Comment #30 on issue 375 by [email protected]: Accept4 on FreeBSD 10 -
sockets stuck in CLOSE
https://code.google.com/p/memcached/issues/detail?id=375
I will try .20 as this was the version where we originally encountered the
problem.
As I already said this bug shows signs of a race condition and/or overflow
(hard to reproduce, no fixed reproduction rule, happens under heavy load).
The following fixes introduced in .22 may have had an impact:
- Fix potential corruption for incr/decr of 0b items
- slab re-balance is not thread-safe in function do_item_get
- Fix potential corruption in hash table expansion
- use item lock instead of global lock when hash expanding
That is of course if the bug is really on memcached's side.
On the FreeBSD side there were not many changes that are related (see
https://www.freebsd.org/releases/10.1R/relnotes.html):
- SA-14:19.tcp 16 September 2014 Denial of Service in TCP packet
processing
- SA-14:08.tcp 30 April 2014 Fix TCP reassembly vulnerability
- A bug in sctp(4) that would allow two listening sockets bound to the same
port has been fixed.
These are all not directly related but changes in that area might have
affected the connection code.
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings
--
---
You received this message because you are subscribed to the Google Groups "memcached" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
