summaryrefslogtreecommitdiff
path: root/src/lib/libc/stdlib (follow)
Commit message (Collapse)AuthorAgeFilesLines
* .Li -> .Vt where appropriate;jmc2022-09-1111-39/+39
| | | | | | from josiah frentsos, tweaked by schwarze ok schwarze
* For putenv(3), return an error if string starts with a '=' character.millert2022-08-082-5/+10
| | | | Both FreeBSD and NetBSD have this behavior. OK deraadt@
* change some 4.4BSD references to earlier releasesjsg2022-08-042-6/+6
| | | | ok schwarze@
* If a command or interface first appeared in PWB/UNIX, UNIX System III orjsg2022-07-252-6/+10
| | | | | | | | | | | | UNIX System V mention it. Only do so in manual pages with a pre-existing HISTORY section. Prompted by the comparison of System V and BSD commands and interfaces in Sun's "System V Enhancements Overview" document. checked against manuals on bitsavers, TUHS archive and CSRG archive CDs ok jmc@ schwarze@
* To figure our whether a large allocation can be grown into theguenther2022-06-301-12/+2
| | | | | | | | | | | following page(s) we've been first mquery()ing for it, mmapp()ing w/o MAP_FIXED if available, and then munmap()ing if there was a race. Instead, just try it directly with mmap(MAP_FIXED | __MAP_NOREPLACE) tested in snaps for weeks ok deraadt@
* system(3) should ignore SIGINT and SIGQUIT until the shell exits.millert2022-05-211-4/+17
| | | | | | | This got broken when system.c was converted from signal(3) to sigaction(2). Also add SIGINT and SIGQUIT to the set of blocked signals and unblock them in the parent after the signal handlers are installed. Based on a diff from Leon Fischer. OK deraadt@
* man pages: add missing commas between subordinate and main clausesnaddy2022-03-312-7/+7
| | | | | | | jmc@ dislikes a comma before "then" in a conditional, so leave those untouched. ok jmc@
* man pages: add missing word, The foo() ... -> The foo() function ...naddy2022-03-291-3/+3
| | | | ok jmc@ schwarze@
* Currently malloc caches a number of free'ed regions up to 128kotto2022-02-261-33/+160
| | | | | | | | | | in size. This cache is indexed by size (in # of pages), so it is very quick to check. Some programs allocate and deallocate larger allocations in a frantic way. Accomodate those programs by also keeping a cache of regions between 128k and 2M, in a cache of variable sized regions. Tested by many in snaps; ok deraadt@
* remove please from manual pagesjsg2022-02-061-3/+3
| | | | ok jmc@ sthen@ millert@
* refer to longindex as an argument, not a field;jmc2022-01-061-3/+3
| | | | | | from uwe@netbsd -r1.22 ok millert
* lsearch(3): reimplement using lfind(3)cheloha2021-12-081-32/+13
| | | | | | | | | | | | | | | | | | | | | | | | lsearch(3) is really just lfind(3) with an additional branch to append the key if lfind(3) fails. If we get rid of the underlying linear_base() function and move the search portion into lfind(3) and the key-copying portion into lsearch(3) we get smaller and simpler code. Misc. notes: - We do not need to keep the historical comment about errno. lsearch(3) is pure computation and does not set errno. That's really all you need to know. The specification reserves no errors, either. - We are using lfind(3) internally now, so it switches from PROTO_DEPRECATED to PROTO_NORMAL in hidden/search.h and needs DEF_WEAK in stdlib/lsearch.c. With advice from guenther@ on symbol housekeeping in libc. Thread: https://marc.info/?l=openbsd-tech&m=163885187632449&w=2 ok millert@
* lsearch(3): append key to array with memmove(3) instead of memcpy(3)cheloha2021-12-071-2/+7
| | | | | | | | | | | | | | | If the key overlaps the end of the array, memcpy(3) mutates the key and copies a corrupted value into the end of the array. If we use memmove(3) instead we at least end up with a clean copy of the key at the end of the array. This is closer to the intended behavior. With input from millert@ and deraadt@. Thread: https://marc.info/?l=openbsd-tech&m=163880307403606&w=2 ok millert@
* bsearch(3): support arrays with more than INT_MAX elementscheloha2021-12-021-2/+3
| | | | | | | | | | | | | | | The "lim" variable needs to be a size_t to match nmemb, otherwise we get undefined behavior when nmemb exceeds INT_MAX. Prompted by a blog post by Joshua Bloch: https://ai.googleblog.com/2006/06/extra-extra-read-all-about-it-nearly.html Fixed by Chris Torek a long time ago: https://svnweb.freebsd.org/csrg/lib/libc/stdlib/bsearch.c?revision=51742&view=markup ok millert@
* For open/openat, if the flags parameter does not contain O_CREAT, thederaadt2021-10-242-6/+6
| | | | | | | | | | | | | 3rd (variadic) mode_t parameter is irrelevant. Many developers in the past have passed mode_t (0, 044, 0644, or such), which might lead future people to copy this broken idiom, and perhaps even believe this parameter has some meaning or implication or application. Delete them all. This comes out of a conversation where tb@ noticed that a strange (but intentional) pledge behaviour is to always knock-out high-bits from mode_t on a number of system calls as a safety factor, and his bewilderment that this appeared to be happening against valid modes (at least visually), but no sorry, they are all irrelevant junk. They could all be 0xdeafbeef. ok millert
* ansijsg2021-10-241-3/+2
| | | | ok mpi@ deraadt@
* Provide realpath(1)kn2021-10-131-2/+3
| | | | | | | A tiny realpath(3) wrapper to make a porter's life easier. Feedback kettenis deraadt cheloha sthen OK cheloha martijn deraadt
* Switch two calls from memset() to explicit_bzero()tb2021-09-191-3/+3
| | | | | | | This matches the documented behavior more obviously and ensures that these aren't optimized away, although this is unlikely. Discussed with deraadt and otto
* * add the missing STANDARDS section as noticed by tb@schwarze2021-09-031-3/+20
| | | | | | * mention that the *optionp input string will be modified * clarify that the array of tokens is expected to be NULL-terminated OK millert@ tb@, and the first half of STANDARDS also OK jmc@
* Make MALLOC_STATS compile again; noted by Omar Polo and Joe Nelsonotto2021-07-231-2/+2
|
* An extra internal consistency check and a missing stats adjustment. ok tb@otto2021-04-091-1/+4
|
* Cache implementation has changed, we do not hold on to an exact numberotto2021-04-091-3/+4
| | | | of pages anymore, but also cache larger regions; ok tb@
* Type-cast getpagesize() from int to size_t for the comparison with d.claudio2021-03-181-2/+2
| | | | | | | getpagesize() will only return positive numbers (there is no negative page size system) and it can not fail. Should fix some compiler warnings seen in -portable projects. OK otto@
* Change the implementation of the malloc cache to keep lists ofotto2021-03-091-152/+118
| | | | | regions of a given size. In snaps for a while, committing since no issues were reported and a wider audience is good. ok deraadt@
* - Make use of the fact that we know how the chunks are aligned, andotto2021-02-252-49/+83
| | | | | | | | | | write 8 bytes at the time by using a uint64_t pointer. For an allocation a max of 4 such uint64_t's are written spread over the allocation. For pages sized and larger, the first page is junked in such a way. - Delayed free of a small chunk checks the corresponiding way. - Pages ending up in the cache are validated upon unmapping or re-use. In snaps for a while
* Some people still argue that rand(3) and random(3) have suitable deterministicderaadt2021-02-122-4/+10
| | | | | | | | | | | use cases, so explain the situation a bit more. Since the 80's, I estimate around 5 algorithm changes, so any chosen seed is unrepeatable UB. +The deterministic sequence algorithm changed a number of times since +original development, is underspecified, and should not be relied upon to +remain consistent between platforms and over time. ok jmc kettenis
* mapalign() only handles allocations >= a page; problem found by and ok semarie@otto2020-11-231-1/+3
|
* make fixed-sized fixed-value mib[] arrays be constderaadt2020-10-121-4/+3
| | | | ok guenther tb millert
* As noted by tb@ previous commit only removed an unused fucntion.otto2020-10-091-4/+9
| | | | | So redo previous commit properly: Use random value for canary bytes; ok tb@.
* Use random value for canary bytes; ok tb@otto2020-10-061-23/+1
|
* For page-sized and larger allocations do not put the pages we'reotto2020-09-061-21/+18
| | | | | | | shaving off into the cache but unamp them. Pages in the cache get re-used and then a future grow of the first allocation will be hampered. Also make realloc a no-op for small shrinkage. ok deraadt@
* This patch fixes one bug and one instance of undesirable behaviour.schwarze2020-05-271-9/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | The bug, present since 4.4BSD, was that a trailing dash in an option group, when the dash is not permitted as an option letter, resulted in the whole option group being returned as an argument, even though the previous option in the group was already parsed as an option: OPTS=abc ./getopt-test -a- -c arg ===>> OPT(a)ARG(-a-)ARG(-c)ARG(arg). Instead, treat the dash as an invalid option and continue parsing options: ===>> OPT(a)ERR(?-)OPT(c)ARG(arg). The undesirable behaviour was that allowing the dash as an option letter only allowed isolated dashes ("-") and trailing dashes in groups ("-a-"), but neither middle dashes in groups ("-a-b"), even though that already partially worked in 4.4BSD, nor leading dashes in groups ("--a"), even though that works on all other BSDs and on glibc. Also, while POSIX does not require that the dash can be used as an option letter at all, arguably, it encourages that letters either be fully supported or not supported at all. It is dubious whether supporting an option letter in some positions but not in others can be considered conforming. This patch makes OpenBSD behaviour identical to FreeBSD and NetBSD, improves compatibility with glibc (except that glibc does not support isolated "-"), improves compatibility with DragonFly (except that DragonFly is buggy when the dash option letter can take an optional argument but that argument is not present), improves compatibility with Illumos and Solaris 11 (except those do not support "-" and mishandle "--a"), and restores 4.4BSD behaviour for "-a-b". In no respect i'm aware of is compatibility with any other systems reduced. For the full rationale, see my mail to tech@ on 30 Mar 2020 14:26:41 +0200. Part of the problem was originally reported by an anonymous coward on tech@ on 12 Mar 2020 03:40:24 +0200, additional analysis was contributed by martijn@, and then the OP sent the final version of the patch i'm now committing on 17 Mar 2020 19:17:56 +0200. No licensing problem here because after the commit, the file does not contain a single word written by the OP. Also, the OP told me in private mail that he intends to publish the patch under the ISC license already contained in the file and that he wishes to be known by the pseudonym "0xef967c36". OK martijn@, and no objection when shown on tech@, but commit delayed to stay clear of the release.
* Minimal maintenance to make this mess slightly less confusing:schwarze2020-04-261-10/+10
| | | | | | | queue -> list; mention "intrusive"; element -> member at one place; delete a bogus remark that maybe referred to a long-gone implementation in VAX assembly code. Much more could be improved, but i don't want to waste too much time here.
* fix the description; from andras farkasjmc2020-04-261-5/+4
| | | | | | ok schwarze kill a Tn while here...
* "eventually" came and went back in 2004.libressl-v3.1.0martijn2020-03-301-3/+1
| | | | OK schwarze@
* correct Research Unix edition "appeared in" use in HISTORYjsg2020-02-081-3/+3
| | | | | | | | | | | | | | | | | Starting from "Combined Table of Contents" in Doug McIlroy's "A Research UNIX Reader" a table of which edition manuals appeared in. Checked against manuals from bitsavers/TUHS and source from TUHS where available. Ingo points out there are cases where something is included but not documented until a later release. bcd(6) v6 v7 printf(3) v2 v4 abort(3) v5 v6 system(3) v6 v7 fmod(3) v5 v6 ok schwarze@
* Document how to make getopt_long(3) process arguments in order and stopstsp2020-01-131-2/+18
| | | | | at the first non-option argument. I had to read source code to figure it out.
* drand48(3) returns values in [0.0, 1.0).tb2019-12-201-3/+3
| | | | | | From j@bitminer.ca with input from Andras Farkas, deraadt, joerg@netbsd "fix however you feel best!" jmc
* The file passed to realpath(3) must exists, adjust man page to newbluhm2019-12-111-3/+3
| | | | | behavior. noticed by hshoexer@; OK beck@
* Add comment line saying S is described vaguely on purpose.otto2019-09-141-2/+3
| | | | Prompted by guenther@
* improve verb-tense for explaining the calling convention of __deraadt2019-07-051-3/+3
| | | | ok guenther jmc
* The last consumer of pre-posix realpath behaviour has stoppedderaadt2019-07-051-34/+3
| | | | | | | | requiring it (sftp-server). Remove the /exists///// behaviour from here. The /nonexistant behaviour remains in the kernel and needs to be shot next. There may be ports fallout, but we doubt it. ok beck djm
* tweak previous; ok guentherjmc2019-06-301-5/+6
|
* Document that getcwd() and realpath() are built on system calls thatguenther2019-06-291-2/+16
| | | | | | | have a different calling convention than the standard function...as seen in kdump output. ok deraadt@ schwarze@
* When system calls indicate an error they return -1, not some arbitraryderaadt2019-06-281-2/+2
| | | | | | value < 0. errno is only updated in this case. Change all (most?) callers of syscalls to follow this better, and let's see if this strictness helps us in the future.
* The C89 standard only requires that atexit(3) returns a non-zero valueschwarze2019-06-271-5/+12
| | | | | | on error, so checking for -1 only is potentially non-portable. Also mention that the C89 standard does not require errno to be set. OK deraadt@ millert@
* Remove old realpath(3), and the userland-vs-kernel realpath verificationderaadt2019-06-171-334/+44
| | | | | | | | code. We now user the simple userland wrapper on top of __realpath(2). The non-POSIX behaviour still remains, that is the next component to fix. From a diff by beck, but I managed to chainsaw it a bit further. Tested in snaps for a couple of days. ok beck
* oops - missing .El in previous;jmc2019-06-151-1/+2
|
* realpath(3) doesn't use lstat(2), readlink(2), or getcwd(3) anymore,deraadt2019-06-151-21/+39
| | | | | | it is a thin wrapper over the syscall __readlink(2). Improve the list of possible errors. ok millert beck jmc
* Complete the ld.so boot cleanup: move most libc initialization fromguenther2019-06-021-7/+2
| | | | | | | | | | | | | | | | | _csu_finish() to _libc_preinit(), which is an .init_array function in shared libc (and mark it INITFIRST) or a .preinit_array function in static libc, grabbing the _dl_cb callback there from ld.so. Then in _csu_finish(), invoke the dl_clean_boot() callback to free ld.so's startup bits before main() gets control. Other cleanups this permits: - move various startup structures into .data.rel.ro - the dl* stubs in libc can be strong and call the callbacks provided via _dl_cb - no longer need to conditionalize dlctl() calls on presence of _DYNAMIC original concept and full diff ok kettenis@ ok deraadt@
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2025-03-19 19:44:46 +0000