linux_dsm_epyc7002/drivers/net/mlx4/eq.c
Roland Dreier 3d73c2884f mlx4_core: Fix section mismatches
Commit ee49bd93 ("mlx4_core: Reset device when internal error is
detected") introduced some section mismatch problems when
CONFIG_HOTPLUG=n, because the error recovery code tears down and
reinitializes the device after everything is loaded, which ends up
calling into lots of code marked __devinit and __devexit from regular
.text.  Fix this by getting rid of these now-incorrect section
markers.

Signed-off-by: Roland Dreier <rolandd@cisco.com>
2007-10-10 15:43:54 -07:00

657 lines
17 KiB
C

/*
* Copyright (c) 2005 Mellanox Technologies. All rights reserved.
* Copyright (c) 2005, 2006, 2007 Cisco Systems, 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/init.h>
#include <linux/interrupt.h>
#include <linux/dma-mapping.h>
#include <linux/mlx4/cmd.h>
#include "mlx4.h"
#include "fw.h"
enum {
MLX4_NUM_ASYNC_EQE = 0x100,
MLX4_NUM_SPARE_EQE = 0x80,
MLX4_EQ_ENTRY_SIZE = 0x20
};
/*
* Must be packed because start is 64 bits but only aligned to 32 bits.
*/
struct mlx4_eq_context {
__be32 flags;
u16 reserved1[3];
__be16 page_offset;
u8 log_eq_size;
u8 reserved2[4];
u8 eq_period;
u8 reserved3;
u8 eq_max_count;
u8 reserved4[3];
u8 intr;
u8 log_page_size;
u8 reserved5[2];
u8 mtt_base_addr_h;
__be32 mtt_base_addr_l;
u32 reserved6[2];
__be32 consumer_index;
__be32 producer_index;
u32 reserved7[4];
};
#define MLX4_EQ_STATUS_OK ( 0 << 28)
#define MLX4_EQ_STATUS_WRITE_FAIL (10 << 28)
#define MLX4_EQ_OWNER_SW ( 0 << 24)
#define MLX4_EQ_OWNER_HW ( 1 << 24)
#define MLX4_EQ_FLAG_EC ( 1 << 18)
#define MLX4_EQ_FLAG_OI ( 1 << 17)
#define MLX4_EQ_STATE_ARMED ( 9 << 8)
#define MLX4_EQ_STATE_FIRED (10 << 8)
#define MLX4_EQ_STATE_ALWAYS_ARMED (11 << 8)
#define MLX4_ASYNC_EVENT_MASK ((1ull << MLX4_EVENT_TYPE_PATH_MIG) | \
(1ull << MLX4_EVENT_TYPE_COMM_EST) | \
(1ull << MLX4_EVENT_TYPE_SQ_DRAINED) | \
(1ull << MLX4_EVENT_TYPE_CQ_ERROR) | \
(1ull << MLX4_EVENT_TYPE_WQ_CATAS_ERROR) | \
(1ull << MLX4_EVENT_TYPE_EEC_CATAS_ERROR) | \
(1ull << MLX4_EVENT_TYPE_PATH_MIG_FAILED) | \
(1ull << MLX4_EVENT_TYPE_WQ_INVAL_REQ_ERROR) | \
(1ull << MLX4_EVENT_TYPE_WQ_ACCESS_ERROR) | \
(1ull << MLX4_EVENT_TYPE_PORT_CHANGE) | \
(1ull << MLX4_EVENT_TYPE_ECC_DETECT) | \
(1ull << MLX4_EVENT_TYPE_SRQ_CATAS_ERROR) | \
(1ull << MLX4_EVENT_TYPE_SRQ_QP_LAST_WQE) | \
(1ull << MLX4_EVENT_TYPE_SRQ_LIMIT) | \
(1ull << MLX4_EVENT_TYPE_CMD))
struct mlx4_eqe {
u8 reserved1;
u8 type;
u8 reserved2;
u8 subtype;
union {
u32 raw[6];
struct {
__be32 cqn;
} __attribute__((packed)) comp;
struct {
u16 reserved1;
__be16 token;
u32 reserved2;
u8 reserved3[3];
u8 status;
__be64 out_param;
} __attribute__((packed)) cmd;
struct {
__be32 qpn;
} __attribute__((packed)) qp;
struct {
__be32 srqn;
} __attribute__((packed)) srq;
struct {
__be32 cqn;
u32 reserved1;
u8 reserved2[3];
u8 syndrome;
} __attribute__((packed)) cq_err;
struct {
u32 reserved1[2];
__be32 port;
} __attribute__((packed)) port_change;
} event;
u8 reserved3[3];
u8 owner;
} __attribute__((packed));
static void eq_set_ci(struct mlx4_eq *eq, int req_not)
{
__raw_writel((__force u32) cpu_to_be32((eq->cons_index & 0xffffff) |
req_not << 31),
eq->doorbell);
/* We still want ordering, just not swabbing, so add a barrier */
mb();
}
static struct mlx4_eqe *get_eqe(struct mlx4_eq *eq, u32 entry)
{
unsigned long off = (entry & (eq->nent - 1)) * MLX4_EQ_ENTRY_SIZE;
return eq->page_list[off / PAGE_SIZE].buf + off % PAGE_SIZE;
}
static struct mlx4_eqe *next_eqe_sw(struct mlx4_eq *eq)
{
struct mlx4_eqe *eqe = get_eqe(eq, eq->cons_index);
return !!(eqe->owner & 0x80) ^ !!(eq->cons_index & eq->nent) ? NULL : eqe;
}
static int mlx4_eq_int(struct mlx4_dev *dev, struct mlx4_eq *eq)
{
struct mlx4_eqe *eqe;
int cqn;
int eqes_found = 0;
int set_ci = 0;
while ((eqe = next_eqe_sw(eq))) {
/*
* Make sure we read EQ entry contents after we've
* checked the ownership bit.
*/
rmb();
switch (eqe->type) {
case MLX4_EVENT_TYPE_COMP:
cqn = be32_to_cpu(eqe->event.comp.cqn) & 0xffffff;
mlx4_cq_completion(dev, cqn);
break;
case MLX4_EVENT_TYPE_PATH_MIG:
case MLX4_EVENT_TYPE_COMM_EST:
case MLX4_EVENT_TYPE_SQ_DRAINED:
case MLX4_EVENT_TYPE_SRQ_QP_LAST_WQE:
case MLX4_EVENT_TYPE_WQ_CATAS_ERROR:
case MLX4_EVENT_TYPE_PATH_MIG_FAILED:
case MLX4_EVENT_TYPE_WQ_INVAL_REQ_ERROR:
case MLX4_EVENT_TYPE_WQ_ACCESS_ERROR:
mlx4_qp_event(dev, be32_to_cpu(eqe->event.qp.qpn) & 0xffffff,
eqe->type);
break;
case MLX4_EVENT_TYPE_SRQ_LIMIT:
case MLX4_EVENT_TYPE_SRQ_CATAS_ERROR:
mlx4_srq_event(dev, be32_to_cpu(eqe->event.srq.srqn) & 0xffffff,
eqe->type);
break;
case MLX4_EVENT_TYPE_CMD:
mlx4_cmd_event(dev,
be16_to_cpu(eqe->event.cmd.token),
eqe->event.cmd.status,
be64_to_cpu(eqe->event.cmd.out_param));
break;
case MLX4_EVENT_TYPE_PORT_CHANGE:
mlx4_dispatch_event(dev, eqe->type, eqe->subtype,
be32_to_cpu(eqe->event.port_change.port) >> 28);
break;
case MLX4_EVENT_TYPE_CQ_ERROR:
mlx4_warn(dev, "CQ %s on CQN %06x\n",
eqe->event.cq_err.syndrome == 1 ?
"overrun" : "access violation",
be32_to_cpu(eqe->event.cq_err.cqn) & 0xffffff);
mlx4_cq_event(dev, be32_to_cpu(eqe->event.cq_err.cqn),
eqe->type);
break;
case MLX4_EVENT_TYPE_EQ_OVERFLOW:
mlx4_warn(dev, "EQ overrun on EQN %d\n", eq->eqn);
break;
case MLX4_EVENT_TYPE_EEC_CATAS_ERROR:
case MLX4_EVENT_TYPE_ECC_DETECT:
default:
mlx4_warn(dev, "Unhandled event %02x(%02x) on EQ %d at index %u\n",
eqe->type, eqe->subtype, eq->eqn, eq->cons_index);
break;
};
++eq->cons_index;
eqes_found = 1;
++set_ci;
/*
* The HCA will think the queue has overflowed if we
* don't tell it we've been processing events. We
* create our EQs with MLX4_NUM_SPARE_EQE extra
* entries, so we must update our consumer index at
* least that often.
*/
if (unlikely(set_ci >= MLX4_NUM_SPARE_EQE)) {
/*
* Conditional on hca_type is OK here because
* this is a rare case, not the fast path.
*/
eq_set_ci(eq, 0);
set_ci = 0;
}
}
eq_set_ci(eq, 1);
return eqes_found;
}
static irqreturn_t mlx4_interrupt(int irq, void *dev_ptr)
{
struct mlx4_dev *dev = dev_ptr;
struct mlx4_priv *priv = mlx4_priv(dev);
int work = 0;
int i;
writel(priv->eq_table.clr_mask, priv->eq_table.clr_int);
for (i = 0; i < MLX4_NUM_EQ; ++i)
work |= mlx4_eq_int(dev, &priv->eq_table.eq[i]);
return IRQ_RETVAL(work);
}
static irqreturn_t mlx4_msi_x_interrupt(int irq, void *eq_ptr)
{
struct mlx4_eq *eq = eq_ptr;
struct mlx4_dev *dev = eq->dev;
mlx4_eq_int(dev, eq);
/* MSI-X vectors always belong to us */
return IRQ_HANDLED;
}
static int mlx4_MAP_EQ(struct mlx4_dev *dev, u64 event_mask, int unmap,
int eq_num)
{
return mlx4_cmd(dev, event_mask, (unmap << 31) | eq_num,
0, MLX4_CMD_MAP_EQ, MLX4_CMD_TIME_CLASS_B);
}
static int mlx4_SW2HW_EQ(struct mlx4_dev *dev, struct mlx4_cmd_mailbox *mailbox,
int eq_num)
{
return mlx4_cmd(dev, mailbox->dma, eq_num, 0, MLX4_CMD_SW2HW_EQ,
MLX4_CMD_TIME_CLASS_A);
}
static int mlx4_HW2SW_EQ(struct mlx4_dev *dev, struct mlx4_cmd_mailbox *mailbox,
int eq_num)
{
return mlx4_cmd_box(dev, 0, mailbox->dma, eq_num, 0, MLX4_CMD_HW2SW_EQ,
MLX4_CMD_TIME_CLASS_A);
}
static void __iomem *mlx4_get_eq_uar(struct mlx4_dev *dev, struct mlx4_eq *eq)
{
struct mlx4_priv *priv = mlx4_priv(dev);
int index;
index = eq->eqn / 4 - dev->caps.reserved_eqs / 4;
if (!priv->eq_table.uar_map[index]) {
priv->eq_table.uar_map[index] =
ioremap(pci_resource_start(dev->pdev, 2) +
((eq->eqn / 4) << PAGE_SHIFT),
PAGE_SIZE);
if (!priv->eq_table.uar_map[index]) {
mlx4_err(dev, "Couldn't map EQ doorbell for EQN 0x%06x\n",
eq->eqn);
return NULL;
}
}
return priv->eq_table.uar_map[index] + 0x800 + 8 * (eq->eqn % 4);
}
static int mlx4_create_eq(struct mlx4_dev *dev, int nent,
u8 intr, struct mlx4_eq *eq)
{
struct mlx4_priv *priv = mlx4_priv(dev);
struct mlx4_cmd_mailbox *mailbox;
struct mlx4_eq_context *eq_context;
int npages;
u64 *dma_list = NULL;
dma_addr_t t;
u64 mtt_addr;
int err = -ENOMEM;
int i;
eq->dev = dev;
eq->nent = roundup_pow_of_two(max(nent, 2));
npages = PAGE_ALIGN(eq->nent * MLX4_EQ_ENTRY_SIZE) / PAGE_SIZE;
eq->page_list = kmalloc(npages * sizeof *eq->page_list,
GFP_KERNEL);
if (!eq->page_list)
goto err_out;
for (i = 0; i < npages; ++i)
eq->page_list[i].buf = NULL;
dma_list = kmalloc(npages * sizeof *dma_list, GFP_KERNEL);
if (!dma_list)
goto err_out_free;
mailbox = mlx4_alloc_cmd_mailbox(dev);
if (IS_ERR(mailbox))
goto err_out_free;
eq_context = mailbox->buf;
for (i = 0; i < npages; ++i) {
eq->page_list[i].buf = dma_alloc_coherent(&dev->pdev->dev,
PAGE_SIZE, &t, GFP_KERNEL);
if (!eq->page_list[i].buf)
goto err_out_free_pages;
dma_list[i] = t;
eq->page_list[i].map = t;
memset(eq->page_list[i].buf, 0, PAGE_SIZE);
}
eq->eqn = mlx4_bitmap_alloc(&priv->eq_table.bitmap);
if (eq->eqn == -1)
goto err_out_free_pages;
eq->doorbell = mlx4_get_eq_uar(dev, eq);
if (!eq->doorbell) {
err = -ENOMEM;
goto err_out_free_eq;
}
err = mlx4_mtt_init(dev, npages, PAGE_SHIFT, &eq->mtt);
if (err)
goto err_out_free_eq;
err = mlx4_write_mtt(dev, &eq->mtt, 0, npages, dma_list);
if (err)
goto err_out_free_mtt;
memset(eq_context, 0, sizeof *eq_context);
eq_context->flags = cpu_to_be32(MLX4_EQ_STATUS_OK |
MLX4_EQ_STATE_ARMED);
eq_context->log_eq_size = ilog2(eq->nent);
eq_context->intr = intr;
eq_context->log_page_size = PAGE_SHIFT - MLX4_ICM_PAGE_SHIFT;
mtt_addr = mlx4_mtt_addr(dev, &eq->mtt);
eq_context->mtt_base_addr_h = mtt_addr >> 32;
eq_context->mtt_base_addr_l = cpu_to_be32(mtt_addr & 0xffffffff);
err = mlx4_SW2HW_EQ(dev, mailbox, eq->eqn);
if (err) {
mlx4_warn(dev, "SW2HW_EQ failed (%d)\n", err);
goto err_out_free_mtt;
}
kfree(dma_list);
mlx4_free_cmd_mailbox(dev, mailbox);
eq->cons_index = 0;
return err;
err_out_free_mtt:
mlx4_mtt_cleanup(dev, &eq->mtt);
err_out_free_eq:
mlx4_bitmap_free(&priv->eq_table.bitmap, eq->eqn);
err_out_free_pages:
for (i = 0; i < npages; ++i)
if (eq->page_list[i].buf)
dma_free_coherent(&dev->pdev->dev, PAGE_SIZE,
eq->page_list[i].buf,
eq->page_list[i].map);
mlx4_free_cmd_mailbox(dev, mailbox);
err_out_free:
kfree(eq->page_list);
kfree(dma_list);
err_out:
return err;
}
static void mlx4_free_eq(struct mlx4_dev *dev,
struct mlx4_eq *eq)
{
struct mlx4_priv *priv = mlx4_priv(dev);
struct mlx4_cmd_mailbox *mailbox;
int err;
int npages = PAGE_ALIGN(MLX4_EQ_ENTRY_SIZE * eq->nent) / PAGE_SIZE;
int i;
mailbox = mlx4_alloc_cmd_mailbox(dev);
if (IS_ERR(mailbox))
return;
err = mlx4_HW2SW_EQ(dev, mailbox, eq->eqn);
if (err)
mlx4_warn(dev, "HW2SW_EQ failed (%d)\n", err);
if (0) {
mlx4_dbg(dev, "Dumping EQ context %02x:\n", eq->eqn);
for (i = 0; i < sizeof (struct mlx4_eq_context) / 4; ++i) {
if (i % 4 == 0)
printk("[%02x] ", i * 4);
printk(" %08x", be32_to_cpup(mailbox->buf + i * 4));
if ((i + 1) % 4 == 0)
printk("\n");
}
}
mlx4_mtt_cleanup(dev, &eq->mtt);
for (i = 0; i < npages; ++i)
pci_free_consistent(dev->pdev, PAGE_SIZE,
eq->page_list[i].buf,
eq->page_list[i].map);
kfree(eq->page_list);
mlx4_bitmap_free(&priv->eq_table.bitmap, eq->eqn);
mlx4_free_cmd_mailbox(dev, mailbox);
}
static void mlx4_free_irqs(struct mlx4_dev *dev)
{
struct mlx4_eq_table *eq_table = &mlx4_priv(dev)->eq_table;
int i;
if (eq_table->have_irq)
free_irq(dev->pdev->irq, dev);
for (i = 0; i < MLX4_NUM_EQ; ++i)
if (eq_table->eq[i].have_irq)
free_irq(eq_table->eq[i].irq, eq_table->eq + i);
}
static int mlx4_map_clr_int(struct mlx4_dev *dev)
{
struct mlx4_priv *priv = mlx4_priv(dev);
priv->clr_base = ioremap(pci_resource_start(dev->pdev, priv->fw.clr_int_bar) +
priv->fw.clr_int_base, MLX4_CLR_INT_SIZE);
if (!priv->clr_base) {
mlx4_err(dev, "Couldn't map interrupt clear register, aborting.\n");
return -ENOMEM;
}
return 0;
}
static void mlx4_unmap_clr_int(struct mlx4_dev *dev)
{
struct mlx4_priv *priv = mlx4_priv(dev);
iounmap(priv->clr_base);
}
int mlx4_map_eq_icm(struct mlx4_dev *dev, u64 icm_virt)
{
struct mlx4_priv *priv = mlx4_priv(dev);
int ret;
/*
* We assume that mapping one page is enough for the whole EQ
* context table. This is fine with all current HCAs, because
* we only use 32 EQs and each EQ uses 64 bytes of context
* memory, or 1 KB total.
*/
priv->eq_table.icm_virt = icm_virt;
priv->eq_table.icm_page = alloc_page(GFP_HIGHUSER);
if (!priv->eq_table.icm_page)
return -ENOMEM;
priv->eq_table.icm_dma = pci_map_page(dev->pdev, priv->eq_table.icm_page, 0,
PAGE_SIZE, PCI_DMA_BIDIRECTIONAL);
if (pci_dma_mapping_error(priv->eq_table.icm_dma)) {
__free_page(priv->eq_table.icm_page);
return -ENOMEM;
}
ret = mlx4_MAP_ICM_page(dev, priv->eq_table.icm_dma, icm_virt);
if (ret) {
pci_unmap_page(dev->pdev, priv->eq_table.icm_dma, PAGE_SIZE,
PCI_DMA_BIDIRECTIONAL);
__free_page(priv->eq_table.icm_page);
}
return ret;
}
void mlx4_unmap_eq_icm(struct mlx4_dev *dev)
{
struct mlx4_priv *priv = mlx4_priv(dev);
mlx4_UNMAP_ICM(dev, priv->eq_table.icm_virt, 1);
pci_unmap_page(dev->pdev, priv->eq_table.icm_dma, PAGE_SIZE,
PCI_DMA_BIDIRECTIONAL);
__free_page(priv->eq_table.icm_page);
}
int mlx4_init_eq_table(struct mlx4_dev *dev)
{
struct mlx4_priv *priv = mlx4_priv(dev);
int err;
int i;
err = mlx4_bitmap_init(&priv->eq_table.bitmap, dev->caps.num_eqs,
dev->caps.num_eqs - 1, dev->caps.reserved_eqs);
if (err)
return err;
for (i = 0; i < ARRAY_SIZE(priv->eq_table.uar_map); ++i)
priv->eq_table.uar_map[i] = NULL;
err = mlx4_map_clr_int(dev);
if (err)
goto err_out_free;
priv->eq_table.clr_mask =
swab32(1 << (priv->eq_table.inta_pin & 31));
priv->eq_table.clr_int = priv->clr_base +
(priv->eq_table.inta_pin < 32 ? 4 : 0);
err = mlx4_create_eq(dev, dev->caps.num_cqs + MLX4_NUM_SPARE_EQE,
(dev->flags & MLX4_FLAG_MSI_X) ? MLX4_EQ_COMP : 0,
&priv->eq_table.eq[MLX4_EQ_COMP]);
if (err)
goto err_out_unmap;
err = mlx4_create_eq(dev, MLX4_NUM_ASYNC_EQE + MLX4_NUM_SPARE_EQE,
(dev->flags & MLX4_FLAG_MSI_X) ? MLX4_EQ_ASYNC : 0,
&priv->eq_table.eq[MLX4_EQ_ASYNC]);
if (err)
goto err_out_comp;
if (dev->flags & MLX4_FLAG_MSI_X) {
static const char *eq_name[] = {
[MLX4_EQ_COMP] = DRV_NAME " (comp)",
[MLX4_EQ_ASYNC] = DRV_NAME " (async)"
};
for (i = 0; i < MLX4_NUM_EQ; ++i) {
err = request_irq(priv->eq_table.eq[i].irq,
mlx4_msi_x_interrupt,
0, eq_name[i], priv->eq_table.eq + i);
if (err)
goto err_out_async;
priv->eq_table.eq[i].have_irq = 1;
}
} else {
err = request_irq(dev->pdev->irq, mlx4_interrupt,
IRQF_SHARED, DRV_NAME, dev);
if (err)
goto err_out_async;
priv->eq_table.have_irq = 1;
}
err = mlx4_MAP_EQ(dev, MLX4_ASYNC_EVENT_MASK, 0,
priv->eq_table.eq[MLX4_EQ_ASYNC].eqn);
if (err)
mlx4_warn(dev, "MAP_EQ for async EQ %d failed (%d)\n",
priv->eq_table.eq[MLX4_EQ_ASYNC].eqn, err);
for (i = 0; i < MLX4_NUM_EQ; ++i)
eq_set_ci(&priv->eq_table.eq[i], 1);
return 0;
err_out_async:
mlx4_free_eq(dev, &priv->eq_table.eq[MLX4_EQ_ASYNC]);
err_out_comp:
mlx4_free_eq(dev, &priv->eq_table.eq[MLX4_EQ_COMP]);
err_out_unmap:
mlx4_unmap_clr_int(dev);
mlx4_free_irqs(dev);
err_out_free:
mlx4_bitmap_cleanup(&priv->eq_table.bitmap);
return err;
}
void mlx4_cleanup_eq_table(struct mlx4_dev *dev)
{
struct mlx4_priv *priv = mlx4_priv(dev);
int i;
mlx4_MAP_EQ(dev, MLX4_ASYNC_EVENT_MASK, 1,
priv->eq_table.eq[MLX4_EQ_ASYNC].eqn);
mlx4_free_irqs(dev);
for (i = 0; i < MLX4_NUM_EQ; ++i)
mlx4_free_eq(dev, &priv->eq_table.eq[i]);
mlx4_unmap_clr_int(dev);
for (i = 0; i < ARRAY_SIZE(priv->eq_table.uar_map); ++i)
if (priv->eq_table.uar_map[i])
iounmap(priv->eq_table.uar_map[i]);
mlx4_bitmap_cleanup(&priv->eq_table.bitmap);
}