Hi Adam,
On Tue, Aug 15, 2023 at 10:48:35PM +0200, Salvatore Bonaccorso wrote:
> Control: tags -1 + upstream
>
> Hi Adam,
>
> On Tue, Aug 15, 2023 at 10:06:16PM +0200, Salvatore Bonaccorso wrote:
> > Hi Adam,
> >
> > On Tue, Aug 15, 2023 at 09:37:36PM +0200, Salvatore Bonaccorso wrote:
> > > Control: tags -1 + confirmed
> > >
> > > Hi Adam,
> > >
> > > On Tue, Aug 15, 2023 at 06:26:59PM +0100, Adam D. Barratt wrote:
> > > > On Sun, 2023-08-13 at 18:21 +0100, Adam D. Barratt wrote:
> > > > > Since the kernels on both the host and guests were upgraded to
> > > > > 5.10.179-5 (from 5.10.179-3), the guests on one of our Ganeti
> > > > > clusters
> > > > > have been reporting as tainted. Looking at dmesg shows the following
> > > > > trace early in boot:
> > > > >
> > > >
> > > > I realised that a little bit more of the surrounding context from dmesg
> > > > might help:
> > > >
> > > > [ 0.093383] kvm-guest: setup PV IPIs
> > > > [ 0.094931] ..TIMER: vector=0x30 apic1=0 pin1=2 apic2=-1 pin2=-1
> > > > [ 0.094968] clocksource: tsc-early: mask: 0xffffffffffffffff
> > > > max_cycles: 0x1e33052aadd, max_idle_ns: 440795310221 ns
> > > > [ 0.094974] Calibrating delay loop (skipped) preset value.. 4190.15
> > > > BogoMIPS (lpj=8380304)
> > > > [ 0.095097] x86/cpu: User Mode Instruction Prevention (UMIP)
> > > > activated
> > > > [ 0.095106] ------------[ cut here ]------------
> > > > [ 0.095107] get of unsupported state
> > > > [ 0.095120] WARNING: CPU: 0 PID: 0 at
> > > > arch/x86/kernel/fpu/xstate.c:973 get_xsave_addr+0x9b/0xb0
> > > > [ 0.095122] Modules linked in:
> > > > [ 0.095126] CPU: 0 PID: 0 Comm: swapper/0 Not tainted
> > > > 5.10.0-24-amd64 #1 Debian 5.10.179-5
> > > > [ 0.095128] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996),
> > > > BIOS 1.14.0-2 04/01/2014
> > > > [ 0.095131] RIP: 0010:get_xsave_addr+0x9b/0xb0
> > > > [ 0.095134] Code: 48 83 c4 08 5b e9 15 80 bc 00 80 3d 8d 7c 80 01 00
> > > > 75 a8 48 c7 c7 97 de eb 94 89 74 24 04 c6 05 79 7c 80 01 01 e8 f5 96 88
> > > > 00 <0f> 0b 8b 74 24 04 eb 89 31 c0 e9 e6 7f bc 00 66 0f 1f 44 00 00 89
> > > > [ 0.095136] RSP: 0000:ffffffff95403ec8 EFLAGS: 00010282
> > > > [ 0.095139] RAX: 0000000000000000 RBX: ffffffff9566a600 RCX:
> > > > ffffffff954b3768
> > > > [ 0.095141] RDX: c0000000ffffefff RSI: 00000000ffffefff RDI:
> > > > 0000000000000247
> > > > [ 0.095142] RBP: ffffffff9566a4a0 R08: 0000000000000000 R09:
> > > > ffffffff95403ce8
> > > > [ 0.095144] R10: ffffffff95403ce0 R11: ffffffff954cb7a8 R12:
> > > > 0000000000000246
> > > > [ 0.095145] R13: 0000000000000000 R14: 0000000000000000 R15:
> > > > 0000000000000000
> > > > [ 0.095148] FS: 0000000000000000(0000) GS:ffff911abec00000(0000)
> > > > knlGS:0000000000000000
> > > > [ 0.095149] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > > > [ 0.095151] CR2: ffff911a95801000 CR3: 0000000013e0a001 CR4:
> > > > 00000000007308b0
> > > > [ 0.095155] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
> > > > 0000000000000000
> > > > [ 0.095156] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7:
> > > > 0000000000000400
> > > > [ 0.095158] Call Trace:
> > > > [ 0.095166] identify_cpu+0x51f/0x540
> > > > [ 0.095171] identify_boot_cpu+0xc/0x94
> > > > [ 0.095174] arch_cpu_finalize_init+0x5/0x47
> > > > [ 0.095177] start_kernel+0x4ec/0x599
> > > > [ 0.095183] secondary_startup_64_no_verify+0xb0/0xbb
> > > > [ 0.095188] ---[ end trace 438682bdad114415 ]---
> > > > [ 0.095288] Last level iTLB entries: 4KB 0, 2MB 0, 4MB 0
> > > > [ 0.095289] Last level dTLB entries: 4KB 0, 2MB 0, 4MB 0, 1GB 0
> > > > [ 0.095296] Spectre V1 : Mitigation: usercopy/swapgs barriers and
> > > > __user pointer sanitization
> > >
> > > Confirmed I can reproduce it.
> >
> > 5.10.191-rc1 does not exibit the problem.
> >
> > Will further investigate.
>
> Quick summary: v5.10.190 upstream exhibit the same problem, so not a
> backporting problem, and v5.10.191-rc1 for the upcoming 5.10.191 seems
> to fix the issue.
This should be fixed by b3607269ff57 ("x86/pkeys: Revert a5eff7259790
("x86/pkeys: Add PKRU value to init_fpstate")")[1] upstream, which is
going to be a pplied in 5.10.191.
[1] https://git.kernel.org/linus/b3607269ff57fd3c9690cb25962c5e4b91a0fd3b
Regards,
Salvatore
