On Thu, 10 Aug 2023 09:41:29 -0500, Jinpu Wang <jinpu.wang@xxxxxxxxx>
wrote:
Hi folks on the list,
We hit following softlockup during test SGX VM with 5.15.125 kernel on
Icelake server:
[ 969.410230] CPU: 64 PID: 35093 Comm: qemu-7.1 Kdump: loaded
Tainted: G W O L 5.15.125-pserver
#5.15.125-1+feature+linux+5.15.y+20230809.0556+bda73c35~deb11 [
969.410233] Hardware name: Dell Inc. PowerEdge R650/0PYXKY, BIOS 1.9.2
11/17/2022 [ 969.410234] RIP: 0010:sgx_vepc_free_page+0x2e/0x80 [
969.410242] Code: 00 41 54 8b 07 48 89 f9 48 c1 e0 05 48 2b 88 b0 a1
45 a0 48 c1 e9 05 48 c1 e1 0c 48 03 88 a8 a1 45 a0 b8 03 00 00 00 0f
01 cf <41> 89 c4 85 c0 74 0f 83 f8 0d 75 19 44 89 e0 41 5c c3 cc cc cc
cc [ 969.410244] RSP: 0018:ff6646e08ce77e20 EFLAGS: 00000202 [
969.410247] RAX: 0000000000000000 RBX: 000000000008001f RCX:
ff6646f9f0ee6000 [ 969.410248] RDX: 000000000000003d RSI:
ff338d6614c00b68 RDI: ff6646f0d930fcc0 [ 969.410249] RBP:
ff338d6563367700 R08: ff6646f0d930fcc0 R09: ffffffffffffffff [
969.410250] R10: 0000000002000000 R11: ff338d65c29f9f10 R12:
ff6646e08ce77e38 [ 969.410252] R13: ff338d65891848e0 R14:
ff338d64884a9c80 R15: 0000000000000000 [ 969.410253] FS:
0000000000000000(0000) GS:ff338da3ff600000(0000)
knlGS:0000000000000000 [ 969.410255] CS: 0010 DS: 0000 ES: 0000 CR0:
0000000080050033 [ 969.410256] CR2: 00007f23a0002806 CR3:
00000024ce60a006 CR4: 0000000000773ee0 [ 969.410257] DR0:
0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [
969.410258] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
0000000000000400 [ 969.410260] PKRU: 55555554 [ 969.410260] Call
Trace: [ 969.410262] <IRQ> [ 969.410264] ?
watchdog_timer_fn+0x1b4/0x210 [ 969.410269] ?
lockup_detector_update_enable+0x50/0x50 [ 969.410272] ?
__hrtimer_run_queues+0x127/0x280 [ 969.410276] ?
hrtimer_interrupt+0xfc/0x210 [ 969.410279] ?
__sysvec_apic_timer_interrupt+0x59/0xd0 [ 969.410281] ?
sysvec_apic_timer_interrupt+0x6d/0x90 [ 969.410285] </IRQ> [
969.410286] <TASK> [ 969.410287] ?
asm_sysvec_apic_timer_interrupt+0x16/0x20 [ 969.410291] ?
sgx_vepc_free_page+0x2e/0x80 [ 969.410293]
sgx_vepc_release+0x65/0x220 [ 969.410296] __fput+0x89/0x250 [
969.410301] task_work_run+0x59/0x90 [ 969.410306]
do_exit+0x337/0x9a0 [ 969.410309] ? do_user_addr_fault+0x1cd/0x660 [
969.410312] __x64_sys_exit+0x17/0x20 [ 969.410314]
do_syscall_64+0x38/0x90 [ 969.410317]
entry_SYSCALL_64_after_hwframe+0x67/0xd1 [ 969.410319] RIP:
0033:0x7f23afe93f56 [ 969.410322] Code: Unable to access opcode bytes
at RIP 0x7f23afe93f2c. [ 969.410323] RSP: 002b:00007f23a5dd6800
EFLAGS: 00000246 ORIG_RAX: 000000000000003c [ 969.410325] RAX:
ffffffffffffffda RBX: 00007f23a5dd7700 RCX: 00007f23afe93f56 [
969.410326] RDX: 000000000000003c RSI: 00000000007fb000 RDI:
0000000000000000 [ 969.410327] RBP: 00007f23a55d7000 R08:
00007f23a5dd66b0 R09: 00007f23a5dd6630 [ 969.410328] R10:
0000000000000008 R11: 0000000000000246 R12: 00007ffe693cbeee [
969.410329] R13: 00007ffe693cbeef R14: 00007f23a5dd68c0 R15:
0000000000802000 [ 969.410331] </TASK>
is it known?
I guess something like this is needed?
diff --git a/arch/x86/kernel/cpu/sgx/virt.c
b/arch/x86/kernel/cpu/sgx/virt.c index 6a77a14eee38..9f863250f08d
100644 --- a/arch/x86/kernel/cpu/sgx/virt.c +++
b/arch/x86/kernel/cpu/sgx/virt.c @@ -204,6 +204,7 @@ static int
sgx_vepc_release(struct inode *inode, struct file *file)
continue; xa_erase(&vepc->page_array,
index); + cond_resched(); } /* @@
-222,6 +223,7 @@ static int sgx_vepc_release(struct inode *inode,
struct file *file)
list_add_tail(&epc_page->list, &secs_pages);
xa_erase(&vepc->page_array, index); + cond_resched();
} /*
Thanks for reporting. This seems reasonable change. Did you test with the
change
Could you submit a patch for review? Or I can do it if you prefer and let
me know the host EPC and VM EPC configuration info for repro.
Thanks
Haitao
