Change '--' to '\-\-' for options and '--' between words to '\(em'
(em-dash).
Signed-off-by: Bjarni Ingi Gislason <bjarniig@rhi.hi.is>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Change '-' to '\-' for the prefix of names to indicate an option.
Signed-off-by: Bjarni Ingi Gislason <bjarniig@rhi.hi.is>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
The host ID might once have been intended to be globally unique,
but that turned out not to feasible.
Reported-by: Rich Felker <dalias@libc.org>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Remove superfluous paragraph macros.
Remove request ".br" if it precedes a line, that begins with a
space, as such lines automatically cause a break.
There is no change in the output from "nroff" and "groff".
###
Examples of warnings from "mandoc -Tlint":
mandoc: bindresvport.3:41:2: WARNING: skipping paragraph macro: PP after SH
mandoc: crypt.3:228:2: WARNING: skipping paragraph macro: PP empty
mandoc: dlinfo.3:151:2: WARNING: skipping paragraph macro: IP empty
mandoc: exec.3:86:2: WARNING: skipping paragraph macro: PP after SS
mandoc: getsubopt.3:45:2: WARNING: skipping paragraph macro: br before text line with leading blank
Signed-off-by: Bjarni Ingi Gislason <bjarniig@rhi.hi.is>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
These are all links that were created several years ago, mainly
when pages were migrated to different sections, in order to
allow the 'man' commands using the old section numbers to work.
However, the plan was always to eventually remove them, after
allowing people who cared to get used to the new section numbers.
Now, after 5+ years in each case, it's time to remove
these links.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
I believe new users should be discouraged from using atoi() and
that its disadvantages should be explained.
I added the information that 0 is returned on error - although C
standard and POSIX say that "If the value of the result cannot be
represented, the behavior is undefined." there are some
interpretations that 0 has to be returned
https://stackoverflow.com/questions/38393162/what-can-i-assume-about-the-behaviour-of-atoi-on-error
and this is also what happens in practice with glibc, musl and
uClibc.
Signed-off-by: Arkadiusz Drabczyk <arkadiusz@drabczyk.org>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Output is from: test-groff -b -e -mandoc -T utf8 -rF0 -t -w w -z
[ "test-groff" is a developmental version of "groff" ]
There is no change in the output of "nroff" and "groff".
####
troff: <fts.3>:50: warning: trailing space
####
troff: <getgrnam.3>:175: warning: trailing space
####
troff: <getpwnam.3>:181: warning: trailing space
####
troff: <rcmd.3>:52: warning: trailing space
troff: <rcmd.3>:57: warning: trailing space
troff: <rcmd.3>:60: warning: trailing space
troff: <rcmd.3>:63: warning: trailing space
troff: <rcmd.3>:69: warning: trailing space
troff: <rcmd.3>:73: warning: trailing space
####
troff: <rexec.3>:48: warning: trailing space
troff: <rexec.3>:51: warning: trailing space
####
troff: <sem_open.3>:36: warning: trailing space
Signed-off-by: Bjarni Ingi Gislason <bjarniig@rhi.hi.is>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
RAND_MAX is for rand(3). POSIX fixes random()'s range at 2^31-1;
RAND_MAX may be smaller on some platforms (even though with glibc
or musl on Linux they are the same).
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
In the few pages where this heading (which is "nonstandard" within
man-pages) is used, it always immediately follows CONFORMING TO
and generally contains information related to standards. Remove
the section heading, thus incorporating AVAILABILITY into
CONFORMING TO.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
EXAMPLES appears to be the wider majority usage across various
projects' manual pages, and is also what is used in the POSIX
manual pages.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
The fact that an FE_UNDERFLOW exception is not raised for
"Range error: result underflow" is intended behavior.
See https://www.sourceware.org/bugzilla/show_bug.cgi?id=6806.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
The function make_message illustrates how to use vsnprintf to
determine the required amount of memory for a specific format and
its arguments.
If make_message is called with a format which will use exactly
INT_MAX characters (excluding '\0'), then the size++ calculation
will overflow the signed integer "size", which is an undefined
behaviour in C.
Since malloc and vsnprintf rightfully take a size_t argument, I
decided to use a size_t variable for size calculation. Therefore,
this patched code uses variables of the same data types as
expected by function arguments.
Proof of concept (tested on Linux/glibc amd64):
int main() { make_message("%647s%2147483000s", "", ""); }
If the code is compiled with address sanitizer (gcc
-fsanitize=address) you can see the following line, assuming that
a signed integer overflow simply leads to INT_MIN:
==3094==WARNING: AddressSanitizer failed to allocate 0xffffffff80000000 bytes
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Remove the text ("rare)" after a note from Vincent Lefèvre:
Subject: [Bug math/13932] dbl-64 pow unexpectedly slow for some inputs
Date: Sat, 23 May 2020 21:31:52 +0000
From: vincent-srcware at vinc17 dot net <sourceware-bugzilla@sourceware.org>
To: mtk.manpages@gmail.comhttps://sourceware.org/bugzilla/show_bug.cgi?id=13932
--- Comment #26 from Vincent Lefèvre <vincent-srcware at vinc17 dot net> ---
(In reply to Michael Kerrisk from comment #25)
> Fix documented for man-pages-5.07.
[...]
> -On 64-bits,
> +Before glibc 2.28,
> .\"
> .\" https://sourceware.org/bugzilla/show_bug.cgi?id=13932
> +on some architectures (e.g., x86-64)
> .BR pow ()
> may be more than 10,000 times slower for some (rare) inputs
> than for other nearby inputs.
[...]
The problematic values are uncommon, but not so rare, in the sense
that they are close to simple values, i.e. are likely to occur in
practice. An example given above: pow(0.999999999999999889, 1.5)
1 and 1.5 are very simple values, which are more likely to occur
in practice than some fixed random value. Then it suffices to have
a small rounding error on 1...
For instance, this is very different from hard-to-round cases of
exp, which are also very slow IMHO, but unless one writes a
specific program for them, no-one should notice the slowness
because such a case would typically occur only once among billions
(I don't remember the accuracy before the slowest path in this
library).
Reported-by: Vincent Lefèvre <vincent-srcware@vinc17.net>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Avoid implying that use of IFUNC is the only way to produce a
symbol with NULL value. Give more scenarios how a symbol may get
NULL value, but explain that in those scenarios dlsym() will fail
with Glibc's ld.so due to an implementation inconsistency.
Signed-off-by: Alexander Monakov <amonakov@ispras.ru>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
This is a sequel to commit baf17bc4f2, addressing the
issues with missing commas in the middle of SEE ALSO lists that
emerged since.
The awk script from the original commit was not working and had to
be slightly modified (s/["]SEE ALSO["]/"?SEE ALSO/), otherwise it
works like a charm. Here's the fixed script and its output just
before this commit:
for f in man*/*; do
awk '
/^.SH "?SEE ALSO/ {
sa=1; print "== " FILENAME " =="; print; next
}
/^\.(PP|SH)/ {
sa=0; no=0; next
}
/^\.BR/ {
if (sa==1) {
print;
if (no == 1)
print "Missing comma in " FILENAME " +" FNR-1; no=0
}
}
/^\.BR .*)$/ {
if (sa==1)
no=1;
next
}
/\.\\"/ {next}
/.*/ {
if (sa==1) {
print; next
}
}
' $f; done | grep Missing
Missing comma in man1/memusage.1 +272
Missing comma in man2/adjtimex.2 +597
Missing comma in man2/adjtimex.2 +598
Missing comma in man2/mkdir.2 +252
Missing comma in man2/sigaction.2 +1045
Missing comma in man2/sigaction.2 +1047
Missing comma in man3/mbsnrtowcs.3 +198
Missing comma in man3/ntp_gettime.3 +142
Missing comma in man3/strcmp.3 +219
Missing comma in man3/strtol.3 +302
Missing comma in man3/wcstombs.3 +120
Missing comma in man7/user_namespaces.7 +1378
Missing comma in man7/xattr.7 +198
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
By way of a hint that the file descriptor returned by dirfd()
could usefully be fed to the *at() APIs.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Both functions behave the same wrt return value, no need to describe
them separately.
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
The Linux man page for ptsname_r, when describing the behaviour
in the error case, is
- not consistent with the future POSIX standard (POSIX Issue 8).
- not consistent with musl libc.
Find attached a patch to
- keep it consistent with what glibc does,
- make it consistent with musl libc,
- make it consistent with the future POSIX standard (POSIX
Issue 8).
Details:
glibc's implementation of ptsname_r, when it fails, returns the
error code as return value AND sets errno. See
https://sourceware.org/git/?p=glibc.git;a=blob;f=login/ptsname.chttps://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/mach/hurd/ptsname.chttps://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/sysv/linux/ptsname.c
musl's implementation of ptsname_r, when it fails, returns the error code
but does NOT set errno. See
https://git.musl-libc.org/cgit/musl/tree/src/misc/pty.c
The proposal to add ptsname_r to POSIX, with text
"If successful, the ptsname_r( ) function shall return zero.
Otherwise, an error number shall be returned to indicate the
error."
has been accepted for inclusion in POSIX Issue 8.
http://austingroupbugs.net/view.php?id=508
Therefore a portable program should look at the return value from
ptsname_r, NOT the errno value. The current text in the man page
suggests to look at the errno value, which is wrong (because of
musl libc) and not future-proof (because of future POSIX).
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Redundant because this is a Section 3 page, and the
text also describes the implementation in glibc.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Since glibc 2.26, posix_spawn (2) function accepts the
POSIX_SPAWN_SETSID flag. This flag has been accepted by POSIX and
should be added to the next major revision. The current support
can be enabled with _GNU_SOURCE.
Upstream commit in glibc.git:
daeb1fa2e1 [BZ 21340] add support for POSIX_SPAWN_SETSID
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Signed-off-by: Olivier Gayot <olivier.gayot@sigexec.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
The implementation of the fork() step in posix_spawn(2) relies on
either fork(2), vfork(2) or clone(2) depending on the version of
the glibc and the arguments passed to posix_spawn(2).
It is sometimes ambiguous whether, when we are mentioning
"fork(2)", we are referring to the fork() step or the actual
fork(2) syscall.
This patch hopefully avoids the ambiguity by replacing confusing
occurrences by "the xxx() step" where appropriate.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Signed-off-by: Olivier Gayot <olivier.gayot@sigexec.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Added a few lines about POSIX_SPAWN_USEVFORK so that it appears
clearly that since glibc 2.24, the flag has no effect.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Signed-off-by: Olivier Gayot <olivier.gayot@sigexec.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Since glibc 2.24, the use of posix_spawn (2) makes an
unconditional call to clone(CLONE_VM | CLONE_VFORK ...) rather
than relying on fork (2) or vfork (2).
As a consequence, the statements regarding the use of the flag
POSIX_SPAWN_USEVFORK and how the function decides whether it
should use fork (2) or vfork (2) are obsolete since glibc 2.24.
This patch makes a distinction in the manual page between glibc
2.24 and older versions.
Upstream commit in glibc.git:
9ff72da471 posix: New Linux posix_spawn{p} implementation
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Signed-off-by: Olivier Gayot <olivier.gayot@sigexec.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
A year cannot only begin with week number 53 of the previous year but
also with week number 52. Year 2011 is an example for this case, as
can be easily seen with GNU date:
$ date -d "jan 1 2011" "+%c %V %G"
Sat Jan 1 00:00:00 2011 52 2010
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Reported-by: Andrew Micallef <andrew.micallef@live.com.au>
Reported-by: Walter Harms <wharms@bfs.de>
Reviewed-by: Andrew Micallef <andrew.micallef@live.com.au>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
* man3/strftime.3 (%C): Describe the meaning of %EC conversion
specification.
(%E): Mention the concept of "era" in description.
(%O): Mention that alternative format is related to numeric
representation.
(%y): Describe the meaning of %Ey conversion specification.
(%Y): Describe the meaning of %EY conversion specification.
(.SH DESCRIPTION): Mention that the behaviour of %E modifier is governed
by ERA locale element and provide ja_JP locale as an example.
Signed-off-by: Eugene Syromyatnikov <evgsyr@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
As it wasn't clear before where this kind of information can be
obtained from.
* man3/strftime.3 (%a, %A, %b, %B, %c, %p, %r, %x, %X): Add information
about the locale elements that can be used to retrieve the relevant
information using nl_langinfo() library call.
Signed-off-by: Eugene Syromyatnikov <evgsyr@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
In many cases, these don't improve readability, and (when stacked)
they sometimes have the side effect of sometimes forcing text
to be justified within a narrow column range.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
One might be tempted to think that realloc() always requests a new
allocation before moving the contents over (at least in the case
where the new size is bigger than the original). This is not the
case; for example, on my system the following program:
#include <stdlib.h>
#include <stdio.h>
#include <unistd.h>
int main(int argc, char *argv[])
{
void *x = malloc(15);
void *y = malloc(32);
printf("x = %p\n", x);
printf("y = %p\n", y);
printf("usable_size(x) = %lu\n", malloc_usable_size(x));
void *z = realloc(x, 24);
printf("z = %p\n", z);
return 0;
}
prints:
x = 0x1b3a010
y = 0x1b3a030
usable_size(x) = 24
z = 0x1b3a010
Signed-off-by: Vegard Nossum <vegard.nossum@oracle.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Admittedly, the POSIX specification for exit() also uses octal.
However, 0xFF immediately indicates the lowest 8 bits to me
whereas I had to think a bit about the octal mask.
Cowritten-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
From an email by Rich Felker:
It came to my attention while reviewing possible breakage with
move to 64-bit time_t that some applications are dereferencing
data in socket control messages (particularly SCM_TIMESTAMP*)
in-place as the message type, rather than memcpy'ing it to
appropriate storage. This necessarily does not work and is not
supportable if the message contains data with greater alignment
requirement than the header. In particular, on 32-bit archs,
cmsghdr has size 12 and alignment 4, but struct timeval and
timespec may have alignment requirement 8.
I found at least ptpd, socat, and ssmping doing this via Debian
Code Search:
https://sources.debian.org/src/ptpd/2.3.1-debian1-4/src/dep/net.c/?hl=1578#L1578https://sources.debian.org/src/socat/1.7.3.3-2/xio-socket.c/?hl=1839#L1839https://sources.debian.org/src/ssmping/0.9.1-3/ssmpngcl.c/?hl=307#L307
and I suspect there are a good deal more out there. On most archs
they won't break, or will visibly break with SIGBUS, but in theory
it's possible that they silently read wrong data and this might
happen on some older and more tiny-embedded-oriented archs.
I think it's clear to someone who understands alignment and who's
thought about it that applications just can't do this, but it
doesn't seem to be documented, and an example in cmsg(3) even
shows access to int payload via *(int *)CMSG_DATA(cmsg) (of course
int is safe because its alignment is <= header alignment, but this
is not mentioned).
Could we add text, and perhaps change the example, to indicate
that in general memcpy needs to be used to copy the payload
to/from a suitable object?
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
This patch is a minor wording fix in getcwd.3 that changes "In the case getcwd()" to "In the case of getcwd()". This patch should apply cleanly to the master branch of the git repository.
Regards,
Mike Salvatore
From 3b68ad225dbaada2b1b55153dc57807b04531cd6 Mon Sep 17 00:00:00 2001
From: Mike Salvatore <mike.salvatore@canonical.com>
Date: Thu, 16 Jan 2020 16:08:08 -0500
Subject: [PATCH] getcwd.3: wfix
Signed-off-by: Mike Salvatore <mike.salvatore@canonical.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Fixes: d8d701003 ("malloc.3: Since glibc 2.29, realloc() is exposed by
defining _DEFAULT_SOURCE")
Signed-off-by: Petr Vorel <pvorel@suse.cz>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
The glibc implementation of getpt has actually never been setting
O_NOCTTY when opening /dev/ptmx or BSD ptys.
Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
In a recent conversation with Mathieu Desnoyers I was reminded
that we haven't written up anything about how deferred
cancellation and asynchronous signal handlers interact. Mathieu
ran into some of this behaviour and I promised to improve the
documentation in this area to point out the potential pitfall.
Thoughts?
8< --- 8< --- 8<
In pthread_setcancelstate.3, pthreads.7, and signal-safety.7 we
describe that if you have an asynchronous signal nesting over a
deferred cancellation region that any cancellation point in the
signal handler may trigger a cancellation that will behave
as-if it was an asynchronous cancellation. This asynchronous
cancellation may have unexpected effects on the consistency of
the application. Therefore care should be taken with asynchronous
signals and deferred cancellation.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Add entries for the new cache geometry values of the auxiliary
vector that got included in the kernel.
Signed-off-by: Raphael Moreira Zinsly <rzinsly@linux.vnet.ibm.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
This requirement on the first digit with the %e format comes from
the ISO C standard. It ensures that all the digits in the output are
significant and forbids output with a precision less than requested.
Signed-off-by: Vincent Lefevre <vincent@vinc17.net>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
The --export-dynamic linker option is not the only way that main's
global symbols may end up in the dynamic symbol table and thus be
used to satisfy symbol reference in a shared object. A symbol
may also be placed into the dynamic symbol table if ld(1)
notices a dependency in another object during the static link.
Verified by experiment; see previous commit.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
The existing text wrongly implied that symbol look up first
occurred in the object and then in main, and did not mention
whether dependencies of main where used for symbol resolution.
Verified by experiment:
$ cat prog.c
#define _GNU_SOURCE
#include <link.h>
#include <dlfcn.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
void /* A function defined in both main and lib_x1 */
prog_x1(void)
{
printf("Called %s::%s\n", __FILE__, __func__);
}
/* The following function is forced into prog's dynamic symbol table
because of the static link-time reference in lib_m1.so */
void /* A function defined in both main and lib_y1 */
prog_y1_exp(void)
{
printf("Called %s::%s\n", __FILE__, __func__);
}
/* The following function is not forced into prog's dynamic symbol table */
void /* A function defined in both main and lib_y1 */
prog_y1_noexp(void)
{
printf("Called %s::%s\n", __FILE__, __func__);
}
static int
callback(struct dl_phdr_info *info, size_t size, void *data)
{
printf("\tName = %s\n", info->dlpi_name);
return 0;
}
int
main(int argc, char *argv[])
{
void *xHandle, *yHandle;
void (*funcp)(void);
char *err;
xHandle = dlopen("./lib_x1.so", RTLD_NOW | RTLD_GLOBAL);
if (xHandle == NULL) {
fprintf(stderr, "dlopen: %s\n", dlerror());
exit(EXIT_FAILURE);
}
yHandle = dlopen("./lib_y1.so", RTLD_NOW | RTLD_GLOBAL);
if (yHandle == NULL) {
fprintf(stderr, "dlopen: %s\n", dlerror());
exit(EXIT_FAILURE);
}
/* Optionally display the link map() */
if (argc > 1) {
printf("Link map as shown from dl_iterate_phdr() callbacks:\n");
dl_iterate_phdr(callback, NULL);
printf("\n");
}
(void) dlerror(); /* Clear dlerror() */
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wpedantic"
funcp = (void (*)(void)) dlsym(yHandle, "y1_enter");
#pragma GCC diagnostic pop
err = dlerror();
if (err != NULL) {
fprintf(stderr, "dlsym: %s", err);
exit(EXIT_FAILURE);
}
(*funcp)();
exit(EXIT_SUCCESS);
}
$ cat lib_m1.c
#include <stdio.h>
void /* A function defined in both lib_m1 and lib_y1 */
m1_y1(void)
{
printf("Called %s::%s\n", __FILE__, __func__);
}
#if 1
void
dummy(void)
{
extern void prog_y1_exp(void);
prog_y1_exp(); /* Forces prog_y1_exp into prog's dynamic symbol table,
so that it will be visible also to lib_y1.so */
}
#endif
$ cat lib_x1.c
#include <stdio.h>
void /* A function defined in both main and lib_x1 */
prog_x1(void)
{
printf("Called %s::%s\n", __FILE__, __func__);
}
void /* A function defined in both lib_x1 and lib_y1 */
x1_y1(void)
{
printf("Called %s::%s\n", __FILE__, __func__);
}
$ cat lib_y1.c
#include <stdio.h>
void /* A function defined in both lib_x1 and lib_y1 */
x1_y1(void)
{
printf("Called %s::%s\n", __FILE__, __func__);
}
void /* A function defined in both main and lib_y1 */
prog_y1_exp(void)
{
printf("Called %s::%s\n", __FILE__, __func__);
}
void /* A function defined in both lib_m1 and lib_y1 */
m1_y1(void)
{
printf("Called %s::%s\n", __FILE__, __func__);
}
void /* A function defined in both main and lib_y1 */
prog_y1_noexp(void)
{
printf("Called %s::%s\n", __FILE__, __func__);
}
void
y1_enter(void)
{
extern void y2(void);
printf("Called %s\n\n", __func__);
prog_x1();
prog_y1_exp();
prog_y1_noexp();
x1_y1();
m1_y1();
y2();
}
$ cat lib_y2.c
#include <stdio.h>
void
y2(void)
{
printf("Called %s::%s\n", __FILE__, __func__);
}
$ cat Build.sh
#!/bin/sh
CFLAGS="-Wno-implicit-function-declaration -Wl,--no-as-needed"
cc $CFLAGS -g -fPIC -shared -o lib_x1.so lib_x1.c
cc $CFLAGS -g -fPIC -shared -o lib_y2.so lib_y2.c
cc $CFLAGS -g -fPIC -shared -o lib_y1.so lib_y1.c ./lib_y2.so
cc $CFLAGS -g -fPIC -shared -o lib_m1.so lib_m1.c
#ED="-Wl,--export-dynamic"
cc $CFLAGS $ED -Wl,--rpath,$PWD -o prog prog.c -ldl lib_m1.so
$ sh Build.sh
$ ./prog x
Link map as shown from dl_iterate_phdr() callbacks:
Name =
Name = linux-vdso.so.1
Name = /lib64/libdl.so.2
Name = /home/mtk/tlpi/code/shlibs/dlopen_sym_res_expt/lib_m1.so
Name = /lib64/libc.so.6
Name = /lib64/ld-linux-x86-64.so.2
Name = ./lib_x1.so
Name = ./lib_y1.so
Name = ./lib_y2.so
Called y1_enter
Called lib_x1.c::prog_x1
Called prog.c::prog_y1_exp
Called lib_y1.c::prog_y1_noexp
Called lib_x1.c::x1_y1
Called lib_m1.c::m1_y1
Called lib_y2.c::y2
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
My earlier commit was in error:
commit 4a1af09bd1
Author: Michael Kerrisk <mtk.manpages@gmail.com>
Date: Sat Mar 14 21:40:35 2015 +0100
dlopen.3: Amend error in description of dlclose() behavior
-If the reference count drops to zero and no other loaded libraries use
-symbols in it, then the dynamic library is unloaded.
+If the reference count drops to zero,
+then the dynamic library is unloaded.
I doubted the removed text, because it provide little clue about
the scenario. The POSIX dlclose(3) specification actually details
the scenario sufficiently:
Although a dlclose() operation is not required to remove
any functions or data objects from the address space,
neither is an implementation prohibited from doing so.
The only restriction on such a removal is that no func‐
tion nor data object shall be removed to which references
have been relocated, until or unless all such references
are removed. For instance, an executable object file that
had been loaded with a dlopen() operation specifying the
RTLD_GLOBAL flag might provide a target for dynamic relo‐
cations performed in the processing of other relocatable
objects—in such environments, an application may assume
that no relocation, once made, shall be undone or remade
unless the executable object file containing the relo‐
cated object has itself been removed.
Verified by experiment:
$ cat openlibs.c # Test program
int
main(int argc, char *argv[])
{
void *libHandle[MAX_LIBS];
int lcnt;
if (argc < 2) {
fprintf(stderr, "Usage: %s lib-path...\n", argv[0]);
exit(EXIT_FAILURE);
}
lcnt = 0;
for (int j = 1; j < argc; j++) {
if (argv[j][0] != '-') {
if (lcnt >= MAX_LIBS) {
fprintf(stderr, "Too many libraries (limit: %d)\n", MAX_LIBS);
exit(EXIT_FAILURE);
}
printf("[%d] Opening %s\n", lcnt, argv[j]);
libHandle[lcnt] = dlopen(argv[j], RTLD_NOW | RTLD_GLOBAL);
if (libHandle[lcnt] == NULL) {
fprintf(stderr, "dlopen: %s\n", dlerror());
exit(EXIT_FAILURE);
}
lcnt++;
} else { /* "-N" closes the Nth handle */
int i = atoi(&argv[j][1]);
printf("Closing handle %d\n", i);
dlclose(libHandle[i]);
}
sleep(1);
printf("\n");
}
printf("Program about to exit\n");
exit(EXIT_SUCCESS);
}
$ cat lib_x1.c
void x1_func(void) { printf("Hello world\n"); }
__attribute__((constructor)) void x1_cstor(void)
{ printf("Called %s\n", __FUNCTION__); }
__attribute__((destructor)) void x1_dstor(void)
{ printf("Called %s\n", __FUNCTION__); }
$ cat lib_y1.c
void y1_func(void) { printf("Hello world\n"); }
__attribute__((constructor)) void y1_cstor(void)
{ printf("Called %s\n", __FUNCTION__); }
__attribute__((destructor)) void y1_dstor(void)
{ printf("Called %s\n", __FUNCTION__); }
static void testref(void) {
/* The following reference, to a symbol in lib_x1.so shows that
RTLD_GLOBAL may pin a library when it might otherwise have been
released with dlclose() */
extern void x1_func(void);
x1_func();
}
$ cc -shared -fPIC -o lib_x1.so lib_x1.c
$ cc -shared -fPIC -o lib_y1.so lib_y1.c
$ cc -o openlibs openlibs.c -ldl
$ LD_LIBRARY_PATH=. ./openlibs lib_x1.so lib_y1.so -0 -1
[0] Opening lib_x1.so
Called x1_cstor
[1] Opening lib_y1.so
Called y1_cstor
Closing handle 0
Closing handle 1
Called y1_dstor
Called x1_dstor
Program about to exit
<end program output>
Note that x1_dstor was called only when handle 1 (lib_y1.so) was closed.
But, if we edit lib_y1 to remove the reference to x1_func(), things are
different:
$ cat lib_y1.c # After editing
void y1_func(void) { printf("Hello world\n"); }
__attribute__((constructor)) void y1_cstor(void)
{ printf("Called %s\n", __FUNCTION__); }
__attribute__((destructor)) void y1_dstor(void)
{ printf("Called %s\n", __FUNCTION__); }
static void testref(void) {
// extern void x1_func(void);
// x1_func();
}
$ cc -shared -fPIC -o lib_y1.so lib_y1.c
$ LD_LIBRARY_PATH=. ./openlibs lib_x1.so lib_y1.so -0 -1
[0] Opening lib_x1.so
Called x1_cstor
[1] Opening lib_y1.so
Called y1_cstor
Closing handle 0
Called x1_dstor
Closing handle 1
Called y1_dstor
Program about to exit
<end program output>
This time, x1_dstor was called when handle 0 (lib_x1.so) was closed.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Here's a program for doing experiments:
/* on_expt_scope_expt.c
(C) Michael Kerrisk, 2019, Licensed GNU GPLv2+
*/
char *tos;
static void
exitFunc(int status, void *p)
{
int efloc;
int *xp = (int *) p;
printf("====== Entered exit handler\n");
printf("&efloc = %p (0x%llx)\n",
(void *) &efloc, (long long) (tos - (char *) &efloc));
printf("xp = %p (value: %d)\n", (void *) xp, *xp);
if (*xp != INIT_VALUE)
printf("It looks like the variable passed to the exit handler "
"has gone out of scope\n");
/* Produce a core dump, which we can examine with GDB to look at the
frames on the stack, if desired */
printf("===\n");
printf("About to abort\n");
abort();
}
static void
recur(int lev, int *xp)
{
int rloc;
int big[65536-12]; /* 12*4 == 48 other bytes allocated on
this stack frame */
tos = (char *) &rloc;
big[0] = lev;
big[0]++;
printf("&rloc = %p (%d) (%d)\n", (void *) &rloc, lev, *xp);
if (lev > 1)
recur(lev - 1, xp);
else {
printf("exit() from recur()\n");
exit(EXIT_SUCCESS);
}
}
int
main(int argc, char *argv[])
{
int lev;
int *xp;
int xx;
if (argc < 2) {
fprintf(stderr, "Usage: %s {s|h} [how]\n", argv[0]);
fprintf(stderr, "\ts => exitFunc() arg is in main() stack\n");
fprintf(stderr, "\th => exitFunc() arg is allocated on heapn");
fprintf(stderr, "\tIf 'how' is not present, then return from main()\n");
fprintf(stderr, "\tIf 'how' is 0, then exit() from main()\n");
fprintf(stderr, "\tIf 'how' is > 0, then make 'how' recursive "
"function calls, and then exit()\n");
exit(EXIT_FAILURE);
}
tos = (char *) &xp;
if (argv[1][0] == 'h') {
xp = malloc(sizeof(int));
if (xp == NULL) {
perror("malloc");
exit(EXIT_FAILURE);
}
printf("Argument for exitFunc() is allocated on heap\n");
} else {
xp = &xx;
printf("Argument for exitFunc() is allocated on stack in main()\n");
}
*xp = INIT_VALUE;
printf("xp = %p (value: %d)\n", (void *) xp, *xp);
printf("===\n");
on_exit(exitFunc, xp);
if (argc == 2) {
printf("return from main\n");
return 0;
}
lev = atoi(argv[2]);
if (lev < 1) {
printf("Calling exit() from main\n");
exit(EXIT_SUCCESS);
} else {
recur(lev, xp);
}
}
Reported-by: Sami Kerola <kerolasa@iki.fi>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>