| Message ID | 20220906161410.2106782-2-oleksandr.savchenko@celeno.com |
|---|---|
| State | Changes Requested |
| Headers | show |
| Series | nl80211: Process RTM event per bss | expand |
On Tue, Sep 06, 2022 at 07:14:05PM +0300, Alexander Savchenko wrote: > This reverts commit 8a9950f7b6ebd4f555575f60d7ae7a98cdb73383. > > Commit 8a9950f7b breaks the virtual AP mode due to non-main interfaces > can't receive UP state. Could you please provide more detail on what exactly is broken and how that could be reproduced in a test? I don't like the idea of just reverting earlier commits (patches 1 and 2) without clear description of how the items fixed in those earlier commits are addressed. Patch 2 is reverting a commit from seven years ago which seems to imply that there has not really been that significant an issue if this comes up only now..
Hi,
Shortly, broken processing RTM netlink events in setup with several bss.
Test steps:
for example we have 2 bss for on one band: wlan2_0, wlan1_0.
1.1) initial state
iw dev
phy#0
Interface wlan2_1
ifindex 13
wdev 0x2
addr 02:1c:51:7d:98:c0
ssid D5G2
type AP
channel 36 (5180 MHz), width: 160 MHz, center1: 5250 MHz
txpower 19.00 dBm
Interface wlan2_0
ifindex 11
wdev 0x1
addr 00:1c:51:7d:98:c2
ssid D5G
type AP
channel 36 (5180 MHz), width: 160 MHz, center1: 5250 MHz
txpower 19.00 dBm
1.2) ip link
11: wlan2_0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq master br0 state UP qlen 1000
link/ether 00:1c:51:7d:98:c2 brd ff:ff:ff:ff:ff:ff
13: wlan2_1: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc mq state DOWN qlen 1000
link/ether 02:1c:51:7d:98:c0 brd ff:ff:ff:ff:ff:ff
2) ifconfig wlan2_1 down
ifconfig wlan2_1 up
2.1) iw dev
phy#0
Interface wlan2_1
ifindex 13
wdev 0x2
addr 02:1c:51:7d:98:c0
type AP
txpower 19.00 dBm
Interface wlan2_0
ifindex 11
wdev 0x1
addr 00:1c:51:7d:98:c2
ssid D5G
type AP
channel 36 (5180 MHz), width: 160 MHz, center1: 5250 MHz
txpower 19.00 dBm
2.2) ip link
11: wlan2_0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq master br0 state UP qlen 1000
link/ether 00:1c:51:7d:98:c2 brd ff:ff:ff:ff:ff:ff
13: wlan2_1: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc mq state DOWN qlen 1000
link/ether 02:1c:51:7d:98:c0 brd ff:ff:ff:ff:ff:ff
You could see that wlan2_1 fell to broken state NO-CARRIER and absent LOWER_UP.
hwsim test for such scenario implemented in patch
`[v2,6/6] tests: hwsim: ap_dynamic: Added test case for VAP restart`
Regards,
Alexander
diff --git a/src/drivers/driver_nl80211.c b/src/drivers/driver_nl80211.c index a4b2d5219..89e9cd112 100644 --- a/src/drivers/driver_nl80211.c +++ b/src/drivers/driver_nl80211.c @@ -1283,20 +1283,11 @@ static void wpa_driver_nl80211_event_rtm_newlink(void *ctx, } if (drv->if_disabled && (ifi->ifi_flags & IFF_UP)) { - namebuf[0] = '\0'; if (if_indextoname(ifi->ifi_index, namebuf) && linux_iface_up(drv->global->ioctl_sock, namebuf) == 0) { wpa_printf(MSG_DEBUG, "nl80211: Ignore interface up " "event since interface %s is down", namebuf); - return; - } - wpa_printf(MSG_DEBUG, "nl80211: Interface up (%s/%s)", - namebuf, ifname); - if (os_strcmp(drv->first_bss->ifname, ifname) != 0) { - wpa_printf(MSG_DEBUG, - "nl80211: Not the main interface (%s) - do not indicate interface up", - drv->first_bss->ifname); } else if (if_nametoindex(drv->first_bss->ifname) == 0) { wpa_printf(MSG_DEBUG, "nl80211: Ignore interface up " "event since interface %s does not exist", @@ -1309,6 +1300,7 @@ static void wpa_driver_nl80211_event_rtm_newlink(void *ctx, /* Re-read MAC address as it may have changed */ nl80211_refresh_mac(drv, ifi->ifi_index, 0); + wpa_printf(MSG_DEBUG, "nl80211: Interface up"); drv->if_disabled = 0; wpa_supplicant_event(drv->ctx, EVENT_INTERFACE_ENABLED, NULL);
This reverts commit 8a9950f7b6ebd4f555575f60d7ae7a98cdb73383. Commit 8a9950f7b breaks the virtual AP mode due to non-main interfaces can't receive UP state. Signed-off-by: Alexander Savchenko <oleksandr.savchenko@celeno.com> --- src/drivers/driver_nl80211.c | 10 +--------- 1 file changed, 1 insertion(+), 9 deletions(-)