eliminate use of cached pid from thread structure

[musl] / src / env / __init_tls.c

diff --git a/src/env/__init_tls.c b/src/env/__init_tls.c
index b19bdb6..6cca968 100644 (file)
--- a/src/env/__init_tls.c
+++ b/src/env/__init_tls.c
@@ -1,70 +1,114 @@
 #include <elf.h>
 #include <limits.h>
+#include <sys/mman.h>
+#include <string.h>
 #include "pthread_impl.h"
 #include "libc.h"
 #include "atomic.h"
+#include "syscall.h"
+
+int __init_tp(void *p)
+{
+       pthread_t td = p;
+       td->self = td;
+       int r = __set_thread_area(TP_ADJ(p));
+       if (r < 0) return -1;
+       if (!r) libc.can_do_threads = 1;
+       libc.has_thread_pointer = 1;
+       td->tid = __syscall(SYS_set_tid_address, &td->tid);
+       td->locale = &libc.global_locale;
+       return 0;
+}
 
 #ifndef SHARED
 
-static void *image;
-static size_t len, size, align;
+static long long builtin_tls[(sizeof(struct pthread) + 64)/sizeof(long long)];
 
-void *__copy_tls(unsigned char *mem)
-{
-       mem += -size & (4*sizeof(size_t)-1);
-       mem += ((uintptr_t)image - (uintptr_t)mem) & (align-1);
-       memcpy(mem, image, len);
-       return mem + size;
-}
+struct tls_image {
+       void *image;
+       size_t len, size, align;
+} __static_tls ATTR_LIBC_VISIBILITY;
+
+#define T __static_tls
 
-static void *simple(void *p)
+void *__copy_tls(unsigned char *mem)
 {
-       *(void **)p = p;
-       return __set_thread_area(p) ? 0 : p;
+       pthread_t td;
+       if (!T.image) return mem;
+       void **dtv = (void *)mem;
+       dtv[0] = (void *)1;
+#ifdef TLS_ABOVE_TP
+       mem += sizeof(void *) * 2;
+       mem += -((uintptr_t)mem + sizeof(struct pthread)) & (T.align-1);
+       td = (pthread_t)mem;
+       mem += sizeof(struct pthread);
+#else
+       mem += libc.tls_size - sizeof(struct pthread);
+       mem -= (uintptr_t)mem & (T.align-1);
+       td = (pthread_t)mem;
+       mem -= T.size;
+#endif
+       td->dtv = dtv;
+       dtv[1] = mem;
+       memcpy(mem, T.image, T.len);
+       return td;
 }
 
-weak_alias(simple, __install_initial_tls);
-
-void *__mmap(void *, size_t, int, int, int, off_t);
-
 #if ULONG_MAX == 0xffffffff
 typedef Elf32_Phdr Phdr;
 #else
 typedef Elf64_Phdr Phdr;
 #endif
 
-#define AUX_CNT 6
-
-void __init_tls(size_t *auxv)
+void __init_tls(size_t *aux)
 {
-       size_t i, aux[AUX_CNT] = { 0 };
-       unsigned char *p, *mem;
-       size_t n, d;
+       unsigned char *p;
+       size_t n;
        Phdr *phdr, *tls_phdr=0;
        size_t base = 0;
+       void *mem;
 
-       for (; auxv[0]; auxv+=2) if (auxv[0]<AUX_CNT) aux[auxv[0]] = auxv[1];
-       p = (void *)aux[AT_PHDR];
-       for (p=(void *)aux[AT_PHDR]; aux[AT_PHNUM]--; p+=aux[AT_PHENT]) {
+       libc.tls_size = sizeof(struct pthread);
+
+       for (p=(void *)aux[AT_PHDR],n=aux[AT_PHNUM]; n; n--,p+=aux[AT_PHENT]) {
                phdr = (void *)p;
                if (phdr->p_type == PT_PHDR)
                        base = aux[AT_PHDR] - phdr->p_vaddr;
                if (phdr->p_type == PT_TLS)
                        tls_phdr = phdr;
        }
-       if (!tls_phdr) return;
 
-       libc.tls_size = size+align+8*sizeof(size_t)+sizeof(struct pthread);
+       if (tls_phdr) {
+               T.image = (void *)(base + tls_phdr->p_vaddr);
+               T.len = tls_phdr->p_filesz;
+               T.size = tls_phdr->p_memsz;
+               T.align = tls_phdr->p_align;
+       }
+
+       T.size += (-T.size - (uintptr_t)T.image) & (T.align-1);
+       if (T.align < 4*sizeof(size_t)) T.align = 4*sizeof(size_t);
+
+       libc.tls_size = 2*sizeof(void *)+T.size+T.align+sizeof(struct pthread);
 
-       image = (void *)(base + tls_phdr->p_vaddr);
-       len = tls_phdr->p_filesz;
-       size = tls_phdr->p_memsz;
-       align = tls_phdr->p_align;
-       mem = __mmap(0, libc.tls_size, PROT_READ|PROT_WRITE,
-               MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
-       if (mem == MAP_FAILED) a_crash();
+       if (libc.tls_size > sizeof builtin_tls) {
+               mem = (void *)__syscall(
+#ifdef SYS_mmap2
+                       SYS_mmap2,
+#else
+                       SYS_mmap,
+#endif
+                       0, libc.tls_size, PROT_READ|PROT_WRITE,
+                       MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
+               /* -4095...-1 cast to void * will crash on dereference anyway,
+                * so don't bloat the init code checking for error codes and
+                * explicitly calling a_crash(). */
+       } else {
+               mem = builtin_tls;
+       }
 
-       if (!__install_initial_tls(__copy_tls(mem))) a_crash();
+       /* Failure to initialize thread pointer is fatal if TLS is used. */
+       if (__init_tp(__copy_tls(mem)) < 0 && tls_phdr)
+               a_crash();
 }
 #else
 void __init_tls(size_t *auxv) { }