Re: [PATCH v5] vnc: don't mess up with iohandlers in the vnc thread

[Peter Lieven <pl@dlh.net>]

On 14.03.2011 10:19, Corentin Chary wrote:
> On Thu, Mar 10, 2011 at 3:13 PM, Corentin Chary
> <corentin.chary@gmail.com>  wrote:
>> The threaded VNC servers messed up with QEMU fd handlers without
>> any kind of locking, and that can cause some nasty race conditions.
>>
>> Using qemu_mutex_lock_iothread() won't work because vnc_dpy_cpy(),
>> which will wait for the current job queue to finish, can be called with
>> the iothread lock held.
>>
>> Instead, we now store the data in a temporary buffer, and use a bottom
>> half to notify the main thread that new data is available.
>>
>> vnc_[un]lock_ouput() is still needed to access VncState members like
>> abort, csock or jobs_buffer.
>>
>> Signed-off-by: Corentin Chary<corentin.chary@gmail.com>
>> ---
>>   ui/vnc-jobs-async.c |   48 +++++++++++++++++++++++++++++-------------------
>>   ui/vnc-jobs.h       |    1 +
>>   ui/vnc.c            |   12 ++++++++++++
>>   ui/vnc.h            |    2 ++
>>   4 files changed, 44 insertions(+), 19 deletions(-)
>>
>> diff --git a/ui/vnc-jobs-async.c b/ui/vnc-jobs-async.c
>> index f596247..4438776 100644
>> --- a/ui/vnc-jobs-async.c
>> +++ b/ui/vnc-jobs-async.c
>> @@ -28,6 +28,7 @@
>>
>>   #include "vnc.h"
>>   #include "vnc-jobs.h"
>> +#include "qemu_socket.h"
>>
>>   /*
>>   * Locking:
>> @@ -155,6 +156,24 @@ void vnc_jobs_join(VncState *vs)
>>          qemu_cond_wait(&queue->cond,&queue->mutex);
>>      }
>>      vnc_unlock_queue(queue);
>> +    vnc_jobs_consume_buffer(vs);
>> +}
>> +
>> +void vnc_jobs_consume_buffer(VncState *vs)
>> +{
>> +    bool flush;
>> +
>> +    vnc_lock_output(vs);
>> +    if (vs->jobs_buffer.offset) {
>> +        vnc_write(vs, vs->jobs_buffer.buffer, vs->jobs_buffer.offset);
>> +        buffer_reset(&vs->jobs_buffer);
>> +    }
>> +    flush = vs->csock != -1&&  vs->abort != true;
>> +    vnc_unlock_output(vs);
>> +
>> +    if (flush) {
>> +      vnc_flush(vs);
>> +    }
>>   }
>>
>>   /*
>> @@ -197,7 +216,6 @@ static int vnc_worker_thread_loop(VncJobQueue *queue)
>>      VncState vs;
>>      int n_rectangles;
>>      int saved_offset;
>> -    bool flush;
>>
>>      vnc_lock_queue(queue);
>>      while (QTAILQ_EMPTY(&queue->jobs)&&  !queue->exit) {
>> @@ -213,6 +231,7 @@ static int vnc_worker_thread_loop(VncJobQueue *queue)
>>
>>      vnc_lock_output(job->vs);
>>      if (job->vs->csock == -1 || job->vs->abort == true) {
>> +        vnc_unlock_output(job->vs);
>>          goto disconnected;
>>      }
>>      vnc_unlock_output(job->vs);
>> @@ -233,10 +252,6 @@ static int vnc_worker_thread_loop(VncJobQueue *queue)
>>
>>          if (job->vs->csock == -1) {
>>              vnc_unlock_display(job->vs->vd);
>> -            /* output mutex must be locked before going to
>> -             * disconnected:
>> -             */
>> -            vnc_lock_output(job->vs);
>>              goto disconnected;
>>          }
>>
>> @@ -254,24 +269,19 @@ static int vnc_worker_thread_loop(VncJobQueue *queue)
>>      vs.output.buffer[saved_offset] = (n_rectangles>>  8)&  0xFF;
>>      vs.output.buffer[saved_offset + 1] = n_rectangles&  0xFF;
>>
>> -    /* Switch back buffers */
>>      vnc_lock_output(job->vs);
>> -    if (job->vs->csock == -1) {
>> -        goto disconnected;
>> +    if (job->vs->csock != -1) {
>> +        buffer_reserve(&job->vs->jobs_buffer, vs.output.offset);
>> +        buffer_append(&job->vs->jobs_buffer, vs.output.buffer,
>> +                      vs.output.offset);
>> +        /* Copy persistent encoding data */
>> +        vnc_async_encoding_end(job->vs,&vs);
>> +
>> +       qemu_bh_schedule(job->vs->bh);
>>      }
>> -
>> -    vnc_write(job->vs, vs.output.buffer, vs.output.offset);
>> -
>> -disconnected:
>> -    /* Copy persistent encoding data */
>> -    vnc_async_encoding_end(job->vs,&vs);
>> -    flush = (job->vs->csock != -1&&  job->vs->abort != true);
>>      vnc_unlock_output(job->vs);
>>
>> -    if (flush) {
>> -        vnc_flush(job->vs);
>> -    }
>> -
>> +disconnected:
>>      vnc_lock_queue(queue);
>>      QTAILQ_REMOVE(&queue->jobs, job, next);
>>      vnc_unlock_queue(queue);
>> diff --git a/ui/vnc-jobs.h b/ui/vnc-jobs.h
>> index b8dab81..4c661f9 100644
>> --- a/ui/vnc-jobs.h
>> +++ b/ui/vnc-jobs.h
>> @@ -40,6 +40,7 @@ void vnc_jobs_join(VncState *vs);
>>
>>   #ifdef CONFIG_VNC_THREAD
>>
>> +void vnc_jobs_consume_buffer(VncState *vs);
>>   void vnc_start_worker_thread(void);
>>   bool vnc_worker_thread_running(void);
>>   void vnc_stop_worker_thread(void);
>> diff --git a/ui/vnc.c b/ui/vnc.c
>> index 610f884..f28873b 100644
>> --- a/ui/vnc.c
>> +++ b/ui/vnc.c
>> @@ -1011,7 +1011,10 @@ static void vnc_disconnect_finish(VncState *vs)
>>
>>   #ifdef CONFIG_VNC_THREAD
>>      qemu_mutex_destroy(&vs->output_mutex);
>> +    qemu_bh_delete(vs->bh);
>> +    buffer_free(&vs->jobs_buffer);
>>   #endif
>> +
>>      for (i = 0; i<  VNC_STAT_ROWS; ++i) {
>>          qemu_free(vs->lossy_rect[i]);
>>      }
>> @@ -1226,6 +1229,14 @@ static long vnc_client_read_plain(VncState *vs)
>>      return ret;
>>   }
>>
>> +#ifdef CONFIG_VNC_THREAD
>> +static void vnc_jobs_bh(void *opaque)
>> +{
>> +    VncState *vs = opaque;
>> +
>> +    vnc_jobs_consume_buffer(vs);
>> +}
>> +#endif
>>
>>   /*
>>   * First function called whenever there is more data to be read from
>> @@ -2525,6 +2536,7 @@ static void vnc_connect(VncDisplay *vd, int csock)
>>
>>   #ifdef CONFIG_VNC_THREAD
>>      qemu_mutex_init(&vs->output_mutex);
>> +    vs->bh = qemu_bh_new(vnc_jobs_bh, vs);
>>   #endif
>>
>>      QTAILQ_INSERT_HEAD(&vd->clients, vs, next);
>> diff --git a/ui/vnc.h b/ui/vnc.h
>> index 8a1e7b9..bca5f87 100644
>> --- a/ui/vnc.h
>> +++ b/ui/vnc.h
>> @@ -283,6 +283,8 @@ struct VncState
>>      VncJob job;
>>   #else
>>      QemuMutex output_mutex;
>> +    QEMUBH *bh;
>> +    Buffer jobs_buffer;
>>   #endif
>>
>>      /* Encoding specific, if you add something here, don't forget to
>> --
>> 1.7.3.4
>>
>>
> Paolo, Anthony, Jan, are you ok with that one ?
> Peter Lieve, could you test that patch ?
i have seen one segfault. unfortunatly, i had no debugger attached.

but whats equally worse during stress test, i managed to trigger oom-killer.
it seems we have a memory leak somewhere....

peter


> Thanks
>