pve-access-control.git
2 weeks agotfa: pass whole webauthn config to 'set_webauthn_config' master
Wolfgang Bumiller [Mon, 25 Jul 2022 11:52:56 +0000 (13:52 +0200)]
tfa: pass whole webauthn config to 'set_webauthn_config'

the field names are being kept compatible

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
3 weeks agoapi: realm sync: avoid separate log line for "remove-vanished" opt
Thomas Lamprecht [Tue, 19 Jul 2022 10:07:58 +0000 (12:07 +0200)]
api: realm sync: avoid separate log line for "remove-vanished" opt

can be a bit confusing and not really necessary, just inline it with
the "syncing user/group" message

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
4 weeks agobump version to 7.2-4
Thomas Lamprecht [Thu, 14 Jul 2022 06:36:58 +0000 (08:36 +0200)]
bump version to 7.2-4

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
4 weeks agoauth key: fix double rotation in clusters
Fabian Grünbichler [Wed, 13 Jul 2022 13:13:59 +0000 (15:13 +0200)]
auth key: fix double rotation in clusters

there is a (hard to trigger) race that can cause a double rotation of
the auth key, with potentially confusing fallout (various processes on
different nodes having an inconsistent view of the current and previous
auth keys, resulting in "random" invalid ticket errors until the next
proper key rotation 24h later).

the underlying cause is that `stat()` calls are excempt from our
otherwise non-cached/direct_io handling of pmxcfs I/O, which allows the
following sequence of events to take place:

LAST: mtime of current auth key

- current epoch advances to LAST + 24h

the following can be arbitrarly interleaved between node A and B:
- LAST+24h node A: pvedaemon/pvestatd on node A calls check_authkey(1)
- LAST+24h node A: it returns 0 (rotation required)
- LAST+24h node A: rotate_key() is called
- LAST+24h node A: cfs_lock_authkey is called
- LAST+24h node B: pvedaemon/pvestatd calls check_authkey(1)
- LAST+24h node B: key is not yet cached in-memory by current process
- LAST+24h node B: key file is opened, stat-ed, read, parsed, and content+mtime
  is cached (the kernel will now cache this stat result for 1s unless
  the path is opened)
- LAST+24h node B: it returns 0 (rotation required)
- LAST+24h node B: rotate_key() is called
- LAST+24h node B: cfs_lock_authkey is called

the following is mutex-ed via a cfs_lock:
- LAST+24h node A: lock is obtained
- LAST+24h node A: check_authkey() is called
- LAST+24h node A: key is stat-ed, mtime is still (correctly) LAST,
  cached mtime and content are returned
- LAST+24h node A: it returns 0 (rotation still required)
- LAST+24h node A: get_pubkey() is called and returns current auth key
- LAST+24h node A: new keypair is generated and persisted
- LAST+24h node A: cfs_lock is released
- LAST+24h node B: changes by node A are processed by pmxcfs
- LAST+24h node B: lock is obtained
- LAST+24h node B: check_authkey() is called
- LAST+24h node B: key is stat-ed, mtime is (incorrectly!) still LAST
  since the stat call is handled by the kernel/page cache, not by
  pmxcfs, cached mtime and content are returned
- LAST+24h node B: it returns 0 (rotation still required)
- LAST+24h node B: get_pubkey() is called and returns either previous or
  key written by node A (depending on whether page cache or pmxcfs
  answers stat call)
- LAST+24h node B: new keypair is generated, key returned by last
  get_pubkey call is written as old key

the end result is that some nodes and process will treat the key
generated by node A as "current", while others will treat the one
generated by nodoe B as "current". both have the same mtime, so the
in-memory cache hash won't be updated unless the service is restarted or
another rotation happens. depending on who generated the ticket and who
attempts validating it, a ticket might be rejected as invalid even
though the generating party would treat it as valid, and time on all
nodes is properly synced.

there seems to be now way for pmxcfs to pro-actively invalidate the page
cache entry safely (since we'd need to do so while writes to the same
path can happen concurrently), so work around by forcing an open/close
at the (stat) call site which does the work for us. regular reads are
not affected since those already bypass the page cache entirely anyway.

thankfully in almost all cases, the following sequence has enough
synchronization overhead baked in to avoid triggering the issue almost
entirely:

- cfs_lock
- generate key
- create tmp file for old key
- write tmp file
- rename tmp file into proper place
- create tmp file for new pub key
- write tmp file
- rename tmp file into proper place
- create tmp file for new priv key
- write tmp file
- rename tmp file into proper place
- release lock

that being said, there has been at least one report where this was
triggered in the wild from time to time.

it is easy to reproduce by increasing the attr_timeout and entry_timeout
fuse settings inside pmxcfs to increase the time stat results are
treated as valid/retained in the page cache:

-----8<-----
 diff --git a/data/src/pmxcfs.c b/data/src/pmxcfs.c
 index d78a248..e3e807b 100644
 --- a/data/src/pmxcfs.c
 +++ b/data/src/pmxcfs.c
 @@ -935,7 +935,7 @@ int main(int argc, char *argv[])

   mkdir(CFSDIR, 0755);

 - char *fa[] = { "-f", "-odefault_permissions", "-oallow_other", NULL};
 + char *fa[] = { "-f", "-odefault_permissions", "-oallow_other", "-oentry_timeout=5", "-oattr_timeout=5", NULL};

   struct fuse_args fuse_args = FUSE_ARGS_INIT(sizeof (fa)/sizeof(gpointer) - 1, fa);

----->8-----

in which case it's even easy to trigger more than double rotation in a
bigger test cluster (stopping all PVE services except for pve-cluster
helps avoiding interference):

on a single node:
$ touch --date yesterday /etc/pve/authkey.pub

in parallel (i.e., via tmux synchronized panes):
-----8<-----
use strict;
use warnings;
use PVE::Cluster;
use PVE::AccessControl;
PVE::Cluster::cfs_update();

# ensure page cache entry is there
PVE::AccessControl::check_authkey(1);
PVE::AccessControl::check_authkey(1);
# now attempt rotation
PVE::AccessControl::rotate_authkey();
----->8-----

Thanks to Wolfgang Bumiller for assistance in triaging and exploring
various avenues of fixing.

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
7 weeks agofix #4074: increase API OpenID code size limit to 2048
Mira Limbeck [Wed, 15 Jun 2022 14:09:50 +0000 (16:09 +0200)]
fix #4074: increase API OpenID code size limit to 2048

Azure AD seems to have a variable authorization code size, depending on
the browser state according to one report in bug #4074 [0].

Sometimes the size is greater than our current limit of 1024, so
increase it to 2048.

The RFC [1] mentions that there is no limit to the code size, but based on
current experience, a size limit of 2048 might be enough for every
current OpenID Connect provider.

[0] https://bugzilla.proxmox.com/show_bug.cgi?id=4074
[1] https://datatracker.ietf.org/doc/html/rfc6749#section-4.1.2

Signed-off-by: Mira Limbeck <m.limbeck@proxmox.com>
[w.bumiller@proxmox.com: bump to 4096]
Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
7 weeks agobump version to 7.2-3
Thomas Lamprecht [Mon, 20 Jun 2022 13:51:19 +0000 (15:51 +0200)]
bump version to 7.2-3

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
7 weeks agoperm check: forbid undefined/empty ACL path
Fabian Grünbichler [Mon, 20 Jun 2022 11:05:12 +0000 (13:05 +0200)]
perm check: forbid undefined/empty ACL path

to detect similar issues to that fixed in the previous commit early on
and without the potential for dangerous side-effects.

root@pam is intentionally still allowed before the check in case such
situations can be triggered by misconfiguration - root@pam can then
still clean up the affected configs via the GUI/API, and not just via
manual editing.

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
7 weeks agoapi2: token: use userid-group as API perm check
Fabian Grünbichler [Mon, 20 Jun 2022 11:05:11 +0000 (13:05 +0200)]
api2: token: use userid-group as API perm check

the previous version using an ACL path of '/access/users/{userid}' was
broken for non-root users, since the '@' character always contained in a
userid is not allowed in ACL paths.

this effectively meant that creating API tokens only worked for:
- root@pam (ACL checks skipped altogether)
- users with User.Modify on '/' with propagation (the roles/privs for
  '/' are propagated to the undefined path in this case)
- users creating their own tokens (first branch of 'or')

the userid-group check is used for all other modifications of user
entities, so it can also be used for creating/modifying/removing API
tokens.

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
2 months agobump version to 7.2-2
Thomas Lamprecht [Fri, 3 Jun 2022 12:02:35 +0000 (14:02 +0200)]
bump version to 7.2-2

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
2 months agoREADME: break long lines
Thomas Lamprecht [Fri, 3 Jun 2022 11:58:46 +0000 (13:58 +0200)]
README: break long lines

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
2 months agotree wide: typo fixes
Thomas Lamprecht [Fri, 3 Jun 2022 11:58:07 +0000 (13:58 +0200)]
tree wide: typo fixes

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
2 months agopermissions: add some more comments
Fabian Grünbichler [Fri, 3 Jun 2022 11:50:49 +0000 (13:50 +0200)]
permissions: add some more comments

for future reference/changes.

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
2 months agopermissions: fix token/user priv intersection
Fabian Grünbichler [Fri, 3 Jun 2022 11:50:48 +0000 (13:50 +0200)]
permissions: fix token/user priv intersection

the token/user priv intersection could only honored user privs that had
the propagation flag set, reducing the scope of the token more than
intended.

the pre-existing test case actually triggered the broken behaviour, but
the expected value matched it so it was not noticed.

Fixes: e8a0cee47bb477162f291e67144ea0c0df47f2ee "rpcenv: improve user/token intersection"
Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
2 months agopermissions: properly merge propagation flag
Fabian Grünbichler [Fri, 3 Jun 2022 11:50:47 +0000 (13:50 +0200)]
permissions: properly merge propagation flag

when multiple roles are defined on a path that share a privilege, this
randomly took the propagation flag for the priv from the last role
encountered. since perl hashes are iterated randomly, this means the
propagation flag was sometimes set correctly, and sometimes not.

note that this propagation flag is only used for display/dumping
purposes, and for intersection with token privs (see next commit).
actual handling of propagation happens on the role level in
PVE::AccessControl::roles().

modified test case (spuriously) fails without the fix.

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
2 months agobump version to 7.2-1
Thomas Lamprecht [Tue, 31 May 2022 11:43:47 +0000 (13:43 +0200)]
bump version to 7.2-1

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
2 months agoaccess check: include user/token id in expired exception
Thomas Lamprecht [Tue, 31 May 2022 11:32:36 +0000 (13:32 +0200)]
access check: include user/token id in expired exception

not that relevant for the user as the daemon auth log already
contains that info, but for token it can be nice.

The API response is always just a plain "401 auth failure" in any
case (expired or wrong creds)

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
2 months agocheck user: re-order enable check first again and adhere noerr flag
Thomas Lamprecht [Tue, 31 May 2022 11:31:23 +0000 (13:31 +0200)]
check user: re-order enable check first again and adhere noerr flag

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
2 months agouser check: fix expiration/enable order
Oguz Bektas [Tue, 31 May 2022 10:58:52 +0000 (12:58 +0200)]
user check: fix expiration/enable order

Signed-off-by: Oguz Bektas <o.bektas@proxmox.com>
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
3 months agobump version to 7.1-8
Thomas Lamprecht [Thu, 28 Apr 2022 15:02:54 +0000 (17:02 +0200)]
bump version to 7.1-8

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
3 months agoadd realm-sync regression test for new 'remove-vanished'
Dominik Csapak [Mon, 28 Mar 2022 12:38:05 +0000 (14:38 +0200)]
add realm-sync regression test for new 'remove-vanished'

by having a test case that does not delete properties, but acls and
entries

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
3 months agoconvert regression tests to new 'remove-vanished' parameter
Dominik Csapak [Mon, 28 Mar 2022 12:38:04 +0000 (14:38 +0200)]
convert regression tests to new 'remove-vanished' parameter

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
3 months agofix #3668: realm-sync: replace 'full' & 'purge' with 'remove-vanished'
Dominik Csapak [Mon, 28 Mar 2022 12:38:03 +0000 (14:38 +0200)]
fix #3668: realm-sync: replace 'full' & 'purge' with 'remove-vanished'

Both, 'full' & 'purge', configure what is removed when an entry
(or property) is not returned anymore from the sync result. For users
it may be thus easier to understand setting a single property without
wondering about the interaction of the two specific ones.

The new 'remove-vanished' is a list of things to remove, namely:
 * 'acl'
 * 'entry'
 * 'properties'.

The old 'full' gets mapped to 'entry;properties' and an old 'purge'
to the 'acl'

Keep the old properties for backwards-compatibility but transform to
the new list on edit. Add a deprecation notice to the description
with a TODO reminder to remove with 8.0

Note, this commit changes two things slightly:
* 'purge' (now remove-vanished -> acl) does now remove ACLs for
  vanished entries even when we keep those entries. Previously those
  were only deleted when we also removed the entries (this can be
  seen by the changed regression test)
* since 'remove-vanished' is empty by default, only the scope must be
  given explicitly on sync or the sync-default. Previously 'full' and
  'purge' must have been configured explicitly (no big deal, since
  the default is to not remove anything)

Bug #3668 is fixed when not selecting 'properties' for the sync, so
that existing (custom) properties are not deleted on update

Suggested-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
[ T: some commit message typos/wording ]
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
3 months agoadd regression tests for realm-sync
Dominik Csapak [Mon, 28 Mar 2022 12:38:02 +0000 (14:38 +0200)]
add regression tests for realm-sync

to fully test the 'end-to-end' sync api call, we have to mock quite
some methods for cluster/rpcenvironment/ldap

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
4 months agorpc env: indentation fixes
Thomas Lamprecht [Wed, 23 Mar 2022 11:24:57 +0000 (12:24 +0100)]
rpc env: indentation fixes

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
4 months agobump version to 7.1-7
Thomas Lamprecht [Mon, 21 Mar 2022 15:15:29 +0000 (16:15 +0100)]
bump version to 7.1-7

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
4 months agoapi: get user: declare token schema
Wolfgang Bumiller [Mon, 21 Mar 2022 14:29:12 +0000 (15:29 +0100)]
api: get user: declare token schema

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
4 months agouserid-group check: distinguish create and update
Fabian Grünbichler [Thu, 17 Mar 2022 08:57:21 +0000 (09:57 +0100)]
userid-group check: distinguish create and update

and check both existing groups and the groups parameter in the update
case. the following user.cfg settings can be used for testing:

user:test@pve:1:0:t::::x:
user:other@pve:1:0:t::::x:
group:test:test@pve::
group:test3:::
role:RealmUserAllocator:Realm.AllocateUser:
role:UserModifier:User.Modify:
acl:0:/access/groups/test:test@pve:UserModifier:
acl:0:/access/groups/test3:@test:UserModifier:
acl:0:/access/realm/pve:test@pve:RealmUserAllocator:

unchanged: the user 'test@pve' can allocate new '@pve' users, but
only if the created user will belong to at least one of 'test'
(direct ACL for that user) or 'test3' (indirect ACL via 'test' group)
groups.

changed: if the user 'test@pve' updates an existing user, they need
to (A) have 'User.Modify' on at least one existing group of that
user, and (B) 'User.Modify' on all of the groups passed in via the
'groups' parameter. A is the general rule for 'allowed to modify
user' across the board, but was missing for this specific variant of
the check. B was the case before, but just checking this without also
checking A allows a user to pull off-limits users into groups that
they can modify, which then in turn allows them to modify those users
via A which is now passing.

for example, without this patch 'test@pve' would be able to add
'other@pve' to either 'test' or 'test3', and then in turn call any of
the API endpoints that require 'User.Modify' on a user's group
(change TFA, change password or delete user if realm is pve, ..).

all the other userid-group checks without group_param set remain
unchanged as well, since $check_existing_user is true in that case.

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
6 months agobump version to 7.1-6
Fabian Grünbichler [Fri, 21 Jan 2022 13:20:58 +0000 (14:20 +0100)]
bump version to 7.1-6

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
6 months agorpcenv->permissions() ensure propagate is always defined
Fabian Grünbichler [Fri, 21 Jan 2022 10:57:30 +0000 (11:57 +0100)]
rpcenv->permissions() ensure propagate is always defined

this shouldn't happen anymore, but a safeguard in case the parser ever
has a bug does not hurt.

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
6 months agorpcenv: skip undef propagation flag in more places
Fabian Grünbichler [Fri, 21 Jan 2022 10:52:08 +0000 (11:52 +0100)]
rpcenv: skip undef propagation flag in more places

these are just cosmetic fixes/safeguards against future bugs -
compute_api_permissions is used to set the 'cap' object to hide parts of
the GUI that are not usable without the corresponding privs in the
backend anyway, and get_effective_permissions is only used to return the
permission tree without a specific path query.

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
6 months agotest: extend intersection tests
Fabian Grünbichler [Fri, 21 Jan 2022 10:23:25 +0000 (11:23 +0100)]
test: extend intersection tests

to check the previous commit's fix.

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
6 months agorpcenv: improve user/token intersection
Fabian Grünbichler [Fri, 21 Jan 2022 10:23:36 +0000 (11:23 +0100)]
rpcenv: improve user/token intersection

this could return undef for the propagation flag instead of 1/0, leading
to confusing displays of permission trees. all the actual checks using
the returned hash check for definedness anyway, so the actual
privileges checked and the displayed ones were not identical.

fixes: 7e8bcaa754432f084e53d9daf13c653a8777d88b
"roles()/permissions(): also return propagate flag"

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
6 months agoget effective permissions: return /sdn path too
Thomas Lamprecht [Fri, 21 Jan 2022 13:19:55 +0000 (14:19 +0100)]
get effective permissions: return /sdn path too

only cosmetic for the user to see what their permission tree looks
like.

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
7 months agofix realm sync permissions
Wolfgang Bumiller [Mon, 20 Dec 2021 10:31:15 +0000 (11:31 +0100)]
fix realm sync permissions

The userid-* permission check variants work on
$param->{userid} directly which does not exist for this
call. Also, they work on the realm of the user being
checked, rather than the realm provided as parameter.

The result was that as non-root user this always failed
with the message "userid '' too short"

Fix this by making the check explicitly work like in the
description.

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
7 months agouse statement fixup
Wolfgang Bumiller [Tue, 14 Dec 2021 08:13:44 +0000 (09:13 +0100)]
use statement fixup

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
8 months agotfa list: account for admin permissions
Wolfgang Bumiller [Mon, 6 Dec 2021 13:33:44 +0000 (14:33 +0100)]
tfa list: account for admin permissions

instead of restricting listing tfa entries of others to
root@pam, perform the same checks the user-list does and
which also reflect the permissions of the api calls actually
operating on those users, so, `User.Modify` on the user (but
also `Sys.Audit`, since it's only a read-operation, just
like the user index API call)

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
8 months agotfa: when modifying others, verify the current user's password
Wolfgang Bumiller [Mon, 6 Dec 2021 13:12:33 +0000 (14:12 +0100)]
tfa: when modifying others, verify the current user's password

this was wrong as it asked for the password of the
to-be-edited user instead, which makes no sense

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
8 months agofix #3768: warn on bad u2f or webauthn settings
Wolfgang Bumiller [Mon, 6 Dec 2021 08:38:01 +0000 (09:38 +0100)]
fix #3768: warn on bad u2f or webauthn settings

but don't bail out of the entire auth process, otherwise
not even totp or recovery keys will work anymore in this
case

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
8 months agofix a 'use of undefined...' warning
Wolfgang Bumiller [Mon, 6 Dec 2021 08:38:00 +0000 (09:38 +0100)]
fix a 'use of undefined...' warning

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
8 months agobump version to 7.1-5
Thomas Lamprecht [Thu, 25 Nov 2021 06:57:45 +0000 (07:57 +0100)]
bump version to 7.1-5

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
8 months agoopenid: fix username-claim fallback
Thomas Lamprecht [Thu, 25 Nov 2021 06:57:10 +0000 (07:57 +0100)]
openid: fix username-claim fallback

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
8 months agobump version to 7.1-4
Wolfgang Bumiller [Mon, 22 Nov 2021 13:04:57 +0000 (14:04 +0100)]
bump version to 7.1-4

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
8 months agofill origin into webauthn config if not provided
Wolfgang Bumiller [Mon, 22 Nov 2021 12:52:57 +0000 (13:52 +0100)]
fill origin into webauthn config if not provided

in order to allow subdomains to work, the wa config should
only specify 'id' and 'rp', the 'origin' gets filled in by
the node

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
8 months agobump version to 7.1-3
Thomas Lamprecht [Fri, 19 Nov 2021 07:12:02 +0000 (08:12 +0100)]
bump version to 7.1-3

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
8 months agod/control: bump versioned dependency to libpve-rs-perl
Thomas Lamprecht [Fri, 19 Nov 2021 08:34:25 +0000 (09:34 +0100)]
d/control: bump versioned dependency to libpve-rs-perl

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
8 months agoopenid: support configuring ACR values
Thomas Lamprecht [Thu, 18 Nov 2021 16:01:04 +0000 (17:01 +0100)]
openid: support configuring ACR values

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
8 months agoopenid: support configuring scopes
Thomas Lamprecht [Thu, 18 Nov 2021 16:00:42 +0000 (17:00 +0100)]
openid: support configuring scopes

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
8 months agoopenid: support configuring the prompt
Thomas Lamprecht [Thu, 18 Nov 2021 13:53:44 +0000 (14:53 +0100)]
openid: support configuring the prompt

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
8 months agoopenid: fix whitespace/indentation
Thomas Lamprecht [Thu, 18 Nov 2021 13:52:52 +0000 (14:52 +0100)]
openid: fix whitespace/indentation

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
8 months agoopenid: allow arbitrary username-claims
Thomas Lamprecht [Thu, 18 Nov 2021 13:24:24 +0000 (14:24 +0100)]
openid: allow arbitrary username-claims

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
8 months agobump version to 7.1-2
Thomas Lamprecht [Wed, 17 Nov 2021 12:45:08 +0000 (13:45 +0100)]
bump version to 7.1-2

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
8 months agod/control: bump versioned dependency to libpve-rs-perl
Thomas Lamprecht [Wed, 17 Nov 2021 12:47:59 +0000 (13:47 +0100)]
d/control: bump versioned dependency to libpve-rs-perl

to ensure we get the incompatible type set for such TFA entries

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
8 months agocatch incompatible tfa entries with a nice error
Wolfgang Bumiller [Wed, 17 Nov 2021 11:34:40 +0000 (12:34 +0100)]
catch incompatible tfa entries with a nice error

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
8 months agobump version to 7.1-1
Thomas Lamprecht [Mon, 15 Nov 2021 14:33:29 +0000 (15:33 +0100)]
bump version to 7.1-1

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
9 months agotfa: fix http 404 in get_tfa_entry
Wolfgang Bumiller [Fri, 12 Nov 2021 09:37:45 +0000 (10:37 +0100)]
tfa: fix http 404 in get_tfa_entry

this produced warnings in the journal and returned code 500
instead

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agobump version to 7.0-7
Thomas Lamprecht [Thu, 11 Nov 2021 17:20:02 +0000 (18:20 +0100)]
bump version to 7.0-7

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
9 months agod/control: break pve-manager (<< 7.0-15)
Thomas Lamprecht [Thu, 11 Nov 2021 17:19:23 +0000 (18:19 +0100)]
d/control: break pve-manager (<< 7.0-15)

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
9 months agoticket: normalize path for verification
Fabian Grünbichler [Thu, 11 Nov 2021 14:07:12 +0000 (15:07 +0100)]
ticket: normalize path for verification

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
9 months agotickets: add tunnel ticket
Fabian Grünbichler [Thu, 11 Nov 2021 14:07:11 +0000 (15:07 +0100)]
tickets: add tunnel ticket

just like VNC ticket, but different prefix to prevent confusion.

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
9 months agotfa: upgrade check: early return if no cluster members
Thomas Lamprecht [Thu, 11 Nov 2021 11:29:58 +0000 (12:29 +0100)]
tfa: upgrade check: early return if no cluster members

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
9 months agotfa: upgrade check: more info in error message(s)
Thomas Lamprecht [Thu, 11 Nov 2021 11:06:42 +0000 (12:06 +0100)]
tfa: upgrade check: more info in error message(s)

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
9 months agotfa: upgrade check: be less strict about version format
Thomas Lamprecht [Thu, 11 Nov 2021 11:05:52 +0000 (12:05 +0100)]
tfa: upgrade check: be less strict about version format

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
9 months agoimplement version checks for tfa
Wolfgang Bumiller [Wed, 10 Nov 2021 14:11:45 +0000 (15:11 +0100)]
implement version checks for tfa

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agomerge old user.cfg keys to tfa config when adding entries
Wolfgang Bumiller [Wed, 10 Nov 2021 12:49:03 +0000 (13:49 +0100)]
merge old user.cfg keys to tfa config when adding entries

this happens when the first new tfa entry is added and the
'keys' entry is replaced by "x"

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
9 months agod/control: add liburi-perl dependency
Wolfgang Bumiller [Wed, 10 Nov 2021 12:49:02 +0000 (13:49 +0100)]
d/control: add liburi-perl dependency

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agocheck enforced realm tfa type in new auth
Wolfgang Bumiller [Wed, 10 Nov 2021 12:49:01 +0000 (13:49 +0100)]
check enforced realm tfa type in new auth

this way we could also add webauthn as a required tfa type
to the realm configs later on

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agoassert tfa/user config lock order
Wolfgang Bumiller [Wed, 10 Nov 2021 12:49:00 +0000 (13:49 +0100)]
assert tfa/user config lock order

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agoset/remove 'x' for tfa keys in user.cfg in new api
Wolfgang Bumiller [Tue, 9 Nov 2021 11:27:05 +0000 (12:27 +0100)]
set/remove 'x' for tfa keys in user.cfg in new api

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agopveum: update tfa delete command
Wolfgang Bumiller [Tue, 9 Nov 2021 11:27:04 +0000 (12:27 +0100)]
pveum: update tfa delete command

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agoupdate tfa cleanup when deleting users
Wolfgang Bumiller [Tue, 9 Nov 2021 11:27:03 +0000 (12:27 +0100)]
update tfa cleanup when deleting users

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agosupport registering yubico otp keys
Wolfgang Bumiller [Tue, 9 Nov 2021 11:27:02 +0000 (12:27 +0100)]
support registering yubico otp keys

In PBS we don't support this, so the current TFA API in rust
does not support this either (although the config does know
about its *existence*).

For now, yubico authentication will be done in perl. Adding
it to rust the rust TFA crate would not make much sense
anyway as we'd likely not want to use the same http client
crate in pve and pbs anyway (since pve is all blocking code
and pbs is async...)

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agoadd pbs-style TFA API implementation
Wolfgang Bumiller [Tue, 9 Nov 2021 11:27:01 +0000 (12:27 +0100)]
add pbs-style TFA API implementation

implements the same api paths as in pbs by forwarding the
api methods to the rust implementation after performing the
product-specific checks

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agomove TFA api path into its own module
Wolfgang Bumiller [Tue, 9 Nov 2021 11:27:00 +0000 (12:27 +0100)]
move TFA api path into its own module

and remove old modification api

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agohandle yubico authentication in new path
Wolfgang Bumiller [Tue, 9 Nov 2021 11:26:59 +0000 (12:26 +0100)]
handle yubico authentication in new path

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agouse PBS-like auth api call flow
Wolfgang Bumiller [Tue, 9 Nov 2021 11:26:58 +0000 (12:26 +0100)]
use PBS-like auth api call flow

The main difference here is that we have no separate api
path for verification. Instead, the ticket api call gets an
optional 'tfa-challenge' parameter.

This is opt-in: old pve-manager UI with new
pve-access-control will still work as expected, but users
won't be able to *update* their TFA settings.

Since the new tfa config parser will build a compatible
in-perl tfa config object as well, the old authentication
code is left unchanged for compatibility and will be removed
with pve-8, where the `new-format` parameter in the ticket
call will change its default to `1`.

The `change_tfa` call will simply die in this commit. It is
removed later when adding the pbs-style TFA API calls.

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agoupdate read_user_tfa_type call
Wolfgang Bumiller [Tue, 9 Nov 2021 11:26:57 +0000 (12:26 +0100)]
update read_user_tfa_type call

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agouse rust parser for TFA config
Wolfgang Bumiller [Tue, 9 Nov 2021 11:26:56 +0000 (12:26 +0100)]
use rust parser for TFA config

Signed-off-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
9 months agoopenid: proxy: only set env var if DC-config property exists
Thomas Lamprecht [Wed, 3 Nov 2021 10:30:05 +0000 (11:30 +0100)]
openid: proxy: only set env var if DC-config property exists

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
9 months agofix #3513: pass configured proxy to OpenID
Fabian Grünbichler [Tue, 13 Jul 2021 08:09:17 +0000 (10:09 +0200)]
fix #3513: pass configured proxy to OpenID

Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
9 months agobump version to 7.0-6
Thomas Lamprecht [Thu, 21 Oct 2021 10:28:58 +0000 (12:28 +0200)]
bump version to 7.0-6

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
9 months agofix user deletion when realm does not enforce TFA
Dominik Csapak [Tue, 19 Oct 2021 11:52:42 +0000 (13:52 +0200)]
fix user deletion when realm does not enforce TFA

here the existance of the user is only interesting if we want to set
data, not if we delete it.

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
10 months agobump version to 7.0-5
Thomas Lamprecht [Mon, 27 Sep 2021 13:50:50 +0000 (15:50 +0200)]
bump version to 7.0-5

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
10 months agoapi: delete user: better communicate partial deletion
Thomas Lamprecht [Mon, 27 Sep 2021 13:46:26 +0000 (15:46 +0200)]
api: delete user: better communicate partial deletion

this is really an edge case and should not happen often in practice,
the time window is small and deletions are not _that_ common, but
still.

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
10 months agoapi: delete user: disable first to avoid surprise on error
Thomas Lamprecht [Mon, 27 Sep 2021 13:32:38 +0000 (15:32 +0200)]
api: delete user: disable first to avoid surprise on error

Write out a config with the user disabled so that it cannot be used
even if deletion fails, why ever that is

Suggested-by: Wolfgang Bumiller <w.bumiller@proxmox.com>
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
10 months agofix #2302: allow deletion of users when realm enforces TFA
Thomas Lamprecht [Mon, 27 Sep 2021 13:31:31 +0000 (15:31 +0200)]
fix #2302: allow deletion of users when realm enforces TFA

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
10 months agoapi: user: indentation & whitspace cleanups
Thomas Lamprecht [Thu, 23 Sep 2021 12:16:50 +0000 (14:16 +0200)]
api: user: indentation & whitspace cleanups

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
11 months agocheck_path: add /sdn/vnets/* path
Alexandre Derumier [Thu, 5 Aug 2021 15:00:44 +0000 (17:00 +0200)]
check_path: add /sdn/vnets/* path

Signed-off-by: Alexandre Derumier <aderumier@odiso.com>
13 months agobump version to 7.0-4
Thomas Lamprecht [Fri, 2 Jul 2021 11:45:51 +0000 (13:45 +0200)]
bump version to 7.0-4

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
13 months agoapi: users: use public regex directly to parse out realm
Thomas Lamprecht [Fri, 2 Jul 2021 11:42:16 +0000 (13:42 +0200)]
api: users: use public regex directly to parse out realm

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
13 months agoapi: users: code-style cleanup and sort when iterating users
Thomas Lamprecht [Fri, 2 Jul 2021 11:41:29 +0000 (13:41 +0200)]
api: users: code-style cleanup and sort when iterating users

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
13 months agoapi: users: s/realmtype/realm-type/ in response
Thomas Lamprecht [Fri, 2 Jul 2021 11:38:46 +0000 (13:38 +0200)]
api: users: s/realmtype/realm-type/ in response

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
13 months agoapi: user: add realmtype to user list
Dominik Csapak [Thu, 1 Jul 2021 12:25:00 +0000 (14:25 +0200)]
api: user: add realmtype to user list

this makes it much easier to determine if a user can e.g.
change a password or tfa, based on realm

Signed-off-by: Dominik Csapak <d.csapak@proxmox.com>
13 months agoimplement OpenID autocreate user feature
Dietmar Maurer [Wed, 30 Jun 2021 06:10:07 +0000 (08:10 +0200)]
implement OpenID autocreate user feature

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
13 months agoapi: implement openid API
Dietmar Maurer [Wed, 30 Jun 2021 06:10:06 +0000 (08:10 +0200)]
api: implement openid API

This moves compute_api_permission() into RPCEnvironment.pm.

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
13 months agodepend on libpve-rs-perl
Dietmar Maurer [Wed, 30 Jun 2021 06:10:05 +0000 (08:10 +0200)]
depend on libpve-rs-perl

13 months agoadd OpenId configuration
Dietmar Maurer [Wed, 30 Jun 2021 06:10:04 +0000 (08:10 +0200)]
add OpenId configuration

13 months agocheck_user_enabled: also check if user is expired
Dietmar Maurer [Wed, 30 Jun 2021 06:10:03 +0000 (08:10 +0200)]
check_user_enabled: also check if user is expired

13 months agobump version to 7.0-3
Thomas Lamprecht [Mon, 21 Jun 2021 08:31:23 +0000 (10:31 +0200)]
bump version to 7.0-3

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
13 months agobuildsys: clean more
Thomas Lamprecht [Thu, 17 Jun 2021 12:43:44 +0000 (14:43 +0200)]
buildsys: clean more

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
13 months agoaccess control: style: register configs in single line each
Thomas Lamprecht [Thu, 17 Jun 2021 12:41:02 +0000 (14:41 +0200)]
access control: style: register configs in single line each

Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>