This commit is the 2/3 part of the fix for the following security
vulnerability:
- CVE-2021-42076 DoS via excess length messages
The issue has been reported by Matthias Gerstner <mgerstner@suse.de>.
Many tools strip trailing whitespaces by default, so after editing a file with
whitespace errors we end up with a bunch of unrelated hunks that need to be
reverted locally.
This could be fixed by configuring each tool to not do this (at least for the
barrier repo), or, simpler, we just sed the problem away and make barrier
whitespace-compliant.
sed commands run:
sed -i 's/[ \t]\+$//' **/*.(cpp|h) **/*CMakeLists.txt
Verified with git diff --ignore-space-change, this shows the empty diff.
The commit a841b28 changed the condition for removing job from processing.
New flag MultiplexerJobStatus::continue_servicing become used
instead of checking pointer for NULL.
However for cases when TCPSocket::newJob() returns nullptr
the behaviour changed: earlier the job was removed, but after change
it is called again, since MultiplexerJobStatus equal to {true, nullptr}
means "run this job again".
This leads to problem with eating CPU and RAM on linux
https://github.com/debauchee/barrier/issues/470
There is similar windows problem, but not sure it is related.
https://github.com/debauchee/barrier/issues/552
Since it looks that the goal of a841b28 was only clarifying
object ownership and not changing job deletion behaviour,
this commit tries to get original behaviour and fix the bugs above
by returning {false, nullptr} instead of {true, nullptr}
when TCPSocket::newJob() returns nullptr.
Its possible poll() will return that a socket is both readable and
writable. When this happens TCPSocket::serviceConnected() is
overwriting the result from doWrite() with the result from doRead()
This can cause a situation where doWrite() tried to notify that we
should stop polling if the socket is writable, but the doRead() result
causes that to be ignored. This results in a tight loop and 100% cpu
usage in SocketMultiplexer::serviceThread() as the poll() call
instantly returns that the socket is writable, but no one cares that
it is. The issue eventually corrects itself with enough mouse
movement/clicks.
- reordered includes to prevent accidental dependencies.
- moved arch specific files into their own folder (easier to include with cmake).
- renamed non-class files to reflect new code style (lower case with underscores).
- using full relative paths for includes (easier for understanding dependencies).
- using #pragma once instead of classic-style header guards.
Povilas Kanapickas
Peter Hutterer
Vasily Galkin
jwestfall
walker0643
Andrew Nelless
XinyuHou
Adam Potolsky
Nigel Armstrong
Nick Bolton