Tiny RCU updates for 4.2
Subsystems (4)
cover Tiny RCU updates for 4.2
Hello!
This series contains updates to Tiny RCU, including a low-probability but quite real race condition.
1. Further shrink Tiny RCU by making empty functions static inlines.
2. Test both RCU-sched and RCU-bh for Tiny RCU, which located the
race condition fixed by #3 below.
3. Correctly handle non-empty Tiny RCU callback list with none ready.
Thanx, Paul
------------------------------------------------------------------------
b/include/linux/rcupdate.h | 4 - b/include/linux/rcutiny.h | 16 ++++ b/include/linux/rcutree.h | 5 + b/kernel/rcu/tiny.c | 38 +---------- b/tools/testing/selftests/rcutorture/configs/rcu/TINY02.boot | 1 5 files changed, 27 insertions(+), 37 deletions(-)
PATCH 3/3 rcu: Correctly handle non-empty Tiny RCU callback list with none ready
In reply to Paul E. McKenneyFrom: "Paul E. McKenney" <[email protected]>
If, at the time __rcu_process_callbacks() is invoked, there are callbacks in Tiny RCU's callback list, but none of them are ready to be invoked, the current list-management code will knit the non-ready callbacks out of the list. This can result in hangs and possibly worse. This commit therefore inserts a check for there being no callbacks that can be invoked immediately.
This bug is unlikely to occur -- you have to get a new callback between the time rcu_sched_qs() or rcu_bh_qs() was called, but before we get to __rcu_process_callbacks(). It was detected by the addition of RCU-bh testing to rcutorture, which in turn was instigated by Iftekhar Ahmed's mutation testing. Although this bug was made much more likely by 915e8a4fe45e (rcu: Remove fastpath from __rcu_process_callbacks()), this
did not cause the bug, but rather made it much more probable. That
said, it takes more than 40 hours of rcutorture testing, on average, for this bug to appear, so this fix cannot be considered an emergency.
--- kernel/rcu/tiny.c | 5 +++++ 1 file changed, 5 insertions(+)
kernel/rcu/tiny.c +5 -0
@@ -137,6 +137,11 @@ static void __rcu_process_callbacks(struct rcu_ctrlblk *rcp)/* Move the ready-to-invoke callbacks to a local list. */ local_irq_save (flags); +if (rcp->donetail == &rcp->rcucblist) { +/* No callbacks ready, so just leave. */ +local_irq_restore (flags); +return ; + }RCU_TRACE (trace_rcu_batch_start (rcp->name,0 , rcp->qlen,-1 )); list = rcp->rcucblist; rcp->rcucblist = *rcp->donetail;
Signature · 1 line
1.8.1.5
PATCH 2/3 rcutorture: Test both RCU-sched and RCU-bh for Tiny RCU
In reply to Paul E. McKenneyFrom: "Paul E. McKenney" <[email protected]>
--- tools/testing/selftests/rcutorture/configs/rcu/TINY02.boot | 1 + 1 file changed, 1 insertion(+)
tools/testing/selftests/rcutorture/configs/rcu/TINY02.boot +1 -0
@@ -1,2 +1,3 @@ rcupdate.rcu_self_test=1 rcupdate.rcu_self_test_bh=1 +rcutorture.torture_type=rcu_bh
Signature · 1 line
1.8.1.5
PATCH 1/3 rcu: Further shrink Tiny RCU by making empty functions static inlines
In reply to Paul E. McKenneyFrom: "Paul E. McKenney" <[email protected]>
The Tiny RCU counterparts to rcu_idle_enter(), rcu_idle_exit(), rcu_irq_enter(), and rcu_irq_exit() are empty functions, but each has EXPORT_SYMBOL_GPL(), which, in kernels built with module support, needlessly consumes some memory. This commit therefore moves these functions to static inlines in rcutiny.h, removing the need for exports.
This won't affect the size of the tiniest kernels, which are likely built without module support, but might help semi-tiny kernels that might include module support.
--- include/linux/rcupdate.h | 4 ---- include/linux/rcutiny.h | 16 ++++++++++++++++ include/linux/rcutree.h | 5 +++++ kernel/rcu/tiny.c | 33 --------------------------------- 4 files changed, 21 insertions(+), 37 deletions(-)
include/linux/rcupdate.h +0 -4
@@ -292,10 +292,6 @@ void rcu_sched_qs(void);void rcu_bh_qs (void );void rcu_check_callbacks (int user);struct notifier_block ; -void rcu_idle_enter (void ); -void rcu_idle_exit (void ); -void rcu_irq_enter (void ); -void rcu_irq_exit (void );int rcu_cpu_notify (struct notifier_block *self,unsigned long action,void *hcpu);
include/linux/rcutiny.h +16 -0
@@ -159,6 +159,22 @@ static inline void rcu_cpu_stall_reset(void) { } +static inline void rcu_idle_enter (void ) +{ +} + +static inline void rcu_idle_exit (void ) +{ +} + +static inline void rcu_irq_enter (void ) +{ +} + +static inline void rcu_irq_exit (void ) +{ +} +static inline void exit_rcu (void ) { }
include/linux/rcutree.h +5 -0
@@ -93,6 +93,11 @@ void rcu_force_quiescent_state(void);void rcu_bh_force_quiescent_state (void );void rcu_sched_force_quiescent_state (void ); +void rcu_idle_enter (void ); +void rcu_idle_exit (void ); +void rcu_irq_enter (void ); +void rcu_irq_exit (void ); +void exit_rcu (void );void rcu_scheduler_starting (void );
kernel/rcu/tiny.c +0 -33
@@ -49,39 +49,6 @@ static void __call_rcu(struct rcu_head *head,#include "tiny_plugin.h" -/* -* Enter idle, which is an extended quiescent state if we have fully -* entered that mode. -*/ -void rcu_idle_enter (void ) -{ -} -EXPORT_SYMBOL_GPL (rcu_idle_enter); - -/* -* Exit an interrupt handler towards idle. -*/ -void rcu_irq_exit (void ) -{ -} -EXPORT_SYMBOL_GPL (rcu_irq_exit); - -/* -* Exit idle, so that we are no longer in an extended quiescent state. -*/ -void rcu_idle_exit (void ) -{ -} -EXPORT_SYMBOL_GPL (rcu_idle_exit); - -/* -* Enter an interrupt handler, moving away from idle. -*/ -void rcu_irq_enter (void ) -{ -} -EXPORT_SYMBOL_GPL (rcu_irq_enter); -#if defined(CONFIG_DEBUG_LOCK_ALLOC ) || defined(CONFIG_RCU_TRACE )/*
Signature · 1 line
1.8.1.5
Re: [PATCH tip/core/rcu 1/3] rcu: Further shrink Tiny RCU by making empty functions static inlines
Quoted 9 lines · Paul E. McKenney
From: "Paul E. McKenney" <[email protected]>
The Tiny RCU counterparts to
rcu_idle_enter(),rcu_idle_exit(),rcu_irq_enter(), andrcu_irq_exit()are empty functions, but each hasEXPORT_SYMBOL_GPL(), which, in kernels built with module support, needlessly consumes some memory. This commit therefore moves these functions to static inlines inrcutiny.h, removing the need for exports.
Quoted 5 lines
This won't affect the size of the tiniest kernels, which are likely built without module support, but might help semi-tiny kernels that might include module support.
Signed-off-by Paul E. McKenney <[email protected]>
Even on kernels that don't include module support, this would shave a few bytes. Callers can't know a function in another .c file is actually empty (without LTO, anyway), so callers have to actually *call* these empty functions, and the functions themselves would need to be emitted. Turning them into static inlines tells the callers that they can avoid emitting any code.
Quoted 111 lines
include/linux/rcupdate.h | 4 ---- include/linux/rcutiny.h | 16 ++++++++++++++++ include/linux/rcutree.h | 5 +++++ kernel/rcu/tiny.c | 33 --------------------------------- 4 files changed, 21 insertions(+), 37 deletions(-)include/linux/rcupdate.h +0
-4@@ -292,10 +292,6 @@ void rcu_sched_qs(void);void rcu_bh_qs (void );void rcu_check_callbacks (int user);struct notifier_block ; -void rcu_idle_enter (void ); -void rcu_idle_exit (void ); -void rcu_irq_enter (void ); -void rcu_irq_exit (void );int rcu_cpu_notify (struct notifier_block *self,unsigned long action,void *hcpu);include/linux/rcutiny.h +16
-0@@ -159,6 +159,22 @@ static inline void rcu_cpu_stall_reset(void) { } +static inline void rcu_idle_enter (void ) +{ +} + +static inline void rcu_idle_exit (void ) +{ +} + +static inline void rcu_irq_enter (void ) +{ +} + +static inline void rcu_irq_exit (void ) +{ +} +static inline void exit_rcu (void ) { }include/linux/rcutree.h +5
-0@@ -93,6 +93,11 @@ void rcu_force_quiescent_state(void);void rcu_bh_force_quiescent_state (void );void rcu_sched_force_quiescent_state (void ); +void rcu_idle_enter (void ); +void rcu_idle_exit (void ); +void rcu_irq_enter (void ); +void rcu_irq_exit (void ); +void exit_rcu (void );void rcu_scheduler_starting (void );kernel/rcu/tiny.c +0
-33@@ -49,39 +49,6 @@ static void __call_rcu(struct rcu_head *head,#include "tiny_plugin.h" -/* -* Enter idle, which is an extended quiescent state if we have fully -* entered that mode. -*/ -void rcu_idle_enter (void ) -{ -} -EXPORT_SYMBOL_GPL (rcu_idle_enter); - -/* -* Exit an interrupt handler towards idle. -*/ -void rcu_irq_exit (void ) -{ -} -EXPORT_SYMBOL_GPL (rcu_irq_exit); - -/* -* Exit idle, so that we are no longer in an extended quiescent state. -*/ -void rcu_idle_exit (void ) -{ -} -EXPORT_SYMBOL_GPL (rcu_idle_exit); - -/* -* Enter an interrupt handler, moving away from idle. -*/ -void rcu_irq_enter (void ) -{ -} -EXPORT_SYMBOL_GPL (rcu_irq_enter); -#if defined(CONFIG_DEBUG_LOCK_ALLOC ) || defined(CONFIG_RCU_TRACE )/* Signature · 1 line
1.8.1.5
Re: [PATCH tip/core/rcu 3/3] rcu: Correctly handle non-empty Tiny RCU callback list with none ready
In reply to Paul E. McKenneyQuoted 19 lines · Paul E. McKenney
From: "Paul E. McKenney" <[email protected]>
If, at the time
__rcu_process_callbacks()is invoked, there are callbacks in Tiny RCU's callback list, but none of them are ready to be invoked, the current list-management code will knit the non-ready callbacks out of the list. This can result in hangs and possibly worse. This commit therefore inserts a check for there being no callbacks that can be invoked immediately.This bug is unlikely to occur -- you have to get a new callback between the time
rcu_sched_qs()orrcu_bh_qs()was called, but before we get to__rcu_process_callbacks(). It was detected by the addition of RCU-bh testing to rcutorture, which in turn was instigated by Iftekhar Ahmed's mutation testing. Although this bug was made much more likely by915e8a4fe45e(rcu: Remove fastpath from__rcu_process_callbacks()), this
did not cause the bug, but rather made it much more probable. That
said, it takes more than 40 hours of rcutorture testing, on average, for this bug to appear, so this fix cannot be considered an emergency.
Quoted 2 lines
Signed-off-by Paul E. McKenney <[email protected]>Cc <[email protected]>
Ouch, subtle.
Quoted 22 lines
kernel/rcu/tiny.c | 5 +++++ 1 file changed, 5 insertions(+)kernel/rcu/tiny.c +5
-0@@ -137,6 +137,11 @@ static void __rcu_process_callbacks(struct rcu_ctrlblk *rcp)/* Move the ready-to-invoke callbacks to a local list. */ local_irq_save (flags); +if (rcp->donetail == &rcp->rcucblist) { +/* No callbacks ready, so just leave. */ +local_irq_restore (flags); +return ; + }RCU_TRACE (trace_rcu_batch_start (rcp->name,0 , rcp->qlen,-1 )); list = rcp->rcucblist; rcp->rcucblist = *rcp->donetail;Signature · 1 line
1.8.1.5
[email protected]
maintainerRe: [PATCH tip/core/rcu 2/3] rcutorture: Test both RCU-sched and RCU-bh for Tiny RCU
In reply to Paul E. McKenneyQuoted 4 lines · Paul E. McKenney
From: "Paul E. McKenney" <[email protected]>
Reported-by "Ahmed, Iftekhar" <[email protected]>Signed-off-by Paul E. McKenney <[email protected]>
Could you elaborate a bit more on this patch (ideally in its commit message)? I see an addition of a command-line parameter to test rcu_bh; is rcu-sched already tested elsewhere by some other config, or does this parameter somehow enable testing both?
Quoted 14 lines
tools/testing/selftests/rcutorture/configs/rcu/TINY02.boot | 1 + 1 file changed, 1 insertion(+)tools/testing/selftests/rcutorture/configs/rcu/TINY02.boot +1
-0@@ -1,2 +1,3 @@ rcupdate.rcu_self_test=1 rcupdate.rcu_self_test_bh=1 +rcutorture.torture_type=rcu_bhSignature · 1 line
1.8.1.5
Paul E. McKenney
Reviewed-by Josh TriplettRe: [PATCH tip/core/rcu 1/3] rcu: Further shrink Tiny RCU by making empty functions static inlines
In reply to [email protected]Quoted 16 lines · On Tue, May 12, 2015 at 05:57:18PM -0700, [email protected] wrote:
Quoted 9 lines · Paul E. McKenney
From: "Paul E. McKenney" <[email protected]>
The Tiny RCU counterparts to
rcu_idle_enter(),rcu_idle_exit(),rcu_irq_enter(), andrcu_irq_exit()are empty functions, but each hasEXPORT_SYMBOL_GPL(), which, in kernels built with module support, needlessly consumes some memory. This commit therefore moves these functions to static inlines inrcutiny.h, removing the need for exports.Quoted 5 lines
This won't affect the size of the tiniest kernels, which are likely built without module support, but might help semi-tiny kernels that might include module support.
Signed-off-by Paul E. McKenney <[email protected]>
Quoted 8 lines
Even on kernels that don't include module support, this would shave a few bytes. Callers can't know a function in another .c file is actually empty (without LTO, anyway), so callers have to actually *call* these empty functions, and the functions themselves would need to be emitted. Turning them into static inlines tells the callers that they can avoid emitting any code.
Reviewed-by Josh Triplett <[email protected]>
Thank you, and good point! The commit log now reads as follows:
rcu: Further shrink Tiny RCU by making empty functions static inlines
The Tiny RCU counterparts to rcu_idle_enter(), rcu_idle_exit(), rcu_irq_enter(), and rcu_irq_exit() are empty functions, but each has EXPORT_SYMBOL_GPL(), which needlessly consumes extra memory, especially in kernels built with module support. This commit therefore moves these functions to static inlines in rcutiny.h, removing the need for exports.
This won't affect the size of the tiniest kernels, which are likely built without module support, but might help semi-tiny kernels that might include module support.
Is that better?
Thanx, Paul
Quoted 112 lines
Quoted 111 lines
include/linux/rcupdate.h | 4 ---- include/linux/rcutiny.h | 16 ++++++++++++++++ include/linux/rcutree.h | 5 +++++ kernel/rcu/tiny.c | 33 --------------------------------- 4 files changed, 21 insertions(+), 37 deletions(-)include/linux/rcupdate.h +0
-4@@ -292,10 +292,6 @@ void rcu_sched_qs(void);void rcu_bh_qs (void );void rcu_check_callbacks (int user);struct notifier_block ; -void rcu_idle_enter (void ); -void rcu_idle_exit (void ); -void rcu_irq_enter (void ); -void rcu_irq_exit (void );int rcu_cpu_notify (struct notifier_block *self,unsigned long action,void *hcpu);include/linux/rcutiny.h +16
-0@@ -159,6 +159,22 @@ static inline void rcu_cpu_stall_reset(void) { } +static inline void rcu_idle_enter (void ) +{ +} + +static inline void rcu_idle_exit (void ) +{ +} + +static inline void rcu_irq_enter (void ) +{ +} + +static inline void rcu_irq_exit (void ) +{ +} +static inline void exit_rcu (void ) { }include/linux/rcutree.h +5
-0@@ -93,6 +93,11 @@ void rcu_force_quiescent_state(void);void rcu_bh_force_quiescent_state (void );void rcu_sched_force_quiescent_state (void ); +void rcu_idle_enter (void ); +void rcu_idle_exit (void ); +void rcu_irq_enter (void ); +void rcu_irq_exit (void ); +void exit_rcu (void );void rcu_scheduler_starting (void );kernel/rcu/tiny.c +0
-33@@ -49,39 +49,6 @@ static void __call_rcu(struct rcu_head *head,#include "tiny_plugin.h" -/* -* Enter idle, which is an extended quiescent state if we have fully -* entered that mode. -*/ -void rcu_idle_enter (void ) -{ -} -EXPORT_SYMBOL_GPL (rcu_idle_enter); - -/* -* Exit an interrupt handler towards idle. -*/ -void rcu_irq_exit (void ) -{ -} -EXPORT_SYMBOL_GPL (rcu_irq_exit); - -/* -* Exit idle, so that we are no longer in an extended quiescent state. -*/ -void rcu_idle_exit (void ) -{ -} -EXPORT_SYMBOL_GPL (rcu_idle_exit); - -/* -* Enter an interrupt handler, moving away from idle. -*/ -void rcu_irq_enter (void ) -{ -} -EXPORT_SYMBOL_GPL (rcu_irq_enter); -#if defined(CONFIG_DEBUG_LOCK_ALLOC ) || defined(CONFIG_RCU_TRACE )/* Signature · 1 line
1.8.1.5
Steven Rostedt
reviewerRe: [PATCH tip/core/rcu 2/3] rcutorture: Test both RCU-sched and RCU-bh for Tiny RCU
In reply to [email protected]Quoted 11 lines · On Tue, 12 May 2015 17:59:29 -0700 [email protected] wrote:
Quoted 4 lines · Paul E. McKenney
From: "Paul E. McKenney" <[email protected]>
Reported-by "Ahmed, Iftekhar" <[email protected]>Signed-off-by Paul E. McKenney <[email protected]>Could you elaborate a bit more on this patch (ideally in its commit message)? I see an addition of a command-line parameter to test
rcu_bh; is rcu-sched already tested elsewhere by some other config, or does this parameter somehow enable testing both?
Agreed. Subject only commits should only be for typo fixes in comments.
-- Steve
Paul E. McKenney
AppliedRe: [PATCH tip/core/rcu 3/3] rcu: Correctly handle non-empty Tiny RCU callback list with none ready
In reply to [email protected]Quoted 20 lines · On Tue, May 12, 2015 at 05:58:21PM -0700, [email protected] wrote:
Quoted 19 lines · Paul E. McKenney
From: "Paul E. McKenney" <[email protected]>
If, at the time
__rcu_process_callbacks()is invoked, there are callbacks in Tiny RCU's callback list, but none of them are ready to be invoked, the current list-management code will knit the non-ready callbacks out of the list. This can result in hangs and possibly worse. This commit therefore inserts a check for there being no callbacks that can be invoked immediately.This bug is unlikely to occur -- you have to get a new callback between the time
rcu_sched_qs()orrcu_bh_qs()was called, but before we get to__rcu_process_callbacks(). It was detected by the addition of RCU-bh testing to rcutorture, which in turn was instigated by Iftekhar Ahmed's mutation testing. Although this bug was made much more likely by915e8a4fe45e(rcu: Remove fastpath from__rcu_process_callbacks()), this
did not cause the bug, but rather made it much more probable. That
said, it takes more than 40 hours of rcutorture testing, on average, for this bug to appear, so this fix cannot be considered an emergency.
Quoted 4 lines
Quoted 2 lines
Signed-off-by Paul E. McKenney <[email protected]>Cc <[email protected]>Ouch, subtle.
Indeed! A bit of a cautionary tale for those who believe that bugs occur only in concurrent code. Of course, they could respond that this bug was in fact due to a concurrent interrupt handler. Still, I must confess that this bug is a bit embarrassing. ;-)
Quoted 1 line
Reviewed-by Josh Triplett <[email protected]>
Thank you, applied!
Thanx, Paul
Quoted 23 lines
Quoted 22 lines
kernel/rcu/tiny.c | 5 +++++ 1 file changed, 5 insertions(+)kernel/rcu/tiny.c +5
-0@@ -137,6 +137,11 @@ static void __rcu_process_callbacks(struct rcu_ctrlblk *rcp)/* Move the ready-to-invoke callbacks to a local list. */ local_irq_save (flags); +if (rcp->donetail == &rcp->rcucblist) { +/* No callbacks ready, so just leave. */ +local_irq_restore (flags); +return ; + }RCU_TRACE (trace_rcu_batch_start (rcp->name,0 , rcp->qlen,-1 )); list = rcp->rcucblist; rcp->rcucblist = *rcp->donetail;Signature · 1 line
1.8.1.5
Re: [PATCH tip/core/rcu 2/3] rcutorture: Test both RCU-sched and RCU-bh for Tiny RCU
In reply to [email protected]Quoted 10 lines · On Tue, May 12, 2015 at 05:59:29PM -0700, [email protected] wrote:
Quoted 4 lines · Paul E. McKenney
From: "Paul E. McKenney" <[email protected]>
Reported-by "Ahmed, Iftekhar" <[email protected]>Signed-off-by Paul E. McKenney <[email protected]>Could you elaborate a bit more on this patch (ideally in its commit message)? I see an addition of a command-line parameter to test
rcu_bh; is rcu-sched already tested elsewhere by some other config, or does this parameter somehow enable testing both?
The commit log now reads as follows, does that help?
Thanx, Paul
------------------------------------------------------------------------
rcutorture: Test both RCU-sched and RCU-bh for Tiny RCU
Tiny RCU supports both RCU-sched and RCU-bh, but only RCU-sched is currently tested by the rcutorture scripts. This commit therefore changes the TINY02 configuration to test RCU-bh, with TINY01 continuing to test RCU-sched.
This shortcoming of the current rcutorture tests was located by mutation testing by Iftekhar. The idea behind mutation testing is to automatically mutate the code under test. If a given mutant is not caught by testing, this is a hint that the testing might need to be improved, as was the case here. Note that this is only a hint because it is possible to mutate the code into something else that still works. For example, a mutation that removes (say) a WARN_ON() will not normally result in a test failure.
This change resulted in the test failure caused by list mishandling, which is fixed by the next commit.
Re: [PATCH tip/core/rcu 2/3] rcutorture: Test both RCU-sched and RCU-bh for Tiny RCU
Quoted 33 lines · Paul E. McKenney
Quoted 10 lines · On Tue, May 12, 2015 at 05:59:29PM -0700, [email protected] wrote:
Quoted 4 lines · Paul E. McKenney
From: "Paul E. McKenney" <[email protected]>
Reported-by "Ahmed, Iftekhar" <[email protected]>Signed-off-by Paul E. McKenney <[email protected]>Could you elaborate a bit more on this patch (ideally in its commit message)? I see an addition of a command-line parameter to test
rcu_bh; is rcu-sched already tested elsewhere by some other config, or does this parameter somehow enable testing both?The commit log now reads as follows, does that help?
Thanx, Paul
------------------------------------------------------------------------
rcutorture: Test both RCU-sched and RCU-bh for Tiny RCU
Tiny RCU supports both RCU-sched and RCU-bh, but only RCU-sched is currently tested by the rcutorture scripts. This commit therefore changes the TINY02 configuration to test RCU-bh, with TINY01 continuing to test RCU-sched.
This shortcoming of the current rcutorture tests was located by mutation testing by Iftekhar. The idea behind mutation testing is to automatically mutate the code under test. If a given mutant is not caught by testing, this is a hint that the testing might need to be improved, as was the case here. Note that this is only a hint because it is possible to mutate the code into something else that still works. For example, a mutation that removes (say) a
WARN_ON()will not normally result in a test failure.
Quoted 6 lines
This change resulted in the test failure caused by list mishandling, which is fixed by the next commit.
Reported-by "Ahmed, Iftekhar" <[email protected]>Signed-off-by Paul E. McKenney <[email protected]>
Much better, thanks. In particular, the information about TINY01 and TINY02 was not obvious from the patch.