Re: [PATCH 2/3] relay: Fix race condition which occurs when readingacross CPUs.

From: Andrea Righi
Date: Thu Jun 12 2008 - 18:58:35 EST


Eduard - Gabriel Munteanu wrote:
@@ -1028,15 +1032,19 @@ static ssize_t relay_file_read_subbufs(struct file *filp, loff_t *ppos,
break;
avail = min(desc->count, avail);
+ /* subbuf_actor may sleep, so release the spinlock for now */
+ spin_unlock_irqrestore(&buf->rw_lock, flags);
ret = subbuf_actor(read_start, buf, avail, desc, actor);
if (desc->error < 0)
break;

If you just break here buf->rw_lock will be unlocked twice. Maybe a
better way is:
if (desc->error < 0)
goto out;
...
} while (desc->count && ret);
spin_unlock_irqrestore(&buf->rw_lock, flags);
out:
mutex_unlock(&filp->f_path.dentry->d_inode->i_mutex);

return desc->written;

+ spin_lock_irqsave(&buf->rw_lock, flags);
if (ret) {
relay_file_read_consume(buf, read_start, ret);
*ppos = relay_file_read_end_pos(buf, read_start, ret);
}
} while (desc->count && ret);
+ spin_unlock_irqrestore(&buf->rw_lock, flags);
mutex_unlock(&filp->f_path.dentry->d_inode->i_mutex);
return desc->written;

-Andrea
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/