summaryrefslogtreecommitdiff
path: root/src/rfkill
diff options
context:
space:
mode:
authorMichal Sekletar <msekleta@redhat.com>2017-01-23 17:12:35 +0100
committerMichal Sekletar <msekleta@redhat.com>2017-04-11 09:22:25 +0200
commite266c068b5597e18b2299f9c9d3ee6cf04198c41 (patch)
treefef3983bff140b995628f4b8c082804377694d64 /src/rfkill
parentc258349f1a56b987fd91a5c136fd15536eec3890 (diff)
service: serialize information about currently executing command
Stored information will help us to resume execution after the daemon-reload. This commit implements following scheme, * On serialization: - we count rank of the currently executing command - we store command type, its rank and command line arguments * On deserialization: - configuration is parsed and loaded - we deserialize stored data, command type, rank and arguments - we look at the given rank in the list and if command there has same arguments then we restore execution at that point - otherwise we search respective command list and we look for command that has the same arguments - if both methods fail we do not do not resume execution at all To better illustrate how does above scheme works, please consider following cases (<<< denotes position where we resume execution after reload) ; Original unit file [Service] ExecStart=/bin/true <<< ExecStart=/bin/false ; Swapped commands ; Second command is not going to be executed [Service] ExecStart=/bin/false ExecStart=/bin/true <<< ; Commands added before ; Same commands are problematic and execution could be restarted at wrong place [Service] ExecStart=/bin/foo ExecStart=/bin/bar ExecStart=/bin/true <<< ExecStart=/bin/false ; Commands added after ; Same commands are not an issue in this case [Service] ExecStart=/bin/true <<< ExecStart=/bin/false ExecStart=/bin/foo ExecStart=/bin/bar ; New commands interleaved with old commands ; Some new commands will be executed while others won't ExecStart=/bin/foo ExecStart=/bin/true <<< ExecStart=/bin/bar ExecStart=/bin/false As you can see, above scheme has some drawbacks. However, in most cases (we assume that in most common case unit file command list is not changed while some other command is running for the same unit) it should cause that systemd does the right thing, which is restoring execution exactly at the point we were before daemon-reload. Fixes #518
Diffstat (limited to 'src/rfkill')
0 files changed, 0 insertions, 0 deletions