On 10/09/2021 11:42, Mark Thomas wrote:
Hi Erik,
Thanks for the report. I'm looking at this now.
I'm testing with a simple index page that references 3 largish images
(~6MB each).
I've found an issue with HTTP/2, sendfile and StackOverflowExcpetion
that I have a local fix for.
With that fix in place, I can see a flow control issue. Somehow, a
stream is getting a larger allocation from the connection control window
than the stream control window. That leads to some internal values
having unexpected (negative) values and things quickly escalate to the
connection closing abruptly from there. I'm currently looking into how
this happens.
I'm not sure if I am seeing a different issue to you or just a different
symptom of the same issue. I'll keep the thread updated with progress.
I found the root cause - there were further concurrency issues in the
connection flow control window management. I've refactored the code to
simplify the approach and (hopefully) make it more robust. I am no
longer able to recreate the issue I was seeing.
The fixes will be in the October release round. If anyone would like to
test this sooner than that, you can build from source or I can make a
test build available on request.
Mark
Mark
On 10/09/2021 11:02, Erik Nilsson wrote:
Hi,
We still have a problem with the http2 flow control and RST_STREAMS
when running nghttp -vnsay <tomcaturl>
I have attached a screenshot and Tomcat debug logging.
/Erik
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
