Con Kolivas
|
11d14620e6
|
Convert to gzip
|
9 years ago |
|
Con Kolivas
|
e3fc6a1e56
|
Selectively compress only large packets greater than one MTU and identify lz4 compatible clients immediately
|
9 years ago |
|
Con Kolivas
|
17cc4411af
|
Fix compilation and send lz4 compatible clients compressed data
|
9 years ago |
|
Con Kolivas
|
950b855b2d
|
Implement backwardly compatible lz4 compression support to be used by passthroughs and nodes
|
9 years ago |
|
Con Kolivas
|
46aef425f2
|
Drop clients when we don't have an upstream pool in passthrough/node mode
|
9 years ago |
|
Con Kolivas
|
3dc070fcc2
|
Do initial parsing of subscribe/auth in node mode
|
9 years ago |
|
Con Kolivas
|
04eaabfaa2
|
Add mining nodes to a linked list and send them the workinfo for now
|
9 years ago |
|
Con Kolivas
|
5ddfcdb14e
|
Identify all node message types in the stratifier
|
9 years ago |
|
Con Kolivas
|
50c9b5ecd2
|
Move message type detection to stratifier
|
9 years ago |
|
Con Kolivas
|
c5718d510e
|
Detect node message type in the generator
|
9 years ago |
|
Con Kolivas
|
b09bb253fc
|
Act as a passthrough in node mode, passing what type of stratum message we're passing through
|
9 years ago |
|
Con Kolivas
|
2eaeb2a96c
|
Don't keep storing shares in redirector mode after we've attempted to redirect the client
|
9 years ago |
|
ckolivas
|
c437d3283b
|
Only try to redirect clients once in redirector mode, acting as a regular passthrough thereafter
|
9 years ago |
|
Con Kolivas
|
e413e0e97b
|
Increase size of address storage for clients to store full inet6 name
|
9 years ago |
|
Con Kolivas
|
b0c30a3f7d
|
Disconnect clients whose sends block for more than 60 seconds
|
9 years ago |
|
Con Kolivas
|
c0e4310c2d
|
Send a dropall to the stratifier when we're in reject mode
|
9 years ago |
|
Con Kolivas
|
99ce264a8e
|
Send a dropall to the stratifier when we're in reject mode
|
9 years ago |
|
Con Kolivas
|
bce4cab176
|
Roll back code for relay node.
|
9 years ago |
|
Con Kolivas
|
b7e35effa5
|
Add backup btcd parsing in node mode
|
10 years ago |
|
Con Kolivas
|
1ba52b3daa
|
Round robin over the redirect URLs in redirector mode, keeping clients from the same IP on the same URL
|
10 years ago |
|
Con Kolivas
|
0dafbe12b4
|
Comments
|
10 years ago |
|
Con Kolivas
|
3ba8e1d6c1
|
Hold an extra reference for the redirect_client message in the redirector
|
10 years ago |
|
Con Kolivas
|
65957df813
|
Issue reconnect to miner submitting valid share in redirector mode
|
10 years ago |
|
Con Kolivas
|
ec9e730b1f
|
Look for unique ID accepted shares in the connector in redirector mode
|
10 years ago |
|
Con Kolivas
|
03edcc03cd
|
Check for valid shares in redirector and redirect for first url for now
|
10 years ago |
|
Con Kolivas
|
b197ba2ff9
|
Maintain a reference to which message we're currently sending to a client to prevent sending interleaved messages should they block
|
10 years ago |
|
Con Kolivas
|
dcf55ff1fb
|
Maintain a reference to which message we're currently sending to a client to prevent sending interleaved messages should they block
|
10 years ago |
|
Con Kolivas
|
cb640b6ab2
|
Check for epoll hangups fully
|
10 years ago |
|
Con Kolivas
|
dcfcae1779
|
Add errno details to write based close
|
10 years ago |
|
Con Kolivas
|
b73f44f00c
|
Tidy up and demote oversize message, adding client id
|
10 years ago |
|
Con Kolivas
|
5365f1a420
|
Increase listen backlog to 8k
|
10 years ago |
|
Con Kolivas
|
882bd693dd
|
Display connector stats every ~60s in passthrough mode
|
10 years ago |
|
Con Kolivas
|
68e44be3ce
|
Check for oversized client message before doing any reads to avoid possibility of exactly the wrong size buffer to ever invalidate the client, adding more info to a downgraded message
|
10 years ago |
|
Con Kolivas
|
8447e3ff55
|
Microoptimise
|
10 years ago |
|
Con Kolivas
|
27d68200ec
|
Add better stats about queued sends consisten with the list changes
|
10 years ago |
|
Con Kolivas
|
ed44843e37
|
Append all new sender_ends in csender instead of only the first
|
10 years ago |
|
Con Kolivas
|
f835697165
|
Microoptimisation
|
10 years ago |
|
Con Kolivas
|
7c62450297
|
Increase listen backlog to speed up reconnects up to the system config limits
|
10 years ago |
|
Con Kolivas
|
83b23d864f
|
Remove decrease of listen backlog which is of questionable utility
|
10 years ago |
|
Con Kolivas
|
f580a82a06
|
Drop the subclients of passthroughs that no longer exist
|
10 years ago |
|
Con Kolivas
|
883d870060
|
Use the simpler read/write calls and make all client sockets non-blocking
|
10 years ago |
|
Con Kolivas
|
3919d18260
|
Rework write path to have no potentially blocking calls and be able to send partial messages
|
10 years ago |
|
Con Kolivas
|
1efe7120c4
|
Use ref_client_by_id in send_client instead of open coding it
|
10 years ago |
|
Con Kolivas
|
ec39117a1c
|
Don't return client if it is invalid in ref_client_by_id in the connector
|
10 years ago |
|
Con Kolivas
|
0395dd052f
|
Use the client reference count in the connector to protect the client fd, closing it only once there are no more references to it
|
10 years ago |
|
Con Kolivas
|
9375362044
|
Revert "Use a unique event structure for each client added to the epoll list"
This reverts commit 8f1336986f.
Unnecessary.
|
10 years ago |
|
Con Kolivas
|
8f1336986f
|
Use a unique event structure for each client added to the epoll list
|
10 years ago |
|
Con Kolivas
|
6a159bf5dc
|
We won't get no error if we are not testing fd -1 sockets any more, and use errno for consistent parsing of socket close errors
|
10 years ago |
|
Con Kolivas
|
62d1ec3f8a
|
Use signal handlers from sender and receiver threads in the connector avoid needing to pthread tryjoin on every message
|
10 years ago |
|
Con Kolivas
|
2bde0bbe3d
|
Avoid trying to parse messages or test the client's error if their fd has already been invalidated
|
10 years ago |
