First of all move PV-only ELF notes inside the XEN_PV conditional; note
that
- HV_START_LOW is dropped altogether, as it was meaningful for 32-bit PV
  only,
- the 32-bit instance of VIRT_BASE is dropped, as it would be dead code
  once inside the conditional,
- while PADDR_OFFSET is not exactly unused for PVH, it defaults to zero
  there, and the hypervisor (or tool stack) complains if it is present
  but VIRT_BASE isn't.
Then have the "supported features" note actually report reality: All
three of the features there are supported and/or applicable only in
certain cases.

Signed-off-by: Jan Beulich <jbeulich@suse.com>
---
PAE_MODE, while in principle meaningful only for 32-bit PV, subtly
affects the start_info structure handed to Dom0. I guess I'd prefer if
that subtlety was first dealt with in the hypervisor, but it may well
turn out that it needs leaving as is.

SUSPEND_CANCEL is questionable as well: It was only ever consumed by
xend.
---
v2: Correct 32-bit non-PV mistakes. Drop HV_START_LOW. Also move
    VIRT_BASE, INIT_P2M, MOD_START_PFN, and PADDR_OFFSET.

--- head.orig/arch/x86/xen/xen-head.S	2023-04-23 21:02:52.000000000 +0200
+++ head/arch/x86/xen/xen-head.S	2023-02-15 15:43:03.000000000 +0100
@@ -90,30 +90,35 @@ SYM_CODE_END(xen_cpu_bringup_again)
 	ELFNOTE(Xen, XEN_ELFNOTE_GUEST_OS,       .asciz "linux")
 	ELFNOTE(Xen, XEN_ELFNOTE_GUEST_VERSION,  .asciz "2.6")
 	ELFNOTE(Xen, XEN_ELFNOTE_XEN_VERSION,    .asciz "xen-3.0")
-#ifdef CONFIG_X86_32
-	ELFNOTE(Xen, XEN_ELFNOTE_VIRT_BASE,      _ASM_PTR __PAGE_OFFSET)
-#else
+#ifdef CONFIG_XEN_PV
 	ELFNOTE(Xen, XEN_ELFNOTE_VIRT_BASE,      _ASM_PTR __START_KERNEL_map)
 	/* Map the p2m table to a 512GB-aligned user address. */
 	ELFNOTE(Xen, XEN_ELFNOTE_INIT_P2M,       .quad (PUD_SIZE * PTRS_PER_PUD))
-#endif
-#ifdef CONFIG_XEN_PV
 	ELFNOTE(Xen, XEN_ELFNOTE_ENTRY,          _ASM_PTR startup_xen)
-#endif
-	ELFNOTE(Xen, XEN_ELFNOTE_HYPERCALL_PAGE, _ASM_PTR hypercall_page)
-	ELFNOTE(Xen, XEN_ELFNOTE_FEATURES,
-		.ascii "!writable_page_tables|pae_pgdir_above_4gb")
-	ELFNOTE(Xen, XEN_ELFNOTE_SUPPORTED_FEATURES,
-		.long (1 << XENFEAT_writable_page_tables) |       \
-		      (1 << XENFEAT_dom0) |                       \
-		      (1 << XENFEAT_linux_rsdp_unrestricted))
+	ELFNOTE(Xen, XEN_ELFNOTE_FEATURES,       .ascii "!writable_page_tables")
 	ELFNOTE(Xen, XEN_ELFNOTE_PAE_MODE,       .asciz "yes")
-	ELFNOTE(Xen, XEN_ELFNOTE_LOADER,         .asciz "generic")
 	ELFNOTE(Xen, XEN_ELFNOTE_L1_MFN_VALID,
 		.quad _PAGE_PRESENT; .quad _PAGE_PRESENT)
-	ELFNOTE(Xen, XEN_ELFNOTE_SUSPEND_CANCEL, .long 1)
 	ELFNOTE(Xen, XEN_ELFNOTE_MOD_START_PFN,  .long 1)
-	ELFNOTE(Xen, XEN_ELFNOTE_HV_START_LOW,   _ASM_PTR __HYPERVISOR_VIRT_START)
 	ELFNOTE(Xen, XEN_ELFNOTE_PADDR_OFFSET,   _ASM_PTR 0)
+# define FEATURES_PV (1 << XENFEAT_writable_page_tables)
+#else
+# define FEATURES_PV 0
+#endif
+#ifdef CONFIG_XEN_PVH
+# define FEATURES_PVH (1 << XENFEAT_linux_rsdp_unrestricted)
+#else
+# define FEATURES_PVH 0
+#endif
+#ifdef CONFIG_XEN_DOM0
+# define FEATURES_DOM0 (1 << XENFEAT_dom0)
+#else
+# define FEATURES_DOM0 0
+#endif
+	ELFNOTE(Xen, XEN_ELFNOTE_HYPERCALL_PAGE, _ASM_PTR hypercall_page)
+	ELFNOTE(Xen, XEN_ELFNOTE_SUPPORTED_FEATURES,
+		.long FEATURES_PV | FEATURES_PVH | FEATURES_DOM0)
+	ELFNOTE(Xen, XEN_ELFNOTE_LOADER,         .asciz "generic")
+	ELFNOTE(Xen, XEN_ELFNOTE_SUSPEND_CANCEL, .long 1)
 
 #endif /*CONFIG_XEN */

On 16.06.23 08:36, Jan Beulich wrote:
> First of all move PV-only ELF notes inside the XEN_PV conditional; note
> that
> - HV_START_LOW is dropped altogether, as it was meaningful for 32-bit PV
>    only,
> - the 32-bit instance of VIRT_BASE is dropped, as it would be dead code
>    once inside the conditional,
> - while PADDR_OFFSET is not exactly unused for PVH, it defaults to zero
>    there, and the hypervisor (or tool stack) complains if it is present
>    but VIRT_BASE isn't.
> Then have the "supported features" note actually report reality: All
> three of the features there are supported and/or applicable only in
> certain cases.
> 
> Signed-off-by: Jan Beulich <jbeulich@suse.com>

Reviewed-by: Juergen Gross <jgross@suse.com>


Juergen