x86/irq: remove leftover code from NUMA_MIGRATE_IRQ_DESC

The original feature of migrating irq_desc dynamic was too fragile and was causing problems: it caused crashes on systems with lots of cards with MSI-X when user-space irq-balancer was enabled. We now have new patches that create irq_desc according to device numa node. This patch removes the leftover bits of the dynamic balancer. [ Impact: remove dead code ] Signed-off-by: Yinghai Lu <yinghai@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Suresh Siddha <suresh.b.siddha@intel.com> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Rusty Russell <rusty@rustcorp.com.au> LKML-Reference: <49F654AF.8000808@kernel.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86/irq: remove leftover code from NUMA_MIGRATE_IRQ_DESC
The original feature of migrating irq_desc dynamic was too fragile and was causing problems: it caused crashes on systems with lots of cards with MSI-X when user-space irq-balancer was enabled. We now have new patches that create irq_desc according to device numa node. This patch removes the leftover bits of the dynamic balancer. [ Impact: remove dead code ] Signed-off-by: Yinghai Lu <yinghai@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Suresh Siddha <suresh.b.siddha@intel.com> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Rusty Russell <rusty@rustcorp.com.au> LKML-Reference: <49F654AF.8000808@kernel.org> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Yinghai Lu · Ingo Molnar
1 parent 9ec4fa271f
Showing 8 changed files with 9 additions and 93 deletions Side-by-side Diff
arch/x86/Kconfig
arch/x86/configs/x86_64_defconfig
arch/x86/kernel/apic/io_apic.c
include/linux/irq.h
kernel/irq/Makefile
kernel/irq/chip.c
kernel/irq/handle.c
kernel/irq/numa_migrate.c
@@ -274,16 +274,6 @@
  
 	  If you don't know what to do here, say N.
  
-config NUMA_MIGRATE_IRQ_DESC
-	bool "Move irq desc when changing irq smp_affinity"
-	depends on SPARSE_IRQ && NUMA
-	depends on BROKEN
-	default n
-	---help---
-	  This enables moving irq_desc to cpu/node that irq will use handled.
-
-	  If you don't know what to do here, say N.
-
 config X86_MPPARSE
 	bool "Enable MPS table" if ACPI
 	default y
@@ -195,7 +195,6 @@
 CONFIG_GENERIC_CLOCKEVENTS_BUILD=y
 CONFIG_SMP=y
 CONFIG_SPARSE_IRQ=y
-# CONFIG_NUMA_MIGRATE_IRQ_DESC is not set
 CONFIG_X86_FIND_SMP_CONFIG=y
 CONFIG_X86_MPPARSE=y
 # CONFIG_X86_ELAN is not set
@@ -148,9 +148,6 @@
 	unsigned move_cleanup_count;
 	u8 vector;
 	u8 move_in_progress : 1;
-#ifdef CONFIG_NUMA_MIGRATE_IRQ_DESC
-	u8 move_desc_pending : 1;
-#endif
 };
  
 /* irq_cfg is indexed by the sum of all RTEs in all I/O APICs. */
@@ -254,8 +251,7 @@
 	return 0;
 }
  
-#ifdef CONFIG_NUMA_MIGRATE_IRQ_DESC
-
+/* for move_irq_desc */
 static void
 init_copy_irq_2_pin(struct irq_cfg *old_cfg, struct irq_cfg *cfg, int cpu)
 {
  
@@ -356,20 +352,8 @@
 		old_desc->chip_data = NULL;
 	}
 }
+/* end for move_irq_desc */
  
-static void
-set_extra_move_desc(struct irq_desc *desc, const struct cpumask *mask)
-{
-	struct irq_cfg *cfg = desc->chip_data;
-
-	if (!cfg->move_in_progress) {
-		/* it means that domain is not changed */
-		if (!cpumask_intersects(desc->affinity, mask))
-			cfg->move_desc_pending = 1;
-	}
-}
-#endif
-
 #else
 static struct irq_cfg *irq_cfg(unsigned int irq)
 {
@@ -378,13 +362,6 @@
  
 #endif
  
-#ifndef CONFIG_NUMA_MIGRATE_IRQ_DESC
-static inline void
-set_extra_move_desc(struct irq_desc *desc, const struct cpumask *mask)
-{
-}
-#endif
-
 struct io_apic {
 	unsigned int index;
 	unsigned int unused[3];
@@ -592,9 +569,6 @@
 	if (assign_irq_vector(irq, cfg, mask))
 		return BAD_APICID;
  
-	/* check that before desc->addinity get updated */
-	set_extra_move_desc(desc, mask);
-
 	cpumask_copy(desc->affinity, mask);
  
 	return apic->cpu_mask_to_apicid_and(desc->affinity, cfg->domain);
@@ -2393,8 +2367,6 @@
 	if (assign_irq_vector(irq, cfg, mask))
 		return;
  
-	set_extra_move_desc(desc, mask);
-
 	dest = apic->cpu_mask_to_apicid_and(cfg->domain, mask);
  
 	irte.vector = cfg->vector;
  
  
  
@@ -2491,34 +2463,14 @@
 	struct irq_cfg *cfg = desc->chip_data;
 	unsigned vector, me;
  
-	if (likely(!cfg->move_in_progress)) {
-#ifdef CONFIG_NUMA_MIGRATE_IRQ_DESC
-		if (likely(!cfg->move_desc_pending))
-			return;
-
-		/* domain has not changed, but affinity did */
-		me = smp_processor_id();
-		if (cpumask_test_cpu(me, desc->affinity)) {
-			*descp = desc = move_irq_desc(desc, me);
-			/* get the new one */
-			cfg = desc->chip_data;
-			cfg->move_desc_pending = 0;
-		}
-#endif
+	if (likely(!cfg->move_in_progress))
 		return;
-	}
  
 	vector = ~get_irq_regs()->orig_ax;
 	me = smp_processor_id();
  
-	if (vector == cfg->vector && cpumask_test_cpu(me, cfg->domain)) {
-#ifdef CONFIG_NUMA_MIGRATE_IRQ_DESC
-		*descp = desc = move_irq_desc(desc, me);
-		/* get the new one */
-		cfg = desc->chip_data;
-#endif
+	if (vector == cfg->vector && cpumask_test_cpu(me, cfg->domain))
 		send_cleanup_vector(cfg);
-	}
 }
 #else
 static inline void irq_complete_move(struct irq_desc **descp) {}
@@ -212,16 +212,6 @@
  
 extern struct irq_desc *irq_to_desc_alloc_cpu(unsigned int irq, int cpu);
  
-static inline struct irq_desc *
-irq_remap_to_desc(unsigned int irq, struct irq_desc *desc)
-{
-#ifdef CONFIG_NUMA_MIGRATE_IRQ_DESC
-	return irq_to_desc(irq);
-#else
-	return desc;
-#endif
-}
-
 /*
  * Migration helpers for obsolete names, they will go away:
  */
@@ -3,6 +3,6 @@
 obj-$(CONFIG_GENERIC_IRQ_PROBE) += autoprobe.o
 obj-$(CONFIG_PROC_FS) += proc.o
 obj-$(CONFIG_GENERIC_PENDING_IRQ) += migration.o
-obj-$(CONFIG_NUMA_MIGRATE_IRQ_DESC) += numa_migrate.o
+obj-$(CONFIG_SPARSE_IRQ) += numa_migrate.o
 obj-$(CONFIG_PM_SLEEP) += pm.o
@@ -359,7 +359,6 @@
  
 	spin_lock(&desc->lock);
 	mask_ack_irq(desc, irq);
-	desc = irq_remap_to_desc(irq, desc);
  
 	if (unlikely(desc->status & IRQ_INPROGRESS))
 		goto out_unlock;
@@ -438,7 +437,6 @@
 	desc->status &= ~IRQ_INPROGRESS;
 out:
 	desc->chip->eoi(irq);
-	desc = irq_remap_to_desc(irq, desc);
  
 	spin_unlock(&desc->lock);
 }
@@ -475,7 +473,6 @@
 		    !desc->action)) {
 		desc->status |= (IRQ_PENDING | IRQ_MASKED);
 		mask_ack_irq(desc, irq);
-		desc = irq_remap_to_desc(irq, desc);
 		goto out_unlock;
 	}
 	kstat_incr_irqs_this_cpu(irq, desc);
@@ -483,7 +480,6 @@
 	/* Start handling the irq */
 	if (desc->chip->ack)
 		desc->chip->ack(irq);
-	desc = irq_remap_to_desc(irq, desc);
  
 	/* Mark the IRQ currently in progress.*/
 	desc->status |= IRQ_INPROGRESS;
  
@@ -544,10 +540,8 @@
 	if (!noirqdebug)
 		note_interrupt(irq, desc, action_ret);
  
-	if (desc->chip->eoi) {
+	if (desc->chip->eoi)
 		desc->chip->eoi(irq);
-		desc = irq_remap_to_desc(irq, desc);
-	}
 }
  
 void
  
@@ -582,10 +576,8 @@
  
 	/* Uninstall? */
 	if (handle == handle_bad_irq) {
-		if (desc->chip != &no_irq_chip) {
+		if (desc->chip != &no_irq_chip)
 			mask_ack_irq(desc, irq);
-			desc = irq_remap_to_desc(irq, desc);
-		}
 		desc->status |= IRQ_DISABLED;
 		desc->depth = 1;
 	}
@@ -458,11 +458,8 @@
 		/*
 		 * No locking required for CPU-local interrupts:
 		 */
-		if (desc->chip->ack) {
+		if (desc->chip->ack)
 			desc->chip->ack(irq);
-			/* get new one */
-			desc = irq_remap_to_desc(irq, desc);
-		}
 		if (likely(!(desc->status & IRQ_DISABLED))) {
 			action_ret = handle_IRQ_event(irq, desc->action);
 			if (!noirqdebug)
  
@@ -473,10 +470,8 @@
 	}
  
 	spin_lock(&desc->lock);
-	if (desc->chip->ack) {
+	if (desc->chip->ack)
 		desc->chip->ack(irq);
-		desc = irq_remap_to_desc(irq, desc);
-	}
 	/*
 	 * REPLAY is when Linux resends an IRQ that was dropped earlier
 	 * WAITING is used by probe to mark irqs that are being tested
@@ -97,9 +97,7 @@
  
 	/* free the old one */
 	free_one_irq_desc(old_desc, desc);
-	spin_unlock(&old_desc->lock);
 	kfree(old_desc);
-	spin_lock(&desc->lock);
  
 	return desc;
...	...	@@ -274,16 +274,6 @@
274	274
275	275	If you don't know what to do here, say N.
276	276
277		-config NUMA_MIGRATE_IRQ_DESC
278		- bool "Move irq desc when changing irq smp_affinity"
279		- depends on SPARSE_IRQ && NUMA
280		- depends on BROKEN
281		- default n
282		- ---help---
283		- This enables moving irq_desc to cpu/node that irq will use handled.
284		-
285		- If you don't know what to do here, say N.
286		-
287	277	config X86_MPPARSE
288	278	bool "Enable MPS table" if ACPI
289	279	default y
...	...	@@ -195,7 +195,6 @@
195	195	CONFIG_GENERIC_CLOCKEVENTS_BUILD=y
196	196	CONFIG_SMP=y
197	197	CONFIG_SPARSE_IRQ=y
198		-# CONFIG_NUMA_MIGRATE_IRQ_DESC is not set
199	198	CONFIG_X86_FIND_SMP_CONFIG=y
200	199	CONFIG_X86_MPPARSE=y
201	200	# CONFIG_X86_ELAN is not set
...	...	@@ -148,9 +148,6 @@
148	148	unsigned move_cleanup_count;
149	149	u8 vector;
150	150	u8 move_in_progress : 1;
151		-#ifdef CONFIG_NUMA_MIGRATE_IRQ_DESC
152		- u8 move_desc_pending : 1;
153		-#endif
154	151	};
155	152
156	153	/* irq_cfg is indexed by the sum of all RTEs in all I/O APICs. */
...	...	@@ -254,8 +251,7 @@
254	251	return 0;
255	252	}
256	253
257		-#ifdef CONFIG_NUMA_MIGRATE_IRQ_DESC
258		-
	254	+/* for move_irq_desc */
259	255	static void
260	256	init_copy_irq_2_pin(struct irq_cfg old_cfg, struct irq_cfg cfg, int cpu)
261	257	{
262	258
...	...	@@ -356,20 +352,8 @@
356	352	old_desc->chip_data = NULL;
357	353	}
358	354	}
	355	+/* end for move_irq_desc */
359	356
360		-static void
361		-set_extra_move_desc(struct irq_desc desc, const struct cpumask mask)
362		-{
363		- struct irq_cfg *cfg = desc->chip_data;
364		-
365		- if (!cfg->move_in_progress) {
366		- /* it means that domain is not changed */
367		- if (!cpumask_intersects(desc->affinity, mask))
368		- cfg->move_desc_pending = 1;
369		- }
370		-}
371		-#endif
372		-
373	357	#else
374	358	static struct irq_cfg *irq_cfg(unsigned int irq)
375	359	{
...	...	@@ -378,13 +362,6 @@
378	362
379	363	#endif
380	364
381		-#ifndef CONFIG_NUMA_MIGRATE_IRQ_DESC
382		-static inline void
383		-set_extra_move_desc(struct irq_desc desc, const struct cpumask mask)
384		-{
385		-}
386		-#endif
387		-
388	365	struct io_apic {
389	366	unsigned int index;
390	367	unsigned int unused[3];
...	...	@@ -592,9 +569,6 @@
592	569	if (assign_irq_vector(irq, cfg, mask))
593	570	return BAD_APICID;
594	571
595		- /* check that before desc->addinity get updated */
596		- set_extra_move_desc(desc, mask);
597		-
598	572	cpumask_copy(desc->affinity, mask);
599	573
600	574	return apic->cpu_mask_to_apicid_and(desc->affinity, cfg->domain);
...	...	@@ -2393,8 +2367,6 @@
2393	2367	if (assign_irq_vector(irq, cfg, mask))
2394	2368	return;
2395	2369
2396		- set_extra_move_desc(desc, mask);
2397		-
2398	2370	dest = apic->cpu_mask_to_apicid_and(cfg->domain, mask);
2399	2371
2400	2372	irte.vector = cfg->vector;
2401	2373
2402	2374
2403	2375
...	...	@@ -2491,34 +2463,14 @@
2491	2463	struct irq_cfg *cfg = desc->chip_data;
2492	2464	unsigned vector, me;
2493	2465
2494		- if (likely(!cfg->move_in_progress)) {
2495		-#ifdef CONFIG_NUMA_MIGRATE_IRQ_DESC
2496		- if (likely(!cfg->move_desc_pending))
2497		- return;
2498		-
2499		- /* domain has not changed, but affinity did */
2500		- me = smp_processor_id();
2501		- if (cpumask_test_cpu(me, desc->affinity)) {
2502		- *descp = desc = move_irq_desc(desc, me);
2503		- /* get the new one */
2504		- cfg = desc->chip_data;
2505		- cfg->move_desc_pending = 0;
2506		- }
2507		-#endif
	2466	+ if (likely(!cfg->move_in_progress))
2508	2467	return;
2509		- }
2510	2468
2511	2469	vector = ~get_irq_regs()->orig_ax;
2512	2470	me = smp_processor_id();
2513	2471
2514		- if (vector == cfg->vector && cpumask_test_cpu(me, cfg->domain)) {
2515		-#ifdef CONFIG_NUMA_MIGRATE_IRQ_DESC
2516		- *descp = desc = move_irq_desc(desc, me);
2517		- /* get the new one */
2518		- cfg = desc->chip_data;
2519		-#endif
	2472	+ if (vector == cfg->vector && cpumask_test_cpu(me, cfg->domain))
2520	2473	send_cleanup_vector(cfg);
2521		- }
2522	2474	}
2523	2475	#else
2524	2476	static inline void irq_complete_move(struct irq_desc **descp) {}
...	...	@@ -212,16 +212,6 @@
212	212
213	213	extern struct irq_desc *irq_to_desc_alloc_cpu(unsigned int irq, int cpu);
214	214
215		-static inline struct irq_desc *
216		-irq_remap_to_desc(unsigned int irq, struct irq_desc *desc)
217		-{
218		-#ifdef CONFIG_NUMA_MIGRATE_IRQ_DESC
219		- return irq_to_desc(irq);
220		-#else
221		- return desc;
222		-#endif
223		-}
224		-
225	215	/*
226	216	* Migration helpers for obsolete names, they will go away:
227	217	*/
...	...	@@ -3,6 +3,6 @@
3	3	obj-$(CONFIG_GENERIC_IRQ_PROBE) += autoprobe.o
4	4	obj-$(CONFIG_PROC_FS) += proc.o
5	5	obj-$(CONFIG_GENERIC_PENDING_IRQ) += migration.o
6		-obj-$(CONFIG_NUMA_MIGRATE_IRQ_DESC) += numa_migrate.o
	6	+obj-$(CONFIG_SPARSE_IRQ) += numa_migrate.o
7	7	obj-$(CONFIG_PM_SLEEP) += pm.o
...	...	@@ -359,7 +359,6 @@
359	359
360	360	spin_lock(&desc->lock);
361	361	mask_ack_irq(desc, irq);
362		- desc = irq_remap_to_desc(irq, desc);
363	362
364	363	if (unlikely(desc->status & IRQ_INPROGRESS))
365	364	goto out_unlock;
...	...	@@ -438,7 +437,6 @@
438	437	desc->status &= ~IRQ_INPROGRESS;
439	438	out:
440	439	desc->chip->eoi(irq);
441		- desc = irq_remap_to_desc(irq, desc);
442	440
443	441	spin_unlock(&desc->lock);
444	442	}
...	...	@@ -475,7 +473,6 @@
475	473	!desc->action)) {
476	474	desc->status \|= (IRQ_PENDING \| IRQ_MASKED);
477	475	mask_ack_irq(desc, irq);
478		- desc = irq_remap_to_desc(irq, desc);
479	476	goto out_unlock;
480	477	}
481	478	kstat_incr_irqs_this_cpu(irq, desc);
...	...	@@ -483,7 +480,6 @@
483	480	/* Start handling the irq */
484	481	if (desc->chip->ack)
485	482	desc->chip->ack(irq);
486		- desc = irq_remap_to_desc(irq, desc);
487	483
488	484	/* Mark the IRQ currently in progress.*/
489	485	desc->status \|= IRQ_INPROGRESS;
490	486
...	...	@@ -544,10 +540,8 @@
544	540	if (!noirqdebug)
545	541	note_interrupt(irq, desc, action_ret);
546	542
547		- if (desc->chip->eoi) {
	543	+ if (desc->chip->eoi)
548	544	desc->chip->eoi(irq);
549		- desc = irq_remap_to_desc(irq, desc);
550		- }
551	545	}
552	546
553	547	void
554	548
...	...	@@ -582,10 +576,8 @@
582	576
583	577	/* Uninstall? */
584	578	if (handle == handle_bad_irq) {
585		- if (desc->chip != &no_irq_chip) {
	579	+ if (desc->chip != &no_irq_chip)
586	580	mask_ack_irq(desc, irq);
587		- desc = irq_remap_to_desc(irq, desc);
588		- }
589	581	desc->status \|= IRQ_DISABLED;
590	582	desc->depth = 1;
591	583	}
...	...	@@ -458,11 +458,8 @@
458	458	/*
459	459	* No locking required for CPU-local interrupts:
460	460	*/
461		- if (desc->chip->ack) {
	461	+ if (desc->chip->ack)
462	462	desc->chip->ack(irq);
463		- /* get new one */
464		- desc = irq_remap_to_desc(irq, desc);
465		- }
466	463	if (likely(!(desc->status & IRQ_DISABLED))) {
467	464	action_ret = handle_IRQ_event(irq, desc->action);
468	465	if (!noirqdebug)
469	466
...	...	@@ -473,10 +470,8 @@
473	470	}
474	471
475	472	spin_lock(&desc->lock);
476		- if (desc->chip->ack) {
	473	+ if (desc->chip->ack)
477	474	desc->chip->ack(irq);
478		- desc = irq_remap_to_desc(irq, desc);
479		- }
480	475	/*
481	476	* REPLAY is when Linux resends an IRQ that was dropped earlier
482	477	* WAITING is used by probe to mark irqs that are being tested
...	...	@@ -97,9 +97,7 @@
97	97
98	98	/* free the old one */
99	99	free_one_irq_desc(old_desc, desc);
100		- spin_unlock(&old_desc->lock);
101	100	kfree(old_desc);
102		- spin_lock(&desc->lock);
103	101
104	102	return desc;
105	103