summaryrefslogtreecommitdiff
path: root/jni/node-inl.h
diff options
context:
space:
mode:
author Zim <zezeozue@google.com> 2020-02-19 01:23:26 +0000
committer Zim <zezeozue@google.com> 2020-02-19 11:07:43 +0000
commita76c349ea76a43d5bb7e4d889bb43cccf72371fc (patch)
treec41e7ba6cf889522d76c6abf61f4bb439580b583 /jni/node-inl.h
parentdab62a2b9b915d78752f925fda06eefa5f2d81c4 (diff)
Invalidate FUSE VFS dentry cache
To improve filesystem performance we can reduce unnecessary requests to the FUSE daemon by enabling VFS caching of dentries after a FUSE_LOOKUP request. To enable correctly, we should ensure that the lower filesystem dentries are not modified outside of the FUSE driver. Unfortunately, we already do this when handling IO requests over binder using the ContentResolver interface. To fix, we should invalidate any FUSE dentries in the VFS cache so that subsequent lookups for those files do not return incorrectly cached values from the VFS. This change just adds support for invalidating FUSE VFS cache dentries, subsequent cl will call it when required. Note that it should never be called in the execution path of a related filesytem operation. See fuse_lowlevel.h for more details. Will ensure that this is never called in the FUSE execution path in a follow up Test: atest FuseDaemonHostTest#testVfsCacheConsistency Bug: 145741152 Change-Id: I54e67f63a29c82a392916fb28d1a7f8f894c685b
Diffstat (limited to 'jni/node-inl.h')
-rw-r--r--jni/node-inl.h5
1 files changed, 5 insertions, 0 deletions
diff --git a/jni/node-inl.h b/jni/node-inl.h
index 76d1ba91b..e21ae5448 100644
--- a/jni/node-inl.h
+++ b/jni/node-inl.h
@@ -160,6 +160,11 @@ class node {
return name_;
}
+ node* GetParent() const {
+ std::lock_guard<std::recursive_mutex> guard(*lock_);
+ return parent_;
+ }
+
inline void AddHandle(handle* h) {
std::lock_guard<std::recursive_mutex> guard(*lock_);
handles_.emplace_back(std::unique_ptr<handle>(h));