Messages in this thread Patch in this message |  | | Date | Mon, 5 Aug 2019 13:53:09 +0200 | From | Peter Zijlstra <> | Subject | Re: [RFC][PATCH 01/13] sched/deadline: Impose global limits on sched_attr::sched_period |
| |
On Fri, Aug 02, 2019 at 06:21:04PM +0100, Alessio Balsini wrote: > Hi Peter, > > This is indeed an important missing piece. > > I think it would be worth having some simple checks, e.g.: > - period_min <= period_max; > - period_min >= (1ULL << DL_SCALE). > > To be even more picky, I'm wondering if it would be a good practice to > fail the write operation if there are already dl-tasks in the system > that violate the new constraint. I'm afraid there is no cheap way of > achieving such check, so, I think we can skip this last tricky thing for > now.
Like so?
--- Subject: sched/deadline: Impose global limits on sched_attr::sched_period From: Peter Zijlstra <peterz@infradead.org> Date: Tue Jul 23 16:01:29 CEST 2019
There are two DoS scenarios with SCHED_DEADLINE related to sched_attr::sched_period:
- since access-control only looks at utilization and density, a very large period can allow a very large runtime, which in turn can incur a very large latency to lower priority tasks.
- for very short periods we can end up spending more time programming the hardware timer than actually running the task.
Mitigate these by imposing limits on the period.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Daniel Bristot de Oliveira <bristot@redhat.com> Cc: Luca Abeni <luca.abeni@santannapisa.it> Cc: Juri Lelli <juri.lelli@redhat.com> Cc: Dmitry Vyukov <dvyukov@google.com> --- include/linux/sched/sysctl.h | 7 ++++ kernel/sched/deadline.c | 61 +++++++++++++++++++++++++++++++++++++++++-- kernel/sysctl.c | 14 +++++++++ 3 files changed, 80 insertions(+), 2 deletions(-)
--- a/include/linux/sched/sysctl.h +++ b/include/linux/sched/sysctl.h @@ -56,6 +56,13 @@ int sched_proc_update_handler(struct ctl extern unsigned int sysctl_sched_rt_period; extern int sysctl_sched_rt_runtime; +extern unsigned int sysctl_sched_dl_period_max; +extern unsigned int sysctl_sched_dl_period_min; + +extern int sched_dl_period_handler(struct ctl_table *table, int write, + void __user *buffer, size_t *lenp, + loff_t *ppos); + #ifdef CONFIG_UCLAMP_TASK extern unsigned int sysctl_sched_uclamp_util_min; extern unsigned int sysctl_sched_uclamp_util_max; --- a/kernel/sched/deadline.c +++ b/kernel/sched/deadline.c @@ -2660,6 +2660,52 @@ void __getparam_dl(struct task_struct *p } /* + * Default limits for DL period; on the top end we guard against small util + * tasks still getting ridiculous long effective runtimes, on the bottom end we + * guard against timer DoS. + */ +unsigned int sysctl_sched_dl_period_max = 1 << 22; /* ~4 seconds */ +unsigned int sysctl_sched_dl_period_min = 100; /* 100 us */ + +int sched_dl_period_handler(struct ctl_table *table, int write, + void __user *buffer, size_t *lenp, + loff_t *ppos) +{ + unsigned int old_min, old_max; + static DEFINE_MUTEX(mutex); + int ret; + + mutex_lock(&mutex); + old_min = sysctl_sched_dl_period_min; + old_max = sysctl_sched_dl_period_max; + + ret = proc_dointvec(table, write, buffer, lenp, ppos); + + if (!ret && write) { + u64 min = (u64)sysctl_sched_dl_period_min * NSEC_PER_USEC; + u64 max = (u64)sysctl_sched_dl_period_man * NSEC_PER_USEC; + + ret = -EINVAL; + if (min < 1ULL << DL_SCALE) + goto undo; + + if (max < min) + goto undo; + + ret = 0; + } + + if (0) { +undo: + sysctl_sched_rt_period = old_period; + sysctl_sched_rt_runtime = old_runtime; + } + mutex_unlock(&mutex); + + return ret; +} + +/* * This function validates the new parameters of a -deadline task. * We ask for the deadline not being zero, and greater or equal * than the runtime, as well as the period of being zero or @@ -2671,6 +2717,8 @@ void __getparam_dl(struct task_struct *p */ bool __checkparam_dl(const struct sched_attr *attr) { + u64 period, max, min; + /* special dl tasks don't actually use any parameter */ if (attr->sched_flags & SCHED_FLAG_SUGOV) return true; @@ -2694,12 +2742,21 @@ bool __checkparam_dl(const struct sched_ attr->sched_period & (1ULL << 63)) return false; + period = attr->sched_period; + if (!period) + period = attr->sched_deadline; + /* runtime <= deadline <= period (if period != 0) */ - if ((attr->sched_period != 0 && - attr->sched_period < attr->sched_deadline) || + if (period < attr->sched_deadline || attr->sched_deadline < attr->sched_runtime) return false; + max = (u64)READ_ONCE(sysctl_sched_dl_period_max) * NSEC_PER_USEC; + min = (u64)READ_ONCE(sysctl_sched_dl_period_min) * NSEC_PER_USEC; + + if (period < min || period > max) + return false; + return true; } --- a/kernel/sysctl.c +++ b/kernel/sysctl.c @@ -443,6 +443,20 @@ static struct ctl_table kern_table[] = { .proc_handler = sched_rt_handler, }, { + .procname = "sched_deadline_period_max_us", + .data = &sysctl_sched_dl_period_max, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = sched_dl_period_handler, + }, + { + .procname = "sched_deadline_period_min_us", + .data = &sysctl_sched_dl_period_min, + .maxlen = sizeof(unsigned int), + .mode = 0644, + .proc_handler = sched_dl_period_handler, + }, + { .procname = "sched_rr_timeslice_ms", .data = &sysctl_sched_rr_timeslice, .maxlen = sizeof(int),
|  |