Web Misc - #1631
Merged
Merged
Web Misc#1631
Conversation
mattiaswal
force-pushed
the
wifi-web
branch
from
September 15, 2026 13:59
85756ff to
38b5d7d
Compare
'iw dev info' reports no SSID for a mesh interface, so the leaf stayed empty in operational status, in the CLI and in the WebUI. Ask wpa_supplicant instead, the one that joined the mesh, so the mesh id is absent until it actually has rather than echoing what was configured. wpa_supplicant prints SSIDs through printf_encode(), so the value needs decoding. Scan results have done that all along; factor it out and use it for both. Signed-off-by: Mattias Walström <lazzer@gmail.com>
admin@jaffa:/> show interface wifi-mesh name : wifi-mesh type : wifi index : 17 mtu : 1500 operational status : up ip forwarding : disabled physical address : 82:0c:43:26:60:00 ipv4 addresses : ipv6 addresses : in-octets : 1181221 out-octets : 743429 mode : mesh-point mesh-id : laser-mesh connected peers : 2
A PHY's hwmon device is named after the radio, so its temperature sensor claimed "radio0" and unique_names() renamed the radio component itself "radio0-1". Interfaces reference their radio by name, so nothing could match an interface to its radio any more: 'show hardware' and the WebUI both lost the link. Make the sensor a child of the radio instead, named radio0-temp, the way the SoC sensor sits under the CPU. Signed-off-by: Mattias Walström <lazzer@gmail.com>
Mesh point interfaces could only be created and inspected through the Advanced YANG tree. Add them to the interface wizard and editor, and show mesh id and peers on the WiFi and interface status pages. Station status gains the BSSID it is associated to, so the access points of a roaming ESS can be told apart. Access point roaming (802.11k/r/v, band steering, OKC) gets its own editor section. The whole wifi container is written in one PUT: a merge can neither drop an unticked presence container nor revert a leaf to its default. Roaming and mesh forwarding stay on the configure page. They are config-only leaves with no operational counterpart, and a status page has no business restating running config as if it were state. Signed-off-by: Mattias Walström <lazzer@gmail.com>
The editor posted the stored mode in a hidden input and only rendered that mode's fields, so an interface created as a station could never become an access point or a mesh point without deleting it or going through the Advanced YANG tree. Render every mode's fields, tagged with the modes they belong to, and let a set of mode radios reveal the matching set. The save already replaces the whole wifi container, so the mode left behind goes away with it. An interface with no mode configured is in scan-only mode and shows no mode-specific field until one is picked. Signed-off-by: Mattias Walström <lazzer@gmail.com>
The inline "+ New" forms for keystore keys and Wi-Fi radios are rendered hidden inside the form they belong to, with required fields. Hiding a control does not take it out of constraint validation, so the browser refused to submit the surrounding form, and could not focus the offending field to explain why. Save WiFi and the Add Interface OK button did nothing at all, with no error anywhere, for Wi-Fi and WireGuard. Disable a block's controls whenever it is hidden and enable them when it opens, so the fields only take part while the user is filling them in. Verified in a browser against the rendered page: before, the editor form reported four hidden blockers and the Add dialog five; after, both submit. Signed-off-by: Mattias Walström <lazzer@gmail.com>
The previous commit disabled the inline "+ New" forms while hidden. The same shape is used by the "+ Add" rows on the keystore, users, routes, DNS, NTP, firewall and hardware pages, and by the DHCP foldouts: fields rendered inside a form but hidden until a button opens them. Those are not broken today, because each Add button sits inside the row it reveals, so nothing pressable is blocked while it is hidden. They break the moment a second submit lands in the same form, and the failure is silent. Disable every block a reveal hook controls, so a hidden required field can never take part in validation. Swept all 24 pages plus the YANG tree fragments in a browser: no page has a hidden field blocking a pressable button, and every block still enables its fields when opened. Signed-off-by: Mattias Walström <lazzer@gmail.com>
mesh_fwding decides whether a node relays for its peers, and nothing reported it: the WebUI had taken to reading the configured value and showing it as state, which says nothing about what the mesh is doing. Add 'get mesh_param' to iw.py, which returns the whole parameter block, and report the flag from it. Absent when the driver has nothing to say, rather than guessed from the YANG default. Signed-off-by: Mattias Walström <lazzer@gmail.com>
The hardware page printed the component name, so a radio temperature read "radio0-temp" under a "RADIO0" heading: the same word twice, and a key that exists only to stay unique. Show the description the collector gave it, and where there is none drop the heading the sensor already sits under, the way 'show hardware' does. "cpu-thermal" under "cpu" reads "Thermal". Signed-off-by: Mattias Walström <lazzer@gmail.com>
Same as the hardware page, but the Overview list has no headings to sit under, so a row says "radio0" rather than "Temperature", which alone would be three identical rows on a three-radio board. The reading is named as well only where one component reports more than one, an SoC with a sensor per die. Signed-off-by: Mattias Walström <lazzer@gmail.com>
The load-time disable sweep disables every control inside the hidden add-hw-row, and reopening the row skips controls nested in a block that is itself still hidden, which the country select is. Picking a radio only cleared .hidden, so the select stayed disabled: required is ignored on a disabled control, the form submitted without country-code, and parseWiFiRadio rejected it with "country code is required". Adding a WiFi radio from Configure > Hardware was impossible. Signed-off-by: Mattias Walström <lazzer@gmail.com>
A form that edits several nodes had to send one request per node, so a later failure left the earlier writes in the candidate. rousette applies a YANG Patch as one edit batch. Signed-off-by: Mattias Walström <lazzer@gmail.com>
The radio and the wifi container went in two requests, and the wifi half was written before the radio form was parsed, so a bad channel left a half-applied edit behind a "save failed" message. Signed-off-by: Mattias Walström <lazzer@gmail.com>
Signed-off-by: Mattias Walström <lazzer@gmail.com>
Signed-off-by: Mattias Walström <lazzer@gmail.com>
The vlans container is merged instead of replaced, so saving the bridge form no longer empties the VLAN table of an 802.1Q bridge. Signed-off-by: Mattias Walström <lazzer@gmail.com>
The presence containers are merged instead of replaced, so re-saving the form keeps whatever the DHCP client container already holds. Signed-off-by: Mattias Walström <lazzer@gmail.com>
The picker option carried data-country but the Edit prefill maps fields by name, so it looked for data-country-code. The select stayed on the placeholder and every Edit-then-Save failed with "country code is required". Signed-off-by: Mattias Walström <lazzer@gmail.com>
The error went to the enclosing form's status span, next to the Save WiFi button far below the radio box, so it looked like nothing happened. Signed-off-by: Mattias Walström <lazzer@gmail.com>
Browsers decode header bytes as ISO-8859-1, so the en dash in "1–196" reached the page as mojibake. Some emitters also concatenated the text into the JSON unescaped, so a quote in an error message broke the event. Signed-off-by: Mattias Walström <lazzer@gmail.com>
The edit for a value-less operation was created as a normal node with no value, which only works for string leaves. Create it opaque, as the DELETE method already does, so an enumeration or boolean leaf can be removed inside a patch. Signed-off-by: Mattias Walström <lazzer@gmail.com>
mattiaswal
force-pushed
the
wifi-web
branch
from
September 15, 2026 14:41
135fa4c to
b2372b7
Compare
mattiaswal
marked this pull request as ready for review
September 15, 2026 14:42
adminurl and path come from whatever the LAN advertises. adminurl went into href through template.URL unchecked, so a javascript: URL got through (CSP stops it running), and path was string-appended to the host, so "@evil" moved the link to another host. Signed-off-by: Mattias Walström <lazzer@gmail.com>
Signed-off-by: Mattias Walström <lazzer@gmail.com>
The YANG tree and /api/data pass the path from the query string, so a ".." segment, a query or fragment went to the device as-is, where nginx normalised it into other locations with the user's credentials. Signed-off-by: Mattias Walström <lazzer@gmail.com>
mattiaswal
force-pushed
the
wifi-web
branch
from
September 16, 2026 05:57
977f5c2 to
d55c100
Compare
troglobit
approved these changes
Sep 16, 2026
troglobit
left a comment
troglobit
left a comment
Contributor
There was a problem hiding this comment.
Nice work, reviewed most of this last night, when we discussed the last three fixes, lgtm!
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Description
Checklist
Tick relevant boxes, this PR is-a or has-a: