locking functions. If an application loads a shared object with
dlopen() and the shared object has an init function which requires
lazy binding, then _rtld_bind is called when the thread is already
inside the dynamic linker. This leads to a recursive acquisition
of the lock, which I was not expecting -- hence the assert failure.
This work-around makes the default locking functions handle recursive
locking. It is NOT the correct fix -- that should be implemented
at the generic locking level rather than in the default locking
functions. I will implement the correct fix in a future commit.
Since the dllockinit() interface will likely need to change, warn
about that in both the man page and the header file.
functions to be used by the dynamic linker. This can be called by
threads packages at start-up time. I will add the call to libc_r
soon.
Also add a default locking method that is used up until dllockinit()
is called. The default method works by blocking SIGVTALRM, SIGPROF,
and SIGALRM in critical sections. It is based on the observation
that most user-space threads packages implement thread preemption
with one of these signals (usually SIGVTALRM).
The dynamic linker has never been reentrant, but it became less
reentrant in revision 1.34 of "src/libexec/rtld-elf/rtld.c".
Starting with that revision, multiple threads each doing lazy
binding could interfere with each other. The usual symptom was
that a symbol was falsely reported as undefined at start-up time.
It was rare but not unseen. This commit fixes it.
takes an int arg and is prototyped in <string.h>. It has the opposite
interface botches to psignal(3) which takes a bogus unsigned arg but is
prototyped in the right place.
This is not the last of the interface problems for strsignal(). We
obtained it from NetBSD, but NetBSD has moved its prototype to
<unistd.h>. strsignal() should return const char *, but it returns
char * for historical reasons. NetBSD declares it as returning
__aconst char, where __aconst is normally empty but can be set to
`const' to give better error checking. glibc-2.1.1 prototypes
strsignal() in <string.h>.
the environment. This allows big ID warnings to be suppressed for
vipw and chpass as well.
Since the environment variable test is only performed for callers
of pw_scan() that do not set pw_big_ids_warning, the test can still
be overriden. Currently, chpass and pwd_mkdb are the only users
of pw_scan() and neither of them overrides the environment variable
test.
readdir_r is not POSIX according to POSIX_SOURCE, bruce says:
> readdir_r() is in the _POSIX_SOURCE section, but is not a POSIX.1-1990
> function. It's POSIX.1-1996 so it should be under a different feature
> test which we don't support yet.
make sure errno is saved so that its contents are cleared unless
necessary.
Submitted by: bde
eischen (Daniel Eischen) added wrappers to protect against cancled
threads orphaning internal resources.
the cancelability code is still a bit fuzzy but works for test
programs of my own, OpenBSD's and some examples from ORA's books.
add readdir_r to both libc and libc_r
add some 'const' attributes to function parameters
Reviewed by: eischen, jasone
happy with how this end up and will re-visit the entire empty field
problem, but this patch solves the NIS problem for now.
Submitted by: Dan Nelson <dan@emsphone.com>
PR: 14865,14984
This fixes some nasty procfs problems for SMP, makes ps(1) run much faster,
and makes ps(1) even less dependent on /proc which will aid chroot and
jails alike.
To disable this facility and revert to previous behaviour:
sysctl -w kern.ps_arg_cache_limit=0
For full details see the current@FreeBSD.org mail-archives.
next try over chroot (descriptor closed). getgrnam() used already handles
endgrent() properly and honors _gr_stayopen. Automatically call
setgroupent(1) when _pw_stayopen is set (for YP/NIS code).
-----------------------------
Most of the userland changes are in libc. For both the alpha
and the i386 setjmp has been changed to accomodate for the
new sigset_t. Internally, libc is mostly rewritten to use the
new syscalls. The exception is in compat-43/sigcompat.c
The POSIX thread library has also been rewritten to use the
new sigset_t. Except, that it currently only handles NSIG
signals instead of the maximum _SIG_MAXSIG. This should not
be a problem because current applications don't use any
signals higher than NSIG.
There are version bumps for the following libraries:
libdialog
libreadline
libc
libc_r
libedit
libftpio
libss
These libraries either a) have one of the modified structures
visible in the interface, or b) use sigset_t internally and
may cause breakage if new binaries are used against libraries
that don't have the sigset_t change. This not an immediate
issue, but will be as soon as applications start using the
new range to its fullest.
NOTE: libncurses already had an version bump and has not been
given one now.
NOTE: doscmd is a real casualty and has been disconnected for
the moment. Reconnection will eventually happen after
doscmd has been fixed. I'm aware that being the last one
to touch it, I'm automaticly promoted to being maintainer.
According to good taste this means that I will receive a
badge which either will be glued or mechanically stapled,
drilled or otherwise violently forced onto me :-)
NOTE: pcvt/vttest cannot be compiled with -traditional. The
change cause sys/types to be included along the way which
contains the const and volatile modifiers. I don't consider
this a solution, but more a workaround.
patch to stop the core dumps while others come up with a better
reviewed patch which may also fix other problems. We do illegal
pointer arithmetic, but it should be OK since FreeBSD only supports
machines with flat address spaces.
Submitted by: bde
should be used from now on for anything security but not auth-related.
Included are updates for all relevant manpages and also to /etc files,
creating a new /var/log/security. Nothing in the system logs to
/var/log/security yet as of the time of this commit.
Reviewed by: rgrimes, imp, chris
- Sort xrefs
- FreeBSD.ORG -> FreeBSD.org
- Be consistent with section names as outlines in mdoc(7)
- Other misc mdoc cleanup.
PR: doc/13144
Submitted by: Alexy M. Zelkin <phantom@cris.net>
When fts_open is used with option FTS_NOCHDIR the full
path entry of type FTS_DP is returned with a trailing
'/' if the final directory is empty.
This fix coresponds to netbsd's __fts13.c v. 1.16
Note: you need to install the current groff tmac macros for these
man pages to format correctly. Specifically, rev 1.21 of
contrib_groff/tmac/doc-syms in -current, or rev 1.17.24 for 3.2-stable
(devname() returned "??" when the lookup failed, but callers expected it
to return NULL). This was fixed in Lite2, but until recently the changes
were only merged into devname.3. A day or two after devname.c was fixed,
devname.3 was made inconsistent again by backing out most of the Lite2
changes.
track.
The $Id$ line is normally at the bottom of the main comment block in the
man page, separated from the rest of the manpage by an empty comment,
like so;
.\" $Id$
.\"
If the immediately preceding comment is a @(#) format ID marker than the
the $Id$ will line up underneath it with no intervening blank lines.
Otherwise, an additional blank line is inserted.
Approved by: bde
function. It was an ill-considered feature. It didn't solve the
problem I wanted it to solve. And it added Yet Another Version
Number that would have to be maintained at every release point.
I'm nuking it now before anybody grows too fond of it.
that counted the number of elements in argv. The counter is incremented
in the next-iteration section of the loop, not the body, so at termination
it's already "counted" the element that failed the continuation test - in
this case the NULL argv terminator.
Noted by: bde
call them. All the execX() libc functions should be vfork() safe now.
Specifically:
- execlp() does the argument count-and-build into a vector from alloca
(like the others) - buildargv() is no longer used (and gone).
- execvp() uses alloca/strcpy rather than strdup().
- the ENOEXEC handler uses alloca rather than malloc.
- a couple of free() calls removed - alloca works on the local stack and
the allocations are freed on function exit (which is why buildargv
wasn't useful - it's alloca() context would disappear on return).
Along the way:
- If alloca() fails (can it?), set errno = ENOMEM explicitly.
- The ENOEXEC recovery routine that trys again with /bin/sh appeared to
not be terminating the new argv[] array for /bin/sh, allowing it to
walk off the end of the list.
I dithered a bit about using alloca() even more as it's most commonly
associated with gcc. However, standalone portable (using malloc) and
machine-specific assembler alloca implementations appear to be available
on just about all the architectures we're likely to want to port to.
alloca will be the least of our problems if ever going to another compiler.
which ones cause us to fail. Now all open errors on the databse file
will cause the next file in the list to be tried.
Submitted by: Arne Henrik Juul <arnej@math.ntnu.no>
PR: 4585
- document that sysctl() and sysctlbyname() return 0 on success
- if the provided buffer is too small, set errno to ENOMEM and return -1
instead of returning ENOMEM.
is actually mounted on "/" can be determined using statfs() and is
in /dev. This fixes fsck operating on the wrong device when the
fs_spec entry is only an alias. The aliased case became more
dangerous when the ROOTSLICE_HUNT hack was committed in mount(8).
ROOTSLICE_HUNT may be unnecessary now.
an unimprovement here. I thought it would be an improvement, as in libkvm,
but here we can access the strings directly.
Use sysctlbyname() instead of sysctl() and trust it to give a nonzero
address if it succeeds.
they cannot mount a filesystem that they cannot see in getvfsbyname().
Part 1 of this is a hack, make vfsisloadable() always return true - the
ultimate decider of whether it's loadable or not is kldload() or mount().
Part 2 of this is to have vfsload() call kldload(2) and return success if
it works. This means that we will use a viable kld module in preference
to an LKM!
Ultimately, the thing to do is remove the hacks to do a vfsload in all the
mount_* commands and let the kernel do it by itself in mount(2).
