arm: Guard ucontext _rtld_global_ro access by SHARED, not PIC macro
Checks
Context |
Check |
Description |
dj/TryBot-apply_patch |
success
|
Patch applied to master at the time it was sent
|
dj/TryBot-32bit |
success
|
Build for i686
|
Commit Message
Due to PIE-by-default, PIC is now defined in more cases. libc.a
does not have _rtld_global_ro, and statically linking setcontext
fails. SHARED is the right condition to use, so that libc.a
references _dl_hwcap instead of _rtld_global_ro.
For static PIE support, the !SHARED case would still have to be made
PIC. This patch does not achieve that.
Fixes commit 23645707f12f2dd9d80b51effb2d9618a7b65565
("Replace --enable-static-pie with --disable-default-pie").
Please not that I have not been able to test/build this change so far.
Thanks,
Florian
---
sysdeps/unix/sysv/linux/arm/getcontext.S | 4 ++--
sysdeps/unix/sysv/linux/arm/setcontext.S | 4 ++--
2 files changed, 4 insertions(+), 4 deletions(-)
Comments
On 12/17/21 01:01, Florian Weimer via Libc-alpha wrote:
> Due to PIE-by-default, PIC is now defined in more cases. libc.a
> does not have _rtld_global_ro, and statically linking setcontext
> fails. SHARED is the right condition to use, so that libc.a
> references _dl_hwcap instead of _rtld_global_ro.
>
> For static PIE support, the !SHARED case would still have to be made
> PIC. This patch does not achieve that.
>
> Fixes commit 23645707f12f2dd9d80b51effb2d9618a7b65565
> ("Replace --enable-static-pie with --disable-default-pie").
>
> Please not that I have not been able to test/build this change so far.
The change looks correct to me; the _rtld_global_ro vs _dl_hwcaps usage
is a SHARED vs static decision and doesn't depend on whether the code is
PIC or not.
So,
Reviewed-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
but please also get Szabolcs' ack too.
Thanks,
Siddhesh
The 12/17/2021 01:20, Siddhesh Poyarekar wrote:
> On 12/17/21 01:01, Florian Weimer via Libc-alpha wrote:
> > Due to PIE-by-default, PIC is now defined in more cases. libc.a
> > does not have _rtld_global_ro, and statically linking setcontext
> > fails. SHARED is the right condition to use, so that libc.a
> > references _dl_hwcap instead of _rtld_global_ro.
> >
> > For static PIE support, the !SHARED case would still have to be made
> > PIC. This patch does not achieve that.
> >
> > Fixes commit 23645707f12f2dd9d80b51effb2d9618a7b65565
> > ("Replace --enable-static-pie with --disable-default-pie").
> >
> > Please not that I have not been able to test/build this change so far.
>
> The change looks correct to me; the _rtld_global_ro vs _dl_hwcaps usage is a
> SHARED vs static decision and doesn't depend on whether the code is PIC or
> not.
>
> So,
>
> Reviewed-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
>
> but please also get Szabolcs' ack too.
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
(i ran the tests and checked the generated code, it looked good)
@@ -50,7 +50,7 @@ ENTRY(__getcontext)
/* Store FP regs. Much of the FP code is copied from arm/setjmp.S. */
-#ifdef PIC
+#ifdef SHARED
ldr r2, 1f
ldr r1, .Lrtld_global_ro
0: add r2, pc, r2
@@ -102,7 +102,7 @@ ENTRY(__getcontext)
END(__getcontext)
-#ifdef PIC
+#ifdef SHARED
1: .long _GLOBAL_OFFSET_TABLE_ - 0b - PC_OFS
.Lrtld_global_ro:
.long C_SYMBOL_NAME(_rtld_global_ro)(GOT)
@@ -32,7 +32,7 @@ ENTRY(__setcontext)
add r0, r0, #UCONTEXT_REGSPACE
/* Restore the VFP registers. Copied from arm/__longjmp.S. */
-#ifdef PIC
+#ifdef SHARED
ldr r2, 1f
ldr r1, .Lrtld_global_ro
0: add r2, pc, r2
@@ -101,7 +101,7 @@ ENTRY(__startcontext)
.fnend
END(__startcontext)
-#ifdef PIC
+#ifdef SHARED
1: .long _GLOBAL_OFFSET_TABLE_ - 0b - PC_OFS
.Lrtld_global_ro:
.long C_SYMBOL_NAME(_rtld_global_ro)(GOT)