Drivers should be able to call cfg80211_sched_scan_results() from atomic
context. However, with the introduction of multiple scheduled scan feature
this requirement was not taken into account resulting in regression shown
below.
[ 119.021594] BUG: scheduling while atomic: irq/47-iwlwifi/517/0x00000200
[ 119.021604] Modules linked in: xt_tcpudp ip6t_rpfilter ipt_REJECT nf_reject_ipv4 ip6t_REJECT nf_reject_ipv6 nf_conntrack_ipv6 nf_defrag_ipv6 xt_conntrack ip_set nfnetlink ebtable_nat ebtable_broute bridge stp llc ip6table_raw ip6table_security ip6table_mangle iptable_raw iptable_nat nf_conntrack_ipv4 nf_defrag_ipv4 nf_nat_ipv4 nf_nat nf_conntrack iptable_security iptable_mangle ebtable_filter ebtables ip6table_filter ip6_tables iptable_filter ip_tables x_tables rfcomm bnep binfmt_misc arc4 iTCO_wdt iTCO_vendor_support uvcvideo videobuf2_vmalloc videobuf2_memops videobuf2_v4l2 videobuf2_core videodev intel_rapl cdc_mbim iwlmvm x86_pkg_temp_thermal intel_powerclamp mac80211 media cdc_wdm btusb coretemp cdc_ncm kvm_intel usbnet mii cdc_acm iwlwifi kvm btintel joydev pcspkr serio_raw cfg80211 snd_hda_codec_hdmi
[ 119.021701] bluetooth lpc_ich snd_hda_codec_realtek snd_hda_codec_generic shpchp sg ecdh_generic snd_hda_intel thinkpad_acpi snd_hda_codec snd_hwdep snd_hda_core snd_pcm snd_timer nvram snd soundcore evdev tpm_tis tpm_tis_core tpm algif_skcipher af_alg crct10dif_pclmul crc32_pclmul crc32c_intel ghash_clmulni_intel rtsx_pci_sdmmc mmc_core aesni_intel aes_x86_64 crypto_simd cryptd glue_helper psmouse i2c_i801 sd_mod ehci_pci ehci_hcd e1000e rtsx_pci mfd_core ptp xhci_pci pps_core xhci_hcd
[ 119.021759] CPU: 1 PID: 517 Comm: irq/47-iwlwifi Not tainted 4.12.0-rc2-t440s-20170522+ #1
[ 119.021763] Hardware name: LENOVO 20AQS03H00/20AQS03H00, BIOS GJET91WW (2.41 ) 09/21/2016
[ 119.021766] Call Trace:
[ 119.021778] ? dump_stack+0x5c/0x84
[ 119.021784] ? __schedule_bug+0x4c/0x70
[ 119.021792] ? __schedule+0x496/0x5c0
[ 119.021798] ? schedule+0x2d/0x80
[ 119.021804] ? schedule_preempt_disabled+0x5/0x10
[ 119.021810] ? __mutex_lock.isra.0+0x18e/0x4c0
[ 119.021817] ? __wake_up+0x2f/0x50
[ 119.021833] ? cfg80211_sched_scan_results+0x19/0x60 [cfg80211]
[ 119.021844] ? cfg80211_sched_scan_results+0x19/0x60 [cfg80211]
[ 119.021859] ? iwl_mvm_rx_lmac_scan_iter_complete_notif+0x17/0x30 [iwlmvm]
[ 119.021869] ? iwl_pcie_rx_handle+0x2a9/0x7e0 [iwlwifi]
[ 119.021878] ? iwl_pcie_irq_handler+0x17c/0x730 [iwlwifi]
[ 119.021884] ? irq_forced_thread_fn+0x60/0x60
[ 119.021887] ? irq_thread_fn+0x16/0x40
[ 119.021892] ? irq_thread+0x109/0x180
[ 119.021896] ? wake_threads_waitq+0x30/0x30
[ 119.021901] ? kthread+0xf2/0x130
[ 119.021905] ? irq_thread_dtor+0x90/0x90
[ 119.021910] ? kthread_create_on_node+0x40/0x40
[ 119.021915] ? ret_from_fork+0x26/0x40
Fixes: b34939b98369 ("cfg80211: add request id to cfg80211_sched_scan_*() api")
Reported-by: Sander Eikelenboom <[email protected]>
Signed-off-by: Arend van Spriel <[email protected]>
---
net/wireless/scan.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/net/wireless/scan.c b/net/wireless/scan.c
index 14d5f0c..9f0901f 100644
--- a/net/wireless/scan.c
+++ b/net/wireless/scan.c
@@ -322,9 +322,9 @@ static void cfg80211_del_sched_scan_req(struct cfg80211_registered_device *rdev,
{
struct cfg80211_sched_scan_request *pos;
- ASSERT_RTNL();
+ WARN_ON_ONCE(!rcu_read_lock_held() && !lockdep_rtnl_is_held());
- list_for_each_entry(pos, &rdev->sched_scan_req_list, list) {
+ list_for_each_entry_rcu(pos, &rdev->sched_scan_req_list, list) {
if (pos->reqid == reqid)
return pos;
}
@@ -398,13 +398,13 @@ void cfg80211_sched_scan_results(struct wiphy *wiphy, u64 reqid)
trace_cfg80211_sched_scan_results(wiphy, reqid);
/* ignore if we're not scanning */
- rtnl_lock();
+ rcu_read_lock();
request = cfg80211_find_sched_scan_req(rdev, reqid);
if (request) {
request->report_results = true;
queue_work(cfg80211_wq, &rdev->sched_scan_res_wk);
}
- rtnl_unlock();
+ rcu_read_unlock();
}
EXPORT_SYMBOL(cfg80211_sched_scan_results);
--
1.9.1
On Tue, 2017-05-23 at 09:58 +0100, Arend van Spriel wrote:
> Drivers should be able to call cfg80211_sched_scan_results() from
> atomic
> context. However, with the introduction of multiple scheduled scan
> feature
> this requirement was not taken into account resulting in regression
> shown
> below.
Applied, thanks.
I did strip out the Modules linked in: ... garbage though :)
johannes