https://issues.apache.org/bugzilla/show_bug.cgi?id=56216
Bug ID: 56216
Summary: multiple reqtimeout_con_cfg created for each request
-> wrong timeout behavior
Product: Apache httpd-2
Version: 2.4.7
Hardware: PC
OS: Linux
Status: NEW
Severity: normal
Priority: P2
Component: mod_reqtimeout
Assignee: [email protected]
Reporter: [email protected]
Created attachment 31365
--> https://issues.apache.org/bugzilla/attachment.cgi?id=31365&action=edit
patch: only create reqtimeout_con_cfg once per connection
while testing mod_reqtimout, I sometimes got strange timeouts i couldn't
explain. After adding some debug code to mod_reqtimeout.c I found out that:
1) reqtimeout_init() is called before each request (not only at the start of
the connection)
2) therefore for each request a new reqtimeout_con_cfg is created and set as
context to the connection config: ap_set_module_config(c->conn_config,
&reqtimeout_module, ccfg) and is added to the input filter:
ap_add_input_filter("reqtimeout", ccfg, NULL, c);
3) For subsequent Requests I got then calls to reqtimeout_filter(), with each
registered reqtimeout_con_cfg. That means during the first Request
reqtimeout_filter() was called with the correct reqtime_con_cfg, in the second
Request reqtimeout_filter() was called twice() one time with the actual
reqtimeout_con_cfg and one time with the one from the first Request. And so
forth.
4) Then one call to reqtimeout_filter() with the obsolete ccfg decided to
detect a timeout, and the request was aborted
It looks to me, it was thought, reqtimeout_init() reqistered in hook
'ap_hook_process_connection' shall only be called once. But since it is called
multiple times, it has to be secured, not to create and register a new
reqtimeout_con_cfg each time. See proposed attached patch.
--
You are receiving this mail because:
You are the assignee for the bug.
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
