A dm-crypt fix for a cpu hotplug crash that switches from using per-cpu

data to a mempool allocation (which offers allocation with cpu locality,
 and there is no inter-cpu communication on slab allocation).
 
 A couple dm-thinp stable fixes to address "out-of-data-space" issues.
 
 A dm-multipath fix for a LOCKDEP warning introduced in 3.15-rc1.
 -----BEGIN PGP SIGNATURE-----
 Version: GnuPG v1
 
 iQEcBAABAgAGBQJTdiI7AAoJEMUj8QotnQNa1fMIAOSGppH4U/VuT1+UMDyabUba
 eXsK8xBUTIDSBuTJ+ljkE5fyvXpn/wvA+b1hTKLhzVkUZ1pCY4pIw1pwpVcw89Bb
 BhktFWRYvcv/MAARDHiMGW5yc6xP319Qm04XN3xbMHx71gxGRwpzb191LSO5S2VR
 0rjXvZZt7WPJe/QPOFUrqyoP7t59LH9hu2/OH/Ic9o5/D0WxbqPEP6X8iJyIs32u
 lNvIQ5r5f3xNzt0VDvEq3sxR3qYhQvLPDMdp0YR67c87fKfsKQj4pkvXltXhY5bM
 wBHFE+NBl6MPzTXNCT9i+p360GXE7B/lY9boochAyE/UEztRq1+oqJOa/dDaBCo=
 =0vk2
 -----END PGP SIGNATURE-----

Merge tag 'dm-3.15-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:
 "A dm-crypt fix for a cpu hotplug crash that switches from using
  per-cpu data to a mempool allocation (which offers allocation with cpu
  locality, and there is no inter-cpu communication on slab allocation).

  A couple dm-thinp stable fixes to address "out-of-data-space" issues.

  A dm-multipath fix for a LOCKDEP warning introduced in 3.15-rc1"

* tag 'dm-3.15-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm mpath: fix lock order inconsistency in multipath_ioctl
  dm thin: add timeout to stop out-of-data-space mode holding IO forever
  dm thin: allow metadata commit if pool is in PM_OUT_OF_DATA_SPACE mode
  dm crypt: fix cpu hotplug crash by removing per-cpu structure
This commit is contained in:
Linus Torvalds 2014-05-21 17:57:31 +09:00
Родитель 31a3fcab11 4cdd2ad780
Коммит 23de4a7af7
3 изменённых файлов: 35 добавлений и 51 удалений

Просмотреть файл

@ -19,7 +19,6 @@
#include <linux/crypto.h>
#include <linux/workqueue.h>
#include <linux/backing-dev.h>
#include <linux/percpu.h>
#include <linux/atomic.h>
#include <linux/scatterlist.h>
#include <asm/page.h>
@ -43,6 +42,7 @@ struct convert_context {
struct bvec_iter iter_out;
sector_t cc_sector;
atomic_t cc_pending;
struct ablkcipher_request *req;
};
/*
@ -111,15 +111,7 @@ struct iv_tcw_private {
enum flags { DM_CRYPT_SUSPENDED, DM_CRYPT_KEY_VALID };
/*
* Duplicated per-CPU state for cipher.
*/
struct crypt_cpu {
struct ablkcipher_request *req;
};
/*
* The fields in here must be read only after initialization,
* changing state should be in crypt_cpu.
* The fields in here must be read only after initialization.
*/
struct crypt_config {
struct dm_dev *dev;
@ -150,12 +142,6 @@ struct crypt_config {
sector_t iv_offset;
unsigned int iv_size;
/*
* Duplicated per cpu state. Access through
* per_cpu_ptr() only.
*/
struct crypt_cpu __percpu *cpu;
/* ESSIV: struct crypto_cipher *essiv_tfm */
void *iv_private;
struct crypto_ablkcipher **tfms;
@ -192,11 +178,6 @@ static void clone_init(struct dm_crypt_io *, struct bio *);
static void kcryptd_queue_crypt(struct dm_crypt_io *io);
static u8 *iv_of_dmreq(struct crypt_config *cc, struct dm_crypt_request *dmreq);
static struct crypt_cpu *this_crypt_config(struct crypt_config *cc)
{
return this_cpu_ptr(cc->cpu);
}
/*
* Use this to access cipher attributes that are the same for each CPU.
*/
@ -903,16 +884,15 @@ static void kcryptd_async_done(struct crypto_async_request *async_req,
static void crypt_alloc_req(struct crypt_config *cc,
struct convert_context *ctx)
{
struct crypt_cpu *this_cc = this_crypt_config(cc);
unsigned key_index = ctx->cc_sector & (cc->tfms_count - 1);
if (!this_cc->req)
this_cc->req = mempool_alloc(cc->req_pool, GFP_NOIO);
if (!ctx->req)
ctx->req = mempool_alloc(cc->req_pool, GFP_NOIO);
ablkcipher_request_set_tfm(this_cc->req, cc->tfms[key_index]);
ablkcipher_request_set_callback(this_cc->req,
ablkcipher_request_set_tfm(ctx->req, cc->tfms[key_index]);
ablkcipher_request_set_callback(ctx->req,
CRYPTO_TFM_REQ_MAY_BACKLOG | CRYPTO_TFM_REQ_MAY_SLEEP,
kcryptd_async_done, dmreq_of_req(cc, this_cc->req));
kcryptd_async_done, dmreq_of_req(cc, ctx->req));
}
/*
@ -921,7 +901,6 @@ static void crypt_alloc_req(struct crypt_config *cc,
static int crypt_convert(struct crypt_config *cc,
struct convert_context *ctx)
{
struct crypt_cpu *this_cc = this_crypt_config(cc);
int r;
atomic_set(&ctx->cc_pending, 1);
@ -932,7 +911,7 @@ static int crypt_convert(struct crypt_config *cc,
atomic_inc(&ctx->cc_pending);
r = crypt_convert_block(cc, ctx, this_cc->req);
r = crypt_convert_block(cc, ctx, ctx->req);
switch (r) {
/* async */
@ -941,7 +920,7 @@ static int crypt_convert(struct crypt_config *cc,
reinit_completion(&ctx->restart);
/* fall through*/
case -EINPROGRESS:
this_cc->req = NULL;
ctx->req = NULL;
ctx->cc_sector++;
continue;
@ -1040,6 +1019,7 @@ static struct dm_crypt_io *crypt_io_alloc(struct crypt_config *cc,
io->sector = sector;
io->error = 0;
io->base_io = NULL;
io->ctx.req = NULL;
atomic_set(&io->io_pending, 0);
return io;
@ -1065,6 +1045,8 @@ static void crypt_dec_pending(struct dm_crypt_io *io)
if (!atomic_dec_and_test(&io->io_pending))
return;
if (io->ctx.req)
mempool_free(io->ctx.req, cc->req_pool);
mempool_free(io, cc->io_pool);
if (likely(!base_io))
@ -1492,8 +1474,6 @@ static int crypt_wipe_key(struct crypt_config *cc)
static void crypt_dtr(struct dm_target *ti)
{
struct crypt_config *cc = ti->private;
struct crypt_cpu *cpu_cc;
int cpu;
ti->private = NULL;
@ -1505,13 +1485,6 @@ static void crypt_dtr(struct dm_target *ti)
if (cc->crypt_queue)
destroy_workqueue(cc->crypt_queue);
if (cc->cpu)
for_each_possible_cpu(cpu) {
cpu_cc = per_cpu_ptr(cc->cpu, cpu);
if (cpu_cc->req)
mempool_free(cpu_cc->req, cc->req_pool);
}
crypt_free_tfms(cc);
if (cc->bs)
@ -1530,9 +1503,6 @@ static void crypt_dtr(struct dm_target *ti)
if (cc->dev)
dm_put_device(ti, cc->dev);
if (cc->cpu)
free_percpu(cc->cpu);
kzfree(cc->cipher);
kzfree(cc->cipher_string);
@ -1588,13 +1558,6 @@ static int crypt_ctr_cipher(struct dm_target *ti,
if (tmp)
DMWARN("Ignoring unexpected additional cipher options");
cc->cpu = __alloc_percpu(sizeof(*(cc->cpu)),
__alignof__(struct crypt_cpu));
if (!cc->cpu) {
ti->error = "Cannot allocate per cpu state";
goto bad_mem;
}
/*
* For compatibility with the original dm-crypt mapping format, if
* only the cipher name is supplied, use cbc-plain.

Просмотреть файл

@ -1566,8 +1566,8 @@ static int multipath_ioctl(struct dm_target *ti, unsigned int cmd,
}
if (m->pg_init_required)
__pg_init_all_paths(m);
spin_unlock_irqrestore(&m->lock, flags);
dm_table_run_md_queue_async(m->ti->table);
spin_unlock_irqrestore(&m->lock, flags);
}
return r ? : __blkdev_driver_ioctl(bdev, mode, cmd, arg);

Просмотреть файл

@ -27,6 +27,7 @@
#define MAPPING_POOL_SIZE 1024
#define PRISON_CELLS 1024
#define COMMIT_PERIOD HZ
#define NO_SPACE_TIMEOUT (HZ * 60)
DECLARE_DM_KCOPYD_THROTTLE_WITH_MODULE_PARM(snapshot_copy_throttle,
"A percentage of time allocated for copy on write");
@ -175,6 +176,7 @@ struct pool {
struct workqueue_struct *wq;
struct work_struct worker;
struct delayed_work waker;
struct delayed_work no_space_timeout;
unsigned long last_commit_jiffies;
unsigned ref_count;
@ -935,7 +937,7 @@ static int commit(struct pool *pool)
{
int r;
if (get_pool_mode(pool) != PM_WRITE)
if (get_pool_mode(pool) >= PM_READ_ONLY)
return -EINVAL;
r = dm_pool_commit_metadata(pool->pmd);
@ -1590,6 +1592,20 @@ static void do_waker(struct work_struct *ws)
queue_delayed_work(pool->wq, &pool->waker, COMMIT_PERIOD);
}
/*
* We're holding onto IO to allow userland time to react. After the
* timeout either the pool will have been resized (and thus back in
* PM_WRITE mode), or we degrade to PM_READ_ONLY and start erroring IO.
*/
static void do_no_space_timeout(struct work_struct *ws)
{
struct pool *pool = container_of(to_delayed_work(ws), struct pool,
no_space_timeout);
if (get_pool_mode(pool) == PM_OUT_OF_DATA_SPACE && !pool->pf.error_if_no_space)
set_pool_mode(pool, PM_READ_ONLY);
}
/*----------------------------------------------------------------*/
struct noflush_work {
@ -1715,6 +1731,9 @@ static void set_pool_mode(struct pool *pool, enum pool_mode new_mode)
pool->process_discard = process_discard;
pool->process_prepared_mapping = process_prepared_mapping;
pool->process_prepared_discard = process_prepared_discard_passdown;
if (!pool->pf.error_if_no_space)
queue_delayed_work(pool->wq, &pool->no_space_timeout, NO_SPACE_TIMEOUT);
break;
case PM_WRITE:
@ -2100,6 +2119,7 @@ static struct pool *pool_create(struct mapped_device *pool_md,
INIT_WORK(&pool->worker, do_worker);
INIT_DELAYED_WORK(&pool->waker, do_waker);
INIT_DELAYED_WORK(&pool->no_space_timeout, do_no_space_timeout);
spin_lock_init(&pool->lock);
bio_list_init(&pool->deferred_flush_bios);
INIT_LIST_HEAD(&pool->prepared_mappings);
@ -2662,6 +2682,7 @@ static void pool_postsuspend(struct dm_target *ti)
struct pool *pool = pt->pool;
cancel_delayed_work(&pool->waker);
cancel_delayed_work(&pool->no_space_timeout);
flush_workqueue(pool->wq);
(void) commit(pool);
}