The Linux kernel uses a long as the return type for this syscall.
As glibc provides no wrapper, use the same types the kernel uses.
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
In the kernel sources (kernel/fork.c::copy_process()), we have:
/*
* sigaltstack should be cleared when sharing the same VM
*/
if ((clone_flags & (CLONE_VM|CLONE_VFORK)) == CLONE_VM)
sas_ss_reset(p);
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
See https://bugzilla.kernel.org/show_bug.cgi?id=12665.
The fix by Thomas Gleixner was in kernel commit
78c9c4dfbf8c04883941445a195276bb4bb92c76.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Linux kernel commit 2a36ab717e8fe678d98f81c14a0b124712719840
(part of 5.10 release) changed sys_membarrier prototype/parameters
and added two new commands [MEMBARRIER_CMD_PRIVATE_EXPEDITED_RSEQ
and MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED_RSEQ]. This
man-pages patch reflects these changes, by mostly copying comments
from the kernel patch into the man-page ([Peter Oskolkov] was also
the author of the kernel change).
[mtk: commit message tweaked]
Signed-off-by: Peter Oskolkov <posk@google.com>
Cowritten-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Linux uses aio_context_t for these syscalls,
and it's the type provided by <linux/aio_abi.h>.
Use it in the SYNOPSIS.
libaio uses 'io_context_t', but that difference is already noted
in NOTES.
[mtk: patch slightly tweaked]
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
The Linux kernel uses long as the return type for this syscall.
As glibc provides no wrapper, use the same type the kernel uses.
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
That comment wrapped on an 80-column terminal.
Divide it into two lines.
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
The Linux kernel uses the following:
kernel/futex.c:3778:
SYSCALL_DEFINE6(futex, u32 __user *, uaddr, int, op, u32, val,
struct __kernel_timespec __user *, utime, u32 __user *, uaddr2,
u32, val3)
Since there is no glibc wrapper, use the same types the kernel uses.
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
[mtk: Applied patch manually]
getdents():
This function has no glibc wrapper.
As such, we should use the same types the Linux kernel uses:
Use 'long' as the return type.
getdents64():
The glibc wrapper uses:
ssize_t getdents64(int, void *, size_t);
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
The text in NOTES doesn't really relate specifically to
the #include, so remove the comment on the #include.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Extend this page with the information about CAP_PERFMON capability
designed to secure performance monitoring and observability
operation in a system according to the principle of least
privilege [1] (POSIX IEEE 1003.1e, 2.2.2.39).
[1] https://sites.google.com/site/fullycapable/, posix_1003.1e-990310.pdf
Signed-off-by: Alexey Budankov <alexey.budankov@linux.intel.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Time namespaces were added in kernel 5.6, but setns() support
for time namespaces was added only starting with kernel 5.8:
commit 76c12881a38aaa83e1eb4ce2fada36c3a732bad4
Author: Christian Brauner <christian.brauner@ubuntu.com>
Date: Mon Jul 6 17:49:11 2020 +0200
nsproxy: support CLONE_NEWTIME with setns()
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Highlight to the reader that if another filter returns a
higher-precedence action value, then the ptracer will not
be notified.
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Due to a userspace breakage, commit 1251201c0d34 ("sched/core: Fix
uclamp ABI bug, clean up and robustify sched_read_attr() ABI logic
and code") changed the semantics of sched_getattr(2) when the
userspace struct is smaller than the kernel struct. Now, any
trailing non-zero data in the kernel structure is ignored when
copying to userspace. We also document the original error code
correctly (it was EFBIG not E2BIG) in the BUGS section.
Ref: 1251201c0d34 ("sched/core: Fix uclamp ABI bug, clean up and
robustify sched_read_attr() ABI logic and code")
Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Around the text:
"Feature Test Macro Requirements for glibc..."
replace ".in -4n/.in" with ".RS -4/.RE".
The latter form is more idiomatic use of man macros.
The nroff output is unchanged.
Reported-by: G. Branden Robinson <g.branden.robinson@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
From the email discussion:
> Hi Alex,
>
> On 9/25/20 9:31 AM, Alejandro Colomar wrote:
>> Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
>> ---
>> man2/seccomp.2 | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/man2/seccomp.2 b/man2/seccomp.2
>> index 58033da1c..d6b856c32 100644
>> --- a/man2/seccomp.2
>> +++ b/man2/seccomp.2
>> @@ -1101,7 +1101,7 @@ install_filter(int syscall_nr, int t_arch, int f_errno)
>> };
>>
>> struct sock_fprog prog = {
>> - .len = (unsigned short) (sizeof(filter) / sizeof(filter[0])),
>> + .len = sizeof(filter) / sizeof(filter[0]),
>> .filter = filter,
>> };
>
> I have a small doubt about this change. With the change,
> there are no compilation warnings.
>
> But, if we change the code to something slightly different:
>
> [[
> size_t x = (sizeof(filter) / sizeof(filter[0]));
> struct sock_fprog prog = {
> .len = x,
> .filter = filter,
> };
> ]]
>
> The "cc -Wconversion" gives us the following warning:
>
> warning: conversion from ‘size_t’ {aka ‘long unsigned int’}
> to ‘short unsigned int’ may change value
>
> Presumably we don't get a warning for an assignment of the form
>
> .len = (sizeof(filter) / sizeof(filter[0]))
>
> because the compiler is smart enough to work out that the
> value of the constant expression is within the range of
> "unsigned short".
>
> Your thoughts?
Hi Michael,
I'd say that the cast doesn't fix any problems at all. It silences a
valid warning, and I'd use a pragma for that (to be more explicit about
the intention of silencing a warning) if I do want -Wconversion enabled
(which usually I don't want, because it's too noisy) and I'm sure that
this won't overflow. I'd limit the use casts to only when I *really*
need to.
I guess that if you enable -O3, the warning will vanish again because
the compiler will optimize away 'x' (but I didn't test).
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Use \(aq to get an unslanted single quote inside monospace code
blocks. Using a simple ' results in a slanted quote inside PDFs.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
These variables are either of an unsigned integer type per POSIX;
or of an integer type per POSIX, that Linux defines as an unsigned integer type.
Print them with 'uintmax_t' instead of 'intmax_t' to avoid
big positive numbers being printed as negative numbers.
Bug report:
From: Konstantin Bukin @ 2020-09-13 15:04 UTC
To: mtk.manpages; +Cc: Konstantin Bukin, linux-man
inode numbers are expected to be positive. Casting them to a signed type
may result in printing negative values. E.g. running example program on
the following file:
$ ls -li test.txt
9280843260537405888 -r--r--r-- 1 kbukin hardware 300 Jul 21 06:36 test.txt
results in the following output:
$ ./example test.txt
ID of containing device: [0,480]
File type: regular file
I-node number: -9165900813172145728
Mode: 100444 (octal)
Link count: 1
Ownership: UID=2743 GID=30
Preferred I/O block size: 32768 bytes
File size: 300 bytes
Blocks allocated: 8
Last status change: Tue Jul 21 06:36:50 2020
Last file access: Sat Sep 12 14:13:38 2020
Last file modification: Tue Jul 21 06:36:50 2020
Such erroneous reporting happens for inode values greater than maximum
value which can be stored in signed long. Casting does not seem to be
necessary here. Printing inode as unsigned long fixes the issue.
Reported-by: Konstantin Bukin <kbukin@gmail.com>
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Various ports that had their own indigenous system calls have
been discontinued. Remove those system calls (none of which had
manual pages!) to a separate part of the page, to avoid
cluttering the main list of system calls.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Some of the links removed in commit 247c654385 should
have been kept, because in some cases there are real system
calls whose wrapper functions are documented in Section 3.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Let's move to the 21st century. Instead of casting system data
types to long/long long/etc. in printf() calls, instead cast to
intmax_t or uintmax_t, the largest available signed/unsigned
integer types.
[mtk: rewrote commit message]
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Use printf()'s '#' flag character to prepend the string "0x".
However, when the number is printed in uppercase, and the prefix
is in lowercase, the string "0x" needs to be manually written.
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
For consistency.
The types are written both with and without the redundant 'int' keyword
all over the man-pages. However, the most used form, by far, is the one
without 'int'.
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Notes: I copied .nf and .fi from futex.2, but they made no visual difference.
What do they actually do?
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
When calling msgrcv() with the MSG_COPY flag, it will report
EINVAL error even we if have disabled CONFIG_CHECKPOINT_RESTORE.
ENOSYS will be reported only if we also specify the IPC_NOWAIT
flag.
[mtk: edited commit message]
Notes from mtk:
The relevant kernel code is this:
[[
#ifdef CONFIG_CHECKPOINT_RESTORE
...
#else
static inline struct msg_msg *prepare_copy(void __user *buf, size_t bufsz)
{
return ERR_PTR(-ENOSYS);
}
...
static long do_msgrcv(int msqid, void __user *buf, size_t bufsz, long
msgtyp, int msgflg,
long (*msg_handler)(void __user *, struct msg_msg *, size_t))
{
...
if (msgflg & MSG_COPY) {
if ((msgflg & MSG_EXCEPT) || !(msgflg & IPC_NOWAIT))
return -EINVAL;
copy = prepare_copy(buf, min_t(size_t, bufsz, ns->msg_ctlmax));
...
}
]]
We'll only hit the ENOSYS error if:
(1) MSG_COPY was specified;
(2) IPC_NOWAIT was not specified; and
(3) CONFIG_CHECKPOINT_RESTORE was not enabled.
Signed-off-by: Yang Xu <xuyang2018.jy@cn.fujitsu.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
The type `struct msgbuf *` is implicitly casted to `const void *`.
Not only that, but the explicit cast to `void *` was slightly
misleading.
Explicitly casting can silence warnings when mistakes are made, so it's
better to remove those casts when possible.
Signed-off-by: Alejandro Colomar <colomar.6.4.3@gmail.com>
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Rather than:
sometype x;
for (x = ....; ...)
use
for (sometype x = ...; ...)
This brings the declaration and use closer together (thus aiding
readability) and also clearly indicates the scope of the loop
counter variable.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Michael Kerrisk