openwrt/toolchain/musl/patches/001-git-2015-06-25.patch

commit 6ba5517a460c6c438f64d69464fdfc3269a4c91a
Author: Rich Felker <dalias@aerifal.cx>
Date:   Thu Jun 25 22:22:00 2015 +0000

    fix local-dynamic model TLS on mips and powerpc
    
    the TLS ABI spec for mips, powerpc, and some other (presently
    unsupported) RISC archs has the return value of __tls_get_addr offset
    by +0x8000 and the result of DTPOFF relocations offset by -0x8000. I
    had previously assumed this part of the ABI was actually just an
    implementation detail, since the adjustments cancel out. however, when
    the local dynamic model is used for accessing TLS that's known to be
    in the same DSO, either of the following may happen:
    
    1. the -0x8000 offset may already be applied to the argument structure
    passed to __tls_get_addr at ld time, without any opportunity for
    runtime relocations.
    
    2. __tls_get_addr may be used with a zero offset argument to obtain a
    base address for the module's TLS, to which the caller then applies
    immediate offsets for individual objects accessed using the local
    dynamic model. since the immediate offsets have the -0x8000 adjustment
    applied to them, the base address they use needs to include the
    +0x8000 offset.
    
    it would be possible, but more complex, to store the pointers in the
    dtv[] array with the +0x8000 offset pre-applied, to avoid the runtime
    cost of adding 0x8000 on each call to __tls_get_addr. this change
    could be made later if measurements show that it would help.

commit ce337daa00e42d4f2d9a4d9ae0ed51b20249d924
Author: Rich Felker <dalias@aerifal.cx>
Date:   Tue Jun 23 04:03:42 2015 +0000

    make dynamic linker work around MAP_FAILED mmap failure on nommu kernels
    
    previously, loading of additional libraries beyond libc/ldso did not
    work on nommu kernels, nor did loading programs via invocation of the
    dynamic linker as a command.

commit a59341420fdedb288d9ff80e73609ae44e9cf258
Author: Rich Felker <dalias@aerifal.cx>
Date:   Tue Jun 23 00:12:25 2015 +0000

    reimplement strverscmp to fix corner cases
    
    this interface is non-standardized and is a GNU invention, and as
    such, our implementation should match the behavior of the GNU
    function. one peculiarity the old implementation got wrong was the
    handling of all-zero digit sequences: they are supposed to compare
    greater than digit sequences of which they are a proper prefix, as in
    009 < 00.
    
    in addition, high bytes were treated with char signedness rather than
    as unsigned. this was wrong regardless of what the GNU function does
    since the resulting order relation varied by arch.
    
    the new strverscmp implementation makes explicit the cases where the
    order differs from what strcmp would produce, of which there are only
    two.

commit 153e952e1a688859d7095345b17e6c1df74a295c
Author: Rich Felker <dalias@aerifal.cx>
Date:   Mon Jun 22 20:33:28 2015 +0000

    fix regression/typo that disabled __simple_malloc when calloc is used
    
    commit ba819787ee93ceae94efd274f7849e317c1bff58 introduced this
    regression. since the __malloc0 weak alias was not properly provided
    by __simple_malloc, use of calloc forced the full malloc to be linked.

commit ba819787ee93ceae94efd274f7849e317c1bff58
Author: Rich Felker <dalias@aerifal.cx>
Date:   Mon Jun 22 18:50:09 2015 +0000

    fix calloc when __simple_malloc implementation is used
    
    previously, calloc's implementation encoded assumptions about the
    implementation of malloc, accessing a size_t word just prior to the
    allocated memory to determine if it was obtained by mmap to optimize
    out the zero-filling. when __simple_malloc is used (static linking a
    program with no realloc/free), it doesn't matter if the result of this
    check is wrong, since all allocations are zero-initialized anyway. but
    the access could be invalid if it crosses a page boundary or if the
    pointer is not sufficiently aligned, which can happen for very small
    allocations.
    
    this patch fixes the issue by moving the zero-fill logic into malloc.c
    with the full malloc, as a new function named __malloc0, which is
    provided by a weak alias to __simple_malloc (which always gives
    zero-filled memory) when the full malloc is not in use.

commit 55d061f031085f24d138664c897791aebe9a2fab
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sat Jun 20 03:01:07 2015 +0000

    provide __stack_chk_fail_local in libc.a
    
    this symbol is needed only on archs where the PLT call ABI is klunky,
    and only for position-independent code compiled with stack protector.
    thus references usually only appear in shared libraries or PIE
    executables, but they can also appear when linking statically if some
    of the object files being linked were built as PIC/PIE.
    
    normally libssp_nonshared.a from the compiler toolchain should provide
    __stack_chk_fail_local, but reportedly it appears prior to -lc in the
    link order, thus failing to satisfy references from libc itself (which
    arise only if libc.a was built as PIC/PIE with stack protector
    enabled).

commit ce3688eca920aa77549323f84e21f33522397115
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sat Jun 20 02:54:30 2015 +0000

    work around mips detached thread exit breakage due to kernel regression
    
    linux kernel commit 46e12c07b3b9603c60fc1d421ff18618241cb081 caused
    the mips syscall mechanism to fail with EFAULT when the userspace
    stack pointer is invalid, breaking __unmapself used for detached
    thread exit. the workaround is to set $sp to a known-valid, readable
    address, and the simplest one to obtain is the address of the current
    function, which is available (per o32 calling convention) in $25.

commit 75eceb3ae824d54e865686c0c538551aeebf3372
Author: Rich Felker <dalias@aerifal.cx>
Date:   Wed Jun 17 17:21:46 2015 +0000

    ignore ENOSYS error from mprotect in pthread_create and dynamic linker
    
    this error simply indicated a system without memory protection (NOMMU)
    and should not cause failure in the caller.

commit 10d0268ccfab9152250eeeed3952ce3fed44131a
Author: Rich Felker <dalias@aerifal.cx>
Date:   Tue Jun 16 15:25:02 2015 +0000

    switch to using trap number 31 for syscalls on sh
    
    nominally the low bits of the trap number on sh are the number of
    syscall arguments, but they have never been used by the kernel, and
    some code making syscalls does not even know the number of arguments
    and needs to pass an arbitrary high number anyway.
    
    sh3/sh4 traditionally used the trap range 16-31 for syscalls, but part
    of this range overlapped with hardware exceptions/interrupts on sh2
    hardware, so an incompatible range 32-47 was chosen for sh2.
    
    using trap number 31 everywhere, since it's in the existing sh3/sh4
    range and does not conflict with sh2 hardware, is a proposed
    unification of the kernel syscall convention that will allow binaries
    to be shared between sh2 and sh3/sh4. if this is not accepted into the
    kernel, we can refit the sh2 target with runtime selection mechanisms
    for the trap number, but doing so would be invasive and would entail
    non-trivial overhead.

commit 3366a99b17847b58f2d8cc52cbb5d65deb824f8a
Author: Rich Felker <dalias@aerifal.cx>
Date:   Tue Jun 16 14:55:06 2015 +0000

    switch sh port's __unmapself to generic version when running on sh2/nommu
    
    due to the way the interrupt and syscall trap mechanism works,
    userspace on sh2 must never set the stack pointer to an invalid value.
    thus, the approach used on most archs, where __unmapself executes with
    no stack for the interval between SYS_munmap and SYS_exit, is not
    viable on sh2.
    
    in order not to pessimize sh3/sh4, the sh asm version of __unmapself
    is not removed. instead it's renamed and redirected through code that
    calls either the generic (safe) __unmapself or the sh3/sh4 asm,
    depending on compile-time and run-time conditions.

commit f9d84554bae0fa17c9a1d724549c4408022228a5
Author: Rich Felker <dalias@aerifal.cx>
Date:   Tue Jun 16 14:28:30 2015 +0000

    add support for sh2 interrupt-masking-based atomics to sh port
    
    the sh2 target is being considered an ISA subset of sh3/sh4, in the
    sense that binaries built for sh2 are intended to be usable on later
    cpu models/kernels with mmu support. so rather than hard-coding
    sh2-specific atomics, the runtime atomic selection mechanisms that was
    already in place has been extended to add sh2 atomics.
    
    at this time, the sh2 atomics are not SMP-compatible; since the ISA
    lacks actual atomic operations, the new code instead masks interrupts
    for the duration of the atomic operation, producing an atomic result
    on single-core. this is only possible because the kernel/hardware does
    not impose protections against userspace doing so. additional changes
    will be needed to support future SMP systems.
    
    care has been taken to avoid producing significant additional code
    size in the case where it's known at compile-time that the target is
    not sh2 and does not need sh2-specific code.

commit 1b0cdc8700d29ef018bf226d74b2b58b23bce91c
Author: Rich Felker <dalias@aerifal.cx>
Date:   Tue Jun 16 07:11:19 2015 +0000

    refactor stdio open file list handling, move it out of global libc struct
    
    functions which open in-memory FILE stream variants all shared a tail
    with __fdopen, adding the FILE structure to stdio's open file list.
    replacing this common tail with a function call reduces code size and
    duplication of logic. the list is also partially encapsulated now.
    
    function signatures were chosen to facilitate tail call optimization
    and reduce the need for additional accessor functions.
    
    with these changes, static linked programs that do not use stdio no
    longer have an open file list at all.

commit f22a9edaf8a6f2ca1d314d18b3785558279a5c03
Author: Rich Felker <dalias@aerifal.cx>
Date:   Tue Jun 16 06:18:00 2015 +0000

    byte-based C locale, phase 3: make MB_CUR_MAX variable to activate code
    
    this patch activates the new byte-based C locale (high bytes treated
    as abstract code unit "characters" rather than decoded as multibyte
    characters) by making the value of MB_CUR_MAX depend on the active
    locale. for the C locale, the LC_CTYPE category pointer is null,
    yielding a value of 1. all other locales yield a value of 4.

commit 16f18d036d9a7bf590ee6eb86785c0a9658220b6
Author: Rich Felker <dalias@aerifal.cx>
Date:   Tue Jun 16 05:35:31 2015 +0000

    byte-based C locale, phase 2: stdio and iconv (multibyte callers)
    
    this patch adjusts libc components which use the multibyte functions
    internally, and which depend on them operating in a particular
    encoding, to make the appropriate locale changes before calling them
    and restore the calling thread's locale afterwards. activating the
    byte-based C locale without these changes would cause regressions in
    stdio and iconv.
    
    in the case of iconv, the current implementation was simply using the
    multibyte functions as UTF-8 conversions. setting a multibyte UTF-8
    locale for the duration of the iconv operation allows the code to
    continue working.
    
    in the case of stdio, POSIX requires that FILE streams have an
    encoding rule bound at the time of setting wide orientation. as long
    as all locales, including the C locale, used the same encoding,
    treating high bytes as UTF-8, there was no need to store an encoding
    rule as part of the stream's state.
    
    a new locale field in the FILE structure points to the locale that
    should be made active during fgetwc/fputwc/ungetwc on the stream. it
    cannot point to the locale active at the time the stream becomes
    oriented, because this locale could be mutable (the global locale) or
    could be destroyed (locale_t objects produced by newlocale) before the
    stream is closed. instead, a pointer to the static C or C.UTF-8 locale
    object added in commit commit aeeac9ca5490d7d90fe061ab72da446c01ddf746
    is used. this is valid since categories other than LC_CTYPE will not
    affect these functions.

commit 1507ebf837334e9e07cfab1ca1c2e88449069a80
Author: Rich Felker <dalias@aerifal.cx>
Date:   Tue Jun 16 04:44:17 2015 +0000

    byte-based C locale, phase 1: multibyte character handling functions
    
    this patch makes the functions which work directly on multibyte
    characters treat the high bytes as individual abstract code units
    rather than as multibyte sequences when MB_CUR_MAX is 1. since
    MB_CUR_MAX is presently defined as a constant 4, all of the new code
    added is dead code, and optimizing compilers' code generation should
    not be affected at all. a future commit will activate the new code.
    
    as abstract code units, bytes 0x80 to 0xff are represented by wchar_t
    values 0xdf80 to 0xdfff, at the end of the surrogates range. this
    ensures that they will never be misinterpreted as Unicode characters,
    and that all wctype functions return false for these "characters"
    without needing locale-specific logic. a high range outside of Unicode
    such as 0x7fffff80 to 0x7fffffff was also considered, but since C11's
    char16_t also needs to be able to represent conversions of these
    bytes, the surrogate range was the natural choice.

commit 38e2f727237230300fea6aff68802db04625fd23
Author: Rich Felker <dalias@aerifal.cx>
Date:   Tue Jun 16 04:21:38 2015 +0000

    fix btowc corner case
    
    btowc is required to interpret its argument by conversion to unsigned
    char, unless the argument is equal to EOF. since the conversion to
    produces a non-character value anyway, we can just unconditionally
    convert, for now.

commit ee59c296d56bf26f49f354d6eb32b4b6d4190188
Author: Szabolcs Nagy <nsz@port70.net>
Date:   Wed Jun 3 10:32:14 2015 +0100

    arm: add vdso support
    
    vdso will be available on arm in linux v4.2, the user-space code
    for it is in kernel commit 8512287a8165592466cb9cb347ba94892e9c56a5

commit e3bc22f1eff87b8f029a6ab31f1a269d69e4b053
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sun Jun 14 01:59:02 2015 +0000

    refactor malloc's expand_heap to share with __simple_malloc
    
    this extends the brk/stack collision protection added to full malloc
    in commit 276904c2f6bde3a31a24ebfa201482601d18b4f9 to also protect the
    __simple_malloc function used in static-linked programs that don't
    reference the free function.
    
    it also extends support for using mmap when brk fails, which full
    malloc got in commit 5446303328adf4b4e36d9fba21848e6feb55fab4, to
    __simple_malloc.
    
    since __simple_malloc may expand the heap by arbitrarily large
    increments, the stack collision detection is enhanced to detect
    interval overlap rather than just proximity of a single address to the
    stack. code size is increased a bit, but this is partly offset by the
    sharing of code between the two malloc implementations, which due to
    linking semantics, both get linked in a program that needs the full
    malloc with realloc/free support.

commit 4ef9b828c1f39553a69e0635ac91f0fcadd6e8c6
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sat Jun 13 20:53:02 2015 +0000

    remove cancellation points in stdio
    
    commit 58165923890865a6ac042fafce13f440ee986fd9 added these optional
    cancellation points on the basis that cancellable stdio could be
    useful, to unblock threads stuck on stdio operations that will never
    complete. however, the only way to ensure that cancellation can
    achieve this is to violate the rules for side effects when
    cancellation is acted upon, discarding knowledge of any partial data
    transfer already completed. our implementation exhibited this behavior
    and was thus non-conforming.
    
    in addition to improving correctness, removing these cancellation
    points moderately reduces code size, and should significantly improve
    performance on i386, where sysenter/syscall instructions can be used
    instead of "int $128" for non-cancellable syscalls.

commit 536c6d5a4205e2a3f161f2983ce1e0ac3082187d
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sat Jun 13 05:17:16 2015 +0000

    fix idiom for setting stdio stream orientation to wide
    
    the old idiom, f->mode |= f->mode+1, was adapted from the idiom for
    setting byte orientation, f->mode |= f->mode-1, but the adaptation was
    incorrect. unless the stream was alreasdy set byte-oriented, this code
    incremented f->mode each time it was executed, which would eventually
    lead to overflow. it could be fixed by changing it to f->mode |= 1,
    but upcoming changes will require slightly more work at the time of
    wide orientation, so it makes sense to just call fwide. as an
    optimization in the single-character functions, fwide is only called
    if the stream is not already wide-oriented.

commit f8f565df467c13248104223f99abf7f37cef7584
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sat Jun 13 04:42:38 2015 +0000

    add printing of null %s arguments as "(null)" in wide printf
    
    this is undefined, but supported in our implementation of the normal
    printf, so for consistency the wide variant should support it too.

commit f9e25d813860d53cd1e9b6145cc63375d2fe2529
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sat Jun 13 04:37:27 2015 +0000

    add %m support to wide printf

commit ec634aad91f57479ef17525e33ed446c780a61f4
Author: Rich Felker <dalias@aerifal.cx>
Date:   Thu Jun 11 05:01:04 2015 +0000

    add sh asm for vfork

commit c30cbcb0a646b1f13a22c645616dce624465b883
Author: Rich Felker <dalias@aerifal.cx>
Date:   Wed Jun 10 02:27:40 2015 +0000

    implement arch-generic version of __unmapself
    
    this can be used to put off writing an asm version of __unmapself for
    new archs, or as a permanent solution on archs where it's not
    practical or even possible to run momentarily with no stack.
    
    the concept here is simple: the caller takes a lock on a global shared
    stack and uses it to make the munmap and exit syscalls. the only trick
    is unlocking, which must be done after the thread exits, and this is
    achieved by using the set_tid_address syscall to have the kernel zero
    and futex-wake the lock word as part of the exit syscall.

commit 276904c2f6bde3a31a24ebfa201482601d18b4f9
Author: Rich Felker <dalias@aerifal.cx>
Date:   Tue Jun 9 20:30:35 2015 +0000

    in malloc, refuse to use brk if it grows into stack
    
    the linux/nommu fdpic ELF loader sets up the brk range to overlap
    entirely with the main thread's stack (but growing from opposite
    ends), so that the resulting failure mode for malloc is not to return
    a null pointer but to start returning pointers to memory that overlaps
    with the caller's stack. needless to say this extremely dangerous and
    makes brk unusable.
    
    since it's non-trivial to detect execution environments that might be
    affected by this kernel bug, and since the severity of the bug makes
    any sort of detection that might yield false-negatives unsafe, we
    instead check the proximity of the brk to the stack pointer each time
    the brk is to be expanded. both the main thread's stack (where the
    real known risk lies) and the calling thread's stack are checked. an
    arbitrary gap distance of 8 MB is imposed, chosen to be larger than
    linux default main-thread stack reservation sizes and larger than any
    reasonable stack configuration on nommu.
    
    the effeciveness of this patch relies on an assumption that the amount
    by which the brk is being grown is smaller than the gap limit, which
    is always true for malloc's use of brk. reliance on this assumption is
    why the check is being done in malloc-specific code and not in __brk.

commit bd1eaceaa3975bd2a2a34e211cff896affaecadf
Author: Rich Felker <dalias@aerifal.cx>
Date:   Tue Jun 9 20:09:27 2015 +0000

    fix spurious errors from pwd/grp functions when nscd backend is absent
    
    for several pwd/grp functions, the only way the caller can distinguish
    between a successful negative result ("no such user/group") and an
    internal error is by clearing errno before the call and checking errno
    afterwards. the nscd backend support code correctly simulated a
    not-found response on systems where such a backend is not running, but
    failed to restore errno.
    
    this commit also fixed an outdated/incorrect comment.

commit 75ce4503950621b11fcc7f1fd1187dbcf3cde312
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sun Jun 7 20:55:23 2015 +0000

    fix regression in pre-v7 arm on kernels with kuser helper removed
    
    the arm atomics/TLS runtime selection code is called from
    __set_thread_area and depends on having libc.auxv and __hwcap
    available. commit 71f099cb7db821c51d8f39dfac622c61e54d794c moved the
    first call to __set_thread_area to the top of dynamic linking stage 3,
    before this data is made available, causing the runtime detection code
    to always see __hwcap as zero and thereby select the atomics/TLS
    implementations based on kuser helper.
    
    upcoming work on superh will use similar runtime detection.
    
    ideally this early-init code should be cleanly refactored and shared
    between the dynamic linker and static-linked startup.

commit 32f3c4f70633488550c29a2444f819aafdf345ff
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sun Jun 7 03:09:16 2015 +0000

    add multiple inclusion guard to locale_impl.h

commit 04b8360adbb6487f61aa0c00e53ec3a90a5a0d29
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sun Jun 7 02:59:49 2015 +0000

    remove redefinition of MB_CUR_MAX in locale_impl.h
    
    unless/until the byte-based C locale is implemented, defining
    MB_CUR_MAX to 1 in the C locale is wrong. no internal code currently
    uses the MB_CUR_MAX macro, but having it defined inconsistently is
    error-prone. applications get the value from stdlib.h and were
    unaffected.

commit 16bf466532d7328e971012b0731ad493b017ad29
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sat Jun 6 18:53:02 2015 +0000

    make static C and C.UTF-8 locales available outside of newlocale

commit 312eea2ea4f4363fb01b73660c08bfcf43dd3bb4
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sat Jun 6 18:20:30 2015 +0000

    remove another invalid skip of locking in ungetwc

commit 3d7e32d28dc9962e9efc1c317c5b44b5b2df3008
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sat Jun 6 18:16:22 2015 +0000

    add macro version of ctype.h isascii function
    
    presumably internal code (ungetwc and fputwc) was written assuming a
    macro implementation existed; otherwise use of isascii is just a
    pessimization.

commit 7e816a6487932cbb3cb71d94b609e50e81f4e5bf
Author: Rich Felker <dalias@aerifal.cx>
Date:   Sat Jun 6 18:11:17 2015 +0000

    remove invalid skip of locking in ungetwc
    
    aside from being invalid, the early check only optimized the error
    case, and likely pessimized the common case by separating the
    two branches on isascii(c) at opposite ends of the function.

commit 63f4b9f18f3674124d8bcb119739fec85e6da005
Author: Timo Teräs <timo.teras@iki.fi>
Date:   Fri Jun 5 10:39:42 2015 +0300

    fix uselocale((locale_t)0) not to modify locale
    
    commit 68630b55c0c7219fe9df70dc28ffbf9efc8021d8 made the new locale to
    be assigned unconditonally resulting in crashes later on.

diff --git a/arch/arm/syscall_arch.h b/arch/arm/syscall_arch.h
index 199ad2a..64461ec 100644
--- a/arch/arm/syscall_arch.h
+++ b/arch/arm/syscall_arch.h
@@ -72,3 +72,7 @@ static inline long __syscall6(long n, long a, long b, long c, long d, long e, lo
 	register long r5 __asm__("r5") = f;
 	__asm_syscall("r"(r7), "0"(r0), "r"(r1), "r"(r2), "r"(r3), "r"(r4), "r"(r5));
 }
+
+#define VDSO_USEFUL
+#define VDSO_CGT_SYM "__vdso_clock_gettime"
+#define VDSO_CGT_VER "LINUX_2.6"
diff --git a/arch/mips/pthread_arch.h b/arch/mips/pthread_arch.h
index f8e35ae..904a248 100644
--- a/arch/mips/pthread_arch.h
+++ b/arch/mips/pthread_arch.h
@@ -13,4 +13,6 @@ static inline struct pthread *__pthread_self()
 #define TLS_ABOVE_TP
 #define TP_ADJ(p) ((char *)(p) + sizeof(struct pthread) + 0x7000)
 
+#define DTP_OFFSET 0x8000
+
 #define CANCEL_REG_IP (3-(union {int __i; char __b;}){1}.__b)
diff --git a/arch/powerpc/pthread_arch.h b/arch/powerpc/pthread_arch.h
index 4115ec8..1cbfc22 100644
--- a/arch/powerpc/pthread_arch.h
+++ b/arch/powerpc/pthread_arch.h
@@ -12,6 +12,8 @@ static inline struct pthread *__pthread_self()
 #define TLS_ABOVE_TP
 #define TP_ADJ(p) ((char *)(p) + sizeof(struct pthread) + 0x7000)
 
+#define DTP_OFFSET 0x8000
+
 // offset of the PC register in mcontext_t, divided by the system wordsize
 // the kernel calls the ip "nip", it's the first saved value after the 32
 // GPRs.
diff --git a/arch/sh/src/__set_thread_area.c b/arch/sh/src/__set_thread_area.c
new file mode 100644
index 0000000..1d3e022
--- /dev/null
+++ b/arch/sh/src/__set_thread_area.c
@@ -0,0 +1,34 @@
+#include "pthread_impl.h"
+#include "libc.h"
+#include "sh_atomic.h"
+#include <elf.h>
+
+/* Also perform sh-specific init */
+
+#define CPU_HAS_LLSC 0x0040
+
+__attribute__((__visibility__("hidden"))) unsigned __sh_atomic_model, __sh_nommu;
+
+int __set_thread_area(void *p)
+{
+	size_t *aux;
+	__asm__ __volatile__ ( "ldc %0, gbr" : : "r"(p) : "memory" );
+#ifndef __SH4A__
+	if (__hwcap & CPU_HAS_LLSC) {
+		__sh_atomic_model = SH_A_LLSC;
+		return 0;
+	}
+#if !defined(__SH3__) && !defined(__SH4__)
+	for (aux=libc.auxv; *aux; aux+=2) {
+		if (*aux != AT_PLATFORM) continue;
+		const char *s = (void *)aux[1];
+		if (s[0]!='s' || s[1]!='h' || s[2]!='2' || s[3]-'0'<10u) break;
+		__sh_atomic_model = SH_A_IMASK;
+		__sh_nommu = 1;
+		return 0;
+	}
+#endif
+	/* __sh_atomic_model = SH_A_GUSA; */ /* 0, default */
+#endif
+	return 0;
+}
diff --git a/arch/sh/src/__unmapself.c b/arch/sh/src/__unmapself.c
new file mode 100644
index 0000000..4df9e7b
--- /dev/null
+++ b/arch/sh/src/__unmapself.c
@@ -0,0 +1,19 @@
+#include "pthread_impl.h"
+
+void __unmapself_sh_mmu(void *, size_t);
+void __unmapself_sh_nommu(void *, size_t);
+
+#if !defined(__SH3__) && !defined(__SH4__)
+#define __unmapself __unmapself_sh_nommu
+#include "../../../src/thread/__unmapself.c"
+#undef __unmapself
+extern __attribute__((__visibility__("hidden"))) unsigned __sh_nommu;
+#else
+#define __sh_nommu 0
+#endif
+
+void __unmapself(void *base, size_t size)
+{
+	if (__sh_nommu) __unmapself_sh_nommu(base, size);
+	else __unmapself_sh_mmu(base, size);
+}
diff --git a/arch/sh/src/atomic.c b/arch/sh/src/atomic.c
index f8c615f..7fd7307 100644
--- a/arch/sh/src/atomic.c
+++ b/arch/sh/src/atomic.c
@@ -1,8 +1,26 @@
 #ifndef __SH4A__
 
+#include "sh_atomic.h"
 #include "atomic.h"
 #include "libc.h"
 
+static inline unsigned mask()
+{
+	unsigned sr;
+	__asm__ __volatile__ ( "\n"
+	"	stc sr,r0 \n"
+	"	mov r0,%0 \n"
+	"	or #0xf0,r0 \n"
+	"	ldc r0,sr \n"
+	: "=&r"(sr) : : "memory", "r0" );
+	return sr;
+}
+
+static inline void unmask(unsigned sr)
+{
+	__asm__ __volatile__ ( "ldc %0,sr" : : "r"(sr) : "memory" );
+}
+
 /* gusa is a hack in the kernel which lets you create a sequence of instructions
  * which will be restarted if the process is preempted in the middle of the
  * sequence. It will do for implementing atomics on non-smp systems. ABI is:
@@ -25,11 +43,17 @@
 	"	mov.l " new ", @" mem "\n" \
 	"1:	mov r1, r15\n"
 
-#define CPU_HAS_LLSC 0x0040
-
 int __sh_cas(volatile int *p, int t, int s)
 {
-	if (__hwcap & CPU_HAS_LLSC) return __sh_cas_llsc(p, t, s);
+	if (__sh_atomic_model == SH_A_LLSC) return __sh_cas_llsc(p, t, s);
+
+	if (__sh_atomic_model == SH_A_IMASK) {
+		unsigned sr = mask();
+		int old = *p;
+		if (old==t) *p = s;
+		unmask(sr);
+		return old;
+	}
 
 	int old;
 	__asm__ __volatile__(
@@ -43,7 +67,15 @@ int __sh_cas(volatile int *p, int t, int s)
 
 int __sh_swap(volatile int *x, int v)
 {
-	if (__hwcap & CPU_HAS_LLSC) return __sh_swap_llsc(x, v);
+	if (__sh_atomic_model == SH_A_LLSC) return __sh_swap_llsc(x, v);
+
+	if (__sh_atomic_model == SH_A_IMASK) {
+		unsigned sr = mask();
+		int old = *x;
+		*x = v;
+		unmask(sr);
+		return old;
+	}
 
 	int old;
 	__asm__ __volatile__(
@@ -55,7 +87,15 @@ int __sh_swap(volatile int *x, int v)
 
 int __sh_fetch_add(volatile int *x, int v)
 {
-	if (__hwcap & CPU_HAS_LLSC) return __sh_fetch_add_llsc(x, v);
+	if (__sh_atomic_model == SH_A_LLSC) return __sh_fetch_add_llsc(x, v);
+
+	if (__sh_atomic_model == SH_A_IMASK) {
+		unsigned sr = mask();
+		int old = *x;
+		*x = old + v;
+		unmask(sr);
+		return old;
+	}
 
 	int old, dummy;
 	__asm__ __volatile__(
@@ -69,7 +109,7 @@ int __sh_fetch_add(volatile int *x, int v)
 
 void __sh_store(volatile int *p, int x)
 {
-	if (__hwcap & CPU_HAS_LLSC) return __sh_store_llsc(p, x);
+	if (__sh_atomic_model == SH_A_LLSC) return __sh_store_llsc(p, x);
 	__asm__ __volatile__(
 		"	mov.l %1, @%0\n"
 		: : "r"(p), "r"(x) : "memory");
@@ -77,7 +117,15 @@ void __sh_store(volatile int *p, int x)
 
 void __sh_and(volatile int *x, int v)
 {
-	if (__hwcap & CPU_HAS_LLSC) return __sh_and_llsc(x, v);
+	if (__sh_atomic_model == SH_A_LLSC) return __sh_and_llsc(x, v);
+
+	if (__sh_atomic_model == SH_A_IMASK) {
+		unsigned sr = mask();
+		int old = *x;
+		*x = old & v;
+		unmask(sr);
+		return;
+	}
 
 	int dummy;
 	__asm__ __volatile__(
@@ -89,7 +137,15 @@ void __sh_and(volatile int *x, int v)
 
 void __sh_or(volatile int *x, int v)
 {
-	if (__hwcap & CPU_HAS_LLSC) return __sh_or_llsc(x, v);
+	if (__sh_atomic_model == SH_A_LLSC) return __sh_or_llsc(x, v);
+
+	if (__sh_atomic_model == SH_A_IMASK) {
+		unsigned sr = mask();
+		int old = *x;
+		*x = old | v;
+		unmask(sr);
+		return;
+	}
 
 	int dummy;
 	__asm__ __volatile__(
diff --git a/arch/sh/src/sh_atomic.h b/arch/sh/src/sh_atomic.h
new file mode 100644
index 0000000..054c2a3
--- /dev/null
+++ b/arch/sh/src/sh_atomic.h
@@ -0,0 +1,15 @@
+#ifndef _SH_ATOMIC_H
+#define _SH_ATOMIC_H
+
+#define SH_A_GUSA 0
+#define SH_A_LLSC 1
+#define SH_A_CAS 2
+#if !defined(__SH3__) && !defined(__SH4__)
+#define SH_A_IMASK 3
+#else
+#define SH_A_IMASK -1LL /* unmatchable by unsigned int */
+#endif
+
+extern __attribute__((__visibility__("hidden"))) unsigned __sh_atomic_model;
+
+#endif
diff --git a/arch/sh/syscall_arch.h b/arch/sh/syscall_arch.h
index 7ee21a5..f63675a 100644
--- a/arch/sh/syscall_arch.h
+++ b/arch/sh/syscall_arch.h
@@ -8,7 +8,7 @@
  */
 #define __asm_syscall(trapno, ...) do {   \
 	__asm__ __volatile__ (                \
-		"trapa #" #trapno "\n"            \
+		"trapa #31\n"            \
 		"or r0, r0\n"                     \
 		"or r0, r0\n"                     \
 		"or r0, r0\n"                     \
diff --git a/include/ctype.h b/include/ctype.h
index cd2e016..7936536 100644
--- a/include/ctype.h
+++ b/include/ctype.h
@@ -64,6 +64,7 @@ int   isascii(int);
 int   toascii(int);
 #define _tolower(a) ((a)|0x20)
 #define _toupper(a) ((a)&0x5f)
+#define isascii(a) (0 ? isascii(a) : (unsigned)(a) < 128)
 
 #endif
 
diff --git a/include/stdlib.h b/include/stdlib.h
index 97ce5a7..d2c911f 100644
--- a/include/stdlib.h
+++ b/include/stdlib.h
@@ -76,7 +76,8 @@ size_t wcstombs (char *__restrict, const wchar_t *__restrict, size_t);
 #define EXIT_FAILURE 1
 #define EXIT_SUCCESS 0
 
-#define MB_CUR_MAX ((size_t)+4)
+size_t __ctype_get_mb_cur_max(void);
+#define MB_CUR_MAX (__ctype_get_mb_cur_max())
 
 #define RAND_MAX (0x7fffffff)
 
diff --git a/src/ctype/__ctype_get_mb_cur_max.c b/src/ctype/__ctype_get_mb_cur_max.c
index d235f4d..8e946fc 100644
--- a/src/ctype/__ctype_get_mb_cur_max.c
+++ b/src/ctype/__ctype_get_mb_cur_max.c
@@ -1,6 +1,7 @@
-#include <stddef.h>
+#include <stdlib.h>
+#include "locale_impl.h"
 
 size_t __ctype_get_mb_cur_max()
 {
-	return 4;
+	return MB_CUR_MAX;
 }
diff --git a/src/ctype/isascii.c b/src/ctype/isascii.c
index 3af0a10..54ad3bf 100644
--- a/src/ctype/isascii.c
+++ b/src/ctype/isascii.c
@@ -1,4 +1,5 @@
 #include <ctype.h>
+#undef isascii
 
 int isascii(int c)
 {
diff --git a/src/env/__stack_chk_fail.c b/src/env/__stack_chk_fail.c
index 47784c6..be0c184 100644
--- a/src/env/__stack_chk_fail.c
+++ b/src/env/__stack_chk_fail.c
@@ -25,4 +25,8 @@ void __stack_chk_fail_local(void)
 	a_crash();
 }
 
+#else
+
+weak_alias(__stack_chk_fail, __stack_chk_fail_local);
+
 #endif
diff --git a/src/internal/libc.h b/src/internal/libc.h
index 6810cd8..98c7535 100644
--- a/src/internal/libc.h
+++ b/src/internal/libc.h
@@ -17,8 +17,6 @@ struct __libc {
 	int secure;
 	volatile int threads_minus_1;
 	size_t *auxv;
-	FILE *ofl_head;
-	volatile int ofl_lock[2];
 	size_t tls_size;
 	size_t page_size;
 	struct __locale_struct global_locale;
diff --git a/src/internal/locale_impl.h b/src/internal/locale_impl.h
index 9b8385e..f5e4d9b 100644
--- a/src/internal/locale_impl.h
+++ b/src/internal/locale_impl.h
@@ -1,3 +1,6 @@
+#ifndef _LOCALE_IMPL_H
+#define _LOCALE_IMPL_H
+
 #include <locale.h>
 #include <stdlib.h>
 #include "libc.h"
@@ -12,6 +15,10 @@ struct __locale_map {
 	const struct __locale_map *next;
 };
 
+extern const struct __locale_map __c_dot_utf8;
+extern const struct __locale_struct __c_locale;
+extern const struct __locale_struct __c_dot_utf8_locale;
+
 const struct __locale_map *__get_locale(int, const char *);
 const char *__mo_lookup(const void *, size_t, const char *);
 const char *__lctrans(const char *, const struct __locale_map *);
@@ -20,9 +27,14 @@ const char *__lctrans_cur(const char *);
 #define LCTRANS(msg, lc, loc) __lctrans(msg, (loc)->cat[(lc)])
 #define LCTRANS_CUR(msg) __lctrans_cur(msg)
 
+#define C_LOCALE ((locale_t)&__c_locale)
+#define UTF8_LOCALE ((locale_t)&__c_dot_utf8_locale)
+
 #define CURRENT_LOCALE (__pthread_self()->locale)
 
 #define CURRENT_UTF8 (!!__pthread_self()->locale->cat[LC_CTYPE])
 
 #undef MB_CUR_MAX
 #define MB_CUR_MAX (CURRENT_UTF8 ? 4 : 1)
+
+#endif
diff --git a/src/internal/pthread_impl.h b/src/internal/pthread_impl.h
index e29f9c8..3890bb5 100644
--- a/src/internal/pthread_impl.h
+++ b/src/internal/pthread_impl.h
@@ -94,6 +94,10 @@ struct __timer {
 #define CANARY canary
 #endif
 
+#ifndef DTP_OFFSET
+#define DTP_OFFSET 0
+#endif
+
 #define SIGTIMER 32
 #define SIGCANCEL 33
 #define SIGSYNCCALL 34
diff --git a/src/internal/sh/syscall.s b/src/internal/sh/syscall.s
index d00712a..331918a 100644
--- a/src/internal/sh/syscall.s
+++ b/src/internal/sh/syscall.s
@@ -13,7 +13,7 @@ __syscall:
 	mov.l @r15, r7
 	mov.l @(4,r15), r0
 	mov.l @(8,r15), r1
-	trapa #22
+	trapa #31
 	or r0, r0
 	or r0, r0
 	or r0, r0
diff --git a/src/internal/stdio_impl.h b/src/internal/stdio_impl.h
index e1325fe..0dd7fb5 100644
--- a/src/internal/stdio_impl.h
+++ b/src/internal/stdio_impl.h
@@ -47,6 +47,7 @@ struct _IO_FILE {
 	unsigned char *shend;
 	off_t shlim, shcnt;
 	FILE *prev_locked, *next_locked;
+	struct __locale_struct *locale;
 };
 
 size_t __stdio_read(FILE *, unsigned char *, size_t);
@@ -75,8 +76,9 @@ int __putc_unlocked(int, FILE *);
 FILE *__fdopen(int, const char *);
 int __fmodeflags(const char *);
 
-#define OFLLOCK() LOCK(libc.ofl_lock)
-#define OFLUNLOCK() UNLOCK(libc.ofl_lock)
+FILE *__ofl_add(FILE *f);
+FILE **__ofl_lock(void);
+void __ofl_unlock(void);
 
 #define feof(f) ((f)->flags & F_EOF)
 #define ferror(f) ((f)->flags & F_ERR)
diff --git a/src/ldso/dynlink.c b/src/ldso/dynlink.c
index 42b056d..d2a7249 100644
--- a/src/ldso/dynlink.c
+++ b/src/ldso/dynlink.c
@@ -337,7 +337,7 @@ static void do_relocs(struct dso *dso, size_t *rel, size_t rel_size, size_t stri
 			*reloc_addr = def.dso->tls_id;
 			break;
 		case REL_DTPOFF:
-			*reloc_addr = tls_val + addend;
+			*reloc_addr = tls_val + addend - DTP_OFFSET;
 			break;
 #ifdef TLS_ABOVE_TP
 		case REL_TPOFF:
@@ -423,6 +423,28 @@ static void reclaim_gaps(struct dso *dso)
 	}
 }
 
+static void *mmap_fixed(void *p, size_t n, int prot, int flags, int fd, off_t off)
+{
+	char *q = mmap(p, n, prot, flags, fd, off);
+	if (q != MAP_FAILED || errno != EINVAL) return q;
+	/* Fallbacks for MAP_FIXED failure on NOMMU kernels. */
+	if (flags & MAP_ANONYMOUS) {
+		memset(p, 0, n);
+		return p;
+	}
+	ssize_t r;
+	if (lseek(fd, off, SEEK_SET) < 0) return MAP_FAILED;
+	for (q=p; n; q+=r, off+=r, n-=r) {
+		r = read(fd, q, n);
+		if (r < 0 && errno != EINTR) return MAP_FAILED;
+		if (!r) {
+			memset(q, 0, n);
+			break;
+		}
+	}
+	return p;
+}
+
 static void *map_library(int fd, struct dso *dso)
 {
 	Ehdr buf[(896+sizeof(Ehdr))/sizeof(Ehdr)];
@@ -524,19 +546,20 @@ static void *map_library(int fd, struct dso *dso)
 		prot = (((ph->p_flags&PF_R) ? PROT_READ : 0) |
 			((ph->p_flags&PF_W) ? PROT_WRITE: 0) |
 			((ph->p_flags&PF_X) ? PROT_EXEC : 0));
-		if (mmap(base+this_min, this_max-this_min, prot, MAP_PRIVATE|MAP_FIXED, fd, off_start) == MAP_FAILED)
+		if (mmap_fixed(base+this_min, this_max-this_min, prot, MAP_PRIVATE|MAP_FIXED, fd, off_start) == MAP_FAILED)
 			goto error;
 		if (ph->p_memsz > ph->p_filesz) {
 			size_t brk = (size_t)base+ph->p_vaddr+ph->p_filesz;
 			size_t pgbrk = brk+PAGE_SIZE-1 & -PAGE_SIZE;
 			memset((void *)brk, 0, pgbrk-brk & PAGE_SIZE-1);
-			if (pgbrk-(size_t)base < this_max && mmap((void *)pgbrk, (size_t)base+this_max-pgbrk, prot, MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) == MAP_FAILED)
+			if (pgbrk-(size_t)base < this_max && mmap_fixed((void *)pgbrk, (size_t)base+this_max-pgbrk, prot, MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) == MAP_FAILED)
 				goto error;
 		}
 	}
 	for (i=0; ((size_t *)(base+dyn))[i]; i+=2)
 		if (((size_t *)(base+dyn))[i]==DT_TEXTREL) {
-			if (mprotect(map, map_len, PROT_READ|PROT_WRITE|PROT_EXEC) < 0)
+			if (mprotect(map, map_len, PROT_READ|PROT_WRITE|PROT_EXEC)
+			    && errno != ENOSYS)
 				goto error;
 			break;
 		}
@@ -927,7 +950,8 @@ static void reloc_all(struct dso *p)
 		do_relocs(p, (void *)(p->base+dyn[DT_RELA]), dyn[DT_RELASZ], 3);
 
 		if (head != &ldso && p->relro_start != p->relro_end &&
-		    mprotect(p->base+p->relro_start, p->relro_end-p->relro_start, PROT_READ) < 0) {
+		    mprotect(p->base+p->relro_start, p->relro_end-p->relro_start, PROT_READ)
+		    && errno != ENOSYS) {
 			error("Error relocating %s: RELRO protection failed: %m",
 				p->name);
 			if (runtime) longjmp(*rtld_fail, 1);
@@ -1078,7 +1102,7 @@ void *__tls_get_new(size_t *v)
 	__block_all_sigs(&set);
 	if (v[0]<=(size_t)self->dtv[0]) {
 		__restore_sigs(&set);
-		return (char *)self->dtv[v[0]]+v[1];
+		return (char *)self->dtv[v[0]]+v[1]+DTP_OFFSET;
 	}
 
 	/* This is safe without any locks held because, if the caller
@@ -1111,7 +1135,7 @@ void *__tls_get_new(size_t *v)
 		if (p->tls_id == v[0]) break;
 	}
 	__restore_sigs(&set);
-	return mem + v[1];
+	return mem + v[1] + DTP_OFFSET;
 }
 
 static void update_tls_size()
@@ -1192,6 +1216,17 @@ _Noreturn void __dls3(size_t *sp)
 	char **argv_orig = argv;
 	char **envp = argv+argc+1;
 
+	/* Find aux vector just past environ[] and use it to initialize
+	 * global data that may be needed before we can make syscalls. */
+	__environ = envp;
+	for (i=argc+1; argv[i]; i++);
+	libc.auxv = auxv = (void *)(argv+i+1);
+	decode_vec(auxv, aux, AUX_CNT);
+	__hwcap = aux[AT_HWCAP];
+	libc.page_size = aux[AT_PAGESZ];
+	libc.secure = ((aux[0]&0x7800)!=0x7800 || aux[AT_UID]!=aux[AT_EUID]
+		|| aux[AT_GID]!=aux[AT_EGID] || aux[AT_SECURE]);
+
 	/* Setup early thread pointer in builtin_tls for ldso/libc itself to
 	 * use during dynamic linking. If possible it will also serve as the
 	 * thread pointer at runtime. */
@@ -1200,25 +1235,11 @@ _Noreturn void __dls3(size_t *sp)
 		a_crash();
 	}
 
-	/* Find aux vector just past environ[] */
-	for (i=argc+1; argv[i]; i++)
-		if (!memcmp(argv[i], "LD_LIBRARY_PATH=", 16))
-			env_path = argv[i]+16;
-		else if (!memcmp(argv[i], "LD_PRELOAD=", 11))
-			env_preload = argv[i]+11;
-	auxv = (void *)(argv+i+1);
-
-	decode_vec(auxv, aux, AUX_CNT);
-
 	/* Only trust user/env if kernel says we're not suid/sgid */
-	if ((aux[0]&0x7800)!=0x7800 || aux[AT_UID]!=aux[AT_EUID]
-	  || aux[AT_GID]!=aux[AT_EGID] || aux[AT_SECURE]) {
-		env_path = 0;
-		env_preload = 0;
-		libc.secure = 1;
+	if (!libc.secure) {
+		env_path = getenv("LD_LIBRARY_PATH");
+		env_preload = getenv("LD_PRELOAD");
 	}
-	libc.page_size = aux[AT_PAGESZ];
-	libc.auxv = auxv;
 
 	/* If the main program was already loaded by the kernel,
 	 * AT_PHDR will point to some location other than the dynamic
diff --git a/src/locale/c_locale.c b/src/locale/c_locale.c
new file mode 100644
index 0000000..77ccf58
--- /dev/null
+++ b/src/locale/c_locale.c
@@ -0,0 +1,15 @@
+#include "locale_impl.h"
+#include <stdint.h>
+
+static const uint32_t empty_mo[] = { 0x950412de, 0, -1, -1, -1 };
+
+const struct __locale_map __c_dot_utf8 = {
+	.map = empty_mo,
+	.map_size = sizeof empty_mo,
+	.name = "C.UTF-8"
+};
+
+const struct __locale_struct __c_locale = { 0 };
+const struct __locale_struct __c_dot_utf8_locale = {
+	.cat[LC_CTYPE] = &__c_dot_utf8
+};
diff --git a/src/locale/iconv.c b/src/locale/iconv.c
index e6121ae..1eeea94 100644
--- a/src/locale/iconv.c
+++ b/src/locale/iconv.c
@@ -5,6 +5,7 @@
 #include <stdlib.h>
 #include <limits.h>
 #include <stdint.h>
+#include "locale_impl.h"
 
 #define UTF_32BE    0300
 #define UTF_16LE    0301
@@ -165,9 +166,12 @@ size_t iconv(iconv_t cd0, char **restrict in, size_t *restrict inb, char **restr
 	int err;
 	unsigned char type = map[-1];
 	unsigned char totype = tomap[-1];
+	locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
 
 	if (!in || !*in || !*inb) return 0;
 
+	*ploc = UTF8_LOCALE;
+
 	for (; *inb; *in+=l, *inb-=l) {
 		c = *(unsigned char *)*in;
 		l = 1;
@@ -431,6 +435,7 @@ size_t iconv(iconv_t cd0, char **restrict in, size_t *restrict inb, char **restr
 			break;
 		}
 	}
+	*ploc = loc;
 	return x;
 ilseq:
 	err = EILSEQ;
@@ -445,5 +450,6 @@ starved:
 	x = -1;
 end:
 	errno = err;
+	*ploc = loc;
 	return x;
 }
diff --git a/src/locale/langinfo.c b/src/locale/langinfo.c
index a1ada24..776b447 100644
--- a/src/locale/langinfo.c
+++ b/src/locale/langinfo.c
@@ -33,7 +33,8 @@ char *__nl_langinfo_l(nl_item item, locale_t loc)
 	int idx = item & 65535;
 	const char *str;
 
-	if (item == CODESET) return "UTF-8";
+	if (item == CODESET)
+		return MB_CUR_MAX==1 ? "UTF-8-CODE-UNITS" : "UTF-8";
 	
 	switch (cat) {
 	case LC_NUMERIC:
diff --git a/src/locale/locale_map.c b/src/locale/locale_map.c
index 4346bb0..c3e5917 100644
--- a/src/locale/locale_map.c
+++ b/src/locale/locale_map.c
@@ -24,14 +24,6 @@ static const char envvars[][12] = {
 	"LC_MESSAGES",
 };
 
-static const uint32_t empty_mo[] = { 0x950412de, 0, -1, -1, -1 };
-
-const struct __locale_map __c_dot_utf8 = {
-	.map = empty_mo,
-	.map_size = sizeof empty_mo,
-	.name = "C.UTF-8"
-};
-
 const struct __locale_map *__get_locale(int cat, const char *val)
 {
 	static int lock[2];
@@ -107,8 +99,8 @@ const struct __locale_map *__get_locale(int cat, const char *val)
 	 * sake of being able to do message translations at the
 	 * application level. */
 	if (!new && (new = malloc(sizeof *new))) {
-		new->map = empty_mo;
-		new->map_size = sizeof empty_mo;
+		new->map = __c_dot_utf8.map;
+		new->map_size = __c_dot_utf8.map_size;
 		memcpy(new->name, val, n);
 		new->name[n] = 0;
 		new->next = loc_head;
diff --git a/src/locale/newlocale.c b/src/locale/newlocale.c
index 89d36b1..f50bbe9 100644
--- a/src/locale/newlocale.c
+++ b/src/locale/newlocale.c
@@ -3,16 +3,9 @@
 #include "locale_impl.h"
 #include "libc.h"
 
-extern const struct __locale_map __c_dot_utf8;
-
-static const struct __locale_struct c_locale = { 0 };
-static const struct __locale_struct c_dot_utf8_locale = {
-	.cat[LC_CTYPE] = &__c_dot_utf8
-};
-
 int __loc_is_allocated(locale_t loc)
 {
-	return loc && loc != &c_locale && loc != &c_dot_utf8_locale;
+	return loc && loc != C_LOCALE && loc != UTF8_LOCALE;
 }
 
 locale_t __newlocale(int mask, const char *name, locale_t loc)
@@ -44,9 +37,9 @@ locale_t __newlocale(int mask, const char *name, locale_t loc)
 	}
 
 	if (!j)
-		return (locale_t)&c_locale;
-	if (j==1 && tmp.cat[LC_CTYPE]==c_dot_utf8_locale.cat[LC_CTYPE])
-		return (locale_t)&c_dot_utf8_locale;
+		return C_LOCALE;
+	if (j==1 && tmp.cat[LC_CTYPE]==&__c_dot_utf8)
+		return UTF8_LOCALE;
 
 	if ((loc = malloc(sizeof *loc))) *loc = tmp;
 
diff --git a/src/locale/uselocale.c b/src/locale/uselocale.c
index b70a0c1..0fc5ecb 100644
--- a/src/locale/uselocale.c
+++ b/src/locale/uselocale.c
@@ -8,9 +8,7 @@ locale_t __uselocale(locale_t new)
 	locale_t old = self->locale;
 	locale_t global = &libc.global_locale;
 
-	if (new == LC_GLOBAL_LOCALE) new = global;
-
-	self->locale = new;
+	if (new) self->locale = new == LC_GLOBAL_LOCALE ? global : new;
 
 	return old == global ? LC_GLOBAL_LOCALE : old;
 }
diff --git a/src/malloc/calloc.c b/src/malloc/calloc.c
index c3dfb47..436c0b0 100644
--- a/src/malloc/calloc.c
+++ b/src/malloc/calloc.c
@@ -1,22 +1,13 @@
 #include <stdlib.h>
 #include <errno.h>
 
+void *__malloc0(size_t);
+
 void *calloc(size_t m, size_t n)
 {
-	void *p;
-	size_t *z;
 	if (n && m > (size_t)-1/n) {
 		errno = ENOMEM;
 		return 0;
 	}
-	n *= m;
-	p = malloc(n);
-	if (!p) return 0;
-	/* Only do this for non-mmapped chunks */
-	if (((size_t *)p)[-1] & 7) {
-		/* Only write words that are not already zero */
-		m = (n + sizeof *z - 1)/sizeof *z;
-		for (z=p; m; m--, z++) if (*z) *z=0;
-	}
-	return p;
+	return __malloc0(n * m);
 }
diff --git a/src/malloc/expand_heap.c b/src/malloc/expand_heap.c
new file mode 100644
index 0000000..d8c0be7
--- /dev/null
+++ b/src/malloc/expand_heap.c
@@ -0,0 +1,72 @@
+#include <limits.h>
+#include <stdint.h>
+#include <errno.h>
+#include <sys/mman.h>
+#include "libc.h"
+#include "syscall.h"
+
+/* This function returns true if the interval [old,new]
+ * intersects the 'len'-sized interval below &libc.auxv
+ * (interpreted as the main-thread stack) or below &b
+ * (the current stack). It is used to defend against
+ * buggy brk implementations that can cross the stack. */
+
+static int traverses_stack_p(uintptr_t old, uintptr_t new)
+{
+	const uintptr_t len = 8<<20;
+	uintptr_t a, b;
+
+	b = (uintptr_t)libc.auxv;
+	a = b > len ? b-len : 0;
+	if (new>a && old<b) return 1;
+
+	b = (uintptr_t)&b;
+	a = b > len ? b-len : 0;
+	if (new>a && old<b) return 1;
+
+	return 0;
+}
+
+void *__mmap(void *, size_t, int, int, int, off_t);
+
+/* Expand the heap in-place if brk can be used, or otherwise via mmap,
+ * using an exponential lower bound on growth by mmap to make
+ * fragmentation asymptotically irrelevant. The size argument is both
+ * an input and an output, since the caller needs to know the size
+ * allocated, which will be larger than requested due to page alignment
+ * and mmap minimum size rules. The caller is responsible for locking
+ * to prevent concurrent calls. */
+
+void *__expand_heap(size_t *pn)
+{
+	static uintptr_t brk;
+	static unsigned mmap_step;
+	size_t n = *pn;
+
+	if (n > SIZE_MAX/2 - PAGE_SIZE) {
+		errno = ENOMEM;
+		return 0;
+	}
+	n += -n & PAGE_SIZE-1;
+
+	if (!brk) {
+		brk = __syscall(SYS_brk, 0);
+		brk += -brk & PAGE_SIZE-1;
+	}
+
+	if (n < SIZE_MAX-brk && !traverses_stack_p(brk, brk+n)
+	    && __syscall(SYS_brk, brk+n)==brk+n) {
+		*pn = n;
+		brk += n;
+		return (void *)(brk-n);
+	}
+
+	size_t min = (size_t)PAGE_SIZE << mmap_step/2;
+	if (n < min) n = min;
+	void *area = __mmap(0, n, PROT_READ|PROT_WRITE,
+		MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
+	if (area == MAP_FAILED) return 0;
+	*pn = n;
+	mmap_step++;
+	return area;
+}
diff --git a/src/malloc/lite_malloc.c b/src/malloc/lite_malloc.c
index 7643fc2..09ac575 100644
--- a/src/malloc/lite_malloc.c
+++ b/src/malloc/lite_malloc.c
@@ -4,43 +4,47 @@
 #include <errno.h>
 #include "libc.h"
 
-uintptr_t __brk(uintptr_t);
-
 #define ALIGN 16
 
+void *__expand_heap(size_t *);
+
 void *__simple_malloc(size_t n)
 {
-	static uintptr_t cur, brk;
-	uintptr_t base, new;
+	static char *cur, *end;
 	static volatile int lock[2];
-	size_t align=1;
+	size_t align=1, pad;
+	void *p;
 
 	if (!n) n++;
-	if (n > SIZE_MAX/2) goto toobig;
-
 	while (align<n && align<ALIGN)
 		align += align;
-	n = n + align - 1 & -align;
 
 	LOCK(lock);
-	if (!cur) cur = brk = __brk(0)+16;
-	base = cur + align-1 & -align;
-	if (n > SIZE_MAX - PAGE_SIZE - base) goto fail;
-	if (base+n > brk) {
-		new = base+n + PAGE_SIZE-1 & -PAGE_SIZE;
-		if (__brk(new) != new) goto fail;
-		brk = new;
-	}
-	cur = base+n;
-	UNLOCK(lock);
 
-	return (void *)base;
+	pad = -(uintptr_t)cur & align-1;
+
+	if (n <= SIZE_MAX/2 + ALIGN) n += pad;
+
+	if (n > end-cur) {
+		size_t m = n;
+		char *new = __expand_heap(&m);
+		if (!new) {
+			UNLOCK(lock);
+			return 0;
+		}
+		if (new != end) {
+			cur = new;
+			n -= pad;
+			pad = 0;
+		}
+		end = new + m;
+	}
 
-fail:
+	p = cur + pad;
+	cur += n;
 	UNLOCK(lock);
-toobig:
-	errno = ENOMEM;
-	return 0;
+	return p;
 }
 
 weak_alias(__simple_malloc, malloc);
+weak_alias(__simple_malloc, __malloc0);
diff --git a/src/malloc/malloc.c b/src/malloc/malloc.c
index d4de2dc..eb68d55 100644
--- a/src/malloc/malloc.c
+++ b/src/malloc/malloc.c
@@ -13,7 +13,6 @@
 #define inline inline __attribute__((always_inline))
 #endif
 
-uintptr_t __brk(uintptr_t);
 void *__mmap(void *, size_t, int, int, int, off_t);
 int __munmap(void *, size_t);
 void *__mremap(void *, size_t, size_t, int, ...);
@@ -31,13 +30,9 @@ struct bin {
 };
 
 static struct {
-	uintptr_t brk;
-	size_t *heap;
 	volatile uint64_t binmap;
 	struct bin bins[64];
-	volatile int brk_lock[2];
 	volatile int free_lock[2];
-	unsigned mmap_step;
 } mal;
 
 
@@ -152,69 +147,52 @@ void __dump_heap(int x)
 }
 #endif
 
+void *__expand_heap(size_t *);
+
 static struct chunk *expand_heap(size_t n)
 {
-	static int init;
+	static int heap_lock[2];
+	static void *end;
+	void *p;
 	struct chunk *w;
-	uintptr_t new;
-
-	lock(mal.brk_lock);
 
-	if (!init) {
-		mal.brk = __brk(0);
-#ifdef SHARED
-		mal.brk = mal.brk + PAGE_SIZE-1 & -PAGE_SIZE;
-#endif
-		mal.brk = mal.brk + 2*SIZE_ALIGN-1 & -SIZE_ALIGN;
-		mal.heap = (void *)mal.brk;
-		init = 1;
-	}
+	/* The argument n already accounts for the caller's chunk
+	 * overhead needs, but if the heap can't be extended in-place,
+	 * we need room for an extra zero-sized sentinel chunk. */
+	n += SIZE_ALIGN;
 
-	if (n > SIZE_MAX - mal.brk - 2*PAGE_SIZE) goto fail;
-	new = mal.brk + n + SIZE_ALIGN + PAGE_SIZE - 1 & -PAGE_SIZE;
-	n = new - mal.brk;
+	lock(heap_lock);
 
-	if (__brk(new) != new) {
-		size_t min = (size_t)PAGE_SIZE << mal.mmap_step/2;
-		n += -n & PAGE_SIZE-1;
-		if (n < min) n = min;
-		void *area = __mmap(0, n, PROT_READ|PROT_WRITE,
-			MAP_PRIVATE|MAP_ANONYMOUS, -1, 0);
-		if (area == MAP_FAILED) goto fail;
+	p = __expand_heap(&n);
+	if (!p) {
+		unlock(heap_lock);
+		return 0;
+	}
 
-		mal.mmap_step++;
-		area = (char *)area + SIZE_ALIGN - OVERHEAD;
-		w = area;
+	/* If not just expanding existing space, we need to make a
+	 * new sentinel chunk below the allocated space. */
+	if (p != end) {
+		/* Valid/safe because of the prologue increment. */
 		n -= SIZE_ALIGN;
+		p = (char *)p + SIZE_ALIGN;
+		w = MEM_TO_CHUNK(p);
 		w->psize = 0 | C_INUSE;
-		w->csize = n | C_INUSE;
-		w = NEXT_CHUNK(w);
-		w->psize = n | C_INUSE;
-		w->csize = 0 | C_INUSE;
-
-		unlock(mal.brk_lock);
-
-		return area;
 	}
 
-	w = MEM_TO_CHUNK(mal.heap);
-	w->psize = 0 | C_INUSE;
-
-	w = MEM_TO_CHUNK(new);
+	/* Record new heap end and fill in footer. */
+	end = (char *)p + n;
+	w = MEM_TO_CHUNK(end);
 	w->psize = n | C_INUSE;
 	w->csize = 0 | C_INUSE;
 
-	w = MEM_TO_CHUNK(mal.brk);
+	/* Fill in header, which may be new or may be replacing a
+	 * zero-size sentinel header at the old end-of-heap. */
+	w = MEM_TO_CHUNK(p);
 	w->csize = n | C_INUSE;
-	mal.brk = new;
-	
-	unlock(mal.brk_lock);
+
+	unlock(heap_lock);
 
 	return w;
-fail:
-	unlock(mal.brk_lock);
-	errno = ENOMEM;
-	return 0;
 }
 
 static int adjust_size(size_t *n)
@@ -378,6 +356,17 @@ void *malloc(size_t n)
 	return CHUNK_TO_MEM(c);
 }
 
+void *__malloc0(size_t n)
+{
+	void *p = malloc(n);
+	if (p && !IS_MMAPPED(MEM_TO_CHUNK(p))) {
+		size_t *z;
+		n = (n + sizeof *z - 1)/sizeof *z;
+		for (z=p; n; n--, z++) if (*z) *z=0;
+	}
+	return p;
+}
+
 void *realloc(void *p, size_t n)
 {
 	struct chunk *self, *next;
diff --git a/src/multibyte/btowc.c b/src/multibyte/btowc.c
index 9d2c3b1..8acd0a2 100644
--- a/src/multibyte/btowc.c
+++ b/src/multibyte/btowc.c
@@ -1,7 +1,10 @@
 #include <stdio.h>
 #include <wchar.h>
+#include <stdlib.h>
+#include "internal.h"
 
 wint_t btowc(int c)
 {
-	return c<128U ? c : EOF;
+	int b = (unsigned char)c;
+	return b<128U ? b : (MB_CUR_MAX==1 && c!=EOF) ? CODEUNIT(c) : WEOF;
 }
diff --git a/src/multibyte/internal.h b/src/multibyte/internal.h
index cc017fa..53d62ed 100644
--- a/src/multibyte/internal.h
+++ b/src/multibyte/internal.h
@@ -23,3 +23,10 @@ extern const uint32_t bittab[];
 
 #define SA 0xc2u
 #define SB 0xf4u
+
+/* Arbitrary encoding for representing code units instead of characters. */
+#define CODEUNIT(c) (0xdfff & (signed char)(c))
+#define IS_CODEUNIT(c) ((unsigned)(c)-0xdf80 < 0x80)
+
+/* Get inline definition of MB_CUR_MAX. */
+#include "locale_impl.h"
diff --git a/src/multibyte/mbrtowc.c b/src/multibyte/mbrtowc.c
index e7b3654..ca7da70 100644
--- a/src/multibyte/mbrtowc.c
+++ b/src/multibyte/mbrtowc.c
@@ -4,6 +4,7 @@
  * unnecessary.
  */
 
+#include <stdlib.h>
 #include <wchar.h>
 #include <errno.h>
 #include "internal.h"
@@ -27,6 +28,7 @@ size_t mbrtowc(wchar_t *restrict wc, const char *restrict src, size_t n, mbstate
 	if (!n) return -2;
 	if (!c) {
 		if (*s < 0x80) return !!(*wc = *s);
+		if (MB_CUR_MAX==1) return (*wc = CODEUNIT(*s)), 1;
 		if (*s-SA > SB-SA) goto ilseq;
 		c = bittab[*s++-SA]; n--;
 	}
diff --git a/src/multibyte/mbsrtowcs.c b/src/multibyte/mbsrtowcs.c
index 3c1343a..e23083d 100644
--- a/src/multibyte/mbsrtowcs.c
+++ b/src/multibyte/mbsrtowcs.c
@@ -7,6 +7,8 @@
 #include <stdint.h>
 #include <wchar.h>
 #include <errno.h>
+#include <string.h>
+#include <stdlib.h>
 #include "internal.h"
 
 size_t mbsrtowcs(wchar_t *restrict ws, const char **restrict src, size_t wn, mbstate_t *restrict st)
@@ -24,6 +26,23 @@ size_t mbsrtowcs(wchar_t *restrict ws, const char **restrict src, size_t wn, mbs
 		}
 	}
 
+	if (MB_CUR_MAX==1) {
+		if (!ws) return strlen((const char *)s);
+		for (;;) {
+			if (!wn) {
+				*src = (const void *)s;
+				return wn0;
+			}
+			if (!*s) break;
+			c = *s++;
+			*ws++ = CODEUNIT(c);
+			wn--;
+		}
+		*ws = 0;
+		*src = 0;
+		return wn0-wn;
+	}
+
 	if (!ws) for (;;) {
 		if (*s-1u < 0x7f && (uintptr_t)s%4 == 0) {
 			while (!(( *(uint32_t*)s | *(uint32_t*)s-0x01010101) & 0x80808080)) {
diff --git a/src/multibyte/mbtowc.c b/src/multibyte/mbtowc.c
index 803d221..71a9506 100644
--- a/src/multibyte/mbtowc.c
+++ b/src/multibyte/mbtowc.c
@@ -4,6 +4,7 @@
  * unnecessary.
  */
 
+#include <stdlib.h>
 #include <wchar.h>
 #include <errno.h>
 #include "internal.h"
@@ -19,6 +20,7 @@ int mbtowc(wchar_t *restrict wc, const char *restrict src, size_t n)
 	if (!wc) wc = &dummy;
 
 	if (*s < 0x80) return !!(*wc = *s);
+	if (MB_CUR_MAX==1) return (*wc = CODEUNIT(*s)), 1;
 	if (*s-SA > SB-SA) goto ilseq;
 	c = bittab[*s++-SA];
 
diff --git a/src/multibyte/wcrtomb.c b/src/multibyte/wcrtomb.c
index 59f733d..ddc37a5 100644
--- a/src/multibyte/wcrtomb.c
+++ b/src/multibyte/wcrtomb.c
@@ -4,8 +4,10 @@
  * unnecessary.
  */
 
+#include <stdlib.h>
 #include <wchar.h>
 #include <errno.h>
+#include "internal.h"
 
 size_t wcrtomb(char *restrict s, wchar_t wc, mbstate_t *restrict st)
 {
@@ -13,6 +15,13 @@ size_t wcrtomb(char *restrict s, wchar_t wc, mbstate_t *restrict st)
 	if ((unsigned)wc < 0x80) {
 		*s = wc;
 		return 1;
+	} else if (MB_CUR_MAX == 1) {
+		if (!IS_CODEUNIT(wc)) {
+			errno = EILSEQ;
+			return -1;
+		}
+		*s = wc;
+		return 1;
 	} else if ((unsigned)wc < 0x800) {
 		*s++ = 0xc0 | (wc>>6);
 		*s = 0x80 | (wc&0x3f);
diff --git a/src/multibyte/wctob.c b/src/multibyte/wctob.c
index d6353ee..4aeda6a 100644
--- a/src/multibyte/wctob.c
+++ b/src/multibyte/wctob.c
@@ -1,8 +1,10 @@
-#include <stdio.h>
 #include <wchar.h>
+#include <stdlib.h>
+#include "internal.h"
 
 int wctob(wint_t c)
 {
 	if (c < 128U) return c;
+	if (MB_CUR_MAX==1 && IS_CODEUNIT(c)) return (unsigned char)c;
 	return EOF;
 }
diff --git a/src/passwd/nscd_query.c b/src/passwd/nscd_query.c
index 69a7815..d38e371 100644
--- a/src/passwd/nscd_query.c
+++ b/src/passwd/nscd_query.c
@@ -32,6 +32,7 @@ FILE *__nscd_query(int32_t req, const char *key, int32_t *buf, size_t len, int *
 		},
 		.msg_iovlen = 2
 	};
+	int errno_save = errno;
 
 	*swap = 0;
 retry:
@@ -50,11 +51,14 @@ retry:
 		return f;
 
 	if (connect(fd, (struct sockaddr*)&addr, sizeof(addr)) < 0) {
-		/* If there isn't a running nscd we return -1 to indicate that
-		 * that is precisely what happened
-		 */
-		if (errno == EACCES || errno == ECONNREFUSED || errno == ENOENT)
+		/* If there isn't a running nscd we simulate a "not found"
+		 * result and the caller is responsible for calling
+		 * fclose on the (unconnected) socket. The value of
+		 * errno must be left unchanged in this case.  */
+		if (errno == EACCES || errno == ECONNREFUSED || errno == ENOENT) {
+			errno = errno_save;
 			return f;
+		}
 		goto error;
 	}
 
diff --git a/src/process/sh/vfork.s b/src/process/sh/vfork.s
new file mode 100644
index 0000000..48cc939
--- /dev/null
+++ b/src/process/sh/vfork.s
@@ -0,0 +1,23 @@
+.global __vfork
+.weak vfork
+.type __vfork,@function
+.type vfork,@function
+__vfork:
+vfork:
+	mov #95, r3
+	add r3, r3
+
+	trapa #31
+	or    r0, r0
+	or    r0, r0
+	or    r0, r0
+	or    r0, r0
+	or    r0, r0
+
+	mov r0, r4
+	mov.l 1f, r0
+2:	braf r0
+	 nop
+	.align 2
+	.hidden __syscall_ret
+1:	.long __syscall_ret@PLT-(2b+4-.)
diff --git a/src/regex/fnmatch.c b/src/regex/fnmatch.c
index 7f6b65f..978fff8 100644
--- a/src/regex/fnmatch.c
+++ b/src/regex/fnmatch.c
@@ -18,6 +18,7 @@
 #include <stdlib.h>
 #include <wchar.h>
 #include <wctype.h>
+#include "locale_impl.h"
 
 #define END 0
 #define UNMATCHABLE -2
@@ -229,7 +230,7 @@ static int fnmatch_internal(const char *pat, size_t m, const char *str, size_t n
 	 * On illegal sequences we may get it wrong, but in that case
 	 * we necessarily have a matching failure anyway. */
 	for (s=endstr; s>str && tailcnt; tailcnt--) {
-		if (s[-1] < 128U) s--;
+		if (s[-1] < 128U || MB_CUR_MAX==1) s--;
 		else while ((unsigned char)*--s-0x80U<0x40 && s>str);
 	}
 	if (tailcnt) return FNM_NOMATCH;
diff --git a/src/signal/sh/restore.s b/src/signal/sh/restore.s
index ab26034..eaedcdf 100644
--- a/src/signal/sh/restore.s
+++ b/src/signal/sh/restore.s
@@ -2,7 +2,7 @@
 .type   __restore, @function
 __restore:
 	mov   #119, r3  !__NR_sigreturn
-	trapa #16
+	trapa #31
 
 	or    r0, r0
 	or    r0, r0
@@ -15,7 +15,7 @@ __restore:
 __restore_rt:
 	mov   #100, r3  !__NR_rt_sigreturn
 	add   #73, r3
-	trapa #16
+	trapa #31
 
 	or    r0, r0
 	or    r0, r0
diff --git a/src/stdio/__fdopen.c b/src/stdio/__fdopen.c
index ef8f47d..8d6ce81 100644
--- a/src/stdio/__fdopen.c
+++ b/src/stdio/__fdopen.c
@@ -54,13 +54,7 @@ FILE *__fdopen(int fd, const char *mode)
 	if (!libc.threaded) f->lock = -1;
 
 	/* Add new FILE to open file list */
-	OFLLOCK();
-	f->next = libc.ofl_head;
-	if (libc.ofl_head) libc.ofl_head->prev = f;
-	libc.ofl_head = f;
-	OFLUNLOCK();
-
-	return f;
+	return __ofl_add(f);
 }
 
 weak_alias(__fdopen, fdopen);
diff --git a/src/stdio/__stdio_exit.c b/src/stdio/__stdio_exit.c
index 716e5f7..191b445 100644
--- a/src/stdio/__stdio_exit.c
+++ b/src/stdio/__stdio_exit.c
@@ -16,8 +16,7 @@ static void close_file(FILE *f)
 void __stdio_exit(void)
 {
 	FILE *f;
-	OFLLOCK();
-	for (f=libc.ofl_head; f; f=f->next) close_file(f);
+	for (f=*__ofl_lock(); f; f=f->next) close_file(f);
 	close_file(__stdin_used);
 	close_file(__stdout_used);
 }
diff --git a/src/stdio/__stdio_read.c b/src/stdio/__stdio_read.c
index 5947344..f8fa6d3 100644
--- a/src/stdio/__stdio_read.c
+++ b/src/stdio/__stdio_read.c
@@ -1,12 +1,5 @@
 #include "stdio_impl.h"
 #include <sys/uio.h>
-#include <pthread.h>
-
-static void cleanup(void *p)
-{
-	FILE *f = p;
-	if (!f->lockcount) __unlockfile(f);
-}
 
 size_t __stdio_read(FILE *f, unsigned char *buf, size_t len)
 {
@@ -16,9 +9,7 @@ size_t __stdio_read(FILE *f, unsigned char *buf, size_t len)
 	};
 	ssize_t cnt;
 
-	pthread_cleanup_push(cleanup, f);
-	cnt = syscall_cp(SYS_readv, f->fd, iov, 2);
-	pthread_cleanup_pop(0);
+	cnt = syscall(SYS_readv, f->fd, iov, 2);
 	if (cnt <= 0) {
 		f->flags |= F_EOF ^ ((F_ERR^F_EOF) & cnt);
 		return cnt;
diff --git a/src/stdio/__stdio_write.c b/src/stdio/__stdio_write.c
index 8c89389..d2d8947 100644
--- a/src/stdio/__stdio_write.c
+++ b/src/stdio/__stdio_write.c
@@ -1,12 +1,5 @@
 #include "stdio_impl.h"
 #include <sys/uio.h>
-#include <pthread.h>
-
-static void cleanup(void *p)
-{
-	FILE *f = p;
-	if (!f->lockcount) __unlockfile(f);
-}
 
 size_t __stdio_write(FILE *f, const unsigned char *buf, size_t len)
 {
@@ -19,9 +12,7 @@ size_t __stdio_write(FILE *f, const unsigned char *buf, size_t len)
 	int iovcnt = 2;
 	ssize_t cnt;
 	for (;;) {
-		pthread_cleanup_push(cleanup, f);
-		cnt = syscall_cp(SYS_writev, f->fd, iov, iovcnt);
-		pthread_cleanup_pop(0);
+		cnt = syscall(SYS_writev, f->fd, iov, iovcnt);
 		if (cnt == rem) {
 			f->wend = f->buf + f->buf_size;
 			f->wpos = f->wbase = f->buf;
@@ -34,11 +25,8 @@ size_t __stdio_write(FILE *f, const unsigned char *buf, size_t len)
 		}
 		rem -= cnt;
 		if (cnt > iov[0].iov_len) {
-			f->wpos = f->wbase = f->buf;
 			cnt -= iov[0].iov_len;
 			iov++; iovcnt--;
-		} else if (iovcnt == 2) {
-			f->wbase += cnt;
 		}
 		iov[0].iov_base = (char *)iov[0].iov_base + cnt;
 		iov[0].iov_len -= cnt;
diff --git a/src/stdio/fclose.c b/src/stdio/fclose.c
index 317b3c9..839d88a 100644
--- a/src/stdio/fclose.c
+++ b/src/stdio/fclose.c
@@ -14,11 +14,11 @@ int fclose(FILE *f)
 	__unlist_locked_file(f);
 
 	if (!(perm = f->flags & F_PERM)) {
-		OFLLOCK();
+		FILE **head = __ofl_lock();
 		if (f->prev) f->prev->next = f->next;
 		if (f->next) f->next->prev = f->prev;
-		if (libc.ofl_head == f) libc.ofl_head = f->next;
-		OFLUNLOCK();
+		if (*head == f) *head = f->next;
+		__ofl_unlock();
 	}
 
 	r = fflush(f);
diff --git a/src/stdio/fflush.c b/src/stdio/fflush.c
index 7bf862a..3f462c8 100644
--- a/src/stdio/fflush.c
+++ b/src/stdio/fflush.c
@@ -35,13 +35,12 @@ int fflush(FILE *f)
 
 	r = __stdout_used ? fflush(__stdout_used) : 0;
 
-	OFLLOCK();
-	for (f=libc.ofl_head; f; f=f->next) {
+	for (f=*__ofl_lock(); f; f=f->next) {
 		FLOCK(f);
 		if (f->wpos > f->wbase) r |= __fflush_unlocked(f);
 		FUNLOCK(f);
 	}
-	OFLUNLOCK();
+	__ofl_unlock();
 	
 	return r;
 }
diff --git a/src/stdio/fgetwc.c b/src/stdio/fgetwc.c
index 8626d54..e455cfe 100644
--- a/src/stdio/fgetwc.c
+++ b/src/stdio/fgetwc.c
@@ -1,8 +1,9 @@
 #include "stdio_impl.h"
+#include "locale_impl.h"
 #include <wchar.h>
 #include <errno.h>
 
-wint_t __fgetwc_unlocked(FILE *f)
+static wint_t __fgetwc_unlocked_internal(FILE *f)
 {
 	mbstate_t st = { 0 };
 	wchar_t wc;
@@ -10,8 +11,6 @@ wint_t __fgetwc_unlocked(FILE *f)
 	unsigned char b;
 	size_t l;
 
-	f->mode |= f->mode+1;
-
 	/* Convert character from buffer if possible */
 	if (f->rpos < f->rend) {
 		l = mbrtowc(&wc, (void *)f->rpos, f->rend - f->rpos, &st);
@@ -39,6 +38,16 @@ wint_t __fgetwc_unlocked(FILE *f)
 	return wc;
 }
 
+wint_t __fgetwc_unlocked(FILE *f)
+{
+	locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
+	if (f->mode <= 0) fwide(f, 1);
+	*ploc = f->locale;
+	wchar_t wc = __fgetwc_unlocked_internal(f);
+	*ploc = loc;
+	return wc;
+}
+
 wint_t fgetwc(FILE *f)
 {
 	wint_t c;
diff --git a/src/stdio/fmemopen.c b/src/stdio/fmemopen.c
index d784960..7c193a5 100644
--- a/src/stdio/fmemopen.c
+++ b/src/stdio/fmemopen.c
@@ -110,11 +110,5 @@ FILE *fmemopen(void *restrict buf, size_t size, const char *restrict mode)
 
 	if (!libc.threaded) f->lock = -1;
 
-	OFLLOCK();
-	f->next = libc.ofl_head;
-	if (libc.ofl_head) libc.ofl_head->prev = f;
-	libc.ofl_head = f;
-	OFLUNLOCK();
-
-	return f;
+	return __ofl_add(f);
 }
diff --git a/src/stdio/fopen.c b/src/stdio/fopen.c
index 07bdb6e..252f082 100644
--- a/src/stdio/fopen.c
+++ b/src/stdio/fopen.c
@@ -18,7 +18,7 @@ FILE *fopen(const char *restrict filename, const char *restrict mode)
 	/* Compute the flags to pass to open() */
 	flags = __fmodeflags(mode);
 
-	fd = sys_open_cp(filename, flags, 0666);
+	fd = sys_open(filename, flags, 0666);
 	if (fd < 0) return 0;
 	if (flags & O_CLOEXEC)
 		__syscall(SYS_fcntl, fd, F_SETFD, FD_CLOEXEC);
diff --git a/src/stdio/fputwc.c b/src/stdio/fputwc.c
index 7b621dd..789fe9c 100644
--- a/src/stdio/fputwc.c
+++ b/src/stdio/fputwc.c
@@ -1,4 +1,5 @@
 #include "stdio_impl.h"
+#include "locale_impl.h"
 #include <wchar.h>
 #include <limits.h>
 #include <ctype.h>
@@ -7,8 +8,10 @@ wint_t __fputwc_unlocked(wchar_t c, FILE *f)
 {
 	char mbc[MB_LEN_MAX];
 	int l;
+	locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
 
-	f->mode |= f->mode+1;
+	if (f->mode <= 0) fwide(f, 1);
+	*ploc = f->locale;
 
 	if (isascii(c)) {
 		c = putc_unlocked(c, f);
@@ -20,6 +23,8 @@ wint_t __fputwc_unlocked(wchar_t c, FILE *f)
 		l = wctomb(mbc, c);
 		if (l < 0 || __fwritex((void *)mbc, l, f) < l) c = WEOF;
 	}
+	if (c==WEOF) f->flags |= F_ERR;
+	*ploc = loc;
 	return c;
 }
 
diff --git a/src/stdio/fputws.c b/src/stdio/fputws.c
index 5723cbc..0ed02f1 100644
--- a/src/stdio/fputws.c
+++ b/src/stdio/fputws.c
@@ -1,23 +1,28 @@
 #include "stdio_impl.h"
+#include "locale_impl.h"
 #include <wchar.h>
 
 int fputws(const wchar_t *restrict ws, FILE *restrict f)
 {
 	unsigned char buf[BUFSIZ];
 	size_t l=0;
+	locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
 
 	FLOCK(f);
 
-	f->mode |= f->mode+1;
+	fwide(f, 1);
+	*ploc = f->locale;
 
 	while (ws && (l = wcsrtombs((void *)buf, (void*)&ws, sizeof buf, 0))+1 > 1)
 		if (__fwritex(buf, l, f) < l) {
 			FUNLOCK(f);
+			*ploc = loc;
 			return -1;
 		}
 
 	FUNLOCK(f);
 
+	*ploc = loc;
 	return l; /* 0 or -1 */
 }
 
diff --git a/src/stdio/fwide.c b/src/stdio/fwide.c
index 8088e7a..8410b15 100644
--- a/src/stdio/fwide.c
+++ b/src/stdio/fwide.c
@@ -1,13 +1,14 @@
-#include <wchar.h>
 #include "stdio_impl.h"
-
-#define SH (8*sizeof(int)-1)
-#define NORMALIZE(x) ((x)>>SH | -((-(x))>>SH))
+#include "locale_impl.h"
 
 int fwide(FILE *f, int mode)
 {
 	FLOCK(f);
-	if (!f->mode) f->mode = NORMALIZE(mode);
+	if (mode) {
+		if (!f->locale) f->locale = MB_CUR_MAX==1
+			? C_LOCALE : UTF8_LOCALE;
+		if (!f->mode) f->mode = mode>0 ? 1 : -1;
+	}
 	mode = f->mode;
 	FUNLOCK(f);
 	return mode;
diff --git a/src/stdio/ofl.c b/src/stdio/ofl.c
new file mode 100644
index 0000000..b143999
--- /dev/null
+++ b/src/stdio/ofl.c
@@ -0,0 +1,16 @@
+#include "stdio_impl.h"
+#include "libc.h"
+
+static FILE *ofl_head;
+static volatile int ofl_lock[2];
+
+FILE **__ofl_lock()
+{
+	LOCK(ofl_lock);
+	return &ofl_head;
+}
+
+void __ofl_unlock()
+{
+	UNLOCK(ofl_lock);
+}
diff --git a/src/stdio/ofl_add.c b/src/stdio/ofl_add.c
new file mode 100644
index 0000000..d7de9f1
--- /dev/null
+++ b/src/stdio/ofl_add.c
@@ -0,0 +1,11 @@
+#include "stdio_impl.h"
+
+FILE *__ofl_add(FILE *f)
+{
+	FILE **head = __ofl_lock();
+	f->next = *head;
+	if (*head) (*head)->prev = f;
+	*head = f;
+	__ofl_unlock();
+	return f;
+}
diff --git a/src/stdio/open_memstream.c b/src/stdio/open_memstream.c
index 9eafdfb..58504c9 100644
--- a/src/stdio/open_memstream.c
+++ b/src/stdio/open_memstream.c
@@ -79,11 +79,5 @@ FILE *open_memstream(char **bufp, size_t *sizep)
 
 	if (!libc.threaded) f->lock = -1;
 
-	OFLLOCK();
-	f->next = libc.ofl_head;
-	if (libc.ofl_head) libc.ofl_head->prev = f;
-	libc.ofl_head = f;
-	OFLUNLOCK();
-
-	return f;
+	return __ofl_add(f);
 }
diff --git a/src/stdio/open_wmemstream.c b/src/stdio/open_wmemstream.c
index 3537030..7ab2c64 100644
--- a/src/stdio/open_wmemstream.c
+++ b/src/stdio/open_wmemstream.c
@@ -81,11 +81,5 @@ FILE *open_wmemstream(wchar_t **bufp, size_t *sizep)
 
 	if (!libc.threaded) f->lock = -1;
 
-	OFLLOCK();
-	f->next = libc.ofl_head;
-	if (libc.ofl_head) libc.ofl_head->prev = f;
-	libc.ofl_head = f;
-	OFLUNLOCK();
-
-	return f;
+	return __ofl_add(f);
 }
diff --git a/src/stdio/ungetwc.c b/src/stdio/ungetwc.c
index 913f716..80d6e20 100644
--- a/src/stdio/ungetwc.c
+++ b/src/stdio/ungetwc.c
@@ -1,4 +1,5 @@
 #include "stdio_impl.h"
+#include "locale_impl.h"
 #include <wchar.h>
 #include <limits.h>
 #include <ctype.h>
@@ -8,21 +9,19 @@ wint_t ungetwc(wint_t c, FILE *f)
 {
 	unsigned char mbc[MB_LEN_MAX];
 	int l=1;
-
-	if (c == WEOF) return c;
-
-	/* Try conversion early so we can fail without locking if invalid */
-	if (!isascii(c) && (l = wctomb((void *)mbc, c)) < 0)
-		return WEOF;
+	locale_t *ploc = &CURRENT_LOCALE, loc = *ploc;
 
 	FLOCK(f);
 
-	f->mode |= f->mode+1;
+	if (f->mode <= 0) fwide(f, 1);
+	*ploc = f->locale;
 
 	if (!f->rpos) __toread(f);
-	if (!f->rpos || f->rpos < f->buf - UNGET + l) {
+	if (!f->rpos || f->rpos < f->buf - UNGET + l || c == WEOF ||
+	    (!isascii(c) && (l = wctomb((void *)mbc, c)) < 0)) {
 		FUNLOCK(f);
-		return EOF;
+		*ploc = loc;
+		return WEOF;
 	}
 
 	if (isascii(c)) *--f->rpos = c;
@@ -31,5 +30,6 @@ wint_t ungetwc(wint_t c, FILE *f)
 	f->flags &= ~F_EOF;
 
 	FUNLOCK(f);
+	*ploc = loc;
 	return c;
 }
diff --git a/src/stdio/vfwprintf.c b/src/stdio/vfwprintf.c
index ebdff00..f06d5ae 100644
--- a/src/stdio/vfwprintf.c
+++ b/src/stdio/vfwprintf.c
@@ -293,7 +293,10 @@ static int wprintf_core(FILE *f, const wchar_t *fmt, va_list *ap, union arg *nl_
 			if ((fl&LEFT_ADJ)) fprintf(f, "%.*s", w-p, "");
 			l=w;
 			continue;
+		case 'm':
+			arg.p = strerror(errno);
 		case 's':
+			if (!arg.p) arg.p = "(null)";
 			bs = arg.p;
 			if (p<0) p = INT_MAX;
 			for (i=l=0; l<p && (i=mbtowc(&wc, bs, MB_LEN_MAX))>0; bs+=i, l++);
@@ -356,7 +359,7 @@ int vfwprintf(FILE *restrict f, const wchar_t *restrict fmt, va_list ap)
 	}
 
 	FLOCK(f);
-	f->mode |= f->mode+1;
+	fwide(f, 1);
 	olderr = f->flags & F_ERR;
 	f->flags &= ~F_ERR;
 	ret = wprintf_core(f, fmt, &ap2, nl_arg, nl_type);
diff --git a/src/stdio/vfwscanf.c b/src/stdio/vfwscanf.c
index ac5c2c2..223aad4 100644
--- a/src/stdio/vfwscanf.c
+++ b/src/stdio/vfwscanf.c
@@ -104,7 +104,7 @@ int vfwscanf(FILE *restrict f, const wchar_t *restrict fmt, va_list ap)
 
 	FLOCK(f);
 
-	f->mode |= f->mode+1;
+	fwide(f, 1);
 
 	for (p=fmt; *p; p++) {
 
diff --git a/src/string/strverscmp.c b/src/string/strverscmp.c
index 6f37cc6..4daf276 100644
--- a/src/string/strverscmp.c
+++ b/src/string/strverscmp.c
@@ -2,40 +2,33 @@
 #include <ctype.h>
 #include <string.h>
 
-int strverscmp(const char *l, const char *r)
+int strverscmp(const char *l0, const char *r0)
 {
-	int haszero=1;
-	while (*l==*r) {
-		if (!*l) return 0;
+	const unsigned char *l = (const void *)l0;
+	const unsigned char *r = (const void *)r0;
+	size_t i, dp, j;
+	int z = 1;
 
-		if (*l=='0') {
-			if (haszero==1) {
-				haszero=0;
-			}
-		} else if (isdigit(*l)) {
-			if (haszero==1) {
-				haszero=2;
-			}
-		} else {
-			haszero=1;
-		}
-		l++; r++;
+	/* Find maximal matching prefix and track its maximal digit
+	 * suffix and whether those digits are all zeros. */
+	for (dp=i=0; l[i]==r[i]; i++) {
+		int c = l[i];
+		if (!c) return 0;
+		if (!isdigit(c)) dp=i+1, z=1;
+		else if (c!='0') z=0;
 	}
-	if (haszero==1 && (*l=='0' || *r=='0')) {
-		haszero=0;
-	}
-	if ((isdigit(*l) && isdigit(*r) ) && haszero) {
-		size_t lenl=0, lenr=0;
-		while (isdigit(l[lenl]) ) lenl++;
-		while (isdigit(r[lenr]) ) lenr++;
-		if (lenl==lenr) {
-			return (*l -  *r);
-		} else if (lenl>lenr) {
-			return 1;
-		} else {
-			return -1;
-		}
-	} else {
-		return (*l -  *r);
+
+	if (l[dp]!='0' && r[dp]!='0') {
+		/* If we're not looking at a digit sequence that began
+		 * with a zero, longest digit string is greater. */
+		for (j=i; isdigit(l[j]); j++)
+			if (!isdigit(r[j])) return 1;
+		if (isdigit(r[j])) return -1;
+	} else if (z && dp<i && (isdigit(l[i]) || isdigit(r[i]))) {
+		/* Otherwise, if common prefix of digit sequence is
+		 * all zeros, digits order less than non-digits. */
+		return (unsigned char)(l[i]-'0') - (unsigned char)(r[i]-'0');
 	}
+
+	return l[i] - r[i];
 }
diff --git a/src/thread/__tls_get_addr.c b/src/thread/__tls_get_addr.c
index 3633396..84a413d 100644
--- a/src/thread/__tls_get_addr.c
+++ b/src/thread/__tls_get_addr.c
@@ -8,9 +8,9 @@ void *__tls_get_addr(size_t *v)
 	__attribute__((__visibility__("hidden")))
 	void *__tls_get_new(size_t *);
 	if (v[0]<=(size_t)self->dtv[0])
-		return (char *)self->dtv[v[0]]+v[1];
+		return (char *)self->dtv[v[0]]+v[1]+DTP_OFFSET;
 	return __tls_get_new(v);
 #else
-	return (char *)self->dtv[1]+v[1];
+	return (char *)self->dtv[1]+v[1]+DTP_OFFSET;
 #endif
 }
diff --git a/src/thread/__unmapself.c b/src/thread/__unmapself.c
index e69de29..1d3bee1 100644
--- a/src/thread/__unmapself.c
+++ b/src/thread/__unmapself.c
@@ -0,0 +1,29 @@
+#include "pthread_impl.h"
+#include "atomic.h"
+#include "syscall.h"
+/* cheat and reuse CRTJMP macro from dynlink code */
+#include "dynlink.h"
+
+static volatile int lock;
+static void *unmap_base;
+static size_t unmap_size;
+static char shared_stack[256];
+
+static void do_unmap()
+{
+	__syscall(SYS_munmap, unmap_base, unmap_size);
+	__syscall(SYS_exit);
+}
+
+void __unmapself(void *base, size_t size)
+{
+	int tid=__pthread_self()->tid;
+	char *stack = shared_stack + sizeof shared_stack;
+	stack -= (uintptr_t)stack % 16;
+	while (lock || a_cas(&lock, 0, tid))
+		a_spin();
+	__syscall(SYS_set_tid_address, &lock);
+	unmap_base = base;
+	unmap_size = size;
+	CRTJMP(do_unmap, stack);
+}
diff --git a/src/thread/mips/__unmapself.s b/src/thread/mips/__unmapself.s
index 9aa0371..ba139dc 100644
--- a/src/thread/mips/__unmapself.s
+++ b/src/thread/mips/__unmapself.s
@@ -2,6 +2,7 @@
 .global __unmapself
 .type   __unmapself,@function
 __unmapself:
+	move $sp, $25
 	li $2, 4091
 	syscall
 	li $4, 0
diff --git a/src/thread/pthread_create.c b/src/thread/pthread_create.c
index de72818..e7df34a 100644
--- a/src/thread/pthread_create.c
+++ b/src/thread/pthread_create.c
@@ -191,8 +191,9 @@ int __pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict att
 	if (!libc.can_do_threads) return ENOSYS;
 	self = __pthread_self();
 	if (!libc.threaded) {
-		for (FILE *f=libc.ofl_head; f; f=f->next)
+		for (FILE *f=*__ofl_lock(); f; f=f->next)
 			init_file_lock(f);
+		__ofl_unlock();
 		init_file_lock(__stdin_used);
 		init_file_lock(__stdout_used);
 		init_file_lock(__stderr_used);
@@ -231,7 +232,8 @@ int __pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict att
 		if (guard) {
 			map = __mmap(0, size, PROT_NONE, MAP_PRIVATE|MAP_ANON, -1, 0);
 			if (map == MAP_FAILED) goto fail;
-			if (__mprotect(map+guard, size-guard, PROT_READ|PROT_WRITE)) {
+			if (__mprotect(map+guard, size-guard, PROT_READ|PROT_WRITE)
+			    && errno != ENOSYS) {
 				__munmap(map, size);
 				goto fail;
 			}
diff --git a/src/thread/sh/__set_thread_area.s b/src/thread/sh/__set_thread_area.s
index d9f1181..e69de29 100644
--- a/src/thread/sh/__set_thread_area.s
+++ b/src/thread/sh/__set_thread_area.s
@@ -1,6 +0,0 @@
-.global __set_thread_area
-.type   __set_thread_area, @function
-__set_thread_area:
-	ldc r4, gbr
-	rts
-	 mov #0, r0
diff --git a/src/thread/sh/__unmapself.s b/src/thread/sh/__unmapself.s
index b34c3c8..0161d53 100644
--- a/src/thread/sh/__unmapself.s
+++ b/src/thread/sh/__unmapself.s
@@ -1,9 +1,9 @@
 .text
-.global __unmapself
-.type   __unmapself, @function
-__unmapself:
+.global __unmapself_sh_mmu
+.type   __unmapself_sh_mmu, @function
+__unmapself_sh_mmu:
 	mov   #91, r3  ! SYS_munmap
-	trapa #18
+	trapa #31
 
 	or    r0, r0
 	or    r0, r0
@@ -13,7 +13,7 @@ __unmapself:
 
 	mov   #1, r3   ! SYS_exit
 	mov   #0, r4
-	trapa #17
+	trapa #31
 
 	or    r0, r0
 	or    r0, r0
diff --git a/src/thread/sh/clone.s b/src/thread/sh/clone.s
index d6c9184..f8ad845 100644
--- a/src/thread/sh/clone.s
+++ b/src/thread/sh/clone.s
@@ -17,7 +17,7 @@ __clone:
 	mov.l @r15,     r6   ! r6 = ptid
 	mov.l @(8,r15), r7   ! r7 = ctid
 	mov.l @(4,r15), r0   ! r0 = tls
-	trapa #21
+	trapa #31
 
 	or r0, r0
 	or r0, r0
@@ -38,7 +38,7 @@ __clone:
 
 	mov   #1, r3   ! __NR_exit
 	mov   r0, r4
-	trapa #17
+	trapa #31
 
 	or   r0, r0
 	or   r0, r0
diff --git a/src/thread/sh/syscall_cp.s b/src/thread/sh/syscall_cp.s
index 6b28ddf..c3cafac 100644
--- a/src/thread/sh/syscall_cp.s
+++ b/src/thread/sh/syscall_cp.s
@@ -31,7 +31,7 @@ L1:	.long __cancel@PLT-(1b-.)
 	mov.l @(4,r15), r7
 	mov.l @(8,r15), r0
 	mov.l @(12,r15), r1
-	trapa #22
+	trapa #31
 
 __cp_end:
 	! work around hardware bug
diff --git a/src/unistd/sh/pipe.s b/src/unistd/sh/pipe.s
index d865ae3..46c4908 100644
--- a/src/unistd/sh/pipe.s
+++ b/src/unistd/sh/pipe.s
@@ -2,7 +2,7 @@
 .type   pipe, @function
 pipe:
 	mov    #42, r3
-	trapa  #17
+	trapa  #31
 
 	! work around hardware bug
 	or     r0, r0