9bc8532d56
container/initdev: mount tmpfs on shm for ro dev
...
Programs expect /dev/shm to be a writable tmpfs.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-26 03:27:07 +09:00
07194c74cb
release: 0.2.0
...
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-26 02:23:59 +09:00
4cf694d2b3
hst: use hsu userid for share path suffix
...
The privileged user is identifier to hakurei through its hsu userid. Using the kernel uid here makes little sense and is a leftover design choice from before hsu was implemented.
Closes #7 .
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-26 02:16:33 +09:00
c9facb746b
hst/config: remove data field, rename dir to home
...
There is no reason to give the home directory special treatment, as this behaviour can be quite confusing. The home directory also does not necessarily require its own mount point, it could be provided by a parent or simply be ephemeral.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-26 00:56:10 +09:00
878b66022e
hst/fsbind: optional ensure source
...
This exposes the BindEnsure flag of BindMountOp.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-26 00:50:23 +09:00
2e0a4795f6
container/initbind: optional ensure host directory
...
This is used for ensuring persistent data directories specific to the container.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-26 00:44:45 +09:00
c328b584c0
hst/fslink: improve string representation
...
This shortens the representation of most common use cases and generally improves readability.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-25 22:52:48 +09:00
9585b35d5b
hst/config: remove symlink field
...
Closes #6 .
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-25 22:23:54 +09:00
26cafe3e80
hst/fs: implement link fstype
...
Symlinks do not require special treatment, and doing this allows placing links in order.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-25 21:57:38 +09:00
125f150784
hst/fs: update doc comments
...
The Type method no longer exists on the interface. Update doc comments to reflect that.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-25 21:11:39 +09:00
0dcac55a0c
hst/config: remove container etc field
...
This no longer needs special treatment since it can be specified as a generic filesystem entry.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-25 19:24:33 +09:00
6d202d73b4
hst/fsbind: optional autoetc behaviour
...
This generalises the special field allowing any special behaviour to be matched from target.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-25 18:38:19 +09:00
1438096339
hst/config: handle filesystem entry targeting root
...
This allows any fstype supported by hst to be directly mounted on sysroot. A special case in internal/app applies the matching entry early and excludes it from path hiding.
Closes #5 .
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-25 17:52:57 +09:00
059164d4fa
hst/fsbind: optional autoroot behaviour
...
This allows autoroot to be configured via Filesystem.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-25 17:44:12 +09:00
8db906ee64
container/dispatcher: remove exit stub test log
...
Turns out testing.T does not like being called in defer.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-25 17:33:35 +09:00
cedfceded5
container/autoroot: remove prefix field
...
This field has been a noop for a long time. Remove it to prevent further confusion.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-25 03:39:20 +09:00
33d2dcce1b
container/initoverlay: internal bypass sysroot prefix
...
This is for supporting overlay mounts for autoroot.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-25 02:42:22 +09:00
2baa2d7063
container/init: measure init behaviour
...
This used to be entirely done via integration tests, with almost no hope of error injection and coverage profile. These tests significantly increase confidence of future work in this area.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-24 04:52:32 +09:00
0166833431
container/dispatcher: start goroutine in dispatcher
...
This allows instrumentation of calls from goroutine without relying on finalizers.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-23 21:58:40 +09:00
b3da3da525
container/init: avoid multiple lastcap calls
...
This reduces the size of []kexpect in the test suite.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-23 11:09:11 +09:00
1b3902df78
container/dispatcher: instrument each goroutine individually
...
Scheduler nondeterminism cannot be accounted for, so do this instead.
There should not be any performance penalty as these calls are optimised out for direct.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-23 11:07:16 +09:00
ea1e3ebae9
container/params: pass fd instead of file
...
The file is very difficult to stub. Pass fd instead as it is the value that is actually useful.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-23 00:16:46 +09:00
1c692bfb79
container/init: call lockOSThread through dispatcher
...
This degrades test performance if not stubbed out.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-22 22:24:14 +09:00
141a18999f
container: move integration test helpers
...
With the new instrumentation it is now possible to run init code outside integration tests.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-22 22:07:19 +09:00
afe23600d2
container/path: use syscall dispatcher
...
This allows path and mount functions to be instrumented.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-22 22:02:21 +09:00
09d2844981
container/init: wrap syscall helper functions
...
This allows tests to stub all kernel behaviour, enabling measurement of all function call arguments and error injection.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-22 19:27:31 +09:00
d500d6e559
system/dbus: share host net ns for abstract
...
Host abstract unix sockets are only accessible when also in the init net ns.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-21 21:55:23 +09:00
5b73316ae0
container/syscall: doc comments from manpages
...
These are pulled straight from the manpages.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-21 00:33:46 +09:00
5d8a2199b6
container/init: op interface valid method
...
Check ops early and eliminate duplicate checks.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-21 00:18:50 +09:00
a1482ecdd0
container/inittmpfs: check path equivalence by value
...
Fixes regression introduced while integrating Absolute.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 20:17:28 +09:00
a07f9ed84c
container/initsymlink: check path equivalence by value
...
Fixes regression introduced while integrating Absolute.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 20:03:02 +09:00
51304b03af
container/initremount: check path equivalence by value
...
Fixes regression introduced while integrating Absolute.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 19:55:51 +09:00
c6397b941f
container/initproc: check path equivalence by value
...
Fixes regression introduced while integrating Absolute.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 19:29:45 +09:00
d65e5f817a
container/initplace: check path equivalence by value
...
Fixes regression introduced while integrating Absolute.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 19:19:27 +09:00
696e593898
container/initoverlay: check path equivalence by value
...
Fixes regression introduced while integrating Absolute.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 17:33:15 +09:00
97ab24feef
container/init: use absolute compare method
...
More checks are also added.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 17:14:36 +09:00
31f0dd36df
absolute: efficient equivalence check method
...
This is more efficient and makes the call site cleaner.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 17:06:38 +09:00
9aec2f46fe
container/initdev: check path equivalence by value
...
Fixes regression introduced while integrating Absolute.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 02:55:45 +09:00
022cc26b2e
container/capability: check CAP_TO_INDEX and CAP_TO_MASK
...
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 02:45:00 +09:00
b4c018da8f
container/autoetc: do not bypass absolute check
...
This can now be done cleanly via path function wrappers.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 02:37:11 +09:00
66f52407d3
container/initmkdir: check path equivalence by value
...
Fixes regression introduced while integrating Absolute.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 02:32:22 +09:00
e463faf649
container/initbind: check path equivalence by value
...
Same problem as autoroot, never updated the checks after integrating Absolute.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 02:22:04 +09:00
375acb476d
container/autoroot: check host path equivalence by value
...
This will never return true otherwise unless the equivalent paths happen to be interned by the caller.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 02:14:39 +09:00
c81c9a9d75
container/init: split setup ops into individual files
...
This significantly increases readability.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 01:28:31 +09:00
339e4080dc
container/ops: move Op type to init file
...
This helps with the eventual separation of all setup ops into individual files.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 01:11:24 +09:00
e0533aaa68
container/autoroot: filter dentry with empty name
...
This is unreachable, but nice to have just in case.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 01:03:49 +09:00
13c7083bc0
container: ptrace protection via Yama LSM
...
This is only a nice to have feature as the init process has no additional privileges and the monitor process was never reachable anyway.
Closes #4 .
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-20 00:43:55 +09:00
6947ff04e0
system/dbus/proc: host abstract only when not binding
...
The test failure seems to be caused by an unrelated bug in xdg-dbus-proxy.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-19 23:39:14 +09:00
140fe21237
container/params: check setup/receive behaviour
...
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-18 22:30:34 +09:00
f52d2c7db6
container/path: check create and mountinfo helpers
...
These can quite easily be checked within the framework. The scanner fault injection might require updating at some point if the implementation changes.
Signed-off-by: Ophestra <cat@gensokyo.uk >
2025-08-18 21:30:28 +09:00
