Skip to contentSkip to conversation

Various patches to improve Secure Boot support

PATCH Superseded Frediano Ziglio 4 participants 34 messages May 29 – Jun 15

v1 of 8 Superseded Latest is v8 →

lore
Share & tools
lore
#0 lore raw

cover Various patches to improve Secure Boot support

These patches improve support for Secure boot.
UEFI CA memory mitigation requires memory pages to be not executable and writable at the same time. So changing permissions and splitting some section is required.
SBAT is the preferred way to revocate executables.
Remove multiboot pieces from EFI executable.

Frediano Ziglio (2):
Align all sections to 4KB
x86: Split .init section to satisfy UEFI CA memory mitigation

Gerald Elder-Vass (1):
Add SBAT section to the PE binary

Roger Pau Monné (3):
x86/efi: discard .text.header for PE binary x86/efi: discard multiboot related entry code for PE binary x86/efi: avoid a relocation in efi_arch_post_exit_boot()

 xen/arch/x86/Makefile       |  4 ++++
 xen/arch/x86/boot/head.S    |  3 ++-
 xen/arch/x86/efi/efi-boot.h |  7 +++++--
 xen/arch/x86/xen.lds.S      | 25 +++++++++++++++----------
 xen/include/xen/xen.lds.h   |  3 ++-
 5 files changed, 28 insertions(+), 14 deletions(-)
Signature · 1 line
2.43.0
#1 lore raw

PATCH 1/6 Add SBAT section to the PE binary

From: Gerald Elder-Vass <[email protected]>

The SBAT section provides a way for the binary to declare a generation id for its upstream source and any vendor changes applied. A compatible loader can then revoke vulnerable binaries by generation, using the binary's declared generation id(s) to determine if it is safe to load.

More information about SBAT is available here:
https://github.com/rhboot/shim/blob/main/SBAT.md

Populate the SBAT section in the Xen binary by using the information in xen/arch/x86/sbat.csv.

On XenServer, the version and release fields are populated by the spec file during the build process.

---
 xen/arch/x86/Makefile     | 4 ++++
 xen/arch/x86/xen.lds.S    | 2 ++
 xen/include/xen/xen.lds.h | 3 ++-
 3 files changed, 8 insertions(+), 1 deletion(-)
xen/arch/x86/Makefile +4 -0
index 47dd6c50fe..a2bdcb6f44 100644
@@ -71,6 +71,7 @@ obj-$(CONFIG_TBOOT) += tboot.o
 obj-y += hpet.o
 obj-$(CONFIG_VM_EVENT) += vm_event.o
 obj-y += xstate.o
+obj-y += sbat_data.o
 
 ifneq ($(CONFIG_PV_SHIM_EXCLUSIVE),y)
 obj-y += domctl.o
@@ -275,6 +276,9 @@ $(obj)/efi.lds: AFLAGS-y += -DEFI
 $(obj)/xen.lds $(obj)/efi.lds: $(src)/xen.lds.S FORCE
 	$(call if_changed_dep,cpp_lds_S)
 
+$(obj)/sbat_data.o: $(src)/sbat.csv
+	$(OBJCOPY) -I binary -O elf64-x86-64 --rename-section .data=.sbat,readonly,data,contents --add-section .note.GNU-stack=/dev/null $(srcdir)/sbat.csv $@
+
 clean-files := \
     include/asm/asm-macros.* \
     $(objtree)/.xen-syms.[0-9]* \
xen/arch/x86/sbat.csv new file +1 -0
new file mode 100644
index 000000000000..1573604e2f10
@@ -0,0 +1,1 @@
+sbat,1,SBAT Version,sbat,1,https://github.com/rhboot/shim/blob/main/SBAT.md
xen/arch/x86/xen.lds.S +2 -0
index b9e888e596..c2b9b5a893 100644
@@ -354,6 +354,8 @@ SECTIONS
   PROVIDE(ALT_START = 0);
   VIRT_START &= 0;
   ALT_START &= 0;
+
+  .sbat (NOLOAD) : { *(.sbat) }
 #elif defined(XEN_BUILD_EFI)
   /*
    * Due to the way EFI support is currently implemented, these two symbols
xen/include/xen/xen.lds.h +2 -1
index ea11e3fb62..c9aa1b7fae 100644
@@ -118,7 +118,8 @@
        *(.comment.*) \
        *(.note.*)
 #else
-#define DISCARD_EFI_SECTIONS
+#define DISCARD_EFI_SECTIONS \
+       *(.sbat)
 #endif
 
 /* Sections to be discarded. */
Signature · 1 line
2.43.0
#2 lore raw

Re: [PATCH 1/6] Add SBAT section to the PE binary

Quoted 16 lines · Frediano Ziglio

From: Gerald Elder-Vass <[email protected]>

The SBAT section provides a way for the binary to declare a generation id for its upstream source and any vendor changes applied. A compatible loader can then revoke vulnerable binaries by generation, using the binary's declared generation id(s) to determine if it is safe to load.

More information about SBAT is available here:
https://github.com/rhboot/shim/blob/main/SBAT.md

Populate the SBAT section in the Xen binary by using the information in xen/arch/x86/sbat.csv.

On XenServer, the version and release fields are populated by the spec file during the build process.

Quoted 7 lines
---
 xen/arch/x86/Makefile     | 4 ++++
 xen/arch/x86/xen.lds.S    | 2 ++
 xen/include/xen/xen.lds.h | 3 ++-
 3 files changed, 8 insertions(+), 1 deletion(-)

This gives the impression of being an entirely new patch, when really the (standalone) patch was already at v4, and my comment there wasn't addressed (perhaps merely by extending the description some). It also feels like there were other pending comments, or else why would this not have gone in long ago? Additionally, ...

xen/arch/x86/Makefile obj-$(CONFIG_TBOOT) += tboot.o · line 74

Quoted 7 lines
xen/arch/x86/Makefile +1 -0
@@ -71,6 +71,7 @@ obj-$(CONFIG_TBOOT) += tboot.o
 obj-y += hpet.o
 obj-$(CONFIG_VM_EVENT) += vm_event.o
 obj-y += xstate.o
+obj-y += sbat_data.o

... like elsewhere: New files' names should prefer dashes over underscores. Question of course is why this isn't simply sbat.o in the first place.

xen/arch/x86/Makefile $(obj)/efi.lds: AFLAGS-y += -DEFI · line 280

Quoted 6 lines
xen/arch/x86/Makefile +2 -0
@@ -275,6 +276,9 @@ $(obj)/efi.lds: AFLAGS-y += -DEFI
 $(obj)/xen.lds $(obj)/efi.lds: $(src)/xen.lds.S FORCE
 	$(call if_changed_dep,cpp_lds_S)
 
+$(obj)/sbat_data.o: $(src)/sbat.csv
+	$(OBJCOPY) -I binary -O elf64-x86-64 --rename-section .data=.sbat,readonly,data,contents --add-section .note.GNU-stack=/dev/null $(srcdir)/sbat.csv $@

That'll be an SHT_PROGBITS .note.GNU-stack, won't it? When it really wants to be SHT_NOTE at least for bleeding edge GNU binutils (see relatively recent changes there).

Jan

#3 lore raw

Re: [PATCH 1/6] Add SBAT section to the PE binary

Quoted 24 lines · Jan Beulich
Quoted 16 lines · Frediano Ziglio

From: Gerald Elder-Vass <[email protected]>

The SBAT section provides a way for the binary to declare a generation id for its upstream source and any vendor changes applied. A compatible loader can then revoke vulnerable binaries by generation, using the binary's declared generation id(s) to determine if it is safe to load.

More information about SBAT is available here:
https://github.com/rhboot/shim/blob/main/SBAT.md

Populate the SBAT section in the Xen binary by using the information in xen/arch/x86/sbat.csv.

On XenServer, the version and release fields are populated by the spec file during the build process.

Quoted 7 lines
---
 xen/arch/x86/Makefile     | 4 ++++
 xen/arch/x86/xen.lds.S    | 2 ++
 xen/include/xen/xen.lds.h | 3 ++-
 3 files changed, 8 insertions(+), 1 deletion(-)
Quoted 5 lines

This gives the impression of being an entirely new patch, when really the (standalone) patch was already at v4, and my comment there wasn't addressed (perhaps merely by extending the description some). It also feels like there were other pending comments, or else why would this not have gone in long ago?

The content of the SBAT table is still pending the security document, and in particular the decision over whether upstream Xen maintains a global revision or not.

Although the global revision could be appended later.

Quoted 11 lines

Additionally, ...

xen/arch/x86/Makefile obj-$(CONFIG_TBOOT) += tboot.o · line 74

Quoted 7 lines
xen/arch/x86/Makefile +1 -0
@@ -71,6 +71,7 @@ obj-$(CONFIG_TBOOT) += tboot.o
 obj-y += hpet.o
 obj-$(CONFIG_VM_EVENT) += vm_event.o
 obj-y += xstate.o
+obj-y += sbat_data.o

... like elsewhere: New files' names should prefer dashes over underscores. Question of course is why this isn't simply sbat.o in the first place.

Also sorted.

Quoted 10 lines

xen/arch/x86/Makefile $(obj)/efi.lds: AFLAGS-y += -DEFI · line 280

Quoted 7 lines
xen/arch/x86/Makefile +2 -0
@@ -275,6 +276,9 @@ $(obj)/efi.lds: AFLAGS-y += -DEFI
 $(obj)/xen.lds $(obj)/efi.lds: $(src)/xen.lds.S FORCE
 	$(call if_changed_dep,cpp_lds_S)
 
+$(obj)/sbat_data.o: $(src)/sbat.csv
+	$(OBJCOPY) -I binary -O elf64-x86-64 --rename-section .data=.sbat,readonly,data,contents --add-section .note.GNU-stack=/dev/null $(srcdir)/sbat.csv $@

That'll be an SHT_PROGBITS .note.GNU-stack, won't it? When it really wants to be SHT_NOTE at least for bleeding edge GNU binutils (see relatively recent changes there).

That was my addition to prevent there being a warning about RWX stacks cascade all the way up the build.

Maybe a better option is to have sbat.S with an .incbin sbat.csv ?

~Andrew

#4 lore raw

Re: [PATCH 1/6] Add SBAT section to the PE binary

Quoted 12 lines · Andrew Cooper
Quoted 10 lines · Jan Beulich

xen/arch/x86/Makefile $(obj)/efi.lds: AFLAGS-y += -DEFI · line 280

Quoted 6 lines · Frediano Ziglio
xen/arch/x86/Makefile +2 -0
@@ -275,6 +276,9 @@ $(obj)/efi.lds: AFLAGS-y += -DEFI
 $(obj)/xen.lds $(obj)/efi.lds: $(src)/xen.lds.S FORCE
 	$(call if_changed_dep,cpp_lds_S)
 
+$(obj)/sbat_data.o: $(src)/sbat.csv
+	$(OBJCOPY) -I binary -O elf64-x86-64 --rename-section .data=.sbat,readonly,data,contents --add-section .note.GNU-stack=/dev/null $(srcdir)/sbat.csv $@

That'll be an SHT_PROGBITS .note.GNU-stack, won't it? When it really wants to be SHT_NOTE at least for bleeding edge GNU binutils (see relatively recent changes there).

Quoted 4 lines

That was my addition to prevent there being a warning about RWX stacks cascade all the way up the build.

Maybe a better option is to have sbat.S with an .incbin sbat.csv ?

Yes, that'll avoid all the custom objcopy-ing.

Jan

#5 lore raw

Re: [PATCH 1/6] Add SBAT section to the PE binary

Quoted 12 lines · Jan Beulich
Quoted 11 lines · Andrew Cooper
Quoted 10 lines · Jan Beulich

xen/arch/x86/Makefile $(obj)/efi.lds: AFLAGS-y += -DEFI · line 280

Quoted 6 lines · Frediano Ziglio
xen/arch/x86/Makefile +2 -0
@@ -275,6 +276,9 @@ $(obj)/efi.lds: AFLAGS-y += -DEFI
 $(obj)/xen.lds $(obj)/efi.lds: $(src)/xen.lds.S FORCE
 	$(call if_changed_dep,cpp_lds_S)
 
+$(obj)/sbat_data.o: $(src)/sbat.csv
+	$(OBJCOPY) -I binary -O elf64-x86-64 --rename-section .data=.sbat,readonly,data,contents --add-section .note.GNU-stack=/dev/null $(srcdir)/sbat.csv $@

That'll be an SHT_PROGBITS .note.GNU-stack, won't it? When it really wants to be SHT_NOTE at least for bleeding edge GNU binutils (see relatively recent changes there).

Quoted 5 lines
Quoted 4 lines

That was my addition to prevent there being a warning about RWX stacks cascade all the way up the build.

Maybe a better option is to have sbat.S with an .incbin sbat.csv ?

Yes, that'll avoid all the custom objcopy-ing.

The objcopy line comes from the instructions in shim-review, not that I think they're great instructions overall.

~Andrew

#6 lore raw

PATCH 2/6 Align all sections to 4KB

In reply to Frediano Ziglio

Required by UEFI CA memory mitigation.

---
 xen/arch/x86/xen.lds.S | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
xen/arch/x86/xen.lds.S +4 -2
index c2b9b5a893..322c116a04 100644
@@ -162,8 +162,8 @@ SECTIONS
        __note_gnu_build_id_end = .;
   } PHDR(note) PHDR(text)
 #elif defined(BUILD_ID_EFI)
-  /* Workaround bug in binutils < 2.36 */
-  . = ALIGN(32);
+  /* align to satisfy UEFI CA memory mitigation */
+  . = ALIGN(PAGE_SIZE);
   DECL_SECTION(.buildid) {
        __note_gnu_build_id_start = .;
        *(.buildid)
@@ -330,6 +330,7 @@ SECTIONS
   __2M_rwdata_end = ALIGN(SECTION_ALIGN);
 
 #ifdef EFI
+  . = ALIGN(PAGE_SIZE);
   .reloc ALIGN(4) : {
     __base_relocs_start = .;
     *(.reloc)
@@ -355,6 +356,7 @@ SECTIONS
   VIRT_START &= 0;
   ALT_START &= 0;
 
+  . = ALIGN(PAGE_SIZE);
   .sbat (NOLOAD) : { *(.sbat) }
 #elif defined(XEN_BUILD_EFI)
   /*
Signature · 1 line
2.43.0
#7 lore raw

Re: [PATCH 2/6] Align all sections to 4KB

Quoted 1 line · Frediano Ziglio

Required by UEFI CA memory mitigation.

Imo this isn't enough for a description: At the very least you want to point at where that requirement is spelled out. Additionally: How much of an image growth is this going to be? Considering that elsewhere it was already said that xen.efi is too large for certain purposes (and hence e.g. debug info wants stripping there), size increase is a potential issue.

Quoted 22 lines
xen/arch/x86/xen.lds.S +3 -2
@@ -162,8 +162,8 @@ SECTIONS
        __note_gnu_build_id_end = .;
   } PHDR(note) PHDR(text)
 #elif defined(BUILD_ID_EFI)
-  /* Workaround bug in binutils < 2.36 */
-  . = ALIGN(32);
+  /* align to satisfy UEFI CA memory mitigation */
+  . = ALIGN(PAGE_SIZE);
   DECL_SECTION(.buildid) {
        __note_gnu_build_id_start = .;
        *(.buildid)
@@ -330,6 +330,7 @@ SECTIONS
   __2M_rwdata_end = ALIGN(SECTION_ALIGN);
 
 #ifdef EFI
+  . = ALIGN(PAGE_SIZE);
   .reloc ALIGN(4) : {
     __base_relocs_start = .;
     *(.reloc)

xen/arch/x86/xen.lds.S SECTIONS · line 362

Quoted 7 lines
xen/arch/x86/xen.lds.S +1 -0
@@ -355,6 +356,7 @@ SECTIONS
   VIRT_START &= 0;
   ALT_START &= 0;
 
+  . = ALIGN(PAGE_SIZE);
   .sbat (NOLOAD) : { *(.sbat) }
 #elif defined(XEN_BUILD_EFI)
   /*

You say "all sections" in the title, yet this is not covering e.g. debug info.

Jan

#8 lore raw

Re: [PATCH 2/6] Align all sections to 4KB

Quoted 4 lines · Jan Beulich
Quoted 1 line · Frediano Ziglio

Required by UEFI CA memory mitigation.

Imo this isn't enough for a description: At the very least you want to point at where that requirement is spelled out.

It is a requirement for NX_COMPAT so the PE can be loaded with W^X perms in the pagetables.

NX_COMPAT is a requirement from shim-review,
https://github.com/rhboot/shim-review#do-you-have-the-nx-bit-set-in-your-shim-if-so-is-your-entire-boot-stack-nx-compatible-and-what-testing-have-you-done-to-ensure-such-compatibility

The discussed exception is (AFAICT) only for re-signing an previously-accepted blobs, and is not usable for new downstream products (which is what Xen is in this context).

~Andrew

#9 lore raw

Re: [PATCH 2/6] Align all sections to 4KB

In reply to Jan Beulich
Quoted 11 lines · Jan Beulich
Quoted 1 line · Frediano Ziglio

Required by UEFI CA memory mitigation.

Imo this isn't enough for a description: At the very least you want to point at where that requirement is spelled out. Additionally: How much of an image growth is this going to be? Considering that elsewhere it was already said that xen.efi is too large for certain purposes (and hence e.g. debug info wants stripping there), size increase is a potential issue.

I'll integrate Andrew comment on the description.
The image will increase but if it's a problem just adding some kbytes it would be a pretty big deal.
If I remember the issue was more if the image was not stripped and in this case it was like 10 times bigger, not just a bit bigger.

Quoted 25 lines
Quoted 25 lines
xen/arch/x86/xen.lds.S +3 -2
@@ -162,8 +162,8 @@ SECTIONS
        __note_gnu_build_id_end = .;
   } PHDR(note) PHDR(text)
 #elif defined(BUILD_ID_EFI)
-  /* Workaround bug in binutils < 2.36 */
-  . = ALIGN(32);
+  /* align to satisfy UEFI CA memory mitigation */
+  . = ALIGN(PAGE_SIZE);
   DECL_SECTION(.buildid) {
        __note_gnu_build_id_start = .;
        *(.buildid)
@@ -330,6 +330,7 @@ SECTIONS
   __2M_rwdata_end = ALIGN(SECTION_ALIGN);
 
 #ifdef EFI
+  . = ALIGN(PAGE_SIZE);
   .reloc ALIGN(4) : {
     __base_relocs_start = .;
     *(.reloc)
@@ -355,6 +356,7 @@ SECTIONS
   VIRT_START &= 0;
   ALT_START &= 0;
 
Quoted 8 lines

xen/arch/x86/xen.lds.S SECTIONS · line 362

Quoted 4 lines
xen/arch/x86/xen.lds.S +1 -0
+  . = ALIGN(PAGE_SIZE);
   .sbat (NOLOAD) : { *(.sbat) }
 #elif defined(XEN_BUILD_EFI)
   /*

You say "all sections" in the title, yet this is not covering e.g. debug info.

I will change to "all loadable sections". debug sections are not loadable so they don't cause an issue.

Quoted 1 line

Jan

Frediano

#10 lore raw

Re: [PATCH 2/6] Align all sections to 4KB

Quoted 35 lines · Frediano Ziglio
Quoted 26 lines · Jan Beulich
Quoted 25 lines · Frediano Ziglio
xen/arch/x86/xen.lds.S +3 -2
@@ -162,8 +162,8 @@ SECTIONS
        __note_gnu_build_id_end = .;
   } PHDR(note) PHDR(text)
 #elif defined(BUILD_ID_EFI)
-  /* Workaround bug in binutils < 2.36 */
-  . = ALIGN(32);
+  /* align to satisfy UEFI CA memory mitigation */
+  . = ALIGN(PAGE_SIZE);
   DECL_SECTION(.buildid) {
        __note_gnu_build_id_start = .;
        *(.buildid)
@@ -330,6 +330,7 @@ SECTIONS
   __2M_rwdata_end = ALIGN(SECTION_ALIGN);
 
 #ifdef EFI
+  . = ALIGN(PAGE_SIZE);
   .reloc ALIGN(4) : {
     __base_relocs_start = .;
     *(.reloc)
@@ -355,6 +356,7 @@ SECTIONS
   VIRT_START &= 0;
   ALT_START &= 0;
 
Quoted 7 lines

xen/arch/x86/xen.lds.S SECTIONS · line 362

Quoted 4 lines
xen/arch/x86/xen.lds.S +1 -0
+  . = ALIGN(PAGE_SIZE);
   .sbat (NOLOAD) : { *(.sbat) }
 #elif defined(XEN_BUILD_EFI)
   /*

You say "all sections" in the title, yet this is not covering e.g. debug info.

Quoted 2 lines

I will change to "all loadable sections". debug sections are not loadable so they don't cause an issue.

Please try to be precise there, as some aspects are subtle. As per my understanding, like .reloc all .debug_* are loadable (and may be loaded). The IMAGE_SCN_MEM_DISCARDABLE flag merely means they can be discarded at a certain point after image loading (for .reloc in particular: after relocations were processed).

Jan

#11 lore raw

Re: [PATCH 2/6] Align all sections to 4KB

Frediano

Quoted 40 lines · Jan Beulich
Quoted 35 lines · Frediano Ziglio
Quoted 26 lines · Jan Beulich
Quoted 25 lines · Frediano Ziglio
xen/arch/x86/xen.lds.S +3 -2
@@ -162,8 +162,8 @@ SECTIONS
        __note_gnu_build_id_end = .;
   } PHDR(note) PHDR(text)
 #elif defined(BUILD_ID_EFI)
-  /* Workaround bug in binutils < 2.36 */
-  . = ALIGN(32);
+  /* align to satisfy UEFI CA memory mitigation */
+  . = ALIGN(PAGE_SIZE);
   DECL_SECTION(.buildid) {
        __note_gnu_build_id_start = .;
        *(.buildid)
@@ -330,6 +330,7 @@ SECTIONS
   __2M_rwdata_end = ALIGN(SECTION_ALIGN);
 
 #ifdef EFI
+  . = ALIGN(PAGE_SIZE);
   .reloc ALIGN(4) : {
     __base_relocs_start = .;
     *(.reloc)
@@ -355,6 +356,7 @@ SECTIONS
   VIRT_START &= 0;
   ALT_START &= 0;
 
Quoted 7 lines

xen/arch/x86/xen.lds.S SECTIONS · line 362

Quoted 4 lines
xen/arch/x86/xen.lds.S +1 -0
+  . = ALIGN(PAGE_SIZE);
   .sbat (NOLOAD) : { *(.sbat) }
 #elif defined(XEN_BUILD_EFI)
   /*

You say "all sections" in the title, yet this is not covering e.g. debug info.

Quoted 2 lines

I will change to "all loadable sections". debug sections are not loadable so they don't cause an issue.

Quoted 7 lines

Please try to be precise there, as some aspects are subtle. As per my understanding, like .reloc all .debug_* are loadable (and may be loaded). The IMAGE_SCN_MEM_DISCARDABLE flag merely means they can be discarded at a certain point after image loading (for .reloc in particular: after relocations were processed).

Jan

No, debug sections are not loadable, for instance, in a random executable I found:

Sections:

Idx Name          Size      VMA               LMA               File off  Algn
  0 .text         0000c7f8  0000000140001000  0000000140001000  00000600  2**4

CONTENTS, ALLOC, LOAD, READONLY, CODE, DATA
1 .data 00000250 000000014000e000 000000014000e000 0000ce00 2**4
CONTENTS, ALLOC, LOAD, DATA
2 .rdata 00002c70 000000014000f000 000000014000f000 0000d200 2**4
CONTENTS, ALLOC, LOAD, READONLY, DATA
3 .pdata 00000654 0000000140012000 0000000140012000 00010000 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
4 .xdata 00000630 0000000140013000 0000000140013000 00010800 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
5 .bss 00001ec0 0000000140014000 0000000140014000 00000000 2**4
ALLOC
6 .idata 00000c88 0000000140016000 0000000140016000 00011000 2**2
CONTENTS, ALLOC, LOAD, DATA
7 .CRT 00000060 0000000140017000 0000000140017000 00011e00 2**2
CONTENTS, ALLOC, LOAD, DATA
8 .tls 00000010 0000000140018000 0000000140018000 00012000 2**2
CONTENTS, ALLOC, LOAD, DATA
9 .reloc 0000009c 0000000140019000 0000000140019000 00012200 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
10 .debug_aranges 00000150 000000014001a000 000000014001a000 00012400 2**0
CONTENTS, READONLY, DEBUGGING
11 .debug_info 0000d5e4 000000014001b000 000000014001b000 00012600 2**0
CONTENTS, READONLY, DEBUGGING
12 .debug_abbrev 000014de 0000000140029000 0000000140029000 0001fc00 2**0
CONTENTS, READONLY, DEBUGGING
13 .debug_line 00001a36 000000014002b000 000000014002b000 00021200 2**0
CONTENTS, READONLY, DEBUGGING
14 .debug_frame 00000f40 000000014002d000 000000014002d000 00022e00 2**0
CONTENTS, READONLY, DEBUGGING
15 .debug_str 000003a0 000000014002e000 000000014002e000 00023e00 2**0
CONTENTS, READONLY, DEBUGGING
16 .debug_line_str 00000a76 000000014002f000 000000014002f000 00024200 2**0
CONTENTS, READONLY, DEBUGGING
17 .debug_loclists 0000174a 0000000140030000 0000000140030000 00024e00 2**0
CONTENTS, READONLY, DEBUGGING
18 .debug_rnglists 0000039c 0000000140032000 0000000140032000 00026600 2**0
CONTENTS, READONLY, DEBUGGING

Maybe "loaded" instead of "loadable" ?

Frediano

#12 lore raw

Re: [PATCH 2/6] Align all sections to 4KB

Quoted 84 lines · Frediano Ziglio

Frediano

Quoted 37 lines · Jan Beulich
Quoted 35 lines · Frediano Ziglio
Quoted 33 lines · Jan Beulich
Quoted 22 lines · Frediano Ziglio
xen/arch/x86/xen.lds.S +3 -2
@@ -162,8 +162,8 @@ SECTIONS
        __note_gnu_build_id_end = .;
   } PHDR(note) PHDR(text)
 #elif defined(BUILD_ID_EFI)
-  /* Workaround bug in binutils < 2.36 */
-  . = ALIGN(32);
+  /* align to satisfy UEFI CA memory mitigation */
+  . = ALIGN(PAGE_SIZE);
   DECL_SECTION(.buildid) {
        __note_gnu_build_id_start = .;
        *(.buildid)
@@ -330,6 +330,7 @@ SECTIONS
   __2M_rwdata_end = ALIGN(SECTION_ALIGN);
 
 #ifdef EFI
+  . = ALIGN(PAGE_SIZE);
   .reloc ALIGN(4) : {
     __base_relocs_start = .;
     *(.reloc)

xen/arch/x86/xen.lds.S SECTIONS · line 362

Quoted 7 lines
xen/arch/x86/xen.lds.S +1 -0
@@ -355,6 +356,7 @@ SECTIONS
   VIRT_START &= 0;
   ALT_START &= 0;
 
+  . = ALIGN(PAGE_SIZE);
   .sbat (NOLOAD) : { *(.sbat) }
 #elif defined(XEN_BUILD_EFI)
   /*

You say "all sections" in the title, yet this is not covering e.g. debug info.

Quoted 8 lines
Quoted 2 lines

I will change to "all loadable sections". debug sections are not loadable so they don't cause an issue.

Please try to be precise there, as some aspects are subtle. As per my understanding, like .reloc all .debug_* are loadable (and may be loaded). The IMAGE_SCN_MEM_DISCARDABLE flag merely means they can be discarded at a certain point after image loading (for .reloc in particular: after relocations were processed).

No, debug sections are not loadable, for instance, in a random executable I found:

Sections:

Idx Name          Size      VMA               LMA               File off  Algn
  0 .text         0000c7f8  0000000140001000  0000000140001000  00000600  2**4

CONTENTS, ALLOC, LOAD, READONLY, CODE, DATA
1 .data 00000250 000000014000e000 000000014000e000 0000ce00 2**4
CONTENTS, ALLOC, LOAD, DATA
2 .rdata 00002c70 000000014000f000 000000014000f000 0000d200 2**4
CONTENTS, ALLOC, LOAD, READONLY, DATA
3 .pdata 00000654 0000000140012000 0000000140012000 00010000 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
4 .xdata 00000630 0000000140013000 0000000140013000 00010800 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
5 .bss 00001ec0 0000000140014000 0000000140014000 00000000 2**4
ALLOC
6 .idata 00000c88 0000000140016000 0000000140016000 00011000 2**2
CONTENTS, ALLOC, LOAD, DATA
7 .CRT 00000060 0000000140017000 0000000140017000 00011e00 2**2
CONTENTS, ALLOC, LOAD, DATA
8 .tls 00000010 0000000140018000 0000000140018000 00012000 2**2
CONTENTS, ALLOC, LOAD, DATA
9 .reloc 0000009c 0000000140019000 0000000140019000 00012200 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
10 .debug_aranges 00000150 000000014001a000 000000014001a000 00012400 2**0
CONTENTS, READONLY, DEBUGGING
11 .debug_info 0000d5e4 000000014001b000 000000014001b000 00012600 2**0
CONTENTS, READONLY, DEBUGGING
12 .debug_abbrev 000014de 0000000140029000 0000000140029000 0001fc00 2**0
CONTENTS, READONLY, DEBUGGING
13 .debug_line 00001a36 000000014002b000 000000014002b000 00021200 2**0
CONTENTS, READONLY, DEBUGGING
14 .debug_frame 00000f40 000000014002d000 000000014002d000 00022e00 2**0
CONTENTS, READONLY, DEBUGGING

Quoted 8 lines

15 .debug_str 000003a0 000000014002e000 000000014002e000 00023e00 2**0
CONTENTS, READONLY, DEBUGGING
16 .debug_line_str 00000a76 000000014002f000 000000014002f000 00024200 2**0
CONTENTS, READONLY, DEBUGGING
17 .debug_loclists 0000174a 0000000140030000 0000000140030000 00024e00 2**0
CONTENTS, READONLY, DEBUGGING
18 .debug_rnglists 0000039c 0000000140032000 0000000140032000 00026600 2**0
CONTENTS, READONLY, DEBUGGING

That's derived from libfd's internal representation, which means nothing at all to the loader processing the image. If your objdump is suitably enabled, try using its -P option.

Jan

#13 lore raw

Re: [PATCH 2/6] Align all sections to 4KB

Quoted 95 lines · Jan Beulich
Quoted 84 lines · Frediano Ziglio

Frediano

Quoted 37 lines · Jan Beulich
Quoted 35 lines · Frediano Ziglio
Quoted 33 lines · Jan Beulich
Quoted 22 lines · Frediano Ziglio
xen/arch/x86/xen.lds.S +3 -2
@@ -162,8 +162,8 @@ SECTIONS
        __note_gnu_build_id_end = .;
   } PHDR(note) PHDR(text)
 #elif defined(BUILD_ID_EFI)
-  /* Workaround bug in binutils < 2.36 */
-  . = ALIGN(32);
+  /* align to satisfy UEFI CA memory mitigation */
+  . = ALIGN(PAGE_SIZE);
   DECL_SECTION(.buildid) {
        __note_gnu_build_id_start = .;
        *(.buildid)
@@ -330,6 +330,7 @@ SECTIONS
   __2M_rwdata_end = ALIGN(SECTION_ALIGN);
 
 #ifdef EFI
+  . = ALIGN(PAGE_SIZE);
   .reloc ALIGN(4) : {
     __base_relocs_start = .;
     *(.reloc)

xen/arch/x86/xen.lds.S SECTIONS · line 362

Quoted 7 lines
xen/arch/x86/xen.lds.S +1 -0
@@ -355,6 +356,7 @@ SECTIONS
   VIRT_START &= 0;
   ALT_START &= 0;
 
+  . = ALIGN(PAGE_SIZE);
   .sbat (NOLOAD) : { *(.sbat) }
 #elif defined(XEN_BUILD_EFI)
   /*

You say "all sections" in the title, yet this is not covering e.g. debug info.

Quoted 8 lines
Quoted 2 lines

I will change to "all loadable sections". debug sections are not loadable so they don't cause an issue.

Please try to be precise there, as some aspects are subtle. As per my understanding, like .reloc all .debug_* are loadable (and may be loaded). The IMAGE_SCN_MEM_DISCARDABLE flag merely means they can be discarded at a certain point after image loading (for .reloc in particular: after relocations were processed).

No, debug sections are not loadable, for instance, in a random executable I found:

Sections:

Idx Name          Size      VMA               LMA               File off  Algn
  0 .text         0000c7f8  0000000140001000  0000000140001000  00000600  2**4

CONTENTS, ALLOC, LOAD, READONLY, CODE, DATA
1 .data 00000250 000000014000e000 000000014000e000 0000ce00 2**4
CONTENTS, ALLOC, LOAD, DATA
2 .rdata 00002c70 000000014000f000 000000014000f000 0000d200 2**4
CONTENTS, ALLOC, LOAD, READONLY, DATA
3 .pdata 00000654 0000000140012000 0000000140012000 00010000 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
4 .xdata 00000630 0000000140013000 0000000140013000 00010800 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
5 .bss 00001ec0 0000000140014000 0000000140014000 00000000 2**4
ALLOC
6 .idata 00000c88 0000000140016000 0000000140016000 00011000 2**2
CONTENTS, ALLOC, LOAD, DATA
7 .CRT 00000060 0000000140017000 0000000140017000 00011e00 2**2
CONTENTS, ALLOC, LOAD, DATA
8 .tls 00000010 0000000140018000 0000000140018000 00012000 2**2
CONTENTS, ALLOC, LOAD, DATA
9 .reloc 0000009c 0000000140019000 0000000140019000 00012200 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
10 .debug_aranges 00000150 000000014001a000 000000014001a000 00012400 2**0
CONTENTS, READONLY, DEBUGGING
11 .debug_info 0000d5e4 000000014001b000 000000014001b000 00012600 2**0
CONTENTS, READONLY, DEBUGGING
12 .debug_abbrev 000014de 0000000140029000 0000000140029000 0001fc00 2**0
CONTENTS, READONLY, DEBUGGING
13 .debug_line 00001a36 000000014002b000 000000014002b000 00021200 2**0
CONTENTS, READONLY, DEBUGGING
14 .debug_frame 00000f40 000000014002d000 000000014002d000 00022e00 2**0
CONTENTS, READONLY, DEBUGGING

Quoted 8 lines

15 .debug_str 000003a0 000000014002e000 000000014002e000 00023e00 2**0
CONTENTS, READONLY, DEBUGGING
16 .debug_line_str 00000a76 000000014002f000 000000014002f000 00024200 2**0
CONTENTS, READONLY, DEBUGGING
17 .debug_loclists 0000174a 0000000140030000 0000000140030000 00024e00 2**0
CONTENTS, READONLY, DEBUGGING
18 .debug_rnglists 0000039c 0000000140032000 0000000140032000 00026600 2**0
CONTENTS, READONLY, DEBUGGING

Quoted 5 lines

That's derived from libfd's internal representation, which means nothing at all to the loader processing the image. If your objdump is suitably enabled, try using its -P option.

Jan

You are right, I got

Section headers (at offset 0x00000188):
# Name paddr vaddr size scnptr relptr lnnoptr nrel nlnno
1 .text 0000c7f8 00001000 0000c800 00000600 00000000 00000000 0 0

            Flags: 60000060: EXECUTE,READ,CODE,INITIALIZED DATA
 2 .data    00000250 0000e000 00000400 0000ce00 00000000 00000000     0     0
            Flags: c0000040: READ,WRITE,INITIALIZED DATA
 3 .rdata   00002c70 0000f000 00002e00 0000d200 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA
 4 .pdata   00000654 00012000 00000800 00010000 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA
 5 .xdata   00000630 00013000 00000800 00010800 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA

6 .bss 00001ec0 00014000 00000000 00000000 00000000 00000000 0 0
Flags: c0000080: READ,WRITE,UNINITIALIZED DATA
7 .idata 00000c88 00016000 00000e00 00011000 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
8 .CRT 00000060 00017000 00000200 00011e00 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
9 .tls 00000010 00018000 00000200 00012000 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
10 .reloc 0000009c 00019000 00000200 00012200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
11 /4 00000150 0001a000 00000200 00012400 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
12 /19 0000d5e4 0001b000 0000d600 00012600 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
13 /31 000014de 00029000 00001600 0001fc00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
14 /45 00001a36 0002b000 00001c00 00021200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
15 /57 00000f40 0002d000 00001000 00022e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
16 /70 000003a0 0002e000 00000400 00023e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
17 /81 00000a76 0002f000 00000c00 00024200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
18 /97 0000174a 00030000 00001800 00024e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
19 /113 0000039c 00032000 00000400 00026600 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA

I suppose I will change to simply "Align some sections to 4KB"

Frediano

#14 lore raw

Re: [PATCH 2/6] Align all sections to 4KB

Quoted 145 lines · Frediano Ziglio
Quoted 95 lines · Jan Beulich
Quoted 84 lines · Frediano Ziglio

Frediano

Quoted 37 lines · Jan Beulich
Quoted 35 lines · Frediano Ziglio
Quoted 33 lines · Jan Beulich
Quoted 22 lines · Frediano Ziglio
xen/arch/x86/xen.lds.S +3 -2
@@ -162,8 +162,8 @@ SECTIONS
        __note_gnu_build_id_end = .;
   } PHDR(note) PHDR(text)
 #elif defined(BUILD_ID_EFI)
-  /* Workaround bug in binutils < 2.36 */
-  . = ALIGN(32);
+  /* align to satisfy UEFI CA memory mitigation */
+  . = ALIGN(PAGE_SIZE);
   DECL_SECTION(.buildid) {
        __note_gnu_build_id_start = .;
        *(.buildid)
@@ -330,6 +330,7 @@ SECTIONS
   __2M_rwdata_end = ALIGN(SECTION_ALIGN);
 
 #ifdef EFI
+  . = ALIGN(PAGE_SIZE);
   .reloc ALIGN(4) : {
     __base_relocs_start = .;
     *(.reloc)

xen/arch/x86/xen.lds.S SECTIONS · line 362

Quoted 7 lines
xen/arch/x86/xen.lds.S +1 -0
@@ -355,6 +356,7 @@ SECTIONS
   VIRT_START &= 0;
   ALT_START &= 0;
 
+  . = ALIGN(PAGE_SIZE);
   .sbat (NOLOAD) : { *(.sbat) }
 #elif defined(XEN_BUILD_EFI)
   /*

You say "all sections" in the title, yet this is not covering e.g. debug info.

Quoted 8 lines
Quoted 2 lines

I will change to "all loadable sections". debug sections are not loadable so they don't cause an issue.

Please try to be precise there, as some aspects are subtle. As per my understanding, like .reloc all .debug_* are loadable (and may be loaded). The IMAGE_SCN_MEM_DISCARDABLE flag merely means they can be discarded at a certain point after image loading (for .reloc in particular: after relocations were processed).

No, debug sections are not loadable, for instance, in a random executable I found:

Sections:

Idx Name          Size      VMA               LMA               File off  Algn
  0 .text         0000c7f8  0000000140001000  0000000140001000  00000600  2**4

CONTENTS, ALLOC, LOAD, READONLY, CODE, DATA
1 .data 00000250 000000014000e000 000000014000e000 0000ce00 2**4
CONTENTS, ALLOC, LOAD, DATA
2 .rdata 00002c70 000000014000f000 000000014000f000 0000d200 2**4
CONTENTS, ALLOC, LOAD, READONLY, DATA
3 .pdata 00000654 0000000140012000 0000000140012000 00010000 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
4 .xdata 00000630 0000000140013000 0000000140013000 00010800 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
5 .bss 00001ec0 0000000140014000 0000000140014000 00000000 2**4
ALLOC
6 .idata 00000c88 0000000140016000 0000000140016000 00011000 2**2
CONTENTS, ALLOC, LOAD, DATA
7 .CRT 00000060 0000000140017000 0000000140017000 00011e00 2**2
CONTENTS, ALLOC, LOAD, DATA
8 .tls 00000010 0000000140018000 0000000140018000 00012000 2**2
CONTENTS, ALLOC, LOAD, DATA
9 .reloc 0000009c 0000000140019000 0000000140019000 00012200 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
10 .debug_aranges 00000150 000000014001a000 000000014001a000 00012400 2**0
CONTENTS, READONLY, DEBUGGING
11 .debug_info 0000d5e4 000000014001b000 000000014001b000 00012600 2**0
CONTENTS, READONLY, DEBUGGING
12 .debug_abbrev 000014de 0000000140029000 0000000140029000 0001fc00 2**0
CONTENTS, READONLY, DEBUGGING
13 .debug_line 00001a36 000000014002b000 000000014002b000 00021200 2**0
CONTENTS, READONLY, DEBUGGING
14 .debug_frame 00000f40 000000014002d000 000000014002d000 00022e00 2**0
CONTENTS, READONLY, DEBUGGING

Quoted 8 lines

15 .debug_str 000003a0 000000014002e000 000000014002e000 00023e00 2**0
CONTENTS, READONLY, DEBUGGING
16 .debug_line_str 00000a76 000000014002f000 000000014002f000 00024200 2**0
CONTENTS, READONLY, DEBUGGING
17 .debug_loclists 0000174a 0000000140030000 0000000140030000 00024e00 2**0
CONTENTS, READONLY, DEBUGGING
18 .debug_rnglists 0000039c 0000000140032000 0000000140032000 00026600 2**0
CONTENTS, READONLY, DEBUGGING

Quoted 5 lines

That's derived from libfd's internal representation, which means nothing at all to the loader processing the image. If your objdump is suitably enabled, try using its -P option.

Jan

You are right, I got

Section headers (at offset 0x00000188):
# Name paddr vaddr size scnptr relptr lnnoptr nrel nlnno
1 .text 0000c7f8 00001000 0000c800 00000600 00000000 00000000 0 0

            Flags: 60000060: EXECUTE,READ,CODE,INITIALIZED DATA
 2 .data    00000250 0000e000 00000400 0000ce00 00000000 00000000     0     0
            Flags: c0000040: READ,WRITE,INITIALIZED DATA
 3 .rdata   00002c70 0000f000 00002e00 0000d200 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA
 4 .pdata   00000654 00012000 00000800 00010000 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA
 5 .xdata   00000630 00013000 00000800 00010800 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA

6 .bss 00001ec0 00014000 00000000 00000000 00000000 00000000 0 0
Flags: c0000080: READ,WRITE,UNINITIALIZED DATA
7 .idata 00000c88 00016000 00000e00 00011000 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
8 .CRT 00000060 00017000 00000200 00011e00 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
9 .tls 00000010 00018000 00000200 00012000 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
10 .reloc 0000009c 00019000 00000200 00012200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
11 /4 00000150 0001a000 00000200 00012400 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
12 /19 0000d5e4 0001b000 0000d600 00012600 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
13 /31 000014de 00029000 00001600 0001fc00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
14 /45 00001a36 0002b000 00001c00 00021200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
15 /57 00000f40 0002d000 00001000 00022e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
16 /70 000003a0 0002e000 00000400 00023e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
17 /81 00000a76 0002f000 00000c00 00024200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
18 /97 0000174a 00030000 00001800 00024e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
19 /113 0000039c 00032000 00000400 00026600 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA

Quoted 1 line

I suppose I will change to simply "Align some sections to 4KB"

"Some" is imo going to be too imprecise. Please qualify which sections you intend to align. If new sections need adding in the future, this then can guide people as to whether those may also need aligning.

Jan

#15 lore raw

Re: [PATCH 2/6] Align all sections to 4KB

Quoted 147 lines · Jan Beulich
Quoted 145 lines · Frediano Ziglio
Quoted 95 lines · Jan Beulich
Quoted 84 lines · Frediano Ziglio

Frediano

Quoted 37 lines · Jan Beulich
Quoted 35 lines · Frediano Ziglio
Quoted 33 lines · Jan Beulich
Quoted 22 lines · Frediano Ziglio
xen/arch/x86/xen.lds.S +3 -2
@@ -162,8 +162,8 @@ SECTIONS
        __note_gnu_build_id_end = .;
   } PHDR(note) PHDR(text)
 #elif defined(BUILD_ID_EFI)
-  /* Workaround bug in binutils < 2.36 */
-  . = ALIGN(32);
+  /* align to satisfy UEFI CA memory mitigation */
+  . = ALIGN(PAGE_SIZE);
   DECL_SECTION(.buildid) {
        __note_gnu_build_id_start = .;
        *(.buildid)
@@ -330,6 +330,7 @@ SECTIONS
   __2M_rwdata_end = ALIGN(SECTION_ALIGN);
 
 #ifdef EFI
+  . = ALIGN(PAGE_SIZE);
   .reloc ALIGN(4) : {
     __base_relocs_start = .;
     *(.reloc)

xen/arch/x86/xen.lds.S SECTIONS · line 362

Quoted 7 lines
xen/arch/x86/xen.lds.S +1 -0
@@ -355,6 +356,7 @@ SECTIONS
   VIRT_START &= 0;
   ALT_START &= 0;
 
+  . = ALIGN(PAGE_SIZE);
   .sbat (NOLOAD) : { *(.sbat) }
 #elif defined(XEN_BUILD_EFI)
   /*

You say "all sections" in the title, yet this is not covering e.g. debug info.

Quoted 8 lines
Quoted 2 lines

I will change to "all loadable sections". debug sections are not loadable so they don't cause an issue.

Please try to be precise there, as some aspects are subtle. As per my understanding, like .reloc all .debug_* are loadable (and may be loaded). The IMAGE_SCN_MEM_DISCARDABLE flag merely means they can be discarded at a certain point after image loading (for .reloc in particular: after relocations were processed).

No, debug sections are not loadable, for instance, in a random executable I found:

Sections:

Idx Name          Size      VMA               LMA               File off  Algn
  0 .text         0000c7f8  0000000140001000  0000000140001000  00000600  2**4

CONTENTS, ALLOC, LOAD, READONLY, CODE, DATA
1 .data 00000250 000000014000e000 000000014000e000 0000ce00 2**4
CONTENTS, ALLOC, LOAD, DATA
2 .rdata 00002c70 000000014000f000 000000014000f000 0000d200 2**4
CONTENTS, ALLOC, LOAD, READONLY, DATA
3 .pdata 00000654 0000000140012000 0000000140012000 00010000 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
4 .xdata 00000630 0000000140013000 0000000140013000 00010800 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
5 .bss 00001ec0 0000000140014000 0000000140014000 00000000 2**4
ALLOC
6 .idata 00000c88 0000000140016000 0000000140016000 00011000 2**2
CONTENTS, ALLOC, LOAD, DATA
7 .CRT 00000060 0000000140017000 0000000140017000 00011e00 2**2
CONTENTS, ALLOC, LOAD, DATA
8 .tls 00000010 0000000140018000 0000000140018000 00012000 2**2
CONTENTS, ALLOC, LOAD, DATA
9 .reloc 0000009c 0000000140019000 0000000140019000 00012200 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
10 .debug_aranges 00000150 000000014001a000 000000014001a000 00012400 2**0
CONTENTS, READONLY, DEBUGGING
11 .debug_info 0000d5e4 000000014001b000 000000014001b000 00012600 2**0
CONTENTS, READONLY, DEBUGGING
12 .debug_abbrev 000014de 0000000140029000 0000000140029000 0001fc00 2**0
CONTENTS, READONLY, DEBUGGING
13 .debug_line 00001a36 000000014002b000 000000014002b000 00021200 2**0
CONTENTS, READONLY, DEBUGGING
14 .debug_frame 00000f40 000000014002d000 000000014002d000 00022e00 2**0
CONTENTS, READONLY, DEBUGGING

Quoted 8 lines

15 .debug_str 000003a0 000000014002e000 000000014002e000 00023e00 2**0
CONTENTS, READONLY, DEBUGGING
16 .debug_line_str 00000a76 000000014002f000 000000014002f000 00024200 2**0
CONTENTS, READONLY, DEBUGGING
17 .debug_loclists 0000174a 0000000140030000 0000000140030000 00024e00 2**0
CONTENTS, READONLY, DEBUGGING
18 .debug_rnglists 0000039c 0000000140032000 0000000140032000 00026600 2**0
CONTENTS, READONLY, DEBUGGING

Quoted 5 lines

That's derived from libfd's internal representation, which means nothing at all to the loader processing the image. If your objdump is suitably enabled, try using its -P option.

Jan

You are right, I got

Section headers (at offset 0x00000188):
# Name paddr vaddr size scnptr relptr lnnoptr nrel nlnno
1 .text 0000c7f8 00001000 0000c800 00000600 00000000 00000000 0 0

            Flags: 60000060: EXECUTE,READ,CODE,INITIALIZED DATA
 2 .data    00000250 0000e000 00000400 0000ce00 00000000 00000000     0     0
            Flags: c0000040: READ,WRITE,INITIALIZED DATA
 3 .rdata   00002c70 0000f000 00002e00 0000d200 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA
 4 .pdata   00000654 00012000 00000800 00010000 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA
 5 .xdata   00000630 00013000 00000800 00010800 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA

6 .bss 00001ec0 00014000 00000000 00000000 00000000 00000000 0 0
Flags: c0000080: READ,WRITE,UNINITIALIZED DATA
7 .idata 00000c88 00016000 00000e00 00011000 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
8 .CRT 00000060 00017000 00000200 00011e00 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
9 .tls 00000010 00018000 00000200 00012000 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
10 .reloc 0000009c 00019000 00000200 00012200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
11 /4 00000150 0001a000 00000200 00012400 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
12 /19 0000d5e4 0001b000 0000d600 00012600 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
13 /31 000014de 00029000 00001600 0001fc00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
14 /45 00001a36 0002b000 00001c00 00021200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
15 /57 00000f40 0002d000 00001000 00022e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
16 /70 000003a0 0002e000 00000400 00023e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
17 /81 00000a76 0002f000 00000c00 00024200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
18 /97 0000174a 00030000 00001800 00024e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
19 /113 0000039c 00032000 00000400 00026600 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA

Quoted 7 lines
Quoted 1 line

I suppose I will change to simply "Align some sections to 4KB"

"Some" is imo going to be too imprecise. Please qualify which sections you intend to align. If new sections need adding in the future, this then can guide people as to whether those may also need aligning.

Jan

The main issue is that sections with different permissions must be in separate sections.
In the case of debug sections they are contiguous and have the same permissions so it's not an issue (although better to strip them off).

Any suggestions on how to describe this?

Frediano

#16 lore raw

Re: [PATCH 2/6] Align all sections to 4KB

Quoted 154 lines · Frediano Ziglio
Quoted 153 lines · Jan Beulich
Quoted 145 lines · Frediano Ziglio
Quoted 95 lines · Jan Beulich
Quoted 84 lines · Frediano Ziglio

Frediano

Quoted 37 lines · Jan Beulich
Quoted 35 lines · Frediano Ziglio
Quoted 33 lines · Jan Beulich
Quoted 22 lines · Frediano Ziglio
xen/arch/x86/xen.lds.S +3 -2
@@ -162,8 +162,8 @@ SECTIONS
        __note_gnu_build_id_end = .;
   } PHDR(note) PHDR(text)
 #elif defined(BUILD_ID_EFI)
-  /* Workaround bug in binutils < 2.36 */
-  . = ALIGN(32);
+  /* align to satisfy UEFI CA memory mitigation */
+  . = ALIGN(PAGE_SIZE);
   DECL_SECTION(.buildid) {
        __note_gnu_build_id_start = .;
        *(.buildid)
@@ -330,6 +330,7 @@ SECTIONS
   __2M_rwdata_end = ALIGN(SECTION_ALIGN);
 
 #ifdef EFI
+  . = ALIGN(PAGE_SIZE);
   .reloc ALIGN(4) : {
     __base_relocs_start = .;
     *(.reloc)

xen/arch/x86/xen.lds.S SECTIONS · line 362

Quoted 7 lines
xen/arch/x86/xen.lds.S +1 -0
@@ -355,6 +356,7 @@ SECTIONS
   VIRT_START &= 0;
   ALT_START &= 0;
 
+  . = ALIGN(PAGE_SIZE);
   .sbat (NOLOAD) : { *(.sbat) }
 #elif defined(XEN_BUILD_EFI)
   /*

You say "all sections" in the title, yet this is not covering e.g. debug info.

Quoted 8 lines
Quoted 2 lines

I will change to "all loadable sections". debug sections are not loadable so they don't cause an issue.

Please try to be precise there, as some aspects are subtle. As per my understanding, like .reloc all .debug_* are loadable (and may be loaded). The IMAGE_SCN_MEM_DISCARDABLE flag merely means they can be discarded at a certain point after image loading (for .reloc in particular: after relocations were processed).

No, debug sections are not loadable, for instance, in a random executable I found:

Sections:

Idx Name          Size      VMA               LMA               File off  Algn
  0 .text         0000c7f8  0000000140001000  0000000140001000  00000600  2**4

CONTENTS, ALLOC, LOAD, READONLY, CODE, DATA
1 .data 00000250 000000014000e000 000000014000e000 0000ce00 2**4
CONTENTS, ALLOC, LOAD, DATA
2 .rdata 00002c70 000000014000f000 000000014000f000 0000d200 2**4
CONTENTS, ALLOC, LOAD, READONLY, DATA
3 .pdata 00000654 0000000140012000 0000000140012000 00010000 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
4 .xdata 00000630 0000000140013000 0000000140013000 00010800 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
5 .bss 00001ec0 0000000140014000 0000000140014000 00000000 2**4
ALLOC
6 .idata 00000c88 0000000140016000 0000000140016000 00011000 2**2
CONTENTS, ALLOC, LOAD, DATA
7 .CRT 00000060 0000000140017000 0000000140017000 00011e00 2**2
CONTENTS, ALLOC, LOAD, DATA
8 .tls 00000010 0000000140018000 0000000140018000 00012000 2**2
CONTENTS, ALLOC, LOAD, DATA
9 .reloc 0000009c 0000000140019000 0000000140019000 00012200 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
10 .debug_aranges 00000150 000000014001a000 000000014001a000 00012400 2**0
CONTENTS, READONLY, DEBUGGING
11 .debug_info 0000d5e4 000000014001b000 000000014001b000 00012600 2**0
CONTENTS, READONLY, DEBUGGING
12 .debug_abbrev 000014de 0000000140029000 0000000140029000 0001fc00 2**0
CONTENTS, READONLY, DEBUGGING
13 .debug_line 00001a36 000000014002b000 000000014002b000 00021200 2**0
CONTENTS, READONLY, DEBUGGING
14 .debug_frame 00000f40 000000014002d000 000000014002d000 00022e00 2**0
CONTENTS, READONLY, DEBUGGING

Quoted 8 lines

15 .debug_str 000003a0 000000014002e000 000000014002e000 00023e00 2**0
CONTENTS, READONLY, DEBUGGING
16 .debug_line_str 00000a76 000000014002f000 000000014002f000 00024200 2**0
CONTENTS, READONLY, DEBUGGING
17 .debug_loclists 0000174a 0000000140030000 0000000140030000 00024e00 2**0
CONTENTS, READONLY, DEBUGGING
18 .debug_rnglists 0000039c 0000000140032000 0000000140032000 00026600 2**0
CONTENTS, READONLY, DEBUGGING

Quoted 5 lines

That's derived from libfd's internal representation, which means nothing at all to the loader processing the image. If your objdump is suitably enabled, try using its -P option.

Jan

You are right, I got

Section headers (at offset 0x00000188):
# Name paddr vaddr size scnptr relptr lnnoptr nrel nlnno
1 .text 0000c7f8 00001000 0000c800 00000600 00000000 00000000 0 0

            Flags: 60000060: EXECUTE,READ,CODE,INITIALIZED DATA
 2 .data    00000250 0000e000 00000400 0000ce00 00000000 00000000     0     0
            Flags: c0000040: READ,WRITE,INITIALIZED DATA
 3 .rdata   00002c70 0000f000 00002e00 0000d200 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA
 4 .pdata   00000654 00012000 00000800 00010000 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA
 5 .xdata   00000630 00013000 00000800 00010800 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA

6 .bss 00001ec0 00014000 00000000 00000000 00000000 00000000 0 0
Flags: c0000080: READ,WRITE,UNINITIALIZED DATA
7 .idata 00000c88 00016000 00000e00 00011000 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
8 .CRT 00000060 00017000 00000200 00011e00 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
9 .tls 00000010 00018000 00000200 00012000 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
10 .reloc 0000009c 00019000 00000200 00012200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
11 /4 00000150 0001a000 00000200 00012400 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
12 /19 0000d5e4 0001b000 0000d600 00012600 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
13 /31 000014de 00029000 00001600 0001fc00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
14 /45 00001a36 0002b000 00001c00 00021200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
15 /57 00000f40 0002d000 00001000 00022e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
16 /70 000003a0 0002e000 00000400 00023e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
17 /81 00000a76 0002f000 00000c00 00024200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
18 /97 0000174a 00030000 00001800 00024e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
19 /113 0000039c 00032000 00000400 00026600 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA

Quoted 1 line

I suppose I will change to simply "Align some sections to 4KB"

"Some" is imo going to be too imprecise. Please qualify which sections you intend to align. If new sections need adding in the future, this then can guide people as to whether those may also need aligning.

Quoted 8 lines
Quoted 1 line

Jan

The main issue is that sections with different permissions must be in separate sections.
In the case of debug sections they are contiguous and have the same permissions so it's not an issue (although better to strip them off).

Any suggestions on how to describe this?

Maybe something like "Align sections of different permissions to force them into separate pages"?

Signature · 3 lines
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab
#17 lore raw

Re: [PATCH 2/6] Align all sections to 4KB

In reply to Frediano Ziglio
Quoted 154 lines · Frediano Ziglio
Quoted 147 lines · Jan Beulich
Quoted 145 lines · Frediano Ziglio
Quoted 95 lines · Jan Beulich
Quoted 84 lines · Frediano Ziglio

Frediano

Quoted 37 lines · Jan Beulich
Quoted 35 lines · Frediano Ziglio
Quoted 33 lines · Jan Beulich
Quoted 22 lines · Frediano Ziglio
xen/arch/x86/xen.lds.S +3 -2
@@ -162,8 +162,8 @@ SECTIONS
        __note_gnu_build_id_end = .;
   } PHDR(note) PHDR(text)
 #elif defined(BUILD_ID_EFI)
-  /* Workaround bug in binutils < 2.36 */
-  . = ALIGN(32);
+  /* align to satisfy UEFI CA memory mitigation */
+  . = ALIGN(PAGE_SIZE);
   DECL_SECTION(.buildid) {
        __note_gnu_build_id_start = .;
        *(.buildid)
@@ -330,6 +330,7 @@ SECTIONS
   __2M_rwdata_end = ALIGN(SECTION_ALIGN);
 
 #ifdef EFI
+  . = ALIGN(PAGE_SIZE);
   .reloc ALIGN(4) : {
     __base_relocs_start = .;
     *(.reloc)

xen/arch/x86/xen.lds.S SECTIONS · line 362

Quoted 7 lines
xen/arch/x86/xen.lds.S +1 -0
@@ -355,6 +356,7 @@ SECTIONS
   VIRT_START &= 0;
   ALT_START &= 0;
 
+  . = ALIGN(PAGE_SIZE);
   .sbat (NOLOAD) : { *(.sbat) }
 #elif defined(XEN_BUILD_EFI)
   /*

You say "all sections" in the title, yet this is not covering e.g. debug info.

Quoted 8 lines
Quoted 2 lines

I will change to "all loadable sections". debug sections are not loadable so they don't cause an issue.

Please try to be precise there, as some aspects are subtle. As per my understanding, like .reloc all .debug_* are loadable (and may be loaded). The IMAGE_SCN_MEM_DISCARDABLE flag merely means they can be discarded at a certain point after image loading (for .reloc in particular: after relocations were processed).

No, debug sections are not loadable, for instance, in a random executable I found:

Sections:

Idx Name          Size      VMA               LMA               File off  Algn
  0 .text         0000c7f8  0000000140001000  0000000140001000  00000600  2**4

CONTENTS, ALLOC, LOAD, READONLY, CODE, DATA
1 .data 00000250 000000014000e000 000000014000e000 0000ce00 2**4
CONTENTS, ALLOC, LOAD, DATA
2 .rdata 00002c70 000000014000f000 000000014000f000 0000d200 2**4
CONTENTS, ALLOC, LOAD, READONLY, DATA
3 .pdata 00000654 0000000140012000 0000000140012000 00010000 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
4 .xdata 00000630 0000000140013000 0000000140013000 00010800 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
5 .bss 00001ec0 0000000140014000 0000000140014000 00000000 2**4
ALLOC
6 .idata 00000c88 0000000140016000 0000000140016000 00011000 2**2
CONTENTS, ALLOC, LOAD, DATA
7 .CRT 00000060 0000000140017000 0000000140017000 00011e00 2**2
CONTENTS, ALLOC, LOAD, DATA
8 .tls 00000010 0000000140018000 0000000140018000 00012000 2**2
CONTENTS, ALLOC, LOAD, DATA
9 .reloc 0000009c 0000000140019000 0000000140019000 00012200 2**2
CONTENTS, ALLOC, LOAD, READONLY, DATA
10 .debug_aranges 00000150 000000014001a000 000000014001a000 00012400 2**0
CONTENTS, READONLY, DEBUGGING
11 .debug_info 0000d5e4 000000014001b000 000000014001b000 00012600 2**0
CONTENTS, READONLY, DEBUGGING
12 .debug_abbrev 000014de 0000000140029000 0000000140029000 0001fc00 2**0
CONTENTS, READONLY, DEBUGGING
13 .debug_line 00001a36 000000014002b000 000000014002b000 00021200 2**0
CONTENTS, READONLY, DEBUGGING
14 .debug_frame 00000f40 000000014002d000 000000014002d000 00022e00 2**0
CONTENTS, READONLY, DEBUGGING

Quoted 8 lines

15 .debug_str 000003a0 000000014002e000 000000014002e000 00023e00 2**0
CONTENTS, READONLY, DEBUGGING
16 .debug_line_str 00000a76 000000014002f000 000000014002f000 00024200 2**0
CONTENTS, READONLY, DEBUGGING
17 .debug_loclists 0000174a 0000000140030000 0000000140030000 00024e00 2**0
CONTENTS, READONLY, DEBUGGING
18 .debug_rnglists 0000039c 0000000140032000 0000000140032000 00026600 2**0
CONTENTS, READONLY, DEBUGGING

Quoted 5 lines

That's derived from libfd's internal representation, which means nothing at all to the loader processing the image. If your objdump is suitably enabled, try using its -P option.

Jan

You are right, I got

Section headers (at offset 0x00000188):
# Name paddr vaddr size scnptr relptr lnnoptr nrel nlnno
1 .text 0000c7f8 00001000 0000c800 00000600 00000000 00000000 0 0

            Flags: 60000060: EXECUTE,READ,CODE,INITIALIZED DATA
 2 .data    00000250 0000e000 00000400 0000ce00 00000000 00000000     0     0
            Flags: c0000040: READ,WRITE,INITIALIZED DATA
 3 .rdata   00002c70 0000f000 00002e00 0000d200 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA
 4 .pdata   00000654 00012000 00000800 00010000 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA
 5 .xdata   00000630 00013000 00000800 00010800 00000000 00000000     0     0
            Flags: 40000040: READ,INITIALIZED DATA

6 .bss 00001ec0 00014000 00000000 00000000 00000000 00000000 0 0
Flags: c0000080: READ,WRITE,UNINITIALIZED DATA
7 .idata 00000c88 00016000 00000e00 00011000 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
8 .CRT 00000060 00017000 00000200 00011e00 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
9 .tls 00000010 00018000 00000200 00012000 00000000 00000000 0 0
Flags: c0000040: READ,WRITE,INITIALIZED DATA
10 .reloc 0000009c 00019000 00000200 00012200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
11 /4 00000150 0001a000 00000200 00012400 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
12 /19 0000d5e4 0001b000 0000d600 00012600 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
13 /31 000014de 00029000 00001600 0001fc00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
14 /45 00001a36 0002b000 00001c00 00021200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
15 /57 00000f40 0002d000 00001000 00022e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
16 /70 000003a0 0002e000 00000400 00023e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
17 /81 00000a76 0002f000 00000c00 00024200 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
18 /97 0000174a 00030000 00001800 00024e00 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA
19 /113 0000039c 00032000 00000400 00026600 00000000 00000000 0 0
Flags: 42000040: DISCARDABLE,READ,INITIALIZED DATA

Quoted 5 lines
Quoted 1 line

I suppose I will change to simply "Align some sections to 4KB"

"Some" is imo going to be too imprecise. Please qualify which sections you intend to align. If new sections need adding in the future, this then can guide people as to whether those may also need aligning.

Quoted 6 lines

The main issue is that sections with different permissions must be in separate sections.
In the case of debug sections they are contiguous and have the same permissions so it's not an issue (although better to strip them off).

Any suggestions on how to describe this?

s/all/relevant/ (or some such) in the subject and then largely what you say above as part of the description.

Jan

#18 lore raw

PATCH 3/6 x86/efi: discard .text.header for PE binary

In reply to Frediano Ziglio

From: Roger Pau Monné <[email protected]>

The multiboot headers are not consumed in the PE binary, hence discard them in the linker script when doing a PE build.

That removes some relocations that otherwise appear due to the usage of the start and __efi64_mb2_start symbols in the multiboot2 header.

No functional change intended.

---
 xen/arch/x86/xen.lds.S | 3 +++
 1 file changed, 3 insertions(+)
xen/arch/x86/xen.lds.S +3 -0
index 322c116a04..907f826ae0 100644
@@ -57,6 +57,9 @@ SECTIONS
   __image_base__ = .;
 #else
   . = __image_base__;
+  /DISCARD/ : {
+    *(.text.header)
+  }
 #endif
 
 #if 0
Signature · 1 line
2.43.0
#19 lore raw

Re: [PATCH 3/6] x86/efi: discard .text.header for PE binary

Quoted 4 lines · Frediano Ziglio

From: Roger Pau Monné <[email protected]>

The multiboot headers are not consumed in the PE binary, hence discard them in the linker script when doing a PE build.

Yet weren't there plans to allow making use of those headers even in xen.efi?

Quoted 9 lines

That removes some relocations that otherwise appear due to the usage of the start and __efi64_mb2_start symbols in the multiboot2 header.

No functional change intended.

---
 xen/arch/x86/xen.lds.S | 3 +++
 1 file changed, 3 insertions(+)

This is lacking your own S-o-b.

xen/arch/x86/xen.lds.S SECTIONS · line 63

Quoted 10 lines
xen/arch/x86/xen.lds.S +3 -0
@@ -57,6 +57,9 @@ SECTIONS
   __image_base__ = .;
 #else
   . = __image_base__;
+  /DISCARD/ : {
+    *(.text.header)
+  }
 #endif

We already have DISCARD_SECTIONS, so I think it needs clarifying why that can't be used / extended.

Jan

#20 lore raw

Re: [PATCH 3/6] x86/efi: discard .text.header for PE binary

Quoted 9 lines · Jan Beulich
Quoted 4 lines · Frediano Ziglio

From: Roger Pau Monné <[email protected]>

The multiboot headers are not consumed in the PE binary, hence discard them in the linker script when doing a PE build.

Yet weren't there plans to allow making use of those headers even in xen.efi?

That method was replaced by different ways of loading. Using multiboot for PE would require some extensions to multiboot as the protocol is not designed to support PE format. There are already ways to use EFI and configuration files that can be used. We (as Citrix) are going (not me personally) to post changes to extend xen_hypervisor and xen_module (used currently only by ARM) to x64. These changes are much smaller, maintainable and coherent.

Quoted 12 lines
Quoted 9 lines

That removes some relocations that otherwise appear due to the usage of the start and __efi64_mb2_start symbols in the multiboot2 header.

No functional change intended.

---
 xen/arch/x86/xen.lds.S | 3 +++
 1 file changed, 3 insertions(+)

This is lacking your own S-o-b.

Quoted 14 lines

xen/arch/x86/xen.lds.S SECTIONS · line 63

Quoted 10 lines
xen/arch/x86/xen.lds.S +3 -0
@@ -57,6 +57,9 @@ SECTIONS
   __image_base__ = .;
 #else
   . = __image_base__;
+  /DISCARD/ : {
+    *(.text.header)
+  }
 #endif

We already have DISCARD_SECTIONS, so I think it needs clarifying why that can't be used / extended.

Yes, this can be done that way. Roger, do you agree on this ?

Quoted 1 line

Jan

Frediano

#21 lore raw

Re: [PATCH 3/6] x86/efi: discard .text.header for PE binary

Quoted 14 lines · Frediano Ziglio
Quoted 4 lines · Jan Beulich
Quoted 4 lines
xen/arch/x86/xen.lds.S +0 -0
@@ -57,6 +57,9 @@ SECTIONS
   __image_base__ = .;
Quoted 9 lines

xen/arch/x86/xen.lds.S SECTIONS · line 63

Quoted 6 lines
xen/arch/x86/xen.lds.S +3 -0
 #else
   . = __image_base__;
+  /DISCARD/ : {
+    *(.text.header)
+  }
 #endif

We already have DISCARD_SECTIONS, so I think it needs clarifying why that can't be used / extended.

Quoted 1 line

Yes, this can be done that way. Roger, do you agree on this ?

You'll need to add a general ARCH_DISCARD_SECTIONS, and have x86 fill in {ELF,EFI}_DISCARD_SECTIONS conditionally.

Also you need to fold this hunk.

docs/hypervisor-guide/x86/how-xen-boots.rst +0 -0
index 8b3229005cca..b6d852050a0c 100644
@@ -82,12 +82,6 @@ When a PEI-capable toolchain is found, the objects are linked together and a

 PE32+ binary is created.  It can be run directly from the EFI shell, and has ``efi_start`` as its entry symbol.

-.. note::
-
-   xen.efi does contain all MB1/MB2/PVH tags included in the rest of the -   build.  However, entry via anything other than the EFI64 protocol is -   unsupported, and won't work.
-

 Boot
 ----

~Andrew

#22 lore raw

Re: [PATCH 3/6] x86/efi: discard .text.header for PE binary

Quoted 16 lines · Andrew Cooper
Quoted 14 lines · Frediano Ziglio
Quoted 13 lines · Jan Beulich

xen/arch/x86/xen.lds.S SECTIONS · line 63

Quoted 10 lines
xen/arch/x86/xen.lds.S +3 -0
@@ -57,6 +57,9 @@ SECTIONS
   __image_base__ = .;
 #else
   . = __image_base__;
+  /DISCARD/ : {
+    *(.text.header)
+  }
 #endif

We already have DISCARD_SECTIONS, so I think it needs clarifying why that can't be used / extended.

Quoted 5 lines
Quoted 1 line

Yes, this can be done that way. Roger, do you agree on this ?

You'll need to add a general ARCH_DISCARD_SECTIONS, and have x86 fill in {ELF,EFI}_DISCARD_SECTIONS conditionally.

I think binutils simply ignore sections not present so this is not necessary. I'll test it.

Quoted 10 lines

Also you need to fold this hunk.

docs/hypervisor-guide/x86/how-xen-boots.rst +0 -0
index 8b3229005cca..b6d852050a0c 100644
@@ -82,12 +82,6 @@ When a PEI-capable toolchain is found, the objects are linked together and a
 PE32+ binary is created.  It can be run directly from the EFI shell, and has
 ``efi_start`` as its entry symbol.

docs/hypervisor-guide/x86/how-xen-boots.rst When a PEI-capable toolchain is found, the objects are linked together and a · line 87

Quoted 10 lines
docs/hypervisor-guide/x86/how-xen-boots.rst +0 -6
 
-.. note::
-
-   xen.efi does contain all MB1/MB2/PVH tags included in the rest of the
-   build.  However, entry via anything other than the EFI64 protocol is
-   unsupported, and won't work.
-
 
 Boot
 ----

Sure, thanks

Quoted 3 lines

~Andrew

Frediano

#23 lore raw

Re: [PATCH 3/6] x86/efi: discard .text.header for PE binary

Quoted 16 lines · Frediano Ziglio
Quoted 15 lines · Andrew Cooper
Quoted 14 lines · Frediano Ziglio
Quoted 13 lines · Jan Beulich

xen/arch/x86/xen.lds.S SECTIONS · line 63

Quoted 10 lines
xen/arch/x86/xen.lds.S +3 -0
@@ -57,6 +57,9 @@ SECTIONS
   __image_base__ = .;
 #else
   . = __image_base__;
+  /DISCARD/ : {
+    *(.text.header)
+  }
 #endif

We already have DISCARD_SECTIONS, so I think it needs clarifying why that can't be used / extended.

Quoted 1 line
Quoted 1 line
Quoted 1 line

Yes, this can be done that way. Roger, do you agree on this ?

FTOAD: I did _not_ suggest to make use of DISCARD_SECTIONS, as that won't work. Instead I did ask that it be clarified in the description why DISCARD_SECTIONS cannot be extended for the purpose here.

Quoted 5 lines
Quoted 2 lines

You'll need to add a general ARCH_DISCARD_SECTIONS, and have x86 fill in {ELF,EFI}_DISCARD_SECTIONS conditionally.

I think binutils simply ignore sections not present so this is not necessary. I'll test it.

But you can't discard .text.header blindly everywhere. The reason it can be discarded on x86, and only for xen.efi, is very much x86-specific.

Jan

#24 lore raw

PATCH 4/6 x86/efi: discard multiboot related entry code for PE binary

In reply to Frediano Ziglio

From: Roger Pau Monné <[email protected]>

The multiboot and PVH entry points are not used in the PE binary, hence discard them in the linker script when doing a PE build.

That removes some relocations that otherwise appear due to the entry point code in head.S not being position independent.

No functional change intended.

---
 xen/arch/x86/boot/head.S | 3 ++-
 xen/arch/x86/xen.lds.S   | 2 ++
 2 files changed, 4 insertions(+), 1 deletion(-)
xen/arch/x86/boot/head.S +2 -1
index 77bb7a9e21..90faf411b9 100644
@@ -152,7 +152,7 @@ vga_text_buffer:
 efi_platform:
         .byte   0
 
-        .section .init.text, "ax", @progbits
+        .section .init.multiboot, "ax", @progbits
 
 early_error: /* Here to improve the disassembly. */
 
@@ -710,6 +710,7 @@ trampoline_setup:
         /* Jump into the relocated trampoline. */
         lret
 
+        .section .init.text, "ax", @progbits
 ENTRY(trampoline_start)
 #include "trampoline.S"
 ENTRY(trampoline_end)
xen/arch/x86/xen.lds.S +2 -0
index 907f826ae0..a5a85e9b8a 100644
@@ -59,6 +59,7 @@ SECTIONS
   . = __image_base__;
   /DISCARD/ : {
     *(.text.header)
+    *(.init.multiboot)
   }
 #endif
 
@@ -200,6 +201,7 @@ SECTIONS
        _sinittext = .;
        *(.init.text)
        *(.text.startup)
+       *(.init.multiboot)
        _einittext = .;
        /*
         * Here are the replacement instructions. The linker sticks them
Signature · 1 line
2.43.0
#25 lore raw

Re: [PATCH 4/6] x86/efi: discard multiboot related entry code for PE binary

Quoted 8 lines · Frediano Ziglio

From: Roger Pau Monné <[email protected]>

The multiboot and PVH entry points are not used in the PE binary, hence discard them in the linker script when doing a PE build.

That removes some relocations that otherwise appear due to the entry point code in head.S not being position independent.

Quoted 7 lines

No functional change intended.

---
 xen/arch/x86/boot/head.S | 3 ++-
 xen/arch/x86/xen.lds.S   | 2 ++
 2 files changed, 4 insertions(+), 1 deletion(-)

S-o-b issue again.

xen/arch/x86/xen.lds.S SECTIONS · line 205

Quoted 6 lines
xen/arch/x86/xen.lds.S +1 -0
@@ -200,6 +201,7 @@ SECTIONS
        _sinittext = .;
        *(.init.text)
        *(.text.startup)
+       *(.init.multiboot)
        _einittext = .;

Aren't you making a needlessly big change to the final image by placing this last rather than first?

Jan

#26 lore raw

Re: [PATCH 4/6] x86/efi: discard multiboot related entry code for PE binary

Quoted 12 lines · Jan Beulich
Quoted 10 lines · Frediano Ziglio

From: Roger Pau Monné <[email protected]>

The multiboot and PVH entry points are not used in the PE binary, hence discard them in the linker script when doing a PE build.

That removes some relocations that otherwise appear due to the entry point code in head.S not being position independent.

No functional change intended.

Quoted 8 lines
Quoted 5 lines
---
 xen/arch/x86/boot/head.S | 3 ++-
 xen/arch/x86/xen.lds.S   | 2 ++
 2 files changed, 4 insertions(+), 1 deletion(-)

S-o-b issue again.

Even if I didn't make any changes to the original patch ?

Quoted 10 lines

xen/arch/x86/xen.lds.S SECTIONS · line 205

Quoted 6 lines
xen/arch/x86/xen.lds.S +1 -0
@@ -200,6 +201,7 @@ SECTIONS
        _sinittext = .;
        *(.init.text)
        *(.text.startup)
+       *(.init.multiboot)
        _einittext = .;

Aren't you making a needlessly big change to the final image by placing this last rather than first?

Yes, I agree, .init.multiboot section should be even before .init.test. I'll change it.

Quoted 1 line

Jan

Frediano

#27 lore raw

Re: [PATCH 4/6] x86/efi: discard multiboot related entry code for PE binary

Quoted 12 lines · Frediano Ziglio
Quoted 11 lines · Jan Beulich
Quoted 10 lines · Frediano Ziglio

From: Roger Pau Monné <[email protected]>

The multiboot and PVH entry points are not used in the PE binary, hence discard them in the linker script when doing a PE build.

That removes some relocations that otherwise appear due to the entry point code in head.S not being position independent.

No functional change intended.

Quoted 8 lines
Quoted 7 lines
Quoted 5 lines
---
 xen/arch/x86/boot/head.S | 3 ++-
 xen/arch/x86/xen.lds.S   | 2 ++
 2 files changed, 4 insertions(+), 1 deletion(-)

S-o-b issue again.

Even if I didn't make any changes to the original patch ?

Yes.  Everything you submit must have your SoB, per the DOC.  It is a declaration that, to the best of your knowledge, the license is compatible.

Honestly, I think this patch would be better merged with the prior one. Both are taking out multiboot content.

~Andrew

#28 lore raw

PATCH 5/6 x86/efi: avoid a relocation in efi_arch_post_exit_boot()

In reply to Frediano Ziglio

From: Roger Pau Monné <[email protected]>

Instead of using the absolute __start_xen address, calculate it as an offset from the current instruction pointer. The relocation would be problematic if the loader has acknowledged the Xen image section attributes, and mapped .init.text with just read and execute permissions.

No functional change intended.

---
 xen/arch/x86/efi/efi-boot.h | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
xen/arch/x86/efi/efi-boot.h +5 -2
index d738b839ee..b983f054b5 100644
@@ -270,7 +270,9 @@ static void __init noreturn efi_arch_post_exit_boot(void)
 
                    /* Jump to higher mappings. */
                    "mov    stack_start(%%rip), %%rsp\n\t"
-                   "movabs $__start_xen, %[rip]\n\t"
+                   "lea    __start_xen(%%rip), %[rip]\n\t"
+                   "add    %[offset], %[rip]\n\t"
+
                    "push   %[cs]\n\t"
                    "push   %[rip]\n\t"
                    "lretq"
@@ -278,7 +280,8 @@ static void __init noreturn efi_arch_post_exit_boot(void)
                      [cr4] "+&r" (cr4)
                    : [cr3] "r" (idle_pg_table),
                      [cs] "i" (__HYPERVISOR_CS),
-                     [ds] "r" (__HYPERVISOR_DS)
+                     [ds] "r" (__HYPERVISOR_DS),
+                     [offset] "r" (__XEN_VIRT_START - xen_phys_start)
                    : "memory" );
     unreachable();
 }
Signature · 1 line
2.43.0
#29 lore raw

Re: [PATCH 5/6] x86/efi: avoid a relocation in efi_arch_post_exit_boot()

Quoted 6 lines · Frediano Ziglio

From: Roger Pau Monné <[email protected]>

Instead of using the absolute __start_xen address, calculate it as an offset from the current instruction pointer. The relocation would be problematic if the loader has acknowledged the Xen image section attributes, and mapped .init.text with just read and execute permissions.

How can the loader respect what .init.text wants when it sees only .init, prior to the subsequent patch? Was the series meant to be ordered differently?

Jan

#30 lore raw

Re: [PATCH 5/6] x86/efi: avoid a relocation in efi_arch_post_exit_boot()

Quoted 10 lines · Jan Beulich
Quoted 6 lines · Frediano Ziglio

From: Roger Pau Monné <[email protected]>

Instead of using the absolute __start_xen address, calculate it as an offset from the current instruction pointer. The relocation would be problematic if the loader has acknowledged the Xen image section attributes, and mapped .init.text with just read and execute permissions.

How can the loader respect what .init.text wants when it sees only .init, prior to the subsequent patch? Was the series meant to be ordered differently?

No.  Xen explodes if you split .init and this relocation is present.

It's the part where Xen undoes the relocation that the UEFI loader did.

Removing this relocation is necessary to make it safe to split .init.

~Andrew

#31 lore raw

Re: [PATCH 5/6] x86/efi: avoid a relocation in efi_arch_post_exit_boot()

Quoted 12 lines · Andrew Cooper
Quoted 10 lines · Jan Beulich
Quoted 6 lines · Frediano Ziglio

From: Roger Pau Monné <[email protected]>

Instead of using the absolute __start_xen address, calculate it as an offset from the current instruction pointer. The relocation would be problematic if the loader has acknowledged the Xen image section attributes, and mapped .init.text with just read and execute permissions.

How can the loader respect what .init.text wants when it sees only .init, prior to the subsequent patch? Was the series meant to be ordered differently?

Quoted 5 lines

No.  Xen explodes if you split .init and this relocation is present.

It's the part where Xen undoes the relocation that the UEFI loader did.

Removing this relocation is necessary to make it safe to split .init.

Okay, so ordering of the series is intentional as is, but the description here is imprecise.

Jan

#32 lore raw

PATCH 6/6 x86: Split .init section to satisfy UEFI CA memory mitigation

In reply to Frediano Ziglio

Currently .init section is both writeable and executable, split data and code to have 2 sections satisfying W^X rule.

---
 xen/arch/x86/xen.lds.S | 12 ++++--------
 1 file changed, 4 insertions(+), 8 deletions(-)
xen/arch/x86/xen.lds.S +4 -8
index a5a85e9b8a..7f892cb2fa 100644
@@ -193,11 +193,7 @@ SECTIONS
   __2M_init_start = .;         /* Start of 2M superpages, mapped RWX (boot only). */
   . = ALIGN(PAGE_SIZE);             /* Init code and data */
   __init_begin = .;
-#ifdef EFI /* EFI wants to merge all of .init.*  ELF doesn't. */
-  DECL_SECTION(.init) {
-#else
   DECL_SECTION(.init.text) {
-#endif
        _sinittext = .;
        *(.init.text)
        *(.text.startup)
@@ -210,12 +206,12 @@ SECTIONS
         */
        *(.altinstr_replacement)
 
-#ifdef EFI /* EFI wants to merge all of .init.*  ELF doesn't. */
-       . = ALIGN(SMP_CACHE_BYTES);
-#else
   } PHDR(text)
-  DECL_SECTION(.init.data) {
+#ifdef EFI
+  /* align to satisfy UEFI CA memory mitigation */
+  . = ALIGN(SECTION_ALIGN);
 #endif
+  DECL_SECTION(.init.data) {
        *(.init.bss.stack_aligned)
        *(.init.data.page_aligned)
 
Signature · 1 line
2.43.0
#33 lore raw

Re: [PATCH 6/6] x86: Split .init section to satisfy UEFI CA memory mitigation

Quoted 4 lines · Frediano Ziglio

Currently .init section is both writeable and executable, split data and code to have 2 sections satisfying W^X rule.

Like for the earlier patch: Reference please to where this requirement is stated.

Quoted 16 lines
xen/arch/x86/xen.lds.S +0 -4
@@ -193,11 +193,7 @@ SECTIONS
   __2M_init_start = .;         /* Start of 2M superpages, mapped RWX (boot only). */
   . = ALIGN(PAGE_SIZE);             /* Init code and data */
   __init_begin = .;
-#ifdef EFI /* EFI wants to merge all of .init.*  ELF doesn't. */
-  DECL_SECTION(.init) {
-#else
   DECL_SECTION(.init.text) {
-#endif
        _sinittext = .;
        *(.init.text)
        *(.text.startup)
@@ -210,12 +206,12 @@ SECTIONS
         */

xen/arch/x86/xen.lds.S SECTIONS · line 212

Quoted 10 lines
xen/arch/x86/xen.lds.S +3 -4
        *(.altinstr_replacement)
 
-#ifdef EFI /* EFI wants to merge all of .init.*  ELF doesn't. */
-       . = ALIGN(SMP_CACHE_BYTES);
-#else
   } PHDR(text)
-  DECL_SECTION(.init.data) {
+#ifdef EFI
+  /* align to satisfy UEFI CA memory mitigation */
+  . = ALIGN(SECTION_ALIGN);

Nit (also for the earlier patch iirc): Comment style.

Jan