Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Alexei Starovoitov]

On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> RCU data structure operations with respect to active programs. For
> example, userspace can update a map->map entry to point to a new map,
> use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> complete, and then drain the old map without fear that BPF programs
> may still be updating it.
> 
> Signed-off-by: Daniel Colascione <dancol@google.com>
> ---
>  include/uapi/linux/bpf.h |  1 +
>  kernel/bpf/syscall.c     | 14 ++++++++++++++
>  2 files changed, 15 insertions(+)
> 
> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> index b7db3261c62d..4365c50e8055 100644
> --- a/include/uapi/linux/bpf.h
> +++ b/include/uapi/linux/bpf.h
> @@ -98,6 +98,7 @@ enum bpf_cmd {
>  	BPF_BTF_LOAD,
>  	BPF_BTF_GET_FD_BY_ID,
>  	BPF_TASK_FD_QUERY,
> +	BPF_SYNCHRONIZE,
>  };
>  
>  enum bpf_map_type {
> diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> index d10ecd78105f..60ec7811846e 100644
> --- a/kernel/bpf/syscall.c
> +++ b/kernel/bpf/syscall.c
> @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, siz
>  	if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
>  		return -EPERM;
>  
> +	if (cmd == BPF_SYNCHRONIZE) {
> +		if (uattr != NULL || size != 0)
> +			return -EINVAL;
> +		err = security_bpf(cmd, NULL, 0);
> +		if (err < 0)
> +			return err;
> +		/* BPF programs are run with preempt disabled, so
> +		 * synchronize_sched is sufficient even with
> +		 * RCU_PREEMPT.
> +		 */
> +		synchronize_sched();
> +		return 0;

I don't think it's necessary. sys_membarrier() can do this already
and some folks use it exactly for this use case.

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Daniel Colascione]

On Fri, Jul 6, 2018 at 7:54 PM, Alexei Starovoitov
<alexei.starovoitov@gmail.com> wrote:
> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
>> BPF_SYNCHRONIZE waits for any BPF programs active at the time of
>> BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
>> RCU data structure operations with respect to active programs. For
>> example, userspace can update a map->map entry to point to a new map,
>> use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
>> complete, and then drain the old map without fear that BPF programs
>> may still be updating it.
>>
>> Signed-off-by: Daniel Colascione <dancol@google.com>
>> ---
>>  include/uapi/linux/bpf.h |  1 +
>>  kernel/bpf/syscall.c     | 14 ++++++++++++++
>>  2 files changed, 15 insertions(+)
>>
>> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
>> index b7db3261c62d..4365c50e8055 100644
>> --- a/include/uapi/linux/bpf.h
>> +++ b/include/uapi/linux/bpf.h
>> @@ -98,6 +98,7 @@ enum bpf_cmd {
>>       BPF_BTF_LOAD,
>>       BPF_BTF_GET_FD_BY_ID,
>>       BPF_TASK_FD_QUERY,
>> +     BPF_SYNCHRONIZE,
>>  };
>>
>>  enum bpf_map_type {
>> diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
>> index d10ecd78105f..60ec7811846e 100644
>> --- a/kernel/bpf/syscall.c
>> +++ b/kernel/bpf/syscall.c
>> @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, siz
>>       if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
>>               return -EPERM;
>>
>> +     if (cmd == BPF_SYNCHRONIZE) {
>> +             if (uattr != NULL || size != 0)
>> +                     return -EINVAL;
>> +             err = security_bpf(cmd, NULL, 0);
>> +             if (err < 0)
>> +                     return err;
>> +             /* BPF programs are run with preempt disabled, so
>> +              * synchronize_sched is sufficient even with
>> +              * RCU_PREEMPT.
>> +              */
>> +             synchronize_sched();
>> +             return 0;
>
> I don't think it's necessary. sys_membarrier() can do this already
> and some folks use it exactly for this use case.

True --- implementation-wise, today. The point of the patch is to sort
out contractual guarantees. membarrier isn't documented to have the
BPF_SYNCHRONIZE effect right now, although it does. If we want to
extend membarrier's contract, great, but that might make it more
expensive than necessary for callers who, well, just want a memory
barrier. It's worthwhile to have a dedicated BPF command for this
task, especially considering the simplicity of the implementation.

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Joel Fernandes]

On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> > RCU data structure operations with respect to active programs. For
> > example, userspace can update a map->map entry to point to a new map,
> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> > complete, and then drain the old map without fear that BPF programs
> > may still be updating it.
> > 
> > Signed-off-by: Daniel Colascione <dancol@google.com>
> > ---
> >  include/uapi/linux/bpf.h |  1 +
> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
> >  2 files changed, 15 insertions(+)
> > 
> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > index b7db3261c62d..4365c50e8055 100644
> > --- a/include/uapi/linux/bpf.h
> > +++ b/include/uapi/linux/bpf.h
> > @@ -98,6 +98,7 @@ enum bpf_cmd {
> >  	BPF_BTF_LOAD,
> >  	BPF_BTF_GET_FD_BY_ID,
> >  	BPF_TASK_FD_QUERY,
> > +	BPF_SYNCHRONIZE,
> >  };
> >  
> >  enum bpf_map_type {
> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> > index d10ecd78105f..60ec7811846e 100644
> > --- a/kernel/bpf/syscall.c
> > +++ b/kernel/bpf/syscall.c
> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, siz
> >  	if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
> >  		return -EPERM;
> >  
> > +	if (cmd == BPF_SYNCHRONIZE) {
> > +		if (uattr != NULL || size != 0)
> > +			return -EINVAL;
> > +		err = security_bpf(cmd, NULL, 0);
> > +		if (err < 0)
> > +			return err;
> > +		/* BPF programs are run with preempt disabled, so
> > +		 * synchronize_sched is sufficient even with
> > +		 * RCU_PREEMPT.
> > +		 */
> > +		synchronize_sched();
> > +		return 0;
> 
> I don't think it's necessary. sys_membarrier() can do this already
> and some folks use it exactly for this use case.

Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
though. No where does the manpage say membarrier should be implemented this
way so what happens if the implementation changes?

Further, membarrier manpage says that a memory barrier should be matched with
a matching barrier. In this use case there is no matching barrier, so it
makes it weirder.

Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
fragile to depend on it for this?

        case MEMBARRIER_CMD_GLOBAL:
                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
                if (tick_nohz_full_enabled())
                        return -EINVAL;
                if (num_online_cpus() > 1)
                        synchronize_sched();
                return 0;


Adding Mathieu as well who I believe is author/maintainer of membarrier.

thanks!

- Joel

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Mathieu Desnoyers]

----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:

> On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
>> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
>> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
>> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
>> > RCU data structure operations with respect to active programs. For
>> > example, userspace can update a map->map entry to point to a new map,
>> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
>> > complete, and then drain the old map without fear that BPF programs
>> > may still be updating it.
>> > 
>> > Signed-off-by: Daniel Colascione <dancol@google.com>
>> > ---
>> >  include/uapi/linux/bpf.h |  1 +
>> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
>> >  2 files changed, 15 insertions(+)
>> > 
>> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
>> > index b7db3261c62d..4365c50e8055 100644
>> > --- a/include/uapi/linux/bpf.h
>> > +++ b/include/uapi/linux/bpf.h
>> > @@ -98,6 +98,7 @@ enum bpf_cmd {
>> >  	BPF_BTF_LOAD,
>> >  	BPF_BTF_GET_FD_BY_ID,
>> >  	BPF_TASK_FD_QUERY,
>> > +	BPF_SYNCHRONIZE,
>> >  };
>> >  
>> >  enum bpf_map_type {
>> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
>> > index d10ecd78105f..60ec7811846e 100644
>> > --- a/kernel/bpf/syscall.c
>> > +++ b/kernel/bpf/syscall.c
>> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
>> > uattr, unsigned int, siz
>> >  	if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
>> >  		return -EPERM;
>> >  
>> > +	if (cmd == BPF_SYNCHRONIZE) {
>> > +		if (uattr != NULL || size != 0)
>> > +			return -EINVAL;
>> > +		err = security_bpf(cmd, NULL, 0);
>> > +		if (err < 0)
>> > +			return err;
>> > +		/* BPF programs are run with preempt disabled, so
>> > +		 * synchronize_sched is sufficient even with
>> > +		 * RCU_PREEMPT.
>> > +		 */
>> > +		synchronize_sched();
>> > +		return 0;
>> 
>> I don't think it's necessary. sys_membarrier() can do this already
>> and some folks use it exactly for this use case.
> 
> Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
> though. No where does the manpage say membarrier should be implemented this
> way so what happens if the implementation changes?
> 
> Further, membarrier manpage says that a memory barrier should be matched with
> a matching barrier. In this use case there is no matching barrier, so it
> makes it weirder.
> 
> Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
> fragile to depend on it for this?
> 
>        case MEMBARRIER_CMD_GLOBAL:
>                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
>                if (tick_nohz_full_enabled())
>                        return -EINVAL;
>                if (num_online_cpus() > 1)
>                        synchronize_sched();
>                return 0;
> 
> 
> Adding Mathieu as well who I believe is author/maintainer of membarrier.

See commit 907565337
"Fix: Disable sys_membarrier when nohz_full is enabled"

"Userspace applications should be allowed to expect the membarrier system
call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
nohz_full CPUs, but synchronize_sched() does not take those into
account."

So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
only care about kernel preempt off critical sections.

Clearly bpf code does not run in user-space, so it would "work".

But the guarantees provided by membarrier are not to synchronize against
preempt off per se. It's just that the current implementation happens to
do that. The point of membarrier is to turn user-space memory barriers
into compiler barriers.

If what you need is to wait for a RCU grace period for whatever RCU flavor
ebpf is using, I would against using membarrier for this. I would rather
recommend adding a dedicated BPF_SYNCHRONIZE so you won't leak
implementation details to user-space, *and* you can eventually change you
RCU implementation for e.g. SRCU in the future if needed.

Thanks,

Mathieu

> 
> thanks!
> 
> - Joel

-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Alexei Starovoitov]

On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
> ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
> 
> > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
> >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> >> > RCU data structure operations with respect to active programs. For
> >> > example, userspace can update a map->map entry to point to a new map,
> >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> >> > complete, and then drain the old map without fear that BPF programs
> >> > may still be updating it.
> >> > 
> >> > Signed-off-by: Daniel Colascione <dancol@google.com>
> >> > ---
> >> >  include/uapi/linux/bpf.h |  1 +
> >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
> >> >  2 files changed, 15 insertions(+)
> >> > 
> >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> >> > index b7db3261c62d..4365c50e8055 100644
> >> > --- a/include/uapi/linux/bpf.h
> >> > +++ b/include/uapi/linux/bpf.h
> >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
> >> >  	BPF_BTF_LOAD,
> >> >  	BPF_BTF_GET_FD_BY_ID,
> >> >  	BPF_TASK_FD_QUERY,
> >> > +	BPF_SYNCHRONIZE,
> >> >  };
> >> >  
> >> >  enum bpf_map_type {
> >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> >> > index d10ecd78105f..60ec7811846e 100644
> >> > --- a/kernel/bpf/syscall.c
> >> > +++ b/kernel/bpf/syscall.c
> >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
> >> > uattr, unsigned int, siz
> >> >  	if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
> >> >  		return -EPERM;
> >> >  
> >> > +	if (cmd == BPF_SYNCHRONIZE) {
> >> > +		if (uattr != NULL || size != 0)
> >> > +			return -EINVAL;
> >> > +		err = security_bpf(cmd, NULL, 0);
> >> > +		if (err < 0)
> >> > +			return err;
> >> > +		/* BPF programs are run with preempt disabled, so
> >> > +		 * synchronize_sched is sufficient even with
> >> > +		 * RCU_PREEMPT.
> >> > +		 */
> >> > +		synchronize_sched();
> >> > +		return 0;
> >> 
> >> I don't think it's necessary. sys_membarrier() can do this already
> >> and some folks use it exactly for this use case.
> > 
> > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
> > though. No where does the manpage say membarrier should be implemented this
> > way so what happens if the implementation changes?
> > 
> > Further, membarrier manpage says that a memory barrier should be matched with
> > a matching barrier. In this use case there is no matching barrier, so it
> > makes it weirder.
> > 
> > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
> > fragile to depend on it for this?
> > 
> >        case MEMBARRIER_CMD_GLOBAL:
> >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
> >                if (tick_nohz_full_enabled())
> >                        return -EINVAL;
> >                if (num_online_cpus() > 1)
> >                        synchronize_sched();
> >                return 0;
> > 
> > 
> > Adding Mathieu as well who I believe is author/maintainer of membarrier.
> 
> See commit 907565337
> "Fix: Disable sys_membarrier when nohz_full is enabled"
> 
> "Userspace applications should be allowed to expect the membarrier system
> call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
> nohz_full CPUs, but synchronize_sched() does not take those into
> account."
> 
> So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
> only care about kernel preempt off critical sections.
> 
> Clearly bpf code does not run in user-space, so it would "work".
> 
> But the guarantees provided by membarrier are not to synchronize against
> preempt off per se. It's just that the current implementation happens to
> do that. The point of membarrier is to turn user-space memory barriers
> into compiler barriers.
> 
> If what you need is to wait for a RCU grace period for whatever RCU flavor
> ebpf is using, I would against using membarrier for this. I would rather
> recommend adding a dedicated BPF_SYNCHRONIZE so you won't leak
> implementation details to user-space, *and* you can eventually change you
> RCU implementation for e.g. SRCU in the future if needed.

The point about future changes to underlying bpf mechanisms is valid.
There is work already on the way to reduce the scope of preempt_off+rcu_lock
that currently lasts the whole prog. We will have new prog types that won't
have such wrappers and will do rcu_lock/unlock and preempt on/off only
when necessary.
So something like BPF_SYNCHRONIZE will break soon, since the kernel cannot have
guarantees on when programs finish. Calling this command BPF_SYNCHRONIZE_PROG
also won't make sense for the same reason.
What we can do it instead is to define synchronization barrier for
programs accessing maps. May be call it something like:
BPF_SYNC_MAP_ACCESS ?
uapi/bpf.h would need to have extensive comment what this barrier is doing.
Implementation should probably call synchronize_rcu() and not play games
with synchronize_sched(), since that's going too much into implementation.
Also should such sys_bpf command be root only?
I'm not sure whether dos attack can be made by spamming synchronize_rcu()
and synchronize_sched() for that matter.

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Mathieu Desnoyers]

----- On Jul 9, 2018, at 5:09 PM, Alexei Starovoitov alexei.starovoitov@gmail.com wrote:

> On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
>> ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
>> 
>> > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
>> >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
>> >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
>> >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
>> >> > RCU data structure operations with respect to active programs. For
>> >> > example, userspace can update a map->map entry to point to a new map,
>> >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
>> >> > complete, and then drain the old map without fear that BPF programs
>> >> > may still be updating it.
>> >> > 
>> >> > Signed-off-by: Daniel Colascione <dancol@google.com>
>> >> > ---
>> >> >  include/uapi/linux/bpf.h |  1 +
>> >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
>> >> >  2 files changed, 15 insertions(+)
>> >> > 
>> >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
>> >> > index b7db3261c62d..4365c50e8055 100644
>> >> > --- a/include/uapi/linux/bpf.h
>> >> > +++ b/include/uapi/linux/bpf.h
>> >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
>> >> >  	BPF_BTF_LOAD,
>> >> >  	BPF_BTF_GET_FD_BY_ID,
>> >> >  	BPF_TASK_FD_QUERY,
>> >> > +	BPF_SYNCHRONIZE,
>> >> >  };
>> >> >  
>> >> >  enum bpf_map_type {
>> >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
>> >> > index d10ecd78105f..60ec7811846e 100644
>> >> > --- a/kernel/bpf/syscall.c
>> >> > +++ b/kernel/bpf/syscall.c
>> >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
>> >> > uattr, unsigned int, siz
>> >> >  	if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
>> >> >  		return -EPERM;
>> >> >  
>> >> > +	if (cmd == BPF_SYNCHRONIZE) {
>> >> > +		if (uattr != NULL || size != 0)
>> >> > +			return -EINVAL;
>> >> > +		err = security_bpf(cmd, NULL, 0);
>> >> > +		if (err < 0)
>> >> > +			return err;
>> >> > +		/* BPF programs are run with preempt disabled, so
>> >> > +		 * synchronize_sched is sufficient even with
>> >> > +		 * RCU_PREEMPT.
>> >> > +		 */
>> >> > +		synchronize_sched();
>> >> > +		return 0;
>> >> 
>> >> I don't think it's necessary. sys_membarrier() can do this already
>> >> and some folks use it exactly for this use case.
>> > 
>> > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
>> > though. No where does the manpage say membarrier should be implemented this
>> > way so what happens if the implementation changes?
>> > 
>> > Further, membarrier manpage says that a memory barrier should be matched with
>> > a matching barrier. In this use case there is no matching barrier, so it
>> > makes it weirder.
>> > 
>> > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
>> > fragile to depend on it for this?
>> > 
>> >        case MEMBARRIER_CMD_GLOBAL:
>> >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
>> >                if (tick_nohz_full_enabled())
>> >                        return -EINVAL;
>> >                if (num_online_cpus() > 1)
>> >                        synchronize_sched();
>> >                return 0;
>> > 
>> > 
>> > Adding Mathieu as well who I believe is author/maintainer of membarrier.
>> 
>> See commit 907565337
>> "Fix: Disable sys_membarrier when nohz_full is enabled"
>> 
>> "Userspace applications should be allowed to expect the membarrier system
>> call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
>> nohz_full CPUs, but synchronize_sched() does not take those into
>> account."
>> 
>> So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
>> only care about kernel preempt off critical sections.
>> 
>> Clearly bpf code does not run in user-space, so it would "work".
>> 
>> But the guarantees provided by membarrier are not to synchronize against
>> preempt off per se. It's just that the current implementation happens to
>> do that. The point of membarrier is to turn user-space memory barriers
>> into compiler barriers.
>> 
>> If what you need is to wait for a RCU grace period for whatever RCU flavor
>> ebpf is using, I would against using membarrier for this. I would rather
>> recommend adding a dedicated BPF_SYNCHRONIZE so you won't leak
>> implementation details to user-space, *and* you can eventually change you
>> RCU implementation for e.g. SRCU in the future if needed.
> 
> The point about future changes to underlying bpf mechanisms is valid.
> There is work already on the way to reduce the scope of preempt_off+rcu_lock
> that currently lasts the whole prog. We will have new prog types that won't
> have such wrappers and will do rcu_lock/unlock and preempt on/off only
> when necessary.
> So something like BPF_SYNCHRONIZE will break soon, since the kernel cannot have
> guarantees on when programs finish. Calling this command BPF_SYNCHRONIZE_PROG
> also won't make sense for the same reason.
> What we can do it instead is to define synchronization barrier for
> programs accessing maps. May be call it something like:
> BPF_SYNC_MAP_ACCESS ?
> uapi/bpf.h would need to have extensive comment what this barrier is doing.
> Implementation should probably call synchronize_rcu() and not play games
> with synchronize_sched(), since that's going too much into implementation.
> Also should such sys_bpf command be root only?
> I'm not sure whether dos attack can be made by spamming synchronize_rcu()
> and synchronize_sched() for that matter.

Adding Paul E. McKenney in CC. He may want to share his thoughts on the matter.

Thanks,

Mathieu


-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Paul E. McKenney]

On Mon, Jul 09, 2018 at 05:35:34PM -0400, Mathieu Desnoyers wrote:
> 
> 
> ----- On Jul 9, 2018, at 5:09 PM, Alexei Starovoitov alexei.starovoitov@gmail.com wrote:
> 
> > On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
> >> ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
> >> 
> >> > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
> >> >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> >> >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> >> >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> >> >> > RCU data structure operations with respect to active programs. For
> >> >> > example, userspace can update a map->map entry to point to a new map,
> >> >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> >> >> > complete, and then drain the old map without fear that BPF programs
> >> >> > may still be updating it.
> >> >> > 
> >> >> > Signed-off-by: Daniel Colascione <dancol@google.com>
> >> >> > ---
> >> >> >  include/uapi/linux/bpf.h |  1 +
> >> >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
> >> >> >  2 files changed, 15 insertions(+)
> >> >> > 
> >> >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> >> >> > index b7db3261c62d..4365c50e8055 100644
> >> >> > --- a/include/uapi/linux/bpf.h
> >> >> > +++ b/include/uapi/linux/bpf.h
> >> >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
> >> >> >  	BPF_BTF_LOAD,
> >> >> >  	BPF_BTF_GET_FD_BY_ID,
> >> >> >  	BPF_TASK_FD_QUERY,
> >> >> > +	BPF_SYNCHRONIZE,
> >> >> >  };
> >> >> >  
> >> >> >  enum bpf_map_type {
> >> >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> >> >> > index d10ecd78105f..60ec7811846e 100644
> >> >> > --- a/kernel/bpf/syscall.c
> >> >> > +++ b/kernel/bpf/syscall.c
> >> >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
> >> >> > uattr, unsigned int, siz
> >> >> >  	if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
> >> >> >  		return -EPERM;
> >> >> >  
> >> >> > +	if (cmd == BPF_SYNCHRONIZE) {
> >> >> > +		if (uattr != NULL || size != 0)
> >> >> > +			return -EINVAL;
> >> >> > +		err = security_bpf(cmd, NULL, 0);
> >> >> > +		if (err < 0)
> >> >> > +			return err;
> >> >> > +		/* BPF programs are run with preempt disabled, so
> >> >> > +		 * synchronize_sched is sufficient even with
> >> >> > +		 * RCU_PREEMPT.
> >> >> > +		 */
> >> >> > +		synchronize_sched();
> >> >> > +		return 0;
> >> >> 
> >> >> I don't think it's necessary. sys_membarrier() can do this already
> >> >> and some folks use it exactly for this use case.
> >> > 
> >> > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
> >> > though. No where does the manpage say membarrier should be implemented this
> >> > way so what happens if the implementation changes?
> >> > 
> >> > Further, membarrier manpage says that a memory barrier should be matched with
> >> > a matching barrier. In this use case there is no matching barrier, so it
> >> > makes it weirder.
> >> > 
> >> > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
> >> > fragile to depend on it for this?
> >> > 
> >> >        case MEMBARRIER_CMD_GLOBAL:
> >> >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
> >> >                if (tick_nohz_full_enabled())
> >> >                        return -EINVAL;
> >> >                if (num_online_cpus() > 1)
> >> >                        synchronize_sched();
> >> >                return 0;
> >> > 
> >> > 
> >> > Adding Mathieu as well who I believe is author/maintainer of membarrier.
> >> 
> >> See commit 907565337
> >> "Fix: Disable sys_membarrier when nohz_full is enabled"
> >> 
> >> "Userspace applications should be allowed to expect the membarrier system
> >> call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
> >> nohz_full CPUs, but synchronize_sched() does not take those into
> >> account."
> >> 
> >> So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
> >> only care about kernel preempt off critical sections.
> >> 
> >> Clearly bpf code does not run in user-space, so it would "work".
> >> 
> >> But the guarantees provided by membarrier are not to synchronize against
> >> preempt off per se. It's just that the current implementation happens to
> >> do that. The point of membarrier is to turn user-space memory barriers
> >> into compiler barriers.
> >> 
> >> If what you need is to wait for a RCU grace period for whatever RCU flavor
> >> ebpf is using, I would against using membarrier for this. I would rather
> >> recommend adding a dedicated BPF_SYNCHRONIZE so you won't leak
> >> implementation details to user-space, *and* you can eventually change you
> >> RCU implementation for e.g. SRCU in the future if needed.
> > 
> > The point about future changes to underlying bpf mechanisms is valid.
> > There is work already on the way to reduce the scope of preempt_off+rcu_lock
> > that currently lasts the whole prog. We will have new prog types that won't
> > have such wrappers and will do rcu_lock/unlock and preempt on/off only
> > when necessary.
> > So something like BPF_SYNCHRONIZE will break soon, since the kernel cannot have
> > guarantees on when programs finish. Calling this command BPF_SYNCHRONIZE_PROG
> > also won't make sense for the same reason.
> > What we can do it instead is to define synchronization barrier for
> > programs accessing maps. May be call it something like:
> > BPF_SYNC_MAP_ACCESS ?
> > uapi/bpf.h would need to have extensive comment what this barrier is doing.
> > Implementation should probably call synchronize_rcu() and not play games
> > with synchronize_sched(), since that's going too much into implementation.
> > Also should such sys_bpf command be root only?
> > I'm not sure whether dos attack can be made by spamming synchronize_rcu()
> > and synchronize_sched() for that matter.
> 
> Adding Paul E. McKenney in CC. He may want to share his thoughts on the matter.

Let's see...

Spamming synchronize_rcu() and synchronize_sched() should be a non-event,
at least aside from the CPUs doing the spamming.  The reason for this
is that a given task can only fire off a single synchronize_sched or
synchronize_rcu() per few milliseconds, so you need a -lot- of tasks
to have much effect, at which point the sheer number of tasks is much
more a problem than the large number of outstanding synchronize_rcu()
or synchronize_sched() invocations.

I very strongly agree that usermode should have a operation that
synchronizes with whatever eBPF uses, rather than something that forces
a specific type of RCU grace period.

Finally, in a few releases, synchronize_sched() will be retiring in favor
of synchronize_rcu(), which will wait on preemption-disabled regions of
code in addition to waiting on RCU read-side critical sections.  Not a
big deal, as I expect to enlist Coccinelle's aid in this.

Did I manage to hit all the high points?

							Thanx, Paul

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Alexei Starovoitov]

On Mon, Jul 09, 2018 at 03:19:03PM -0700, Paul E. McKenney wrote:
> On Mon, Jul 09, 2018 at 05:35:34PM -0400, Mathieu Desnoyers wrote:
> > 
> > 
> > ----- On Jul 9, 2018, at 5:09 PM, Alexei Starovoitov alexei.starovoitov@gmail.com wrote:
> > 
> > > On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
> > >> ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
> > >> 
> > >> > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
> > >> >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> > >> >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> > >> >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> > >> >> > RCU data structure operations with respect to active programs. For
> > >> >> > example, userspace can update a map->map entry to point to a new map,
> > >> >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> > >> >> > complete, and then drain the old map without fear that BPF programs
> > >> >> > may still be updating it.
> > >> >> > 
> > >> >> > Signed-off-by: Daniel Colascione <dancol@google.com>
> > >> >> > ---
> > >> >> >  include/uapi/linux/bpf.h |  1 +
> > >> >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
> > >> >> >  2 files changed, 15 insertions(+)
> > >> >> > 
> > >> >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > >> >> > index b7db3261c62d..4365c50e8055 100644
> > >> >> > --- a/include/uapi/linux/bpf.h
> > >> >> > +++ b/include/uapi/linux/bpf.h
> > >> >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
> > >> >> >  	BPF_BTF_LOAD,
> > >> >> >  	BPF_BTF_GET_FD_BY_ID,
> > >> >> >  	BPF_TASK_FD_QUERY,
> > >> >> > +	BPF_SYNCHRONIZE,
> > >> >> >  };
> > >> >> >  
> > >> >> >  enum bpf_map_type {
> > >> >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> > >> >> > index d10ecd78105f..60ec7811846e 100644
> > >> >> > --- a/kernel/bpf/syscall.c
> > >> >> > +++ b/kernel/bpf/syscall.c
> > >> >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
> > >> >> > uattr, unsigned int, siz
> > >> >> >  	if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
> > >> >> >  		return -EPERM;
> > >> >> >  
> > >> >> > +	if (cmd == BPF_SYNCHRONIZE) {
> > >> >> > +		if (uattr != NULL || size != 0)
> > >> >> > +			return -EINVAL;
> > >> >> > +		err = security_bpf(cmd, NULL, 0);
> > >> >> > +		if (err < 0)
> > >> >> > +			return err;
> > >> >> > +		/* BPF programs are run with preempt disabled, so
> > >> >> > +		 * synchronize_sched is sufficient even with
> > >> >> > +		 * RCU_PREEMPT.
> > >> >> > +		 */
> > >> >> > +		synchronize_sched();
> > >> >> > +		return 0;
> > >> >> 
> > >> >> I don't think it's necessary. sys_membarrier() can do this already
> > >> >> and some folks use it exactly for this use case.
> > >> > 
> > >> > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
> > >> > though. No where does the manpage say membarrier should be implemented this
> > >> > way so what happens if the implementation changes?
> > >> > 
> > >> > Further, membarrier manpage says that a memory barrier should be matched with
> > >> > a matching barrier. In this use case there is no matching barrier, so it
> > >> > makes it weirder.
> > >> > 
> > >> > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
> > >> > fragile to depend on it for this?
> > >> > 
> > >> >        case MEMBARRIER_CMD_GLOBAL:
> > >> >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
> > >> >                if (tick_nohz_full_enabled())
> > >> >                        return -EINVAL;
> > >> >                if (num_online_cpus() > 1)
> > >> >                        synchronize_sched();
> > >> >                return 0;
> > >> > 
> > >> > 
> > >> > Adding Mathieu as well who I believe is author/maintainer of membarrier.
> > >> 
> > >> See commit 907565337
> > >> "Fix: Disable sys_membarrier when nohz_full is enabled"
> > >> 
> > >> "Userspace applications should be allowed to expect the membarrier system
> > >> call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
> > >> nohz_full CPUs, but synchronize_sched() does not take those into
> > >> account."
> > >> 
> > >> So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
> > >> only care about kernel preempt off critical sections.
> > >> 
> > >> Clearly bpf code does not run in user-space, so it would "work".
> > >> 
> > >> But the guarantees provided by membarrier are not to synchronize against
> > >> preempt off per se. It's just that the current implementation happens to
> > >> do that. The point of membarrier is to turn user-space memory barriers
> > >> into compiler barriers.
> > >> 
> > >> If what you need is to wait for a RCU grace period for whatever RCU flavor
> > >> ebpf is using, I would against using membarrier for this. I would rather
> > >> recommend adding a dedicated BPF_SYNCHRONIZE so you won't leak
> > >> implementation details to user-space, *and* you can eventually change you
> > >> RCU implementation for e.g. SRCU in the future if needed.
> > > 
> > > The point about future changes to underlying bpf mechanisms is valid.
> > > There is work already on the way to reduce the scope of preempt_off+rcu_lock
> > > that currently lasts the whole prog. We will have new prog types that won't
> > > have such wrappers and will do rcu_lock/unlock and preempt on/off only
> > > when necessary.
> > > So something like BPF_SYNCHRONIZE will break soon, since the kernel cannot have
> > > guarantees on when programs finish. Calling this command BPF_SYNCHRONIZE_PROG
> > > also won't make sense for the same reason.
> > > What we can do it instead is to define synchronization barrier for
> > > programs accessing maps. May be call it something like:
> > > BPF_SYNC_MAP_ACCESS ?
> > > uapi/bpf.h would need to have extensive comment what this barrier is doing.
> > > Implementation should probably call synchronize_rcu() and not play games
> > > with synchronize_sched(), since that's going too much into implementation.
> > > Also should such sys_bpf command be root only?
> > > I'm not sure whether dos attack can be made by spamming synchronize_rcu()
> > > and synchronize_sched() for that matter.
> > 
> > Adding Paul E. McKenney in CC. He may want to share his thoughts on the matter.
> 
> Let's see...
> 
> Spamming synchronize_rcu() and synchronize_sched() should be a non-event,
> at least aside from the CPUs doing the spamming.  The reason for this
> is that a given task can only fire off a single synchronize_sched or
> synchronize_rcu() per few milliseconds, so you need a -lot- of tasks
> to have much effect, at which point the sheer number of tasks is much
> more a problem than the large number of outstanding synchronize_rcu()
> or synchronize_sched() invocations.
> 
> I very strongly agree that usermode should have a operation that
> synchronizes with whatever eBPF uses, rather than something that forces
> a specific type of RCU grace period.
> 
> Finally, in a few releases, synchronize_sched() will be retiring in favor
> of synchronize_rcu(), which will wait on preemption-disabled regions of
> code in addition to waiting on RCU read-side critical sections.  Not a
> big deal, as I expect to enlist Coccinelle's aid in this.
> 
> Did I manage to hit all the high points?

Thanks. It's ok for new cmd being unpriv then.

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Paul E. McKenney]

On Mon, Jul 09, 2018 at 03:19:55PM -0700, Alexei Starovoitov wrote:
> On Mon, Jul 09, 2018 at 03:19:03PM -0700, Paul E. McKenney wrote:
> > On Mon, Jul 09, 2018 at 05:35:34PM -0400, Mathieu Desnoyers wrote:
> > > 
> > > 
> > > ----- On Jul 9, 2018, at 5:09 PM, Alexei Starovoitov alexei.starovoitov@gmail.com wrote:
> > > 
> > > > On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
> > > >> ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
> > > >> 
> > > >> > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
> > > >> >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> > > >> >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> > > >> >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> > > >> >> > RCU data structure operations with respect to active programs. For
> > > >> >> > example, userspace can update a map->map entry to point to a new map,
> > > >> >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> > > >> >> > complete, and then drain the old map without fear that BPF programs
> > > >> >> > may still be updating it.
> > > >> >> > 
> > > >> >> > Signed-off-by: Daniel Colascione <dancol@google.com>
> > > >> >> > ---
> > > >> >> >  include/uapi/linux/bpf.h |  1 +
> > > >> >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
> > > >> >> >  2 files changed, 15 insertions(+)
> > > >> >> > 
> > > >> >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > > >> >> > index b7db3261c62d..4365c50e8055 100644
> > > >> >> > --- a/include/uapi/linux/bpf.h
> > > >> >> > +++ b/include/uapi/linux/bpf.h
> > > >> >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
> > > >> >> >  	BPF_BTF_LOAD,
> > > >> >> >  	BPF_BTF_GET_FD_BY_ID,
> > > >> >> >  	BPF_TASK_FD_QUERY,
> > > >> >> > +	BPF_SYNCHRONIZE,
> > > >> >> >  };
> > > >> >> >  
> > > >> >> >  enum bpf_map_type {
> > > >> >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> > > >> >> > index d10ecd78105f..60ec7811846e 100644
> > > >> >> > --- a/kernel/bpf/syscall.c
> > > >> >> > +++ b/kernel/bpf/syscall.c
> > > >> >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
> > > >> >> > uattr, unsigned int, siz
> > > >> >> >  	if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
> > > >> >> >  		return -EPERM;
> > > >> >> >  
> > > >> >> > +	if (cmd == BPF_SYNCHRONIZE) {
> > > >> >> > +		if (uattr != NULL || size != 0)
> > > >> >> > +			return -EINVAL;
> > > >> >> > +		err = security_bpf(cmd, NULL, 0);
> > > >> >> > +		if (err < 0)
> > > >> >> > +			return err;
> > > >> >> > +		/* BPF programs are run with preempt disabled, so
> > > >> >> > +		 * synchronize_sched is sufficient even with
> > > >> >> > +		 * RCU_PREEMPT.
> > > >> >> > +		 */
> > > >> >> > +		synchronize_sched();
> > > >> >> > +		return 0;
> > > >> >> 
> > > >> >> I don't think it's necessary. sys_membarrier() can do this already
> > > >> >> and some folks use it exactly for this use case.
> > > >> > 
> > > >> > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
> > > >> > though. No where does the manpage say membarrier should be implemented this
> > > >> > way so what happens if the implementation changes?
> > > >> > 
> > > >> > Further, membarrier manpage says that a memory barrier should be matched with
> > > >> > a matching barrier. In this use case there is no matching barrier, so it
> > > >> > makes it weirder.
> > > >> > 
> > > >> > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
> > > >> > fragile to depend on it for this?
> > > >> > 
> > > >> >        case MEMBARRIER_CMD_GLOBAL:
> > > >> >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
> > > >> >                if (tick_nohz_full_enabled())
> > > >> >                        return -EINVAL;
> > > >> >                if (num_online_cpus() > 1)
> > > >> >                        synchronize_sched();
> > > >> >                return 0;
> > > >> > 
> > > >> > 
> > > >> > Adding Mathieu as well who I believe is author/maintainer of membarrier.
> > > >> 
> > > >> See commit 907565337
> > > >> "Fix: Disable sys_membarrier when nohz_full is enabled"
> > > >> 
> > > >> "Userspace applications should be allowed to expect the membarrier system
> > > >> call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
> > > >> nohz_full CPUs, but synchronize_sched() does not take those into
> > > >> account."
> > > >> 
> > > >> So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
> > > >> only care about kernel preempt off critical sections.
> > > >> 
> > > >> Clearly bpf code does not run in user-space, so it would "work".
> > > >> 
> > > >> But the guarantees provided by membarrier are not to synchronize against
> > > >> preempt off per se. It's just that the current implementation happens to
> > > >> do that. The point of membarrier is to turn user-space memory barriers
> > > >> into compiler barriers.
> > > >> 
> > > >> If what you need is to wait for a RCU grace period for whatever RCU flavor
> > > >> ebpf is using, I would against using membarrier for this. I would rather
> > > >> recommend adding a dedicated BPF_SYNCHRONIZE so you won't leak
> > > >> implementation details to user-space, *and* you can eventually change you
> > > >> RCU implementation for e.g. SRCU in the future if needed.
> > > > 
> > > > The point about future changes to underlying bpf mechanisms is valid.
> > > > There is work already on the way to reduce the scope of preempt_off+rcu_lock
> > > > that currently lasts the whole prog. We will have new prog types that won't
> > > > have such wrappers and will do rcu_lock/unlock and preempt on/off only
> > > > when necessary.
> > > > So something like BPF_SYNCHRONIZE will break soon, since the kernel cannot have
> > > > guarantees on when programs finish. Calling this command BPF_SYNCHRONIZE_PROG
> > > > also won't make sense for the same reason.
> > > > What we can do it instead is to define synchronization barrier for
> > > > programs accessing maps. May be call it something like:
> > > > BPF_SYNC_MAP_ACCESS ?
> > > > uapi/bpf.h would need to have extensive comment what this barrier is doing.
> > > > Implementation should probably call synchronize_rcu() and not play games
> > > > with synchronize_sched(), since that's going too much into implementation.
> > > > Also should such sys_bpf command be root only?
> > > > I'm not sure whether dos attack can be made by spamming synchronize_rcu()
> > > > and synchronize_sched() for that matter.
> > > 
> > > Adding Paul E. McKenney in CC. He may want to share his thoughts on the matter.
> > 
> > Let's see...
> > 
> > Spamming synchronize_rcu() and synchronize_sched() should be a non-event,
> > at least aside from the CPUs doing the spamming.  The reason for this
> > is that a given task can only fire off a single synchronize_sched or
> > synchronize_rcu() per few milliseconds, so you need a -lot- of tasks
> > to have much effect, at which point the sheer number of tasks is much
> > more a problem than the large number of outstanding synchronize_rcu()
> > or synchronize_sched() invocations.
> > 
> > I very strongly agree that usermode should have a operation that
> > synchronizes with whatever eBPF uses, rather than something that forces
> > a specific type of RCU grace period.
> > 
> > Finally, in a few releases, synchronize_sched() will be retiring in favor
> > of synchronize_rcu(), which will wait on preemption-disabled regions of
> > code in addition to waiting on RCU read-side critical sections.  Not a
> > big deal, as I expect to enlist Coccinelle's aid in this.
> > 
> > Did I manage to hit all the high points?
> 
> Thanks. It's ok for new cmd being unpriv then.

Yes.  After all, users can cause RCU far more pain by doing close(open())
in a tight loop.  ;-)

							Thanx, Paul

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Daniel Colascione]

On Mon, Jul 9, 2018 at 2:09 PM, Alexei Starovoitov
<alexei.starovoitov@gmail.com> wrote:
> On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
>> ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
>>
>> > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
>> >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
>> >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
>> >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
>> >> > RCU data structure operations with respect to active programs. For
>> >> > example, userspace can update a map->map entry to point to a new map,
>> >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
>> >> > complete, and then drain the old map without fear that BPF programs
>> >> > may still be updating it.
>> >> >
>> >> > Signed-off-by: Daniel Colascione <dancol@google.com>
>> >> > ---
>> >> >  include/uapi/linux/bpf.h |  1 +
>> >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
>> >> >  2 files changed, 15 insertions(+)
>> >> >
>> >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
>> >> > index b7db3261c62d..4365c50e8055 100644
>> >> > --- a/include/uapi/linux/bpf.h
>> >> > +++ b/include/uapi/linux/bpf.h
>> >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
>> >> >          BPF_BTF_LOAD,
>> >> >          BPF_BTF_GET_FD_BY_ID,
>> >> >          BPF_TASK_FD_QUERY,
>> >> > +        BPF_SYNCHRONIZE,
>> >> >  };
>> >> >
>> >> >  enum bpf_map_type {
>> >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
>> >> > index d10ecd78105f..60ec7811846e 100644
>> >> > --- a/kernel/bpf/syscall.c
>> >> > +++ b/kernel/bpf/syscall.c
>> >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
>> >> > uattr, unsigned int, siz
>> >> >          if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
>> >> >                  return -EPERM;
>> >> >
>> >> > +        if (cmd == BPF_SYNCHRONIZE) {
>> >> > +                if (uattr != NULL || size != 0)
>> >> > +                        return -EINVAL;
>> >> > +                err = security_bpf(cmd, NULL, 0);
>> >> > +                if (err < 0)
>> >> > +                        return err;
>> >> > +                /* BPF programs are run with preempt disabled, so
>> >> > +                 * synchronize_sched is sufficient even with
>> >> > +                 * RCU_PREEMPT.
>> >> > +                 */
>> >> > +                synchronize_sched();
>> >> > +                return 0;
>> >>
>> >> I don't think it's necessary. sys_membarrier() can do this already
>> >> and some folks use it exactly for this use case.
>> >
>> > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
>> > though. No where does the manpage say membarrier should be implemented this
>> > way so what happens if the implementation changes?
>> >
>> > Further, membarrier manpage says that a memory barrier should be matched with
>> > a matching barrier. In this use case there is no matching barrier, so it
>> > makes it weirder.
>> >
>> > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
>> > fragile to depend on it for this?
>> >
>> >        case MEMBARRIER_CMD_GLOBAL:
>> >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
>> >                if (tick_nohz_full_enabled())
>> >                        return -EINVAL;
>> >                if (num_online_cpus() > 1)
>> >                        synchronize_sched();
>> >                return 0;
>> >
>> >
>> > Adding Mathieu as well who I believe is author/maintainer of membarrier.
>>
>> See commit 907565337
>> "Fix: Disable sys_membarrier when nohz_full is enabled"
>>
>> "Userspace applications should be allowed to expect the membarrier system
>> call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
>> nohz_full CPUs, but synchronize_sched() does not take those into
>> account."
>>
>> So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
>> only care about kernel preempt off critical sections.
>>
>> Clearly bpf code does not run in user-space, so it would "work".
>>
>> But the guarantees provided by membarrier are not to synchronize against
>> preempt off per se. It's just that the current implementation happens to
>> do that. The point of membarrier is to turn user-space memory barriers
>> into compiler barriers.
>>
>> If what you need is to wait for a RCU grace period for whatever RCU flavor
>> ebpf is using, I would against using membarrier for this. I would rather
>> recommend adding a dedicated BPF_SYNCHRONIZE so you won't leak
>> implementation details to user-space, *and* you can eventually change you
>> RCU implementation for e.g. SRCU in the future if needed.
>
> The point about future changes to underlying bpf mechanisms is valid.
> There is work already on the way to reduce the scope of preempt_off+rcu_lock
> that currently lasts the whole prog. We will have new prog types that won't
> have such wrappers and will do rcu_lock/unlock and preempt on/off only
> when necessary.
> So something like BPF_SYNCHRONIZE will break soon, since the kernel cannot have
> guarantees on when programs finish. Calling this command BPF_SYNCHRONIZE_PROG
> also won't make sense for the same reason.
> What we can do it instead is to define synchronization barrier for
> programs accessing maps. May be call it something like:
> BPF_SYNC_MAP_ACCESS ?

I'm not sure what you're proposing. In the case the commit message
describes, a user-space program that wants to "drain" a map needs to
be confident that the map won't change under it, even across multiple
bpf system calls on that map. One way of doing that is to ensure that
nothing that could possibly hold a reference to that map is still
running. Are you proposing some kind of refcount-draining approach?
Simple locking won't work, since BPF programs can't block, and I don't
see right now how a simple barrier would help.

> Also should such sys_bpf command be root only?
> I'm not sure whether dos attack can be made by spamming synchronize_rcu()
> and synchronize_sched() for that matter.

membarrier being globally available seems not to have caused any kind
of DoS catastrophe.

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Alexei Starovoitov]

On Mon, Jul 09, 2018 at 02:36:37PM -0700, Daniel Colascione wrote:
> On Mon, Jul 9, 2018 at 2:09 PM, Alexei Starovoitov
> <alexei.starovoitov@gmail.com> wrote:
> > On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
> >> ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
> >>
> >> > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
> >> >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> >> >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> >> >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> >> >> > RCU data structure operations with respect to active programs. For
> >> >> > example, userspace can update a map->map entry to point to a new map,
> >> >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> >> >> > complete, and then drain the old map without fear that BPF programs
> >> >> > may still be updating it.
> >> >> >
> >> >> > Signed-off-by: Daniel Colascione <dancol@google.com>
> >> >> > ---
> >> >> >  include/uapi/linux/bpf.h |  1 +
> >> >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
> >> >> >  2 files changed, 15 insertions(+)
> >> >> >
> >> >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> >> >> > index b7db3261c62d..4365c50e8055 100644
> >> >> > --- a/include/uapi/linux/bpf.h
> >> >> > +++ b/include/uapi/linux/bpf.h
> >> >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
> >> >> >          BPF_BTF_LOAD,
> >> >> >          BPF_BTF_GET_FD_BY_ID,
> >> >> >          BPF_TASK_FD_QUERY,
> >> >> > +        BPF_SYNCHRONIZE,
> >> >> >  };
> >> >> >
> >> >> >  enum bpf_map_type {
> >> >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> >> >> > index d10ecd78105f..60ec7811846e 100644
> >> >> > --- a/kernel/bpf/syscall.c
> >> >> > +++ b/kernel/bpf/syscall.c
> >> >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
> >> >> > uattr, unsigned int, siz
> >> >> >          if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
> >> >> >                  return -EPERM;
> >> >> >
> >> >> > +        if (cmd == BPF_SYNCHRONIZE) {
> >> >> > +                if (uattr != NULL || size != 0)
> >> >> > +                        return -EINVAL;
> >> >> > +                err = security_bpf(cmd, NULL, 0);
> >> >> > +                if (err < 0)
> >> >> > +                        return err;
> >> >> > +                /* BPF programs are run with preempt disabled, so
> >> >> > +                 * synchronize_sched is sufficient even with
> >> >> > +                 * RCU_PREEMPT.
> >> >> > +                 */
> >> >> > +                synchronize_sched();
> >> >> > +                return 0;
> >> >>
> >> >> I don't think it's necessary. sys_membarrier() can do this already
> >> >> and some folks use it exactly for this use case.
> >> >
> >> > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
> >> > though. No where does the manpage say membarrier should be implemented this
> >> > way so what happens if the implementation changes?
> >> >
> >> > Further, membarrier manpage says that a memory barrier should be matched with
> >> > a matching barrier. In this use case there is no matching barrier, so it
> >> > makes it weirder.
> >> >
> >> > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
> >> > fragile to depend on it for this?
> >> >
> >> >        case MEMBARRIER_CMD_GLOBAL:
> >> >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
> >> >                if (tick_nohz_full_enabled())
> >> >                        return -EINVAL;
> >> >                if (num_online_cpus() > 1)
> >> >                        synchronize_sched();
> >> >                return 0;
> >> >
> >> >
> >> > Adding Mathieu as well who I believe is author/maintainer of membarrier.
> >>
> >> See commit 907565337
> >> "Fix: Disable sys_membarrier when nohz_full is enabled"
> >>
> >> "Userspace applications should be allowed to expect the membarrier system
> >> call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
> >> nohz_full CPUs, but synchronize_sched() does not take those into
> >> account."
> >>
> >> So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
> >> only care about kernel preempt off critical sections.
> >>
> >> Clearly bpf code does not run in user-space, so it would "work".
> >>
> >> But the guarantees provided by membarrier are not to synchronize against
> >> preempt off per se. It's just that the current implementation happens to
> >> do that. The point of membarrier is to turn user-space memory barriers
> >> into compiler barriers.
> >>
> >> If what you need is to wait for a RCU grace period for whatever RCU flavor
> >> ebpf is using, I would against using membarrier for this. I would rather
> >> recommend adding a dedicated BPF_SYNCHRONIZE so you won't leak
> >> implementation details to user-space, *and* you can eventually change you
> >> RCU implementation for e.g. SRCU in the future if needed.
> >
> > The point about future changes to underlying bpf mechanisms is valid.
> > There is work already on the way to reduce the scope of preempt_off+rcu_lock
> > that currently lasts the whole prog. We will have new prog types that won't
> > have such wrappers and will do rcu_lock/unlock and preempt on/off only
> > when necessary.
> > So something like BPF_SYNCHRONIZE will break soon, since the kernel cannot have
> > guarantees on when programs finish. Calling this command BPF_SYNCHRONIZE_PROG
> > also won't make sense for the same reason.
> > What we can do it instead is to define synchronization barrier for
> > programs accessing maps. May be call it something like:
> > BPF_SYNC_MAP_ACCESS ?
> 
> I'm not sure what you're proposing. In the case the commit message
> describes, a user-space program that wants to "drain" a map needs to
> be confident that the map won't change under it, even across multiple
> bpf system calls on that map. One way of doing that is to ensure that
> nothing that could possibly hold a reference to that map is still
> running. Are you proposing some kind of refcount-draining approach?
> Simple locking won't work, since BPF programs can't block, and I don't
> see right now how a simple barrier would help.

I'm proposing few changes for your patch:
s/BPF_SYNCHRONIZE/BPF_SYNC_MAP_ACCESS/
and s/synchronize_sched/synchronize_rcu/
with detailed comment in uapi/bpf.h that has an example why folks
would want to use this new cmd.
I think the bpf maps will be rcu protected for foreseeable future
even when rcu_read_lock/unlock will be done by the programs instead of
kernel wrappers.

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Daniel Colascione]

On Mon, Jul 9, 2018 at 3:10 PM, Alexei Starovoitov
<alexei.starovoitov@gmail.com> wrote:
> On Mon, Jul 09, 2018 at 02:36:37PM -0700, Daniel Colascione wrote:
>> On Mon, Jul 9, 2018 at 2:09 PM, Alexei Starovoitov
>> <alexei.starovoitov@gmail.com> wrote:
>> > On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
>> >> ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
>> >>
>> >> > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
>> >> >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
>> >> >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
>> >> >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
>> >> >> > RCU data structure operations with respect to active programs. For
>> >> >> > example, userspace can update a map->map entry to point to a new map,
>> >> >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
>> >> >> > complete, and then drain the old map without fear that BPF programs
>> >> >> > may still be updating it.
>> >> >> >
>> >> >> > Signed-off-by: Daniel Colascione <dancol@google.com>
>> >> >> > ---
>> >> >> >  include/uapi/linux/bpf.h |  1 +
>> >> >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
>> >> >> >  2 files changed, 15 insertions(+)
>> >> >> >
>> >> >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
>> >> >> > index b7db3261c62d..4365c50e8055 100644
>> >> >> > --- a/include/uapi/linux/bpf.h
>> >> >> > +++ b/include/uapi/linux/bpf.h
>> >> >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
>> >> >> >          BPF_BTF_LOAD,
>> >> >> >          BPF_BTF_GET_FD_BY_ID,
>> >> >> >          BPF_TASK_FD_QUERY,
>> >> >> > +        BPF_SYNCHRONIZE,
>> >> >> >  };
>> >> >> >
>> >> >> >  enum bpf_map_type {
>> >> >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
>> >> >> > index d10ecd78105f..60ec7811846e 100644
>> >> >> > --- a/kernel/bpf/syscall.c
>> >> >> > +++ b/kernel/bpf/syscall.c
>> >> >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
>> >> >> > uattr, unsigned int, siz
>> >> >> >          if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
>> >> >> >                  return -EPERM;
>> >> >> >
>> >> >> > +        if (cmd == BPF_SYNCHRONIZE) {
>> >> >> > +                if (uattr != NULL || size != 0)
>> >> >> > +                        return -EINVAL;
>> >> >> > +                err = security_bpf(cmd, NULL, 0);
>> >> >> > +                if (err < 0)
>> >> >> > +                        return err;
>> >> >> > +                /* BPF programs are run with preempt disabled, so
>> >> >> > +                 * synchronize_sched is sufficient even with
>> >> >> > +                 * RCU_PREEMPT.
>> >> >> > +                 */
>> >> >> > +                synchronize_sched();
>> >> >> > +                return 0;
>> >> >>
>> >> >> I don't think it's necessary. sys_membarrier() can do this already
>> >> >> and some folks use it exactly for this use case.
>> >> >
>> >> > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
>> >> > though. No where does the manpage say membarrier should be implemented this
>> >> > way so what happens if the implementation changes?
>> >> >
>> >> > Further, membarrier manpage says that a memory barrier should be matched with
>> >> > a matching barrier. In this use case there is no matching barrier, so it
>> >> > makes it weirder.
>> >> >
>> >> > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
>> >> > fragile to depend on it for this?
>> >> >
>> >> >        case MEMBARRIER_CMD_GLOBAL:
>> >> >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
>> >> >                if (tick_nohz_full_enabled())
>> >> >                        return -EINVAL;
>> >> >                if (num_online_cpus() > 1)
>> >> >                        synchronize_sched();
>> >> >                return 0;
>> >> >
>> >> >
>> >> > Adding Mathieu as well who I believe is author/maintainer of membarrier.
>> >>
>> >> See commit 907565337
>> >> "Fix: Disable sys_membarrier when nohz_full is enabled"
>> >>
>> >> "Userspace applications should be allowed to expect the membarrier system
>> >> call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
>> >> nohz_full CPUs, but synchronize_sched() does not take those into
>> >> account."
>> >>
>> >> So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
>> >> only care about kernel preempt off critical sections.
>> >>
>> >> Clearly bpf code does not run in user-space, so it would "work".
>> >>
>> >> But the guarantees provided by membarrier are not to synchronize against
>> >> preempt off per se. It's just that the current implementation happens to
>> >> do that. The point of membarrier is to turn user-space memory barriers
>> >> into compiler barriers.
>> >>
>> >> If what you need is to wait for a RCU grace period for whatever RCU flavor
>> >> ebpf is using, I would against using membarrier for this. I would rather
>> >> recommend adding a dedicated BPF_SYNCHRONIZE so you won't leak
>> >> implementation details to user-space, *and* you can eventually change you
>> >> RCU implementation for e.g. SRCU in the future if needed.
>> >
>> > The point about future changes to underlying bpf mechanisms is valid.
>> > There is work already on the way to reduce the scope of preempt_off+rcu_lock
>> > that currently lasts the whole prog. We will have new prog types that won't
>> > have such wrappers and will do rcu_lock/unlock and preempt on/off only
>> > when necessary.
>> > So something like BPF_SYNCHRONIZE will break soon, since the kernel cannot have
>> > guarantees on when programs finish. Calling this command BPF_SYNCHRONIZE_PROG
>> > also won't make sense for the same reason.
>> > What we can do it instead is to define synchronization barrier for
>> > programs accessing maps. May be call it something like:
>> > BPF_SYNC_MAP_ACCESS ?
>>
>> I'm not sure what you're proposing. In the case the commit message
>> describes, a user-space program that wants to "drain" a map needs to
>> be confident that the map won't change under it, even across multiple
>> bpf system calls on that map. One way of doing that is to ensure that
>> nothing that could possibly hold a reference to that map is still
>> running. Are you proposing some kind of refcount-draining approach?
>> Simple locking won't work, since BPF programs can't block, and I don't
>> see right now how a simple barrier would help.
>
> I'm proposing few changes for your patch:
> s/BPF_SYNCHRONIZE/BPF_SYNC_MAP_ACCESS/
> and s/synchronize_sched/synchronize_rcu/
> with detailed comment in uapi/bpf.h that has an example why folks
> would want to use this new cmd.

Thanks for clarifying.

> I think the bpf maps will be rcu protected for foreseeable future
> even when rcu_read_lock/unlock will be done by the programs instead of
> kernel wrappers.

Can we guarantee that we always obtain a map reference and dispose of
that reference inside the same critical section? If so, can BPF
programs then disable preemption for as long as they'd like?

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Alexei Starovoitov]

On Mon, Jul 09, 2018 at 03:21:43PM -0700, Daniel Colascione wrote:
> On Mon, Jul 9, 2018 at 3:10 PM, Alexei Starovoitov
> <alexei.starovoitov@gmail.com> wrote:
> > On Mon, Jul 09, 2018 at 02:36:37PM -0700, Daniel Colascione wrote:
> >> On Mon, Jul 9, 2018 at 2:09 PM, Alexei Starovoitov
> >> <alexei.starovoitov@gmail.com> wrote:
> >> > On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
> >> >> ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
> >> >>
> >> >> > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
> >> >> >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> >> >> >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> >> >> >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> >> >> >> > RCU data structure operations with respect to active programs. For
> >> >> >> > example, userspace can update a map->map entry to point to a new map,
> >> >> >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> >> >> >> > complete, and then drain the old map without fear that BPF programs
> >> >> >> > may still be updating it.
> >> >> >> >
> >> >> >> > Signed-off-by: Daniel Colascione <dancol@google.com>
> >> >> >> > ---
> >> >> >> >  include/uapi/linux/bpf.h |  1 +
> >> >> >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
> >> >> >> >  2 files changed, 15 insertions(+)
> >> >> >> >
> >> >> >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> >> >> >> > index b7db3261c62d..4365c50e8055 100644
> >> >> >> > --- a/include/uapi/linux/bpf.h
> >> >> >> > +++ b/include/uapi/linux/bpf.h
> >> >> >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
> >> >> >> >          BPF_BTF_LOAD,
> >> >> >> >          BPF_BTF_GET_FD_BY_ID,
> >> >> >> >          BPF_TASK_FD_QUERY,
> >> >> >> > +        BPF_SYNCHRONIZE,
> >> >> >> >  };
> >> >> >> >
> >> >> >> >  enum bpf_map_type {
> >> >> >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> >> >> >> > index d10ecd78105f..60ec7811846e 100644
> >> >> >> > --- a/kernel/bpf/syscall.c
> >> >> >> > +++ b/kernel/bpf/syscall.c
> >> >> >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
> >> >> >> > uattr, unsigned int, siz
> >> >> >> >          if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
> >> >> >> >                  return -EPERM;
> >> >> >> >
> >> >> >> > +        if (cmd == BPF_SYNCHRONIZE) {
> >> >> >> > +                if (uattr != NULL || size != 0)
> >> >> >> > +                        return -EINVAL;
> >> >> >> > +                err = security_bpf(cmd, NULL, 0);
> >> >> >> > +                if (err < 0)
> >> >> >> > +                        return err;
> >> >> >> > +                /* BPF programs are run with preempt disabled, so
> >> >> >> > +                 * synchronize_sched is sufficient even with
> >> >> >> > +                 * RCU_PREEMPT.
> >> >> >> > +                 */
> >> >> >> > +                synchronize_sched();
> >> >> >> > +                return 0;
> >> >> >>
> >> >> >> I don't think it's necessary. sys_membarrier() can do this already
> >> >> >> and some folks use it exactly for this use case.
> >> >> >
> >> >> > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
> >> >> > though. No where does the manpage say membarrier should be implemented this
> >> >> > way so what happens if the implementation changes?
> >> >> >
> >> >> > Further, membarrier manpage says that a memory barrier should be matched with
> >> >> > a matching barrier. In this use case there is no matching barrier, so it
> >> >> > makes it weirder.
> >> >> >
> >> >> > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
> >> >> > fragile to depend on it for this?
> >> >> >
> >> >> >        case MEMBARRIER_CMD_GLOBAL:
> >> >> >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
> >> >> >                if (tick_nohz_full_enabled())
> >> >> >                        return -EINVAL;
> >> >> >                if (num_online_cpus() > 1)
> >> >> >                        synchronize_sched();
> >> >> >                return 0;
> >> >> >
> >> >> >
> >> >> > Adding Mathieu as well who I believe is author/maintainer of membarrier.
> >> >>
> >> >> See commit 907565337
> >> >> "Fix: Disable sys_membarrier when nohz_full is enabled"
> >> >>
> >> >> "Userspace applications should be allowed to expect the membarrier system
> >> >> call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
> >> >> nohz_full CPUs, but synchronize_sched() does not take those into
> >> >> account."
> >> >>
> >> >> So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
> >> >> only care about kernel preempt off critical sections.
> >> >>
> >> >> Clearly bpf code does not run in user-space, so it would "work".
> >> >>
> >> >> But the guarantees provided by membarrier are not to synchronize against
> >> >> preempt off per se. It's just that the current implementation happens to
> >> >> do that. The point of membarrier is to turn user-space memory barriers
> >> >> into compiler barriers.
> >> >>
> >> >> If what you need is to wait for a RCU grace period for whatever RCU flavor
> >> >> ebpf is using, I would against using membarrier for this. I would rather
> >> >> recommend adding a dedicated BPF_SYNCHRONIZE so you won't leak
> >> >> implementation details to user-space, *and* you can eventually change you
> >> >> RCU implementation for e.g. SRCU in the future if needed.
> >> >
> >> > The point about future changes to underlying bpf mechanisms is valid.
> >> > There is work already on the way to reduce the scope of preempt_off+rcu_lock
> >> > that currently lasts the whole prog. We will have new prog types that won't
> >> > have such wrappers and will do rcu_lock/unlock and preempt on/off only
> >> > when necessary.
> >> > So something like BPF_SYNCHRONIZE will break soon, since the kernel cannot have
> >> > guarantees on when programs finish. Calling this command BPF_SYNCHRONIZE_PROG
> >> > also won't make sense for the same reason.
> >> > What we can do it instead is to define synchronization barrier for
> >> > programs accessing maps. May be call it something like:
> >> > BPF_SYNC_MAP_ACCESS ?
> >>
> >> I'm not sure what you're proposing. In the case the commit message
> >> describes, a user-space program that wants to "drain" a map needs to
> >> be confident that the map won't change under it, even across multiple
> >> bpf system calls on that map. One way of doing that is to ensure that
> >> nothing that could possibly hold a reference to that map is still
> >> running. Are you proposing some kind of refcount-draining approach?
> >> Simple locking won't work, since BPF programs can't block, and I don't
> >> see right now how a simple barrier would help.
> >
> > I'm proposing few changes for your patch:
> > s/BPF_SYNCHRONIZE/BPF_SYNC_MAP_ACCESS/
> > and s/synchronize_sched/synchronize_rcu/
> > with detailed comment in uapi/bpf.h that has an example why folks
> > would want to use this new cmd.
> 
> Thanks for clarifying.
> 
> > I think the bpf maps will be rcu protected for foreseeable future
> > even when rcu_read_lock/unlock will be done by the programs instead of
> > kernel wrappers.
> 
> Can we guarantee that we always obtain a map reference and dispose of
> that reference inside the same critical section? 

yep. the verifier will guarantee that.

> If so, can BPF
> programs then disable preemption for as long as they'd like?

you mean after the finish? no. only while running.
The verifier will match things like lookup/release, lock/unlock, preempt on/off
and will make sure there is no dangling preempt disable after program returns.

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Chenbo Feng]

On Mon, Jul 9, 2018 at 3:34 PM Alexei Starovoitov
<alexei.starovoitov@gmail.com> wrote:
>
> On Mon, Jul 09, 2018 at 03:21:43PM -0700, Daniel Colascione wrote:
> > On Mon, Jul 9, 2018 at 3:10 PM, Alexei Starovoitov
> > <alexei.starovoitov@gmail.com> wrote:
> > > On Mon, Jul 09, 2018 at 02:36:37PM -0700, Daniel Colascione wrote:
> > >> On Mon, Jul 9, 2018 at 2:09 PM, Alexei Starovoitov
> > >> <alexei.starovoitov@gmail.com> wrote:
> > >> > On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
> > >> >> ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
> > >> >>
> > >> >> > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
> > >> >> >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> > >> >> >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> > >> >> >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> > >> >> >> > RCU data structure operations with respect to active programs. For
> > >> >> >> > example, userspace can update a map->map entry to point to a new map,
> > >> >> >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> > >> >> >> > complete, and then drain the old map without fear that BPF programs
> > >> >> >> > may still be updating it.
> > >> >> >> >
> > >> >> >> > Signed-off-by: Daniel Colascione <dancol@google.com>
> > >> >> >> > ---
> > >> >> >> >  include/uapi/linux/bpf.h |  1 +
> > >> >> >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
> > >> >> >> >  2 files changed, 15 insertions(+)
> > >> >> >> >
> > >> >> >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > >> >> >> > index b7db3261c62d..4365c50e8055 100644
> > >> >> >> > --- a/include/uapi/linux/bpf.h
> > >> >> >> > +++ b/include/uapi/linux/bpf.h
> > >> >> >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
> > >> >> >> >          BPF_BTF_LOAD,
> > >> >> >> >          BPF_BTF_GET_FD_BY_ID,
> > >> >> >> >          BPF_TASK_FD_QUERY,
> > >> >> >> > +        BPF_SYNCHRONIZE,
> > >> >> >> >  };
> > >> >> >> >
> > >> >> >> >  enum bpf_map_type {
> > >> >> >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> > >> >> >> > index d10ecd78105f..60ec7811846e 100644
> > >> >> >> > --- a/kernel/bpf/syscall.c
> > >> >> >> > +++ b/kernel/bpf/syscall.c
> > >> >> >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
> > >> >> >> > uattr, unsigned int, siz
> > >> >> >> >          if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
> > >> >> >> >                  return -EPERM;
> > >> >> >> >
> > >> >> >> > +        if (cmd == BPF_SYNCHRONIZE) {
> > >> >> >> > +                if (uattr != NULL || size != 0)
> > >> >> >> > +                        return -EINVAL;
> > >> >> >> > +                err = security_bpf(cmd, NULL, 0);
> > >> >> >> > +                if (err < 0)
> > >> >> >> > +                        return err;
> > >> >> >> > +                /* BPF programs are run with preempt disabled, so
> > >> >> >> > +                 * synchronize_sched is sufficient even with
> > >> >> >> > +                 * RCU_PREEMPT.
> > >> >> >> > +                 */
> > >> >> >> > +                synchronize_sched();
> > >> >> >> > +                return 0;
> > >> >> >>
> > >> >> >> I don't think it's necessary. sys_membarrier() can do this already
> > >> >> >> and some folks use it exactly for this use case.
> > >> >> >
> > >> >> > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
> > >> >> > though. No where does the manpage say membarrier should be implemented this
> > >> >> > way so what happens if the implementation changes?
> > >> >> >
> > >> >> > Further, membarrier manpage says that a memory barrier should be matched with
> > >> >> > a matching barrier. In this use case there is no matching barrier, so it
> > >> >> > makes it weirder.
> > >> >> >
> > >> >> > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
> > >> >> > fragile to depend on it for this?
> > >> >> >
> > >> >> >        case MEMBARRIER_CMD_GLOBAL:
> > >> >> >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
> > >> >> >                if (tick_nohz_full_enabled())
> > >> >> >                        return -EINVAL;
> > >> >> >                if (num_online_cpus() > 1)
> > >> >> >                        synchronize_sched();
> > >> >> >                return 0;
> > >> >> >
> > >> >> >
> > >> >> > Adding Mathieu as well who I believe is author/maintainer of membarrier.
> > >> >>
> > >> >> See commit 907565337
> > >> >> "Fix: Disable sys_membarrier when nohz_full is enabled"
> > >> >>
> > >> >> "Userspace applications should be allowed to expect the membarrier system
> > >> >> call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
> > >> >> nohz_full CPUs, but synchronize_sched() does not take those into
> > >> >> account."
> > >> >>
> > >> >> So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
> > >> >> only care about kernel preempt off critical sections.
> > >> >>
> > >> >> Clearly bpf code does not run in user-space, so it would "work".
> > >> >>
> > >> >> But the guarantees provided by membarrier are not to synchronize against
> > >> >> preempt off per se. It's just that the current implementation happens to
> > >> >> do that. The point of membarrier is to turn user-space memory barriers
> > >> >> into compiler barriers.
> > >> >>
> > >> >> If what you need is to wait for a RCU grace period for whatever RCU flavor
> > >> >> ebpf is using, I would against using membarrier for this. I would rather
> > >> >> recommend adding a dedicated BPF_SYNCHRONIZE so you won't leak
> > >> >> implementation details to user-space, *and* you can eventually change you
> > >> >> RCU implementation for e.g. SRCU in the future if needed.
> > >> >
> > >> > The point about future changes to underlying bpf mechanisms is valid.
> > >> > There is work already on the way to reduce the scope of preempt_off+rcu_lock
> > >> > that currently lasts the whole prog. We will have new prog types that won't
> > >> > have such wrappers and will do rcu_lock/unlock and preempt on/off only
> > >> > when necessary.
> > >> > So something like BPF_SYNCHRONIZE will break soon, since the kernel cannot have
> > >> > guarantees on when programs finish. Calling this command BPF_SYNCHRONIZE_PROG
> > >> > also won't make sense for the same reason.
> > >> > What we can do it instead is to define synchronization barrier for
> > >> > programs accessing maps. May be call it something like:
> > >> > BPF_SYNC_MAP_ACCESS ?
> > >>
> > >> I'm not sure what you're proposing. In the case the commit message
> > >> describes, a user-space program that wants to "drain" a map needs to
> > >> be confident that the map won't change under it, even across multiple
> > >> bpf system calls on that map. One way of doing that is to ensure that
> > >> nothing that could possibly hold a reference to that map is still
> > >> running. Are you proposing some kind of refcount-draining approach?
> > >> Simple locking won't work, since BPF programs can't block, and I don't
> > >> see right now how a simple barrier would help.
> > >
> > > I'm proposing few changes for your patch:
> > > s/BPF_SYNCHRONIZE/BPF_SYNC_MAP_ACCESS/
> > > and s/synchronize_sched/synchronize_rcu/
> > > with detailed comment in uapi/bpf.h that has an example why folks
> > > would want to use this new cmd.
> >
> > Thanks for clarifying.
> >
> > > I think the bpf maps will be rcu protected for foreseeable future
> > > even when rcu_read_lock/unlock will be done by the programs instead of
> > > kernel wrappers.
> >
> > Can we guarantee that we always obtain a map reference and dispose of
> > that reference inside the same critical section?
>
> yep. the verifier will guarantee that.
>
> > If so, can BPF
> > programs then disable preemption for as long as they'd like?
>
> you mean after the finish? no. only while running.
> The verifier will match things like lookup/release, lock/unlock, preempt on/off
> and will make sure there is no dangling preempt disable after program returns.
>
It might be a silly question but does this new bpf program type
provide a way to customize where to hold the rcu_lock/unlock and
enable/disable preemption when creating the program? Since the
BPF_SYNC_MAP_ACCESS cmd sounds not very useful if it only protect each
single map access instead of the whole program. For example, sometimes
we have to read the same map multiple times when running a eBPF
program and if the userspace changed the map value between two reads,
it may cause troubles. It would be great if we can have a RCU lock on
a block of eBPF program and in that way I think BPF_SYNC_MAP_ACCESS
cmd should be enough in handling userspace kernel racing problems.

Thanks
Chenbo Feng

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Alexei Starovoitov]

On Mon, Jul 09, 2018 at 10:25:04PM -0700, Chenbo Feng wrote:
> On Mon, Jul 9, 2018 at 3:34 PM Alexei Starovoitov
> <alexei.starovoitov@gmail.com> wrote:
> >
> > On Mon, Jul 09, 2018 at 03:21:43PM -0700, Daniel Colascione wrote:
> > > On Mon, Jul 9, 2018 at 3:10 PM, Alexei Starovoitov
> > > <alexei.starovoitov@gmail.com> wrote:
> > > > On Mon, Jul 09, 2018 at 02:36:37PM -0700, Daniel Colascione wrote:
> > > >> On Mon, Jul 9, 2018 at 2:09 PM, Alexei Starovoitov
> > > >> <alexei.starovoitov@gmail.com> wrote:
> > > >> > On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
> > > >> >> ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
> > > >> >>
> > > >> >> > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
> > > >> >> >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> > > >> >> >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> > > >> >> >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> > > >> >> >> > RCU data structure operations with respect to active programs. For
> > > >> >> >> > example, userspace can update a map->map entry to point to a new map,
> > > >> >> >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> > > >> >> >> > complete, and then drain the old map without fear that BPF programs
> > > >> >> >> > may still be updating it.
> > > >> >> >> >
> > > >> >> >> > Signed-off-by: Daniel Colascione <dancol@google.com>
> > > >> >> >> > ---
> > > >> >> >> >  include/uapi/linux/bpf.h |  1 +
> > > >> >> >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
> > > >> >> >> >  2 files changed, 15 insertions(+)
> > > >> >> >> >
> > > >> >> >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > > >> >> >> > index b7db3261c62d..4365c50e8055 100644
> > > >> >> >> > --- a/include/uapi/linux/bpf.h
> > > >> >> >> > +++ b/include/uapi/linux/bpf.h
> > > >> >> >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
> > > >> >> >> >          BPF_BTF_LOAD,
> > > >> >> >> >          BPF_BTF_GET_FD_BY_ID,
> > > >> >> >> >          BPF_TASK_FD_QUERY,
> > > >> >> >> > +        BPF_SYNCHRONIZE,
> > > >> >> >> >  };
> > > >> >> >> >
> > > >> >> >> >  enum bpf_map_type {
> > > >> >> >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> > > >> >> >> > index d10ecd78105f..60ec7811846e 100644
> > > >> >> >> > --- a/kernel/bpf/syscall.c
> > > >> >> >> > +++ b/kernel/bpf/syscall.c
> > > >> >> >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
> > > >> >> >> > uattr, unsigned int, siz
> > > >> >> >> >          if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
> > > >> >> >> >                  return -EPERM;
> > > >> >> >> >
> > > >> >> >> > +        if (cmd == BPF_SYNCHRONIZE) {
> > > >> >> >> > +                if (uattr != NULL || size != 0)
> > > >> >> >> > +                        return -EINVAL;
> > > >> >> >> > +                err = security_bpf(cmd, NULL, 0);
> > > >> >> >> > +                if (err < 0)
> > > >> >> >> > +                        return err;
> > > >> >> >> > +                /* BPF programs are run with preempt disabled, so
> > > >> >> >> > +                 * synchronize_sched is sufficient even with
> > > >> >> >> > +                 * RCU_PREEMPT.
> > > >> >> >> > +                 */
> > > >> >> >> > +                synchronize_sched();
> > > >> >> >> > +                return 0;
> > > >> >> >>
> > > >> >> >> I don't think it's necessary. sys_membarrier() can do this already
> > > >> >> >> and some folks use it exactly for this use case.
> > > >> >> >
> > > >> >> > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
> > > >> >> > though. No where does the manpage say membarrier should be implemented this
> > > >> >> > way so what happens if the implementation changes?
> > > >> >> >
> > > >> >> > Further, membarrier manpage says that a memory barrier should be matched with
> > > >> >> > a matching barrier. In this use case there is no matching barrier, so it
> > > >> >> > makes it weirder.
> > > >> >> >
> > > >> >> > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
> > > >> >> > fragile to depend on it for this?
> > > >> >> >
> > > >> >> >        case MEMBARRIER_CMD_GLOBAL:
> > > >> >> >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
> > > >> >> >                if (tick_nohz_full_enabled())
> > > >> >> >                        return -EINVAL;
> > > >> >> >                if (num_online_cpus() > 1)
> > > >> >> >                        synchronize_sched();
> > > >> >> >                return 0;
> > > >> >> >
> > > >> >> >
> > > >> >> > Adding Mathieu as well who I believe is author/maintainer of membarrier.
> > > >> >>
> > > >> >> See commit 907565337
> > > >> >> "Fix: Disable sys_membarrier when nohz_full is enabled"
> > > >> >>
> > > >> >> "Userspace applications should be allowed to expect the membarrier system
> > > >> >> call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
> > > >> >> nohz_full CPUs, but synchronize_sched() does not take those into
> > > >> >> account."
> > > >> >>
> > > >> >> So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
> > > >> >> only care about kernel preempt off critical sections.
> > > >> >>
> > > >> >> Clearly bpf code does not run in user-space, so it would "work".
> > > >> >>
> > > >> >> But the guarantees provided by membarrier are not to synchronize against
> > > >> >> preempt off per se. It's just that the current implementation happens to
> > > >> >> do that. The point of membarrier is to turn user-space memory barriers
> > > >> >> into compiler barriers.
> > > >> >>
> > > >> >> If what you need is to wait for a RCU grace period for whatever RCU flavor
> > > >> >> ebpf is using, I would against using membarrier for this. I would rather
> > > >> >> recommend adding a dedicated BPF_SYNCHRONIZE so you won't leak
> > > >> >> implementation details to user-space, *and* you can eventually change you
> > > >> >> RCU implementation for e.g. SRCU in the future if needed.
> > > >> >
> > > >> > The point about future changes to underlying bpf mechanisms is valid.
> > > >> > There is work already on the way to reduce the scope of preempt_off+rcu_lock
> > > >> > that currently lasts the whole prog. We will have new prog types that won't
> > > >> > have such wrappers and will do rcu_lock/unlock and preempt on/off only
> > > >> > when necessary.
> > > >> > So something like BPF_SYNCHRONIZE will break soon, since the kernel cannot have
> > > >> > guarantees on when programs finish. Calling this command BPF_SYNCHRONIZE_PROG
> > > >> > also won't make sense for the same reason.
> > > >> > What we can do it instead is to define synchronization barrier for
> > > >> > programs accessing maps. May be call it something like:
> > > >> > BPF_SYNC_MAP_ACCESS ?
> > > >>
> > > >> I'm not sure what you're proposing. In the case the commit message
> > > >> describes, a user-space program that wants to "drain" a map needs to
> > > >> be confident that the map won't change under it, even across multiple
> > > >> bpf system calls on that map. One way of doing that is to ensure that
> > > >> nothing that could possibly hold a reference to that map is still
> > > >> running. Are you proposing some kind of refcount-draining approach?
> > > >> Simple locking won't work, since BPF programs can't block, and I don't
> > > >> see right now how a simple barrier would help.
> > > >
> > > > I'm proposing few changes for your patch:
> > > > s/BPF_SYNCHRONIZE/BPF_SYNC_MAP_ACCESS/
> > > > and s/synchronize_sched/synchronize_rcu/
> > > > with detailed comment in uapi/bpf.h that has an example why folks
> > > > would want to use this new cmd.
> > >
> > > Thanks for clarifying.
> > >
> > > > I think the bpf maps will be rcu protected for foreseeable future
> > > > even when rcu_read_lock/unlock will be done by the programs instead of
> > > > kernel wrappers.
> > >
> > > Can we guarantee that we always obtain a map reference and dispose of
> > > that reference inside the same critical section?
> >
> > yep. the verifier will guarantee that.
> >
> > > If so, can BPF
> > > programs then disable preemption for as long as they'd like?
> >
> > you mean after the finish? no. only while running.
> > The verifier will match things like lookup/release, lock/unlock, preempt on/off
> > and will make sure there is no dangling preempt disable after program returns.
> >
> It might be a silly question but does this new bpf program type
> provide a way to customize where to hold the rcu_lock/unlock and
> enable/disable preemption when creating the program? Since the
> BPF_SYNC_MAP_ACCESS cmd sounds not very useful if it only protect each
> single map access instead of the whole program. For example, sometimes
> we have to read the same map multiple times when running a eBPF
> program and if the userspace changed the map value between two reads,
> it may cause troubles. It would be great if we can have a RCU lock on
> a block of eBPF program and in that way I think BPF_SYNC_MAP_ACCESS
> cmd should be enough in handling userspace kernel racing problems.

we need to make sure we have detailed description of BPF_SYNC_MAP_ACCESS
in uapi/bpf.h, since I feel the confusion regarding its usage is starting already.
This new cmd will only make sense for map-in-map type of maps.
Expecting that BPF_SYNC_MAP_ACCESS is somehow implies the end of
the program or doing some other map synchronization is not correct.
Commit log of this patch got it right:
"""
 For example, userspace can update a map->map entry to point to a new map,
 use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
 complete, and then drain the old map without fear that BPF programs
 may still be updating it.
"""
Now consider two cases:
1. single program does two lookups (inner and outer map) and the program is called
  twice (as trigger on two events or it was tail-called or through prog_array)
2. future single program without rcu wrappers does:
   rcu_lock + outer lookup + inner lookup + rcu_unlock + .. some other bpf stuff .. +
   rcu_lock + outer lookup + inner lookup + rcu_unlock + .. yet another bpf stuff ..
   All of the above as part of single program that is called once.
These two cases are the same from the point of view of BPF_SYNC_MAP_ACCESS
and completion of this cmd does not guarantee that program finished.
In both cases the user space will be correct to start draining the inner map
it replaced after BPF_SYNC_MAP_ACCESS is done
(assuming that implementation does synchronize_rcu I mentioned earlier)

Preemption on/off as done today by prog wrappers or in the future explicitly
by the prog code is orthogonal to this new cmd and map-in-map draining.

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Lorenzo Colitti]

On Wed, Jul 11, 2018 at 8:52 AM Alexei Starovoitov
<alexei.starovoitov@gmail.com> wrote:
>
> we need to make sure we have detailed description of BPF_SYNC_MAP_ACCESS
> in uapi/bpf.h, since I feel the confusion regarding its usage is starting already.
> This new cmd will only make sense for map-in-map type of maps.
> Expecting that BPF_SYNC_MAP_ACCESS is somehow implies the end of
> the program or doing some other map synchronization is not correct.
> Commit log of this patch got it right:
> """
>  For example, userspace can update a map->map entry to point to a new map,
>  use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
>  complete, and then drain the old map without fear that BPF programs
>  may still be updating it.
> """

+1 for detailed documentation. For example, consider what happens if
we have two map fds, one active and one standby, and a map-in-map with
one element that contains a pointer to the currently-active map fd.
The kernel program might do:

=====
const int current_map_key = 1;
void *current_map = bpf_map_lookup_elem(outer_map, &current_map_key);

int stats_key = 42;
uint64_t *stats_value = bpf_map_lookup_elem(current_map, &stats_key);
__sync_fetch_and_add(&stats_value, 1);
=====

If a userspace does:

1. Write new fd to outer_map[1].
2. Call BPF_SYNC_MAP_ACCESS.
3. Start deleting everything in the old map.

How can we guarantee that the __sync_fetch_and_add will not add to the
old map? If it does, we'll lose data. Will the verifier automatically
hold the RCU lock for as long as a pointer to an inner map is valid?

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Alexei Starovoitov]

On Wed, Jul 11, 2018 at 11:46:19AM +0900, Lorenzo Colitti wrote:
> On Wed, Jul 11, 2018 at 8:52 AM Alexei Starovoitov
> <alexei.starovoitov@gmail.com> wrote:
> >
> > we need to make sure we have detailed description of BPF_SYNC_MAP_ACCESS
> > in uapi/bpf.h, since I feel the confusion regarding its usage is starting already.
> > This new cmd will only make sense for map-in-map type of maps.
> > Expecting that BPF_SYNC_MAP_ACCESS is somehow implies the end of
> > the program or doing some other map synchronization is not correct.
> > Commit log of this patch got it right:
> > """
> >  For example, userspace can update a map->map entry to point to a new map,
> >  use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> >  complete, and then drain the old map without fear that BPF programs
> >  may still be updating it.
> > """
> 
> +1 for detailed documentation. For example, consider what happens if
> we have two map fds, one active and one standby, and a map-in-map with
> one element that contains a pointer to the currently-active map fd.

yes. that's exactly the use case that folks use.

> The kernel program might do:
> 
> =====
> const int current_map_key = 1;
> void *current_map = bpf_map_lookup_elem(outer_map, &current_map_key);
> 
> int stats_key = 42;
> uint64_t *stats_value = bpf_map_lookup_elem(current_map, &stats_key);
> __sync_fetch_and_add(&stats_value, 1);
> =====
> 
> If a userspace does:
> 
> 1. Write new fd to outer_map[1].
> 2. Call BPF_SYNC_MAP_ACCESS.
> 3. Start deleting everything in the old map.
> 
> How can we guarantee that the __sync_fetch_and_add will not add to the
> old map? 

without any changes to the kernel sys_membarrier will work.
And that's what folks use already.
BPF_SYNC_MAP_ACCESS implemented via synchronize_rcu() will work
as well whether in the current implementation where rcu_lock/unlock
is done outside of the program and in the future when
rcu_lock/unlock are called by the program itself.

> Will the verifier automatically
> hold the RCU lock for as long as a pointer to an inner map is valid?

the verifier will guarantee the equivalency of future explicit
lock/unlock by the program vs current situation of implicit
lock/unlock by the kernel.
The verifier will track that bpf_map_lookup_elem() is done
after rcu_lock and that the value returned by this helper is
not accessed after rcu_unlock. Baby steps of dataflow analysis.

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Joel Fernandes]

On Tue, Jul 10, 2018 at 08:40:19PM -0700, Alexei Starovoitov wrote:
[..]
> > The kernel program might do:
> > 
> > =====
> > const int current_map_key = 1;
> > void *current_map = bpf_map_lookup_elem(outer_map, &current_map_key);
> > 
> > int stats_key = 42;
> > uint64_t *stats_value = bpf_map_lookup_elem(current_map, &stats_key);
> > __sync_fetch_and_add(&stats_value, 1);
> > =====
> > 
> > If a userspace does:
> > 
> > 1. Write new fd to outer_map[1].
> > 2. Call BPF_SYNC_MAP_ACCESS.
> > 3. Start deleting everything in the old map.
> > 
> > How can we guarantee that the __sync_fetch_and_add will not add to the
> > old map? 
> 
> without any changes to the kernel sys_membarrier will work.
> And that's what folks use already.
> BPF_SYNC_MAP_ACCESS implemented via synchronize_rcu() will work
> as well whether in the current implementation where rcu_lock/unlock
> is done outside of the program and in the future when
> rcu_lock/unlock are called by the program itself.

Cool Alexei and Lorenzo, sounds great to me. Daniel want to send a follow up
patch with BPF_SYNC_MAP_ACCESS changes then?

> > Will the verifier automatically
> > hold the RCU lock for as long as a pointer to an inner map is valid?
> 
> the verifier will guarantee the equivalency of future explicit
> lock/unlock by the program vs current situation of implicit
> lock/unlock by the kernel.
> The verifier will track that bpf_map_lookup_elem() is done
> after rcu_lock and that the value returned by this helper is
> not accessed after rcu_unlock. Baby steps of dataflow analysis.

Nice!

By the way just curious I was briefly going through kernel/bpf/arraymap.c.
How are you protecting against load-store tearing of values of array map
updates/lookups?

For example, if userspace reads an array map at a particular index, while
another CPU is updating it, then userspace can read partial values /
half-updated values right? Since rcu_read_lock is in use, I was hoping to
find something like rcu_assign_pointer there to protect readers against
concurrent updates.  Thanks for any clarification.

Regards,

- Joel

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Daniel Colascione]

On Sat, Jul 14, 2018 at 11:18 AM, Joel Fernandes <joel@joelfernandes.org> wrote:
> On Tue, Jul 10, 2018 at 08:40:19PM -0700, Alexei Starovoitov wrote:
> [..]
>> > The kernel program might do:
>> >
>> > =====
>> > const int current_map_key = 1;
>> > void *current_map = bpf_map_lookup_elem(outer_map, &current_map_key);
>> >
>> > int stats_key = 42;
>> > uint64_t *stats_value = bpf_map_lookup_elem(current_map, &stats_key);
>> > __sync_fetch_and_add(&stats_value, 1);
>> > =====
>> >
>> > If a userspace does:
>> >
>> > 1. Write new fd to outer_map[1].
>> > 2. Call BPF_SYNC_MAP_ACCESS.
>> > 3. Start deleting everything in the old map.
>> >
>> > How can we guarantee that the __sync_fetch_and_add will not add to the
>> > old map?
>>
>> without any changes to the kernel sys_membarrier will work.
>> And that's what folks use already.
>> BPF_SYNC_MAP_ACCESS implemented via synchronize_rcu() will work
>> as well whether in the current implementation where rcu_lock/unlock
>> is done outside of the program and in the future when
>> rcu_lock/unlock are called by the program itself.
>
> Cool Alexei and Lorenzo, sounds great to me. Daniel want to send a follow up
> patch with BPF_SYNC_MAP_ACCESS changes then?

Will do. Mind if I just mine this thread for the doc comment?

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Joel Fernandes]

On Mon, Jul 16, 2018 at 08:29:47AM -0700, Daniel Colascione wrote:
> On Sat, Jul 14, 2018 at 11:18 AM, Joel Fernandes <joel@joelfernandes.org> wrote:
> > On Tue, Jul 10, 2018 at 08:40:19PM -0700, Alexei Starovoitov wrote:
> > [..]
> >> > The kernel program might do:
> >> >
> >> > =====
> >> > const int current_map_key = 1;
> >> > void *current_map = bpf_map_lookup_elem(outer_map, &current_map_key);
> >> >
> >> > int stats_key = 42;
> >> > uint64_t *stats_value = bpf_map_lookup_elem(current_map, &stats_key);
> >> > __sync_fetch_and_add(&stats_value, 1);
> >> > =====
> >> >
> >> > If a userspace does:
> >> >
> >> > 1. Write new fd to outer_map[1].
> >> > 2. Call BPF_SYNC_MAP_ACCESS.
> >> > 3. Start deleting everything in the old map.
> >> >
> >> > How can we guarantee that the __sync_fetch_and_add will not add to the
> >> > old map?
> >>
> >> without any changes to the kernel sys_membarrier will work.
> >> And that's what folks use already.
> >> BPF_SYNC_MAP_ACCESS implemented via synchronize_rcu() will work
> >> as well whether in the current implementation where rcu_lock/unlock
> >> is done outside of the program and in the future when
> >> rcu_lock/unlock are called by the program itself.
> >
> > Cool Alexei and Lorenzo, sounds great to me. Daniel want to send a follow up
> > patch with BPF_SYNC_MAP_ACCESS changes then?
> 
> Will do. Mind if I just mine this thread for the doc comment?

Do you mean the changelog? Yes I believe you could use the discussion in this
thread for the rationale as Alexei described.

thanks,

- Joel

[PATCH v2] Add BPF_SYNCHRONIZE_MAPS bpf(2) command

[Daniel Colascione]

BPF_SYNCHRONIZE_MAPS waits for the release of any references to a BPF
map made by a BPF program that is running at the time the
BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command is
to provide a means for userspace to replace a BPF map with another,
newer version, then ensure that no component is still using the "old"
map before manipulating the "old" map in some way.

Signed-off-by: Daniel Colascione <dancol@google.com>
---
 include/uapi/linux/bpf.h |  9 +++++++++
 kernel/bpf/syscall.c     | 13 +++++++++++++
 2 files changed, 22 insertions(+)

diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
index b7db3261c62d..5b27e9117d3e 100644
--- a/include/uapi/linux/bpf.h
+++ b/include/uapi/linux/bpf.h
@@ -75,6 +75,14 @@ struct bpf_lpm_trie_key {
 	__u8	data[0];	/* Arbitrary size */
 };
 
+/* BPF_SYNCHRONIZE_MAPS waits for the release of any references to a
+ * BPF map made by a BPF program that is running at the time the
+ * BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command
+ * is to provide a means for userspace to replace a BPF map with
+ * another, newer version, then ensure that no component is still
+ * using the "old" map before manipulating the "old" map in some way.
+ */
+
 /* BPF syscall commands, see bpf(2) man-page for details. */
 enum bpf_cmd {
 	BPF_MAP_CREATE,
@@ -98,6 +106,7 @@ enum bpf_cmd {
 	BPF_BTF_LOAD,
 	BPF_BTF_GET_FD_BY_ID,
 	BPF_TASK_FD_QUERY,
+	BPF_SYNCHRONIZE_MAPS,
 };
 
 enum bpf_map_type {
diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index a31a1ba0f8ea..8bbd1a5d01d1 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c
@@ -2274,6 +2274,19 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, siz
 	if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
 		return -EPERM;
 
+	if (cmd == BPF_SYNCHRONIZE_MAPS) {
+		if (uattr != NULL || size != 0)
+			return -EINVAL;
+		err = security_bpf(cmd, NULL, 0);
+		if (err < 0)
+			return err;
+		/* BPF programs always enter a critical section while
+		 * they have a map reference outstanding.
+		 */
+		synchronize_rcu();
+		return 0;
+	}
+
 	err = bpf_check_uarg_tail_zero(uattr, sizeof(attr), size);
 	if (err)
 		return err;
-- 
2.18.0.233.g985f88cf7e-goog

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Daniel Colascione]

[-- Attachment #1: Type: text/plain, Size: 659 bytes --]

On Sat, Jul 14, 2018 at 11:18 AM, Joel Fernandes <joel@joelfernandes.org>
wrote:

> By the way just curious I was briefly going through kernel/bpf/arraymap.c.
> How are you protecting against load-store tearing of values of array map
> updates/lookups?
>
> For example, if userspace reads an array map at a particular index, while
> another CPU is updating it, then userspace can read partial values /
> half-updated values right? Since rcu_read_lock is in use, I was hoping to
> find something like rcu_assign_pointer there to protect readers against
> concurrent updates.  Thanks for any clarification.


I'm also curious about the answer to this question.

[-- Attachment #2: Type: text/html, Size: 1002 bytes --]

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Joel Fernandes]

On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
> ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
> 
> > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
> >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> >> > RCU data structure operations with respect to active programs. For
> >> > example, userspace can update a map->map entry to point to a new map,
> >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> >> > complete, and then drain the old map without fear that BPF programs
> >> > may still be updating it.
> >> > 
> >> > Signed-off-by: Daniel Colascione <dancol@google.com>
> >> > ---
> >> >  include/uapi/linux/bpf.h |  1 +
> >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
> >> >  2 files changed, 15 insertions(+)
> >> > 
> >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> >> > index b7db3261c62d..4365c50e8055 100644
> >> > --- a/include/uapi/linux/bpf.h
> >> > +++ b/include/uapi/linux/bpf.h
> >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
> >> >  	BPF_BTF_LOAD,
> >> >  	BPF_BTF_GET_FD_BY_ID,
> >> >  	BPF_TASK_FD_QUERY,
> >> > +	BPF_SYNCHRONIZE,
> >> >  };
> >> >  
> >> >  enum bpf_map_type {
> >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> >> > index d10ecd78105f..60ec7811846e 100644
> >> > --- a/kernel/bpf/syscall.c
> >> > +++ b/kernel/bpf/syscall.c
> >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
> >> > uattr, unsigned int, siz
> >> >  	if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
> >> >  		return -EPERM;
> >> >  
> >> > +	if (cmd == BPF_SYNCHRONIZE) {
> >> > +		if (uattr != NULL || size != 0)
> >> > +			return -EINVAL;
> >> > +		err = security_bpf(cmd, NULL, 0);
> >> > +		if (err < 0)
> >> > +			return err;
> >> > +		/* BPF programs are run with preempt disabled, so
> >> > +		 * synchronize_sched is sufficient even with
> >> > +		 * RCU_PREEMPT.
> >> > +		 */
> >> > +		synchronize_sched();
> >> > +		return 0;
> >> 
> >> I don't think it's necessary. sys_membarrier() can do this already
> >> and some folks use it exactly for this use case.
> > 
> > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
> > though. No where does the manpage say membarrier should be implemented this
> > way so what happens if the implementation changes?
> > 
> > Further, membarrier manpage says that a memory barrier should be matched with
> > a matching barrier. In this use case there is no matching barrier, so it
> > makes it weirder.
> > 
> > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
> > fragile to depend on it for this?
> > 
> >        case MEMBARRIER_CMD_GLOBAL:
> >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
> >                if (tick_nohz_full_enabled())
> >                        return -EINVAL;
> >                if (num_online_cpus() > 1)
> >                        synchronize_sched();
> >                return 0;
> > 
> > 
> > Adding Mathieu as well who I believe is author/maintainer of membarrier.
> 
> See commit 907565337
> "Fix: Disable sys_membarrier when nohz_full is enabled"
> 
> "Userspace applications should be allowed to expect the membarrier system
> call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
> nohz_full CPUs, but synchronize_sched() does not take those into
> account."
> 
> So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
> only care about kernel preempt off critical sections.

Mathieu, Thanks a lot for your reply. I understand what you said and agree
with you. Slight OT, but I tried to go back to first principles and
understand how membarrier() uses synchronize_sched() for the "slow path" and
it didn't make immediate sense to me. Let me clarify my dillema..

My understanding is membarrier's MEMBARRIER_CMD_GLOBAL will employ
synchronize_sched to make sure all other CPUs aren't executing anymore in an
section of usercode that happen to be accessing memory that was written to
before the membarrier call was made. To do this, the system call will use
synchronize_sched to try to guarantee that all user-mode execution that
started before the membarrier call would be completed when the membarrier
call returns. This guarantees that without using a real memory barrier on the
"fast path", things work just fine and everyone wins.

But, going through RCU code, I see that a "RCU-sched quiecent state" on a CPU
may be reached when the CPU receives a timer tick while executing in user
mode:

void rcu_check_callbacks(int user)
{
	trace_rcu_utilization(TPS("Start scheduler-tick"));
	increment_cpu_stall_ticks();
	if (user || rcu_is_cpu_rrupt_from_idle()) {
[...]
		rcu_sched_qs();
		rcu_bh_qs();

The problem I see is the CPU could be executing usermode code at the time of
the RCU sched-QS. This IMO is enough reason for synchronize_sched() to
return, because the CPU in question just reported a QS (assuming all other
CPUs also happen to do so if they needed to).

Then I am wondering how does the membarrier call even work, the tick could
very well have interrupted the CPU while it was executing usermode code in
the middle of a set of instructions performing memory accesses. Reporting a
quiescent state at such an inopportune time would cause the membarrier call
to prematurely return, no? Sorry if I missed something.

The other question I have is about the whole "nohz-full doesn't work" thing.
I didn't fully understand why. RCU is already tracking the state of nohz-full
CPUs because the rcu dynticks code in (kernel/rcu/tree.c) monitors
transitions to and from usermode even if the timer tick is turned off. So why
would it not work?

thanks a lot!

 - Joel

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Paul E. McKenney]

On Mon, Jul 09, 2018 at 10:13:47PM -0700, Joel Fernandes wrote:
> On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
> > ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
> > 
> > > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
> > >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> > >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> > >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> > >> > RCU data structure operations with respect to active programs. For
> > >> > example, userspace can update a map->map entry to point to a new map,
> > >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> > >> > complete, and then drain the old map without fear that BPF programs
> > >> > may still be updating it.
> > >> > 
> > >> > Signed-off-by: Daniel Colascione <dancol@google.com>
> > >> > ---
> > >> >  include/uapi/linux/bpf.h |  1 +
> > >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
> > >> >  2 files changed, 15 insertions(+)
> > >> > 
> > >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > >> > index b7db3261c62d..4365c50e8055 100644
> > >> > --- a/include/uapi/linux/bpf.h
> > >> > +++ b/include/uapi/linux/bpf.h
> > >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
> > >> >  	BPF_BTF_LOAD,
> > >> >  	BPF_BTF_GET_FD_BY_ID,
> > >> >  	BPF_TASK_FD_QUERY,
> > >> > +	BPF_SYNCHRONIZE,
> > >> >  };
> > >> >  
> > >> >  enum bpf_map_type {
> > >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> > >> > index d10ecd78105f..60ec7811846e 100644
> > >> > --- a/kernel/bpf/syscall.c
> > >> > +++ b/kernel/bpf/syscall.c
> > >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
> > >> > uattr, unsigned int, siz
> > >> >  	if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
> > >> >  		return -EPERM;
> > >> >  
> > >> > +	if (cmd == BPF_SYNCHRONIZE) {
> > >> > +		if (uattr != NULL || size != 0)
> > >> > +			return -EINVAL;
> > >> > +		err = security_bpf(cmd, NULL, 0);
> > >> > +		if (err < 0)
> > >> > +			return err;
> > >> > +		/* BPF programs are run with preempt disabled, so
> > >> > +		 * synchronize_sched is sufficient even with
> > >> > +		 * RCU_PREEMPT.
> > >> > +		 */
> > >> > +		synchronize_sched();
> > >> > +		return 0;
> > >> 
> > >> I don't think it's necessary. sys_membarrier() can do this already
> > >> and some folks use it exactly for this use case.
> > > 
> > > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
> > > though. No where does the manpage say membarrier should be implemented this
> > > way so what happens if the implementation changes?
> > > 
> > > Further, membarrier manpage says that a memory barrier should be matched with
> > > a matching barrier. In this use case there is no matching barrier, so it
> > > makes it weirder.
> > > 
> > > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
> > > fragile to depend on it for this?
> > > 
> > >        case MEMBARRIER_CMD_GLOBAL:
> > >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
> > >                if (tick_nohz_full_enabled())
> > >                        return -EINVAL;
> > >                if (num_online_cpus() > 1)
> > >                        synchronize_sched();
> > >                return 0;
> > > 
> > > 
> > > Adding Mathieu as well who I believe is author/maintainer of membarrier.
> > 
> > See commit 907565337
> > "Fix: Disable sys_membarrier when nohz_full is enabled"
> > 
> > "Userspace applications should be allowed to expect the membarrier system
> > call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
> > nohz_full CPUs, but synchronize_sched() does not take those into
> > account."
> > 
> > So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
> > only care about kernel preempt off critical sections.
> 
> Mathieu, Thanks a lot for your reply. I understand what you said and agree
> with you. Slight OT, but I tried to go back to first principles and
> understand how membarrier() uses synchronize_sched() for the "slow path" and
> it didn't make immediate sense to me. Let me clarify my dillema..
> 
> My understanding is membarrier's MEMBARRIER_CMD_GLOBAL will employ
> synchronize_sched to make sure all other CPUs aren't executing anymore in an
> section of usercode that happen to be accessing memory that was written to
> before the membarrier call was made. To do this, the system call will use
> synchronize_sched to try to guarantee that all user-mode execution that
> started before the membarrier call would be completed when the membarrier
> call returns. This guarantees that without using a real memory barrier on the
> "fast path", things work just fine and everyone wins.
> 
> But, going through RCU code, I see that a "RCU-sched quiecent state" on a CPU
> may be reached when the CPU receives a timer tick while executing in user
> mode:
> 
> void rcu_check_callbacks(int user)
> {
> 	trace_rcu_utilization(TPS("Start scheduler-tick"));
> 	increment_cpu_stall_ticks();
> 	if (user || rcu_is_cpu_rrupt_from_idle()) {
> [...]
> 		rcu_sched_qs();
> 		rcu_bh_qs();
> 
> The problem I see is the CPU could be executing usermode code at the time of
> the RCU sched-QS. This IMO is enough reason for synchronize_sched() to
> return, because the CPU in question just reported a QS (assuming all other
> CPUs also happen to do so if they needed to).

This scenario will have inserted the needed smp_mb() into the userspace
instruction execution stream, as is required by the sys_membarrier
use cases.

> Then I am wondering how does the membarrier call even work, the tick could
> very well have interrupted the CPU while it was executing usermode code in
> the middle of a set of instructions performing memory accesses. Reporting a
> quiescent state at such an inopportune time would cause the membarrier call
> to prematurely return, no? Sorry if I missed something.

One way to think of sys_membarrier() is as something that promotes a
barrier() to an smp_mb().  This barrier then separates the target CPU's
accesses that the caller saw before the sys_membarrier() from that same
CPU's accesses that the caller will see after the sys_membarrier().

> The other question I have is about the whole "nohz-full doesn't work" thing.
> I didn't fully understand why. RCU is already tracking the state of nohz-full
> CPUs because the rcu dynticks code in (kernel/rcu/tree.c) monitors
> transitions to and from usermode even if the timer tick is turned off. So why
> would it not work?

In the nohz_full case, there is no need for sys_membarrier()'s call to
synchronize_sched() to interact directly with the nohz_full CPU.  It
can instead look at the target CPU's dyntick-idle state, and that state
would potentially have been set in the dim distant past, thus having
no effect on the target CPU's current execution.

							Thanx, Paul

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Joel Fernandes]

On Tue, Jul 10, 2018 at 09:42:12AM -0700, Paul E. McKenney wrote:
> On Mon, Jul 09, 2018 at 10:13:47PM -0700, Joel Fernandes wrote:
> > On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
> > > ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
> > > 
> > > > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
> > > >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> > > >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> > > >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> > > >> > RCU data structure operations with respect to active programs. For
> > > >> > example, userspace can update a map->map entry to point to a new map,
> > > >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> > > >> > complete, and then drain the old map without fear that BPF programs
> > > >> > may still be updating it.
> > > >> > 
> > > >> > Signed-off-by: Daniel Colascione <dancol@google.com>
> > > >> > ---
> > > >> >  include/uapi/linux/bpf.h |  1 +
> > > >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
> > > >> >  2 files changed, 15 insertions(+)
> > > >> > 
> > > >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > > >> > index b7db3261c62d..4365c50e8055 100644
> > > >> > --- a/include/uapi/linux/bpf.h
> > > >> > +++ b/include/uapi/linux/bpf.h
> > > >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
> > > >> >  	BPF_BTF_LOAD,
> > > >> >  	BPF_BTF_GET_FD_BY_ID,
> > > >> >  	BPF_TASK_FD_QUERY,
> > > >> > +	BPF_SYNCHRONIZE,
> > > >> >  };
> > > >> >  
> > > >> >  enum bpf_map_type {
> > > >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> > > >> > index d10ecd78105f..60ec7811846e 100644
> > > >> > --- a/kernel/bpf/syscall.c
> > > >> > +++ b/kernel/bpf/syscall.c
> > > >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
> > > >> > uattr, unsigned int, siz
> > > >> >  	if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
> > > >> >  		return -EPERM;
> > > >> >  
> > > >> > +	if (cmd == BPF_SYNCHRONIZE) {
> > > >> > +		if (uattr != NULL || size != 0)
> > > >> > +			return -EINVAL;
> > > >> > +		err = security_bpf(cmd, NULL, 0);
> > > >> > +		if (err < 0)
> > > >> > +			return err;
> > > >> > +		/* BPF programs are run with preempt disabled, so
> > > >> > +		 * synchronize_sched is sufficient even with
> > > >> > +		 * RCU_PREEMPT.
> > > >> > +		 */
> > > >> > +		synchronize_sched();
> > > >> > +		return 0;
> > > >> 
> > > >> I don't think it's necessary. sys_membarrier() can do this already
> > > >> and some folks use it exactly for this use case.
> > > > 
> > > > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
> > > > though. No where does the manpage say membarrier should be implemented this
> > > > way so what happens if the implementation changes?
> > > > 
> > > > Further, membarrier manpage says that a memory barrier should be matched with
> > > > a matching barrier. In this use case there is no matching barrier, so it
> > > > makes it weirder.
> > > > 
> > > > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
> > > > fragile to depend on it for this?
> > > > 
> > > >        case MEMBARRIER_CMD_GLOBAL:
> > > >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
> > > >                if (tick_nohz_full_enabled())
> > > >                        return -EINVAL;
> > > >                if (num_online_cpus() > 1)
> > > >                        synchronize_sched();
> > > >                return 0;
> > > > 
> > > > 
> > > > Adding Mathieu as well who I believe is author/maintainer of membarrier.
> > > 
> > > See commit 907565337
> > > "Fix: Disable sys_membarrier when nohz_full is enabled"
> > > 
> > > "Userspace applications should be allowed to expect the membarrier system
> > > call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
> > > nohz_full CPUs, but synchronize_sched() does not take those into
> > > account."
> > > 
> > > So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
> > > only care about kernel preempt off critical sections.
> > 
> > Mathieu, Thanks a lot for your reply. I understand what you said and agree
> > with you. Slight OT, but I tried to go back to first principles and
> > understand how membarrier() uses synchronize_sched() for the "slow path" and
> > it didn't make immediate sense to me. Let me clarify my dillema..
> > 
> > My understanding is membarrier's MEMBARRIER_CMD_GLOBAL will employ
> > synchronize_sched to make sure all other CPUs aren't executing anymore in an
> > section of usercode that happen to be accessing memory that was written to
> > before the membarrier call was made. To do this, the system call will use
> > synchronize_sched to try to guarantee that all user-mode execution that
> > started before the membarrier call would be completed when the membarrier
> > call returns. This guarantees that without using a real memory barrier on the
> > "fast path", things work just fine and everyone wins.
> > 
> > But, going through RCU code, I see that a "RCU-sched quiecent state" on a CPU
> > may be reached when the CPU receives a timer tick while executing in user
> > mode:
> > 
> > void rcu_check_callbacks(int user)
> > {
> > 	trace_rcu_utilization(TPS("Start scheduler-tick"));
> > 	increment_cpu_stall_ticks();
> > 	if (user || rcu_is_cpu_rrupt_from_idle()) {
> > [...]
> > 		rcu_sched_qs();
> > 		rcu_bh_qs();
> > 
> > The problem I see is the CPU could be executing usermode code at the time of
> > the RCU sched-QS. This IMO is enough reason for synchronize_sched() to
> > return, because the CPU in question just reported a QS (assuming all other
> > CPUs also happen to do so if they needed to).
> 
> This scenario will have inserted the needed smp_mb() into the userspace
> instruction execution stream, as is required by the sys_membarrier
> use cases.

Oh ok, that makes sense!

> > Then I am wondering how does the membarrier call even work, the tick could
> > very well have interrupted the CPU while it was executing usermode code in
> > the middle of a set of instructions performing memory accesses. Reporting a
> > quiescent state at such an inopportune time would cause the membarrier call
> > to prematurely return, no? Sorry if I missed something.
> 
> One way to think of sys_membarrier() is as something that promotes a
> barrier() to an smp_mb().  This barrier then separates the target CPU's
> accesses that the caller saw before the sys_membarrier() from that same
> CPU's accesses that the caller will see after the sys_membarrier().

Got it!

> > The other question I have is about the whole "nohz-full doesn't work" thing.
> > I didn't fully understand why. RCU is already tracking the state of nohz-full
> > CPUs because the rcu dynticks code in (kernel/rcu/tree.c) monitors
> > transitions to and from usermode even if the timer tick is turned off. So why
> > would it not work?
> 
> In the nohz_full case, there is no need for sys_membarrier()'s call to
> synchronize_sched() to interact directly with the nohz_full CPU.  It
> can instead look at the target CPU's dyntick-idle state, and that state
> would potentially have been set in the dim distant past, thus having
> no effect on the target CPU's current execution.

In nohz-idle case though, there's nothing to promote the barrier() to
smp_mb() if you were to purely look at the dynticks-idle state on the
nohz-full CPU executing in user mode?

So then it makes sense to me now that nohz-full needs something to IPI that
CPU inorder to enforce the needed memory barrier and pure synchronize_sched()
wouldn't work. So then makes me think the expedited versions of
synchronize_sched should be able to do the job but I could off on a different
track..

Thanks a lot,

-Joel

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Paul E. McKenney]

On Tue, Jul 10, 2018 at 09:57:44AM -0700, Joel Fernandes wrote:
> On Tue, Jul 10, 2018 at 09:42:12AM -0700, Paul E. McKenney wrote:
> > On Mon, Jul 09, 2018 at 10:13:47PM -0700, Joel Fernandes wrote:
> > > On Sun, Jul 08, 2018 at 04:54:38PM -0400, Mathieu Desnoyers wrote:
> > > > ----- On Jul 7, 2018, at 4:33 PM, Joel Fernandes joelaf@google.com wrote:
> > > > 
> > > > > On Fri, Jul 06, 2018 at 07:54:28PM -0700, Alexei Starovoitov wrote:
> > > > >> On Fri, Jul 06, 2018 at 06:56:16PM -0700, Daniel Colascione wrote:
> > > > >> > BPF_SYNCHRONIZE waits for any BPF programs active at the time of
> > > > >> > BPF_SYNCHRONIZE to complete, allowing userspace to ensure atomicity of
> > > > >> > RCU data structure operations with respect to active programs. For
> > > > >> > example, userspace can update a map->map entry to point to a new map,
> > > > >> > use BPF_SYNCHRONIZE to wait for any BPF programs using the old map to
> > > > >> > complete, and then drain the old map without fear that BPF programs
> > > > >> > may still be updating it.
> > > > >> > 
> > > > >> > Signed-off-by: Daniel Colascione <dancol@google.com>
> > > > >> > ---
> > > > >> >  include/uapi/linux/bpf.h |  1 +
> > > > >> >  kernel/bpf/syscall.c     | 14 ++++++++++++++
> > > > >> >  2 files changed, 15 insertions(+)
> > > > >> > 
> > > > >> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > > > >> > index b7db3261c62d..4365c50e8055 100644
> > > > >> > --- a/include/uapi/linux/bpf.h
> > > > >> > +++ b/include/uapi/linux/bpf.h
> > > > >> > @@ -98,6 +98,7 @@ enum bpf_cmd {
> > > > >> >  	BPF_BTF_LOAD,
> > > > >> >  	BPF_BTF_GET_FD_BY_ID,
> > > > >> >  	BPF_TASK_FD_QUERY,
> > > > >> > +	BPF_SYNCHRONIZE,
> > > > >> >  };
> > > > >> >  
> > > > >> >  enum bpf_map_type {
> > > > >> > diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
> > > > >> > index d10ecd78105f..60ec7811846e 100644
> > > > >> > --- a/kernel/bpf/syscall.c
> > > > >> > +++ b/kernel/bpf/syscall.c
> > > > >> > @@ -2272,6 +2272,20 @@ SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *,
> > > > >> > uattr, unsigned int, siz
> > > > >> >  	if (sysctl_unprivileged_bpf_disabled && !capable(CAP_SYS_ADMIN))
> > > > >> >  		return -EPERM;
> > > > >> >  
> > > > >> > +	if (cmd == BPF_SYNCHRONIZE) {
> > > > >> > +		if (uattr != NULL || size != 0)
> > > > >> > +			return -EINVAL;
> > > > >> > +		err = security_bpf(cmd, NULL, 0);
> > > > >> > +		if (err < 0)
> > > > >> > +			return err;
> > > > >> > +		/* BPF programs are run with preempt disabled, so
> > > > >> > +		 * synchronize_sched is sufficient even with
> > > > >> > +		 * RCU_PREEMPT.
> > > > >> > +		 */
> > > > >> > +		synchronize_sched();
> > > > >> > +		return 0;
> > > > >> 
> > > > >> I don't think it's necessary. sys_membarrier() can do this already
> > > > >> and some folks use it exactly for this use case.
> > > > > 
> > > > > Alexei, the use of sys_membarrier for this purpose seems kind of weird to me
> > > > > though. No where does the manpage say membarrier should be implemented this
> > > > > way so what happens if the implementation changes?
> > > > > 
> > > > > Further, membarrier manpage says that a memory barrier should be matched with
> > > > > a matching barrier. In this use case there is no matching barrier, so it
> > > > > makes it weirder.
> > > > > 
> > > > > Lastly, sys_membarrier seems will not work on nohz-full systems, so its a bit
> > > > > fragile to depend on it for this?
> > > > > 
> > > > >        case MEMBARRIER_CMD_GLOBAL:
> > > > >                /* MEMBARRIER_CMD_GLOBAL is not compatible with nohz_full. */
> > > > >                if (tick_nohz_full_enabled())
> > > > >                        return -EINVAL;
> > > > >                if (num_online_cpus() > 1)
> > > > >                        synchronize_sched();
> > > > >                return 0;
> > > > > 
> > > > > 
> > > > > Adding Mathieu as well who I believe is author/maintainer of membarrier.
> > > > 
> > > > See commit 907565337
> > > > "Fix: Disable sys_membarrier when nohz_full is enabled"
> > > > 
> > > > "Userspace applications should be allowed to expect the membarrier system
> > > > call with MEMBARRIER_CMD_SHARED command to issue memory barriers on
> > > > nohz_full CPUs, but synchronize_sched() does not take those into
> > > > account."
> > > > 
> > > > So AFAIU you'd want to re-use membarrier to issue synchronize_sched, and you
> > > > only care about kernel preempt off critical sections.
> > > 
> > > Mathieu, Thanks a lot for your reply. I understand what you said and agree
> > > with you. Slight OT, but I tried to go back to first principles and
> > > understand how membarrier() uses synchronize_sched() for the "slow path" and
> > > it didn't make immediate sense to me. Let me clarify my dillema..
> > > 
> > > My understanding is membarrier's MEMBARRIER_CMD_GLOBAL will employ
> > > synchronize_sched to make sure all other CPUs aren't executing anymore in an
> > > section of usercode that happen to be accessing memory that was written to
> > > before the membarrier call was made. To do this, the system call will use
> > > synchronize_sched to try to guarantee that all user-mode execution that
> > > started before the membarrier call would be completed when the membarrier
> > > call returns. This guarantees that without using a real memory barrier on the
> > > "fast path", things work just fine and everyone wins.
> > > 
> > > But, going through RCU code, I see that a "RCU-sched quiecent state" on a CPU
> > > may be reached when the CPU receives a timer tick while executing in user
> > > mode:
> > > 
> > > void rcu_check_callbacks(int user)
> > > {
> > > 	trace_rcu_utilization(TPS("Start scheduler-tick"));
> > > 	increment_cpu_stall_ticks();
> > > 	if (user || rcu_is_cpu_rrupt_from_idle()) {
> > > [...]
> > > 		rcu_sched_qs();
> > > 		rcu_bh_qs();
> > > 
> > > The problem I see is the CPU could be executing usermode code at the time of
> > > the RCU sched-QS. This IMO is enough reason for synchronize_sched() to
> > > return, because the CPU in question just reported a QS (assuming all other
> > > CPUs also happen to do so if they needed to).
> > 
> > This scenario will have inserted the needed smp_mb() into the userspace
> > instruction execution stream, as is required by the sys_membarrier
> > use cases.
> 
> Oh ok, that makes sense!
> 
> > > Then I am wondering how does the membarrier call even work, the tick could
> > > very well have interrupted the CPU while it was executing usermode code in
> > > the middle of a set of instructions performing memory accesses. Reporting a
> > > quiescent state at such an inopportune time would cause the membarrier call
> > > to prematurely return, no? Sorry if I missed something.
> > 
> > One way to think of sys_membarrier() is as something that promotes a
> > barrier() to an smp_mb().  This barrier then separates the target CPU's
> > accesses that the caller saw before the sys_membarrier() from that same
> > CPU's accesses that the caller will see after the sys_membarrier().
> 
> Got it!
> 
> > > The other question I have is about the whole "nohz-full doesn't work" thing.
> > > I didn't fully understand why. RCU is already tracking the state of nohz-full
> > > CPUs because the rcu dynticks code in (kernel/rcu/tree.c) monitors
> > > transitions to and from usermode even if the timer tick is turned off. So why
> > > would it not work?
> > 
> > In the nohz_full case, there is no need for sys_membarrier()'s call to
> > synchronize_sched() to interact directly with the nohz_full CPU.  It
> > can instead look at the target CPU's dyntick-idle state, and that state
> > would potentially have been set in the dim distant past, thus having
> > no effect on the target CPU's current execution.
> 
> In nohz-idle case though, there's nothing to promote the barrier() to
> smp_mb() if you were to purely look at the dynticks-idle state on the
> nohz-full CPU executing in user mode?
> 
> So then it makes sense to me now that nohz-full needs something to IPI that
> CPU inorder to enforce the needed memory barrier and pure synchronize_sched()
> wouldn't work. So then makes me think the expedited versions of
> synchronize_sched should be able to do the job but I could off on a different
> track..

The problem is that the expedited versions also check the dyntick-idle
state and don't touch idle (or nohz_full usermode) CPUs.  This is by
design for the battery-powered embedded use case.  ;-)

							Thanx, Paul

> Thanks a lot,
> 
> -Joel
> 
> 

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Joel Fernandes]

On Tue, Jul 10, 2018 at 10:12:29AM -0700, Paul E. McKenney wrote:
[..]
> > > > The other question I have is about the whole "nohz-full doesn't work" thing.
> > > > I didn't fully understand why. RCU is already tracking the state of nohz-full
> > > > CPUs because the rcu dynticks code in (kernel/rcu/tree.c) monitors
> > > > transitions to and from usermode even if the timer tick is turned off. So why
> > > > would it not work?
> > > 
> > > In the nohz_full case, there is no need for sys_membarrier()'s call to
> > > synchronize_sched() to interact directly with the nohz_full CPU.  It
> > > can instead look at the target CPU's dyntick-idle state, and that state
> > > would potentially have been set in the dim distant past, thus having
> > > no effect on the target CPU's current execution.
> > 
> > In nohz-idle case though, there's nothing to promote the barrier() to
> > smp_mb() if you were to purely look at the dynticks-idle state on the
> > nohz-full CPU executing in user mode?
> > 
> > So then it makes sense to me now that nohz-full needs something to IPI that
> > CPU inorder to enforce the needed memory barrier and pure synchronize_sched()
> > wouldn't work. So then makes me think the expedited versions of
> > synchronize_sched should be able to do the job but I could off on a different
> > track..
> 
> The problem is that the expedited versions also check the dyntick-idle
> state and don't touch idle (or nohz_full usermode) CPUs.  This is by
> design for the battery-powered embedded use case.  ;-)

Oh ok! ;)

I guess there's also a MEMBARRIER_CMD_GLOBAL_EXPEDITED which seems to IPI
CPUs (I'm guessing regardless of dynticks state) and execute smp_mb within
the IPI so userspace can fallback to using that incase MEMBARRIER_CMD_GLOBAL
returns -EINVAL.

thanks!

-Joel

Re: [RFC] Add BPF_SYNCHRONIZE bpf(2) command

[Paul E. McKenney]

On Tue, Jul 10, 2018 at 10:29:57AM -0700, Joel Fernandes wrote:
> On Tue, Jul 10, 2018 at 10:12:29AM -0700, Paul E. McKenney wrote:
> [..]
> > > > > The other question I have is about the whole "nohz-full doesn't work" thing.
> > > > > I didn't fully understand why. RCU is already tracking the state of nohz-full
> > > > > CPUs because the rcu dynticks code in (kernel/rcu/tree.c) monitors
> > > > > transitions to and from usermode even if the timer tick is turned off. So why
> > > > > would it not work?
> > > > 
> > > > In the nohz_full case, there is no need for sys_membarrier()'s call to
> > > > synchronize_sched() to interact directly with the nohz_full CPU.  It
> > > > can instead look at the target CPU's dyntick-idle state, and that state
> > > > would potentially have been set in the dim distant past, thus having
> > > > no effect on the target CPU's current execution.
> > > 
> > > In nohz-idle case though, there's nothing to promote the barrier() to
> > > smp_mb() if you were to purely look at the dynticks-idle state on the
> > > nohz-full CPU executing in user mode?
> > > 
> > > So then it makes sense to me now that nohz-full needs something to IPI that
> > > CPU inorder to enforce the needed memory barrier and pure synchronize_sched()
> > > wouldn't work. So then makes me think the expedited versions of
> > > synchronize_sched should be able to do the job but I could off on a different
> > > track..
> > 
> > The problem is that the expedited versions also check the dyntick-idle
> > state and don't touch idle (or nohz_full usermode) CPUs.  This is by
> > design for the battery-powered embedded use case.  ;-)
> 
> Oh ok! ;)
> 
> I guess there's also a MEMBARRIER_CMD_GLOBAL_EXPEDITED which seems to IPI
> CPUs (I'm guessing regardless of dynticks state) and execute smp_mb within
> the IPI so userspace can fallback to using that incase MEMBARRIER_CMD_GLOBAL
> returns -EINVAL.

Yes, and this avoids IPIing idle CPUs via the ->mm checks.  But it will
IPI nohz_full CPUs in that same process, as it must for correctness.

							Thanx, Paul

Re: [PATCH v2] Add BPF_SYNCHRONIZE_MAPS bpf(2) command

[Alexei Starovoitov]

On Thu, Jul 26, 2018 at 7:51 PM, Daniel Colascione <dancol@google.com> wrote:
> BPF_SYNCHRONIZE_MAPS waits for the release of any references to a BPF
> map made by a BPF program that is running at the time the
> BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command is
> to provide a means for userspace to replace a BPF map with another,
> newer version, then ensure that no component is still using the "old"
> map before manipulating the "old" map in some way.
>
> Signed-off-by: Daniel Colascione <dancol@google.com>
> ---
>  include/uapi/linux/bpf.h |  9 +++++++++
>  kernel/bpf/syscall.c     | 13 +++++++++++++
>  2 files changed, 22 insertions(+)
>
> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> index b7db3261c62d..5b27e9117d3e 100644
> --- a/include/uapi/linux/bpf.h
> +++ b/include/uapi/linux/bpf.h
> @@ -75,6 +75,14 @@ struct bpf_lpm_trie_key {
>         __u8    data[0];        /* Arbitrary size */
>  };
>
> +/* BPF_SYNCHRONIZE_MAPS waits for the release of any references to a
> + * BPF map made by a BPF program that is running at the time the
> + * BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command

that doesn't sound right to me.
such command won't wait for the release of the references.
in case of map-in-map the program does not hold
the references to inner map (only to outer map).

> + * is to provide a means for userspace to replace a BPF map with
> + * another, newer version, then ensure that no component is still
> + * using the "old" map before manipulating the "old" map in some way.
> + */

that's correct, but the whole paragraph still reads too
'generic' which will lead to confusion,
whereas the use case is map-in-map only.
I think bpf_sync_inner_map or
bpf_sync_map_in_map would be better
choices for the name.

btw i don't have reliable internet access at the moment,
so pls excuse the delays.

Re: [PATCH v2] Add BPF_SYNCHRONIZE_MAPS bpf(2) command

[Daniel Colascione]

[-- Attachment #1: Type: text/plain, Size: 2236 bytes --]

On Sun, Jul 29, 2018 at 8:51 AM, Alexei Starovoitov <
alexei.starovoitov@gmail.com> wrote:

> On Thu, Jul 26, 2018 at 7:51 PM, Daniel Colascione <dancol@google.com>
> wrote:
> > BPF_SYNCHRONIZE_MAPS waits for the release of any references to a BPF
> > map made by a BPF program that is running at the time the
> > BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command is
> > to provide a means for userspace to replace a BPF map with another,
> > newer version, then ensure that no component is still using the "old"
> > map before manipulating the "old" map in some way.
> >
> > Signed-off-by: Daniel Colascione <dancol@google.com>
> > ---
> >  include/uapi/linux/bpf.h |  9 +++++++++
> >  kernel/bpf/syscall.c     | 13 +++++++++++++
> >  2 files changed, 22 insertions(+)
> >
> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > index b7db3261c62d..5b27e9117d3e 100644
> > --- a/include/uapi/linux/bpf.h
> > +++ b/include/uapi/linux/bpf.h
> > @@ -75,6 +75,14 @@ struct bpf_lpm_trie_key {
> >         __u8    data[0];        /* Arbitrary size */
> >  };
> >
> > +/* BPF_SYNCHRONIZE_MAPS waits for the release of any references to a
> > + * BPF map made by a BPF program that is running at the time the
> > + * BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command
>
> that doesn't sound right to me.
> such command won't wait for the release of the references.
> in case of map-in-map the program does not hold
> the references to inner map (only to outer map).
>

Well, today that's the guarantee it provides. :-) I figured it'd be simpler
to explain the guarantee this way.

How about "waits for the release of any reference to a map obtained from
another map"?


>
> > + * is to provide a means for userspace to replace a BPF map with
> > + * another, newer version, then ensure that no component is still
> > + * using the "old" map before manipulating the "old" map in some way.
> > + */
>
> that's correct, but the whole paragraph still reads too
> 'generic' which will lead to confusion,
> whereas the use case is map-in-map only.
> I think bpf_sync_inner_map or
> bpf_sync_map_in_map would be better
> choices for the name.


I worry that a name like that would be too specific.

[-- Attachment #2: Type: text/html, Size: 3163 bytes --]

Re: [PATCH v2] Add BPF_SYNCHRONIZE_MAPS bpf(2) command

[Joel Fernandes]

On Sun, Jul 29, 2018 at 06:51:18PM +0300, Alexei Starovoitov wrote:
> On Thu, Jul 26, 2018 at 7:51 PM, Daniel Colascione <dancol@google.com> wrote:
> > BPF_SYNCHRONIZE_MAPS waits for the release of any references to a BPF
> > map made by a BPF program that is running at the time the
> > BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command is
> > to provide a means for userspace to replace a BPF map with another,
> > newer version, then ensure that no component is still using the "old"
> > map before manipulating the "old" map in some way.
> >
> > Signed-off-by: Daniel Colascione <dancol@google.com>
> > ---
> >  include/uapi/linux/bpf.h |  9 +++++++++
> >  kernel/bpf/syscall.c     | 13 +++++++++++++
> >  2 files changed, 22 insertions(+)
> >
> > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > index b7db3261c62d..5b27e9117d3e 100644
> > --- a/include/uapi/linux/bpf.h
> > +++ b/include/uapi/linux/bpf.h
> > @@ -75,6 +75,14 @@ struct bpf_lpm_trie_key {
> >         __u8    data[0];        /* Arbitrary size */
> >  };
> >
> > +/* BPF_SYNCHRONIZE_MAPS waits for the release of any references to a
> > + * BPF map made by a BPF program that is running at the time the
> > + * BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command
> 
> that doesn't sound right to me.
> such command won't wait for the release of the references.
> in case of map-in-map the program does not hold
> the references to inner map (only to outer map).

I didn't follow this completely.

The userspace program is using the inner map per your description of the
algorithm for using map-in-map to solve the race conditions that this patch
is trying to address:

If you don't mind, I copy-pasted it below from your netdev post:

if you use map-in-map you don't need extra boolean map.
0. bpf prog can do
   inner_map = lookup(map_in_map, key=0);
   lookup(inner_map, your_real_key);
1. user space writes into map_in_map[0] <- FD of new map
2. some cpus are using old inner map and some a new
3. user space does sys_membarrier(CMD_GLOBAL) which will do synchronize_sched()
   which in CONFIG_PREEMPT_NONE=y servers is the same as synchronize_rcu()
   which will guarantee that progs finished.
4. scan old inner map

In step 2, as you mentioned there are CPUs using different inner maps. So
could you clarify how the synchronize_rcu mechanism will even work if you're
now saying "program does not hold references to the inner maps"? 

Thanks!

- Joel

Re: [PATCH v2] Add BPF_SYNCHRONIZE_MAPS bpf(2) command

[Joel Fernandes]

On Mon, Jul 30, 2018 at 07:01:22PM -0700, Joel Fernandes wrote:
> On Sun, Jul 29, 2018 at 06:51:18PM +0300, Alexei Starovoitov wrote:
> > On Thu, Jul 26, 2018 at 7:51 PM, Daniel Colascione <dancol@google.com> wrote:
> > > BPF_SYNCHRONIZE_MAPS waits for the release of any references to a BPF
> > > map made by a BPF program that is running at the time the
> > > BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command is
> > > to provide a means for userspace to replace a BPF map with another,
> > > newer version, then ensure that no component is still using the "old"
> > > map before manipulating the "old" map in some way.
> > >
> > > Signed-off-by: Daniel Colascione <dancol@google.com>
> > > ---
> > >  include/uapi/linux/bpf.h |  9 +++++++++
> > >  kernel/bpf/syscall.c     | 13 +++++++++++++
> > >  2 files changed, 22 insertions(+)
> > >
> > > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> > > index b7db3261c62d..5b27e9117d3e 100644
> > > --- a/include/uapi/linux/bpf.h
> > > +++ b/include/uapi/linux/bpf.h
> > > @@ -75,6 +75,14 @@ struct bpf_lpm_trie_key {
> > >         __u8    data[0];        /* Arbitrary size */
> > >  };
> > >
> > > +/* BPF_SYNCHRONIZE_MAPS waits for the release of any references to a
> > > + * BPF map made by a BPF program that is running at the time the
> > > + * BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command
> > 
> > that doesn't sound right to me.
> > such command won't wait for the release of the references.
> > in case of map-in-map the program does not hold
> > the references to inner map (only to outer map).
> 
> I didn't follow this completely.
> 
> The userspace program is using the inner map per your description of the

Sorry just to correct myself, here I meant "The kernel eBPF program is using
the inner map on multiple CPUs" instead of "userspace".

thanks,

- Joel





> algorithm for using map-in-map to solve the race conditions that this patch
> is trying to address:
> 
> If you don't mind, I copy-pasted it below from your netdev post:
> 
> if you use map-in-map you don't need extra boolean map.
> 0. bpf prog can do
>    inner_map = lookup(map_in_map, key=0);
>    lookup(inner_map, your_real_key);
> 1. user space writes into map_in_map[0] <- FD of new map
> 2. some cpus are using old inner map and some a new
> 3. user space does sys_membarrier(CMD_GLOBAL) which will do synchronize_sched()
>    which in CONFIG_PREEMPT_NONE=y servers is the same as synchronize_rcu()
>    which will guarantee that progs finished.
> 4. scan old inner map
> 
> In step 2, as you mentioned there are CPUs using different inner maps. So
> could you clarify how the synchronize_rcu mechanism will even work if you're
> now saying "program does not hold references to the inner maps"? 
> 
> Thanks!
> 
> - Joel
> 

Re: [PATCH v2] Add BPF_SYNCHRONIZE_MAPS bpf(2) command

[Y Song]

On Mon, Jul 30, 2018 at 7:06 PM, Joel Fernandes <joel@joelfernandes.org> wrote:
> On Mon, Jul 30, 2018 at 07:01:22PM -0700, Joel Fernandes wrote:
>> On Sun, Jul 29, 2018 at 06:51:18PM +0300, Alexei Starovoitov wrote:
>> > On Thu, Jul 26, 2018 at 7:51 PM, Daniel Colascione <dancol@google.com> wrote:
>> > > BPF_SYNCHRONIZE_MAPS waits for the release of any references to a BPF
>> > > map made by a BPF program that is running at the time the
>> > > BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command is
>> > > to provide a means for userspace to replace a BPF map with another,
>> > > newer version, then ensure that no component is still using the "old"
>> > > map before manipulating the "old" map in some way.
>> > >
>> > > Signed-off-by: Daniel Colascione <dancol@google.com>
>> > > ---
>> > >  include/uapi/linux/bpf.h |  9 +++++++++
>> > >  kernel/bpf/syscall.c     | 13 +++++++++++++
>> > >  2 files changed, 22 insertions(+)
>> > >
>> > > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
>> > > index b7db3261c62d..5b27e9117d3e 100644
>> > > --- a/include/uapi/linux/bpf.h
>> > > +++ b/include/uapi/linux/bpf.h
>> > > @@ -75,6 +75,14 @@ struct bpf_lpm_trie_key {
>> > >         __u8    data[0];        /* Arbitrary size */
>> > >  };
>> > >
>> > > +/* BPF_SYNCHRONIZE_MAPS waits for the release of any references to a
>> > > + * BPF map made by a BPF program that is running at the time the
>> > > + * BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command
>> >
>> > that doesn't sound right to me.
>> > such command won't wait for the release of the references.
>> > in case of map-in-map the program does not hold
>> > the references to inner map (only to outer map).
>>
>> I didn't follow this completely.
>>
>> The userspace program is using the inner map per your description of the
>
> Sorry just to correct myself, here I meant "The kernel eBPF program is using
> the inner map on multiple CPUs" instead of "userspace".
>
> thanks,
>
> - Joel
>
>
>
>
>
>> algorithm for using map-in-map to solve the race conditions that this patch
>> is trying to address:
>>
>> If you don't mind, I copy-pasted it below from your netdev post:
>>
>> if you use map-in-map you don't need extra boolean map.
>> 0. bpf prog can do
>>    inner_map = lookup(map_in_map, key=0);
>>    lookup(inner_map, your_real_key);
>> 1. user space writes into map_in_map[0] <- FD of new map
>> 2. some cpus are using old inner map and some a new
>> 3. user space does sys_membarrier(CMD_GLOBAL) which will do synchronize_sched()
>>    which in CONFIG_PREEMPT_NONE=y servers is the same as synchronize_rcu()
>>    which will guarantee that progs finished.
>> 4. scan old inner map
>>
>> In step 2, as you mentioned there are CPUs using different inner maps. So
>> could you clarify how the synchronize_rcu mechanism will even work if you're
>> now saying "program does not hold references to the inner maps"?

The program only held references to the outer maps, and the outer map
held references to the inner maps. The user space program can add/remove
the inner map for a particular outer map while the prog <-> outer-map
relationship is not changed.

>>
>> Thanks!
>>
>> - Joel
>>

Re: [PATCH v2] Add BPF_SYNCHRONIZE_MAPS bpf(2) command

[Joel Fernandes]

On Mon, Jul 30, 2018 at 09:03:18PM -0700, Y Song wrote:
> On Mon, Jul 30, 2018 at 7:06 PM, Joel Fernandes <joel@joelfernandes.org> wrote:
> > On Mon, Jul 30, 2018 at 07:01:22PM -0700, Joel Fernandes wrote:
> >> On Sun, Jul 29, 2018 at 06:51:18PM +0300, Alexei Starovoitov wrote:
> >> > On Thu, Jul 26, 2018 at 7:51 PM, Daniel Colascione <dancol@google.com> wrote:
> >> > > BPF_SYNCHRONIZE_MAPS waits for the release of any references to a BPF
> >> > > map made by a BPF program that is running at the time the
> >> > > BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command is
> >> > > to provide a means for userspace to replace a BPF map with another,
> >> > > newer version, then ensure that no component is still using the "old"
> >> > > map before manipulating the "old" map in some way.
> >> > >
> >> > > Signed-off-by: Daniel Colascione <dancol@google.com>
> >> > > ---
> >> > >  include/uapi/linux/bpf.h |  9 +++++++++
> >> > >  kernel/bpf/syscall.c     | 13 +++++++++++++
> >> > >  2 files changed, 22 insertions(+)
> >> > >
> >> > > diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
> >> > > index b7db3261c62d..5b27e9117d3e 100644
> >> > > --- a/include/uapi/linux/bpf.h
> >> > > +++ b/include/uapi/linux/bpf.h
> >> > > @@ -75,6 +75,14 @@ struct bpf_lpm_trie_key {
> >> > >         __u8    data[0];        /* Arbitrary size */
> >> > >  };
> >> > >
> >> > > +/* BPF_SYNCHRONIZE_MAPS waits for the release of any references to a
> >> > > + * BPF map made by a BPF program that is running at the time the
> >> > > + * BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command
> >> >
> >> > that doesn't sound right to me.
> >> > such command won't wait for the release of the references.
> >> > in case of map-in-map the program does not hold
> >> > the references to inner map (only to outer map).
> >>
> >> I didn't follow this completely.
> >>
> >> The userspace program is using the inner map per your description of the
> >> algorithm for using map-in-map to solve the race conditions that this patch
> >> is trying to address:
> >>
> >> If you don't mind, I copy-pasted it below from your netdev post:
> >>
> >> if you use map-in-map you don't need extra boolean map.
> >> 0. bpf prog can do
> >>    inner_map = lookup(map_in_map, key=0);
> >>    lookup(inner_map, your_real_key);
> >> 1. user space writes into map_in_map[0] <- FD of new map
> >> 2. some cpus are using old inner map and some a new
> >> 3. user space does sys_membarrier(CMD_GLOBAL) which will do synchronize_sched()
> >>    which in CONFIG_PREEMPT_NONE=y servers is the same as synchronize_rcu()
> >>    which will guarantee that progs finished.
> >> 4. scan old inner map
> >>
> >> In step 2, as you mentioned there are CPUs using different inner maps. So
> >> could you clarify how the synchronize_rcu mechanism will even work if you're
> >> now saying "program does not hold references to the inner maps"?
> 
> The program only held references to the outer maps, and the outer map
> held references to the inner maps. The user space program can add/remove
> the inner map for a particular outer map while the prog <-> outer-map
> relationship is not changed.

My definition of "reference" in this context is protection by rcu_read_lock.

So I was concerned the above map-in-map access isn't protected as such when
Alexei said "program doesn't have reference on inner map" in the above steps.
Maybe I misunderstood what is the meaning of reference here.

To make the map-in-map thing to work for Chenbo/Lorenzo's usecase, both the
access of outer map at key=0 and the inner map have to protect by
rcu_read_lock so that the membarrier call will work.

So basically step 0 in the steps above should be rcu_read_lock protected to
satisfy Chenbo/Lorenzo's usecase.

I know today the entire program is run as preempt disabled (unless something
changed) so this shouldn't be a problem, but in the future if the verifier is
doing similar things at a finer grainer level, then the above has to be
taken into consideration.

Does that make sense or am I missing something?

thanks,

- Joel

Re: [PATCH v2] Add BPF_SYNCHRONIZE_MAPS bpf(2) command

[Daniel Borkmann]

On 07/31/2018 11:56 PM, Joel Fernandes wrote:
> On Mon, Jul 30, 2018 at 09:03:18PM -0700, Y Song wrote:
>> On Mon, Jul 30, 2018 at 7:06 PM, Joel Fernandes <joel@joelfernandes.org> wrote:
>>> On Mon, Jul 30, 2018 at 07:01:22PM -0700, Joel Fernandes wrote:
>>>> On Sun, Jul 29, 2018 at 06:51:18PM +0300, Alexei Starovoitov wrote:
>>>>> On Thu, Jul 26, 2018 at 7:51 PM, Daniel Colascione <dancol@google.com> wrote:
>>>>>> BPF_SYNCHRONIZE_MAPS waits for the release of any references to a BPF
>>>>>> map made by a BPF program that is running at the time the
>>>>>> BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command is
>>>>>> to provide a means for userspace to replace a BPF map with another,
>>>>>> newer version, then ensure that no component is still using the "old"
>>>>>> map before manipulating the "old" map in some way.
>>>>>>
>>>>>> Signed-off-by: Daniel Colascione <dancol@google.com>
>>>>>> ---
>>>>>>  include/uapi/linux/bpf.h |  9 +++++++++
>>>>>>  kernel/bpf/syscall.c     | 13 +++++++++++++
>>>>>>  2 files changed, 22 insertions(+)
>>>>>>
>>>>>> diff --git a/include/uapi/linux/bpf.h b/include/uapi/linux/bpf.h
>>>>>> index b7db3261c62d..5b27e9117d3e 100644
>>>>>> --- a/include/uapi/linux/bpf.h
>>>>>> +++ b/include/uapi/linux/bpf.h
>>>>>> @@ -75,6 +75,14 @@ struct bpf_lpm_trie_key {
>>>>>>         __u8    data[0];        /* Arbitrary size */
>>>>>>  };
>>>>>>
>>>>>> +/* BPF_SYNCHRONIZE_MAPS waits for the release of any references to a
>>>>>> + * BPF map made by a BPF program that is running at the time the
>>>>>> + * BPF_SYNCHRONIZE_MAPS command is issued. The purpose of this command
>>>>>
>>>>> that doesn't sound right to me.
>>>>> such command won't wait for the release of the references.
>>>>> in case of map-in-map the program does not hold
>>>>> the references to inner map (only to outer map).
>>>>
>>>> I didn't follow this completely.
>>>>
>>>> The userspace program is using the inner map per your description of the
>>>> algorithm for using map-in-map to solve the race conditions that this patch
>>>> is trying to address:
>>>>
>>>> If you don't mind, I copy-pasted it below from your netdev post:
>>>>
>>>> if you use map-in-map you don't need extra boolean map.
>>>> 0. bpf prog can do
>>>>    inner_map = lookup(map_in_map, key=0);
>>>>    lookup(inner_map, your_real_key);
>>>> 1. user space writes into map_in_map[0] <- FD of new map
>>>> 2. some cpus are using old inner map and some a new
>>>> 3. user space does sys_membarrier(CMD_GLOBAL) which will do synchronize_sched()
>>>>    which in CONFIG_PREEMPT_NONE=y servers is the same as synchronize_rcu()
>>>>    which will guarantee that progs finished.
>>>> 4. scan old inner map
>>>>
>>>> In step 2, as you mentioned there are CPUs using different inner maps. So
>>>> could you clarify how the synchronize_rcu mechanism will even work if you're
>>>> now saying "program does not hold references to the inner maps"?
>>
>> The program only held references to the outer maps, and the outer map
>> held references to the inner maps. The user space program can add/remove
>> the inner map for a particular outer map while the prog <-> outer-map
>> relationship is not changed.
> 
> My definition of "reference" in this context is protection by rcu_read_lock.
> 
> So I was concerned the above map-in-map access isn't protected as such when
> Alexei said "program doesn't have reference on inner map" in the above steps.
> Maybe I misunderstood what is the meaning of reference here.
> 
> To make the map-in-map thing to work for Chenbo/Lorenzo's usecase, both the
> access of outer map at key=0 and the inner map have to protect by
> rcu_read_lock so that the membarrier call will work.
> 
> So basically step 0 in the steps above should be rcu_read_lock protected to
> satisfy Chenbo/Lorenzo's usecase.
> 
> I know today the entire program is run as preempt disabled (unless something
> changed) so this shouldn't be a problem, but in the future if the verifier is
> doing similar things at a finer grainer level, then the above has to be
> taken into consideration.
> 
> Does that make sense or am I missing something?

All BPF programs are required to run under rcu_read_lock today, so that
assumption holds. Should this ever change in future, then this constraint
of course needs to be taken into consideration.

Thanks,
Daniel