summaryrefslogtreecommitdiff
path: root/daemon/gdm-session.c
Commit message (Collapse)AuthorAgeFilesLines
* daemon: Allow Wayland sessions on non-primary seatsmsizanoen12023-05-081-4/+1
|
* gdm-session: Plug memory leaksAlessandro Bono2023-04-281-1/+1
| | | | | The function is supposed to return const char *. Every caller is not freeing the returned string. Don't duplicate it.
* gdm-session: Add missing guards in public functionsAlessandro Bono2022-10-291-7/+40
|
* daemon: Infer session type from desktop file if user has no saved session typeinfer-session-type-when-neededRay Strode2021-09-291-0/+2
| | | | | | | | | | The accountsservice user cache file can specify a session type associated with the saved session. This is optional though. If one isn't specified GDM needs to figure out the session type based on the list of preferred session types for the system and the session file itself. It was failing to do the latter, though. This commit fixes that.
* daemon: Don't update session type if no saved sessionRay Strode2021-09-061-10/+13
| | | | | | | | | At the moment we always set the session type when the session name is read. But users don't always have a session type specified. If they don't, then don't set the session type. https://gitlab.gnome.org/GNOME/gdm/-/merge_requests/153
* session: Fix gdm_session_is_wayland_sessionRay Strode2021-07-281-8/+6
| | | | | | | | | | gdm_session_is_wayland_session has two problems: 1. it's leaking the full patch returned 2. It's not honoring the supported-session-types preference ordered list when determining which session to pick. This commit fixes both issues.
* session: Fix operators when computing session dirsRay Strode2021-07-281-2/+3
| | | | | | | The loop that computes the list of session directories to check used the wrong operator, so it was computing the list incorrectly. This commit fixes that.
* session: support new accountsservice Session and SessionType propsRay Strode2021-07-221-23/+62
| | | | | | | | | | At the moment the user's session is stored in a property called "XSession". This is pretty weird if the user is using wayland. AccountService now supports a more generic property "Session" and a related "SessionType" property to replace "XSession". This commit switches GDM over to use the new properties.
* daemon: Provide more flexibility for configuring display serverRay Strode2021-07-221-51/+58
| | | | | | | | | | | | | | | | | | | There's currently a way to disable wayland, but no way to disable Xorg. We currently prefer wayland if it's not disabled, but have no way to prefer Xorg without disabling wayland entirely. There's currently no way use legacy Xorg support at all if user display server support is enabled at a build time. This commit adds more flexibility to display server selection. It adds two new keys: XorgEnable and and PreferredDisplayServer. XorgEnable=false disables Xorg support entirely on seat 0. PreferredDisplayServer can be set to "wayland", "xorg", "legacy-xorg" or "none" to select which display server is used by default. If it's set to "wayland", it will fall back to "xorg". If it's set to "xorg" it will fall back to "wayland".
* session: Emit session-opened-failed on session failuresMarco Trevisan (Treviño)2021-07-091-1/+13
| | | | | | | | We used the same signal (session-start-failed) when we had both session starting and opening failures. This may lead to wrong behaviors, so let's introduce another signal to notify such state.
* session: Only emit verification-complete on reauth or after session is openedMarco Trevisan (Treviño)2021-07-091-11/+11
| | | | | | | | | | | | | | | | | | | We're currently emitting verification-complete signal as soon as we got a positive authentication callback from the worker, however at this point we've not opened the session yet, and if that fails we'll end up in a scenario where gdm assumed that we're logging in, and so its greeter is shut down and never restarted on its VT until we manually switch back and forth from it. Meanwhile the session opening failure is never exposed to the user. So, do not emit verification-completed on logins, until we've opened the session. However, re-order the session-opened callback events so that we keep notifying verification-complete before than session-opened. This basically reverts commit c4292870 (again). Fixes: https://gitlab.gnome.org/GNOME/gdm/-/issues/712
* session: Initialize DBus error domain before resolving errorsBenjamin Berg2021-03-151-0/+3
| | | | | | | | | | | We would not initialize the DBus error domain before we retrieved the first error, but only did so to compare the error after receiving them. This means that the first error we received will not be resolved correctly, while all subsequent ones are resolved. Fix this by calling GDM_SESSION_WORKER_ERROR from gdm_session_class_init and add gdm_dbus_error_ensure to make sure this can never be optimized away.
* session: Threat PAM max retries error as service unavailableMarco Trevisan (Treviño)2021-02-161-1/+4
| | | | | | | | | | | | | | | | When a service has returned PAM_MAXTRIES error code we should both emit the error message and also ensure that this service isn't marked as available again for the current session, to prevent us to restart the authentication of it. An example of this can be the PAM fprintd module that will return us a max retries error when the user tried to use his finger more times than configured. In such scenario we want to both prompt the error to the user and prevent that this module is used again if the user can still use other authentication methods. So add an error specialization for PAM_MAXTRIES and threat it as service unavailable error when emitting it.
* session: Don't leak remote greeter interfaceRay Strode2020-11-021-0/+1
| | | | | | | | | | | | XDMCP login screens get a "Remote Geeter Interface" exported over the bus connection (so the login window can provide a Disconnect button). This interface is getting leaked when the session object is disposed, leaving the bus connection itself undisposed, which causes an fd leak. This commit plugs the interface leak, and thus the fd leak.
* session: Always use separate session bus for greeter sessionsBenjamin Berg2020-04-221-8/+18
| | | | | | | | | | | | | | | This is a workaround for the fact that we currently need to run multiple greeter sessions in multi-seat environments that use the same user. We should not be doing this in the first place. Doing this effectively prevents GNOME from using a systemd startup, which would cause relevant processes to be outside of the session scope preventing lookups of the logind session from the PID. Instead, we really should be running each of the greeter session as a separate (dynamic) user. But lacking that, this workaround should get multi-seat support up and running again for the time being. See: #526
* gdm-sessions: Remove unnecessary cast to GDestroyNotifyRobert Mader2019-08-131-1/+1
|
* Allow sessions to register with GDMIain Lane2019-05-291-4/+43
| | | | | | | | | | | | | | | | | | | | | | Recording when sessions start, for Wayland → Xorg fallback or transitioning to the user session, is currently done with timeouts. This isn't ideal, because on some very slow machines the timeout can be hit before the session has had a chance to fail: if gnome-session takes more than 3 seconds to fail then the session will be considered to have exited rather than failed, and so we don't do Xorg fallback. We can do this more reliably if we allow sessions to optionally register themselves with GDM. Then we will know when they've started, so can shut down the greeter or fall back to Xorg as appropriate. The mechanism is that they specify X-GDM-SessionRegisters=true in their file, and then call RegsterSession on the DisplayManager interface on the bus (added in the previous commit) to say that they've started up. If X-GDM-SessionRegisters is missing or false, GDM will call the same method for them after 10 seconds. Closes: #483
* session: Don't allow greeter operations on an running sessionIain Lane2019-02-071-0/+75
| | | | | | | If a client has a reference to a session that starts running, refuse to allow further operations on the session. CVE-2019-3825
* manager,session: Add some debugging around starting reauthenticationIain Lane2019-02-061-4/+21
| | | | | | | | There's a bug right now dealing with timed login and reauthentication, but it's not clear what's going on by looking at the logs. This commit sprinkles some more logging throughout the code, to make the bug easier to track.
* Use G_PARAM_STATIC_STRIGS on propertiesNiels De Graef2019-01-071-13/+13
| | | | This prevents strings from being unnecessarily copied.
* session: use G_DECLARE_FINAL_TYPENiels De Graef2018-12-251-299/+293
|
* session: make sure conversation is closed before its freedRay Strode2018-08-281-0/+4
| | | | | | | | | | Right now if the session-worker crashes, we fail to close the worker's dbus connection. This commit addresses the problem by making sure the conversation is explicitly closed at the time it's freed. Closes https://gitlab.gnome.org/GNOME/gdm/issues/388
* Search sessions in XDG_DATA_DIRSJan Tojnar2018-08-271-0/+17
|
* Revert "session: don't call gdm_session_defaults_changed from setup"Ray Strode2018-02-191-0/+2
| | | | | | | | | | This reverts commit 572a19324b75cc1f1b2db4908e2d7c9f06e4e335. It turns out we need this call for more than just the session type, we also need to it to inform the greeter about the default session to use. https://bugzilla.gnome.org/show_bug.cgi?id=793609
* session: don't call gdm_session_defaults_changed from setupRay Strode2017-10-311-2/+0
| | | | | | | It's only needed to update the session type, which we do earlier in the function anyway. https://bugzilla.gnome.org/show_bug.cgi?id=789722
* daemon: add ChoiceList PAM extensionRay Strode2017-10-201-0/+92
| | | | | | | | | | | | | | | | This commit adds one PAM extension, a "Choice List" using the new PAM_BINARY_PROMPT protocol added in the previous commit. The PAM module sends a list of (key, row text) pairs, and GDM ferries the request to gnome-shell using a new user verifier sub-interface. gnome-shell should present the list to the user and pass back the corresponding key, which GDM ferries back to the PAM module. Note this commit is only the daemon side. A subsequent commit will add the libgdm API needed for gnome-shell to actually deal with this new PAM extension. https://bugzilla.gnome.org/show_bug.cgi?id=788851
* daemon: introduce pam extension mechanismRay Strode2017-10-201-2/+47
| | | | | | | | | | | | | | | | | This abuses PAM_BINARY_PROMPT for our own nefarious purposes. The way it works is GDM advertises what "extensions" it supports with the environment variable, GDM_SUPPORTED_PAM_EXTENSIONS (a space separated list of reverse dns notation names). PAM services that support this protocol, will read the environment variable, and check for extension strings they support. They then know that sending PAM_BINARY_PROMPT won't blow up, and know what format to use for the binary data. The type field of the structure is the index of the string from the environment variable. This commit is just foundation work. It doesn't actually add any extensions. https://bugzilla.gnome.org/show_bug.cgi?id=788851
* session: add new Initialize methodRay Strode2017-10-201-169/+42
| | | | | | | | | | | | | Right now the worker interface has three methods for starting a conversation: Setup, SetupForUser, SetupForProgram Each of these method calls take a large number of overlapping arguments. Extending these argument lists is painful and breaks upgrades. This commit adds a new, fourth call, Initialize, which supercedes the others and just takes a vardict, which is much more extensible. https://bugzilla.gnome.org/show_bug.cgi?id=788851
* Use standard exit codes.Robert Ancell2017-10-031-1/+1
| | | | | | | | | Use EXIT_ defines for readibility. There were some exit codes > 1, but they don't seem to be checked by any of the parent process code. This does mean that the logs might have changed, but modern logging techniques have probably made this obsolete. https://bugzilla.gnome.org/show_bug.cgi?id=788307
* gdm-sessions: force a session bus for non-seat0 sessionRay Strode2017-09-201-1/+6
| | | | | | | | | | | | | | | | Eventually, our software should become "multi-seat aware", where it takes into account multiple seats at a time for the user (even if it's just putting up a dialog saying "user is busy" on all but one seat). We're not there yet. And user bus currently breaks XDMCP (which should really spawn session with its own separate user, but again, we're not there yet). This commit changes GDM to start a session bus for all non-seat0 displays, as a near-term workaround. https://bugzilla.gnome.org/show_bug.cgi?id=787943
* gdm-session: always pass down user saved languageMario Sanchez Prada2017-09-181-2/+1
| | | | | | | | | | | | | | | | There's a bug when changing the language of the login screen to a different one than the user session's language, which causes the new language to be used both in the login screen AND the user session if you simply logout and then re-login without rebooting. Restarting the machine after that point normalizes things, getting the new language used only in the login screen but not in the user session. Making sure that the user language is always saved by the session seems without checking whether is the same than the default language fixes this. Based on a patch by Felipe Erias Morandeira <femorandeira@igalia.com> https://bugzilla.gnome.org/show_bug.cgi?id=786422
* gdm-session: Calculate fallback session name without ".desktop"Iain Lane2017-09-121-12/+5
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | When the user has not ever selected a session from the selector, we will log them into the `fallback' session, which is calculated by looking at all installed sessions and picking the first one in the list. There is a bug that the visually selected session presented in the greeter can be different from the fallback session calculated inside GDM. This happens because the two sort sessions differently. Inside GDM we calculate the fallback session by comparing the full basename of the file: >>> GLib.strcmp0("gnome-xorg.desktop", "gnome.desktop") -1 Whereas gnome-shell calls gdm_get_session_ids(), which returns IDs without the ".desktop" extension. Those are then sorted: >>> GLib.strcmp0("gnome-xorg", "gnome") 45 which gives a different ordering. Fix this bug by calculating the fallback session in the same way, by removing ".desktop" from the filename. https://bugs.launchpad.net/ubuntu/+source/gnome-session/+bug/1705157 https://bugzilla.gnome.org/show_bug.cgi?id=787304
* session: ensure wayland paths get searched even with ↵Daniel van Vugt2017-07-171-1/+0
| | | | | | | | | | | | | | | | | | | | --disable-user-display-server get_system_session_dirs is function that says which directories to search for session desktop files. The returned directories may be xsession dirs, wayland-session dirs, or both, depending on configuration. A bug in this function means an unused NULL sentinal value from the xsession list of directories is getting copied into the middle of the output list, when --disable-user-display-server is passed to configure. This commit corrects the bug, by dropping the unused sentinal value. Launchpad-Bug: 1704050 https://bugzilla.gnome.org/show_bug.cgi?id=784891
* session: update session type when session is readRay Strode2017-05-041-7/+9
| | | | | | | | | | | | | | | We currently update the session type (to either wayland or x11) when we first start the PAM conversation and later when the username is set (if the user is not set at the same time as the PAM conversation). There's a race that means the session won't necessarly have been read from accountsservice at these points. This commit changes the code to instead update the session type in result to the session actually getting read. https://bugzilla.gnome.org/show_bug.cgi?id=781825
* session: update session search dirs when ignore-wayland changesRay Strode2017-03-251-13/+21
| | | | | | | | | | | Right now the first time the session search dirs are requested, they are computed and cached. That leaves them stale if the ignore-wayland flag changes, which breaks fallback mode in some cases. This commit fix the problem by always computing the session search dirs in full, with no caching. https://bugzilla.gnome.org/show_bug.cgi?id=780518
* session: honor ignore-wayland flagRay Strode2017-03-031-2/+4
| | | | | | | | | | | commit 5b1e625417c274823c66a552f37d912986fc34d8 changed how the default session path was computed. It inadvertently dropped support for the ignore-wayland flag, breaking fall back to X in scenarios where wayland doesn't work. This commit fixes the code to again honor ignore-wayland. https://bugzilla.gnome.org/show_bug.cgi?id=779338
* daemon: add knob to disable starting X server as userRay Strode2017-02-271-0/+17
| | | | | | | | | | | | Some deployments need to be able to turn off running X servers as a user for backward compatibility. This commit adds some #ifdef goop to that end. Note wayland requires running as the user, so this option, merely deprioritizes wayland sessions under X sessions. https://bugzilla.gnome.org/show_bug.cgi?id=779338
* gdm-session: change how session search path is computedRay Strode2017-02-271-8/+13
| | | | | | | | The current mechanism of assuming the wayland-sessions path is the first element of the array isn't very conducive to reprioritizing wayland sessions. This commit changes the code to use a GArray instead. https://bugzilla.gnome.org/show_bug.cgi?id=779338
* gdm-session: update session type when the session defaults changeRay Strode2016-02-091-2/+2
| | | | | | | | | | | | If a user has never logged in before and they login using autologin, then we don't select the correct session type if they end up with a wayland session. This is because we fail to initialize the session type based on the default session type. This commit changes the code to update the session type anytime the session defaults are reset. https://bugzilla.gnome.org/show_bug.cgi?id=761785
* session: free conversation hash when disposing sessionRay Strode2016-01-221-0/+3
| | | | | | | We're currently leaking the hash table when disposing the session, this commit fixes that. https://bugzilla.gnome.org/show_bug.cgi?id=754814
* session: keep session object alive while establishing credentialsRay Strode2016-01-221-2/+5
| | | | | | | | | | | The only reference to session objects gets cleaned up when verification-complete is emitted, which happens in the middle of the establish_credentials handler. This commit makes sure the session object stays alive until the handler completes to prevent a crash. https://bugzilla.gnome.org/show_bug.cgi?id=754814
* gdm-session: emit verification-complete even for loginsRay Strode2016-01-211-8/+7
| | | | | | | | | | Right now we only emit verification-complete when the a user successfully reauthenticates. We should also do it when they successfully initially authenticate. This commit fixes that. https://bugzilla.gnome.org/show_bug.cgi?id=754814
* session: make sure greeter is GNOME session even if gnome isn't installedRay Strode2016-01-211-9/+15
| | | | | | | | | | | | | | | We figure out the XDG_CURRENT_DESKTOP by looking at the session that's use for login. For the greeter, we don't use a session, at all, but instead a directly specified program. Still, we're determining the XDG_CURRENT_DESKTOP for the greeter based on what default session gets picked. On KDE only setups, the greeter will get marked as a KDE session which will prevent gnome-settings-daemon from starting properly. This commit changes the code so that greeter sessions, don't rely on the default session desktop file. https://bugzilla.gnome.org/show_bug.cgi?id=760943
* session: Crash better when no sessions are installedMichael Catanzaro2016-01-121-0/+4
| | | | | | | | If no session desktop file can be found in the search path, we are doomed. Crash nicely with g_error rather than continuing on to call g_sequence_get with an invalid GSequenceIter. https://bugzilla.gnome.org/show_bug.cgi?id=760548
* session: explain rationale between display modes and simplify getterRay Strode2016-01-121-19/+20
| | | | | | | | | | | | | gdm_session_get_display_mode is longer and more confusing than it needs to be. This commit, consolidates code paths together and adds a more descriptive comment explaining why NEW_VT display mode is used for everything but non-seat0 seats. Based on a patch and bug report by Michael Catanzaro <mcatanzaro@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=760415
* session: Cancel worker proxy async ops when freeing conversationsRui Matos2015-11-151-13/+23
| | | | | | | | We need to cancel ongoing async ops for worker proxies when freeing conversations or we'll crash when the completion handler runs and we access free'd memory. https://bugzilla.gnome.org/show_bug.cgi?id=758032
* session: disconnect signals from worker proxy when conversation is freedRay Strode2015-11-151-1/+22
| | | | | | | We don't want an outstanding reference on the worker proxy to lead to signal handlers getting dispatched after the conversation is freed. https://bugzilla.gnome.org/show_bug.cgi?id=758032
* session: always update session type when reading saved sessionRay Strode2015-11-091-2/+2
| | | | | | | | We previously only did it if the session name was different, but now it's possible to have sessions of both types with the same name. https://bugzilla.gnome.org/show_bug.cgi?id=757715
* session: give wayland sessions priorityRay Strode2015-11-091-7/+6
| | | | | | | If two sessions have the same name, give the wayland one priority, unless wayland sessions are being ignored. https://bugzilla.gnome.org/show_bug.cgi?id=757715
* session: add property for ignoring wayland sessionsRay Strode2015-11-061-1/+43
| | | | | | | | | | | We're going to want to be able to exclude wayland sessions from the search path if the manager deems it necessary. This commit adds a property to GdmSession to give the manager that ability. https://bugzilla.gnome.org/show_bug.cgi?id=757715
View Lorry Controller Status