Commit Graph

8794 Commits

Author SHA1 Message Date
Mateusz Guzik
2847cfce54 amd64: remove stale attribution for memmove work
While the routine started as expanded bcopy, it is now entirely rewritten.

Sponsored by:	The FreeBSD Foundation
2018-11-30 00:47:36 +00:00
Mateusz Guzik
dd219e5ea5 amd64: tidy up copying backwards in memmove
For non-ERMS case the code used handle possible trailing bytes with
movsb first and then followed it up with movsq. This also happened
to alter how calculations were done for other cases.

Handle the tail with regular movs, just like when copying forward.
Use leaq to calculate the right offset from the get go, instead of
doing separate add and sub.

This adjusts the offset for non-rep cases so that they can be used
to handle the tail.

The routine is still a work in progress.

Sponsored by:	The FreeBSD Foundation
2018-11-30 00:45:10 +00:00
Dag-Erling Smørgrav
e86caa0bd0 Style cleanup. 2018-11-27 09:41:47 +00:00
Ed Maste
cac83db9b7 revert r340640 "libc: forcibly disable BIND_NOW"
When immediate bind mode is requested, as of r340675 rtld processes
irelocs in PLT immediately after other PLT relocs.  That addresses the
libc + BIND_NOW startup crash the workaround is no longer needed.

PR:		233333
2018-11-26 13:56:19 +00:00
Yuri Pankov
63cbe8d1d9 regexec: fix processing multibyte strings.
Matcher function incorrectly assumed that moffset that we get from
findmust is in bytes. Fix this by introducing a stepback function,
taking short path if MB_CUR_MAX is 1, and going back byte-by-byte,
checking if we have a legal character sequence otherwise.

PR:		153502
Reviewed by:	pfg, kevans
Approved by:	kib (mentor, implicit)
Differential revision:	https://reviews.freebsd.org/D18297
2018-11-23 15:49:18 +00:00
Edward Tomasz Napierala
3a9cdc5132 Make sysconf(_SC_PAGESIZE) return the value from getpagesize(3).
That avoids a syscall - getpagesize(3) gets the value from the ELF
aux strings.

Reviewed by:	kib
MFC after:	2 weeks
Sponsored by:	DARPA, AFRL
Differential Revision:	https://reviews.freebsd.org/D17989
2018-11-19 18:23:17 +00:00
Ed Maste
00bf4d2515 libc: forcibly disable BIND_NOW
Building libc WITH_BIND_NOW results in segfault at process start.  For
now force BIND_NOW off until the root cause can be identified and fixed.

PR:		233333
Sponsored by:	The FreeBSD Foundation
2018-11-19 18:12:39 +00:00
Mateusz Guzik
088ac3ef4b amd64: handle small memset buffers with overlapping stores
Instead of jumping to locations which store the exact number of bytes,
use displacement to move the destination.

In particular the following clears an area between 8-16 (inclusive)
branch-free:

movq    %r10,(%rdi)
movq    %r10,-8(%rdi,%rcx)

For instance for rcx of 10 the second line is rdi + 10 - 8 = rdi + 2.
Writing 8 bytes starting at that offset overlaps with 6 bytes written
previously and writes 2 new, giving 10 in total.

Provides a nice win for smaller stores. Other ones are erratic depending
on the microarchitecture.

General idea taken from NetBSD (restricted use of the trick) and bionic
string functions (use for various ranges like in this patch).

Reviewed by:	kib (previous version)
Sponsored by:	The FreeBSD Foundation
Differential Revision:	https://reviews.freebsd.org/D17660
2018-11-16 00:44:22 +00:00
Mateusz Guzik
ad2ff705a4 amd64: sync up libc memset with the kernel version
- tidy up memset to have rax set earlier for small sizes
- finish the tail in memset with an overlapping store
- align memset buffers to 16 bytes before using rep stos

Sponsored by:	The FreeBSD Foundation
2018-11-15 20:28:35 +00:00
Mateusz Guzik
6fff634455 amd64: convert libc bzero to a C func to avoid future bloat
Reviewed by:	kib (previous version)
Sponsored by:	The FreeBSD Foundation
Differential Revision:	https://reviews.freebsd.org/D17549
2018-11-15 20:20:39 +00:00
Alan Somers
a14a34ef62 fcntl.2: document an additional error condition
MFC after:	2 weeks
2018-11-15 16:13:25 +00:00
Konstantin Belousov
1c4ca77890 Add d_off support for multiple filesystems.
The d_off field has been added to the dirent structure recently.
Currently filesystems don't support this feature.  Support has been
added and tested for zfs, ufs, ext2fs, fdescfs, msdosfs and unionfs.
A stub implementation is available for cd9660, nandfs, udf and
pseudofs but hasn't been tested.

Motivation for this feature: our usecase is for a userspace nfs server
(nfs-ganesha) with zfs.  At the moment we cache direntry offsets by
calling lseek once per entry, with this patch we can get the offset
directly from getdirentries(2) calls which provides a significant
speedup.

Submitted by:	Jack Halford <jack@gandi.net>
Reviewed by:	mckusick, pfg, rmacklem (previous versions)
Sponsored by:	Gandi.net
MFC after:	1 week
Differential revision:	https://reviews.freebsd.org/D17917
2018-11-14 14:18:35 +00:00
Yuri Pankov
98f3acfaaf Fix WITHOUT_ICONV build after r340276.
Reported by:	olivier
Approved by:	kib (mentor, implicit)
2018-11-14 09:06:15 +00:00
Konstantin Belousov
44748aa487 Create namespace for the symbols added during 13-CURRENT cycle.
Sponsored by:	The FreeBSD Foundation
MFC after:	1 week
2018-11-12 19:12:14 +00:00
Konstantin Belousov
5b1fb8ec66 First draft of documentation for AT/O_BENEATH handling of the absolute
paths.

It was decided that committing the code and drafting of the man page
update is better than allowing the code to rot until wordsmithing
happens.

Reviewed by:	jilles (previous version)
Discussed with:	brooks, jilles, emaste
Sponsored by:	The FreeBSD Foundation
MFC after:	1 week
Differential revision:	https://reviews.freebsd.org/D17714
2018-11-11 01:46:48 +00:00
Edward Tomasz Napierala
8d113f4aec Don't call stat(2) on nsswitch.conf(5) every time nsdispatch(3)
and dependent functions (eg getpwname(3)) get called. This can
improve performance of binaries that perform a lot of name
lookups, such as gssd(8). It also matches documented behaviour
of Linux and Solaris.

The old code is left in place, should anyone need it, guarded
by #ifdef NS_REREAD_CONF.

Reviewed by:	imp, bcr
MFC after:	2 weeks
Relnotes:	yes
Sponsored by:	DARPA, AFRL
Differential Revision:	https://reviews.freebsd.org/D17934
2018-11-10 23:07:46 +00:00
Yuri Pankov
9eb7d595e1 Reset persistent mbstates when rune locale encoding changes.
This was shown to be a problem by side effect of now-enabled test case,
which was going through C, en_US.UTF-8, ja_JP.SJIS, and ja_JP.eucJP,
and failing eventually as data in mbrtowc's mbstate, that was
perfectly correct for en_US.UTF-8 was treated as incorrect for
ja_JP.SJIS, failing the entire test case.

This makes the persistent mbstates to be per ctype-component,
and not per-locale so we could easily reset the mbstates when
only LC_CTYPE is changed.

Reviewed by:	bapt, pfg
Approved by:	kib (mentor, implicit)
Differential Revision:	https://reviews.freebsd.org/D17796
2018-11-09 03:32:53 +00:00
Yuri Pankov
dd7c41a378 Add hybrid C.UTF-8 locale being identical to default C locale except
that it uses the same ctype maps and functions as other UTF-8 locales.

Reviewed by:	bapt, cem, eadler
Approved by:	kib (mentor, implicit)
Differential Revision:	https://reviews.freebsd.org/D17833
2018-11-04 22:13:22 +00:00
Yuri Pankov
32a8ec8bdd strptime: make %k and %l specifiers match their description in
strftime(3), and allow them to process space-padded input.

PR:		230720
Submitted by:	rlittle@inetco.com (original version)
Approved by:	kib (mentor, implicit)
Differential Revision:	https://reviews.freebsd.org/D17761
2018-11-03 23:37:13 +00:00
Conrad Meyer
78c2a9806e kern_poll: Restore explanatory comment removed in r177374
The comment isn't stale.  The check is bogus in the sense that poll(2)
does not require pollfd entries to be unique in fd space, so there is no
reason there cannot be more pollfd entries than open or even allowed
fds.  The check is mostly a seatbelt against accidental misuse or
abuse.  FD_SETSIZE, while usually unrelated to poll, is used as an
arbitrary floor for systems with very low kern.maxfilesperproc.

Additionally, document this possible EINVAL condition in the poll.2
manual.

No functional change.

Reviewed by:	markj
Sponsored by:	Dell EMC Isilon
Differential Revision:	https://reviews.freebsd.org/D17671
2018-11-01 23:46:23 +00:00
Yuri Pankov
281c29899f Connect libc/tests/time to the build, adding test cases for strptime()
issues fixed recently, and disabling the failing ones (mostly due to TZ
parsing differences with NetBSD).

Reviewed by:	ngie
Approved by:	kib (mentor)
Differential Revision:	https://reviews.freebsd.org/D17546
2018-10-30 02:37:23 +00:00
Konstantin Belousov
1d160e1e9c Convert amd64_get/set_fs/gsbase to ifunc.
Note that this is the first use of ifuncs in our userspace.

Sponsored by:	The FreeBSD Foundation
MFC after:	1 month
2018-10-30 00:11:30 +00:00
Warner Losh
f64bccc6d9 Bump .Dd forgotten in last commit. 2018-10-28 03:02:09 +00:00
Warner Losh
5669c6748d Note that the kenrel doesn't keep track daylight savings time, nor
timezone offset. These values are generally zero.

While one still theoreticall could set these values, that's almost
never done. Users wishing to have an offset between the time of day
clock hardware and UTC use adjkerntz(8) instead.

localtime(3) should be used to find these values for the current
timezone.
2018-10-28 02:58:22 +00:00
Konstantin Belousov
4f77f48884 Implement O_BENEATH and AT_BENEATH.
Flags prevent open(2) and *at(2) vfs syscalls name lookup from
escaping the starting directory.  Supposedly the interface is similar
to the same proposed Linux flags.

Reviewed by:	jilles (code, previous version of manpages), 0mp (manpages)
Discussed with:	allanjude, emaste, jonathan
Sponsored by:	The FreeBSD Foundation
Differential revision:	https://reviews.freebsd.org/D17547
2018-10-25 22:16:34 +00:00
Mark Johnston
b8e4cdda35 Clarify slightly the interaction between wait*() and pdfork().
There are multiple ways to wait for any child process to return a
status (e.g., waitpid(-1, ...), waitid(P_ALL, ...)), so don't be so
specific.

MFC after:	1 week
Sponsored by:	The FreeBSD Foundation
2018-10-24 18:42:13 +00:00
Cy Schubert
3ad12d2782 Follow up on r331936. gets_s(3) will also fail in the same way that
gets(3) does. This was missed in r331936.

Reported by:	emaste@
MFC after:	3 days
2018-10-23 14:16:02 +00:00
Poul-Henning Kamp
01652e9c8e Update example to something people less than 40 years old have heard about. 2018-10-21 07:30:26 +00:00
Ed Maste
6492be46b8 ANSIfy libc/quad
Some of these routines exist in both lib/libc/quad/ and sys/libkern/.
r325988 ANSIfied sys/libkern.  Update libc/quad to match.

PR:		223641
Reported by:	bde
Sponsored by:	The FreeBSD Foundation
2018-10-21 00:20:40 +00:00
Ed Maste
05bc9aa78f mcount: tidy up ANSIfication
libc/gmon's mcount was ANSIfied in r124180, with libkern following over
a decade later, in r325988, but some minor discrepancies remained.
Update libc/gmon's mexitcount to an ANSI C function definition, and use
(void) for libkern-only functions that take no arguments.

Reported by:	bde
2018-10-20 22:39:35 +00:00
Mariusz Zaborski
27eff71458 Add link to the setproctitle_fast function.
MFC after:	1 week
2018-10-20 22:25:57 +00:00
Conrad Meyer
243e09430a Embedded chacha: Distinguish via dedicated macro
Set embedding expectations via CHACHA_EMBED macro rather than _KERNEL
definition.

No functional change.

Sponsored by:	Dell EMC Isilon
2018-10-20 22:14:29 +00:00
Conrad Meyer
0a3a36cd94 getentropy(3): Trap non-API errnos from getrandom(2) and abort
Additionally, reconcile our abort behavior with arc4random(3).  Unlike
SIGABRT, SIGKILL cannot be caught by the user program.  These failures
are fatal conditions and should not return to the caller, as they did in
the instance that resulted in D17049.

While here, fix some minor typos in a comment.

Reviewed by:	delphij
Sponsored by:	Dell EMC Isilon
Differential Revision:	https://reviews.freebsd.org/D17050
2018-10-20 21:45:17 +00:00
Yuri Pankov
4644f9bef6 Add -b/-l options to localedef(1) to specify output endianness and use
it appropriately when building share/ctypedef and share/colldef.

This makes the resulting locale data in EL->EB (amd64->powerpc64) cross
build and in the native EB build match.  Revert the changes done to libc
in r308170 as they are no longer needed.

PR:		231965
Reviewed by:	bapt, emaste, sbruno, 0mp
Approved by:	kib (mentor)
Differential Revision:	https://reviews.freebsd.org/D17603
2018-10-20 20:51:05 +00:00
Mark Johnston
4ac9afdd2c Fix formatting.
MFC after:	1 week
Sponsored by:	The FreeBSD Foundation
2018-10-20 17:48:05 +00:00
Yuri Pankov
eb144aa007 strptime: fix parsing of tm_year when both %C and %y appear in the
format string in arbitrary order.  This makes the related test cases in
lib/libc/tests/time (not yet connected to the build) pass.

While here, don't error on negative tm_year value based on the
APPLICATION USAGE in
http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/time.h.html
(glibc does the same):

tm_year is a signed value; therefore, years before 1900 may be represented.

Approved by:	re (gjb), kib (mentor)
Differential Revision:	https://reviews.freebsd.org/D17550
2018-10-17 14:51:43 +00:00
Edward Tomasz Napierala
8bb6c72e63 Get rid unneccessary useless calls to lseek(2) from getgrent(3).
Looks like this:

 fstatat(AT_FDCWD,"/etc/nsswitch.conf",{ mode=-rw-r--r-- ,inode=2167001,size=390,blksize=32768 },0x0) = 0 (0x0)
 open("/etc/group",O_RDONLY|O_CLOEXEC,0666)      = 3 (0x3)
 fstat(3,{ mode=-rw-r--r-- ,inode=2166927,size=919,blksize=32768 }) = 0 (0x0)
-lseek(3,0x0,SEEK_CUR)                           = 0 (0x0)
-lseek(3,0x0,SEEK_SET)                           = 0 (0x0)
 read(3,"# $FreeBSD: release/10.0.0/etc/g"...,32768) = 919 (0x397)
 close(3)                                        = 0 (0x0)

Reviewed by:	kib
Approved by:	re (gjb)
MFC after:	2 weeks
Sponsored by:	DARPA, AFRL
Differential Revision:	https://reviews.freebsd.org/D17571
2018-10-16 18:17:07 +00:00
Gleb Smirnoff
9004dbddea Avoid OOB reads in memmem(3).
commit 51bdcdc424bd7169c8cccdc2de7cad17f5ea0f70
Author: Alexander Monakov <amonakov@ispras.ru>
Date:   Fri Jun 30 00:35:33 2017 +0300

    fix OOB reads in Xbyte_memmem

    Reported by Leah Neukirchen.

Reviewed by:	emaste
Approved by:	re (kib)
2018-10-15 20:20:57 +00:00
Edward Tomasz Napierala
d3450d02c7 Prevent flex(1) from generating calls to isatty(3) - and, in turn,
TIOCGETA ioctls - when parsing nsswitch.conf(5).

Reviewed by:	imp, markj
Approved by:	re (gjb)
MFC after:	2 weeks
Sponsored by:	DARPA, AFRL
Differential Revision:	https://reviews.freebsd.org/D17564
2018-10-15 19:48:36 +00:00
Edward Tomasz Napierala
1d7da823b4 Don't call dlopen(3) for built-in NSS types - "cache", "compat",
"dns", "files", "db", and "nis". It saves some path lookups during
binary startup.

Reviewed by:	markj
Approved by:	re (gjb, kib)
MFC after:	2 weeks
Sponsored by:	DARPA, AFRL
Differential Revision:	https://reviews.freebsd.org/D17557
2018-10-15 17:50:02 +00:00
Edward Tomasz Napierala
edbedaf4dc Add .Xrs to kqueue(2) from pdfork(2) and procdesc(4), to make EVFILT_PROCDESC
easier to find.

Approved by:	re (rgrimes)
MFC after:	2 weeks
Sponsored by:	DARPA, AFRL
2018-10-14 18:42:54 +00:00
Mateusz Guzik
9c7d70ee7d amd64: convert libc bcopy to a C func to avoid future bloat
The function is of limited use and is an almost a direct clone of
memmove/memcpy (with arguments swapped). Introduction of ERMS variants
of string routines would mean avoidable growth of libc.

bcopy will get redefined to a __builtin_memmove later on with this
symbol only left for compatibility.

Reviewed by:	kib
Approved by:	re (gjb)
Sponsored by:	The FreeBSD Foundation
Differential Revision:	https://reviews.freebsd.org/D17539
2018-10-13 21:17:28 +00:00
Mateusz Guzik
1e52ba8c62 amd64: import updated kernel memmove to libc
bcopy is left alone as it is expected to be converted to a C func.

Due to header mess ALIGN_TEXT is temporarily defined explicitly in memmove.S

Reviewed by:	kib
Approved by:	re (gjb)
Sponsored by:	The FreeBSD Foundation
Differential Revision:	https://reviews.freebsd.org/D17538
2018-10-13 21:15:47 +00:00
Yuri Pankov
cbc0062992 strptime: disallow zero hour for %I (defined by POSIX as [01,12]) and %l
(extension, defined in strftime(3) as 1-12).

Approved by:	re (gjb), kib (mentor)
Differential Revision:	https://reviews.freebsd.org/D17543
2018-10-13 16:25:28 +00:00
Allan Jude
c452913091 Document that sendfile(2) can return ENOTCAPABLE
PR:		232207
Submitted by:	Enji Cooper <yaneurabeya@gmail.com>
Approved by:	re (rgrimes)
2018-10-13 02:20:16 +00:00
Konstantin Belousov
92cbfb6395 Disallow zero day of month from strptime("%d").
It is required by POSIX, specified in our man page, and followed by
Linux.

PR:	232072
Reported by:	miguel_tete17@hotmail.com
Sponsored by:	The FreeBSD Foundation
Approved by:	re (gjb)
MFC after:	1 week
2018-10-08 18:45:40 +00:00
Mateusz Guzik
167374a162 amd64: import updated kernel memset to libc
See r339205 for details.

An unused ERMS support is retained in the macro. It will be activated
after ifunc support lands.

Reviewed by:    kib
Approved by:    re (gjb)
Sponsored by:   The FreeBSD Foundation
Differential Revision:	https://reviews.freebsd.org/D17405
2018-10-05 19:27:42 +00:00
Mateusz Guzik
7e02ad0769 amd64: reimplement libc memset and bzero with kernel memset
This is a depessimization, see r334537 for an explanation. Routines
remain significantly slower than they have to be.

bzero was removed from the kernel but remains in libc. Macroify to
accommodate differences to memset (no return value, always setting to 0).

The bzero.S file is left in place due to libc build magic which pulls in
a C variant if a matching .S file is missing.

Reviewed by:	kib
Approved by:	re (gjb)
Differential Revision:	https://reviews.freebsd.org/D17355
2018-10-01 20:39:17 +00:00
Mateusz Guzik
275c893dab amd64: remove unnecessary cld from libc memcpy/bcopy
The ABI specifies the direction forward on function call, making
the cld instruction redundant.

Approved by:	re (kib)
2018-09-29 07:40:52 +00:00
Mateusz Guzik
5bbde333cd amd64: reimplement libc memcmp and bcmp with kernel memcmp
Both are significantly slower than hand-coded loops. See r338963 for
kernel commit.

bcmp differs from memcmp by always returning 1 when a difference is
found, as opposed to going for a value bigger or lower than 0
depending on what it is. This means it can do less work. For now the
code is duplicated and modified. This will get deduplicated after
another round of optimization when memcmp will get a longer-term form.

Both tested with the glibc suite. While the suite does not have a test
for bcmp, I created a wrapper routine which verified that values match
(0 vs 0, 1 vs non-zero).

Reviewed by:	kib
Approved by:	re (gjb)
Sponsored by:	The FreeBSD Foundation
Differential Revision:	https://reviews.freebsd.org/D17336
2018-09-27 17:08:29 +00:00