Re: [PATCH 1/3] getrandom.2: new manpage

[Michael Kerrisk (man-pages)]

Hi Heinrich,

On Fri, Oct 3, 2014 at 2:15 AM, Heinrich Schuchardt <xypron.glpk@xxxxxx> wrote:
> Kernel 3.17 introduces a new system call getrandom(2).
>
> The man page in this patch is based on the commit message by
> Theodore Ts'o <tytso@xxxxxxx> and suggestion by
> Michael Kerrisk <mtk.manpages@xxxxxxxxx>.

No word from Ted so far...

I've added LKML to CC, because I think it's worth getting wider review
of the page at this point.

I've done some further editing on the page, and pushed the current
version in public branch:
http://git.kernel.org/cgit/docs/man-pages/man-pages.git/log/?h=draft_getrandom

Could you please review the current version, appended below. Note that
there are a couple of FIXMEs there. Also, could you pay special
attention to the changes in these commits, to make sure I have not
injected any errors:

pick 0ef180e getrandom.2: Add a sentence to clarify the default behavior...
pick 62342ef getrandom.2: Reword GRND_NONBLOCK description
pick 0c90d3d getrandom.2: Reword GRND_RANDOM description

Cheers,

Michael

.\" Copyright (C) 2014, Theodore Ts'o <tytso@xxxxxxx>
.\" Copyright (C) 2014, Heinrich Schuchardt <xypron.glpk@xxxxxx>
.\"
.\" %%%LICENSE_START(VERBATIM)
.\" Permission is granted to make and distribute verbatim copies of this
.\" manual provided the copyright notice and this permission notice are
.\" preserved on all copies.
.\"
.\" Permission is granted to copy and distribute modified versions of
.\" this manual under the conditions for verbatim copying, provided that
.\" the entire resulting derived work is distributed under the terms of
.\" a permission notice identical to this one.
.\"
.\" Since the Linux kernel and libraries are constantly changing, this
.\" manual page may be incorrect or out-of-date.  The author(s) assume.
.\" no responsibility for errors or omissions, or for damages resulting.
.\" from the use of the information contained herein.  The author(s) may.
.\" not have taken the same level of care in the production of this.
.\" manual, which is licensed free of charge, as they might when working.
.\" professionally.
.\"
.\" Formatted or processed versions of this manual, if unaccompanied by
.\" the source, must acknowledge the copyright and authors of this work.
.\" %%%LICENSE_END

.TH GETRANDOM 2 2014-10-03 "Linux" "Linux Programmer's Manual"
.SH NAME
getrandom \- obtain a series of random bytes
.SH SYNOPSIS
.B #include <linux/random.h>
.sp
.BI "int getrandom(void *"buf ", size_t " buflen ", unsigned int " flags );
.SH DESCRIPTION
The system call
.BR getrandom ()
fills the buffer pointed to by
.I buf
with up to
.I buflen
random bytes.
These can be used to seed user-space random number generators
or for other cryptographic purposes.
.PP
.BR getrandom ()
relies on entropy gathered from device drivers and other sources of
environmental noise.
Unnecessarily reading large quantities of data will have a negative impact
on other users of the
.I /dev/random
and
.I /dev/urandom
devices.
Therefore
.BR getrandom ()
should not be used for Monte Carlo simulations or other
programs/algorithms which are doing probabilistic sampling.

.\" FIXME is the following paragraph correct?
By default,
.BR getrandom ()
draws entropy from the
.IR /dev/urandom
pool, and, if that pool has been initialized and
.IR buflen
is less than or equal to 256 (see NOTES, below),
then the call never blocks when drawing from that pool
and always returns the number of bytes requested in
.IR buflen .
This behavior can be changed via the
.I flags
argument.

The
.I flags
argument is a bit mask that can contain zero or more of the following values
ORed together:
.TP
.B GRND_RANDOM
If this bit is set, then random bytes are drawn from the
.I /dev/random
pool instead of the
.I /dev/urandom
pool.
The
.I /dev/random
pool is limited based on the entropy that can be obtained from environmental
noise.
If the number of available bytes in
.I /dev/random
is less than requested in
.IR buflen ,
the call returns just the available random bytes.
If no random byte is available, the response will depend on the
presence of
.B GRND_NONBLOCK
in the
.I flags
argument.
.TP
.B GRND_NONBLOCK
By default, if there is no random byte available at all,
.BR getrandom ()
blocks until data is available.
If the
.B GRND_NONBLOCK
flag is set, then
.BR getrandom ()
instead immediately returns -1 with
.I errno
set to
.BR EAGAIN .
.SH RETURN VALUE
On success,
.BR getrandom ()
returns the number of bytes that were copied to the buffer
.IR buf .
This may be less than the number of bytes requested via
.I buflen
if insufficient entropy was present in the
.IR /dev/random
pool, or if the system call was interrupted by a signal.
.PP
On error, -1 is returned, and
.I errno
is set appropriately.
.SH ERRORS
.TP
.B EINVAL
An invalid flag was specified in
.IR flags .
.TP
.B EFAULT
The address referred to by
.I buf
is outside the accessible address space.
.TP
.B EAGAIN
The requested entropy was not available, and
.BR getrandom ()
would have blocked if the
.B GRND_NONBLOCK
flag was not set.
.TP
.B EINTR
While blocked waiting for entropy, the call was interrupted by a signal
handler; see the description of how interrupted
.BR read (2)
calls on "slow" devices are handled with and without the
.B SA_RESTART
flag in the
.BR signal (7)
man page.
.SH VERSIONS
.BR getrandom ()
was introduced in version 3.17 of the Linux kernel.
.SH CONFORMING TO
This system call is Linux-specific.
.SH NOTES
.SS Interruption by a signal handler
.\" FIXME Here, I think there needs to be an opening paragraph that describes
.\" the cases where getrandom() can block. This should cover the cases with
.\" GRND_RANDOM and without GRND_RANDOM. Reading the existing page, I am
.\" still not completely confident that I know what the cases are.
The reaction of
.BR getrandom ()
in case of an interruption of a blocking call by a signal
when reading from
.I /dev/urandom
.RB ( GRND_RANDOM
is not set)
depends on the initialization state of the entropy buffer
and on the request size
.IR buflen .
If the entropy is not yet initialized or the request size is large
.RI ( buflen "\ >\ 256),"
.B EINTR
will be returned.
If the entropy pool has been initialized and the request size is small
.RI ( buflen "\ <=\ 256),"
.BR getrandom ()
will not return
.BR EINTR .
Instead, it will return all of the bytes that have been requested.
.PP
When reading from
.I /dev/random
.RB ( GRND_RANDOM
is set)
these guarantees do
.I not
apply.
.PP
Calling
.BR getrandom ()
to read
.I /dev/urandom
for small values (<=\ 256) of
.I buflen
is the preferred mode of usage.
.PP
The special treatment of small values of
.I buflen
was designed for compatibility with
OpenBSD's
.BR getentropy ()
system call.
.PP
The user of
.BR getrandom ()
.I must
always check the return value,
to determine whether either an error occurred
or fewer bytes than requested were returned.
In the case where
.B GRND_RANDOM
is not specified and
.I buflen
is less than or equal to 256,
a return of fewer bytes than requested should never happen,
but the careful user-space code should check for this anyway!
.SS Choice of random device
Unless you are doing long-term key generation (and perhaps not even
then), you probably shouldn't be using
.B GRND_RANDOM.
The cryptographic algorithms used for
.I /dev/urandom
are quite conservative, and so should be sufficient for all purposes.
The disadvantage of
.B GRND_RANDOM
is that it can block.
Furthermore, dealing with partially fulfilled
.BR getrandom ()
requests increases code complexity.
.SS Emulating OpenBSD's getentropy()
The
.BR getentropy ()
system call in OpenBSD can be emulated using the following
function:

.in +4n
.nf
int
getentropy(void *buf, size_t buflen)
{
    int ret;

    if (buflen > 256)
        goto failure;
    ret = getrandom(buf, buflen, 0);
    if (ret < 0)
        return ret;
    if (ret == buflen)
        return 0;
failure:
    errno = EIO;
    return -1;
}
.fi
.in
.SH SEE ALSO
.BR random (4),
.BR urandom (4)
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/