Creating a new login with SDR mechanism set to AES overrides old DES3 key instead of creating a new one
Categories
(NSS :: Libraries, defect, P3)
Tracking
(firefox-esr128 unaffected, firefox-esr140 unaffected, firefox140 unaffected, firefox141 wontfix, firefox142 wontfix, firefox143 fixed)
| Tracking | Status | |
|---|---|---|
| firefox-esr128 | --- | unaffected |
| firefox-esr140 | --- | unaffected |
| firefox140 | --- | unaffected |
| firefox141 | --- | wontfix |
| firefox142 | --- | wontfix |
| firefox143 | --- | fixed |
People
(Reporter: maltejur, Assigned: maltejur)
References
(Regression)
Details
(Keywords: regression)
Attachments
(1 file)
I've found that the changes from Bug 198090 currently have a rather big bug. To reproduce:
- Set
security.sdr.mechanismto0 - Create a login
- Set
security.sdr.mechanismto1 - Create another login
- Observe the login list
Only the second login will appear. That is due to a new AES key being generated with the same key ID as the old DES3 key. This results in the old key just being overwritten when the second login gets created. That is because, contrary to what I believed when I wrote the original patch, the keys are only keyed by their ID, not by the key type. The fix would probably be to either key the keys by their type as well, or to use a distinct key ID for each mechanism.
| Assignee | ||
Comment 1•9 months ago
|
||
Updated•9 months ago
|
Comment 2•9 months ago
|
||
Set release status flags based on info from the regressing bug 198090
Comment 3•9 months ago
|
||
We are in our last week of beta for 141, is that a bug we need a fix for in 141? Thanks
Comment 4•9 months ago
|
||
Also could we get a severity set on this bug? Thanks
Comment 5•9 months ago
|
||
This would only affect users who have manually flipped the security.sdr.mechanism pref. We don't need a fix in 141.
Updated•8 months ago
|
Updated•8 months ago
|
Updated•8 months ago
|
Comment 7•8 months ago
|
||
Updated•8 months ago
|
Comment 8•8 months ago
|
||
It looks like the windows build doesn't like the syntax for the gtest changes. I think we can just remove the cast.
| Assignee | ||
Comment 9•8 months ago
|
||
(In reply to Robert Relyea from comment #8)
It looks like the windows build doesn't like the syntax for the gtest changes. I think we can just remove the cast.
Looking at it in Bug 1981518.
Description
•