Re: [PATCH, RFC] improved hacks to allow -rt to run kernbench on POWER

Score:

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

To: Benjamin Herrenschmidt <benh@...>

Cc: <linux-kernel@...>, <tony@...>, <paulus@...>, <dino@...>, <tytso@...>, <dvhltc@...>, <niv@...>, <antonb@...>, <rostedt@...>

Subject: Re: [PATCH, RFC] improved hacks to allow -rt to run kernbench on POWER

Date: Friday, November 9, 2007 - 6:07 pm

On Sat, Nov 10, 2007 at 07:52:04AM +1100, Benjamin Herrenschmidt wrote:
quoted text> > diff -urpNa -X dontdiff linux-2.6.23.1-rt4/arch/powerpc/kernel/process.c linux-2.6.23.1-rt4-fix/arch/powerpc/kernel/process.c
> > --- linux-2.6.23.1-rt4/arch/powerpc/kernel/process.c	2007-10-12 09:43:44.000000000 -0700
> > +++ linux-2.6.23.1-rt4-fix/arch/powerpc/kernel/process.c	2007-11-08 20:33:59.000000000 -0800
> > @@ -245,6 +245,8 @@ struct task_struct *__switch_to(struct t
> >  	struct thread_struct *new_thread, *old_thread;
> >  	unsigned long flags;
> >  	struct task_struct *last;
> > +	struct ppc64_tlb_batch *batch;
> > +	int hadbatch;
> >  
> >  #ifdef CONFIG_SMP
> >  	/* avoid complexity of lazy save/restore of fpu
> > @@ -325,6 +327,16 @@ struct task_struct *__switch_to(struct t
> >  	}
> >  #endif
> >  
> > +	batch = &get_cpu_var(ppc64_tlb_batch);
> > +	if (batch->active) {
> > +		hadbatch = 1;
> > +		if (batch->index) {
> > +			__flush_tlb_pending(batch);
> > +		}
> > +		batch->active = 0;
> > +	}
> > +	put_cpu_var(ppc64_tlb_batch);
> > +
> >  	local_irq_save(flags);
> >  
> >  	account_system_vtime(current);
> > @@ -335,6 +347,12 @@ struct task_struct *__switch_to(struct t
> >  
> >  	local_irq_restore(flags);
> >  
> > +	if (hadbatch) {
> > +		batch = &get_cpu_var(ppc64_tlb_batch);
> > +		batch->active = 1;
> > +		put_cpu_var(ppc64_tlb_batch);
> > +	}
> > +
> >  	return last;
> >  }
> 
> I doubt we can schedule within __switch_to() (can somebody confirm
> this ?), in which case, you can just use __get_cpu_var() and avoid
> the put, thus saving a handful of cycles in the code above.

Thank you for looking this over!

I tried this, and it seemed to work.

quoted text> >  /* export that to outside world */
> >  struct device_node *of_chosen;
> > diff -urpNa -X dontdiff linux-2.6.23.1-rt4/arch/powerpc/mm/fault.c linux-2.6.23.1-rt4-fix/arch/powerpc/mm/fault.c
> > --- linux-2.6.23.1-rt4/arch/powerpc/mm/fault.c	2007-10-27 22:20:57.000000000 -0700
> > +++ linux-2.6.23.1-rt4-fix/arch/powerpc/mm/fault.c	2007-10-28 13:49:07.000000000 -0700
> > @@ -301,6 +301,7 @@ good_area:
> >  		if (get_pteptr(mm, address, &ptep, &pmdp)) {
> >  			spinlock_t *ptl = pte_lockptr(mm, pmdp);
> >  			spin_lock(ptl);
> > +			preempt_disable();
> >  			if (pte_present(*ptep)) {
> >  				struct page *page = pte_page(*ptep);
> >  
> > @@ -310,10 +311,12 @@ good_area:
> >  				}
> >  				pte_update(ptep, 0, _PAGE_HWEXEC);
> >  				_tlbie(address);
> > +				preempt_enable();
> >  				pte_unmap_unlock(ptep, ptl);
> >  				up_read(&mm->mmap_sem);
> >  				return 0;
> >  			}
> > +			preempt_enable();
> >  			pte_unmap_unlock(ptep, ptl);
> >  		}
> >  #endif
> 
> 
> The patch above will have to be rebased following a fix I did that makes
> _tlbie() takes a context ID argument. I don't actually think
> preempt_disable/enable is necessary here. I don't think it matters if we
> preempt here, but I suppose better safe than sorry.... (this is 44x
> code).

Removed this as well, also seemed to work.  Please note, however, that
this is just running kernbench.  But this did seem to get rid of some
of the warnings as well.  ;-)  Now only have the xics_startup() warning.

quoted text> Overall, looks fine !

I bet that there are more gotchas lurking in there somewhere, but in
the meantime here is the updated patch.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---

 arch/powerpc/kernel/process.c        |   16 ++++++++++++++++
 arch/powerpc/kernel/prom.c           |    2 +-
 arch/powerpc/mm/tlb_64.c             |    5 ++++-
 arch/powerpc/platforms/pseries/eeh.c |    2 +-
 drivers/of/base.c                    |    2 +-
 include/asm-powerpc/tlb.h            |    5 ++++-
 include/asm-powerpc/tlbflush.h       |   15 ++++++++++-----
 7 files changed, 37 insertions(+), 10 deletions(-)

diff -urpNa -X dontdiff linux-2.6.23.1-rt4/arch/powerpc/kernel/process.c linux-2.6.23.1-rt4-fix/arch/powerpc/kernel/process.c
--- linux-2.6.23.1-rt4/arch/powerpc/kernel/process.c	2007-10-12 09:43:44.000000000 -0700
+++ linux-2.6.23.1-rt4-fix/arch/powerpc/kernel/process.c	2007-11-09 13:24:46.000000000 -0800
@@ -245,6 +245,8 @@ struct task_struct *__switch_to(struct t
 	struct thread_struct *new_thread, *old_thread;
 	unsigned long flags;
 	struct task_struct *last;
+	struct ppc64_tlb_batch *batch;
+	int hadbatch;
 
 #ifdef CONFIG_SMP
 	/* avoid complexity of lazy save/restore of fpu
@@ -325,6 +327,15 @@ struct task_struct *__switch_to(struct t
 	}
 #endif
 
+	batch = &__get_cpu_var(ppc64_tlb_batch);
+	if (batch->active) {
+		hadbatch = 1;
+		if (batch->index) {
+			__flush_tlb_pending(batch);
+		}
+		batch->active = 0;
+	}
+
 	local_irq_save(flags);
 
 	account_system_vtime(current);
@@ -335,6 +346,11 @@ struct task_struct *__switch_to(struct t
 
 	local_irq_restore(flags);
 
+	if (hadbatch) {
+		batch = &__get_cpu_var(ppc64_tlb_batch);
+		batch->active = 1;
+	}
+
 	return last;
 }
 
diff -urpNa -X dontdiff linux-2.6.23.1-rt4/arch/powerpc/kernel/prom.c linux-2.6.23.1-rt4-fix/arch/powerpc/kernel/prom.c
--- linux-2.6.23.1-rt4/arch/powerpc/kernel/prom.c	2007-10-12 09:43:44.000000000 -0700
+++ linux-2.6.23.1-rt4-fix/arch/powerpc/kernel/prom.c	2007-10-28 13:37:23.000000000 -0700
@@ -80,7 +80,7 @@ struct boot_param_header *initial_boot_p
 
 extern struct device_node *allnodes;	/* temporary while merging */
 
-extern rwlock_t devtree_lock;	/* temporary while merging */
+extern raw_rwlock_t devtree_lock;	/* temporary while merging */
 
 /* export that to outside world */
 struct device_node *of_chosen;
diff -urpNa -X dontdiff linux-2.6.23.1-rt4/arch/powerpc/mm/tlb_64.c linux-2.6.23.1-rt4-fix/arch/powerpc/mm/tlb_64.c
--- linux-2.6.23.1-rt4/arch/powerpc/mm/tlb_64.c	2007-10-27 22:20:57.000000000 -0700
+++ linux-2.6.23.1-rt4-fix/arch/powerpc/mm/tlb_64.c	2007-11-08 16:49:04.000000000 -0800
@@ -133,7 +133,7 @@ void pgtable_free_tlb(struct mmu_gather 
 void hpte_need_flush(struct mm_struct *mm, unsigned long addr,
 		     pte_t *ptep, unsigned long pte, int huge)
 {
-	struct ppc64_tlb_batch *batch = &__get_cpu_var(ppc64_tlb_batch);
+	struct ppc64_tlb_batch *batch = &get_cpu_var(ppc64_tlb_batch);
 	unsigned long vsid, vaddr;
 	unsigned int psize;
 	real_pte_t rpte;
@@ -180,6 +180,7 @@ void hpte_need_flush(struct mm_struct *m
 	 */
 	if (!batch->active) {
 		flush_hash_page(vaddr, rpte, psize, 0);
+		put_cpu_var(ppc64_tlb_batch);
 		return;
 	}
 
@@ -212,12 +213,14 @@ void hpte_need_flush(struct mm_struct *m
 	 */
 	if (machine_is(celleb)) {
 		__flush_tlb_pending(batch);
+		put_cpu_var(ppc64_tlb_batch);
 		return;
 	}
 #endif /* CONFIG_PREEMPT_RT */
 
 	if (i >= PPC64_TLB_BATCH_NR)
 		__flush_tlb_pending(batch);
+	put_cpu_var(ppc64_tlb_batch);
 }
 
 /*
diff -urpNa -X dontdiff linux-2.6.23.1-rt4/arch/powerpc/platforms/pseries/eeh.c linux-2.6.23.1-rt4-fix/arch/powerpc/platforms/pseries/eeh.c
--- linux-2.6.23.1-rt4/arch/powerpc/platforms/pseries/eeh.c	2007-10-12 09:43:44.000000000 -0700
+++ linux-2.6.23.1-rt4-fix/arch/powerpc/platforms/pseries/eeh.c	2007-10-28 15:43:54.000000000 -0700
@@ -97,7 +97,7 @@ int eeh_subsystem_enabled;
 EXPORT_SYMBOL(eeh_subsystem_enabled);
 
 /* Lock to avoid races due to multiple reports of an error */
-static DEFINE_SPINLOCK(confirm_error_lock);
+static DEFINE_RAW_SPINLOCK(confirm_error_lock);
 
 /* Buffer for reporting slot-error-detail rtas calls. Its here
  * in BSS, and not dynamically alloced, so that it ends up in
diff -urpNa -X dontdiff linux-2.6.23.1-rt4/drivers/of/base.c linux-2.6.23.1-rt4-fix/drivers/of/base.c
--- linux-2.6.23.1-rt4/drivers/of/base.c	2007-10-12 09:43:44.000000000 -0700
+++ linux-2.6.23.1-rt4-fix/drivers/of/base.c	2007-10-28 13:38:36.000000000 -0700
@@ -25,7 +25,7 @@ struct device_node *allnodes;
 /* use when traversing tree through the allnext, child, sibling,
  * or parent members of struct device_node.
  */
-DEFINE_RWLOCK(devtree_lock);
+DEFINE_RAW_RWLOCK(devtree_lock);
 
 int of_n_addr_cells(struct device_node *np)
 {
diff -urpNa -X dontdiff linux-2.6.23.1-rt4/include/asm-powerpc/tlbflush.h linux-2.6.23.1-rt4-fix/include/asm-powerpc/tlbflush.h
--- linux-2.6.23.1-rt4/include/asm-powerpc/tlbflush.h	2007-10-12 09:43:44.000000000 -0700
+++ linux-2.6.23.1-rt4-fix/include/asm-powerpc/tlbflush.h	2007-11-08 17:11:18.000000000 -0800
@@ -109,18 +109,23 @@ extern void hpte_need_flush(struct mm_st
 
 static inline void arch_enter_lazy_mmu_mode(void)
 {
-	struct ppc64_tlb_batch *batch = &__get_cpu_var(ppc64_tlb_batch);
+	struct ppc64_tlb_batch *batch = &get_cpu_var(ppc64_tlb_batch);
 
 	batch->active = 1;
+	put_cpu_var(ppc64_tlb_batch);
 }
 
 static inline void arch_leave_lazy_mmu_mode(void)
 {
-	struct ppc64_tlb_batch *batch = &__get_cpu_var(ppc64_tlb_batch);
+	struct ppc64_tlb_batch *batch = &get_cpu_var(ppc64_tlb_batch);
 
-	if (batch->index)
-		__flush_tlb_pending(batch);
-	batch->active = 0;
+	if (batch->active) {
+		if (batch->index) {
+			__flush_tlb_pending(batch);
+		}
+		batch->active = 0;
+	}
+	put_cpu_var(ppc64_tlb_batch);
 }
 
 #define arch_flush_lazy_mmu_mode()      do {} while (0)
diff -urpNa -X dontdiff linux-2.6.23.1-rt4/include/asm-powerpc/tlb.h linux-2.6.23.1-rt4-fix/include/asm-powerpc/tlb.h
--- linux-2.6.23.1-rt4/include/asm-powerpc/tlb.h	2007-10-12 09:43:44.000000000 -0700
+++ linux-2.6.23.1-rt4-fix/include/asm-powerpc/tlb.h	2007-10-28 11:36:05.000000000 -0700
@@ -44,8 +44,11 @@ static inline void tlb_flush(struct mmu_
 	 * pages are going to be freed and we really don't want to have a CPU
 	 * access a freed page because it has a stale TLB
 	 */
-	if (tlbbatch->index)
+	if (tlbbatch->index) {
+		preempt_disable();
 		__flush_tlb_pending(tlbbatch);
+		preempt_enable();
+	}
 
 	pte_free_finish();
 }
-
unsubscribe noticeTo unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:

[PATCH, RFC] improved hacks to allow -rt to run kernbench on..., Paul E. McKenney, (Fri Nov 9, 2:10 pm)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Benjamin Herrenschmidt, (Fri Nov 9, 4:52 pm)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Nick Piggin, (Fri Nov 9, 9:18 pm)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Paul E. McKenney, (Fri Nov 9, 6:07 pm)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Benjamin Herrenschmidt, (Sat Nov 10, 11:59 pm)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Steven Rostedt, (Sun Nov 11, 10:45 am)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Benjamin Herrenschmidt, (Sun Nov 11, 4:48 pm)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Paul E. McKenney, (Mon Nov 12, 3:15 pm)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Benjamin Herrenschmidt, (Mon Nov 12, 5:12 pm)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Peter Zijlstra, (Tue Nov 13, 5:10 am)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Benjamin Herrenschmidt, (Tue Nov 13, 5:43 am)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Peter Zijlstra, (Tue Nov 13, 6:05 am)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Benjamin Herrenschmidt, (Tue Nov 13, 7:09 am)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Peter Zijlstra, (Tue Nov 13, 6:59 am)

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbenc..., Paul E. McKenney, (Mon Nov 12, 1:05 pm)


linux-kernel:
Chuck Ebbert	Wanted: simple, safe x86 stack overflow detection
Alan Cox	Re: ndiswrapper and GPL-only symbols redux
Yinghai Lu	[PATCH 03/42] x86: remove irq_vectors_limits
Greg Kroah-Hartman	[PATCH 001/196] Chinese: Add the known_regression URI to the HOWTO
git:
しらいしななこ	Re: [ANNOUNCE] GIT 1.5.4
Jan Wielemaker	git filter-branch --subdirectory-filter, still a mistery
Pierre Habouzit	[PATCH] guilt(1): Obvious bashisms fixed.
Christopher Faylor	Re: First cut at git port to Cygwin
openbsd-misc:
Thilo Pfennig	OpenBSD project goals
Marco Peereboom	Re: Real men don't attack straw men
Daniel Hazelton	Re: Wasting our Freedom
Luke Bakken	Re: No Blob without Puffy
linux-netdev:
Julius Volz	[PATCHv3 19/24] IVPS: Disable sync daemon for IPv6 connections
Paul Moore	[RFC PATCH v4 04/14] selinux: Fix missing calls to netlbl_skbuff_err()
Dave Jones	odd RTL8139 quirk.
Patrick McHardy	[NET_SCHED 04/15]: act_api: use nlmsg_parse


How to exec user process in kernel mode.	1 day ago	Linux kernel
[IPSEC]IPSEC_MANUAL_REQID_MAX	1 day ago	Linux kernel
help in UDP catching module..	2 days ago	Linux kernel
Is there anything like Real-time drivers?	4 days ago	Linux general
ns16550 serail console in Linux 2.6.19	4 days ago	Linux general
what class should i use to register my devices	4 days ago	Linux kernel
reset bios pasword toshiba	5 days ago	Hardware
Analysis of Process Scheduling	6 days ago	Linux kernel
RT Kernel and SSH Server Panics	6 days ago	Linux kernel
Resetting the bios password for Toshiba Laptop	6 days ago	Hardware

Re: [PATCH, RFC] improved hacks to allow -rt to run kernbench on POWER

Online users