Skip to contentSkip to conversation

ppc: add "-z notext" flag to disable diagnostic

PATCH Superseded Bill Wendling 7 participants 13 messages Aug 12, 2021 – Aug 27, 2021 PowerPC

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ppc: add "-z notext" flag to disable diagnostic

The "-z notext" flag disables reporting an error if DT_TEXTREL is set on PPC with CONFIG=kdump:

ld.lld: error: can't create dynamic relocation R_PPC64_ADDR64 against
local symbol in readonly segment; recompile object files with -fPIC or pass '-Wl,-z,notext' to allow text relocations in the output

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

defined in built-in.a(arch/powerpc/kernel/misc.o) referenced by arch/powerpc/kernel/misc.o:(.text+0x20) in archive

built-in.a

The BFD linker disables this by default (though it's configurable in current versions). LLD enables this by default. So we add the flag to keep LLD from emitting the error.

---
The output of the "get_maintainer.pl" run just in case no one believes me. :-)

$ ./scripts/get_maintainer.pl arch/powerpc/Makefile
Michael Ellerman <[email protected]> (supporter:LINUX FOR POWERPC (32-BIT AND 64-BIT))
Benjamin Herrenschmidt <[email protected]> (reviewer:LINUX FOR POWERPC (32-BIT AND 64-BIT))
Paul Mackerras <[email protected]> (reviewer:LINUX FOR POWERPC (32-BIT AND 64-BIT))
Nathan Chancellor <[email protected]> (supporter:CLANG/LLVM BUILD SUPPORT)
Nick Desaulniers <[email protected]> (supporter:CLANG/LLVM BUILD SUPPORT)
[email protected] (open list:LINUX FOR POWERPC (32-BIT AND 64-BIT))
[email protected] (open list)
[email protected] (open list:CLANG/LLVM BUILD SUPPORT)
---
 arch/powerpc/Makefile | 1 +
 1 file changed, 1 insertion(+)
arch/powerpc/Makefile +1 -0
index 6505d66f1193..17a9fbf9b789 100644
@@ -122,6 +122,7 @@ endif
 
 LDFLAGS_vmlinux-y := -Bstatic
 LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) := -pie
+LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) += -z notext
 LDFLAGS_vmlinux	:= $(LDFLAGS_vmlinux-y)
 
 ifdef CONFIG_PPC64
Signature · 1 line
2.33.0.rc1.237.g0d66db33f3-goog
Reviewed-by on v1 1/1
#1 lore raw
Quoted 9 lines · Bill Wendling

The "-z notext" flag disables reporting an error if DT_TEXTREL is set on PPC with CONFIG=kdump:

ld.lld: error: can't create dynamic relocation R_PPC64_ADDR64 against
local symbol in readonly segment; recompile object files with -fPIC or pass '-Wl,-z,notext' to allow text relocations in the output

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

defined in built-in.a(arch/powerpc/kernel/misc.o) referenced by arch/powerpc/kernel/misc.o:(.text+0x20) in archive

Quoted 7 lines

built-in.a

The BFD linker disables this by default (though it's configurable in current versions). LLD enables this by default. So we add the flag to keep LLD from emitting the error.

Quoted 3 lines

---
The output of the "get_maintainer.pl" run just in case no one believes me. :-)

LOL!

Quoted 29 lines
$ ./scripts/get_maintainer.pl arch/powerpc/Makefile
Michael Ellerman <[email protected]> (supporter:LINUX FOR POWERPC (32-BIT AND 64-BIT))
Benjamin Herrenschmidt <[email protected]> (reviewer:LINUX FOR POWERPC (32-BIT AND 64-BIT))
Paul Mackerras <[email protected]> (reviewer:LINUX FOR POWERPC (32-BIT AND 64-BIT))
Nathan Chancellor <[email protected]> (supporter:CLANG/LLVM BUILD SUPPORT)
Nick Desaulniers <[email protected]> (supporter:CLANG/LLVM BUILD SUPPORT)
[email protected] (open list:LINUX FOR POWERPC (32-BIT AND 64-BIT))
[email protected] (open list)
[email protected] (open list:CLANG/LLVM BUILD SUPPORT)
---
 arch/powerpc/Makefile | 1 +
 1 file changed, 1 insertion(+)
arch/powerpc/Makefile +1 -1
index 6505d66f1193..17a9fbf9b789 100644
@@ -122,6 +122,7 @@ endif
 
 LDFLAGS_vmlinux-y := -Bstatic
 LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) := -pie
+LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) += -z notext
 LDFLAGS_vmlinux        := $(LDFLAGS_vmlinux-y)
 
 ifdef CONFIG_PPC64
--

2.33.0.rc1.237.g0d66db33f3-goog

Signature · 2 lines
Thanks,
~Nick Desaulniers
#2 lore raw
In reply to Bill Wendling
Quoted 6 lines · Bill Wendling

The "-z notext" flag disables reporting an error if DT_TEXTREL is set on PPC with CONFIG=kdump:

ld.lld: error: can't create dynamic relocation R_PPC64_ADDR64 against
local symbol in readonly segment; recompile object files with -fPIC or pass '-Wl,-z,notext' to allow text relocations in the output

Quoted 7 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

defined in built-in.a(arch/powerpc/kernel/misc.o) referenced by arch/powerpc/kernel/misc.o:(.text+0x20) in archive

built-in.a

The BFD linker disables this by default (though it's configurable in current versions). LLD enables this by default. So we add the flag to keep LLD from emitting the error.

You didn't provide a huge amount of context but I was able to reproduce a similar set of errors with pseries_le_defconfig and

make ARCH=powerpc CROSS_COMPILE=powerpc64-linux-gnu- CC="ccache clang-11" LD=ld.lld-11 AR=llvm-ar-11 -j4 vmlinux

I also checked the manpage, and indeed the system ld does not issue this warning/error by default.

arch/powerpc/Makefile endif · line 125

Quoted 13 lines
---
 arch/powerpc/Makefile | 1 +
 1 file changed, 1 insertion(+)
arch/powerpc/Makefile +1 -0
index 6505d66f1193..17a9fbf9b789 100644
@@ -122,6 +122,7 @@ endif
 
 LDFLAGS_vmlinux-y := -Bstatic
 LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) := -pie
+LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) += -z notext

Is there any reason this should be gated on CONFIG_RELOCATABLE? (I tried without it and got different but possibly related linker errors...)

Also, is this behaviour new?

Kind regards,
Daniel

Quoted 5 lines
arch/powerpc/Makefile +0 -0
 LDFLAGS_vmlinux	:= $(LDFLAGS_vmlinux-y)
 
 ifdef CONFIG_PPC64
Signature · 1 line
2.33.0.rc1.237.g0d66db33f3-goog
#3 lore raw
Quoted 13 lines · Daniel Axtens
Quoted 10 lines · Bill Wendling

The "-z notext" flag disables reporting an error if DT_TEXTREL is set on PPC with CONFIG=kdump:

ld.lld: error: can't create dynamic relocation R_PPC64_ADDR64 against
local symbol in readonly segment; recompile object files with -fPIC or pass '-Wl,-z,notext' to allow text relocations in the output

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

defined in built-in.a(arch/powerpc/kernel/misc.o) referenced by arch/powerpc/kernel/misc.o:(.text+0x20) in archive

built-in.a

Quoted 7 lines
Quoted 3 lines

The BFD linker disables this by default (though it's configurable in current versions). LLD enables this by default. So we add the flag to keep LLD from emitting the error.

You didn't provide a huge amount of context but I was able to reproduce a similar set of errors with pseries_le_defconfig and

My apologies for the lack of context. We discovered this issue when moving to use LLD instead of BFD for our linker. This change suggested by Fangrui Song. Here's his comments from
https://github.com/ClangBuiltLinux/linux/issues/811. (Nick Desaulniers added a "Link" tag with his review.)

```
The text relocations are real. lld can link .tmp_vmlinux1 smoothly if
you pass -z notext. Though, it will still be insightful to investigate
where these text relocations come from. I believe there are only 2
categories.

(a) For a CONFIG_HAVE_ARCH_PREL32_RELOCATIONS=y build (x86 and arm64
select the option by default), ___ksymtab+* sections (non-SHF_WRITE)
contains entries relocated by PC relative relocations. These entries
do not need dynamic relocations.

out/powerpc64le/.config generated by ppc64le_defconfig does not enable
CONFIG_HAVE_ARCH_PREL32_RELOCATIONS=y.

% readelf -r out/x86_64/entry/entry_64.o
...
Relocation section '.rela___ksymtab+native_load_gs_index' at offset
0x6460 contains 2 entries:
    Offset             Info             Type               Symbol's
Value  Symbol's Name + Addend
0000000000000000  0000007a00000002 R_X86_64_PC32
0000000000000ea0 native_load_gs_index + 0
0000000000000004  0000001d00000002 R_X86_64_PC32
0000000000000000 __ksymtab_strings + 0

include/asm-generic/export.h

ld.lld: error: can't create dynamic relocation R_PPC64_ADDR64 against
local symbol in readonly segment; recompile object files with -fPIC or
pass '-Wl
,-z,notext' to allow text relocations in the output
>>> defined in init/built-in.a(do_mounts.o)
>>> referenced by do_mounts.c
>>>               do_mounts.o:(__ksymtab_name_to_dev_t) in archive init/built-in.a

(b) R_PPC64_ADDR64 in __mcount_loc sections.

ld.lld: error: can't create dynamic relocation R_PPC64_ADDR64 against
local symbol in readonly segment; recompile object files with -fPIC or
pass '-Wl
,-z,notext' to allow text relocations in the output
>>> defined in init/built-in.a(do_mounts_rd.o)
>>> referenced by do_mounts_rd.c
>>>               do_mounts_rd.o:(__mcount_loc+0x0) in archive init/built-in.a

This section is generated by ./scripts/recordmcount
"init/do_mounts_rd.o". The tool hard codes R_PPC64_ADDR64.

% grep PPC64 scripts/recordmcount.c
        case EM_PPC64:  reltype = R_PPC64_ADDR64; break;
```
Quoted 5 lines

make ARCH=powerpc CROSS_COMPILE=powerpc64-linux-gnu- CC="ccache clang-11" LD=ld.lld-11 AR=llvm-ar-11 -j4 vmlinux

I also checked the manpage, and indeed the system ld does not issue this warning/error by default.

It should be noted that the BFD linker can emit this warning/error if binutils is configured with "--enable-textrel-check={warning,error,yes}".

Quoted 10 lines
Quoted 10 lines
---
 arch/powerpc/Makefile | 1 +
 1 file changed, 1 insertion(+)
arch/powerpc/Makefile +0 -0
index 6505d66f1193..17a9fbf9b789 100644
@@ -122,6 +122,7 @@ endif
 
Quoted 7 lines

arch/powerpc/Makefile endif · line 125

Quoted 3 lines
arch/powerpc/Makefile +1 -0
 LDFLAGS_vmlinux-y := -Bstatic
 LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) := -pie
+LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) += -z notext

Is there any reason this should be gated on CONFIG_RELOCATABLE? (I tried without it and got different but possibly related linker errors...)

My understanding is that '-z notext' allows (i.e. doesn't emit warnings/errors for) relocations against read-only segments. So it's not really relevant to non-relocatable builds.

Unrelated question: Should the "-pie" flag be added with "+= -pie" (note the plus sign)?

Quoted 2 lines

Also, is this behaviour new?

I don't believe so. It's only when we wanted to use LLD that it was noticed.

BTW, this patch should more properly be attributed to Fangrui Song. I can send a follow-up patch that reflects this and adds more context to the commit message.

-bw

Quoted 8 lines

Kind regards,
Daniel

Quoted 5 lines
arch/powerpc/Makefile +0 -1
 LDFLAGS_vmlinux      := $(LDFLAGS_vmlinux-y)
 
 ifdef CONFIG_PPC64
--

2.33.0.rc1.237.g0d66db33f3-goog

#4 lore raw
Quoted 4 lines · Bill Wendling

BTW, this patch should more properly be attributed to Fangrui Song. I can send a follow-up patch that reflects this and adds more context to the commit message.

Sounds good to me. The TL;DR is that LLD has a different implicit default for -z text vs -z notext than BFD. We can emulate the behavior or BFD by simply being explicit about -z notext always.

Or we can dig through why there are relocations in read only sections, fix those, then enable -z text for all linkers. My recommendation would be get the thing building, then go digging time permitting.

Signature · 2 lines
Thanks,
~Nick Desaulniers
#5 lore raw
Quoted 3 lines · Nick Desaulniers

Or we can dig through why there are relocations in read only sections, fix those, then enable -z text for all linkers. My recommendation would be get the thing building, then go digging time permitting.

It is not always a bug. You can get much more efficient code if you have text relocations than if you don't. This "read-only" memory is perfectly writable until after relocation, a la relro.

But you no doubt will find some non-optimalities (or even straight out bugs) if you build with -ztext sometimes :-)

Segher

#6 lore raw
In reply to Bill Wendling
Quoted 2 lines · Bill Wendling
Quoted 1 line · Daniel Axtens

Bill Wendling <[email protected]> writes:

...

Quoted 9 lines
Quoted 9 lines
Quoted 9 lines
arch/powerpc/Makefile +1 -0
index 6505d66f1193..17a9fbf9b789 100644
@@ -122,6 +122,7 @@ endif
 
 LDFLAGS_vmlinux-y := -Bstatic
 LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) := -pie
+LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) += -z notext

...

Quoted 3 lines

Unrelated question: Should the "-pie" flag be added with "+= -pie" (note the plus sign)?

I noticed that too.

It's been like that since the original relocatable support was added in 2008, commit 549e8152de80 ("powerpc: Make the 64-bit kernel as a position-independent executable"), which did:

-LDFLAGS_vmlinux := -Bstatic
+LDFLAGS_vmlinux-yy := -Bstatic +LDFLAGS_vmlinux-$(CONFIG_PPC64)$(CONFIG_RELOCATABLE) := -pie
+LDFLAGS_vmlinux := $(LDFLAGS_vmlinux-yy)

There's no mention of those flags in the change log. But the way it's written suggests the intention was to not pass -Bstatic for relocatable builds, otherwise it could have been more simply:

+LDFLAGS_vmlinux-$(CONFIG_PPC64)$(CONFIG_RELOCATABLE) := -pie
+LDFLAGS_vmlinux := -Bstatic $(LDFLAGS_vmlinux-yy)

So I think it was deliberate to not use +=, but whether that's actually correct I can't say. Maybe in the past -Bstatic and -pie were incompatible?

cheers

#7 lore raw
In reply to Daniel Axtens
Quoted 34 lines · Daniel Axtens
Quoted 6 lines · Bill Wendling

The "-z notext" flag disables reporting an error if DT_TEXTREL is set on PPC with CONFIG=kdump:

ld.lld: error: can't create dynamic relocation R_PPC64_ADDR64 against
local symbol in readonly segment; recompile object files with -fPIC or pass '-Wl,-z,notext' to allow text relocations in the output

Quoted 7 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

defined in built-in.a(arch/powerpc/kernel/misc.o) referenced by arch/powerpc/kernel/misc.o:(.text+0x20) in archive

built-in.a

The BFD linker disables this by default (though it's configurable in current versions). LLD enables this by default. So we add the flag to keep LLD from emitting the error.

You didn't provide a huge amount of context but I was able to reproduce a similar set of errors with pseries_le_defconfig and

make ARCH=powerpc CROSS_COMPILE=powerpc64-linux-gnu- CC="ccache clang-11" LD=ld.lld-11 AR=llvm-ar-11 -j4 vmlinux

I also checked the manpage, and indeed the system ld does not issue this warning/error by default.

Quoted 10 lines
---
 arch/powerpc/Makefile | 1 +
 1 file changed, 1 insertion(+)
arch/powerpc/Makefile +0 -0
index 6505d66f1193..17a9fbf9b789 100644
@@ -122,6 +122,7 @@ endif
 
Quoted 8 lines

arch/powerpc/Makefile endif · line 125

Quoted 3 lines
arch/powerpc/Makefile +1 -0
 LDFLAGS_vmlinux-y := -Bstatic
 LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) := -pie
+LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) += -z notext

Is there any reason this should be gated on CONFIG_RELOCATABLE? (I tried without it and got different but possibly related linker errors...)

Also, is this behaviour new?

This is a longstanding behavior.

https://maskray.me/blog/2020-12-19-lld-and-gnu-linker-incompatibilities
See "Text relocations"

.o files used to link .tmp_vmlinux.kallsyms1 have many R_PPC64_ADDR64 relocations in non-SHF_WRITE sections. There are many text relocations (e.g. in .rela___ksymtab_gpl+* and .rela__mcount_loc sections) in a -pie link and are disallowed by LLD:

ld.lld: error: can't create dynamic relocation R_PPC64_ADDR64 against local symbol in readonly segment; recompile object files with -fPIC or pass '-Wl,-z,notext' to allow text relocations in the output

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

defined in arch/powerpc/kernel/head_64.o
referenced by arch/powerpc/kernel/head_64.o:(__restart_table+0x10)

Newer GNU ld configured with --enable-textrel-check=error will report an error as well:

ld/ld-new: read-only segment has dynamic relocations

Text relocations are considered very awful by linker developers. binutils 2.35 added --enable-textrel-check={no,warn,error}
https://sourceware.org/bugzilla/show_bug.cgi?id=20824

I can imagine that in the future some Linux distributions (especially those focusing on security) will default their binutils to use --enable-textrel-check={no,warn,error}. CONFIG_RELOCATABLE build will break sooner or later.

In -no-pie links, R_PPC64_ADDR64 relocations are link-time constants. There are no text relocations, therefore no need for -z notext.

Quoted 13 lines

Kind regards,
Daniel

Quoted 5 lines

LDFLAGS_vmlinux := $(LDFLAGS_vmlinux-y)

ifdef CONFIG_PPC64
--
2.33.0.rc1.237.g0d66db33f3-goog

Signature · 3 lines
You received this message because you are subscribed to the Google Groups "Clang Built Linux" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].
To view this discussion on the web visit https://groups.google.com/d/msgid/clang-built-linux/87sfzde8lk.fsf%40linkitivity.dja.id.au.
#8 lore raw
Quoted 1 line · Fangrui Song

Text relocations are considered very awful by linker developers.

By very few linker developers.

Quoted 2 lines

binutils 2.35 added --enable-textrel-check={no,warn,error}
https://sourceware.org/bugzilla/show_bug.cgi?id=20824

Yes, some people wanted the default to be configurable. So now we have a default default that is sane, so most people get to reap the benefits of having defaults at all, but we also allow other people to shoot themselves (and people who have to deal with them) in the foot. "Progress". Changing the defaults should be a one-time event, only done when the benefits strongly outweigh the costs. Defaults should never be configurable (by the user).

Quoted 3 lines

I can imagine that in the future some Linux distributions (especially those focusing on security) will default their binutils to use --enable-textrel-check={no,warn,error}.

How would this be a benefit to security?

Quoted 1 line

In -no-pie links, R_PPC64_ADDR64 relocations are link-time constants.

Where "link" includes dynamic links as well. There are no constants.

Quoted 1 line

There are no text relocations, therefore no need for -z notext.

This is a choice by the compiler, nothing more. It saves some process startup time, and allows slightly more maps to be shared by processes that run the same images. But it is a tradeoff, so it might change; and of course it is not an ABI requirement.

Segher

#9 lore raw
Quoted 5 lines · Segher Boessenkool
Quoted 1 line · Fangrui Song

Text relocations are considered very awful by linker developers.

By very few linker developers.

https://groups.google.com/g/generic-abi/c/Ckq19PfLxyk/m/uW29sgkoAgAJ
Ali Bahrami: "My opinion is that no one wants text relocations, but not labeling them if they do exist doesn't seem right. I find the presence of DF_TEXTREL very interesting when diagnosing various issues."

https://gcc.gnu.org/legacy-ml/gcc/2016-04/msg00138.html
( "So why not simply create 'dynamic text relocations' then? Is that possible with a pure linker change?" )
Cary Coutant: "Ugh. Besides being a bad idea from a performance point of view, it's not even always possible to do. Depending on the architecture, a direct reference from an executable to a variable in a shared library may not have the necessary reach."

binutils-gdb commit "Add linker option: --warn-shared-textrel to produce warnings when adding a DT_TEXTREL to a shared object." Nick Clifton

https://www.openwall.com/lists/musl/2020/09/26/3
Szabolcs Nagy: "nice. and gcc passes -z text for static pie code so that case should not end up with text rels."

Someone wrote "Overall, the overhead of processing text relocations can cause serious performance degradation." in Solaris' Linker and Libraries Guide.

Me :)
(I wrote lld/ELF 9.0.0~13.0.0 release notes and filed dozen of GNU ld/gold bugs/feature requests)

Quoted 10 lines
Quoted 2 lines

binutils 2.35 added --enable-textrel-check={no,warn,error}
https://sourceware.org/bugzilla/show_bug.cgi?id=20824

Yes, some people wanted the default to be configurable. So now we have a default default that is sane, so most people get to reap the benefits of having defaults at all, but we also allow other people to shoot themselves (and people who have to deal with them) in the foot. "Progress". Changing the defaults should be a one-time event, only done when the benefits strongly outweigh the costs. Defaults should never be configurable (by the user).

ld.lld has such a non-configurable model and thus caught the issue (which the patch intends to address).

Future --enable-textrel-check={yes,error} configured GNU ld will be similar.

Quoted 5 lines
Quoted 3 lines

I can imagine that in the future some Linux distributions (especially those focusing on security) will default their binutils to use --enable-textrel-check={no,warn,error}.

How would this be a benefit to security?

https://flameeyes.blog/2016/01/16/textrels-text-relocations-and-their-impact-on-hardening-techniques/

https://github.com/golang/go/issues/9210
Android: "libexample.so has text relocations. This is wasting memory and prevents security hardening. Please fix."

FWIW I contributed a glibc patch allowing TEXTREL to co-exist with ifunc. It requires temporary mapping the text segment W^X.

Quoted 3 lines
Quoted 1 line

In -no-pie links, R_PPC64_ADDR64 relocations are link-time constants.

Where "link" includes dynamic links as well. There are no constants.

"Link-time" usually refers to the processing of the static linker.

Dynamic links can use load-time or run-time.

Quoted 9 lines
Quoted 1 line

There are no text relocations, therefore no need for -z notext.

This is a choice by the compiler, nothing more. It saves some process startup time, and allows slightly more maps to be shared by processes that run the same images. But it is a tradeoff, so it might change; and of course it is not an ABI requirement.

Segher

Text relocations are generally awful.
GNU ld and gold's traditional "add DF_TEXTREL on-demand" behavior made such user errors easy to make.
I understand that kernels are special applications where we apply relocations once and many user-space objection can be less of a concern/ignored.
However, the Linux kernel is already in a position where many linker options are controlled and thus should specify -z notext to make the intention explicit, or fix the problems (I think x86-64 is good; that said, powerpc
has a higher cost using PC-relative instructions so pay the oneshot relocation time cost probably isn't a bad choice)

--
宋方睿

#10 lore raw

Hi!

Quoted 6 lines · Fāng-ruì Sòng
Quoted 4 lines · Segher Boessenkool
Quoted 1 line · Fangrui Song

Text relocations are considered very awful by linker developers.

Quoted 7 lines
Quoted 1 line

By very few linker developers.

https://groups.google.com/g/generic-abi/c/Ckq19PfLxyk/m/uW29sgkoAgAJ
Ali Bahrami: "My opinion is that no one wants text relocations, but not labeling them if they do exist doesn't seem right. I find the presence of DF_TEXTREL very interesting when diagnosing various issues."

I don't know who that is, and that post has no context.

Quoted 7 lines

https://gcc.gnu.org/legacy-ml/gcc/2016-04/msg00138.html
( "So why not simply create 'dynamic text relocations' then? Is that possible with a pure linker change?" )
Cary Coutant: "Ugh. Besides being a bad idea from a performance point of view, it's not even always possible to do. Depending on the architecture, a direct reference from an executable to a variable in a shared library may not have the necessary reach."

That is about a very specific kind of relocation.

Quoted 3 lines

binutils-gdb commit "Add linker option: --warn-shared-textrel to produce warnings when adding a DT_TEXTREL to a shared object." Nick Clifton

That does not say text relocations are bad. Of course you want to know if they are there, for various reasons, like, if they are disallowed on some systems.

Quoted 3 lines

https://www.openwall.com/lists/musl/2020/09/26/3
Szabolcs Nagy: "nice. and gcc passes -z text for static pie code so that case should not end up with text rels."

That does not say text relocations are bad.

Quoted 3 lines

Someone wrote "Overall, the overhead of processing text relocations can cause serious performance degradation." in Solaris' Linker and Libraries Guide.

In process startup, sure. And it can make those processes run faster as well. That is the tradeoff with *all* relocations; you can make any code without any relocations. Relocations are a tradeoff, like most things.

Quoted 3 lines
Quoted 1 line

How would this be a benefit to security?

https://flameeyes.blog/2016/01/16/textrels-text-relocations-and-their-impact-on-hardening-techniques/

This means that those "hardening techniques" have some serious weaknesses, that is all. And hardening is not part of security anyway; it is impact mitigation.

Quoted 2 lines

FWIW I contributed a glibc patch allowing TEXTREL to co-exist with ifunc. It requires temporary mapping the text segment W^X.

What does W^X mean here? It normally means no mapping is both writable and executable at the same time.

Quoted 8 lines
Quoted 6 lines
Quoted 1 line

There are no text relocations, therefore no need for -z notext.

This is a choice by the compiler, nothing more. It saves some process startup time, and allows slightly more maps to be shared by processes that run the same images. But it is a tradeoff, so it might change; and of course it is not an ABI requirement.

Text relocations are generally awful.

Great arguments, thanks! :-P

Quoted 2 lines

GNU ld and gold's traditional "add DF_TEXTREL on-demand" behavior made such user errors easy to make.

That has no bearing on if text relocations are useful or not.

Quoted 9 lines

I understand that kernels are special applications where we apply relocations once and many user-space objection can be less of a concern/ignored.
However, the Linux kernel is already in a position where many linker options are controlled and thus should specify -z notext to make the intention explicit, or fix the problems (I think x86-64 is good; that said, powerpc
has a higher cost using PC-relative instructions so pay the oneshot relocation time cost probably isn't a bad choice)

I have no idea what you mean.

Segher

#11 lore raw

PATCH v2 1/1 ppc: add "-z notext" flag to disable diagnostic

In reply to Bill Wendling

From: Fangrui Song <[email protected]>

Object files used to link .tmp_vmlinux.kallsyms1 have many R_PPC64_ADDR64 relocations in non-SHF_WRITE sections. There are many text relocations (e.g. in .rela___ksymtab_gpl+* and .rela__mcount_loc sections) in a -pie link and are disallowed by LLD:

ld.lld: error: can't create dynamic relocation R_PPC64_ADDR64 against local symbol in readonly segment; recompile object files with -fPIC or pass '-Wl,-z,notext' to allow text relocations in the output

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

defined in arch/powerpc/kernel/head_64.o
referenced by arch/powerpc/kernel/head_64.o:(__restart_table+0x10)

Newer GNU ld configured with "--enable-textrel-check=error" will report an error as well:

  $ ld-new -EL -m elf64lppc -pie ... -o .tmp_vmlinux.kallsyms1 ...
  ld-new: read-only segment has dynamic relocations

Add "-z notext" to suppress the errors. Non-CONFIG_RELOCATABLE builds use the default -no-pie mode and thus R_PPC64_ADDR64 relocations can be resolved at link-time.

---
v2:
- Assign "Fangrui Song" as the proper author.
- Improve the commit message to add more context.
- Appending tags from original patch's review.

---
 arch/powerpc/Makefile | 1 +
 1 file changed, 1 insertion(+)
arch/powerpc/Makefile +1 -0
index 6505d66f1193..17a9fbf9b789 100644
@@ -122,6 +122,7 @@ endif
 
 LDFLAGS_vmlinux-y := -Bstatic
 LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) := -pie
+LDFLAGS_vmlinux-$(CONFIG_RELOCATABLE) += -z notext
 LDFLAGS_vmlinux	:= $(LDFLAGS_vmlinux-y)
 
 ifdef CONFIG_PPC64
Signature · 1 line
2.33.0.rc1.237.g0d66db33f3-goog
#12 lore raw
Quoted 10 lines · Bill Wendling

Object files used to link .tmp_vmlinux.kallsyms1 have many R_PPC64_ADDR64 relocations in non-SHF_WRITE sections. There are many text relocations (e.g. in .rela___ksymtab_gpl+* and .rela__mcount_loc sections) in a -pie link and are disallowed by LLD:

ld.lld: error: can't create dynamic relocation R_PPC64_ADDR64 against local symbol in readonly segment; recompile object files with -fPIC or pass '-Wl,-z,notext' to allow text relocations in the output

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

defined in arch/powerpc/kernel/head_64.o
referenced by arch/powerpc/kernel/head_64.o:(__restart_table+0x10)

[...]

Applied to powerpc/next.

[1/1] ppc: add "-z notext" flag to disable diagnostic

      https://git.kernel.org/powerpc/c/0355785313e2191be4e1108cdbda94ddb0238c48

cheers