commit-hurd
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] tst-cancel4: Make blocking on write more portable


From: Adhemerval Zanella
Subject: Re: [PATCH] tst-cancel4: Make blocking on write more portable
Date: Tue, 7 Jul 2020 18:08:31 -0300
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101 Thunderbird/68.8.0

On 05/07/2020 15:11, Samuel Thibault wrote:
> diff --git a/nptl/tst-cancel4-common.h b/nptl/tst-cancel4-common.h
> index c8763cacba..f99785c4f2 100644
> --- a/nptl/tst-cancel4-common.h
> +++ b/nptl/tst-cancel4-common.h
> @@ -62,7 +62,7 @@ static pthread_barrier_t b2;
>  
>  #define WRITE_BUFFER_SIZE 16384
>  
> -/* Set the send buffer of socket S to 1 byte so any send operation
> +/* Set the send buffer of socket S to 1 byte and fill it so any send 
> operation
>     done with WRITE_BUFFER_SIZE bytes will force syscall blocking.  */
>  static void
>  set_socket_buffer (int s)
> @@ -74,6 +74,23 @@ set_socket_buffer (int s)
>                   sizeof (val)) == 0);
>    TEST_VERIFY_EXIT (getsockopt (s, SOL_SOCKET, SO_SNDBUF, &val, &len) == 0);
>    TEST_VERIFY_EXIT (val < WRITE_BUFFER_SIZE);
> +
> +  char buf = '\0';
> +  struct timeval timeout = { .tv_sec = 0, .tv_usec = 0 };
> +
> +  while (1)
> +    {
> +      fd_set set;
> +      int ret;
> +
> +      FD_ZERO (&set);
> +      FD_SET (s, &set);
> +      ret = select (s + 1, NULL, &set, NULL, &timeout);
> +      if (!ret)
> +     break;
> +
> +      write (s, &buf, 1);
> +    }
>  }

Unfortunately this does not do what you are intending to do on Linux and 
the tests only pass because of BZ#12683 where cancellation does not act on 
write with side-effects.

For instance on kernel 5.4.0-40-generic the syscall sequence is:

[...]
select(4, NULL, [3], NULL, {tv_sec=0, tv_usec=0}) = 1 (out [3], left {tv_sec=0, 
tv_usec=0})
write(3, "\0", 1)                       = 1
select(4, NULL, [3], NULL, {tv_sec=0, tv_usec=0}) = 1 (out [3], left {tv_sec=0, 
tv_usec=0})
write(3, "\0", 1)                       = 1
select(4, NULL, [3], NULL, {tv_sec=0, tv_usec=0}) = 0 (Timeout)
[...]
clone(child_stack=0x7fd3ee4f4fb0, 
flags=CLONE_VM|CLONE_FS|CLONE_FILES|CLONE_SIGHAND|CLONE_THREAD|CLONE_SYSVSEM|CLONE_SETTLS|CLONE_PARENT_SETTID|CLONE_CHILD_CLEARTIDstrace:
 Process 2641554 attached
, parent_tid=[2641554], tls=0x7fd3ee4f5700, child_tidptr=0x7fd3ee4f59d0) = 
2641554
[...]
[pid 2641554] write(3, 
"\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 16384 
<unfinished ...>
[...]
[pid 2641553] tgkill(2641553, 2641554, SIGRTMIN) = 0
[pid 2641554] write(1, "write (3, buf, 16384) returned 2"..., 40write (3, buf, 
16384) returned 2240 (0)
[...]

So it only fills 2 bytes on the socket buffer, but later the write *does*
return with side-effects indicating that it has at least written 2240 bytes.

I haven't investigate further on Linux kernel the iteration between
SOL_SOCKET, buffer filling and blocking operations.  But it does show that
is indeed system specific and partial write might still occur even with
clever tricks.  I am also not sure if this is something specific to 
socketpair and if using a named socket might have different results.  

Are you trying to make the tests behave better on Hurd? If so, in how
exactly Hurd behaves different than Linux in this case?



reply via email to

[Prev in Thread] Current Thread [Next in Thread]