Fix endless loop in udp_output() on 32 bit architectures.
As CMSG_ALIGN() can overflow the integer value, do the length check
with and without alignment.
from Acts1631; OK mvs@
rpki-client: do not leak cert if auth_insert() fails
If certid reaches CERTID_MAX or the depth exceeds the maximum cert depth,
cert is not inserted into the auth tree. In this case we need to free the
cert.
Reported by Ties de Kock
ok claudio
Don't deallocate the Tx and Rx rings in dwxe_down() and only allocate
them in dwxe_up() if they haven't been allocated already. This avoids
memory allocations that might sleep in DVACT_RESUME.
ok deraadt@
Fix the OM3 link length for SFP modules.
SFF-8472 defines the byte 19 length in units of 10 meters, the same as
the OM1 and OM2 lengths, so scale it by 10 instead of 20.
ok deraadt@ dlg@
fix a mistake i made with the scope of futex addresses.
only futexes pointing at shared memory mappings should allow
coordination between processes.
Wouter Spee reported a preforking gunicorn server would with a futex
in a COW mapping across a fork had futex wakeups delivered to the
wrong process.
this is a slightly different fix to the one they provided, but
semantically equivalent.
ok kettenis@ deraadt@
Update to pkgconf 3.0.11
no concerns by sthen and jca
* Security fixes:
- Package paths are accepted only in queries made directly on the command
line, not in Requires or Requires.private. This prevents a crafted pc(5)
file from causing infinite recursion by requiring itself through a path.
See GHSA-qcp5-98c5-hvh7.
- Limit the bytecode produced when a variable is repeatedly redefined in
terms of itself. This prevents a crafted pc(5) file from causing
exponential memory growth through self-referential variable expansion.
See GHSA-fr7f-m8jv-4qcm.
* Documentation and command-line help:
- List the existing --debug and --license-file options in --help output.
Patch by Andrej730.
https://github.com/pkgconf/pkgconf/blob/pkgconf-3.0.11/NEWS
rpki-client: fix doc comment for carepo_increment()
ca.c 1.17 changed carepo_increment() to a void function, so it no longer
returns the counter.
Pointed out by Ties de Kock
sndiod: Handle -1 return value from snprintf(3)
Wrap snprintf(3) in a new xsnprintf() function that aborts if -1 is
returned. Use the value returned by xsnprintf() to determine the
expected output string size (i.e. the number of bytes that would have
been written if the buffer size was unlimited).
Fix buffer overflow in printf %s and %c formats for UTF-8 chars.
Awk allocates its buffer based on the length in characters instead
of bytes, which can lead to a buffer overflow when formatting UTF-8
characters in a UTF-8 locale.
From Martin Chang
Fix buffer overflow in string() when storing escaped unicode.
Only two bytes were allocated but escaped unicode can require up to 5
(four plus a NUL).
From Martin Chang
ieee80211_input_hwdecrypt() must free the mbuf before returning NULL.
Like ieee80211_decrypt(), ieee80211_input_hwdecrypt() is responsible for
freeing discarded frames, but failed to do so. This bug was introduced by
me in ieee80211_input.c,v r1.216 2020/05/15 when adding this function.
This sounds very bad, but my own impact analysis is as follows:
Only drivers which offload TKIP/CCMP decryption to hardware can be affected.
These are: athn, iwm, iwn, iwx, mtw, mwx, qwx, qwz, ral, run, urtwn, and wpi.
The following drivers filter duplicate data frames in hardware or in the
driver itself: athn, iwm, iwn, iwx, qwx, qwz, wpi, and urtwn.
In which case the leak won't trigger for regular data frames.
Encrypted multicast or management (PMF) frames might also trigger the leak
where drivers offload the group cipher to hardware: iwx, run, qwx, and qwz.
Under normal conditions such frames are rare enough for a leak to go unnoticed.
[30 lines not shown]
Use the generic timer counter to implement getsecs() in efiboot.
On the HP OmniBook X, the notify function of the periodic UEFI timer event
is never called, so ticks stays at 0 and the boot prompt timeout never
expires. Reading CNTVCT_EL0 directly avoids depending on firmware timer
events. Fall back to the timer event ticks if CNTFRQ_EL0 isn't set. Apply
the same Cortex-A73 errata 858921 workaround as agtimer(4).
Help, testing, and ok kettenis@
Support for the IBM FireGL-UX boards.
The last piece of the puzzle was to pass the right flags at the first init
microcode call, in order to correctly completely initialize its own "softc".
This makes character drawing operation point to the right offset within the
frame buffer.
Thanks to Daniel Dickman for noticing the microcode had a specific code path
when some bits (which we did not set) in the init flags were set.
Battery notifications only require the low byte, mask the rest.
This silences unknown battery notification messages on some machines.
ok mlarkin@, deraadt@
_shutdown group members can run shutdown with their own environment
including PATH, which is not good. Reuse the (extremely, maybe too restricted?)
restricted_environ[] used by the the wall(1) run.
from Acts1631