cgroup_namespaces.7: Remove redundant use of 'sh -c' in shell session

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

atoi.3: tfix

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

getlogin.3: tfix

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

ttyslot.3: tfix

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

keyctl.2: tfix

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

cgroup_namespaces.7: wfix

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

cgroups.7: Complete partial sentence re kernel boot options and 'nsdelegate'

The intended text was hidden elsewhere in the source of the
page as a comment.

https://bugzilla.kernel.org/show_bug.cgi?id=201029

Reported-by: Mike Weilgart <mike.weilgart@verticalsysadmin.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

sched.7: SEE ALSO: add ps(1) and top(1)

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

cgroups.7: Minor wording fix

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

cgroups.7: wfix: use "threads" consistently

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

cgroups.7: Add more detail on v2 'cpu' controller and realtime threads

Explicitly note the scheduling policies that are relevant for the
v2 'cpu' controller.

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

cgroups.7: Minor wording fix

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

cgroups.7: Rework discussion of writing to cgroup.type file

In particular, it is possible to write "threaded" to a
cgroup.type file if the current type is "domain threaded".
Previously, the text had implied that this was not possible.
Verified by experiment on Linux 4.15 and 4.19-rc.

Reported-by: Leah Hanson <lhanson@pivotal.io>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

pid_namespaces.7: Note a detail of /proc/PID/ns/pid_for_children behavior

After clone(CLONE_NEWPID), /proc/PID/ns/pid_for_children is empty
until the first child is created. Verified by experiment.

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

pid_namespaces.7: Note that a process can do unshare(CLONE_NEWPID) only once

(See the recent commit to the unshare(2) manual page.)

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

xcrypt.3: Warn folks not to use these functions

There is not an acceptable reason to use these functions ever in
new code.  For example, just observe the implementation of the
KDF:

/*
 * Turn password into DES key
 */
void
passwd2des_internal (char *pw, char *key)
{
  int i;

  memset (key, 0, 8);
  for (i = 0; *pw && i < 8; ++i)
    key[i] ^= *pw++ << 1;

  des_setparity (key);
}

This kind of nonsense isn't okay in the year 2017. Therefore, we
enlighten our poor users.

[Note from mtk: I think Jason knows that of which he talks.]

Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

capabilities.7: Update URL for location of POSIX.1e draft standard

Reported-by: Allison Randal <allison@lohutok.net>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

unshare.2: Minor tweaks to Tycho Andersen's patch

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

unshare.2: Note EINVAL when unsharing pid ns twice

The kernel doesn't allow unsharing a pid NS if it has previously been
unshared, per this check in copy_pid_ns:

        if (task_active_pid_ns(current) != old_ns)
                return ERR_PTR(-EINVAL);

so let's note that.

Signed-off-by: Tycho Andersen <tycho@tycho.ws>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

stat.2: ERRORS: ENOENT can occur where a path component is a dangling symlink

See https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=909789

Reported-by: Alessandro Vesely <vesely@tana.it>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

stat.2: Minor fix: split distinct ENOENT cases into separate list entries

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

open.2: Minor fix: split distinct ENOENT cases into separate list entries

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

sched.7: SEE ALSO: add chcpu(1), lscpu(1)

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

cgroups.7: Minor clarification to remove possible ambiguity

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

epoll_ctl.2: Minor rewording: split a long paragraph

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

malloc.3: Since glibc 2.29, realloc() is exposed by defining _DEFAULT_SOURCE

Info gleaned from glilic NEWS file.

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

capabilities.7: Ambient capabilities do not trigger secure-execution mode

Reported-by: Pierre Chifflier <pollux@debian.org>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

fcntl.2: Actual pipe capacity may in practice be less than nominal capacity

The number of bytes that can be written to the pipe may be less
(sometimes substantially less) than the nominal capacity. This
was confirmed with some testing. For example, when writing
4097-byte blocks to a pipe with 65536 byte capacity, only 45066
bytes could be written (i.e., 20470 bytes less than 65536).

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

clone.2: Clarify the discussion of threads and signals

And explicitly introduce the terms "process-directed" and
"thread-directed" signals.

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

signal.7: SEE ALSO: add clone(2)

Because of the discussion of trheads and signals in clone(2)/

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

clone.2: Rework discussion of threads and signals

The discussion is phrased in terms of signals send using kill(2),
but applies equally to a signal sent by the kernel.

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

memfd_create.2: _GNU_SOURCE is required

The memfd_create function and its corresponding constants have
required _GNU_SOURCE for as long as they've been in glibc.

Signed-off-by: Joseph C. Sible <josephcsible@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

cgroups.7: tfix

Reported-by: Mike Weilgart <mike.weilgart@verticalsysadmin.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

mount.2: tfix

Signed-off-by: Jakub Wilk <jwilk@jwilk.net>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

connect.2: tfix

Signed-off-by: Jakub Wilk <jwilk@jwilk.net>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

reboot.2: Johannes Liebermann <johanan.liebermann@gmail.com>

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

fnmatch.3: Clarify "shell wildcard pattern"

There's a detailed explanation in glob(7), so reuse the same text
glob(3) uses to redirect the reader, rather than inlining a short
explanation.

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

malloc.3: Add reference to glibc MallocInternals wiki

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

socket.2: Minor tweaks to Tobias Klauser's patch

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

socket.2: Document AF_XDP

Document AF_XDP added in Linux 4.18.

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

connect.2: srcfix: rewrap source lines

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

connect.2: Document error semantics of nonblocking UNIX domain sockets

connect(2) on a nonblocking UNIX domain socket when the receive
queue is full results in EAGAIN [1]. This is unlike other
connection-based socket families that return EINPROGRESS as
already documented.

mtk: confirmed with some light testing. And in
net/unix/af_unix.c::unix_stream_connect(), we have:

        if (unix_recvq_full(other)) {
                err = -EAGAIN;
                if (!timeo)
                        goto out_unlock;

Signed-off-by: Benjamin Peterson <benjamin@python.org>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

mmap.2: Minor rewording

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

mmap.2: Explicitly state that the fd can be closed

I regularly see excessive fd usage bugs (or even leaks) caused by
people who think they need to keep the fd open as long as the
mapping exists.

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

pivot_root.2: Document EINVAL if root is rootfs

Per the comment on the pivot_root syscall in fs/namespace.c:

    Also, the current root cannot be on the 'rootfs'
    (initial ramfs) filesystem.  See
    Documentation/filesystems/ramfs-rootfs-initramfs.txt
    for alternatives in this situation.

Signed-off-by: Joseph C. Sible <josephcsible@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

trunc.3: Make the description a little clearer

The double negative is a little confusing, but required. But try
to make the semantics a little clearer in the wording (which is
now closer to the wording in the C standard).

Reported-by: Eric Benton <erbenton@comcast.net>
Reported-by: G. Branden Robinson <g.branden.robinson@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

trunc.3: Emphasize that the return value is a floating-point number

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

wait.2: Add some cross references to core(5)

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

sigaltstack.2: tfix

MINSTKSZ is not defined anywhere, MINSIGSTKSZ seems valid instead.

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

core.5: Add cross reference to vdso(7) where "virtaul DSO" is mentioned

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

mount.2: Add MS_STRICTATIME to list of flags that can be used in remount

Reported-by: Simone Piccardi <piccardi@truelite.it>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

mount.2: Clarify that per-superblock flags are shared during remount

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

mount.2: Attempts to change MS_SILENT setting during remount are silently ignored

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

mount.2: Clearly distinguish per-mount-point vs per-superblock mount flags

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

mount.2: wfix

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

mount.2: MS_SILENT is ignored when changing propagation type

MS_SILENT can be specified when changing propagation type,
but is ignored, as far as I can see from reading the code.
(The flags are passed to do_change_type(), which, as well
as the propagation flags, allows MS_REC and MS_SILENT
(in flags_to_propagation_type()), but does noting with
MS_SILENT. (Linux 4.17 source)

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

mount.2: wfix

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

mount.2: Minor rewording

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

mount.2: Remove crufty sentence about MS_BIND + MS_REMOUNT

This sentence is left over from an earlier rewrite of the text,
and the relevant details are covered a few paragraphs later.

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

mount.2: srcfix

mount.2: srcfix: update FIXME

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

mount.2: tfix

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

mount.2: SEE ALSO: add ioctl_iflags(2)

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

mount.2: srcfix: add FIXME

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

mount.2: Mandatory locking also now requires CONFIG_MANDATORY_FILE_LOCKING

This is so since Linux 4.5. See the definition of may_mandlock()
in fs/namespace.c.

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

mount.2: ffix

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

mount.2: ffix

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

mount.2: wfix

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

rename.2: Add feature test macro for renameat2()

The glibc wrapper for renameat2() was added in glibc 2.28 and
requires _GNU_SOURCE.

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

proc.5: tfix

Remove duplicated word.

Signed-off-by: Jakub Wilk <jwilk@jwilk.net>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

strcpy.3: Use "destination" consistently (instead of "target" sometimes)

Reported-by: Frank Theile <ftheile@grundfos.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

proc.5: ffix

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

proc.5: Reword /proc/PID/fdinfo timerfd field descriptions as a hanging list

And a few other minor tweaks to the text.

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

proc.5: Add a few details on /rpoc/PID/fdinfo timerfd

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

proc.5: Minor tweaks to Lukas Wermeister's patch

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

proc.5: Document fdinfo format for timerfd

Extended information for timerfd file descriptors in
/proc/[pid]/fdinfo was added in commit af9c4957cf21 ("timerfd:
Implement show_fdinfo method", 2014-07-16), to support
checkpoint/restore for such file descriptors (see also the
TFD_IOC_SET_TICKS ioctl which is documented in timerfd_create.2).

Signed-off-by: Lucas Werkmeister <mail@lucaswerkmeister.de>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

user_namespaces.7: wfix

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

io_submit.2: Minor tweaks to Adam Manzanares' patch

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

io_submit.2: ffix

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

io_submit.2: Document IOCB_FLAG_IOPRIO

The newly added IOCB_FLAG_IOPRIO aio_flag introduces new behaviors
and return values.

The details of this new feature are posted here:
https://lkml.org/lkml/2018/5/22/809

Signed-off-by: Adam Manzanares <adam.manzanares@wdc.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

proc.5: srcfix: rewrap source lines

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

proc.5: Describe ambiguities in /proc/<pid>/maps

Pathname escaping is not done properly in /proc/<pid>/maps;
because of this, different pathnames may appear the same
(verified by experiment and reading the source code).

Further details from Elvira about the relevant location in
the kernel code:

    show_map_vma() from fs/proc/task_mmu.c uses seq_file_path()
    from fs/seq_file.c to print the dentry name, which in turn
    calls seq_path() from the same file.  seq_path() uses
    d_path() from fs/d_path.c to get the path name; this is
    where the " (deleted)" part comes from. This is followed by
    mangling the string with mangle_path() (fs/seq_file.c); this
    function only replaces those characters that were supplied
    in the "esc" argument and does not bother with escaping
    anything else ('\\', for example).  The value of this
    argument comes without modifications from the initial call
    of seq_file_path() by show_map_vma(), and that is "\n".

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

perf_event_open.2: Document the PERF_EVENT_IOC_QUERY_BPF ioctl

The PERF_EVENT_IOC_QUERY_BPF ioctl was introduced in Linux 4.16.

Signed-off-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

perf_event_open.2: wfix

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

perf_event_open.2: Document the PERF_EVENT_IOC_MODIFY_ATTRIBUTES ioctl

The PERF_EVENT_IOC_MODIFY_ATTRIBUTES ioctl was introduced in
Linux 4.17.  It currently only works on breakpoint events.

Signed-off-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

perf_event_open.2: Document the PERF_EVENT_IOC_PAUSE_OUTPUT ioctl

The PERF_EVENT_IOC_PAUSE_OUTPUT ioctl was introduced in Linux 4.7.

I've have this patch for a long time, I apologize for the delay
in getting it submitted.  I've made some minor changes to the
original patch proposed by Wang Nan.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Reviewed-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

perf_event_open.2: Clarify exclude_idle

It turns out no one is really sure what the perf_event_open.2
exclude_idle field is supposed to do, and a recent thread on the
linux-kernel list:
[RFC] perf/core: what is exclude_idle supposed to do
did not really clarify things.

I think the following adjustment to the page clarifies things
at least a little.

Signed-off-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

perf_event_open.2: Fix prctl behavior description

Some discussion on the linux-perf-users list has turned up that
the perf_event_open.2 description of how
PR_TASK_PERF_EVENTS_ENABLE / PR_TASK_PERF_EVENTS_DISABLE prctl()
works is misleading.

The descriptions were based on the tools/perf/design.txt document
which describes behavior that was removed in 082ff5a2767a06 (prior
to 2.6.31, the first release with perf_event_open support).

I have written some tests in my perf_event_tests testsuite that
verifies the behavior of prctl() in this case.

Signed-off-by: Vince Weaver <vincent.weaver@maine.edu>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

namespaces.7: tfix

Signed-off-by: Jakub Wilk <jwilk@jwilk.net>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

bpf.2: srcfix: fix typo in kernel version revied for JIT compiler source

Reported-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

bpf.2: Minor tweaks to Tobias Klauser's patch

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

bpf.2: Update JIT support list for Linux 4.18

JIT support for x86-32 was during the Linux 4.18 release cycle.
Also correct the entry for MIPS (only MIPS64 is supported).

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

proc.5: tfix

Reported-by: Stan Schwertly <stan@schwertly.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

vmsplice.2: Minor tweaks to Andrei's patch

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

vmsplice.2: Note that vmsplice can splice pages from pipe to memory

The man page notes that vmsplice() can splice pages from memory
to a pipe, but it can work in the other direction as well.

Cc: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Andrei Vagin <avagin@openvz.org>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

resolv.conf.5: Document no-reload (RES_NPRELOAD) option

Signed-off-by: Nikola Forró <nforro@redhat.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

namespaces.7: tfix

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

rename.2: renameat2() now has a glibc wrapper; remove mention of syscall(2)

The glibc wrapper was added in glibc 2.28

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

statx.2: statx() now has a glibc wrapper; remove mention of syscall(2)

The glibc wrapper was added in glibc 2.28

Signed-off-by: Tobias Klauser <tklauser@distanz.ch>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>

strcmp.3: tfix

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