Re: [PATCH net] net: stmmac: Disable Tx queues when reconfiguring the interface

On Wed, Nov 24, 2021 at 01:10:55PM +0100, Yannick Vignon wrote:

On 11/24/2021 11:06 AM, Vladimir Oltean wrote:

quoted

On Tue, Nov 23, 2021 at 08:07:51PM -0800, Jakub Kicinski wrote:

quoted

On Tue, 23 Nov 2021 19:54:48 +0100 Yannick Vignon wrote:

quoted

From: Yannick Vignon <redacted>

The Tx queues were not disabled in cases where the driver needed to stop
the interface to apply a new configuration. This could result in a kernel
panic when doing any of the 3 following actions:
* reconfiguring the number of queues (ethtool -L)
* reconfiguring the size of the ring buffers (ethtool -G)
* installing/removing an XDP program (ip l set dev ethX xdp)

Prevent the panic by making sure netif_tx_disable is called when stopping
an interface.

Without this patch, the following kernel panic can be observed when loading
an XDP program:

Unable to handle kernel paging request at virtual address ffff80001238d040
[....]
  Call trace:
   dwmac4_set_addr+0x8/0x10
   dev_hard_start_xmit+0xe4/0x1ac
   sch_direct_xmit+0xe8/0x39c
   __dev_queue_xmit+0x3ec/0xaf0
   dev_queue_xmit+0x14/0x20
[...]
[ end trace 0000000000000002 ]---

Fixes: 78cb988d36b6 ("net: stmmac: Add initial XDP support")
Signed-off-by: Yannick Vignon <redacted>

Fixes tag: Fixes: 78cb988d36b6 ("net: stmmac: Add initial XDP support")
Has these problem(s):
	- Target SHA1 does not exist

You caught him backporting! Although I agree, things sent to the "net"
tree should also be tested against the "net" tree.

That would be more like forward-porting in this case, since I first fixed
the issue on an older 5.10 kernel :)

XDP for stmmac isn't in v5.10, so backporting is what it is.