Archive-Team
/
zfs
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 103 Wiki Activity

Prevent race condition in dnode_dest (#10101) 

dnode_special_close() waits for the refcount of dn_holds to go to zero
without holding the dn_mtx. dnode_rele_and_unlock() does the final
remove to dn_holds with dn_mtx being held:

	refs = zfs_refcount_remove(&dn->dn_holds, tag);
	mutex_exit(&dn->dn_mtx);

So, there is a race condition after the remove until dn_mtx is
dropped. During that time, dnode_destroy() can get called, which ends
up in dnode_dest() calling mutex_destroy() and a panic since the lock
is still held.

This change adds a condvar to wait for the final dnode_rele_and_unlock()
to release the dn_mtx before calling dnode_destroy().

Reviewed-by: Paul Dagnelie <pcd@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Matthew Ahrens <mahrens@delphix.com>
Signed-off-by: John Poduska <jpoduska@datto.com>
Closes #7814
Closes #10101
This commit is contained in:
John Poduska 2020-03-12 13:25:56 -04:00 committed by GitHub
parent 1e9231ada8
commit e6b28efccc
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 15 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
include/sys/dmu_impl.h
View File

@ -164,6 +164,7 @@ extern "C" {
* dn_dirty_txg * dn_dirty_txg
* dd_assigned_tx * dd_assigned_tx
* dn_notxholds * dn_notxholds
* dn_nodnholds
* dn_dirtyctx * dn_dirtyctx
* dn_dirtyctx_firstset * dn_dirtyctx_firstset
* (dn_phys copy fields?) * (dn_phys copy fields?)

1
include/sys/dnode.h
View File

@ -332,6 +332,7 @@ struct dnode {
uint64_t dn_assigned_txg; uint64_t dn_assigned_txg;
uint64_t dn_dirty_txg; /* txg dnode was last dirtied */ uint64_t dn_dirty_txg; /* txg dnode was last dirtied */
kcondvar_t dn_notxholds; kcondvar_t dn_notxholds;
kcondvar_t dn_nodnholds;
enum dnode_dirtycontext dn_dirtyctx; enum dnode_dirtycontext dn_dirtyctx;
void *dn_dirtyctx_firstset; /* dbg: contents meaningless */ void *dn_dirtyctx_firstset; /* dbg: contents meaningless */

19
module/zfs/dnode.c
View File

@ -119,6 +119,7 @@ dnode_cons(void *arg, void *unused, int kmflag)
mutex_init(&dn->dn_mtx, NULL, MUTEX_DEFAULT, NULL); mutex_init(&dn->dn_mtx, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&dn->dn_dbufs_mtx, NULL, MUTEX_DEFAULT, NULL); mutex_init(&dn->dn_dbufs_mtx, NULL, MUTEX_DEFAULT, NULL);
cv_init(&dn->dn_notxholds, NULL, CV_DEFAULT, NULL); cv_init(&dn->dn_notxholds, NULL, CV_DEFAULT, NULL);
cv_init(&dn->dn_nodnholds, NULL, CV_DEFAULT, NULL);
/* /*
* Every dbuf has a reference, and dropping a tracked reference is * Every dbuf has a reference, and dropping a tracked reference is
@ -183,6 +184,7 @@ dnode_dest(void *arg, void *unused)
mutex_destroy(&dn->dn_mtx); mutex_destroy(&dn->dn_mtx);
mutex_destroy(&dn->dn_dbufs_mtx); mutex_destroy(&dn->dn_dbufs_mtx);
cv_destroy(&dn->dn_notxholds); cv_destroy(&dn->dn_notxholds);
cv_destroy(&dn->dn_nodnholds);
zfs_refcount_destroy(&dn->dn_holds); zfs_refcount_destroy(&dn->dn_holds);
zfs_refcount_destroy(&dn->dn_tx_holds); zfs_refcount_destroy(&dn->dn_tx_holds);
ASSERT(!list_link_active(&dn->dn_link)); ASSERT(!list_link_active(&dn->dn_link));
@ -1171,13 +1173,15 @@ dnode_special_close(dnode_handle_t *dnh)
dnode_t *dn = dnh->dnh_dnode; dnode_t *dn = dnh->dnh_dnode;
/* /*
* Wait for final references to the dnode to clear. This can * Ensure dnode_rele_and_unlock() has released dn_mtx, after final
* only happen if the arc is asynchronously evicting state that * zfs_refcount_remove()
* has a hold on this dnode while we are trying to evict this
* dnode.
*/ */
while (zfs_refcount_count(&dn->dn_holds) > 0) mutex_enter(&dn->dn_mtx);
delay(1); if (zfs_refcount_count(&dn->dn_holds) > 0)
cv_wait(&dn->dn_nodnholds, &dn->dn_mtx);
mutex_exit(&dn->dn_mtx);
ASSERT3U(zfs_refcount_count(&dn->dn_holds), ==, 0);
ASSERT(dn->dn_dbuf == NULL || ASSERT(dn->dn_dbuf == NULL ||
dmu_buf_get_user(&dn->dn_dbuf->db) == NULL); dmu_buf_get_user(&dn->dn_dbuf->db) == NULL);
zrl_add(&dnh->dnh_zrlock); zrl_add(&dnh->dnh_zrlock);
@ -1606,7 +1610,10 @@ dnode_rele_and_unlock(dnode_t *dn, void *tag, boolean_t evicting)
dnode_handle_t *dnh = dn->dn_handle; dnode_handle_t *dnh = dn->dn_handle;
refs = zfs_refcount_remove(&dn->dn_holds, tag); refs = zfs_refcount_remove(&dn->dn_holds, tag);
if (refs == 0)
cv_broadcast(&dn->dn_nodnholds);
mutex_exit(&dn->dn_mtx); mutex_exit(&dn->dn_mtx);
/* dnode could get destroyed at this point, so don't use it anymore */
/* /*
* It's unsafe to release the last hold on a dnode by dnode_rele() or * It's unsafe to release the last hold on a dnode by dnode_rele() or