Hi Michal,
> On Dec 8, 2021, at 12:54 AM, Michal Hocko <mhocko@xxxxxxxx> wrote:
>
> Alexey,
> this is still not finalized but it would really help if you could give
> it a spin on your setup. I still have to think about how to transition
> from a memoryless node to standard node (in hotplug code). Also there
> might be other surprises on the way.
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index c5952749ad40..8ed8db2ccb13 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -6382,7 +6382,11 @@ static void __build_all_zonelists(void *data)
> if (self && !node_online(self->node_id)) {
> build_zonelists(self);
> } else {
> - for_each_online_node(nid) {
> + /*
> + * All possible nodes have pgdat preallocated
> + * free_area_init
> + */
> + for_each_node(nid) {
> pg_data_t *pgdat = NODE_DATA(nid);
>
> build_zonelists(pgdat);
> @@ -8032,8 +8036,32 @@ void __init free_area_init(unsigned long *max_zone_pfn)
> /* Initialise every node */
> mminit_verify_pageflags_layout();
> setup_nr_node_ids();
> - for_each_online_node(nid) {
> - pg_data_t *pgdat = NODE_DATA(nid);
> + for_each_node(nid) {
> + pg_data_t *pgdat;
> +
> + if (!node_online(nid)) {
> + pr_warn("Node %d uninitialized by the platform. Please report with boot dmesg.\n", nid);
> +
> + /* Allocator not initialized yet */
> + pgdat = memblock_alloc(sizeof(*pgdat), SMP_CACHE_BYTES);
> + if (!pgdat) {
> + pr_err("Cannot allocate %zuB for node %d.\n",
> + sizeof(*pgdat), nid);
> + continue;
> + }
> + /* TODO do we need this for memoryless nodes */
> + pgdat->per_cpu_nodestats = alloc_percpu(struct per_cpu_nodestat);
> + arch_refresh_nodedata(nid, pgdat);
> + free_area_init_memoryless_node(nid);
> + /*
> + * not marking this node online because we do not want to
> + * confuse userspace by sysfs files/directories for node
> + * without any memory attached to it (see topology_init)
> + */
> + continue;
> + }
> +
> + pgdat = NODE_DATA(nid);
> free_area_init_node(nid);
>
> /* Any memory on that node */
After applying this patch, kernel panics in early boot with:
[ 0.081838] Initmem setup node 0 [mem 0x0000000000001000-0x000000007fffffff]
[ 0.081842] Initmem setup node 1 [mem 0x0000000080000000-0x000000013fffffff]
[ 0.081844] Node 2 uninitialized by the platform. Please report with boot dmesg.
[ 0.081877] BUG: kernel NULL pointer dereference, address: 0000000000000000
[ 0.081879] #PF: supervisor read access in kernel mode
[ 0.081882] #PF: error_code(0x0000) - not-present page
[ 0.081884] PGD 0 P4D 0
[ 0.081887] Oops: 0000 [#1] SMP PTI
[ 0.081890] CPU: 0 PID: 0 Comm: swapper Not tainted 5.15.0+ #33
[ 0.081893] Hardware name: VMware, Inc. VMware7,1/440BX Desktop Reference Platform, BIOS VMW
[ 0.081896] RIP: 0010:pcpu_alloc+0x330/0x850
[ 0.081903] Code: c7 c7 e4 38 5b 82 e8 5f b5 60 00 81 7d ac c0 0c 00 00 0f 85 f1 04 00 00 48
[ 0.081906] RSP: 0000:ffffffff82003dc0 EFLAGS: 00010046
[ 0.081909] RAX: 0000000000000000 RBX: 0000000000000003 RCX: 0000000000000cc0
[ 0.081911] RDX: 0000000000000003 RSI: 0000000000000006 RDI: ffffffff825b38e4
[ 0.081913] RBP: ffffffff82003e40 R08: ffff88813ffb7480 R09: 0000000000001000
[ 0.081915] R10: 0000000000001000 R11: 000000013ffff000 R12: 0000000000000001
[ 0.081917] R13: 0000000001a2c000 R14: 0000000000000000 R15: 0000000000000003
[ 0.081919] FS: 0000000000000000(0000) GS:ffffffff822ee000(0000) knlGS:0000000000000000
[ 0.081921] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 0.081923] CR2: 0000000000000000 CR3: 000000000200a000 CR4: 00000000000606b0
[ 0.081946] Call Trace:
[ 0.081951] __alloc_percpu+0x15/0x20
[ 0.081954] free_area_init+0x270/0x300
[ 0.081960] zone_sizes_init+0x44/0x46
[ 0.081965] paging_init+0x23/0x25
[ 0.081969] setup_arch+0x5aa/0x668
[ 0.081973] start_kernel+0x53/0x5b6
[ 0.081978] x86_64_start_reservations+0x24/0x26
[ 0.081983] x86_64_start_kernel+0x70/0x74
[ 0.081986] secondary_startup_64_no_verify+0xb0/0xbb
[ 0.081991] Modules linked in:
[ 0.081993] CR2: 0000000000000000
[ 0.081996] random: get_random_bytes called from oops_exit+0x39/0x60 with crng_init=0
pcpu_alloc+0x330 is
/root/linux-5.15.0/mm/percpu.c:1833
if (list_empty(&pcpu_chunk_lists[pcpu_free_slot])) {
359e: 48 63 05 00 00 00 00 movslq 0x0(%rip),%rax # 35a5 <pcpu_alloc+0x325>
35a1: R_X86_64_PC32 .data..ro_after_init+0x5c
35a5: 48 c1 e0 04 shl $0x4,%rax
35a9: 48 03 05 00 00 00 00 add 0x0(%rip),%rax # 35b0 <pcpu_alloc+0x330>
35ac: R_X86_64_PC32 pcpu_chunk_lists-0x4
list_empty():
/root/linux-5.15.0/./include/linux/list.h:282
return READ_ONCE(head->next) == head;
35b0: 48 8b 10 mov (%rax),%rdx <— rax == 0
free_area_init() -> /* added by patch */ alloc_percpu() -> pcpu_alloc():
/*
* No space left. Create a new chunk. We don't want multiple
* tasks to create chunks simultaneously. Serialize and create iff
* there's still no empty chunk after grabbing the mutex.
*/
if (is_atomic) {
err = "atomic alloc failed, no space left";
goto fail;
}
if (list_empty(&pcpu_chunk_lists[pcpu_free_slot])) { <— &pcpu_chunk_lists[pcpu_free_slot]) == NULL
chunk = pcpu_create_chunk(pcpu_gfp);
if (!chunk) {
err = "failed to allocate new chunk";
goto fail;
}
spin_lock_irqsave(&pcpu_lock, flags);
pcpu_chunk_relocate(chunk, -1);
} else {
This patch calls alloc_percpu() from setup_arch() while percpu allocator is not yet initialized (before setup_per_cpu_areas()).
Thanks,
—Alexey
