A-Team_Digital_Solutions/android_kernel_motorola_sm6375
Watch
2
0
Fork
You've already forked android_kernel_motorola_sm6375
 0
mirror of https://github.com/BobTheBlinker/android_kernel_motorola_sm6375.git synced 2026-10-15 01:01:56 -04:00
Code Issues Projects Releases Packages Wiki Activity

qcacld-3.0: fix a potential spinlock lockup issue

Browse source 
When time stamping RX packets, spinlock
host_target_sync_lock will be hold if it's
now in TSF capturing state; if TSF-captured
IRQ happened at this moment, it will also
try to get host_target_sync_lock; if it's
handled on the same CPU, lockup will happen.

In the use-case, to update the current host
time, it's no need to get the spinlock in
this TSF-captured IRQ context, so move
host-time updating out of the lock.

Change-Id: I87205d5935bd2063c80ce7cf767cbc36dde55236
CRs-Fixed: 2057693
This commit is contained in:
Yu Wang 2017-06-08 12:15:42 +08:00 • committed by Gerrit - the friendly Code Review server
commit 70757d54d1
1 changed files with 12 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

14
core/hdd/src/wlan_hdd_tsf.c
View file

@ -422,11 +422,21 @@ static void hdd_update_timestamp(hdd_adapter_t *adapter,
if (!adapter)
return;
/* host time is updated in IRQ context, it's always before target time,
* and so no need to try update last_host_time at present;
* since the interval of capturing TSF
* (WLAN_HDD_CAPTURE_TSF_INTERVAL_SEC) is long enough, host and target
* time are updated in pairs, and one by one, we can return here to
* avoid requiring spin lock, and to speed up the IRQ processing.
*/
if (host_time > 0) {
adapter->cur_host_time = host_time;
return;
}
qdf_spin_lock_bh(&adapter->host_target_sync_lock);
if (target_time > 0)
adapter->cur_target_time = target_time;
if (host_time > 0)
adapter->cur_host_time = host_time;
sync_status = hdd_check_timestamp_status(adapter->last_target_time,
adapter->last_host_time,