Bug #16066: client: FAILED assert(root_ancestor->qtree == __null) - CephFS - Ceph

Actions

Copy link

Bug #16066

closed

client: FAILED assert(root_ancestor->qtree == __null)

Added by John Spray almost 8 years ago. Updated about 5 years ago.

Status:

Resolved

Priority:

High

Assignee:

Category:

Target version:

% Done:

Source:

other

Tags:

Backport:

jewel

Regression:

Severity:

3 - minor

Reviewed:

Affected Versions:

ceph-qa-suite:

Component(FS):

Client

Labels (FS):

quotas

Pull request ID:

Crash signature (v1):

Crash signature (v2):

Description

Seen on test branch where client quota was enabled by default (https://github.com/ceph/ceph/pull/9346) -- presumably this is an existing bug that we didn't notice because we don't usually have client quota enabled during testing.

Assertion: client/Client.cc: 12008: FAILED assert(root_ancestor->qtree == __null)
ceph version 10.2.0-1432-g387a52d (387a52daef4fba042dfd041a54544090a3754abf)
 1: (ceph::__ceph_assert_fail(char const*, char const*, int, char const*)+0x85) [0x55e99159efa5]
 2: (Client::get_quota_root(Inode*)+0x7b2) [0x55e9913c4cb2]
 3: (Client::check_quota_condition(Inode*, std::function<bool (Inode const&)>)+0x3d) [0x55e9913c530d]
 4: (Client::is_quota_files_exceeded(Inode*)+0x5e) [0x55e9913c53be]
 5: (Client::_mkdir(Inode*, char const*, unsigned int, int, int, boost::intrusive_ptr<Inode>*)+0xd3) [0x55e9913f2b23]
 6: (Client::ll_mkdir(Inode*, char const*, unsigned int, stat*, Inode**, int, int)+0xfb) [0x55e9913f353b]
 7: (()+0x19ad72) [0x55e991396d72]
 8: (()+0x16bdb) [0x7f62351bebdb]
 9: (()+0x13471) [0x7f62351bb471]
 10: (()+0x7dc5) [0x7f6234614dc5]
 11: (clone()+0x6d) [0x7f62334faced]

Related issues 1 (0 open — 1 closed)

Actions

Copy link

Updated by John Spray almost 8 years ago

http://pulpito.ceph.com/jspray-2016-05-28_13:42:42-fs-wip-jcsp-testing-20160527b---basic-mira/220171
+other instances in the same suite run

Actions

Copy link

Updated by Zheng Yan almost 8 years ago

Status changed from New to 12

The quota code expects that directory is alway connected to the FS hierarchy. But this is not true, we can create disconnected directory inode by

Client1> mkdir testdir
Client1> cd testdir
Client1> setfattr -n ceph.quota.max_files -n 100 .
Client1> mkdir dir1

Client2> cd testdir/dir1

Client1> mv dir1 dir2
Client1> mkdir dir1

Client2> stat ../dir1
Client2> seq 1 200 | xargs touch // ceph-fuse crashes

This issue is difficult to handle. For now, we can use lookup-parent request to reconnect the directory. The long-term solution should be something like snaprealm.

Actions

Copy link