Re: [PATCH 5.15 v2 0/3] Support static calls with LLVM-built kernels

[Greg KH <greg@kroah.com>]

On Sat, Mar 30, 2024 at 06:57:06AM -0300, Thadeu Lima de Souza Cascardo wrote:
> On Fri, Mar 29, 2024 at 01:50:11PM +0100, Greg KH wrote:
> > On Mon, Mar 18, 2024 at 10:39:04AM -0300, Thadeu Lima de Souza Cascardo wrote:
> > > Otherwise, we see warnings like this:
> > > 
> > > [    0.000000][    T0] ------------[ cut here ]------------
> > > [    0.000000][    T0] unexpected static_call insn opcode 0xf at kvm_vcpu_reload_apic_access_page+0x17/0x30
> > > [    0.000000][    T0] WARNING: CPU: 0 PID: 0 at arch/x86/kernel/static_call.c:88 __static_call_validate+0x68/0x70
> > > [    0.000000][    T0] Modules linked in:
> > > [    0.000000][    T0] CPU: 0 PID: 0 Comm: swapper Not tainted 5.15.151-00083-gf200c7260296 #68 fe3cb25cf78cb710722bb5acd1cadddd35172924
> > > [    0.000000][    T0] RIP: 0010:__static_call_validate+0x68/0x70
> > > [    0.000000][    T0] Code: 0f b6 4a 04 81 f1 c0 00 00 00 09 c1 74 cc 80 3d be 2c 02 02 00 75 c3 c6 05 b5 2c 02 02 01 48 c7 c7 38 4f c3 82 e8 e8 c8 09 00 <0f> 0b c3 00 00 cc cc 00 53 48 89 fb 48 63 15 31 71 06 02
> > > e8 b0 b8
> > > [    0.000000][    T0] RSP: 0000:ffffffff82e03e70 EFLAGS: 00010046 ORIG_RAX: 0000000000000000
> > > [    0.000000][    T0] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000002
> > > [    0.000000][    T0] RDX: 0000000000000000 RSI: ffffffff82e03ce0 RDI: 0000000000000001
> > > [    0.000000][    T0] RBP: 0000000000000001 R08: 00000000ffffffff R09: ffffffff82eaab70
> > > [    0.000000][    T0] R10: ffffffff82e2e900 R11: 205d305420202020 R12: ffffffff82e51960
> > > [    0.000000][    T0] R13: ffffffff81038987 R14: ffffffff81038987 R15: 0000000000000001
> > > [    0.000000][    T0] FS:  0000000000000000(0000) GS:ffffffff83726000(0000) knlGS:0000000000000000
> > > [    0.000000][    T0] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> > > [    0.000000][    T0] CR2: ffff888000014be8 CR3: 00000000037b2000 CR4: 00000000000000a0
> > > [    0.000000][    T0] Call Trace:
> > > [    0.000000][    T0]  <TASK>
> > > [    0.000000][    T0]  ? __warn+0x75/0xe0
> > > [    0.000000][    T0]  ? report_bug+0x81/0xe0
> > > [    0.000000][    T0]  ? kvm_vcpu_reload_apic_access_page+0x17/0x30
> > > [    0.000000][    T0]  ? kvm_vcpu_reload_apic_access_page+0x17/0x30
> > > [    0.000000][    T0]  ? early_fixup_exception+0x44/0xa0
> > > [    0.000000][    T0]  ? early_idt_handler_common+0x2f/0x40
> > > [    0.000000][    T0]  ? kvm_vcpu_reload_apic_access_page+0x17/0x30
> > > [    0.000000][    T0]  ? kvm_vcpu_reload_apic_access_page+0x17/0x30
> > > [    0.000000][    T0]  ? __static_call_validate+0x68/0x70
> > > [    0.000000][    T0]  ? arch_static_call_transform+0x5c/0x90
> > > [    0.000000][    T0]  ? __static_call_init+0x1ec/0x230
> > > [    0.000000][    T0]  ? static_call_init+0x32/0x70
> > > [    0.000000][    T0]  ? setup_arch+0x36/0x4f0
> > > [    0.000000][    T0]  ? start_kernel+0x67/0x400
> > > [    0.000000][    T0]  ? secondary_startup_64_no_verify+0xb1/0xbb
> > > [    0.000000][    T0]  </TASK>
> > > [    0.000000][    T0] ---[ end trace 8c8589c01f370686 ]---
> > > 
> > > 
> > > 
> > > Peter Zijlstra (3):
> > >   x86/alternatives: Introduce int3_emulate_jcc()
> > >   x86/alternatives: Teach text_poke_bp() to patch Jcc.d32 instructions
> > >   x86/static_call: Add support for Jcc tail-calls
> > > 
> > >  arch/x86/include/asm/text-patching.h | 31 +++++++++++++++
> > >  arch/x86/kernel/alternative.c        | 56 +++++++++++++++++++++++-----
> > >  arch/x86/kernel/kprobes/core.c       | 38 ++++---------------
> > >  arch/x86/kernel/static_call.c        | 49 ++++++++++++++++++++++--
> > >  4 files changed, 132 insertions(+), 42 deletions(-)
> > 
> > Why is there a v2 series here?  Are the ones I just took not correct?
> > 
> > confused,
> > 
> > greg k-h
> 
> Because Sasha questioned the presence of the first 2 patches in the series
> while they were not backported to 6.1. Then, I looked at the 6.1 backport for
> reference and determined they were not really necessary if I picked the same
> changes that the 6.1 backport applied.

So is what I queued up correct or not?

still confused,

greg k-h