| Message ID | 20240712005629.2980-1-inwardvessel@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | srcu: faster srcu gp seq wrap-around | expand |
Hi, On Thu, Jul 11, 2024 at 05:56:29PM -0700, JP Kobryn wrote: > Using a higher value for the initial gp sequence counter values allows for > wrapping to occur faster. It can help with surfacing any issues that may be > happening as a result of the wrap around. > > Signed-off-by: JP Kobryn <inwardvessel@gmail.com> > --- > include/linux/rcupdate.h | 3 +++ > include/linux/srcutree.h | 15 ++++++++++++++- > kernel/rcu/rcu.h | 3 --- > kernel/rcu/srcutree.c | 6 ++++-- > 4 files changed, 21 insertions(+), 6 deletions(-) > > diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h > index dfd2399f2cde..c148674b9e09 100644 > --- a/include/linux/rcupdate.h > +++ b/include/linux/rcupdate.h > @@ -35,6 +35,9 @@ > #define ULONG_CMP_GE(a, b) (ULONG_MAX / 2 >= (a) - (b)) > #define ULONG_CMP_LT(a, b) (ULONG_MAX / 2 < (a) - (b)) > > +#define RCU_SEQ_CTR_SHIFT 2 > +#define RCU_SEQ_STATE_MASK ((1 << RCU_SEQ_CTR_SHIFT) - 1) > + > /* Exported common interfaces */ > void call_rcu(struct rcu_head *head, rcu_callback_t func); > void rcu_barrier_tasks(void); > diff --git a/include/linux/srcutree.h b/include/linux/srcutree.h > index 8f3f72480e78..ed57598394de 100644 > --- a/include/linux/srcutree.h > +++ b/include/linux/srcutree.h > @@ -129,10 +129,23 @@ struct srcu_struct { > #define SRCU_STATE_SCAN1 1 > #define SRCU_STATE_SCAN2 2 > > +/* > + * Values for initializing gp sequence fields. Higher values allow wrap arounds to > + * occur earlier. > + * The second value with state is useful in the case of static initialization of > + * srcu_usage where srcu_gp_seq_needed is expected to have some state value in its > + * lower bits (or else it will appear to be already initialized within > + * the call check_init_srcu_struct()). > + */ > +#define SRCU_GP_SEQ_INITIAL_VAL ((0UL - 100UL) << RCU_SEQ_CTR_SHIFT) > +#define SRCU_GP_SEQ_INITIAL_VAL_WITH_STATE (SRCU_GP_SEQ_INITIAL_VAL - 1) > + > #define __SRCU_USAGE_INIT(name) \ > { \ > .lock = __SPIN_LOCK_UNLOCKED(name.lock), \ > - .srcu_gp_seq_needed = -1UL, \ As srcu_gp_seq* counters are initialized before first use, if we do not modify this file, we can move the SRCU_GP_SEQ_INITIAL_VAL and SRCU_GP_SEQ_INITIAL_VAL_WITH_STATE macros to srcutree.c. We then also don't need to move RCU_SEQ_CTR_SHIFT and RCU_SEQ_STATE_MASK from rcu.h (keeping them internal to RCU). Is there a scenario where that won't work? > + .srcu_gp_seq = SRCU_GP_SEQ_INITIAL_VAL, \ > + .srcu_gp_seq_needed = SRCU_GP_SEQ_INITIAL_VAL_WITH_STATE, \ > + .srcu_gp_seq_needed_exp = SRCU_GP_SEQ_INITIAL_VAL, \ > .work = __DELAYED_WORK_INITIALIZER(name.work, NULL, 0), \ > } > > diff --git a/kernel/rcu/rcu.h b/kernel/rcu/rcu.h > index 38238e595a61..2bfed9855d67 100644 > --- a/kernel/rcu/rcu.h > +++ b/kernel/rcu/rcu.h > @@ -54,9 +54,6 @@ > * grace-period sequence number. > */ > > -#define RCU_SEQ_CTR_SHIFT 2 > -#define RCU_SEQ_STATE_MASK ((1 << RCU_SEQ_CTR_SHIFT) - 1) > - > /* Low-order bit definition for polled grace-period APIs. */ > #define RCU_GET_STATE_COMPLETED 0x1 > > diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c > index bc4b58b0204e..2209dd0589de 100644 > --- a/kernel/rcu/srcutree.c > +++ b/kernel/rcu/srcutree.c > @@ -247,7 +247,7 @@ static int init_srcu_struct_fields(struct srcu_struct *ssp, bool is_static) > mutex_init(&ssp->srcu_sup->srcu_cb_mutex); > mutex_init(&ssp->srcu_sup->srcu_gp_mutex); > ssp->srcu_idx = 0; > - ssp->srcu_sup->srcu_gp_seq = 0; > + ssp->srcu_sup->srcu_gp_seq = SRCU_GP_SEQ_INITIAL_VAL; > ssp->srcu_sup->srcu_barrier_seq = 0; > mutex_init(&ssp->srcu_sup->srcu_barrier_mutex); > atomic_set(&ssp->srcu_sup->srcu_barrier_cpu_cnt, 0); > @@ -258,7 +258,7 @@ static int init_srcu_struct_fields(struct srcu_struct *ssp, bool is_static) > if (!ssp->sda) > goto err_free_sup; > init_srcu_struct_data(ssp); > - ssp->srcu_sup->srcu_gp_seq_needed_exp = 0; > + ssp->srcu_sup->srcu_gp_seq_needed_exp = SRCU_GP_SEQ_INITIAL_VAL; > ssp->srcu_sup->srcu_last_gp_end = ktime_get_mono_fast_ns(); > if (READ_ONCE(ssp->srcu_sup->srcu_size_state) == SRCU_SIZE_SMALL && SRCU_SIZING_IS_INIT()) { > if (!init_srcu_struct_nodes(ssp, GFP_ATOMIC)) > @@ -267,6 +267,8 @@ static int init_srcu_struct_fields(struct srcu_struct *ssp, bool is_static) > } > ssp->srcu_sup->srcu_ssp = ssp; > smp_store_release(&ssp->srcu_sup->srcu_gp_seq_needed, 0); /* Init done. */ Remove this line? - Neeraj > + smp_store_release(&ssp->srcu_sup->srcu_gp_seq_needed, > + SRCU_GP_SEQ_INITIAL_VAL); /* Init done. */ > return 0; > > err_free_sda: > -- > 2.45.2
diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h index dfd2399f2cde..c148674b9e09 100644 --- a/include/linux/rcupdate.h +++ b/include/linux/rcupdate.h @@ -35,6 +35,9 @@ #define ULONG_CMP_GE(a, b) (ULONG_MAX / 2 >= (a) - (b)) #define ULONG_CMP_LT(a, b) (ULONG_MAX / 2 < (a) - (b)) +#define RCU_SEQ_CTR_SHIFT 2 +#define RCU_SEQ_STATE_MASK ((1 << RCU_SEQ_CTR_SHIFT) - 1) + /* Exported common interfaces */ void call_rcu(struct rcu_head *head, rcu_callback_t func); void rcu_barrier_tasks(void); diff --git a/include/linux/srcutree.h b/include/linux/srcutree.h index 8f3f72480e78..ed57598394de 100644 --- a/include/linux/srcutree.h +++ b/include/linux/srcutree.h @@ -129,10 +129,23 @@ struct srcu_struct { #define SRCU_STATE_SCAN1 1 #define SRCU_STATE_SCAN2 2 +/* + * Values for initializing gp sequence fields. Higher values allow wrap arounds to + * occur earlier. + * The second value with state is useful in the case of static initialization of + * srcu_usage where srcu_gp_seq_needed is expected to have some state value in its + * lower bits (or else it will appear to be already initialized within + * the call check_init_srcu_struct()). + */ +#define SRCU_GP_SEQ_INITIAL_VAL ((0UL - 100UL) << RCU_SEQ_CTR_SHIFT) +#define SRCU_GP_SEQ_INITIAL_VAL_WITH_STATE (SRCU_GP_SEQ_INITIAL_VAL - 1) + #define __SRCU_USAGE_INIT(name) \ { \ .lock = __SPIN_LOCK_UNLOCKED(name.lock), \ - .srcu_gp_seq_needed = -1UL, \ + .srcu_gp_seq = SRCU_GP_SEQ_INITIAL_VAL, \ + .srcu_gp_seq_needed = SRCU_GP_SEQ_INITIAL_VAL_WITH_STATE, \ + .srcu_gp_seq_needed_exp = SRCU_GP_SEQ_INITIAL_VAL, \ .work = __DELAYED_WORK_INITIALIZER(name.work, NULL, 0), \ } diff --git a/kernel/rcu/rcu.h b/kernel/rcu/rcu.h index 38238e595a61..2bfed9855d67 100644 --- a/kernel/rcu/rcu.h +++ b/kernel/rcu/rcu.h @@ -54,9 +54,6 @@ * grace-period sequence number. */ -#define RCU_SEQ_CTR_SHIFT 2 -#define RCU_SEQ_STATE_MASK ((1 << RCU_SEQ_CTR_SHIFT) - 1) - /* Low-order bit definition for polled grace-period APIs. */ #define RCU_GET_STATE_COMPLETED 0x1 diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c index bc4b58b0204e..2209dd0589de 100644 --- a/kernel/rcu/srcutree.c +++ b/kernel/rcu/srcutree.c @@ -247,7 +247,7 @@ static int init_srcu_struct_fields(struct srcu_struct *ssp, bool is_static) mutex_init(&ssp->srcu_sup->srcu_cb_mutex); mutex_init(&ssp->srcu_sup->srcu_gp_mutex); ssp->srcu_idx = 0; - ssp->srcu_sup->srcu_gp_seq = 0; + ssp->srcu_sup->srcu_gp_seq = SRCU_GP_SEQ_INITIAL_VAL; ssp->srcu_sup->srcu_barrier_seq = 0; mutex_init(&ssp->srcu_sup->srcu_barrier_mutex); atomic_set(&ssp->srcu_sup->srcu_barrier_cpu_cnt, 0); @@ -258,7 +258,7 @@ static int init_srcu_struct_fields(struct srcu_struct *ssp, bool is_static) if (!ssp->sda) goto err_free_sup; init_srcu_struct_data(ssp); - ssp->srcu_sup->srcu_gp_seq_needed_exp = 0; + ssp->srcu_sup->srcu_gp_seq_needed_exp = SRCU_GP_SEQ_INITIAL_VAL; ssp->srcu_sup->srcu_last_gp_end = ktime_get_mono_fast_ns(); if (READ_ONCE(ssp->srcu_sup->srcu_size_state) == SRCU_SIZE_SMALL && SRCU_SIZING_IS_INIT()) { if (!init_srcu_struct_nodes(ssp, GFP_ATOMIC)) @@ -267,6 +267,8 @@ static int init_srcu_struct_fields(struct srcu_struct *ssp, bool is_static) } ssp->srcu_sup->srcu_ssp = ssp; smp_store_release(&ssp->srcu_sup->srcu_gp_seq_needed, 0); /* Init done. */ + smp_store_release(&ssp->srcu_sup->srcu_gp_seq_needed, + SRCU_GP_SEQ_INITIAL_VAL); /* Init done. */ return 0; err_free_sda:
Using a higher value for the initial gp sequence counter values allows for wrapping to occur faster. It can help with surfacing any issues that may be happening as a result of the wrap around. Signed-off-by: JP Kobryn <inwardvessel@gmail.com> --- include/linux/rcupdate.h | 3 +++ include/linux/srcutree.h | 15 ++++++++++++++- kernel/rcu/rcu.h | 3 --- kernel/rcu/srcutree.c | 6 ++++-- 4 files changed, 21 insertions(+), 6 deletions(-)