[deliverable/linux.git] / drivers / net / ethernet / mellanox / mlx5 / core / pagealloc.c

/*
 * Copyright (c) 2013-2015, 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 <asm-generic/kmap_types.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/mlx5/driver.h>
#include <linux/mlx5/cmd.h>
#include "mlx5_core.h"

enum {
	MLX5_PAGES_CANT_GIVE	= 0,
	MLX5_PAGES_GIVE		= 1,
	MLX5_PAGES_TAKE		= 2
};

enum {
	MLX5_BOOT_PAGES		= 1,
	MLX5_INIT_PAGES		= 2,
	MLX5_POST_INIT_PAGES	= 3
};

struct mlx5_pages_req {
	struct mlx5_core_dev *dev;
	u16	func_id;
	s32	npages;
	struct work_struct work;
};

struct fw_page {
	struct rb_node		rb_node;
	u64			addr;
	struct page	       *page;
	u16			func_id;
	unsigned long		bitmask;
	struct list_head	list;
	unsigned		free_count;
};

struct mlx5_query_pages_inbox {
	struct mlx5_inbox_hdr	hdr;
	u8			rsvd[8];
};

struct mlx5_query_pages_outbox {
	struct mlx5_outbox_hdr	hdr;
	__be16			rsvd;
	__be16			func_id;
	__be32			num_pages;
};

struct mlx5_manage_pages_inbox {
	struct mlx5_inbox_hdr	hdr;
	__be16			rsvd;
	__be16			func_id;
	__be32			num_entries;
	__be64			pas[0];
};

struct mlx5_manage_pages_outbox {
	struct mlx5_outbox_hdr	hdr;
	__be32			num_entries;
	u8			rsvd[4];
	__be64			pas[0];
};

enum {
	MAX_RECLAIM_TIME_MSECS	= 5000,
};

enum {
	MLX5_MAX_RECLAIM_TIME_MILI	= 5000,
	MLX5_NUM_4K_IN_PAGE		= PAGE_SIZE / MLX5_ADAPTER_PAGE_SIZE,
};

static int insert_page(struct mlx5_core_dev *dev, u64 addr, struct page *page, u16 func_id)
{
	struct rb_root *root = &dev->priv.page_root;
	struct rb_node **new = &root->rb_node;
	struct rb_node *parent = NULL;
	struct fw_page *nfp;
	struct fw_page *tfp;
	int i;

	while (*new) {
		parent = *new;
		tfp = rb_entry(parent, struct fw_page, rb_node);
		if (tfp->addr < addr)
			new = &parent->rb_left;
		else if (tfp->addr > addr)
			new = &parent->rb_right;
		else
			return -EEXIST;
	}

	nfp = kzalloc(sizeof(*nfp), GFP_KERNEL);
	if (!nfp)
		return -ENOMEM;

	nfp->addr = addr;
	nfp->page = page;
	nfp->func_id = func_id;
	nfp->free_count = MLX5_NUM_4K_IN_PAGE;
	for (i = 0; i < MLX5_NUM_4K_IN_PAGE; i++)
		set_bit(i, &nfp->bitmask);

	rb_link_node(&nfp->rb_node, parent, new);
	rb_insert_color(&nfp->rb_node, root);
	list_add(&nfp->list, &dev->priv.free_list);

	return 0;
}

static struct fw_page *find_fw_page(struct mlx5_core_dev *dev, u64 addr)
{
	struct rb_root *root = &dev->priv.page_root;
	struct rb_node *tmp = root->rb_node;
	struct fw_page *result = NULL;
	struct fw_page *tfp;

	while (tmp) {
		tfp = rb_entry(tmp, struct fw_page, rb_node);
		if (tfp->addr < addr) {
			tmp = tmp->rb_left;
		} else if (tfp->addr > addr) {
			tmp = tmp->rb_right;
		} else {
			result = tfp;
			break;
		}
	}

	return result;
}

static int mlx5_cmd_query_pages(struct mlx5_core_dev *dev, u16 *func_id,
				s32 *npages, int boot)
{
	struct mlx5_query_pages_inbox	in;
	struct mlx5_query_pages_outbox	out;
	int err;

	memset(&in, 0, sizeof(in));
	memset(&out, 0, sizeof(out));
	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_PAGES);
	in.hdr.opmod = boot ? cpu_to_be16(MLX5_BOOT_PAGES) : cpu_to_be16(MLX5_INIT_PAGES);

	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
	if (err)
		return err;

	if (out.hdr.status)
		return mlx5_cmd_status_to_err(&out.hdr);

	*npages = be32_to_cpu(out.num_pages);
	*func_id = be16_to_cpu(out.func_id);

	return err;
}

static int alloc_4k(struct mlx5_core_dev *dev, u64 *addr)
{
	struct fw_page *fp;
	unsigned n;

	if (list_empty(&dev->priv.free_list))
		return -ENOMEM;

	fp = list_entry(dev->priv.free_list.next, struct fw_page, list);
	n = find_first_bit(&fp->bitmask, 8 * sizeof(fp->bitmask));
	if (n >= MLX5_NUM_4K_IN_PAGE) {
		mlx5_core_warn(dev, "alloc 4k bug\n");
		return -ENOENT;
	}
	clear_bit(n, &fp->bitmask);
	fp->free_count--;
	if (!fp->free_count)
		list_del(&fp->list);

	*addr = fp->addr + n * MLX5_ADAPTER_PAGE_SIZE;

	return 0;
}

#define MLX5_U64_4K_PAGE_MASK ((~(u64)0U) << PAGE_SHIFT)

static void free_4k(struct mlx5_core_dev *dev, u64 addr)
{
	struct fw_page *fwp;
	int n;

	fwp = find_fw_page(dev, addr & MLX5_U64_4K_PAGE_MASK);
	if (!fwp) {
		mlx5_core_warn(dev, "page not found\n");
		return;
	}

	n = (addr & ~MLX5_U64_4K_PAGE_MASK) >> MLX5_ADAPTER_PAGE_SHIFT;
	fwp->free_count++;
	set_bit(n, &fwp->bitmask);
	if (fwp->free_count == MLX5_NUM_4K_IN_PAGE) {
		rb_erase(&fwp->rb_node, &dev->priv.page_root);
		if (fwp->free_count != 1)
			list_del(&fwp->list);
		dma_unmap_page(&dev->pdev->dev, addr & MLX5_U64_4K_PAGE_MASK,
			       PAGE_SIZE, DMA_BIDIRECTIONAL);
		__free_page(fwp->page);
		kfree(fwp);
	} else if (fwp->free_count == 1) {
		list_add(&fwp->list, &dev->priv.free_list);
	}
}

static int alloc_system_page(struct mlx5_core_dev *dev, u16 func_id)
{
	struct page *page;
	u64 addr;
	int err;
	int nid = dev_to_node(&dev->pdev->dev);

	page = alloc_pages_node(nid, GFP_HIGHUSER, 0);
	if (!page) {
		mlx5_core_warn(dev, "failed to allocate page\n");
		return -ENOMEM;
	}
	addr = dma_map_page(&dev->pdev->dev, page, 0,
			    PAGE_SIZE, DMA_BIDIRECTIONAL);
	if (dma_mapping_error(&dev->pdev->dev, addr)) {
		mlx5_core_warn(dev, "failed dma mapping page\n");
		err = -ENOMEM;
		goto out_alloc;
	}
	err = insert_page(dev, addr, page, func_id);
	if (err) {
		mlx5_core_err(dev, "failed to track allocated page\n");
		goto out_mapping;
	}

	return 0;

out_mapping:
	dma_unmap_page(&dev->pdev->dev, addr, PAGE_SIZE, DMA_BIDIRECTIONAL);

out_alloc:
	__free_page(page);

	return err;
}
static int give_pages(struct mlx5_core_dev *dev, u16 func_id, int npages,
		      int notify_fail)
{
	struct mlx5_manage_pages_inbox *in;
	struct mlx5_manage_pages_outbox out;
	struct mlx5_manage_pages_inbox *nin;
	int inlen;
	u64 addr;
	int err;
	int i;

	inlen = sizeof(*in) + npages * sizeof(in->pas[0]);
	in = mlx5_vzalloc(inlen);
	if (!in) {
		mlx5_core_warn(dev, "vzalloc failed %d\n", inlen);
		return -ENOMEM;
	}
	memset(&out, 0, sizeof(out));

	for (i = 0; i < npages; i++) {
retry:
		err = alloc_4k(dev, &addr);
		if (err) {
			if (err == -ENOMEM)
				err = alloc_system_page(dev, func_id);
			if (err)
				goto out_4k;

			goto retry;
		}
		in->pas[i] = cpu_to_be64(addr);
	}

	in->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_MANAGE_PAGES);
	in->hdr.opmod = cpu_to_be16(MLX5_PAGES_GIVE);
	in->func_id = cpu_to_be16(func_id);
	in->num_entries = cpu_to_be32(npages);
	err = mlx5_cmd_exec(dev, in, inlen, &out, sizeof(out));
	if (err) {
		mlx5_core_warn(dev, "func_id 0x%x, npages %d, err %d\n",
			       func_id, npages, err);
		goto out_alloc;
	}
	dev->priv.fw_pages += npages;

	if (out.hdr.status) {
		err = mlx5_cmd_status_to_err(&out.hdr);
		if (err) {
			mlx5_core_warn(dev, "func_id 0x%x, npages %d, status %d\n",
				       func_id, npages, out.hdr.status);
			goto out_alloc;
		}
	}

	mlx5_core_dbg(dev, "err %d\n", err);

	goto out_free;

out_alloc:
	if (notify_fail) {
		nin = kzalloc(sizeof(*nin), GFP_KERNEL);
		if (!nin) {
			mlx5_core_warn(dev, "allocation failed\n");
			goto out_4k;
		}
		memset(&out, 0, sizeof(out));
		nin->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_MANAGE_PAGES);
		nin->hdr.opmod = cpu_to_be16(MLX5_PAGES_CANT_GIVE);
		if (mlx5_cmd_exec(dev, nin, sizeof(*nin), &out, sizeof(out)))
			mlx5_core_warn(dev, "page notify failed\n");
		kfree(nin);
	}

out_4k:
	for (i--; i >= 0; i--)
		free_4k(dev, be64_to_cpu(in->pas[i]));
out_free:
	kvfree(in);
	return err;
}

static int reclaim_pages(struct mlx5_core_dev *dev, u32 func_id, int npages,
			 int *nclaimed)
{
	struct mlx5_manage_pages_inbox   in;
	struct mlx5_manage_pages_outbox *out;
	int num_claimed;
	int outlen;
	u64 addr;
	int err;
	int i;

	if (nclaimed)
		*nclaimed = 0;

	memset(&in, 0, sizeof(in));
	outlen = sizeof(*out) + npages * sizeof(out->pas[0]);
	out = mlx5_vzalloc(outlen);
	if (!out)
		return -ENOMEM;

	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_MANAGE_PAGES);
	in.hdr.opmod = cpu_to_be16(MLX5_PAGES_TAKE);
	in.func_id = cpu_to_be16(func_id);
	in.num_entries = cpu_to_be32(npages);
	mlx5_core_dbg(dev, "npages %d, outlen %d\n", npages, outlen);
	err = mlx5_cmd_exec(dev, &in, sizeof(in), out, outlen);
	if (err) {
		mlx5_core_err(dev, "failed reclaiming pages\n");
		goto out_free;
	}
	dev->priv.fw_pages -= npages;

	if (out->hdr.status) {
		err = mlx5_cmd_status_to_err(&out->hdr);
		goto out_free;
	}

	num_claimed = be32_to_cpu(out->num_entries);
	if (nclaimed)
		*nclaimed = num_claimed;

	for (i = 0; i < num_claimed; i++) {
		addr = be64_to_cpu(out->pas[i]);
		free_4k(dev, addr);
	}

out_free:
	kvfree(out);
	return err;
}

static void pages_work_handler(struct work_struct *work)
{
	struct mlx5_pages_req *req = container_of(work, struct mlx5_pages_req, work);
	struct mlx5_core_dev *dev = req->dev;
	int err = 0;

	if (req->npages < 0)
		err = reclaim_pages(dev, req->func_id, -1 * req->npages, NULL);
	else if (req->npages > 0)
		err = give_pages(dev, req->func_id, req->npages, 1);

	if (err)
		mlx5_core_warn(dev, "%s fail %d\n",
			       req->npages < 0 ? "reclaim" : "give", err);

	kfree(req);
}

void mlx5_core_req_pages_handler(struct mlx5_core_dev *dev, u16 func_id,
				 s32 npages)
{
	struct mlx5_pages_req *req;

	req = kzalloc(sizeof(*req), GFP_ATOMIC);
	if (!req) {
		mlx5_core_warn(dev, "failed to allocate pages request\n");
		return;
	}

	req->dev = dev;
	req->func_id = func_id;
	req->npages = npages;
	INIT_WORK(&req->work, pages_work_handler);
	queue_work(dev->priv.pg_wq, &req->work);
}

int mlx5_satisfy_startup_pages(struct mlx5_core_dev *dev, int boot)
{
	u16 uninitialized_var(func_id);
	s32 uninitialized_var(npages);
	int err;

	err = mlx5_cmd_query_pages(dev, &func_id, &npages, boot);
	if (err)
		return err;

	mlx5_core_dbg(dev, "requested %d %s pages for func_id 0x%x\n",
		      npages, boot ? "boot" : "init", func_id);

	return give_pages(dev, func_id, npages, 0);
}

enum {
	MLX5_BLKS_FOR_RECLAIM_PAGES = 12
};

static int optimal_reclaimed_pages(void)
{
	struct mlx5_cmd_prot_block *block;
	struct mlx5_cmd_layout *lay;
	int ret;

	ret = (sizeof(lay->out) + MLX5_BLKS_FOR_RECLAIM_PAGES * sizeof(block->data) -
	       sizeof(struct mlx5_manage_pages_outbox)) /
	       FIELD_SIZEOF(struct mlx5_manage_pages_outbox, pas[0]);

	return ret;
}

int mlx5_reclaim_startup_pages(struct mlx5_core_dev *dev)
{
	unsigned long end = jiffies + msecs_to_jiffies(MAX_RECLAIM_TIME_MSECS);
	struct fw_page *fwp;
	struct rb_node *p;
	int nclaimed = 0;
	int err;

	do {
		p = rb_first(&dev->priv.page_root);
		if (p) {
			fwp = rb_entry(p, struct fw_page, rb_node);
			err = reclaim_pages(dev, fwp->func_id,
					    optimal_reclaimed_pages(),
					    &nclaimed);
			if (err) {
				mlx5_core_warn(dev, "failed reclaiming pages (%d)\n",
					       err);
				return err;
			}
			if (nclaimed)
				end = jiffies + msecs_to_jiffies(MAX_RECLAIM_TIME_MSECS);
		}
		if (time_after(jiffies, end)) {
			mlx5_core_warn(dev, "FW did not return all pages. giving up...\n");
			break;
		}
	} while (p);

	return 0;
}

void mlx5_pagealloc_init(struct mlx5_core_dev *dev)
{
	dev->priv.page_root = RB_ROOT;
	INIT_LIST_HEAD(&dev->priv.free_list);
}

void mlx5_pagealloc_cleanup(struct mlx5_core_dev *dev)
{
	/* nothing */
}

int mlx5_pagealloc_start(struct mlx5_core_dev *dev)
{
	dev->priv.pg_wq = create_singlethread_workqueue("mlx5_page_allocator");
	if (!dev->priv.pg_wq)
		return -ENOMEM;

	return 0;
}

void mlx5_pagealloc_stop(struct mlx5_core_dev *dev)
{
	destroy_workqueue(dev->priv.pg_wq);
}
Commit	Line	Data
e126ba97	1	/*
302bdf68	2	* Copyright (c) 2013-2015, Mellanox Technologies. All rights reserved.
e126ba97 EC	3	*
	4	* This software is available to you under a choice of one of two
	5	* licenses. You may choose to be licensed under the terms of the GNU
	6	* General Public License (GPL) Version 2, available from the file
	7	* COPYING in the main directory of this source tree, or the
	8	* OpenIB.org BSD license below:
	9	*
	10	* Redistribution and use in source and binary forms, with or
	11	* without modification, are permitted provided that the following
	12	* conditions are met:
	13	*
	14	* - Redistributions of source code must retain the above
	15	* copyright notice, this list of conditions and the following
	16	* disclaimer.
	17	*
	18	* - Redistributions in binary form must reproduce the above
	19	* copyright notice, this list of conditions and the following
	20	* disclaimer in the documentation and/or other materials
	21	* provided with the distribution.
	22	*
	23	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
	24	* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
	25	* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
	26	* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
	27	* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
	28	* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
	29	* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
	30	* SOFTWARE.
	31	*/
	32
	33	#include <asm-generic/kmap_types.h>
	34	#include <linux/kernel.h>
	35	#include <linux/module.h>
	36	#include <linux/mlx5/driver.h>
	37	#include <linux/mlx5/cmd.h>
	38	#include "mlx5_core.h"
	39
	40	enum {
	41	MLX5_PAGES_CANT_GIVE = 0,
	42	MLX5_PAGES_GIVE = 1,
	43	MLX5_PAGES_TAKE = 2
	44	};
	45
0a324f31 ML	46	enum {
	47	MLX5_BOOT_PAGES = 1,
	48	MLX5_INIT_PAGES = 2,
	49	MLX5_POST_INIT_PAGES = 3
	50	};
	51
e126ba97 EC	52	struct mlx5_pages_req {
e126ba97 EC	53	struct mlx5_core_dev *dev;
f241e749	54	u16 func_id;
0a324f31	55	s32 npages;
e126ba97 EC	56	struct work_struct work;
	57	};
	58
	59	struct fw_page {
bf0bf77f EC	60	struct rb_node rb_node;
	61	u64 addr;
	62	struct page *page;
	63	u16 func_id;
	64	unsigned long bitmask;
	65	struct list_head list;
	66	unsigned free_count;
e126ba97 EC	67	};
	68
	69	struct mlx5_query_pages_inbox {
	70	struct mlx5_inbox_hdr hdr;
	71	u8 rsvd[8];
	72	};
	73
	74	struct mlx5_query_pages_outbox {
	75	struct mlx5_outbox_hdr hdr;
0a324f31	76	__be16 rsvd;
e126ba97	77	__be16 func_id;
0a324f31	78	__be32 num_pages;
e126ba97 EC	79	};
	80
	81	struct mlx5_manage_pages_inbox {
	82	struct mlx5_inbox_hdr hdr;
0a324f31	83	__be16 rsvd;
e126ba97	84	__be16 func_id;
0a324f31	85	__be32 num_entries;
e126ba97 EC	86	__be64 pas[0];
	87	};
	88
	89	struct mlx5_manage_pages_outbox {
	90	struct mlx5_outbox_hdr hdr;
0a324f31 ML	91	__be32 num_entries;
0a324f31 ML	92	u8 rsvd[4];
e126ba97 EC	93	__be64 pas[0];
	94	};
	95
dabed0e6 EC	96	enum {
	97	MAX_RECLAIM_TIME_MSECS = 5000,
	98	};
	99
bf0bf77f EC	100	enum {
bf0bf77f EC	101	MLX5_MAX_RECLAIM_TIME_MILI = 5000,
05bdb2ab	102	MLX5_NUM_4K_IN_PAGE = PAGE_SIZE / MLX5_ADAPTER_PAGE_SIZE,
bf0bf77f EC	103	};
bf0bf77f EC	104
e126ba97 EC	105	static int insert_page(struct mlx5_core_dev dev, u64 addr, struct page page, u16 func_id)
	106	{
	107	struct rb_root *root = &dev->priv.page_root;
	108	struct rb_node **new = &root->rb_node;
	109	struct rb_node *parent = NULL;
	110	struct fw_page *nfp;
	111	struct fw_page *tfp;
bf0bf77f	112	int i;
e126ba97 EC	113
	114	while (*new) {
	115	parent = *new;
	116	tfp = rb_entry(parent, struct fw_page, rb_node);
	117	if (tfp->addr < addr)
	118	new = &parent->rb_left;
	119	else if (tfp->addr > addr)
	120	new = &parent->rb_right;
	121	else
	122	return -EEXIST;
	123	}
	124
bf0bf77f	125	nfp = kzalloc(sizeof(*nfp), GFP_KERNEL);
e126ba97 EC	126	if (!nfp)
	127	return -ENOMEM;
	128
	129	nfp->addr = addr;
	130	nfp->page = page;
	131	nfp->func_id = func_id;
bf0bf77f EC	132	nfp->free_count = MLX5_NUM_4K_IN_PAGE;
	133	for (i = 0; i < MLX5_NUM_4K_IN_PAGE; i++)
	134	set_bit(i, &nfp->bitmask);
e126ba97 EC	135
	136	rb_link_node(&nfp->rb_node, parent, new);
	137	rb_insert_color(&nfp->rb_node, root);
bf0bf77f	138	list_add(&nfp->list, &dev->priv.free_list);
e126ba97 EC	139
	140	return 0;
	141	}
	142
bf0bf77f	143	static struct fw_page find_fw_page(struct mlx5_core_dev dev, u64 addr)
e126ba97 EC	144	{
	145	struct rb_root *root = &dev->priv.page_root;
	146	struct rb_node *tmp = root->rb_node;
bf0bf77f	147	struct fw_page *result = NULL;
e126ba97 EC	148	struct fw_page *tfp;
	149
	150	while (tmp) {
	151	tfp = rb_entry(tmp, struct fw_page, rb_node);
	152	if (tfp->addr < addr) {
	153	tmp = tmp->rb_left;
	154	} else if (tfp->addr > addr) {
	155	tmp = tmp->rb_right;
	156	} else {
bf0bf77f	157	result = tfp;
e126ba97 EC	158	break;
	159	}
	160	}
	161
	162	return result;
	163	}
	164
	165	static int mlx5_cmd_query_pages(struct mlx5_core_dev dev, u16 func_id,
0a324f31	166	s32 *npages, int boot)
e126ba97 EC	167	{
	168	struct mlx5_query_pages_inbox in;
	169	struct mlx5_query_pages_outbox out;
	170	int err;
	171
	172	memset(&in, 0, sizeof(in));
	173	memset(&out, 0, sizeof(out));
	174	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_QUERY_PAGES);
0a324f31 ML	175	in.hdr.opmod = boot ? cpu_to_be16(MLX5_BOOT_PAGES) : cpu_to_be16(MLX5_INIT_PAGES);
0a324f31 ML	176
e126ba97 EC	177	err = mlx5_cmd_exec(dev, &in, sizeof(in), &out, sizeof(out));
	178	if (err)
	179	return err;
	180
	181	if (out.hdr.status)
	182	return mlx5_cmd_status_to_err(&out.hdr);
	183
0a324f31	184	*npages = be32_to_cpu(out.num_pages);
e126ba97 EC	185	*func_id = be16_to_cpu(out.func_id);
	186
	187	return err;
	188	}
	189
bf0bf77f EC	190	static int alloc_4k(struct mlx5_core_dev dev, u64 addr)
	191	{
	192	struct fw_page *fp;
	193	unsigned n;
	194
24e42754	195	if (list_empty(&dev->priv.free_list))
bf0bf77f	196	return -ENOMEM;
bf0bf77f EC	197
	198	fp = list_entry(dev->priv.free_list.next, struct fw_page, list);
	199	n = find_first_bit(&fp->bitmask, 8 * sizeof(fp->bitmask));
	200	if (n >= MLX5_NUM_4K_IN_PAGE) {
	201	mlx5_core_warn(dev, "alloc 4k bug\n");
	202	return -ENOENT;
	203	}
	204	clear_bit(n, &fp->bitmask);
	205	fp->free_count--;
	206	if (!fp->free_count)
	207	list_del(&fp->list);
	208
05bdb2ab	209	addr = fp->addr + n MLX5_ADAPTER_PAGE_SIZE;
bf0bf77f EC	210
	211	return 0;
	212	}
	213
59d2d18c HL	214	#define MLX5_U64_4K_PAGE_MASK ((~(u64)0U) << PAGE_SHIFT)
59d2d18c HL	215
bf0bf77f EC	216	static void free_4k(struct mlx5_core_dev *dev, u64 addr)
	217	{
	218	struct fw_page *fwp;
	219	int n;
	220
59d2d18c	221	fwp = find_fw_page(dev, addr & MLX5_U64_4K_PAGE_MASK);
bf0bf77f EC	222	if (!fwp) {
	223	mlx5_core_warn(dev, "page not found\n");
	224	return;
	225	}
	226
59d2d18c	227	n = (addr & ~MLX5_U64_4K_PAGE_MASK) >> MLX5_ADAPTER_PAGE_SHIFT;
bf0bf77f EC	228	fwp->free_count++;
	229	set_bit(n, &fwp->bitmask);
	230	if (fwp->free_count == MLX5_NUM_4K_IN_PAGE) {
	231	rb_erase(&fwp->rb_node, &dev->priv.page_root);
2b136d02 EC	232	if (fwp->free_count != 1)
2b136d02 EC	233	list_del(&fwp->list);
59d2d18c HL	234	dma_unmap_page(&dev->pdev->dev, addr & MLX5_U64_4K_PAGE_MASK,
59d2d18c HL	235	PAGE_SIZE, DMA_BIDIRECTIONAL);
bf0bf77f EC	236	__free_page(fwp->page);
	237	kfree(fwp);
	238	} else if (fwp->free_count == 1) {
	239	list_add(&fwp->list, &dev->priv.free_list);
	240	}
	241	}
	242
	243	static int alloc_system_page(struct mlx5_core_dev *dev, u16 func_id)
	244	{
	245	struct page *page;
	246	u64 addr;
	247	int err;
ad189106	248	int nid = dev_to_node(&dev->pdev->dev);
bf0bf77f	249
ad189106	250	page = alloc_pages_node(nid, GFP_HIGHUSER, 0);
bf0bf77f EC	251	if (!page) {
	252	mlx5_core_warn(dev, "failed to allocate page\n");
	253	return -ENOMEM;
	254	}
	255	addr = dma_map_page(&dev->pdev->dev, page, 0,
	256	PAGE_SIZE, DMA_BIDIRECTIONAL);
	257	if (dma_mapping_error(&dev->pdev->dev, addr)) {
	258	mlx5_core_warn(dev, "failed dma mapping page\n");
	259	err = -ENOMEM;
	260	goto out_alloc;
	261	}
	262	err = insert_page(dev, addr, page, func_id);
	263	if (err) {
	264	mlx5_core_err(dev, "failed to track allocated page\n");
	265	goto out_mapping;
	266	}
	267
	268	return 0;
	269
	270	out_mapping:
	271	dma_unmap_page(&dev->pdev->dev, addr, PAGE_SIZE, DMA_BIDIRECTIONAL);
	272
	273	out_alloc:
	274	__free_page(page);
	275
	276	return err;
	277	}
e126ba97 EC	278	static int give_pages(struct mlx5_core_dev *dev, u16 func_id, int npages,
	279	int notify_fail)
	280	{
	281	struct mlx5_manage_pages_inbox *in;
	282	struct mlx5_manage_pages_outbox out;
952f5f6e	283	struct mlx5_manage_pages_inbox *nin;
e126ba97 EC	284	int inlen;
	285	u64 addr;
	286	int err;
	287	int i;
	288
	289	inlen = sizeof(in) + npages sizeof(in->pas[0]);
	290	in = mlx5_vzalloc(inlen);
	291	if (!in) {
	292	mlx5_core_warn(dev, "vzalloc failed %d\n", inlen);
	293	return -ENOMEM;
	294	}
	295	memset(&out, 0, sizeof(out));
	296
	297	for (i = 0; i < npages; i++) {
bf0bf77f EC	298	retry:
bf0bf77f EC	299	err = alloc_4k(dev, &addr);
e126ba97	300	if (err) {
bf0bf77f EC	301	if (err == -ENOMEM)
	302	err = alloc_system_page(dev, func_id);
	303	if (err)
	304	goto out_4k;
	305
	306	goto retry;
e126ba97 EC	307	}
	308	in->pas[i] = cpu_to_be64(addr);
	309	}
	310
	311	in->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_MANAGE_PAGES);
	312	in->hdr.opmod = cpu_to_be16(MLX5_PAGES_GIVE);
	313	in->func_id = cpu_to_be16(func_id);
0a324f31	314	in->num_entries = cpu_to_be32(npages);
e126ba97	315	err = mlx5_cmd_exec(dev, in, inlen, &out, sizeof(out));
e126ba97	316	if (err) {
1a91de28 JP	317	mlx5_core_warn(dev, "func_id 0x%x, npages %d, err %d\n",
1a91de28 JP	318	func_id, npages, err);
e126ba97 EC	319	goto out_alloc;
	320	}
	321	dev->priv.fw_pages += npages;
	322
	323	if (out.hdr.status) {
	324	err = mlx5_cmd_status_to_err(&out.hdr);
	325	if (err) {
1a91de28 JP	326	mlx5_core_warn(dev, "func_id 0x%x, npages %d, status %d\n",
1a91de28 JP	327	func_id, npages, out.hdr.status);
e126ba97 EC	328	goto out_alloc;
	329	}
	330	}
	331
	332	mlx5_core_dbg(dev, "err %d\n", err);
	333
	334	goto out_free;
	335
	336	out_alloc:
	337	if (notify_fail) {
952f5f6e EC	338	nin = kzalloc(sizeof(*nin), GFP_KERNEL);
	339	if (!nin) {
	340	mlx5_core_warn(dev, "allocation failed\n");
bf0bf77f	341	goto out_4k;
952f5f6e	342	}
e126ba97	343	memset(&out, 0, sizeof(out));
952f5f6e EC	344	nin->hdr.opcode = cpu_to_be16(MLX5_CMD_OP_MANAGE_PAGES);
	345	nin->hdr.opmod = cpu_to_be16(MLX5_PAGES_CANT_GIVE);
	346	if (mlx5_cmd_exec(dev, nin, sizeof(*nin), &out, sizeof(out)))
	347	mlx5_core_warn(dev, "page notify failed\n");
	348	kfree(nin);
e126ba97	349	}
952f5f6e	350
bf0bf77f EC	351	out_4k:
	352	for (i--; i >= 0; i--)
	353	free_4k(dev, be64_to_cpu(in->pas[i]));
e126ba97	354	out_free:
479163f4	355	kvfree(in);
e126ba97 EC	356	return err;
	357	}
	358
	359	static int reclaim_pages(struct mlx5_core_dev *dev, u32 func_id, int npages,
	360	int *nclaimed)
	361	{
	362	struct mlx5_manage_pages_inbox in;
	363	struct mlx5_manage_pages_outbox *out;
e126ba97 EC	364	int num_claimed;
	365	int outlen;
	366	u64 addr;
	367	int err;
	368	int i;
	369
dabed0e6 EC	370	if (nclaimed)
	371	*nclaimed = 0;
	372
e126ba97 EC	373	memset(&in, 0, sizeof(in));
	374	outlen = sizeof(out) + npages sizeof(out->pas[0]);
	375	out = mlx5_vzalloc(outlen);
	376	if (!out)
	377	return -ENOMEM;
	378
	379	in.hdr.opcode = cpu_to_be16(MLX5_CMD_OP_MANAGE_PAGES);
	380	in.hdr.opmod = cpu_to_be16(MLX5_PAGES_TAKE);
	381	in.func_id = cpu_to_be16(func_id);
0a324f31	382	in.num_entries = cpu_to_be32(npages);
e126ba97 EC	383	mlx5_core_dbg(dev, "npages %d, outlen %d\n", npages, outlen);
	384	err = mlx5_cmd_exec(dev, &in, sizeof(in), out, outlen);
	385	if (err) {
1a91de28	386	mlx5_core_err(dev, "failed reclaiming pages\n");
e126ba97 EC	387	goto out_free;
	388	}
	389	dev->priv.fw_pages -= npages;
	390
	391	if (out->hdr.status) {
	392	err = mlx5_cmd_status_to_err(&out->hdr);
	393	goto out_free;
	394	}
	395
0a324f31	396	num_claimed = be32_to_cpu(out->num_entries);
e126ba97 EC	397	if (nclaimed)
	398	*nclaimed = num_claimed;
	399
	400	for (i = 0; i < num_claimed; i++) {
	401	addr = be64_to_cpu(out->pas[i]);
bf0bf77f	402	free_4k(dev, addr);
e126ba97 EC	403	}
	404
	405	out_free:
479163f4	406	kvfree(out);
e126ba97 EC	407	return err;
	408	}
	409
	410	static void pages_work_handler(struct work_struct *work)
	411	{
	412	struct mlx5_pages_req *req = container_of(work, struct mlx5_pages_req, work);
	413	struct mlx5_core_dev *dev = req->dev;
	414	int err = 0;
	415
	416	if (req->npages < 0)
	417	err = reclaim_pages(dev, req->func_id, -1 * req->npages, NULL);
	418	else if (req->npages > 0)
	419	err = give_pages(dev, req->func_id, req->npages, 1);
	420
	421	if (err)
1a91de28 JP	422	mlx5_core_warn(dev, "%s fail %d\n",
1a91de28 JP	423	req->npages < 0 ? "reclaim" : "give", err);
e126ba97 EC	424
	425	kfree(req);
	426	}
	427
	428	void mlx5_core_req_pages_handler(struct mlx5_core_dev *dev, u16 func_id,
0a324f31	429	s32 npages)
e126ba97 EC	430	{
	431	struct mlx5_pages_req *req;
	432
	433	req = kzalloc(sizeof(*req), GFP_ATOMIC);
	434	if (!req) {
	435	mlx5_core_warn(dev, "failed to allocate pages request\n");
	436	return;
	437	}
	438
	439	req->dev = dev;
	440	req->func_id = func_id;
	441	req->npages = npages;
	442	INIT_WORK(&req->work, pages_work_handler);
	443	queue_work(dev->priv.pg_wq, &req->work);
	444	}
	445
cd23b14b	446	int mlx5_satisfy_startup_pages(struct mlx5_core_dev *dev, int boot)
e126ba97	447	{
e126ba97	448	u16 uninitialized_var(func_id);
0a324f31	449	s32 uninitialized_var(npages);
e126ba97 EC	450	int err;
e126ba97 EC	451
0a324f31	452	err = mlx5_cmd_query_pages(dev, &func_id, &npages, boot);
e126ba97 EC	453	if (err)
	454	return err;
	455
0a324f31 ML	456	mlx5_core_dbg(dev, "requested %d %s pages for func_id 0x%x\n",
0a324f31 ML	457	npages, boot ? "boot" : "init", func_id);
e126ba97	458
0a324f31	459	return give_pages(dev, func_id, npages, 0);
e126ba97 EC	460	}
e126ba97 EC	461
4e3d677b ML	462	enum {
	463	MLX5_BLKS_FOR_RECLAIM_PAGES = 12
	464	};
	465
e126ba97 EC	466	static int optimal_reclaimed_pages(void)
	467	{
	468	struct mlx5_cmd_prot_block *block;
	469	struct mlx5_cmd_layout *lay;
	470	int ret;
	471
4e3d677b ML	472	ret = (sizeof(lay->out) + MLX5_BLKS_FOR_RECLAIM_PAGES * sizeof(block->data) -
	473	sizeof(struct mlx5_manage_pages_outbox)) /
	474	FIELD_SIZEOF(struct mlx5_manage_pages_outbox, pas[0]);
e126ba97 EC	475
	476	return ret;
	477	}
	478
	479	int mlx5_reclaim_startup_pages(struct mlx5_core_dev *dev)
	480	{
dabed0e6	481	unsigned long end = jiffies + msecs_to_jiffies(MAX_RECLAIM_TIME_MSECS);
e126ba97 EC	482	struct fw_page *fwp;
e126ba97 EC	483	struct rb_node *p;
dabed0e6	484	int nclaimed = 0;
e126ba97 EC	485	int err;
	486
	487	do {
	488	p = rb_first(&dev->priv.page_root);
	489	if (p) {
	490	fwp = rb_entry(p, struct fw_page, rb_node);
dabed0e6 EC	491	err = reclaim_pages(dev, fwp->func_id,
	492	optimal_reclaimed_pages(),
	493	&nclaimed);
e126ba97	494	if (err) {
1a91de28 JP	495	mlx5_core_warn(dev, "failed reclaiming pages (%d)\n",
1a91de28 JP	496	err);
e126ba97 EC	497	return err;
e126ba97 EC	498	}
dabed0e6 EC	499	if (nclaimed)
dabed0e6 EC	500	end = jiffies + msecs_to_jiffies(MAX_RECLAIM_TIME_MSECS);
e126ba97 EC	501	}
	502	if (time_after(jiffies, end)) {
	503	mlx5_core_warn(dev, "FW did not return all pages. giving up...\n");
	504	break;
	505	}
	506	} while (p);
	507
	508	return 0;
	509	}
	510
	511	void mlx5_pagealloc_init(struct mlx5_core_dev *dev)
	512	{
	513	dev->priv.page_root = RB_ROOT;
bf0bf77f	514	INIT_LIST_HEAD(&dev->priv.free_list);
e126ba97 EC	515	}
	516
	517	void mlx5_pagealloc_cleanup(struct mlx5_core_dev *dev)
	518	{
	519	/* nothing */
	520	}
	521
	522	int mlx5_pagealloc_start(struct mlx5_core_dev *dev)
	523	{
	524	dev->priv.pg_wq = create_singlethread_workqueue("mlx5_page_allocator");
	525	if (!dev->priv.pg_wq)
	526	return -ENOMEM;
	527
	528	return 0;
	529	}
	530
	531	void mlx5_pagealloc_stop(struct mlx5_core_dev *dev)
	532	{
	533	destroy_workqueue(dev->priv.pg_wq);
	534	}