android/android_kernel_samsung_sm8650
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

ANDROID: sched/uclamp: Don't enable uclamp_is_used static key by in-kernel requests

Browse Source 
We do have now in-kernel users of uclamp to implement inheritance. The
static_branch_enable() path unconditionally holds the cpus_read_lock()
which might_sleep(). The path in binder that implements inheritance
happens from in_atomic() context which leads to a splat like this one:

	[  147.529960] BUG: sleeping function called from invalid context at include/linux/percpu-rwsem.h:56
	[  147.530196] in_atomic(): 1, irqs_disabled(): 0, non_block: 0, pid: 2586, name: RenderThread
	[  147.530410] INFO: lockdep is turned off.
	[  147.530518] Preemption disabled at:
	[  147.530521] [<ffffffc008ca2cec>] binder_proc_transaction+0x78/0x41c
	[  147.530793] CPU: 8 PID: 2586 Comm: RenderThread Tainted: G S      W  O      5.15.76-android14-5-00086-gc01afe5d262f #1
	[  147.531214] Call trace:
	[  147.531288] dump_backtrace+0xe8/0x134
	[  147.531444] show_stack+0x1c/0x4c
	[  147.531598] dump_stack_lvl+0x74/0x94
	[  147.531766] dump_stack+0x14/0x3c
	[  147.531920] ___might_sleep+0x210/0x230
	[  147.532094] __might_sleep+0x54/0x84
	[  147.532259] cpus_read_lock+0x2c/0x160
	[  147.532429] static_key_enable+0x1c/0x34
	[  147.532608] __sched_setscheduler+0x2a8/0x99c
	[  147.532802] sched_setattr_nocheck+0x1c/0x24
	[  147.532994] binder_do_set_priority+0x31c/0x4a4
	[  147.533195] binder_transaction_priority+0x200/0x3f4
	[  147.533413] binder_proc_transaction+0x220/0x41c
	[  147.533618] binder_transaction+0x1df0/0x234c
	[  147.533812] binder_thread_write+0xd84/0x2398
	[  147.534007] binder_ioctl_write_read+0x19c/0xb28
	[  147.534212] binder_ioctl+0x344/0x1a3c
	[  147.534382] __arm64_sys_ioctl+0x94/0xc8
	[  147.534561] invoke_syscall+0x44/0xf8
	[  147.534729] el0_svc_common+0xc8/0x10c
	[  147.534900] do_el0_svc+0x20/0x28
	[  147.535053] el0_svc+0x58/0xe0
	[  147.535198] el0t_64_sync_handler+0x7c/0xe4
	[  147.535386] el0t_64_sync+0x188/0x18c

Prevent enabling the lock for !user initiated sched_setattr()
operations. Generally we don't expect in-kernel uclamp users.

Bug: 259145692
Signed-off-by: Qais Yousef <qyousef@google.com>
Change-Id: Iac5be139b5ffd39f5e1c0431ce253133d81b98cf
This commit is contained in:
Qais Yousef 2022-11-25 15:23:17 +00:00 committed by Treehugger Robot
parent 2b25d535d0
commit b57e3c1d99
1 changed files with 14 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
kernel/sched/core.c
View File

@ -1880,7 +1880,7 @@ static int sysctl_sched_uclamp_handler(struct ctl_table *table, int write,
#endif
static int uclamp_validate(struct task_struct *p,
const struct sched_attr *attr)
const struct sched_attr *attr, bool user)
{
int util_min = p->uclamp_req[UCLAMP_MIN].value;
int util_max = p->uclamp_req[UCLAMP_MAX].value;
@ -1905,11 +1905,19 @@ static int uclamp_validate(struct task_struct *p,
/*
* We have valid uclamp attributes; make sure uclamp is enabled.
*
* We need to do that here, because enabling static branches is a
* blocking operation which obviously cannot be done while holding
* We need to do that here, because enabling static branches is
* a blocking operation which obviously cannot be done while holding
* scheduler locks.
*
* We only enable the static key if this was initiated by user space
* request. There should be no in-kernel users of uclamp except to
* implement things like inheritance like in binder. These in-kernel
* callers can rightfully be called be sometimes in_atomic() context
* which is invalid context to enable the key in. The enabling path
* unconditionally holds the cpus_read_lock() which might_sleep().
*/
static_branch_enable(&sched_uclamp_used);
if (user)
static_branch_enable(&sched_uclamp_used);
return 0;
}
@ -2050,7 +2058,7 @@ static void __init init_uclamp(void)
static inline void uclamp_rq_inc(struct rq *rq, struct task_struct *p) { }
static inline void uclamp_rq_dec(struct rq *rq, struct task_struct *p) { }
static inline int uclamp_validate(struct task_struct *p,
const struct sched_attr *attr)
const struct sched_attr *attr, bool user)
{
return -EOPNOTSUPP;
}
@ -7649,7 +7657,7 @@ static int __sched_setscheduler(struct task_struct *p,
/* Update task specific "requested" clamps */
if (attr->sched_flags & SCHED_FLAG_UTIL_CLAMP) {
retval = uclamp_validate(p, attr);
retval = uclamp_validate(p, attr, user);
if (retval)
return retval;
}