The FreeBSD-native arc4random_buf implementation falls back to weak sources of
entropy if the sysctl fails. Remove these dangerous fallbacks by overriding
locally.
Unfortunately, pthread_atfork() is also broken on FreeBSD (at least 9 and 10)
if a program does not link to -lthr. Callbacks registered with pthread_atfork()
simply fail silently. So, it is not always possible to detect a PID wraparound.
I wish we could do better.
This improves arc4random_buf's safety compared to the native FreeBSD
implementation.
Tested on FreeBSD 9 and 10.
The old man page had a lot of useful information, but it was all mixed
together which made it difficult to reference. The main theme in this
commit is that the sections are more focused:
* DESCRIPTION describes the overall behavior
* RETURN VALUES describes what it may return (including implementation
defined values)
* EXAMPLES shows why we recently started an audit on malloc and realloc
usage in the tree.
* Added CAVEATS which describes what is implementation defined, gotchas
and security implications of misusing these functions
* Added IDIOMS which describes how these functions should or
should not be used
The MALLOC_OPTIONS section was left unchanged. Function names were
added to DIAGNOSTICS and STANDARDS. The MALLOC_OPTIONS and DIAGNOSTICS
sections were pushed down in the page so more pertinent information is
higher up.
This has gone through several revisions thanks to input from deraadt@
and schwarze@. Ingo also helped with some of the mandoc formatting.
OK schwarze@ (as far as it is a good starting point and the code
snippets look ok)
in case something wants to create massive amounts of environment, like
a bit more than 1/4 of a 32-bit address space. unrealistic -- but why
audit one code path, and not treat others the same? then you have to
re-engage everytime you see the code. read the news, that isn't what
developers do. At least if the code paths look the same, there is hope,
because they are easier to verify for correctness. developers need
to give other developers a chance to want to care.
Mention that invalid bases do set EINVAL (as required by POSIX);
this part of the change uses part of an earlier patch by millert@.
Minor mdoc(7) cleanup and sync between the two pages while here.
Feedback and ok jmc@ and millert@.
required by the C standard when called with an invalid base:
strtoll(), strtoimax(), strtoul(), strtoull(), and strtoumax().
Same behaviour for strtoq() and strtouq() even though not standardized.
No functional change in strtol(), it was the only one already correct.
While here, simplify the conditional expression for checking the base
and sync whitespace and comments among the six files.
ok millert@
to hold the malloc lock across mmap syscalls in all cases. dropping it
allows another thread to access the existing chunk cache if necessary.
could be improved to be a bit more aggressive, but i've been testing this
simple diff for some time now with good results.
This enables support for the new getrandom(2) syscall in Linux 3.17.
If the call exists and fails, return a failure in getentropy(2) emulation as
well. This adds a EINTR check in case the urandom pool is not initialized.
Tested on Fedora Rawhide with 3.17rc0 and Ubuntu 14.04
ok deraadt@
circular lists. Amazingly, they managed to extend the requirements to no
longer match the behavior of the VAX instructions they were modeled after,
so the trivial VAX ASM versions have to go. Nice job breaking it, X/Open!
Based on a diff from enh (at) google.com
ok miod@
Most assembly blocks remain inactive if OPENSSL_NO_ASM is not defined,
only enabling inline assembly, but the RSA / RC4-5 blocks (used only in
amd64 systems) turn on implicitly. Guard these two as well.
This simplifies enabling just inline ASM in portable, no effective
change in OpenBSD.
the details are under embargo. The original plan was to wait for the
embargo to lift, but we've been waiting for quite some time, and there's no
indication of when or even if it will end. No sense in dragging this out
any longer.
The SRP code has never been enabled in OpenBSD, though I understand it is
in use by some other people. However, in light of this and other issues,
we're officially saying SRP is outside the scope of libressl. (For now.)
Move <sys/mman.h> and raise(SIGKILL) calls to OS-specific headers.
On OpenBSD, move thread_private.h as well to arc4random.h.
On Windows, use TerminateProcess on getentropy failure.
ok deraadt@
6.11.5 - Storage-class specifiers:
The placement of a storage-class specifier other than at the
beginning of the declaration specifiers in a declaration is
an obsolescent feature.
Diff from Jean-Philippe Ouellet (jean-philippe (at) ouellet.biz)
it shows up in libraries. Even the system call is probably not finalized.
Bit dissapointed it has turned out to be a descriptor-less read() with
EINVAL and EINTR error conditions, but we can work with it.
into one if a system has an awesome getentropy(). In that case it
is valid to totally throw away the rsx state in the child. If the
getentropy() is not very good and has a lazy reseed operation, this
combining is a bad idea, and the reseed should probably continue to
use the "something old, something new" mix. _rs_allocate() can
accomodate either method, but not on the fly.
ok matthew
1. Use "len" parameter instead of sizeof(*rs).
2. Simplify the atfork handler to be strictly async signal safe by
simply writing to a global volatile sig_atomic_t object, and then
checking for this in _rs_forkdetect(). (Idea from discussions with
Szabolcs Nagy and Rich Felker.)
3. Use memset(rs, 0, sizeof(*rs)) to match OpenBSD's MAP_INHERIT_ZERO
fork semantics to avoid any skew in behavior across platforms.
ok deraadt
bcook