ipvs: fix races around est_mutex and est_cpulist
Subsystems (4)
Julian Anastasov
maintainerPATCH v4 1/1 ipvs: fix races around est_mutex and est_cpulist
Sashiko reports for races and possible crash around the usage of est_cpulist_valid and sysctl_est_cpulist. The problem is that we do not lock est_mutex in some places which can lead to wrong write ordering and as result problems when calling cpumask_weight() and cpumask_empty().
Fix them by moving the est_max_threads read/write under locked est_mutex. Do the same for one ip_vs_est_reload_start() call to protect the cpumask_empty() usage of sysctl_est_cpulist.
To remove the chance of deadlock while stopping the estimation kthreads, keep the data structure for kthread 0 even after last estimator is removed and do not hold mutexes while stopping this task. Now we will use a new flag 'needed' to know when kthread 0 should run. The kthreads above 0 do not use mutexes, so stop them under est_mutex because their kthread data still can be destroyed if they do not serve estimators. Now all kthreads will be started by the est_reload_work to properly serialize the stop/start for kthread 0.
Reduce the use of service_mutex in ip_vs_est_calc_phase() because under est_mutex we can safely walk est_kt_arr to stop the kthreads above slot 0.
Finally, fix use-after-free for kd->est_row in ip_vs_est_calc_phase(). est->ktrow should simply switch to a delay value while estimator is linked to est_temp_list.
--- include/net/ip_vs.h | 3 +- net/netfilter/ipvs/ip_vs_ctl.c | 37 +++++++++++---- net/netfilter/ipvs/ip_vs_est.c | 83 ++++++++++++++++++++-------------- 3 files changed, 79 insertions(+), 44 deletions(-)
include/net/ip_vs.h +2 -1
@@ -491,6 +491,7 @@ struct ip_vs_est_kt_data {DECLARE_BITMAP (avail,IPVS_EST_NTICKS );/* tick has space for ests */ unsigned long est_timer;/* estimation timer (jiffies) */ struct ip_vs_stats *calc_stats;/* Used for calculation */ +int needed;/* task is needed */ int tick_len[IPVS_EST_NTICKS ];/* est count */ int id;/* ktid per netns */ int chain_max;/* max ests per tick chain */
@@ -1884,7 +1885,7 @@ int ip_vs_start_estimator(struct netns_ipvs *ipvs, struct ip_vs_stats *stats);void ip_vs_stop_estimator (struct netns_ipvs *ipvs,struct ip_vs_stats *stats);void ip_vs_zero_estimator (struct ip_vs_stats *stats);void ip_vs_read_estimator (struct ip_vs_kstats *dst,struct ip_vs_stats *stats); -void ip_vs_est_reload_start (struct netns_ipvs *ipvs); +void ip_vs_est_reload_start (struct netns_ipvs *ipvs,bool restart);int ip_vs_est_kthread_start (struct netns_ipvs *ipvs,struct ip_vs_est_kt_data *kd);void ip_vs_est_kthread_stop (struct ip_vs_est_kt_data *kd);
net/netfilter/ipvs/ip_vs_ctl.c +29 -8
@@ -261,12 +261,28 @@ static void est_reload_work_handler(struct work_struct *work)if (!kd)continue ;/* New config ? Stop kthread tasks */ -if (genid != genid_done) -ip_vs_est_kthread_stop (kd); +if (genid != genid_done) { +if (!id) { +/* Only we can stop kt 0 but not under mutex */ +mutex_unlock (&ipvs->est_mutex); +ip_vs_est_kthread_stop (kd); +mutex_lock (&ipvs->est_mutex); +if (!READ_ONCE (ipvs->enable)) + goto unlock; +/* kd for kt 0 is never destroyed */ + }else { +ip_vs_est_kthread_stop (kd); + } + }if (!kd->task && !ip_vs_est_stopped (ipvs)) { +bool start; +/* Do not start kthreads above 0 in calc phase */ -if ((!id || !ipvs->est_calc_phase) && -ip_vs_est_kthread_start (ipvs, kd) <0 ) +if (id) + start = !ipvs->est_calc_phase; +else + start = kd->needed; +if (start &&ip_vs_est_kthread_start (ipvs, kd) <0 ) repeat = true; } }
@@ -1812,11 +1828,16 @@ ip_vs_add_service(struct netns_ipvs *ipvs, struct ip_vs_service_user_kern *u, *svc_p = svc;if (!READ_ONCE (ipvs->enable)) { +mutex_lock (&ipvs->est_mutex); +/* Now there is a service - full throttle */ WRITE_ONCE (ipvs->enable,1 ); + ipvs->est_max_threads =ip_vs_est_max_threads (ipvs); +/* Start estimation for first time */ -ip_vs_est_reload_start (ipvs); +ip_vs_est_reload_start (ipvs, true); +mutex_unlock (&ipvs->est_mutex); }return 0 ;
@@ -2337,7 +2358,7 @@ static int ipvs_proc_est_cpumask_set(const struct ctl_table *table,/* est_max_threads may depend on cpulist size */ ipvs->est_max_threads =ip_vs_est_max_threads (ipvs); ipvs->est_calc_phase =1 ; -ip_vs_est_reload_start (ipvs); +ip_vs_est_reload_start (ipvs, true); unlock:mutex_unlock (&ipvs->est_mutex);
@@ -2417,7 +2438,7 @@ static int ipvs_proc_est_nice(const struct ctl_table *table, int write,mutex_lock (&ipvs->est_mutex);if (*valp != val) { *valp = val; -ip_vs_est_reload_start (ipvs); +ip_vs_est_reload_start (ipvs, true); }mutex_unlock (&ipvs->est_mutex); }
@@ -2444,7 +2465,7 @@ static int ipvs_proc_run_estimation(const struct ctl_table *table, int write,mutex_lock (&ipvs->est_mutex);if (*valp != val) { *valp = val; -ip_vs_est_reload_start (ipvs); +ip_vs_est_reload_start (ipvs, true); }mutex_unlock (&ipvs->est_mutex); }
net/netfilter/ipvs/ip_vs_est.c +48 -35
184 lines not on this page · raw message
Signature · 1 line
2.53.0
Julian Anastasov
maintainerHello,
Quoted 30 lines · Julian Anastasov
Sashiko reports for races and possible crash around the usage of
est_cpulist_validandsysctl_est_cpulist. The problem is that we do not lockest_mutexin some places which can lead to wrong write ordering and as result problems when callingcpumask_weight()andcpumask_empty().Fix them by moving the
est_max_threadsread/write under lockedest_mutex. Do the same for oneip_vs_est_reload_start()call to protect thecpumask_empty()usage ofsysctl_est_cpulist.To remove the chance of deadlock while stopping the estimation kthreads, keep the data structure for kthread 0 even after last estimator is removed and do not hold mutexes while stopping this task. Now we will use a new flag 'needed' to know when kthread 0 should run. The kthreads above 0 do not use mutexes, so stop them under
est_mutexbecause their kthread data still can be destroyed if they do not serve estimators. Now all kthreads will be started by theest_reload_workto properly serialize the stop/start for kthread 0.Reduce the use of
service_mutexinip_vs_est_calc_phase()because underest_mutexwe can safely walkest_kt_arrto stop the kthreads above slot 0.Finally, fix use-after-free for
kd->est_rowinip_vs_est_calc_phase().est->ktrowshould simply switch to a delay value while estimator is linked toest_temp_list.
Quoted 6 lines
Fixes f0be83d54217 ("ipvs: add est_cpulist and est_nice sysctl vars")Signed-off-by Julian Anastasov <[email protected]>
Sashiko found another missing mutex lock [1].
I'll send new patch version...
pw-bot: changes-requested
[1] https://sashiko.dev/#/patchset/20260425103918.7447-1-ja%40ssi.bg
Regards
--
Julian Anastasov <[email protected]>