Fix pre-SI_SUB_CPU initialization of per-CPU counters.

r336020 introduced pcpu_page_alloc(), replacing page_alloc() as the
backend allocator for PCPU UMA zones.  Unlike page_alloc(), it does
not honour malloc(9) flags such as M_ZERO or M_NODUMP, so fix that.

r336020 also changed counter(9) to initialize each counter using a
CPU_FOREACH() loop instead of an SMP rendezvous.  Before SI_SUB_CPU,
smp_rendezvous() will only execute the callback on the current CPU
(i.e., CPU 0), so only one counter gets zeroed.  The rest are zeroed
by virtue of the fact that UMA gratuitously zeroes slabs when importing
them into a zone.

Prior to SI_SUB_CPU, all_cpus is clear, so with r336020 we weren't
zeroing vm_cnt counters during boot: the CPU_FOREACH() loop had no
effect, and pcpu_page_alloc() didn't honour M_ZERO.  Fix this by
iterating over the full range of CPU IDs when zeroing counters,
ignoring whether the corresponding bits in all_cpus are set.

Reported and tested by:	pho (previous version)
Reviewed by:		kib (previous version)
Differential Revision:	https://reviews.freebsd.org/D16190
This commit is contained in:
Mark Johnston 2018-07-10 00:18:12 +00:00
parent bdea3adca6
commit 013072f04c
2 changed files with 7 additions and 22 deletions

View File

@ -43,15 +43,6 @@ __FBSDID("$FreeBSD$");
#define IN_SUBR_COUNTER_C
#include <sys/counter.h>
static void
counter_u64_zero_sync(counter_u64_t c)
{
int cpu;
CPU_FOREACH(cpu)
*(uint64_t*)zpcpu_get_cpu(c, cpu) = 0;
}
void
counter_u64_zero(counter_u64_t c)
{
@ -69,13 +60,8 @@ counter_u64_fetch(counter_u64_t c)
counter_u64_t
counter_u64_alloc(int flags)
{
counter_u64_t r;
r = uma_zalloc_pcpu(pcpu_zone_64, flags);
if (r != NULL)
counter_u64_zero_sync(r);
return (r);
return (uma_zalloc_pcpu(pcpu_zone_64, flags | M_ZERO));
}
void

View File

@ -1186,13 +1186,12 @@ pcpu_page_alloc(uma_zone_t zone, vm_size_t bytes, int domain, uint8_t *pflag,
struct pcpu *pc;
#endif
TAILQ_INIT(&alloctail);
MPASS(bytes == (mp_maxid+1)*PAGE_SIZE);
*pflag = UMA_SLAB_KERNEL;
MPASS(bytes == (mp_maxid + 1) * PAGE_SIZE);
TAILQ_INIT(&alloctail);
flags = VM_ALLOC_SYSTEM | VM_ALLOC_WIRED | VM_ALLOC_NOOBJ |
((wait & M_WAITOK) != 0 ? VM_ALLOC_WAITOK :
VM_ALLOC_NOWAIT);
malloc2vm_flags(wait);
*pflag = UMA_SLAB_KERNEL;
for (cpu = 0; cpu <= mp_maxid; cpu++) {
if (CPU_ABSENT(cpu)) {
p = vm_page_alloc(NULL, 0, flags);
@ -2328,10 +2327,10 @@ uma_zalloc_pcpu_arg(uma_zone_t zone, void *udata, int flags)
MPASS(zone->uz_flags & UMA_ZONE_PCPU);
#endif
item = uma_zalloc_arg(zone, udata, flags &~ M_ZERO);
item = uma_zalloc_arg(zone, udata, flags & ~M_ZERO);
if (item != NULL && (flags & M_ZERO)) {
#ifdef SMP
CPU_FOREACH(i)
for (i = 0; i <= mp_maxid; i++)
bzero(zpcpu_get_cpu(item, i), zone->uz_size);
#else
bzero(item, zone->uz_size);