getrlimit.2: Various minor wording fixes

Mainly: use complete sentence to introduce each limit.

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

unix.7: tfix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

recv.2: SEE ALSO: add ip(7), ipv6(7), tcp(7), udp(7), unix(7)

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

send.2: SEE ALSO: add ipv6(7), socket(7), unix(7)

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

capabilities.7: tfix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

capabilities.7: Add another case for CAP_DAC_READ_SEARCH

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

link.2: When using linkat() AT_EMPTY_PATH, 'olddirfd' must not be a directory

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

capabilities.7: Add some more operations governed by CAP_SYS_ADMIN

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

capabilities.7: Add a few more operations covered by CAP_SYS_ADMIN

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

scandir.3: Minor reworking of text for example program

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

scandir.3: Improve logic of the example program

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

scandir.3: Fix errors in example program

See http://bugs.debian.org/848231.

Reported-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

prctl.2: wfix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

prctl.2: wfix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

prctl.2: Clarify that the ambient capability set is per-thread

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

execve.2: File capabilities can be ignored for the same reasons as set-UID/set-GID

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

execve.2: wfix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

capabilities.7: Refer to execve(2) for the reasons that file capabilities may be ignored

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

capabilities.7: ffix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

execve.2: The 'no_new_privs' bit inhibits transformations of the effective IDs

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

execve.2: Rework text describing when effective IDs aren't transformed by execve()

Mainly in preparation for addition of information about
no_new_privs bit.

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

prctl.2: The no_new_privs setting is per-thread (not per-process)

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

prctl.2: Add reference to seccomp(2) on discussion of PR_SET_NO_NEW_PRIVS

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

prctl.2: Mention /proc/PID/status 'NoNewPrivs' field

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

proc.5: Document /proc/PID/status 'NoNewPrivs' filed

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

namespaces.7: tfix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

filesystems.5: SEE ALSO: add btrfs(5), nfs(5), tmpfs(5)

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

random.4: Note that entropY-avail will be a number in the range 0..4096

Reported-by: Jon Jensen <jon@endpoint.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

random.7: SEE ALSO: add getauxval(3)

A small hint to the reader that some random bytes arrive
in the auxiliary vector.

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

sleep.3: Note that sleep() is implemented via nanosleep(2)

See https://bugzilla.kernel.org/show_bug.cgi?id=73371.

Reported-by: Mike Frysiner <vapier@gentoo.org>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

sleep.3: Convert BUGS text to "Portability notes" subsection

The existing text is not a bug, as such.

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

sleep.3: Note that sleep() sleeps for a real-time number of seconds

See https://bugzilla.kernel.org/show_bug.cgi?id=73371.

Reported-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

sleep.3: DESCRIPTION: minor reworking

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

sleep.3: wfix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

clock_getres.2: Note POSIX.1 requirements re relative time services and CLOCK_REALTIME

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

exec.3: execl() and execle() were not async-signal-safe before glibc 2.24

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

aio_suspend.3: Note that the glibc implementation is not async-signal-safe

See https://sourceware.org/bugzilla/show_bug.cgi?id=13172

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

sem_post.3, setjmp.3: SEE ALSO: add signal-safety(7)

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fork.2, sigaction.2, signal.2: Amend cross references to point to signal-safety(7)

The discussion of async-signal-safe functions has moved
to a new page.

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

signal-safety.7: Note async-signal-safety problems caused by pthread_atfork()

See https://bugzilla.kernel.org/show_bug.cgi?id=25292

And there will likely be future changes as well.
Quoting http://www.opengroup.org/austin/aardvark/latest/xshbug3.txt:

    COMMENT                          Enhancement Request Number 15
    rajani.g.k:xxxxxx               Defect in XSH 2.4.3 (rdvk#  6)
    {GKRFORK012009}                   Thu, 8 Jan 2009 07:41:10 GMT

    [...]

    As per this section, XSH P1529, L49389-49402, it is possible
    that multithreaded libraries could be used by single threaded
    applications. In which case, atfork handlers are essential for
    the libraries to protect their internal state during fork. As
    explained further P1530, L49403-49404, pthread_atfork
    functions are mainly required to acquire/release mutex locks,
    for protecting the applications/libraries from fork() calls.
    C-library needs to as well have an atfork handler which
    acquires all the required locks to protect its memory state
    across fork().

    The acquire/release mutex calls themselves are aync-signal
    unsafe operations. Use of them makes pthread_atfork handlers
    async-signal unsafe which in turn makes fork() async-signal
    unsafe when called by an application which is multi threaded,
    or which is linked to a library which is multi threaded.

    Action:

    Need clarification with respect to
    1. Is it correct to list fork as an async-signal safe
    interface, in a multi threaded scenario?

    2. Can the implementation be allowed to not call the atfor
    handlers, when fork is called from a signal handler? If the
    atfork handlers are not going to be called when fork is called
    in the signal handler, then they can not be called, even if
    fork is called in the newly created child before exec.

    3. If only async-signal safe functions are to be called from
    pthread_atfork handlers, then how will multi-threaded librarie
    protect themselves by the fork calls, made by single threaded
    applications linked to them?

Reported-by: KASAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

signal-safety.7: Note glibc deviations from POSIX requirements

See https://bugzilla.kernel.org/show_bug.cgi?id=25292

Reported-by: KASAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

signal-safety.7, signal.7: Migrate discussion of async-signal-safe functions to new page

Also:
* add more detail on changes across standards
* provide proper section cross references in function references

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

Changes.old: tfix in 4.09 changes

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

Start of man-pages-4.10: updating Changes and Changes.old

Start of man-pages-4.10: updating .Announce and .lsm files

Start of man-pages-4.10: renaming .Announce and .lsm files

Ready for 4.09

Removed trailing white space at end of lines

Changes: Expanded tabs

bind.2, chmod.2, chown.2, chroot.2, clock_getres.2, clone.2, connect.2, dup.2, fallocate.2, get_mempolicy.2, getpeername.2, getpriority.2, getsockname.2, getsockopt.2, gettimeofday.2, ioctl_ficlonerange.2, ioctl_fideduperange.2, kill.2, mbind.2, mmap.2, mount.2, mprotect.2, nfsservctl.2, nice.2, open.2, perf_event_open.2, pipe.2, pkey_alloc.2, prctl.2, ptrace.2, quotactl.2, remap_file_pages.2, sched_setscheduler.2, set_mempolicy.2, signal.2, signalfd.2, swapon.2, sync_file_range.2, syscalls.2, timer_create.2, timerfd_create.2, utime.2, utimensat.2, wait.2, atof.3, ctime.3, errno.3, fclose.3, fflush.3, insque.3, malloc_get_state.3, mallopt.3, mbsnrtowcs.3, mq_close.3, mq_open.3, mq_receive.3, mq_send.3, printf.3, pthread_attr_init.3, pthread_create.3, pthread_setaffinity_np.3, ptsname.3, remainder.3, strtod.3, tgamma.3, timegm.3, tmpnam.3, ttyname.3, console_ioctl.4, elf.5, filesystems.5, proc.5, utmp.5, capabilities.7, cgroups.7, credentials.7, ddp.7, feature_test_macros.7, fifo.7, inotify.7, libc.7, mount_namespaces.7, namespaces.7, netlink.7, pid_namespaces.7, pkeys.7, shm_overview.7, standards.7, uri.7, user_namespaces.7: tstamp

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

Changes: Ready for 4.09

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: srcfix: FIXME

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: ffix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

filesystems.5: SEE ALSO: add fuse(4)

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: Tweaks

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: Address FIXMEs

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: Remove FUSE_* constant values

No need to show the values of the FUSE_* constants, I believe.

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: Add list of undocumented messages

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: Add list of FOPEN_* flags

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: srcfix: FIXME

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: Add SEE ALSO section

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: Add CONFORMING TO section

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: Repair ENODEV description

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: srcfix: FIXME

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: srcfix: FIXME

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: Small rewording in FUSE_INIT

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: Repair wording in EINVAL error text

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: ffix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: fuse_entry_out: rework discussion of uniqueness of nodeid + generation

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: Various minor edits to Keno Fischer's patch

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: srcfix: rewrap source lines

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: Kill information about FUSE header license

I'd say this is fair use (many other man pages do the same).
So, no need to clutter the man page source with this information.

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fuse.4: New page describing /dev/fuse

This is my writeup of a basic description of /dev/fuse after
playing with it for a few hours today. It is of course woefully
incomplete, and since I neither have a use case nor am working
on this code, I will not be in a position to expand it in the
near future. However, I'm hoping this could still serve as a
handy reference for others looking at this interface.

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

namespaces.7: Adjust example program to show device major and minor numbers

Reported-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

namespaces.7: Comparisons between fstat()ed files should be on st_dev+st_ino

The FDs returned by NS_GET_USERNS and NS_GET_PAREENT must be
tested by comparing to both the 'st_dev' and 'st_ino' fields
returned by fstat(2).

Reported-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

namespaces.7: Add ENOTTY error() for ioctl namespace operations

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

namespaces.7: ffix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

namespaces.7: Tweaks to text on ioctl() operations

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

mount_namespaces.7: Minor clean-up (note a forward reference in the text)

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

mount_namespaces.7, user_namespaces.7: Migrate subsection on mount restrictions to mount_namespaces(7)

This section material in the user_namespaces(7) page was written
before the creation of the mount_namespaces(7) manual page.
Nowadays, this material properly belongs in the newer page.

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

mount_namespaces.7: SEE ALSO: add user_namespaces(7)

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

user_namespaces.7: Add reference to namespaces(7) for NS_GET_USERNS operation

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

user_namespaces.7: Add reference to namespaces(7) for NS_GET_PARENT operation

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

pid_namespaces.7: Refer to namespaces(7) for information about NS_GET_PARENT

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

namespaces.7: Document the NS_GET_USERNS and NS_GET_PARENT ioctl() operations

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

prctl.2: Minor tweaks to Eugene Syromyatnikov's patch

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

prctl.2: Some additional details regarding PR_GET_UNALIGNED operation

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

prctl.2: Minor tweaks to Eugene Syromyatnikov's patch

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

prctl.2: Note the output buffer size for PR_GET_TID_ADDRESS operation on x32/n32

Add note regarding the size of the output user space buffer
for PR_GET_TID_ADDRESS operation on x32/n32.

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

execve.2: tfix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

pid_namespaces.7: tfix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

user_namespaces.7: ffix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

namespaces.7: ffix

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

setjmp.3: _BSD_SOURCE must be *explicitly* defined to get BSD setjmp() semantics

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

setjmp.3: Repair changes to description of pre-2.19 glibc default behavior

The changes in commit 2ae4c26dec0497 were a mistake.
The System V behavior was always the default for setjmp(3).

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

feature_test_macros.7: Further tweaks to _REENTRANT description

Reviewed-by: Zack Weinberg <zackw@panix.com>
Reported-by: Zack Weinberg <zackw@panix.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

feature_test_macros.7: Note that "cc -pthread" defines _REENTRANT

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

feature_test_macros.7: Improve the text on _REENTRANT/_THREAD_SAFE deprecation

[mtk] I did a little code spelunking and found the following:

1. In glibc 1.09 (tagged 1995-03-02 in the git history),
   __USE_REENTRANT, _THREAD_SAFE, and _REENTRANT do not appear.

2. In glibc-1.93 (tagged 1996-08-29 in the git history),
   __USE_REENTRANT governs the exposure of some "_r()"
   functions from about a dozen header files.  However, it is
   defined in <features.h> via

        #if defined (__USE_GNU) || defined (__USE_MISC)
        #define __USE_REENTRANT 1
        #endif

   _REENTRANT and _THREAD_SAFE solely govern declarations in
   <stdio.h>, where they expose declarations of a few "unlocked"
   stdio functions and use #define to redirect a few stdio
   function names to "locked" versions.

3. THREAD_SAFE and _REENTRANT first appear in the git logs
   1996-05-09.

4. About 9 months later, glibc 2.0.1 arrives on 1997-02-04
   (timestamp and tarball taken from
   https://ftp.gnu.org/gnu/libc/, since there is no tag in the
   git history; casual inspection of the logs suggests the
   glibc 2.0 release was about a week earlier.

   By now we have the following in <features.h>:

        #if defined _REENTRANT || defined _THREAD_SAFE
        #define __USE_REENTRANT 1
        #endif

   And _THREAD_SAFE, and _REENTRANT do not appear appear in
   other headers. However, by now, __USE_REENTRANT governs only
   the declarations of tmpnam_r() and getlogin_r()

In other words, the window of time where _REENTRANT and
_THREAD_SAFE did anything much in glibc was quite short, IIUC.

Cowritten-by: Zack Weinberg <zackw@panix.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

getlogin.3: Remove deprecated _REENTRANT from FTM requirements for getlogin_r()

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

feature_test_macros.7: Note that _REENTRANT and _THREAD_SAFE are now deprecated

Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>