Re: [RFC PATCH] x86, numa: always initialize all possible nodes
On Mon 11-02-19 14:49:09, Ingo Molnar wrote: > > * Michal Hocko wrote: > > > On Thu 24-01-19 11:10:50, Dave Hansen wrote: > > > On 1/24/19 6:17 AM, Michal Hocko wrote: > > > > and nr_cpus set to 4. The underlying reason is tha the device is bound > > > > to node 2 which doesn't have any memory and init_cpu_to_node only > > > > initializes memory-less nodes for possible cpus which nr_cpus restrics. > > > > This in turn means that proper zonelists are not allocated and the page > > > > allocator blows up. > > > > > > This looks OK to me. > > > > > > Could we add a few DEBUG_VM checks that *look* for these invalid > > > zonelists? Or, would our existing list debugging have caught this? > > > > Currently we simply blow up because those zonelists are NULL. I do not > > think we have a way to check whether an existing zonelist is actually > > _correct_ other thatn check it for NULL. But what would we do in the > > later case? > > > > > Basically, is this bug also a sign that we need better debugging around > > > this? > > > > My earlier patch had a debugging printk to display the zonelists and > > that might be worthwhile I guess. Basically something like this > > > > diff --git a/mm/page_alloc.c b/mm/page_alloc.c > > index 2e097f336126..c30d59f803fb 100644 > > --- a/mm/page_alloc.c > > +++ b/mm/page_alloc.c > > @@ -5259,6 +5259,11 @@ static void build_zonelists(pg_data_t *pgdat) > > > > build_zonelists_in_node_order(pgdat, node_order, nr_nodes); > > build_thisnode_zonelists(pgdat); > > + > > + pr_info("node[%d] zonelist: ", pgdat->node_id); > > + for_each_zone_zonelist(zone, z, > > >node_zonelists[ZONELIST_FALLBACK], MAX_NR_ZONES-1) > > + pr_cont("%d:%s ", zone_to_nid(zone), zone->name); > > + pr_cont("\n"); > > } > > Looks like this patch fell through the cracks - any update on this? I was waiting for some feedback. As there were no complains about the above debugging output I will make it a separate patch and post both patches later this week. I just have to go through my backlog pile after vacation. -- Michal Hocko SUSE Labs
Re: [RFC PATCH] x86, numa: always initialize all possible nodes
* Michal Hocko wrote: > On Thu 24-01-19 11:10:50, Dave Hansen wrote: > > On 1/24/19 6:17 AM, Michal Hocko wrote: > > > and nr_cpus set to 4. The underlying reason is tha the device is bound > > > to node 2 which doesn't have any memory and init_cpu_to_node only > > > initializes memory-less nodes for possible cpus which nr_cpus restrics. > > > This in turn means that proper zonelists are not allocated and the page > > > allocator blows up. > > > > This looks OK to me. > > > > Could we add a few DEBUG_VM checks that *look* for these invalid > > zonelists? Or, would our existing list debugging have caught this? > > Currently we simply blow up because those zonelists are NULL. I do not > think we have a way to check whether an existing zonelist is actually > _correct_ other thatn check it for NULL. But what would we do in the > later case? > > > Basically, is this bug also a sign that we need better debugging around > > this? > > My earlier patch had a debugging printk to display the zonelists and > that might be worthwhile I guess. Basically something like this > > diff --git a/mm/page_alloc.c b/mm/page_alloc.c > index 2e097f336126..c30d59f803fb 100644 > --- a/mm/page_alloc.c > +++ b/mm/page_alloc.c > @@ -5259,6 +5259,11 @@ static void build_zonelists(pg_data_t *pgdat) > > build_zonelists_in_node_order(pgdat, node_order, nr_nodes); > build_thisnode_zonelists(pgdat); > + > + pr_info("node[%d] zonelist: ", pgdat->node_id); > + for_each_zone_zonelist(zone, z, > >node_zonelists[ZONELIST_FALLBACK], MAX_NR_ZONES-1) > + pr_cont("%d:%s ", zone_to_nid(zone), zone->name); > + pr_cont("\n"); > } Looks like this patch fell through the cracks - any update on this? Thanks, Ingo
Re: [RFC PATCH] x86, numa: always initialize all possible nodes
On Fri, Jan 25, 2019 at 11:40:23AM +0100, Michal Hocko wrote: > On Thu 24-01-19 19:51:44, Mike Rapoport wrote: > > On Thu, Jan 24, 2019 at 03:17:27PM +0100, Michal Hocko wrote: > > > a friendly ping for this. Does anybody see any problem with this > > > approach? > > > > FWIW, it looks fine to me. > > > > It'd just be nice to have a few more words in the changelog about *how* the > > x86 init was reworked ;-) > > Heh, I thought it was there but nope... It probably just existed in my > head. Sorry about that. What about the following paragraphs added? > " > The new code relies on the arch specific initialization to allocate all > possible NUMA nodes (including memory less) - numa_register_memblks in > this case. Generic code then initializes both zonelists > (__build_all_zonelists) > and allocator internals (free_area_init_nodes) for all non-null pgdats > rather than online ones. > > For the x86 specific part also do not make new node online in alloc_node_data > because this is too early to know that. numa_register_memblks knows that > a node has some memory so it can make the node online appropriately. > init_memory_less_node hack can be safely removed altogether now. > " LGTM, thanks! > -- > Michal Hocko > SUSE Labs > -- Sincerely yours, Mike.
Re: [RFC PATCH] x86, numa: always initialize all possible nodes
On Thu 24-01-19 11:10:50, Dave Hansen wrote: > On 1/24/19 6:17 AM, Michal Hocko wrote: > > and nr_cpus set to 4. The underlying reason is tha the device is bound > > to node 2 which doesn't have any memory and init_cpu_to_node only > > initializes memory-less nodes for possible cpus which nr_cpus restrics. > > This in turn means that proper zonelists are not allocated and the page > > allocator blows up. > > This looks OK to me. > > Could we add a few DEBUG_VM checks that *look* for these invalid > zonelists? Or, would our existing list debugging have caught this? Currently we simply blow up because those zonelists are NULL. I do not think we have a way to check whether an existing zonelist is actually _correct_ other thatn check it for NULL. But what would we do in the later case? > Basically, is this bug also a sign that we need better debugging around > this? My earlier patch had a debugging printk to display the zonelists and that might be worthwhile I guess. Basically something like this diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 2e097f336126..c30d59f803fb 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -5259,6 +5259,11 @@ static void build_zonelists(pg_data_t *pgdat) build_zonelists_in_node_order(pgdat, node_order, nr_nodes); build_thisnode_zonelists(pgdat); + + pr_info("node[%d] zonelist: ", pgdat->node_id); + for_each_zone_zonelist(zone, z, >node_zonelists[ZONELIST_FALLBACK], MAX_NR_ZONES-1) + pr_cont("%d:%s ", zone_to_nid(zone), zone->name); + pr_cont("\n"); } #ifdef CONFIG_HAVE_MEMORYLESS_NODES -- Michal Hocko SUSE Labs
Re: [RFC PATCH] x86, numa: always initialize all possible nodes
On Thu 24-01-19 19:51:44, Mike Rapoport wrote: > On Thu, Jan 24, 2019 at 03:17:27PM +0100, Michal Hocko wrote: > > a friendly ping for this. Does anybody see any problem with this > > approach? > > FWIW, it looks fine to me. > > It'd just be nice to have a few more words in the changelog about *how* the > x86 init was reworked ;-) Heh, I thought it was there but nope... It probably just existed in my head. Sorry about that. What about the following paragraphs added? " The new code relies on the arch specific initialization to allocate all possible NUMA nodes (including memory less) - numa_register_memblks in this case. Generic code then initializes both zonelists (__build_all_zonelists) and allocator internals (free_area_init_nodes) for all non-null pgdats rather than online ones. For the x86 specific part also do not make new node online in alloc_node_data because this is too early to know that. numa_register_memblks knows that a node has some memory so it can make the node online appropriately. init_memory_less_node hack can be safely removed altogether now. " -- Michal Hocko SUSE Labs
Re: [RFC PATCH] x86, numa: always initialize all possible nodes
On 1/24/19 6:17 AM, Michal Hocko wrote: > and nr_cpus set to 4. The underlying reason is tha the device is bound > to node 2 which doesn't have any memory and init_cpu_to_node only > initializes memory-less nodes for possible cpus which nr_cpus restrics. > This in turn means that proper zonelists are not allocated and the page > allocator blows up. This looks OK to me. Could we add a few DEBUG_VM checks that *look* for these invalid zonelists? Or, would our existing list debugging have caught this? Basically, is this bug also a sign that we need better debugging around this?
Re: [RFC PATCH] x86, numa: always initialize all possible nodes
On Thu, Jan 24, 2019 at 03:17:27PM +0100, Michal Hocko wrote: > a friendly ping for this. Does anybody see any problem with this > approach? FWIW, it looks fine to me. It'd just be nice to have a few more words in the changelog about *how* the x86 init was reworked ;-) > On Mon 14-01-19 09:24:16, Michal Hocko wrote: > > From: Michal Hocko > > > > Pingfan Liu has reported the following splat > > [5.772742] BUG: unable to handle kernel paging request at > > 2088 > > [5.773618] PGD 0 P4D 0 > > [5.773618] Oops: [#1] SMP NOPTI > > [5.773618] CPU: 2 PID: 1 Comm: swapper/0 Not tainted 4.20.0-rc1+ #3 > > [5.773618] Hardware name: Dell Inc. PowerEdge R7425/02MJ3T, BIOS 1.4.3 > > 06/29/2018 > > [5.773618] RIP: 0010:__alloc_pages_nodemask+0xe2/0x2a0 > > [5.773618] Code: 00 00 44 89 ea 80 ca 80 41 83 f8 01 44 0f 44 ea 89 da > > c1 ea 08 83 e2 01 88 54 24 20 48 8b 54 24 08 48 85 d2 0f 85 46 01 00 00 > > <3b> 77 08 0f 82 3d 01 00 00 48 89 f8 44 89 ea 48 89 > > e1 44 89 e6 89 > > [5.773618] RSP: 0018:aa65fb20 EFLAGS: 00010246 > > [5.773618] RAX: RBX: 006012c0 RCX: > > > > [5.773618] RDX: RSI: 0002 RDI: > > 2080 > > [5.773618] RBP: 006012c0 R08: R09: > > 0002 > > [5.773618] R10: 006080c0 R11: 0002 R12: > > > > [5.773618] R13: 0001 R14: R15: > > 0002 > > [5.773618] FS: () GS:8c69afe0() > > knlGS: > > [5.773618] CS: 0010 DS: ES: CR0: 80050033 > > [5.773618] CR2: 2088 CR3: 00087e00a000 CR4: > > 003406e0 > > [5.773618] Call Trace: > > [5.773618] new_slab+0xa9/0x570 > > [5.773618] ___slab_alloc+0x375/0x540 > > [5.773618] ? pinctrl_bind_pins+0x2b/0x2a0 > > [5.773618] __slab_alloc+0x1c/0x38 > > [5.773618] __kmalloc_node_track_caller+0xc8/0x270 > > [5.773618] ? pinctrl_bind_pins+0x2b/0x2a0 > > [5.773618] devm_kmalloc+0x28/0x60 > > [5.773618] pinctrl_bind_pins+0x2b/0x2a0 > > [5.773618] really_probe+0x73/0x420 > > [5.773618] driver_probe_device+0x115/0x130 > > [5.773618] __driver_attach+0x103/0x110 > > [5.773618] ? driver_probe_device+0x130/0x130 > > [5.773618] bus_for_each_dev+0x67/0xc0 > > [5.773618] ? klist_add_tail+0x3b/0x70 > > [5.773618] bus_add_driver+0x41/0x260 > > [5.773618] ? pcie_port_setup+0x4d/0x4d > > [5.773618] driver_register+0x5b/0xe0 > > [5.773618] ? pcie_port_setup+0x4d/0x4d > > [5.773618] do_one_initcall+0x4e/0x1d4 > > [5.773618] ? init_setup+0x25/0x28 > > [5.773618] kernel_init_freeable+0x1c1/0x26e > > [5.773618] ? loglevel+0x5b/0x5b > > [5.773618] ? rest_init+0xb0/0xb0 > > [5.773618] kernel_init+0xa/0x110 > > [5.773618] ret_from_fork+0x22/0x40 > > [5.773618] Modules linked in: > > [5.773618] CR2: 2088 > > [5.773618] ---[ end trace 1030c9120a03d081 ]--- > > > > with his AMD machine with the following topology > > NUMA node0 CPU(s): 0,8,16,24 > > NUMA node1 CPU(s): 2,10,18,26 > > NUMA node2 CPU(s): 4,12,20,28 > > NUMA node3 CPU(s): 6,14,22,30 > > NUMA node4 CPU(s): 1,9,17,25 > > NUMA node5 CPU(s): 3,11,19,27 > > NUMA node6 CPU(s): 5,13,21,29 > > NUMA node7 CPU(s): 7,15,23,31 > > > > [0.007418] Early memory node ranges > > [0.007419] node 1: [mem 0x1000-0x0008efff] > > [0.007420] node 1: [mem 0x0009-0x0009] > > [0.007422] node 1: [mem 0x0010-0x5c3d6fff] > > [0.007422] node 1: [mem 0x643df000-0x68ff7fff] > > [0.007423] node 1: [mem 0x6c528000-0x6fff] > > [0.007424] node 1: [mem 0x0001-0x00047fff] > > [0.007425] node 5: [mem 0x00048000-0x00087eff] > > > > and nr_cpus set to 4. The underlying reason is tha the device is bound > > to node 2 which doesn't have any memory and init_cpu_to_node only > > initializes memory-less nodes for possible cpus which nr_cpus restrics. > > This in turn means that proper zonelists are not allocated and the page > > allocator blows up. > > > > Fix the issue by reworking how x86 initializes the memory less nodes. > > The current implementation is hacked into the workflow and it doesn't > > allow any flexibility. There is init_memory_less_node called for each > > offline node that has a CPU as already mentioned above. This will make > > sure that we will have a new online node without any memory. Much later > > on we build a zone list for this node and things seem to work, except > > they do not (e.g. due to nr_cpus). Not to mention that it doesn't really > > make much sense to consider an empty node as
Re: [RFC PATCH] x86, numa: always initialize all possible nodes
a friendly ping for this. Does anybody see any problem with this approach? On Mon 14-01-19 09:24:16, Michal Hocko wrote: > From: Michal Hocko > > Pingfan Liu has reported the following splat > [5.772742] BUG: unable to handle kernel paging request at 2088 > [5.773618] PGD 0 P4D 0 > [5.773618] Oops: [#1] SMP NOPTI > [5.773618] CPU: 2 PID: 1 Comm: swapper/0 Not tainted 4.20.0-rc1+ #3 > [5.773618] Hardware name: Dell Inc. PowerEdge R7425/02MJ3T, BIOS 1.4.3 > 06/29/2018 > [5.773618] RIP: 0010:__alloc_pages_nodemask+0xe2/0x2a0 > [5.773618] Code: 00 00 44 89 ea 80 ca 80 41 83 f8 01 44 0f 44 ea 89 da c1 > ea 08 83 e2 01 88 54 24 20 48 8b 54 24 08 48 85 d2 0f 85 46 01 00 00 <3b> 77 > 08 0f 82 3d 01 00 00 48 89 f8 44 89 ea 48 89 > e1 44 89 e6 89 > [5.773618] RSP: 0018:aa65fb20 EFLAGS: 00010246 > [5.773618] RAX: RBX: 006012c0 RCX: > > [5.773618] RDX: RSI: 0002 RDI: > 2080 > [5.773618] RBP: 006012c0 R08: R09: > 0002 > [5.773618] R10: 006080c0 R11: 0002 R12: > > [5.773618] R13: 0001 R14: R15: > 0002 > [5.773618] FS: () GS:8c69afe0() > knlGS: > [5.773618] CS: 0010 DS: ES: CR0: 80050033 > [5.773618] CR2: 2088 CR3: 00087e00a000 CR4: > 003406e0 > [5.773618] Call Trace: > [5.773618] new_slab+0xa9/0x570 > [5.773618] ___slab_alloc+0x375/0x540 > [5.773618] ? pinctrl_bind_pins+0x2b/0x2a0 > [5.773618] __slab_alloc+0x1c/0x38 > [5.773618] __kmalloc_node_track_caller+0xc8/0x270 > [5.773618] ? pinctrl_bind_pins+0x2b/0x2a0 > [5.773618] devm_kmalloc+0x28/0x60 > [5.773618] pinctrl_bind_pins+0x2b/0x2a0 > [5.773618] really_probe+0x73/0x420 > [5.773618] driver_probe_device+0x115/0x130 > [5.773618] __driver_attach+0x103/0x110 > [5.773618] ? driver_probe_device+0x130/0x130 > [5.773618] bus_for_each_dev+0x67/0xc0 > [5.773618] ? klist_add_tail+0x3b/0x70 > [5.773618] bus_add_driver+0x41/0x260 > [5.773618] ? pcie_port_setup+0x4d/0x4d > [5.773618] driver_register+0x5b/0xe0 > [5.773618] ? pcie_port_setup+0x4d/0x4d > [5.773618] do_one_initcall+0x4e/0x1d4 > [5.773618] ? init_setup+0x25/0x28 > [5.773618] kernel_init_freeable+0x1c1/0x26e > [5.773618] ? loglevel+0x5b/0x5b > [5.773618] ? rest_init+0xb0/0xb0 > [5.773618] kernel_init+0xa/0x110 > [5.773618] ret_from_fork+0x22/0x40 > [5.773618] Modules linked in: > [5.773618] CR2: 2088 > [5.773618] ---[ end trace 1030c9120a03d081 ]--- > > with his AMD machine with the following topology > NUMA node0 CPU(s): 0,8,16,24 > NUMA node1 CPU(s): 2,10,18,26 > NUMA node2 CPU(s): 4,12,20,28 > NUMA node3 CPU(s): 6,14,22,30 > NUMA node4 CPU(s): 1,9,17,25 > NUMA node5 CPU(s): 3,11,19,27 > NUMA node6 CPU(s): 5,13,21,29 > NUMA node7 CPU(s): 7,15,23,31 > > [0.007418] Early memory node ranges > [0.007419] node 1: [mem 0x1000-0x0008efff] > [0.007420] node 1: [mem 0x0009-0x0009] > [0.007422] node 1: [mem 0x0010-0x5c3d6fff] > [0.007422] node 1: [mem 0x643df000-0x68ff7fff] > [0.007423] node 1: [mem 0x6c528000-0x6fff] > [0.007424] node 1: [mem 0x0001-0x00047fff] > [0.007425] node 5: [mem 0x00048000-0x00087eff] > > and nr_cpus set to 4. The underlying reason is tha the device is bound > to node 2 which doesn't have any memory and init_cpu_to_node only > initializes memory-less nodes for possible cpus which nr_cpus restrics. > This in turn means that proper zonelists are not allocated and the page > allocator blows up. > > Fix the issue by reworking how x86 initializes the memory less nodes. > The current implementation is hacked into the workflow and it doesn't > allow any flexibility. There is init_memory_less_node called for each > offline node that has a CPU as already mentioned above. This will make > sure that we will have a new online node without any memory. Much later > on we build a zone list for this node and things seem to work, except > they do not (e.g. due to nr_cpus). Not to mention that it doesn't really > make much sense to consider an empty node as online because we just > consider this node whenever we want to iterate nodes to use and empty > node is obviously not the best candidate. This is all just too fragile. > > Reported-by: Pingfan Liu > Tested-by: Pingfan Liu > Signed-off-by: Michal Hocko > --- > > Hi, > I am sending this as an RFC because I am not sure this is the proper way > to go myself. I am especially not sure
Re: [RFC PATCH] x86, numa: always initialize all possible nodes
[...] > > > > I would appreciate a help with those architectures because I couldn't > > really grasp how the memoryless nodes are really initialized there. E.g. > > ppc only seem to call setup_node_data for online nodes but I couldn't > > find any special treatment for nodes without any memory. > > We have a somewhat dubious hack in our hotplug code, see: > > e67e02a544e9 ("powerpc/pseries: Fix cpu hotplug crash with memoryless nodes") > > Which basically onlines the node when we hotplug a CPU into it. > This bug should be related with the present state of numa node during boot time. On PowerNV and PSeries, the boot code seems not to bring up all nodes if memoryless. Then it can not avoid this bug. Thanks, Pingfan
Re: [RFC PATCH] x86, numa: always initialize all possible nodes
On Mon 14-01-19 21:26:39, Michael Ellerman wrote: > Michal Hocko writes: > > > From: Michal Hocko > > > > Pingfan Liu has reported the following splat > > [5.772742] BUG: unable to handle kernel paging request at > > 2088 > > [5.773618] PGD 0 P4D 0 > > [5.773618] Oops: [#1] SMP NOPTI > > [5.773618] CPU: 2 PID: 1 Comm: swapper/0 Not tainted 4.20.0-rc1+ #3 > > [5.773618] Hardware name: Dell Inc. PowerEdge R7425/02MJ3T, BIOS 1.4.3 > > 06/29/2018 > > [5.773618] RIP: 0010:__alloc_pages_nodemask+0xe2/0x2a0 > > [5.773618] Code: 00 00 44 89 ea 80 ca 80 41 83 f8 01 44 0f 44 ea 89 da > > c1 ea 08 83 e2 01 88 54 24 20 48 8b 54 24 08 48 85 d2 0f 85 46 01 00 00 > > <3b> 77 08 0f 82 3d 01 00 00 48 89 f8 44 89 ea 48 89 > > e1 44 89 e6 89 > > [5.773618] RSP: 0018:aa65fb20 EFLAGS: 00010246 > > [5.773618] RAX: RBX: 006012c0 RCX: > > > > [5.773618] RDX: RSI: 0002 RDI: > > 2080 > > [5.773618] RBP: 006012c0 R08: R09: > > 0002 > > [5.773618] R10: 006080c0 R11: 0002 R12: > > > > [5.773618] R13: 0001 R14: R15: > > 0002 > > [5.773618] FS: () GS:8c69afe0() > > knlGS: > > [5.773618] CS: 0010 DS: ES: CR0: 80050033 > > [5.773618] CR2: 2088 CR3: 00087e00a000 CR4: > > 003406e0 > > [5.773618] Call Trace: > > [5.773618] new_slab+0xa9/0x570 > > [5.773618] ___slab_alloc+0x375/0x540 > > [5.773618] ? pinctrl_bind_pins+0x2b/0x2a0 > > [5.773618] __slab_alloc+0x1c/0x38 > > [5.773618] __kmalloc_node_track_caller+0xc8/0x270 > > [5.773618] ? pinctrl_bind_pins+0x2b/0x2a0 > > [5.773618] devm_kmalloc+0x28/0x60 > > [5.773618] pinctrl_bind_pins+0x2b/0x2a0 > > [5.773618] really_probe+0x73/0x420 > > [5.773618] driver_probe_device+0x115/0x130 > > [5.773618] __driver_attach+0x103/0x110 > > [5.773618] ? driver_probe_device+0x130/0x130 > > [5.773618] bus_for_each_dev+0x67/0xc0 > > [5.773618] ? klist_add_tail+0x3b/0x70 > > [5.773618] bus_add_driver+0x41/0x260 > > [5.773618] ? pcie_port_setup+0x4d/0x4d > > [5.773618] driver_register+0x5b/0xe0 > > [5.773618] ? pcie_port_setup+0x4d/0x4d > > [5.773618] do_one_initcall+0x4e/0x1d4 > > [5.773618] ? init_setup+0x25/0x28 > > [5.773618] kernel_init_freeable+0x1c1/0x26e > > [5.773618] ? loglevel+0x5b/0x5b > > [5.773618] ? rest_init+0xb0/0xb0 > > [5.773618] kernel_init+0xa/0x110 > > [5.773618] ret_from_fork+0x22/0x40 > > [5.773618] Modules linked in: > > [5.773618] CR2: 2088 > > [5.773618] ---[ end trace 1030c9120a03d081 ]--- > > > > with his AMD machine with the following topology > > NUMA node0 CPU(s): 0,8,16,24 > > NUMA node1 CPU(s): 2,10,18,26 > > NUMA node2 CPU(s): 4,12,20,28 > > NUMA node3 CPU(s): 6,14,22,30 > > NUMA node4 CPU(s): 1,9,17,25 > > NUMA node5 CPU(s): 3,11,19,27 > > NUMA node6 CPU(s): 5,13,21,29 > > NUMA node7 CPU(s): 7,15,23,31 > > > > [0.007418] Early memory node ranges > > [0.007419] node 1: [mem 0x1000-0x0008efff] > > [0.007420] node 1: [mem 0x0009-0x0009] > > [0.007422] node 1: [mem 0x0010-0x5c3d6fff] > > [0.007422] node 1: [mem 0x643df000-0x68ff7fff] > > [0.007423] node 1: [mem 0x6c528000-0x6fff] > > [0.007424] node 1: [mem 0x0001-0x00047fff] > > [0.007425] node 5: [mem 0x00048000-0x00087eff] > > > > and nr_cpus set to 4. The underlying reason is tha the device is bound > > to node 2 which doesn't have any memory and init_cpu_to_node only > > initializes memory-less nodes for possible cpus which nr_cpus restrics. > > This in turn means that proper zonelists are not allocated and the page > > allocator blows up. > > > > Fix the issue by reworking how x86 initializes the memory less nodes. > > The current implementation is hacked into the workflow and it doesn't > > allow any flexibility. There is init_memory_less_node called for each > > offline node that has a CPU as already mentioned above. This will make > > sure that we will have a new online node without any memory. Much later > > on we build a zone list for this node and things seem to work, except > > they do not (e.g. due to nr_cpus). Not to mention that it doesn't really > > make much sense to consider an empty node as online because we just > > consider this node whenever we want to iterate nodes to use and empty > > node is obviously not the best candidate. This is all just too fragile. > > > > Reported-by: Pingfan Liu > > Tested-by: Pingfan Liu > >
Re: [RFC PATCH] x86, numa: always initialize all possible nodes
Michal Hocko writes: > From: Michal Hocko > > Pingfan Liu has reported the following splat > [5.772742] BUG: unable to handle kernel paging request at 2088 > [5.773618] PGD 0 P4D 0 > [5.773618] Oops: [#1] SMP NOPTI > [5.773618] CPU: 2 PID: 1 Comm: swapper/0 Not tainted 4.20.0-rc1+ #3 > [5.773618] Hardware name: Dell Inc. PowerEdge R7425/02MJ3T, BIOS 1.4.3 > 06/29/2018 > [5.773618] RIP: 0010:__alloc_pages_nodemask+0xe2/0x2a0 > [5.773618] Code: 00 00 44 89 ea 80 ca 80 41 83 f8 01 44 0f 44 ea 89 da c1 > ea 08 83 e2 01 88 54 24 20 48 8b 54 24 08 48 85 d2 0f 85 46 01 00 00 <3b> 77 > 08 0f 82 3d 01 00 00 48 89 f8 44 89 ea 48 89 > e1 44 89 e6 89 > [5.773618] RSP: 0018:aa65fb20 EFLAGS: 00010246 > [5.773618] RAX: RBX: 006012c0 RCX: > > [5.773618] RDX: RSI: 0002 RDI: > 2080 > [5.773618] RBP: 006012c0 R08: R09: > 0002 > [5.773618] R10: 006080c0 R11: 0002 R12: > > [5.773618] R13: 0001 R14: R15: > 0002 > [5.773618] FS: () GS:8c69afe0() > knlGS: > [5.773618] CS: 0010 DS: ES: CR0: 80050033 > [5.773618] CR2: 2088 CR3: 00087e00a000 CR4: > 003406e0 > [5.773618] Call Trace: > [5.773618] new_slab+0xa9/0x570 > [5.773618] ___slab_alloc+0x375/0x540 > [5.773618] ? pinctrl_bind_pins+0x2b/0x2a0 > [5.773618] __slab_alloc+0x1c/0x38 > [5.773618] __kmalloc_node_track_caller+0xc8/0x270 > [5.773618] ? pinctrl_bind_pins+0x2b/0x2a0 > [5.773618] devm_kmalloc+0x28/0x60 > [5.773618] pinctrl_bind_pins+0x2b/0x2a0 > [5.773618] really_probe+0x73/0x420 > [5.773618] driver_probe_device+0x115/0x130 > [5.773618] __driver_attach+0x103/0x110 > [5.773618] ? driver_probe_device+0x130/0x130 > [5.773618] bus_for_each_dev+0x67/0xc0 > [5.773618] ? klist_add_tail+0x3b/0x70 > [5.773618] bus_add_driver+0x41/0x260 > [5.773618] ? pcie_port_setup+0x4d/0x4d > [5.773618] driver_register+0x5b/0xe0 > [5.773618] ? pcie_port_setup+0x4d/0x4d > [5.773618] do_one_initcall+0x4e/0x1d4 > [5.773618] ? init_setup+0x25/0x28 > [5.773618] kernel_init_freeable+0x1c1/0x26e > [5.773618] ? loglevel+0x5b/0x5b > [5.773618] ? rest_init+0xb0/0xb0 > [5.773618] kernel_init+0xa/0x110 > [5.773618] ret_from_fork+0x22/0x40 > [5.773618] Modules linked in: > [5.773618] CR2: 2088 > [5.773618] ---[ end trace 1030c9120a03d081 ]--- > > with his AMD machine with the following topology > NUMA node0 CPU(s): 0,8,16,24 > NUMA node1 CPU(s): 2,10,18,26 > NUMA node2 CPU(s): 4,12,20,28 > NUMA node3 CPU(s): 6,14,22,30 > NUMA node4 CPU(s): 1,9,17,25 > NUMA node5 CPU(s): 3,11,19,27 > NUMA node6 CPU(s): 5,13,21,29 > NUMA node7 CPU(s): 7,15,23,31 > > [0.007418] Early memory node ranges > [0.007419] node 1: [mem 0x1000-0x0008efff] > [0.007420] node 1: [mem 0x0009-0x0009] > [0.007422] node 1: [mem 0x0010-0x5c3d6fff] > [0.007422] node 1: [mem 0x643df000-0x68ff7fff] > [0.007423] node 1: [mem 0x6c528000-0x6fff] > [0.007424] node 1: [mem 0x0001-0x00047fff] > [0.007425] node 5: [mem 0x00048000-0x00087eff] > > and nr_cpus set to 4. The underlying reason is tha the device is bound > to node 2 which doesn't have any memory and init_cpu_to_node only > initializes memory-less nodes for possible cpus which nr_cpus restrics. > This in turn means that proper zonelists are not allocated and the page > allocator blows up. > > Fix the issue by reworking how x86 initializes the memory less nodes. > The current implementation is hacked into the workflow and it doesn't > allow any flexibility. There is init_memory_less_node called for each > offline node that has a CPU as already mentioned above. This will make > sure that we will have a new online node without any memory. Much later > on we build a zone list for this node and things seem to work, except > they do not (e.g. due to nr_cpus). Not to mention that it doesn't really > make much sense to consider an empty node as online because we just > consider this node whenever we want to iterate nodes to use and empty > node is obviously not the best candidate. This is all just too fragile. > > Reported-by: Pingfan Liu > Tested-by: Pingfan Liu > Signed-off-by: Michal Hocko > --- > > Hi, > I am sending this as an RFC because I am not sure this is the proper way > to go myself. I am especially not sure about other architectures > supporting memoryless nodes (ppc and ia64 AFAICS or are there more?). > > I would
[RFC PATCH] x86, numa: always initialize all possible nodes
From: Michal Hocko Pingfan Liu has reported the following splat [5.772742] BUG: unable to handle kernel paging request at 2088 [5.773618] PGD 0 P4D 0 [5.773618] Oops: [#1] SMP NOPTI [5.773618] CPU: 2 PID: 1 Comm: swapper/0 Not tainted 4.20.0-rc1+ #3 [5.773618] Hardware name: Dell Inc. PowerEdge R7425/02MJ3T, BIOS 1.4.3 06/29/2018 [5.773618] RIP: 0010:__alloc_pages_nodemask+0xe2/0x2a0 [5.773618] Code: 00 00 44 89 ea 80 ca 80 41 83 f8 01 44 0f 44 ea 89 da c1 ea 08 83 e2 01 88 54 24 20 48 8b 54 24 08 48 85 d2 0f 85 46 01 00 00 <3b> 77 08 0f 82 3d 01 00 00 48 89 f8 44 89 ea 48 89 e1 44 89 e6 89 [5.773618] RSP: 0018:aa65fb20 EFLAGS: 00010246 [5.773618] RAX: RBX: 006012c0 RCX: [5.773618] RDX: RSI: 0002 RDI: 2080 [5.773618] RBP: 006012c0 R08: R09: 0002 [5.773618] R10: 006080c0 R11: 0002 R12: [5.773618] R13: 0001 R14: R15: 0002 [5.773618] FS: () GS:8c69afe0() knlGS: [5.773618] CS: 0010 DS: ES: CR0: 80050033 [5.773618] CR2: 2088 CR3: 00087e00a000 CR4: 003406e0 [5.773618] Call Trace: [5.773618] new_slab+0xa9/0x570 [5.773618] ___slab_alloc+0x375/0x540 [5.773618] ? pinctrl_bind_pins+0x2b/0x2a0 [5.773618] __slab_alloc+0x1c/0x38 [5.773618] __kmalloc_node_track_caller+0xc8/0x270 [5.773618] ? pinctrl_bind_pins+0x2b/0x2a0 [5.773618] devm_kmalloc+0x28/0x60 [5.773618] pinctrl_bind_pins+0x2b/0x2a0 [5.773618] really_probe+0x73/0x420 [5.773618] driver_probe_device+0x115/0x130 [5.773618] __driver_attach+0x103/0x110 [5.773618] ? driver_probe_device+0x130/0x130 [5.773618] bus_for_each_dev+0x67/0xc0 [5.773618] ? klist_add_tail+0x3b/0x70 [5.773618] bus_add_driver+0x41/0x260 [5.773618] ? pcie_port_setup+0x4d/0x4d [5.773618] driver_register+0x5b/0xe0 [5.773618] ? pcie_port_setup+0x4d/0x4d [5.773618] do_one_initcall+0x4e/0x1d4 [5.773618] ? init_setup+0x25/0x28 [5.773618] kernel_init_freeable+0x1c1/0x26e [5.773618] ? loglevel+0x5b/0x5b [5.773618] ? rest_init+0xb0/0xb0 [5.773618] kernel_init+0xa/0x110 [5.773618] ret_from_fork+0x22/0x40 [5.773618] Modules linked in: [5.773618] CR2: 2088 [5.773618] ---[ end trace 1030c9120a03d081 ]--- with his AMD machine with the following topology NUMA node0 CPU(s): 0,8,16,24 NUMA node1 CPU(s): 2,10,18,26 NUMA node2 CPU(s): 4,12,20,28 NUMA node3 CPU(s): 6,14,22,30 NUMA node4 CPU(s): 1,9,17,25 NUMA node5 CPU(s): 3,11,19,27 NUMA node6 CPU(s): 5,13,21,29 NUMA node7 CPU(s): 7,15,23,31 [0.007418] Early memory node ranges [0.007419] node 1: [mem 0x1000-0x0008efff] [0.007420] node 1: [mem 0x0009-0x0009] [0.007422] node 1: [mem 0x0010-0x5c3d6fff] [0.007422] node 1: [mem 0x643df000-0x68ff7fff] [0.007423] node 1: [mem 0x6c528000-0x6fff] [0.007424] node 1: [mem 0x0001-0x00047fff] [0.007425] node 5: [mem 0x00048000-0x00087eff] and nr_cpus set to 4. The underlying reason is tha the device is bound to node 2 which doesn't have any memory and init_cpu_to_node only initializes memory-less nodes for possible cpus which nr_cpus restrics. This in turn means that proper zonelists are not allocated and the page allocator blows up. Fix the issue by reworking how x86 initializes the memory less nodes. The current implementation is hacked into the workflow and it doesn't allow any flexibility. There is init_memory_less_node called for each offline node that has a CPU as already mentioned above. This will make sure that we will have a new online node without any memory. Much later on we build a zone list for this node and things seem to work, except they do not (e.g. due to nr_cpus). Not to mention that it doesn't really make much sense to consider an empty node as online because we just consider this node whenever we want to iterate nodes to use and empty node is obviously not the best candidate. This is all just too fragile. Reported-by: Pingfan Liu Tested-by: Pingfan Liu Signed-off-by: Michal Hocko --- Hi, I am sending this as an RFC because I am not sure this is the proper way to go myself. I am especially not sure about other architectures supporting memoryless nodes (ppc and ia64 AFAICS or are there more?). I would appreciate a help with those architectures because I couldn't really grasp how the memoryless nodes are really initialized there. E.g. ppc only seem to call setup_node_data for online nodes but I couldn't find any special treatment