[udp] make response message cache configurable #590
Conversation
hasheddan
changed the title
WalkthroughThe changes introduce a new caching mechanism for CoAP messages in the Changes
Poem
📜 Recent review detailsConfiguration used: CodeRabbit UI 📒 Files selected for processing (1)
🚧 Files skipped from review as they are similar to previous changes (1)
Thank you for using CodeRabbit. We offer it for free to the OSS community and would appreciate your support in helping us grow. If you find it useful, would you consider giving us a shout-out on your favorite social media? 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments. CodeRabbit Commands (Invoked using PR comments)
Other keywords and placeholders
CodeRabbit Configuration File (
|
hasheddan
force-pushed
the
udp/modular-cache
branch
from
c447896
to
b1f7350
Compare
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
| // getResponseFromCache gets a message from the response message cache. | ||
| func (cc *Conn) getResponseFromCache(mid int32, resp *pool.Message) (bool, error) { | ||
| return cc.responseMsgCache.Load(strconv.Itoa(int(mid)), resp) | ||
| } | ||
|
|
||
| // addResponseToCache adds a message to the response message cache. | ||
| func (cc *Conn) addResponseToCache(resp *pool.Message) error { | ||
| marshaledResp, err := resp.MarshalWithEncoder(coder.DefaultCoder) | ||
| if err != nil { | ||
| return err | ||
| } | ||
| cacheMsg := make([]byte, len(marshaledResp)) | ||
| copy(cacheMsg, marshaledResp) | ||
| cc.responseMsgCache.LoadOrStore(cc.responseMsgCacheID(resp.MessageID()), cache.NewElement(cacheMsg, time.Now().Add(ExchangeLifetime), nil)) | ||
| return nil | ||
| } | ||
|
|
||
| func (cc *Conn) getResponseFromCache(mid int32, resp *pool.Message) (bool, error) { | ||
| cachedResp := cc.responseMsgCache.Load(cc.responseMsgCacheID(mid)) | ||
| if cachedResp == nil { | ||
| return false, nil | ||
| } | ||
| if rawMsg := cachedResp.Data(); len(rawMsg) > 0 { | ||
| _, err := resp.UnmarshalWithDecoder(coder.DefaultCoder, rawMsg) | ||
| if err != nil { | ||
| return false, err | ||
| } | ||
| return true, nil | ||
| } | ||
| return false, nil | ||
| return cc.responseMsgCache.Store(strconv.Itoa(int(resp.MessageID())), resp) |
There was a problem hiding this comment.
Potential cache key collision due to Message ID reuse across different remote addresses
Using only the Message ID (MID) as the cache key may lead to collisions when the same MID is used by different remote endpoints. According to the CoAP specification, MIDs are unique per endpoint but can be reused over time and across different clients. To prevent cache key collisions and ensure correct message handling, consider incorporating the remote address (e.g., IP address and port) into the cache key.
Apply this diff to modify the cache key to include the remote address:
func (cc *Conn) getResponseFromCache(mid int32, resp *pool.Message) (bool, error) {
+ key := cc.RemoteAddr().String() + ":" + strconv.Itoa(int(mid))
- return cc.responseMsgCache.Load(strconv.Itoa(int(mid)), resp)
+ return cc.responseMsgCache.Load(key, resp)
}
func (cc *Conn) addResponseToCache(resp *pool.Message) error {
+ key := cc.RemoteAddr().String() + ":" + strconv.Itoa(int(resp.MessageID()))
- return cc.responseMsgCache.Store(strconv.Itoa(int(resp.MessageID())), resp)
+ return cc.responseMsgCache.Store(key, resp)
}You'll also need to update calls to these functions to ensure the remote address is available when constructing the cache key.
📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| // getResponseFromCache gets a message from the response message cache. | |
| func (cc *Conn) getResponseFromCache(mid int32, resp *pool.Message) (bool, error) { | |
| return cc.responseMsgCache.Load(strconv.Itoa(int(mid)), resp) | |
| } | |
| // addResponseToCache adds a message to the response message cache. | |
| func (cc *Conn) addResponseToCache(resp *pool.Message) error { | |
| marshaledResp, err := resp.MarshalWithEncoder(coder.DefaultCoder) | |
| if err != nil { | |
| return err | |
| } | |
| cacheMsg := make([]byte, len(marshaledResp)) | |
| copy(cacheMsg, marshaledResp) | |
| cc.responseMsgCache.LoadOrStore(cc.responseMsgCacheID(resp.MessageID()), cache.NewElement(cacheMsg, time.Now().Add(ExchangeLifetime), nil)) | |
| return nil | |
| } | |
| func (cc *Conn) getResponseFromCache(mid int32, resp *pool.Message) (bool, error) { | |
| cachedResp := cc.responseMsgCache.Load(cc.responseMsgCacheID(mid)) | |
| if cachedResp == nil { | |
| return false, nil | |
| } | |
| if rawMsg := cachedResp.Data(); len(rawMsg) > 0 { | |
| _, err := resp.UnmarshalWithDecoder(coder.DefaultCoder, rawMsg) | |
| if err != nil { | |
| return false, err | |
| } | |
| return true, nil | |
| } | |
| return false, nil | |
| return cc.responseMsgCache.Store(strconv.Itoa(int(resp.MessageID())), resp) | |
| // getResponseFromCache gets a message from the response message cache. | |
| func (cc *Conn) getResponseFromCache(mid int32, resp *pool.Message) (bool, error) { | |
| key := cc.RemoteAddr().String() + ":" + strconv.Itoa(int(mid)) | |
| return cc.responseMsgCache.Load(key, resp) | |
| } | |
| // addResponseToCache adds a message to the response message cache. | |
| func (cc *Conn) addResponseToCache(resp *pool.Message) error { | |
| key := cc.RemoteAddr().String() + ":" + strconv.Itoa(int(resp.MessageID())) | |
| return cc.responseMsgCache.Store(key, resp) | |
| } |
There was a problem hiding this comment.
I agree. Collisions could occur – could you please use responseMsgCacheID(resp.MessageID()) as the key?
There was a problem hiding this comment.
@jkralik, certainly! We can reintroduce the responseMsgCacheID function and use it as the key to prevent collisions. Here are the necessary changes:
First, add the responseMsgCacheID function back into your code:
func (cc *Conn) responseMsgCacheID(msgID int32) string {
return fmt.Sprintf("resp-%v-%d", cc.RemoteAddr(), msgID)
}Then, update the getResponseFromCache and addResponseToCache functions to use this key:
// getResponseFromCache gets a message from the response message cache.
func (cc *Conn) getResponseFromCache(mid int32, resp *pool.Message) (bool, error) {
return cc.responseMsgCache.Load(cc.responseMsgCacheID(mid), resp)
}
// addResponseToCache adds a message to the response message cache.
func (cc *Conn) addResponseToCache(resp *pool.Message) error {
return cc.responseMsgCache.Store(cc.responseMsgCacheID(resp.MessageID()), resp)
}This change ensures that the cache key includes both the remote address and the message ID, preventing collisions across different remote addresses.
There was a problem hiding this comment.
@jkralik I intentionally made this update because, if using DTLS 1.2 Connection IDs, the underlying remote IP address / port can change and the connection can continue. Items in the cache would then no longer be accessible because we would attempt to access with a different key. In the event that we are not using DTLS, or using DTLS without Connection IDs, I don't believe that Conns should be used across multiple endpoints, and I believe there are issues other than just the cache keys here if they are. Is there a scenario in which a Conn is being used across endpoints, thus meaning that a responseMsgCache would not be local to a single endpoint?
There was a problem hiding this comment.
Seems like the humans are having a chat. I'll hop back into my burrow for now. If you need me again, just tag @coderabbitai in a new comment, and I'll come hopping out!
Modifies UDP conn implementation to allow for a custom response message cache to be provided. This change is implemented in such a way that API compatibility with previous v3 releases is preserved. Users may now provide an option to supply their own cache implementation. Structured messages, rather than their serialized representation, are now passed to the cache implementation to allow for caching decisions to be made in the cache implementation. For example, it may be desirable to skip caching blockwise message responses if the entire underlying data being transferred is also cached. The cache implementation is responsible for cloning messages or otherwise ensuring that it is not storing data that may subsequently be modified. Signed-off-by: Daniel Mangum <[email protected]>
hasheddan
force-pushed
the
udp/modular-cache
branch
from
b1f7350
to
b80b7a8
Compare
Modifies UDP conn implementation to allow for a custom response message cache to be provided. This change is implemented in such a way that API compatibility with previous v3 releases is preserved. Users may now provide an option to supply their own cache implementation. Structured messages, rather than their serialized representation, are now passed to the cache implementation to allow for caching decisions to be made in the cache implementation. For example, it may be desirable to skip caching blockwise message responses if the entire underlying data being transferred is also cached. The cache implementation is responsible for cloning messages or otherwise ensuring that it is not storing data that may subsequently be modified.
Replaces #589
Summary by CodeRabbit
New Features
Improvements