Re: [PATCH v2 3/7] powerpc: use task_pid_nr() for TID allocation

[Alastair D'Silva]

On Mon, 2018-05-07 at 19:37 +0200, Frederic Barrat wrote:
> 
> Le 18/04/2018 à 03:08, Alastair D'Silva a écrit :
> > From: Alastair D'Silva 
> > 
> > The current implementation of TID allocation, using a global IDR,
> > may
> > result in an errant process starving the system of available TIDs.
> > Instead, use task_pid_nr(), as mentioned by the original author.
> > The
> > scenario described which prevented it's use is not applicable, as
> > set_thread_tidr can only be called after the task struct has been
> > populated.
> 
> 
> Here is how I understand what's going to happen if 2 threads are
> using 
> the same TIDR value, which is possible with this patch (if unlikely):
> 
> 1. waking up the wrong thread is not really a problem, as threads
> have 
> to handle spurious wake up from the 'wait' instruction anyway, and
> must 
> be using some other condition to know when to loop around the 'wait' 
> instruction.
> 
> 2. missing the right thread: if the wrong thread is on a CPU, and a 
> wake_host_thread/as_notify is sent, the core will see a matching
> thread 
> and will accept the command. The (open)capi adapter won't send an 
> interrupt. The wrong thread is awaken, which is not a problem as 
> discussed above. As the right thread to notify is not running, no
> harm 
> is done either: as soon as the thread runs, it's supposed to check
> its 
> condition (which will be met) or call 'wait', but 'wait' immediately 
> returns when called the first time after a thread is scheduled.
> 
> So I believe we are ok. But I think it requires a huge comment with
> the 
> above (at the minimum) :-)
> 
> With a comment:
> Reviewed-by: Frederic Barrat 
> 
>Fred
> 

Good point, I'll add this in the next revision.


-- 
Alastair D'Silva
Open Source Developer
Linux Technology Centre, IBM Australiamob: 0423 762 819

--
To unsubscribe from this list: send the line "unsubscribe linux-doc" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH v2 3/7] powerpc: use task_pid_nr() for TID allocation

[Frederic Barrat]

Le 18/04/2018 à 03:08, Alastair D'Silva a écrit :

From: Alastair D'Silva 

The current implementation of TID allocation, using a global IDR, may
result in an errant process starving the system of available TIDs.
Instead, use task_pid_nr(), as mentioned by the original author. The
scenario described which prevented it's use is not applicable, as
set_thread_tidr can only be called after the task struct has been
populated.



Here is how I understand what's going to happen if 2 threads are using 
the same TIDR value, which is possible with this patch (if unlikely):


1. waking up the wrong thread is not really a problem, as threads have 
to handle spurious wake up from the 'wait' instruction anyway, and must 
be using some other condition to know when to loop around the 'wait' 
instruction.


2. missing the right thread: if the wrong thread is on a CPU, and a 
wake_host_thread/as_notify is sent, the core will see a matching thread 
and will accept the command. The (open)capi adapter won't send an 
interrupt. The wrong thread is awaken, which is not a problem as 
discussed above. As the right thread to notify is not running, no harm 
is done either: as soon as the thread runs, it's supposed to check its 
condition (which will be met) or call 'wait', but 'wait' immediately 
returns when called the first time after a thread is scheduled.


So I believe we are ok. But I think it requires a huge comment with the 
above (at the minimum) :-)


With a comment:
Reviewed-by: Frederic Barrat 

  Fred




Signed-off-by: Alastair D'Silva 
---
  arch/powerpc/include/asm/switch_to.h |  1 -
  arch/powerpc/kernel/process.c| 97 +---
  2 files changed, 1 insertion(+), 97 deletions(-)

diff --git a/arch/powerpc/include/asm/switch_to.h 
b/arch/powerpc/include/asm/switch_to.h
index be8c9fa23983..5b03d8a82409 100644
--- a/arch/powerpc/include/asm/switch_to.h
+++ b/arch/powerpc/include/asm/switch_to.h
@@ -94,6 +94,5 @@ static inline void clear_task_ebb(struct task_struct *t)
  extern int set_thread_uses_vas(void);

  extern int set_thread_tidr(struct task_struct *t);
-extern void clear_thread_tidr(struct task_struct *t);

  #endif /* _ASM_POWERPC_SWITCH_TO_H */
diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c
index 3b00da47699b..87f047fd2762 100644
--- a/arch/powerpc/kernel/process.c
+++ b/arch/powerpc/kernel/process.c
@@ -1496,103 +1496,12 @@ int set_thread_uses_vas(void)
  }

  #ifdef CONFIG_PPC64
-static DEFINE_SPINLOCK(vas_thread_id_lock);
-static DEFINE_IDA(vas_thread_ida);
-
-/*
- * We need to assign a unique thread id to each thread in a process.
- *
- * This thread id, referred to as TIDR, and separate from the Linux's tgid,
- * is intended to be used to direct an ASB_Notify from the hardware to the
- * thread, when a suitable event occurs in the system.
- *
- * One such event is a "paste" instruction in the context of Fast Thread
- * Wakeup (aka Core-to-core wake up in the Virtual Accelerator Switchboard
- * (VAS) in POWER9.
- *
- * To get a unique TIDR per process we could simply reuse task_pid_nr() but
- * the problem is that task_pid_nr() is not yet available copy_thread() is
- * called. Fixing that would require changing more intrusive arch-neutral
- * code in code path in copy_process()?.
- *
- * Further, to assign unique TIDRs within each process, we need an atomic
- * field (or an IDR) in task_struct, which again intrudes into the arch-
- * neutral code. So try to assign globally unique TIDRs for now.
- *
- * NOTE: TIDR 0 indicates that the thread does not need a TIDR value.
- *  For now, only threads that expect to be notified by the VAS
- *  hardware need a TIDR value and we assign values > 0 for those.
- */
-#define MAX_THREAD_CONTEXT ((1 << 16) - 1)
-static int assign_thread_tidr(void)
-{
-   int index;
-   int err;
-   unsigned long flags;
-
-again:
-   if (!ida_pre_get(_thread_ida, GFP_KERNEL))
-   return -ENOMEM;
-
-   spin_lock_irqsave(_thread_id_lock, flags);
-   err = ida_get_new_above(_thread_ida, 1, );
-   spin_unlock_irqrestore(_thread_id_lock, flags);
-
-   if (err == -EAGAIN)
-   goto again;
-   else if (err)
-   return err;
-
-   if (index > MAX_THREAD_CONTEXT) {
-   spin_lock_irqsave(_thread_id_lock, flags);
-   ida_remove(_thread_ida, index);
-   spin_unlock_irqrestore(_thread_id_lock, flags);
-   return -ENOMEM;
-   }
-
-   return index;
-}
-
-static void free_thread_tidr(int id)
-{
-   unsigned long flags;
-
-   spin_lock_irqsave(_thread_id_lock, flags);
-   ida_remove(_thread_ida, id);
-   spin_unlock_irqrestore(_thread_id_lock, flags);
-}
-
-/*
- * Clear any TIDR value assigned to this thread.
- */
-void clear_thread_tidr(struct task_struct *t)
-{
-   if (!t->thread.tidr)
-   

Re: [PATCH v2 3/7] powerpc: use task_pid_nr() for TID allocation

[Andrew Donnellan]

On 25/04/18 07:12, Sukadev Bhattiprolu wrote:

Yes. Like with PIDR, was trying to assign TIDR initially to all threads.
But since only a subset of threads need/use TIDR, we can assign the
value later (when set_thread_tidr() is called). So we should be able to
use task_pid_nr() then.


OK. Alastair has also confirmed with me that truncating the pid to a u16 
should be safe, so therefore:


Reviewed-by: Andrew Donnellan 


--
Andrew Donnellan  OzLabs, ADL Canberra
andrew.donnel...@au1.ibm.com  IBM Australia Limited

--
To unsubscribe from this list: send the line "unsubscribe linux-doc" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH v2 3/7] powerpc: use task_pid_nr() for TID allocation

[Sukadev Bhattiprolu]

Andrew Donnellan [andrew.donnel...@au1.ibm.com] wrote:
> [+ Sukadev, Christophe]
> 
> On 18/04/18 11:08, Alastair D'Silva wrote:
> > From: Alastair D'Silva 
> > 
> > The current implementation of TID allocation, using a global IDR, may
> > result in an errant process starving the system of available TIDs.
> > Instead, use task_pid_nr(), as mentioned by the original author. The
> > scenario described which prevented it's use is not applicable, as
> > set_thread_tidr can only be called after the task struct has been
> > populated.
> > 
> > Signed-off-by: Alastair D'Silva 
> 
> So it's too late in the evening for me to completely get my head around
> what's going on here enough to give my Reviewed-by:, but my current thinking
> is:
> 
> - In the first version of the patch to add TIDR support
> (https://patchwork.ozlabs.org/patch/799494/), it was originally proposed to
> call assign_thread_id() (as it was then called) from copy_thread()
> 
> - The comment block documents the reason why we can't use task_pid_nr() but
> assumes that we're trying to assign a TIDR from within copy_thread()
> 
> - The final patch that was accepted
> (https://patchwork.ozlabs.org/patch/835552/,
> ec233ede4c8654894610ea54f4dae7adc954ac62) instead sets the TIDR to 0 from
> copy_thread(), so the original reasoning regarding not using task_pid_nr()
> within copy_thread() is no longer applicable.
> 
> Sukadev: does this sound right?

Yes. Like with PIDR, was trying to assign TIDR initially to all threads.
But since only a subset of threads need/use TIDR, we can assign the
value later (when set_thread_tidr() is called). So we should be able to
use task_pid_nr() then.

Sukadev

--
To unsubscribe from this list: send the line "unsubscribe linux-doc" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH v2 3/7] powerpc: use task_pid_nr() for TID allocation

[Alastair D'Silva]

From: Alastair D'Silva 

The current implementation of TID allocation, using a global IDR, may
result in an errant process starving the system of available TIDs.
Instead, use task_pid_nr(), as mentioned by the original author. The
scenario described which prevented it's use is not applicable, as
set_thread_tidr can only be called after the task struct has been
populated.

Signed-off-by: Alastair D'Silva 
---
 arch/powerpc/include/asm/switch_to.h |  1 -
 arch/powerpc/kernel/process.c| 97 +---
 2 files changed, 1 insertion(+), 97 deletions(-)

diff --git a/arch/powerpc/include/asm/switch_to.h 
b/arch/powerpc/include/asm/switch_to.h
index be8c9fa23983..5b03d8a82409 100644
--- a/arch/powerpc/include/asm/switch_to.h
+++ b/arch/powerpc/include/asm/switch_to.h
@@ -94,6 +94,5 @@ static inline void clear_task_ebb(struct task_struct *t)
 extern int set_thread_uses_vas(void);
 
 extern int set_thread_tidr(struct task_struct *t);
-extern void clear_thread_tidr(struct task_struct *t);
 
 #endif /* _ASM_POWERPC_SWITCH_TO_H */
diff --git a/arch/powerpc/kernel/process.c b/arch/powerpc/kernel/process.c
index 3b00da47699b..87f047fd2762 100644
--- a/arch/powerpc/kernel/process.c
+++ b/arch/powerpc/kernel/process.c
@@ -1496,103 +1496,12 @@ int set_thread_uses_vas(void)
 }
 
 #ifdef CONFIG_PPC64
-static DEFINE_SPINLOCK(vas_thread_id_lock);
-static DEFINE_IDA(vas_thread_ida);
-
-/*
- * We need to assign a unique thread id to each thread in a process.
- *
- * This thread id, referred to as TIDR, and separate from the Linux's tgid,
- * is intended to be used to direct an ASB_Notify from the hardware to the
- * thread, when a suitable event occurs in the system.
- *
- * One such event is a "paste" instruction in the context of Fast Thread
- * Wakeup (aka Core-to-core wake up in the Virtual Accelerator Switchboard
- * (VAS) in POWER9.
- *
- * To get a unique TIDR per process we could simply reuse task_pid_nr() but
- * the problem is that task_pid_nr() is not yet available copy_thread() is
- * called. Fixing that would require changing more intrusive arch-neutral
- * code in code path in copy_process()?.
- *
- * Further, to assign unique TIDRs within each process, we need an atomic
- * field (or an IDR) in task_struct, which again intrudes into the arch-
- * neutral code. So try to assign globally unique TIDRs for now.
- *
- * NOTE: TIDR 0 indicates that the thread does not need a TIDR value.
- *  For now, only threads that expect to be notified by the VAS
- *  hardware need a TIDR value and we assign values > 0 for those.
- */
-#define MAX_THREAD_CONTEXT ((1 << 16) - 1)
-static int assign_thread_tidr(void)
-{
-   int index;
-   int err;
-   unsigned long flags;
-
-again:
-   if (!ida_pre_get(_thread_ida, GFP_KERNEL))
-   return -ENOMEM;
-
-   spin_lock_irqsave(_thread_id_lock, flags);
-   err = ida_get_new_above(_thread_ida, 1, );
-   spin_unlock_irqrestore(_thread_id_lock, flags);
-
-   if (err == -EAGAIN)
-   goto again;
-   else if (err)
-   return err;
-
-   if (index > MAX_THREAD_CONTEXT) {
-   spin_lock_irqsave(_thread_id_lock, flags);
-   ida_remove(_thread_ida, index);
-   spin_unlock_irqrestore(_thread_id_lock, flags);
-   return -ENOMEM;
-   }
-
-   return index;
-}
-
-static void free_thread_tidr(int id)
-{
-   unsigned long flags;
-
-   spin_lock_irqsave(_thread_id_lock, flags);
-   ida_remove(_thread_ida, id);
-   spin_unlock_irqrestore(_thread_id_lock, flags);
-}
-
-/*
- * Clear any TIDR value assigned to this thread.
- */
-void clear_thread_tidr(struct task_struct *t)
-{
-   if (!t->thread.tidr)
-   return;
-
-   if (!cpu_has_feature(CPU_FTR_P9_TIDR)) {
-   WARN_ON_ONCE(1);
-   return;
-   }
-
-   mtspr(SPRN_TIDR, 0);
-   free_thread_tidr(t->thread.tidr);
-   t->thread.tidr = 0;
-}
-
-void arch_release_task_struct(struct task_struct *t)
-{
-   clear_thread_tidr(t);
-}
-
 /*
  * Assign a unique TIDR (thread id) for task @t and set it in the thread
  * structure. For now, we only support setting TIDR for 'current' task.
  */
 int set_thread_tidr(struct task_struct *t)
 {
-   int rc;
-
if (!cpu_has_feature(CPU_FTR_P9_TIDR))
return -EINVAL;
 
@@ -1602,11 +1511,7 @@ int set_thread_tidr(struct task_struct *t)
if (t->thread.tidr)
return 0;
 
-   rc = assign_thread_tidr();
-   if (rc < 0)
-   return rc;
-
-   t->thread.tidr = rc;
+   t->thread.tidr = (u16)task_pid_nr(t);
mtspr(SPRN_TIDR, t->thread.tidr);
 
return 0;
-- 
2.14.3

--
To unsubscribe from this list: send the line "unsubscribe linux-doc" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html