Skip site navigation (1)Skip section navigation (2) 
Date:      Thu, 20 Sep 2018 15:45:12 +0000 (UTC)
From:      Mark Johnston <markj@FreeBSD.org>
To:        src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org
Subject:   svn commit: r338830 - head/sys/vm
Message-ID:  <201809201545.w8KFjCr5081308@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help 
Author: markj
Date: Thu Sep 20 15:45:12 2018
New Revision: 338830
URL: https://svnweb.freebsd.org/changeset/base/338830

Log:
  Change the domain selection policy in kmem_back().
  
  Ensure that pages backing the same virtual large page come from the
  same physical domain, as kmem_malloc_domain() does.
  
  PR:		231038
  Reviewed by:	alc, kib
  Approved by:	re (gjb)
  Sponsored by:	The FreeBSD Foundation
  Differential Revision:	https://reviews.freebsd.org/D17248

Modified:
  head/sys/vm/vm_kern.c

Modified: head/sys/vm/vm_kern.c
==============================================================================
--- head/sys/vm/vm_kern.c	Thu Sep 20 13:32:40 2018	(r338829)
+++ head/sys/vm/vm_kern.c	Thu Sep 20 15:45:12 2018	(r338830)
@@ -122,11 +122,12 @@ SYSCTL_ULONG(_vm, OID_AUTO, max_kernel_address, CTLFLA
     "Max kernel address");
 
 #if VM_NRESERVLEVEL > 0
-#define	KVA_QUANTUM	(1 << (VM_LEVEL_0_ORDER + PAGE_SHIFT))
+#define	KVA_QUANTUM_SHIFT	(VM_LEVEL_0_ORDER + PAGE_SHIFT)
 #else
 /* On non-superpage architectures want large import sizes. */
-#define	KVA_QUANTUM	(PAGE_SIZE * 1024)
+#define	KVA_QUANTUM_SHIFT	(10 + PAGE_SHIFT)
 #endif
+#define	KVA_QUANTUM		(1 << KVA_QUANTUM_SHIFT)
 
 /*
  *	kva_alloc:
@@ -416,9 +417,10 @@ kmem_malloc(vm_size_t size, int flags)
 }
 
 /*
- *	kmem_back:
+ *	kmem_back_domain:
  *
- *	Allocate physical pages for the specified virtual address range.
+ *	Allocate physical pages from the specified domain for the specified
+ *	virtual address range.
  */
 int
 kmem_back_domain(int domain, vm_object_t object, vm_offset_t addr,
@@ -479,24 +481,39 @@ retry:
 	return (KERN_SUCCESS);
 }
 
+/*
+ *	kmem_back:
+ *
+ *	Allocate physical pages for the specified virtual address range.
+ */
 int
 kmem_back(vm_object_t object, vm_offset_t addr, vm_size_t size, int flags)
 {
-	struct vm_domainset_iter di;
-	int domain;
-	int ret;
+	vm_offset_t end, next, start;
+	int domain, rv;
 
 	KASSERT(object == kernel_object,
 	    ("kmem_back: only supports kernel object."));
 
-	vm_domainset_iter_malloc_init(&di, kernel_object, &domain, &flags);
-	do {
-		ret = kmem_back_domain(domain, object, addr, size, flags);
-		if (ret == KERN_SUCCESS)
+	for (start = addr, end = addr + size; addr < end; addr = next) {
+		/*
+		 * We must ensure that pages backing a given large virtual page
+		 * all come from the same physical domain.
+		 */
+		if (vm_ndomains > 1) {
+			domain = (addr >> KVA_QUANTUM_SHIFT) % vm_ndomains;
+			next = roundup2(addr + 1, KVA_QUANTUM);
+			if (next > end || next < start)
+				next = end;
+		} else
+			next = end;
+		rv = kmem_back_domain(domain, object, addr, next - addr, flags);
+		if (rv != KERN_SUCCESS) {
+			kmem_unback(object, start, addr - start);
 			break;
-	} while (vm_domainset_iter_malloc(&di, &domain, &flags) == 0);
-
-	return (ret);
+		}
+	}
+	return (rv);
 }
 
 /*

Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201809201545.w8KFjCr5081308>