Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1603328 > unrolled thread

Re: [locking/lockdep] 383776fa75: INFO: trying to register non-static key.

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [locking/lockdep] 383776fa75:  INFO: trying to register  non-static key. Peter Zijlstra <peterz@infradead.org> - 2017-03-17 15:00 +0100
    Re: [locking/lockdep] 383776fa75:  INFO: trying to register  non-static key. Sebastian Andrzej Siewior <bigeasy@linutronix.de> - 2017-03-17 15:50 +0100
      Re: [locking/lockdep] 383776fa75:  INFO: trying to register  non-static key. Peter Zijlstra <peterz@infradead.org> - 2017-03-20 12:50 +0100
        Re: [locking/lockdep] 383776fa75:  INFO: trying to register  non-static key. Borislav Petkov <bp@alien8.de> - 2017-03-25 19:20 +0100
        [tip:locking/core] lockdep: Fix per-cpu static objects tip-bot for Peter Zijlstra <tipbot@zytor.com> - 2017-03-26 11:50 +0200
        [tip:locking/core] lockdep: Fix per-cpu static objects tip-bot for Peter Zijlstra <tipbot@zytor.com> - 2017-03-26 15:30 +0200

#1603328 — Re: [locking/lockdep] 383776fa75: INFO: trying to register non-static key.

On Fri, Mar 17, 2017 at 02:07:05AM +0800, kernel test robot wrote:

>     locking/lockdep: Handle statically initialized PER_CPU locks properly

> [   11.712266] INFO: trying to register non-static key.

Blergh; so the problem is that when we assign can_addr to lock->key, we
can, upon using a different subclass, reach static_obj(lock->key), which
will fail on the can_addr.

One way to fix this would be to redefine the canonical address as the
per-cpu address for a specific cpu; the below hard codes cpu0, but I'm
not sure we want to rely on cpu0 being a valid cpu.


---
diff --git a/kernel/module.c b/kernel/module.c
index 5ef618133849..bdd9d62ce08c 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -682,8 +682,10 @@ bool __is_module_percpu_address(unsigned long addr, unsigned long *can_addr)
 			void *va = (void *)addr;
 
 			if (va >= start && va < start + mod->percpu_size) {
-				if (can_addr)
+				if (can_addr) {
 					*can_addr = (unsigned long) (va - start);
+					*can_addr += (unsigned long)per_cpu_ptr(mod->percpu, 0);
+				}
 				preempt_enable();
 				return true;
 			}
diff --git a/mm/percpu.c b/mm/percpu.c
index e30c995f2b7b..a5d7b7477888 100644
--- a/mm/percpu.c
+++ b/mm/percpu.c
@@ -1296,8 +1296,10 @@ bool __is_kernel_percpu_address(unsigned long addr, unsigned long *can_addr)
 		void *va = (void *)addr;
 
 		if (va >= start && va < start + static_size) {
-			if (can_addr)
+			if (can_addr) {
 				*can_addr = (unsigned long) (va - start);
+				*can_addr += (unsigned long)per_cpu_ptr(base, 0);
+			}
 			return true;
 		}
 	}

[toc] | [next] | [standalone]

#1603368

On 2017-03-17 14:41:09 [+0100], Peter Zijlstra wrote:
> On Fri, Mar 17, 2017 at 02:07:05AM +0800, kernel test robot wrote:
> 
> >     locking/lockdep: Handle statically initialized PER_CPU locks properly
> 
> > [   11.712266] INFO: trying to register non-static key.
> 
> Blergh; so the problem is that when we assign can_addr to lock->key, we
> can, upon using a different subclass, reach static_obj(lock->key), which
> will fail on the can_addr.
> 
> One way to fix this would be to redefine the canonical address as the
> per-cpu address for a specific cpu; the below hard codes cpu0, but I'm
> not sure we want to rely on cpu0 being a valid cpu.

This solves two problems: The one reported by the bot. The other thing,
that is fixed by the patch, is that the first PER-CPU variable built-in
will return 0 for can_addr and so will the first variable in every
module. As far as I understand it, this should be unique and having the
same value for multiple different variables does not look too good :)
So adding the offset from CPU0 sounds good.

Sebastian

[toc] | [prev] | [next] | [standalone]

#1604428

On Fri, Mar 17, 2017 at 03:41:41PM +0100, Sebastian Andrzej Siewior wrote:
> On 2017-03-17 14:41:09 [+0100], Peter Zijlstra wrote:
> > On Fri, Mar 17, 2017 at 02:07:05AM +0800, kernel test robot wrote:
> > 
> > >     locking/lockdep: Handle statically initialized PER_CPU locks properly
> > 
> > > [   11.712266] INFO: trying to register non-static key.
> > 
> > Blergh; so the problem is that when we assign can_addr to lock->key, we
> > can, upon using a different subclass, reach static_obj(lock->key), which
> > will fail on the can_addr.
> > 
> > One way to fix this would be to redefine the canonical address as the
> > per-cpu address for a specific cpu; the below hard codes cpu0, but I'm
> > not sure we want to rely on cpu0 being a valid cpu.
> 
> This solves two problems: The one reported by the bot. The other thing,
> that is fixed by the patch, is that the first PER-CPU variable built-in
> will return 0 for can_addr and so will the first variable in every
> module. As far as I understand it, this should be unique and having the
> same value for multiple different variables does not look too good :)
> So adding the offset from CPU0 sounds good.

Right; so how about something liek this?

---
Subject: lockdep: Fix per-cpu static objects
From: Peter Zijlstra <peterz@infradead.org>
Date: Mon Mar 20 12:26:55 CET 2017

Since commit:

  383776fa7527 ("locking/lockdep: Handle statically initialized PER_CPU locks properly")

we try to collapse per-cpu locks into a single class by giving them
all the same key. For this key we choose the canonical address of the
per-cpu object, which would be the offset into the per-cpu area.

This has two problems:

 - there is a case where we run !0 lock->key through static_obj() and
   expect this to pass; it doesn't for canonical pointers.

 - 0 is a valid canonical address.

Cure both issues by redefining the canonical address as the address of
the per-cpu variable on the boot CPU.

Since I didn't want to rely on CPU0 being the boot-cpu, or even
existing at all, track the boot CPU in a variable.

Fixes: 383776fa7527 ("locking/lockdep: Handle statically initialized PER_CPU locks properly")
Reported-by: kernel test robot <fengguang.wu@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
---
--- a/include/linux/smp.h
+++ b/include/linux/smp.h
@@ -120,6 +120,13 @@ extern unsigned int setup_max_cpus;
 extern void __init setup_nr_cpu_ids(void);
 extern void __init smp_init(void);
 
+extern int __boot_cpu_id;
+
+static inline int boot_cpu_id(void)
+{
+	return __boot_cpu_id;
+}
+
 #else /* !SMP */
 
 static inline void smp_send_stop(void) { }
@@ -158,6 +165,11 @@ static inline void smp_init(void) { up_l
 static inline void smp_init(void) { }
 #endif
 
+static inline int boot_cpu_id(void)
+{
+	return 0;
+}
+
 #endif /* !SMP */
 
 /*
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -1125,6 +1125,8 @@ core_initcall(cpu_hotplug_pm_sync_init);
 
 #endif /* CONFIG_PM_SLEEP_SMP */
 
+int __boot_cpu_id;
+
 #endif /* CONFIG_SMP */
 
 /* Boot processor state steps */
@@ -1815,6 +1817,10 @@ void __init boot_cpu_init(void)
 	set_cpu_active(cpu, true);
 	set_cpu_present(cpu, true);
 	set_cpu_possible(cpu, true);
+
+#ifdef CONFIG_SMP
+	__boot_cpu_id = cpu;
+#endif
 }
 
 /*
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -682,8 +682,11 @@ bool __is_module_percpu_address(unsigned
 			void *va = (void *)addr;
 
 			if (va >= start && va < start + mod->percpu_size) {
-				if (can_addr)
+				if (can_addr) {
 					*can_addr = (unsigned long) (va - start);
+					*can_addr += (unsigned long)
+						per_cpu_ptr(mod->percpu, boot_cpu_id());
+				}
 				preempt_enable();
 				return true;
 			}
--- a/mm/percpu.c
+++ b/mm/percpu.c
@@ -1296,8 +1296,11 @@ bool __is_kernel_percpu_address(unsigned
 		void *va = (void *)addr;
 
 		if (va >= start && va < start + static_size) {
-			if (can_addr)
+			if (can_addr) {
 				*can_addr = (unsigned long) (va - start);
+				*can_addr += (unsigned long)
+					per_cpu_ptr(base, boot_cpu_id());
+			}
 			return true;
 		}
 	}

[toc] | [prev] | [next] | [standalone]

#1609202

On Mon, Mar 20, 2017 at 12:41:08PM +0100, Peter Zijlstra wrote:
> Subject: lockdep: Fix per-cpu static objects
> From: Peter Zijlstra <peterz@infradead.org>
> Date: Mon Mar 20 12:26:55 CET 2017
> 
> Since commit:
> 
>   383776fa7527 ("locking/lockdep: Handle statically initialized PER_CPU locks properly")
> 
> we try to collapse per-cpu locks into a single class by giving them
> all the same key. For this key we choose the canonical address of the
> per-cpu object, which would be the offset into the per-cpu area.
> 
> This has two problems:
> 
>  - there is a case where we run !0 lock->key through static_obj() and
>    expect this to pass; it doesn't for canonical pointers.
> 
>  - 0 is a valid canonical address.
> 
> Cure both issues by redefining the canonical address as the address of
> the per-cpu variable on the boot CPU.
> 
> Since I didn't want to rely on CPU0 being the boot-cpu, or even
> existing at all, track the boot CPU in a variable.
> 
> Fixes: 383776fa7527 ("locking/lockdep: Handle statically initialized PER_CPU locks properly")
> Reported-by: kernel test robot <fengguang.wu@intel.com>
> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>

Tested-by: Borislav Petkov <bp@suse.de>

-- 
Regards/Gruss,
    Boris.

Good mailing practices for 400: avoid top-posting and trim the reply.

[toc] | [prev] | [next] | [standalone]

#1609309 — [tip:locking/core] lockdep: Fix per-cpu static objects

Commit-ID:  4ef9ea9524c0015e551d062ac75ab9fd04365277
Gitweb:     http://git.kernel.org/tip/4ef9ea9524c0015e551d062ac75ab9fd04365277
Author:     Peter Zijlstra <peterz@infradead.org>
AuthorDate: Mon, 20 Mar 2017 12:26:55 +0100
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Sun, 26 Mar 2017 11:27:42 +0200

lockdep: Fix per-cpu static objects

Since commit 383776fa7527 ("locking/lockdep: Handle statically initialized
PER_CPU locks properly") we try to collapse per-cpu locks into a single
class by giving them all the same key. For this key we choose the canonical
address of the per-cpu object, which would be the offset into the per-cpu
area.

This has two problems:

 - there is a case where we run !0 lock->key through static_obj() and
   expect this to pass; it doesn't for canonical pointers.

 - 0 is a valid canonical address.

Cure both issues by redefining the canonical address as the address of the
per-cpu variable on the boot CPU.

Since I didn't want to rely on CPU0 being the boot-cpu, or even existing at
all, track the boot CPU in a variable.

Fixes: 383776fa7527 ("locking/lockdep: Handle statically initialized PER_CPU locks properly")
Reported-by: kernel test robot <fengguang.wu@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Borislav Petkov <bp@suse.de>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: linux-mm@kvack.org
Cc: wfg@linux.intel.com
Cc: kernel test robot <fengguang.wu@intel.com>
Cc: LKP <lkp@01.org>
Link: http://lkml.kernel.org/r/20170320114108.kbvcsuepem45j5cr@hirez.programming.kicks-ass.net
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

---
 include/linux/smp.h | 12 ++++++++++++
 kernel/cpu.c        |  6 ++++++
 kernel/module.c     |  5 ++++-
 mm/percpu.c         |  5 ++++-
 4 files changed, 26 insertions(+), 2 deletions(-)

diff --git a/include/linux/smp.h b/include/linux/smp.h
index 8e0cb7a..8d41185 100644
--- a/include/linux/smp.h
+++ b/include/linux/smp.h
@@ -120,6 +120,13 @@ extern unsigned int setup_max_cpus;
 extern void __init setup_nr_cpu_ids(void);
 extern void __init smp_init(void);
 
+extern int __boot_cpu_id;
+
+static inline int boot_cpu_id(void)
+{
+	return __boot_cpu_id;
+}
+
 #else /* !SMP */
 
 static inline void smp_send_stop(void) { }
@@ -158,6 +165,11 @@ static inline void smp_init(void) { up_late_init(); }
 static inline void smp_init(void) { }
 #endif
 
+static inline int boot_cpu_id(void)
+{
+	return 0;
+}
+
 #endif /* !SMP */
 
 /*
diff --git a/kernel/cpu.c b/kernel/cpu.c
index f7c0632..a795725 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -1125,6 +1125,8 @@ core_initcall(cpu_hotplug_pm_sync_init);
 
 #endif /* CONFIG_PM_SLEEP_SMP */
 
+int __boot_cpu_id;
+
 #endif /* CONFIG_SMP */
 
 /* Boot processor state steps */
@@ -1815,6 +1817,10 @@ void __init boot_cpu_init(void)
 	set_cpu_active(cpu, true);
 	set_cpu_present(cpu, true);
 	set_cpu_possible(cpu, true);
+
+#ifdef CONFIG_SMP
+	__boot_cpu_id = cpu;
+#endif
 }
 
 /*
diff --git a/kernel/module.c b/kernel/module.c
index 5ef6181..043ca86 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -682,8 +682,11 @@ bool __is_module_percpu_address(unsigned long addr, unsigned long *can_addr)
 			void *va = (void *)addr;
 
 			if (va >= start && va < start + mod->percpu_size) {
-				if (can_addr)
+				if (can_addr) {
 					*can_addr = (unsigned long) (va - start);
+					*can_addr += (unsigned long)
+						per_cpu_ptr(mod->percpu, boot_cpu_id());
+				}
 				preempt_enable();
 				return true;
 			}
diff --git a/mm/percpu.c b/mm/percpu.c
index 7d3b728..c95b475 100644
--- a/mm/percpu.c
+++ b/mm/percpu.c
@@ -1293,8 +1293,11 @@ bool __is_kernel_percpu_address(unsigned long addr, unsigned long *can_addr)
 		void *va = (void *)addr;
 
 		if (va >= start && va < start + static_size) {
-			if (can_addr)
+			if (can_addr) {
 				*can_addr = (unsigned long) (va - start);
+				*can_addr += (unsigned long)
+					per_cpu_ptr(base, boot_cpu_id());
+			}
 			return true;
 		}
 	}

[toc] | [prev] | [next] | [standalone]

#1609328 — [tip:locking/core] lockdep: Fix per-cpu static objects

Commit-ID:  8ce371f9846ef1e8b3cc8f6865766cb5c1f17e40
Gitweb:     http://git.kernel.org/tip/8ce371f9846ef1e8b3cc8f6865766cb5c1f17e40
Author:     Peter Zijlstra <peterz@infradead.org>
AuthorDate: Mon, 20 Mar 2017 12:26:55 +0100
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Sun, 26 Mar 2017 15:09:45 +0200

lockdep: Fix per-cpu static objects

Since commit 383776fa7527 ("locking/lockdep: Handle statically initialized
PER_CPU locks properly") we try to collapse per-cpu locks into a single
class by giving them all the same key. For this key we choose the canonical
address of the per-cpu object, which would be the offset into the per-cpu
area.

This has two problems:

 - there is a case where we run !0 lock->key through static_obj() and
   expect this to pass; it doesn't for canonical pointers.

 - 0 is a valid canonical address.

Cure both issues by redefining the canonical address as the address of the
per-cpu variable on the boot CPU.

Since I didn't want to rely on CPU0 being the boot-cpu, or even existing at
all, track the boot CPU in a variable.

Fixes: 383776fa7527 ("locking/lockdep: Handle statically initialized PER_CPU locks properly")
Reported-by: kernel test robot <fengguang.wu@intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Tested-by: Borislav Petkov <bp@suse.de>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: linux-mm@kvack.org
Cc: wfg@linux.intel.com
Cc: kernel test robot <fengguang.wu@intel.com>
Cc: LKP <lkp@01.org>
Link: http://lkml.kernel.org/r/20170320114108.kbvcsuepem45j5cr@hirez.programming.kicks-ass.net
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

---
 include/linux/smp.h | 12 ++++++++++++
 kernel/cpu.c        |  6 ++++++
 kernel/module.c     |  6 +++++-
 mm/percpu.c         |  5 ++++-
 4 files changed, 27 insertions(+), 2 deletions(-)

diff --git a/include/linux/smp.h b/include/linux/smp.h
index 8e0cb7a..68123c1 100644
--- a/include/linux/smp.h
+++ b/include/linux/smp.h
@@ -120,6 +120,13 @@ extern unsigned int setup_max_cpus;
 extern void __init setup_nr_cpu_ids(void);
 extern void __init smp_init(void);
 
+extern int __boot_cpu_id;
+
+static inline int get_boot_cpu_id(void)
+{
+	return __boot_cpu_id;
+}
+
 #else /* !SMP */
 
 static inline void smp_send_stop(void) { }
@@ -158,6 +165,11 @@ static inline void smp_init(void) { up_late_init(); }
 static inline void smp_init(void) { }
 #endif
 
+static inline int get_boot_cpu_id(void)
+{
+	return 0;
+}
+
 #endif /* !SMP */
 
 /*
diff --git a/kernel/cpu.c b/kernel/cpu.c
index f7c0632..a795725 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -1125,6 +1125,8 @@ core_initcall(cpu_hotplug_pm_sync_init);
 
 #endif /* CONFIG_PM_SLEEP_SMP */
 
+int __boot_cpu_id;
+
 #endif /* CONFIG_SMP */
 
 /* Boot processor state steps */
@@ -1815,6 +1817,10 @@ void __init boot_cpu_init(void)
 	set_cpu_active(cpu, true);
 	set_cpu_present(cpu, true);
 	set_cpu_possible(cpu, true);
+
+#ifdef CONFIG_SMP
+	__boot_cpu_id = cpu;
+#endif
 }
 
 /*
diff --git a/kernel/module.c b/kernel/module.c
index 5ef6181..6d99880 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -682,8 +682,12 @@ bool __is_module_percpu_address(unsigned long addr, unsigned long *can_addr)
 			void *va = (void *)addr;
 
 			if (va >= start && va < start + mod->percpu_size) {
-				if (can_addr)
+				if (can_addr) {
 					*can_addr = (unsigned long) (va - start);
+					*can_addr += (unsigned long)
+						per_cpu_ptr(mod->percpu,
+							    get_boot_cpu_id());
+				}
 				preempt_enable();
 				return true;
 			}
diff --git a/mm/percpu.c b/mm/percpu.c
index 7d3b728..bd74167 100644
--- a/mm/percpu.c
+++ b/mm/percpu.c
@@ -1293,8 +1293,11 @@ bool __is_kernel_percpu_address(unsigned long addr, unsigned long *can_addr)
 		void *va = (void *)addr;
 
 		if (va >= start && va < start + static_size) {
-			if (can_addr)
+			if (can_addr) {
 				*can_addr = (unsigned long) (va - start);
+				*can_addr += (unsigned long)
+					per_cpu_ptr(base, get_boot_cpu_id());
+			}
 			return true;
 		}
 	}

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web