext *ucontext; + + PDBG("%s ib_pd %p\n", __FUNCTION__, pd); + if (attrs->qp_type != IB_QPT_RC) + return ERR_PTR(-EINVAL); + php = to_iwch_pd(pd); + rhp = php->rhp; + schp = get_chp(rhp, ((struct iwch_cq *) attrs->send_cq)->cq.cqid); + rchp = get_chp(rhp, ((struct iwch_cq *) attrs->recv_cq)->cq.cqid); + if (!schp || !rchp) + return ERR_PTR(-EINVAL); + + /* The RQT size must be # of entries + 1 rounded up to a power of two */ + rqsize = roundup_pow_of_two(attrs->cap.max_recv_wr); + if (rqsize == attrs->cap.max_recv_wr) + rqsize = roundup_pow_of_two(attrs->cap.max_recv_wr+1); + + /* T3 doesn't support RQT depth < 16 */ + if (rqsize < 16) + rqsize = 16; + + if (rqsize > T3_MAX_RQ_SIZE) + return ERR_PTR(-EINVAL); + + /* + * NOTE: The SQ and total WQ sizes don't need to be + * a power of two. However, all the code assumes + * they are. EG: Q_FREECNT() and friends. + */ + sqsize = roundup_pow_of_two(attrs->cap.max_send_wr); + wqsize = roundup_pow_of_two(rqsize + sqsize); + PDBG("%s wqsize %d sqsize %d rqsize %d\n", __FUNCTION__, + wqsize, sqsize, rqsize); + qhp = kzalloc(sizeof(*qhp), GFP_KERNEL); + if (!qhp) + return ERR_PTR(-ENOMEM); + qhp->wq.size_log2 = ilog2(wqsize); + qhp->wq.rq_size_log2 = ilog2(rqsize); + qhp->wq.sq_size_log2 = ilog2(sqsize); + ucontext = pd->uobject ? to_iwch_ucontext(pd->uobject->context) : NULL; + if (cxio_create_qp(&rhp->rdev, !udata, &qhp->wq, + ucontext ? &ucontext->uctx : &rhp->rdev.uctx)) { + kfree(qhp); + return ERR_PTR(-ENOMEM); + } + attrs->cap.max_recv_wr = rqsize - 1; + attrs->cap.max_send_wr = sqsize; + qhp->rhp = rhp; + qhp->attr.pd = php->pdid; + qhp->attr.scq = ((struct iwch_cq *) attrs->send_cq)->cq.cqid; + qhp->attr.rcq = ((struct iwch_cq *) attrs->recv_cq)->cq.cqid; + qhp->attr.sq_num_entries = attrs->cap.max_send_wr; + qhp->attr.rq_num_entries = attrs->cap.max_recv_wr; + qhp->attr.sq_max_sges = attrs->cap.max_send_sge; + qhp->attr.sq_max_sges_rdma_write = attrs->cap.max_send_sge; + qhp->attr.rq_max_sges = attrs->cap.max_recv_sge; + qhp->attr.state = IWCH_QP_STATE_IDLE; + qhp->attr.next_state = IWCH_QP_STATE_IDLE; + + /* + * XXX - These don't get passed in from the openib user + * at create time. The CM sets them via a QP modify. + * Need to fix... I think the CM should + */ + qhp->attr.enable_rdma_read = 1; + qhp->attr.enable_rdma_write = 1; + qhp->attr.enable_bind = 1; + qhp->attr.max_ord = 1; + qhp->attr.max_ird = 1; + + spin_lock_init(&qhp->lock); + init_waitqueue_head(&qhp->wait); + atomic_set(&qhp->refcnt, 1); + insert_handle(rhp, &rhp->qpidr, qhp, qhp->wq.qpid); + + if (udata) { + + struct iwch_mm_entry *mm1, *mm2; + + mm1 = kmalloc(sizeof *mm1, GFP_KERNEL); + if (!mm1) { + iwch_destroy_qp(&qhp->ibqp); + return ERR_PTR(-ENOMEM); + } + + mm2 = kmalloc(sizeof *mm2, GFP_KERNEL); + if (!mm2) { + kfree(mm1); + iwch_destroy_qp(&qhp->ibqp); + return ERR_PTR(-ENOMEM); + } + + uresp.qpid = qhp->wq.qpid; + uresp.size_log2 = qhp->wq.size_log2; + uresp.sq_size_log2 = qhp->wq.sq_size_log2; + uresp.rq_size_log2 = qhp->wq.rq_size_log2; + spin_lock(&ucontext->mmap_lock); + uresp.key = ucontext->key; + ucontext->key += PAGE_SIZE; + uresp.db_key = ucontext->key; + ucontext->key += PAGE_SIZE; + spin_unlock(&ucontext->mmap_lock); + if (ib_copy_to_udata(udata, &uresp, sizeof (uresp))) { + kfree(mm1); + kfree(mm2); + iwch_destroy_qp(&qhp->ibqp); + return ERR_PTR(-EFAULT); + } + mm1->key = uresp.key; + mm1->addr = virt_to_phys(qhp->wq.queue); + mm1->len = PAGE_ALIGN(wqsize * sizeof (union t3_wr)); + insert_mmap(ucontext, mm1); + mm2->key = uresp.db_key; + mm2->addr = qhp->wq.udb & PAGE_MASK; + mm2->len = PAGE_SIZE; + insert_mmap(ucontext, mm2); + } + qhp->ibqp.qp_num = qhp->wq.qpid; + init_timer(&(qhp->timer)); + PDBG("%s sq_num_entries %d, rq_num_entries %d " + "qpid 0x%0x qhp %p dma_addr 0x%llx size %d\n", + __FUNCTION__, qhp->attr.sq_num_entries, qhp->attr.rq_num_entries, + qhp->wq.qpid, qhp, (unsigned long long) qhp->wq.dma_addr, + 1 << qhp->wq.size_log2); + return &qhp->ibqp; +} + +static int iwch_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr, + int attr_mask, struct ib_udata *udata) +{ + struct iwch_dev *rhp; + struct iwch_qp *qhp; + enum iwch_qp_attr_mask mask = 0; + struct iwch_qp_attributes attrs; + + PDBG("%s ib_qp %p\n", __FUNCTION__, ibqp); + + /* iwarp does not support the RTR state */ + if ((attr_mask & IB_QP_STATE) && (attr->qp_state == IB_QPS_RTR)) + attr_mask &= ~IB_QP_STATE; + + /* Make sure we still have something left to do */ + if (!attr_mask) + return 0; + + memset(&attrs, 0, sizeof attrs); + qhp = to_iwch_qp(ibqp); + rhp = qhp->rhp; + + attrs.next_state = iwch_convert_state(attr->qp_state); + attrs.enable_rdma_read = (attr->qp_access_flags & + IB_ACCESS_REMOTE_READ) ? 1 : 0; + attrs.enable_rdma_write = (attr->qp_access_flags & + IB_ACCESS_REMOTE_WRITE) ? 1 : 0; + attrs.enable_bind = (attr->qp_access_flags & IB_ACCESS_MW_BIND) ? 1 : 0; + + + mask |= (attr_mask & IB_QP_STATE) ? IWCH_QP_ATTR_NEXT_STATE : 0; + mask |= (attr_mask & IB_QP_ACCESS_FLAGS) ? + (IWCH_QP_ATTR_ENABLE_RDMA_READ | + IWCH_QP_ATTR_ENABLE_RDMA_WRITE | + IWCH_QP_ATTR_ENABLE_RDMA_BIND) : 0; + + return iwch_modify_qp(rhp, qhp, mask, &attrs, 0); +} + +void iwch_qp_add_ref(struct ib_qp *qp) +{ + PDBG("%s ib_qp %p\n", __FUNCTION__, qp); + atomic_inc(&(to_iwch_qp(qp)->refcnt)); +} + +void iwch_qp_rem_ref(struct ib_qp *qp) +{ + PDBG("%s ib_qp %p\n", __FUNCTION__, qp); + if (atomic_dec_and_test(&(to_iwch_qp(qp)->refcnt))) + wake_up(&(to_iwch_qp(qp)->wait)); +} + +struct ib_qp *iwch_get_qp(struct ib_device *dev, int qpn) +{ + PDBG("%s ib_dev %p qpn 0x%x\n", __FUNCTION__, dev, qpn); + return (struct ib_qp *)get_qhp(to_iwch_dev(dev), qpn); +} + + +static int iwch_query_pkey(struct ib_device *ibdev, + u8 port, u16 index, u16 * pkey) +{ + PDBG("%s ibdev %p\n", __FUNCTION__, ibdev); + *pkey = 0; + return 0; +} + +static int iwch_query_gid(struct ib_device *ibdev, u8 port, + int index, union ib_gid *gid) +{ + struct iwch_dev *dev; + + PDBG("%s ibdev %p, port %d, index %d, gid %p\n", + __FUNCTION__, ibdev, port, index, gid); + dev = to_iwch_dev(ibdev); + BUG_ON(port == 0 || port > 2); + memset(&(gid->raw[0]), 0, sizeof(gid->raw)); + memcpy(&(gid->raw[0]), dev->rdev.port_info.lldevs[port-1]->dev_addr, 6); + return 0; +} + +static int iwch_query_device(struct ib_device *ibdev, + struct ib_device_attr *props) +{ + + struct iwch_dev *dev; + PDBG("%s ibdev %p\n", __FUNCTION__, ibdev); + + dev = to_iwch_dev(ibdev); + memset(props, 0, sizeof *props); + memcpy(&props->sys_image_guid, dev->rdev.t3cdev_p->lldev->dev_addr, 6); + props->device_cap_flags = dev->device_cap_flags; + props->vendor_id = (u32)dev->rdev.rnic_info.pdev->vendor; + props->vendor_part_id = (u32)dev->rdev.rnic_info.pdev->device; + props->max_mr_size = ~0ull; + props->max_qp = dev->attr.max_qps; + props->max_qp_wr = dev->attr.max_wrs; + props->max_sge = dev->attr.max_sge_per_wr; + props->max_sge_rd = 1; + props->max_qp_rd_atom = dev->attr.max_rdma_reads_per_qp; + props->max_cq = dev->attr.max_cqs; + props->max_cqe = dev->attr.max_cqes_per_cq; + props->max_mr = dev->attr.max_mem_regs; + props->max_pd = dev->attr.max_pds; + props->local_ca_ack_delay = 0; + + return 0; +} + +static int iwch_query_port(struct ib_device *ibdev, + u8 port, struct ib_port_attr *props) +{ + PDBG("%s ibdev %p\n", __FUNCTION__, ibdev); + props->max_mtu = IB_MTU_4096; + props->lid = 0; + props->lmc = 0; + props->sm_lid = 0; + props->sm_sl = 0; + props->state = IB_PORT_ACTIVE; + props->phys_state = 0; + props->port_cap_flags = + IB_PORT_CM_SUP | + IB_PORT_SNMP_TUNNEL_SUP | + IB_PORT_REINIT_SUP | + IB_PORT_DEVICE_MGMT_SUP | + IB_PORT_VENDOR_CLASS_SUP | IB_PORT_BOOT_MGMT_SUP; + props->gid_tbl_len = 1; + props->pkey_tbl_len = 1; + props->qkey_viol_cntr = 0; + props->active_width = 2; + props->active_speed = 2; + props->max_msg_sz = -1; + + return 0; +} + +static ssize_t show_rev(struct class_device *cdev, char *buf) +{ + struct iwch_dev *dev = container_of(cdev, struct iwch_dev, + ibdev.class_dev); + PDBG("%s class dev 0x%p\n", __FUNCTION__, cdev); + return sprintf(buf, "%d\n", dev->rdev.t3cdev_p->type); +} + +static ssize_t show_fw_ver(struct class_device *cdev, char *buf) +{ + struct iwch_dev *dev = container_of(cdev, struct iwch_dev, + ibdev.class_dev); + struct ethtool_drvinfo info; + struct net_device *lldev = dev->rdev.t3cdev_p->lldev; + + PDBG("%s class dev 0x%p\n", __FUNCTION__, cdev); + lldev->ethtool_ops->get_drvinfo(lldev, &info); + return sprintf(buf, "%s\n", info.fw_version); +} + +static ssize_t show_hca(struct class_device *cdev, char *buf) +{ + struct iwch_dev *dev = container_of(cdev, struct iwch_dev, + ibdev.class_dev); + struct ethtool_drvinfo info; + struct net_device *lldev = dev->rdev.t3cdev_p->lldev; + + PDBG("%s class dev 0x%p\n", __FUNCTION__, cdev); + lldev->ethtool_ops->get_drvinfo(lldev, &info); + return sprintf(buf, "%s\n", info.driver); +} + +static ssize_t show_board(struct class_device *cdev, char *buf) +{ + struct iwch_dev *dev = container_of(cdev, struct iwch_dev, + ibdev.class_dev); + PDBG("%s class dev 0x%p\n", __FUNCTION__, dev); + return sprintf(buf, "%x.%x\n", dev->rdev.rnic_info.pdev->vendor, + dev->rdev.rnic_info.pdev->device); +} + +static CLASS_DEVICE_ATTR(hw_rev, S_IRUGO, show_rev, NULL); +static CLASS_DEVICE_ATTR(fw_ver, S_IRUGO, show_fw_ver, NULL); +static CLASS_DEVICE_ATTR(hca_type, S_IRUGO, show_hca, NULL); +static CLASS_DEVICE_ATTR(board_id, S_IRUGO, show_board, NULL); + +static struct class_device_attribute *iwch_class_attributes[] = { + &class_device_attr_hw_rev, + &class_device_attr_fw_ver, + &class_device_attr_hca_type, + &class_device_attr_board_id +}; + +int iwch_register_device(struct iwch_dev *dev) +{ + int ret; + int i; + + PDBG("%s iwch_dev %p\n", __FUNCTION__, dev); + strlcpy(dev->ibdev.name, "cxgb3_%d", IB_DEVICE_NAME_MAX); + memset(&dev->ibdev.node_guid, 0, sizeof(dev->ibdev.node_guid)); + memcpy(&dev->ibdev.node_guid, dev->rdev.t3cdev_p->lldev->dev_addr, 6); + dev->ibdev.owner = THIS_MODULE; + dev->device_cap_flags = + (IB_DEVICE_ZERO_STAG | + IB_DEVICE_SEND_W_INV | IB_DEVICE_MEM_WINDOW); + + dev->ibdev.uverbs_cmd_mask = + (1ull << IB_USER_VERBS_CMD_GET_CONTEXT) | + (1ull << IB_USER_VERBS_CMD_QUERY_DEVICE) | + (1ull << IB_USER_VERBS_CMD_QUERY_PORT) | + (1ull << IB_USER_VERBS_CMD_ALLOC_PD) | + (1ull << IB_USER_VERBS_CMD_DEALLOC_PD) | + (1ull << IB_USER_VERBS_CMD_REG_MR) | + (1ull << IB_USER_VERBS_CMD_DEREG_MR) | + (1ull << IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL) | + (1ull << IB_USER_VERBS_CMD_CREATE_CQ) | + (1ull << IB_USER_VERBS_CMD_DESTROY_CQ) | + (1ull << IB_USER_VERBS_CMD_REQ_NOTIFY_CQ) | + (1ull << IB_USER_VERBS_CMD_CREATE_QP) | + (1ull << IB_USER_VERBS_CMD_MODIFY_QP) | + (1ull << IB_USER_VERBS_CMD_POLL_CQ) | + (1ull << IB_USER_VERBS_CMD_DESTROY_QP) | + (1ull << IB_USER_VERBS_CMD_POST_SEND) | + (1ull << IB_USER_VERBS_CMD_POST_RECV); + dev->ibdev.node_type = RDMA_NODE_RNIC; + memcpy(dev->ibdev.node_desc, IWCH_NODE_DESC, sizeof(IWCH_NODE_DESC)); + dev->ibdev.phys_port_cnt = dev->rdev.port_info.nports; + dev->ibdev.dma_device = &(dev->rdev.rnic_info.pdev->dev); + dev->ibdev.class_dev.dev = &(dev->rdev.rnic_info.pdev->dev); + dev->ibdev.query_device = iwch_query_device; + dev->ibdev.query_port = iwch_query_port; + dev->ibdev.modify_port = iwch_modify_port; + dev->ibdev.query_pkey = iwch_query_pkey; + dev->ibdev.query_gid = iwch_query_gid; + dev->ibdev.alloc_ucontext = iwch_alloc_ucontext; + dev->ibdev.dealloc_ucontext = iwch_dealloc_ucontext; + dev->ibdev.mmap = iwch_mmap; + dev->ibdev.alloc_pd = iwch_allocate_pd; + dev->ibdev.dealloc_pd = iwch_deallocate_pd; + dev->ibdev.create_ah = iwch_ah_create; + dev->ibdev.destroy_ah = iwch_ah_destroy; + dev->ibdev.create_qp = iwch_create_qp; + dev->ibdev.modify_qp = iwch_ib_modify_qp; + dev->ibdev.destroy_qp = iwch_destroy_qp; + dev->ibdev.create_cq = iwch_create_cq; + dev->ibdev.destroy_cq = iwch_destroy_cq; + dev->ibdev.resize_cq = iwch_resize_cq; + dev->ibdev.poll_cq = iwch_poll_cq; + dev->ibdev.get_dma_mr = iwch_get_dma_mr; + dev->ibdev.reg_phys_mr = iwch_register_phys_mem; + dev->ibdev.rereg_phys_mr = iwch_reregister_phys_mem; + dev->ibdev.reg_user_mr = iwch_reg_user_mr; + dev->ibdev.dereg_mr = iwch_dereg_mr; + dev->ibdev.alloc_mw = iwch_alloc_mw; + dev->ibdev.bind_mw = iwch_bind_mw; + dev->ibdev.dealloc_mw = iwch_dealloc_mw; + + dev->ibdev.attach_mcast = iwch_multicast_attach; + dev->ibdev.detach_mcast = iwch_multicast_detach; + dev->ibdev.process_mad = iwch_process_mad; + + dev->ibdev.req_notify_cq = iwch_arm_cq; + dev->ibdev.post_send = iwch_post_send; + dev->ibdev.post_recv = iwch_post_receive; + + + dev->ibdev.iwcm = + (struct iw_cm_verbs *) kmalloc(sizeof(struct iw_cm_verbs), + GFP_KERNEL); + dev->ibdev.iwcm->connect = iwch_connect; + dev->ibdev.iwcm->accept = iwch_accept_cr; + dev->ibdev.iwcm->reject = iwch_reject_cr; + dev->ibdev.iwcm->create_listen = iwch_create_listen; + dev->ibdev.iwcm->destroy_listen = iwch_destroy_listen; + dev->ibdev.iwcm->add_ref = iwch_qp_add_ref; + dev->ibdev.iwcm->rem_ref = iwch_qp_rem_ref; + dev->ibdev.iwcm->get_qp = iwch_get_qp; + + ret = ib_register_device(&dev->ibdev); + if (ret) + goto bail1; + + for (i = 0; i < ARRAY_SIZE(iwch_class_attributes); ++i) { + ret = class_device_create_file(&dev->ibdev.class_dev, + iwch_class_attributes[i]); + if (ret) { + goto bail2; + } + } + return 0; +bail2: + ib_unregister_device(&dev->ibdev); +bail1: + return ret; +} + +void iwch_unregister_device(struct iwch_dev *dev) +{ + int i; + + PDBG("%s iwch_dev %p\n", __FUNCTION__, dev); + for (i = 0; i < ARRAY_SIZE(iwch_class_attributes); ++i) + class_device_remove_file(&dev->ibdev.class_dev, + iwch_class_attributes[i]); + ib_unregister_device(&dev->ibdev); + return; +} |