2018-03-29 20:07:13 -06:00
|
|
|
/* SPDX-License-Identifier: GPL-2.0 */
|
|
|
|
/* Copyright(c) 2015 Intel Corporation. All rights reserved. */
|
2015-08-10 21:07:07 -06:00
|
|
|
#include <linux/device.h>
|
2015-08-10 21:07:06 -06:00
|
|
|
#include <linux/io.h>
|
2018-08-17 16:47:04 -06:00
|
|
|
#include <linux/kasan.h>
|
2015-08-17 08:00:35 -06:00
|
|
|
#include <linux/memory_hotplug.h>
|
2018-08-15 12:22:16 -06:00
|
|
|
#include <linux/mm.h>
|
|
|
|
#include <linux/pfn_t.h>
|
2017-09-08 17:11:43 -06:00
|
|
|
#include <linux/swap.h>
|
|
|
|
#include <linux/swapops.h>
|
2018-08-15 12:22:16 -06:00
|
|
|
#include <linux/types.h>
|
2018-05-16 12:46:08 -06:00
|
|
|
#include <linux/wait_bit.h>
|
2018-08-15 12:22:16 -06:00
|
|
|
#include <linux/xarray.h>
|
2015-08-10 21:07:06 -06:00
|
|
|
|
2018-08-15 12:22:16 -06:00
|
|
|
static DEFINE_XARRAY(pgmap_array);
|
2016-01-15 17:56:19 -07:00
|
|
|
|
2019-06-26 06:27:10 -06:00
|
|
|
#ifdef CONFIG_DEV_PAGEMAP_OPS
|
|
|
|
DEFINE_STATIC_KEY_FALSE(devmap_managed_key);
|
|
|
|
EXPORT_SYMBOL(devmap_managed_key);
|
|
|
|
static atomic_t devmap_managed_enable;
|
|
|
|
|
2019-08-18 03:05:56 -06:00
|
|
|
static void devmap_managed_enable_put(void)
|
2019-06-26 06:27:10 -06:00
|
|
|
{
|
|
|
|
if (atomic_dec_and_test(&devmap_managed_enable))
|
|
|
|
static_branch_disable(&devmap_managed_key);
|
|
|
|
}
|
|
|
|
|
2019-08-18 03:05:56 -06:00
|
|
|
static int devmap_managed_enable_get(struct dev_pagemap *pgmap)
|
2019-06-26 06:27:10 -06:00
|
|
|
{
|
2019-06-26 06:27:14 -06:00
|
|
|
if (!pgmap->ops || !pgmap->ops->page_free) {
|
2019-06-26 06:27:10 -06:00
|
|
|
WARN(1, "Missing page_free method\n");
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (atomic_inc_return(&devmap_managed_enable) == 1)
|
|
|
|
static_branch_enable(&devmap_managed_key);
|
2019-08-18 03:05:56 -06:00
|
|
|
return 0;
|
2019-06-26 06:27:10 -06:00
|
|
|
}
|
|
|
|
#else
|
2019-08-18 03:05:56 -06:00
|
|
|
static int devmap_managed_enable_get(struct dev_pagemap *pgmap)
|
2019-06-26 06:27:10 -06:00
|
|
|
{
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
2019-08-18 03:05:56 -06:00
|
|
|
static void devmap_managed_enable_put(void)
|
|
|
|
{
|
|
|
|
}
|
2019-06-26 06:27:10 -06:00
|
|
|
#endif /* CONFIG_DEV_PAGEMAP_OPS */
|
|
|
|
|
2018-08-15 12:22:16 -06:00
|
|
|
static void pgmap_array_delete(struct resource *res)
|
2017-09-06 17:24:13 -06:00
|
|
|
{
|
2018-08-15 12:22:16 -06:00
|
|
|
xa_store_range(&pgmap_array, PHYS_PFN(res->start), PHYS_PFN(res->end),
|
|
|
|
NULL, GFP_KERNEL);
|
2017-09-06 17:24:13 -06:00
|
|
|
synchronize_rcu();
|
2016-01-15 17:56:19 -07:00
|
|
|
}
|
|
|
|
|
2017-12-29 00:54:04 -07:00
|
|
|
static unsigned long pfn_first(struct dev_pagemap *pgmap)
|
2016-01-15 17:56:49 -07:00
|
|
|
{
|
2019-07-18 16:58:33 -06:00
|
|
|
return PHYS_PFN(pgmap->res.start) +
|
2019-06-26 06:27:13 -06:00
|
|
|
vmem_altmap_offset(pgmap_altmap(pgmap));
|
2016-01-15 17:56:49 -07:00
|
|
|
}
|
|
|
|
|
2017-12-29 00:54:04 -07:00
|
|
|
static unsigned long pfn_end(struct dev_pagemap *pgmap)
|
2016-01-15 17:56:49 -07:00
|
|
|
{
|
2017-12-29 00:54:04 -07:00
|
|
|
const struct resource *res = &pgmap->res;
|
2016-01-15 17:56:49 -07:00
|
|
|
|
|
|
|
return (res->start + resource_size(res)) >> PAGE_SHIFT;
|
|
|
|
}
|
|
|
|
|
2018-02-06 20:34:11 -07:00
|
|
|
static unsigned long pfn_next(unsigned long pfn)
|
|
|
|
{
|
|
|
|
if (pfn % 1024 == 0)
|
|
|
|
cond_resched();
|
|
|
|
return pfn + 1;
|
|
|
|
}
|
|
|
|
|
2016-01-15 17:56:49 -07:00
|
|
|
#define for_each_device_pfn(pfn, map) \
|
2018-02-06 20:34:11 -07:00
|
|
|
for (pfn = pfn_first(map); pfn < pfn_end(map); pfn = pfn_next(pfn))
|
2016-01-15 17:56:49 -07:00
|
|
|
|
2019-06-26 06:27:14 -06:00
|
|
|
static void dev_pagemap_kill(struct dev_pagemap *pgmap)
|
|
|
|
{
|
|
|
|
if (pgmap->ops && pgmap->ops->kill)
|
|
|
|
pgmap->ops->kill(pgmap);
|
|
|
|
else
|
|
|
|
percpu_ref_kill(pgmap->ref);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void dev_pagemap_cleanup(struct dev_pagemap *pgmap)
|
|
|
|
{
|
|
|
|
if (pgmap->ops && pgmap->ops->cleanup) {
|
|
|
|
pgmap->ops->cleanup(pgmap);
|
|
|
|
} else {
|
|
|
|
wait_for_completion(&pgmap->done);
|
|
|
|
percpu_ref_exit(pgmap->ref);
|
|
|
|
}
|
2019-08-08 15:43:49 -06:00
|
|
|
/*
|
|
|
|
* Undo the pgmap ref assignment for the internal case as the
|
|
|
|
* caller may re-enable the same pgmap.
|
|
|
|
*/
|
|
|
|
if (pgmap->ref == &pgmap->internal_ref)
|
|
|
|
pgmap->ref = NULL;
|
2019-06-26 06:27:14 -06:00
|
|
|
}
|
|
|
|
|
2019-08-18 03:05:57 -06:00
|
|
|
void memunmap_pages(struct dev_pagemap *pgmap)
|
2015-08-17 08:00:35 -06:00
|
|
|
{
|
2017-12-29 00:54:04 -07:00
|
|
|
struct resource *res = &pgmap->res;
|
2017-04-28 11:23:37 -06:00
|
|
|
unsigned long pfn;
|
2018-12-28 01:36:22 -07:00
|
|
|
int nid;
|
2017-04-28 11:23:37 -06:00
|
|
|
|
2019-06-26 06:27:14 -06:00
|
|
|
dev_pagemap_kill(pgmap);
|
2017-12-29 00:54:04 -07:00
|
|
|
for_each_device_pfn(pfn, pgmap)
|
2017-04-28 11:23:37 -06:00
|
|
|
put_page(pfn_to_page(pfn));
|
2019-06-26 06:27:14 -06:00
|
|
|
dev_pagemap_cleanup(pgmap);
|
2016-01-15 17:56:19 -07:00
|
|
|
|
2015-08-17 08:00:35 -06:00
|
|
|
/* pages are dead and unused, undo the arch mapping */
|
2019-07-18 16:58:33 -06:00
|
|
|
nid = page_to_nid(pfn_to_page(PHYS_PFN(res->start)));
|
2018-12-28 01:36:22 -07:00
|
|
|
|
mm: fix devm_memremap_pages crash, use mem_hotplug_{begin, done}
Both arch_add_memory() and arch_remove_memory() expect a single threaded
context.
For example, arch/x86/mm/init_64.c::kernel_physical_mapping_init() does
not hold any locks over this check and branch:
if (pgd_val(*pgd)) {
pud = (pud_t *)pgd_page_vaddr(*pgd);
paddr_last = phys_pud_init(pud, __pa(vaddr),
__pa(vaddr_end),
page_size_mask);
continue;
}
pud = alloc_low_page();
paddr_last = phys_pud_init(pud, __pa(vaddr), __pa(vaddr_end),
page_size_mask);
The result is that two threads calling devm_memremap_pages()
simultaneously can end up colliding on pgd initialization. This leads
to crash signatures like the following where the loser of the race
initializes the wrong pgd entry:
BUG: unable to handle kernel paging request at ffff888ebfff0000
IP: memcpy_erms+0x6/0x10
PGD 2f8e8fc067 PUD 0 /* <---- Invalid PUD */
Oops: 0000 [#1] SMP DEBUG_PAGEALLOC
CPU: 54 PID: 3818 Comm: systemd-udevd Not tainted 4.6.7+ #13
task: ffff882fac290040 ti: ffff882f887a4000 task.ti: ffff882f887a4000
RIP: memcpy_erms+0x6/0x10
[..]
Call Trace:
? pmem_do_bvec+0x205/0x370 [nd_pmem]
? blk_queue_enter+0x3a/0x280
pmem_rw_page+0x38/0x80 [nd_pmem]
bdev_read_page+0x84/0xb0
Hold the standard memory hotplug mutex over calls to
arch_{add,remove}_memory().
Fixes: 41e94a851304 ("add devm_memremap_pages")
Link: http://lkml.kernel.org/r/148357647831.9498.12606007370121652979.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-01-10 17:57:36 -07:00
|
|
|
mem_hotplug_begin();
|
2018-12-28 01:35:01 -07:00
|
|
|
if (pgmap->type == MEMORY_DEVICE_PRIVATE) {
|
2019-07-18 16:58:33 -06:00
|
|
|
pfn = PHYS_PFN(res->start);
|
2018-12-28 01:35:01 -07:00
|
|
|
__remove_pages(page_zone(pfn_to_page(pfn)), pfn,
|
2019-07-18 16:58:33 -06:00
|
|
|
PHYS_PFN(resource_size(res)), NULL);
|
2018-12-28 01:35:01 -07:00
|
|
|
} else {
|
2019-07-18 16:58:33 -06:00
|
|
|
arch_remove_memory(nid, res->start, resource_size(res),
|
2019-06-26 06:27:13 -06:00
|
|
|
pgmap_altmap(pgmap));
|
2019-07-18 16:58:33 -06:00
|
|
|
kasan_remove_zero_shadow(__va(res->start), resource_size(res));
|
2018-12-28 01:35:01 -07:00
|
|
|
}
|
mm: fix devm_memremap_pages crash, use mem_hotplug_{begin, done}
Both arch_add_memory() and arch_remove_memory() expect a single threaded
context.
For example, arch/x86/mm/init_64.c::kernel_physical_mapping_init() does
not hold any locks over this check and branch:
if (pgd_val(*pgd)) {
pud = (pud_t *)pgd_page_vaddr(*pgd);
paddr_last = phys_pud_init(pud, __pa(vaddr),
__pa(vaddr_end),
page_size_mask);
continue;
}
pud = alloc_low_page();
paddr_last = phys_pud_init(pud, __pa(vaddr), __pa(vaddr_end),
page_size_mask);
The result is that two threads calling devm_memremap_pages()
simultaneously can end up colliding on pgd initialization. This leads
to crash signatures like the following where the loser of the race
initializes the wrong pgd entry:
BUG: unable to handle kernel paging request at ffff888ebfff0000
IP: memcpy_erms+0x6/0x10
PGD 2f8e8fc067 PUD 0 /* <---- Invalid PUD */
Oops: 0000 [#1] SMP DEBUG_PAGEALLOC
CPU: 54 PID: 3818 Comm: systemd-udevd Not tainted 4.6.7+ #13
task: ffff882fac290040 ti: ffff882f887a4000 task.ti: ffff882f887a4000
RIP: memcpy_erms+0x6/0x10
[..]
Call Trace:
? pmem_do_bvec+0x205/0x370 [nd_pmem]
? blk_queue_enter+0x3a/0x280
pmem_rw_page+0x38/0x80 [nd_pmem]
bdev_read_page+0x84/0xb0
Hold the standard memory hotplug mutex over calls to
arch_{add,remove}_memory().
Fixes: 41e94a851304 ("add devm_memremap_pages")
Link: http://lkml.kernel.org/r/148357647831.9498.12606007370121652979.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-01-10 17:57:36 -07:00
|
|
|
mem_hotplug_done();
|
2017-02-24 15:55:45 -07:00
|
|
|
|
2019-07-18 16:58:33 -06:00
|
|
|
untrack_pfn(NULL, PHYS_PFN(res->start), resource_size(res));
|
2018-08-15 12:22:16 -06:00
|
|
|
pgmap_array_delete(res);
|
2019-08-18 03:05:55 -06:00
|
|
|
WARN_ONCE(pgmap->altmap.alloc, "failed to free all reserved pages\n");
|
2019-08-18 03:05:56 -06:00
|
|
|
devmap_managed_enable_put();
|
2016-01-15 17:56:19 -07:00
|
|
|
}
|
2019-08-18 03:05:57 -06:00
|
|
|
EXPORT_SYMBOL_GPL(memunmap_pages);
|
|
|
|
|
|
|
|
static void devm_memremap_pages_release(void *data)
|
|
|
|
{
|
|
|
|
memunmap_pages(data);
|
|
|
|
}
|
2016-01-15 17:56:19 -07:00
|
|
|
|
2019-06-26 06:27:14 -06:00
|
|
|
static void dev_pagemap_percpu_release(struct percpu_ref *ref)
|
|
|
|
{
|
|
|
|
struct dev_pagemap *pgmap =
|
|
|
|
container_of(ref, struct dev_pagemap, internal_ref);
|
|
|
|
|
|
|
|
complete(&pgmap->done);
|
|
|
|
}
|
|
|
|
|
2019-08-18 03:05:57 -06:00
|
|
|
/*
|
|
|
|
* Not device managed version of dev_memremap_pages, undone by
|
|
|
|
* memunmap_pages(). Please use dev_memremap_pages if you have a struct
|
|
|
|
* device available.
|
2016-01-15 17:56:22 -07:00
|
|
|
*/
|
2019-08-18 03:05:57 -06:00
|
|
|
void *memremap_pages(struct dev_pagemap *pgmap, int nid)
|
2015-08-17 08:00:35 -06:00
|
|
|
{
|
2018-02-06 20:34:11 -07:00
|
|
|
struct resource *res = &pgmap->res;
|
2018-10-26 16:07:52 -06:00
|
|
|
struct dev_pagemap *conflict_pgmap;
|
2019-05-13 18:21:26 -06:00
|
|
|
struct mhp_restrictions restrictions = {
|
|
|
|
/*
|
|
|
|
* We do not want any optional features only our own memmap
|
2019-07-18 16:58:33 -06:00
|
|
|
*/
|
2019-06-26 06:27:13 -06:00
|
|
|
.altmap = pgmap_altmap(pgmap),
|
2019-05-13 18:21:26 -06:00
|
|
|
};
|
2016-09-07 09:51:21 -06:00
|
|
|
pgprot_t pgprot = PAGE_KERNEL;
|
2019-08-18 03:05:57 -06:00
|
|
|
int error, is_ram;
|
2019-06-26 06:27:10 -06:00
|
|
|
bool need_devmap_managed = true;
|
2016-03-09 15:08:13 -07:00
|
|
|
|
2019-06-26 06:27:07 -06:00
|
|
|
switch (pgmap->type) {
|
|
|
|
case MEMORY_DEVICE_PRIVATE:
|
|
|
|
if (!IS_ENABLED(CONFIG_DEVICE_PRIVATE)) {
|
|
|
|
WARN(1, "Device private memory not supported\n");
|
|
|
|
return ERR_PTR(-EINVAL);
|
|
|
|
}
|
2019-06-26 06:27:11 -06:00
|
|
|
if (!pgmap->ops || !pgmap->ops->migrate_to_ram) {
|
|
|
|
WARN(1, "Missing migrate_to_ram method\n");
|
|
|
|
return ERR_PTR(-EINVAL);
|
|
|
|
}
|
2019-06-26 06:27:07 -06:00
|
|
|
break;
|
|
|
|
case MEMORY_DEVICE_FS_DAX:
|
|
|
|
if (!IS_ENABLED(CONFIG_ZONE_DEVICE) ||
|
|
|
|
IS_ENABLED(CONFIG_FS_DAX_LIMITED)) {
|
|
|
|
WARN(1, "File system DAX not supported\n");
|
|
|
|
return ERR_PTR(-EINVAL);
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
case MEMORY_DEVICE_DEVDAX:
|
|
|
|
case MEMORY_DEVICE_PCI_P2PDMA:
|
2019-06-26 06:27:10 -06:00
|
|
|
need_devmap_managed = false;
|
2019-06-26 06:27:07 -06:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
WARN(1, "Invalid pgmap type %d\n", pgmap->type);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2019-06-26 06:27:14 -06:00
|
|
|
if (!pgmap->ref) {
|
|
|
|
if (pgmap->ops && (pgmap->ops->kill || pgmap->ops->cleanup))
|
|
|
|
return ERR_PTR(-EINVAL);
|
|
|
|
|
|
|
|
init_completion(&pgmap->done);
|
|
|
|
error = percpu_ref_init(&pgmap->internal_ref,
|
|
|
|
dev_pagemap_percpu_release, 0, GFP_KERNEL);
|
|
|
|
if (error)
|
|
|
|
return ERR_PTR(error);
|
|
|
|
pgmap->ref = &pgmap->internal_ref;
|
|
|
|
} else {
|
|
|
|
if (!pgmap->ops || !pgmap->ops->kill || !pgmap->ops->cleanup) {
|
|
|
|
WARN(1, "Missing reference count teardown definition\n");
|
|
|
|
return ERR_PTR(-EINVAL);
|
|
|
|
}
|
2019-06-13 16:56:33 -06:00
|
|
|
}
|
mm, devm_memremap_pages: fix shutdown handling
The last step before devm_memremap_pages() returns success is to allocate
a release action, devm_memremap_pages_release(), to tear the entire setup
down. However, the result from devm_add_action() is not checked.
Checking the error from devm_add_action() is not enough. The api
currently relies on the fact that the percpu_ref it is using is killed by
the time the devm_memremap_pages_release() is run. Rather than continue
this awkward situation, offload the responsibility of killing the
percpu_ref to devm_memremap_pages_release() directly. This allows
devm_memremap_pages() to do the right thing relative to init failures and
shutdown.
Without this change we could fail to register the teardown of
devm_memremap_pages(). The likelihood of hitting this failure is tiny as
small memory allocations almost always succeed. However, the impact of
the failure is large given any future reconfiguration, or disable/enable,
of an nvdimm namespace will fail forever as subsequent calls to
devm_memremap_pages() will fail to setup the pgmap_radix since there will
be stale entries for the physical address range.
An argument could be made to require that the ->kill() operation be set in
the @pgmap arg rather than passed in separately. However, it helps code
readability, tracking the lifetime of a given instance, to be able to grep
the kill routine directly at the devm_memremap_pages() call site.
Link: http://lkml.kernel.org/r/154275558526.76910.7535251937849268605.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Fixes: e8d513483300 ("memremap: change devm_memremap_pages interface...")
Reviewed-by: "Jérôme Glisse" <jglisse@redhat.com>
Reported-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 01:34:57 -07:00
|
|
|
|
2019-06-26 06:27:10 -06:00
|
|
|
if (need_devmap_managed) {
|
2019-08-18 03:05:56 -06:00
|
|
|
error = devmap_managed_enable_get(pgmap);
|
2019-06-26 06:27:10 -06:00
|
|
|
if (error)
|
|
|
|
return ERR_PTR(error);
|
|
|
|
}
|
|
|
|
|
2019-07-18 16:58:33 -06:00
|
|
|
conflict_pgmap = get_dev_pagemap(PHYS_PFN(res->start), NULL);
|
2018-07-26 17:37:15 -06:00
|
|
|
if (conflict_pgmap) {
|
2019-08-18 03:05:57 -06:00
|
|
|
WARN(1, "Conflicting mapping in same section\n");
|
2018-07-26 17:37:15 -06:00
|
|
|
put_dev_pagemap(conflict_pgmap);
|
2019-06-13 16:56:33 -06:00
|
|
|
error = -ENOMEM;
|
|
|
|
goto err_array;
|
2018-07-26 17:37:15 -06:00
|
|
|
}
|
|
|
|
|
2019-07-18 16:58:33 -06:00
|
|
|
conflict_pgmap = get_dev_pagemap(PHYS_PFN(res->end), NULL);
|
2018-07-26 17:37:15 -06:00
|
|
|
if (conflict_pgmap) {
|
2019-08-18 03:05:57 -06:00
|
|
|
WARN(1, "Conflicting mapping in same section\n");
|
2018-07-26 17:37:15 -06:00
|
|
|
put_dev_pagemap(conflict_pgmap);
|
2019-06-13 16:56:33 -06:00
|
|
|
error = -ENOMEM;
|
|
|
|
goto err_array;
|
2018-07-26 17:37:15 -06:00
|
|
|
}
|
|
|
|
|
2019-07-18 16:58:33 -06:00
|
|
|
is_ram = region_intersects(res->start, resource_size(res),
|
2016-03-14 16:15:51 -06:00
|
|
|
IORESOURCE_SYSTEM_RAM, IORES_DESC_NONE);
|
2015-08-17 08:00:35 -06:00
|
|
|
|
2018-12-28 01:34:54 -07:00
|
|
|
if (is_ram != REGION_DISJOINT) {
|
|
|
|
WARN_ONCE(1, "%s attempted on %s region %pr\n", __func__,
|
|
|
|
is_ram == REGION_MIXED ? "mixed" : "ram", res);
|
mm, devm_memremap_pages: fix shutdown handling
The last step before devm_memremap_pages() returns success is to allocate
a release action, devm_memremap_pages_release(), to tear the entire setup
down. However, the result from devm_add_action() is not checked.
Checking the error from devm_add_action() is not enough. The api
currently relies on the fact that the percpu_ref it is using is killed by
the time the devm_memremap_pages_release() is run. Rather than continue
this awkward situation, offload the responsibility of killing the
percpu_ref to devm_memremap_pages_release() directly. This allows
devm_memremap_pages() to do the right thing relative to init failures and
shutdown.
Without this change we could fail to register the teardown of
devm_memremap_pages(). The likelihood of hitting this failure is tiny as
small memory allocations almost always succeed. However, the impact of
the failure is large given any future reconfiguration, or disable/enable,
of an nvdimm namespace will fail forever as subsequent calls to
devm_memremap_pages() will fail to setup the pgmap_radix since there will
be stale entries for the physical address range.
An argument could be made to require that the ->kill() operation be set in
the @pgmap arg rather than passed in separately. However, it helps code
readability, tracking the lifetime of a given instance, to be able to grep
the kill routine directly at the devm_memremap_pages() call site.
Link: http://lkml.kernel.org/r/154275558526.76910.7535251937849268605.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Fixes: e8d513483300 ("memremap: change devm_memremap_pages interface...")
Reviewed-by: "Jérôme Glisse" <jglisse@redhat.com>
Reported-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Logan Gunthorpe <logang@deltatee.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Balbir Singh <bsingharora@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2018-12-28 01:34:57 -07:00
|
|
|
error = -ENXIO;
|
|
|
|
goto err_array;
|
2015-08-17 08:00:35 -06:00
|
|
|
}
|
|
|
|
|
2018-08-15 12:22:16 -06:00
|
|
|
error = xa_err(xa_store_range(&pgmap_array, PHYS_PFN(res->start),
|
|
|
|
PHYS_PFN(res->end), pgmap, GFP_KERNEL));
|
2016-01-15 17:56:19 -07:00
|
|
|
if (error)
|
2018-08-15 12:22:16 -06:00
|
|
|
goto err_array;
|
2016-01-15 17:56:19 -07:00
|
|
|
|
2015-08-17 08:00:35 -06:00
|
|
|
if (nid < 0)
|
2015-10-05 18:35:55 -06:00
|
|
|
nid = numa_mem_id();
|
2015-08-17 08:00:35 -06:00
|
|
|
|
2019-07-18 16:58:33 -06:00
|
|
|
error = track_pfn_remap(NULL, &pgprot, PHYS_PFN(res->start), 0,
|
|
|
|
resource_size(res));
|
2016-09-07 09:51:21 -06:00
|
|
|
if (error)
|
|
|
|
goto err_pfn_remap;
|
|
|
|
|
mm: fix devm_memremap_pages crash, use mem_hotplug_{begin, done}
Both arch_add_memory() and arch_remove_memory() expect a single threaded
context.
For example, arch/x86/mm/init_64.c::kernel_physical_mapping_init() does
not hold any locks over this check and branch:
if (pgd_val(*pgd)) {
pud = (pud_t *)pgd_page_vaddr(*pgd);
paddr_last = phys_pud_init(pud, __pa(vaddr),
__pa(vaddr_end),
page_size_mask);
continue;
}
pud = alloc_low_page();
paddr_last = phys_pud_init(pud, __pa(vaddr), __pa(vaddr_end),
page_size_mask);
The result is that two threads calling devm_memremap_pages()
simultaneously can end up colliding on pgd initialization. This leads
to crash signatures like the following where the loser of the race
initializes the wrong pgd entry:
BUG: unable to handle kernel paging request at ffff888ebfff0000
IP: memcpy_erms+0x6/0x10
PGD 2f8e8fc067 PUD 0 /* <---- Invalid PUD */
Oops: 0000 [#1] SMP DEBUG_PAGEALLOC
CPU: 54 PID: 3818 Comm: systemd-udevd Not tainted 4.6.7+ #13
task: ffff882fac290040 ti: ffff882f887a4000 task.ti: ffff882f887a4000
RIP: memcpy_erms+0x6/0x10
[..]
Call Trace:
? pmem_do_bvec+0x205/0x370 [nd_pmem]
? blk_queue_enter+0x3a/0x280
pmem_rw_page+0x38/0x80 [nd_pmem]
bdev_read_page+0x84/0xb0
Hold the standard memory hotplug mutex over calls to
arch_{add,remove}_memory().
Fixes: 41e94a851304 ("add devm_memremap_pages")
Link: http://lkml.kernel.org/r/148357647831.9498.12606007370121652979.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-01-10 17:57:36 -07:00
|
|
|
mem_hotplug_begin();
|
2018-12-28 01:35:01 -07:00
|
|
|
|
|
|
|
/*
|
|
|
|
* For device private memory we call add_pages() as we only need to
|
|
|
|
* allocate and initialize struct page for the device memory. More-
|
|
|
|
* over the device memory is un-accessible thus we do not want to
|
|
|
|
* create a linear mapping for the memory like arch_add_memory()
|
|
|
|
* would do.
|
|
|
|
*
|
|
|
|
* For all other device memory types, which are accessible by
|
|
|
|
* the CPU, we do want the linear mapping and thus use
|
|
|
|
* arch_add_memory().
|
|
|
|
*/
|
|
|
|
if (pgmap->type == MEMORY_DEVICE_PRIVATE) {
|
2019-07-18 16:58:33 -06:00
|
|
|
error = add_pages(nid, PHYS_PFN(res->start),
|
|
|
|
PHYS_PFN(resource_size(res)), &restrictions);
|
2018-12-28 01:35:01 -07:00
|
|
|
} else {
|
2019-07-18 16:58:33 -06:00
|
|
|
error = kasan_add_zero_shadow(__va(res->start), resource_size(res));
|
2018-12-28 01:35:01 -07:00
|
|
|
if (error) {
|
|
|
|
mem_hotplug_done();
|
|
|
|
goto err_kasan;
|
|
|
|
}
|
|
|
|
|
2019-07-18 16:58:33 -06:00
|
|
|
error = arch_add_memory(nid, res->start, resource_size(res),
|
2019-05-13 18:21:26 -06:00
|
|
|
&restrictions);
|
2018-12-28 01:35:01 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
if (!error) {
|
|
|
|
struct zone *zone;
|
|
|
|
|
|
|
|
zone = &NODE_DATA(nid)->node_zones[ZONE_DEVICE];
|
2019-07-18 16:58:33 -06:00
|
|
|
move_pfn_range_to_zone(zone, PHYS_PFN(res->start),
|
|
|
|
PHYS_PFN(resource_size(res)), restrictions.altmap);
|
2018-08-17 16:47:04 -06:00
|
|
|
}
|
|
|
|
|
mm: fix devm_memremap_pages crash, use mem_hotplug_{begin, done}
Both arch_add_memory() and arch_remove_memory() expect a single threaded
context.
For example, arch/x86/mm/init_64.c::kernel_physical_mapping_init() does
not hold any locks over this check and branch:
if (pgd_val(*pgd)) {
pud = (pud_t *)pgd_page_vaddr(*pgd);
paddr_last = phys_pud_init(pud, __pa(vaddr),
__pa(vaddr_end),
page_size_mask);
continue;
}
pud = alloc_low_page();
paddr_last = phys_pud_init(pud, __pa(vaddr), __pa(vaddr_end),
page_size_mask);
The result is that two threads calling devm_memremap_pages()
simultaneously can end up colliding on pgd initialization. This leads
to crash signatures like the following where the loser of the race
initializes the wrong pgd entry:
BUG: unable to handle kernel paging request at ffff888ebfff0000
IP: memcpy_erms+0x6/0x10
PGD 2f8e8fc067 PUD 0 /* <---- Invalid PUD */
Oops: 0000 [#1] SMP DEBUG_PAGEALLOC
CPU: 54 PID: 3818 Comm: systemd-udevd Not tainted 4.6.7+ #13
task: ffff882fac290040 ti: ffff882f887a4000 task.ti: ffff882f887a4000
RIP: memcpy_erms+0x6/0x10
[..]
Call Trace:
? pmem_do_bvec+0x205/0x370 [nd_pmem]
? blk_queue_enter+0x3a/0x280
pmem_rw_page+0x38/0x80 [nd_pmem]
bdev_read_page+0x84/0xb0
Hold the standard memory hotplug mutex over calls to
arch_{add,remove}_memory().
Fixes: 41e94a851304 ("add devm_memremap_pages")
Link: http://lkml.kernel.org/r/148357647831.9498.12606007370121652979.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2017-01-10 17:57:36 -07:00
|
|
|
mem_hotplug_done();
|
2016-01-15 17:56:19 -07:00
|
|
|
if (error)
|
|
|
|
goto err_add_memory;
|
2015-08-17 08:00:35 -06:00
|
|
|
|
2018-10-26 16:07:52 -06:00
|
|
|
/*
|
|
|
|
* Initialization of the pages has been deferred until now in order
|
|
|
|
* to allow us to do the work while not holding the hotplug lock.
|
|
|
|
*/
|
|
|
|
memmap_init_zone_device(&NODE_DATA(nid)->node_zones[ZONE_DEVICE],
|
2019-07-18 16:58:33 -06:00
|
|
|
PHYS_PFN(res->start),
|
|
|
|
PHYS_PFN(resource_size(res)), pgmap);
|
2018-10-26 16:07:52 -06:00
|
|
|
percpu_ref_get_many(pgmap->ref, pfn_end(pgmap) - pfn_first(pgmap));
|
2015-08-17 08:00:35 -06:00
|
|
|
return __va(res->start);
|
2016-01-15 17:56:19 -07:00
|
|
|
|
|
|
|
err_add_memory:
|
2019-07-18 16:58:33 -06:00
|
|
|
kasan_remove_zero_shadow(__va(res->start), resource_size(res));
|
2018-08-17 16:47:04 -06:00
|
|
|
err_kasan:
|
2019-07-18 16:58:33 -06:00
|
|
|
untrack_pfn(NULL, PHYS_PFN(res->start), resource_size(res));
|
2016-09-07 09:51:21 -06:00
|
|
|
err_pfn_remap:
|
2018-08-15 12:22:16 -06:00
|
|
|
pgmap_array_delete(res);
|
|
|
|
err_array:
|
2019-06-26 06:27:14 -06:00
|
|
|
dev_pagemap_kill(pgmap);
|
|
|
|
dev_pagemap_cleanup(pgmap);
|
2019-08-18 03:05:56 -06:00
|
|
|
devmap_managed_enable_put();
|
2016-01-15 17:56:19 -07:00
|
|
|
return ERR_PTR(error);
|
2015-08-17 08:00:35 -06:00
|
|
|
}
|
2019-08-18 03:05:57 -06:00
|
|
|
EXPORT_SYMBOL_GPL(memremap_pages);
|
|
|
|
|
|
|
|
/**
|
|
|
|
* devm_memremap_pages - remap and provide memmap backing for the given resource
|
|
|
|
* @dev: hosting device for @res
|
|
|
|
* @pgmap: pointer to a struct dev_pagemap
|
|
|
|
*
|
|
|
|
* Notes:
|
|
|
|
* 1/ At a minimum the res and type members of @pgmap must be initialized
|
|
|
|
* by the caller before passing it to this function
|
|
|
|
*
|
|
|
|
* 2/ The altmap field may optionally be initialized, in which case
|
|
|
|
* PGMAP_ALTMAP_VALID must be set in pgmap->flags.
|
|
|
|
*
|
|
|
|
* 3/ The ref field may optionally be provided, in which pgmap->ref must be
|
|
|
|
* 'live' on entry and will be killed and reaped at
|
|
|
|
* devm_memremap_pages_release() time, or if this routine fails.
|
|
|
|
*
|
|
|
|
* 4/ res is expected to be a host memory range that could feasibly be
|
|
|
|
* treated as a "System RAM" range, i.e. not a device mmio range, but
|
|
|
|
* this is not enforced.
|
|
|
|
*/
|
|
|
|
void *devm_memremap_pages(struct device *dev, struct dev_pagemap *pgmap)
|
|
|
|
{
|
|
|
|
int error;
|
|
|
|
void *ret;
|
|
|
|
|
|
|
|
ret = memremap_pages(pgmap, dev_to_node(dev));
|
|
|
|
if (IS_ERR(ret))
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
error = devm_add_action_or_reset(dev, devm_memremap_pages_release,
|
|
|
|
pgmap);
|
|
|
|
if (error)
|
|
|
|
return ERR_PTR(error);
|
|
|
|
return ret;
|
|
|
|
}
|
2018-12-28 01:34:50 -07:00
|
|
|
EXPORT_SYMBOL_GPL(devm_memremap_pages);
|
2016-01-15 17:56:22 -07:00
|
|
|
|
2019-06-13 16:56:21 -06:00
|
|
|
void devm_memunmap_pages(struct device *dev, struct dev_pagemap *pgmap)
|
|
|
|
{
|
|
|
|
devm_release_action(dev, devm_memremap_pages_release, pgmap);
|
|
|
|
}
|
|
|
|
EXPORT_SYMBOL_GPL(devm_memunmap_pages);
|
|
|
|
|
2016-01-15 17:56:22 -07:00
|
|
|
unsigned long vmem_altmap_offset(struct vmem_altmap *altmap)
|
|
|
|
{
|
|
|
|
/* number of pfns from base where pfn_to_page() is valid */
|
2019-06-26 06:27:13 -06:00
|
|
|
if (altmap)
|
|
|
|
return altmap->reserve + altmap->free;
|
|
|
|
return 0;
|
2016-01-15 17:56:22 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
void vmem_altmap_free(struct vmem_altmap *altmap, unsigned long nr_pfns)
|
|
|
|
{
|
|
|
|
altmap->alloc -= nr_pfns;
|
|
|
|
}
|
|
|
|
|
2017-12-29 00:54:00 -07:00
|
|
|
/**
|
|
|
|
* get_dev_pagemap() - take a new live reference on the dev_pagemap for @pfn
|
|
|
|
* @pfn: page frame number to lookup page_map
|
|
|
|
* @pgmap: optional known pgmap that already has a reference
|
|
|
|
*
|
2017-12-29 00:54:01 -07:00
|
|
|
* If @pgmap is non-NULL and covers @pfn it will be returned as-is. If @pgmap
|
|
|
|
* is non-NULL but does not cover @pfn the reference to it will be released.
|
2017-12-29 00:54:00 -07:00
|
|
|
*/
|
|
|
|
struct dev_pagemap *get_dev_pagemap(unsigned long pfn,
|
|
|
|
struct dev_pagemap *pgmap)
|
|
|
|
{
|
|
|
|
resource_size_t phys = PFN_PHYS(pfn);
|
|
|
|
|
|
|
|
/*
|
2017-12-29 00:54:01 -07:00
|
|
|
* In the cached case we're already holding a live reference.
|
2017-12-29 00:54:00 -07:00
|
|
|
*/
|
2017-12-29 00:54:01 -07:00
|
|
|
if (pgmap) {
|
2017-12-29 00:54:04 -07:00
|
|
|
if (phys >= pgmap->res.start && phys <= pgmap->res.end)
|
2017-12-29 00:54:01 -07:00
|
|
|
return pgmap;
|
|
|
|
put_dev_pagemap(pgmap);
|
2017-12-29 00:54:00 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
/* fall back to slow path lookup */
|
|
|
|
rcu_read_lock();
|
2018-08-15 12:22:16 -06:00
|
|
|
pgmap = xa_load(&pgmap_array, PHYS_PFN(phys));
|
2017-12-29 00:54:00 -07:00
|
|
|
if (pgmap && !percpu_ref_tryget_live(pgmap->ref))
|
|
|
|
pgmap = NULL;
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
return pgmap;
|
|
|
|
}
|
2018-05-16 12:46:08 -06:00
|
|
|
EXPORT_SYMBOL_GPL(get_dev_pagemap);
|
2017-09-08 17:11:46 -06:00
|
|
|
|
2018-05-16 12:46:08 -06:00
|
|
|
#ifdef CONFIG_DEV_PAGEMAP_OPS
|
|
|
|
void __put_devmap_managed_page(struct page *page)
|
2017-09-08 17:11:46 -06:00
|
|
|
{
|
|
|
|
int count = page_ref_dec_return(page);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If refcount is 1 then page is freed and refcount is stable as nobody
|
|
|
|
* holds a reference on the page.
|
|
|
|
*/
|
|
|
|
if (count == 1) {
|
|
|
|
/* Clear Active bit in case of parallel mark_page_accessed */
|
|
|
|
__ClearPageActive(page);
|
|
|
|
__ClearPageWaiters(page);
|
|
|
|
|
2017-09-08 17:11:54 -06:00
|
|
|
mem_cgroup_uncharge(page);
|
2017-09-08 17:11:46 -06:00
|
|
|
|
2019-08-13 16:37:07 -06:00
|
|
|
/*
|
|
|
|
* When a device_private page is freed, the page->mapping field
|
|
|
|
* may still contain a (stale) mapping value. For example, the
|
|
|
|
* lower bits of page->mapping may still identify the page as
|
|
|
|
* an anonymous page. Ultimately, this entire field is just
|
|
|
|
* stale and wrong, and it will cause errors if not cleared.
|
|
|
|
* One example is:
|
|
|
|
*
|
|
|
|
* migrate_vma_pages()
|
|
|
|
* migrate_vma_insert_page()
|
|
|
|
* page_add_new_anon_rmap()
|
|
|
|
* __page_set_anon_rmap()
|
|
|
|
* ...checks page->mapping, via PageAnon(page) call,
|
|
|
|
* and incorrectly concludes that the page is an
|
|
|
|
* anonymous page. Therefore, it incorrectly,
|
|
|
|
* silently fails to set up the new anon rmap.
|
|
|
|
*
|
|
|
|
* For other types of ZONE_DEVICE pages, migration is either
|
|
|
|
* handled differently or not done at all, so there is no need
|
|
|
|
* to clear page->mapping.
|
|
|
|
*/
|
|
|
|
if (is_device_private_page(page))
|
|
|
|
page->mapping = NULL;
|
|
|
|
|
2019-06-26 06:27:12 -06:00
|
|
|
page->pgmap->ops->page_free(page);
|
2017-09-08 17:11:46 -06:00
|
|
|
} else if (!count)
|
|
|
|
__put_page(page);
|
|
|
|
}
|
2018-07-26 17:37:22 -06:00
|
|
|
EXPORT_SYMBOL(__put_devmap_managed_page);
|
2018-05-16 12:46:08 -06:00
|
|
|
#endif /* CONFIG_DEV_PAGEMAP_OPS */
|