linux-kernel - Re: [PATCH -v2 0/9] mm: make movable onlining suck less

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: <20170411123801.GM6729@dhcp22.suse.cz>
Date:   Tue, 11 Apr 2017 14:38:01 +0200
From:   Michal Hocko <mhocko@...nel.org>
To:     Igor Mammedov <imammedo@...hat.com>
Cc:     linux-mm@...ck.org, Andrew Morton <akpm@...ux-foundation.org>,
        Mel Gorman <mgorman@...e.de>, Vlastimil Babka <vbabka@...e.cz>,
        Andrea Arcangeli <aarcange@...hat.com>,
        Jerome Glisse <jglisse@...hat.com>,
        Reza Arbab <arbab@...ux.vnet.ibm.com>,
        Yasuaki Ishimatsu <yasu.isimatu@...il.com>,
        qiuxishi@...wei.com, Kani Toshimitsu <toshi.kani@....com>,
        slaoub@...il.com, Joonsoo Kim <js1304@...il.com>,
        Andi Kleen <ak@...ux.intel.com>,
        David Rientjes <rientjes@...gle.com>,
        Daniel Kiper <daniel.kiper@...cle.com>,
        Vitaly Kuznetsov <vkuznets@...hat.com>,
        LKML <linux-kernel@...r.kernel.org>,
        Dan Williams <dan.j.williams@...il.com>,
        Heiko Carstens <heiko.carstens@...ibm.com>,
        Lai Jiangshan <laijs@...fujitsu.com>,
        Martin Schwidefsky <schwidefsky@...ibm.com>,
        Tobias Regnery <tobias.regnery@...il.com>
Subject: Re: [PATCH -v2 0/9] mm: make movable onlining suck less

On Tue 11-04-17 13:38:16, Michal Hocko wrote:
> On Tue 11-04-17 13:01:43, Michal Hocko wrote:
> > On Tue 11-04-17 11:59:31, Igor Mammedov wrote:
> > > On Tue, 11 Apr 2017 11:23:07 +0200
> > > Michal Hocko <mhocko@...nel.org> wrote:
> > > 
> > > > On Tue 11-04-17 08:38:34, Igor Mammedov wrote:
> > > > > for issue2:
> > > > > -enable-kvm -m 2G,slots=4,maxmem=4G -smp 4 -numa node -numa node \
> > > > > -drive if=virtio,file=disk.img -kernel bzImage -append 'root=/dev/vda1' \
> > > > > -object memory-backend-ram,id=mem1,size=256M -object memory-backend-ram,id=mem0,size=256M \
> > > > > -device pc-dimm,id=dimm1,memdev=mem1,slot=1,node=0 -device pc-dimm,id=dimm0,memdev=mem0,slot=0,node=1  
> > > > 
> > > > I must be doing something wrong here...
> > > > qemu-system-x86_64 -enable-kvm -monitor telnet:127.0.0.1:9999,server,nowait -net nic -net user,hostfwd=tcp:127.0.0.1:5555-:22 -serial file:test.qcow_serial.log -enable-kvm -m 2G,slots=4,maxmem=4G -smp 4 -numa node -numa node -object memory-backend-ram,id=mem1,size=256M -object memory-backend-ram,id=mem0,size=256M -device pc-dimm,id=dimm1,memdev=mem1,slot=1,node=0 -device pc-dimm,id=dimm0,memdev=mem0,slot=0,node=1 -drive file=test.qcow,if=ide,index=0
> > > > 
> > > > for i in $(seq 0 3)
> > > > do
> > > > 	sh probe_memblock.sh $i
> > > > done
> > >
> > > dimm to node mapping comes from ACPI subsystem (_PXM object in memory device),
> > > which adds memory blocks automatically on hotplug.
> > 
> > Hmm, memory_probe_store relies on memory_add_physaddr_to_nid which in
> > turn relies on numa_meminfo. I am not familiar with the intialization
> > and got lost in in the code rather quickly but I assumed this should get
> > the proper information from the ACPI subsystem. I will have to double
> > check.
> > 
> > > you probably don't have ACPI_HOTPLUG_MEMORY config option enabled.
> > 
> > Yes that is the case and enabling it made all 4 memblocks available
> > and associated with the proper node
> > # ls -l /sys/devices/system/memory/memory3?/node*
> > lrwxrwxrwx 1 root root 0 Apr 11 12:56 /sys/devices/system/memory/memory32/node0 -> ../../node/node0
> > lrwxrwxrwx 1 root root 0 Apr 11 12:56 /sys/devices/system/memory/memory33/node0 -> ../../node/node0
> > lrwxrwxrwx 1 root root 0 Apr 11 12:56 /sys/devices/system/memory/memory34/node1 -> ../../node/node1
> > lrwxrwxrwx 1 root root 0 Apr 11 12:56 /sys/devices/system/memory/memory35/node1 -> ../../node/node1
> > 
> > # grep . /sys/devices/system/memory/memory3?/valid_zones
> > /sys/devices/system/memory/memory32/valid_zones:Normal Movable
> > /sys/devices/system/memory/memory33/valid_zones:Normal Movable
> > /sys/devices/system/memory/memory34/valid_zones:Normal Movable
> > /sys/devices/system/memory/memory35/valid_zones:Normal Movable
> > 
> > I can even reproduce your problem
> > # echo online_movable > /sys/devices/system/memory/memory33/state
> > # echo online > /sys/devices/system/memory/memory32/state
> > # grep . /sys/devices/system/memory/memory3?/valid_zones
> > /sys/devices/system/memory/memory32/valid_zones:Movable
> > /sys/devices/system/memory/memory33/valid_zones:Movable
> > /sys/devices/system/memory/memory34/valid_zones:Normal Movable
> > /sys/devices/system/memory/memory35/valid_zones:Normal Movable
> > 
> > I will investigate this
> 
> Dang, guess what. It is a similar type bug I've fixed in
> show_valid_zones [1] already.
> 
> [1] http://lkml.kernel.org/r/20170410152228.GF4618@dhcp22.suse.cz
> ---
> diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
> index ec2f987ec549..410c7ccb74fb 100644
> --- a/include/linux/mmzone.h
> +++ b/include/linux/mmzone.h
> @@ -541,7 +541,7 @@ static inline bool zone_intersects(struct zone *zone,
>  {
>  	if (zone->zone_start_pfn <= start_pfn && start_pfn < zone_end_pfn(zone))
>  		return true;
> -	if (start_pfn + nr_pages > start_pfn && !zone_is_empty(zone))
> +	if (start_pfn + nr_pages > zone->zone_start_pfn && !zone_is_empty(zone))
>  		return true;
>  	return false;
>  }
> 
> I have decided to make it more readable and did zone_is_empty check
> first. Everything is in my git tree attempts/rewrite-mem_hotplug branch.
> I have to test it but I believe this is the culprit here.

OK, tested and it seems to be fixed. Thanks again for your testing and
the kvm configuration which made my testing much easier (probing and
adding areas from the qemu monitor was just PITA)!
-- 
Michal Hocko
SUSE Labs