Changes

Jump to navigation Jump to search
1,260 bytes added ,  08:51, 30 November 2015
no edit summary
Line 4: Line 4:  
* Ability to check CRIU features via [[RPC]]
 
* Ability to check CRIU features via [[RPC]]
 
* New zdtm.py test suite
 
* New zdtm.py test suite
* C/R of read-only bind mounts
  −
* C/R of IPv6 routes and iptables rules
  −
* C/R of ip rules (it ip tool supports such)
   
* Pre-dump and pre-restore [[action scripts]]
 
* Pre-dump and pre-restore [[action scripts]]
 
* The "info" action in [[CRIT]] showing stats about image file
 
* The "info" action in [[CRIT]] showing stats about image file
Line 12: Line 9:  
* Python API -- pycriu
 
* Python API -- pycriu
 
* Ability to add custom paths to irmap scan
 
* Ability to add custom paths to irmap scan
* C/R of ignore_routes_with_linkdown netns devconf
+
* C/R of
 +
** read-only bind mounts
 +
** IPv6 routes and iptables rules
 +
** ip rules (it ip tool supports such)
 +
** ignore_routes_with_linkdown netns devconf
 +
** empty bridges in netns
 +
** FILTER mode of seccomp
 +
** IP_FREEBIND socket option
    
=== Optimizations/improvements ===
 
=== Optimizations/improvements ===
Line 24: Line 28:  
* Improved page-server dump speed by keeping control over the Nagle algorithm  
 
* Improved page-server dump speed by keeping control over the Nagle algorithm  
 
* Read pages.img in more optimal manner rather than page-by-page
 
* Read pages.img in more optimal manner rather than page-by-page
 +
* Less "Error"-s in logs, that actually don't lead to errors
 +
* Slightly faster /proc/pid/status parsing
 +
* Dead/live-locks on internal criu locks now emits a warning into logs
    
=== Fixes ===
 
=== Fixes ===
Line 45: Line 52:  
* Unsupported filesystems silently failed the dump
 
* Unsupported filesystems silently failed the dump
 
* External tmpfs (and some other) mounts generated tarballs with their contents
 
* External tmpfs (and some other) mounts generated tarballs with their contents
 +
* Privately mapped files were picked from wrong mount namespace
 +
* Controlling tty could be restored on wrong tty end
 +
* Tmpfs mount of sub-namespace was restored from wrong image file
 +
* Potential stack overflow in libcriu
 +
* Partially-restored tasks could be left after failed restore
 +
* In-container TCP connection sometimes failed to restore
 +
* Race in sending SIGSTOP vs dump might cause dump to fail
 +
* Post-restore actions could generate stats files in wrong directories
 +
* Freeze-cgroup didn't take sub-cgroups' tasks into account
 +
* Tentative state in IPv6 sockets binding prevented socket from being bound immediately
 +
* Restoring from images with files pointing to /proc file of dead tasks could crash
 +
* Tasks with STOP in queue (i.e. -- not ''yet'' stopped) were CONT-ed in case of --leave-running dump
 +
* Stopped task with one more STOP in queue caused dump to stuck
 +
* If parent task left the MNT namespace it created for children restore could BUG()
 +
* Link-local IPv6 addresses sometimes failed to bind() at restore
    
=== Security ===
 
=== Security ===
 
* Service run as root could allow users to violate ptrace policies
 
* Service run as root could allow users to violate ptrace policies
 
* Service run as root could give users access to privileged files and directories
 
* Service run as root could give users access to privileged files and directories

Navigation menu