+++ /dev/null
-Some notes to help future porters. Replace 'ARCH' with whatever arch
-you are hacking on.
-
-====================
-=== Config Files ===
-====================
-- create extra/Configs/Config.ARCH
- See the other arch files for some good examples. powerpc/sparc/alpha
- should be pretty simple templates.
-- add ARCH to the 'Target Architecture' list in extra/Configs/Config.in
-- Initially you will want to disable shared libraries, since making
- the shared library loader work requires you first have basic architecture
- support working. Thus you should add ARCH_HAS_NO_SHARED and
- ARCH_HAS_NO_LDSO to Config.ARCH's TARGET_ARCH
-
-====================
-=== libc sysdeps ===
-====================
-(note: if glibc has already been ported to your arch, you can usually just
- copy a lot of files from them rather than coding from scratch)
-- create libc/sysdeps/linux/ARCH
-- copy Makefile and Makefile.arch from libc/sysdeps/linux/i386/
-- set CSRC and SSRC to nothing in Makefile.arch for now
-
-- create crt1.S which defines the _start function ... you will probably want
- to clear the frame pointer to make gdb happy, and then you will want to call
- the funcion __uClibc_main() which takes these parameters:
- __uClibc_main(main(), argc, argv, _init(), _fini())
- Initially if you wish to make things easier on yourself, you can disable the
- UCLIBC_CTOR_DTOR option and just set the init/fini arguments to NULL.
- glibc generally stores this function in libc/sysdeps/ARCH/elf/start.S
-
-- create these additional files in ARCH/bits/
-
- (template versions can be found in common/bits/ for you to tweak)
- endian.h fcntl.h setjmp.h stackinfo.h uClibc_arch_features.h wordsize.h
-
- kernel_types.h should be created based upon linux asm-ARCH/posix_types.h
-
- copy linux asm-ARCH/stat.h to bits/kernel_stat.h
-
- create syscalls.h based upon linux's unistd.h / glibc's sysdeps.h ... really
- you just want to define the _syscall[0-6] macros. It is important that
- these syscalls should be PIC safe (or you should provide a PIC and non-PIC
- version) if you wish to properly support shared libraries.
-
-- at this point, you should have enough to generate a working HELLO WORLD
- static binary
-
-- if you want UCLIBC_CTOR_DTOR support, you will need to create crti.S and
- crtn.S files which define function prologues/epilogues.
-
-- for a more stable static port, you will need to create these files (and
- update the Makefile.arch values accordingly)
- __longjmp bsd-_setjmp bsd-setjmp brk clone setjmp syscall vfork
- usually these are written in assembler, but you may be able to cheat and
- write them in C ... see other ports for more information
-
-====================
-=== pthread deps ===
-====================
-
-TODO: nptl / linuxthreads / linuxthreads.old
-
-====================
-=== ldso sysdeps ===
-====================
-- elf.h - presumably you've already taught binutils all about the random ELF
- relocations your arch needs, so now you need to make sure the defines exist
- for uClibc. make sure the EM_### define exists and all of the R_###_###
- reloc defines.
-
-- enable ldso/shared options in your extra/Configs/Config.ARCH file
-- you will need to create the following files in ldso/ldso/ARCH/
- dl-debug.h dl-startup.h dl-syscalls.h dl-sysdep.h elfinterp.c resolve.S
-
-- dl-debug.h: define string versions of all the relocations of your arch in the
- _dl_reltypes_tab array ... the index should match the actual reloc type, so
- if the value of say R_X86_64_PC16 is 13, then "R_X86_64_PC16" better be at
- index 13 of the array
-
-- dl-startup.h:
- - define the _start function which should call _dl_start which takes just one
- parameter ... a pointer to argc (usually on the stack)
- glibc stores this function in libc/sysdeps/ARCH/dl-machine.h as RTLD_START
- - define the GET_ARGV() macro which calculates the value of argv based upon
- the parameter passed to _dl_start (usually it's simply just ARGS+1)
- - define PERFORM_BOOTSTRAP_RELOC() macro which will handle just the relocs
- that the ldso itself will generate
-
-- dl-syscalls.h:
- if you wrote your bits/syscalls.h file correctly in the libc step above, you
- can simply copy this file from another arch and be done ... otherwise you
- will have to define the syscall[0-6] macros again, but this time setting
- _dl_errno instead of just errno
-
-- dl-sysdep.h:
- misc cruft goes in here ... you want to:
- - either define or undefine ELF_USES_RELOCA
- - define the INIT_GOT macro
- - define MAGIC1 to the EM_### value your ELF arch uses
- - define ELF_TARGET to a string name for your arch
- - define the do_rem() macro
- - define misc ALIGN macro's
- - define elf_machine_type_class() macro
- - define the inline functions elf_machine_dynamic, elf_machine_load_address,
- and elf_machine_relative
- glibc stores a bunch of these values in libc/sysdeps/ARCH/dl-machine.h
-
-- elfinterp.c:
- define all the relocation functions ... it's best if you just copy from
- another arch which uses the same type of relocations (REL or RELA) and
- start from there.
-
-- resolve.S:
- front end of lazy relocation ... define the _dl_linux_resolve symbol which
- is called by a PLT entry which has yet to be setup ... you will want to:
- - set up arguments for _dl_linux_resolver()
- - call _dl_linux_resolver()
- - clean up after call
- - jump to function address now stored in PLT
- glibc stores this function in libc/sysdeps/ARCH/dl-trampoline.S
-
-- utils/ldd.c - if you want support for ldso cache files (spoiler: you do),
- then you'll need to teach ldd a little. generally, the fallback code
- should be smart and "just work", but you should be explicit. just pop
- it open and add an appropriate ifdef for your arch and set MATCH_MACHINE()
- and ELFCLASSM. there are plenty examples and you're (hopefully) smart.
-
-====================
-=== Misc Cruft ===
-====================
-- MAINTAINERS - presumably you're going to submit this code back to mainline
- and since you're the only one who cares about this arch (right now), you
- should add yourself to the toplevel MAINTAINERS file. do it.
+++ /dev/null
-.\" $OpenBSD: arc4random.3,v 1.19 2005/07/17 08:50:55 jaredy Exp $
-.\"
-.\" All rights reserved.
-.\"
-.\" Redistribution and use in source and binary forms, with or without
-.\" modification, are permitted provided that the following conditions
-.\" are met:
-.\" 1. Redistributions of source code must retain the above copyright
-.\" notice, this list of conditions and the following disclaimer.
-.\" 2. Redistributions in binary form must reproduce the above copyright
-.\" notice, this list of conditions and the following disclaimer in the
-.\" documentation and/or other materials provided with the distribution.
-.\" 3. All advertising materials mentioning features or use of this software
-.\" must display the following acknowledgement:
-.\" This product includes software developed by Niels Provos.
-.\" 4. The name of the author may not be used to endorse or promote products
-.\" derived from this software without specific prior written permission.
-.\"
-.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
-.\" IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
-.\" OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
-.\" IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
-.\" INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
-.\" NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-.\" DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-.\" THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-.\" (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
-.\" THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-.\"
-.\" Manual page, using -mandoc macros
-.\"
-.Dd April 15, 1997
-.Dt ARC4RANDOM 3
-.Os
-.Sh NAME
-.Nm arc4random ,
-.Nm arc4random_stir ,
-.Nm arc4random_addrandom
-.Nd arc4 random number generator
-.Sh SYNOPSIS
-.Fd #include <stdlib.h>
-.Ft uint32_t
-.Fn arc4random "void"
-.Ft void
-.Fn arc4random_stir "void"
-.Ft void
-.Fn arc4random_addrandom "u_char *dat" "int datlen"
-.Sh DESCRIPTION
-The
-.Fn arc4random
-function provides a high quality 32-bit pseudo-random
-number very quickly.
-.Fn arc4random
-seeds itself on a regular basis from the kernel strong random number
-subsystem described in
-.Xr random 4 .
-On each call, an ARC4 generator is used to generate a new result.
-The
-.Fn arc4random
-function uses the ARC4 cipher key stream generator,
-which uses 8*8 8-bit S-Boxes.
-The S-Boxes can be in about (2**1700) states.
-.Pp
-.Fn arc4random
-fits into a middle ground not covered by other subsystems such as
-the strong, slow, and resource expensive random
-devices described in
-.Xr random 4
-versus the fast but poor quality interfaces described in
-.Xr rand 3 ,
-.Xr random 3 ,
-and
-.Xr drand48 3 .
-.Pp
-The
-.Fn arc4random_stir
-function reads data from a pseudo-random device, usually
-.Pa /dev/urandom,
-and uses it to permute the S-Boxes via
-.Fn arc4random_addrandom .
-.Pp
-There is no need to call
-.Fn arc4random_stir
-before using
-.Fn arc4random ,
-since
-.Fn arc4random
-automatically initializes itself.
-.Sh SEE ALSO
-.Xr rand 3 ,
-.Xr rand48 3 ,
-.Xr random 3
-.Sh HISTORY
-An algorithm called
-.Pa RC4
-was designed by RSA Data Security, Inc.
-It was considered a trade secret.
-Because it was a trade secret, it obviously could not be patented.
-A clone of this was posted anonymously to USENET and confirmed to
-be equivalent by several sources who had access to the original cipher.
-Because of the trade secret situation, RSA Data Security, Inc. can do
-nothing about the release of the ARC4 algorithm.
-Since
-.Pa RC4
-used to be a trade secret, the cipher is now referred to as
-.Pa ARC4 .
-.Pp
-These functions first appeared in
-.Ox 2.1 .
+++ /dev/null
-.TH LDCONFIG 8 2005-08-15 uClibc "Linux Programmer's Manual"
-.SH NAME
-ldconfig \- updates symlinks and cache for shared libraries
-.SH SYNOPSIS
-.B ldconfig
-[
-.B -DvqnNX
-] [
-.B -f conf
-] [
-.B -C cache
-] [
-.B -r root
-]
-.B dir ...
-.br
-.B ldconfig -l
-[
-.B -Dv
-]
-.B lib ...
-.br
-.B ldconfig -p
-.SH DESCRIPTION
-.B ldconfig
-creates the necessary links and cache to the most recent shared libraries
-found in the directories specified on the command line, in the file
-\fI/etc/ld.so.conf\fR, and in the default trusted directories (\fI/lib\fR and
-\fI/usr/lib\fR). The cache is used by the run-time linker.
-.B ldconfig
-checks the header and file names of the libraries it encounters when
-determining which versions should have their links updated.
-
-.B Note:
-Some features may not exist depending on how uClibc was built
-.SH OPTIONS
-.TP
-.B -C cache
-use specified \fIcache\fR instead of default
-.TP
-.B -D
-debug mode, do not update links
-.TP
-.B -f conf
-use specified \fIconf\fR instead of default
-.TP
-.B -l
-library mode, manually link libraries
-.TP
-.B -n
-do not process standard trusted directories
-.TP
-.B -N
-do not update the library cache
-.TP
-.B -p
-print the current library cache
-.TP
-.B -q
-quiet mode, do not print warnings you should actually be reading
-.TP
-.B -r root
-chroot to \fIroot\fR before running
-.TP
-.B -v
-verbose mode, print things as we go, and generally be annoying
-.TP
-.B -X
-do not update the library links
-.SH BUGS
-Probably, make sure you complain when you find them:
-.br
-http://bugs.uclibc.org/
+++ /dev/null
-.TH LDD 1 2005-08-15 uClibc "Linux Programmer's Manual"
-.SH NAME
-ldd \- Print shared library dependencies
-.SH SYNOPSIS
-.B ldd
-[
-.B OPTIONS
-]
-.B FILE ...
-.SH DESCRIPTION
-Prints shared library dependencies.
-.SH OPTIONS
-None actually, we lied about that part.
-.SH BUGS
-Probably, make sure you complain when you find them:
-.br
-http://bugs.uclibc.org/
--- /dev/null
+Some notes to help future porters. Replace 'ARCH' with whatever arch
+you are hacking on.
+
+====================
+=== Config Files ===
+====================
+- create extra/Configs/Config.ARCH
+ See the other arch files for some good examples. powerpc/sparc/alpha
+ should be pretty simple templates.
+- add ARCH to the 'Target Architecture' list in extra/Configs/Config.in
+- Initially you will want to disable shared libraries, since making
+ the shared library loader work requires you first have basic architecture
+ support working. Thus you should add ARCH_HAS_NO_SHARED and
+ ARCH_HAS_NO_LDSO to Config.ARCH's TARGET_ARCH
+
+====================
+=== libc sysdeps ===
+====================
+(note: if glibc has already been ported to your arch, you can usually just
+ copy a lot of files from them rather than coding from scratch)
+- create libc/sysdeps/linux/ARCH
+- copy Makefile and Makefile.arch from libc/sysdeps/linux/i386/
+- set CSRC and SSRC to nothing in Makefile.arch for now
+
+- create crt1.S which defines the _start function ... you will probably want
+ to clear the frame pointer to make gdb happy, and then you will want to call
+ the funcion __uClibc_main() which takes these parameters:
+ __uClibc_main(main(), argc, argv, _init(), _fini())
+ Initially if you wish to make things easier on yourself, you can disable the
+ UCLIBC_CTOR_DTOR option and just set the init/fini arguments to NULL.
+ glibc generally stores this function in libc/sysdeps/ARCH/elf/start.S
+
+- create these additional files in ARCH/bits/
+
+ (template versions can be found in common/bits/ for you to tweak)
+ endian.h fcntl.h setjmp.h stackinfo.h uClibc_arch_features.h wordsize.h
+
+ kernel_types.h should be created based upon linux asm-ARCH/posix_types.h
+
+ copy linux asm-ARCH/stat.h to bits/kernel_stat.h
+
+ create syscalls.h based upon linux's unistd.h / glibc's sysdeps.h ... really
+ you just want to define the _syscall[0-6] macros. It is important that
+ these syscalls should be PIC safe (or you should provide a PIC and non-PIC
+ version) if you wish to properly support shared libraries.
+
+- at this point, you should have enough to generate a working HELLO WORLD
+ static binary
+
+- if you want UCLIBC_CTOR_DTOR support, you will need to create crti.S and
+ crtn.S files which define function prologues/epilogues.
+
+- for a more stable static port, you will need to create these files (and
+ update the Makefile.arch values accordingly)
+ __longjmp bsd-_setjmp bsd-setjmp brk clone setjmp syscall vfork
+ usually these are written in assembler, but you may be able to cheat and
+ write them in C ... see other ports for more information
+
+====================
+=== ldso sysdeps ===
+====================
+- elf.h - presumably you've already taught binutils all about the random ELF
+ relocations your arch needs, so now you need to make sure the defines exist
+ for uClibc. make sure the EM_### define exists and all of the R_###_###
+ reloc defines.
+
+- enable ldso/shared options in your extra/Configs/Config.ARCH file
+- you will need to create the following files in ldso/ldso/ARCH/
+ dl-startup.h dl-syscalls.h dl-sysdep.h elfinterp.c resolve.S
+
+- dl-startup.h:
+ - define the _start function which should call _dl_start which takes just one
+ parameter ... a pointer to argc (usually on the stack)
+ glibc stores this function in sysdeps/ARCH/dl-machine.h as RTLD_START
+ - define the GET_ARGV() macro which calculates the value of argv based upon
+ the parameter passed to _dl_start (usually it's simply just ARGS+1)
+ - define PERFORM_BOOTSTRAP_RELOC() macro which will handle just the relocs
+ that the ldso itself will generate
+
+- dl-syscalls.h:
+ if you wrote your bits/syscalls.h file correctly in the libc step above, you
+ can simply copy this file from another arch and be done ... otherwise you
+ will have to define the syscall[0-6] macros again, but this time setting
+ _dl_errno instead of just errno
+
+- dl-sysdep.h:
+ misc cruft goes in here ... you want to:
+ - either define or undefine ELF_USES_RELOCA
+ - define the INIT_GOT macro
+ - define MAGIC1 to the EM_### value your ELF arch uses
+ - define ELF_TARGET to a string name for your arch
+ - define the do_rem() macro
+ - define misc ALIGN macro's
+ - define elf_machine_type_class() macro
+ - define the inline functions elf_machine_dynamic, elf_machine_load_address,
+ and elf_machine_relative
+ glibc stores a bunch of these values in sysdeps/ARCH/dl-machine.h
+
+- elfinterp.c:
+ define all the relocation functions ... it's best if you just copy from
+ another arch which uses the same type of relocations (REL or RELA) and
+ start from there.
+
+- resolve.S:
+ front end of lazy relocation ... define the _dl_linux_resolve symbol which
+ is called by a PLT entry which has yet to be setup ... you will want to:
+ - set up arguments for _dl_linux_resolver()
+ - call _dl_linux_resolver()
+ - clean up after call
+ - jump to function address now stored in PLT
+ glibc stores this function in sysdeps/ARCH/dl-trampoline.S
+
+- utils/ldd.c - if you want support for ldso cache files (spoiler: you do),
+ then you'll need to teach ldd a little. generally, the fallback code
+ should be smart and "just work", but you should be explicit. just pop
+ it open and add an appropriate ifdef for your arch and set MATCH_MACHINE()
+ and ELFCLASSM. there are plenty examples and you're (hopefully) smart.
+
+====================
+=== Misc Cruft ===
+====================
+- MAINTAINERS - presumably you're going to submit this code back to mainline
+ and since you're the only one who cares about this arch (right now), you
+ should add yourself to the toplevel MAINTAINERS file. do it.
+++ /dev/null
-/* Small test program for probing how various math functions
- * with specific operands set floating point exceptions
- */
-
-#define _ISOC99_SOURCE 1
-#define _GNU_SOURCE 1
-
-#include <stdint.h>
-#include <math.h>
-#include <fenv.h>
-#include <stdio.h>
-
-int main(int argc, char **argv)
-{
- float largest, small, t, inf_float;
-
- largest = small = 1;
- while (1) {
- t = largest + small;
- /* optimizations may make plain "t == largest" unreliable */
- if (memcmp(&t, &largest, sizeof(float)) == 0)
- break;
- if (isfinite(t)) {
- largest = t;
- small *= 2;
- continue;
- }
- small /= 2;
- }
- inf_float = largest + largest;
- //printf("%.40g ", largest);
- //printf("[%llx]\n", (long long) (*(uint32_t *)&largest));
-
- feclearexcept(FE_ALL_EXCEPT);
-
- //t = 1.0 / 0.0; // simple test: FE_DIVBYZERO
- //t = nextafterf(largest, 1); // glibc 2.8: no math exceptions raised
- //t = nextafterf(largest, largest); // glibc 2.8: no math exceptions raised
- //t = nextafterf(largest, inf_float); // glibc 2.8: FE_INEXACT FE_OVERFLOW
-
-#define PREX(ex) do { if (fetestexcept(ex)) printf(#ex " "); } while(0)
-#ifdef FE_INEXACT
- PREX(FE_INEXACT);
-#endif
-#ifdef FE_DIVBYZERO
- PREX(FE_DIVBYZERO);
-#endif
-#ifdef FE_UNDERFLOW
- PREX(FE_UNDERFLOW);
-#endif
-#ifdef FE_OVERFLOW
- PREX(FE_OVERFLOW);
-#endif
-#ifdef FE_INVALID
- PREX(FE_INVALID);
-#endif
- if (fetestexcept(FE_ALL_EXCEPT))
- printf("\n");
- else
- printf("no math exceptions raised\n");
-
- printf("%.40g\n", t);
- return 0;
-}
+++ /dev/null
- All what you never wanted to know about sigaction(),
- struct sigaction, and sigset_t.
-
-
-Before vda started messing with sigset_t, struct sigaction
-and sigaction() functions, things looked this way:
-
-
- Structures
-
-MIPS:
-
-Ignoring bogus "#if defined(__mips__) ..." block in
-libc/sysdeps/linux/common/bits/kernel_sigaction.h
-and using
-libc/sysdeps/linux/mips/bits/kernel_sigaction.h
-as an authoritative source:
-
-HAVE_SA_RESTORER is #defined
-struct old_kernel_sigaction {
- unsigned sa_flags;
- sighandler_t k_sa_handler;
- unsigned long sa_mask;
- unsigned pad0[3]; /* reserved, keep size constant */
- /* Abi says here follows reserved int[2] */
- void (*sa_restorer)(void);
-#if (_MIPS_SZPTR < 64)
- /* For 32 bit code we have to pad struct sigaction to get
- * constant size for the ABI */
- int pad1[1]; /* reserved */
-#endif
-};
-struct kernel_sigaction {
- unsigned int sa_flags;
- sighandler_t k_sa_handler;
- kernel_sigset_t sa_mask;
- void (*sa_restorer)(void);
- int s_resv[1]; /* reserved */
-};
-struct sigaction {
- unsigned sa_flags;
- sighandler_t sa_handler;
- sigset_t sa_mask;
- /* The ABI says here are two unused ints following. */
- /* Restore handler. */
- void (*sa_restorer)(void);
-#if _MIPS_SZPTR < 64
- int sa_resv[1];
-#endif
-};
-
-IA64:
-
-Has no old_sigaction. What a relief.
-
-struct kernel_sigaction {
- sighandler_t k_sa_handler;
- unsigned long sa_flags;
- sigset_t sa_mask;
-};
-struct sigaction {
- sighandler_t sa_handler;
- unsigned long sa_flags;
- sigset_t sa_mask;
-};
-
-Alpha:
-
-struct old_kernel_sigaction {
- sighandler_t k_sa_handler;
- unsigned long sa_mask;
- unsigned sa_flags;
-};
-struct kernel_sigaction {
- sighandler_t k_sa_handler;
- unsigned sa_flags;
- sigset_t sa_mask;
-};
-struct sigaction {
- sighandler_t sa_handler;
- sigset_t sa_mask;
- unsigned sa_flags;
-};
-
-HPPA:
-
-struct kernel_sigaction {
- sighandler_t k_sa_handler;
- unsigned long sa_flags;
- sigset_t sa_mask;
-};
-struct sigaction {
- sighandler_t sa_handler;
- unsigned long sa_flags;
- sigset_t sa_mask;
-};
-
-The rest, kernel side:
-
-HAVE_SA_RESTORER #defined
-struct old_kernel_sigaction {
- sighandler_t k_sa_handler;
- unsigned long sa_mask;
- unsigned long sa_flags;
- void (*sa_restorer)(void);
-};
-struct kernel_sigaction {
- sighandler_t k_sa_handler;
- unsigned long sa_flags;
- void (*sa_restorer)(void);
- sigset_t sa_mask;
-};
-
-On userspace side, Sparc has special struct sigaction:
-
-struct sigaction {
- sighandler_t sa_handler;
- sigset_t sa_mask;
- unsigned long sa_flags;
- void (*sa_restorer)(void); /* Not used by Linux/Sparc */
-};
-
-And finally the rest has:
-
-struct sigaction {
- sighandler_t sa_handler;
- sigset_t sa_mask;
- int sa_flags;
- void (*sa_restorer)(void);
-};
-
-Userspace sigset_t was uniformly defined as vector of longs
-big enough to hold 1024 (!) bits - carried over from glibc.
-Since the only arch whose struct kernel_sigaction contains sa_mask
-not as a last member is MIPS, MIPS has special kernel_sigset_t,
-which is an array of longs long enough for 128 bits.
-Other arches still used userspace sigset_t in struct kernel_sigaction,
-but it did not really matter because overlong kernel_sigaction
-does not hurt in sigaction() [explained below].
-On kernel side, all arches define _NSIG to 65 (meaning
-there are 64 signals, 1..64) except MIPS, which define it to 129.
-
-
- Functions
-
-sigaction() [libc function] usually has two kernel_sigaction's
-on stack and copy (userspace) struct sigaction members into
-first one, executes syscall, then pulls out the result from
-second one. This accomodates differences in layouts of structs.
-
-The only typically present quirk is what to do with sa_restorer.
-
- libc/sysdeps/linux/arm/sigaction.c
-
-if HAVE_SA_RESTORER and (sa_flags & SA_RESTORER) is not set,
-sets sa_restorer to
-(flags & SA_SIGINFO) ? __default_rt_sa_restorer : __default_sa_restorer,
-and sets SA_RESTORER,
-otherwise passes it as-is. Which is kinda strange, because AFAICS
-HAVE_SA_RESTORER is *not* defined for ARM.
-
- libc/sysdeps/linux/i386/sigaction.c
-
-Forcibly sets SA_RESTORER and sa_restorer:
-kact.sa_flags = act->sa_flags | SA_RESTORER;
-kact.sa_restorer = ((act->sa_flags & SA_SIGINFO) ? &restore_rt : &restore);
-
- libc/sysdeps/linux/x86_64/sigaction.c
-
-Forcibly sets SA_RESTORER and sa_restorer:
-kact.sa_flags = act->sa_flags | SA_RESTORER;
-kact.sa_restorer = &restore_rt;
-
- libc/sysdeps/linux/mips/sigaction.c
-
-# ifdef HAVE_SA_RESTORER
-# if _MIPS_SIM == _ABIO32
- kact.sa_restorer = act->sa_restorer;
-# else
- kact.sa_restorer = &restore_rt;
-# endif
-# endif
-No confusion here, HAVE_SA_RESTORER is #defined for MIPS
-
- libc/sysdeps/linux/avr32/sigaction.c
-
-if (kact.sa_flags & SA_RESTORER) {
- kact.sa_restorer = act->sa_restorer;
-} else {
- kact.sa_restorer = __default_rt_sa_restorer;
- kact.sa_flags |= SA_RESTORER;
-}
-Does not check HAVE_SA_RESTORER, but avr32 falls
-in "completely ordinary" category on both kernel and
-userspace sides, and those have it defined.
-
- libc/sysdeps/linux/xtensa/sigaction.c
-
-if (kact.sa_flags & SA_RESTORER) {
- kact.sa_restorer = act->sa_restorer;
-} else {
- kact.sa_restorer = __default_sa_restorer;
- kact.sa_flags |= SA_RESTORER;
-}
-Thus, similar to avr32.
-
- libc/signal/sigaction.c (i.e. the all other arches)
-
-# ifdef HAVE_SA_RESTORER
- kact.sa_restorer = act->sa_restorer;
-# endif
-Plain translation, just sa_restorer copy is protected
-by HAVE_SA_RESTORER #define check. Looks like here
-HAVE_SA_RESTORER will be undef'ed only for IA64,
-Alpha an HPPA.
-
-
- Proposed overhaul past 0.9.30
-
-Since we can define libc-side structures at will:
-make sigset_t and struct sigaction identical on kernel side and libc side
-within each arch. If arches do not need special handling of sa_restorer,
-then sigaction() can directly use passed struct sigaction as-is.
-Otherwise, a copy is still needed, although sigaction() might have
-just one struct kernel_sigaction on stack and use it both for passing
-data to kernel and for receiving it back. Might save a few bytes.
-
-To this effect:
-
-* Make sigset_t size match kernel side on all arches.
- This is easy since all arches have 64 signals and only MIPS has 128.
-
-* Modify libc/sysdeps/linux/$ARCH/bits/sigaction.h
- so that its struct sigaction matches kernel's. If sa_restorer
- field is present in libc but is missing in kernel_sigaction,
- add it at the bottom in order to not mess up kernel_sigaction layout.
-
-* Modify libc/sysdeps/linux/$ARCH/sigaction.c
- to implement the logic above. In "common" pseudo-arch
- (libc/signal/sigaction.c file),
- we would not even need to do any copying, as described above.
-
-* Document discovered arch quirks while debugging this mess.
-
-* struct old_kernel_sigaction can't be disposed of in a similar way,
- we need to have userspace struct sigaction unchanged regardless
- whether we use "old" or "new" kernel sigaction() syscall.
- It's moot anyway because "old" one is long unused, it's from
- pre-2.2 kernels.
projects / uclibc-ng.git / commitdiff