diff options
Diffstat (limited to 'kernel')
| -rw-r--r-- | kernel/bpf/core.c | 30 | ||||
| -rw-r--r-- | kernel/bpf/verifier.c | 70 | ||||
| -rw-r--r-- | kernel/sched/core.c | 3 | ||||
| -rw-r--r-- | kernel/sched/deadline.c | 98 | ||||
| -rw-r--r-- | kernel/sched/rt.c | 148 | ||||
| -rw-r--r-- | kernel/sched/sched.h | 1 | ||||
| -rw-r--r-- | kernel/time/hrtimer.c | 5 |
7 files changed, 338 insertions, 17 deletions
diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c index 3fd76cf0c21e..eb52d11fdaa7 100644 --- a/kernel/bpf/core.c +++ b/kernel/bpf/core.c @@ -256,6 +256,7 @@ noinline u64 __bpf_call_base(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5) } EXPORT_SYMBOL_GPL(__bpf_call_base); +#ifndef CONFIG_BPF_JIT_ALWAYS_ON /** * __bpf_prog_run - run eBPF program on a given context * @ctx: is the data we are operating on @@ -443,7 +444,7 @@ select_insn: DST = tmp; CONT; ALU_MOD_X: - if (unlikely(SRC == 0)) + if (unlikely((u32)SRC == 0)) return 0; tmp = (u32) DST; DST = do_div(tmp, (u32) SRC); @@ -462,7 +463,7 @@ select_insn: DST = div64_u64(DST, SRC); CONT; ALU_DIV_X: - if (unlikely(SRC == 0)) + if (unlikely((u32)SRC == 0)) return 0; tmp = (u32) DST; do_div(tmp, (u32) SRC); @@ -517,7 +518,7 @@ select_insn: struct bpf_map *map = (struct bpf_map *) (unsigned long) BPF_R2; struct bpf_array *array = container_of(map, struct bpf_array, map); struct bpf_prog *prog; - u64 index = BPF_R3; + u32 index = BPF_R3; if (unlikely(index >= array->map.max_entries)) goto out; @@ -725,6 +726,13 @@ load_byte: return 0; } +#else +static unsigned int __bpf_prog_ret0(void *ctx, const struct bpf_insn *insn) +{ + return 0; +} +#endif + bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp) { @@ -771,9 +779,23 @@ static int bpf_check_tail_call(const struct bpf_prog *fp) */ int bpf_prog_select_runtime(struct bpf_prog *fp) { +#ifndef CONFIG_BPF_JIT_ALWAYS_ON fp->bpf_func = (void *) __bpf_prog_run; - +#else + fp->bpf_func = (void *) __bpf_prog_ret0; +#endif + + /* eBPF JITs can rewrite the program in case constant + * blinding is active. However, in case of error during + * blinding, bpf_int_jit_compile() must always return a + * valid program, which in this case would simply not + * be JITed, but falls back to the interpreter. + */ bpf_int_jit_compile(fp); +#ifdef CONFIG_BPF_JIT_ALWAYS_ON + if (!fp->jited) + return -ENOTSUPP; +#endif bpf_prog_lock_ro(fp); /* The tail call compatibility check can only be done at diff --git a/kernel/bpf/verifier.c b/kernel/bpf/verifier.c index 014c2d759916..c14003840bc5 100644 --- a/kernel/bpf/verifier.c +++ b/kernel/bpf/verifier.c @@ -191,6 +191,7 @@ struct bpf_insn_aux_data { enum bpf_reg_type ptr_type; /* pointer type for load/store insns */ struct bpf_map *map_ptr; /* pointer for call insn into lookup_elem */ }; + bool seen; /* this insn was processed by the verifier */ }; #define MAX_USED_MAPS 64 /* max number of maps accessed by one eBPF program */ @@ -682,6 +683,13 @@ static bool is_pointer_value(struct verifier_env *env, int regno) } } +static bool is_ctx_reg(struct verifier_env *env, int regno) +{ + const struct reg_state *reg = &env->cur_state.regs[regno]; + + return reg->type == PTR_TO_CTX; +} + /* check whether memory at (regno + off) is accessible for t = (read | write) * if t==write, value_regno is a register which value is stored into memory * if t==read, value_regno is a register which will receive the value from memory @@ -778,6 +786,12 @@ static int check_xadd(struct verifier_env *env, struct bpf_insn *insn) return -EACCES; } + if (is_ctx_reg(env, insn->dst_reg)) { + verbose("BPF_XADD stores into R%d context is not allowed\n", + insn->dst_reg); + return -EACCES; + } + /* check whether atomic_add can read the memory */ err = check_mem_access(env, insn->dst_reg, insn->off, BPF_SIZE(insn->code), BPF_READ, -1); @@ -1164,6 +1178,11 @@ static int check_alu_op(struct verifier_env *env, struct bpf_insn *insn) return -EINVAL; } + if (opcode == BPF_ARSH && BPF_CLASS(insn->code) != BPF_ALU64) { + verbose("BPF_ARSH not supported for 32 bit ALU\n"); + return -EINVAL; + } + if ((opcode == BPF_LSH || opcode == BPF_RSH || opcode == BPF_ARSH) && BPF_SRC(insn->code) == BPF_K) { int size = BPF_CLASS(insn->code) == BPF_ALU64 ? 64 : 32; @@ -1793,6 +1812,7 @@ static int do_check(struct verifier_env *env) print_bpf_insn(env, insn); } + env->insn_aux_data[insn_idx].seen = true; if (class == BPF_ALU || class == BPF_ALU64) { err = check_alu_op(env, insn); if (err) @@ -1902,6 +1922,12 @@ static int do_check(struct verifier_env *env) if (err) return err; + if (is_ctx_reg(env, insn->dst_reg)) { + verbose("BPF_ST stores into R%d context is not allowed\n", + insn->dst_reg); + return -EACCES; + } + /* check that memory (dst_reg + off) is writeable */ err = check_mem_access(env, insn->dst_reg, insn->off, BPF_SIZE(insn->code), BPF_WRITE, @@ -1988,6 +2014,7 @@ process_bpf_exit: return err; insn_idx++; + env->insn_aux_data[insn_idx].seen = true; } else { verbose("invalid BPF_LD mode\n"); return -EINVAL; @@ -2125,6 +2152,7 @@ static int adjust_insn_aux_data(struct verifier_env *env, u32 prog_len, u32 off, u32 cnt) { struct bpf_insn_aux_data *new_data, *old_data = env->insn_aux_data; + int i; if (cnt == 1) return 0; @@ -2134,6 +2162,8 @@ static int adjust_insn_aux_data(struct verifier_env *env, u32 prog_len, memcpy(new_data, old_data, sizeof(struct bpf_insn_aux_data) * off); memcpy(new_data + off + cnt - 1, old_data + off, sizeof(struct bpf_insn_aux_data) * (prog_len - off - cnt + 1)); + for (i = off; i < off + cnt - 1; i++) + new_data[i].seen = true; env->insn_aux_data = new_data; vfree(old_data); return 0; @@ -2152,6 +2182,25 @@ static struct bpf_prog *bpf_patch_insn_data(struct verifier_env *env, u32 off, return new_prog; } +/* The verifier does more data flow analysis than llvm and will not explore + * branches that are dead at run time. Malicious programs can have dead code + * too. Therefore replace all dead at-run-time code with nops. + */ +static void sanitize_dead_code(struct verifier_env *env) +{ + struct bpf_insn_aux_data *aux_data = env->insn_aux_data; + struct bpf_insn nop = BPF_MOV64_REG(BPF_REG_0, BPF_REG_0); + struct bpf_insn *insn = env->prog->insnsi; + const int insn_cnt = env->prog->len; + int i; + + for (i = 0; i < insn_cnt; i++) { + if (aux_data[i].seen) + continue; + memcpy(insn + i, &nop, sizeof(nop)); + } +} + /* convert load instructions that access fields of 'struct __sk_buff' * into sequence of instructions that access fields of 'struct sk_buff' */ @@ -2218,6 +2267,24 @@ static int fixup_bpf_calls(struct verifier_env *env) int i, cnt, delta = 0; for (i = 0; i < insn_cnt; i++, insn++) { + if (insn->code == (BPF_ALU | BPF_MOD | BPF_X) || + insn->code == (BPF_ALU | BPF_DIV | BPF_X)) { + /* due to JIT bugs clear upper 32-bits of src register + * before div/mod operation + */ + insn_buf[0] = BPF_MOV32_REG(insn->src_reg, insn->src_reg); + insn_buf[1] = *insn; + cnt = 2; + new_prog = bpf_patch_insn_data(env, i + delta, insn_buf, cnt); + if (!new_prog) + return -ENOMEM; + + delta += cnt - 1; + env->prog = prog = new_prog; + insn = new_prog->insnsi + i + delta; + continue; + } + if (insn->code != (BPF_JMP | BPF_CALL)) continue; @@ -2371,6 +2438,9 @@ skip_full_check: free_states(env); if (ret == 0) + sanitize_dead_code(env); + + if (ret == 0) /* program is valid, convert *(u32*)(ctx + off) accesses */ ret = convert_ctx_accesses(env); diff --git a/kernel/sched/core.c b/kernel/sched/core.c index 8db48e263f8e..41a8fdb1436c 100644 --- a/kernel/sched/core.c +++ b/kernel/sched/core.c @@ -2289,6 +2289,7 @@ void __dl_clear_params(struct task_struct *p) dl_se->dl_period = 0; dl_se->flags = 0; dl_se->dl_bw = 0; + dl_se->dl_density = 0; dl_se->dl_throttled = 0; dl_se->dl_new = 1; @@ -2367,6 +2368,7 @@ static void __sched_fork(unsigned long clone_flags, struct task_struct *p) init_dl_task_timer(&p->dl); __dl_clear_params(p); + init_rt_schedtune_timer(&p->rt); INIT_LIST_HEAD(&p->rt.run_list); p->rt.timeout = 0; p->rt.time_slice = sched_rr_timeslice; @@ -3991,6 +3993,7 @@ __setparam_dl(struct task_struct *p, const struct sched_attr *attr) dl_se->dl_period = attr->sched_period ?: dl_se->dl_deadline; dl_se->flags = attr->sched_flags; dl_se->dl_bw = to_ratio(dl_se->dl_period, dl_se->dl_runtime); + dl_se->dl_density = to_ratio(dl_se->dl_deadline, dl_se->dl_runtime); /* * Changing the parameters of a task is 'tricky' and we're not doing diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c index 105f6cdac6fc..188c8388a63f 100644 --- a/kernel/sched/deadline.c +++ b/kernel/sched/deadline.c @@ -502,13 +502,84 @@ static bool dl_entity_overflow(struct sched_dl_entity *dl_se, } /* - * When a -deadline entity is queued back on the runqueue, its runtime and - * deadline might need updating. + * Revised wakeup rule [1]: For self-suspending tasks, rather then + * re-initializing task's runtime and deadline, the revised wakeup + * rule adjusts the task's runtime to avoid the task to overrun its + * density. * - * The policy here is that we update the deadline of the entity only if: - * - the current deadline is in the past, - * - using the remaining runtime with the current deadline would make - * the entity exceed its bandwidth. + * Reasoning: a task may overrun the density if: + * runtime / (deadline - t) > dl_runtime / dl_deadline + * + * Therefore, runtime can be adjusted to: + * runtime = (dl_runtime / dl_deadline) * (deadline - t) + * + * In such way that runtime will be equal to the maximum density + * the task can use without breaking any rule. + * + * [1] Luca Abeni, Giuseppe Lipari, and Juri Lelli. 2015. Constant + * bandwidth server revisited. SIGBED Rev. 11, 4 (January 2015), 19-24. + */ +static void +update_dl_revised_wakeup(struct sched_dl_entity *dl_se, struct rq *rq) +{ + u64 laxity = dl_se->deadline - rq_clock(rq); + + /* + * If the task has deadline < period, and the deadline is in the past, + * it should already be throttled before this check. + * + * See update_dl_entity() comments for further details. + */ + WARN_ON(dl_time_before(dl_se->deadline, rq_clock(rq))); + + dl_se->runtime = (dl_se->dl_density * laxity) >> 20; +} + +/* + * Regarding the deadline, a task with implicit deadline has a relative + * deadline == relative period. A task with constrained deadline has a + * relative deadline <= relative period. + * + * We support constrained deadline tasks. However, there are some restrictions + * applied only for tasks which do not have an implicit deadline. See + * update_dl_entity() to know more about such restrictions. + * + * The dl_is_implicit() returns true if the task has an implicit deadline. + */ +static inline bool dl_is_implicit(struct sched_dl_entity *dl_se) +{ + return dl_se->dl_deadline == dl_se->dl_period; +} + +/* + * When a deadline entity is placed in the runqueue, its runtime and deadline + * might need to be updated. This is done by a CBS wake up rule. There are two + * different rules: 1) the original CBS; and 2) the Revisited CBS. + * + * When the task is starting a new period, the Original CBS is used. In this + * case, the runtime is replenished and a new absolute deadline is set. + * + * When a task is queued before the begin of the next period, using the + * remaining runtime and deadline could make the entity to overflow, see + * dl_entity_overflow() to find more about runtime overflow. When such case + * is detected, the runtime and deadline need to be updated. + * + * If the task has an implicit deadline, i.e., deadline == period, the Original + * CBS is applied. the runtime is replenished and a new absolute deadline is + * set, as in the previous cases. + * + * However, the Original CBS does not work properly for tasks with + * deadline < period, which are said to have a constrained deadline. By + * applying the Original CBS, a constrained deadline task would be able to run + * runtime/deadline in a period. With deadline < period, the task would + * overrun the runtime/period allowed bandwidth, breaking the admission test. + * + * In order to prevent this misbehave, the Revisited CBS is used for + * constrained deadline tasks when a runtime overflow is detected. In the + * Revisited CBS, rather than replenishing & setting a new absolute deadline, + * the remaining runtime of the task is reduced to avoid runtime overflow. + * Please refer to the comments update_dl_revised_wakeup() function to find + * more about the Revised CBS rule. */ static void update_dl_entity(struct sched_dl_entity *dl_se, struct sched_dl_entity *pi_se) @@ -530,6 +601,14 @@ static void update_dl_entity(struct sched_dl_entity *dl_se, if (dl_time_before(dl_se->deadline, rq_clock(rq)) || dl_entity_overflow(dl_se, pi_se, rq_clock(rq))) { + + if (unlikely(!dl_is_implicit(dl_se) && + !dl_time_before(dl_se->deadline, rq_clock(rq)) && + !dl_se->dl_boosted)){ + update_dl_revised_wakeup(dl_se, rq); + return; + } + dl_se->deadline = rq_clock(rq) + pi_se->dl_deadline; dl_se->runtime = pi_se->dl_runtime; } @@ -1054,11 +1133,6 @@ static void dequeue_dl_entity(struct sched_dl_entity *dl_se) __dequeue_dl_entity(dl_se); } -static inline bool dl_is_constrained(struct sched_dl_entity *dl_se) -{ - return dl_se->dl_deadline < dl_se->dl_period; -} - static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags) { struct task_struct *pi_task = rt_mutex_get_top_task(p); @@ -1090,7 +1164,7 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags) * If that is the case, the task will be throttled and * the replenishment timer will be set to the next period. */ - if (!p->dl.dl_throttled && dl_is_constrained(&p->dl)) + if (!p->dl.dl_throttled && !dl_is_implicit(&p->dl)) dl_check_constrained_dl(&p->dl); /* diff --git a/kernel/sched/rt.c b/kernel/sched/rt.c index 8025828ff4e0..baad64b94f15 100644 --- a/kernel/sched/rt.c +++ b/kernel/sched/rt.c @@ -9,6 +9,7 @@ #include <linux/slab.h> #include <linux/irq_work.h> #include <trace/events/sched.h> +#include <linux/hrtimer.h> #include "tune.h" @@ -980,6 +981,70 @@ static int sched_rt_runtime_exceeded(struct rt_rq *rt_rq) return 0; } +#define RT_SCHEDTUNE_INTERVAL 50000000ULL + +static enum hrtimer_restart rt_schedtune_timer(struct hrtimer *timer) +{ + struct sched_rt_entity *rt_se = container_of(timer, + struct sched_rt_entity, + schedtune_timer); + struct task_struct *p = rt_task_of(rt_se); + struct rq *rq = task_rq(p); + + raw_spin_lock(&rq->lock); + + /* + * Nothing to do if: + * - task has switched runqueues + * - task isn't RT anymore + */ + if (rq != task_rq(p) || (p->sched_class != &rt_sched_class)) + goto out; + + /* + * If task got enqueued back during callback time, it means we raced + * with the enqueue on another cpu, that's Ok, just do nothing as + * enqueue path would have tried to cancel us and we shouldn't run + * Also check the schedtune_enqueued flag as class-switch on a + * sleeping task may have already canceled the timer and done dq + */ + if (p->on_rq || !rt_se->schedtune_enqueued) + goto out; + + /* + * RT task is no longer active, cancel boost + */ + rt_se->schedtune_enqueued = false; + schedtune_dequeue_task(p, cpu_of(rq)); + cpufreq_update_this_cpu(rq, SCHED_CPUFREQ_RT); +out: + raw_spin_unlock(&rq->lock); + + /* + * This can free the task_struct if no more references. + */ + put_task_struct(p); + + return HRTIMER_NORESTART; +} + +void init_rt_schedtune_timer(struct sched_rt_entity *rt_se) +{ + struct hrtimer *timer = &rt_se->schedtune_timer; + + hrtimer_init(timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); + timer->function = rt_schedtune_timer; + rt_se->schedtune_enqueued = false; +} + +static void start_schedtune_timer(struct sched_rt_entity *rt_se) +{ + struct hrtimer *timer = &rt_se->schedtune_timer; + + hrtimer_start(timer, ns_to_ktime(RT_SCHEDTUNE_INTERVAL), + HRTIMER_MODE_REL_PINNED); +} + /* * Update the current task's runtime statistics. Skip current tasks that * are not in our scheduling class. @@ -1386,7 +1451,32 @@ enqueue_task_rt(struct rq *rq, struct task_struct *p, int flags) if (!task_current(rq, p) && p->nr_cpus_allowed > 1) enqueue_pushable_task(rq, p); + if (!schedtune_task_boost(p)) + return; + + /* + * If schedtune timer is active, that means a boost was already + * done, just cancel the timer so that deboost doesn't happen. + * Otherwise, increase the boost. If an enqueued timer was + * cancelled, put the task reference. + */ + if (hrtimer_try_to_cancel(&rt_se->schedtune_timer) == 1) + put_task_struct(p); + + /* + * schedtune_enqueued can be true in the following situation: + * enqueue_task_rt grabs rq lock before timer fires + * or before its callback acquires rq lock + * schedtune_enqueued can be false if timer callback is running + * and timer just released rq lock, or if the timer finished + * running and canceling the boost + */ + if (rt_se->schedtune_enqueued) + return; + + rt_se->schedtune_enqueued = true; schedtune_enqueue_task(p, cpu_of(rq)); + cpufreq_update_this_cpu(rq, SCHED_CPUFREQ_RT); } static void dequeue_task_rt(struct rq *rq, struct task_struct *p, int flags) @@ -1398,7 +1488,19 @@ static void dequeue_task_rt(struct rq *rq, struct task_struct *p, int flags) dec_hmp_sched_stats_rt(rq, p); dequeue_pushable_task(rq, p); + + if (!rt_se->schedtune_enqueued) + return; + + if (flags == DEQUEUE_SLEEP) { + get_task_struct(p); + start_schedtune_timer(rt_se); + return; + } + + rt_se->schedtune_enqueued = false; schedtune_dequeue_task(p, cpu_of(rq)); + cpufreq_update_this_cpu(rq, SCHED_CPUFREQ_RT); } /* @@ -1472,6 +1574,32 @@ task_may_not_preempt(struct task_struct *task, int cpu) task_thread_info(task)->preempt_count & SOFTIRQ_MASK)); } +/* + * Perform a schedtune dequeue and cancelation of boost timers if needed. + * Should be called only with the rq->lock held. + */ +static void schedtune_dequeue_rt(struct rq *rq, struct task_struct *p) +{ + struct sched_rt_entity *rt_se = &p->rt; + + BUG_ON(!raw_spin_is_locked(&rq->lock)); + + if (!rt_se->schedtune_enqueued) + return; + + /* + * Incase of class change cancel any active timers. If an enqueued + * timer was cancelled, put the task ref. + */ + if (hrtimer_try_to_cancel(&rt_se->schedtune_timer) == 1) + put_task_struct(p); + + /* schedtune_enqueued is true, deboost it */ + rt_se->schedtune_enqueued = false; + schedtune_dequeue_task(p, task_cpu(p)); + cpufreq_update_this_cpu(rq, SCHED_CPUFREQ_RT); +} + static int select_task_rq_rt(struct task_struct *p, int cpu, int sd_flag, int flags, int sibling_count_hint) @@ -1546,6 +1674,19 @@ select_task_rq_rt(struct task_struct *p, int cpu, int sd_flag, int flags, rcu_read_unlock(); out: + /* + * If previous CPU was different, make sure to cancel any active + * schedtune timers and deboost. + */ + if (task_cpu(p) != cpu) { + unsigned long fl; + struct rq *prq = task_rq(p); + + raw_spin_lock_irqsave(&prq->lock, fl); + schedtune_dequeue_rt(prq, p); + raw_spin_unlock_irqrestore(&prq->lock, fl); + } + return cpu; } @@ -2402,6 +2543,13 @@ static void rq_offline_rt(struct rq *rq) static void switched_from_rt(struct rq *rq, struct task_struct *p) { /* + * On class switch from rt, always cancel active schedtune timers, + * this handles the cases where we switch class for a task that is + * already rt-dequeued but has a running timer. + */ + schedtune_dequeue_rt(rq, p); + + /* * If there are other RT tasks then we will reschedule * and the scheduling of the other RT tasks will handle * the balancing. But if we are the last RT task diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h index ca2294d06f44..6ef3a59612a9 100644 --- a/kernel/sched/sched.h +++ b/kernel/sched/sched.h @@ -2199,6 +2199,7 @@ extern void resched_cpu(int cpu); extern struct rt_bandwidth def_rt_bandwidth; extern void init_rt_bandwidth(struct rt_bandwidth *rt_b, u64 period, u64 runtime); +extern void init_rt_schedtune_timer(struct sched_rt_entity *rt_se); extern struct dl_bandwidth def_dl_bandwidth; extern void init_dl_bandwidth(struct dl_bandwidth *dl_b, u64 period, u64 runtime); diff --git a/kernel/time/hrtimer.c b/kernel/time/hrtimer.c index e7c2392666cb..0605755c246f 100644 --- a/kernel/time/hrtimer.c +++ b/kernel/time/hrtimer.c @@ -312,7 +312,7 @@ EXPORT_SYMBOL_GPL(__ktime_divns); */ ktime_t ktime_add_safe(const ktime_t lhs, const ktime_t rhs) { - ktime_t res = ktime_add(lhs, rhs); + ktime_t res = ktime_add_unsafe(lhs, rhs); /* * We use KTIME_SEC_MAX here, the maximum timeout which we can @@ -669,7 +669,9 @@ static void hrtimer_reprogram(struct hrtimer *timer, static inline void hrtimer_init_hres(struct hrtimer_cpu_base *base) { base->expires_next.tv64 = KTIME_MAX; + base->hang_detected = 0; base->hres_active = 0; + base->next_timer = NULL; } /* @@ -1587,6 +1589,7 @@ static void init_hrtimers_cpu(int cpu) timerqueue_init_head(&cpu_base->clock_base[i].active); } + cpu_base->active_bases = 0; cpu_base->cpu = cpu; hrtimer_init_hres(cpu_base); } |