fix misalignment of dtv in static-linked programs with odd-sized TLS

[musl] / src / env / __init_tls.c

diff --git a/src/env/__init_tls.c b/src/env/__init_tls.c
index a4704f0..73551e6 100644 (file)
--- a/src/env/__init_tls.c
+++ b/src/env/__init_tls.c
@@ -2,53 +2,67 @@
 #include <limits.h>
 #include <sys/mman.h>
 #include <string.h>
+#include <stddef.h>
 #include "pthread_impl.h"
 #include "libc.h"
 #include "atomic.h"
+#include "syscall.h"
+
+#ifndef SHARED
+static
+#endif
+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;
+       td->tid = __syscall(SYS_set_tid_address, &td->tid);
+       td->locale = &libc.global_locale;
+       td->robust_list.head = &td->robust_list.head;
+       return 0;
+}
 
 #ifndef SHARED
 
-static void *image;
-static size_t len, size, align;
+static struct builtin_tls {
+       char c;
+       struct pthread pt;
+       void *space[16];
+} builtin_tls[1];
+#define MIN_TLS_ALIGN offsetof(struct builtin_tls, pt)
+
+struct tls_image {
+       void *image;
+       size_t len, size, align;
+} __static_tls;
+
+#define T __static_tls
 
 void *__copy_tls(unsigned char *mem)
 {
        pthread_t td;
-       if (!image) return mem;
+       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)) & (align-1);
+       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 & (align-1);
+       mem -= (uintptr_t)mem & (T.align-1);
        td = (pthread_t)mem;
-       mem -= size;
+       mem -= T.size;
 #endif
-       td->dtv = dtv;
+       td->dtv = td->dtv_copy = dtv;
        dtv[1] = mem;
-       memcpy(mem, image, len);
+       memcpy(mem, T.image, T.len);
        return td;
 }
 
-void *__tls_get_addr(size_t *v)
-{
-       return (char *)__pthread_self()->dtv[1]+v[1];
-}
-
-static void *simple(void *p)
-{
-       *(void **)p = p;
-       return __set_thread_area(TP_ADJ(p)) ? 0 : p;
-}
-
-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
@@ -57,12 +71,11 @@ typedef Elf64_Phdr Phdr;
 
 void __init_tls(size_t *aux)
 {
-       unsigned char *p, *mem;
+       unsigned char *p;
        size_t n;
        Phdr *phdr, *tls_phdr=0;
        size_t base = 0;
-
-       libc.tls_size = sizeof(struct pthread);
+       void *mem;
 
        for (p=(void *)aux[AT_PHDR],n=aux[AT_PHNUM]; n; n--,p+=aux[AT_PHENT]) {
                phdr = (void *)p;
@@ -71,21 +84,38 @@ void __init_tls(size_t *aux)
                if (phdr->p_type == PT_TLS)
                        tls_phdr = phdr;
        }
-       if (!tls_phdr) return;
 
-       image = (void *)(base + tls_phdr->p_vaddr);
-       len = tls_phdr->p_filesz;
-       size = tls_phdr->p_memsz;
-       align = tls_phdr->p_align;
+       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 < MIN_TLS_ALIGN) T.align = MIN_TLS_ALIGN;
 
-       size += (-size - (uintptr_t)image) & (align-1);
-       if (align < 4*sizeof(size_t)) align = 4*sizeof(size_t);
+       libc.tls_size = 2*sizeof(void *)+T.size+T.align+sizeof(struct pthread)
+               + MIN_TLS_ALIGN-1 & -MIN_TLS_ALIGN;
 
-       libc.tls_size = 2*sizeof(void *)+size+align+sizeof(struct pthread);
+       if (libc.tls_size > sizeof builtin_tls) {
+#ifndef SYS_mmap2
+#define SYS_mmap2 SYS_mmap
+#endif
+               mem = (void *)__syscall(
+                       SYS_mmap2,
+                       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;
+       }
 
-       mem = __mmap(0, libc.tls_size, PROT_READ|PROT_WRITE,
-               MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
-       if (!__install_initial_tls(__copy_tls(mem))) a_crash();
+       /* Failure to initialize thread pointer is always fatal. */
+       if (__init_tp(__copy_tls(mem)) < 0)
+               a_crash();
 }
 #else
 void __init_tls(size_t *auxv) { }