Re: User domain starts with a crash loop when memory configured is above 500GB

[Juergen Gross <jgross@xxxxxxxx>]

On 18.09.23 17:34, Robert Polasek wrote:
> Hi everybody,
>
> I have a server with 760GB of RAM. I have only domain 0 running there with 16GB 
> of ram assigned to it.
> Here is a configuration for my user domain:
>
> name = "node01"
> kernel = "/boot/vmlinuz-5.15.0-82-generic"
> root = "/dev/xvda"
> memory = 614400
> maxmem = 614400
> vcpus = 32
> maxvcpus = 32
> disk = ['file:/vserver/images/node01.img,xvda,w']
> vif = ['bridge=virbr0,mac=00:16:3e:01:01:02']
> iommu = "soft"
> swiotlb = "force"
> pci_permissive = 1
> pci = 
> ['0000:3e:00.0','0000:3f:00.0','0000:40:00.0','0000:41:00.0','0000:b1:00.0','0000:b2:00.0']
> nics = 1
> dhcp = "off"
> ip = "192.168.122.15"
> netmask = "255.255.255.0"
> gateway = "192.168.122.1"
> hostname = "node01"
>
> extra="3"
>
> When I try to start the domain, it spins in a crash loop with following error 
> messages:
> [ 6864.140170] WARNING: CPU: 2 PID: 266 at arch/x86/xen/multicalls.c:102 
> xen_mc_flush+0x197/0x200
> [ 6864.140183] Modules linked in:
> [ 6864.140190] CPU: 2 PID: 266 Comm: xen-balloon Tainted: G      D W        
>   5.15.0-82-generic #91-Ubuntu
> [ 6864.140203] RIP: e030:xen_mc_flush+0x197/0x200
> [ 6864.140212] Code: 77 65 89 c0 48 c1 e0 05 48 05 00 20 00 81 ff d0 0f 1f 00 49 
> 89 45 18 48 85 c0 0f 89 17 ff ff ff 45 8b 4d 00 41 bf 01 00 00 00 <0f> 0b 48 c7 
> c7 f0 8e 5b 82 44 89 ca 44 89 fe 45 31 f6 65 8b 0d e8
> [ 6864.140234] RSP: e02b:ffffc90041027b88 EFLAGS: 00010002
> [ 6864.140243] RAX: 0000000000000001 RBX: 0000000000000040 RCX: 0000000000000000
> [ 6864.140253] RDX: 0000000000000000 RSI: 0000000000000002 RDI: ffff89009809e310
> [ 6864.140264] RBP: ffffc90041027bb8 R08: ffff888168dc0000 R09: 0000000000000002
> [ 6864.140275] R10: 0000000000000200 R11: ffff8900980b7690 R12: 0000000000000000
> [ 6864.140286] R13: ffff89009809e300 R14: 0000000000000002 R15: 0000000000000001
> [ 6864.140303] FS:  0000000000000000(0000) GS:ffff890098080000(0000) 
> knlGS:0000000000000000
> [ 6864.140315] CS:  10000e030 DS: 0000 ES: 0000 CR0: 0000000080050033
> [ 6864.140324] CR2: 0000000000000000 CR3: 0000000002e10000 CR4: 0000000000050660
> [ 6864.140339] Call Trace:
> [ 6864.140344]  <TASK>
> [ 6864.140349]  ? __raw_callee_save_xen_make_pte+0x15/0x27
> [ 6864.140359]  xen_mc_issue+0x61/0x80
> [ 6864.140367]  xen_alloc_pte+0xd8/0x290
> [ 6864.140376]  pmd_populate_kernel.constprop.0+0x4b/0xa0
> [ 6864.140387]  vmemmap_pmd_populate+0x69/0x79
> [ 6864.140395]  vmemmap_populate_basepages+0x68/0xb3
> [ 6864.140405]  vmemmap_populate+0x2a/0xa9
> [ 6864.140412]  __populate_section_memmap+0x3c/0x57
> [ 6864.140422]  sparse_add_section+0x12b/0x1dc
> [ 6864.140431]  __add_pages+0xac/0x150
> [ 6864.140440]  add_pages+0x17/0x70
> [ 6864.140447]  arch_add_memory+0x45/0x60
> [ 6864.140455]  add_memory_resource+0x12c/0x320
> [ 6864.140467]  reserve_additional_memory+0x10f/0x160
> [ 6864.140476]  balloon_thread+0x337/0x500
> [ 6864.140483]  ? wait_woken+0x70/0x70
> [ 6864.140492]  ? reserve_additional_memory+0x160/0x160
> [ 6864.140501]  kthread+0x127/0x150
> [ 6864.140509]  ? set_kthread_struct+0x50/0x50
> [ 6864.140518]  ret_from_fork+0x1f/0x30
> [ 6864.140528]  </TASK>
> [ 6864.140533] ---[ end trace 3bca9737718a46b2 ]---
> [ 6864.140541] 1 of 2 multicall(s) failed: cpu 2
> [ 6864.140549]   call  2: op=26 arg=[ffff89009809eb10] result=-22
>
> Any suggestion what I am doing wrong? There should be plenty of RAM to start 
> 600GB domain. I can start  user domain with 500GB no problem. Thank you in 
> advance for your help and suggestions.
I think your kernel has been configured with CONFIG_XEN_512GB.

You should try to add "xen_512gb_limit=0" to your guest's command line.

Even if this is fixing your boot issue, the guest shouldn't show the error
you are seeing.


Juergen

Attachment: OpenPGP_0xB0DE9DD628BF132F.asc
Description: OpenPGP public key

Attachment: OpenPGP_signature.asc
Description: OpenPGP digital signature