A sysfs reader can race with a device reset or removal, attempting to
read device state when the device is not actually present.

This is the same sort of panic as observed in commit 4224cfd7fb65
("net-sysfs: add check for netdevice being present to speed_show"):

     [exception RIP: qed_get_current_link+17]
  #8 [ffffb9e4f2907c48] qede_get_link_ksettings at ffffffffc07a994a [qede]
  #9 [ffffb9e4f2907cd8] __rh_call_get_link_ksettings at ffffffff992b01a3
 #10 [ffffb9e4f2907d38] __ethtool_get_link_ksettings at ffffffff992b04e4
 #11 [ffffb9e4f2907d90] duplex_show at ffffffff99260300
 #12 [ffffb9e4f2907e38] dev_attr_show at ffffffff9905a01c
 #13 [ffffb9e4f2907e50] sysfs_kf_seq_show at ffffffff98e0145b
 #14 [ffffb9e4f2907e68] seq_read at ffffffff98d902e3
 #15 [ffffb9e4f2907ec8] vfs_read at ffffffff98d657d1
 #16 [ffffb9e4f2907f00] ksys_read at ffffffff98d65c3f
 #17 [ffffb9e4f2907f38] do_syscall_64 at ffffffff98a052fb

 crash> struct net_device.state ffff9a9d21336000
   state = 5,

state 5 is __LINK_STATE_START (0b1) and __LINK_STATE_NOCARRIER (0b100).
The device is not present, note lack of __LINK_STATE_PRESENT (0b10).

Resolve by adding the same netif_device_present() check to duplex_show.

Fixes: d519e17e2d01 ("net: export device speed and duplex via sysfs")
Signed-off-by: Jamie Bainbridge <jamie.bainbridge@gmail.com>
---
v2: Restrict patch to just required path and describe problem in more
    detail as suggested by Johannes Berg. Improve commit message format
    as suggested by Shigeru Yoshida.
v3: Use earlier Fixes commit as suggested by Paolo Abeni.
v4: Typo as suggested by Simon Horman.
---
 net/core/net-sysfs.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/core/net-sysfs.c b/net/core/net-sysfs.c
index 0e2084ce7b7572bff458ed7e02358d9258c74628..22801d165d852a6578ca625b9674090519937be5 100644
--- a/net/core/net-sysfs.c
+++ b/net/core/net-sysfs.c
@@ -261,7 +261,7 @@ static ssize_t duplex_show(struct device *dev,
 	if (!rtnl_trylock())
 		return restart_syscall();
 
-	if (netif_running(netdev)) {
+	if (netif_running(netdev) && netif_device_present(netdev)) {
 		struct ethtool_link_ksettings cmd;
 
 		if (!__ethtool_get_link_ksettings(netdev, &cmd)) {
-- 
2.39.2

On Tue,  6 Aug 2024 16:35:27 +1000 Jamie Bainbridge wrote:
> A sysfs reader can race with a device reset or removal, attempting to
> read device state when the device is not actually present.

True, but..

> -	if (netif_running(netdev)) {
> +	if (netif_running(netdev) && netif_device_present(netdev)) {
>  		struct ethtool_link_ksettings cmd;
>  
>  		if (!__ethtool_get_link_ksettings(netdev, &cmd)) {

..there are more callers of __ethtool_get_link_ksettings() and only 
a fraction of them have something resembling a presence check in
their path. Can we put the check inside __ethtool_get_link_ksettings()
itself?

On Fri, 9 Aug 2024 at 01:10, Jakub Kicinski <kuba@kernel.org> wrote:
>
> On Tue,  6 Aug 2024 16:35:27 +1000 Jamie Bainbridge wrote:
> > A sysfs reader can race with a device reset or removal, attempting to
> > read device state when the device is not actually present.
>
> True, but..
>
> > -     if (netif_running(netdev)) {
> > +     if (netif_running(netdev) && netif_device_present(netdev)) {
> >               struct ethtool_link_ksettings cmd;
> >
> >               if (!__ethtool_get_link_ksettings(netdev, &cmd)) {
>
> ..there are more callers of __ethtool_get_link_ksettings() and only
> a fraction of them have something resembling a presence check in
> their path. Can we put the check inside __ethtool_get_link_ksettings()
> itself?

No worries. iiuc that would also mean reverting commit 4224cfd7fb65
("net-sysfs: add check for netdevice being present to speed_show")
because the check is being centralised. Should I do that in the same
patch, or a separate revert patch?