nptl: Add rseq auxvals

Get the rseq feature size and alignment requirement from the auxiliary
vector for use inside the dynamic loader. Use '__rseq_size' directly to
store the feature size. If the main thread registration fails or is
disabled by tunable, reset the value to 0.

This will be used in the TLS block allocator to compute the size and
alignment of the rseq area block for the extended ABI support.

Signed-off-by: Michael Jeanson <mjeanson@efficios.com>
Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Reviewed-by: Florian Weimer <fweimer@redhat.com>

sysdeps/nptl/dl-tls_init_tp.c

@@ -46,6 +46,8 @@ rtld_mutex_dummy (pthread_mutex_t *lock)
 
 const unsigned int __rseq_flags;
 
+size_t _rseq_align attribute_hidden;
+
 void
 __tls_pre_init_tp (void)
 {
@@ -99,10 +101,14 @@ __tls_init_tp (void)
   }
 
   {
-    bool do_rseq = true;
-    do_rseq = TUNABLE_GET (rseq, int, NULL);
-    if (rseq_register_current_thread (pd, do_rseq))
-      _rseq_size = RSEQ_AREA_SIZE_INITIAL_USED;
+    /* If the registration fails or is disabled by tunable, the public
+       '__rseq_size' will be set to '0' regardless of the feature size of the
+       allocated rseq area.  An rseq area of at least 32 bytes is always
+       allocated since application code is allowed to check the status of the
+       rseq registration by reading the content of the 'cpu_id' field.  */
+    bool do_rseq = TUNABLE_GET (rseq, int, NULL);
+    if (!rseq_register_current_thread (pd, do_rseq))
+      _rseq_size = 0;
 
 #ifdef RSEQ_SIG
     /* This should be a compile-time constant, but the current

sysdeps/unix/sysv/linux/dl-parse_auxv.h

@@ -21,6 +21,7 @@
 #include <fpu_control.h>
 #include <ldsodefs.h>
 #include <link.h>
+#include <rseq-internal.h>
 
 typedef ElfW(Addr) dl_parse_auxv_t[AT_MINSIGSTKSZ + 1];
 
@@ -59,5 +60,17 @@ void _dl_parse_auxv (ElfW(auxv_t) *av, dl_parse_auxv_t auxv_values)
     GLRO(dl_sysinfo) = auxv_values[AT_SYSINFO];
 #endif
 
+  /* Get the rseq feature size, with a minimum of RSEQ_AREA_SIZE_INITIAL_USED
+     (20) for kernels that don't have AT_RSEQ_FEATURE_SIZE.  Limit the feature
+     size to RSEQ_AREA_SIZE_MAX_USED (28) which fits the rseq area in 'struct
+     pthread' and represents the maximum feature size of currently released
+     kernels.  Since no kernels currently cross the 32 bytes of the original
+     ABI, the semantics of a feature size of 32 or more are still undetermined.
+     */
+  _rseq_size = MIN (MAX (auxv_values[AT_RSEQ_FEATURE_SIZE],
+                         RSEQ_AREA_SIZE_INITIAL_USED),
+		    RSEQ_AREA_SIZE_MAX_USED);
+  _rseq_align = MAX (auxv_values[AT_RSEQ_ALIGN], RSEQ_MIN_ALIGN);
+
   DL_PLATFORM_AUXV
 }

sysdeps/unix/sysv/linux/rseq-internal.h

@@ -25,13 +25,31 @@
 #include <stdio.h>
 #include <sys/rseq.h>
 
-/* 32 is the initially required value for the area size.  The
-   actually used rseq size may be less (20 bytes initially).  */
+/* Minimum size of the rseq area allocation required by the syscall.  The
+   actually used rseq feature size may be less (20 bytes initially).  */
 #define RSEQ_AREA_SIZE_INITIAL 32
+
+/* Minimum used feature size of the rseq area.  */
 #define RSEQ_AREA_SIZE_INITIAL_USED 20
 
-/* The variables are in .data.relro but are not yet write-protected.  */
+/* Maximum currently used feature size of the rseq area.  */
+#define RSEQ_AREA_SIZE_MAX_USED 28
+
+/* Minimum alignment of the rseq area.  */
+#define RSEQ_MIN_ALIGN 32
+
+/* Alignment requirement of the rseq area.
+   Populated from the auxiliary vector with a minimum of '32'.
+   In .data.relro but not yet write-protected.  */
+extern size_t _rseq_align attribute_hidden;
+
+/* Size of the active features in the rseq area.
+   Populated from the auxiliary vector with a minimum of '20'.
+   In .data.relro but not yet write-protected.  */
 extern unsigned int _rseq_size attribute_hidden;
+
+/* Offset from the thread pointer to the rseq area.
+   In .data.relro but not yet write-protected.  */
 extern ptrdiff_t _rseq_offset attribute_hidden;
 
 #ifdef RSEQ_SIG

sysdeps/unix/sysv/linux/tst-rseq.c

@@ -38,13 +38,15 @@ static void
 do_rseq_main_test (void)
 {
   struct pthread *pd = THREAD_SELF;
+  size_t rseq_feature_size = MIN (MAX (getauxval (AT_RSEQ_FEATURE_SIZE),
+                                       RSEQ_AREA_SIZE_INITIAL_USED),
+                                  RSEQ_AREA_SIZE_MAX_USED);
 
   TEST_VERIFY_EXIT (rseq_thread_registered ());
   TEST_COMPARE (__rseq_flags, 0);
   TEST_VERIFY ((char *) __thread_pointer () + __rseq_offset
                == (char *) &pd->rseq_area);
-  /* The current implementation only supports the initial size.  */
-  TEST_COMPARE (__rseq_size, 20);
+  TEST_COMPARE (__rseq_size, rseq_feature_size);
 }
 
 static void

sysdeps/unix/sysv/linux/tst-rseq.h

@@ -23,6 +23,7 @@
 #include <syscall.h>
 #include <sys/rseq.h>
 #include <tls.h>
+#include <rseq-internal.h>
 
 static inline bool
 rseq_thread_registered (void)