linux_dsm_epyc7002/drivers/infiniband/hw/qib/qib_user_pages.c
Christoph Lameter bc3e53f682 mm: distinguish between mlocked and pinned pages
Some kernel components pin user space memory (infiniband and perf) (by
increasing the page count) and account that memory as "mlocked".

The difference between mlocking and pinning is:

A. mlocked pages are marked with PG_mlocked and are exempt from
   swapping. Page migration may move them around though.
   They are kept on a special LRU list.

B. Pinned pages cannot be moved because something needs to
   directly access physical memory. They may not be on any
   LRU list.

I recently saw an mlockalled process where mm->locked_vm became
bigger than the virtual size of the process (!) because some
memory was accounted for twice:

Once when the page was mlocked and once when the Infiniband
layer increased the refcount because it needt to pin the RDMA
memory.

This patch introduces a separate counter for pinned pages and
accounts them seperately.

Signed-off-by: Christoph Lameter <cl@linux.com>
Cc: Mike Marciniszyn <infinipath@qlogic.com>
Cc: Roland Dreier <roland@kernel.org>
Cc: Sean Hefty <sean.hefty@intel.com>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2011-10-31 17:30:46 -07:00

158 lines
4.4 KiB
C

/*
* Copyright (c) 2006, 2007, 2008, 2009 QLogic Corporation. All rights reserved.
* Copyright (c) 2003, 2004, 2005, 2006 PathScale, 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/mm.h>
#include <linux/device.h>
#include "qib.h"
static void __qib_release_user_pages(struct page **p, size_t num_pages,
int dirty)
{
size_t i;
for (i = 0; i < num_pages; i++) {
if (dirty)
set_page_dirty_lock(p[i]);
put_page(p[i]);
}
}
/*
* Call with current->mm->mmap_sem held.
*/
static int __qib_get_user_pages(unsigned long start_page, size_t num_pages,
struct page **p, struct vm_area_struct **vma)
{
unsigned long lock_limit;
size_t got;
int ret;
lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
if (num_pages > lock_limit && !capable(CAP_IPC_LOCK)) {
ret = -ENOMEM;
goto bail;
}
for (got = 0; got < num_pages; got += ret) {
ret = get_user_pages(current, current->mm,
start_page + got * PAGE_SIZE,
num_pages - got, 1, 1,
p + got, vma);
if (ret < 0)
goto bail_release;
}
current->mm->pinned_vm += num_pages;
ret = 0;
goto bail;
bail_release:
__qib_release_user_pages(p, got, 0);
bail:
return ret;
}
/**
* qib_map_page - a safety wrapper around pci_map_page()
*
* A dma_addr of all 0's is interpreted by the chip as "disabled".
* Unfortunately, it can also be a valid dma_addr returned on some
* architectures.
*
* The powerpc iommu assigns dma_addrs in ascending order, so we don't
* have to bother with retries or mapping a dummy page to insure we
* don't just get the same mapping again.
*
* I'm sure we won't be so lucky with other iommu's, so FIXME.
*/
dma_addr_t qib_map_page(struct pci_dev *hwdev, struct page *page,
unsigned long offset, size_t size, int direction)
{
dma_addr_t phys;
phys = pci_map_page(hwdev, page, offset, size, direction);
if (phys == 0) {
pci_unmap_page(hwdev, phys, size, direction);
phys = pci_map_page(hwdev, page, offset, size, direction);
/*
* FIXME: If we get 0 again, we should keep this page,
* map another, then free the 0 page.
*/
}
return phys;
}
/**
* qib_get_user_pages - lock user pages into memory
* @start_page: the start page
* @num_pages: the number of pages
* @p: the output page structures
*
* This function takes a given start page (page aligned user virtual
* address) and pins it and the following specified number of pages. For
* now, num_pages is always 1, but that will probably change at some point
* (because caller is doing expected sends on a single virtually contiguous
* buffer, so we can do all pages at once).
*/
int qib_get_user_pages(unsigned long start_page, size_t num_pages,
struct page **p)
{
int ret;
down_write(&current->mm->mmap_sem);
ret = __qib_get_user_pages(start_page, num_pages, p, NULL);
up_write(&current->mm->mmap_sem);
return ret;
}
void qib_release_user_pages(struct page **p, size_t num_pages)
{
if (current->mm) /* during close after signal, mm can be NULL */
down_write(&current->mm->mmap_sem);
__qib_release_user_pages(p, num_pages, 1);
if (current->mm) {
current->mm->pinned_vm -= num_pages;
up_write(&current->mm->mmap_sem);
}
}