linux/drivers/infiniband/hw/usnic/usnic_uiom.c
Davidlohr Bueso f808c13fd3 lib/interval_tree: fast overlap detection
Allow interval trees to quickly check for overlaps to avoid unnecesary
tree lookups in interval_tree_iter_first().

As of this patch, all interval tree flavors will require using a
'rb_root_cached' such that we can have the leftmost node easily
available.  While most users will make use of this feature, those with
special functions (in addition to the generic insert, delete, search
calls) will avoid using the cached option as they can do funky things
with insertions -- for example, vma_interval_tree_insert_after().

[jglisse@redhat.com: fix deadlock from typo vm_lock_anon_vma()]
  Link: http://lkml.kernel.org/r/20170808225719.20723-1-jglisse@redhat.com
Link: http://lkml.kernel.org/r/20170719014603.19029-12-dave@stgolabs.net
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Doug Ledford <dledford@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Christian Benvenuti <benve@cisco.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-08 18:26:49 -07:00

603 lines
15 KiB
C

/*
* Copyright (c) 2005 Topspin Communications. All rights reserved.
* Copyright (c) 2005 Mellanox Technologies. All rights reserved.
* Copyright (c) 2013 Cisco Systems. 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
* 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/dma-mapping.h>
#include <linux/sched/signal.h>
#include <linux/sched/mm.h>
#include <linux/hugetlb.h>
#include <linux/iommu.h>
#include <linux/workqueue.h>
#include <linux/list.h>
#include <linux/pci.h>
#include "usnic_log.h"
#include "usnic_uiom.h"
#include "usnic_uiom_interval_tree.h"
static struct workqueue_struct *usnic_uiom_wq;
#define USNIC_UIOM_PAGE_CHUNK \
((PAGE_SIZE - offsetof(struct usnic_uiom_chunk, page_list)) /\
((void *) &((struct usnic_uiom_chunk *) 0)->page_list[1] - \
(void *) &((struct usnic_uiom_chunk *) 0)->page_list[0]))
static void usnic_uiom_reg_account(struct work_struct *work)
{
struct usnic_uiom_reg *umem = container_of(work,
struct usnic_uiom_reg, work);
down_write(&umem->mm->mmap_sem);
umem->mm->locked_vm -= umem->diff;
up_write(&umem->mm->mmap_sem);
mmput(umem->mm);
kfree(umem);
}
static int usnic_uiom_dma_fault(struct iommu_domain *domain,
struct device *dev,
unsigned long iova, int flags,
void *token)
{
usnic_err("Device %s iommu fault domain 0x%pK va 0x%lx flags 0x%x\n",
dev_name(dev),
domain, iova, flags);
return -ENOSYS;
}
static void usnic_uiom_put_pages(struct list_head *chunk_list, int dirty)
{
struct usnic_uiom_chunk *chunk, *tmp;
struct page *page;
struct scatterlist *sg;
int i;
dma_addr_t pa;
list_for_each_entry_safe(chunk, tmp, chunk_list, list) {
for_each_sg(chunk->page_list, sg, chunk->nents, i) {
page = sg_page(sg);
pa = sg_phys(sg);
if (dirty)
set_page_dirty_lock(page);
put_page(page);
usnic_dbg("pa: %pa\n", &pa);
}
kfree(chunk);
}
}
static int usnic_uiom_get_pages(unsigned long addr, size_t size, int writable,
int dmasync, struct list_head *chunk_list)
{
struct page **page_list;
struct scatterlist *sg;
struct usnic_uiom_chunk *chunk;
unsigned long locked;
unsigned long lock_limit;
unsigned long cur_base;
unsigned long npages;
int ret;
int off;
int i;
int flags;
dma_addr_t pa;
unsigned int gup_flags;
if (!can_do_mlock())
return -EPERM;
INIT_LIST_HEAD(chunk_list);
page_list = (struct page **) __get_free_page(GFP_KERNEL);
if (!page_list)
return -ENOMEM;
npages = PAGE_ALIGN(size + (addr & ~PAGE_MASK)) >> PAGE_SHIFT;
down_write(&current->mm->mmap_sem);
locked = npages + current->mm->locked_vm;
lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
if ((locked > lock_limit) && !capable(CAP_IPC_LOCK)) {
ret = -ENOMEM;
goto out;
}
flags = IOMMU_READ | IOMMU_CACHE;
flags |= (writable) ? IOMMU_WRITE : 0;
gup_flags = FOLL_WRITE;
gup_flags |= (writable) ? 0 : FOLL_FORCE;
cur_base = addr & PAGE_MASK;
ret = 0;
while (npages) {
ret = get_user_pages(cur_base,
min_t(unsigned long, npages,
PAGE_SIZE / sizeof(struct page *)),
gup_flags, page_list, NULL);
if (ret < 0)
goto out;
npages -= ret;
off = 0;
while (ret) {
chunk = kmalloc(sizeof(*chunk) +
sizeof(struct scatterlist) *
min_t(int, ret, USNIC_UIOM_PAGE_CHUNK),
GFP_KERNEL);
if (!chunk) {
ret = -ENOMEM;
goto out;
}
chunk->nents = min_t(int, ret, USNIC_UIOM_PAGE_CHUNK);
sg_init_table(chunk->page_list, chunk->nents);
for_each_sg(chunk->page_list, sg, chunk->nents, i) {
sg_set_page(sg, page_list[i + off],
PAGE_SIZE, 0);
pa = sg_phys(sg);
usnic_dbg("va: 0x%lx pa: %pa\n",
cur_base + i*PAGE_SIZE, &pa);
}
cur_base += chunk->nents * PAGE_SIZE;
ret -= chunk->nents;
off += chunk->nents;
list_add_tail(&chunk->list, chunk_list);
}
ret = 0;
}
out:
if (ret < 0)
usnic_uiom_put_pages(chunk_list, 0);
else
current->mm->locked_vm = locked;
up_write(&current->mm->mmap_sem);
free_page((unsigned long) page_list);
return ret;
}
static void usnic_uiom_unmap_sorted_intervals(struct list_head *intervals,
struct usnic_uiom_pd *pd)
{
struct usnic_uiom_interval_node *interval, *tmp;
long unsigned va, size;
list_for_each_entry_safe(interval, tmp, intervals, link) {
va = interval->start << PAGE_SHIFT;
size = ((interval->last - interval->start) + 1) << PAGE_SHIFT;
while (size > 0) {
/* Workaround for RH 970401 */
usnic_dbg("va 0x%lx size 0x%lx", va, PAGE_SIZE);
iommu_unmap(pd->domain, va, PAGE_SIZE);
va += PAGE_SIZE;
size -= PAGE_SIZE;
}
}
}
static void __usnic_uiom_reg_release(struct usnic_uiom_pd *pd,
struct usnic_uiom_reg *uiomr,
int dirty)
{
int npages;
unsigned long vpn_start, vpn_last;
struct usnic_uiom_interval_node *interval, *tmp;
int writable = 0;
LIST_HEAD(rm_intervals);
npages = PAGE_ALIGN(uiomr->length + uiomr->offset) >> PAGE_SHIFT;
vpn_start = (uiomr->va & PAGE_MASK) >> PAGE_SHIFT;
vpn_last = vpn_start + npages - 1;
spin_lock(&pd->lock);
usnic_uiom_remove_interval(&pd->root, vpn_start,
vpn_last, &rm_intervals);
usnic_uiom_unmap_sorted_intervals(&rm_intervals, pd);
list_for_each_entry_safe(interval, tmp, &rm_intervals, link) {
if (interval->flags & IOMMU_WRITE)
writable = 1;
list_del(&interval->link);
kfree(interval);
}
usnic_uiom_put_pages(&uiomr->chunk_list, dirty & writable);
spin_unlock(&pd->lock);
}
static int usnic_uiom_map_sorted_intervals(struct list_head *intervals,
struct usnic_uiom_reg *uiomr)
{
int i, err;
size_t size;
struct usnic_uiom_chunk *chunk;
struct usnic_uiom_interval_node *interval_node;
dma_addr_t pa;
dma_addr_t pa_start = 0;
dma_addr_t pa_end = 0;
long int va_start = -EINVAL;
struct usnic_uiom_pd *pd = uiomr->pd;
long int va = uiomr->va & PAGE_MASK;
int flags = IOMMU_READ | IOMMU_CACHE;
flags |= (uiomr->writable) ? IOMMU_WRITE : 0;
chunk = list_first_entry(&uiomr->chunk_list, struct usnic_uiom_chunk,
list);
list_for_each_entry(interval_node, intervals, link) {
iter_chunk:
for (i = 0; i < chunk->nents; i++, va += PAGE_SIZE) {
pa = sg_phys(&chunk->page_list[i]);
if ((va >> PAGE_SHIFT) < interval_node->start)
continue;
if ((va >> PAGE_SHIFT) == interval_node->start) {
/* First page of the interval */
va_start = va;
pa_start = pa;
pa_end = pa;
}
WARN_ON(va_start == -EINVAL);
if ((pa_end + PAGE_SIZE != pa) &&
(pa != pa_start)) {
/* PAs are not contiguous */
size = pa_end - pa_start + PAGE_SIZE;
usnic_dbg("va 0x%lx pa %pa size 0x%zx flags 0x%x",
va_start, &pa_start, size, flags);
err = iommu_map(pd->domain, va_start, pa_start,
size, flags);
if (err) {
usnic_err("Failed to map va 0x%lx pa %pa size 0x%zx with err %d\n",
va_start, &pa_start, size, err);
goto err_out;
}
va_start = va;
pa_start = pa;
pa_end = pa;
}
if ((va >> PAGE_SHIFT) == interval_node->last) {
/* Last page of the interval */
size = pa - pa_start + PAGE_SIZE;
usnic_dbg("va 0x%lx pa %pa size 0x%zx flags 0x%x\n",
va_start, &pa_start, size, flags);
err = iommu_map(pd->domain, va_start, pa_start,
size, flags);
if (err) {
usnic_err("Failed to map va 0x%lx pa %pa size 0x%zx with err %d\n",
va_start, &pa_start, size, err);
goto err_out;
}
break;
}
if (pa != pa_start)
pa_end += PAGE_SIZE;
}
if (i == chunk->nents) {
/*
* Hit last entry of the chunk,
* hence advance to next chunk
*/
chunk = list_first_entry(&chunk->list,
struct usnic_uiom_chunk,
list);
goto iter_chunk;
}
}
return 0;
err_out:
usnic_uiom_unmap_sorted_intervals(intervals, pd);
return err;
}
struct usnic_uiom_reg *usnic_uiom_reg_get(struct usnic_uiom_pd *pd,
unsigned long addr, size_t size,
int writable, int dmasync)
{
struct usnic_uiom_reg *uiomr;
unsigned long va_base, vpn_start, vpn_last;
unsigned long npages;
int offset, err;
LIST_HEAD(sorted_diff_intervals);
/*
* Intel IOMMU map throws an error if a translation entry is
* changed from read to write. This module may not unmap
* and then remap the entry after fixing the permission
* b/c this open up a small windows where hw DMA may page fault
* Hence, make all entries to be writable.
*/
writable = 1;
va_base = addr & PAGE_MASK;
offset = addr & ~PAGE_MASK;
npages = PAGE_ALIGN(size + offset) >> PAGE_SHIFT;
vpn_start = (addr & PAGE_MASK) >> PAGE_SHIFT;
vpn_last = vpn_start + npages - 1;
uiomr = kmalloc(sizeof(*uiomr), GFP_KERNEL);
if (!uiomr)
return ERR_PTR(-ENOMEM);
uiomr->va = va_base;
uiomr->offset = offset;
uiomr->length = size;
uiomr->writable = writable;
uiomr->pd = pd;
err = usnic_uiom_get_pages(addr, size, writable, dmasync,
&uiomr->chunk_list);
if (err) {
usnic_err("Failed get_pages vpn [0x%lx,0x%lx] err %d\n",
vpn_start, vpn_last, err);
goto out_free_uiomr;
}
spin_lock(&pd->lock);
err = usnic_uiom_get_intervals_diff(vpn_start, vpn_last,
(writable) ? IOMMU_WRITE : 0,
IOMMU_WRITE,
&pd->root,
&sorted_diff_intervals);
if (err) {
usnic_err("Failed disjoint interval vpn [0x%lx,0x%lx] err %d\n",
vpn_start, vpn_last, err);
goto out_put_pages;
}
err = usnic_uiom_map_sorted_intervals(&sorted_diff_intervals, uiomr);
if (err) {
usnic_err("Failed map interval vpn [0x%lx,0x%lx] err %d\n",
vpn_start, vpn_last, err);
goto out_put_intervals;
}
err = usnic_uiom_insert_interval(&pd->root, vpn_start, vpn_last,
(writable) ? IOMMU_WRITE : 0);
if (err) {
usnic_err("Failed insert interval vpn [0x%lx,0x%lx] err %d\n",
vpn_start, vpn_last, err);
goto out_unmap_intervals;
}
usnic_uiom_put_interval_set(&sorted_diff_intervals);
spin_unlock(&pd->lock);
return uiomr;
out_unmap_intervals:
usnic_uiom_unmap_sorted_intervals(&sorted_diff_intervals, pd);
out_put_intervals:
usnic_uiom_put_interval_set(&sorted_diff_intervals);
out_put_pages:
usnic_uiom_put_pages(&uiomr->chunk_list, 0);
spin_unlock(&pd->lock);
out_free_uiomr:
kfree(uiomr);
return ERR_PTR(err);
}
void usnic_uiom_reg_release(struct usnic_uiom_reg *uiomr, int closing)
{
struct mm_struct *mm;
unsigned long diff;
__usnic_uiom_reg_release(uiomr->pd, uiomr, 1);
mm = get_task_mm(current);
if (!mm) {
kfree(uiomr);
return;
}
diff = PAGE_ALIGN(uiomr->length + uiomr->offset) >> PAGE_SHIFT;
/*
* We may be called with the mm's mmap_sem already held. This
* can happen when a userspace munmap() is the call that drops
* the last reference to our file and calls our release
* method. If there are memory regions to destroy, we'll end
* up here and not be able to take the mmap_sem. In that case
* we defer the vm_locked accounting to the system workqueue.
*/
if (closing) {
if (!down_write_trylock(&mm->mmap_sem)) {
INIT_WORK(&uiomr->work, usnic_uiom_reg_account);
uiomr->mm = mm;
uiomr->diff = diff;
queue_work(usnic_uiom_wq, &uiomr->work);
return;
}
} else
down_write(&mm->mmap_sem);
current->mm->locked_vm -= diff;
up_write(&mm->mmap_sem);
mmput(mm);
kfree(uiomr);
}
struct usnic_uiom_pd *usnic_uiom_alloc_pd(void)
{
struct usnic_uiom_pd *pd;
void *domain;
pd = kzalloc(sizeof(*pd), GFP_KERNEL);
if (!pd)
return ERR_PTR(-ENOMEM);
pd->domain = domain = iommu_domain_alloc(&pci_bus_type);
if (!domain) {
usnic_err("Failed to allocate IOMMU domain");
kfree(pd);
return ERR_PTR(-ENOMEM);
}
iommu_set_fault_handler(pd->domain, usnic_uiom_dma_fault, NULL);
spin_lock_init(&pd->lock);
INIT_LIST_HEAD(&pd->devs);
return pd;
}
void usnic_uiom_dealloc_pd(struct usnic_uiom_pd *pd)
{
iommu_domain_free(pd->domain);
kfree(pd);
}
int usnic_uiom_attach_dev_to_pd(struct usnic_uiom_pd *pd, struct device *dev)
{
struct usnic_uiom_dev *uiom_dev;
int err;
uiom_dev = kzalloc(sizeof(*uiom_dev), GFP_ATOMIC);
if (!uiom_dev)
return -ENOMEM;
uiom_dev->dev = dev;
err = iommu_attach_device(pd->domain, dev);
if (err)
goto out_free_dev;
if (!iommu_capable(dev->bus, IOMMU_CAP_CACHE_COHERENCY)) {
usnic_err("IOMMU of %s does not support cache coherency\n",
dev_name(dev));
err = -EINVAL;
goto out_detach_device;
}
spin_lock(&pd->lock);
list_add_tail(&uiom_dev->link, &pd->devs);
pd->dev_cnt++;
spin_unlock(&pd->lock);
return 0;
out_detach_device:
iommu_detach_device(pd->domain, dev);
out_free_dev:
kfree(uiom_dev);
return err;
}
void usnic_uiom_detach_dev_from_pd(struct usnic_uiom_pd *pd, struct device *dev)
{
struct usnic_uiom_dev *uiom_dev;
int found = 0;
spin_lock(&pd->lock);
list_for_each_entry(uiom_dev, &pd->devs, link) {
if (uiom_dev->dev == dev) {
found = 1;
break;
}
}
if (!found) {
usnic_err("Unable to free dev %s - not found\n",
dev_name(dev));
spin_unlock(&pd->lock);
return;
}
list_del(&uiom_dev->link);
pd->dev_cnt--;
spin_unlock(&pd->lock);
return iommu_detach_device(pd->domain, dev);
}
struct device **usnic_uiom_get_dev_list(struct usnic_uiom_pd *pd)
{
struct usnic_uiom_dev *uiom_dev;
struct device **devs;
int i = 0;
spin_lock(&pd->lock);
devs = kcalloc(pd->dev_cnt + 1, sizeof(*devs), GFP_ATOMIC);
if (!devs) {
devs = ERR_PTR(-ENOMEM);
goto out;
}
list_for_each_entry(uiom_dev, &pd->devs, link) {
devs[i++] = uiom_dev->dev;
}
out:
spin_unlock(&pd->lock);
return devs;
}
void usnic_uiom_free_dev_list(struct device **devs)
{
kfree(devs);
}
int usnic_uiom_init(char *drv_name)
{
if (!iommu_present(&pci_bus_type)) {
usnic_err("IOMMU required but not present or enabled. USNIC QPs will not function w/o enabling IOMMU\n");
return -EPERM;
}
usnic_uiom_wq = create_workqueue(drv_name);
if (!usnic_uiom_wq) {
usnic_err("Unable to alloc wq for drv %s\n", drv_name);
return -ENOMEM;
}
return 0;
}
void usnic_uiom_fini(void)
{
flush_workqueue(usnic_uiom_wq);
destroy_workqueue(usnic_uiom_wq);
}