c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2026-05-24 19:28:53 +03:00
Code Issues Packages Projects Releases Wiki Activity

dmu_objset_from_ds must be called with dp_config_rwlock held

Browse Source 
The normal lock order is that the dp_config_rwlock must be held before
the ds_opening_lock.  For example, dmu_objset_hold() does this.
However, dmu_objset_open_impl() is called with the ds_opening_lock held,
and if the dp_config_rwlock is not already held, it will attempt to
acquire it.  This may lead to deadlock, since the lock order is
reversed.

Looking at all the callers of dmu_objset_open_impl() (which is
principally the callers of dmu_objset_from_ds()), almost all callers
already have the dp_config_rwlock.  However, there are a few places in
the send and receive code paths that do not.  For example:
dsl_crypto_populate_key_nvlist, send_cb, dmu_recv_stream,
receive_write_byref, redact_traverse_thread.

This commit resolves the problem by requiring all callers ot
dmu_objset_from_ds() to hold the dp_config_rwlock.  In most cases, the
code has been restructured such that we call dmu_objset_from_ds()
earlier on in the send and receive processes, when we already have the
dp_config_rwlock, and save the objset_t until we need it in the middle
of the send or receive (similar to what we already do with the
dsl_dataset_t).  Thus we do not need to acquire the dp_config_rwlock in
many new places.

I also cleaned up code in dmu_redact_snap() and send_traverse_thread().

Reviewed-by: Paul Dagnelie <pcd@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Paul Zuchowski <pzuchowski@datto.com>
Signed-off-by: Matthew Ahrens <mahrens@delphix.com>
Closes #9662
Closes #10115
This commit is contained in:
Matthew Ahrens
2020-03-12 10:55:02 -07:00
committed by GitHub
parent fa130e010c
commit 0fdd6106bb
7 changed files with 101 additions and 115 deletions
Download Patch File Download Diff File Expand all files Collapse all files
module/zfs/dmu_objset.c
+9 -18
View File
@@ -414,6 +414,11 @@ dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock));
ASSERT(!BP_IS_REDACTED(bp));
/*
* We need the pool config lock to get properties.
*/
ASSERT(ds == NULL || dsl_pool_config_held(ds->ds_dir->dd_pool));
/*
* The $ORIGIN dataset (if it exists) doesn't have an associated
* objset, so there's no reason to open it. The $ORIGIN dataset
@@ -503,20 +508,8 @@ dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
* checksum/compression/copies.
*/
if (ds != NULL) {
boolean_t needlock = B_FALSE;
os->os_encrypted = (ds->ds_dir->dd_crypto_obj != 0);
/*
* Note: it's valid to open the objset if the dataset is
* long-held, in which case the pool_config lock will not
* be held.
*/
if (!dsl_pool_config_held(dmu_objset_pool(os))) {
needlock = B_TRUE;
dsl_pool_config_enter(dmu_objset_pool(os), FTAG);
}
err = dsl_prop_register(ds,
zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE),
primary_cache_changed_cb, os);
@@ -579,8 +572,6 @@ dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
smallblk_changed_cb, os);
}
}
if (needlock)
dsl_pool_config_exit(dmu_objset_pool(os), FTAG);
if (err != 0) {
arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf);
kmem_free(os, sizeof (objset_t));
@@ -650,11 +641,11 @@ dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp)
int err = 0;
/*
* We shouldn't be doing anything with dsl_dataset_t's unless the
* pool_config lock is held, or the dataset is long-held.
* We need the pool_config lock to manipulate the dsl_dataset_t.
* Even if the dataset is long-held, we need the pool_config lock
* to open the objset, as it needs to get properties.
*/
ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool) ||
dsl_dataset_long_held(ds));
ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool));
mutex_enter(&ds->ds_opening_lock);
if (ds->ds_objset == NULL) {