Skip to contentSkip to conversation

x86/linker: add a reloc section to ELF linker script

PATCH In review Roger Pau Monne 3 participants 15 messages Jun 27, 2019

lore
Share & tools
lore
#0 lore raw

PATCH v3 1/3 x86/linker: add a reloc section to ELF linker script

if the hypervisor has been built with EFI support (ie: multiboot2). This allows to position the .reloc section correctly in the output binary.

Note that for the ELF output format the .reloc section is moved before .bss because the data it contains is read-only, so it belongs with the other sections containing read-only data.

---
Cc: Jan Beulich <[email protected]>
Cc: Andrew Cooper <[email protected]>
Cc: Wei Liu <[email protected]>
---
Changes since v2:
- Fix subject to mention the section is added to the linker script.
- Fix commit message to note .reloc is added together with the rest of the read-only sections.

Changes since v1:
- Move the .reloc section position in the output binary only for the ELF output format.

---
 xen/arch/x86/xen.lds.S | 8 ++++++++
 1 file changed, 8 insertions(+)
xen/arch/x86/xen.lds.S +8 -0
index 98a99444c2..19aa4332cf 100644
@@ -175,6 +175,14 @@ SECTIONS
   } :text
 #endif
 #endif
+
+#if defined(XEN_BUILD_EFI) && !defined(EFI)
+  . = ALIGN(4);
+  DECL_SECTION(.reloc) {
+    *(.reloc)
+  } :text
+#endif
+
   _erodata = .;
 
   . = ALIGN(SECTION_ALIGN);
Signature · 7 lines
2.20.1 (Apple Git-117)


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
#1 lore raw

PATCH v3 2/3 xen/link: handle .init.rodata.cst* sections in the linker script

Note that those sections when not prefixed with .init are already handled by the more general .rodata.* matching pattern in the .rodata output section.

---
Cc: Stefano Stabellini <[email protected]>
Cc: Julien Grall <[email protected]>
Cc: Jan Beulich <[email protected]>
Cc: Andrew Cooper <[email protected]>
Cc: Wei Liu <[email protected]>

---
 xen/arch/arm/xen.lds.S | 1 +
 xen/arch/x86/xen.lds.S | 1 +
 2 files changed, 2 insertions(+)
xen/arch/arm/xen.lds.S +1 -0
index e664c4441a..b636d9f223 100644
@@ -156,6 +156,7 @@ SECTIONS
        *(.init.rodata)
        *(.init.rodata.rel)
        *(.init.rodata.str*)
+       *(.init.rodata.cst*)
 
        . = ALIGN(POINTER_ALIGN);
        __setup_start = .;
xen/arch/x86/xen.lds.S +1 -0
index 19aa4332cf..d0c7fbc37b 100644
@@ -216,6 +216,7 @@ SECTIONS
        *(.init.rodata)
        *(.init.rodata.rel)
        *(.init.rodata.str*)
+       *(.init.rodata.cst*)
 
        . = ALIGN(POINTER_ALIGN);
        __setup_start = .;
Signature · 7 lines
2.20.1 (Apple Git-117)


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
#2 lore raw

Re: [Xen-devel] [PATCH v3 2/3] xen/link: handle .init.rodata.cst* sections in the linker script

Quoted 5 lines · Roger Pau Monne

Note that those sections when not prefixed with .init are already handled by the more general .rodata.* matching pattern in the .rodata output section.

So, in hindsight, we'll never get .cst in .data, because of the "constant" in the name, which rules out a lot of my first attemt.

As this does resolve the problem, Acked-by: Andrew Cooper <[email protected]>

However ...

Quoted 16 lines

---
Cc: Stefano Stabellini <[email protected]>
Cc: Julien Grall <[email protected]>
Cc: Jan Beulich <[email protected]>
Cc: Andrew Cooper <[email protected]>
Cc: Wei Liu <[email protected]>

---
 xen/arch/arm/xen.lds.S | 1 +
 xen/arch/x86/xen.lds.S | 1 +
 2 files changed, 2 insertions(+)
xen/arch/arm/xen.lds.S +0 -0
index e664c4441a..b636d9f223 100644

xen/arch/arm/xen.lds.S SECTIONS · line 159

Quoted 4 lines
xen/arch/arm/xen.lds.S +1 -0
@@ -156,6 +156,7 @@ SECTIONS
        *(.init.rodata)
        *(.init.rodata.rel)
        *(.init.rodata.str*)
+       *(.init.rodata.cst*)

... .init is just a grouping prefix, so I'd recommend that we treat .init.rodata in exactly the same way as we treat .rodata, so I'd suggest turning this into

*(.init.rodata)
*(.init.rodata.*)

to match the regular .rodata.

This can easily be adjusted on commit, to save yet-another-round.

~Andrew

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#3 lore raw

Re: [Xen-devel] [PATCH v3 2/3] xen/link: handle .init.rodata.cst* sections in the linker script

Quoted 14 lines
Quoted 1 line
Quoted 1 line · On 27.06.19 at 12:53, <[email protected]> wrote:

xen/arch/arm/xen.lds.S SECTIONS · line 159

Quoted 7 lines · Roger Pau Monne
xen/arch/arm/xen.lds.S +1 -0
@@ -156,6 +156,7 @@ SECTIONS
        *(.init.rodata)
        *(.init.rodata.rel)
        *(.init.rodata.str*)
+       *(.init.rodata.cst*)

... .init is just a grouping prefix, so I'd recommend that we treat .init.rodata in exactly the same way as we treat .rodata, so I'd suggest turning this into

Quoted 4 lines

*(.init.rodata)
*(.init.rodata.*)

to match the regular .rodata.

Or, as suggested elsewhere, make .rodata use less wide matching, like we do for .init.rodata.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#4 lore raw

Re: [Xen-devel] [PATCH v3 2/3] xen/link: handle .init.rodata.cst* sections in the linker script

Quoted 14 lines · Jan Beulich
Quoted 14 lines
Quoted 1 line
Quoted 1 line · On 27.06.19 at 12:53, <[email protected]> wrote:

xen/arch/arm/xen.lds.S SECTIONS · line 159

Quoted 7 lines · Roger Pau Monne
xen/arch/arm/xen.lds.S +1 -0
@@ -156,6 +156,7 @@ SECTIONS
        *(.init.rodata)
        *(.init.rodata.rel)
        *(.init.rodata.str*)
+       *(.init.rodata.cst*)

... .init is just a grouping prefix, so I'd recommend that we treat .init.rodata in exactly the same way as we treat .rodata, so I'd suggest turning this into

Quoted 7 lines
Quoted 4 lines

*(.init.rodata)
*(.init.rodata.*)

to match the regular .rodata.

Or, as suggested elsewhere, make .rodata use less wide matching, like we do for .init.rodata.

I'm happy to handle .init.rodata subsections as Xen currently handles .rodata subsections. There are no custom subsections explicitly added to either .rodata or .init.rodata, and if we start adding such subsections explicitly the linker script will likely need modifications anyway to mark the start and end of those subsections in the final binary, or the default placement of the wider wildcard will be already fine.

Thanks, Roger.

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#5 lore raw

PATCH v3 3/3 x86: check for multiboot{1, 2} header presence

In reply to Roger Pau Monne

After building the hypervisor binary. Note that the check is performed by searching for the magic header value at the start of the binary.

---
Cc: Jan Beulich <[email protected]>
Cc: Andrew Cooper <[email protected]>
Cc: Wei Liu <[email protected]>
---
Changes since v2:
- Use a variable to store the intermediate file name.
- Remove the intermediate file in the clean target.
- Add intermediate file to gitignore.

Changes since v1:
- Use an intermediate file to perform the header checks.

---
 .gitignore            | 1 +
 xen/arch/x86/Makefile | 9 +++++++--
 2 files changed, 8 insertions(+), 2 deletions(-)
.gitignore +1 -0
index a77cbff02c..56bcb64837 100644
@@ -278,6 +278,7 @@ tools/xentrace/xentrace
 xen/.banner
 xen/.config
 xen/.config.old
+xen/.xen
 xen/System.map
 xen/arch/x86/asm-macros.i
 xen/arch/x86/boot/mkelf32
xen/arch/x86/Makefile +7 -2
index 8a8d8f060f..94e6c9aee3 100644
@@ -99,9 +99,14 @@ endif
 syms-warn-dup-y := --warn-dup
 syms-warn-dup-$(CONFIG_SUPPRESS_DUPLICATE_SYMBOL_WARNINGS) :=
 
+$(TARGET): TMP = $(@D)/.$(@F)
 $(TARGET): $(TARGET)-syms $(efi-y) boot/mkelf32
-	./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TARGET) $(XEN_IMG_OFFSET) \
+	./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TMP) $(XEN_IMG_OFFSET) \
 	               `$(NM) $(TARGET)-syms | sed -ne 's/^\([^ ]*\) . __2M_rwdata_end$$/0x\1/p'`
+	# Check for multiboot{1,2} headers
+	od -t x4 -N 8192 $(TMP) | grep 1badb002 > /dev/null
+	od -t x4 -N 32768 $(TMP) | grep e85250d6 > /dev/null
+	mv $(TMP) $(TARGET)
 
 ALL_OBJS := $(BASEDIR)/arch/x86/boot/built_in.o $(BASEDIR)/arch/x86/efi/built_in.o $(ALL_OBJS)
 
@@ -249,7 +254,7 @@ efi/mkreloc: efi/mkreloc.c
 clean::
 	rm -f asm-offsets.s *.lds boot/*.o boot/*~ boot/core boot/mkelf32
 	rm -f asm-macros.i $(BASEDIR)/include/asm-x86/asm-macros.*
-	rm -f $(BASEDIR)/.xen-syms.[0-9]* boot/.*.d
+	rm -f $(BASEDIR)/.xen-syms.[0-9]* boot/.*.d $(BASEDIR)/.xen
 	rm -f $(BASEDIR)/.xen.efi.[0-9]* efi/*.efi efi/mkreloc
 	rm -f boot/cmdline.S boot/reloc.S boot/*.lnk boot/*.bin
 	rm -f note.o
Signature · 7 lines
2.20.1 (Apple Git-117)


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Acked-by on v3 3/3
#6 lore raw

Re: [Xen-devel] [PATCH v3 3/3] x86: check for multiboot{1, 2} header presence

Quoted 5 lines
Quoted 1 line
Quoted 1 line · On 27.06.19 at 11:33, <[email protected]> wrote:

After building the hypervisor binary. Note that the check is performed by searching for the magic header value at the start of the binary.

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#7 lore raw

Re: [Xen-devel] [PATCH v3 3/3] x86: check for multiboot{1, 2} header presence

In reply to Roger Pau Monne

xen/arch/x86/Makefile endif · line 102

Quoted 9 lines · Roger Pau Monne
xen/arch/x86/Makefile +1 -0
index 8a8d8f060f..94e6c9aee3 100644
@@ -99,9 +99,14 @@ endif
 syms-warn-dup-y := --warn-dup
 syms-warn-dup-$(CONFIG_SUPPRESS_DUPLICATE_SYMBOL_WARNINGS) :=
 
+$(TARGET): TMP = $(@D)/.$(@F)

I'd suggest giving this a .elf32 suffix to make it clear which pass of the build it comes from.

xen/arch/x86/Makefile endif · line 108

Quoted 7 lines
xen/arch/x86/Makefile +4 -1
 $(TARGET): $(TARGET)-syms $(efi-y) boot/mkelf32
-	./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TARGET) $(XEN_IMG_OFFSET) \
+	./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TMP) $(XEN_IMG_OFFSET) \
 	               `$(NM) $(TARGET)-syms | sed -ne 's/^\([^ ]*\) . __2M_rwdata_end$$/0x\1/p'`
+	# Check for multiboot{1,2} headers
+	od -t x4 -N 8192 $(TMP) | grep 1badb002 > /dev/null ||
+	od -t x4 -N 32768 $(TMP) | grep e85250d6 > /dev/null

This works, but

Makefile:104: recipe for target '/local/xen.git/xen/xen' failed

Isn't helpful to identify what went wrong.  Sadly, we can't use $(error ...) in a shell snippet, but:

andrewcoop@andrewcoop:/local/xen.git/xen$ git d
xen/arch/x86/Makefile +5 -4
index 94e6c9aee3..a1d6605a8b 100644
@@ -99,13 +99,14 @@ endif
 syms-warn-dup-y := --warn-dup
 syms-warn-dup-$(CONFIG_SUPPRESS_DUPLICATE_SYMBOL_WARNINGS) :=
 
-$(TARGET): TMP = $(@D)/.$(@F)
+$(TARGET): TMP = $(@D)/.$(@F).elf32
 $(TARGET): $(TARGET)-syms $(efi-y) boot/mkelf32
        ./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TMP) $(XEN_IMG_OFFSET) \
                       `$(NM) $(TARGET)-syms | sed -ne 's/^\([^ ]*\) . __2M_rwdata_end$$/0x\1/p'`
-       # Check for multiboot{1,2} headers
-       od -t x4 -N 8192 $(TMP) | grep 1badb002 > /dev/null
-       od -t x4 -N 32768 $(TMP) | grep e85250d6 > /dev/null
+       od -t x4 -N 8192 $(TMP)  | grep 1badb002 > /dev/null || \
+               { echo "No Multiboot1 header found"; false; }
+       od -t x4 -N 32768 $(TMP) | grep e85250d6 > /dev/null || \
+               { echo "No Multiboot2 header found"; false; }
        mv $(TMP) $(TARGET)
 
 ALL_OBJS := $(BASEDIR)/arch/x86/boot/built_in.o $(BASEDIR)/arch/x86/efi/built_in.o $(ALL_OBJS)

results in:

No Multiboot1 header found
Makefile:104: recipe for target '/local/xen.git/xen/xen' failed make[2]: *** [/local/xen.git/xen/xen] Error 1 Makefile:136: recipe for target '/local/xen.git/xen/xen' failed make[1]: *** [/local/xen.git/xen/xen] Error 2 Makefile:45: recipe for target 'build' failed
make: *** [build] Error 2

Which is far more clear.

Thoughts?

~Andrew

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#8 lore raw

Re: [Xen-devel] [PATCH v3 3/3] x86: check for multiboot{1, 2} header presence

Quoted 25 lines
Quoted 1 line
Quoted 1 line · On 27.06.19 at 13:51, <[email protected]> wrote:

xen/arch/x86/Makefile endif · line 102

Quoted 9 lines · Roger Pau Monne
xen/arch/x86/Makefile +1 -0
index 8a8d8f060f..94e6c9aee3 100644
@@ -99,9 +99,14 @@ endif
 syms-warn-dup-y := --warn-dup
 syms-warn-dup-$(CONFIG_SUPPRESS_DUPLICATE_SYMBOL_WARNINGS) :=
 
+$(TARGET): TMP = $(@D)/.$(@F)

I'd suggest giving this a .elf32 suffix to make it clear which pass of the build it comes from.

xen/arch/x86/Makefile endif · line 108

Quoted 7 lines
xen/arch/x86/Makefile +4 -1
 $(TARGET): $(TARGET)-syms $(efi-y) boot/mkelf32
-	./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TARGET) $(XEN_IMG_OFFSET) \
+	./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TMP) $(XEN_IMG_OFFSET) \
 	               `$(NM) $(TARGET)-syms | sed -ne 's/^\([^ ]*\) . __2M_rwdata_end$$/0x\1/p'`
+	# Check for multiboot{1,2} headers
+	od -t x4 -N 8192 $(TMP) | grep 1badb002 > /dev/null ||
+	od -t x4 -N 32768 $(TMP) | grep e85250d6 > /dev/null

This works, but

Quoted 4 lines

Makefile:104: recipe for target '/local/xen.git/xen/xen' failed

Isn't helpful to identify what went wrong. Sadly, we can't use $(error ...) in a shell snippet,

I think we could:

       $(if $(shell od -t x4 -N 8192 $(TMP) | grep 1badb002),,$(error ...)exit 1)

But I admit I didn't check whether it is well defined that only one of the last two operands of $(if ) get actually evaluated.

Quoted 38 lines

but:

andrewcoop@andrewcoop:/local/xen.git/xen$ git d
xen/arch/x86/Makefile +5 -4
index 94e6c9aee3..a1d6605a8b 100644
@@ -99,13 +99,14 @@ endif
 syms-warn-dup-y := --warn-dup
 syms-warn-dup-$(CONFIG_SUPPRESS_DUPLICATE_SYMBOL_WARNINGS) :=
 
-$(TARGET): TMP = $(@D)/.$(@F)
+$(TARGET): TMP = $(@D)/.$(@F).elf32
 $(TARGET): $(TARGET)-syms $(efi-y) boot/mkelf32
        ./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TMP) $(XEN_IMG_OFFSET) \
                       `$(NM) $(TARGET)-syms | sed -ne 's/^\([^ ]*\) . __2M_rwdata_end$$/0x\1/p'`
-       # Check for multiboot{1,2} headers
-       od -t x4 -N 8192 $(TMP) | grep 1badb002 > /dev/null
-       od -t x4 -N 32768 $(TMP) | grep e85250d6 > /dev/null
+       od -t x4 -N 8192 $(TMP)  | grep 1badb002 > /dev/null || \
+               { echo "No Multiboot1 header found"; false; }
+       od -t x4 -N 32768 $(TMP) | grep e85250d6 > /dev/null || \
+               { echo "No Multiboot2 header found"; false; }
        mv $(TMP) $(TARGET)
 
 ALL_OBJS := $(BASEDIR)/arch/x86/boot/built_in.o 

$(BASEDIR)/arch/x86/efi/built_in.o $(ALL_OBJS)

results in:

No Multiboot1 header found
Makefile:104: recipe for target '/local/xen.git/xen/xen' failed make[2]: *** [/local/xen.git/xen/xen] Error 1 Makefile:136: recipe for target '/local/xen.git/xen/xen' failed make[1]: *** [/local/xen.git/xen/xen] Error 2 Makefile:45: recipe for target 'build' failed
make: *** [build] Error 2

Quoted 3 lines

Which is far more clear.

Thoughts?

Good idea. The only further request I have is to add >&2 to the echo commands (unless we go the $(error ) route anyway).

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#9 lore raw

Re: [Xen-devel] [PATCH v3 3/3] x86: check for multiboot{1, 2} header presence

Quoted 25 lines · Jan Beulich
Quoted 25 lines
Quoted 1 line
Quoted 1 line · On 27.06.19 at 13:51, <[email protected]> wrote:

xen/arch/x86/Makefile endif · line 102

Quoted 9 lines · Roger Pau Monne
xen/arch/x86/Makefile +1 -0
index 8a8d8f060f..94e6c9aee3 100644
@@ -99,9 +99,14 @@ endif
 syms-warn-dup-y := --warn-dup
 syms-warn-dup-$(CONFIG_SUPPRESS_DUPLICATE_SYMBOL_WARNINGS) :=
 
+$(TARGET): TMP = $(@D)/.$(@F)

I'd suggest giving this a .elf32 suffix to make it clear which pass of the build it comes from.

xen/arch/x86/Makefile endif · line 108

Quoted 7 lines
xen/arch/x86/Makefile +4 -1
 $(TARGET): $(TARGET)-syms $(efi-y) boot/mkelf32
-	./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TARGET) $(XEN_IMG_OFFSET) \
+	./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TMP) $(XEN_IMG_OFFSET) \
 	               `$(NM) $(TARGET)-syms | sed -ne 's/^\([^ ]*\) . __2M_rwdata_end$$/0x\1/p'`
+	# Check for multiboot{1,2} headers
+	od -t x4 -N 8192 $(TMP) | grep 1badb002 > /dev/null ||
+	od -t x4 -N 32768 $(TMP) | grep e85250d6 > /dev/null

This works, but

Makefile:104: recipe for target '/local/xen.git/xen/xen' failed

Quoted 8 lines
Quoted 2 lines

Isn't helpful to identify what went wrong. Sadly, we can't use $(error ...) in a shell snippet,

I think we could:

       $(if $(shell od -t x4 -N 8192 $(TMP) | grep 1badb002),,$(error ...)exit 1)

But I admit I didn't check whether it is well defined that only one of the last two operands of $(if ) get actually evaluated.

Of the two options, I think the || { ; false } is clearer to follow.

Quoted 39 lines
Quoted 39 lines

but:

andrewcoop@andrewcoop:/local/xen.git/xen$ git d
xen/arch/x86/Makefile +5 -4
index 94e6c9aee3..a1d6605a8b 100644
@@ -99,13 +99,14 @@ endif
 syms-warn-dup-y := --warn-dup
 syms-warn-dup-$(CONFIG_SUPPRESS_DUPLICATE_SYMBOL_WARNINGS) :=
 
-$(TARGET): TMP = $(@D)/.$(@F)
+$(TARGET): TMP = $(@D)/.$(@F).elf32
 $(TARGET): $(TARGET)-syms $(efi-y) boot/mkelf32
        ./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TMP) $(XEN_IMG_OFFSET) \
                       `$(NM) $(TARGET)-syms | sed -ne 's/^\([^ ]*\) . __2M_rwdata_end$$/0x\1/p'`
-       # Check for multiboot{1,2} headers
-       od -t x4 -N 8192 $(TMP) | grep 1badb002 > /dev/null
-       od -t x4 -N 32768 $(TMP) | grep e85250d6 > /dev/null
+       od -t x4 -N 8192 $(TMP)  | grep 1badb002 > /dev/null || \
+               { echo "No Multiboot1 header found"; false; }
+       od -t x4 -N 32768 $(TMP) | grep e85250d6 > /dev/null || \
+               { echo "No Multiboot2 header found"; false; }
        mv $(TMP) $(TARGET)
 
 ALL_OBJS := $(BASEDIR)/arch/x86/boot/built_in.o 

$(BASEDIR)/arch/x86/efi/built_in.o $(ALL_OBJS)

results in:

No Multiboot1 header found
Makefile:104: recipe for target '/local/xen.git/xen/xen' failed make[2]: *** [/local/xen.git/xen/xen] Error 1 Makefile:136: recipe for target '/local/xen.git/xen/xen' failed make[1]: *** [/local/xen.git/xen/xen] Error 2 Makefile:45: recipe for target 'build' failed
make: *** [build] Error 2

Quoted 5 lines
Quoted 3 lines

Which is far more clear.

Thoughts?

Good idea. The only further request I have is to add >&2 to the echo commands.

Done.

~Andrew

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#10 lore raw

Re: [Xen-devel] [PATCH v3 3/3] x86: check for multiboot{1, 2} header presence

In reply to Andrew Cooper
Quoted 9 lines · Andrew Cooper
Quoted 8 lines · Roger Pau Monne
xen/arch/x86/Makefile +0 -0
index 8a8d8f060f..94e6c9aee3 100644
@@ -99,9 +99,14 @@ endif
 syms-warn-dup-y := --warn-dup
 syms-warn-dup-$(CONFIG_SUPPRESS_DUPLICATE_SYMBOL_WARNINGS) :=
 
Quoted 4 lines

xen/arch/x86/Makefile endif · line 102

Quoted 1 line
xen/arch/x86/Makefile +1 -0
+$(TARGET): TMP = $(@D)/.$(@F)

I'd suggest giving this a .elf32 suffix to make it clear which pass of the build it comes from.

That's fine, please also adjust the ignored list and the clean target.

Quoted 50 lines

xen/arch/x86/Makefile endif · line 108

Quoted 7 lines
xen/arch/x86/Makefile +4 -1
 $(TARGET): $(TARGET)-syms $(efi-y) boot/mkelf32
-	./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TARGET) $(XEN_IMG_OFFSET) \
+	./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TMP) $(XEN_IMG_OFFSET) \
 	               `$(NM) $(TARGET)-syms | sed -ne 's/^\([^ ]*\) . __2M_rwdata_end$$/0x\1/p'`
+	# Check for multiboot{1,2} headers
+	od -t x4 -N 8192 $(TMP) | grep 1badb002 > /dev/null ||
+	od -t x4 -N 32768 $(TMP) | grep e85250d6 > /dev/null

This works, but

Makefile:104: recipe for target '/local/xen.git/xen/xen' failed

Isn't helpful to identify what went wrong.  Sadly, we can't use $(error ...) in a shell snippet, but:

andrewcoop@andrewcoop:/local/xen.git/xen$ git d
xen/arch/x86/Makefile +5 -4
index 94e6c9aee3..a1d6605a8b 100644
@@ -99,13 +99,14 @@ endif
 syms-warn-dup-y := --warn-dup
 syms-warn-dup-$(CONFIG_SUPPRESS_DUPLICATE_SYMBOL_WARNINGS) :=
 
-$(TARGET): TMP = $(@D)/.$(@F)
+$(TARGET): TMP = $(@D)/.$(@F).elf32
 $(TARGET): $(TARGET)-syms $(efi-y) boot/mkelf32
        ./boot/mkelf32 $(notes_phdrs) $(TARGET)-syms $(TMP) $(XEN_IMG_OFFSET) \
                       `$(NM) $(TARGET)-syms | sed -ne 's/^\([^ ]*\) . __2M_rwdata_end$$/0x\1/p'`
-       # Check for multiboot{1,2} headers
-       od -t x4 -N 8192 $(TMP) | grep 1badb002 > /dev/null
-       od -t x4 -N 32768 $(TMP) | grep e85250d6 > /dev/null
+       od -t x4 -N 8192 $(TMP)  | grep 1badb002 > /dev/null || \
+               { echo "No Multiboot1 header found"; false; }
+       od -t x4 -N 32768 $(TMP) | grep e85250d6 > /dev/null || \
+               { echo "No Multiboot2 header found"; false; }
        mv $(TMP) $(TARGET)
 
 ALL_OBJS := $(BASEDIR)/arch/x86/boot/built_in.o $(BASEDIR)/arch/x86/efi/built_in.o $(ALL_OBJS)

results in:

No Multiboot1 header found
Makefile:104: recipe for target '/local/xen.git/xen/xen' failed make[2]: *** [/local/xen.git/xen/xen] Error 1 Makefile:136: recipe for target '/local/xen.git/xen/xen' failed make[1]: *** [/local/xen.git/xen/xen] Error 2 Makefile:45: recipe for target 'build' failed
make: *** [build] Error 2

Quoted 3 lines

Which is far more clear.

Thoughts?

Thanks, is indeed better. I also agree with Jan on the redirection to stderr.

Roger.

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#11 lore raw
In reply to Roger Pau Monne
Quoted 9 lines · Roger Pau Monne

if the hypervisor has been built with EFI support (ie: multiboot2). This allows to position the .reloc section correctly in the output binary.

Note that for the ELF output format the .reloc section is moved before .bss because the data it contains is read-only, so it belongs with the other sections containing read-only data.

I have to admit that I'm still confused as to why we need this in the first place.

The ELF build is linked to a fixed virtual address, irrespective of whether grub loads it via MB1 or MB2 and/or with EFI details.

i.e. the non-EFI build shouldn't have any remaining relocations by the time it is fully linked.

Or am I missing something?

~Andrew

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#12 lore raw
Quoted 14 lines · Andrew Cooper
Quoted 9 lines · Roger Pau Monne

if the hypervisor has been built with EFI support (ie: multiboot2). This allows to position the .reloc section correctly in the output binary.

Note that for the ELF output format the .reloc section is moved before .bss because the data it contains is read-only, so it belongs with the other sections containing read-only data.

I have to admit that I'm still confused as to why we need this in the first place.

Quoted 7 lines

The ELF build is linked to a fixed virtual address, irrespective of whether grub loads it via MB1 or MB2 and/or with EFI details.

i.e. the non-EFI build shouldn't have any remaining relocations by the time it is fully linked.

Or am I missing something?

Right, but there's code that depends on the symbols defined in .reloc (__base_relocs_start/__base_relocs_end), so unless those symbols are defined the linker will throw a missing symbols error on the final link step.

I could add .reloc to the discarded sections list and create the __base_relocs_start and __base_relocs_end symbols on the linker script maybe, but I'm not sure that's any better than just having the dummy .reloc section.

Or another option would be to compile the units that use those symbols twice, one for the ELF build and one for the PE build, but again that doesn't seem much better IMO.

Thanks, Roger.

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#13 lore raw
Quoted 17 lines · Roger Pau Monné
Quoted 16 lines · Andrew Cooper
Quoted 9 lines · Roger Pau Monne

if the hypervisor has been built with EFI support (ie: multiboot2). This allows to position the .reloc section correctly in the output binary.

Note that for the ELF output format the .reloc section is moved before .bss because the data it contains is read-only, so it belongs with the other sections containing read-only data.

I have to admit that I'm still confused as to why we need this in the first place.

The ELF build is linked to a fixed virtual address, irrespective of whether grub loads it via MB1 or MB2 and/or with EFI details.

Quoted 8 lines
Quoted 4 lines

i.e. the non-EFI build shouldn't have any remaining relocations by the time it is fully linked.

Or am I missing something?

Right, but there's code that depends on the symbols defined in .reloc (__base_relocs_start/__base_relocs_end), so unless those symbols are defined the linker will throw a missing symbols error on the final link step.

Ok.  I can certainly accept that this is how the code currently functions.

Quoted 8 lines

I could add .reloc to the discarded sections list and create the __base_relocs_start and __base_relocs_end symbols on the linker script maybe, but I'm not sure that's any better than just having the dummy .reloc section.

Or another option would be to compile the units that use those symbols twice, one for the ELF build and one for the PE build, but again that doesn't seem much better IMO.

So the bug here is that efi_arch_relocate_image() isn't inside an #ifdef EFI (or is it XEN_BUILD_EFI?)

And the reason #ifdef-ing it won't work is because we have a single pass of extra objects to link into the main Xen to add EFI support.

I think the proper longterm fix is to have CONFIG_EFI (suitably guarded on toolchain support), seeing as it is common across our two binaries, and the extra bits for the real EFI build then become rather smaller.

However, it is definitely not fair to lump this fix on you for this series, so given a comment explaining that this isn't used in the ELF build, but needs to be present for compatibility with the EFI build, I'll be ok with the patch in this form.

Again, a comment can be fixed up on commit if everyone is happy with this approach.

~Andrew

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

Acked-by on v3 1/3
#14 lore raw
In reply to Roger Pau Monne
Quoted 10 lines
Quoted 1 line
Quoted 1 line · On 27.06.19 at 11:33, <[email protected]> wrote:

if the hypervisor has been built with EFI support (ie: multiboot2). This allows to position the .reloc section correctly in the output binary.

Note that for the ELF output format the .reloc section is moved before .bss because the data it contains is read-only, so it belongs with the other sections containing read-only data.

Assuming you've addressed Andrew's concerns in your reply,

(But Andrew, please confirm.)

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel