Databases
StackOverflow2013
Postsdbo
POThread synchronization issue: possible race, misuse of volatile, cache coherency?
 

Note that there are some explanatory texts on larger screens.

plurals
  1. POThread synchronization issue: possible race, misuse of volatile, cache coherency?
    primarykey
    data
    text
    <p>I am getting acquainted with <strong>pthreads</strong> programming; the following code is a simple producer/consumer design where data is put/fetched from a global list. The problem is: the data pointer in the consumer function is tried to be freed twice. Beside, if I add a <code>printf()</code> at the beginning of the loop, everything seems ok... What am I doing wrong? I suspect a misuse of the <code>volatile</code> keyword, or something hidden by the cache... Unless it's just a design issue (which probably is :p).</p> <p>Thanks for your insights.</p> <p>Note: <code>malloc()/free()</code> is thread-safe on my system. I am compiling with <code>$ gcc -pthread -O0</code> which should, I guess, reduce possible design errors due to misuse of <code>volatile</code>. Finally, we don't care in this code snippet with running out of memory (in case of more producer than consumer).</p> <p><strong>Edit:</strong> Changed code to a single list head.</p> <pre><code>#include &lt;pthread.h&gt; #include &lt;stdlib.h&gt; #include &lt;string.h&gt; pthread_mutex_t lock; pthread_cond_t new_data; struct data { int i; struct data *next; }; struct data *list_head = NULL; void *consume(void *arg) { struct data *data; while (1) { pthread_mutex_lock(&amp;lock); while (list_head == NULL) { pthread_cond_wait(&amp;new_data, &amp;lock); } data = list_head; list_head = list_head-&gt;next; pthread_mutex_unlock(&amp;lock); free(data); } return NULL; } void *produce(void *arg) { struct data *data; while (1) { data = malloc(sizeof(struct data)); pthread_mutex_lock(&amp;lock); data-&gt;next = list_head; list_head = data; pthread_mutex_unlock(&amp;lock); pthread_cond_signal(&amp;new_data); } return NULL; } int main() { pthread_t tid[2]; int i; pthread_mutex_init(&amp;lock, NULL); pthread_cond_init(&amp;new_data, NULL); pthread_create(&amp;tid[0], NULL, consume, NULL); pthread_create(&amp;tid[1], NULL, produce, NULL); for (i = 0; i &lt; 2; i++) { pthread_join(tid[i], NULL); } } </code></pre> <p>And the output:</p> <pre><code>$ ./a.out *** glibc detected *** ./a.out: double free or corruption (fasttop): 0x00007f5870109000 *** </code></pre>
    singulars
    1. PO
      singulars
      1. PTAnswer
    1. This table or related slice is empty.
    1. PTQuestion
    1. USBenoit
    1. USBenoit
    plurals
    1. PL
      singulars
      1. LTLinked
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. PO
      singulars
      1. PTAnswer
    2. PO
      singulars
      1. PTAnswer
    3. PO
      singulars
      1. PTAnswer
    1. VO
      singulars
      1. POThread synchronization issue: possible race, misuse of volatile, cache coherency?
      1. This table or related slice is empty.
      1. VTUpMod
    2. VO
      singulars
      1. POThread synchronization issue: possible race, misuse of volatile, cache coherency?
      1. This table or related slice is empty.
      1. VTUpMod
    3. VO
      singulars
      1. POThread synchronization issue: possible race, misuse of volatile, cache coherency?
      1. This table or related slice is empty.
      1. VTUpMod
    1. COYou should remove all use of `volatile` in this code. The lock generates the correct memory barriers and such for your platform -- `volatile` is for communication with memory mapped devices and such, not for synchronization. `volatile` prevents reordering or read/write removal by the compiler, but not by the CPU.
      singulars
      1. POThread synchronization issue: possible race, misuse of volatile, cache coherency?
      1. USBilly ONeal
    2. COAs an aside, I'd also swap the `pthread_cond_signal` and `pthread_mutex_unlock`.
      singulars
      1. POThread synchronization issue: possible race, misuse of volatile, cache coherency?
      1. USuser7116
    3. CO@sixlettervariables I have seen both permutations on different code samples, I can't figure out what's the best practice... I guess that's because the lock is still hold and even though the waiting thread is released from waiting, it still won't be able to get the lock since it has not been released yet?
      singulars
      1. POThread synchronization issue: possible race, misuse of volatile, cache coherency?
      1. USBenoit
 

Querying!

 
Guidance
A row detail

Detail views are divided into sections. All the information in the data section comes from columns in the selected row. The other sections display data from other, related rows.

Related data can be related in a to-one or a to-many fashion. Captions of data related in a to-many fashion link to a list view showing a filtered view of the table.

Try moving around until you find a non-empty to-many entry and click on the label to get to one. You can move back to the root by clicking on the database name in the header.

SQuiL has stopped working due to an internal error.

If you are curious you may find further information in the browser console, which is accessible through the devtools (F12).

Reload