[pve-devel] [PATCH v3 pve-storage] fix #6900: correctly detect PBS API tokens in storage plugin

[pve-devel] [PATCH v3 pve-storage] fix #6900: correctly detect PBS API tokens in storage plugin

[Robert Obkircher]

The PBS storage plugin used PVE code to detect if an API token was
entered in the username field. This lead to bad requests for some
valid PBS tokens which are not valid PVE tokens. Examples are
"root@pam!1234" and "root@pam!_-".

Relax the token pattern to allow token names and realms that start
with numbers or underscores. Also allow single character token names,
which are allowed on the backend even though they can't be created
through the PBS Web UI.

Signed-off-by: Robert Obkircher <r.obkircher@proxmox.com>
---
 src/PVE/Storage/PBSPlugin.pm | 19 +++++++++++++++++--
 1 file changed, 17 insertions(+), 2 deletions(-)

diff --git a/src/PVE/Storage/PBSPlugin.pm b/src/PVE/Storage/PBSPlugin.pm
index 5842004..17b2e2d 100644
--- a/src/PVE/Storage/PBSPlugin.pm
+++ b/src/PVE/Storage/PBSPlugin.pm
@@ -701,6 +701,21 @@ my sub snapshot_files_encrypted {
     return $any && $all;
 }
 
+# SAFE_ID_REGEX_STR: &str = r"(?:[A-Za-z0-9_][A-Za-z0-9._\-]*)";
+my $safe_id_regex = qr/(?:[A-Za-z0-9_][A-Za-z0-9\._\-]*)/;
+
+# TOKEN_NAME_REGEX_STR: = SAFE_ID_REGEX_STR
+my $token_name_regex = $safe_id_regex;
+
+# USER_NAME_REGEX_STR: &str = r"(?:[^\s:/[:cntrl:]]+)";
+my $user_name_regex = qr/(?:[^\s:\/\p{PosixCntrl}]+)/;
+
+# USER_ID_REGEX_STR: &str = concatcp!(USER_NAME_REGEX_STR, r"@", SAFE_ID_REGEX_STR);
+my $user_id_regex = qr/${user_name_regex}\@${safe_id_regex}/;
+
+# APITOKEN_ID_REGEX_STR: &str = concatcp!(USER_ID_REGEX_STR, r"!", TOKEN_NAME_REGEX_STR);
+my $apitoken_id_regex = qr/${user_id_regex}\!${token_name_regex}/;
+
 # TODO: use a client with native rust/proxmox-backup bindings to profit from
 # API schema checks and types
 my sub pbs_api_connect {
@@ -710,8 +725,8 @@ my sub pbs_api_connect {
 
     my $user = $scfg->{username} // 'root@pam';
 
-    if (my $tokenid = PVE::AccessControl::pve_verify_tokenid($user, 1)) {
-        $params->{apitoken} = "PBSAPIToken=${tokenid}:${password}";
+    if ($user =~ qr/^${apitoken_id_regex}$/) {
+        $params->{apitoken} = "PBSAPIToken=${user}:${password}";
     } else {
         $params->{password} = $password;
         $params->{username} = $user;
-- 
2.47.3



_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel

Re: [pve-devel] [PATCH v3 pve-storage] fix #6900: correctly detect PBS API tokens in storage plugin

[Thomas Lamprecht]

Am 20.11.25 um 13:10 schrieb Robert Obkircher:
> The PBS storage plugin used PVE code to detect if an API token was
> entered in the username field. This lead to bad requests for some
> valid PBS tokens which are not valid PVE tokens. Examples are
> "root@pam!1234" and "root@pam!_-".
> 
> Relax the token pattern to allow token names and realms that start
> with numbers or underscores. Also allow single character token names,
> which are allowed on the backend even though they can't be created
> through the PBS Web UI.
> 
> Signed-off-by: Robert Obkircher <r.obkircher@proxmox.com>
> ---
Please add a short changelog for new patch revisions in this area.
Placing it here ensures that it will be visible for review, but doesn't
gets picked up by git am, as that info should not contain anything relevant
for the persistent git history, rather just meta info for helping review.

FWICT it would spell out something like:

changelog v2 -> v3:
* drop importing unused PVE::Auth::Plugin module

in this case.

>  src/PVE/Storage/PBSPlugin.pm | 19 +++++++++++++++++--
>  1 file changed, 17 insertions(+), 2 deletions(-)
> 
> diff --git a/src/PVE/Storage/PBSPlugin.pm b/src/PVE/Storage/PBSPlugin.pm
> index 5842004..17b2e2d 100644
> --- a/src/PVE/Storage/PBSPlugin.pm
> +++ b/src/PVE/Storage/PBSPlugin.pm
> @@ -701,6 +701,21 @@ my sub snapshot_files_encrypted {
>      return $any && $all;
>  }
>  
> +# SAFE_ID_REGEX_STR: &str = r"(?:[A-Za-z0-9_][A-Za-z0-9._\-]*)";

IMO there not much of a point in having the rust code source as literal
comment here, that might change and then one might not be able to draw
that assumption anymore.

Rather, just explicitly state where they come from and why we do not
reuse the PVE specific regex here, something like:

# We cannot use the PVE API token regexes as we're stricter in PVE, so some tokens that would
# be valid for PBS would get rejected. Adapt over the PBS ones from <mod/file/...>

> +my $safe_id_regex = qr/(?:[A-Za-z0-9_][A-Za-z0-9\._\-]*)/;
> +
> +# TOKEN_NAME_REGEX_STR: = SAFE_ID_REGEX_STR
> +my $token_name_regex = $safe_id_regex;
> +
> +# USER_NAME_REGEX_STR: &str = r"(?:[^\s:/[:cntrl:]]+)";
> +my $user_name_regex = qr/(?:[^\s:\/\p{PosixCntrl}]+)/;
> +
> +# USER_ID_REGEX_STR: &str = concatcp!(USER_NAME_REGEX_STR, r"@", SAFE_ID_REGEX_STR);
> +my $user_id_regex = qr/${user_name_regex}\@${safe_id_regex}/;
> +
> +# APITOKEN_ID_REGEX_STR: &str = concatcp!(USER_ID_REGEX_STR, r"!", TOKEN_NAME_REGEX_STR);
> +my $apitoken_id_regex = qr/${user_id_regex}\!${token_name_regex}/;
> +
>  # TODO: use a client with native rust/proxmox-backup bindings to profit from
>  # API schema checks and types
>  my sub pbs_api_connect {
> @@ -710,8 +725,8 @@ my sub pbs_api_connect {
>  
>      my $user = $scfg->{username} // 'root@pam';
>  
> -    if (my $tokenid = PVE::AccessControl::pve_verify_tokenid($user, 1)) {
> -        $params->{apitoken} = "PBSAPIToken=${tokenid}:${password}";
> +    if ($user =~ qr/^${apitoken_id_regex}$/) {
> +        $params->{apitoken} = "PBSAPIToken=${user}:${password}";
>      } else {
>          $params->{password} = $password;
>          $params->{username} = $user;



_______________________________________________
pve-devel mailing list
pve-devel@lists.proxmox.com
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel