summaryrefslogtreecommitdiff
path: root/src/thread
diff options
context:
space:
mode:
authorRich Felker <dalias@aerifal.cx>2019-02-17 23:22:27 -0500
committerRich Felker <dalias@aerifal.cx>2019-02-18 21:01:16 -0500
commit9d44b6460ab603487dab4d916342d9ba4467e6b9 (patch)
treec7aa27a062fe7847972b204ced082217b5e8b0ad /src/thread
parent805288929fdf511b4044cf07c59e02e2eaa9c546 (diff)
downloadmusl-9d44b6460ab603487dab4d916342d9ba4467e6b9.tar.gz
install dynamic tls synchronously at dlopen, streamline access
previously, dynamic loading of new libraries with thread-local storage allocated the storage needed for all existing threads at load-time, precluding late failure that can't be handled, but left installation in existing threads to take place lazily on first access. this imposed an additional memory access and branch on every dynamic tls access, and imposed a requirement, which was not actually met, that the dynamic tlsdesc asm functions preserve all call-clobbered registers before calling C code to to install new dynamic tls on first access. the x86[_64] versions of this code wrongly omitted saving and restoring of fpu/vector registers, assuming the compiler would not generate anything using them in the called C code. the arm and aarch64 versions saved known existing registers, but failed to be future-proof against expansion of the register file. now that we track live threads in a list, it's possible to install the new dynamic tls for each thread at dlopen time. for the most part, synchronization is not needed, because if a thread has not synchronized with completion of the dlopen, there is no way it can meaningfully request access to a slot past the end of the old dtv, which remains valid for accessing slots which already existed. however, it is necessary to ensure that, if a thread sees its new dtv pointer, it sees correct pointers in each of the slots that existed prior to the dlopen. my understanding is that, on most real-world coherency architectures including all the ones we presently support, a built-in consume order guarantees this; however, don't rely on that. instead, the SYS_membarrier syscall is used to ensure that all threads see the stores to the slots of their new dtv prior to the installation of the new dtv. if it is not supported, the same is implemented in userspace via signals, using the same mechanism as __synccall. the __tls_get_addr function, variants, and dynamic tlsdesc asm functions are all updated to remove the fallback paths for claiming new dynamic tls, and are now all branch-free.
Diffstat (limited to 'src/thread')
-rw-r--r--src/thread/__tls_get_addr.c7
-rw-r--r--src/thread/i386/tls.s8
-rw-r--r--src/thread/pthread_create.c2
3 files changed, 3 insertions, 14 deletions
diff --git a/src/thread/__tls_get_addr.c b/src/thread/__tls_get_addr.c
index d7afdabd..19524fe0 100644
--- a/src/thread/__tls_get_addr.c
+++ b/src/thread/__tls_get_addr.c
@@ -1,12 +1,7 @@
-#include <stddef.h>
#include "pthread_impl.h"
void *__tls_get_addr(tls_mod_off_t *v)
{
pthread_t self = __pthread_self();
- if (v[0] <= self->dtv[0])
- return (void *)(self->dtv[v[0]] + v[1]);
- return __tls_get_new(v);
+ return (void *)(self->dtv[v[0]] + v[1]);
}
-
-weak_alias(__tls_get_addr, __tls_get_new);
diff --git a/src/thread/i386/tls.s b/src/thread/i386/tls.s
index 76d5d462..6e4c4cb9 100644
--- a/src/thread/i386/tls.s
+++ b/src/thread/i386/tls.s
@@ -4,14 +4,6 @@
___tls_get_addr:
mov %gs:4,%edx
mov (%eax),%ecx
- cmp %ecx,(%edx)
- jc 1f
mov 4(%eax),%eax
add (%edx,%ecx,4),%eax
ret
-1: push %eax
-.weak __tls_get_new
-.hidden __tls_get_new
- call __tls_get_new
- pop %edx
- ret
diff --git a/src/thread/pthread_create.c b/src/thread/pthread_create.c
index cec82157..0142b347 100644
--- a/src/thread/pthread_create.c
+++ b/src/thread/pthread_create.c
@@ -15,6 +15,7 @@ weak_alias(dummy_0, __release_ptc);
weak_alias(dummy_0, __pthread_tsd_run_dtors);
weak_alias(dummy_0, __do_orphaned_stdio_locks);
weak_alias(dummy_0, __dl_thread_cleanup);
+weak_alias(dummy_0, __dl_prepare_for_threads);
void __tl_lock(void)
{
@@ -235,6 +236,7 @@ int __pthread_create(pthread_t *restrict res, const pthread_attr_t *restrict att
init_file_lock(__stderr_used);
__syscall(SYS_rt_sigprocmask, SIG_UNBLOCK, SIGPT_SET, 0, _NSIG/8);
self->tsd = (void **)__pthread_tsd_main;
+ __dl_prepare_for_threads();
libc.threaded = 1;
}
if (attrp && !c11) attr = *attrp;