Skip to content
Projects
Groups
Snippets
Help
Loading...
Help
Support
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
N
neoppod
Project overview
Project overview
Details
Activity
Releases
Repository
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Issues
0
Issues
0
List
Boards
Labels
Milestones
Merge Requests
0
Merge Requests
0
CI / CD
CI / CD
Pipelines
Jobs
Schedules
Analytics
Analytics
CI / CD
Repository
Value Stream
Wiki
Wiki
Snippets
Snippets
Members
Members
Collapse sidebar
Close sidebar
Activity
Graph
Create a new issue
Jobs
Commits
Issue Boards
Open sidebar
Levin Zimmermann
neoppod
Commits
412c35a2
Commit
412c35a2
authored
Aug 22, 2017
by
Kirill Smelkov
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
.
parent
ebbd4b8f
Changes
1
Hide whitespace changes
Inline
Side-by-side
Showing
1 changed file
with
68 additions
and
26 deletions
+68
-26
go/neo/client/cache.go
go/neo/client/cache.go
+68
-26
No files found.
go/neo/client/cache.go
View file @
412c35a2
...
@@ -22,6 +22,7 @@ package client
...
@@ -22,6 +22,7 @@ package client
import
(
import
(
"sync"
"sync"
"unsafe"
"lab.nexedi.com/kirr/neo/go/zodb"
"lab.nexedi.com/kirr/neo/go/zodb"
)
)
...
@@ -37,8 +38,7 @@ type Cache struct {
...
@@ -37,8 +38,7 @@ type Cache struct {
oidDir
map
[
zodb
.
Oid
]
*
oidEntry
oidDir
map
[
zodb
.
Oid
]
*
oidEntry
bytesUsed
int64
size
int64
// in bytes
lru
revListHead
lru
revListHead
}
}
...
@@ -95,6 +95,10 @@ type revListHead struct {
...
@@ -95,6 +95,10 @@ type revListHead struct {
next
,
prev
*
revCacheEntry
next
,
prev
*
revCacheEntry
}
}
func
(
h
*
revListHead
)
rceFromInLRU
()
(
rce
*
revCacheEntry
)
{
return
(
*
revCacheEntry
)(
unsafe
.
Pointer
(
h
)
-
unsafe
.
OffsetOf
(
rce
.
inLRU
))
}
// XXX -> to ctor?
// XXX -> to ctor?
func
(
h
*
revListHead
)
init
()
{
func
(
h
*
revListHead
)
init
()
{
h
.
next
=
h
h
.
next
=
h
...
@@ -118,6 +122,29 @@ func (a *revListHead) MoveBefore(b *revListHead) {
...
@@ -118,6 +122,29 @@ func (a *revListHead) MoveBefore(b *revListHead) {
a
.
prev
.
next
=
a
a
.
prev
.
next
=
a
}
}
// XXX doc
func
(
oce
*
oidCacheEntry
)
newRevEntry
(
before
zodb
.
Tid
)
*
revCacheEntry
{
rce
:=
&
revCacheEntry
{
parent
:
oce
,
serial
:
0
,
before
:
before
,
ready
:
make
(
chan
struct
{})}
}
rce
.
inLRU
.
init
()
return
rce
}
// XXX doc; must be called with oce lock held
func
(
oce
*
oidCacheEntry
)
del
(
rce
*
revCacheEntry
)
{
for
i
,
r
:=
range
rce
.
revv
{
if
r
==
rce
{
rce
.
revv
=
append
(
rce
.
revv
[
:
i
],
rce
.
revv
[
i
+
1
:
])
return
}
}
panic
(
"rce not found"
)
}
// cleaner is the process that cleans cache by evicting less-needed entries.
// cleaner is the process that cleans cache by evicting less-needed entries.
func
(
c
*
cache
)
cleaner
()
{
func
(
c
*
cache
)
cleaner
()
{
...
@@ -144,23 +171,23 @@ func (c *cache) Load(xid zodb.Xid) (data []byte, tid Tid, err error) {
...
@@ -144,23 +171,23 @@ func (c *cache) Load(xid zodb.Xid) (data []byte, tid Tid, err error) {
// exit with cacheEntry locked and cache.before read consistently
// exit with cacheEntry locked and cache.before read consistently
c
.
mu
.
RLock
()
c
.
mu
.
RLock
()
entry
:=
c
.
entryMap
[
xid
.
Oid
]
oce
:=
c
.
entryMap
[
xid
.
Oid
]
cacheBefore
:=
c
.
before
cacheBefore
:=
c
.
before
cacheMemUsed
:=
c
.
memUsed
cacheMemUsed
:=
c
.
memUsed
if
entry
!=
nil
{
if
oce
!=
nil
{
entry
.
Lock
()
oce
.
Lock
()
c
.
mu
.
RUnlock
()
c
.
mu
.
RUnlock
()
}
else
{
}
else
{
c
.
mu
.
RUnlock
()
c
.
mu
.
RUnlock
()
c
.
mu
.
Lock
()
c
.
mu
.
Lock
()
entry
=
c
.
entryMap
[
xid
.
Oid
]
oce
=
c
.
entryMap
[
xid
.
Oid
]
if
entry
==
nil
{
if
oce
==
nil
{
entry
=
&
cacheEntry
{}
oce
=
&
cacheEntry
{}
c
.
entryMap
[
xid
.
Oid
]
=
entry
c
.
entryMap
[
xid
.
Oid
]
=
oce
}
}
cacheBefore
=
c
.
before
// reload c.before for correctness
cacheBefore
=
c
.
before
// reload c.before for correctness
entry
.
Lock
()
oce
.
Lock
()
c
.
mu
.
Unlock
()
c
.
mu
.
Unlock
()
}
}
...
@@ -169,9 +196,9 @@ func (c *cache) Load(xid zodb.Xid) (data []byte, tid Tid, err error) {
...
@@ -169,9 +196,9 @@ func (c *cache) Load(xid zodb.Xid) (data []byte, tid Tid, err error) {
// before -> revCacheEntry
// before -> revCacheEntry
if
xid
.
TidBefore
{
if
xid
.
TidBefore
{
l
:=
len
(
entry
.
revv
)
l
:=
len
(
oce
.
revv
)
i
:=
sort
.
Search
(
l
,
func
(
i
int
)
bool
{
i
:=
sort
.
Search
(
l
,
func
(
i
int
)
bool
{
before
:=
entry
.
revv
[
i
]
.
before
before
:=
oce
.
revv
[
i
]
.
before
if
before
==
zodb
.
TidMax
{
if
before
==
zodb
.
TidMax
{
before
=
cacheBefore
before
=
cacheBefore
}
}
...
@@ -181,13 +208,13 @@ func (c *cache) Load(xid zodb.Xid) (data []byte, tid Tid, err error) {
...
@@ -181,13 +208,13 @@ func (c *cache) Load(xid zodb.Xid) (data []byte, tid Tid, err error) {
switch
{
switch
{
// not found - tid > max(revv.before) - create new max entry
// not found - tid > max(revv.before) - create new max entry
case
i
==
l
:
case
i
==
l
:
rce
=
&
revCacheEntry
{
serial
:
0
,
before
:
before
}
rce
=
oce
.
newRevEntry
(
xid
.
Tid
)
if
rce
.
before
==
cacheBefore
{
if
rce
.
before
==
cacheBefore
{
// XXX better do this when the entry becomes loaded ?
// XXX better do this when the entry becomes loaded ?
rce
.
before
=
zodb
.
TidMax
// XXX vs concurrent invalidations?
rce
.
before
=
zodb
.
TidMax
// XXX vs concurrent invalidations?
}
}
rceNew
=
true
rceNew
=
true
entry
.
revv
=
append
(
entry
.
revv
,
rce
)
oce
.
revv
=
append
(
oce
.
revv
,
rce
)
// found:
// found:
// tid <= revv[i].before
// tid <= revv[i].before
...
@@ -195,13 +222,13 @@ func (c *cache) Load(xid zodb.Xid) (data []byte, tid Tid, err error) {
...
@@ -195,13 +222,13 @@ func (c *cache) Load(xid zodb.Xid) (data []byte, tid Tid, err error) {
// exact match - we already have it
// exact match - we already have it
case
xid
.
Tid
==
revv
[
i
]
.
before
:
case
xid
.
Tid
==
revv
[
i
]
.
before
:
rce
=
entry
.
revv
[
i
]
rce
=
oce
.
revv
[
i
]
// if outside [serial, before) - insert new entry
// if outside [serial, before) - insert new entry
case
!
(
revv
[
i
]
.
serial
!=
0
&&
revv
[
i
]
.
serial
<=
xid
.
Tid
)
:
case
!
(
revv
[
i
]
.
serial
!=
0
&&
revv
[
i
]
.
serial
<=
xid
.
Tid
)
:
rce
=
&
revCacheEntry
{
serial
:
0
,
before
:
xid
.
Tid
}
rce
=
oce
.
newRevEntry
(
xid
.
Tid
)
rceNew
=
true
rceNew
=
true
entry
.
revv
=
append
(
entry
.
revv
[
:
i
],
rce
,
entry
.
revv
[
i
:
])
oce
.
revv
=
append
(
oce
.
revv
[
:
i
],
rce
,
oce
.
revv
[
i
:
])
}
}
// XXX serial -> revCacheEntry
// XXX serial -> revCacheEntry
...
@@ -209,15 +236,12 @@ func (c *cache) Load(xid zodb.Xid) (data []byte, tid Tid, err error) {
...
@@ -209,15 +236,12 @@ func (c *cache) Load(xid zodb.Xid) (data []byte, tid Tid, err error) {
// TODO
// TODO
}
}
if
rceNew
{
oce
.
Unlock
()
rce
.
ready
=
make
(
chan
struct
{})
}
entry
.
Unlock
()
// entry was already in cache - use it
// entry was already in cache - use it
if
!
rceNew
{
if
!
rceNew
{
<-
rce
.
ready
<-
rce
.
ready
// XXX update lru
return
rce
.
data
,
rce
.
serial
,
rce
.
err
return
rce
.
data
,
rce
.
serial
,
rce
.
err
}
}
...
@@ -228,14 +252,32 @@ func (c *cache) Load(xid zodb.Xid) (data []byte, tid Tid, err error) {
...
@@ -228,14 +252,32 @@ func (c *cache) Load(xid zodb.Xid) (data []byte, tid Tid, err error) {
rce
.
err
=
err
rce
.
err
=
err
close
(
ready
)
close
(
ready
)
entry
.
Lock
()
oce
.
Lock
()
// XXX merge with adjacent entries in revv
// XXX merge with adjacent entries in revv
entry
.
Unlock
()
oce
.
Unlock
()
c
.
Lock
()
c
.
Lock
()
c
.
memUsed
+=
len
(
data
)
c
.
size
+=
len
(
data
)
// XXX GC if memUsed > threthold ?
if
c
.
size
>
c
.
sizeTarget
{
->
run
gc
}
c
.
Unlock
()
c
.
Unlock
()
return
rce
.
data
,
rce
.
serial
,
rce
.
err
return
rce
.
data
,
rce
.
serial
,
rce
.
err
}
}
func
(
c
*
cache
)
gc
(
...
)
{
c
.
lruMu
.
Lock
()
revh
:=
c
.
lru
.
next
c
.
lruMu
.
Unlock
()
for
;
revh
!=
&
lru
;
revh
=
revh
.
next
{
// .next after .delete - ok ?
rce
:=
revh
.
rceFromInLRU
()
oce
:=
rce
.
parent
oce
.
Lock
()
oce
.
del
(
rce
)
oce
.
Unlock
()
}
}
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment