Add malloc_domainset(9) and _domainset variants to other allocator KPIs.

Remove malloc_domain(9) and most other _domain KPIs added in r327900.
The new functions allow the caller to specify a general NUMA domain
selection policy, rather than specifically requesting an allocation from
a specific domain.  The latter policy tends to interact poorly with
M_WAITOK, resulting in situations where a caller is blocked indefinitely
because the specified domain is depleted.  Most existing consumers of
the _domain KPIs are converted to instead use a DOMAINSET_PREF() policy,
in which we fall back to other domains to satisfy the allocation
request.

This change also defines a set of DOMAINSET_FIXED() policies, which
only permit allocations from the specified domain.

Discussed with:	gallatin, jeff
Reported and tested by:	pho (previous version)
MFC after:	2 weeks
Sponsored by:	The FreeBSD Foundation
Differential Revision:	https://reviews.freebsd.org/D17418

2 files changed, 29 insertions, 41 deletions

diff --git a/sys/dev/hwpmc/hwpmc_logging.c b/sys/dev/hwpmc/hwpmc_logging.c
index 8764ac9e922c..e76aa9b710eb 100644
--- a/sys/dev/hwpmc/hwpmc_logging.c
+++ b/sys/dev/hwpmc/hwpmc_logging.c
@@ -41,6 +41,7 @@ __FBSDID("$FreeBSD$");
 
 #include <sys/param.h>
 #include <sys/capsicum.h>
+#include <sys/domainset.h>
 #include <sys/file.h>
 #include <sys/kernel.h>
 #include <sys/kthread.h>
@@ -1231,8 +1232,8 @@ pmclog_process_userlog(struct pmc_owner *po, struct pmc_op_writelog *wl)
 void
 pmclog_initialize()
 {
-	int domain;
 	struct pmclog_buffer *plb;
+	int domain, ncpus, total;
 
 	if (pmclog_buffer_size <= 0 || pmclog_buffer_size > 16*1024) {
 		(void) printf("hwpmc: tunable logbuffersize=%d must be "
@@ -1253,16 +1254,17 @@ pmclog_initialize()
 		pmclog_buffer_size = PMC_LOG_BUFFER_SIZE;
 	}
 	for (domain = 0; domain < vm_ndomains; domain++) {
-		int ncpus = pmc_dom_hdrs[domain]->pdbh_ncpus;
-		int total = ncpus*pmc_nlogbuffers_pcpu;
+		ncpus = pmc_dom_hdrs[domain]->pdbh_ncpus;
+		total = ncpus * pmc_nlogbuffers_pcpu;
 
-		plb = malloc_domain(sizeof(struct pmclog_buffer)*total, M_PMC, domain, M_WAITOK|M_ZERO);
+		plb = malloc_domainset(sizeof(struct pmclog_buffer) * total,
+		    M_PMC, DOMAINSET_PREF(domain), M_WAITOK | M_ZERO);
 		pmc_dom_hdrs[domain]->pdbh_plbs = plb;
-		for (int i = 0; i < total; i++, plb++) {
+		for (; total > 0; total--, plb++) {
 			void *buf;
 
-			buf = malloc_domain(1024 * pmclog_buffer_size, M_PMC, domain,
-								M_WAITOK|M_ZERO);
+			buf = malloc_domainset(1024 * pmclog_buffer_size, M_PMC,
+			    DOMAINSET_PREF(domain), M_WAITOK | M_ZERO);
 			PMCLOG_INIT_BUFFER_DESCRIPTOR(plb, buf, domain);
 			pmc_plb_rele_unlocked(plb);
 		}
diff --git a/sys/dev/hwpmc/hwpmc_mod.c b/sys/dev/hwpmc/hwpmc_mod.c
index 1fad25c55702..f94f645b2e39 100644
--- a/sys/dev/hwpmc/hwpmc_mod.c
+++ b/sys/dev/hwpmc/hwpmc_mod.c
@@ -36,6 +36,7 @@
 __FBSDID("$FreeBSD$");
 
 #include <sys/param.h>
+#include <sys/domainset.h>
 #include <sys/eventhandler.h>
 #include <sys/gtaskqueue.h>
 #include <sys/jail.h>
@@ -78,14 +79,6 @@ __FBSDID("$FreeBSD$");
 
 #include "hwpmc_soft.h"
 
-#ifdef NUMA
-#define NDOMAINS vm_ndomains
-#else
-#define NDOMAINS 1
-#define malloc_domain(size, type, domain, flags) malloc((size), (type), (flags))
-#define free_domain(addr, type) free(addr, type)
-#endif
-
 #define PMC_EPOCH_ENTER() struct epoch_tracker pmc_et; epoch_enter_preempt(global_epoch_preempt, &pmc_et)
 #define PMC_EPOCH_EXIT() epoch_exit_preempt(global_epoch_preempt, &pmc_et)
 
@@ -5643,15 +5636,16 @@ pmc_initialize(void)
 			continue;
 		pc = pcpu_find(cpu);
 		domain = pc->pc_domain;
-		sb = malloc_domain(sizeof(struct pmc_samplebuffer) +
-			pmc_nsamples * sizeof(struct pmc_sample), M_PMC, domain,
-		    M_WAITOK|M_ZERO);
+		sb = malloc_domainset(sizeof(struct pmc_samplebuffer) +
+		    pmc_nsamples * sizeof(struct pmc_sample), M_PMC,
+		    DOMAINSET_PREF(domain), M_WAITOK | M_ZERO);
 
 		KASSERT(pmc_pcpu[cpu] != NULL,
 		    ("[pmc,%d] cpu=%d Null per-cpu data", __LINE__, cpu));
 
-		sb->ps_callchains = malloc_domain(pmc_callchaindepth * pmc_nsamples *
-			sizeof(uintptr_t), M_PMC, domain, M_WAITOK|M_ZERO);
+		sb->ps_callchains = malloc_domainset(pmc_callchaindepth *
+		    pmc_nsamples * sizeof(uintptr_t), M_PMC,
+		    DOMAINSET_PREF(domain), M_WAITOK | M_ZERO);
 
 		for (n = 0, ps = sb->ps_samples; n < pmc_nsamples; n++, ps++)
 			ps->ps_pc = sb->ps_callchains +
@@ -5659,35 +5653,27 @@ pmc_initialize(void)
 
 		pmc_pcpu[cpu]->pc_sb[PMC_HR] = sb;
 
-		sb = malloc_domain(sizeof(struct pmc_samplebuffer) +
-			pmc_nsamples * sizeof(struct pmc_sample), M_PMC, domain,
-		    M_WAITOK|M_ZERO);
-
-		KASSERT(pmc_pcpu[cpu] != NULL,
-		    ("[pmc,%d] cpu=%d Null per-cpu data", __LINE__, cpu));
-
-		sb->ps_callchains = malloc_domain(pmc_callchaindepth * pmc_nsamples *
-			sizeof(uintptr_t), M_PMC, domain, M_WAITOK|M_ZERO);
+		sb = malloc_domainset(sizeof(struct pmc_samplebuffer) +
+		    pmc_nsamples * sizeof(struct pmc_sample), M_PMC,
+		    DOMAINSET_PREF(domain), M_WAITOK | M_ZERO);
 
+		sb->ps_callchains = malloc_domainset(pmc_callchaindepth *
+		    pmc_nsamples * sizeof(uintptr_t), M_PMC,
+		    DOMAINSET_PREF(domain), M_WAITOK | M_ZERO);
 		for (n = 0, ps = sb->ps_samples; n < pmc_nsamples; n++, ps++)
 			ps->ps_pc = sb->ps_callchains +
 			    (n * pmc_callchaindepth);
 
 		pmc_pcpu[cpu]->pc_sb[PMC_SR] = sb;
 
-		sb = malloc_domain(sizeof(struct pmc_samplebuffer) +
-			pmc_nsamples * sizeof(struct pmc_sample), M_PMC, domain,
-		    M_WAITOK|M_ZERO);
-
-		KASSERT(pmc_pcpu[cpu] != NULL,
-		    ("[pmc,%d] cpu=%d Null per-cpu data", __LINE__, cpu));
-
-		sb->ps_callchains = malloc_domain(pmc_callchaindepth * pmc_nsamples *
-		    sizeof(uintptr_t), M_PMC, domain, M_WAITOK|M_ZERO);
-
+		sb = malloc_domainset(sizeof(struct pmc_samplebuffer) +
+		    pmc_nsamples * sizeof(struct pmc_sample), M_PMC,
+		    DOMAINSET_PREF(domain), M_WAITOK | M_ZERO);
+		sb->ps_callchains = malloc_domainset(pmc_callchaindepth *
+		    pmc_nsamples * sizeof(uintptr_t), M_PMC,
+		    DOMAINSET_PREF(domain), M_WAITOK | M_ZERO);
 		for (n = 0, ps = sb->ps_samples; n < pmc_nsamples; n++, ps++)
-			ps->ps_pc = sb->ps_callchains +
-			    (n * pmc_callchaindepth);
+			ps->ps_pc = sb->ps_callchains + n * pmc_callchaindepth;
 
 		pmc_pcpu[cpu]->pc_sb[PMC_UR] = sb;
 	}