linux/drivers/net/ethernet/mellanox/mlx4/pd.c

287 lines
7.1 KiB
C
Raw Normal View History

/*
* Copyright (c) 2006, 2007 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2005 Mellanox Technologies. 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/errno.h>
#include <linux/export.h>
#include <linux/io-mapping.h>
#include <asm/page.h>
#include "mlx4.h"
#include "icm.h"
enum {
MLX4_NUM_RESERVED_UARS = 8
};
int mlx4_pd_alloc(struct mlx4_dev *dev, u32 *pdn)
{
struct mlx4_priv *priv = mlx4_priv(dev);
*pdn = mlx4_bitmap_alloc(&priv->pd_bitmap);
if (*pdn == -1)
return -ENOMEM;
return 0;
}
EXPORT_SYMBOL_GPL(mlx4_pd_alloc);
void mlx4_pd_free(struct mlx4_dev *dev, u32 pdn)
{
mlx4_core: Roll back round robin bitmap allocation commit for CQs, SRQs, and MPTs Commit f4ec9e9 "mlx4_core: Change bitmap allocator to work in round-robin fashion" introduced round-robin allocation (via bitmap) for all resources which allocate via a bitmap. Round robin allocation is desirable for mcgs, counters, pd's, UARs, and xrcds. These are simply numbers, with no involvement of ICM memory mapping. Round robin is required for QPs, since we had a problem with immediate reuse of a 24-bit QP number (commit f4ec9e9). However, for other resources which use the bitmap allocator and involve mapping ICM memory -- MPTs, CQs, SRQs -- round-robin is not desirable. What happens in these cases is the following: ICM memory is allocated and mapped in chunks of 256K. Since the resource allocation index goes up monotonically, the allocator will eventually require mapping a new chunk. Now, chunks are also unmapped when their reference count goes back to zero. Thus, if a single app is running and starts/exits frequently we will have the following situation: When the app starts, a new chunk must be allocated and mapped. When the app exits, the chunk reference count goes back to zero, and the chunk is unmapped and freed. Therefore, the app must pay the cost of allocation and mapping of ICM memory each time it runs (although the price is paid only when allocating the initial entry in the new chunk). For apps which allocate MPTs/SRQs/CQs and which operate as described above, this presented a performance problem. We therefore roll back the round-robin allocator modification for MPTs, CQs, SRQs. Reported-by: Matthew Finlay <matt@mellanox.com> Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-12-08 14:50:17 +00:00
mlx4_bitmap_free(&mlx4_priv(dev)->pd_bitmap, pdn, MLX4_USE_RR);
}
EXPORT_SYMBOL_GPL(mlx4_pd_free);
int __mlx4_xrcd_alloc(struct mlx4_dev *dev, u32 *xrcdn)
{
struct mlx4_priv *priv = mlx4_priv(dev);
*xrcdn = mlx4_bitmap_alloc(&priv->xrcd_bitmap);
if (*xrcdn == -1)
return -ENOMEM;
return 0;
}
int mlx4_xrcd_alloc(struct mlx4_dev *dev, u32 *xrcdn)
{
u64 out_param;
int err;
if (mlx4_is_mfunc(dev)) {
err = mlx4_cmd_imm(dev, 0, &out_param,
RES_XRCD, RES_OP_RESERVE,
MLX4_CMD_ALLOC_RES,
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_WRAPPED);
if (err)
return err;
*xrcdn = get_param_l(&out_param);
return 0;
}
return __mlx4_xrcd_alloc(dev, xrcdn);
}
EXPORT_SYMBOL_GPL(mlx4_xrcd_alloc);
void __mlx4_xrcd_free(struct mlx4_dev *dev, u32 xrcdn)
{
mlx4_core: Roll back round robin bitmap allocation commit for CQs, SRQs, and MPTs Commit f4ec9e9 "mlx4_core: Change bitmap allocator to work in round-robin fashion" introduced round-robin allocation (via bitmap) for all resources which allocate via a bitmap. Round robin allocation is desirable for mcgs, counters, pd's, UARs, and xrcds. These are simply numbers, with no involvement of ICM memory mapping. Round robin is required for QPs, since we had a problem with immediate reuse of a 24-bit QP number (commit f4ec9e9). However, for other resources which use the bitmap allocator and involve mapping ICM memory -- MPTs, CQs, SRQs -- round-robin is not desirable. What happens in these cases is the following: ICM memory is allocated and mapped in chunks of 256K. Since the resource allocation index goes up monotonically, the allocator will eventually require mapping a new chunk. Now, chunks are also unmapped when their reference count goes back to zero. Thus, if a single app is running and starts/exits frequently we will have the following situation: When the app starts, a new chunk must be allocated and mapped. When the app exits, the chunk reference count goes back to zero, and the chunk is unmapped and freed. Therefore, the app must pay the cost of allocation and mapping of ICM memory each time it runs (although the price is paid only when allocating the initial entry in the new chunk). For apps which allocate MPTs/SRQs/CQs and which operate as described above, this presented a performance problem. We therefore roll back the round-robin allocator modification for MPTs, CQs, SRQs. Reported-by: Matthew Finlay <matt@mellanox.com> Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-12-08 14:50:17 +00:00
mlx4_bitmap_free(&mlx4_priv(dev)->xrcd_bitmap, xrcdn, MLX4_USE_RR);
}
void mlx4_xrcd_free(struct mlx4_dev *dev, u32 xrcdn)
{
u64 in_param = 0;
int err;
if (mlx4_is_mfunc(dev)) {
set_param_l(&in_param, xrcdn);
err = mlx4_cmd(dev, in_param, RES_XRCD,
RES_OP_RESERVE, MLX4_CMD_FREE_RES,
MLX4_CMD_TIME_CLASS_A, MLX4_CMD_WRAPPED);
if (err)
mlx4_warn(dev, "Failed to release xrcdn %d\n", xrcdn);
} else
__mlx4_xrcd_free(dev, xrcdn);
}
EXPORT_SYMBOL_GPL(mlx4_xrcd_free);
int mlx4_init_pd_table(struct mlx4_dev *dev)
{
struct mlx4_priv *priv = mlx4_priv(dev);
return mlx4_bitmap_init(&priv->pd_bitmap, dev->caps.num_pds,
(1 << NOT_MASKED_PD_BITS) - 1,
dev->caps.reserved_pds, 0);
}
void mlx4_cleanup_pd_table(struct mlx4_dev *dev)
{
mlx4_bitmap_cleanup(&mlx4_priv(dev)->pd_bitmap);
}
int mlx4_init_xrcd_table(struct mlx4_dev *dev)
{
struct mlx4_priv *priv = mlx4_priv(dev);
return mlx4_bitmap_init(&priv->xrcd_bitmap, (1 << 16),
(1 << 16) - 1, dev->caps.reserved_xrcds + 1, 0);
}
void mlx4_cleanup_xrcd_table(struct mlx4_dev *dev)
{
mlx4_bitmap_cleanup(&mlx4_priv(dev)->xrcd_bitmap);
}
int mlx4_uar_alloc(struct mlx4_dev *dev, struct mlx4_uar *uar)
{
int offset;
uar->index = mlx4_bitmap_alloc(&mlx4_priv(dev)->uar_table.bitmap);
if (uar->index == -1)
return -ENOMEM;
if (mlx4_is_slave(dev))
offset = uar->index % ((int) pci_resource_len(dev->pdev, 2) /
dev->caps.uar_page_size);
else
offset = uar->index;
uar->pfn = (pci_resource_start(dev->pdev, 2) >> PAGE_SHIFT) + offset;
uar->map = NULL;
return 0;
}
EXPORT_SYMBOL_GPL(mlx4_uar_alloc);
void mlx4_uar_free(struct mlx4_dev *dev, struct mlx4_uar *uar)
{
mlx4_core: Roll back round robin bitmap allocation commit for CQs, SRQs, and MPTs Commit f4ec9e9 "mlx4_core: Change bitmap allocator to work in round-robin fashion" introduced round-robin allocation (via bitmap) for all resources which allocate via a bitmap. Round robin allocation is desirable for mcgs, counters, pd's, UARs, and xrcds. These are simply numbers, with no involvement of ICM memory mapping. Round robin is required for QPs, since we had a problem with immediate reuse of a 24-bit QP number (commit f4ec9e9). However, for other resources which use the bitmap allocator and involve mapping ICM memory -- MPTs, CQs, SRQs -- round-robin is not desirable. What happens in these cases is the following: ICM memory is allocated and mapped in chunks of 256K. Since the resource allocation index goes up monotonically, the allocator will eventually require mapping a new chunk. Now, chunks are also unmapped when their reference count goes back to zero. Thus, if a single app is running and starts/exits frequently we will have the following situation: When the app starts, a new chunk must be allocated and mapped. When the app exits, the chunk reference count goes back to zero, and the chunk is unmapped and freed. Therefore, the app must pay the cost of allocation and mapping of ICM memory each time it runs (although the price is paid only when allocating the initial entry in the new chunk). For apps which allocate MPTs/SRQs/CQs and which operate as described above, this presented a performance problem. We therefore roll back the round-robin allocator modification for MPTs, CQs, SRQs. Reported-by: Matthew Finlay <matt@mellanox.com> Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il> Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com> Signed-off-by: David S. Miller <davem@davemloft.net>
2013-12-08 14:50:17 +00:00
mlx4_bitmap_free(&mlx4_priv(dev)->uar_table.bitmap, uar->index, MLX4_USE_RR);
}
EXPORT_SYMBOL_GPL(mlx4_uar_free);
int mlx4_bf_alloc(struct mlx4_dev *dev, struct mlx4_bf *bf, int node)
{
struct mlx4_priv *priv = mlx4_priv(dev);
struct mlx4_uar *uar;
int err = 0;
int idx;
if (!priv->bf_mapping)
return -ENOMEM;
mutex_lock(&priv->bf_mutex);
if (!list_empty(&priv->bf_list))
uar = list_entry(priv->bf_list.next, struct mlx4_uar, bf_list);
else {
if (mlx4_bitmap_avail(&priv->uar_table.bitmap) < MLX4_NUM_RESERVED_UARS) {
err = -ENOMEM;
goto out;
}
uar = kmalloc_node(sizeof(*uar), GFP_KERNEL, node);
if (!uar) {
uar = kmalloc(sizeof(*uar), GFP_KERNEL);
if (!uar) {
err = -ENOMEM;
goto out;
}
}
err = mlx4_uar_alloc(dev, uar);
if (err)
goto free_kmalloc;
uar->map = ioremap(uar->pfn << PAGE_SHIFT, PAGE_SIZE);
if (!uar->map) {
err = -ENOMEM;
goto free_uar;
}
uar->bf_map = io_mapping_map_wc(priv->bf_mapping, uar->index << PAGE_SHIFT);
if (!uar->bf_map) {
err = -ENOMEM;
goto unamp_uar;
}
uar->free_bf_bmap = 0;
list_add(&uar->bf_list, &priv->bf_list);
}
bf->uar = uar;
idx = ffz(uar->free_bf_bmap);
uar->free_bf_bmap |= 1 << idx;
bf->uar = uar;
bf->offset = 0;
bf->buf_size = dev->caps.bf_reg_size / 2;
bf->reg = uar->bf_map + idx * dev->caps.bf_reg_size;
if (uar->free_bf_bmap == (1 << dev->caps.bf_regs_per_page) - 1)
list_del_init(&uar->bf_list);
goto out;
unamp_uar:
bf->uar = NULL;
iounmap(uar->map);
free_uar:
mlx4_uar_free(dev, uar);
free_kmalloc:
kfree(uar);
out:
mutex_unlock(&priv->bf_mutex);
return err;
}
EXPORT_SYMBOL_GPL(mlx4_bf_alloc);
void mlx4_bf_free(struct mlx4_dev *dev, struct mlx4_bf *bf)
{
struct mlx4_priv *priv = mlx4_priv(dev);
int idx;
if (!bf->uar || !bf->uar->bf_map)
return;
mutex_lock(&priv->bf_mutex);
idx = (bf->reg - bf->uar->bf_map) / dev->caps.bf_reg_size;
bf->uar->free_bf_bmap &= ~(1 << idx);
if (!bf->uar->free_bf_bmap) {
if (!list_empty(&bf->uar->bf_list))
list_del(&bf->uar->bf_list);
io_mapping_unmap(bf->uar->bf_map);
iounmap(bf->uar->map);
mlx4_uar_free(dev, bf->uar);
kfree(bf->uar);
} else if (list_empty(&bf->uar->bf_list))
list_add(&bf->uar->bf_list, &priv->bf_list);
mutex_unlock(&priv->bf_mutex);
}
EXPORT_SYMBOL_GPL(mlx4_bf_free);
int mlx4_init_uar_table(struct mlx4_dev *dev)
{
if (dev->caps.num_uars <= 128) {
mlx4_err(dev, "Only %d UAR pages (need more than 128)\n",
dev->caps.num_uars);
mlx4_err(dev, "Increase firmware log2_uar_bar_megabytes?\n");
return -ENODEV;
}
return mlx4_bitmap_init(&mlx4_priv(dev)->uar_table.bitmap,
dev->caps.num_uars, dev->caps.num_uars - 1,
dev->caps.reserved_uars, 0);
}
void mlx4_cleanup_uar_table(struct mlx4_dev *dev)
{
mlx4_bitmap_cleanup(&mlx4_priv(dev)->uar_table.bitmap);
}