Re: BUG: spinlock bad magic in tun_do_read

On 2018年05月09日 10:50, Cong Wang wrote:

On Mon, May 7, 2018 at 11:04 PM, Eric Dumazet [off-list ref] wrote:

quoted

On 05/07/2018 10:54 PM, Cong Wang wrote:

quoted

Yeah, we should return early before hitting this uninitialized ptr ring...
Something like:

diff --git a/drivers/net/tun.c b/drivers/net/tun.c
index ef33950a45d9..638c87a95247 100644
--- a/drivers/net/tun.c
+++ b/drivers/net/tun.c

@@ -2128,6 +2128,9 @@ static void *tun_ring_recv(struct tun_file

*tfile, int noblock, int *err)
         void *ptr = NULL;
         int error = 0;

+       if (!tfile->tx_ring.queue)
+               goto out;
+

Or, checking if tun is detached...

tx_ring was properly initialized when first ptr_ring_consume() at line 2131 was attempted.

The bug happens later at line 2143 , after a schedule() call, line 2155

So a single check at function prologue wont solve the case the thread had to sleep,
then some uninit happened.

Very good point. RTNL lock is supposed to protect cleanup path, but I don't
think we can acquire RTNL for tun_chr_read_iter() path...

I think the root cause is we try to initialize ptr ring during TUNSETIFF 
since the length depends on the dev->tx_queue_len and try to destroy it 
when device is gone. We can solve this by initializing a zero size 
ptr_ring during open() and resize if necessary. Then there no need for 
any workaround like memset and checking against NULL.

Let me try to cook a patch for this.

Thanks