Implement a different fix for the use-after-free in filt_procattach
from rev 1.207.
Instead of shuffling locks around take a process reference in this
code to prevent the process from exiting under us while sleeping
for the kqueue_ps_list_lock rwlock. This is the same solution that
was used for a similar issue in fork1.
Reported by Acts1631 and OpenAI
Patch provided by OpenAI
from claudio@; OK mvs@
this is errata/8.0/022_kqueue.patch.sig
Implement a different fix for the use-after-free in filt_procattach
from rev 1.207.
Instead of shuffling locks around take a process reference in this
code to prevent the process from exiting under us while sleeping
for the kqueue_ps_list_lock rwlock. This is the same solution that
was used for a similar issue in fork1.
Reported by Acts1631 and OpenAI
Patch provided by OpenAI
from claudio@; OK mvs@
this is errata/7.9/057_kqueue.patch.sig
Do not allow TIOCSCTTY, TIOCNOTTY, or TIOCSPGRP on the pty master. Return
ENOTTY for all 3 of these. They could create incoherent session states,
leading to dangerous UAF reported by openai (unfortunately their proposed
patch was batshit crazy and made the situation worse, due to not identifying
this as root cause, and writing code to support the condition).
from deraadt@ with claudio@ various versions; OK kettenis@ kirill@ and others
this is errata/8.0/021_pty.patch.sig
Do not allow TIOCSCTTY, TIOCNOTTY, or TIOCSPGRP on the pty master. Return
ENOTTY for all 3 of these. They could create incoherent session states,
leading to dangerous UAF reported by openai (unfortunately their proposed
patch was batshit crazy and made the situation worse, due to not identifying
this as root cause, and writing code to support the condition).
from deraadt@ with claudio@ various versions; OK kettenis@ kirill@ and others
this is errata/7.9/056_pty.patch.sig
Update to Blender 4.5.16.
- Build atomic_ops without libatomic; drop atomic from WANTLIB.
- Use the system Eigen (math/eigen3).
- Garbage collect the ONLY_FOR_ARCHS comments; 32-bit archs
are not coming back.
from Brad, thanks
libcrypto: fix a long-standing quirk in CCM mode
For some reason it was decided that in CCM mode the caller should not be
allowed to set the tag length before setting the tag, which is the natural
order. This required various contortions in downstream, at least in Ruby
and rust-openssl.
This pulls in 9cca7be1 (before license change) and makes a trivial change
functionally equivalent to (but different from) 67c81ec3 from OpenSSL.
From William T. Nelson
libcrypto: add missing AEAD flag to AES-CCM
The AES-CCM ciphers aren't marked as AEAD ciphers, so openssl enc allows
them even though it shouldn't. OpenSSL has set the flag since adding CCM
support to libssl in e75c5a79.
From William T. Nelson