mei: ignore client writing state during cb completion
authorAlexander Usyskin <alexander.usyskin@intel.com>
Tue, 1 Apr 2014 20:50:41 +0000 (23:50 +0300)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 16 Apr 2014 19:24:16 +0000 (12:24 -0700)
Ignore client writing state during cb completion to fix a memory
leak.

When moving cbs to the completion list we should not look at
writing_state as this state can be already overwritten by next
write, the fact that a cb is on the write waiting list means
that it was already written to the HW and we can safely complete it.

Same pays for wait in poll handler, we do not have to check the state
wake is done after completion list processing.

Cc: stable@vger.kernel.org
Signed-off-by: Alexander Usyskin <alexander.usyskin@intel.com>
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/misc/mei/interrupt.c
drivers/misc/mei/main.c

index 29b5af8efb71f33ad000578f08493128fef69806..4e3cba6da3f5cccd32d4ee623f91fe5c03655816 100644 (file)
@@ -455,8 +455,7 @@ int mei_irq_write_handler(struct mei_device *dev, struct mei_cl_cb *cmpl_list)
 
                cl->status = 0;
                list_del(&cb->list);
-               if (MEI_WRITING == cl->writing_state &&
-                   cb->fop_type == MEI_FOP_WRITE &&
+               if (cb->fop_type == MEI_FOP_WRITE &&
                    cl != &dev->iamthif_cl) {
                        cl_dbg(dev, cl, "MEI WRITE COMPLETE\n");
                        cl->writing_state = MEI_WRITE_COMPLETE;
index b35594dbf52f49bf034e94254b85d70d0a019756..147413145c97d8a4af71e6a5b555a99cbc23c959 100644 (file)
@@ -644,8 +644,7 @@ static unsigned int mei_poll(struct file *file, poll_table *wait)
                goto out;
        }
 
-       if (MEI_WRITE_COMPLETE == cl->writing_state)
-               mask |= (POLLIN | POLLRDNORM);
+       mask |= (POLLIN | POLLRDNORM);
 
 out:
        mutex_unlock(&dev->device_lock);