Bug #4023: kclient: d_revalidate is abusing d_parent - CephFS - Ceph

Actions

Copy link

Bug #4023

open

kclient: d_revalidate is abusing d_parent

Added by Sage Weil about 11 years ago. Updated over 4 years ago.

Status:

New

Priority:

Normal

Assignee:

Category:

Correctness/Safety

Target version:

% Done:

Source:

Community (dev)

Tags:

Backport:

Regression:

Severity:

3 - minor

Reviewed:

Affected Versions:

ceph-qa-suite:

Component(FS):

kceph

Labels (FS):

Pull request ID:

Crash signature (v1):

Crash signature (v2):

Description

See Viro's email to linux-fsdevel, http://marc.info/?l=linux-fsdevel&m=135968126020360&w=2 .

We probably need to link the ceph_dentry_info to struct on the parent inode (properly refcounted) when its validity is chained to the parent's cap. This is sort of duplicating d_parent, but the current code is abusing it and racy.

Actions

Copy link

Updated by Sage Weil about 11 years ago

Assignee set to Sage Weil

Actions

Copy link

Updated by Greg Farnum about 10 years ago

Priority changed from Normal to High

Is this still a problem?

Actions

Copy link

Updated by Zheng Yan about 10 years ago

The race still exists, but I don't think it's big problem. Because even if ceph_get_dentry_parent_inode() returns a wrong parent inode, ceph_d_revalidate() still works.

Actions

Copy link