Re: [PATCH] i386: Use pthread_barrier for synchronization on tst-bz21269

[Adhemerval Zanella Netto <adhemerval.zanella@linaro.org>]

On 20/02/23 19:55, DJ Delorie wrote:
> 
> The resulting test has two calls to pthread_barrier_wait() in the
> subthread but three calls in main().  These will quickly get out of
> sync.
> 
> 
> Adhemerval Zanella via Libc-alpha <libc-alpha@sourceware.org> writes:
>> -   - C11 atomics instead of plain access.
>> +   - Use pthread_barrier instead of atomic and futexes.
> 
> Is this true relative to the original testcase?  Still, merely a
> comment, so OK.

The original refers to the kernel tests in here.

> 
>> -#include <stdatomic.h>
>> -
>>  #include <asm/ldt.h>
>> -#include <linux/futex.h>
> 
> Ok.
> 
>> +#include <support/xsignal.h>
> 
> Ok.
> 
>> -static int
>> -futex (int *uaddr, int futex_op, int val, void *timeout, int *uaddr2,
>> -	int val3)
>> -{
>> -  return syscall (SYS_futex, uaddr, futex_op, val, timeout, uaddr2, val3);
>> -}
> 
> We remove all calls to futex, so no longer need this.  Ok.
> 
>> -  TEST_VERIFY_EXIT (sigaction (sig, &sa, 0) == 0);
>> +  xsigaction (sig, &sa, 0);
> 
> Ok.
> 
>> -/* Possible values of futex:
>> -   0: thread is idle.
>> -   1: thread armed.
>> -   2: thread should clear LDT entry 0.
>> -   3: thread should exit.  */
>> -static atomic_uint ftx;
> 
> Ok.
> 
>> +static pthread_barrier_t barrier;
> 
> New, ok.
> 
>>  static void *
>>  threadproc (void *ctx)
>>  {
>> -  while (1)
>> +  for (int i = 0; i < 5; i++)
> 
> This matches the loop in main.  Ok.  In the future, a #define loop limit
> would be appropriate, to prevent these getting out of sync.  Or a
> comment that it has to match main().

Ack.

> 
>>      {
>> -      futex ((int *) &ftx, FUTEX_WAIT, 1, NULL, NULL, 0);
>> -      while (atomic_load (&ftx) != 2)
>> -	{
>> -	  if (atomic_load (&ftx) >= 3)
>> -	    return NULL;
>> -	}
>> +      xpthread_barrier_wait (&barrier);
> 
> First barrier, ok.
> 
>>        /* clear LDT entry 0.  */
>>        const struct user_desc desc = { 0 };
>>        xmodify_ldt (1, &desc, sizeof (desc));
> 
> Leave the stuff the thread is actually testing ;-)
> 
>> -      /* If ftx == 2, set it to zero,  If ftx == 100, quit.  */
>> -      if (atomic_fetch_add (&ftx, -2) != 2)
>> -	return NULL;
>> +      /* Wait for 'ss' set in main thread.  */
>> +      xpthread_barrier_wait (&barrier);
> 
> Second barrier, ok.
> 
>>      }
>> +
>> +  return NULL;
> 
> Ok, moved from above.
> 
>>  }
>>  
>>  
>> @@ -180,6 +162,8 @@ do_test (void)
>>    /* Some kernels send SIGBUS instead.  */
>>    xsethandler (SIGBUS, sigsegv_handler, 0);
>>  
>> +  xpthread_barrier_init (&barrier, NULL, 2);
> 
> Initialize; must have two callers.  Ok - main and thread both call.
> 
>>    thread = xpthread_create (0, threadproc, 0);
>>  
>>    asm volatile ("mov %%ss, %0" : "=rm" (orig_ss));
>> @@ -190,8 +174,7 @@ do_test (void)
>>  	continue;
>>  
>>        /* Make sure the thread is ready after the last test. */
>> -      while (atomic_load (&ftx) != 0)
>> -	;
>> +      xpthread_barrier_wait (&barrier);
> 
> First barrier, ok.
> 
>>        struct user_desc desc = {
>>  	.entry_number       = 0,
>> @@ -207,28 +190,21 @@ do_test (void)
>>  
>>        xmodify_ldt (0x11, &desc, sizeof (desc));
>>  
>> -      /* Arm the thread.  */
>> -      ftx = 1;
>> -      futex ((int*) &ftx, FUTEX_WAKE, 0, NULL, NULL, 0);
>> +      xpthread_barrier_wait (&barrier);
> 
> Second barrier, ok.
>   
>>        asm volatile ("mov %0, %%ss" : : "r" (0x7));
>>  
>> -      /* Fire up thread modify_ldt call.  */
>> -      atomic_store (&ftx, 2);
>> -
>> -      while (atomic_load (&ftx) != 0)
>> -	;
>> +      xpthread_barrier_wait (&barrier);
> 
> Third barrier?  This puts main and the thread out of sync.

Indeed, this is wrong. I will remove it.

> 
>>        /* On success, modify_ldt will segfault us synchronously and we will
>>  	 escape via siglongjmp.  */
>>        support_record_failure ();
>>      }
>>  
>> -  atomic_store (&ftx, 100);
>> -  futex ((int*) &ftx, FUTEX_WAKE, 0, NULL, NULL, 0);
>> -
> 
> Ok.
> 
>>    xpthread_join (thread);
>>  
>> +  xpthread_barrier_destroy (&barrier);
>> +
> 
> Ok.
>