From: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
To: dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org
Cc: linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
Subject: [PATCH v2 19/49] IB/hfi1: add module init hooks
Date: Mon, 15 Jun 2015 09:27:41 -0400 [thread overview]
Message-ID: <20150615132741.29741.83775.stgit@phlsvslse11.ph.intel.com> (raw)
In-Reply-To: <20150615132434.29741.35258.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
Signed-off-by: Andrew Friedley <andrew.friedley-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Arthur Kepner <arthur.kepner-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Brendan Cunningham <brendan.cunningham-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Brian Welty <brian.welty-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Caz Yokoyama <caz.yokoyama-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dean Luick <dean.luick-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Easwar Hariharan <easwar.hariharan-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Harish Chegondi <harish.chegondi-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Ira Weiny <ira.weiny-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Jim Snow <jim.m.snow-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: John Gregor <john.a.gregor-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Jubin John <jubin.john-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Kaike Wan <kaike.wan-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Kevin Pine <kevin.pine-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Kyle Liddell <kyle.liddell-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Mike Marciniszyn <mike.marciniszyn-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Mitko Haralanov <mitko.haralanov-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Ravi Krishnaswamy <ravi.krishnaswamy-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Sadanand Warrier <sadanand.warrier-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Sanath Kumar <sanath.s.kumar-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Sudeep Dutt <sudeep.dutt-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
Signed-off-by: Vlad Danushevsky <vladimir.danusevsky-ral2JQCrhuEAvxtiuMwx3w@public.gmane.org>
---
drivers/infiniband/hw/hfi1/init.c | 1709 +++++++++++++++++++++++++++++++++++++
1 file changed, 1709 insertions(+)
create mode 100644 drivers/infiniband/hw/hfi1/init.c
diff --git a/drivers/infiniband/hw/hfi1/init.c b/drivers/infiniband/hw/hfi1/init.c
new file mode 100644
index 0000000..99eb9f1
--- /dev/null
+++ b/drivers/infiniband/hw/hfi1/init.c
@@ -0,0 +1,1709 @@
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license. When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * 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.
+ * - Neither the name of Intel Corporation nor the names of its
+ * contributors may be used to endorse or promote products derived
+ * from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include <linux/pci.h>
+#include <linux/netdevice.h>
+#include <linux/vmalloc.h>
+#include <linux/delay.h>
+#include <linux/idr.h>
+#include <linux/module.h>
+#include <linux/printk.h>
+#include <linux/hrtimer.h>
+
+#include "hfi.h"
+#include "device.h"
+#include "common.h"
+#include "mad.h"
+#include "sdma.h"
+#include "debugfs.h"
+#include "verbs.h"
+
+#undef pr_fmt
+#define pr_fmt(fmt) DRIVER_NAME ": " fmt
+
+/*
+ * min buffers we want to have per context, after driver
+ */
+#define HFI1_MIN_USER_CTXT_BUFCNT 7
+
+#define HFI1_MIN_HDRQ_EGRBUF_CNT 2
+#define HFI1_MIN_EAGER_BUFFER_SIZE (4 * 1024) /* 4KB */
+#define HFI1_MAX_EAGER_BUFFER_SIZE (256 * 1024) /* 256KB */
+
+/*
+ * Number of receive contexts we are configured to use (to allow for more pio
+ * buffers per ctxt, etc.) Zero means use chip value.
+ */
+uint num_rcv_contexts;
+module_param_named(num_rcv_contexts, num_rcv_contexts, uint, S_IRUGO);
+MODULE_PARM_DESC(num_rcv_contexts, "Set max number of receive contexts to use");
+
+u8 krcvqs[RXE_NUM_DATA_VL];
+int krcvqsset;
+module_param_array(krcvqs, byte, &krcvqsset, S_IRUGO);
+MODULE_PARM_DESC(krcvqs, "Array of the number of kernel receive queues by VL");
+
+/* computed based on above array */
+unsigned n_krcvqs;
+
+static unsigned hfi1_rcvarr_split = 25;
+module_param_named(rcvarr_split, hfi1_rcvarr_split, uint, S_IRUGO);
+MODULE_PARM_DESC(rcvarr_split, "Percent of context's RcvArray entries used for Eager buffers");
+
+static uint eager_buffer_size = (2 << 20); /* 2MB */
+module_param(eager_buffer_size, uint, S_IRUGO);
+MODULE_PARM_DESC(eager_buffer_size, "Size of the eager buffers, default: 2MB");
+
+static uint rcvhdrcnt = 2048; /* 2x the max eager buffer count */
+module_param_named(rcvhdrcnt, rcvhdrcnt, uint, S_IRUGO);
+MODULE_PARM_DESC(rcvhdrcnt, "Receive header queue count (default 2048)");
+
+static uint hfi1_hdrq_entsize = 32;
+module_param_named(hdrq_entsize, hfi1_hdrq_entsize, uint, S_IRUGO);
+MODULE_PARM_DESC(hdrq_entsize, "Size of header queue entries: 2 - 8B, 16 - 64B (default), 32 - 128B");
+
+unsigned int user_credit_return_threshold = 33; /* default is 33% */
+module_param(user_credit_return_threshold, uint, S_IRUGO);
+MODULE_PARM_DESC(user_credit_return_theshold, "Credit return threshold for user send contexts, return when unreturned credits passes this many blocks (in percent of allocated blocks, 0 is off)");
+
+static inline u64 encode_rcv_header_entry_size(u16);
+
+static struct idr hfi1_unit_table;
+u32 hfi1_cpulist_count;
+unsigned long *hfi1_cpulist;
+
+/*
+ * Common code for creating the receive context array.
+ */
+int hfi1_create_ctxts(struct hfi1_devdata *dd)
+{
+ unsigned i;
+ int ret;
+ int local_node_id = pcibus_to_node(dd->pcidev->bus);
+
+ if (local_node_id < 0)
+ local_node_id = numa_node_id();
+ dd->assigned_node_id = local_node_id;
+
+ dd->rcd = kcalloc(dd->num_rcv_contexts, sizeof(*dd->rcd), GFP_KERNEL);
+ if (!dd->rcd) {
+ dd_dev_err(dd,
+ "Unable to allocate receive context array, failing\n");
+ goto nomem;
+ }
+
+ /* create one or more kernel contexts */
+ for (i = 0; i < dd->first_user_ctxt; ++i) {
+ struct hfi1_pportdata *ppd;
+ struct hfi1_ctxtdata *rcd;
+
+ ppd = dd->pport + (i % dd->num_pports);
+ rcd = hfi1_create_ctxtdata(ppd, i);
+ if (!rcd) {
+ dd_dev_err(dd,
+ "Unable to allocate kernel receive context, failing\n");
+ goto nomem;
+ }
+ /*
+ * Set up the kernel context flags here and now because they
+ * use default values for all receive side memories. User
+ * contexts will be handled as they are created.
+ */
+ rcd->flags = HFI1_CAP_KGET(MULTI_PKT_EGR) |
+ HFI1_CAP_KGET(NODROP_RHQ_FULL) |
+ HFI1_CAP_KGET(NODROP_EGR_FULL) |
+ HFI1_CAP_KGET(DMA_RTAIL);
+ rcd->seq_cnt = 1;
+
+ rcd->sc = sc_alloc(dd, SC_ACK, rcd->rcvhdrqentsize, dd->node);
+ if (!rcd->sc) {
+ dd_dev_err(dd,
+ "Unable to allocate kernel send context, failing\n");
+ dd->rcd[rcd->ctxt] = NULL;
+ hfi1_free_ctxtdata(dd, rcd);
+ goto nomem;
+ }
+
+ ret = hfi1_init_ctxt(rcd->sc);
+ if (ret < 0) {
+ dd_dev_err(dd,
+ "Failed to setup kernel receive context, failing\n");
+ sc_free(rcd->sc);
+ dd->rcd[rcd->ctxt] = NULL;
+ hfi1_free_ctxtdata(dd, rcd);
+ ret = -EFAULT;
+ goto bail;
+ }
+ }
+
+ return 0;
+nomem:
+ ret = -ENOMEM;
+bail:
+ kfree(dd->rcd);
+ dd->rcd = NULL;
+ return ret;
+}
+
+/*
+ * Common code for user and kernel context setup.
+ */
+struct hfi1_ctxtdata *hfi1_create_ctxtdata(struct hfi1_pportdata *ppd, u32 ctxt)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ struct hfi1_ctxtdata *rcd;
+ unsigned kctxt_ngroups = 0;
+ u32 base;
+
+ if (dd->rcv_entries.nctxt_extra >
+ dd->num_rcv_contexts - dd->first_user_ctxt)
+ kctxt_ngroups = (dd->rcv_entries.nctxt_extra -
+ (dd->num_rcv_contexts - dd->first_user_ctxt));
+ rcd = kzalloc(sizeof(*rcd), GFP_KERNEL);
+ if (rcd) {
+ u32 rcvtids, max_entries;
+
+ dd_dev_info(dd, "%s: setting up context %u\n", __func__, ctxt);
+
+ INIT_LIST_HEAD(&rcd->qp_wait_list);
+ rcd->ppd = ppd;
+ rcd->dd = dd;
+ rcd->cnt = 1;
+ rcd->ctxt = ctxt;
+ dd->rcd[ctxt] = rcd;
+ rcd->numa_id = numa_node_id();
+ rcd->rcv_array_groups = dd->rcv_entries.ngroups;
+
+ spin_lock_init(&rcd->exp_lock);
+
+ /*
+ * Calculate the context's RcvArray entry starting point.
+ * We do this here because we have to take into account all
+ * the RcvArray entries that previous context would have
+ * taken and we have to account for any extra groups
+ * assigned to the kernel or user contexts.
+ */
+ if (ctxt < dd->first_user_ctxt) {
+ if (ctxt < kctxt_ngroups) {
+ base = ctxt * (dd->rcv_entries.ngroups + 1);
+ rcd->rcv_array_groups++;
+ } else
+ base = kctxt_ngroups +
+ (ctxt * dd->rcv_entries.ngroups);
+ } else {
+ u16 ct = ctxt - dd->first_user_ctxt;
+
+ base = ((dd->n_krcv_queues * dd->rcv_entries.ngroups) +
+ kctxt_ngroups);
+ if (ct < dd->rcv_entries.nctxt_extra) {
+ base += ct * (dd->rcv_entries.ngroups + 1);
+ rcd->rcv_array_groups++;
+ } else
+ base += dd->rcv_entries.nctxt_extra +
+ (ct * dd->rcv_entries.ngroups);
+ }
+ rcd->eager_base = base * dd->rcv_entries.group_size;
+
+ /* Validate and initialize Rcv Hdr Q variables */
+ if (rcvhdrcnt & HDRQ_INCREMENT) {
+ dd_dev_err(dd,
+ "ctxt%u: header queue count %d must be divisible by %d\n",
+ rcd->ctxt, rcvhdrcnt, HDRQ_INCREMENT);
+ goto bail;
+ }
+ rcd->rcvhdrq_cnt = rcvhdrcnt;
+ rcd->rcvhdrqentsize = hfi1_hdrq_entsize;
+ /*
+ * Simple Eager buffer allocation: we have already pre-allocated
+ * the number of RcvArray entry groups. Each ctxtdata structure
+ * holds the number of groups for that context.
+ *
+ * To follow CSR requirements and maintain cacheline alignment,
+ * make sure all sizes and bases are multiples of group_size.
+ *
+ * The expected entry count is what is left after assigning
+ * eager.
+ */
+ max_entries = rcd->rcv_array_groups *
+ dd->rcv_entries.group_size;
+ rcvtids = ((max_entries * hfi1_rcvarr_split) / 100);
+ rcd->egrbufs.count = round_down(rcvtids,
+ dd->rcv_entries.group_size);
+ if (rcd->egrbufs.count > MAX_EAGER_ENTRIES) {
+ dd_dev_err(dd, "ctxt%u: requested too many RcvArray entries.\n",
+ rcd->ctxt);
+ rcd->egrbufs.count = MAX_EAGER_ENTRIES;
+ }
+ dd_dev_info(dd, "ctxt%u: max Eager buffer RcvArray entries: %u\n",
+ rcd->ctxt, rcd->egrbufs.count);
+
+ /*
+ * Allocate array that will hold the eager buffer accounting
+ * data.
+ * This will allocate the maximum possible buffer count based
+ * on the value of the RcvArray split parameter.
+ * The resulting value will be rounded down to the closest
+ * multiple of dd->rcv_entries.group_size.
+ */
+ rcd->egrbufs.buffers = kzalloc(sizeof(*rcd->egrbufs.buffers) *
+ rcd->egrbufs.count, GFP_KERNEL);
+ if (!rcd->egrbufs.buffers)
+ goto bail;
+ rcd->egrbufs.rcvtids = kzalloc(sizeof(*rcd->egrbufs.rcvtids) *
+ rcd->egrbufs.count, GFP_KERNEL);
+ if (!rcd->egrbufs.rcvtids)
+ goto bail;
+ rcd->egrbufs.size = eager_buffer_size;
+ /*
+ * The size of the buffers programmed into the RcvArray
+ * entries needs to be big enough to handle the highest
+ * MTU supported.
+ */
+ if (rcd->egrbufs.size < hfi1_max_mtu) {
+ rcd->egrbufs.size = __roundup_pow_of_two(hfi1_max_mtu);
+ dd_dev_info(dd,
+ "ctxt%u: eager bufs size too small. Adjusting to %zu\n",
+ rcd->ctxt, rcd->egrbufs.size);
+ }
+ rcd->egrbufs.rcvtid_size = HFI1_MAX_EAGER_BUFFER_SIZE;
+
+ if (ctxt < dd->first_user_ctxt) { /* N/A for PSM contexts */
+ rcd->opstats = kzalloc(sizeof(*rcd->opstats),
+ GFP_KERNEL);
+ if (!rcd->opstats) {
+ dd_dev_err(dd,
+ "ctxt%u: Unable to allocate per ctxt stats buffer\n",
+ rcd->ctxt);
+ goto bail;
+ }
+ }
+ }
+ return rcd;
+bail:
+ kfree(rcd->opstats);
+ kfree(rcd->egrbufs.rcvtids);
+ kfree(rcd->egrbufs.buffers);
+ kfree(rcd);
+ return NULL;
+}
+
+/*
+ * Convert a receive header entry size that to the encoding used in the CSR.
+ *
+ * Return a zero if the given size is invalid.
+ */
+static inline u64 encode_rcv_header_entry_size(u16 size)
+{
+ /* there are only 3 valid receive header entry sizes */
+ if (size == 2)
+ return 1;
+ if (size == 16)
+ return 2;
+ else if (size == 32)
+ return 4;
+ return 0; /* invalid */
+}
+
+/*
+ * Select the largest ccti value over all SLs to determine the intra-
+ * packet gap for the link.
+ *
+ * called with cca_timer_lock held (to protect access to cca_timer
+ * array), and rcu_read_lock() (to protect access to cc_state).
+ */
+void set_link_ipg(struct hfi1_pportdata *ppd)
+{
+ struct hfi1_devdata *dd = ppd->dd;
+ struct cc_state *cc_state;
+ int i;
+ u16 cce, ccti_limit, max_ccti = 0;
+ u16 shift, mult;
+ u64 src;
+ u32 current_egress_rate; /* Mbits /sec */
+ u32 max_pkt_time;
+ /*
+ * max_pkt_time is the maximum packet egress time in units
+ * of the fabric clock period 1/(805 MHz).
+ */
+
+ cc_state = get_cc_state(ppd);
+
+ if (cc_state == NULL)
+ /*
+ * This should _never_ happen - rcu_read_lock() is held,
+ * and set_link_ipg() should not be called if cc_state
+ * is NULL.
+ */
+ return;
+
+ for (i = 0; i < OPA_MAX_SLS; i++) {
+ u16 ccti = ppd->cca_timer[i].ccti;
+
+ if (ccti > max_ccti)
+ max_ccti = ccti;
+ }
+
+ ccti_limit = cc_state->cct.ccti_limit;
+ if (max_ccti > ccti_limit)
+ max_ccti = ccti_limit;
+
+ cce = cc_state->cct.entries[max_ccti].entry;
+ shift = (cce & 0xc000) >> 14;
+ mult = (cce & 0x3fff);
+
+ current_egress_rate = active_egress_rate(ppd);
+
+ max_pkt_time = egress_cycles(ppd->ibmaxlen, current_egress_rate);
+
+ src = (max_pkt_time >> shift) * mult;
+
+ src &= SEND_STATIC_RATE_CONTROL_CSR_SRC_RELOAD_SMASK;
+ src <<= SEND_STATIC_RATE_CONTROL_CSR_SRC_RELOAD_SHIFT;
+
+ write_csr(dd, SEND_STATIC_RATE_CONTROL, src);
+}
+
+static enum hrtimer_restart cca_timer_fn(struct hrtimer *t)
+{
+ struct cca_timer *cca_timer;
+ struct hfi1_pportdata *ppd;
+ int sl;
+ u16 ccti, ccti_timer, ccti_min;
+ struct cc_state *cc_state;
+
+ cca_timer = container_of(t, struct cca_timer, hrtimer);
+ ppd = cca_timer->ppd;
+ sl = cca_timer->sl;
+
+ rcu_read_lock();
+
+ cc_state = get_cc_state(ppd);
+
+ if (cc_state == NULL) {
+ rcu_read_unlock();
+ return HRTIMER_NORESTART;
+ }
+
+ /*
+ * 1) decrement ccti for SL
+ * 2) calculate IPG for link (set_link_ipg())
+ * 3) restart timer, unless ccti is at min value
+ */
+
+ ccti_min = cc_state->cong_setting.entries[sl].ccti_min;
+ ccti_timer = cc_state->cong_setting.entries[sl].ccti_timer;
+
+ spin_lock(&ppd->cca_timer_lock);
+
+ ccti = cca_timer->ccti;
+
+ if (ccti > ccti_min) {
+ cca_timer->ccti--;
+ set_link_ipg(ppd);
+ }
+
+ spin_unlock(&ppd->cca_timer_lock);
+
+ rcu_read_unlock();
+
+ if (ccti > ccti_min) {
+ unsigned long nsec = 1024 * ccti_timer;
+ /* ccti_timer is in units of 1.024 usec */
+ hrtimer_forward_now(t, ns_to_ktime(nsec));
+ return HRTIMER_RESTART;
+ }
+ return HRTIMER_NORESTART;
+}
+
+/*
+ * Common code for initializing the physical port structure.
+ */
+void hfi1_init_pportdata(struct pci_dev *pdev, struct hfi1_pportdata *ppd,
+ struct hfi1_devdata *dd, u8 hw_pidx, u8 port)
+{
+ int i, size;
+ uint default_pkey_idx;
+
+ ppd->dd = dd;
+ ppd->hw_pidx = hw_pidx;
+ ppd->port = port; /* IB port number, not index */
+
+ default_pkey_idx = 1;
+
+ ppd->pkeys[default_pkey_idx] = DEFAULT_P_KEY;
+ if (loopback) {
+ hfi1_early_err(&pdev->dev,
+ "Faking data partition 0x8001 in idx %u\n",
+ !default_pkey_idx);
+ ppd->pkeys[!default_pkey_idx] = 0x8001;
+ }
+
+ INIT_WORK(&ppd->link_vc_work, handle_verify_cap);
+ INIT_WORK(&ppd->link_up_work, handle_link_up);
+ INIT_WORK(&ppd->link_down_work, handle_link_down);
+ INIT_WORK(&ppd->freeze_work, handle_freeze);
+ INIT_WORK(&ppd->link_downgrade_work, handle_link_downgrade);
+ INIT_WORK(&ppd->sma_message_work, handle_sma_message);
+ mutex_init(&ppd->hls_lock);
+ spin_lock_init(&ppd->sdma_alllock);
+ spin_lock_init(&ppd->qsfp_info.qsfp_lock);
+
+ ppd->sm_trap_qp = 0x0;
+ ppd->sa_qp = 0x1;
+
+ ppd->hfi1_wq = NULL;
+
+ spin_lock_init(&ppd->cca_timer_lock);
+
+ for (i = 0; i < OPA_MAX_SLS; i++) {
+ hrtimer_init(&ppd->cca_timer[i].hrtimer, CLOCK_MONOTONIC,
+ HRTIMER_MODE_REL);
+ ppd->cca_timer[i].ppd = ppd;
+ ppd->cca_timer[i].sl = i;
+ ppd->cca_timer[i].ccti = 0;
+ ppd->cca_timer[i].hrtimer.function = cca_timer_fn;
+ }
+
+ ppd->cc_max_table_entries = IB_CC_TABLE_CAP_DEFAULT;
+
+ spin_lock_init(&ppd->cc_state_lock);
+ spin_lock_init(&ppd->cc_log_lock);
+ size = sizeof(struct cc_state);
+ RCU_INIT_POINTER(ppd->cc_state, kzalloc(size, GFP_KERNEL));
+ if (!rcu_dereference(ppd->cc_state))
+ goto bail;
+ return;
+
+bail:
+
+ hfi1_early_err(&pdev->dev,
+ "Congestion Control Agent disabled for port %d\n", port);
+}
+
+/*
+ * Do initialization for device that is only needed on
+ * first detect, not on resets.
+ */
+static int loadtime_init(struct hfi1_devdata *dd)
+{
+ return 0;
+}
+
+/**
+ * init_after_reset - re-initialize after a reset
+ * @dd: the hfi1_ib device
+ *
+ * sanity check at least some of the values after reset, and
+ * ensure no receive or transmit (explicitly, in case reset
+ * failed
+ */
+static int init_after_reset(struct hfi1_devdata *dd)
+{
+ int i;
+
+ /*
+ * Ensure chip does no sends or receives, tail updates, or
+ * pioavail updates while we re-initialize. This is mostly
+ * for the driver data structures, not chip registers.
+ */
+ for (i = 0; i < dd->num_rcv_contexts; i++)
+ hfi1_rcvctrl(dd, HFI1_RCVCTRL_CTXT_DIS |
+ HFI1_RCVCTRL_INTRAVAIL_DIS |
+ HFI1_RCVCTRL_TAILUPD_DIS, i);
+ pio_send_control(dd, PSC_GLOBAL_DISABLE);
+ for (i = 0; i < dd->num_send_contexts; i++)
+ sc_disable(dd->send_contexts[i].sc);
+
+ return 0;
+}
+
+static void enable_chip(struct hfi1_devdata *dd)
+{
+ u32 rcvmask;
+ u32 i;
+
+ /* enable PIO send */
+ pio_send_control(dd, PSC_GLOBAL_ENABLE);
+
+ /*
+ * Enable kernel ctxts' receive and receive interrupt.
+ * Other ctxts done as user opens and initializes them.
+ */
+ rcvmask = HFI1_RCVCTRL_CTXT_ENB | HFI1_RCVCTRL_INTRAVAIL_ENB;
+ for (i = 0; i < dd->first_user_ctxt; ++i) {
+ rcvmask |= HFI1_CAP_KGET_MASK(dd->rcd[i]->flags, DMA_RTAIL) ?
+ HFI1_RCVCTRL_TAILUPD_ENB : HFI1_RCVCTRL_TAILUPD_DIS;
+ if (!HFI1_CAP_KGET_MASK(dd->rcd[i]->flags, MULTI_PKT_EGR))
+ rcvmask |= HFI1_RCVCTRL_ONE_PKT_EGR_ENB;
+ if (HFI1_CAP_KGET_MASK(dd->rcd[i]->flags, NODROP_RHQ_FULL))
+ rcvmask |= HFI1_RCVCTRL_NO_RHQ_DROP_ENB;
+ if (HFI1_CAP_KGET_MASK(dd->rcd[i]->flags, NODROP_EGR_FULL))
+ rcvmask |= HFI1_RCVCTRL_NO_EGR_DROP_ENB;
+ hfi1_rcvctrl(dd, rcvmask, i);
+ sc_enable(dd->rcd[i]->sc);
+ }
+}
+
+/**
+ * create_workqueues - create per port workqueues
+ * @dd: the hfi1_ib device
+ */
+static int create_workqueues(struct hfi1_devdata *dd)
+{
+ int pidx;
+ struct hfi1_pportdata *ppd;
+
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ ppd = dd->pport + pidx;
+ if (!ppd->hfi1_wq) {
+ char wq_name[8]; /* 3 + 2 + 1 + 1 + 1 */
+
+ snprintf(wq_name, sizeof(wq_name), "hfi%d_%d",
+ dd->unit, pidx);
+ ppd->hfi1_wq =
+ create_singlethread_workqueue(wq_name);
+ if (!ppd->hfi1_wq)
+ goto wq_error;
+ }
+ }
+ return 0;
+wq_error:
+ pr_err("create_singlethread_workqueue failed for port %d\n",
+ pidx + 1);
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ ppd = dd->pport + pidx;
+ if (ppd->hfi1_wq) {
+ destroy_workqueue(ppd->hfi1_wq);
+ ppd->hfi1_wq = NULL;
+ }
+ }
+ return -ENOMEM;
+}
+
+/**
+ * hfi1_init - do the actual initialization sequence on the chip
+ * @dd: the hfi1_ib device
+ * @reinit: re-initializing, so don't allocate new memory
+ *
+ * Do the actual initialization sequence on the chip. This is done
+ * both from the init routine called from the PCI infrastructure, and
+ * when we reset the chip, or detect that it was reset internally,
+ * or it's administratively re-enabled.
+ *
+ * Memory allocation here and in called routines is only done in
+ * the first case (reinit == 0). We have to be careful, because even
+ * without memory allocation, we need to re-write all the chip registers
+ * TIDs, etc. after the reset or enable has completed.
+ */
+int hfi1_init(struct hfi1_devdata *dd, int reinit)
+{
+ int ret = 0, pidx, lastfail = 0;
+ unsigned i, len;
+ struct hfi1_ctxtdata *rcd;
+ struct hfi1_pportdata *ppd;
+
+ /* Set up recv low level handlers */
+ dd->normal_rhf_rcv_functions[RHF_RCV_TYPE_EXPECTED] =
+ kdeth_process_expected;
+ dd->normal_rhf_rcv_functions[RHF_RCV_TYPE_EAGER] =
+ kdeth_process_eager;
+ dd->normal_rhf_rcv_functions[RHF_RCV_TYPE_IB] = process_receive_ib;
+ dd->normal_rhf_rcv_functions[RHF_RCV_TYPE_ERROR] =
+ process_receive_error;
+ dd->normal_rhf_rcv_functions[RHF_RCV_TYPE_BYPASS] =
+ process_receive_bypass;
+ dd->normal_rhf_rcv_functions[RHF_RCV_TYPE_INVALID5] =
+ process_receive_invalid;
+ dd->normal_rhf_rcv_functions[RHF_RCV_TYPE_INVALID6] =
+ process_receive_invalid;
+ dd->normal_rhf_rcv_functions[RHF_RCV_TYPE_INVALID7] =
+ process_receive_invalid;
+ dd->rhf_rcv_function_map = dd->normal_rhf_rcv_functions;
+
+ /* Set up send low level handlers */
+ dd->process_pio_send = hfi1_verbs_send_pio;
+ dd->process_dma_send = hfi1_verbs_send_dma;
+ dd->pio_inline_send = pio_copy;
+
+ if (is_a0(dd)) {
+ atomic_set(&dd->drop_packet, DROP_PACKET_ON);
+ dd->do_drop = 1;
+ } else {
+ atomic_set(&dd->drop_packet, DROP_PACKET_OFF);
+ dd->do_drop = 0;
+ }
+
+ /* make sure the link is not "up" */
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ ppd = dd->pport + pidx;
+ ppd->linkup = 0;
+ }
+
+ if (reinit)
+ ret = init_after_reset(dd);
+ else
+ ret = loadtime_init(dd);
+ if (ret)
+ goto done;
+
+ /* dd->rcd can be NULL if early initialization failed */
+ for (i = 0; dd->rcd && i < dd->first_user_ctxt; ++i) {
+ /*
+ * Set up the (kernel) rcvhdr queue and egr TIDs. If doing
+ * re-init, the simplest way to handle this is to free
+ * existing, and re-allocate.
+ * Need to re-create rest of ctxt 0 ctxtdata as well.
+ */
+ rcd = dd->rcd[i];
+ if (!rcd)
+ continue;
+
+ rcd->do_interrupt = &handle_receive_interrupt;
+
+ lastfail = hfi1_create_rcvhdrq(dd, rcd);
+ if (!lastfail)
+ lastfail = hfi1_setup_eagerbufs(rcd);
+ if (lastfail)
+ dd_dev_err(dd,
+ "failed to allocate kernel ctxt's rcvhdrq and/or egr bufs\n");
+ }
+ if (lastfail)
+ ret = lastfail;
+
+ /* Allocate enough memory for user event notification. */
+ len = ALIGN(dd->chip_rcv_contexts * HFI1_MAX_SHARED_CTXTS *
+ sizeof(*dd->events), PAGE_SIZE);
+ dd->events = vmalloc_user(len);
+ if (!dd->events)
+ dd_dev_err(dd, "Failed to allocate user events page\n");
+ /*
+ * Allocate a page for device and port status.
+ * Page will be shared amongst all user processes.
+ */
+ dd->status = vmalloc_user(PAGE_SIZE);
+ if (!dd->status)
+ dd_dev_err(dd, "Failed to allocate dev status page\n");
+ else
+ dd->freezelen = PAGE_SIZE - (sizeof(*dd->status) -
+ sizeof(dd->status->freezemsg));
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ ppd = dd->pport + pidx;
+ if (dd->status)
+ /* Currently, we only have one port */
+ ppd->statusp = &dd->status->port;
+
+ set_mtu(ppd);
+ }
+
+ /* enable chip even if we have an error, so we can debug cause */
+ enable_chip(dd);
+
+ ret = hfi1_cq_init(dd);
+done:
+ /*
+ * Set status even if port serdes is not initialized
+ * so that diags will work.
+ */
+ if (dd->status)
+ dd->status->dev |= HFI1_STATUS_CHIP_PRESENT |
+ HFI1_STATUS_INITTED;
+ if (!ret) {
+ /* enable all interrupts from the chip */
+ set_intr_state(dd, 1);
+
+ /* chip is OK for user apps; mark it as initialized */
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ ppd = dd->pport + pidx;
+
+ /* initialize the qsfp if it exists
+ * Requires interrupts to be enabled so we are notified
+ * when the QSFP completes reset, and has
+ * to be done before bringing up the SERDES
+ */
+ init_qsfp(ppd);
+
+ /* start the serdes - must be after interrupts are
+ enabled so we are notified when the link goes up */
+ lastfail = bringup_serdes(ppd);
+ if (lastfail)
+ dd_dev_info(dd,
+ "Failed to bring up port %u\n",
+ ppd->port);
+
+ /*
+ * Set status even if port serdes is not initialized
+ * so that diags will work.
+ */
+ if (ppd->statusp)
+ *ppd->statusp |= HFI1_STATUS_CHIP_PRESENT |
+ HFI1_STATUS_INITTED;
+ if (!ppd->link_speed_enabled)
+ continue;
+ }
+ }
+
+ /* if ret is non-zero, we probably should do some cleanup here... */
+ return ret;
+}
+
+static inline struct hfi1_devdata *__hfi1_lookup(int unit)
+{
+ return idr_find(&hfi1_unit_table, unit);
+}
+
+struct hfi1_devdata *hfi1_lookup(int unit)
+{
+ struct hfi1_devdata *dd;
+ unsigned long flags;
+
+ spin_lock_irqsave(&hfi1_devs_lock, flags);
+ dd = __hfi1_lookup(unit);
+ spin_unlock_irqrestore(&hfi1_devs_lock, flags);
+
+ return dd;
+}
+
+/*
+ * Stop the timers during unit shutdown, or after an error late
+ * in initialization.
+ */
+static void stop_timers(struct hfi1_devdata *dd)
+{
+ struct hfi1_pportdata *ppd;
+ int pidx;
+
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ ppd = dd->pport + pidx;
+ if (ppd->led_override_timer.data) {
+ del_timer_sync(&ppd->led_override_timer);
+ atomic_set(&ppd->led_override_timer_active, 0);
+ }
+ }
+}
+
+/**
+ * shutdown_device - shut down a device
+ * @dd: the hfi1_ib device
+ *
+ * This is called to make the device quiet when we are about to
+ * unload the driver, and also when the device is administratively
+ * disabled. It does not free any data structures.
+ * Everything it does has to be setup again by hfi1_init(dd, 1)
+ */
+static void shutdown_device(struct hfi1_devdata *dd)
+{
+ struct hfi1_pportdata *ppd;
+ unsigned pidx;
+ int i;
+
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ ppd = dd->pport + pidx;
+
+ ppd->linkup = 0;
+ if (ppd->statusp)
+ *ppd->statusp &= ~(HFI1_STATUS_IB_CONF |
+ HFI1_STATUS_IB_READY);
+ }
+ dd->flags &= ~HFI1_INITTED;
+
+ /* mask interrupts, but not errors */
+ set_intr_state(dd, 0);
+
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ ppd = dd->pport + pidx;
+ for (i = 0; i < dd->num_rcv_contexts; i++)
+ hfi1_rcvctrl(dd, HFI1_RCVCTRL_TAILUPD_DIS |
+ HFI1_RCVCTRL_CTXT_DIS |
+ HFI1_RCVCTRL_INTRAVAIL_DIS |
+ HFI1_RCVCTRL_PKEY_DIS |
+ HFI1_RCVCTRL_ONE_PKT_EGR_DIS, i);
+ /*
+ * Gracefully stop all sends allowing any in progress to
+ * trickle out first.
+ */
+ for (i = 0; i < dd->num_send_contexts; i++)
+ sc_flush(dd->send_contexts[i].sc);
+ }
+
+ /*
+ * Enough for anything that's going to trickle out to have actually
+ * done so.
+ */
+ udelay(20);
+
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ ppd = dd->pport + pidx;
+
+ /* disable all contexts */
+ for (i = 0; i < dd->num_send_contexts; i++)
+ sc_disable(dd->send_contexts[i].sc);
+ /* disable the send device */
+ pio_send_control(dd, PSC_GLOBAL_DISABLE);
+
+ /*
+ * Clear SerdesEnable.
+ * We can't count on interrupts since we are stopping.
+ */
+ hfi1_quiet_serdes(ppd);
+
+ if (ppd->hfi1_wq) {
+ destroy_workqueue(ppd->hfi1_wq);
+ ppd->hfi1_wq = NULL;
+ }
+ }
+ sdma_exit(dd);
+}
+
+/**
+ * hfi1_free_ctxtdata - free a context's allocated data
+ * @dd: the hfi1_ib device
+ * @rcd: the ctxtdata structure
+ *
+ * free up any allocated data for a context
+ * This should not touch anything that would affect a simultaneous
+ * re-allocation of context data, because it is called after hfi1_mutex
+ * is released (and can be called from reinit as well).
+ * It should never change any chip state, or global driver state.
+ */
+void hfi1_free_ctxtdata(struct hfi1_devdata *dd, struct hfi1_ctxtdata *rcd)
+{
+ unsigned e;
+
+ if (!rcd)
+ return;
+
+ if (rcd->rcvhdrq) {
+ dma_free_coherent(&dd->pcidev->dev, rcd->rcvhdrq_size,
+ rcd->rcvhdrq, rcd->rcvhdrq_phys);
+ rcd->rcvhdrq = NULL;
+ if (rcd->rcvhdrtail_kvaddr) {
+ dma_free_coherent(&dd->pcidev->dev, PAGE_SIZE,
+ (void *)rcd->rcvhdrtail_kvaddr,
+ rcd->rcvhdrqtailaddr_phys);
+ rcd->rcvhdrtail_kvaddr = NULL;
+ }
+ }
+
+ /* all the RcvArray entries should have been cleared by now */
+ kfree(rcd->egrbufs.rcvtids);
+
+ for (e = 0; e < rcd->egrbufs.alloced; e++) {
+ if (rcd->egrbufs.buffers[e].phys)
+ dma_free_coherent(&dd->pcidev->dev,
+ rcd->egrbufs.buffers[e].len,
+ rcd->egrbufs.buffers[e].addr,
+ rcd->egrbufs.buffers[e].phys);
+ }
+ kfree(rcd->egrbufs.buffers);
+
+ sc_free(rcd->sc);
+ vfree(rcd->physshadow);
+ vfree(rcd->tid_pg_list);
+ vfree(rcd->user_event_mask);
+ vfree(rcd->subctxt_uregbase);
+ vfree(rcd->subctxt_rcvegrbuf);
+ vfree(rcd->subctxt_rcvhdr_base);
+ kfree(rcd->tidusemap);
+ kfree(rcd->opstats);
+ kfree(rcd);
+}
+
+void hfi1_free_devdata(struct hfi1_devdata *dd)
+{
+ unsigned long flags;
+
+ spin_lock_irqsave(&hfi1_devs_lock, flags);
+ idr_remove(&hfi1_unit_table, dd->unit);
+ list_del(&dd->list);
+ spin_unlock_irqrestore(&hfi1_devs_lock, flags);
+ hfi1_dbg_ibdev_exit(&dd->verbs_dev);
+ rcu_barrier(); /* wait for rcu callbacks to complete */
+ free_percpu(dd->int_counter);
+ ib_dealloc_device(&dd->verbs_dev.ibdev);
+}
+
+/*
+ * Allocate our primary per-unit data structure. Must be done via verbs
+ * allocator, because the verbs cleanup process both does cleanup and
+ * free of the data structure.
+ * "extra" is for chip-specific data.
+ *
+ * Use the idr mechanism to get a unit number for this unit.
+ */
+struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev, size_t extra)
+{
+ unsigned long flags;
+ struct hfi1_devdata *dd;
+ int ret;
+
+ dd = (struct hfi1_devdata *)ib_alloc_device(sizeof(*dd) + extra);
+ if (!dd)
+ return ERR_PTR(-ENOMEM);
+ /* extra is * number of ports */
+ dd->num_pports = extra / sizeof(struct hfi1_pportdata);
+ dd->pport = (struct hfi1_pportdata *)(dd + 1);
+
+ INIT_LIST_HEAD(&dd->list);
+ dd->node = dev_to_node(&pdev->dev);
+ if (dd->node < 0)
+ dd->node = 0;
+ idr_preload(GFP_KERNEL);
+ spin_lock_irqsave(&hfi1_devs_lock, flags);
+
+ ret = idr_alloc(&hfi1_unit_table, dd, 0, 0, GFP_NOWAIT);
+ if (ret >= 0) {
+ dd->unit = ret;
+ list_add(&dd->list, &hfi1_dev_list);
+ }
+
+ spin_unlock_irqrestore(&hfi1_devs_lock, flags);
+ idr_preload_end();
+
+ if (ret < 0) {
+ hfi1_early_err(&pdev->dev,
+ "Could not allocate unit ID: error %d\n", -ret);
+ goto bail;
+ }
+ /*
+ * Initialize all locks for the device. This needs to be as early as
+ * possible so locks are usable.
+ */
+ spin_lock_init(&dd->sc_lock);
+ spin_lock_init(&dd->sendctrl_lock);
+ spin_lock_init(&dd->rcvctrl_lock);
+ spin_lock_init(&dd->uctxt_lock);
+ spin_lock_init(&dd->hfi1_diag_trans_lock);
+ spin_lock_init(&dd->sc_init_lock);
+ spin_lock_init(&dd->dc8051_lock);
+ mutex_init(&dd->qsfp_i2c_mutex);
+ seqlock_init(&dd->sc2vl_lock);
+ spin_lock_init(&dd->sde_map_lock);
+ init_waitqueue_head(&dd->event_queue);
+
+ dd->int_counter = alloc_percpu(u64);
+ if (!dd->int_counter) {
+ ret = -ENOMEM;
+ hfi1_early_err(&pdev->dev,
+ "Could not allocate per-cpu int_counter\n");
+ goto bail;
+ }
+
+ if (!hfi1_cpulist_count) {
+ u32 count = num_online_cpus();
+
+ hfi1_cpulist = kzalloc(BITS_TO_LONGS(count) *
+ sizeof(long), GFP_KERNEL);
+ if (hfi1_cpulist)
+ hfi1_cpulist_count = count;
+ else
+ hfi1_early_err(
+ &pdev->dev,
+ "Could not alloc cpulist info, cpu affinity might be wrong\n");
+ }
+ hfi1_dbg_ibdev_init(&dd->verbs_dev);
+ return dd;
+
+bail:
+ if (!list_empty(&dd->list))
+ list_del_init(&dd->list);
+ ib_dealloc_device(&dd->verbs_dev.ibdev);
+ return ERR_PTR(ret);
+}
+
+/*
+ * Called from freeze mode handlers, and from PCI error
+ * reporting code. Should be paranoid about state of
+ * system and data structures.
+ */
+void hfi1_disable_after_error(struct hfi1_devdata *dd)
+{
+ if (dd->flags & HFI1_INITTED) {
+ u32 pidx;
+
+ dd->flags &= ~HFI1_INITTED;
+ if (dd->pport)
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ struct hfi1_pportdata *ppd;
+
+ ppd = dd->pport + pidx;
+ if (dd->flags & HFI1_PRESENT)
+ set_link_state(ppd, HLS_DN_DISABLE);
+
+ if (ppd->statusp)
+ *ppd->statusp &= ~HFI1_STATUS_IB_READY;
+ }
+ }
+
+ /*
+ * Mark as having had an error for driver, and also
+ * for /sys and status word mapped to user programs.
+ * This marks unit as not usable, until reset.
+ */
+ if (dd->status)
+ dd->status->dev |= HFI1_STATUS_HWERROR;
+}
+
+static void remove_one(struct pci_dev *);
+static int init_one(struct pci_dev *, const struct pci_device_id *);
+
+#define DRIVER_LOAD_MSG "Intel " DRIVER_NAME " loaded: "
+#define PFX DRIVER_NAME ": "
+
+static const struct pci_device_id hfi1_pci_tbl[] = {
+ { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL0) },
+ { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL1) },
+ { PCI_DEVICE(PCI_VENDOR_ID_INTEL, PCI_DEVICE_ID_INTEL2) },
+ { 0, }
+};
+
+MODULE_DEVICE_TABLE(pci, hfi1_pci_tbl);
+
+static struct pci_driver hfi1_pci_driver = {
+ .name = DRIVER_NAME,
+ .probe = init_one,
+ .remove = remove_one,
+ .id_table = hfi1_pci_tbl,
+ .err_handler = &hfi1_pci_err_handler,
+};
+
+static void __init compute_krcvqs(void)
+{
+ int i;
+
+ for (i = 0; i < krcvqsset; i++)
+ n_krcvqs += krcvqs[i];
+}
+
+/*
+ * Do all the generic driver unit- and chip-independent memory
+ * allocation and initialization.
+ */
+static int __init hfi1_mod_init(void)
+{
+ int ret;
+
+ ret = dev_init();
+ if (ret)
+ goto bail;
+
+ /* validate max MTU before any devices start */
+ if (!valid_opa_mtu(hfi1_max_mtu))
+ hfi1_max_mtu = HFI1_DEFAULT_MAX_MTU;
+ /* valid CUs run from 1-128 in powers of 2 */
+ if (hfi1_cu > 128 || !is_power_of_2(hfi1_cu))
+ hfi1_cu = 1;
+ /* valid credit return threshold is 0-100, variable is unsigned */
+ if (user_credit_return_threshold > 100)
+ user_credit_return_threshold = 100;
+
+ compute_krcvqs();
+ /* sanitize receive interrupt count, time must wait until after
+ the hardware type is known */
+ if (rcv_intr_count > RCV_HDR_HEAD_COUNTER_MASK)
+ rcv_intr_count = RCV_HDR_HEAD_COUNTER_MASK;
+ /* reject invalid combinations */
+ if (rcv_intr_count == 0 && rcv_intr_timeout == 0) {
+ pr_err("Invalid mode: both receive interrupt count and available timeout are zero - setting interrupt count to 1\n");
+ rcv_intr_count = 1;
+ }
+ if (rcv_intr_count > 1 && rcv_intr_timeout == 0) {
+ /*
+ * Avoid indefinite packet delivery by requiring a timeout
+ * if count is > 1.
+ */
+ pr_err("Invalid mode: receive interrupt count greater than 1 and available timeout is zero - setting available timeout to 1\n");
+ rcv_intr_timeout = 1;
+ }
+ if (rcv_intr_dynamic && !(rcv_intr_count > 1 && rcv_intr_timeout > 0)) {
+ /*
+ * The dynamic algorithm expects a non-zero timeout
+ * and a count > 1.
+ */
+ pr_err("Invalid mode: dynamic receive interrupt mitigation with invalid count and timeout - turning dynamic off\n");
+ rcv_intr_dynamic = 0;
+ }
+
+ /* sanitize link CRC options */
+ link_crc_mask &= SUPPORTED_CRCS;
+
+ /*
+ * These must be called before the driver is registered with
+ * the PCI subsystem.
+ */
+ idr_init(&hfi1_unit_table);
+
+ hfi1_dbg_init();
+ ret = pci_register_driver(&hfi1_pci_driver);
+ if (ret < 0) {
+ pr_err("Unable to register driver: error %d\n", -ret);
+ goto bail_dev;
+ }
+ goto bail; /* all OK */
+
+bail_dev:
+ hfi1_dbg_exit();
+ idr_destroy(&hfi1_unit_table);
+ dev_cleanup();
+bail:
+ return ret;
+}
+
+module_init(hfi1_mod_init);
+
+/*
+ * Do the non-unit driver cleanup, memory free, etc. at unload.
+ */
+static void __exit hfi1_mod_cleanup(void)
+{
+ pci_unregister_driver(&hfi1_pci_driver);
+ hfi1_dbg_exit();
+ hfi1_cpulist_count = 0;
+ kfree(hfi1_cpulist);
+
+ idr_destroy(&hfi1_unit_table);
+ dispose_firmware(); /* asymmetric with obtain_firmware() */
+ dev_cleanup();
+}
+
+module_exit(hfi1_mod_cleanup);
+
+/* this can only be called after a successful initialization */
+static void cleanup_device_data(struct hfi1_devdata *dd)
+{
+ int ctxt;
+ int pidx;
+ struct hfi1_ctxtdata **tmp;
+ unsigned long flags;
+
+ /* users can't do anything more with chip */
+ for (pidx = 0; pidx < dd->num_pports; ++pidx) {
+ struct hfi1_pportdata *ppd = &dd->pport[pidx];
+ struct cc_state *cc_state;
+ int i;
+
+ if (ppd->statusp)
+ *ppd->statusp &= ~HFI1_STATUS_CHIP_PRESENT;
+
+ for (i = 0; i < OPA_MAX_SLS; i++)
+ hrtimer_cancel(&ppd->cca_timer[i].hrtimer);
+
+ spin_lock(&ppd->cc_state_lock);
+ cc_state = get_cc_state(ppd);
+ rcu_assign_pointer(ppd->cc_state, NULL);
+ spin_unlock(&ppd->cc_state_lock);
+
+ if (cc_state)
+ call_rcu(&cc_state->rcu, cc_state_reclaim);
+ }
+
+ free_credit_return(dd);
+
+ /*
+ * Free any resources still in use (usually just kernel contexts)
+ * at unload; we do for ctxtcnt, because that's what we allocate.
+ * We acquire lock to be really paranoid that rcd isn't being
+ * accessed from some interrupt-related code (that should not happen,
+ * but best to be sure).
+ */
+ spin_lock_irqsave(&dd->uctxt_lock, flags);
+ tmp = dd->rcd;
+ dd->rcd = NULL;
+ spin_unlock_irqrestore(&dd->uctxt_lock, flags);
+ for (ctxt = 0; tmp && ctxt < dd->num_rcv_contexts; ctxt++) {
+ struct hfi1_ctxtdata *rcd = tmp[ctxt];
+
+ tmp[ctxt] = NULL; /* debugging paranoia */
+ if (rcd) {
+ hfi1_clear_tids(rcd);
+ hfi1_free_ctxtdata(dd, rcd);
+ }
+ }
+ kfree(tmp);
+ /* must follow rcv context free - need to remove rcv's hooks */
+ for (ctxt = 0; ctxt < dd->num_send_contexts; ctxt++)
+ sc_free(dd->send_contexts[ctxt].sc);
+ dd->num_send_contexts = 0;
+ kfree(dd->send_contexts);
+ dd->send_contexts = NULL;
+ kfree(dd->boardname);
+ vfree(dd->events);
+ vfree(dd->status);
+ hfi1_cq_exit(dd);
+}
+
+/*
+ * Clean up on unit shutdown, or error during unit load after
+ * successful initialization.
+ */
+static void postinit_cleanup(struct hfi1_devdata *dd)
+{
+ hfi1_start_cleanup(dd);
+
+ hfi1_pcie_ddcleanup(dd);
+ hfi1_pcie_cleanup(dd->pcidev);
+
+ cleanup_device_data(dd);
+
+ hfi1_free_devdata(dd);
+}
+
+static int init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+ int ret = 0, j, pidx, initfail;
+ struct hfi1_devdata *dd = NULL;
+
+ /* First, lock the non-writable module parameters */
+ HFI1_CAP_LOCK();
+
+ /* Validate some global module parameters */
+ if (rcvhdrcnt <= HFI1_MIN_HDRQ_EGRBUF_CNT) {
+ hfi1_early_err(&pdev->dev, "Header queue count too small\n");
+ ret = -EINVAL;
+ goto bail;
+ }
+ /* use the encoding function as a sanitization check */
+ if (!encode_rcv_header_entry_size(hfi1_hdrq_entsize)) {
+ hfi1_early_err(&pdev->dev, "Invalid HdrQ Entry size %u\n",
+ hfi1_hdrq_entsize);
+ goto bail;
+ }
+
+ /* The receive eager buffer size must be set before the receive
+ * contexts are created.
+ *
+ * Set the eager buffer size. Validate that it falls in a range
+ * allowed by the hardware - all powers of 2 between the min and
+ * max. The maximum valid MTU is within the eager buffer range
+ * so we do not need to cap the max_mtu by an eager buffer size
+ * setting.
+ */
+ if (eager_buffer_size) {
+ if (!is_power_of_2(eager_buffer_size))
+ eager_buffer_size =
+ roundup_pow_of_two(eager_buffer_size);
+ eager_buffer_size =
+ clamp_val(eager_buffer_size,
+ MIN_EAGER_BUFFER * 8,
+ MAX_EAGER_BUFFER_TOTAL);
+ hfi1_early_info(&pdev->dev, "Eager buffer size %u\n",
+ eager_buffer_size);
+ } else {
+ hfi1_early_err(&pdev->dev, "Invalid Eager buffer size of 0\n");
+ ret = -EINVAL;
+ goto bail;
+ }
+
+ /* restrict value of hfi1_rcvarr_split */
+ hfi1_rcvarr_split = clamp_val(hfi1_rcvarr_split, 0, 100);
+
+ ret = hfi1_pcie_init(pdev, ent);
+ if (ret)
+ goto bail;
+
+ /*
+ * Do device-specific initialization, function table setup, dd
+ * allocation, etc.
+ */
+ switch (ent->device) {
+ case PCI_DEVICE_ID_INTEL0:
+ case PCI_DEVICE_ID_INTEL1:
+ case PCI_DEVICE_ID_INTEL2:
+ dd = hfi1_init_dd(pdev, ent);
+ break;
+ default:
+ hfi1_early_err(&pdev->dev,
+ "Failing on unknown Intel deviceid 0x%x\n",
+ ent->device);
+ ret = -ENODEV;
+ }
+
+ if (IS_ERR(dd))
+ ret = PTR_ERR(dd);
+ if (ret)
+ goto clean_bail; /* error already printed */
+
+ ret = create_workqueues(dd);
+ if (ret)
+ goto clean_bail;
+
+ /* do the generic initialization */
+ initfail = hfi1_init(dd, 0);
+
+ ret = hfi1_register_ib_device(dd);
+
+ /*
+ * Now ready for use. this should be cleared whenever we
+ * detect a reset, or initiate one. If earlier failure,
+ * we still create devices, so diags, etc. can be used
+ * to determine cause of problem.
+ */
+ if (!initfail && !ret)
+ dd->flags |= HFI1_INITTED;
+
+ j = hfi1_device_create(dd);
+ if (j)
+ dd_dev_err(dd, "Failed to create /dev devices: %d\n", -j);
+
+ if (initfail || ret) {
+ stop_timers(dd);
+ flush_workqueue(ib_wq);
+ for (pidx = 0; pidx < dd->num_pports; ++pidx)
+ hfi1_quiet_serdes(dd->pport + pidx);
+ if (!j)
+ hfi1_device_remove(dd);
+ if (!ret)
+ hfi1_unregister_ib_device(dd);
+ postinit_cleanup(dd);
+ if (initfail)
+ ret = initfail;
+ goto bail; /* everything already cleaned */
+ }
+
+ sdma_start(dd);
+
+ return 0;
+
+clean_bail:
+ hfi1_pcie_cleanup(pdev);
+bail:
+ return ret;
+}
+
+static void remove_one(struct pci_dev *pdev)
+{
+ struct hfi1_devdata *dd = pci_get_drvdata(pdev);
+
+ /* unregister from IB core */
+ hfi1_unregister_ib_device(dd);
+
+ /*
+ * Disable the IB link, disable interrupts on the device,
+ * clear dma engines, etc.
+ */
+ shutdown_device(dd);
+
+ stop_timers(dd);
+
+ /* wait until all of our (qsfp) queue_work() calls complete */
+ flush_workqueue(ib_wq);
+
+ hfi1_device_remove(dd);
+
+ postinit_cleanup(dd);
+}
+
+/**
+ * hfi1_create_rcvhdrq - create a receive header queue
+ * @dd: the hfi1_ib device
+ * @rcd: the context data
+ *
+ * This must be contiguous memory (from an i/o perspective), and must be
+ * DMA'able (which means for some systems, it will go through an IOMMU,
+ * or be forced into a low address range).
+ */
+int hfi1_create_rcvhdrq(struct hfi1_devdata *dd, struct hfi1_ctxtdata *rcd)
+{
+ unsigned amt;
+ u64 reg;
+
+ if (!rcd->rcvhdrq) {
+ dma_addr_t phys_hdrqtail;
+ gfp_t gfp_flags;
+
+ /*
+ * rcvhdrqentsize is in DWs, so we have to convert to bytes
+ * (* sizeof(u32)).
+ */
+ amt = ALIGN(rcd->rcvhdrq_cnt * rcd->rcvhdrqentsize *
+ sizeof(u32), PAGE_SIZE);
+
+ gfp_flags = (rcd->ctxt >= dd->first_user_ctxt) ?
+ GFP_USER : GFP_KERNEL;
+ rcd->rcvhdrq = dma_zalloc_coherent(
+ &dd->pcidev->dev, amt, &rcd->rcvhdrq_phys,
+ gfp_flags | __GFP_COMP);
+
+ if (!rcd->rcvhdrq) {
+ dd_dev_err(dd,
+ "attempt to allocate %d bytes for ctxt %u rcvhdrq failed\n",
+ amt, rcd->ctxt);
+ goto bail;
+ }
+
+ /* Event mask is per device now and is in hfi1_devdata */
+ /*if (rcd->ctxt >= dd->first_user_ctxt) {
+ rcd->user_event_mask = vmalloc_user(PAGE_SIZE);
+ if (!rcd->user_event_mask)
+ goto bail_free_hdrq;
+ }*/
+
+ if (HFI1_CAP_KGET_MASK(rcd->flags, DMA_RTAIL)) {
+ rcd->rcvhdrtail_kvaddr = dma_zalloc_coherent(
+ &dd->pcidev->dev, PAGE_SIZE, &phys_hdrqtail,
+ gfp_flags);
+ if (!rcd->rcvhdrtail_kvaddr)
+ goto bail_free;
+ rcd->rcvhdrqtailaddr_phys = phys_hdrqtail;
+ }
+
+ rcd->rcvhdrq_size = amt;
+ }
+ /*
+ * These values are per-context:
+ * RcvHdrCnt
+ * RcvHdrEntSize
+ * RcvHdrSize
+ */
+ reg = ((u64)(rcd->rcvhdrq_cnt >> HDRQ_SIZE_SHIFT)
+ & RCV_HDR_CNT_CNT_MASK)
+ << RCV_HDR_CNT_CNT_SHIFT;
+ write_kctxt_csr(dd, rcd->ctxt, RCV_HDR_CNT, reg);
+ reg = (encode_rcv_header_entry_size(rcd->rcvhdrqentsize)
+ & RCV_HDR_ENT_SIZE_ENT_SIZE_MASK)
+ << RCV_HDR_ENT_SIZE_ENT_SIZE_SHIFT;
+ write_kctxt_csr(dd, rcd->ctxt, RCV_HDR_ENT_SIZE, reg);
+ reg = (dd->rcvhdrsize & RCV_HDR_SIZE_HDR_SIZE_MASK)
+ << RCV_HDR_SIZE_HDR_SIZE_SHIFT;
+ write_kctxt_csr(dd, rcd->ctxt, RCV_HDR_SIZE, reg);
+ return 0;
+
+bail_free:
+ dd_dev_err(dd,
+ "attempt to allocate 1 page for ctxt %u rcvhdrqtailaddr failed\n",
+ rcd->ctxt);
+ vfree(rcd->user_event_mask);
+ rcd->user_event_mask = NULL;
+ dma_free_coherent(&dd->pcidev->dev, amt, rcd->rcvhdrq,
+ rcd->rcvhdrq_phys);
+ rcd->rcvhdrq = NULL;
+bail:
+ return -ENOMEM;
+}
+
+/**
+ * allocate eager buffers, both kernel and user contexts.
+ * @rcd: the context we are setting up.
+ *
+ * Allocate the eager TID buffers and program them into hip.
+ * They are no longer completely contiguous, we do multiple allocation
+ * calls. Otherwise we get the OOM code involved, by asking for too
+ * much per call, with disastrous results on some kernels.
+ */
+int hfi1_setup_eagerbufs(struct hfi1_ctxtdata *rcd)
+{
+ struct hfi1_devdata *dd = rcd->dd;
+ u32 max_entries, egrtop, alloced_bytes = 0, idx = 0;
+ gfp_t gfp_flags;
+ u16 order;
+ int ret = 0;
+ u16 round_mtu = roundup_pow_of_two(hfi1_max_mtu);
+
+ /*
+ * GFP_USER, but without GFP_FS, so buffer cache can be
+ * coalesced (we hope); otherwise, even at order 4,
+ * heavy filesystem activity makes these fail, and we can
+ * use compound pages.
+ */
+ gfp_flags = __GFP_WAIT | __GFP_IO | __GFP_COMP;
+
+ /*
+ * The minimum size of the eager buffers is a groups of MTU-sized
+ * buffers.
+ * The global eager_buffer_size parameter is checked against the
+ * theoretical lower limit of the value. Here, we check against the
+ * MTU.
+ */
+ if (rcd->egrbufs.size < (round_mtu * dd->rcv_entries.group_size))
+ rcd->egrbufs.size = round_mtu * dd->rcv_entries.group_size;
+ /*
+ * If using one-pkt-per-egr-buffer, lower the eager buffer
+ * size to the max MTU (page-aligned).
+ */
+ if (!HFI1_CAP_KGET_MASK(rcd->flags, MULTI_PKT_EGR))
+ rcd->egrbufs.rcvtid_size = round_mtu;
+
+ /*
+ * Eager buffers sizes of 1MB or less require smaller TID sizes
+ * to satisfy the "multiple of 8 RcvArray entries" requirement.
+ */
+ if (rcd->egrbufs.size <= (1 << 20))
+ rcd->egrbufs.rcvtid_size = max((unsigned long)round_mtu,
+ rounddown_pow_of_two(rcd->egrbufs.size / 8));
+
+ while (alloced_bytes < rcd->egrbufs.size &&
+ rcd->egrbufs.alloced < rcd->egrbufs.count) {
+ rcd->egrbufs.buffers[idx].addr =
+ dma_zalloc_coherent(&dd->pcidev->dev,
+ rcd->egrbufs.rcvtid_size,
+ &rcd->egrbufs.buffers[idx].phys,
+ gfp_flags);
+ if (rcd->egrbufs.buffers[idx].addr) {
+ rcd->egrbufs.buffers[idx].len =
+ rcd->egrbufs.rcvtid_size;
+ rcd->egrbufs.rcvtids[rcd->egrbufs.alloced].addr =
+ rcd->egrbufs.buffers[idx].addr;
+ rcd->egrbufs.rcvtids[rcd->egrbufs.alloced].phys =
+ rcd->egrbufs.buffers[idx].phys;
+ rcd->egrbufs.alloced++;
+ alloced_bytes += rcd->egrbufs.rcvtid_size;
+ idx++;
+ } else {
+ u32 new_size, i, j;
+ u64 offset = 0;
+
+ /*
+ * Fail the eager buffer allocation if:
+ * - we are already using the lowest acceptable size
+ * - we are using one-pkt-per-egr-buffer (this implies
+ * that we are accepting only one size)
+ */
+ if (rcd->egrbufs.rcvtid_size == round_mtu ||
+ !HFI1_CAP_KGET_MASK(rcd->flags, MULTI_PKT_EGR)) {
+ dd_dev_err(dd, "ctxt%u: Failed to allocate eager buffers\n",
+ rcd->ctxt);
+ goto bail_rcvegrbuf_phys;
+ }
+
+ new_size = rcd->egrbufs.rcvtid_size / 2;
+
+ /*
+ * If the first attempt to allocate memory failed, don't
+ * fail everything but continue with the next lower
+ * size.
+ */
+ if (idx == 0) {
+ rcd->egrbufs.rcvtid_size = new_size;
+ continue;
+ }
+
+ /*
+ * Re-partition already allocated buffers to a smaller
+ * size.
+ */
+ rcd->egrbufs.alloced = 0;
+ for (i = 0, j = 0, offset = 0; j < idx; i++) {
+ if (i >= rcd->egrbufs.count)
+ break;
+ rcd->egrbufs.rcvtids[i].phys =
+ rcd->egrbufs.buffers[j].phys + offset;
+ rcd->egrbufs.rcvtids[i].addr =
+ rcd->egrbufs.buffers[j].addr + offset;
+ rcd->egrbufs.alloced++;
+ if ((rcd->egrbufs.buffers[j].phys + offset +
+ new_size) ==
+ (rcd->egrbufs.buffers[j].phys +
+ rcd->egrbufs.buffers[j].len)) {
+ j++;
+ offset = 0;
+ } else
+ offset += new_size;
+ }
+ rcd->egrbufs.rcvtid_size = new_size;
+ }
+ }
+ rcd->egrbufs.numbufs = idx;
+ rcd->egrbufs.size = alloced_bytes;
+
+ dd_dev_info(dd, "ctxt%u: Alloced %u rcv tid entries @ %uKB, total %zuKB\n",
+ rcd->ctxt, rcd->egrbufs.alloced, rcd->egrbufs.rcvtid_size,
+ rcd->egrbufs.size);
+
+ /*
+ * Set the contexts rcv array head update threshold to the closest
+ * power of 2 (so we can use a mask instead of modulo) below half
+ * the allocated entries.
+ */
+ rcd->egrbufs.threshold =
+ rounddown_pow_of_two(rcd->egrbufs.alloced / 2);
+ /*
+ * Compute the expected RcvArray entry base. This is done after
+ * allocating the eager buffers in order to maximize the
+ * expected RcvArray entries for the context.
+ */
+ max_entries = rcd->rcv_array_groups * dd->rcv_entries.group_size;
+ egrtop = roundup(rcd->egrbufs.alloced, dd->rcv_entries.group_size);
+ rcd->expected_count = max_entries - egrtop;
+ if (rcd->expected_count > MAX_TID_PAIR_ENTRIES * 2)
+ rcd->expected_count = MAX_TID_PAIR_ENTRIES * 2;
+
+ rcd->expected_base = rcd->eager_base + egrtop;
+ dd_dev_info(dd, "ctxt%u: eager:%u, exp:%u, egrbase:%u, expbase:%u\n",
+ rcd->ctxt, rcd->egrbufs.alloced, rcd->expected_count,
+ rcd->eager_base, rcd->expected_base);
+
+ if (!hfi1_rcvbuf_validate(rcd->egrbufs.rcvtid_size, PT_EAGER, &order)) {
+ dd_dev_err(dd, "ctxt%u: current Eager buffer size is invalid %u\n",
+ rcd->ctxt, rcd->egrbufs.rcvtid_size);
+ ret = -EINVAL;
+ goto bail;
+ }
+
+ for (idx = 0; idx < rcd->egrbufs.alloced; idx++) {
+ hfi1_put_tid(dd, rcd->eager_base + idx, PT_EAGER,
+ rcd->egrbufs.rcvtids[idx].phys, order);
+ cond_resched();
+ }
+ goto bail;
+
+bail_rcvegrbuf_phys:
+ for (idx = 0; idx < rcd->egrbufs.alloced &&
+ rcd->egrbufs.buffers[idx].addr;
+ idx++) {
+ dma_free_coherent(&dd->pcidev->dev,
+ rcd->egrbufs.buffers[idx].len,
+ rcd->egrbufs.buffers[idx].addr,
+ rcd->egrbufs.buffers[idx].phys);
+ rcd->egrbufs.buffers[idx].addr = NULL;
+ rcd->egrbufs.buffers[idx].phys = 0;
+ rcd->egrbufs.buffers[idx].len = 0;
+ }
+bail:
+ return ret;
+}
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
next prev parent reply other threads:[~2015-06-15 13:27 UTC|newest]
Thread overview: 64+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-06-15 13:25 [PATCH v2 00/49] Add OPA gen1 driver Mike Marciniszyn
[not found] ` <20150615132434.29741.35258.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
2015-06-15 13:26 ` [PATCH v2 01/49] IB/core: Add OPA Port header definitions Mike Marciniszyn
[not found] ` <20150615132603.29741.75037.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
2015-06-15 17:11 ` Hefty, Sean
[not found] ` <1828884A29C6694DAF28B7E6B8A82373A8FF5771-P5GAC/sN6hkd3b2yrw5b5LfspsVTdybXVpNB7YpNyf8@public.gmane.org>
2015-06-17 12:08 ` Marciniszyn, Mike
[not found] ` <32E1700B9017364D9B60AED9960492BC2575145F-RjuIdWtd+YbTXloPLtfHfbfspsVTdybXVpNB7YpNyf8@public.gmane.org>
2015-06-17 12:12 ` Hal Rosenstock
[not found] ` <55816432.906-LDSdmyG8hGV8YrgS2mwiifqBs+8SCbDb@public.gmane.org>
2015-06-17 15:32 ` Hefty, Sean
2015-06-15 13:26 ` [PATCH v2 02/49] IB/hfi1: add chip specific support part1 Mike Marciniszyn
2015-06-15 13:26 ` [PATCH v2 03/49] IB/hfi1: add chip specific support part2 Mike Marciniszyn
2015-06-15 13:26 ` [PATCH v2 04/49] IB/hfi1: add chip specific support part3 Mike Marciniszyn
2015-06-15 13:26 ` [PATCH v2 05/49] IB/hfi1: add chip specific support part4 Mike Marciniszyn
2015-06-15 13:26 ` [PATCH v2 06/49] IB/hfi1: add chip register definitions Mike Marciniszyn
2015-06-15 13:26 ` [PATCH v2 07/49] IB/hfi1: add chip specific headers Mike Marciniszyn
2015-06-15 13:26 ` [PATCH v2 08/49] IB/hfi1: add common header file definitions Mike Marciniszyn
2015-06-15 13:26 ` [PATCH v2 09/49] IB/hfi1: add completion queue processing Mike Marciniszyn
2015-06-15 13:26 ` [PATCH v2 10/49] IB/hfi1: add debugfs handling Mike Marciniszyn
2015-06-15 13:26 ` [PATCH v2 11/49] IB/hfi1: add char device instantiation code Mike Marciniszyn
2015-06-15 13:27 ` [PATCH v2 12/49] IB/hfi1: add diagnostic hooks Mike Marciniszyn
2015-06-15 13:27 ` [PATCH v2 13/49] IB/hfi1: add dma operation hooks Mike Marciniszyn
2015-06-15 13:27 ` [PATCH v2 14/49] IB/hfi1: add low lower receive functions Mike Marciniszyn
2015-06-15 13:27 ` [PATCH v2 15/49] IB/hfi1: add eeprom hooks Mike Marciniszyn
2015-06-15 13:27 ` [PATCH v2 16/49] IB/hfi1: add PSM driver control/data path Mike Marciniszyn
2015-06-15 13:27 ` [PATCH v2 17/49] IB/hfi1: add firmware hooks Mike Marciniszyn
2015-06-15 13:27 ` [PATCH v2 18/49] IB/hfi1: add general hfi header file Mike Marciniszyn
2015-06-15 13:27 ` Mike Marciniszyn [this message]
2015-06-15 13:27 ` [PATCH v2 20/49] IB/hfi1: add interrupt hooks Mike Marciniszyn
2015-06-15 13:27 ` [PATCH v2 21/49] IB/hfi1: add progress delay/restart hooks Mike Marciniszyn
2015-06-15 13:27 ` [PATCH v2 22/49] IB/hfi1: add rkey/lkey validation Mike Marciniszyn
2015-06-15 13:28 ` [PATCH v2 23/49] IB/hfi1: add OPA mad handling part1 Mike Marciniszyn
2015-06-15 13:28 ` [PATCH v2 24/49] IB/hfi1: add OPA mad handling part2 Mike Marciniszyn
2015-06-15 13:28 ` [PATCH v2 25/49] IB/hfi1: add local mad header Mike Marciniszyn
[not found] ` <20150615132813.29741.3528.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
2015-06-15 23:27 ` Hefty, Sean
[not found] ` <1828884A29C6694DAF28B7E6B8A82373A8FF5B18-P5GAC/sN6hkd3b2yrw5b5LfspsVTdybXVpNB7YpNyf8@public.gmane.org>
2015-06-17 11:51 ` Marciniszyn, Mike
2015-06-15 13:28 ` [PATCH v2 26/49] IB/hfi1: add user/kernel memory sharing hooks Mike Marciniszyn
2015-06-15 13:28 ` [PATCH v2 27/49] IB/hfi1: add memory region handling Mike Marciniszyn
2015-06-15 13:28 ` [PATCH v2 28/49] IB/hfi1: add misc OPA defines Mike Marciniszyn
2015-06-15 13:28 ` [PATCH v2 29/49] IB/hfi1: add pcie routines Mike Marciniszyn
2015-06-15 13:28 ` [PATCH v2 30/49] IB/hfi1: add pio handling Mike Marciniszyn
2015-06-15 13:28 ` [PATCH v2 31/49] IB/hfi1: add platform config definitions Mike Marciniszyn
2015-06-15 13:28 ` [PATCH v2 32/49] IB/hfi1: add qp handling Mike Marciniszyn
2015-06-15 13:28 ` [PATCH v2 33/49] IB/hfi1: add qsfp handling Mike Marciniszyn
2015-06-15 13:29 ` [PATCH v2 34/49] IB/hfi1: add RC QP handling Mike Marciniszyn
2015-06-15 13:29 ` [PATCH v2 35/49] IB/hfi1: add routines for RC/UC Mike Marciniszyn
2015-06-15 13:29 ` [PATCH v2 36/49] IB/hfi1: add sdma routines Mike Marciniszyn
2015-06-15 13:29 ` [PATCH v2 37/49] IB/hfi1: add sdma header file Mike Marciniszyn
2015-06-15 13:29 ` [PATCH v2 38/49] IB/hfi1: add SRQ handling Mike Marciniszyn
2015-06-15 13:29 ` [PATCH v2 39/49] IB/hfi1: add sysfs routines Mike Marciniszyn
[not found] ` <20150615132929.29741.87559.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
2015-06-15 19:14 ` Or Gerlitz
[not found] ` <CAJ3xEMiiqY04q4pnjRSm=7UBnS_jQJP-KhAkYEvu7FAu-F98sA-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2015-06-15 19:20 ` Or Gerlitz
[not found] ` <CAJ3xEMiJ_0pFUvPJN7ov50v0hOZVsN7qC27ZbG5+pfmVuZQuMw-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2015-06-17 11:48 ` Marciniszyn, Mike
[not found] ` <32E1700B9017364D9B60AED9960492BC257513D9-RjuIdWtd+YbTXloPLtfHfbfspsVTdybXVpNB7YpNyf8@public.gmane.org>
2015-06-17 15:50 ` Or Gerlitz
[not found] ` <CAJ3xEMjdhyk2OOfeCV4sZMF=Fh06H-hJn_2y85V0eBLYBBgPiA-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org>
2015-07-08 22:32 ` Marciniszyn, Mike
[not found] ` <32E1700B9017364D9B60AED9960492BC2575DB49-RjuIdWtd+YbTXloPLtfHfbfspsVTdybXVpNB7YpNyf8@public.gmane.org>
2015-07-08 23:22 ` ira.weiny
2015-06-15 13:29 ` [PATCH v2 40/49] IB/hfi1: add tracepoint debug routines Mike Marciniszyn
2015-06-15 13:29 ` [PATCH v2 41/49] IB/hfi1: add QSFP twsi routines Mike Marciniszyn
2015-06-15 13:29 ` [PATCH v2 42/49] IB/hfi1: add UC QP handling Mike Marciniszyn
2015-06-15 13:29 ` [PATCH v2 43/49] IB/hfi1: add UD " Mike Marciniszyn
2015-06-15 13:29 ` [PATCH v2 44/49] IB/hfi1: add low level page locking Mike Marciniszyn
2015-06-15 13:30 ` [PATCH v2 45/49] IB/hfi1: add PSM sdma hooks Mike Marciniszyn
2015-06-15 13:30 ` [PATCH v2 46/49] IB/hfi1: add general verbs handling Mike Marciniszyn
2015-06-15 13:30 ` [PATCH v2 47/49] IB/hfi1: add multicast routines Mike Marciniszyn
[not found] ` <20150615133012.29741.61892.stgit-K+u1se/DcYrLESAwzcoQNrvm/XP+8Wra@public.gmane.org>
2015-06-15 17:24 ` Christoph Lameter
[not found] ` <alpine.DEB.2.11.1506151222560.20941-wcBtFHqTun5QOdAKl3ChDw@public.gmane.org>
2015-06-17 12:14 ` Marciniszyn, Mike
2015-06-15 13:30 ` [PATCH v2 48/49] IB/hfi1: add driver make/config files Mike Marciniszyn
2015-06-15 13:30 ` [PATCH v2 49/49] IB/core: Add opa driver to kbuild Mike Marciniszyn
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20150615132741.29741.83775.stgit@phlsvslse11.ph.intel.com \
--to=mike.marciniszyn-ral2jqcrhueavxtiumwx3w@public.gmane.org \
--cc=dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org \
--cc=linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.