Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > linux.kernel > #1519244
| From | James Simmons <jsimmons@infradead.org> |
|---|---|
| Newsgroups | linux.kernel |
| Subject | [PATCH 13/35] staging: lustre: llite: lookup master inode by ilookup5_nowait |
| Date | 2016-11-10 18:50 +0100 |
| Message-ID | <sC1E7-QH-55@gated-at.bofh.it> (permalink) |
| References | <sC1up-MX-3@gated-at.bofh.it> |
| Organization | linux.* mail to news gateway |
From: wang di <di.wang@intel.com>
Do not lookup master inode by ilookup5, instead it should
use ilookup5_nowait, otherwise it will cause dead lock,
1. Client1 send chmod req to the MDT0, then on MDT0, it
enqueues master and all of its slaves lock, (mdt_attr_set()
->mdt_lock_slaves()), after gets master and stripe0 lock,
it will send the enqueue request(for stripe1) to MDT1, then
MDT1 finds the lock has been granted to client2. Then MDT1
sends blocking ast to client2.
2. At the same time, client2 tries to unlink the striped
dir (rm -rf striped_dir), and during lookup, it will hold
the master inode of the striped directory, whose inode state
is NEW, then tries to revalidate all of its slaves,
(ll_prep_inode()->ll_iget()->ll_read_inode2()->
ll_update_inode().). And it will be blocked on the server
side because of 1.
3. Then the client get the blocking_ast request, cancel the
lock, but being blocked by ilookup5 in ll_md_blocking_ast(),
because the inode state is still NEW.
Signed-off-by: wang di <di.wang@intel.com>
Intel-bug-id: https://jira.hpdd.intel.com/browse/LU-5344
Reviewed-on: http://review.whamcloud.com/16066
Reviewed-by: John L. Hammond <john.hammond@intel.com>
Reviewed-by: Lai Siyao <lai.siyao@intel.com>
Reviewed-by: James Simmons <uja.ornl@yahoo.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
Signed-off-by: James Simmons <jsimmons@infradead.org>
---
drivers/staging/lustre/lustre/llite/namei.c | 36 ++++++++++++++++++++++++---
1 files changed, 32 insertions(+), 4 deletions(-)
diff --git a/drivers/staging/lustre/lustre/llite/namei.c b/drivers/staging/lustre/lustre/llite/namei.c
index c268f32..9ea43e1 100644
--- a/drivers/staging/lustre/lustre/llite/namei.c
+++ b/drivers/staging/lustre/lustre/llite/namei.c
@@ -286,10 +286,38 @@ int ll_md_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc,
hash = cl_fid_build_ino(&lli->lli_pfid,
ll_need_32bit_api(ll_i2sbi(inode)));
-
- master_inode = ilookup5(inode->i_sb, hash,
- ll_test_inode_by_fid,
- (void *)&lli->lli_pfid);
+ /*
+ * Do not lookup the inode with ilookup5,
+ * otherwise it will cause dead lock,
+ *
+ * 1. Client1 send chmod req to the MDT0, then
+ * on MDT0, it enqueues master and all of its
+ * slaves lock, (mdt_attr_set() ->
+ * mdt_lock_slaves()), after gets master and
+ * stripe0 lock, it will send the enqueue req
+ * (for stripe1) to MDT1, then MDT1 finds the
+ * lock has been granted to client2. Then MDT1
+ * sends blocking ast to client2.
+ *
+ * 2. At the same time, client2 tries to unlink
+ * the striped dir (rm -rf striped_dir), and
+ * during lookup, it will hold the master inode
+ * of the striped directory, whose inode state
+ * is NEW, then tries to revalidate all of its
+ * slaves, (ll_prep_inode()->ll_iget()->
+ * ll_read_inode2()-> ll_update_inode().). And
+ * it will be blocked on the server side because
+ * of 1.
+ *
+ * 3. Then the client get the blocking_ast req,
+ * cancel the lock, but being blocked if using
+ * ->ilookup5()), because master inode state is
+ * NEW.
+ */
+ master_inode = ilookup5_nowait(inode->i_sb,
+ hash,
+ ll_test_inode_by_fid,
+ (void *)&lli->lli_pfid);
if (master_inode) {
ll_invalidate_negative_children(master_inode);
iput(master_inode);
--
1.7.1
Back to linux.kernel | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread
[PATCH 00/35] second batch of missing lustre 2.8 patches James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 04/35] staging: lustre: lov: init LOV stripe type beforehand James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 31/35] staging: lustre: obdclass: add export for lprocfs_stats_alloc_one() James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 30/35] staging: lustre: llite: ll_write_begin/end not passing on errors James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 14/35] staging: lustre: nrs: serialize executions of nrs_policy_stop James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 19/35] staging: lustre: obdclass: remove structure holes to reduce memory James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 15/35] staging: lustre: llite: tar restore fails for HSM released files. James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 20/35] staging: lustre: ptlrpc: Move IT_* definitions to lustre_idl.h James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 26/35] staging: lustre: mdc: remove console spew from mdc_ioc_fid2path James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 35/35] staging: lustre: idl: clean up file attribute flags James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 22/35] staging: lustre: llite: clear dir stripe md in ll_iget James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 01/35] staging: lustre: hsm: Use file lease to implement migration James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 10/35] staging: lustre: lmv: revalidate the dentry for striped dir James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 28/35] staging: lustre: osc: Remove remains of osc_ast_guard James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 03/35] staging: lustre: ptlrpc: mbits is sent within ptlrpc_body James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 27/35] staging: lustre: ptlrpc: reset imp_replay_cursor James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 25/35] staging: lustre: osc: Do not merge extents with partial pages James Simmons <jsimmons@infradead.org> - 2016-11-10 18:40 +0100 [PATCH 16/35] staging: lustre: llite: support SELinux context labelling James Simmons <jsimmons@infradead.org> - 2016-11-10 18:50 +0100 [PATCH 11/35] staging: lustre: ptlrpc: race at req processing James Simmons <jsimmons@infradead.org> - 2016-11-10 18:50 +0100 [PATCH 06/35] staging: lustre: osc: Performance tune for LRU James Simmons <jsimmons@infradead.org> - 2016-11-10 18:50 +0100 [PATCH 17/35] staging: lustre: obd: Remove dead code in precleanup James Simmons <jsimmons@infradead.org> - 2016-11-10 18:50 +0100 [PATCH 13/35] staging: lustre: llite: lookup master inode by ilookup5_nowait James Simmons <jsimmons@infradead.org> - 2016-11-10 18:50 +0100 [PATCH 18/35] staging: lustre: osc: fix max_dirty_mb tunable setting limit James Simmons <jsimmons@infradead.org> - 2016-11-10 18:50 +0100 [PATCH 24/35] staging: lustre: osc: osc_extent should hold refcount to osc_object James Simmons <jsimmons@infradead.org> - 2016-11-10 18:50 +0100 [PATCH 23/35] staging: lustre: ldlm: improve lock timeout messages James Simmons <jsimmons@infradead.org> - 2016-11-10 18:50 +0100 [PATCH 05/35] staging: lustre: llog: fix wrong offset in llog_process_thread() James Simmons <jsimmons@infradead.org> - 2016-11-10 18:50 +0100 [PATCH 02/35] staging: lustre: obd: rename obd_unpackmd() to md_unpackmd() James Simmons <jsimmons@infradead.org> - 2016-11-10 19:10 +0100 [PATCH 07/35] staging: lustre: lov: avoid infinite loop in lsm_alloc_plain() James Simmons <jsimmons@infradead.org> - 2016-11-10 19:10 +0100 [PATCH 09/35] staging: lustre: mgc: IR log failure should not stop mount James Simmons <jsimmons@infradead.org> - 2016-11-10 19:10 +0100
csiph-web