linux-kernel - Re: [PATCH] mm/alloc: fallback to first node if the wanted node offline

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20181206082806.GB1286@dhcp22.suse.cz>
Date:   Thu, 6 Dec 2018 09:28:06 +0100
From:   Michal Hocko <mhocko@...nel.org>
To:     Pingfan Liu <kernelfans@...il.com>
Cc:     Vlastimil Babka <vbabka@...e.cz>, linux-mm@...ck.org,
        linux-kernel@...r.kernel.org,
        Andrew Morton <akpm@...ux-foundation.org>,
        Mike Rapoport <rppt@...ux.vnet.ibm.com>,
        Bjorn Helgaas <bhelgaas@...gle.com>,
        Jonathan Cameron <Jonathan.Cameron@...wei.com>
Subject: Re: [PATCH] mm/alloc: fallback to first node if the wanted node
 offline

On Thu 06-12-18 11:07:33, Pingfan Liu wrote:
> On Wed, Dec 5, 2018 at 5:40 PM Vlastimil Babka <vbabka@...e.cz> wrote:
> >
> > On 12/5/18 10:29 AM, Pingfan Liu wrote:
> > >> [    0.007418] Early memory node ranges
> > >> [    0.007419]   node   1: [mem 0x0000000000001000-0x000000000008efff]
> > >> [    0.007420]   node   1: [mem 0x0000000000090000-0x000000000009ffff]
> > >> [    0.007422]   node   1: [mem 0x0000000000100000-0x000000005c3d6fff]
> > >> [    0.007422]   node   1: [mem 0x00000000643df000-0x0000000068ff7fff]
> > >> [    0.007423]   node   1: [mem 0x000000006c528000-0x000000006fffffff]
> > >> [    0.007424]   node   1: [mem 0x0000000100000000-0x000000047fffffff]
> > >> [    0.007425]   node   5: [mem 0x0000000480000000-0x000000087effffff]
> > >>
> > >> There is clearly no node2. Where did the driver get the node2 from?
> >
> > I don't understand these tables too much, but it seems the other nodes
> > exist without them:
> >
> > [    0.007393] SRAT: PXM 2 -> APIC 0x20 -> Node 2
> >
> > Maybe the nodes are hotplugable or something?
> >
> I also not sure about it, and just have a hurry look at acpi spec. I
> will reply it on another email, and Cced some acpi guys about it
> 
> > > Since using nr_cpus=4 , the node2 is not be instanced by x86 initalizing code.
> >
> > Indeed, nr_cpus seems to restrict what nodes we allocate and populate
> > zonelists for.
> 
> Yes, in init_cpu_to_node(),  since nr_cpus limits the possible cpu,
> which affects the loop for_each_possible_cpu(cpu) and skip the node2
> in this case.

THanks for pointing this out. It made my life easier. So It think the
bug is that we call init_memory_less_node from this path. I suspect
numa_register_memblks is the right place to do this. So I admit I
am not 100% sure but could you give this a try please?

diff --git a/arch/x86/mm/numa.c b/arch/x86/mm/numa.c
index 1308f5408bf7..4575ae4d5449 100644
--- a/arch/x86/mm/numa.c
+++ b/arch/x86/mm/numa.c
@@ -527,6 +527,19 @@ static void __init numa_clear_kernel_node_hotplug(void)
 	}
 }
 
+static void __init init_memory_less_node(int nid)
+{
+	unsigned long zones_size[MAX_NR_ZONES] = {0};
+	unsigned long zholes_size[MAX_NR_ZONES] = {0};
+
+	free_area_init_node(nid, zones_size, 0, zholes_size);
+
+	/*
+	 * All zonelists will be built later in start_kernel() after per cpu
+	 * areas are initialized.
+	 */
+}
+
 static int __init numa_register_memblks(struct numa_meminfo *mi)
 {
 	unsigned long uninitialized_var(pfn_align);
@@ -592,6 +605,8 @@ static int __init numa_register_memblks(struct numa_meminfo *mi)
 			continue;
 
 		alloc_node_data(nid);
+		if (!end)
+			init_memory_less_node(nid);
 	}
 
 	/* Dump memblock with node info and return. */
@@ -721,21 +736,6 @@ void __init x86_numa_init(void)
 	numa_init(dummy_numa_init);
 }
 
-static void __init init_memory_less_node(int nid)
-{
-	unsigned long zones_size[MAX_NR_ZONES] = {0};
-	unsigned long zholes_size[MAX_NR_ZONES] = {0};
-
-	/* Allocate and initialize node data. Memory-less node is now online.*/
-	alloc_node_data(nid);
-	free_area_init_node(nid, zones_size, 0, zholes_size);
-
-	/*
-	 * All zonelists will be built later in start_kernel() after per cpu
-	 * areas are initialized.
-	 */
-}
-
 /*
  * Setup early cpu_to_node.
  *
@@ -763,9 +763,6 @@ void __init init_cpu_to_node(void)
 		if (node == NUMA_NO_NODE)
 			continue;
 
-		if (!node_online(node))
-			init_memory_less_node(node);
-
 		numa_set_node(cpu, node);
 	}
 }
-- 
Michal Hocko
SUSE Labs