NetBSD Problem Report #57511
From www@netbsd.org Fri Jul 7 08:17:57 2023
Return-Path: <www@netbsd.org>
Received: from mail.netbsd.org (mail.netbsd.org [199.233.217.200])
(using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits))
(Client CN "mail.NetBSD.org", Issuer "mail.NetBSD.org CA" (not verified))
by mollari.NetBSD.org (Postfix) with ESMTPS id 9F5421A923D
for <gnats-bugs@gnats.NetBSD.org>; Fri, 7 Jul 2023 08:17:57 +0000 (UTC)
Message-Id: <20230707081756.18F281A923E@mollari.NetBSD.org>
Date: Fri, 7 Jul 2023 08:17:56 +0000 (UTC)
From: rokuyama.rk@gmail.com
Reply-To: rokuyama.rk@gmail.com
To: gnats-bugs@NetBSD.org
Subject: ld.elf_so(1) does not support secure PLT for alpha
X-Send-Pr-Version: www-1.0
>Number: 57511
>Notify-List: riastradh@NetBSD.org
>Category: port-alpha
>Synopsis: ld.elf_so(1) does not support secure PLT for alpha
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: port-alpha-maintainer
>State: analyzed
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Fri Jul 07 08:20:00 +0000 2023
>Closed-Date:
>Last-Modified: Sun Apr 20 06:25:01 +0000 2025
>Originator: Rin Okuyama
>Release: 10.99.4
>Organization:
Internet Initiative Japan Inc.
>Environment:
NetBSD ds10 10.99.4 NetBSD 10.99.4 (GENERIC-$Revision: 1.418 $) #6: Wed Jun 28 19:19:37 JST 2023 rin@latipes:/home/rin/repos/src/sys/arch/alpha/compile/DS10 alpha
>Description:
Dynamic-linked binaries for alpha built with binutils 2.39 crash
due to SIGSEGV in __start, trying to jump into NULL.
This is worked around by forcibly disabling secure PLT option,
introduced by binutils 2.39 for us.
With this option, write permission to .plt section is dropped,
and we need to parse GOT differently as before, see
``if (elf64_alpha_use_secureplt)'' blocks in
external/gpl3/binutils/dist/bfd/elf64-alpha.c.
Also note that there is no official System V (or ELF) ABI
definitions published for alpha.
>How-To-Repeat:
Try to run any dynamically-linked binaries generated by binutils
2.39 on alpha. Then, it will crash due to SIGSEGV.
>Fix:
Add support for secure PLT for ld.elf_so(1) on alpha.
>Release-Note:
>Audit-Trail:
From: "Rin Okuyama" <rin@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc:
Subject: PR/57511 CVS commit: src/external/gpl3/binutils/dist/bfd
Date: Fri, 7 Jul 2023 08:33:25 +0000
Module Name: src
Committed By: rin
Date: Fri Jul 7 08:33:25 UTC 2023
Modified Files:
src/external/gpl3/binutils/dist/bfd: elf64-alpha.c
Log Message:
PR port-alpha/57511 Temporally disable secure PLT for alpha
until support is added to ld.elf_so(1).
To generate a diff of this commit:
cvs rdiff -u -r1.10 -r1.11 src/external/gpl3/binutils/dist/bfd/elf64-alpha.c
Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.
From: "Rin Okuyama" <rin@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc:
Subject: PR/57511 CVS commit: src/doc
Date: Fri, 7 Jul 2023 08:39:24 +0000
Module Name: src
Committed By: rin
Date: Fri Jul 7 08:39:24 UTC 2023
Modified Files:
src/doc: HACKS
Log Message:
doc/HACKS: Document PR port-alpha/57511
To generate a diff of this commit:
cvs rdiff -u -r1.236 -r1.237 src/doc/HACKS
Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.
From: "Rin Okuyama" <rin@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc:
Subject: PR/57511 CVS commit: src/external/gpl3/gdb/dist/bfd
Date: Thu, 17 Aug 2023 07:40:20 +0000
Module Name: src
Committed By: rin
Date: Thu Aug 17 07:40:20 UTC 2023
Modified Files:
src/external/gpl3/gdb/dist/bfd: configure
Log Message:
gdb/bfd: Disable secure PLT for NetBSD/alpha, lost during merge
Until our ld.elf_so supports it, see PR port-alpha/57511
This does not affect platforms other than alpha.
XXX
mknative later
To generate a diff of this commit:
cvs rdiff -u -r1.16 -r1.17 src/external/gpl3/gdb/dist/bfd/configure
Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.
From: Taylor R Campbell <riastradh@NetBSD.org>
To: rokuyama.rk@gmail.com, thorpej@NetBSD.org
Cc: gnats-bugs@NetBSD.org, netbsd-bugs@NetBSD.org
Subject: Re: port-alpha/57511: ld.elf_so(1) does not support secure PLT for alpha
Date: Thu, 17 Apr 2025 19:29:46 +0000
This is a multi-part message in MIME format.
--=_k+R2A0jPCfl/fFvU84G4E4n02s6AxFtx
The attached patch addresses PR 57511 and PR 57717 by implementing
support for secureplt in ld.elf_so on alpha and flipping it back on by
default (as upstream has it) in binutils.
I also took the liberty of leaving some more explanatory comments
about what's going on in rtld_start.S, based on examination of PLT
headers actually generated by binutils.
Missing: automatic tests of programs built with `-Wl,--secureplt',
`-Wl,--no-secureplt', and `-Wl,--no-secureplt,--traditional-format',
so we make sure all three paths remain exercised.
Review welcome! I hope my accent isn't showing through too much in
the assembly code as a non-native Alpha speaker.
--=_k+R2A0jPCfl/fFvU84G4E4n02s6AxFtx
Content-Type: text/plain; charset="ISO-8859-1"; name="pr57511-alphasecureplt"
Content-Transfer-Encoding: quoted-printable
Content-Disposition: attachment; filename="pr57511-alphasecureplt.patch"
# HG changeset patch
# User Taylor R Campbell <riastradh@NetBSD.org>
# Date 1744917745 0
# Thu Apr 17 19:22:25 2025 +0000
# Branch trunk
# Node ID ff8aa06d8f252223c352e0560098b25e54967e80
# Parent 438b14ea036dd2bd2e1109f5e08f7d74109e6fae
# EXP-Topic riastradh-pr57511-alphasecureplt
ld.elf_so on alpha: Add support for secureplt.
Enable secureplt by default in binutils now that ld.elf_so can handle
it.
PR port-alpha/57511: ld.elf_so(1) does not support secure PLT for alpha
PR port-alpha/57717: Alpha linker generates LOAD segments (and
generates warnings about them)
diff -r 438b14ea036d -r ff8aa06d8f25 doc/HACKS
--- a/doc/HACKS Thu Apr 17 13:43:59 2025 +0000
+++ b/doc/HACKS Thu Apr 17 19:22:25 2025 +0000
@@ -949,17 +949,6 @@ descr
kernel cannot execute /sbin/init.
kcah
=20
-hack Disable secure PLT for alpha
-cdate Fri Jul 7 08:33:25 UTC 2023
-who rin
-port alpha
-file src/external/gpl3/binutils/dist/bfd/elf64-alpha.c: 1.11
-pr port-alpha/57511
-descr
- Temporally disable secure PLT support, introduced by
- binutils 2.39 for us. We need its support to ld.elf_so(1).
-kcah
-
hack openssl: Disable ec_nistp_64_gcc_128 on aarch64eb
cdate Sat Jul 8 23:54:27 UTC 2023
who rin
diff -r 438b14ea036d -r ff8aa06d8f25 external/gpl3/binutils/dist/bfd/elf64-=
alpha.c
--- a/external/gpl3/binutils/dist/bfd/elf64-alpha.c Thu Apr 17 13:43:59 202=
5 +0000
+++ b/external/gpl3/binutils/dist/bfd/elf64-alpha.c Thu Apr 17 19:22:25 202=
5 +0000
@@ -83,7 +83,7 @@
=20
/* Set by ld emulation. Putting this into the link_info or hash structure
is simply working too hard. */
-#if defined(USE_SECUREPLT) && 0 /* XXX port-alpha/57511 */
+#if defined(USE_SECUREPLT)
bool elf64_alpha_use_secureplt =3D true;
#else
bool elf64_alpha_use_secureplt =3D false;
diff -r 438b14ea036d -r ff8aa06d8f25 libexec/ld.elf_so/arch/alpha/alpha_rel=
oc.c
--- a/libexec/ld.elf_so/arch/alpha/alpha_reloc.c Thu Apr 17 13:43:59 2025 +=
0000
+++ b/libexec/ld.elf_so/arch/alpha/alpha_reloc.c Thu Apr 17 19:22:25 2025 +=
0000
@@ -78,6 +78,7 @@
#define adbg(x) /* nothing */
#endif
=20
+void _rtld_bind_start_secureplt(void);
void _rtld_bind_start(void);
void _rtld_bind_start_old(void);
void _rtld_relocate_nonplt_self(Elf_Dyn *, Elf_Addr);
@@ -91,7 +92,19 @@ void
uint32_t word0;
=20
/*
- * The PLTGOT on the Alpha looks like this:
+ * If we're using Alpha secureplt, the PLTGOT points to the
+ * .got.plt section. Just fill in the rtld binding stub and
+ * we're done -- we're not writing to instruction memory, so no
+ * imb needed.
+ */
+ if (obj->secureplt) {
+ obj->pltgot[0] =3D (Elf_Addr) _rtld_bind_start_secureplt;
+ obj->pltgot[1] =3D (Elf_Addr) obj;
+ return;
+ }
+
+ /*
+ * The non-secureplt PLTGOT on the Alpha looks like this:
*
* PLT HEADER
* .
@@ -418,7 +431,8 @@ static inline int
*
* Note if the shared object uses the old PLT format, then
* we cannot patch up the PLT safely, and so we skip it
- * in that case[*].
+ * in that case[*]. And if the shared object has a read-only
+ * secureplt, then we also skip it.
*
* [*] Actually, if we're not doing lazy-binding, then
* we *can* (and do) patch up this PLT entry; the PLTGOT
@@ -426,6 +440,10 @@ static inline int
* so this test will fail as it would for the new PLT
* entry format.
*/
+ if (obj->secureplt) {
+ rdbg((" secureplt format"));
+ goto out;
+ }
if (obj->pltgot[2] =3D=3D (Elf_Addr) &_rtld_bind_start_old) {
rdbg((" old PLT format"));
goto out;
diff -r 438b14ea036d -r ff8aa06d8f25 libexec/ld.elf_so/arch/alpha/rtld_star=
t.S
--- a/libexec/ld.elf_so/arch/alpha/rtld_start.S Thu Apr 17 13:43:59 2025 +0=
000
+++ b/libexec/ld.elf_so/arch/alpha/rtld_start.S Thu Apr 17 19:22:25 2025 +0=
000
@@ -92,7 +92,7 @@ END(_rtld_start)
lda sp, -168(sp) ; \
stq ra, 0(sp) ; \
stq v0, 8(sp) ; \
- stq t0, 16(sp) ; \
+ stq t0, 16(sp) /* XXX t0-t7 necessary? */ ; \
stq t1, 24(sp) ; \
stq t2, 32(sp) ; \
stq t3, 40(sp) ; \
@@ -106,7 +106,7 @@ END(_rtld_start)
stq a3, 104(sp) ; \
stq a4, 112(sp) ; \
stq a5, 120(sp) ; \
- stq t8, 128(sp) ; \
+ stq t8, 128(sp) /* XXX t8-t11 necessary? */ ; \
stq t9, 136(sp) ; \
stq t10, 144(sp) ; \
stq t11, 152(sp) ; \
@@ -119,7 +119,7 @@ END(_rtld_start)
br t0, 1f ; \
1: LDGP(t0)
=20
-#define RTLD_BIND_START_EPILOGUE \
+#define RTLD_BIND_START_EPILOGUE(imb) \
/* Move the destination address into position. */ \
mov v0, pv ; \
\
@@ -157,7 +157,77 @@ 1: LDGP(t0)
jmp zero, (pv)
=20
/*
- * Lazy binding entry point, called via PLT.
+ * _rtld_bind_start_secureplt(_rtld_bind_start_secureplt@pv, obj@at,
+ * (sizeof(Elf_Rela)*index)@t11)
+ *
+ * Lazy binding entry point, called via PLT with read-only
+ * secureplt, when DT_ALPHA_PLTRO is set. The PLT itself looks
+ * something like this:
+ *
+ * _PROCEDURE_LINKAGE_TABLE_:
+ * subq pv, at, t11 // t11 :=3D pv - ent0 =3D 4*index
+ * s4subq t11, t11, t11 // t11 :=3D 12*index
+ * addq t11, t11, t11 // t11 :=3D 24*index
+ * // =3D sizeof(Elf_Rela)*index
+ * ldah at, ...(at) // at :=3D PLTGOT
+ * lda at, ...(at)
+ * ldq pv, 0(at) // pv :=3D PLTGOT[0]
+ * // =3D _rtld_bind_start_secureplt
+ * ldq at, 8(at) // at :=3D PLTGOT[1]
+ * // =3D obj
+ * jmp (pv)
+ * 0: br at, _PROCEDURE_LINKAGE_TABLE_ // at :=3D ent0
+ * ent0: br 0b // pv - ent0 =3D 0 =3D 4*index
+ * ent1: br 0b // pv - ent0 =3D 4 =3D 4*index
+ * ent2: br 0b // pv - ent0 =3D 8 =3D 4*index
+ * ...
+ */
+NESTED_NOPROFILE(_rtld_bind_start_secureplt, 0, 168, ra, 0, 0)
+
+ RTLD_BIND_START_PROLOGUE
+
+ /* Set up the arguments for _rtld_bind. */
+ mov at_reg, a0
+ mov t11, a1
+
+ CALL(_rtld_bind)
+
+ RTLD_BIND_START_EPILOGUE(/* no text writes, so no imb */)
+
+END(_rtld_bind_start_secureplt)
+
+/*
+ * _rtld_bind_start(_rtld_bind_start@pv, &PLTGOT[2]@pv,
+ * (ent0 + 4*(index + 1))@at)
+ *
+ * Lazy binding entry point, called via PLT with read/write
+ * non-secureplt, when DT_ALPHA_PLTRO is not set. The PLT itself
+ * looks something like this at program startup, with PLTGOT (an
+ * array of 64-bit Elf_Addr) pointing at _PROCEDURE_LINKAGE_TABLE_
+ * and PLTGOT[2] and PLTGOT[3] initialized by _rtld_setup_pltgot:
+ *
+ * _PROCEDURE_LINKAGE_TABLE_:
+ * br pv, .Lref // pv :=3D .Lref
+ * .Lref: ldq pv, 12(pv) // pv :=3D PLTGOT[2]
+ * =3D _rtld_bind_start
+ * unop // no-op for alignment
+ * jmp pv, (pv) // pv :=3D &PLTGOT[2]
+ * .qword (_rtld_bind_start) // PLTGOT[2]
+ * .qword (object pointer) // PLTGOT[3]
+ * ent0: br at, _PROCEDURE_LINKAGE_TABLE_
+ * unop // space for adjusted stub
+ * unop // space for adjusted stub
+ * ent1: br at, _PROCEDURE_LINKAGE_TABLE_
+ * unop
+ * unop
+ * ent2: br at, _PROCEDURE_LINKAGE_TABLE_
+ * unop
+ * unop
+ * ...
+ *
+ * Note: Distance from &PLTGOT[2] (pv) to ent[0] + 4 (at) is 20
+ * bytes, and each ent[index] + 4 (at) after that is separated by
+ * 3 instructions, i.e., 12 bytes.
*/
NESTED_NOPROFILE(_rtld_bind_start, 0, 168, ra, 0, 0)
=20
@@ -171,13 +241,37 @@ NESTED_NOPROFILE(_rtld_bind_start, 0, 16
=20
CALL(_rtld_bind)
=20
- RTLD_BIND_START_EPILOGUE
+ RTLD_BIND_START_EPILOGUE(imb)
=20
END(_rtld_bind_start)
=20
/*
- * Lazy binding entry point, called via PLT. This version is for the
- * old PLT entry format.
+ * _rtld_bind_start_old(&PLTGOT[2]@pv, (sizeof(Elf_Rela)*index)@at)
+ *
+ * Lazy binding entry point, called via PLT. This version is for
+ * the old PLT entry format, for which the PLT looks something
+ * like this at program startup, with PLTGOT (an array of 64-bit
+ * Elf_Addr) pointing at _PROCEDURE_LINKAGE_TABLE_, and PLTGOT[2]
+ * and PLTGOT[3] initialized by _rtld_setup_pltgot:
+ *
+ * _PROCEDURE_LINKAGE_TABLE_:
+ * br pv, 1f // pv :=3D .Lref
+ * .Lref: ldq pv, 12(pv) // pv :=3D PLTGOT[2]
+ * =3D _rtld_bind_start
+ * unop // no-op for alignment
+ * jmp pv, (pv) // pv :=3D &PLTGOT[2]
+ * .qword (_rtld_bind_start) // PLTGOT[2]
+ * .qword (object pointer) // PLTGOT[3]
+ * ent0: ldah at, 0 // at :=3D 24*0
+ * lda at, 0(at) // =3D sizeof(Elf_Rela)*index
+ * br _PROCEDURE_LINKAGE_TABLE_
+ * ent1: ldah at, 0 // at :=3D 24*1
+ * lda at, 24(at) // =3D sizeof(Elf_Rela)*index
+ * br _PROCEDURE_LINKAGE_TABLE_
+ * ent3: ldah at, 0 // at :=3D 24*2
+ * lda at, 48(at) // =3D sizeof(Elf_Rela)*index
+ * br _PROCEDURE_LINKAGE_TABLE_
+ * ...
*/
NESTED_NOPROFILE(_rtld_bind_start_old, 0, 168, ra, 0, 0)
=20
@@ -189,6 +283,6 @@ NESTED_NOPROFILE(_rtld_bind_start_old, 0
=20
CALL(_rtld_bind)
=20
- RTLD_BIND_START_EPILOGUE
+ RTLD_BIND_START_EPILOGUE(imb)
=20
END(_rtld_bind_start_old)
diff -r 438b14ea036d -r ff8aa06d8f25 libexec/ld.elf_so/headers.c
--- a/libexec/ld.elf_so/headers.c Thu Apr 17 13:43:59 2025 +0000
+++ b/libexec/ld.elf_so/headers.c Thu Apr 17 19:22:25 2025 +0000
@@ -347,6 +347,11 @@ void
break;
#endif
=20
+#ifdef __alpha__
+ case DT_ALPHA_PLTRO:
+ obj->secureplt =3D (dynp->d_un.d_val !=3D 0);
+ break;
+#endif
#ifdef __mips__
case DT_MIPS_LOCAL_GOTNO:
obj->local_gotno =3D dynp->d_un.d_val;
diff -r 438b14ea036d -r ff8aa06d8f25 libexec/ld.elf_so/rtld.h
--- a/libexec/ld.elf_so/rtld.h Thu Apr 17 13:43:59 2025 +0000
+++ b/libexec/ld.elf_so/rtld.h Thu Apr 17 19:22:25 2025 +0000
@@ -218,6 +218,9 @@ typedef struct Struct_Obj_Entry {
dlopen'ed */
phdr_loaded:1, /* Phdr is loaded and doesn't need to
* be freed. */
+#ifdef __alpha__
+ secureplt:1, /* True if PLT is read-only format */
+#endif
#if defined(__HAVE_TLS_VARIANT_I) || defined(__HAVE_TLS_VARIANT_II)
tls_static:1, /* True if static TLS offset
* has been allocated */
diff -r 438b14ea036d -r ff8aa06d8f25 sys/arch/alpha/include/elf_machdep.h
--- a/sys/arch/alpha/include/elf_machdep.h Thu Apr 17 13:43:59 2025 +0000
+++ b/sys/arch/alpha/include/elf_machdep.h Thu Apr 17 19:22:25 2025 +0000
@@ -74,4 +74,7 @@
=20
#define R_TYPE(name) __CONCAT(R_ALPHA_,name)
=20
+/* Alpha dynamic tags */
+#define DT_ALPHA_PLTRO 0x70000000 /* secure (read-only) PLT */
+
#endif /* _ALPHA_ELF_MACHDEP_H_ */
--=_k+R2A0jPCfl/fFvU84G4E4n02s6AxFtx--
State-Changed-From-To: open->analyzed
State-Changed-By: riastradh@NetBSD.org
State-Changed-When: Thu, 17 Apr 2025 22:10:35 +0000
State-Changed-Why:
patch posted
From: Jason Thorpe <thorpej@me.com>
To: Taylor Campbell <riastradh@NetBSD.org>
Cc: Rin Okuyama <rokuyama.rk@gmail.com>,
Jason Thorpe <thorpej@NetBSD.org>,
"gnats-bugs@netbsd.org" <gnats-bugs@NetBSD.org>,
"netbsd-bugs@netbsd.org" <netbsd-bugs@NetBSD.org>
Subject: Re: port-alpha/57511: ld.elf_so(1) does not support secure PLT for
alpha
Date: Fri, 18 Apr 2025 09:00:17 -0700
> On Apr 17, 2025, at 12:29=E2=80=AFPM, Taylor R Campbell =
<riastradh@NetBSD.org> wrote:
>=20
> The attached patch addresses PR 57511 and PR 57717 by implementing
> support for secureplt in ld.elf_so on alpha and flipping it back on by
> default (as upstream has it) in binutils.
>=20
> I also took the liberty of leaving some more explanatory comments
> about what's going on in rtld_start.S, based on examination of PLT
> headers actually generated by binutils.
>=20
> Missing: automatic tests of programs built with `-Wl,--secureplt',
> `-Wl,--no-secureplt', and `-Wl,--no-secureplt,--traditional-format',
> so we make sure all three paths remain exercised.
>=20
> Review welcome! I hope my accent isn't showing through too much in
> the assembly code as a non-native Alpha speaker.
> <pr57511-alphasecureplt.patch>
Looks good to me. I do have concerns about the performance implications =
of large programs that end up with multiple GOTs, but that=E2=80=99s an =
issue with the format, not with the change itself.
-- thorpej
From: "Taylor R Campbell" <riastradh@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc:
Subject: PR/57511 CVS commit: src
Date: Fri, 18 Apr 2025 17:56:50 +0000
Module Name: src
Committed By: riastradh
Date: Fri Apr 18 17:56:50 UTC 2025
Modified Files:
src/libexec/ld.elf_so: headers.c rtld.h
src/libexec/ld.elf_so/arch/alpha: alpha_reloc.c rtld_start.S
src/sys/arch/alpha/include: elf_machdep.h
Log Message:
ld.elf_so on alpha: Add support for secureplt.
ok thorpej
PR port-alpha/57511: ld.elf_so(1) does not support secure PLT for alpha
PR port-alpha/57717: Alpha linker generates LOAD segments (and
generates warnings about them)
To generate a diff of this commit:
cvs rdiff -u -r1.73 -r1.74 src/libexec/ld.elf_so/headers.c
cvs rdiff -u -r1.147 -r1.148 src/libexec/ld.elf_so/rtld.h
cvs rdiff -u -r1.45 -r1.46 src/libexec/ld.elf_so/arch/alpha/alpha_reloc.c
cvs rdiff -u -r1.16 -r1.17 src/libexec/ld.elf_so/arch/alpha/rtld_start.S
cvs rdiff -u -r1.15 -r1.16 src/sys/arch/alpha/include/elf_machdep.h
Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.
From: Taylor R Campbell <riastradh@NetBSD.org>
To: Jason Thorpe <thorpej@me.com>
Cc: Rin Okuyama <rokuyama.rk@gmail.com>,
Jason Thorpe <thorpej@NetBSD.org>,
gnats-bugs@NetBSD.org, netbsd-bugs@NetBSD.org
Subject: Re: port-alpha/57511: ld.elf_so(1) does not support secure PLT for
alpha
Date: Fri, 18 Apr 2025 18:20:18 +0000
> Date: Fri, 18 Apr 2025 09:00:17 -0700
> From: Jason Thorpe <thorpej@me.com>
>
> Looks good to me. I do have concerns about the performance
> implications of large programs that end up with multiple GOTs, but
> that's an issue with the format, not with the change itself.
Thanks! I split it up into two commits to make it easier to pull them
up separately if we want: one to add support to ld.elf_so, the other
to remove our local hack disabling secureplt by default in binutils
(and we could always put -Wl,--no-secureplt in bsd.*.mk in a third
commit, of course).
How is the multiple-GOT issue different from other architectures?
From: "Taylor R Campbell" <riastradh@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc:
Subject: PR/57511 CVS commit: src
Date: Fri, 18 Apr 2025 18:21:17 +0000
Module Name: src
Committed By: riastradh
Date: Fri Apr 18 18:21:17 UTC 2025
Modified Files:
src/doc: HACKS
src/external/gpl3/binutils/dist/bfd: elf64-alpha.c
Log Message:
binutils: Remove local hack to disable secureplt by default on alpha.
ld.elf_so can handle it now, so there's no need to patch it out of
binutils. Users can always use -Wl,--no-secureplt to disable it if
they want.
PR port-alpha/57511: ld.elf_so(1) does not support secure PLT for alpha
PR port-alpha/57717: Alpha linker generates LOAD segments (and
generates warnings about them)
To generate a diff of this commit:
cvs rdiff -u -r1.250 -r1.251 src/doc/HACKS
cvs rdiff -u -r1.12 -r1.13 src/external/gpl3/binutils/dist/bfd/elf64-alpha.c
Please note that diffs are not public domain; they are subject to the
copyright notices on the relevant files.
From: Rin Okuyama <rokuyama.rk@gmail.com>
To: Taylor R Campbell <riastradh@NetBSD.org>, Jason Thorpe <thorpej@me.com>
Cc: gnats-bugs@NetBSD.org, netbsd-bugs@NetBSD.org
Subject: Re: port-alpha/57511: ld.elf_so(1) does not support secure PLT for
alpha
Date: Sun, 20 Apr 2025 15:23:52 +0900
On 2025/04/19 3:20, Taylor R Campbell wrote:
>> Date: Fri, 18 Apr 2025 09:00:17 -0700
>> From: Jason Thorpe <thorpej@me.com>
>>
>> Looks good to me. I do have concerns about the performance
>> implications of large programs that end up with multiple GOTs, but
>> that's an issue with the format, not with the change itself.
>
> Thanks! I split it up into two commits to make it easier to pull them
> up separately if we want: one to add support to ld.elf_so, the other
> to remove our local hack disabling secureplt by default in binutils
> (and we could always put -Wl,--no-secureplt in bsd.*.mk in a third
> commit, of course).
>
> How is the multiple-GOT issue different from other architectures?
Thanks for your works!!
-current as of yesterday works just fine on my AlphaStation DS10:
no regression for full ATF run, and some pkgsrc's successfully build.
rin
>Unformatted:
(Contact us)
$NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026
The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.
Home