[PATCH] [STABLE] 6.12.23: arch/x86/coco/tdx/tdx.c build error

[PATCH] [STABLE] 6.12.23: arch/x86/coco/tdx/tdx.c build error

[Cal Peake]

[-- Attachment #1: Type: text/plain, Size: 1833 bytes --]

Hi,

6.12.23 fails to build for me with the error:

  CC      arch/x86/coco/tdx/tdx.o
arch/x86/coco/tdx/tdx.c: In function ‘tdx_early_init’:
arch/x86/coco/tdx/tdx.c:1080:19: error: ‘struct pv_irq_ops’ has no member named ‘safe_halt’
 1080 |         pv_ops.irq.safe_halt = tdx_safe_halt;
      |                   ^
arch/x86/coco/tdx/tdx.c:1081:19: error: ‘struct pv_irq_ops’ has no member named ‘halt’
 1081 |         pv_ops.irq.halt = tdx_halt;
      |                   ^
make[5]: *** [scripts/Makefile.build:229: arch/x86/coco/tdx/tdx.o] Error 1
make[4]: *** [scripts/Makefile.build:478: arch/x86/coco/tdx] Error 2
make[3]: *** [scripts/Makefile.build:478: arch/x86/coco] Error 2
make[2]: *** [scripts/Makefile.build:478: arch/x86] Error 2
make[1]: *** [/home/cal/tmp/volatile/linux-6.12.23/Makefile:1944: .] Error 2
make: *** [Makefile:224: __sub-make] Error 2

Those struct members are only available when CONFIG_PARAVIRT_XXL is 
defined.

The below patch adds a check for that macro and gets the build to 
completion.

My config is attached for reference.

-- 
Cal Peake

Signed-off-by: Cal Peake <cp@absolutedigital.net>

--- ./arch/x86/coco/tdx/tdx.c	2025-04-10 08:39:41.000000000 -0400
+++ ./arch/x86/coco/tdx/tdx.c	2025-04-12 15:42:35.275284746 -0400
@@ -1066,7 +1066,7 @@
 
 	x86_platform.guest.enc_kexec_begin	     = tdx_kexec_begin;
 	x86_platform.guest.enc_kexec_finish	     = tdx_kexec_finish;
-
+#ifdef CONFIG_PARAVIRT_XXL
 	/*
 	 * Avoid "sti;hlt" execution in TDX guests as HLT induces a #VE that
 	 * will enable interrupts before HLT TDCALL invocation if executed
@@ -1079,7 +1079,7 @@
 	 */
 	pv_ops.irq.safe_halt = tdx_safe_halt;
 	pv_ops.irq.halt = tdx_halt;
-
+#endif
 	/*
 	 * TDX intercepts the RDMSR to read the X2APIC ID in the parallel
 	 * bringup low level code. That raises #VE which cannot be handled

[-- Attachment #2: config-6.12.23.gz --]
[-- Type: application/octet-stream, Size: 64546 bytes --]

Re: [STABLE] 6.12.23: vfio error (was: arch/x86/coco/tdx/tdx.c build error)

[Cal Peake]

Well, this isn't good...

After booting the new kernel, I get errors trying to do vfio passthrough:

qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.3, no available reset mechanism.
qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.2, no available reset mechanism.
qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.1, no available reset mechanism.
qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.0, no available reset mechanism.

Anyone have any ideas?

-- 
Cal Peake

Re: [STABLE] 6.12.23: vfio error (was: arch/x86/coco/tdx/tdx.c build error)

[Cal Peake]

On Sun, 13 Apr 2025, Cal Peake wrote:

> After booting the new kernel, I get errors trying to do vfio passthrough:
> 
> qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.3, no available reset mechanism.
> qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.2, no available reset mechanism.
> qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.1, no available reset mechanism.
> qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.0, no available reset mechanism.

This was resolved by Alex Williamson, with his patch:

https://lore.kernel.org/lkml/20250414211828.3530741-1-alex.williamson@redhat.com/

Thank you, Alex!

-- 
Cal Peake

Re: [STABLE] 6.12.23: vfio error (was: arch/x86/coco/tdx/tdx.c build error)

[Greg Kroah-Hartman]

On Mon, Apr 14, 2025 at 07:55:06PM -0400, Cal Peake wrote:
> On Sun, 13 Apr 2025, Cal Peake wrote:
> 
> > After booting the new kernel, I get errors trying to do vfio passthrough:
> > 
> > qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.3, no available reset mechanism.
> > qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.2, no available reset mechanism.
> > qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.1, no available reset mechanism.
> > qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.0, no available reset mechanism.
> 
> This was resolved by Alex Williamson, with his patch:
> 
> https://lore.kernel.org/lkml/20250414211828.3530741-1-alex.williamson@redhat.com/

Is this in Linus's tree yet?

thanks,

greg k-h

Re: [STABLE] 6.12.23: vfio error (was: arch/x86/coco/tdx/tdx.c build error)

[Cal Peake]

On Thu, 17 Apr 2025, Greg Kroah-Hartman wrote:

> On Mon, Apr 14, 2025 at 07:55:06PM -0400, Cal Peake wrote:
> > On Sun, 13 Apr 2025, Cal Peake wrote:
> > 
> > > After booting the new kernel, I get errors trying to do vfio passthrough:
> > > 
> > > qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.3, no available reset mechanism.
> > > qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.2, no available reset mechanism.
> > > qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.1, no available reset mechanism.
> > > qemu-8.2-system-x86_64: vfio: Cannot reset device 0000:14:00.0, no available reset mechanism.
> > 
> > This was resolved by Alex Williamson, with his patch:
> > 
> > https://lore.kernel.org/lkml/20250414211828.3530741-1-alex.williamson@redhat.com/
> 
> Is this in Linus's tree yet?

Adding Bjorn Helgaas and quoting from their message[1]:

 > Applied with Kevin's reviewed-by to pci/for-linus for v6.15, thanks,
 > and sorry for the breakage.

-- 
Cal Peake

[1] https://lore.kernel.org/lkml/20250415203929.GA34692@bhelgaas/