reset_all_ctrls() and resctrl_arch_update_domains() use on_each_cpu_mask()
to call rdt_ctrl_update() on potentially one CPU from each domain.
But this means rdt_ctrl_update() needs to figure out which domain to
apply changes to. Doing so requires a search of all domains in a resource,
which can only be done safely if cpus_lock is held. Both callers do hold
this lock, but there isn't a way for a function called on another CPU
via IPI to verify this.
Commit c0d848fcb09d ("x86/resctrl: Remove lockdep annotation that triggers
false positive") removed the incorrect assertions.
Add the target domain to the msr_param structure and
call rdt_ctrl_update() for each domain separately using
smp_call_function_single(). This means that rdt_ctrl_update() doesn't
need to search for the domain and get_domain_from_cpu() can safely assert
that the cpus_lock is held since the remaining callers do not use IPI.
Signed-off-by: Tony Luck <tony.luck@intel.com>
---
arch/x86/kernel/cpu/resctrl/internal.h | 2 ++
arch/x86/kernel/cpu/resctrl/core.c | 17 ++++------
arch/x86/kernel/cpu/resctrl/ctrlmondata.c | 40 +++++------------------
arch/x86/kernel/cpu/resctrl/rdtgroup.c | 14 +++-----
4 files changed, 21 insertions(+), 52 deletions(-)
diff --git a/arch/x86/kernel/cpu/resctrl/internal.h b/arch/x86/kernel/cpu/resctrl/internal.h
index c99f26ebe7a6..bc999471f072 100644
--- a/arch/x86/kernel/cpu/resctrl/internal.h
+++ b/arch/x86/kernel/cpu/resctrl/internal.h
@@ -378,11 +378,13 @@ static inline struct rdt_hw_domain *resctrl_to_arch_dom(struct rdt_domain *r)
/**
* struct msr_param - set a range of MSRs from a domain
* @res: The resource to use
+ * @dom: The domain to update
* @low: Beginning index from base MSR
* @high: End index
*/
struct msr_param {
struct rdt_resource *res;
+ struct rdt_domain *dom;
u32 low;
u32 high;
};
diff --git a/arch/x86/kernel/cpu/resctrl/core.c b/arch/x86/kernel/cpu/resctrl/core.c
index 83e40341583e..acf52aa185e0 100644
--- a/arch/x86/kernel/cpu/resctrl/core.c
+++ b/arch/x86/kernel/cpu/resctrl/core.c
@@ -362,6 +362,8 @@ struct rdt_domain *get_domain_from_cpu(int cpu, struct rdt_resource *r)
{
struct rdt_domain *d;
+ lockdep_assert_cpus_held();
+
list_for_each_entry(d, &r->domains, list) {
/* Find the domain that contains this CPU */
if (cpumask_test_cpu(cpu, &d->cpu_mask))
@@ -378,19 +380,11 @@ u32 resctrl_arch_get_num_closid(struct rdt_resource *r)
void rdt_ctrl_update(void *arg)
{
+ struct rdt_hw_resource *hw_res;
struct msr_param *m = arg;
- struct rdt_hw_resource *hw_res = resctrl_to_arch_res(m->res);
- struct rdt_resource *r = m->res;
- int cpu = smp_processor_id();
- struct rdt_domain *d;
- d = get_domain_from_cpu(cpu, r);
- if (d) {
- hw_res->msr_update(d, m, r);
- return;
- }
- pr_warn_once("cpu %d not found in any domain for resource %s\n",
- cpu, r->name);
+ hw_res = resctrl_to_arch_res(m->res);
+ hw_res->msr_update(m->dom, m, m->res);
}
/*
@@ -463,6 +457,7 @@ static int domain_setup_ctrlval(struct rdt_resource *r, struct rdt_domain *d)
hw_dom->ctrl_val = dc;
setup_default_ctrlval(r, dc);
+ m.dom = d;
m.low = 0;
m.high = hw_res->num_closid;
hw_res->msr_update(d, &m, r);
diff --git a/arch/x86/kernel/cpu/resctrl/ctrlmondata.c b/arch/x86/kernel/cpu/resctrl/ctrlmondata.c
index 7997b47743a2..a3a0fd80daa8 100644
--- a/arch/x86/kernel/cpu/resctrl/ctrlmondata.c
+++ b/arch/x86/kernel/cpu/resctrl/ctrlmondata.c
@@ -272,22 +272,6 @@ static u32 get_config_index(u32 closid, enum resctrl_conf_type type)
}
}
-static bool apply_config(struct rdt_hw_domain *hw_dom,
- struct resctrl_staged_config *cfg, u32 idx,
- cpumask_var_t cpu_mask)
-{
- struct rdt_domain *dom = &hw_dom->d_resctrl;
-
- if (cfg->new_ctrl != hw_dom->ctrl_val[idx]) {
- cpumask_set_cpu(cpumask_any(&dom->cpu_mask), cpu_mask);
- hw_dom->ctrl_val[idx] = cfg->new_ctrl;
-
- return true;
- }
-
- return false;
-}
-
int resctrl_arch_update_one(struct rdt_resource *r, struct rdt_domain *d,
u32 closid, enum resctrl_conf_type t, u32 cfg_val)
{
@@ -302,6 +286,7 @@ int resctrl_arch_update_one(struct rdt_resource *r, struct rdt_domain *d,
hw_dom->ctrl_val[idx] = cfg_val;
msr_param.res = r;
+ msr_param.dom = d;
msr_param.low = idx;
msr_param.high = idx + 1;
hw_res->msr_update(d, &msr_param, r);
@@ -315,27 +300,27 @@ int resctrl_arch_update_domains(struct rdt_resource *r, u32 closid)
struct rdt_hw_domain *hw_dom;
struct msr_param msr_param;
enum resctrl_conf_type t;
- cpumask_var_t cpu_mask;
struct rdt_domain *d;
+ int cpu;
u32 idx;
/* Walking r->domains, ensure it can't race with cpuhp */
lockdep_assert_cpus_held();
- if (!zalloc_cpumask_var(&cpu_mask, GFP_KERNEL))
- return -ENOMEM;
-
- msr_param.res = NULL;
list_for_each_entry(d, &r->domains, list) {
hw_dom = resctrl_to_arch_dom(d);
+ msr_param.res = NULL;
+ msr_param.dom = d;
for (t = 0; t < CDP_NUM_TYPES; t++) {
cfg = &hw_dom->d_resctrl.staged_config[t];
if (!cfg->have_new_ctrl)
continue;
idx = get_config_index(closid, t);
- if (!apply_config(hw_dom, cfg, idx, cpu_mask))
+ if (cfg->new_ctrl == hw_dom->ctrl_val[idx])
continue;
+ hw_dom->ctrl_val[idx] = cfg->new_ctrl;
+ cpu = cpumask_any(&d->cpu_mask);
if (!msr_param.res) {
msr_param.low = idx;
@@ -346,17 +331,10 @@ int resctrl_arch_update_domains(struct rdt_resource *r, u32 closid)
msr_param.high = max(msr_param.high, idx + 1);
}
}
+ if (msr_param.res)
+ smp_call_function_single(cpu, rdt_ctrl_update, &msr_param, 1);
}
- if (cpumask_empty(cpu_mask))
- goto done;
-
- /* Update resource control msr on all the CPUs. */
- on_each_cpu_mask(cpu_mask, rdt_ctrl_update, &msr_param, 1);
-
-done:
- free_cpumask_var(cpu_mask);
-
return 0;
}
diff --git a/arch/x86/kernel/cpu/resctrl/rdtgroup.c b/arch/x86/kernel/cpu/resctrl/rdtgroup.c
index 011e17efb1a6..da4f13db4161 100644
--- a/arch/x86/kernel/cpu/resctrl/rdtgroup.c
+++ b/arch/x86/kernel/cpu/resctrl/rdtgroup.c
@@ -2813,16 +2813,13 @@ static int reset_all_ctrls(struct rdt_resource *r)
struct rdt_hw_resource *hw_res = resctrl_to_arch_res(r);
struct rdt_hw_domain *hw_dom;
struct msr_param msr_param;
- cpumask_var_t cpu_mask;
struct rdt_domain *d;
+ int cpu;
int i;
/* Walking r->domains, ensure it can't race with cpuhp */
lockdep_assert_cpus_held();
- if (!zalloc_cpumask_var(&cpu_mask, GFP_KERNEL))
- return -ENOMEM;
-
msr_param.res = r;
msr_param.low = 0;
msr_param.high = hw_res->num_closid;
@@ -2834,17 +2831,14 @@ static int reset_all_ctrls(struct rdt_resource *r)
*/
list_for_each_entry(d, &r->domains, list) {
hw_dom = resctrl_to_arch_dom(d);
- cpumask_set_cpu(cpumask_any(&d->cpu_mask), cpu_mask);
+ cpu = cpumask_any(&d->cpu_mask);
for (i = 0; i < hw_res->num_closid; i++)
hw_dom->ctrl_val[i] = r->default_ctrl;
+ msr_param.dom = d;
+ smp_call_function_single(cpu, rdt_ctrl_update, &msr_param, 1);
}
- /* Update CBM on all the CPUs in cpu_mask */
- on_each_cpu_mask(cpu_mask, rdt_ctrl_update, &msr_param, 1);
-
- free_cpumask_var(cpu_mask);
-
return 0;
}
--
2.43.0Hi Tony,
On 28/02/2024 19:36, Tony Luck wrote:
> reset_all_ctrls() and resctrl_arch_update_domains() use on_each_cpu_mask()
> to call rdt_ctrl_update() on potentially one CPU from each domain.
>
> But this means rdt_ctrl_update() needs to figure out which domain to
> apply changes to. Doing so requires a search of all domains in a resource,
> which can only be done safely if cpus_lock is held. Both callers do hold
> this lock, but there isn't a way for a function called on another CPU
> via IPI to verify this.
>
> Commit c0d848fcb09d ("x86/resctrl: Remove lockdep annotation that triggers
> false positive") removed the incorrect assertions.
>
> Add the target domain to the msr_param structure and
> call rdt_ctrl_update() for each domain separately using
> smp_call_function_single(). This means that rdt_ctrl_update() doesn't
> need to search for the domain and get_domain_from_cpu() can safely assert
> that the cpus_lock is held since the remaining callers do not use IPI.
> diff --git a/arch/x86/kernel/cpu/resctrl/core.c b/arch/x86/kernel/cpu/resctrl/core.c
> index 83e40341583e..acf52aa185e0 100644
> --- a/arch/x86/kernel/cpu/resctrl/core.c
> +++ b/arch/x86/kernel/cpu/resctrl/core.c
> @@ -362,6 +362,8 @@ struct rdt_domain *get_domain_from_cpu(int cpu, struct rdt_resource *r)
> {
> struct rdt_domain *d;
>
> + lockdep_assert_cpus_held();
(Huzzah!)
> list_for_each_entry(d, &r->domains, list) {
> /* Find the domain that contains this CPU */
> if (cpumask_test_cpu(cpu, &d->cpu_mask))
> diff --git a/arch/x86/kernel/cpu/resctrl/ctrlmondata.c b/arch/x86/kernel/cpu/resctrl/ctrlmondata.c
> index 7997b47743a2..a3a0fd80daa8 100644
> --- a/arch/x86/kernel/cpu/resctrl/ctrlmondata.c
> +++ b/arch/x86/kernel/cpu/resctrl/ctrlmondata.c
> @@ -315,27 +300,27 @@ int resctrl_arch_update_domains(struct rdt_resource *r, u32 closid)
[..]
> - msr_param.res = NULL;
> list_for_each_entry(d, &r->domains, list) {
> hw_dom = resctrl_to_arch_dom(d);
> + msr_param.res = NULL;
> + msr_param.dom = d;
> for (t = 0; t < CDP_NUM_TYPES; t++) {
> cfg = &hw_dom->d_resctrl.staged_config[t];
> if (!cfg->have_new_ctrl)
> continue;
>
> idx = get_config_index(closid, t);
> - if (!apply_config(hw_dom, cfg, idx, cpu_mask))
> + if (cfg->new_ctrl == hw_dom->ctrl_val[idx])
> continue;
> + hw_dom->ctrl_val[idx] = cfg->new_ctrl;
> + cpu = cpumask_any(&d->cpu_mask);
If this CPU is part of the domain, then cpumask_any() could chose a different CPU leading
to an IPI, where this CPU could have done the work itself.
smp_call_function_any() has the logic to try and run on this CPU if possible.
As d->cpu_mask is still valid when you call smp_call_function_single(), could we use
smp_call_function_any() instead?
> if (!msr_param.res) {
> msr_param.low = idx;
> @@ -346,17 +331,10 @@ int resctrl_arch_update_domains(struct rdt_resource *r, u32 closid)
> msr_param.high = max(msr_param.high, idx + 1);
> }
> }
> + if (msr_param.res)
> + smp_call_function_single(cpu, rdt_ctrl_update, &msr_param, 1);
> }
> diff --git a/arch/x86/kernel/cpu/resctrl/rdtgroup.c b/arch/x86/kernel/cpu/resctrl/rdtgroup.c
> index 011e17efb1a6..da4f13db4161 100644
> --- a/arch/x86/kernel/cpu/resctrl/rdtgroup.c
> +++ b/arch/x86/kernel/cpu/resctrl/rdtgroup.c
> @@ -2813,16 +2813,13 @@ static int reset_all_ctrls(struct rdt_resource *r)
> @@ -2834,17 +2831,14 @@ static int reset_all_ctrls(struct rdt_resource *r)
> */
> list_for_each_entry(d, &r->domains, list) {
> hw_dom = resctrl_to_arch_dom(d);
> - cpumask_set_cpu(cpumask_any(&d->cpu_mask), cpu_mask);
> + cpu = cpumask_any(&d->cpu_mask);
Same comment about picking an unlucky CPU here, smp_call_function_any() can improve our
chances.
> for (i = 0; i < hw_res->num_closid; i++)
> hw_dom->ctrl_val[i] = r->default_ctrl;
> + msr_param.dom = d;
> + smp_call_function_single(cpu, rdt_ctrl_update, &msr_param, 1);
> }
Still:
Reviewed-by: James Morse <james.morse@arm.com>
Thanks!
James
> > + hw_dom->ctrl_val[idx] = cfg->new_ctrl; > > + cpu = cpumask_any(&d->cpu_mask); > > If this CPU is part of the domain, then cpumask_any() could chose a different CPU leading > to an IPI, where this CPU could have done the work itself. > smp_call_function_any() has the logic to try and run on this CPU if possible. > > As d->cpu_mask is still valid when you call smp_call_function_single(), could we use > smp_call_function_any() instead? I'd thought that cpumask_any() was smarter. But I see that it dumbly just picks the first CPU in the mask. Great suggestion, will add to next version. >> - cpumask_set_cpu(cpumask_any(&d->cpu_mask), cpu_mask); >> + cpu = cpumask_any(&d->cpu_mask); > > Same comment about picking an unlucky CPU here, smp_call_function_any() can improve our > chances. Will fix here too. > Still: > Reviewed-by: James Morse <james.morse@arm.com> Thanks! -Tony
Hi Tony,
On 2/28/2024 11:36 AM, Tony Luck wrote:
> reset_all_ctrls() and resctrl_arch_update_domains() use on_each_cpu_mask()
> to call rdt_ctrl_update() on potentially one CPU from each domain.
>
> But this means rdt_ctrl_update() needs to figure out which domain to
> apply changes to. Doing so requires a search of all domains in a resource,
> which can only be done safely if cpus_lock is held. Both callers do hold
> this lock, but there isn't a way for a function called on another CPU
> via IPI to verify this.
>
> Commit c0d848fcb09d ("x86/resctrl: Remove lockdep annotation that triggers
> false positive") removed the incorrect assertions.
>
> Add the target domain to the msr_param structure and
> call rdt_ctrl_update() for each domain separately using
> smp_call_function_single(). This means that rdt_ctrl_update() doesn't
> need to search for the domain and get_domain_from_cpu() can safely assert
> that the cpus_lock is held since the remaining callers do not use IPI.
>
> Signed-off-by: Tony Luck <tony.luck@intel.com>
> ---
...
> @@ -315,27 +300,27 @@ int resctrl_arch_update_domains(struct rdt_resource *r, u32 closid)
> struct rdt_hw_domain *hw_dom;
> struct msr_param msr_param;
> enum resctrl_conf_type t;
> - cpumask_var_t cpu_mask;
> struct rdt_domain *d;
> + int cpu;
> u32 idx;
>
> /* Walking r->domains, ensure it can't race with cpuhp */
> lockdep_assert_cpus_held();
>
> - if (!zalloc_cpumask_var(&cpu_mask, GFP_KERNEL))
> - return -ENOMEM;
> -
> - msr_param.res = NULL;
> list_for_each_entry(d, &r->domains, list) {
> hw_dom = resctrl_to_arch_dom(d);
> + msr_param.res = NULL;
> + msr_param.dom = d;
> for (t = 0; t < CDP_NUM_TYPES; t++) {
> cfg = &hw_dom->d_resctrl.staged_config[t];
> if (!cfg->have_new_ctrl)
> continue;
>
> idx = get_config_index(closid, t);
> - if (!apply_config(hw_dom, cfg, idx, cpu_mask))
> + if (cfg->new_ctrl == hw_dom->ctrl_val[idx])
> continue;
> + hw_dom->ctrl_val[idx] = cfg->new_ctrl;
> + cpu = cpumask_any(&d->cpu_mask);
>
cpu only needs to be assigned once. How about initializing cpu to
nr_cpu_ids at the same time msr_param.res and msr_param.dom is
initialized and only assign it when msr_param.res is assigned?
I think that will be more robust.
If you agree and do this then please feel free to add:
Reviewed-by: Reinette Chatre <reinette.chatre@intel.com>
Reinette
> > + cpu = cpumask_any(&d->cpu_mask); > > > > cpu only needs to be assigned once. How about initializing cpu to > nr_cpu_ids at the same time msr_param.res and msr_param.dom is > initialized and only assign it when msr_param.res is assigned? > I think that will be more robust. > > If you agree and do this then please feel free to add: > Reviewed-by: Reinette Chatre <reinette.chatre@intel.com> Reinette, I agree. I'll move the assignment to "cpu" outside the CDP_NUM_TYPES loop. Thanks for the review. -Tony
Hi Tony, On 3/4/2024 4:17 PM, Luck, Tony wrote: >>> + cpu = cpumask_any(&d->cpu_mask); >>> >> >> cpu only needs to be assigned once. How about initializing cpu to >> nr_cpu_ids at the same time msr_param.res and msr_param.dom is >> initialized and only assign it when msr_param.res is assigned? >> I think that will be more robust. >> >> If you agree and do this then please feel free to add: >> Reviewed-by: Reinette Chatre <reinette.chatre@intel.com> > > Reinette, > > I agree. I'll move the assignment to "cpu" outside the CDP_NUM_TYPES > loop. > If I understand correctly that would always look for a valid "cpu" even when none is needed. Not quite what I proposed but should work. Just some wasted cycles in a non critical path. Reinette
> If I understand correctly that would always look for a valid "cpu"
> even when none is needed. Not quite what I proposed but should
> work. Just some wasted cycles in a non critical path.
Reinette,
Sorry for misunderstanding. You are right. Assignment to msr_param.dom
can also be deferred to the same point. Like this:
list_for_each_entry(d, &r->domains, list) {
hw_dom = resctrl_to_arch_dom(d);
msr_param.res = NULL;
for (t = 0; t < CDP_NUM_TYPES; t++) {
cfg = &hw_dom->d_resctrl.staged_config[t];
if (!cfg->have_new_ctrl)
continue;
idx = get_config_index(closid, t);
if (cfg->new_ctrl == hw_dom->ctrl_val[idx])
continue;
hw_dom->ctrl_val[idx] = cfg->new_ctrl;
if (!msr_param.res) {
msr_param.low = idx;
msr_param.high = msr_param.low + 1;
msr_param.res = r;
msr_param.dom = d;
cpu = cpumask_any(&d->cpu_mask);
} else {
msr_param.low = min(msr_param.low, idx);
msr_param.high = max(msr_param.high, idx + 1);
}
}
if (msr_param.res)
smp_call_function_single(cpu, rdt_ctrl_update, &msr_param, 1);
}
-Tony
Hi Tony,
On 3/5/2024 8:37 AM, Luck, Tony wrote:
>> If I understand correctly that would always look for a valid "cpu"
>> even when none is needed. Not quite what I proposed but should
>> work. Just some wasted cycles in a non critical path.
>
> Reinette,
>
> Sorry for misunderstanding. You are right. Assignment to msr_param.dom
> can also be deferred to the same point. Like this:
>
> list_for_each_entry(d, &r->domains, list) {
> hw_dom = resctrl_to_arch_dom(d);
> msr_param.res = NULL;
> for (t = 0; t < CDP_NUM_TYPES; t++) {
> cfg = &hw_dom->d_resctrl.staged_config[t];
> if (!cfg->have_new_ctrl)
> continue;
>
> idx = get_config_index(closid, t);
> if (cfg->new_ctrl == hw_dom->ctrl_val[idx])
> continue;
> hw_dom->ctrl_val[idx] = cfg->new_ctrl;
>
> if (!msr_param.res) {
> msr_param.low = idx;
> msr_param.high = msr_param.low + 1;
> msr_param.res = r;
> msr_param.dom = d;
> cpu = cpumask_any(&d->cpu_mask);
> } else {
> msr_param.low = min(msr_param.low, idx);
> msr_param.high = max(msr_param.high, idx + 1);
> }
> }
> if (msr_param.res)
> smp_call_function_single(cpu, rdt_ctrl_update, &msr_param, 1);
> }
This looks good to me.
Thank you very much.
Reinette
© 2016 - 2026 Red Hat, Inc.