Revision d7a702f0b1033cf402fef65bd6395072738f0844 authored by Matt Fleming on 06 August 2015, 12:12:43 UTC, committed by Ingo Molnar on 12 August 2015, 09:37:23 UTC
Tony reports that booting his 144-cpu machine with maxcpus=10 triggers the following WARN_ON(): [ 21.045727] WARNING: CPU: 8 PID: 647 at arch/x86/kernel/cpu/perf_event_intel_cqm.c:1267 intel_cqm_cpu_prepare+0x75/0x90() [ 21.045744] CPU: 8 PID: 647 Comm: systemd-udevd Not tainted 4.2.0-rc4 #1 [ 21.045745] Hardware name: Intel Corporation BRICKLAND/BRICKLAND, BIOS BRHSXSD1.86B.0066.R00.1506021730 06/02/2015 [ 21.045747] 0000000000000000 0000000082771b09 ffff880856333ba8 ffffffff81669b67 [ 21.045748] 0000000000000000 0000000000000000 ffff880856333be8 ffffffff8107b02a [ 21.045750] ffff88085b789800 ffff88085f68a020 ffffffff819e2470 000000000000000a [ 21.045750] Call Trace: [ 21.045757] [<ffffffff81669b67>] dump_stack+0x45/0x57 [ 21.045759] [<ffffffff8107b02a>] warn_slowpath_common+0x8a/0xc0 [ 21.045761] [<ffffffff8107b15a>] warn_slowpath_null+0x1a/0x20 [ 21.045762] [<ffffffff81036725>] intel_cqm_cpu_prepare+0x75/0x90 [ 21.045764] [<ffffffff81036872>] intel_cqm_cpu_notifier+0x42/0x160 [ 21.045767] [<ffffffff8109a33d>] notifier_call_chain+0x4d/0x80 [ 21.045769] [<ffffffff8109a44e>] __raw_notifier_call_chain+0xe/0x10 [ 21.045770] [<ffffffff8107b538>] _cpu_up+0xe8/0x190 [ 21.045771] [<ffffffff8107b65a>] cpu_up+0x7a/0xa0 [ 21.045774] [<ffffffff8165e920>] cpu_subsys_online+0x40/0x90 [ 21.045777] [<ffffffff81433b37>] device_online+0x67/0x90 [ 21.045778] [<ffffffff81433bea>] online_store+0x8a/0xa0 [ 21.045782] [<ffffffff81430e78>] dev_attr_store+0x18/0x30 [ 21.045785] [<ffffffff8126b6ba>] sysfs_kf_write+0x3a/0x50 [ 21.045786] [<ffffffff8126ad40>] kernfs_fop_write+0x120/0x170 [ 21.045789] [<ffffffff811f0b77>] __vfs_write+0x37/0x100 [ 21.045791] [<ffffffff811f38b8>] ? __sb_start_write+0x58/0x110 [ 21.045795] [<ffffffff81296d2d>] ? security_file_permission+0x3d/0xc0 [ 21.045796] [<ffffffff811f1279>] vfs_write+0xa9/0x190 [ 21.045797] [<ffffffff811f2075>] SyS_write+0x55/0xc0 [ 21.045800] [<ffffffff81067300>] ? do_page_fault+0x30/0x80 [ 21.045804] [<ffffffff816709ae>] entry_SYSCALL_64_fastpath+0x12/0x71 [ 21.045805] ---[ end trace fe228b836d8af405 ]--- The root cause is that CPU_UP_PREPARE is completely the wrong notifier action from which to access cpu_data(), because smp_store_cpu_info() won't have been executed by the target CPU at that point, which in turn means that ->x86_cache_max_rmid and ->x86_cache_occ_scale haven't been filled out. Instead let's invoke our handler from CPU_STARTING and rename it appropriately. Reported-by: Tony Luck <tony.luck@intel.com> Signed-off-by: Matt Fleming <matt.fleming@intel.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Cc: Ashok Raj <ashok.raj@intel.com> Cc: Kanaka Juvva <kanaka.d.juvva@intel.com> Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Vikas Shivappa <vikas.shivappa@intel.com> Link: http://lkml.kernel.org/r/1438863163-14083-1-git-send-email-matt@codeblueprint.co.uk Signed-off-by: Ingo Molnar <mingo@kernel.org>
1 parent dbc72b7
blk-mq.h
#ifndef INT_BLK_MQ_H
#define INT_BLK_MQ_H
struct blk_mq_tag_set;
struct blk_mq_ctx {
struct {
spinlock_t lock;
struct list_head rq_list;
} ____cacheline_aligned_in_smp;
unsigned int cpu;
unsigned int index_hw;
unsigned int last_tag ____cacheline_aligned_in_smp;
/* incremented at dispatch time */
unsigned long rq_dispatched[2];
unsigned long rq_merged;
/* incremented at completion time */
unsigned long ____cacheline_aligned_in_smp rq_completed[2];
struct request_queue *queue;
struct kobject kobj;
} ____cacheline_aligned_in_smp;
void __blk_mq_complete_request(struct request *rq);
void blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx, bool async);
void blk_mq_freeze_queue(struct request_queue *q);
void blk_mq_free_queue(struct request_queue *q);
void blk_mq_clone_flush_request(struct request *flush_rq,
struct request *orig_rq);
int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr);
void blk_mq_wake_waiters(struct request_queue *q);
/*
* CPU hotplug helpers
*/
struct blk_mq_cpu_notifier;
void blk_mq_init_cpu_notifier(struct blk_mq_cpu_notifier *notifier,
int (*fn)(void *, unsigned long, unsigned int),
void *data);
void blk_mq_register_cpu_notifier(struct blk_mq_cpu_notifier *notifier);
void blk_mq_unregister_cpu_notifier(struct blk_mq_cpu_notifier *notifier);
void blk_mq_cpu_init(void);
void blk_mq_enable_hotplug(void);
void blk_mq_disable_hotplug(void);
/*
* CPU -> queue mappings
*/
extern unsigned int *blk_mq_make_queue_map(struct blk_mq_tag_set *set);
extern int blk_mq_update_queue_map(unsigned int *map, unsigned int nr_queues);
extern int blk_mq_hw_queue_to_node(unsigned int *map, unsigned int);
/*
* sysfs helpers
*/
extern int blk_mq_sysfs_register(struct request_queue *q);
extern void blk_mq_sysfs_unregister(struct request_queue *q);
extern void blk_mq_rq_timed_out(struct request *req, bool reserved);
void blk_mq_release(struct request_queue *q);
/*
* Basic implementation of sparser bitmap, allowing the user to spread
* the bits over more cachelines.
*/
struct blk_align_bitmap {
unsigned long word;
unsigned long depth;
} ____cacheline_aligned_in_smp;
static inline struct blk_mq_ctx *__blk_mq_get_ctx(struct request_queue *q,
unsigned int cpu)
{
return per_cpu_ptr(q->queue_ctx, cpu);
}
/*
* This assumes per-cpu software queueing queues. They could be per-node
* as well, for instance. For now this is hardcoded as-is. Note that we don't
* care about preemption, since we know the ctx's are persistent. This does
* mean that we can't rely on ctx always matching the currently running CPU.
*/
static inline struct blk_mq_ctx *blk_mq_get_ctx(struct request_queue *q)
{
return __blk_mq_get_ctx(q, get_cpu());
}
static inline void blk_mq_put_ctx(struct blk_mq_ctx *ctx)
{
put_cpu();
}
struct blk_mq_alloc_data {
/* input parameter */
struct request_queue *q;
gfp_t gfp;
bool reserved;
/* input & output parameter */
struct blk_mq_ctx *ctx;
struct blk_mq_hw_ctx *hctx;
};
static inline void blk_mq_set_alloc_data(struct blk_mq_alloc_data *data,
struct request_queue *q, gfp_t gfp, bool reserved,
struct blk_mq_ctx *ctx,
struct blk_mq_hw_ctx *hctx)
{
data->q = q;
data->gfp = gfp;
data->reserved = reserved;
data->ctx = ctx;
data->hctx = hctx;
}
static inline bool blk_mq_hw_queue_mapped(struct blk_mq_hw_ctx *hctx)
{
return hctx->nr_ctx && hctx->tags;
}
#endif
![swh spinner](/static/img/swh-spinner.gif)
Computing file changes ...