linux_dsm_epyc7002/drivers/xen/gntalloc.c
Michal Hocko 0ee931c4e3 mm: treewide: remove GFP_TEMPORARY allocation flag
GFP_TEMPORARY was introduced by commit e12ba74d8f ("Group short-lived
and reclaimable kernel allocations") along with __GFP_RECLAIMABLE.  It's
primary motivation was to allow users to tell that an allocation is
short lived and so the allocator can try to place such allocations close
together and prevent long term fragmentation.  As much as this sounds
like a reasonable semantic it becomes much less clear when to use the
highlevel GFP_TEMPORARY allocation flag.  How long is temporary? Can the
context holding that memory sleep? Can it take locks? It seems there is
no good answer for those questions.

The current implementation of GFP_TEMPORARY is basically GFP_KERNEL |
__GFP_RECLAIMABLE which in itself is tricky because basically none of
the existing caller provide a way to reclaim the allocated memory.  So
this is rather misleading and hard to evaluate for any benefits.

I have checked some random users and none of them has added the flag
with a specific justification.  I suspect most of them just copied from
other existing users and others just thought it might be a good idea to
use without any measuring.  This suggests that GFP_TEMPORARY just
motivates for cargo cult usage without any reasoning.

I believe that our gfp flags are quite complex already and especially
those with highlevel semantic should be clearly defined to prevent from
confusion and abuse.  Therefore I propose dropping GFP_TEMPORARY and
replace all existing users to simply use GFP_KERNEL.  Please note that
SLAB users with shrinkers will still get __GFP_RECLAIMABLE heuristic and
so they will be placed properly for memory fragmentation prevention.

I can see reasons we might want some gfp flag to reflect shorterm
allocations but I propose starting from a clear semantic definition and
only then add users with proper justification.

This was been brought up before LSF this year by Matthew [1] and it
turned out that GFP_TEMPORARY really doesn't have a clear semantic.  It
seems to be a heuristic without any measured advantage for most (if not
all) its current users.  The follow up discussion has revealed that
opinions on what might be temporary allocation differ a lot between
developers.  So rather than trying to tweak existing users into a
semantic which they haven't expected I propose to simply remove the flag
and start from scratch if we really need a semantic for short term
allocations.

[1] http://lkml.kernel.org/r/20170118054945.GD18349@bombadil.infradead.org

[akpm@linux-foundation.org: fix typo]
[akpm@linux-foundation.org: coding-style fixes]
[sfr@canb.auug.org.au: drm/i915: fix up]
  Link: http://lkml.kernel.org/r/20170816144703.378d4f4d@canb.auug.org.au
Link: http://lkml.kernel.org/r/20170728091904.14627-1-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Neil Brown <neilb@suse.de>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-09-13 18:53:16 -07:00

613 lines
15 KiB
C

/******************************************************************************
* gntalloc.c
*
* Device for creating grant references (in user-space) that may be shared
* with other domains.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
/*
* This driver exists to allow userspace programs in Linux to allocate kernel
* memory that will later be shared with another domain. Without this device,
* Linux userspace programs cannot create grant references.
*
* How this stuff works:
* X -> granting a page to Y
* Y -> mapping the grant from X
*
* 1. X uses the gntalloc device to allocate a page of kernel memory, P.
* 2. X creates an entry in the grant table that says domid(Y) can access P.
* This is done without a hypercall unless the grant table needs expansion.
* 3. X gives the grant reference identifier, GREF, to Y.
* 4. Y maps the page, either directly into kernel memory for use in a backend
* driver, or via a the gntdev device to map into the address space of an
* application running in Y. This is the first point at which Xen does any
* tracking of the page.
* 5. A program in X mmap()s a segment of the gntalloc device that corresponds
* to the shared page, and can now communicate with Y over the shared page.
*
*
* NOTE TO USERSPACE LIBRARIES:
* The grant allocation and mmap()ing are, naturally, two separate operations.
* You set up the sharing by calling the create ioctl() and then the mmap().
* Teardown requires munmap() and either close() or ioctl().
*
* WARNING: Since Xen does not allow a guest to forcibly end the use of a grant
* reference, this device can be used to consume kernel memory by leaving grant
* references mapped by another domain when an application exits. Therefore,
* there is a global limit on the number of pages that can be allocated. When
* all references to the page are unmapped, it will be freed during the next
* grant operation.
*/
#define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt
#include <linux/atomic.h>
#include <linux/module.h>
#include <linux/miscdevice.h>
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/slab.h>
#include <linux/fs.h>
#include <linux/device.h>
#include <linux/mm.h>
#include <linux/uaccess.h>
#include <linux/types.h>
#include <linux/list.h>
#include <linux/highmem.h>
#include <xen/xen.h>
#include <xen/page.h>
#include <xen/grant_table.h>
#include <xen/gntalloc.h>
#include <xen/events.h>
static int limit = 1024;
module_param(limit, int, 0644);
MODULE_PARM_DESC(limit, "Maximum number of grants that may be allocated by "
"the gntalloc device");
static LIST_HEAD(gref_list);
static DEFINE_MUTEX(gref_mutex);
static int gref_size;
struct notify_info {
uint16_t pgoff:12; /* Bits 0-11: Offset of the byte to clear */
uint16_t flags:2; /* Bits 12-13: Unmap notification flags */
int event; /* Port (event channel) to notify */
};
/* Metadata on a grant reference. */
struct gntalloc_gref {
struct list_head next_gref; /* list entry gref_list */
struct list_head next_file; /* list entry file->list, if open */
struct page *page; /* The shared page */
uint64_t file_index; /* File offset for mmap() */
unsigned int users; /* Use count - when zero, waiting on Xen */
grant_ref_t gref_id; /* The grant reference number */
struct notify_info notify; /* Unmap notification */
};
struct gntalloc_file_private_data {
struct list_head list;
uint64_t index;
};
struct gntalloc_vma_private_data {
struct gntalloc_gref *gref;
int users;
int count;
};
static void __del_gref(struct gntalloc_gref *gref);
static void do_cleanup(void)
{
struct gntalloc_gref *gref, *n;
list_for_each_entry_safe(gref, n, &gref_list, next_gref) {
if (!gref->users)
__del_gref(gref);
}
}
static int add_grefs(struct ioctl_gntalloc_alloc_gref *op,
uint32_t *gref_ids, struct gntalloc_file_private_data *priv)
{
int i, rc, readonly;
LIST_HEAD(queue_gref);
LIST_HEAD(queue_file);
struct gntalloc_gref *gref, *next;
readonly = !(op->flags & GNTALLOC_FLAG_WRITABLE);
for (i = 0; i < op->count; i++) {
gref = kzalloc(sizeof(*gref), GFP_KERNEL);
if (!gref) {
rc = -ENOMEM;
goto undo;
}
list_add_tail(&gref->next_gref, &queue_gref);
list_add_tail(&gref->next_file, &queue_file);
gref->users = 1;
gref->file_index = op->index + i * PAGE_SIZE;
gref->page = alloc_page(GFP_KERNEL|__GFP_ZERO);
if (!gref->page) {
rc = -ENOMEM;
goto undo;
}
/* Grant foreign access to the page. */
rc = gnttab_grant_foreign_access(op->domid,
xen_page_to_gfn(gref->page),
readonly);
if (rc < 0)
goto undo;
gref_ids[i] = gref->gref_id = rc;
}
/* Add to gref lists. */
mutex_lock(&gref_mutex);
list_splice_tail(&queue_gref, &gref_list);
list_splice_tail(&queue_file, &priv->list);
mutex_unlock(&gref_mutex);
return 0;
undo:
mutex_lock(&gref_mutex);
gref_size -= (op->count - i);
list_for_each_entry_safe(gref, next, &queue_file, next_file) {
list_del(&gref->next_file);
__del_gref(gref);
}
/* It's possible for the target domain to map the just-allocated grant
* references by blindly guessing their IDs; if this is done, then
* __del_gref will leave them in the queue_gref list. They need to be
* added to the global list so that we can free them when they are no
* longer referenced.
*/
if (unlikely(!list_empty(&queue_gref)))
list_splice_tail(&queue_gref, &gref_list);
mutex_unlock(&gref_mutex);
return rc;
}
static void __del_gref(struct gntalloc_gref *gref)
{
if (gref->notify.flags & UNMAP_NOTIFY_CLEAR_BYTE) {
uint8_t *tmp = kmap(gref->page);
tmp[gref->notify.pgoff] = 0;
kunmap(gref->page);
}
if (gref->notify.flags & UNMAP_NOTIFY_SEND_EVENT) {
notify_remote_via_evtchn(gref->notify.event);
evtchn_put(gref->notify.event);
}
gref->notify.flags = 0;
if (gref->gref_id) {
if (gnttab_query_foreign_access(gref->gref_id))
return;
if (!gnttab_end_foreign_access_ref(gref->gref_id, 0))
return;
gnttab_free_grant_reference(gref->gref_id);
}
gref_size--;
list_del(&gref->next_gref);
if (gref->page)
__free_page(gref->page);
kfree(gref);
}
/* finds contiguous grant references in a file, returns the first */
static struct gntalloc_gref *find_grefs(struct gntalloc_file_private_data *priv,
uint64_t index, uint32_t count)
{
struct gntalloc_gref *rv = NULL, *gref;
list_for_each_entry(gref, &priv->list, next_file) {
if (gref->file_index == index && !rv)
rv = gref;
if (rv) {
if (gref->file_index != index)
return NULL;
index += PAGE_SIZE;
count--;
if (count == 0)
return rv;
}
}
return NULL;
}
/*
* -------------------------------------
* File operations.
* -------------------------------------
*/
static int gntalloc_open(struct inode *inode, struct file *filp)
{
struct gntalloc_file_private_data *priv;
priv = kzalloc(sizeof(*priv), GFP_KERNEL);
if (!priv)
goto out_nomem;
INIT_LIST_HEAD(&priv->list);
filp->private_data = priv;
pr_debug("%s: priv %p\n", __func__, priv);
return 0;
out_nomem:
return -ENOMEM;
}
static int gntalloc_release(struct inode *inode, struct file *filp)
{
struct gntalloc_file_private_data *priv = filp->private_data;
struct gntalloc_gref *gref;
pr_debug("%s: priv %p\n", __func__, priv);
mutex_lock(&gref_mutex);
while (!list_empty(&priv->list)) {
gref = list_entry(priv->list.next,
struct gntalloc_gref, next_file);
list_del(&gref->next_file);
gref->users--;
if (gref->users == 0)
__del_gref(gref);
}
kfree(priv);
mutex_unlock(&gref_mutex);
return 0;
}
static long gntalloc_ioctl_alloc(struct gntalloc_file_private_data *priv,
struct ioctl_gntalloc_alloc_gref __user *arg)
{
int rc = 0;
struct ioctl_gntalloc_alloc_gref op;
uint32_t *gref_ids;
pr_debug("%s: priv %p\n", __func__, priv);
if (copy_from_user(&op, arg, sizeof(op))) {
rc = -EFAULT;
goto out;
}
gref_ids = kcalloc(op.count, sizeof(gref_ids[0]), GFP_KERNEL);
if (!gref_ids) {
rc = -ENOMEM;
goto out;
}
mutex_lock(&gref_mutex);
/* Clean up pages that were at zero (local) users but were still mapped
* by remote domains. Since those pages count towards the limit that we
* are about to enforce, removing them here is a good idea.
*/
do_cleanup();
if (gref_size + op.count > limit) {
mutex_unlock(&gref_mutex);
rc = -ENOSPC;
goto out_free;
}
gref_size += op.count;
op.index = priv->index;
priv->index += op.count * PAGE_SIZE;
mutex_unlock(&gref_mutex);
rc = add_grefs(&op, gref_ids, priv);
if (rc < 0)
goto out_free;
/* Once we finish add_grefs, it is unsafe to touch the new reference,
* since it is possible for a concurrent ioctl to remove it (by guessing
* its index). If the userspace application doesn't provide valid memory
* to write the IDs to, then it will need to close the file in order to
* release - which it will do by segfaulting when it tries to access the
* IDs to close them.
*/
if (copy_to_user(arg, &op, sizeof(op))) {
rc = -EFAULT;
goto out_free;
}
if (copy_to_user(arg->gref_ids, gref_ids,
sizeof(gref_ids[0]) * op.count)) {
rc = -EFAULT;
goto out_free;
}
out_free:
kfree(gref_ids);
out:
return rc;
}
static long gntalloc_ioctl_dealloc(struct gntalloc_file_private_data *priv,
void __user *arg)
{
int i, rc = 0;
struct ioctl_gntalloc_dealloc_gref op;
struct gntalloc_gref *gref, *n;
pr_debug("%s: priv %p\n", __func__, priv);
if (copy_from_user(&op, arg, sizeof(op))) {
rc = -EFAULT;
goto dealloc_grant_out;
}
mutex_lock(&gref_mutex);
gref = find_grefs(priv, op.index, op.count);
if (gref) {
/* Remove from the file list only, and decrease reference count.
* The later call to do_cleanup() will remove from gref_list and
* free the memory if the pages aren't mapped anywhere.
*/
for (i = 0; i < op.count; i++) {
n = list_entry(gref->next_file.next,
struct gntalloc_gref, next_file);
list_del(&gref->next_file);
gref->users--;
gref = n;
}
} else {
rc = -EINVAL;
}
do_cleanup();
mutex_unlock(&gref_mutex);
dealloc_grant_out:
return rc;
}
static long gntalloc_ioctl_unmap_notify(struct gntalloc_file_private_data *priv,
void __user *arg)
{
struct ioctl_gntalloc_unmap_notify op;
struct gntalloc_gref *gref;
uint64_t index;
int pgoff;
int rc;
if (copy_from_user(&op, arg, sizeof(op)))
return -EFAULT;
index = op.index & ~(PAGE_SIZE - 1);
pgoff = op.index & (PAGE_SIZE - 1);
mutex_lock(&gref_mutex);
gref = find_grefs(priv, index, 1);
if (!gref) {
rc = -ENOENT;
goto unlock_out;
}
if (op.action & ~(UNMAP_NOTIFY_CLEAR_BYTE|UNMAP_NOTIFY_SEND_EVENT)) {
rc = -EINVAL;
goto unlock_out;
}
/* We need to grab a reference to the event channel we are going to use
* to send the notify before releasing the reference we may already have
* (if someone has called this ioctl twice). This is required so that
* it is possible to change the clear_byte part of the notification
* without disturbing the event channel part, which may now be the last
* reference to that event channel.
*/
if (op.action & UNMAP_NOTIFY_SEND_EVENT) {
if (evtchn_get(op.event_channel_port)) {
rc = -EINVAL;
goto unlock_out;
}
}
if (gref->notify.flags & UNMAP_NOTIFY_SEND_EVENT)
evtchn_put(gref->notify.event);
gref->notify.flags = op.action;
gref->notify.pgoff = pgoff;
gref->notify.event = op.event_channel_port;
rc = 0;
unlock_out:
mutex_unlock(&gref_mutex);
return rc;
}
static long gntalloc_ioctl(struct file *filp, unsigned int cmd,
unsigned long arg)
{
struct gntalloc_file_private_data *priv = filp->private_data;
switch (cmd) {
case IOCTL_GNTALLOC_ALLOC_GREF:
return gntalloc_ioctl_alloc(priv, (void __user *)arg);
case IOCTL_GNTALLOC_DEALLOC_GREF:
return gntalloc_ioctl_dealloc(priv, (void __user *)arg);
case IOCTL_GNTALLOC_SET_UNMAP_NOTIFY:
return gntalloc_ioctl_unmap_notify(priv, (void __user *)arg);
default:
return -ENOIOCTLCMD;
}
return 0;
}
static void gntalloc_vma_open(struct vm_area_struct *vma)
{
struct gntalloc_vma_private_data *priv = vma->vm_private_data;
if (!priv)
return;
mutex_lock(&gref_mutex);
priv->users++;
mutex_unlock(&gref_mutex);
}
static void gntalloc_vma_close(struct vm_area_struct *vma)
{
struct gntalloc_vma_private_data *priv = vma->vm_private_data;
struct gntalloc_gref *gref, *next;
int i;
if (!priv)
return;
mutex_lock(&gref_mutex);
priv->users--;
if (priv->users == 0) {
gref = priv->gref;
for (i = 0; i < priv->count; i++) {
gref->users--;
next = list_entry(gref->next_gref.next,
struct gntalloc_gref, next_gref);
if (gref->users == 0)
__del_gref(gref);
gref = next;
}
kfree(priv);
}
mutex_unlock(&gref_mutex);
}
static const struct vm_operations_struct gntalloc_vmops = {
.open = gntalloc_vma_open,
.close = gntalloc_vma_close,
};
static int gntalloc_mmap(struct file *filp, struct vm_area_struct *vma)
{
struct gntalloc_file_private_data *priv = filp->private_data;
struct gntalloc_vma_private_data *vm_priv;
struct gntalloc_gref *gref;
int count = vma_pages(vma);
int rv, i;
if (!(vma->vm_flags & VM_SHARED)) {
pr_err("%s: Mapping must be shared\n", __func__);
return -EINVAL;
}
vm_priv = kmalloc(sizeof(*vm_priv), GFP_KERNEL);
if (!vm_priv)
return -ENOMEM;
mutex_lock(&gref_mutex);
pr_debug("%s: priv %p,%p, page %lu+%d\n", __func__,
priv, vm_priv, vma->vm_pgoff, count);
gref = find_grefs(priv, vma->vm_pgoff << PAGE_SHIFT, count);
if (gref == NULL) {
rv = -ENOENT;
pr_debug("%s: Could not find grant reference",
__func__);
kfree(vm_priv);
goto out_unlock;
}
vm_priv->gref = gref;
vm_priv->users = 1;
vm_priv->count = count;
vma->vm_private_data = vm_priv;
vma->vm_flags |= VM_DONTEXPAND | VM_DONTDUMP;
vma->vm_ops = &gntalloc_vmops;
for (i = 0; i < count; i++) {
gref->users++;
rv = vm_insert_page(vma, vma->vm_start + i * PAGE_SIZE,
gref->page);
if (rv)
goto out_unlock;
gref = list_entry(gref->next_file.next,
struct gntalloc_gref, next_file);
}
rv = 0;
out_unlock:
mutex_unlock(&gref_mutex);
return rv;
}
static const struct file_operations gntalloc_fops = {
.owner = THIS_MODULE,
.open = gntalloc_open,
.release = gntalloc_release,
.unlocked_ioctl = gntalloc_ioctl,
.mmap = gntalloc_mmap
};
/*
* -------------------------------------
* Module creation/destruction.
* -------------------------------------
*/
static struct miscdevice gntalloc_miscdev = {
.minor = MISC_DYNAMIC_MINOR,
.name = "xen/gntalloc",
.fops = &gntalloc_fops,
};
static int __init gntalloc_init(void)
{
int err;
if (!xen_domain())
return -ENODEV;
err = misc_register(&gntalloc_miscdev);
if (err != 0) {
pr_err("Could not register misc gntalloc device\n");
return err;
}
pr_debug("Created grant allocation device at %d,%d\n",
MISC_MAJOR, gntalloc_miscdev.minor);
return 0;
}
static void __exit gntalloc_exit(void)
{
misc_deregister(&gntalloc_miscdev);
}
module_init(gntalloc_init);
module_exit(gntalloc_exit);
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Carter Weatherly <carter.weatherly@jhuapl.edu>, "
"Daniel De Graaf <dgdegra@tycho.nsa.gov>");
MODULE_DESCRIPTION("User-space grant reference allocator driver");