cpumask: more cpumask updates
Subsystems (12)
cover cpumask: more cpumask updates
One big change to irq_desc and a few smaller maintenance type updates.
cpumask: update irq_desc to use cpumask_var_t
cpumask: Use topology_core_cpumask()/topology_thread_cpumask()
cpumask: convert misc driver functions
cpumask: convert drivers/net/sfc
cpumask: convert other misc kernel functions
Based on tip/cpus4096: v2.6.28-5703-g1d1a70e
(Ingo - I added a blank line before the From: part and that seems to separate it from the Subject. Hopefully your git-am will still recognizes it - I did try a git-am here and it seemed to work ok.)
PATCH 1/5 cpumask: update irq_desc to use cpumask_var_t
Impact: reduce memory usage, use new cpumask API.
Replace the affinity and pending_masks with cpumask_var_t's. This adds to the significant size reduction done with the SPARSE_IRQS changes.
The added functions (init_alloc_desc_masks & init_copy_desc_masks) are in the include file so they can be inlined (and optimized out for the !CONFIG_CPUMASKS_OFFSTACK case.) [Naming chosen to be consistent with the other init*irq functions, as well as the backwards arg declaration of "from, to" instead of the more common "to, from" standard.]
Includes a slight change to the declaration of struct irq_desc to embed the pending_mask within ifdef(CONFIG_SMP) to be consistent with other references, and some small changes to Xen.
Tested: sparse/non-sparse/cpumask_offstack/non-cpumask_offstack/nonuma/nosmp on x86_64
--- arch/x86/kernel/io_apic.c | 20 +++++------ arch/x86/kernel/irq_32.c | 2 - arch/x86/kernel/irq_64.c | 2 - drivers/xen/events.c | 4 +- include/linux/irq.h | 81 ++++++++++++++++++++++++++++++++++++++++++++-- kernel/irq/chip.c | 5 ++ kernel/irq/handle.c | 26 +++++++------- kernel/irq/manage.c | 12 +++--- kernel/irq/migration.c | 12 +++--- kernel/irq/numa_migrate.c | 12 ++++++ kernel/irq/proc.c | 4 +- 11 files changed, 135 insertions(+), 45 deletions(-)
arch/x86/kernel/io_apic.c +10 -10
@@ -358,7 +358,7 @@ set_extra_move_desc(struct irq_desc *desif (!cfg->move_in_progress) {/* it means that domain is not changed */ -if (!cpumask_intersects (&desc->affinity, mask)) +if (!cpumask_intersects (desc->affinity, mask)) cfg->move_desc_pending =1 ; } }
@@ -581,9 +581,9 @@ set_desc_affinity(struct irq_desc *desc,if (assign_irq_vector (irq, cfg, mask))return BAD_APICID ; -cpumask_and (&desc->affinity, cfg->domain, mask); +cpumask_and (desc->affinity, cfg->domain, mask);set_extra_move_desc (desc, mask); -return cpu_mask_to_apicid_and (&desc->affinity, cpu_online_mask); +return cpu_mask_to_apicid_and (desc->affinity, cpu_online_mask); } static void
@@ -2385,7 +2385,7 @@ migrate_ioapic_irq_desc(struct irq_descif (cfg->move_in_progress)send_cleanup_vector (cfg); -cpumask_copy (&desc->affinity, mask); +cpumask_copy (desc->affinity, mask); }static int migrate_irq_remapped_level_desc (struct irq_desc *desc)
@@ -2407,11 +2407,11 @@ static int migrate_irq_remapped_level_de }/* everthing is clear. we have right of way */ -migrate_ioapic_irq_desc (desc, &desc->pending_mask); +migrate_ioapic_irq_desc (desc, desc->pending_mask); ret =0 ; desc->status &= ~IRQ_MOVE_PENDING ; -cpumask_clear (&desc->pending_mask); +cpumask_clear (desc->pending_mask); unmask:unmask_IO_APIC_irq_desc (desc);
@@ -2436,7 +2436,7 @@ static void ir_irq_migration(struct workcontinue ; } - desc->chip->set_affinity (irq, &desc->pending_mask); + desc->chip->set_affinity (irq, desc->pending_mask);spin_unlock_irqrestore (&desc->lock, flags); } }
@@ -2450,7 +2450,7 @@ static void set_ir_ioapic_affinity_irq_d {if (desc->status &IRQ_LEVEL ) { desc->status |=IRQ_MOVE_PENDING ; -cpumask_copy (&desc->pending_mask, mask); +cpumask_copy (desc->pending_mask, mask);migrate_irq_remapped_level_desc (desc);return ; }
@@ -2518,7 +2518,7 @@ static void irq_complete_move(struct irq/* domain has not changed, but affinity did */ me =smp_processor_id (); -if (cpu_isset (me, desc->affinity)) { +if (cpumask_test_cpu (me, desc->affinity)) { *descp = desc =move_irq_desc (desc, me);/* get the new one */ cfg = desc->chip_data;
@@ -4041,7 +4041,7 @@ void __init setup_ioapic_dest(void)*/ if (desc->status & (IRQ_NO_BALANCING |IRQ_AFFINITY_SET )) - mask = &desc->affinity; + mask = desc->affinity;else mask =TARGET_CPUS ;
arch/x86/kernel/irq_32.c +1 -1
@@ -248,7 +248,7 @@ void fixup_irqs(void)if (irq ==2 )continue ; - affinity = &desc->affinity; + affinity = desc->affinity;if (cpumask_any_and (affinity, cpu_online_mask) >= nr_cpu_ids) {printk ("Breaking affinity for irq %i\n" , irq); affinity = cpu_all_mask;
arch/x86/kernel/irq_64.c +1 -1
@@ -100,7 +100,7 @@ void fixup_irqs(void)/* interrupt's are disabled at this point */ spin_lock (&desc->lock); - affinity = &desc->affinity; + affinity = desc->affinity;if (!irq_has_action (irq) ||cpumask_equal (affinity, cpu_online_mask)) {spin_unlock (&desc->lock);
drivers/xen/events.c +2 -2
@@ -125,7 +125,7 @@ static void bind_evtchn_to_cpu(unsignedBUG_ON (irq ==-1 );#ifdef CONFIG_SMP -irq_to_desc (irq)->affinity =cpumask_of_cpu (cpu); +cpumask_copy (irq_to_desc (irq)->affinity,cpumask_of (cpu));#endif __clear_bit (chn, cpu_evtchn_mask[cpu_evtchn[chn]]);
@@ -142,7 +142,7 @@ static void init_evtchn_cpu_bindings(voi/* By default all event channels notify CPU#0. */ for_each_irq_desc (i, desc) { - desc->affinity =cpumask_of_cpu (0 ); +cpumask_copy (desc->affinity,cpumask_of (0 )); }#endif
include/linux/irq.h +78 -3
93 lines not on this page · raw message
kernel/irq/chip.c +4 -1
@@ -46,7 +46,10 @@ void dynamic_irq_init(unsigned int irq) desc->irq_count =0 ; desc->irqs_unhandled =0 ;#ifdef CONFIG_SMP -cpumask_setall (&desc->affinity); +cpumask_setall (desc->affinity); +#ifdef CONFIG_GENERIC_PENDING_IRQ +cpumask_clear (desc->pending_mask); +#endif #endif spin_unlock_irqrestore (&desc->lock, flags); }
kernel/irq/handle.c +14 -12
@@ -64,9 +64,6 @@ static struct irq_desc irq_desc_init = { .handle_irq = handle_bad_irq, .depth = 1, .lock = __SPIN_LOCK_UNLOCKED(irq_desc_init.lock), -#ifdef CONFIG_SMP - .affinity =CPU_MASK_ALL -#endif };void init_kstat_irqs (struct irq_desc *desc,int cpu,int nr)
@@ -88,6 +85,8 @@ void init_kstat_irqs(struct irq_desc *destatic void init_one_irq_desc (int irq,struct irq_desc *desc,int cpu) { +int node =cpu_to_node (cpu); +memcpy (desc, &irq_desc_init,sizeof (struct irq_desc ));spin_lock_init (&desc->lock);
@@ -101,6 +100,10 @@ static void init_one_irq_desc(int irq, sprintk (KERN_ERR "can not alloc kstat_irqs\n" );BUG_ON (1 ); } +if (!init_alloc_desc_masks (desc, node, false)) { +printk (KERN_ERR "can not alloc irq_desc cpumasks\n" ); +BUG_ON (1 ); + }arch_init_chip_data (desc, cpu); }
@@ -119,9 +122,6 @@ static struct irq_desc irq_desc_legacy[N .handle_irq = handle_bad_irq, .depth = 1, .lock = __SPIN_LOCK_UNLOCKED(irq_desc_init.lock), -#ifdef CONFIG_SMP - .affinity = CPU_MASK_ALL -#endif } };
@@ -141,7 +141,7 @@ int __init early_irq_init(void) desc[i].irq = i; desc[i].kstat_irqs = kstat_irqs_legacy[i];lockdep_set_class (&desc[i].lock, &irq_desc_lock_class); - +init_alloc_desc_masks (&desc[i],0 , true); irq_desc_ptrs[i] = desc + i; }
@@ -188,6 +188,10 @@ struct irq_desc *irq_to_desc_alloc_cpu(uprintk (KERN_ERR "can not alloc irq_desc\n" );BUG_ON (1 ); } +if (!init_alloc_desc_masks (desc, node, false)) { +printk (KERN_ERR "can not alloc irq_desc cpumasks\n" ); +BUG_ON (1 ); + }init_one_irq_desc (irq, desc, cpu); irq_desc_ptrs[irq] = desc;
@@ -207,9 +211,6 @@ struct irq_desc irq_desc[NR_IRQS] __cach .handle_irq = handle_bad_irq, .depth = 1, .lock = __SPIN_LOCK_UNLOCKED(irq_desc->lock), -#ifdef CONFIG_SMP - .affinity = CPU_MASK_ALL -#endif } };
@@ -222,9 +223,10 @@ int __init early_irq_init(void) desc = irq_desc; count =ARRAY_SIZE (irq_desc); -for (i =0 ; i < count; i++) +for (i =0 ; i < count; i++) { desc[i].irq = i; - +init_alloc_desc_masks (&desc[i],0 , true); + }return arch_early_irq_init (); }
kernel/irq/manage.c +6 -6
@@ -98,14 +98,14 @@ int irq_set_affinity(unsigned int irq, c#ifdef CONFIG_GENERIC_PENDING_IRQ if (desc->status &IRQ_MOVE_PCNTXT || desc->status &IRQ_DISABLED ) { -cpumask_copy (&desc->affinity, cpumask); +cpumask_copy (desc->affinity, cpumask); desc->chip->set_affinity (irq, cpumask); }else { desc->status |=IRQ_MOVE_PENDING ; -cpumask_copy (&desc->pending_mask, cpumask); +cpumask_copy (desc->pending_mask, cpumask); }#else -cpumask_copy (&desc->affinity, cpumask); +cpumask_copy (desc->affinity, cpumask); desc->chip->set_affinity (irq, cpumask);#endif desc->status |=IRQ_AFFINITY_SET ;
@@ -127,16 +127,16 @@ int do_irq_select_affinity(unsigned int* one of the targets is online. */ if (desc->status & (IRQ_AFFINITY_SET |IRQ_NO_BALANCING )) { -if (cpumask_any_and (&desc->affinity, cpu_online_mask) +if (cpumask_any_and (desc->affinity, cpu_online_mask) < nr_cpu_ids) goto set_affinity;else desc->status &= ~IRQ_AFFINITY_SET ; } -cpumask_and (&desc->affinity, cpu_online_mask, irq_default_affinity); +cpumask_and (desc->affinity, cpu_online_mask, irq_default_affinity); set_affinity: - desc->chip->set_affinity (irq, &desc->affinity); + desc->chip->set_affinity (irq, desc->affinity);return 0 ; }
kernel/irq/migration.c +6 -6
@@ -18,7 +18,7 @@ void move_masked_irq(int irq) desc->status &= ~IRQ_MOVE_PENDING ; -if (unlikely (cpumask_empty (&desc->pending_mask))) +if (unlikely (cpumask_empty (desc->pending_mask)))return ;if (!desc->chip->set_affinity)
@@ -38,13 +38,13 @@ void move_masked_irq(int irq)* For correct operation this depends on the caller * masking the irqs. */ -if (likely (cpumask_any_and (&desc->pending_mask, cpu_online_mask) +if (likely (cpumask_any_and (desc->pending_mask, cpu_online_mask) < nr_cpu_ids)) { -cpumask_and (&desc->affinity, - &desc->pending_mask, cpu_online_mask); - desc->chip->set_affinity (irq, &desc->affinity); +cpumask_and (desc->affinity, + desc->pending_mask, cpu_online_mask); + desc->chip->set_affinity (irq, desc->affinity); } -cpumask_clear (&desc->pending_mask); +cpumask_clear (desc->pending_mask); }void move_native_irq (int irq)
kernel/irq/numa_migrate.c +11 -1
@@ -46,6 +46,7 @@ static void init_copy_one_irq_desc(int i desc->cpu = cpu;lockdep_set_class (&desc->lock, &irq_desc_lock_class);init_copy_kstat_irqs (old_desc, desc, cpu, nr_cpu_ids); +init_copy_desc_masks (old_desc, desc);arch_init_copy_chip_data (old_desc, desc, cpu); }
@@ -76,11 +77,20 @@ static struct irq_desc *__real_move_irq_ node =cpu_to_node (cpu); desc =kzalloc_node (sizeof (*desc),GFP_ATOMIC , node);if (!desc) { -printk (KERN_ERR "irq %d: can not get new irq_desc for migration.\n" , irq); +printk (KERN_ERR "irq %d: can not get new irq_desc " +"for migration.\n" , irq);/* still use old one */ desc = old_desc; goto out_unlock; } +if (!init_alloc_desc_masks (desc, node, false)) { +printk (KERN_ERR "irq %d: can not get new irq_desc cpumask " +"for migration.\n" , irq); +/* still use old one */ +kfree (desc); + desc = old_desc; + goto out_unlock; + }init_copy_one_irq_desc (irq, old_desc, desc, cpu); irq_desc_ptrs[irq] = desc;
kernel/irq/proc.c +2 -2
@@ -20,11 +20,11 @@ static struct proc_dir_entry *root_irq_dstatic int irq_affinity_proc_show (struct seq_file *m,void *v) {struct irq_desc *desc =irq_to_desc ((long )m->private); -const struct cpumask *mask = &desc->affinity; +const struct cpumask *mask = desc->affinity;#ifdef CONFIG_GENERIC_PENDING_IRQ if (desc->status &IRQ_MOVE_PENDING ) - mask = &desc->pending_mask; + mask = desc->pending_mask;#endif seq_cpumask (m, mask);seq_putc (m,'\n' );
Re: [PATCH 1/5] cpumask: update irq_desc to use cpumask_var_t
On Wed, Jan 7, 2009 at 11:58 AM, Mike Travis <[email protected]> wrote:
| --- linux-2.6-for-ingo.orig/kernel/irq/numa_migrate.c
| +++ linux-2.6-for-ingo/kernel/irq/numa_migrate.c
| @@ -46,6 +46,7 @@ static void init_copy_one_irq_desc(int i
| desc->cpu = cpu;
| lockdep_set_class(&desc->lock, &irq_desc_lock_class);
| init_copy_kstat_irqs(old_desc, desc, cpu, nr_cpu_ids);
|+ init_copy_desc_masks(old_desc, desc);
| arch_init_copy_chip_data(old_desc, desc, cpu);
| }
|
|@@ -76,11 +77,20 @@ static struct irq_desc *__real_move_irq_
| node = cpu_to_node(cpu);
| desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node);
| if (!desc) {
|- printk(KERN_ERR "irq %d: can not get new irq_desc for migration.\n", irq);
|+ printk(KERN_ERR "irq %d: can not get new irq_desc "
|+ "for migration.\n", irq);
| /* still use old one */
| desc = old_desc;
| goto out_unlock;
| }
|+ if (!init_alloc_desc_masks(desc, node, false)) {
|+ printk(KERN_ERR "irq %d: can not get new irq_desc cpumask "
|+ "for migration.\n", irq);
|+ /* still use old one */
|+ kfree(desc);
|+ desc = old_desc;
|+ goto out_unlock;
|+ }
| init_copy_one_irq_desc(irq, old_desc, desc, cpu);desc new mask_var (allocated) aka the pointer is overwritten here... you may need to calling move init_alloc_desc_masks() into init_copy_one_irq_desc()
| irq_desc_ptrs[irq] = desc;
YH
Re: [PATCH 1/5] cpumask: update irq_desc to use cpumask_var_t
Quoted 32 lines · Yinghai Lu
On Wed, Jan 7, 2009 at 11:58 AM, Mike Travis <[email protected]> wrote:
| --- linux-2.6-for-ingo.orig/kernel/irq/numa_migrate.c | +++ linux-2.6-for-ingo/kernel/irq/numa_migrate.c | @@ -46,6 +46,7 @@ static void init_copy_one_irq_desc(int i | desc->cpu = cpu; | lockdep_set_class(&desc->lock, &irq_desc_lock_class); | init_copy_kstat_irqs(old_desc, desc, cpu, nr_cpu_ids); |+ init_copy_desc_masks(old_desc, desc); | arch_init_copy_chip_data(old_desc, desc, cpu); | } | |@@ -76,11 +77,20 @@ static struct irq_desc *__real_move_irq_ | node = cpu_to_node(cpu); | desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node); | if (!desc) { |- printk(KERN_ERR "irq %d: can not get new irq_desc for migration.\n", irq); |+ printk(KERN_ERR "irq %d: can not get new irq_desc " |+ "for migration.\n", irq); | /* still use old one */ | desc = old_desc; | goto out_unlock; | } |+ if (!init_alloc_desc_masks(desc, node, false)) { |+ printk(KERN_ERR "irq %d: can not get new irq_desc cpumask " |+ "for migration.\n", irq); |+ /* still use old one */ |+ kfree(desc); |+ desc = old_desc; |+ goto out_unlock; |+ } | init_copy_one_irq_desc(irq, old_desc, desc, cpu);
Quoted 3 lines
desc new
mask_var(allocated) aka the pointer is overwritten here... you may need to calling moveinit_alloc_desc_masks()intoinit_copy_one_irq_desc()
Wouldn't this in init_copy_one_irq_desc() take care of that?
@@ -46,6 +46,7 @@ static void init_copy_one_irq_desc(int i
desc->cpu = cpu;
lockdep_set_class(&desc->lock, &irq_desc_lock_class);
init_copy_kstat_irqs(old_desc, desc, cpu, nr_cpu_ids);
+ init_copy_desc_masks(old_desc, desc);
arch_init_copy_chip_data(old_desc, desc, cpu);where:
static inline void init_copy_desc_masks(struct irq_desc *old_desc,
struct irq_desc *new_desc)
{
#ifdef CONFIG_CPUMASKS_OFFSTACK
cpumask_copy(new_desc->affinity, old_desc->affinity);#ifdef CONFIG_GENERIC_PENDING_IRQ
cpumask_copy(new_desc->pending_mask, old_desc->pending_mask);#endif
#endif
}
In other words if the masks are not a cpumask[1] but instead a *cpumask pointer, then the old masks are copied to the new desc/
Or am I missing your point?
Thanks!
Mike
Quoted 9 lines
| irq_desc_ptrs[irq] = desc;
YH
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 1/5] cpumask: update irq_desc to use cpumask_var_t
Quoted 56 lines · Mike Travis
Quoted 32 lines · Yinghai Lu
On Wed, Jan 7, 2009 at 11:58 AM, Mike Travis <[email protected]> wrote:
| --- linux-2.6-for-ingo.orig/kernel/irq/numa_migrate.c | +++ linux-2.6-for-ingo/kernel/irq/numa_migrate.c | @@ -46,6 +46,7 @@ static void init_copy_one_irq_desc(int i | desc->cpu = cpu; | lockdep_set_class(&desc->lock, &irq_desc_lock_class); | init_copy_kstat_irqs(old_desc, desc, cpu, nr_cpu_ids); |+ init_copy_desc_masks(old_desc, desc); | arch_init_copy_chip_data(old_desc, desc, cpu); | } | |@@ -76,11 +77,20 @@ static struct irq_desc *__real_move_irq_ | node = cpu_to_node(cpu); | desc = kzalloc_node(sizeof(*desc), GFP_ATOMIC, node); | if (!desc) { |- printk(KERN_ERR "irq %d: can not get new irq_desc for migration.\n", irq); |+ printk(KERN_ERR "irq %d: can not get new irq_desc " |+ "for migration.\n", irq); | /* still use old one */ | desc = old_desc; | goto out_unlock; | } |+ if (!init_alloc_desc_masks(desc, node, false)) { |+ printk(KERN_ERR "irq %d: can not get new irq_desc cpumask " |+ "for migration.\n", irq); |+ /* still use old one */ |+ kfree(desc); |+ desc = old_desc; |+ goto out_unlock; |+ } | init_copy_one_irq_desc(irq, old_desc, desc, cpu);Quoted 3 lines
desc new
mask_var(allocated) aka the pointer is overwritten here... you may need to calling moveinit_alloc_desc_masks()intoinit_copy_one_irq_desc()Wouldn't this in
init_copy_one_irq_desc()take care of that?kernel/irq/numa_migrate.c +1
-0@@ -46,6 +46,7 @@ static void init_copy_one_irq_desc(int i desc->cpu = cpu;lockdep_set_class (&desc->lock, &irq_desc_lock_class);init_copy_kstat_irqs (old_desc, desc, cpu, nr_cpu_ids); +init_copy_desc_masks (old_desc, desc);arch_init_copy_chip_data (old_desc, desc, cpu);where:
static inline void init_copy_desc_masks(struct
irq_desc*old_desc,
structirq_desc*new_desc)
{#ifdef CONFIG_CPUMASKS_OFFSTACK cpumask_copy(new_desc->affinity, old_desc->affinity);#ifdef CONFIG_GENERIC_PENDING_IRQ cpumask_copy(new_desc->pending_mask, old_desc->pending_mask);
Quoted 8 lines
#endif
#endif
}In other words if the masks are not a cpumask[1] but instead a *cpumask pointer, then the old masks are copied to the new desc/
Or am I missing your point?
static void init_copy_one_irq_desc(int irq, struct irq_desc *old_desc,
struct irq_desc *desc, int cpu)
{
memcpy(desc, old_desc, sizeof(struct irq_desc));will overwrite new_desc->affinity and pending_mask
YH
Re: [PATCH 1/5] cpumask: update irq_desc to use cpumask_var_t
Yinghai Lu wrote:
...
Quoted 11 lines
Quoted 1 line
Or am I missing your point?
static void init_copy_one_irq_desc(int irq, struct
irq_desc*old_desc,
structirq_desc*desc, int cpu){ memcpy(desc, old_desc, sizeof(struct irq_desc));will overwrite
new_desc->affinityandpending_maskYH
Yup, apparently I was missing your point (banging on head once more).
Thanks for spotting this!
Ingo - what's the state of tip/cpus4096? Shall I push this as an append patch, or redo the original so bisectability works.
Also, Yinghai - would you know of a straight forward way to test
the irq migration? I have taken cpus offline and back online
but wasn't sure if any irq's were being moved off of cpu 0 which cannot be taken offline.
Thanks,
Mike
Re: [PATCH 1/5] cpumask: update irq_desc to use cpumask_var_t
Quoted 17 lines · Mike Travis
Yinghai Lu wrote:
...Quoted 10 lines
Quoted 1 line
Or am I missing your point?
static void init_copy_one_irq_desc(int irq, struct
irq_desc*old_desc,
structirq_desc*desc, int cpu){ memcpy(desc, old_desc, sizeof(struct irq_desc));will overwrite
new_desc->affinityandpending_maskYH
Yup, apparently I was missing your point (banging on head once more).
Thanks for spotting this!
Quoted 7 lines
Ingo - what's the state of tip/cpus4096? Shall I push this as an append patch, or redo the original so bisectability works.
Also, Yinghai - would you know of a straight forward way to test
the irq migration? I have taken cpus offline and back online
but wasn't sure if any irq's were being moved off of cpu 0 which cannot be taken offline.
use echo "mask" > /proc/irq/xxx/smp_affinity to change affinity
and check /proc/interrupts and in /sys/log/messages i will said allocate new irq desc on some other node/cpu.
YH
Re: [PATCH 1/5] cpumask: update irq_desc to use cpumask_var_t: fix
In reply to Yinghai LuYinghai Lu wrote:
...
Quoted 9 lines
static void init_copy_one_irq_desc(int irq, struct
irq_desc*old_desc,
structirq_desc*desc, int cpu){ memcpy(desc, old_desc, sizeof(struct irq_desc));will overwrite
new_desc->affinityandpending_maskYH
I've appended this to the patchset:
Subject: cpumask: fix bug in use cpumask_var_t in irq_desc
Impact: fix bug where new irq_desc uses old cpumask pointers which are freed.
As Yinghai pointed out, init_copy_one_irq_desc() copies the old desc to the new desc overwriting the cpumask pointers. Since the old_desc and the cpumask pointers are freed, then memory corruption will occur if these old pointers are used.
Move the allocation of these pointers to after the copy.
--- include/linux/irq.h | 9 +++++++-- kernel/irq/handle.c | 8 +------- kernel/irq/numa_migrate.c | 13 ++++++++----- 3 files changed, 16 insertions(+), 14 deletions(-)
include/linux/irq.h +7 -2
@@ -426,15 +426,18 @@ extern int set_irq_msi(unsigned int irq,/** * init_alloc_desc_masks - allocate cpumasks for irq_desc * @desc: pointer to irq_desc struct +* @cpu: cpu which will be handling the cpumasks * @boot: true if need bootmem * * Allocates affinity and pending_mask cpumask if required. * Returns true if successful (or not required). * Side effect: affinity has all bits set, pending_mask has all bits clear. */ -static inline bool init_alloc_desc_masks (struct irq_desc *desc,int node, +static inline bool init_alloc_desc_masks (struct irq_desc *desc,int cpu,bool boot) { +int node; +if (boot) {alloc_bootmem_cpumask_var (&desc->affinity);cpumask_setall (desc->affinity);
@@ -446,6 +449,8 @@ static inline bool init_alloc_desc_masksreturn true; } + node =cpu_to_node (cpu); +if (!alloc_cpumask_var_node (&desc->affinity,GFP_ATOMIC , node))return false;cpumask_setall (desc->affinity);
@@ -484,7 +489,7 @@ static inline void init_copy_desc_masks(#else /* !CONFIG_SMP */ -static inline bool init_alloc_desc_masks (struct irq_desc *desc,int node, +static inline bool init_alloc_desc_masks (struct irq_desc *desc,int cpu,bool boot) {return true;
kernel/irq/handle.c +1 -7
@@ -85,8 +85,6 @@ void init_kstat_irqs(struct irq_desc *destatic void init_one_irq_desc (int irq,struct irq_desc *desc,int cpu) { -int node =cpu_to_node (cpu); -memcpy (desc, &irq_desc_init,sizeof (struct irq_desc ));spin_lock_init (&desc->lock);
@@ -100,7 +98,7 @@ static void init_one_irq_desc(int irq, sprintk (KERN_ERR "can not alloc kstat_irqs\n" );BUG_ON (1 ); } -if (!init_alloc_desc_masks (desc, node, false)) { +if (!init_alloc_desc_masks (desc, cpu, false)) {printk (KERN_ERR "can not alloc irq_desc cpumasks\n" );BUG_ON (1 ); }
@@ -188,10 +186,6 @@ struct irq_desc *irq_to_desc_alloc_cpu(uprintk (KERN_ERR "can not alloc irq_desc\n" );BUG_ON (1 ); } -if (!init_alloc_desc_masks (desc, node, false)) { -printk (KERN_ERR "can not alloc irq_desc cpumasks\n" ); -BUG_ON (1 ); - }init_one_irq_desc (irq, desc, cpu); irq_desc_ptrs[irq] = desc;
kernel/irq/numa_migrate.c +8 -5
@@ -38,16 +38,22 @@ static void free_kstat_irqs(struct irq_d old_desc->kstat_irqs =NULL ; } -static void init_copy_one_irq_desc (int irq,struct irq_desc *old_desc, +static bool init_copy_one_irq_desc (int irq,struct irq_desc *old_desc,struct irq_desc *desc,int cpu) {memcpy (desc, old_desc,sizeof (struct irq_desc )); +if (!init_alloc_desc_masks (desc, cpu, false)) { +printk (KERN_ERR "irq %d: can not get new irq_desc cpumask " +"for migration.\n" , irq); +return false; + }spin_lock_init (&desc->lock); desc->cpu = cpu;lockdep_set_class (&desc->lock, &irq_desc_lock_class);init_copy_kstat_irqs (old_desc, desc, cpu, nr_cpu_ids);init_copy_desc_masks (old_desc, desc);arch_init_copy_chip_data (old_desc, desc, cpu); +return true; }static void free_one_irq_desc (struct irq_desc *old_desc,struct irq_desc *desc)
@@ -83,15 +89,12 @@ static struct irq_desc *__real_move_irq_ desc = old_desc; goto out_unlock; } -if (!init_alloc_desc_masks (desc, node, false)) { -printk (KERN_ERR "irq %d: can not get new irq_desc cpumask " -"for migration.\n" , irq); +if (!init_copy_one_irq_desc (irq, old_desc, desc, cpu)) {/* still use old one */ kfree (desc); desc = old_desc; goto out_unlock; } -init_copy_one_irq_desc (irq, old_desc, desc, cpu); irq_desc_ptrs[irq] = desc;
PATCH 2/5 cpumask: Use topology_core_cpumask()/topology_thread_cpumask()
In reply to Mike TravisFrom: Rusty Russell <[email protected]>
Impact: reduce stack usage, use new cpumask API.
This actually uses topology_core_cpumask() and topology_thread_cpumask(), removing the only users of topology_core_siblings() and topology_thread_siblings()
--- Documentation/cputopology.txt | 6 +++--- drivers/base/topology.c | 35 +++++++++++++++++------------------ drivers/net/sfc/efx.c | 4 ++-- include/linux/topology.h | 6 ++++++ 4 files changed, 28 insertions(+), 23 deletions(-)
Documentation/cputopology.txt +3 -3
@@ -18,11 +18,11 @@ For an architecture to support this feat these macros in include/asm-XXX/topology.h: #define topology_physical_package_id(cpu) #define topology_core_id(cpu) -#define topology_thread_siblings(cpu) -#define topology_core_siblings(cpu) +#define topology_thread_cpumask(cpu) +#define topology_core_cpumask(cpu) The type of **_id is int. -The type of siblings is cpumask_t. +The type of siblings is (const) struct cpumask *. To be consistent on all architectures, include/linux/topology.h provides default definitions for any of the above macros that are
drivers/base/topology.c +17 -18
@@ -30,7 +30,10 @@#include <linux/module.h> #include <linux/topology.h> -#define define_one_ro (_name) \ +#define define_one_ro_named (_name, _func) \ +static SYSDEV_ATTR(_name, 0444, _func, NULL) + +#define define_one_ro (_name) \ static SYSDEV_ATTR(_name, 0444, show_##_name, NULL)#define define_id_show_func (name) \
@@ -41,8 +44,8 @@ static ssize_t show_##name(struct sys_de return sprintf(buf, "%d\n", topology_##name(cpu)); \ } -#if defined(topology_thread_siblings) || defined(topology_core_siblings) -static ssize_t show_cpumap (int type,cpumask_t *mask,char *buf) +#if defined(topology_thread_cpumask) || defined(topology_core_cpumask) +static ssize_t show_cpumap (int type,const struct cpumask *mask,char *buf) {ptrdiff_t len =PTR_ALIGN (buf +PAGE_SIZE -1 ,PAGE_SIZE ) - buf;int n =0 ;
@@ -64,7 +67,7 @@ static ssize_t show_##name(struct sys_de struct sysdev_attribute *attr, char *buf) \ { \unsigned int cpu = dev->id; \ - return show_cpumap(0, &(topology_##name(cpu)), buf); \ + return show_cpumap(0, topology_##name(cpu), buf); \ }#define define_siblings_show_list (name) \
@@ -73,7 +76,7 @@ static ssize_t show_##name##_list(struct char *buf) \ { \unsigned int cpu = dev->id; \ - return show_cpumap(1, &(topology_##name(cpu)), buf); \ + return show_cpumap(1, topology_##name(cpu), buf); \ }#else
@@ -81,9 +84,7 @@ static ssize_t show_##name##_list(struct static ssize_t show_##name(struct sys_device *dev, \ struct sysdev_attribute *attr, char *buf) \ { \ - unsigned int cpu = dev->id; \ - cpumask_t mask = topology_##name(cpu); \ - return show_cpumap(0, &mask, buf); \ + return show_cpumap(0, topology_##name(dev->id), buf); \ }#define define_siblings_show_list (name) \
@@ -91,9 +92,7 @@ static ssize_t show_##name##_list(struct struct sysdev_attribute *attr, \ char *buf) \ { \ - unsigned int cpu = dev->id; \ - cpumask_t mask = topology_##name(cpu); \ - return show_cpumap(1, &mask, buf); \ + return show_cpumap(1, topology_##name(dev->id), buf); \ }#endif
@@ -106,13 +105,13 @@ define_one_ro(physical_package_id);define_id_show_func (core_id);define_one_ro (core_id); -define_siblings_show_func (thread_siblings); -define_one_ro (thread_siblings); -define_one_ro (thread_siblings_list); - -define_siblings_show_func (core_siblings); -define_one_ro (core_siblings); -define_one_ro (core_siblings_list); +define_siblings_show_func (thread_cpumask); +define_one_ro_named (thread_siblings, show_thread_cpumask); +define_one_ro_named (thread_siblings_list, show_thread_cpumask_list); + +define_siblings_show_func (core_cpumask); +define_one_ro_named (core_siblings, show_core_cpumask); +define_one_ro_named (core_siblings_list, show_core_cpumask_list);static struct attribute *default_attrs[] = { &attr_physical_package_id.attr,
drivers/net/sfc/efx.c +2 -2
@@ -863,8 +863,8 @@ static int efx_wanted_rx_queues(void)for_each_online_cpu (cpu) {if (!cpu_isset (cpu, core_mask)) { ++count; -cpus_or (core_mask, core_mask, -topology_core_siblings (cpu)); +cpumask_or (&core_mask, &core_mask, +topology_core_cpumask (cpu)); } }
include/linux/topology.h +6 -0
@@ -193,5 +193,11 @@ int arch_update_cpu_topology(void);#ifndef topology_core_siblings#define topology_core_siblings (cpu) cpumask_of_cpu(cpu)#endif +#ifndef topology_thread_cpumask +#define topology_thread_cpumask (cpu) cpumask_of(cpu) +#endif +#ifndef topology_core_cpumask +#define topology_core_cpumask (cpu) cpumask_of(cpu) +#endif #endif /* _LINUX_TOPOLOGY_H */
Re: [PATCH 2/5] cpumask: Use topology_core_cpumask()/topology_thread_cpumask()
Quoted 11 lines · Mike Travis
From: Rusty Russell <[email protected]>
Impact: reduce stack usage, use new cpumask API.
This actually uses
topology_core_cpumask()andtopology_thread_cpumask(), removing the only users oftopology_core_siblings()andtopology_thread_siblings()Signed-off-by Rusty Russell <[email protected]>Signed-off-by Mike Travis <[email protected]>
Ben.
Signature · 3 lines
Ben Hutchings, Senior Software Engineer, Solarflare Communications Not speaking for my employer; that's the marketing department's job. They asked us to note that Solarflare product names are trademarked.
PATCH 3/5 cpumask: convert misc driver functions
In reply to Mike TravisFrom: Rusty Russell <[email protected]>
Impact: use new cpumask API.
Convert misc driver functions to use struct cpumask.
To Do:
- Convert iucv_buffer_cpumask to cpumask_var_t.
--- drivers/base/cpu.c | 2 +- drivers/misc/sgi-xp/xpc_main.c | 2 +- drivers/oprofile/buffer_sync.c | 22 ++++++++++++++++++---- drivers/oprofile/buffer_sync.h | 4 ++++ drivers/oprofile/oprof.c | 9 ++++++++- drivers/xen/manage.c | 2 +- net/iucv/iucv.c | 28 ++++++++++++++++++---------- 7 files changed, 51 insertions(+), 18 deletions(-)
drivers/base/cpu.c +1 -1
@@ -107,7 +107,7 @@ static SYSDEV_ATTR(crash_notes, 0400, sh/* * Print cpu online, possible, present, and system maps */ -static ssize_t print_cpus_map (char *buf,cpumask_t *map) +static ssize_t print_cpus_map (char *buf,const struct cpumask *map) {int n =cpulist_scnprintf (buf,PAGE_SIZE -2 , map);
drivers/misc/sgi-xp/xpc_main.c +1 -1
@@ -318,7 +318,7 @@ xpc_hb_checker(void *ignore)/* this thread was marked active by xpc_hb_init() */ -set_cpus_allowed_ptr (current, &cpumask_of_cpu (XPC_HB_CHECK_CPU )); +set_cpus_allowed_ptr (current,cpumask_of (XPC_HB_CHECK_CPU ));/* set our heartbeating to other partitions into motion */ xpc_hb_check_timeout = jiffies + (xpc_hb_check_interval *HZ );
drivers/oprofile/buffer_sync.c +18 -4
@@ -37,7 +37,7 @@static LIST_HEAD (dying_tasks);static LIST_HEAD (dead_tasks); -static cpumask_t marked_cpus =CPU_MASK_NONE ; +static cpumask_var_t marked_cpus;static DEFINE_SPINLOCK (task_mortuary);static void process_task_mortuary (void );
@@ -489,10 +489,10 @@ static void mark_done(int cpu) {int i; -cpu_set (cpu, marked_cpus); +cpumask_set_cpu (cpu, marked_cpus);for_each_online_cpu (i) { -if (!cpu_isset (i, marked_cpus)) +if (!cpumask_test_cpu (i, marked_cpus))return ; }
@@ -501,7 +501,7 @@ static void mark_done(int cpu)*/ process_task_mortuary (); -cpus_clear (marked_cpus); +cpumask_clear (marked_cpus); }
@@ -595,6 +595,20 @@ void sync_buffer(int cpu)mutex_unlock (&buffer_mutex); } +int __init buffer_sync_init (void ) +{ +if (!alloc_cpumask_var (&marked_cpus,GFP_KERNEL )) +return -ENOMEM ; + +cpumask_clear (marked_cpus); +return 0 ; +} + +void __exit buffer_sync_cleanup(void) +{ +free_cpumask_var (marked_cpus); +} +/* The function can be used to add a buffer worth of data directly to * the kernel buffer. The buffer is assumed to be a circular buffer. * Take the entries from index start and end at index end, wrapping
drivers/oprofile/buffer_sync.h +4 -0
@@ -19,4 +19,8 @@ void sync_stop(void);/* sync the given CPU's buffer */ void sync_buffer (int cpu); +/* initialize/destroy the buffer system. */ +int buffer_sync_init (void ); +void buffer_sync_cleanup (void ); +#endif /* OPROFILE_BUFFER_SYNC_H */
drivers/oprofile/oprof.c +8 -1
@@ -183,6 +183,10 @@ static int __init oprofile_init(void) {int err; + err =buffer_sync_init (); +if (err) +return err; + err =oprofile_arch_init (&oprofile_ops);if (err <0 || timer) {
@@ -191,8 +195,10 @@ static int __init oprofile_init(void) } err =oprofilefs_register (); -if (err) +if (err) {oprofile_arch_exit (); +buffer_sync_cleanup (); + }return err; }
@@ -202,6 +208,7 @@ static void __exit oprofile_exit(void) {oprofilefs_unregister ();oprofile_arch_exit (); +buffer_sync_cleanup (); }
drivers/xen/manage.c +1 -1
@@ -100,7 +100,7 @@ static void do_suspend(void)/* XXX use normal device tree? */ xenbus_suspend (); - err =stop_machine (xen_suspend, &cancelled, &cpumask_of_cpu (0 )); + err =stop_machine (xen_suspend, &cancelled,cpumask_of (0 ));if (err) {printk (KERN_ERR "failed to start xen_suspend: %d\n" , err); goto out;
net/iucv/iucv.c +18 -10
@@ -492,15 +492,14 @@ static void iucv_setmask_mp(void)* * Allow iucv interrupts on a single cpu. */ -static void iucv_setmask_up (void ) +static void iucv_setmask_up (struct cpumask *cpumask) { -cpumask_t cpumask;int cpu;/* Disable all cpu but the first in cpu_irq_cpumask. */ - cpumask = iucv_irq_cpumask; -cpu_clear (first_cpu (iucv_irq_cpumask), cpumask); -for_each_cpu_mask_nr (cpu, cpumask) +cpumask_copy (cpumask, iucv_irq_cpumask); +cpumask_clear_cpu (cpumask_first (iucv_irq_cpumask), cpumask); +for_each_cpu (cpu, cpumask)smp_call_function_single (cpu, iucv_block_cpu,NULL ,1 ); }
@@ -558,7 +557,7 @@ static void iucv_disable(void)static int __cpuinit iucv_cpu_notify (struct notifier_block *self,unsigned long action,void *hcpu) { -cpumask_t cpumask; +cpumask_var_t cpumask;long cpu = (long ) hcpu;switch (action) {
@@ -593,15 +592,20 @@ static int __cpuinit iucv_cpu_notify(strbreak ;case CPU_DOWN_PREPARE :case CPU_DOWN_PREPARE_FROZEN : - cpumask = iucv_buffer_cpumask; -cpu_clear (cpu, cpumask); -if (cpus_empty (cpumask)) +if (!alloc_cpumask_var (&cpumask,GFP_KERNEL )) +return NOTIFY_BAD ; +cpumask_copy (cpumask, &iucv_buffer_cpumask); +cpumask_clear_cpu (cpu, cpumask); +if (cpumask_empty (cpumask)) {/* Can't offline last IUCV enabled cpu. */ +free_cpumask_var (cpumask);return NOTIFY_BAD ; + }smp_call_function_single (cpu, iucv_retrieve_cpu,NULL ,1 );if (cpus_empty (iucv_irq_cpumask))smp_call_function_single (first_cpu (iucv_buffer_cpumask), iucv_allow_cpu,NULL ,1 ); +free_cpumask_var (cpumask);break ; }return NOTIFY_OK ;
@@ -686,9 +690,12 @@ static void iucv_cleanup_queue(void)int iucv_register (struct iucv_handler *handler,int smp) {int rc; +cpumask_var_t scratch;if (!iucv_available)return -ENOSYS ; +if (!alloc_cpumask_var (&scratch,GFP_KERNEL )) +return -ENOMEM ;mutex_lock (&iucv_register_mutex);if (!smp) iucv_nonsmp_handler++;
@@ -697,7 +704,7 @@ int iucv_register(struct iucv_handler *hif (rc) goto out_mutex; }else if (!smp && iucv_nonsmp_handler ==1 ) -iucv_setmask_up (); +iucv_setmask_up (scratch);INIT_LIST_HEAD (&handler->paths);spin_lock_bh (&iucv_table_lock);
@@ -706,6 +713,7 @@ int iucv_register(struct iucv_handler *h rc =0 ; out_mutex:mutex_unlock (&iucv_register_mutex); +free_cpumask_var (scratch);return rc; }EXPORT_SYMBOL (iucv_register);
Re: [PATCH 3/5] cpumask: convert misc driver functions
Quoted 5 lines · On Thursday 08 January 2009 06:28:35 Mike Travis wrote:
From: Rusty Russell <[email protected]>
Impact: use new cpumask API.
Convert misc driver functions to use struct cpumask.
...
Quoted 12 lines
net/iucv/iucv.c +5
-3@@ -593,15 +592,20 @@ static int __cpuinit iucv_cpu_notify(strbreak ;case CPU_DOWN_PREPARE :case CPU_DOWN_PREPARE_FROZEN : - cpumask = iucv_buffer_cpumask; -cpu_clear (cpu, cpumask); -if (cpus_empty (cpumask)) +if (!alloc_cpumask_var (&cpumask,GFP_KERNEL )) +return NOTIFY_BAD ; +cpumask_copy (cpumask, &iucv_buffer_cpumask); +cpumask_clear_cpu (cpu, cpumask); +if (cpumask_empty (cpumask)) {
net/iucv/iucv.c static int __cpuinit iucv_cpu_notify(str · line 609
Quoted 10 lines
net/iucv/iucv.c +3
-0/* Can't offline last IUCV enabled cpu. */ +free_cpumask_var (cpumask);return NOTIFY_BAD ; + }smp_call_function_single (cpu, iucv_retrieve_cpu,NULL ,1 );if (cpus_empty (iucv_irq_cpumask))smp_call_function_single (first_cpu (iucv_buffer_cpumask), iucv_allow_cpu,NULL ,1 ); +free_cpumask_var (cpumask);break ;
Minor nit: this can be done better, actually:
@@ -596,9 +594,7 @@ static int __cpuinit iucv_cpu_notify(str
case CPU_DOWN_PREPARE_FROZEN:
if (!iucv_path_table)
break;
- cpumask = iucv_buffer_cpumask;
- cpu_clear(cpu, cpumask);
- if (cpus_empty(cpumask))
+ if (cpumask_any_but(&iucv_buffer_cpumask, cpu) >= nr_cpu_ids)
/* Can't offline last IUCV enabled cpu. */ return NOTIFY_BAD;
smp_call_function_single(cpu, iucv_retrieve_cpu, NULL, 1);
Cheers,
Rusty.
Re: [PATCH 3/5] cpumask: convert misc driver functions
In reply to Mike Travis+void __exit buffer_sync_cleanup(void)
+{
+ free_cpumask_var(marked_cpus);
+}
This breaks ia64 with CONFIG_OPROFILE=y
buffer_sync_cleanup' referenced in section .init.text' of arch/ia64/oprofile/built-in.o: defined in discarded section `.exit.text' of arch/ia64/oprofile/built-in.o
make: *** [.tmp_vmlinux1] Error 1
-Tony
Re: [PATCH 3/5] cpumask: convert misc driver functions
Quoted 5 lines · Tony Luck
drivers/oprofile/buffer_sync.c +4
-0+void __exit buffer_sync_cleanup(void) +{ +free_cpumask_var (marked_cpus); +}
Quoted 8 lines
This breaks ia64 with
CONFIG_OPROFILE=y
buffer_sync_cleanup' referenced in section.init.text' ofarch/ia64/oprofile/built-in.o: defined in discarded section `.exit.text' ofarch/ia64/oprofile/built-in.o
make: *** [.tmp_vmlinux1] Error 1-Tony
Hi Tony,
Hmm, that is a tricky error. I'll see what can be done to eliminate it.
Thanks!
Mike
Re: [PATCH 3/5] cpumask: convert misc driver functions
In reply to Tony LuckQuoted 5 lines · Tony Luck
(quoted diff) +4
-0+void __exit buffer_sync_cleanup(void) +{ + free_cpumask_var(marked_cpus); +}
Quoted 8 lines
This breaks ia64 with
CONFIG_OPROFILE=y
buffer_sync_cleanup' referenced in section.init.text' ofarch/ia64/oprofile/built-in.o: defined in discarded section `.exit.text' ofarch/ia64/oprofile/built-in.o
make: *** [.tmp_vmlinux1] Error 1-Tony
Hi,
It's a bit hackish but I couldn't think of another workaround. And curiously, it doesn't affect x86?
Also, CPUMASK_OFFSTACK=n for ia64 so this is really a nop. (I haven't been able to turn it on for ia64.)
Thanks,
Mike
---
Subject: ia64: fix build error when OPROFILE enabled.
Impact: Fix build error.
Add a buffer_sync_cleanup_init() which duplicates the buffer_sync_cleanup() function but is placed in the .init.text section. This eliminates this build error:
buffer_sync_cleanup' referenced in section .init.text' of arch/ia64/oprofile/built-in.o: defined in discarded section `.exit.text' of arch/ia64/oprofile/built-in.o
--- drivers/oprofile/buffer_sync.c | 6 ++++++ drivers/oprofile/buffer_sync.h | 1 + drivers/oprofile/oprof.c | 2 +- 3 files changed, 8 insertions(+), 1 deletion(-)
drivers/oprofile/buffer_sync.c +6 -0
@@ -579,6 +579,12 @@ void __exit buffer_sync_cleanup(void)free_cpumask_var (marked_cpus); } +/* same as above but in the __init section */ +void __init buffer_sync_cleanup_init (void ) +{ +free_cpumask_var (marked_cpus); +} +/* The function can be used to add a buffer worth of data directly to * the kernel buffer. The buffer is assumed to be a circular buffer. * Take the entries from index start and end at index end, wrapping
drivers/oprofile/buffer_sync.h +1 -0
@@ -22,5 +22,6 @@ void sync_buffer(int cpu);/* initialize/destroy the buffer system. */ int buffer_sync_init (void );void buffer_sync_cleanup (void ); +void buffer_sync_cleanup_init (void );#endif /* OPROFILE_BUFFER_SYNC_H */
drivers/oprofile/oprof.c +1 -1
@@ -197,7 +197,7 @@ static int __init oprofile_init(void) err =oprofilefs_register ();if (err) {oprofile_arch_exit (); -buffer_sync_cleanup (); +buffer_sync_cleanup_init (); }return err;
Re: [PATCH 3/5] cpumask: convert misc driver functions
In reply to Mike TravisQuoted 21 lines · Mike Travis
From: Rusty Russell <[email protected]>
Impact: use new cpumask API.
Convert misc driver functions to use struct cpumask.
To Do:
- Convertiucv_buffer_cpumasktocpumask_var_t.Signed-off-by Rusty Russell <[email protected]>Signed-off-by Mike Travis <[email protected]>Acked-by Dean Nelson <[email protected]>Cc Robert Richter <[email protected]>Cc Jeremy Fitzhardinge <[email protected]>Cc Chris Wright <[email protected]>Cc Ursula Braun <[email protected]>
Quoted 6 lines
--- drivers/base/cpu.c | 2 +- drivers/misc/sgi-xp/xpc_main.c | 2 +- drivers/oprofile/buffer_sync.c | 22 ++++++++++++++++++---- drivers/oprofile/buffer_sync.h | 4 ++++ drivers/oprofile/oprof.c | 9 ++++++++-
I will send out a delta patch that changes the initialization.
-Robert
Quoted 3 lines
drivers/xen/manage.c | 2 +- net/iucv/iucv.c | 28 ++++++++++++++++++---------- 7 files changed, 51 insertions(+), 18 deletions(-)
Signature · 3 lines
Advanced Micro Devices, Inc. Operating System Research Center email: [email protected]
PATCH 1/1 cpumask: modifiy oprofile initialization
Delta patch to f7df8ed164996cd2c6aca9674388be6ef78d8b37 for tip/cpus4096.
Moved initialization to sync_start()/sync_stop(). No changes needed in buffer_sync.h and oprof.c anymore.
--- drivers/oprofile/buffer_sync.c | 20 ++++++-------------- drivers/oprofile/buffer_sync.h | 4 ---- drivers/oprofile/oprof.c | 9 +-------- 3 files changed, 7 insertions(+), 26 deletions(-)
drivers/oprofile/buffer_sync.c +6 -14
@@ -154,6 +154,10 @@ int sync_start(void) {int err; +if (!alloc_cpumask_var (&marked_cpus,GFP_KERNEL )) +return -ENOMEM ; +cpumask_clear (marked_cpus); +start_cpu_work (); err =task_handoff_register (&task_free_nb);
@@ -179,6 +183,7 @@ out2:task_handoff_unregister (&task_free_nb); out1:end_sync (); +free_cpumask_var (marked_cpus); goto out; }
@@ -190,6 +195,7 @@ void sync_stop(void)profile_event_unregister (PROFILE_TASK_EXIT , &task_exit_nb);task_handoff_unregister (&task_free_nb);end_sync (); +free_cpumask_var (marked_cpus); }
@@ -565,20 +571,6 @@ void sync_buffer(int cpu)mutex_unlock (&buffer_mutex); } -int __init buffer_sync_init (void ) -{ -if (!alloc_cpumask_var (&marked_cpus,GFP_KERNEL )) -return -ENOMEM ; - -cpumask_clear (marked_cpus); -return 0 ; -} - -void __exit buffer_sync_cleanup(void) -{ -free_cpumask_var (marked_cpus); -} -/* The function can be used to add a buffer worth of data directly to * the kernel buffer. The buffer is assumed to be a circular buffer. * Take the entries from index start and end at index end, wrapping
drivers/oprofile/buffer_sync.h +0 -4
@@ -19,8 +19,4 @@ void sync_stop(void);/* sync the given CPU's buffer */ void sync_buffer (int cpu); -/* initialize/destroy the buffer system. */ -int buffer_sync_init (void ); -void buffer_sync_cleanup (void ); -#endif /* OPROFILE_BUFFER_SYNC_H */
drivers/oprofile/oprof.c +1 -8
@@ -183,10 +183,6 @@ static int __init oprofile_init(void) {int err; - err =buffer_sync_init (); -if (err) -return err; - err =oprofile_arch_init (&oprofile_ops);if (err <0 || timer) {
@@ -195,10 +191,8 @@ static int __init oprofile_init(void) } err =oprofilefs_register (); -if (err) { +if (err)oprofile_arch_exit (); -buffer_sync_cleanup (); - }return err; }
@@ -208,7 +202,6 @@ static void __exit oprofile_exit(void) {oprofilefs_unregister ();oprofile_arch_exit (); -buffer_sync_cleanup (); }
Signature · 1 line
1.6.1
Ingo Molnar
AppliedRe: [PATCH] cpumask: modifiy oprofile initialization
Quoted 6 lines · Robert Richter
Delta patch to
f7df8ed164996cd2c6aca9674388be6ef78d8b37for tip/cpus4096.Moved initialization to sync_start()/sync_stop(). No changes needed in
buffer_sync.handoprof.canymore.
Quoted 8 lines
Cc Rusty Russell <[email protected]>Cc Mike Travis <[email protected]>Signed-off-by Robert Richter <[email protected]>--- drivers/oprofile/buffer_sync.c | 20 ++++++-------------- drivers/oprofile/buffer_sync.h | 4 ---- drivers/oprofile/oprof.c | 9 +-------- 3 files changed, 7 insertions(+), 26 deletions(-)
applied to tip/cpus4096, thanks Robert!
Ingo
Re: [PATCH] cpumask: modifiy oprofile initialization
Quoted 17 lines · Ingo Molnar
Quoted 6 lines · Robert Richter
Delta patch to
f7df8ed164996cd2c6aca9674388be6ef78d8b37for tip/cpus4096.Moved initialization to sync_start()/sync_stop(). No changes needed in
buffer_sync.handoprof.canymore.Quoted 8 lines
Cc Rusty Russell <[email protected]>Cc Mike Travis <[email protected]>Signed-off-by Robert Richter <[email protected]>--- drivers/oprofile/buffer_sync.c | 20 ++++++-------------- drivers/oprofile/buffer_sync.h | 4 ---- drivers/oprofile/oprof.c | 9 +-------- 3 files changed, 7 insertions(+), 26 deletions(-)
Quoted 3 lines
applied to tip/cpus4096, thanks Robert!
Ingo
Hi Ingo,
Doesn't this need to be pushed upstream as the error was discovered in linux-next on ia64?
Thanks,
Mike
Re: [PATCH] cpumask: modifiy oprofile initialization
Quoted 18 lines · Mike Travis
Quoted 17 lines · Ingo Molnar
Quoted 14 lines · Robert Richter
Delta patch to
f7df8ed164996cd2c6aca9674388be6ef78d8b37for tip/cpus4096.Moved initialization to sync_start()/sync_stop(). No changes needed in
buffer_sync.handoprof.canymore.Cc Rusty Russell <[email protected]>Cc Mike Travis <[email protected]>Signed-off-by Robert Richter <[email protected]>--- drivers/oprofile/buffer_sync.c | 20 ++++++-------------- drivers/oprofile/buffer_sync.h | 4 ---- drivers/oprofile/oprof.c | 9 +-------- 3 files changed, 7 insertions(+), 26 deletions(-)
Quoted 8 lines
Quoted 3 lines
applied to tip/cpus4096, thanks Robert!
Ingo
Hi Ingo,
Doesn't this need to be pushed upstream as the error was discovered in linux-next on ia64?
no, upstream -git does not have these changes yet, they came into linux-next via tip:cpus4096.
Ingo
Re: [PATCH] cpumask: modifiy oprofile initialization
In reply to Ingo MolnarQuoted 17 lines · Ingo Molnar
Quoted 6 lines · Robert Richter
Delta patch to
f7df8ed164996cd2c6aca9674388be6ef78d8b37for tip/cpus4096.Moved initialization to sync_start()/sync_stop(). No changes needed in
buffer_sync.handoprof.canymore.Quoted 8 lines
Cc Rusty Russell <[email protected]>Cc Mike Travis <[email protected]>Signed-off-by Robert Richter <[email protected]>--- drivers/oprofile/buffer_sync.c | 20 ++++++-------------- drivers/oprofile/buffer_sync.h | 4 ---- drivers/oprofile/oprof.c | 9 +-------- 3 files changed, 7 insertions(+), 26 deletions(-)
Quoted 3 lines
applied to tip/cpus4096, thanks Robert!
Ingo
Built and tested on an Altix 4700.
Cheers,
Mike
Re: [PATCH] cpumask: modifiy oprofile initialization
In reply to Robert RichterQuoted 5 lines · Robert Richter
Delta patch to
f7df8ed164996cd2c6aca9674388be6ef78d8b37for tip/cpus4096.Moved initialization to sync_start()/sync_stop(). No changes needed in
buffer_sync.handoprof.canymore.
Great, thanks Robert!
Are you pushing this via some upstream queue so I can drop it from mine?
Thanks,
Mike
Quoted 110 lines
Cc Rusty Russell <[email protected]>Cc Mike Travis <[email protected]>Signed-off-by Robert Richter <[email protected]>--- drivers/oprofile/buffer_sync.c | 20 ++++++-------------- drivers/oprofile/buffer_sync.h | 4 ---- drivers/oprofile/oprof.c | 9 +-------- 3 files changed, 7 insertions(+), 26 deletions(-)drivers/oprofile/buffer_sync.c +6
-14@@ -154,6 +154,10 @@ int sync_start(void) {int err; +if (!alloc_cpumask_var (&marked_cpus,GFP_KERNEL )) +return -ENOMEM ; +cpumask_clear (marked_cpus); +start_cpu_work (); err =task_handoff_register (&task_free_nb);@@ -179,6 +183,7 @@ out2:task_handoff_unregister (&task_free_nb); out1:end_sync (); +free_cpumask_var (marked_cpus); goto out; }@@ -190,6 +195,7 @@ void sync_stop(void)profile_event_unregister (PROFILE_TASK_EXIT , &task_exit_nb);task_handoff_unregister (&task_free_nb);end_sync (); +free_cpumask_var (marked_cpus); }@@ -565,20 +571,6 @@ void sync_buffer(int cpu)mutex_unlock (&buffer_mutex); } -int __init buffer_sync_init (void ) -{ -if (!alloc_cpumask_var (&marked_cpus,GFP_KERNEL )) -return -ENOMEM ; - -cpumask_clear (marked_cpus); -return 0 ; -} - -void __exit buffer_sync_cleanup(void) -{ -free_cpumask_var (marked_cpus); -} -/* The function can be used to add a buffer worth of data directly to * the kernel buffer. The buffer is assumed to be a circular buffer. * Take the entries from index start and end at index end, wrapping drivers/oprofile/buffer_sync.h +0
-4@@ -19,8 +19,4 @@ void sync_stop(void);/* sync the given CPU's buffer */ void sync_buffer (int cpu); -/* initialize/destroy the buffer system. */ -int buffer_sync_init (void ); -void buffer_sync_cleanup (void ); -#endif /* OPROFILE_BUFFER_SYNC_H */ drivers/oprofile/oprof.c +1
-8@@ -183,10 +183,6 @@ static int __init oprofile_init(void) {int err; - err =buffer_sync_init (); -if (err) -return err; - err =oprofile_arch_init (&oprofile_ops);if (err <0 || timer) {@@ -195,10 +191,8 @@ static int __init oprofile_init(void) } err =oprofilefs_register (); -if (err) { +if (err)oprofile_arch_exit (); -buffer_sync_cleanup (); - }return err; }@@ -208,7 +202,6 @@ static void __exit oprofile_exit(void) {oprofilefs_unregister ();oprofile_arch_exit (); -buffer_sync_cleanup (); }
Re: [PATCH] cpumask: modifiy oprofile initialization
Quoted 10 lines · Mike Travis
Quoted 5 lines · Robert Richter
Delta patch to
f7df8ed164996cd2c6aca9674388be6ef78d8b37for tip/cpus4096.Moved initialization to sync_start()/sync_stop(). No changes needed in
buffer_sync.handoprof.canymore.Great, thanks Robert!
Are you pushing this via some upstream queue so I can drop it from mine?
Mike, it is in the tip tree commit id 4c50d9e (tip/cpus4096).
-Robert
Signature · 3 lines
Advanced Micro Devices, Inc. Operating System Research Center email: [email protected]
PATCH 4/5 cpumask: convert drivers/net/sfc
In reply to Mike TravisFrom: Rusty Russell <[email protected]>
Impact: reduce stack usage, use new cpumask API.
Remove a cpumask from the stack. Ben Hutchings indicated that printing a warning and returning 1 was acceptable for the corner case where allocation fails.
--- drivers/net/sfc/efx.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)
drivers/net/sfc/efx.c +11 -4
@@ -854,20 +854,27 @@ static void efx_fini_io(struct efx_nic ** interrupts across them. */ static int efx_wanted_rx_queues (void ) { -cpumask_t core_mask; +cpumask_var_t core_mask;int count;int cpu; -cpus_clear (core_mask); +if (!alloc_cpumask_var (&core_mask,GFP_KERNEL )) { +printk (KERN_WARNING +"efx.c: allocation failure, irq balancing hobbled\n" ); +return 1 ; + } + +cpumask_clear (core_mask); count =0 ;for_each_online_cpu (cpu) { -if (!cpu_isset (cpu, core_mask)) { +if (!cpumask_test_cpu (cpu, core_mask)) { ++count; -cpumask_or (&core_mask, &core_mask, +cpumask_or (core_mask, core_mask,topology_core_cpumask (cpu)); } } +free_cpumask_var (core_mask);return count; }
Re: [PATCH 4/5] cpumask: convert drivers/net/sfc
Quoted 4 lines · Mike Travis
From: Rusty Russell <[email protected]>
Impact: reduce stack usage, use new cpumask API.
Quoted 8 lines
Remove a cpumask from the stack. Ben Hutchings indicated that printing a warning and returning 1 was acceptable for the corner case where allocation fails.
Signed-off-by Rusty Russell <[email protected]>Signed-off-by Mike Travis <[email protected]>Cc Ben Hutchings <[email protected]>
This seems to work.
Quoted 10 lines
--- drivers/net/sfc/efx.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)drivers/net/sfc/efx.c +0
-0@@ -854,20 +854,27 @@ static void efx_fini_io(struct efx_nic ** interrupts across them. */ static int efx_wanted_rx_queues (void )
drivers/net/sfc/efx.c static void efx_fini_io(struct efx_nic * · line 863
Quoted 10 lines
drivers/net/sfc/efx.c +4
-2{ -cpumask_t core_mask; +cpumask_var_t core_mask;int count;int cpu; -cpus_clear (core_mask); +if (!alloc_cpumask_var (&core_mask,GFP_KERNEL )) { +printk (KERN_WARNING +"efx.c: allocation failure, irq balancing hobbled\n" );
[...]
This is an exceedingly unlikely error case so the error message is not that important, but if you don't mind re-spinning this then please make it "sfc: RSS disabled due to allocation failure\n".
(Device-related log messages should also include the device address, but I don't want to add a device parameter to the function just for that. Maybe we should call this function just once at module init time since the result is not device-specific.)
Ben.
A message that is not in the archive (sent off-list or not yet received): [email protected]
Re: [PATCH 4/5] cpumask: convert drivers/net/sfc [PATCH supplied]
Quoted 44 lines · Rusty Russell
Quoted 39 lines · On Tuesday 20 January 2009 03:38:30 Ben Hutchings wrote:
Quoted 4 lines · Mike Travis
From: Rusty Russell <[email protected]>
Impact: reduce stack usage, use new cpumask API.
Quoted 8 lines
Remove a cpumask from the stack. Ben Hutchings indicated that printing a warning and returning 1 was acceptable for the corner case where allocation fails.
Signed-off-by Rusty Russell <[email protected]>Signed-off-by Mike Travis <[email protected]>Cc Ben Hutchings <[email protected]>Acked-by Ben Hutchings <[email protected]>This seems to work.
Quoted 20 lines
--- drivers/net/sfc/efx.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)drivers/net/sfc/efx.c +4
-2@@ -854,20 +854,27 @@ static void efx_fini_io(struct efx_nic ** interrupts across them. */ static int efx_wanted_rx_queues (void ) { -cpumask_t core_mask; +cpumask_var_t core_mask;int count;int cpu; -cpus_clear (core_mask); +if (!alloc_cpumask_var (&core_mask,GFP_KERNEL )) { +printk (KERN_WARNING +"efx.c: allocation failure, irq balancing hobbled\n" );[...]
Quoted 3 lines
This is an exceedingly unlikely error case so the error message is not that important, but if you don't mind re-spinning this then please make it "sfc: RSS disabled due to allocation failure\n".
Quoted 6 lines
OK, I'm away from all my machines, at linux.conf.au. Mike, can you change this if you haven't already?
Thanks,
Rusty.
Subject: sfc: modify allocation error message
Impact: change error message when alloc_cpumask_var fails.
--- drivers/net/sfc/efx.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/sfc/efx.c +2 -2
@@ -858,9 +858,9 @@ static int efx_wanted_rx_queues(void)int count;int cpu; -if (!alloc_cpumask_var (&core_mask,GFP_KERNEL )) { +if (unlikely (!alloc_cpumask_var (&core_mask,GFP_KERNEL ))) {printk (KERN_WARNING -"efx.c: allocation failure, irq balancing hobbled\n" ); +"sfc: RSS disabled due to allocation failure\n" );return 1 ; }
PATCH 1/1 sfc: modify allocation error message
In reply to a message not in the archiveQuoted 44 lines · Rusty Russell
Quoted 39 lines · On Tuesday 20 January 2009 03:38:30 Ben Hutchings wrote:
Quoted 4 lines · Mike Travis
From: Rusty Russell <[email protected]>
Impact: reduce stack usage, use new cpumask API.
Quoted 8 lines
Remove a cpumask from the stack. Ben Hutchings indicated that printing a warning and returning 1 was acceptable for the corner case where allocation fails.
Signed-off-by Rusty Russell <[email protected]>Signed-off-by Mike Travis <[email protected]>Cc Ben Hutchings <[email protected]>Acked-by Ben Hutchings <[email protected]>This seems to work.
Quoted 20 lines
--- drivers/net/sfc/efx.c | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-)drivers/net/sfc/efx.c +4
-2@@ -854,20 +854,27 @@ static void efx_fini_io(struct efx_nic ** interrupts across them. */ static int efx_wanted_rx_queues (void ) { -cpumask_t core_mask; +cpumask_var_t core_mask;int count;int cpu; -cpus_clear (core_mask); +if (!alloc_cpumask_var (&core_mask,GFP_KERNEL )) { +printk (KERN_WARNING +"efx.c: allocation failure, irq balancing hobbled\n" );[...]
Quoted 3 lines
This is an exceedingly unlikely error case so the error message is not that important, but if you don't mind re-spinning this then please make it "sfc: RSS disabled due to allocation failure\n".
Quoted 6 lines
OK, I'm away from all my machines, at linux.conf.au. Mike, can you change this if you haven't already?
Thanks,
Rusty.
Subject: sfc: modify allocation error message
Impact: change error message when alloc_cpumask_var fails.
--- drivers/net/sfc/efx.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/net/sfc/efx.c +2 -2
@@ -858,9 +858,9 @@ static int efx_wanted_rx_queues(void)int count;int cpu; -if (!alloc_cpumask_var (&core_mask,GFP_KERNEL )) { +if (unlikely (!alloc_cpumask_var (&core_mask,GFP_KERNEL ))) {printk (KERN_WARNING -"efx.c: allocation failure, irq balancing hobbled\n" ); +"sfc: RSS disabled due to allocation failure\n" );return 1 ; }
Re: [PATCH] sfc: modify allocation error message
On Mon, 2009-01-26 at 07:04 -0800, Mike Travis wrote:
[...]
Quoted 6 lines
Subject: sfc: modify allocation error message
Impact: change error message when
alloc_cpumask_varfails.Signed-off-by Mike Travis <[email protected]>Cc Ben Hutchings <[email protected]>
Quoted 18 lines
--
drivers/net/sfc/efx.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)drivers/net/sfc/efx.c +2
-2@@ -858,9 +858,9 @@ static int efx_wanted_rx_queues(void)int count;int cpu; -if (!alloc_cpumask_var (&core_mask,GFP_KERNEL )) { +if (unlikely (!alloc_cpumask_var (&core_mask,GFP_KERNEL ))) {printk (KERN_WARNING -"efx.c: allocation failure, irq balancing hobbled\n" ); +"sfc: RSS disabled due to allocation failure\n" );return 1 ; }
Signature · 3 lines
Ben Hutchings, Senior Software Engineer, Solarflare Communications Not speaking for my employer; that's the marketing department's job. They asked us to note that Solarflare product names are trademarked.
PATCH 5/5 cpumask: convert other misc kernel functions
In reply to Mike TravisFrom: Rusty Russell <[email protected]>
Impact: use new cpumask API.
Convert other misc kernel functions to use struct cpumask.
--- lib/smp_processor_id.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
lib/smp_processor_id.c +1 -1
@@ -22,7 +22,7 @@ notrace unsigned int debug_smp_processor* Kernel threads bound to a single CPU can safely use * smp_processor_id(): */ -if (cpus_equal (current->cpus_allowed,cpumask_of_cpu (this_cpu))) +if (cpumask_equal (¤t->cpus_allowed,cpumask_of (this_cpu))) goto out;/*
Quoted 9 lines · Mike Travis
One big change to
irq_descand a few smaller maintenance type updates.cpumask: update
irq_descto usecpumask_var_t
cpumask: Use topology_core_cpumask()/topology_thread_cpumask()
cpumask: convert misc driver functions
cpumask: convertdrivers/net/sfc
cpumask: convert other misc kernel functions
Quoted 7 lines
Based on tip/cpus4096: v2.6.28-5703-g1d1a70e
(Ingo - I added a blank line before the From: part and that seems to separate it from the Subject. Hopefully your git-am will still recognizes it - I did try a git-am here and it seemed to work ok.)
Signed-off-by Mike Travis <[email protected]>
Since you have it in a Git tree, why dont you send me a pull request?
I might be more picky about pull requests in general (because it's harder to fix up a pulled tree than to fix up commits that come from emails - and it's also more work for you if i refuse to pull than if i have some comments over email-patches), but it's a much better (more scalable) workflow in the long run than the email path.
Ingo
Quoted 20 lines · Ingo Molnar
Quoted 8 lines · Mike Travis
One big change to
irq_descand a few smaller maintenance type updates.cpumask: update
irq_descto usecpumask_var_t
cpumask: Use topology_core_cpumask()/topology_thread_cpumask()
cpumask: convert misc driver functions
cpumask: convertdrivers/net/sfc
cpumask: convert other misc kernel functionsQuoted 7 lines
Based on tip/cpus4096: v2.6.28-5703-g1d1a70e
(Ingo - I added a blank line before the From: part and that seems to separate it from the Subject. Hopefully your git-am will still recognizes it - I did try a git-am here and it seemed to work ok.)
Signed-off-by Mike Travis <[email protected]>Since you have it in a Git tree, why dont you send me a pull request?
Quoted 7 lines
I might be more picky about pull requests in general (because it's harder to fix up a pulled tree than to fix up commits that come from emails - and it's also more work for you if i refuse to pull than if i have some comments over email-patches), but it's a much better (more scalable) workflow in the long run than the email path.
Ingo
Ok, will do.
Thanks,
Mike