All of lore.kernel.org
 help / color / mirror / Atom feed
* [Adeos-main] [PATCH] Detect leaking stalls of topmost domain - v2
@ 2008-09-21 10:11 Jan Kiszka
  2008-09-28 15:27 ` Philippe Gerum
  0 siblings, 1 reply; 2+ messages in thread
From: Jan Kiszka @ 2008-09-21 10:11 UTC (permalink / raw)
  To: adeos-main

[-- Attachment #1: Type: text/plain, Size: 5559 bytes --]

Here comes an improved version of my debugging helper for leaking
top-most domain stalls. This version is now also NMI-safe, no longer
causing false positives from this context.

The instrumentation can detect infamous bug pattern like this:

function()
{
	stall_topmost_domain();
	...
	if (condition)
		return;
	...
	unstall_topmost_domain();
}

The result is often a locked-up system, specifically the root domain no
longer receives IRQs. Unless you find the bug quickly by code
inspection, debugging/instrumenting can take quite some time.

Also, this patch can help with these patterns:

	stall_topmost_domain();
	...
	blocking_or_rescheduling_in_lowprio_domain();
	...
	unstall_topmost_domain();

One example of this was just fixed in Xenomai's cleanup code.

To catch such issues earlier, I therefore propose the following
extension of ipipe_check_context. It is based on the assumption that the
topmost domain should never be stalled when lower domains execute that
check. This specifically takes care of not breaking Xenomai's IRQ shield
(a mid-prio domain that intentionally blocks Linux IRQs).


---
 include/linux/hardirq.h      |   19 +++++++++++++++++--
 include/linux/ipipe.h        |   20 ++++++++++++++++++++
 include/linux/ipipe_percpu.h |    1 +
 kernel/ipipe/core.c          |   21 ++++++++++++++++-----
 4 files changed, 54 insertions(+), 7 deletions(-)

Index: b/kernel/ipipe/core.c
===================================================================
--- a/kernel/ipipe/core.c
+++ b/kernel/ipipe/core.c
@@ -1561,13 +1561,16 @@ void __init ipipe_init_proc(void)
 #ifdef CONFIG_IPIPE_DEBUG_CONTEXT
 
 DEFINE_PER_CPU(int, ipipe_percpu_context_check) = { 1 };
+DEFINE_PER_CPU(int, ipipe_saved_context_check_state);
 
 void ipipe_check_context(struct ipipe_domain *border_ipd)
 {
 	/* Note: We don't make the per_cpu access atomic. We assume that code
 	   which temporarily disables the check does this in atomic context
 	   only. */
-	if (likely(ipipe_current_domain->priority <= border_ipd->priority) ||
+	if (likely(ipipe_current_domain->priority <= border_ipd->priority &&
+		   !test_bit(IPIPE_STALL_FLAG,
+			     &ipipe_head_cpudom_var(status))) ||
 	    !per_cpu(ipipe_percpu_context_check, ipipe_processor_id()))
 		return;
 
@@ -1575,10 +1578,18 @@ void ipipe_check_context(struct ipipe_do
 
 	ipipe_trace_panic_freeze();
 	ipipe_set_printk_sync(ipipe_current_domain);
-	printk(KERN_ERR "I-pipe: Detected illicit call from domain '%s'\n"
-	       KERN_ERR "        into a service reserved for domain '%s' and "
-			"below.\n",
-	       ipipe_current_domain->name, border_ipd->name);
+
+	if (ipipe_current_domain->priority > border_ipd->priority)
+		printk(KERN_ERR "I-pipe: Detected illicit call from domain "
+				"'%s'\n"
+		       KERN_ERR "        into a service reserved for domain "
+				"'%s' and below.\n",
+		       ipipe_current_domain->name, border_ipd->name);
+	else
+		printk(KERN_ERR "I-pipe: Detected stalled topmost domain, "
+				"probably caused by a bug.\n"
+				"        A critical section may have been "
+				"left unterminated.\n");
 	dump_stack();
 	ipipe_trace_panic_dump();
 }
Index: b/include/linux/hardirq.h
===================================================================
--- a/include/linux/hardirq.h
+++ b/include/linux/hardirq.h
@@ -161,7 +161,22 @@ extern void irq_enter(void);
  */
 extern void irq_exit(void);
 
-#define nmi_enter()		do { if (ipipe_root_domain_p) { lockdep_off(); __irq_enter(); } } while (0)
-#define nmi_exit()		do { if (ipipe_root_domain_p) { __irq_exit(); lockdep_on(); } } while (0)
+#define nmi_enter() 					\
+	do {						\
+		ipipe_nmi_enter();			\
+		if (ipipe_root_domain_p) {		\
+			lockdep_off();			\
+			__irq_enter();			\
+		}					\
+	} while (0)
+
+#define nmi_exit()					\
+	do {						\
+		if (ipipe_root_domain_p) {		\
+			__irq_exit();			\
+			lockdep_on();			\
+		}					\
+		ipipe_nmi_exit();			\
+	} while (0)
 
 #endif /* LINUX_HARDIRQ_H */
Index: b/include/linux/ipipe.h
===================================================================
--- a/include/linux/ipipe.h
+++ b/include/linux/ipipe.h
@@ -566,6 +566,22 @@ static inline void ipipe_context_check_o
 		per_cpu(ipipe_percpu_context_check, cpu) = 0;
 }
 
+static inline void ipipe_nmi_enter(void)
+{
+	int cpu = ipipe_processor_id();
+
+	per_cpu(ipipe_saved_context_check_state, cpu) =
+		ipipe_disable_context_check(cpu);
+}
+
+static inline void ipipe_nmi_exit(void)
+{
+	int cpu = ipipe_processor_id();
+
+	ipipe_restore_context_check
+		(cpu, per_cpu(ipipe_saved_context_check_state, cpu));
+}
+
 #else	/* !CONFIG_IPIPE_DEBUG_CONTEXT */
 
 static inline int ipipe_disable_context_check(int cpu)
@@ -577,6 +593,10 @@ static inline void ipipe_restore_context
 
 static inline void ipipe_context_check_off(void) { }
 
+static inline void ipipe_nmi_enter(void) { }
+
+static inline void ipipe_nmi_exit(void) { }
+
 #endif	/* !CONFIG_IPIPE_DEBUG_CONTEXT */
 
 #endif	/* !__LINUX_IPIPE_H */
Index: b/include/linux/ipipe_percpu.h
===================================================================
--- a/include/linux/ipipe_percpu.h
+++ b/include/linux/ipipe_percpu.h
@@ -58,6 +58,7 @@ DECLARE_PER_CPU(struct ipipe_domain *, i
 
 #ifdef CONFIG_IPIPE_DEBUG_CONTEXT
 DECLARE_PER_CPU(int, ipipe_percpu_context_check);
+DECLARE_PER_CPU(int, ipipe_saved_context_check_state);
 #endif
 
 #define ipipe_percpu(var, cpu)		per_cpu(var, cpu)


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 257 bytes --]

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [Adeos-main] [PATCH] Detect leaking stalls of topmost domain - v2
  2008-09-21 10:11 [Adeos-main] [PATCH] Detect leaking stalls of topmost domain - v2 Jan Kiszka
@ 2008-09-28 15:27 ` Philippe Gerum
  0 siblings, 0 replies; 2+ messages in thread
From: Philippe Gerum @ 2008-09-28 15:27 UTC (permalink / raw)
  To: Jan Kiszka; +Cc: adeos-main

Jan Kiszka wrote:
> Here comes an improved version of my debugging helper for leaking
> top-most domain stalls. This version is now also NMI-safe, no longer
> causing false positives from this context.
>

Applied, thanks.

> The instrumentation can detect infamous bug pattern like this:
> 
> function()
> {
> 	stall_topmost_domain();
> 	...
> 	if (condition)
> 		return;
> 	...
> 	unstall_topmost_domain();
> }
> 
> The result is often a locked-up system, specifically the root domain no
> longer receives IRQs. Unless you find the bug quickly by code
> inspection, debugging/instrumenting can take quite some time.
> 
> Also, this patch can help with these patterns:
> 
> 	stall_topmost_domain();
> 	...
> 	blocking_or_rescheduling_in_lowprio_domain();
> 	...
> 	unstall_topmost_domain();
> 
> One example of this was just fixed in Xenomai's cleanup code.
> 
> To catch such issues earlier, I therefore propose the following
> extension of ipipe_check_context. It is based on the assumption that the
> topmost domain should never be stalled when lower domains execute that
> check. This specifically takes care of not breaking Xenomai's IRQ shield
> (a mid-prio domain that intentionally blocks Linux IRQs).
> 
> 
> ---
>  include/linux/hardirq.h      |   19 +++++++++++++++++--
>  include/linux/ipipe.h        |   20 ++++++++++++++++++++
>  include/linux/ipipe_percpu.h |    1 +
>  kernel/ipipe/core.c          |   21 ++++++++++++++++-----
>  4 files changed, 54 insertions(+), 7 deletions(-)
> 
> Index: b/kernel/ipipe/core.c
> ===================================================================
> --- a/kernel/ipipe/core.c
> +++ b/kernel/ipipe/core.c
> @@ -1561,13 +1561,16 @@ void __init ipipe_init_proc(void)
>  #ifdef CONFIG_IPIPE_DEBUG_CONTEXT
>  
>  DEFINE_PER_CPU(int, ipipe_percpu_context_check) = { 1 };
> +DEFINE_PER_CPU(int, ipipe_saved_context_check_state);
>  
>  void ipipe_check_context(struct ipipe_domain *border_ipd)
>  {
>  	/* Note: We don't make the per_cpu access atomic. We assume that code
>  	   which temporarily disables the check does this in atomic context
>  	   only. */
> -	if (likely(ipipe_current_domain->priority <= border_ipd->priority) ||
> +	if (likely(ipipe_current_domain->priority <= border_ipd->priority &&
> +		   !test_bit(IPIPE_STALL_FLAG,
> +			     &ipipe_head_cpudom_var(status))) ||
>  	    !per_cpu(ipipe_percpu_context_check, ipipe_processor_id()))
>  		return;
>  
> @@ -1575,10 +1578,18 @@ void ipipe_check_context(struct ipipe_do
>  
>  	ipipe_trace_panic_freeze();
>  	ipipe_set_printk_sync(ipipe_current_domain);
> -	printk(KERN_ERR "I-pipe: Detected illicit call from domain '%s'\n"
> -	       KERN_ERR "        into a service reserved for domain '%s' and "
> -			"below.\n",
> -	       ipipe_current_domain->name, border_ipd->name);
> +
> +	if (ipipe_current_domain->priority > border_ipd->priority)
> +		printk(KERN_ERR "I-pipe: Detected illicit call from domain "
> +				"'%s'\n"
> +		       KERN_ERR "        into a service reserved for domain "
> +				"'%s' and below.\n",
> +		       ipipe_current_domain->name, border_ipd->name);
> +	else
> +		printk(KERN_ERR "I-pipe: Detected stalled topmost domain, "
> +				"probably caused by a bug.\n"
> +				"        A critical section may have been "
> +				"left unterminated.\n");
>  	dump_stack();
>  	ipipe_trace_panic_dump();
>  }
> Index: b/include/linux/hardirq.h
> ===================================================================
> --- a/include/linux/hardirq.h
> +++ b/include/linux/hardirq.h
> @@ -161,7 +161,22 @@ extern void irq_enter(void);
>   */
>  extern void irq_exit(void);
>  
> -#define nmi_enter()		do { if (ipipe_root_domain_p) { lockdep_off(); __irq_enter(); } } while (0)
> -#define nmi_exit()		do { if (ipipe_root_domain_p) { __irq_exit(); lockdep_on(); } } while (0)
> +#define nmi_enter() 					\
> +	do {						\
> +		ipipe_nmi_enter();			\
> +		if (ipipe_root_domain_p) {		\
> +			lockdep_off();			\
> +			__irq_enter();			\
> +		}					\
> +	} while (0)
> +
> +#define nmi_exit()					\
> +	do {						\
> +		if (ipipe_root_domain_p) {		\
> +			__irq_exit();			\
> +			lockdep_on();			\
> +		}					\
> +		ipipe_nmi_exit();			\
> +	} while (0)
>  
>  #endif /* LINUX_HARDIRQ_H */
> Index: b/include/linux/ipipe.h
> ===================================================================
> --- a/include/linux/ipipe.h
> +++ b/include/linux/ipipe.h
> @@ -566,6 +566,22 @@ static inline void ipipe_context_check_o
>  		per_cpu(ipipe_percpu_context_check, cpu) = 0;
>  }
>  
> +static inline void ipipe_nmi_enter(void)
> +{
> +	int cpu = ipipe_processor_id();
> +
> +	per_cpu(ipipe_saved_context_check_state, cpu) =
> +		ipipe_disable_context_check(cpu);
> +}
> +
> +static inline void ipipe_nmi_exit(void)
> +{
> +	int cpu = ipipe_processor_id();
> +
> +	ipipe_restore_context_check
> +		(cpu, per_cpu(ipipe_saved_context_check_state, cpu));
> +}
> +
>  #else	/* !CONFIG_IPIPE_DEBUG_CONTEXT */
>  
>  static inline int ipipe_disable_context_check(int cpu)
> @@ -577,6 +593,10 @@ static inline void ipipe_restore_context
>  
>  static inline void ipipe_context_check_off(void) { }
>  
> +static inline void ipipe_nmi_enter(void) { }
> +
> +static inline void ipipe_nmi_exit(void) { }
> +
>  #endif	/* !CONFIG_IPIPE_DEBUG_CONTEXT */
>  
>  #endif	/* !__LINUX_IPIPE_H */
> Index: b/include/linux/ipipe_percpu.h
> ===================================================================
> --- a/include/linux/ipipe_percpu.h
> +++ b/include/linux/ipipe_percpu.h
> @@ -58,6 +58,7 @@ DECLARE_PER_CPU(struct ipipe_domain *, i
>  
>  #ifdef CONFIG_IPIPE_DEBUG_CONTEXT
>  DECLARE_PER_CPU(int, ipipe_percpu_context_check);
> +DECLARE_PER_CPU(int, ipipe_saved_context_check_state);
>  #endif
>  
>  #define ipipe_percpu(var, cpu)		per_cpu(var, cpu)
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Adeos-main mailing list
> Adeos-main@domain.hid
> https://mail.gna.org/listinfo/adeos-main


-- 
Philippe.


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2008-09-28 15:27 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2008-09-21 10:11 [Adeos-main] [PATCH] Detect leaking stalls of topmost domain - v2 Jan Kiszka
2008-09-28 15:27 ` Philippe Gerum

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.