c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2026-05-25 11:47:43 +03:00
Code Issues Packages Projects Releases Wiki Activity

Illumos 5056 - ZFS deadlock on db_mtx and dn_holds

Browse Source 
5056 ZFS deadlock on db_mtx and dn_holds
Author: Justin Gibbs <justing@spectralogic.com>
Reviewed by: Will Andrews <willa@spectralogic.com>
Reviewed by: Matt Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Approved by: Dan McDonald <danmcd@omniti.com>

References:
  https://www.illumos.org/issues/5056
  https://github.com/illumos/illumos-gate/commit/bc9014e

Porting Notes:

sa_handle_get_from_db():
  - the original patch includes an otherwise unmentioned fix for a
    possible usage of an uninitialised variable

dmu_objset_open_impl():
  - Under Illumos list_link_init() is the same as filling a list_node_t
    with NULLs, so they don't notice if they miss doing list_link_init()
    on a zero'd containing structure (e.g. allocated with kmem_zalloc as
    here). Under Linux, not so much: an uninitialised list_node_t goes
    "Boom!" some time later when it's used or destroyed.

dmu_objset_evict_dbufs():
  - reduce stack usage using kmem_alloc()

Ported-by: Chris Dunlop <chris@onthe.net.au>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
This commit is contained in:
Justin T. Gibbs
2015-04-02 14:44:32 +11:00
committed by Brian Behlendorf
parent d683ddbb72
commit 0c66c32d1d
35 changed files with 645 additions and 316 deletions
Download Patch File Download Diff File Expand all files Collapse all files
module/zfs/zap.c
+18 -17
View File
@@ -21,6 +21,7 @@
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012, 2014 by Delphix. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
*/
/*
@@ -52,7 +53,6 @@ int fzap_default_block_shift = 14; /* 16k blocksize */
extern inline zap_phys_t *zap_f_phys(zap_t *zap);
static void zap_leaf_pageout(dmu_buf_t *db, void *vl);
static uint64_t zap_allocate_blocks(zap_t *zap, int nblocks);
void
@@ -81,7 +81,7 @@ fzap_upgrade(zap_t *zap, dmu_tx_t *tx, zap_flags_t flags)
ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
zap->zap_ismicro = FALSE;
(void) dmu_buf_update_user(zap->zap_dbuf, zap, zap, zap_evict);
zap->zap_dbu.dbu_evict_func = zap_evict;
mutex_init(&zap->zap_f.zap_num_entries_mtx, 0, 0, 0);
zap->zap_f.zap_block_shift = highbit64(zap->zap_dbuf->db_size) - 1;
@@ -387,11 +387,20 @@ zap_allocate_blocks(zap_t *zap, int nblocks)
return (newblk);
}
static void
zap_leaf_pageout(void *dbu)
{
zap_leaf_t *l = dbu;
rw_destroy(&l->l_rwlock);
kmem_free(l, sizeof (zap_leaf_t));
}
static zap_leaf_t *
zap_create_leaf(zap_t *zap, dmu_tx_t *tx)
{
void *winner;
zap_leaf_t *l = kmem_alloc(sizeof (zap_leaf_t), KM_SLEEP);
zap_leaf_t *l = kmem_zalloc(sizeof (zap_leaf_t), KM_SLEEP);
ASSERT(RW_WRITE_HELD(&zap->zap_rwlock));
@@ -403,7 +412,8 @@ zap_create_leaf(zap_t *zap, dmu_tx_t *tx)
VERIFY(0 == dmu_buf_hold(zap->zap_objset, zap->zap_object,
l->l_blkid << FZAP_BLOCK_SHIFT(zap), NULL, &l->l_dbuf,
DMU_READ_NO_PREFETCH));
winner = dmu_buf_set_user(l->l_dbuf, l, zap_leaf_pageout);
dmu_buf_init_user(&l->l_dbu, zap_leaf_pageout, &l->l_dbuf);
winner = dmu_buf_set_user(l->l_dbuf, &l->l_dbu);
ASSERT(winner == NULL);
dmu_buf_will_dirty(l->l_dbuf, tx);
@@ -435,16 +445,6 @@ zap_put_leaf(zap_leaf_t *l)
dmu_buf_rele(l->l_dbuf, NULL);
}
_NOTE(ARGSUSED(0))
static void
zap_leaf_pageout(dmu_buf_t *db, void *vl)
{
zap_leaf_t *l = vl;
rw_destroy(&l->l_rwlock);
kmem_free(l, sizeof (zap_leaf_t));
}
static zap_leaf_t *
zap_open_leaf(uint64_t blkid, dmu_buf_t *db)
{
@@ -452,19 +452,20 @@ zap_open_leaf(uint64_t blkid, dmu_buf_t *db)
ASSERT(blkid != 0);
l = kmem_alloc(sizeof (zap_leaf_t), KM_SLEEP);
l = kmem_zalloc(sizeof (zap_leaf_t), KM_SLEEP);
rw_init(&l->l_rwlock, NULL, RW_DEFAULT, NULL);
rw_enter(&l->l_rwlock, RW_WRITER);
l->l_blkid = blkid;
l->l_bs = highbit64(db->db_size) - 1;
l->l_dbuf = db;
winner = dmu_buf_set_user(db, l, zap_leaf_pageout);
dmu_buf_init_user(&l->l_dbu, zap_leaf_pageout, &l->l_dbuf);
winner = dmu_buf_set_user(db, &l->l_dbu);
rw_exit(&l->l_rwlock);
if (winner != NULL) {
/* someone else set it first */
zap_leaf_pageout(NULL, l);
zap_leaf_pageout(&l->l_dbu);
l = winner;
}