RFC: deprecating crypto usage in secret-service
smcv at collabora.com
Tue Sep 1 12:23:12 UTC 2020
On Sun, 30 Aug 2020 at 22:34:24 -0600, Thayne wrote:
> How about having some way to signal to D-Bus that a message is
> sensitive so should be stored in non-swappable memory?
Sorry, this is unlikely to be implementable. dbus-daemon (or dbus-broker
or whatever other message bus implementation is in use) cannot know
what flags a message has until it has already received at least part of
the message, at which point it's very likely to be too late to put that
message in non-swappable memory.
The message bus is already relatively complex, and I would prefer not to
add additional platform-specific complexity by trying to prevent certain
messages from being swapped out.
The design of D-Bus is such that clients can cause the message bus to
allocate large amounts of memory, and this cannot be prevented without
dropping messages on the floor, which is not something that clients
can generally detect or recover from. (Yes, this is a design flaw;
unfortunately, solving this would not be a backwards-compatible change,
and D-Bus' single most important feature is interoperability). As a
result, asking for message bus features that involve much-more-limited
resources like locked memory pages is unlikely to succeed.
More information about the xdg