Skip to contentSkip to conversation

x86/entry/vdso: clean up the vdso build, vdso updates

PATCH Applied H. Peter Anvin 10 participants 50 messages Dec 16, 2025 – Feb 10

Subsystems (10)
lore
Share & tools
lore

H. Peter Anvin

author maintainer
#0 lore raw

v4 cover x86/entry/vdso: clean up the vdso build, vdso updates

This patchset cleans up the vdso build by building the 32- and 64-bit vdsos in separate directories, moving the vdso2c tool to arch/x86/tools, and by merging common code and especially Makefile rules between the 32- and 64-bit vdsos to the greatest extent possible.

Be more strict about sanitizing and standardizing the vdso build options.

Disable CFI for the vdso until the kernel adds user space IBT support.

Modernize the DWARF generation vdso32/sigreturn.S.

Add macros to cleanly call system calls from vdso source code.

Add the GNU_PROPERTY and GNU_STACK PHDRs to the vdso.

When using int $0x80 (old 32-bit hardware or FRED-capable 64-bit hardware) skip the stack stuff in the 32-bit kernel entry code and call int $0x80 directly when used from C code.

Changes from v3 to v4:

- Improved description of patch 01/10.
- Split out the move of vdso2c to tools into a separate patch.
- Remove unrelated formatting changes from vdso-layout.lds.S.
- Fix *-x32.o being missing from "targets".
- Rebased onto v6.19-rc1.

Changes from v2 to v3:

In arch/x86/include/asm/vdso/sys_call.h:
- remove stray comment from prototyping (Uros Biszjak)
- remove VDSO_SYSCALL6() since it would require special handling on 32 bits with frame pointers and is currently unused. (Uros Biszjak)
- indent nested preprocessor directives.

Changes from v1 to v2:

Too many to count - much of the patchset has been reworked

Patches:

x86/entry/vdso: rename vdso_image_* to vdso*_image x86/entry/vdso: move vdso2c to arch/x86/tools
x86/entry/vdso: refactor the vdso build
x86/entry/vdso32: don't rely on int80_landing_pad for adjusting ip x86/entry/vdso32: remove SYSCALL_ENTER_KERNEL macro in sigreturn.S x86/entry/vdso32: remove open-coded DWARF in sigreturn.S x86/entry/vdso: include GNU_PROPERTY and GNU_STACK PHDRs x86/vdso: abstract out vdso system call internals x86/cpufeature: replace X86_FEATURE_SYSENTER32 with X86_FEATURE_SYSFAST32 x86/entry/vdso32: when using int $0x80, use it directly

---
 arch/x86/Kconfig.cpufeatures                       |   8 +
 arch/x86/Makefile                                  |   2 +-
 arch/x86/entry/syscall_32.c                        |   2 +-
 arch/x86/entry/vdso/.gitignore                     |  11 +-
 arch/x86/entry/vdso/Makefile                       | 162 +--------------------
 arch/x86/entry/vdso/common/Makefile.include        |  89 +++++++++++
 arch/x86/entry/vdso/{vdso-note.S => common/note.S} |   5 +-
 arch/x86/entry/vdso/{ => common}/vclock_gettime.c  |   0
 arch/x86/entry/vdso/{ => common}/vdso-layout.lds.S |  38 +++--
 arch/x86/entry/vdso/{ => common}/vgetcpu.c         |   0
 arch/x86/entry/vdso/vdso32/Makefile                |  24 +++
 arch/x86/entry/vdso/vdso32/note.S                  |  19 +--
 arch/x86/entry/vdso/vdso32/sigreturn.S             | 152 +++++--------------
 arch/x86/entry/vdso/vdso32/system_call.S           |  22 ++-
 arch/x86/entry/vdso/vdso32/vclock_gettime.c        |   5 +-
 arch/x86/entry/vdso/vdso32/vdso32.lds.S            |   2 +-
 arch/x86/entry/vdso/vdso32/vgetcpu.c               |   4 +-
 arch/x86/entry/vdso/vdso64/Makefile                |  46 ++++++
 arch/x86/entry/vdso/vdso64/note.S                  |   1 +
 arch/x86/entry/vdso/vdso64/vclock_gettime.c        |   1 +
 .../entry/vdso/{vdso.lds.S => vdso64/vdso64.lds.S} |   2 +-
 arch/x86/entry/vdso/{ => vdso64}/vdsox32.lds.S     |   2 +-
 arch/x86/entry/vdso/vdso64/vgetcpu.c               |   1 +
 .../entry/vdso/{ => vdso64}/vgetrandom-chacha.S    |   0
 arch/x86/entry/vdso/{ => vdso64}/vgetrandom.c      |   2 +-
 arch/x86/entry/vdso/{ => vdso64}/vsgx.S            |   0
 arch/x86/entry/vdso/vma.c                          |  24 ++-
 arch/x86/include/asm/cpufeatures.h                 |   2 +-
 arch/x86/include/asm/dwarf2.h                      |   1 +
 arch/x86/include/asm/elf.h                         |   2 +-
 arch/x86/include/asm/vdso.h                        |   6 +-
 arch/x86/include/asm/vdso/gettimeofday.h           | 108 +-------------
 arch/x86/include/asm/vdso/sys_call.h               | 105 +++++++++++++
 arch/x86/kernel/asm-offsets.c                      |   6 +
 arch/x86/kernel/cpu/centaur.c                      |   3 -
 arch/x86/kernel/cpu/common.c                       |   8 +
 arch/x86/kernel/cpu/intel.c                        |   4 +-
 arch/x86/kernel/cpu/zhaoxin.c                      |   4 +-
 arch/x86/kernel/fred.c                             |   2 +-
 arch/x86/kernel/process_64.c                       |   6 +-
 arch/x86/kernel/signal_32.c                        |   4 +-
 arch/x86/tools/Makefile                            |  15 +-
 arch/x86/{entry/vdso => tools}/vdso2c.c            |   0
 arch/x86/{entry/vdso => tools}/vdso2c.h            |   0
 arch/x86/xen/setup.c                               |  28 ++--
 arch/x86/xen/smp_pv.c                              |   5 +-
 arch/x86/xen/xen-ops.h                             |   1 -
 47 files changed, 444 insertions(+), 490 deletions(-)
author maintainer
#1 lore raw

PATCH v4 1/10 x86/entry/vdso: rename vdso_image_* to vdso*_image

The vdso .so files are named vdso*.so. These structures are binary images and descriptions of these files, so it is more consistent for them to have a naming that more directly mirrors the filenames.

It is also very slightly more compact (by one character...) and simplifies the Makefile just a little bit.

---
 arch/x86/entry/syscall_32.c    |  2 +-
 arch/x86/entry/vdso/.gitignore | 11 ++++-------
 arch/x86/entry/vdso/Makefile   |  8 ++++----
 arch/x86/entry/vdso/vma.c      | 10 +++++-----
 arch/x86/include/asm/elf.h     |  2 +-
 arch/x86/include/asm/vdso.h    |  6 +++---
 arch/x86/kernel/process_64.c   |  6 +++---
 arch/x86/kernel/signal_32.c    |  4 ++--
 8 files changed, 23 insertions(+), 26 deletions(-)
arch/x86/entry/syscall_32.c +1 -1
index a67a644d0cfe..8e829575e12f 100644
@@ -319,7 +319,7 @@ __visible noinstr bool do_fast_syscall_32(struct pt_regs *regs)
 	 * convention.  Adjust regs so it looks like we entered using int80.
 	 */
 	unsigned long landing_pad = (unsigned long)current->mm->context.vdso +
-					vdso_image_32.sym_int80_landing_pad;
+					vdso32_image.sym_int80_landing_pad;
 
 	/*
 	 * SYSENTER loses EIP, and even SYSCALL32 needs us to skip forward
arch/x86/entry/vdso/.gitignore +4 -7
index 37a6129d597b..eb60859dbcbf 100644
@@ -1,8 +1,5 @@
 # SPDX-License-Identifier: GPL-2.0-only
-vdso.lds
-vdsox32.lds
-vdso32-syscall-syms.lds
-vdso32-sysenter-syms.lds
-vdso32-int80-syms.lds
-vdso-image-*.c
-vdso2c
+*.lds
+*.so
+*.so.dbg
+vdso*-image.c
arch/x86/entry/vdso/Makefile +4 -4
index f247f5f5cb44..7f833026d5b2 100644
@@ -16,9 +16,9 @@ vobjs-$(CONFIG_X86_SGX)	+= vsgx.o
 obj-y						+= vma.o extable.o
 
 # vDSO images to build:
-obj-$(CONFIG_X86_64)				+= vdso-image-64.o
-obj-$(CONFIG_X86_X32_ABI)			+= vdso-image-x32.o
-obj-$(CONFIG_COMPAT_32)				+= vdso-image-32.o vdso32-setup.o
+obj-$(CONFIG_X86_64)				+= vdso64-image.o
+obj-$(CONFIG_X86_X32_ABI)			+= vdsox32-image.o
+obj-$(CONFIG_COMPAT_32)				+= vdso32-image.o vdso32-setup.o
 
 vobjs := $(addprefix $(obj)/, $(vobjs-y))
 vobjs32 := $(addprefix $(obj)/, $(vobjs32-y))
@@ -44,7 +44,7 @@ hostprogs += vdso2c
 quiet_cmd_vdso2c = VDSO2C  $@
       cmd_vdso2c = $(obj)/vdso2c $< $(<:%.dbg=%) $@
 
-$(obj)/vdso-image-%.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(obj)/vdso2c FORCE
+$(obj)/vdso%-image.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(obj)/vdso2c FORCE
 	$(call if_changed,vdso2c)
 
 #
arch/x86/entry/vdso/vma.c +5 -5
index afe105b2f907..8f98c2d7c7a9 100644
@@ -65,7 +65,7 @@ static vm_fault_t vdso_fault(const struct vm_special_mapping *sm,
 static void vdso_fix_landing(const struct vdso_image *image,
 		struct vm_area_struct *new_vma)
 {
-	if (in_ia32_syscall() && image == &vdso_image_32) {
+	if (in_ia32_syscall() && image == &vdso32_image) {
 		struct pt_regs *regs = current_pt_regs();
 		unsigned long vdso_land = image->sym_int80_landing_pad;
 		unsigned long old_land_addr = vdso_land +
@@ -230,7 +230,7 @@ static int load_vdso32(void)
 	if (vdso32_enabled != 1)  /* Other values all mean "disabled" */
 		return 0;
 
-	return map_vdso(&vdso_image_32, 0);
+	return map_vdso(&vdso32_image, 0);
 }
 
 int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
@@ -239,7 +239,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 		if (!vdso64_enabled)
 			return 0;
 
-		return map_vdso(&vdso_image_64, 0);
+		return map_vdso(&vdso64_image, 0);
 	}
 
 	return load_vdso32();
@@ -252,7 +252,7 @@ int compat_arch_setup_additional_pages(struct linux_binprm *bprm,
 	if (IS_ENABLED(CONFIG_X86_X32_ABI) && x32) {
 		if (!vdso64_enabled)
 			return 0;
-		return map_vdso(&vdso_image_x32, 0);
+		return map_vdso(&vdsox32_image, 0);
 	}
 
 	if (IS_ENABLED(CONFIG_IA32_EMULATION))
@@ -267,7 +267,7 @@ bool arch_syscall_is_vdso_sigreturn(struct pt_regs *regs)
 	const struct vdso_image *image = current->mm->context.vdso_image;
 	unsigned long vdso = (unsigned long) current->mm->context.vdso;
 
-	if (in_ia32_syscall() && image == &vdso_image_32) {
+	if (in_ia32_syscall() && image == &vdso32_image) {
 		if (regs->ip == vdso + image->sym_vdso32_sigreturn_landing_pad ||
 		    regs->ip == vdso + image->sym_vdso32_rt_sigreturn_landing_pad)
 			return true;
arch/x86/include/asm/elf.h +1 -1
index 6c8fdc96be7e..2ba5f166e58f 100644
@@ -361,7 +361,7 @@ else if (IS_ENABLED(CONFIG_IA32_EMULATION))				\
 
 #define VDSO_ENTRY							\
 	((unsigned long)current->mm->context.vdso +			\
-	 vdso_image_32.sym___kernel_vsyscall)
+	 vdso32_image.sym___kernel_vsyscall)
 
 struct linux_binprm;
 
arch/x86/include/asm/vdso.h +3 -3
index b7253ef3205a..e8afbe9faa5b 100644
@@ -27,9 +27,9 @@ struct vdso_image {
 	long sym_vdso32_rt_sigreturn_landing_pad;
 };
 
-extern const struct vdso_image vdso_image_64;
-extern const struct vdso_image vdso_image_x32;
-extern const struct vdso_image vdso_image_32;
+extern const struct vdso_image vdso64_image;
+extern const struct vdso_image vdsox32_image;
+extern const struct vdso_image vdso32_image;
 
 extern int __init init_vdso_image(const struct vdso_image *image);
 
arch/x86/kernel/process_64.c +3 -3
index 432c0a004c60..08e72f429870 100644
@@ -941,14 +941,14 @@ long do_arch_prctl_64(struct task_struct *task, int option, unsigned long arg2)
 #ifdef CONFIG_CHECKPOINT_RESTORE
 # ifdef CONFIG_X86_X32_ABI
 	case ARCH_MAP_VDSO_X32:
-		return prctl_map_vdso(&vdso_image_x32, arg2);
+		return prctl_map_vdso(&vdsox32_image, arg2);
 # endif
 # ifdef CONFIG_IA32_EMULATION
 	case ARCH_MAP_VDSO_32:
-		return prctl_map_vdso(&vdso_image_32, arg2);
+		return prctl_map_vdso(&vdso32_image, arg2);
 # endif
 	case ARCH_MAP_VDSO_64:
-		return prctl_map_vdso(&vdso_image_64, arg2);
+		return prctl_map_vdso(&vdso64_image, arg2);
 #endif
 #ifdef CONFIG_ADDRESS_MASKING
 	case ARCH_GET_UNTAG_MASK:
arch/x86/kernel/signal_32.c +2 -2
index 42bbc42bd350..e55cf19e68fe 100644
@@ -282,7 +282,7 @@ int ia32_setup_frame(struct ksignal *ksig, struct pt_regs *regs)
 		/* Return stub is in 32bit vsyscall page */
 		if (current->mm->context.vdso)
 			restorer = current->mm->context.vdso +
-				vdso_image_32.sym___kernel_sigreturn;
+				vdso32_image.sym___kernel_sigreturn;
 		else
 			restorer = &frame->retcode;
 	}
@@ -368,7 +368,7 @@ int ia32_setup_rt_frame(struct ksignal *ksig, struct pt_regs *regs)
 		restorer = ksig->ka.sa.sa_restorer;
 	else
 		restorer = current->mm->context.vdso +
-			vdso_image_32.sym___kernel_rt_sigreturn;
+			vdso32_image.sym___kernel_rt_sigreturn;
 	unsafe_put_user(ptr_to_compat(restorer), &frame->pretcode, Efault);
 
 	/*
Signature · 1 line
2.52.0
merge bot Applied
#2 lore raw

[tip: x86/entry] x86/entry/vdso: Rename vdso_image_* to vdso*_image

The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     93d73005bff4f600696ce30e366e742c3373b13d
Gitweb:        https://git.kernel.org/tip/93d73005bff4f600696ce30e366e742c3373b13d
Author:        H. Peter Anvin <[email protected]>
AuthorDate:    Tue, 16 Dec 2025 13:25:55 -08:00
Committer:     Dave Hansen <[email protected]>
CommitterDate: Tue, 13 Jan 2026 15:33:20 -08:00

x86/entry/vdso: Rename vdso_image_* to vdso*_image

The vdso .so files are named vdso*.so. These structures are binary images and descriptions of these files, so it is more consistent for them to have a naming that more directly mirrors the filenames.

It is also very slightly more compact (by one character...) and simplifies the Makefile just a little bit.

---
 arch/x86/entry/syscall_32.c    |  2 +-
 arch/x86/entry/vdso/.gitignore | 11 ++++-------
 arch/x86/entry/vdso/Makefile   |  8 ++++----
 arch/x86/entry/vdso/vma.c      | 10 +++++-----
 arch/x86/include/asm/elf.h     |  2 +-
 arch/x86/include/asm/vdso.h    |  6 +++---
 arch/x86/kernel/process_64.c   |  6 +++---
 arch/x86/kernel/signal_32.c    |  4 ++--
 8 files changed, 23 insertions(+), 26 deletions(-)
arch/x86/entry/syscall_32.c +1 -1
index a67a644..8e82957 100644
@@ -319,7 +319,7 @@ __visible noinstr bool do_fast_syscall_32(struct pt_regs *regs)
 	 * convention.  Adjust regs so it looks like we entered using int80.
 	 */
 	unsigned long landing_pad = (unsigned long)current->mm->context.vdso +
-					vdso_image_32.sym_int80_landing_pad;
+					vdso32_image.sym_int80_landing_pad;
 
 	/*
 	 * SYSENTER loses EIP, and even SYSCALL32 needs us to skip forward
arch/x86/entry/vdso/.gitignore +4 -7
index 37a6129..eb60859 100644
@@ -1,8 +1,5 @@
 # SPDX-License-Identifier: GPL-2.0-only
-vdso.lds
-vdsox32.lds
-vdso32-syscall-syms.lds
-vdso32-sysenter-syms.lds
-vdso32-int80-syms.lds
-vdso-image-*.c
-vdso2c
+*.lds
+*.so
+*.so.dbg
+vdso*-image.c
arch/x86/entry/vdso/Makefile +4 -4
index f247f5f..7f83302 100644
@@ -16,9 +16,9 @@ vobjs-$(CONFIG_X86_SGX)	+= vsgx.o
 obj-y						+= vma.o extable.o
 
 # vDSO images to build:
-obj-$(CONFIG_X86_64)				+= vdso-image-64.o
-obj-$(CONFIG_X86_X32_ABI)			+= vdso-image-x32.o
-obj-$(CONFIG_COMPAT_32)				+= vdso-image-32.o vdso32-setup.o
+obj-$(CONFIG_X86_64)				+= vdso64-image.o
+obj-$(CONFIG_X86_X32_ABI)			+= vdsox32-image.o
+obj-$(CONFIG_COMPAT_32)				+= vdso32-image.o vdso32-setup.o
 
 vobjs := $(addprefix $(obj)/, $(vobjs-y))
 vobjs32 := $(addprefix $(obj)/, $(vobjs32-y))
@@ -44,7 +44,7 @@ hostprogs += vdso2c
 quiet_cmd_vdso2c = VDSO2C  $@
       cmd_vdso2c = $(obj)/vdso2c $< $(<:%.dbg=%) $@
 
-$(obj)/vdso-image-%.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(obj)/vdso2c FORCE
+$(obj)/vdso%-image.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(obj)/vdso2c FORCE
 	$(call if_changed,vdso2c)
 
 #
arch/x86/entry/vdso/vma.c +5 -5
index afe105b..8f98c2d 100644
@@ -65,7 +65,7 @@ static vm_fault_t vdso_fault(const struct vm_special_mapping *sm,
 static void vdso_fix_landing(const struct vdso_image *image,
 		struct vm_area_struct *new_vma)
 {
-	if (in_ia32_syscall() && image == &vdso_image_32) {
+	if (in_ia32_syscall() && image == &vdso32_image) {
 		struct pt_regs *regs = current_pt_regs();
 		unsigned long vdso_land = image->sym_int80_landing_pad;
 		unsigned long old_land_addr = vdso_land +
@@ -230,7 +230,7 @@ static int load_vdso32(void)
 	if (vdso32_enabled != 1)  /* Other values all mean "disabled" */
 		return 0;
 
-	return map_vdso(&vdso_image_32, 0);
+	return map_vdso(&vdso32_image, 0);
 }
 
 int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
@@ -239,7 +239,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 		if (!vdso64_enabled)
 			return 0;
 
-		return map_vdso(&vdso_image_64, 0);
+		return map_vdso(&vdso64_image, 0);
 	}
 
 	return load_vdso32();
@@ -252,7 +252,7 @@ int compat_arch_setup_additional_pages(struct linux_binprm *bprm,
 	if (IS_ENABLED(CONFIG_X86_X32_ABI) && x32) {
 		if (!vdso64_enabled)
 			return 0;
-		return map_vdso(&vdso_image_x32, 0);
+		return map_vdso(&vdsox32_image, 0);
 	}
 
 	if (IS_ENABLED(CONFIG_IA32_EMULATION))
@@ -267,7 +267,7 @@ bool arch_syscall_is_vdso_sigreturn(struct pt_regs *regs)
 	const struct vdso_image *image = current->mm->context.vdso_image;
 	unsigned long vdso = (unsigned long) current->mm->context.vdso;
 
-	if (in_ia32_syscall() && image == &vdso_image_32) {
+	if (in_ia32_syscall() && image == &vdso32_image) {
 		if (regs->ip == vdso + image->sym_vdso32_sigreturn_landing_pad ||
 		    regs->ip == vdso + image->sym_vdso32_rt_sigreturn_landing_pad)
 			return true;
arch/x86/include/asm/elf.h +1 -1
index 6c8fdc9..2ba5f16 100644
@@ -361,7 +361,7 @@ else if (IS_ENABLED(CONFIG_IA32_EMULATION))				\
 
 #define VDSO_ENTRY							\
 	((unsigned long)current->mm->context.vdso +			\
-	 vdso_image_32.sym___kernel_vsyscall)
+	 vdso32_image.sym___kernel_vsyscall)
 
 struct linux_binprm;
 
arch/x86/include/asm/vdso.h +3 -3
index b7253ef..e8afbe9 100644
@@ -27,9 +27,9 @@ struct vdso_image {
 	long sym_vdso32_rt_sigreturn_landing_pad;
 };
 
-extern const struct vdso_image vdso_image_64;
-extern const struct vdso_image vdso_image_x32;
-extern const struct vdso_image vdso_image_32;
+extern const struct vdso_image vdso64_image;
+extern const struct vdso_image vdsox32_image;
+extern const struct vdso_image vdso32_image;
 
 extern int __init init_vdso_image(const struct vdso_image *image);
 
arch/x86/kernel/process_64.c +3 -3
index 432c0a0..08e72f4 100644
@@ -941,14 +941,14 @@ long do_arch_prctl_64(struct task_struct *task, int option, unsigned long arg2)
 #ifdef CONFIG_CHECKPOINT_RESTORE
 # ifdef CONFIG_X86_X32_ABI
 	case ARCH_MAP_VDSO_X32:
-		return prctl_map_vdso(&vdso_image_x32, arg2);
+		return prctl_map_vdso(&vdsox32_image, arg2);
 # endif
 # ifdef CONFIG_IA32_EMULATION
 	case ARCH_MAP_VDSO_32:
-		return prctl_map_vdso(&vdso_image_32, arg2);
+		return prctl_map_vdso(&vdso32_image, arg2);
 # endif
 	case ARCH_MAP_VDSO_64:
-		return prctl_map_vdso(&vdso_image_64, arg2);
+		return prctl_map_vdso(&vdso64_image, arg2);
 #endif
 #ifdef CONFIG_ADDRESS_MASKING
 	case ARCH_GET_UNTAG_MASK:
arch/x86/kernel/signal_32.c +2 -2
index 42bbc42..e55cf19 100644
@@ -282,7 +282,7 @@ int ia32_setup_frame(struct ksignal *ksig, struct pt_regs *regs)
 		/* Return stub is in 32bit vsyscall page */
 		if (current->mm->context.vdso)
 			restorer = current->mm->context.vdso +
-				vdso_image_32.sym___kernel_sigreturn;
+				vdso32_image.sym___kernel_sigreturn;
 		else
 			restorer = &frame->retcode;
 	}
@@ -368,7 +368,7 @@ int ia32_setup_rt_frame(struct ksignal *ksig, struct pt_regs *regs)
 		restorer = ksig->ka.sa.sa_restorer;
 	else
 		restorer = current->mm->context.vdso +
-			vdso_image_32.sym___kernel_rt_sigreturn;
+			vdso32_image.sym___kernel_rt_sigreturn;
 	unsafe_put_user(ptr_to_compat(restorer), &frame->pretcode, Efault);
 
 	/*
author maintainer
#3 lore raw

PATCH v4 2/10 x86/entry/vdso: move vdso2c to arch/x86/tools

In reply to H. Peter Anvin

It is generally better to build tools in arch/x86/tools to keep host cflags proliferation down, and to reduce makefile sequencing issues. Move the vdso build tool vdso2c into arch/x86/tools in preparation for refactoring the vdso makefiles.

---
 arch/x86/Makefile                       |  2 +-
 arch/x86/entry/vdso/Makefile            |  7 +++----
 arch/x86/tools/Makefile                 | 15 ++++++++++-----
 arch/x86/{entry/vdso => tools}/vdso2c.c |  0
 arch/x86/{entry/vdso => tools}/vdso2c.h |  0
 5 files changed, 14 insertions(+), 10 deletions(-)
 rename arch/x86/{entry/vdso => tools}/vdso2c.c (100%)
 rename arch/x86/{entry/vdso => tools}/vdso2c.h (100%)
arch/x86/Makefile +1 -1
index 1d403a3612ea..9ab7522ced18 100644
@@ -252,7 +252,7 @@ endif
 
 
 archscripts: scripts_basic
-	$(Q)$(MAKE) $(build)=arch/x86/tools relocs
+	$(Q)$(MAKE) $(build)=arch/x86/tools relocs vdso2c
 
 ###
 # Syscall table generation
arch/x86/entry/vdso/Makefile +3 -4
index 7f833026d5b2..3d9b09f00c70 100644
@@ -38,13 +38,12 @@ VDSO_LDFLAGS_vdso.lds = -m elf_x86_64 -soname linux-vdso.so.1 \
 $(obj)/vdso64.so.dbg: $(obj)/vdso.lds $(vobjs) FORCE
 	$(call if_changed,vdso_and_check)
 
-HOST_EXTRACFLAGS += -I$(srctree)/tools/include -I$(srctree)/include/uapi -I$(srctree)/arch/$(SUBARCH)/include/uapi
-hostprogs += vdso2c
+VDSO2C = $(objtree)/arch/x86/tools/vdso2c
 
 quiet_cmd_vdso2c = VDSO2C  $@
-      cmd_vdso2c = $(obj)/vdso2c $< $(<:%.dbg=%) $@
+      cmd_vdso2c = $(VDSO2C) $< $(<:%.dbg=%) $@
 
-$(obj)/vdso%-image.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(obj)/vdso2c FORCE
+$(obj)/vdso%-image.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(VDSO2C) FORCE
 	$(call if_changed,vdso2c)
 
 #
arch/x86/tools/Makefile +10 -5
index 7278e2545c35..39a183fffd04 100644
@@ -38,9 +38,14 @@ $(obj)/insn_decoder_test.o: $(srctree)/tools/arch/x86/lib/insn.c $(srctree)/tool
 
 $(obj)/insn_sanity.o: $(srctree)/tools/arch/x86/lib/insn.c $(srctree)/tools/arch/x86/lib/inat.c $(srctree)/tools/arch/x86/include/asm/inat_types.h $(srctree)/tools/arch/x86/include/asm/inat.h $(srctree)/tools/arch/x86/include/asm/insn.h $(objtree)/arch/x86/lib/inat-tables.c
 
-HOST_EXTRACFLAGS += -I$(srctree)/tools/include
-hostprogs	+= relocs
-relocs-objs     := relocs_32.o relocs_64.o relocs_common.o
-PHONY += relocs
-relocs: $(obj)/relocs
+HOST_EXTRACFLAGS += -I$(srctree)/tools/include -I$(srctree)/include/uapi \
+		    -I$(srctree)/arch/$(SUBARCH)/include/uapi
+
+hostprogs	+= relocs vdso2c
+relocs-objs	:= relocs_32.o relocs_64.o relocs_common.o
+
+always-y	:= $(hostprogs)
+
+PHONY += $(hostprogs)
+$(hostprogs): %: $(obj)/%
 	@:
arch/x86/tools/vdso2c.c renamed from arch/x86/entry/vdso/vdso2c.c +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vdso2c.c
rename to arch/x86/tools/vdso2c.c
arch/x86/tools/vdso2c.h renamed from arch/x86/entry/vdso/vdso2c.h +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vdso2c.h
rename to arch/x86/tools/vdso2c.h
Signature · 1 line
2.52.0
merge bot Applied
#4 lore raw

[tip: x86/entry] x86/entry/vdso: Move vdso2c to arch/x86/tools

The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     a76108d05ee13cddb72b620752a80b2c3e87aee1
Gitweb:        https://git.kernel.org/tip/a76108d05ee13cddb72b620752a80b2c3e87aee1
Author:        H. Peter Anvin <[email protected]>
AuthorDate:    Tue, 16 Dec 2025 13:25:56 -08:00
Committer:     Dave Hansen <[email protected]>
CommitterDate: Tue, 13 Jan 2026 15:33:20 -08:00

x86/entry/vdso: Move vdso2c to arch/x86/tools

It is generally better to build tools in arch/x86/tools to keep host cflags proliferation down, and to reduce makefile sequencing issues. Move the vdso build tool vdso2c into arch/x86/tools in preparation for refactoring the vdso makefiles.

---
 arch/x86/Makefile            |   2 +-
 arch/x86/entry/vdso/Makefile |   7 +-
 arch/x86/entry/vdso/vdso2c.c | 233 +----------------------------------
 arch/x86/entry/vdso/vdso2c.h | 208 +------------------------------
 arch/x86/tools/Makefile      |  15 +-
 arch/x86/tools/vdso2c.c      | 233 ++++++++++++++++++++++++++++++++++-
 arch/x86/tools/vdso2c.h      | 208 ++++++++++++++++++++++++++++++-
 7 files changed, 455 insertions(+), 451 deletions(-)
 delete mode 100644 arch/x86/entry/vdso/vdso2c.c
 delete mode 100644 arch/x86/entry/vdso/vdso2c.h
 create mode 100644 arch/x86/tools/vdso2c.c
 create mode 100644 arch/x86/tools/vdso2c.h
arch/x86/Makefile +1 -1
index 1d403a3..9ab7522 100644
@@ -252,7 +252,7 @@ endif
 
 
 archscripts: scripts_basic
-	$(Q)$(MAKE) $(build)=arch/x86/tools relocs
+	$(Q)$(MAKE) $(build)=arch/x86/tools relocs vdso2c
 
 ###
 # Syscall table generation
arch/x86/entry/vdso/Makefile +3 -4
index 7f83302..3d9b09f 100644
@@ -38,13 +38,12 @@ VDSO_LDFLAGS_vdso.lds = -m elf_x86_64 -soname linux-vdso.so.1 \
 $(obj)/vdso64.so.dbg: $(obj)/vdso.lds $(vobjs) FORCE
 	$(call if_changed,vdso_and_check)
 
-HOST_EXTRACFLAGS += -I$(srctree)/tools/include -I$(srctree)/include/uapi -I$(srctree)/arch/$(SUBARCH)/include/uapi
-hostprogs += vdso2c
+VDSO2C = $(objtree)/arch/x86/tools/vdso2c
 
 quiet_cmd_vdso2c = VDSO2C  $@
-      cmd_vdso2c = $(obj)/vdso2c $< $(<:%.dbg=%) $@
+      cmd_vdso2c = $(VDSO2C) $< $(<:%.dbg=%) $@
 
-$(obj)/vdso%-image.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(obj)/vdso2c FORCE
+$(obj)/vdso%-image.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(VDSO2C) FORCE
 	$(call if_changed,vdso2c)
 
 #
arch/x86/entry/vdso/vdso2c.c deleted +0 -233

233 lines not on this page · raw message

arch/x86/entry/vdso/vdso2c.h deleted +0 -208

208 lines not on this page · raw message

arch/x86/tools/Makefile +10 -5
index 7278e25..39a183f 100644
@@ -38,9 +38,14 @@ $(obj)/insn_decoder_test.o: $(srctree)/tools/arch/x86/lib/insn.c $(srctree)/tool
 
 $(obj)/insn_sanity.o: $(srctree)/tools/arch/x86/lib/insn.c $(srctree)/tools/arch/x86/lib/inat.c $(srctree)/tools/arch/x86/include/asm/inat_types.h $(srctree)/tools/arch/x86/include/asm/inat.h $(srctree)/tools/arch/x86/include/asm/insn.h $(objtree)/arch/x86/lib/inat-tables.c
 
-HOST_EXTRACFLAGS += -I$(srctree)/tools/include
-hostprogs	+= relocs
-relocs-objs     := relocs_32.o relocs_64.o relocs_common.o
-PHONY += relocs
-relocs: $(obj)/relocs
+HOST_EXTRACFLAGS += -I$(srctree)/tools/include -I$(srctree)/include/uapi \
+		    -I$(srctree)/arch/$(SUBARCH)/include/uapi
+
+hostprogs	+= relocs vdso2c
+relocs-objs	:= relocs_32.o relocs_64.o relocs_common.o
+
+always-y	:= $(hostprogs)
+
+PHONY += $(hostprogs)
+$(hostprogs): %: $(obj)/%
 	@:
arch/x86/tools/vdso2c.c new file +233 -0

233 lines not on this page · raw message

arch/x86/tools/vdso2c.h new file +208 -0

208 lines not on this page · raw message

maintainer
#5 lore raw

PATCH 1/1 x86/entry/vdso: Add vdso2c to .gitignore

From: "Borislav Petkov (AMD)" <[email protected]>
Date: Tue, 27 Jan 2026 23:09:13 +0100

The commit

a76108d05ee1 ("x86/entry/vdso: Move vdso2c to arch/x86/tools")

moved vdso2c to arch/x86/tools/ and commit

93d73005bff4 ("x86/entry/vdso: Rename vdso_image_* to vdso*_image")

renamed .so files but also dropped vdso2c from
arch/x86/entry/vdso/.gitignore.

It should've moved it to arch/x86/tools/.gitignore instead.

Do that.

---
 arch/x86/tools/.gitignore | 1 +
 1 file changed, 1 insertion(+)
arch/x86/tools/.gitignore +1 -0
index d36dc7cf9115..51d5c22b38d7 100644
@@ -1,2 +1,3 @@
 # SPDX-License-Identifier: GPL-2.0-only
 relocs
+vdso2c
Signature · 8 lines
2.51.0


-- 
Regards/Gruss,
    Boris.

https://people.kernel.org/tglx/notes-about-netiquette
#6 lore raw

[tip: x86/entry] x86/entry/vdso: Add vdso2c to .gitignore

The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     ce9b1c10c3f1c723c3cc7b63aa8331fdb6c57a04
Gitweb:        https://git.kernel.org/tip/ce9b1c10c3f1c723c3cc7b63aa8331fdb6c57a04
Author:        Borislav Petkov (AMD) <[email protected]>
AuthorDate:    Tue, 27 Jan 2026 23:09:13 +01:00
Committer:     Borislav Petkov (AMD) <[email protected]>
CommitterDate: Tue, 27 Jan 2026 23:27:51 +01:00

x86/entry/vdso: Add vdso2c to .gitignore

The commit

a76108d05ee1 ("x86/entry/vdso: Move vdso2c to arch/x86/tools")

moved vdso2c to arch/x86/tools/ and commit

93d73005bff4 ("x86/entry/vdso: Rename vdso_image_* to vdso*_image")

renamed .so files but also dropped vdso2c from
arch/x86/entry/vdso/.gitignore.

It should've moved it to arch/x86/tools/.gitignore instead.

Do that.

---
 arch/x86/tools/.gitignore | 1 +
 1 file changed, 1 insertion(+)
arch/x86/tools/.gitignore +1 -0
index d36dc7c..51d5c22 100644
@@ -1,2 +1,3 @@
 # SPDX-License-Identifier: GPL-2.0-only
 relocs
+vdso2c
author maintainer
#7 lore raw

PATCH v4 3/10 x86/entry/vdso: refactor the vdso build

In reply to H. Peter Anvin

- Separate out the vdso sources into common, vdso32, and vdso64 directories.
- Build the 32- and 64-bit vdsos in their respective subdirectories; this greatly simplifies the build flags handling.
- Unify the mangling of Makefile flags between the 32- and 64-bit vdso code as much as possible; all common rules are put in arch/x86/entry/vdso/common/Makefile.include. The remaining is very simple for 32 bits; the 64-bit one is only slightly more complicated because it contains the x32 generation rule.
- Define __DISABLE_EXPORTS when building the vdso. This need seems to have been masked by different ordering compile flags before.
- Change CONFIG_X86_64 to BUILD_VDSO32_64 in vdso32/system_call.S, to make it compatible with including fake_32bit_build.h.
- The -fcf-protection= option was "leaking" from the kernel build, for reasons that was not clear to me. Futhermore, several distributions ship with it set to a default value other than "-fcf-protection=none". Make it match the configuration options for *user space*.

Note that this patch may seem large, but the vast majority of it is simply code movement.

---
 arch/x86/entry/vdso/Makefile                  | 161 +-----------------
 arch/x86/entry/vdso/common/Makefile.include   |  89 ++++++++++
 .../entry/vdso/{vdso-note.S => common/note.S} |   5 +-
 .../entry/vdso/{ => common}/vclock_gettime.c  |   0
 .../entry/vdso/{ => common}/vdso-layout.lds.S |   0
 arch/x86/entry/vdso/{ => common}/vgetcpu.c    |   0
 arch/x86/entry/vdso/vdso32/Makefile           |  24 +++
 arch/x86/entry/vdso/vdso32/note.S             |  19 +--
 arch/x86/entry/vdso/vdso32/system_call.S      |   2 +-
 arch/x86/entry/vdso/vdso32/vclock_gettime.c   |   5 +-
 arch/x86/entry/vdso/vdso32/vdso32.lds.S       |   2 +-
 arch/x86/entry/vdso/vdso32/vgetcpu.c          |   4 +-
 arch/x86/entry/vdso/vdso64/Makefile           |  46 +++++
 arch/x86/entry/vdso/vdso64/note.S             |   1 +
 arch/x86/entry/vdso/vdso64/vclock_gettime.c   |   1 +
 .../vdso/{vdso.lds.S => vdso64/vdso64.lds.S}  |   2 +-
 .../x86/entry/vdso/{ => vdso64}/vdsox32.lds.S |   2 +-
 arch/x86/entry/vdso/vdso64/vgetcpu.c          |   1 +
 .../vdso/{ => vdso64}/vgetrandom-chacha.S     |   0
 arch/x86/entry/vdso/{ => vdso64}/vgetrandom.c |   2 +-
 arch/x86/entry/vdso/{ => vdso64}/vsgx.S       |   0
 21 files changed, 180 insertions(+), 186 deletions(-)
 create mode 100644 arch/x86/entry/vdso/common/Makefile.include
 rename arch/x86/entry/vdso/{vdso-note.S => common/note.S} (62%)
 rename arch/x86/entry/vdso/{ => common}/vclock_gettime.c (100%)
 rename arch/x86/entry/vdso/{ => common}/vdso-layout.lds.S (100%)
 rename arch/x86/entry/vdso/{ => common}/vgetcpu.c (100%)
 create mode 100644 arch/x86/entry/vdso/vdso32/Makefile
 create mode 100644 arch/x86/entry/vdso/vdso64/Makefile
 create mode 100644 arch/x86/entry/vdso/vdso64/note.S
 create mode 100644 arch/x86/entry/vdso/vdso64/vclock_gettime.c
 rename arch/x86/entry/vdso/{vdso.lds.S => vdso64/vdso64.lds.S} (94%)
 rename arch/x86/entry/vdso/{ => vdso64}/vdsox32.lds.S (92%)
 create mode 100644 arch/x86/entry/vdso/vdso64/vgetcpu.c
 rename arch/x86/entry/vdso/{ => vdso64}/vgetrandom-chacha.S (100%)
 rename arch/x86/entry/vdso/{ => vdso64}/vgetrandom.c (91%)
 rename arch/x86/entry/vdso/{ => vdso64}/vsgx.S (100%)
arch/x86/entry/vdso/Makefile +6 -155
index 3d9b09f00c70..987b43fd4cd3 100644
@@ -3,159 +3,10 @@
 # Building vDSO images for x86.
 #
 
-# Include the generic Makefile to check the built vDSO:
-include $(srctree)/lib/vdso/Makefile.include
+# Regular kernel objects
+obj-y				:= vma.o extable.o
+obj-$(CONFIG_COMPAT_32)		+= vdso32-setup.o
 
-# Files to link into the vDSO:
-vobjs-y := vdso-note.o vclock_gettime.o vgetcpu.o vgetrandom.o vgetrandom-chacha.o
-vobjs32-y := vdso32/note.o vdso32/system_call.o vdso32/sigreturn.o
-vobjs32-y += vdso32/vclock_gettime.o vdso32/vgetcpu.o
-vobjs-$(CONFIG_X86_SGX)	+= vsgx.o
-
-# Files to link into the kernel:
-obj-y						+= vma.o extable.o
-
-# vDSO images to build:
-obj-$(CONFIG_X86_64)				+= vdso64-image.o
-obj-$(CONFIG_X86_X32_ABI)			+= vdsox32-image.o
-obj-$(CONFIG_COMPAT_32)				+= vdso32-image.o vdso32-setup.o
-
-vobjs := $(addprefix $(obj)/, $(vobjs-y))
-vobjs32 := $(addprefix $(obj)/, $(vobjs32-y))
-
-$(obj)/vdso.o: $(obj)/vdso.so
-
-targets += vdso.lds $(vobjs-y)
-targets += vdso32/vdso32.lds $(vobjs32-y)
-
-targets += $(foreach x, 64 x32 32, vdso-image-$(x).c vdso$(x).so vdso$(x).so.dbg)
-
-CPPFLAGS_vdso.lds += -P -C
-
-VDSO_LDFLAGS_vdso.lds = -m elf_x86_64 -soname linux-vdso.so.1 \
-			-z max-page-size=4096
-
-$(obj)/vdso64.so.dbg: $(obj)/vdso.lds $(vobjs) FORCE
-	$(call if_changed,vdso_and_check)
-
-VDSO2C = $(objtree)/arch/x86/tools/vdso2c
-
-quiet_cmd_vdso2c = VDSO2C  $@
-      cmd_vdso2c = $(VDSO2C) $< $(<:%.dbg=%) $@
-
-$(obj)/vdso%-image.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(VDSO2C) FORCE
-	$(call if_changed,vdso2c)
-
-#
-# Don't omit frame pointers for ease of userspace debugging, but do
-# optimize sibling calls.
-#
-CFL := $(PROFILING) -mcmodel=small -fPIC -O2 -fasynchronous-unwind-tables -m64 \
-       $(filter -g%,$(KBUILD_CFLAGS)) -fno-stack-protector \
-       -fno-omit-frame-pointer -foptimize-sibling-calls \
-       -DDISABLE_BRANCH_PROFILING -DBUILD_VDSO
-
-ifdef CONFIG_MITIGATION_RETPOLINE
-ifneq ($(RETPOLINE_VDSO_CFLAGS),)
-  CFL += $(RETPOLINE_VDSO_CFLAGS)
-endif
-endif
-
-$(vobjs): KBUILD_CFLAGS := $(filter-out $(PADDING_CFLAGS) $(CC_FLAGS_LTO) $(CC_FLAGS_CFI) $(RANDSTRUCT_CFLAGS) $(KSTACK_ERASE_CFLAGS) $(GCC_PLUGINS_CFLAGS) $(RETPOLINE_CFLAGS),$(KBUILD_CFLAGS)) $(CFL)
-$(vobjs): KBUILD_AFLAGS += -DBUILD_VDSO
-
-#
-# vDSO code runs in userspace and -pg doesn't help with profiling anyway.
-#
-CFLAGS_REMOVE_vclock_gettime.o = -pg
-CFLAGS_REMOVE_vdso32/vclock_gettime.o = -pg
-CFLAGS_REMOVE_vgetcpu.o = -pg
-CFLAGS_REMOVE_vdso32/vgetcpu.o = -pg
-CFLAGS_REMOVE_vsgx.o = -pg
-CFLAGS_REMOVE_vgetrandom.o = -pg
-
-#
-# X32 processes use x32 vDSO to access 64bit kernel data.
-#
-# Build x32 vDSO image:
-# 1. Compile x32 vDSO as 64bit.
-# 2. Convert object files to x32.
-# 3. Build x32 VDSO image with x32 objects, which contains 64bit codes
-# so that it can reach 64bit address space with 64bit pointers.
-#
-
-CPPFLAGS_vdsox32.lds = $(CPPFLAGS_vdso.lds)
-VDSO_LDFLAGS_vdsox32.lds = -m elf32_x86_64 -soname linux-vdso.so.1 \
-			   -z max-page-size=4096
-
-# x32-rebranded versions
-vobjx32s-y := $(vobjs-y:.o=-x32.o)
-
-# same thing, but in the output directory
-vobjx32s := $(addprefix $(obj)/, $(vobjx32s-y))
-
-# Convert 64bit object file to x32 for x32 vDSO.
-quiet_cmd_x32 = X32     $@
-      cmd_x32 = $(OBJCOPY) -O elf32-x86-64 $< $@
-
-$(obj)/%-x32.o: $(obj)/%.o FORCE
-	$(call if_changed,x32)
-
-targets += vdsox32.lds $(vobjx32s-y)
-
-$(obj)/%.so: OBJCOPYFLAGS := -S --remove-section __ex_table
-$(obj)/%.so: $(obj)/%.so.dbg FORCE
-	$(call if_changed,objcopy)
-
-$(obj)/vdsox32.so.dbg: $(obj)/vdsox32.lds $(vobjx32s) FORCE
-	$(call if_changed,vdso_and_check)
-
-CPPFLAGS_vdso32/vdso32.lds = $(CPPFLAGS_vdso.lds)
-VDSO_LDFLAGS_vdso32.lds = -m elf_i386 -soname linux-gate.so.1
-
-KBUILD_AFLAGS_32 := $(filter-out -m64,$(KBUILD_AFLAGS)) -DBUILD_VDSO
-$(obj)/vdso32.so.dbg: KBUILD_AFLAGS = $(KBUILD_AFLAGS_32)
-$(obj)/vdso32.so.dbg: asflags-$(CONFIG_X86_64) += -m32
-
-KBUILD_CFLAGS_32 := $(filter-out -m64,$(KBUILD_CFLAGS))
-KBUILD_CFLAGS_32 := $(filter-out -mcmodel=kernel,$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out -fno-pic,$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out -mfentry,$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(RANDSTRUCT_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(KSTACK_ERASE_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(GCC_PLUGINS_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(RETPOLINE_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(CC_FLAGS_LTO),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(CC_FLAGS_CFI),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(PADDING_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 += -m32 -msoft-float -mregparm=0 -fpic
-KBUILD_CFLAGS_32 += -fno-stack-protector
-KBUILD_CFLAGS_32 += $(call cc-option, -foptimize-sibling-calls)
-KBUILD_CFLAGS_32 += -fno-omit-frame-pointer
-KBUILD_CFLAGS_32 += -DDISABLE_BRANCH_PROFILING
-KBUILD_CFLAGS_32 += -DBUILD_VDSO
-
-ifdef CONFIG_MITIGATION_RETPOLINE
-ifneq ($(RETPOLINE_VDSO_CFLAGS),)
-  KBUILD_CFLAGS_32 += $(RETPOLINE_VDSO_CFLAGS)
-endif
-endif
-
-$(obj)/vdso32.so.dbg: KBUILD_CFLAGS = $(KBUILD_CFLAGS_32)
-
-$(obj)/vdso32.so.dbg: $(obj)/vdso32/vdso32.lds $(vobjs32) FORCE
-	$(call if_changed,vdso_and_check)
-
-#
-# The DSO images are built using a special linker script.
-#
-quiet_cmd_vdso = VDSO    $@
-      cmd_vdso = $(LD) -o $@ \
-		       $(VDSO_LDFLAGS) $(VDSO_LDFLAGS_$(filter %.lds,$(^F))) \
-		       -T $(filter %.lds,$^) $(filter %.o,$^)
-
-VDSO_LDFLAGS = -shared --hash-style=both --build-id=sha1 --no-undefined \
-	$(call ld-option, --eh-frame-hdr) -Bsymbolic -z noexecstack
-
-quiet_cmd_vdso_and_check = VDSO    $@
-      cmd_vdso_and_check = $(cmd_vdso); $(cmd_vdso_check)
+# vDSO directories
+obj-$(CONFIG_X86_64)		+= vdso64/
+obj-$(CONFIG_COMPAT_32)		+= vdso32/
arch/x86/entry/vdso/common/Makefile.include new file +89 -0

89 lines not on this page · raw message

arch/x86/entry/vdso/common/note.S renamed from arch/x86/entry/vdso/vdso-note.S +4 -1
similarity index 62%
rename from arch/x86/entry/vdso/vdso-note.S
rename to arch/x86/entry/vdso/common/note.S
index 79423170118f..2cbd39939dc6 100644
@@ -1,13 +1,16 @@
+/* SPDX-License-Identifier: GPL-2.0 */
 /*
  * This supplies .note.* sections to go into the PT_NOTE inside the vDSO text.
  * Here we can supply some information useful to userland.
  */
 
 #include <linux/build-salt.h>
-#include <linux/uts.h>
 #include <linux/version.h>
 #include <linux/elfnote.h>
 
+/* Ideally this would use UTS_NAME, but using a quoted string here
+   doesn't work. Remember to change this when changing the
+   kernel's name. */
 ELFNOTE_START(Linux, 0, "a")
 	.long LINUX_VERSION_CODE
 ELFNOTE_END
arch/x86/entry/vdso/common/vclock_gettime.c renamed from arch/x86/entry/vdso/vclock_gettime.c +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vclock_gettime.c
rename to arch/x86/entry/vdso/common/vclock_gettime.c
arch/x86/entry/vdso/common/vdso-layout.lds.S renamed from arch/x86/entry/vdso/vdso-layout.lds.S +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vdso-layout.lds.S
rename to arch/x86/entry/vdso/common/vdso-layout.lds.S
arch/x86/entry/vdso/common/vgetcpu.c renamed from arch/x86/entry/vdso/vgetcpu.c +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vgetcpu.c
rename to arch/x86/entry/vdso/common/vgetcpu.c
arch/x86/entry/vdso/vdso32/Makefile new file +24 -0
new file mode 100644
index 000000000000..add6afb484ba
@@ -0,0 +1,24 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# 32-bit vDSO images for x86.
+#
+
+# The vDSOs built in this directory
+vdsos-y			:= 32
+
+# Files to link into the vDSO:
+vobjs-y			:= note.o vclock_gettime.o vgetcpu.o
+vobjs-y			+= system_call.o sigreturn.o
+
+# Compilation flags
+flags-y			:= -DBUILD_VDSO32 -m32 -mregparm=0
+flags-$(CONFIG_X86_64)	+= -include $(src)/fake_32bit_build.h
+flags-remove-y          := -m64
+
+# The location of this include matters!
+include $(src)/../common/Makefile.include
+
+# Linker options for the vdso
+VDSO_LDFLAGS_32		:= -m elf_i386 -soname linux-gate.so.1
+
+$(obj)/vdso32.so.dbg: $(vobjs)
arch/x86/entry/vdso/vdso32/note.S +1 -18
index 2cbd39939dc6..62d8aa51ce99 100644
@@ -1,18 +1 @@
-/* SPDX-License-Identifier: GPL-2.0 */
-/*
- * This supplies .note.* sections to go into the PT_NOTE inside the vDSO text.
- * Here we can supply some information useful to userland.
- */
-
-#include <linux/build-salt.h>
-#include <linux/version.h>
-#include <linux/elfnote.h>
-
-/* Ideally this would use UTS_NAME, but using a quoted string here
-   doesn't work. Remember to change this when changing the
-   kernel's name. */
-ELFNOTE_START(Linux, 0, "a")
-	.long LINUX_VERSION_CODE
-ELFNOTE_END
-
-BUILD_SALT
+#include "common/note.S"
arch/x86/entry/vdso/vdso32/system_call.S +1 -1
index d33c6513fd2c..2a15634bbe75 100644
@@ -52,7 +52,7 @@ __kernel_vsyscall:
 	#define SYSENTER_SEQUENCE	"movl %esp, %ebp; sysenter"
 	#define SYSCALL_SEQUENCE	"movl %ecx, %ebp; syscall"
 
-#ifdef CONFIG_X86_64
+#ifdef BUILD_VDSO32_64
 	/* If SYSENTER (Intel) or SYSCALL32 (AMD) is available, use it. */
 	ALTERNATIVE_2 "", SYSENTER_SEQUENCE, X86_FEATURE_SYSENTER32, \
 	                  SYSCALL_SEQUENCE,  X86_FEATURE_SYSCALL32
arch/x86/entry/vdso/vdso32/vclock_gettime.c +1 -4
index 86981decfea8..1481f0021b9f 100644
@@ -1,4 +1 @@
-// SPDX-License-Identifier: GPL-2.0
-#define BUILD_VDSO32
-#include "fake_32bit_build.h"
-#include "../vclock_gettime.c"
+#include "common/vclock_gettime.c"
arch/x86/entry/vdso/vdso32/vdso32.lds.S +1 -1
index 8a3be07006bb..8a853543fc0d 100644
@@ -11,7 +11,7 @@
 
 #define BUILD_VDSO32
 
-#include "../vdso-layout.lds.S"
+#include "common/vdso-layout.lds.S"
 
 /* The ELF entry point can be used to set the AT_SYSINFO value.  */
 ENTRY(__kernel_vsyscall);
arch/x86/entry/vdso/vdso32/vgetcpu.c +1 -3
index 3a9791f5e998..00cc8325a020 100644
@@ -1,3 +1 @@
-// SPDX-License-Identifier: GPL-2.0
-#include "fake_32bit_build.h"
-#include "../vgetcpu.c"
+#include "common/vgetcpu.c"
arch/x86/entry/vdso/vdso64/Makefile new file +46 -0

46 lines not on this page · raw message

arch/x86/entry/vdso/vdso64/note.S new file +1 -0
new file mode 100644
index 000000000000..62d8aa51ce99
@@ -0,0 +1 @@
+#include "common/note.S"
arch/x86/entry/vdso/vdso64/vclock_gettime.c new file +1 -0
new file mode 100644
index 000000000000..1481f0021b9f
@@ -0,0 +1 @@
+#include "common/vclock_gettime.c"
arch/x86/entry/vdso/vdso64/vdso64.lds.S renamed from arch/x86/entry/vdso/vdso.lds.S +1 -1
similarity index 94%
rename from arch/x86/entry/vdso/vdso.lds.S
rename to arch/x86/entry/vdso/vdso64/vdso64.lds.S
index 0bab5f4af6d1..5ce3f2b6373a 100644
@@ -9,7 +9,7 @@
 
 #define BUILD_VDSO64
 
-#include "vdso-layout.lds.S"
+#include "common/vdso-layout.lds.S"
 
 /*
  * This controls what userland symbols we export from the vDSO.
arch/x86/entry/vdso/vdso64/vdsox32.lds.S renamed from arch/x86/entry/vdso/vdsox32.lds.S +1 -1
similarity index 92%
rename from arch/x86/entry/vdso/vdsox32.lds.S
rename to arch/x86/entry/vdso/vdso64/vdsox32.lds.S
index 16a8050a4fb6..3dbd20c8dacc 100644
@@ -9,7 +9,7 @@
 
 #define BUILD_VDSOX32
 
-#include "vdso-layout.lds.S"
+#include "common/vdso-layout.lds.S"
 
 /*
  * This controls what userland symbols we export from the vDSO.
arch/x86/entry/vdso/vdso64/vgetcpu.c new file +1 -0
new file mode 100644
index 000000000000..00cc8325a020
@@ -0,0 +1 @@
+#include "common/vgetcpu.c"
arch/x86/entry/vdso/vdso64/vgetrandom-chacha.S renamed from arch/x86/entry/vdso/vgetrandom-chacha.S +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vgetrandom-chacha.S
rename to arch/x86/entry/vdso/vdso64/vgetrandom-chacha.S
arch/x86/entry/vdso/vdso64/vgetrandom.c renamed from arch/x86/entry/vdso/vgetrandom.c +1 -1
similarity index 91%
rename from arch/x86/entry/vdso/vgetrandom.c
rename to arch/x86/entry/vdso/vdso64/vgetrandom.c
index 430862b8977c..6a95d36b12d9 100644
@@ -4,7 +4,7 @@
  */
 #include <linux/types.h>
 
-#include "../../../../lib/vdso/getrandom.c"
+#include "lib/vdso/getrandom.c"
 
 ssize_t __vdso_getrandom(void *buffer, size_t len, unsigned int flags, void *opaque_state, size_t opaque_len)
 {
arch/x86/entry/vdso/vdso64/vsgx.S renamed from arch/x86/entry/vdso/vsgx.S +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vsgx.S
rename to arch/x86/entry/vdso/vdso64/vsgx.S
Signature · 1 line
2.52.0
merge bot Applied
#8 lore raw

[tip: x86/entry] x86/entry/vdso: Refactor the vdso build

The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     693c819fedcdcabfda7488e2d5e355a84c2fd1b0
Gitweb:        https://git.kernel.org/tip/693c819fedcdcabfda7488e2d5e355a84c2fd1b0
Author:        H. Peter Anvin <[email protected]>
AuthorDate:    Tue, 16 Dec 2025 13:25:57 -08:00
Committer:     Dave Hansen <[email protected]>
CommitterDate: Tue, 13 Jan 2026 15:35:09 -08:00

x86/entry/vdso: Refactor the vdso build

- Separate out the vdso sources into common, vdso32, and vdso64 directories.
- Build the 32- and 64-bit vdsos in their respective subdirectories; this greatly simplifies the build flags handling.
- Unify the mangling of Makefile flags between the 32- and 64-bit vdso code as much as possible; all common rules are put in arch/x86/entry/vdso/common/Makefile.include. The remaining is very simple for 32 bits; the 64-bit one is only slightly more complicated because it contains the x32 generation rule.
- Define __DISABLE_EXPORTS when building the vdso. This need seems to have been masked by different ordering compile flags before.
- Change CONFIG_X86_64 to BUILD_VDSO32_64 in vdso32/system_call.S, to make it compatible with including fake_32bit_build.h.
- The -fcf-protection= option was "leaking" from the kernel build, for reasons that was not clear to me. Furthermore, several distributions ship with it set to a default value other than "-fcf-protection=none". Make it match the configuration options for *user space*.

Note that this patch may seem large, but the vast majority of it is simply code movement.

---
 arch/x86/entry/vdso/Makefile                   | 161 +--------------
 arch/x86/entry/vdso/common/Makefile.include    |  89 ++++++++-
 arch/x86/entry/vdso/common/note.S              |  18 ++-
 arch/x86/entry/vdso/common/vclock_gettime.c    |  77 +++++++-
 arch/x86/entry/vdso/common/vdso-layout.lds.S   | 101 +++++++++-
 arch/x86/entry/vdso/common/vgetcpu.c           |  22 ++-
 arch/x86/entry/vdso/vclock_gettime.c           |  77 +-------
 arch/x86/entry/vdso/vdso-layout.lds.S          | 101 +---------
 arch/x86/entry/vdso/vdso-note.S                |  15 +-
 arch/x86/entry/vdso/vdso.lds.S                 |  37 +---
 arch/x86/entry/vdso/vdso32/Makefile            |  24 ++-
 arch/x86/entry/vdso/vdso32/note.S              |  19 +--
 arch/x86/entry/vdso/vdso32/system_call.S       |   2 +-
 arch/x86/entry/vdso/vdso32/vclock_gettime.c    |   5 +-
 arch/x86/entry/vdso/vdso32/vdso32.lds.S        |   2 +-
 arch/x86/entry/vdso/vdso32/vgetcpu.c           |   4 +-
 arch/x86/entry/vdso/vdso64/Makefile            |  46 ++++-
 arch/x86/entry/vdso/vdso64/note.S              |   1 +-
 arch/x86/entry/vdso/vdso64/vclock_gettime.c    |   1 +-
 arch/x86/entry/vdso/vdso64/vdso64.lds.S        |  37 +++-
 arch/x86/entry/vdso/vdso64/vdsox32.lds.S       |  27 ++-
 arch/x86/entry/vdso/vdso64/vgetcpu.c           |   1 +-
 arch/x86/entry/vdso/vdso64/vgetrandom-chacha.S | 178 ++++++++++++++++-
 arch/x86/entry/vdso/vdso64/vgetrandom.c        |  15 +-
 arch/x86/entry/vdso/vdso64/vsgx.S              | 150 +++++++++++++-
 arch/x86/entry/vdso/vdsox32.lds.S              |  27 +--
 arch/x86/entry/vdso/vgetcpu.c                  |  22 +--
 arch/x86/entry/vdso/vgetrandom-chacha.S        | 178 +----------------
 arch/x86/entry/vdso/vgetrandom.c               |  15 +-
 arch/x86/entry/vdso/vsgx.S                     | 150 +-------------
 30 files changed, 798 insertions(+), 804 deletions(-)
 create mode 100644 arch/x86/entry/vdso/common/Makefile.include
 create mode 100644 arch/x86/entry/vdso/common/note.S
 create mode 100644 arch/x86/entry/vdso/common/vclock_gettime.c
 create mode 100644 arch/x86/entry/vdso/common/vdso-layout.lds.S
 create mode 100644 arch/x86/entry/vdso/common/vgetcpu.c
 delete mode 100644 arch/x86/entry/vdso/vclock_gettime.c
 delete mode 100644 arch/x86/entry/vdso/vdso-layout.lds.S
 delete mode 100644 arch/x86/entry/vdso/vdso-note.S
 delete mode 100644 arch/x86/entry/vdso/vdso.lds.S
 create mode 100644 arch/x86/entry/vdso/vdso32/Makefile
 create mode 100644 arch/x86/entry/vdso/vdso64/Makefile
 create mode 100644 arch/x86/entry/vdso/vdso64/note.S
 create mode 100644 arch/x86/entry/vdso/vdso64/vclock_gettime.c
 create mode 100644 arch/x86/entry/vdso/vdso64/vdso64.lds.S
 create mode 100644 arch/x86/entry/vdso/vdso64/vdsox32.lds.S
 create mode 100644 arch/x86/entry/vdso/vdso64/vgetcpu.c
 create mode 100644 arch/x86/entry/vdso/vdso64/vgetrandom-chacha.S
 create mode 100644 arch/x86/entry/vdso/vdso64/vgetrandom.c
 create mode 100644 arch/x86/entry/vdso/vdso64/vsgx.S
 delete mode 100644 arch/x86/entry/vdso/vdsox32.lds.S
 delete mode 100644 arch/x86/entry/vdso/vgetcpu.c
 delete mode 100644 arch/x86/entry/vdso/vgetrandom-chacha.S
 delete mode 100644 arch/x86/entry/vdso/vgetrandom.c
 delete mode 100644 arch/x86/entry/vdso/vsgx.S
arch/x86/entry/vdso/Makefile +6 -155
index 3d9b09f..987b43f 100644
@@ -3,159 +3,10 @@
 # Building vDSO images for x86.
 #
 
-# Include the generic Makefile to check the built vDSO:
-include $(srctree)/lib/vdso/Makefile.include
+# Regular kernel objects
+obj-y				:= vma.o extable.o
+obj-$(CONFIG_COMPAT_32)		+= vdso32-setup.o
 
-# Files to link into the vDSO:
-vobjs-y := vdso-note.o vclock_gettime.o vgetcpu.o vgetrandom.o vgetrandom-chacha.o
-vobjs32-y := vdso32/note.o vdso32/system_call.o vdso32/sigreturn.o
-vobjs32-y += vdso32/vclock_gettime.o vdso32/vgetcpu.o
-vobjs-$(CONFIG_X86_SGX)	+= vsgx.o
-
-# Files to link into the kernel:
-obj-y						+= vma.o extable.o
-
-# vDSO images to build:
-obj-$(CONFIG_X86_64)				+= vdso64-image.o
-obj-$(CONFIG_X86_X32_ABI)			+= vdsox32-image.o
-obj-$(CONFIG_COMPAT_32)				+= vdso32-image.o vdso32-setup.o
-
-vobjs := $(addprefix $(obj)/, $(vobjs-y))
-vobjs32 := $(addprefix $(obj)/, $(vobjs32-y))
-
-$(obj)/vdso.o: $(obj)/vdso.so
-
-targets += vdso.lds $(vobjs-y)
-targets += vdso32/vdso32.lds $(vobjs32-y)
-
-targets += $(foreach x, 64 x32 32, vdso-image-$(x).c vdso$(x).so vdso$(x).so.dbg)
-
-CPPFLAGS_vdso.lds += -P -C
-
-VDSO_LDFLAGS_vdso.lds = -m elf_x86_64 -soname linux-vdso.so.1 \
-			-z max-page-size=4096
-
-$(obj)/vdso64.so.dbg: $(obj)/vdso.lds $(vobjs) FORCE
-	$(call if_changed,vdso_and_check)
-
-VDSO2C = $(objtree)/arch/x86/tools/vdso2c
-
-quiet_cmd_vdso2c = VDSO2C  $@
-      cmd_vdso2c = $(VDSO2C) $< $(<:%.dbg=%) $@
-
-$(obj)/vdso%-image.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(VDSO2C) FORCE
-	$(call if_changed,vdso2c)
-
-#
-# Don't omit frame pointers for ease of userspace debugging, but do
-# optimize sibling calls.
-#
-CFL := $(PROFILING) -mcmodel=small -fPIC -O2 -fasynchronous-unwind-tables -m64 \
-       $(filter -g%,$(KBUILD_CFLAGS)) -fno-stack-protector \
-       -fno-omit-frame-pointer -foptimize-sibling-calls \
-       -DDISABLE_BRANCH_PROFILING -DBUILD_VDSO
-
-ifdef CONFIG_MITIGATION_RETPOLINE
-ifneq ($(RETPOLINE_VDSO_CFLAGS),)
-  CFL += $(RETPOLINE_VDSO_CFLAGS)
-endif
-endif
-
-$(vobjs): KBUILD_CFLAGS := $(filter-out $(PADDING_CFLAGS) $(CC_FLAGS_LTO) $(CC_FLAGS_CFI) $(RANDSTRUCT_CFLAGS) $(KSTACK_ERASE_CFLAGS) $(GCC_PLUGINS_CFLAGS) $(RETPOLINE_CFLAGS),$(KBUILD_CFLAGS)) $(CFL)
-$(vobjs): KBUILD_AFLAGS += -DBUILD_VDSO
-
-#
-# vDSO code runs in userspace and -pg doesn't help with profiling anyway.
-#
-CFLAGS_REMOVE_vclock_gettime.o = -pg
-CFLAGS_REMOVE_vdso32/vclock_gettime.o = -pg
-CFLAGS_REMOVE_vgetcpu.o = -pg
-CFLAGS_REMOVE_vdso32/vgetcpu.o = -pg
-CFLAGS_REMOVE_vsgx.o = -pg
-CFLAGS_REMOVE_vgetrandom.o = -pg
-
-#
-# X32 processes use x32 vDSO to access 64bit kernel data.
-#
-# Build x32 vDSO image:
-# 1. Compile x32 vDSO as 64bit.
-# 2. Convert object files to x32.
-# 3. Build x32 VDSO image with x32 objects, which contains 64bit codes
-# so that it can reach 64bit address space with 64bit pointers.
-#
-
-CPPFLAGS_vdsox32.lds = $(CPPFLAGS_vdso.lds)
-VDSO_LDFLAGS_vdsox32.lds = -m elf32_x86_64 -soname linux-vdso.so.1 \
-			   -z max-page-size=4096
-
-# x32-rebranded versions
-vobjx32s-y := $(vobjs-y:.o=-x32.o)
-
-# same thing, but in the output directory
-vobjx32s := $(addprefix $(obj)/, $(vobjx32s-y))
-
-# Convert 64bit object file to x32 for x32 vDSO.
-quiet_cmd_x32 = X32     $@
-      cmd_x32 = $(OBJCOPY) -O elf32-x86-64 $< $@
-
-$(obj)/%-x32.o: $(obj)/%.o FORCE
-	$(call if_changed,x32)
-
-targets += vdsox32.lds $(vobjx32s-y)
-
-$(obj)/%.so: OBJCOPYFLAGS := -S --remove-section __ex_table
-$(obj)/%.so: $(obj)/%.so.dbg FORCE
-	$(call if_changed,objcopy)
-
-$(obj)/vdsox32.so.dbg: $(obj)/vdsox32.lds $(vobjx32s) FORCE
-	$(call if_changed,vdso_and_check)
-
-CPPFLAGS_vdso32/vdso32.lds = $(CPPFLAGS_vdso.lds)
-VDSO_LDFLAGS_vdso32.lds = -m elf_i386 -soname linux-gate.so.1
-
-KBUILD_AFLAGS_32 := $(filter-out -m64,$(KBUILD_AFLAGS)) -DBUILD_VDSO
-$(obj)/vdso32.so.dbg: KBUILD_AFLAGS = $(KBUILD_AFLAGS_32)
-$(obj)/vdso32.so.dbg: asflags-$(CONFIG_X86_64) += -m32
-
-KBUILD_CFLAGS_32 := $(filter-out -m64,$(KBUILD_CFLAGS))
-KBUILD_CFLAGS_32 := $(filter-out -mcmodel=kernel,$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out -fno-pic,$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out -mfentry,$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(RANDSTRUCT_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(KSTACK_ERASE_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(GCC_PLUGINS_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(RETPOLINE_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(CC_FLAGS_LTO),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(CC_FLAGS_CFI),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(PADDING_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 += -m32 -msoft-float -mregparm=0 -fpic
-KBUILD_CFLAGS_32 += -fno-stack-protector
-KBUILD_CFLAGS_32 += $(call cc-option, -foptimize-sibling-calls)
-KBUILD_CFLAGS_32 += -fno-omit-frame-pointer
-KBUILD_CFLAGS_32 += -DDISABLE_BRANCH_PROFILING
-KBUILD_CFLAGS_32 += -DBUILD_VDSO
-
-ifdef CONFIG_MITIGATION_RETPOLINE
-ifneq ($(RETPOLINE_VDSO_CFLAGS),)
-  KBUILD_CFLAGS_32 += $(RETPOLINE_VDSO_CFLAGS)
-endif
-endif
-
-$(obj)/vdso32.so.dbg: KBUILD_CFLAGS = $(KBUILD_CFLAGS_32)
-
-$(obj)/vdso32.so.dbg: $(obj)/vdso32/vdso32.lds $(vobjs32) FORCE
-	$(call if_changed,vdso_and_check)
-
-#
-# The DSO images are built using a special linker script.
-#
-quiet_cmd_vdso = VDSO    $@
-      cmd_vdso = $(LD) -o $@ \
-		       $(VDSO_LDFLAGS) $(VDSO_LDFLAGS_$(filter %.lds,$(^F))) \
-		       -T $(filter %.lds,$^) $(filter %.o,$^)
-
-VDSO_LDFLAGS = -shared --hash-style=both --build-id=sha1 --no-undefined \
-	$(call ld-option, --eh-frame-hdr) -Bsymbolic -z noexecstack
-
-quiet_cmd_vdso_and_check = VDSO    $@
-      cmd_vdso_and_check = $(cmd_vdso); $(cmd_vdso_check)
+# vDSO directories
+obj-$(CONFIG_X86_64)		+= vdso64/
+obj-$(CONFIG_COMPAT_32)		+= vdso32/
arch/x86/entry/vdso/common/Makefile.include new file +89 -0

89 lines not on this page · raw message

arch/x86/entry/vdso/common/note.S new file +18 -0
new file mode 100644
index 0000000..2cbd399
@@ -0,0 +1,18 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * This supplies .note.* sections to go into the PT_NOTE inside the vDSO text.
+ * Here we can supply some information useful to userland.
+ */
+
+#include <linux/build-salt.h>
+#include <linux/version.h>
+#include <linux/elfnote.h>
+
+/* Ideally this would use UTS_NAME, but using a quoted string here
+   doesn't work. Remember to change this when changing the
+   kernel's name. */
+ELFNOTE_START(Linux, 0, "a")
+	.long LINUX_VERSION_CODE
+ELFNOTE_END
+
+BUILD_SALT
arch/x86/entry/vdso/common/vclock_gettime.c new file +77 -0

77 lines not on this page · raw message

arch/x86/entry/vdso/common/vdso-layout.lds.S new file +101 -0

101 lines not on this page · raw message

arch/x86/entry/vdso/common/vgetcpu.c new file +22 -0
new file mode 100644
index 0000000..e464030
@@ -0,0 +1,22 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright 2006 Andi Kleen, SUSE Labs.
+ *
+ * Fast user context implementation of getcpu()
+ */
+
+#include <linux/kernel.h>
+#include <linux/getcpu.h>
+#include <asm/segment.h>
+#include <vdso/processor.h>
+
+notrace long
+__vdso_getcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *unused)
+{
+	vdso_read_cpunode(cpu, node);
+
+	return 0;
+}
+
+long getcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *tcache)
+	__attribute__((weak, alias("__vdso_getcpu")));
arch/x86/entry/vdso/vclock_gettime.c deleted +0 -77

77 lines not on this page · raw message

arch/x86/entry/vdso/vdso-layout.lds.S deleted +0 -101

101 lines not on this page · raw message

arch/x86/entry/vdso/vdso-note.S deleted +0 -15
deleted file mode 100644
index 7942317..0000000
@@ -1,15 +0,0 @@
-/*
- * This supplies .note.* sections to go into the PT_NOTE inside the vDSO text.
- * Here we can supply some information useful to userland.
- */
-
-#include <linux/build-salt.h>
-#include <linux/uts.h>
-#include <linux/version.h>
-#include <linux/elfnote.h>
-
-ELFNOTE_START(Linux, 0, "a")
-	.long LINUX_VERSION_CODE
-ELFNOTE_END
-
-BUILD_SALT
arch/x86/entry/vdso/vdso.lds.S deleted +0 -37
deleted file mode 100644
index 0bab5f4..0000000
@@ -1,37 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0 */
-/*
- * Linker script for 64-bit vDSO.
- * We #include the file to define the layout details.
- *
- * This file defines the version script giving the user-exported symbols in
- * the DSO.
- */
-
-#define BUILD_VDSO64
-
-#include "vdso-layout.lds.S"
-
-/*
- * This controls what userland symbols we export from the vDSO.
- */
-VERSION {
-	LINUX_2.6 {
-	global:
-		clock_gettime;
-		__vdso_clock_gettime;
-		gettimeofday;
-		__vdso_gettimeofday;
-		getcpu;
-		__vdso_getcpu;
-		time;
-		__vdso_time;
-		clock_getres;
-		__vdso_clock_getres;
-#ifdef CONFIG_X86_SGX
-		__vdso_sgx_enter_enclave;
-#endif
-		getrandom;
-		__vdso_getrandom;
-	local: *;
-	};
-}
arch/x86/entry/vdso/vdso32/Makefile new file +24 -0
new file mode 100644
index 0000000..add6afb
@@ -0,0 +1,24 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# 32-bit vDSO images for x86.
+#
+
+# The vDSOs built in this directory
+vdsos-y			:= 32
+
+# Files to link into the vDSO:
+vobjs-y			:= note.o vclock_gettime.o vgetcpu.o
+vobjs-y			+= system_call.o sigreturn.o
+
+# Compilation flags
+flags-y			:= -DBUILD_VDSO32 -m32 -mregparm=0
+flags-$(CONFIG_X86_64)	+= -include $(src)/fake_32bit_build.h
+flags-remove-y          := -m64
+
+# The location of this include matters!
+include $(src)/../common/Makefile.include
+
+# Linker options for the vdso
+VDSO_LDFLAGS_32		:= -m elf_i386 -soname linux-gate.so.1
+
+$(obj)/vdso32.so.dbg: $(vobjs)
arch/x86/entry/vdso/vdso32/note.S +1 -18
index 2cbd399..62d8aa5 100644
@@ -1,18 +1 @@
-/* SPDX-License-Identifier: GPL-2.0 */
-/*
- * This supplies .note.* sections to go into the PT_NOTE inside the vDSO text.
- * Here we can supply some information useful to userland.
- */
-
-#include <linux/build-salt.h>
-#include <linux/version.h>
-#include <linux/elfnote.h>
-
-/* Ideally this would use UTS_NAME, but using a quoted string here
-   doesn't work. Remember to change this when changing the
-   kernel's name. */
-ELFNOTE_START(Linux, 0, "a")
-	.long LINUX_VERSION_CODE
-ELFNOTE_END
-
-BUILD_SALT
+#include "common/note.S"
arch/x86/entry/vdso/vdso32/system_call.S +1 -1
index d33c651..2a15634 100644
@@ -52,7 +52,7 @@ __kernel_vsyscall:
 	#define SYSENTER_SEQUENCE	"movl %esp, %ebp; sysenter"
 	#define SYSCALL_SEQUENCE	"movl %ecx, %ebp; syscall"
 
-#ifdef CONFIG_X86_64
+#ifdef BUILD_VDSO32_64
 	/* If SYSENTER (Intel) or SYSCALL32 (AMD) is available, use it. */
 	ALTERNATIVE_2 "", SYSENTER_SEQUENCE, X86_FEATURE_SYSENTER32, \
 	                  SYSCALL_SEQUENCE,  X86_FEATURE_SYSCALL32
arch/x86/entry/vdso/vdso32/vclock_gettime.c +1 -4
index 86981de..1481f00 100644
@@ -1,4 +1 @@
-// SPDX-License-Identifier: GPL-2.0
-#define BUILD_VDSO32
-#include "fake_32bit_build.h"
-#include "../vclock_gettime.c"
+#include "common/vclock_gettime.c"
arch/x86/entry/vdso/vdso32/vdso32.lds.S +1 -1
index 8a3be07..8a85354 100644
@@ -11,7 +11,7 @@
 
 #define BUILD_VDSO32
 
-#include "../vdso-layout.lds.S"
+#include "common/vdso-layout.lds.S"
 
 /* The ELF entry point can be used to set the AT_SYSINFO value.  */
 ENTRY(__kernel_vsyscall);
arch/x86/entry/vdso/vdso32/vgetcpu.c +1 -3
index 3a9791f..00cc832 100644
@@ -1,3 +1 @@
-// SPDX-License-Identifier: GPL-2.0
-#include "fake_32bit_build.h"
-#include "../vgetcpu.c"
+#include "common/vgetcpu.c"
arch/x86/entry/vdso/vdso64/Makefile new file +46 -0

46 lines not on this page · raw message

arch/x86/entry/vdso/vdso64/note.S new file +1 -0
new file mode 100644
index 0000000..62d8aa5
@@ -0,0 +1 @@
+#include "common/note.S"
arch/x86/entry/vdso/vdso64/vclock_gettime.c new file +1 -0
new file mode 100644
index 0000000..1481f00
@@ -0,0 +1 @@
+#include "common/vclock_gettime.c"
arch/x86/entry/vdso/vdso64/vdso64.lds.S new file +37 -0
new file mode 100644
index 0000000..5ce3f2b
@@ -0,0 +1,37 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Linker script for 64-bit vDSO.
+ * We #include the file to define the layout details.
+ *
+ * This file defines the version script giving the user-exported symbols in
+ * the DSO.
+ */
+
+#define BUILD_VDSO64
+
+#include "common/vdso-layout.lds.S"
+
+/*
+ * This controls what userland symbols we export from the vDSO.
+ */
+VERSION {
+	LINUX_2.6 {
+	global:
+		clock_gettime;
+		__vdso_clock_gettime;
+		gettimeofday;
+		__vdso_gettimeofday;
+		getcpu;
+		__vdso_getcpu;
+		time;
+		__vdso_time;
+		clock_getres;
+		__vdso_clock_getres;
+#ifdef CONFIG_X86_SGX
+		__vdso_sgx_enter_enclave;
+#endif
+		getrandom;
+		__vdso_getrandom;
+	local: *;
+	};
+}
arch/x86/entry/vdso/vdso64/vdsox32.lds.S new file +27 -0
new file mode 100644
index 0000000..3dbd20c
@@ -0,0 +1,27 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Linker script for x32 vDSO.
+ * We #include the file to define the layout details.
+ *
+ * This file defines the version script giving the user-exported symbols in
+ * the DSO.
+ */
+
+#define BUILD_VDSOX32
+
+#include "common/vdso-layout.lds.S"
+
+/*
+ * This controls what userland symbols we export from the vDSO.
+ */
+VERSION {
+	LINUX_2.6 {
+	global:
+		__vdso_clock_gettime;
+		__vdso_gettimeofday;
+		__vdso_getcpu;
+		__vdso_time;
+		__vdso_clock_getres;
+	local: *;
+	};
+}
arch/x86/entry/vdso/vdso64/vgetcpu.c new file +1 -0
new file mode 100644
index 0000000..00cc832
@@ -0,0 +1 @@
+#include "common/vgetcpu.c"
arch/x86/entry/vdso/vdso64/vgetrandom-chacha.S new file +178 -0

178 lines not on this page · raw message

arch/x86/entry/vdso/vdso64/vgetrandom.c new file +15 -0
new file mode 100644
index 0000000..6a95d36
@@ -0,0 +1,15 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2022-2024 Jason A. Donenfeld <[email protected]>. All Rights Reserved.
+ */
+#include <linux/types.h>
+
+#include "lib/vdso/getrandom.c"
+
+ssize_t __vdso_getrandom(void *buffer, size_t len, unsigned int flags, void *opaque_state, size_t opaque_len)
+{
+	return __cvdso_getrandom(buffer, len, flags, opaque_state, opaque_len);
+}
+
+ssize_t getrandom(void *, size_t, unsigned int, void *, size_t)
+	__attribute__((weak, alias("__vdso_getrandom")));
arch/x86/entry/vdso/vdso64/vsgx.S new file +150 -0

150 lines not on this page · raw message

arch/x86/entry/vdso/vdsox32.lds.S deleted +0 -27
deleted file mode 100644
index 16a8050..0000000
@@ -1,27 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0 */
-/*
- * Linker script for x32 vDSO.
- * We #include the file to define the layout details.
- *
- * This file defines the version script giving the user-exported symbols in
- * the DSO.
- */
-
-#define BUILD_VDSOX32
-
-#include "vdso-layout.lds.S"
-
-/*
- * This controls what userland symbols we export from the vDSO.
- */
-VERSION {
-	LINUX_2.6 {
-	global:
-		__vdso_clock_gettime;
-		__vdso_gettimeofday;
-		__vdso_getcpu;
-		__vdso_time;
-		__vdso_clock_getres;
-	local: *;
-	};
-}
arch/x86/entry/vdso/vgetcpu.c deleted +0 -22
deleted file mode 100644
index e464030..0000000
@@ -1,22 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0-only
-/*
- * Copyright 2006 Andi Kleen, SUSE Labs.
- *
- * Fast user context implementation of getcpu()
- */
-
-#include <linux/kernel.h>
-#include <linux/getcpu.h>
-#include <asm/segment.h>
-#include <vdso/processor.h>
-
-notrace long
-__vdso_getcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *unused)
-{
-	vdso_read_cpunode(cpu, node);
-
-	return 0;
-}
-
-long getcpu(unsigned *cpu, unsigned *node, struct getcpu_cache *tcache)
-	__attribute__((weak, alias("__vdso_getcpu")));
arch/x86/entry/vdso/vgetrandom-chacha.S deleted +0 -178

178 lines not on this page · raw message

arch/x86/entry/vdso/vgetrandom.c deleted +0 -15
deleted file mode 100644
index 430862b..0000000
@@ -1,15 +0,0 @@
-// SPDX-License-Identifier: GPL-2.0-only
-/*
- * Copyright (C) 2022-2024 Jason A. Donenfeld <[email protected]>. All Rights Reserved.
- */
-#include <linux/types.h>
-
-#include "../../../../lib/vdso/getrandom.c"
-
-ssize_t __vdso_getrandom(void *buffer, size_t len, unsigned int flags, void *opaque_state, size_t opaque_len)
-{
-	return __cvdso_getrandom(buffer, len, flags, opaque_state, opaque_len);
-}
-
-ssize_t getrandom(void *, size_t, unsigned int, void *, size_t)
-	__attribute__((weak, alias("__vdso_getrandom")));
arch/x86/entry/vdso/vsgx.S deleted +0 -150

150 lines not on this page · raw message

author maintainer
#9 lore raw

PATCH v4 4/10 x86/entry/vdso32: don't rely on int80_landing_pad for adjusting ip

In reply to H. Peter Anvin

There is no fundamental reason to use the int80_landing_pad symbol to adjust ip when moving the vdso. If ip falls within the vdso, and the vdso is moved, we should change the ip accordingly, regardless of mode or location within the vdso. This *currently* can only happen on 32 bits, but there isn't any reason not to do so generically.

Note that if this is ever possible from a vdso-internal call, then the user space stack will also needed to be adjusted (as well as the shadow stack, if enabled.) Fortunately this is not currently the case.

At the moment, we don't even consider other threads when moving the vdso. The assumption is that it is only used by process freeze/thaw for migration, where this is not an issue.

---
 arch/x86/entry/vdso/vma.c | 16 ++++++----------
 1 file changed, 6 insertions(+), 10 deletions(-)
arch/x86/entry/vdso/vma.c +6 -10
index 8f98c2d7c7a9..e7fd7517370f 100644
@@ -65,16 +65,12 @@ static vm_fault_t vdso_fault(const struct vm_special_mapping *sm,
 static void vdso_fix_landing(const struct vdso_image *image,
 		struct vm_area_struct *new_vma)
 {
-	if (in_ia32_syscall() && image == &vdso32_image) {
-		struct pt_regs *regs = current_pt_regs();
-		unsigned long vdso_land = image->sym_int80_landing_pad;
-		unsigned long old_land_addr = vdso_land +
-			(unsigned long)current->mm->context.vdso;
-
-		/* Fixing userspace landing - look at do_fast_syscall_32 */
-		if (regs->ip == old_land_addr)
-			regs->ip = new_vma->vm_start + vdso_land;
-	}
+	struct pt_regs *regs = current_pt_regs();
+	unsigned long ipoffset = regs->ip -
+		(unsigned long)current->mm->context.vdso;
+
+	if (ipoffset < image->size)
+		regs->ip = new_vma->vm_start + ipoffset;
 }
 
 static int vdso_mremap(const struct vm_special_mapping *sm,
Signature · 1 line
2.52.0
merge bot Applied
#10 lore raw

[tip: x86/entry] x86/entry/vdso32: Don't rely on int80_landing_pad for adjusting ip

The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     6e150b71019f386a021004fafea9ef7189bc6aea
Gitweb:        https://git.kernel.org/tip/6e150b71019f386a021004fafea9ef7189bc6aea
Author:        H. Peter Anvin <[email protected]>
AuthorDate:    Tue, 16 Dec 2025 13:25:58 -08:00
Committer:     Dave Hansen <[email protected]>
CommitterDate: Tue, 13 Jan 2026 16:37:58 -08:00

x86/entry/vdso32: Don't rely on int80_landing_pad for adjusting ip

There is no fundamental reason to use the int80_landing_pad symbol to adjust ip when moving the vdso. If ip falls within the vdso, and the vdso is moved, we should change the ip accordingly, regardless of mode or location within the vdso. This *currently* can only happen on 32 bits, but there isn't any reason not to do so generically.

Note that if this is ever possible from a vdso-internal call, then the user space stack will also needed to be adjusted (as well as the shadow stack, if enabled.) Fortunately this is not currently the case.

At the moment, we don't even consider other threads when moving the vdso. The assumption is that it is only used by process freeze/thaw for migration, where this is not an issue.

---
 arch/x86/entry/vdso/vma.c | 16 ++++++----------
 1 file changed, 6 insertions(+), 10 deletions(-)
arch/x86/entry/vdso/vma.c +6 -10
index 8f98c2d..e7fd751 100644
@@ -65,16 +65,12 @@ static vm_fault_t vdso_fault(const struct vm_special_mapping *sm,
 static void vdso_fix_landing(const struct vdso_image *image,
 		struct vm_area_struct *new_vma)
 {
-	if (in_ia32_syscall() && image == &vdso32_image) {
-		struct pt_regs *regs = current_pt_regs();
-		unsigned long vdso_land = image->sym_int80_landing_pad;
-		unsigned long old_land_addr = vdso_land +
-			(unsigned long)current->mm->context.vdso;
-
-		/* Fixing userspace landing - look at do_fast_syscall_32 */
-		if (regs->ip == old_land_addr)
-			regs->ip = new_vma->vm_start + vdso_land;
-	}
+	struct pt_regs *regs = current_pt_regs();
+	unsigned long ipoffset = regs->ip -
+		(unsigned long)current->mm->context.vdso;
+
+	if (ipoffset < image->size)
+		regs->ip = new_vma->vm_start + ipoffset;
 }
 
 static int vdso_mremap(const struct vm_special_mapping *sm,
author maintainer
#11 lore raw

PATCH v4 5/10 x86/entry/vdso32: remove SYSCALL_ENTER_KERNEL macro in sigreturn.S

In reply to H. Peter Anvin

A macro SYSCALL_ENTER_KERNEL was defined in sigreturn.S, with the ability of overriding it. The override capability, however, is not used anywhere, and the macro name is potentially confusing because it seems to imply that sysenter/syscall could be used here, which is NOT
true: the sigreturn system calls MUST use int $0x80.

---
 arch/x86/entry/vdso/vdso32/sigreturn.S | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)
arch/x86/entry/vdso/vdso32/sigreturn.S +2 -6
index 1bd068f72d4c..965900c6763b 100644
@@ -3,10 +3,6 @@
 #include <asm/unistd_32.h>
 #include <asm/asm-offsets.h>
 
-#ifndef SYSCALL_ENTER_KERNEL
-#define	SYSCALL_ENTER_KERNEL	int $0x80
-#endif
-
 	.text
 	.globl __kernel_sigreturn
 	.type __kernel_sigreturn,@function
@@ -16,7 +12,7 @@ __kernel_sigreturn:
 .LSTART_sigreturn:
 	popl %eax		/* XXX does this mean it needs unwind info? */
 	movl $__NR_sigreturn, %eax
-	SYSCALL_ENTER_KERNEL
+	int $0x80
 .LEND_sigreturn:
 SYM_INNER_LABEL(vdso32_sigreturn_landing_pad, SYM_L_GLOBAL)
 	nop
@@ -28,7 +24,7 @@ SYM_INNER_LABEL(vdso32_sigreturn_landing_pad, SYM_L_GLOBAL)
 __kernel_rt_sigreturn:
 .LSTART_rt_sigreturn:
 	movl $__NR_rt_sigreturn, %eax
-	SYSCALL_ENTER_KERNEL
+	int $0x80
 .LEND_rt_sigreturn:
 SYM_INNER_LABEL(vdso32_rt_sigreturn_landing_pad, SYM_L_GLOBAL)
 	nop
Signature · 1 line
2.52.0
merge bot Applied
#12 lore raw

[tip: x86/entry] x86/entry/vdso32: Remove SYSCALL_ENTER_KERNEL macro in sigreturn.S

The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     98d3e996513ad00b7824ea3bece506fc645547dd
Gitweb:        https://git.kernel.org/tip/98d3e996513ad00b7824ea3bece506fc645547dd
Author:        H. Peter Anvin <[email protected]>
AuthorDate:    Tue, 16 Dec 2025 13:25:59 -08:00
Committer:     Dave Hansen <[email protected]>
CommitterDate: Tue, 13 Jan 2026 16:37:58 -08:00

x86/entry/vdso32: Remove SYSCALL_ENTER_KERNEL macro in sigreturn.S

A macro SYSCALL_ENTER_KERNEL was defined in sigreturn.S, with the ability of overriding it. The override capability, however, is not used anywhere, and the macro name is potentially confusing because it seems to imply that sysenter/syscall could be used here, which is NOT
true: the sigreturn system calls MUST use int $0x80.

---
 arch/x86/entry/vdso/vdso32/sigreturn.S | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)
arch/x86/entry/vdso/vdso32/sigreturn.S +2 -6
index 1bd068f..965900c 100644
@@ -3,10 +3,6 @@
 #include <asm/unistd_32.h>
 #include <asm/asm-offsets.h>
 
-#ifndef SYSCALL_ENTER_KERNEL
-#define	SYSCALL_ENTER_KERNEL	int $0x80
-#endif
-
 	.text
 	.globl __kernel_sigreturn
 	.type __kernel_sigreturn,@function
@@ -16,7 +12,7 @@ __kernel_sigreturn:
 .LSTART_sigreturn:
 	popl %eax		/* XXX does this mean it needs unwind info? */
 	movl $__NR_sigreturn, %eax
-	SYSCALL_ENTER_KERNEL
+	int $0x80
 .LEND_sigreturn:
 SYM_INNER_LABEL(vdso32_sigreturn_landing_pad, SYM_L_GLOBAL)
 	nop
@@ -28,7 +24,7 @@ SYM_INNER_LABEL(vdso32_sigreturn_landing_pad, SYM_L_GLOBAL)
 __kernel_rt_sigreturn:
 .LSTART_rt_sigreturn:
 	movl $__NR_rt_sigreturn, %eax
-	SYSCALL_ENTER_KERNEL
+	int $0x80
 .LEND_rt_sigreturn:
 SYM_INNER_LABEL(vdso32_rt_sigreturn_landing_pad, SYM_L_GLOBAL)
 	nop
author maintainer
#13 lore raw

PATCH v4 6/10 x86/entry/vdso32: remove open-coded DWARF in sigreturn.S

In reply to H. Peter Anvin

The vdso32 sigreturn.S contains open-coded DWARF bytecode, which includes a hack for gdb to not try to step back to a previous call instruction when backtracing from a signal handler.

Neither of those are necessary anymore: the backtracing issue is handled by ".cfi_entry simple" and ".cfi_signal_frame", both of which have been supported for a very long time now, which allows the remaining frame to be built using regular .cfi annotations.

Add a few more register offsets to the signal frame just for good measure.

Replace the nop on fallthrough of the system call (which should never, ever happen) with a ud2a trap.

---
 arch/x86/entry/vdso/vdso32/sigreturn.S | 146 ++++++-------------------
 arch/x86/include/asm/dwarf2.h          |   1 +
 arch/x86/kernel/asm-offsets.c          |   6 +
 3 files changed, 39 insertions(+), 114 deletions(-)
arch/x86/entry/vdso/vdso32/sigreturn.S +32 -114
index 965900c6763b..25b0ac4b4bfe 100644
@@ -1,136 +1,54 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 #include <linux/linkage.h>
 #include <asm/unistd_32.h>
+#include <asm/dwarf2.h>
 #include <asm/asm-offsets.h>
 
+.macro STARTPROC_SIGNAL_FRAME sc
+	CFI_STARTPROC	simple
+	CFI_SIGNAL_FRAME
+	/* -4 as pretcode has already been popped */
+	CFI_DEF_CFA	esp,	\sc - 4
+	CFI_OFFSET	eip,    IA32_SIGCONTEXT_ip
+	CFI_OFFSET	eax,    IA32_SIGCONTEXT_ax
+	CFI_OFFSET	ebx,    IA32_SIGCONTEXT_bx
+	CFI_OFFSET	ecx,    IA32_SIGCONTEXT_cx
+	CFI_OFFSET	edx,    IA32_SIGCONTEXT_dx
+	CFI_OFFSET	esp,    IA32_SIGCONTEXT_sp
+	CFI_OFFSET	ebp,    IA32_SIGCONTEXT_bp
+	CFI_OFFSET	esi,    IA32_SIGCONTEXT_si
+	CFI_OFFSET	edi,    IA32_SIGCONTEXT_di
+	CFI_OFFSET	es,     IA32_SIGCONTEXT_es
+	CFI_OFFSET	cs,     IA32_SIGCONTEXT_cs
+	CFI_OFFSET	ss,     IA32_SIGCONTEXT_ss
+	CFI_OFFSET	ds,     IA32_SIGCONTEXT_ds
+	CFI_OFFSET	eflags, IA32_SIGCONTEXT_flags
+.endm
+
 	.text
 	.globl __kernel_sigreturn
 	.type __kernel_sigreturn,@function
-	nop /* this guy is needed for .LSTARTFDEDLSI1 below (watch for HACK) */
 	ALIGN
 __kernel_sigreturn:
-.LSTART_sigreturn:
-	popl %eax		/* XXX does this mean it needs unwind info? */
+	STARTPROC_SIGNAL_FRAME IA32_SIGFRAME_sigcontext
+	popl %eax
+	CFI_ADJUST_CFA_OFFSET -4
 	movl $__NR_sigreturn, %eax
 	int $0x80
-.LEND_sigreturn:
 SYM_INNER_LABEL(vdso32_sigreturn_landing_pad, SYM_L_GLOBAL)
-	nop
-	.size __kernel_sigreturn,.-.LSTART_sigreturn
+	ud2a
+	CFI_ENDPROC
+	.size __kernel_sigreturn,.-__kernel_sigreturn
 
 	.globl __kernel_rt_sigreturn
 	.type __kernel_rt_sigreturn,@function
 	ALIGN
 __kernel_rt_sigreturn:
-.LSTART_rt_sigreturn:
+	STARTPROC_SIGNAL_FRAME IA32_RT_SIGFRAME_sigcontext
 	movl $__NR_rt_sigreturn, %eax
 	int $0x80
-.LEND_rt_sigreturn:
 SYM_INNER_LABEL(vdso32_rt_sigreturn_landing_pad, SYM_L_GLOBAL)
-	nop
-	.size __kernel_rt_sigreturn,.-.LSTART_rt_sigreturn
-	.previous
-
-	.section .eh_frame,"a",@progbits
-.LSTARTFRAMEDLSI1:
-	.long .LENDCIEDLSI1-.LSTARTCIEDLSI1
-.LSTARTCIEDLSI1:
-	.long 0			/* CIE ID */
-	.byte 1			/* Version number */
-	.string "zRS"		/* NUL-terminated augmentation string */
-	.uleb128 1		/* Code alignment factor */
-	.sleb128 -4		/* Data alignment factor */
-	.byte 8			/* Return address register column */
-	.uleb128 1		/* Augmentation value length */
-	.byte 0x1b		/* DW_EH_PE_pcrel|DW_EH_PE_sdata4. */
-	.byte 0			/* DW_CFA_nop */
-	.align 4
-.LENDCIEDLSI1:
-	.long .LENDFDEDLSI1-.LSTARTFDEDLSI1 /* Length FDE */
-.LSTARTFDEDLSI1:
-	.long .LSTARTFDEDLSI1-.LSTARTFRAMEDLSI1 /* CIE pointer */
-	/* HACK: The dwarf2 unwind routines will subtract 1 from the
-	   return address to get an address in the middle of the
-	   presumed call instruction.  Since we didn't get here via
-	   a call, we need to include the nop before the real start
-	   to make up for it.  */
-	.long .LSTART_sigreturn-1-.	/* PC-relative start address */
-	.long .LEND_sigreturn-.LSTART_sigreturn+1
-	.uleb128 0			/* Augmentation */
-	/* What follows are the instructions for the table generation.
-	   We record the locations of each register saved.  This is
-	   complicated by the fact that the "CFA" is always assumed to
-	   be the value of the stack pointer in the caller.  This means
-	   that we must define the CFA of this body of code to be the
-	   saved value of the stack pointer in the sigcontext.  Which
-	   also means that there is no fixed relation to the other
-	   saved registers, which means that we must use DW_CFA_expression
-	   to compute their addresses.  It also means that when we
-	   adjust the stack with the popl, we have to do it all over again.  */
-
-#define do_cfa_expr(offset)						\
-	.byte 0x0f;			/* DW_CFA_def_cfa_expression */	\
-	.uleb128 1f-0f;			/*   length */			\
-0:	.byte 0x74;			/*     DW_OP_breg4 */		\
-	.sleb128 offset;		/*      offset */		\
-	.byte 0x06;			/*     DW_OP_deref */		\
-1:
-
-#define do_expr(regno, offset)						\
-	.byte 0x10;			/* DW_CFA_expression */		\
-	.uleb128 regno;			/*   regno */			\
-	.uleb128 1f-0f;			/*   length */			\
-0:	.byte 0x74;			/*     DW_OP_breg4 */		\
-	.sleb128 offset;		/*       offset */		\
-1:
-
-	do_cfa_expr(IA32_SIGCONTEXT_sp+4)
-	do_expr(0, IA32_SIGCONTEXT_ax+4)
-	do_expr(1, IA32_SIGCONTEXT_cx+4)
-	do_expr(2, IA32_SIGCONTEXT_dx+4)
-	do_expr(3, IA32_SIGCONTEXT_bx+4)
-	do_expr(5, IA32_SIGCONTEXT_bp+4)
-	do_expr(6, IA32_SIGCONTEXT_si+4)
-	do_expr(7, IA32_SIGCONTEXT_di+4)
-	do_expr(8, IA32_SIGCONTEXT_ip+4)
-
-	.byte 0x42	/* DW_CFA_advance_loc 2 -- nop; popl eax. */
-
-	do_cfa_expr(IA32_SIGCONTEXT_sp)
-	do_expr(0, IA32_SIGCONTEXT_ax)
-	do_expr(1, IA32_SIGCONTEXT_cx)
-	do_expr(2, IA32_SIGCONTEXT_dx)
-	do_expr(3, IA32_SIGCONTEXT_bx)
-	do_expr(5, IA32_SIGCONTEXT_bp)
-	do_expr(6, IA32_SIGCONTEXT_si)
-	do_expr(7, IA32_SIGCONTEXT_di)
-	do_expr(8, IA32_SIGCONTEXT_ip)
-
-	.align 4
-.LENDFDEDLSI1:
-
-	.long .LENDFDEDLSI2-.LSTARTFDEDLSI2 /* Length FDE */
-.LSTARTFDEDLSI2:
-	.long .LSTARTFDEDLSI2-.LSTARTFRAMEDLSI1 /* CIE pointer */
-	/* HACK: See above wrt unwind library assumptions.  */
-	.long .LSTART_rt_sigreturn-1-.	/* PC-relative start address */
-	.long .LEND_rt_sigreturn-.LSTART_rt_sigreturn+1
-	.uleb128 0			/* Augmentation */
-	/* What follows are the instructions for the table generation.
-	   We record the locations of each register saved.  This is
-	   slightly less complicated than the above, since we don't
-	   modify the stack pointer in the process.  */
-
-	do_cfa_expr(IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_sp)
-	do_expr(0, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_ax)
-	do_expr(1, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_cx)
-	do_expr(2, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_dx)
-	do_expr(3, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_bx)
-	do_expr(5, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_bp)
-	do_expr(6, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_si)
-	do_expr(7, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_di)
-	do_expr(8, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_ip)
-
-	.align 4
-.LENDFDEDLSI2:
+	ud2a
+	CFI_ENDPROC
+	.size __kernel_rt_sigreturn,.-__kernel_rt_sigreturn
 	.previous
arch/x86/include/asm/dwarf2.h +1 -0
index 302e11b15da8..09c9684d3ad6 100644
@@ -20,6 +20,7 @@
 #define CFI_RESTORE_STATE	.cfi_restore_state
 #define CFI_UNDEFINED		.cfi_undefined
 #define CFI_ESCAPE		.cfi_escape
+#define CFI_SIGNAL_FRAME	.cfi_signal_frame
 
 #ifndef BUILD_VDSO
 	/*
arch/x86/kernel/asm-offsets.c +6 -0
index 25fcde525c68..081816888f7a 100644
@@ -63,8 +63,14 @@ static void __used common(void)
 	OFFSET(IA32_SIGCONTEXT_bp, sigcontext_32, bp);
 	OFFSET(IA32_SIGCONTEXT_sp, sigcontext_32, sp);
 	OFFSET(IA32_SIGCONTEXT_ip, sigcontext_32, ip);
+	OFFSET(IA32_SIGCONTEXT_es, sigcontext_32, es);
+	OFFSET(IA32_SIGCONTEXT_cs, sigcontext_32, cs);
+	OFFSET(IA32_SIGCONTEXT_ss, sigcontext_32, ss);
+	OFFSET(IA32_SIGCONTEXT_ds, sigcontext_32, ds);
+	OFFSET(IA32_SIGCONTEXT_flags, sigcontext_32, flags);
 
 	BLANK();
+	OFFSET(IA32_SIGFRAME_sigcontext,    sigframe_ia32,    sc);
 	OFFSET(IA32_RT_SIGFRAME_sigcontext, rt_sigframe_ia32, uc.uc_mcontext);
 #endif
 
Signature · 1 line
2.52.0
merge bot Applied
#14 lore raw

[tip: x86/entry] x86/entry/vdso32: Remove open-coded DWARF in sigreturn.S

The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     884961618ee51307cc63ab620a0bdd710fa0b0af
Gitweb:        https://git.kernel.org/tip/884961618ee51307cc63ab620a0bdd710fa0b0af
Author:        H. Peter Anvin <[email protected]>
AuthorDate:    Tue, 16 Dec 2025 13:26:00 -08:00
Committer:     Dave Hansen <[email protected]>
CommitterDate: Tue, 13 Jan 2026 16:37:58 -08:00

x86/entry/vdso32: Remove open-coded DWARF in sigreturn.S

The vdso32 sigreturn.S contains open-coded DWARF bytecode, which includes a hack for gdb to not try to step back to a previous call instruction when backtracing from a signal handler.

Neither of those are necessary anymore: the backtracing issue is handled by ".cfi_entry simple" and ".cfi_signal_frame", both of which have been supported for a very long time now, which allows the remaining frame to be built using regular .cfi annotations.

Add a few more register offsets to the signal frame just for good measure.

Replace the nop on fallthrough of the system call (which should never, ever happen) with a ud2a trap.

---
 arch/x86/entry/vdso/vdso32/sigreturn.S | 146 +++++-------------------
 arch/x86/include/asm/dwarf2.h          |   1 +-
 arch/x86/kernel/asm-offsets.c          |   6 +-
 3 files changed, 39 insertions(+), 114 deletions(-)
arch/x86/entry/vdso/vdso32/sigreturn.S +32 -114
index 965900c..25b0ac4 100644
@@ -1,136 +1,54 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 #include <linux/linkage.h>
 #include <asm/unistd_32.h>
+#include <asm/dwarf2.h>
 #include <asm/asm-offsets.h>
 
+.macro STARTPROC_SIGNAL_FRAME sc
+	CFI_STARTPROC	simple
+	CFI_SIGNAL_FRAME
+	/* -4 as pretcode has already been popped */
+	CFI_DEF_CFA	esp,	\sc - 4
+	CFI_OFFSET	eip,    IA32_SIGCONTEXT_ip
+	CFI_OFFSET	eax,    IA32_SIGCONTEXT_ax
+	CFI_OFFSET	ebx,    IA32_SIGCONTEXT_bx
+	CFI_OFFSET	ecx,    IA32_SIGCONTEXT_cx
+	CFI_OFFSET	edx,    IA32_SIGCONTEXT_dx
+	CFI_OFFSET	esp,    IA32_SIGCONTEXT_sp
+	CFI_OFFSET	ebp,    IA32_SIGCONTEXT_bp
+	CFI_OFFSET	esi,    IA32_SIGCONTEXT_si
+	CFI_OFFSET	edi,    IA32_SIGCONTEXT_di
+	CFI_OFFSET	es,     IA32_SIGCONTEXT_es
+	CFI_OFFSET	cs,     IA32_SIGCONTEXT_cs
+	CFI_OFFSET	ss,     IA32_SIGCONTEXT_ss
+	CFI_OFFSET	ds,     IA32_SIGCONTEXT_ds
+	CFI_OFFSET	eflags, IA32_SIGCONTEXT_flags
+.endm
+
 	.text
 	.globl __kernel_sigreturn
 	.type __kernel_sigreturn,@function
-	nop /* this guy is needed for .LSTARTFDEDLSI1 below (watch for HACK) */
 	ALIGN
 __kernel_sigreturn:
-.LSTART_sigreturn:
-	popl %eax		/* XXX does this mean it needs unwind info? */
+	STARTPROC_SIGNAL_FRAME IA32_SIGFRAME_sigcontext
+	popl %eax
+	CFI_ADJUST_CFA_OFFSET -4
 	movl $__NR_sigreturn, %eax
 	int $0x80
-.LEND_sigreturn:
 SYM_INNER_LABEL(vdso32_sigreturn_landing_pad, SYM_L_GLOBAL)
-	nop
-	.size __kernel_sigreturn,.-.LSTART_sigreturn
+	ud2a
+	CFI_ENDPROC
+	.size __kernel_sigreturn,.-__kernel_sigreturn
 
 	.globl __kernel_rt_sigreturn
 	.type __kernel_rt_sigreturn,@function
 	ALIGN
 __kernel_rt_sigreturn:
-.LSTART_rt_sigreturn:
+	STARTPROC_SIGNAL_FRAME IA32_RT_SIGFRAME_sigcontext
 	movl $__NR_rt_sigreturn, %eax
 	int $0x80
-.LEND_rt_sigreturn:
 SYM_INNER_LABEL(vdso32_rt_sigreturn_landing_pad, SYM_L_GLOBAL)
-	nop
-	.size __kernel_rt_sigreturn,.-.LSTART_rt_sigreturn
-	.previous
-
-	.section .eh_frame,"a",@progbits
-.LSTARTFRAMEDLSI1:
-	.long .LENDCIEDLSI1-.LSTARTCIEDLSI1
-.LSTARTCIEDLSI1:
-	.long 0			/* CIE ID */
-	.byte 1			/* Version number */
-	.string "zRS"		/* NUL-terminated augmentation string */
-	.uleb128 1		/* Code alignment factor */
-	.sleb128 -4		/* Data alignment factor */
-	.byte 8			/* Return address register column */
-	.uleb128 1		/* Augmentation value length */
-	.byte 0x1b		/* DW_EH_PE_pcrel|DW_EH_PE_sdata4. */
-	.byte 0			/* DW_CFA_nop */
-	.align 4
-.LENDCIEDLSI1:
-	.long .LENDFDEDLSI1-.LSTARTFDEDLSI1 /* Length FDE */
-.LSTARTFDEDLSI1:
-	.long .LSTARTFDEDLSI1-.LSTARTFRAMEDLSI1 /* CIE pointer */
-	/* HACK: The dwarf2 unwind routines will subtract 1 from the
-	   return address to get an address in the middle of the
-	   presumed call instruction.  Since we didn't get here via
-	   a call, we need to include the nop before the real start
-	   to make up for it.  */
-	.long .LSTART_sigreturn-1-.	/* PC-relative start address */
-	.long .LEND_sigreturn-.LSTART_sigreturn+1
-	.uleb128 0			/* Augmentation */
-	/* What follows are the instructions for the table generation.
-	   We record the locations of each register saved.  This is
-	   complicated by the fact that the "CFA" is always assumed to
-	   be the value of the stack pointer in the caller.  This means
-	   that we must define the CFA of this body of code to be the
-	   saved value of the stack pointer in the sigcontext.  Which
-	   also means that there is no fixed relation to the other
-	   saved registers, which means that we must use DW_CFA_expression
-	   to compute their addresses.  It also means that when we
-	   adjust the stack with the popl, we have to do it all over again.  */
-
-#define do_cfa_expr(offset)						\
-	.byte 0x0f;			/* DW_CFA_def_cfa_expression */	\
-	.uleb128 1f-0f;			/*   length */			\
-0:	.byte 0x74;			/*     DW_OP_breg4 */		\
-	.sleb128 offset;		/*      offset */		\
-	.byte 0x06;			/*     DW_OP_deref */		\
-1:
-
-#define do_expr(regno, offset)						\
-	.byte 0x10;			/* DW_CFA_expression */		\
-	.uleb128 regno;			/*   regno */			\
-	.uleb128 1f-0f;			/*   length */			\
-0:	.byte 0x74;			/*     DW_OP_breg4 */		\
-	.sleb128 offset;		/*       offset */		\
-1:
-
-	do_cfa_expr(IA32_SIGCONTEXT_sp+4)
-	do_expr(0, IA32_SIGCONTEXT_ax+4)
-	do_expr(1, IA32_SIGCONTEXT_cx+4)
-	do_expr(2, IA32_SIGCONTEXT_dx+4)
-	do_expr(3, IA32_SIGCONTEXT_bx+4)
-	do_expr(5, IA32_SIGCONTEXT_bp+4)
-	do_expr(6, IA32_SIGCONTEXT_si+4)
-	do_expr(7, IA32_SIGCONTEXT_di+4)
-	do_expr(8, IA32_SIGCONTEXT_ip+4)
-
-	.byte 0x42	/* DW_CFA_advance_loc 2 -- nop; popl eax. */
-
-	do_cfa_expr(IA32_SIGCONTEXT_sp)
-	do_expr(0, IA32_SIGCONTEXT_ax)
-	do_expr(1, IA32_SIGCONTEXT_cx)
-	do_expr(2, IA32_SIGCONTEXT_dx)
-	do_expr(3, IA32_SIGCONTEXT_bx)
-	do_expr(5, IA32_SIGCONTEXT_bp)
-	do_expr(6, IA32_SIGCONTEXT_si)
-	do_expr(7, IA32_SIGCONTEXT_di)
-	do_expr(8, IA32_SIGCONTEXT_ip)
-
-	.align 4
-.LENDFDEDLSI1:
-
-	.long .LENDFDEDLSI2-.LSTARTFDEDLSI2 /* Length FDE */
-.LSTARTFDEDLSI2:
-	.long .LSTARTFDEDLSI2-.LSTARTFRAMEDLSI1 /* CIE pointer */
-	/* HACK: See above wrt unwind library assumptions.  */
-	.long .LSTART_rt_sigreturn-1-.	/* PC-relative start address */
-	.long .LEND_rt_sigreturn-.LSTART_rt_sigreturn+1
-	.uleb128 0			/* Augmentation */
-	/* What follows are the instructions for the table generation.
-	   We record the locations of each register saved.  This is
-	   slightly less complicated than the above, since we don't
-	   modify the stack pointer in the process.  */
-
-	do_cfa_expr(IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_sp)
-	do_expr(0, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_ax)
-	do_expr(1, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_cx)
-	do_expr(2, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_dx)
-	do_expr(3, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_bx)
-	do_expr(5, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_bp)
-	do_expr(6, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_si)
-	do_expr(7, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_di)
-	do_expr(8, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_ip)
-
-	.align 4
-.LENDFDEDLSI2:
+	ud2a
+	CFI_ENDPROC
+	.size __kernel_rt_sigreturn,.-__kernel_rt_sigreturn
 	.previous
arch/x86/include/asm/dwarf2.h +1 -0
index 302e11b..09c9684 100644
@@ -20,6 +20,7 @@
 #define CFI_RESTORE_STATE	.cfi_restore_state
 #define CFI_UNDEFINED		.cfi_undefined
 #define CFI_ESCAPE		.cfi_escape
+#define CFI_SIGNAL_FRAME	.cfi_signal_frame
 
 #ifndef BUILD_VDSO
 	/*
arch/x86/kernel/asm-offsets.c +6 -0
index 25fcde5..0818168 100644
@@ -63,8 +63,14 @@ static void __used common(void)
 	OFFSET(IA32_SIGCONTEXT_bp, sigcontext_32, bp);
 	OFFSET(IA32_SIGCONTEXT_sp, sigcontext_32, sp);
 	OFFSET(IA32_SIGCONTEXT_ip, sigcontext_32, ip);
+	OFFSET(IA32_SIGCONTEXT_es, sigcontext_32, es);
+	OFFSET(IA32_SIGCONTEXT_cs, sigcontext_32, cs);
+	OFFSET(IA32_SIGCONTEXT_ss, sigcontext_32, ss);
+	OFFSET(IA32_SIGCONTEXT_ds, sigcontext_32, ds);
+	OFFSET(IA32_SIGCONTEXT_flags, sigcontext_32, flags);
 
 	BLANK();
+	OFFSET(IA32_SIGFRAME_sigcontext,    sigframe_ia32,    sc);
 	OFFSET(IA32_RT_SIGFRAME_sigcontext, rt_sigframe_ia32, uc.uc_mcontext);
 #endif
 
author maintainer
#15 lore raw

PATCH v4 7/10 x86/entry/vdso: include GNU_PROPERTY and GNU_STACK PHDRs

In reply to H. Peter Anvin

Currently the vdso doesn't include .note.gnu.property or a GNU noexec stack annotation (the -z noexecstack in the linker script is ineffective because we specify PHDRs explicitly.)

The motivation is that the dynamic linker currently do not check these.

However, this is a weak excuse: the vdso*.so are also supposed to be usable at link libraries, and there is no reason why the dynamic linker might not want or need to check these in the future, so add them back in -- it is trivial enough.

Use symbolic constants for the PHDR permission flags.

[ v4: drop unrelated formatting changes ]

---
 arch/x86/entry/vdso/common/vdso-layout.lds.S | 38 ++++++++++++--------
 1 file changed, 23 insertions(+), 15 deletions(-)
arch/x86/entry/vdso/common/vdso-layout.lds.S +23 -15
index ec1ac191a057..a1e30be3e83d 100644
@@ -47,18 +47,18 @@ SECTIONS
 		*(.gnu.linkonce.b.*)
 	}						:text
 
-	/*
-	 * Discard .note.gnu.property sections which are unused and have
-	 * different alignment requirement from vDSO note sections.
-	 */
-	/DISCARD/ : {
+	.note.gnu.property : {
 		*(.note.gnu.property)
-	}
-	.note		: { *(.note.*) }		:text	:note
-
-	.eh_frame_hdr	: { *(.eh_frame_hdr) }		:text	:eh_frame_hdr
-	.eh_frame	: { KEEP (*(.eh_frame)) }	:text
+	}					:text :note :gnu_property
+	.note		: {
+		*(.note*)
+	}					:text :note
 
+	.eh_frame_hdr	: { *(.eh_frame_hdr) }	:text :eh_frame_hdr
+	.eh_frame	: {
+		KEEP (*(.eh_frame))
+		*(.eh_frame.*)
+	}					:text
 
 	/*
 	 * Text is well-separated from actual data: there's plenty of
@@ -87,15 +87,23 @@ SECTIONS
  * Very old versions of ld do not recognize this name token; use the constant.
  */
 #define PT_GNU_EH_FRAME	0x6474e550
+#define PT_GNU_STACK	0x6474e551
+#define PT_GNU_PROPERTY	0x6474e553
 
 /*
  * We must supply the ELF program headers explicitly to get just one
  * PT_LOAD segment, and set the flags explicitly to make segments read-only.
- */
+*/
+#define PF_R	FLAGS(4)
+#define PF_RW	FLAGS(6)
+#define PF_RX	FLAGS(5)
+
 PHDRS
 {
-	text		PT_LOAD		FLAGS(5) FILEHDR PHDRS; /* PF_R|PF_X */
-	dynamic		PT_DYNAMIC	FLAGS(4);		/* PF_R */
-	note		PT_NOTE		FLAGS(4);		/* PF_R */
-	eh_frame_hdr	PT_GNU_EH_FRAME;
+	text		PT_LOAD		PF_RX FILEHDR PHDRS;
+	dynamic		PT_DYNAMIC	PF_R;
+	note		PT_NOTE		PF_R;
+	eh_frame_hdr	PT_GNU_EH_FRAME PF_R;
+	gnu_stack	PT_GNU_STACK	PF_RW;
+	gnu_property	PT_GNU_PROPERTY	PF_R;
 }
Signature · 1 line
2.52.0
#16 lore raw

Re: [PATCH v4 07/10] x86/entry/vdso: include GNU_PROPERTY and GNU_STACK PHDRs

Quoted 56 lines · H. Peter Anvin

Currently the vdso doesn't include .note.gnu.property or a GNU noexec stack annotation (the -z noexecstack in the linker script is ineffective because we specify PHDRs explicitly.)

The motivation is that the dynamic linker currently do not check these.

However, this is a weak excuse: the vdso*.so are also supposed to be usable at link libraries, and there is no reason why the dynamic linker might not want or need to check these in the future, so add them back in -- it is trivial enough.

Use symbolic constants for the PHDR permission flags.

[ v4: drop unrelated formatting changes ]

---
 arch/x86/entry/vdso/common/vdso-layout.lds.S | 38 ++++++++++++--------
 1 file changed, 23 insertions(+), 15 deletions(-)
arch/x86/entry/vdso/common/vdso-layout.lds.S +10 -10
index ec1ac191a057..a1e30be3e83d 100644
@@ -47,18 +47,18 @@ SECTIONS
                *(.gnu.linkonce.b.*)
        }                                               :text
 
-       /*
-        * Discard .note.gnu.property sections which are unused and have
-        * different alignment requirement from vDSO note sections.
-        */
-       /DISCARD/ : {
+       .note.gnu.property : {
                *(.note.gnu.property)
-       }
-       .note           : { *(.note.*) }                :text   :note
-
-       .eh_frame_hdr   : { *(.eh_frame_hdr) }          :text   :eh_frame_hdr
-       .eh_frame       : { KEEP (*(.eh_frame)) }       :text
+       }                                       :text :note :gnu_property
+       .note           : {
+               *(.note*)
+       }                                       :text :note
 
+       .eh_frame_hdr   : { *(.eh_frame_hdr) }  :text :eh_frame_hdr
+       .eh_frame       : {
+               KEEP (*(.eh_frame))
+               *(.eh_frame.*)
+       }                                       :text
 
        /*
         * Text is well-separated from actual data: there's plenty of

arch/x86/entry/vdso/common/vdso-layout.lds.S SECTIONS · line 91

Quoted 5 lines
arch/x86/entry/vdso/common/vdso-layout.lds.S +2 -0
@@ -87,15 +87,23 @@ SECTIONS
  * Very old versions of ld do not recognize this name token; use the constant.
  */
 #define PT_GNU_EH_FRAME        0x6474e550
+#define PT_GNU_STACK   0x6474e551
+#define PT_GNU_PROPERTY        0x6474e553

Do we even still support the old linkers that need these constants?

Brian Gerst

#17 lore raw

Re: [PATCH v4 07/10] x86/entry/vdso: include GNU_PROPERTY and GNU_STACK PHDRs

Quoted 55 lines · Brian Gerst
Quoted 53 lines · H. Peter Anvin

Currently the vdso doesn't include .note.gnu.property or a GNU noexec stack annotation (the -z noexecstack in the linker script is ineffective because we specify PHDRs explicitly.)

The motivation is that the dynamic linker currently do not check these.

However, this is a weak excuse: the vdso*.so are also supposed to be usable at link libraries, and there is no reason why the dynamic linker might not want or need to check these in the future, so add them back in -- it is trivial enough.

Use symbolic constants for the PHDR permission flags.

[ v4: drop unrelated formatting changes ]

---
 arch/x86/entry/vdso/common/vdso-layout.lds.S | 38 ++++++++++++--------
 1 file changed, 23 insertions(+), 15 deletions(-)
arch/x86/entry/vdso/common/vdso-layout.lds.S +10 -10
index ec1ac191a057..a1e30be3e83d 100644
@@ -47,18 +47,18 @@ SECTIONS
                *(.gnu.linkonce.b.*)
        }                                               :text
 
-       /*
-        * Discard .note.gnu.property sections which are unused and have
-        * different alignment requirement from vDSO note sections.
-        */
-       /DISCARD/ : {
+       .note.gnu.property : {
                *(.note.gnu.property)
-       }
-       .note           : { *(.note.*) }                :text   :note
-
-       .eh_frame_hdr   : { *(.eh_frame_hdr) }          :text   :eh_frame_hdr
-       .eh_frame       : { KEEP (*(.eh_frame)) }       :text
+       }                                       :text :note :gnu_property
+       .note           : {
+               *(.note*)
+       }                                       :text :note
 
+       .eh_frame_hdr   : { *(.eh_frame_hdr) }  :text :eh_frame_hdr
+       .eh_frame       : {
+               KEEP (*(.eh_frame))
+               *(.eh_frame.*)
+       }                                       :text
 
Quoted 10 lines

arch/x86/entry/vdso/common/vdso-layout.lds.S SECTIONS · line 91

Quoted 8 lines
arch/x86/entry/vdso/common/vdso-layout.lds.S +2 -0
        /*
         * Text is well-separated from actual data: there's plenty of
@@ -87,15 +87,23 @@ SECTIONS
  * Very old versions of ld do not recognize this name token; use the constant.
  */
 #define PT_GNU_EH_FRAME        0x6474e550
+#define PT_GNU_STACK   0x6474e551
+#define PT_GNU_PROPERTY        0x6474e553

Do we even still support the old linkers that need these constants?

Digging into the binutils source, PT_GNU_EH_FRAME and PT_GNU_STACK were added to the parser around bintils-2.15. PT_GNU_PROPERTY was added in binutils-2.38, which is newer than the minimum supported version of binutils-2.30.

Probably better to just leave them then.

Brian Gerst

merge bot Applied
#18 lore raw

[tip: x86/entry] x86/entry/vdso: Include GNU_PROPERTY and GNU_STACK PHDRs

In reply to H. Peter Anvin
The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     8717b02b8c030dc0c4b55781b59e88def0a1a92f
Gitweb:        https://git.kernel.org/tip/8717b02b8c030dc0c4b55781b59e88def0a1a92f
Author:        H. Peter Anvin <[email protected]>
AuthorDate:    Tue, 16 Dec 2025 13:26:01 -08:00
Committer:     Dave Hansen <[email protected]>
CommitterDate: Tue, 13 Jan 2026 16:37:58 -08:00

x86/entry/vdso: Include GNU_PROPERTY and GNU_STACK PHDRs

Currently the vdso doesn't include .note.gnu.property or a GNU noexec stack annotation (the -z noexecstack in the linker script is ineffective because we specify PHDRs explicitly.)

The motivation is that the dynamic linker currently do not check these.

However, this is a weak excuse: the vdso*.so are also supposed to be usable at link libraries, and there is no reason why the dynamic linker might not want or need to check these in the future, so add them back in -- it is trivial enough.

Use symbolic constants for the PHDR permission flags.

[ v4: drop unrelated formatting changes ]

---
 arch/x86/entry/vdso/common/vdso-layout.lds.S | 38 +++++++++++--------
 1 file changed, 23 insertions(+), 15 deletions(-)
arch/x86/entry/vdso/common/vdso-layout.lds.S +23 -15
index ec1ac19..a1e30be 100644
@@ -47,18 +47,18 @@ SECTIONS
 		*(.gnu.linkonce.b.*)
 	}						:text
 
-	/*
-	 * Discard .note.gnu.property sections which are unused and have
-	 * different alignment requirement from vDSO note sections.
-	 */
-	/DISCARD/ : {
+	.note.gnu.property : {
 		*(.note.gnu.property)
-	}
-	.note		: { *(.note.*) }		:text	:note
-
-	.eh_frame_hdr	: { *(.eh_frame_hdr) }		:text	:eh_frame_hdr
-	.eh_frame	: { KEEP (*(.eh_frame)) }	:text
+	}					:text :note :gnu_property
+	.note		: {
+		*(.note*)
+	}					:text :note
 
+	.eh_frame_hdr	: { *(.eh_frame_hdr) }	:text :eh_frame_hdr
+	.eh_frame	: {
+		KEEP (*(.eh_frame))
+		*(.eh_frame.*)
+	}					:text
 
 	/*
 	 * Text is well-separated from actual data: there's plenty of
@@ -87,15 +87,23 @@ SECTIONS
  * Very old versions of ld do not recognize this name token; use the constant.
  */
 #define PT_GNU_EH_FRAME	0x6474e550
+#define PT_GNU_STACK	0x6474e551
+#define PT_GNU_PROPERTY	0x6474e553
 
 /*
  * We must supply the ELF program headers explicitly to get just one
  * PT_LOAD segment, and set the flags explicitly to make segments read-only.
- */
+*/
+#define PF_R	FLAGS(4)
+#define PF_RW	FLAGS(6)
+#define PF_RX	FLAGS(5)
+
 PHDRS
 {
-	text		PT_LOAD		FLAGS(5) FILEHDR PHDRS; /* PF_R|PF_X */
-	dynamic		PT_DYNAMIC	FLAGS(4);		/* PF_R */
-	note		PT_NOTE		FLAGS(4);		/* PF_R */
-	eh_frame_hdr	PT_GNU_EH_FRAME;
+	text		PT_LOAD		PF_RX FILEHDR PHDRS;
+	dynamic		PT_DYNAMIC	PF_R;
+	note		PT_NOTE		PF_R;
+	eh_frame_hdr	PT_GNU_EH_FRAME PF_R;
+	gnu_stack	PT_GNU_STACK	PF_RW;
+	gnu_property	PT_GNU_PROPERTY	PF_R;
 }
author maintainer
#19 lore raw

PATCH v4 8/10 x86/vdso: abstract out vdso system call internals

In reply to H. Peter Anvin

Abstract out the calling of true system calls from the vdso into macros.

It has been a very long time since gcc did not allow %ebx or %ebp in inline asm in 32-bit PIC mode; remove the corresponding hacks.

Remove the use of memory output constraints in gettimeofday.h in favor of "memory" clobbers. The resulting code is identical for the current use cases, as the system call is usually a terminal fallback anyway, and it merely complicates the macroization.

This patch adds only a handful of more lines of code than it removes, and in fact could be made substantially smaller by removing the macros for the argument counts that aren't currently used, however, it seems better to be general from the start.

[ v3: remove stray comment from prototyping; remove VDSO_SYSCALL6()
since it would require special handling on 32 bits and is currently unused. (Uros Biszjak)

Indent nested preprocessor directives. ]

---
 arch/x86/include/asm/vdso/gettimeofday.h | 108 ++---------------------
 arch/x86/include/asm/vdso/sys_call.h     | 103 +++++++++++++++++++++
 2 files changed, 111 insertions(+), 100 deletions(-)
 create mode 100644 arch/x86/include/asm/vdso/sys_call.h
arch/x86/include/asm/vdso/gettimeofday.h +8 -100
index 73b2e7ee8f0f..3cf214cc4a75 100644
@@ -18,6 +18,7 @@
 #include <asm/msr.h>
 #include <asm/pvclock.h>
 #include <clocksource/hyperv_timer.h>
+#include <asm/vdso/sys_call.h>
 
 #define VDSO_HAS_TIME 1
 
@@ -53,130 +54,37 @@ extern struct ms_hyperv_tsc_page hvclock_page
 	__attribute__((visibility("hidden")));
 #endif
 
-#ifndef BUILD_VDSO32
-
 static __always_inline
 long clock_gettime_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
 {
-	long ret;
-
-	asm ("syscall" : "=a" (ret), "=m" (*_ts) :
-	     "0" (__NR_clock_gettime), "D" (_clkid), "S" (_ts) :
-	     "rcx", "r11");
-
-	return ret;
+	return VDSO_SYSCALL2(clock_gettime,64,_clkid,_ts);
 }
 
 static __always_inline
 long gettimeofday_fallback(struct __kernel_old_timeval *_tv,
 			   struct timezone *_tz)
 {
-	long ret;
-
-	asm("syscall" : "=a" (ret) :
-	    "0" (__NR_gettimeofday), "D" (_tv), "S" (_tz) : "memory");
-
-	return ret;
+	return VDSO_SYSCALL2(gettimeofday,,_tv,_tz);
 }
 
 static __always_inline
 long clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
 {
-	long ret;
-
-	asm ("syscall" : "=a" (ret), "=m" (*_ts) :
-	     "0" (__NR_clock_getres), "D" (_clkid), "S" (_ts) :
-	     "rcx", "r11");
-
-	return ret;
+	return VDSO_SYSCALL2(clock_getres,_time64,_clkid,_ts);
 }
 
-#else
-
-static __always_inline
-long clock_gettime_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
-{
-	long ret;
-
-	asm (
-		"mov %%ebx, %%edx \n"
-		"mov %[clock], %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret), "=m" (*_ts)
-		: "0" (__NR_clock_gettime64), [clock] "g" (_clkid), "c" (_ts)
-		: "edx");
-
-	return ret;
-}
+#ifndef CONFIG_X86_64
 
 static __always_inline
 long clock_gettime32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)
 {
-	long ret;
-
-	asm (
-		"mov %%ebx, %%edx \n"
-		"mov %[clock], %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret), "=m" (*_ts)
-		: "0" (__NR_clock_gettime), [clock] "g" (_clkid), "c" (_ts)
-		: "edx");
-
-	return ret;
-}
-
-static __always_inline
-long gettimeofday_fallback(struct __kernel_old_timeval *_tv,
-			   struct timezone *_tz)
-{
-	long ret;
-
-	asm(
-		"mov %%ebx, %%edx \n"
-		"mov %2, %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret)
-		: "0" (__NR_gettimeofday), "g" (_tv), "c" (_tz)
-		: "memory", "edx");
-
-	return ret;
+	return VDSO_SYSCALL2(clock_gettime,,_clkid,_ts);
 }
 
 static __always_inline long
-clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
-{
-	long ret;
-
-	asm (
-		"mov %%ebx, %%edx \n"
-		"mov %[clock], %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret), "=m" (*_ts)
-		: "0" (__NR_clock_getres_time64), [clock] "g" (_clkid), "c" (_ts)
-		: "edx");
-
-	return ret;
-}
-
-static __always_inline
-long clock_getres32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)
+clock_getres32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)
 {
-	long ret;
-
-	asm (
-		"mov %%ebx, %%edx \n"
-		"mov %[clock], %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret), "=m" (*_ts)
-		: "0" (__NR_clock_getres), [clock] "g" (_clkid), "c" (_ts)
-		: "edx");
-
-	return ret;
+	return VDSO_SYSCALL2(clock_getres,,_clkid,_ts);
 }
 
 #endif
arch/x86/include/asm/vdso/sys_call.h new file +103 -0

103 lines not on this page · raw message

Signature · 1 line
2.52.0
Acked-by on v4 8/10
#20 lore raw

Re: [PATCH v4 08/10] x86/vdso: abstract out vdso system call internals

Quoted 17 lines · H. Peter Anvin

Abstract out the calling of true system calls from the vdso into macros.

It has been a very long time since gcc did not allow %ebx or %ebp in inline asm in 32-bit PIC mode; remove the corresponding hacks.

Remove the use of memory output constraints in gettimeofday.h in favor of "memory" clobbers. The resulting code is identical for the current use cases, as the system call is usually a terminal fallback anyway, and it merely complicates the macroization.

This patch adds only a handful of more lines of code than it removes, and in fact could be made substantially smaller by removing the macros for the argument counts that aren't currently used, however, it seems better to be general from the start.

Quoted 7 lines

[ v3: remove stray comment from prototyping; remove VDSO_SYSCALL6()
since it would require special handling on 32 bits and is currently unused. (Uros Biszjak)

Indent nested preprocessor directives. ]

Quoted 268 lines
---
 arch/x86/include/asm/vdso/gettimeofday.h | 108 ++---------------------
 arch/x86/include/asm/vdso/sys_call.h     | 103 +++++++++++++++++++++
 2 files changed, 111 insertions(+), 100 deletions(-)
 create mode 100644 arch/x86/include/asm/vdso/sys_call.h
arch/x86/include/asm/vdso/gettimeofday.h +8 -100
index 73b2e7ee8f0f..3cf214cc4a75 100644
@@ -18,6 +18,7 @@
 #include <asm/msr.h>
 #include <asm/pvclock.h>
 #include <clocksource/hyperv_timer.h>
+#include <asm/vdso/sys_call.h>
 
 #define VDSO_HAS_TIME 1
 
@@ -53,130 +54,37 @@ extern struct ms_hyperv_tsc_page hvclock_page
        __attribute__((visibility("hidden")));
 #endif
 
-#ifndef BUILD_VDSO32
-
 static __always_inline
 long clock_gettime_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
 {
-       long ret;
-
-       asm ("syscall" : "=a" (ret), "=m" (*_ts) :
-            "0" (__NR_clock_gettime), "D" (_clkid), "S" (_ts) :
-            "rcx", "r11");
-
-       return ret;
+       return VDSO_SYSCALL2(clock_gettime,64,_clkid,_ts);
 }
 
 static __always_inline
 long gettimeofday_fallback(struct __kernel_old_timeval *_tv,
                           struct timezone *_tz)
 {
-       long ret;
-
-       asm("syscall" : "=a" (ret) :
-           "0" (__NR_gettimeofday), "D" (_tv), "S" (_tz) : "memory");
-
-       return ret;
+       return VDSO_SYSCALL2(gettimeofday,,_tv,_tz);
 }
 
 static __always_inline
 long clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
 {
-       long ret;
-
-       asm ("syscall" : "=a" (ret), "=m" (*_ts) :
-            "0" (__NR_clock_getres), "D" (_clkid), "S" (_ts) :
-            "rcx", "r11");
-
-       return ret;
+       return VDSO_SYSCALL2(clock_getres,_time64,_clkid,_ts);
 }
 
-#else
-
-static __always_inline
-long clock_gettime_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
-{
-       long ret;
-
-       asm (
-               "mov %%ebx, %%edx \n"
-               "mov %[clock], %%ebx \n"
-               "call __kernel_vsyscall \n"
-               "mov %%edx, %%ebx \n"
-               : "=a" (ret), "=m" (*_ts)
-               : "0" (__NR_clock_gettime64), [clock] "g" (_clkid), "c" (_ts)
-               : "edx");
-
-       return ret;
-}
+#ifndef CONFIG_X86_64
 
 static __always_inline
 long clock_gettime32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)
 {
-       long ret;
-
-       asm (
-               "mov %%ebx, %%edx \n"
-               "mov %[clock], %%ebx \n"
-               "call __kernel_vsyscall \n"
-               "mov %%edx, %%ebx \n"
-               : "=a" (ret), "=m" (*_ts)
-               : "0" (__NR_clock_gettime), [clock] "g" (_clkid), "c" (_ts)
-               : "edx");
-
-       return ret;
-}
-
-static __always_inline
-long gettimeofday_fallback(struct __kernel_old_timeval *_tv,
-                          struct timezone *_tz)
-{
-       long ret;
-
-       asm(
-               "mov %%ebx, %%edx \n"
-               "mov %2, %%ebx \n"
-               "call __kernel_vsyscall \n"
-               "mov %%edx, %%ebx \n"
-               : "=a" (ret)
-               : "0" (__NR_gettimeofday), "g" (_tv), "c" (_tz)
-               : "memory", "edx");
-
-       return ret;
+       return VDSO_SYSCALL2(clock_gettime,,_clkid,_ts);
 }
 
 static __always_inline long
-clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
-{
-       long ret;
-
-       asm (
-               "mov %%ebx, %%edx \n"
-               "mov %[clock], %%ebx \n"
-               "call __kernel_vsyscall \n"
-               "mov %%edx, %%ebx \n"
-               : "=a" (ret), "=m" (*_ts)
-               : "0" (__NR_clock_getres_time64), [clock] "g" (_clkid), "c" (_ts)
-               : "edx");
-
-       return ret;
-}
-
-static __always_inline
-long clock_getres32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)
+clock_getres32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)
 {
-       long ret;
-
-       asm (
-               "mov %%ebx, %%edx \n"
-               "mov %[clock], %%ebx \n"
-               "call __kernel_vsyscall \n"
-               "mov %%edx, %%ebx \n"
-               : "=a" (ret), "=m" (*_ts)
-               : "0" (__NR_clock_getres), [clock] "g" (_clkid), "c" (_ts)
-               : "edx");
-
-       return ret;
+       return VDSO_SYSCALL2(clock_getres,,_clkid,_ts);
 }
 
 #endif
arch/x86/include/asm/vdso/sys_call.h new file +103 -1
new file mode 100644
index 000000000000..dcfd17c6dd57
@@ -0,0 +1,103 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Macros for issuing an inline system call from the vDSO.
+ */
+
+#ifndef X86_ASM_VDSO_SYS_CALL_H
+#define X86_ASM_VDSO_SYS_CALL_H
+
+#include <linux/compiler.h>
+#include <asm/cpufeatures.h>
+#include <asm/alternative.h>
+
+#ifdef CONFIG_X86_64
+# define __sys_instr   "syscall"
+# define __sys_clobber "rcx", "r11", "memory"
+# define __sys_nr(x,y) __NR_ ## x
+# define __sys_reg1    "rdi"
+# define __sys_reg2    "rsi"
+# define __sys_reg3    "rdx"
+# define __sys_reg4    "r10"
+# define __sys_reg5    "r8"
+#else
+# define __sys_instr   "call __kernel_vsyscall"
+# define __sys_clobber "memory"
+# define __sys_nr(x,y) __NR_ ## x ## y
+# define __sys_reg1    "ebx"
+# define __sys_reg2    "ecx"
+# define __sys_reg3    "edx"
+# define __sys_reg4    "esi"
+# define __sys_reg5    "edi"
+#endif
+
+/*
+ * Example usage:
+ *
+ * result = VDSO_SYSCALL3(foo,64,x,y,z);
+ *
+ * ... calls foo(x,y,z) on 64 bits, and foo64(x,y,z) on 32 bits.
+ *
+ * VDSO_SYSCALL6() is currently missing, because it would require
+ * special handling for %ebp on 32 bits when the vdso is compiled with
+ * frame pointers enabled (the default on 32 bits.) Add it as a special
+ * case when and if it becomes necessary.
+ */
+#define _VDSO_SYSCALL(name,suf32,...)                                  \
+       ({                                                              \
+               long _sys_num_ret = __sys_nr(name,suf32);               \
+               asm_inline volatile(                                    \
+                       __sys_instr                                     \
+                       : "+a" (_sys_num_ret)                           \
+                       : __VA_ARGS__                                   \
+                       : __sys_clobber);                               \
+               _sys_num_ret;                                           \
+       })
+
+#define VDSO_SYSCALL0(name,suf32)                                      \
+       _VDSO_SYSCALL(name,suf32)
+#define VDSO_SYSCALL1(name,suf32,a1)                                   \
+       ({                                                              \
+               register long _sys_arg1 asm(__sys_reg1) = (long)(a1);   \
+               _VDSO_SYSCALL(name,suf32,                               \
+                             "r" (_sys_arg1));                         \
+       })
+#define VDSO_SYSCALL2(name,suf32,a1,a2)                                \
+       ({                                                              \
+               register long _sys_arg1 asm(__sys_reg1) = (long)(a1);   \
+               register long _sys_arg2 asm(__sys_reg2) = (long)(a2);   \
+               _VDSO_SYSCALL(name,suf32,                               \
+                             "r" (_sys_arg1), "r" (_sys_arg2));        \
+       })
+#define VDSO_SYSCALL3(name,suf32,a1,a2,a3)                             \
+       ({                                                              \
+               register long _sys_arg1 asm(__sys_reg1) = (long)(a1);   \
+               register long _sys_arg2 asm(__sys_reg2) = (long)(a2);   \
+               register long _sys_arg3 asm(__sys_reg3) = (long)(a3);   \
+               _VDSO_SYSCALL(name,suf32,                               \
+                             "r" (_sys_arg1), "r" (_sys_arg2),         \
+                             "r" (_sys_arg3));                         \
+       })
+#define VDSO_SYSCALL4(name,suf32,a1,a2,a3,a4)                          \
+       ({                                                              \
+               register long _sys_arg1 asm(__sys_reg1) = (long)(a1);   \
+               register long _sys_arg2 asm(__sys_reg2) = (long)(a2);   \
+               register long _sys_arg3 asm(__sys_reg3) = (long)(a3);   \
+               register long _sys_arg4 asm(__sys_reg4) = (long)(a4);   \
+               _VDSO_SYSCALL(name,suf32,                               \
+                             "r" (_sys_arg1), "r" (_sys_arg2),         \
+                             "r" (_sys_arg3), "r" (_sys_arg4));        \
+       })
+#define VDSO_SYSCALL5(name,suf32,a1,a2,a3,a4,a5)                       \
+       ({                                                              \
+               register long _sys_arg1 asm(__sys_reg1) = (long)(a1);   \
+               register long _sys_arg2 asm(__sys_reg2) = (long)(a2);   \
+               register long _sys_arg3 asm(__sys_reg3) = (long)(a3);   \
+               register long _sys_arg4 asm(__sys_reg4) = (long)(a4);   \
+               register long _sys_arg5 asm(__sys_reg5) = (long)(a5);   \
+               _VDSO_SYSCALL(name,suf32,                               \
+                             "r" (_sys_arg1), "r" (_sys_arg2),         \
+                             "r" (_sys_arg3), "r" (_sys_arg4),         \
+                             "r" (_sys_arg5));                         \
+       })
+
+#endif /* X86_VDSO_SYS_CALL_H */
--

2.52.0

merge bot Applied
#21 lore raw

[tip: x86/entry] x86/vdso: Abstract out vdso system call internals

In reply to H. Peter Anvin
The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     a0636d4c3ad0da0cd6069eb6fef5d2b7d3449378
Gitweb:        https://git.kernel.org/tip/a0636d4c3ad0da0cd6069eb6fef5d2b7d3449378
Author:        H. Peter Anvin <[email protected]>
AuthorDate:    Tue, 16 Dec 2025 13:26:02 -08:00
Committer:     Dave Hansen <[email protected]>
CommitterDate: Tue, 13 Jan 2026 16:37:58 -08:00

x86/vdso: Abstract out vdso system call internals

Abstract out the calling of true system calls from the vdso into macros.

It has been a very long time since gcc did not allow %ebx or %ebp in inline asm in 32-bit PIC mode; remove the corresponding hacks.

Remove the use of memory output constraints in gettimeofday.h in favor of "memory" clobbers. The resulting code is identical for the current use cases, as the system call is usually a terminal fallback anyway, and it merely complicates the macroization.

This patch adds only a handful of more lines of code than it removes, and in fact could be made substantially smaller by removing the macros for the argument counts that aren't currently used, however, it seems better to be general from the start.

[ v3: remove stray comment from prototyping; remove VDSO_SYSCALL6()
since it would require special handling on 32 bits and is currently unused. (Uros Biszjak)

Indent nested preprocessor directives. ]

---
 arch/x86/include/asm/vdso/gettimeofday.h | 108 +---------------------
 arch/x86/include/asm/vdso/sys_call.h     | 103 +++++++++++++++++++++-
 2 files changed, 111 insertions(+), 100 deletions(-)
 create mode 100644 arch/x86/include/asm/vdso/sys_call.h
arch/x86/include/asm/vdso/gettimeofday.h +8 -100
index 73b2e7e..3cf214c 100644
@@ -18,6 +18,7 @@
 #include <asm/msr.h>
 #include <asm/pvclock.h>
 #include <clocksource/hyperv_timer.h>
+#include <asm/vdso/sys_call.h>
 
 #define VDSO_HAS_TIME 1
 
@@ -53,130 +54,37 @@ extern struct ms_hyperv_tsc_page hvclock_page
 	__attribute__((visibility("hidden")));
 #endif
 
-#ifndef BUILD_VDSO32
-
 static __always_inline
 long clock_gettime_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
 {
-	long ret;
-
-	asm ("syscall" : "=a" (ret), "=m" (*_ts) :
-	     "0" (__NR_clock_gettime), "D" (_clkid), "S" (_ts) :
-	     "rcx", "r11");
-
-	return ret;
+	return VDSO_SYSCALL2(clock_gettime,64,_clkid,_ts);
 }
 
 static __always_inline
 long gettimeofday_fallback(struct __kernel_old_timeval *_tv,
 			   struct timezone *_tz)
 {
-	long ret;
-
-	asm("syscall" : "=a" (ret) :
-	    "0" (__NR_gettimeofday), "D" (_tv), "S" (_tz) : "memory");
-
-	return ret;
+	return VDSO_SYSCALL2(gettimeofday,,_tv,_tz);
 }
 
 static __always_inline
 long clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
 {
-	long ret;
-
-	asm ("syscall" : "=a" (ret), "=m" (*_ts) :
-	     "0" (__NR_clock_getres), "D" (_clkid), "S" (_ts) :
-	     "rcx", "r11");
-
-	return ret;
+	return VDSO_SYSCALL2(clock_getres,_time64,_clkid,_ts);
 }
 
-#else
-
-static __always_inline
-long clock_gettime_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
-{
-	long ret;
-
-	asm (
-		"mov %%ebx, %%edx \n"
-		"mov %[clock], %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret), "=m" (*_ts)
-		: "0" (__NR_clock_gettime64), [clock] "g" (_clkid), "c" (_ts)
-		: "edx");
-
-	return ret;
-}
+#ifndef CONFIG_X86_64
 
 static __always_inline
 long clock_gettime32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)
 {
-	long ret;
-
-	asm (
-		"mov %%ebx, %%edx \n"
-		"mov %[clock], %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret), "=m" (*_ts)
-		: "0" (__NR_clock_gettime), [clock] "g" (_clkid), "c" (_ts)
-		: "edx");
-
-	return ret;
-}
-
-static __always_inline
-long gettimeofday_fallback(struct __kernel_old_timeval *_tv,
-			   struct timezone *_tz)
-{
-	long ret;
-
-	asm(
-		"mov %%ebx, %%edx \n"
-		"mov %2, %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret)
-		: "0" (__NR_gettimeofday), "g" (_tv), "c" (_tz)
-		: "memory", "edx");
-
-	return ret;
+	return VDSO_SYSCALL2(clock_gettime,,_clkid,_ts);
 }
 
 static __always_inline long
-clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
-{
-	long ret;
-
-	asm (
-		"mov %%ebx, %%edx \n"
-		"mov %[clock], %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret), "=m" (*_ts)
-		: "0" (__NR_clock_getres_time64), [clock] "g" (_clkid), "c" (_ts)
-		: "edx");
-
-	return ret;
-}
-
-static __always_inline
-long clock_getres32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)
+clock_getres32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)
 {
-	long ret;
-
-	asm (
-		"mov %%ebx, %%edx \n"
-		"mov %[clock], %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret), "=m" (*_ts)
-		: "0" (__NR_clock_getres), [clock] "g" (_clkid), "c" (_ts)
-		: "edx");
-
-	return ret;
+	return VDSO_SYSCALL2(clock_getres,,_clkid,_ts);
 }
 
 #endif
arch/x86/include/asm/vdso/sys_call.h new file +103 -0

103 lines not on this page · raw message

author maintainer
#22 lore raw

PATCH v4 9/10 x86/cpufeature: replace X86_FEATURE_SYSENTER32 with X86_FEATURE_SYSFAST32

In reply to H. Peter Anvin

In most cases, the use of "fast 32-bit system call" depends either on X86_FEATURE_SEP or X86_FEATURE_SYSENTER32 || X86_FEATURE_SYSCALL32. However, nearly all the logic for both is identical.

Define X86_FEATURE_SYSFAST32 which indicates that *either* SYSENTER32 or SYSCALL32 should be used, for either 32- or 64-bit kernels. This defaults to SYSENTER; use SYSCALL if the SYSCALL32 bit is also set.

As this removes ALL existing uses of X86_FEATURE_SYSENTER32, which is a kernel-only synthetic feature bit, simply remove it and replace it with X86_FEATURE_SYSFAST32.

This leaves an unused alternative for a true 32-bit kernel, but that should really not matter in any way.

The clearing of X86_FEATURE_SYSCALL32 can be removed once the patches for automatically clearing disabled features has been merged.

---
 arch/x86/Kconfig.cpufeatures             |  8 +++++++
 arch/x86/entry/vdso/vdso32/system_call.S |  8 ++-----
 arch/x86/include/asm/cpufeatures.h       |  2 +-
 arch/x86/kernel/cpu/centaur.c            |  3 ---
 arch/x86/kernel/cpu/common.c             |  8 +++++++
 arch/x86/kernel/cpu/intel.c              |  4 +---
 arch/x86/kernel/cpu/zhaoxin.c            |  4 +---
 arch/x86/kernel/fred.c                   |  2 +-
 arch/x86/xen/setup.c                     | 28 +++++++++++++++---------
 arch/x86/xen/smp_pv.c                    |  5 ++---
 arch/x86/xen/xen-ops.h                   |  1 -
 11 files changed, 42 insertions(+), 31 deletions(-)
arch/x86/Kconfig.cpufeatures +8 -0
index 733d5aff2456..423ac795baa7 100644
@@ -56,6 +56,10 @@ config X86_REQUIRED_FEATURE_MOVBE
 	def_bool y
 	depends on MATOM
 
+config X86_REQUIRED_FEATURE_SYSFAST32
+	def_bool y
+	depends on X86_64 && !X86_FRED
+
 config X86_REQUIRED_FEATURE_CPUID
 	def_bool y
 	depends on X86_64
@@ -120,6 +124,10 @@ config X86_DISABLED_FEATURE_CENTAUR_MCR
 	def_bool y
 	depends on X86_64
 
+config X86_DISABLED_FEATURE_SYSCALL32
+	def_bool y
+	depends on !X86_64
+
 config X86_DISABLED_FEATURE_PCID
 	def_bool y
 	depends on !X86_64
arch/x86/entry/vdso/vdso32/system_call.S +2 -6
index 2a15634bbe75..7b1c0f16e511 100644
@@ -52,13 +52,9 @@ __kernel_vsyscall:
 	#define SYSENTER_SEQUENCE	"movl %esp, %ebp; sysenter"
 	#define SYSCALL_SEQUENCE	"movl %ecx, %ebp; syscall"
 
-#ifdef BUILD_VDSO32_64
 	/* If SYSENTER (Intel) or SYSCALL32 (AMD) is available, use it. */
-	ALTERNATIVE_2 "", SYSENTER_SEQUENCE, X86_FEATURE_SYSENTER32, \
-	                  SYSCALL_SEQUENCE,  X86_FEATURE_SYSCALL32
-#else
-	ALTERNATIVE "", SYSENTER_SEQUENCE, X86_FEATURE_SEP
-#endif
+	ALTERNATIVE_2 "", SYSENTER_SEQUENCE, X86_FEATURE_SYSFAST32, \
+			  SYSCALL_SEQUENCE,  X86_FEATURE_SYSCALL32
 
 	/* Enter using int $0x80 */
 	int	$0x80
arch/x86/include/asm/cpufeatures.h +1 -1
index c3b53beb1300..63b0f9aa9b3e 100644
@@ -84,7 +84,7 @@
 #define X86_FEATURE_PEBS		( 3*32+12) /* "pebs" Precise-Event Based Sampling */
 #define X86_FEATURE_BTS			( 3*32+13) /* "bts" Branch Trace Store */
 #define X86_FEATURE_SYSCALL32		( 3*32+14) /* syscall in IA32 userspace */
-#define X86_FEATURE_SYSENTER32		( 3*32+15) /* sysenter in IA32 userspace */
+#define X86_FEATURE_SYSFAST32		( 3*32+15) /* sysenter/syscall in IA32 userspace */
 #define X86_FEATURE_REP_GOOD		( 3*32+16) /* "rep_good" REP microcode works well */
 #define X86_FEATURE_AMD_LBR_V2		( 3*32+17) /* "amd_lbr_v2" AMD Last Branch Record Extension Version 2 */
 #define X86_FEATURE_CLEAR_CPU_BUF	( 3*32+18) /* Clear CPU buffers using VERW */
arch/x86/kernel/cpu/centaur.c +0 -3
index a3b55db35c96..9833f837141c 100644
@@ -102,9 +102,6 @@ static void early_init_centaur(struct cpuinfo_x86 *c)
 	    (c->x86 >= 7))
 		set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
 
-#ifdef CONFIG_X86_64
-	set_cpu_cap(c, X86_FEATURE_SYSENTER32);
-#endif
 	if (c->x86_power & (1 << 8)) {
 		set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
 		set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC);
arch/x86/kernel/cpu/common.c +8 -0
index e7ab22fce3b5..1c3261cae40c 100644
@@ -1068,6 +1068,9 @@ void get_cpu_cap(struct cpuinfo_x86 *c)
 	init_scattered_cpuid_features(c);
 	init_speculation_control(c);
 
+	if (IS_ENABLED(CONFIG_X86_64) || cpu_has(c, X86_FEATURE_SEP))
+		set_cpu_cap(c, X86_FEATURE_SYSFAST32);
+
 	/*
 	 * Clear/Set all flags overridden by options, after probe.
 	 * This needs to happen each time we re-probe, which may happen
@@ -1813,6 +1816,11 @@ static void __init early_identify_cpu(struct cpuinfo_x86 *c)
 	 * that it can't be enabled in 32-bit mode.
 	 */
 	setup_clear_cpu_cap(X86_FEATURE_PCID);
+
+	/*
+	 * Never use SYSCALL on a 32-bit kernel
+	 */
+	setup_clear_cpu_cap(X86_FEATURE_SYSCALL32);
 #endif
 
 	/*
arch/x86/kernel/cpu/intel.c +1 -3
index 98ae4c37c93e..646ff33c4651 100644
@@ -236,9 +236,7 @@ static void early_init_intel(struct cpuinfo_x86 *c)
 		clear_cpu_cap(c, X86_FEATURE_PSE);
 	}
 
-#ifdef CONFIG_X86_64
-	set_cpu_cap(c, X86_FEATURE_SYSENTER32);
-#else
+#ifndef CONFIG_X86_64
 	/* Netburst reports 64 bytes clflush size, but does IO in 128 bytes */
 	if (c->x86 == 15 && c->x86_cache_alignment == 64)
 		c->x86_cache_alignment = 128;
arch/x86/kernel/cpu/zhaoxin.c +1 -3
index 89b1c8a70fe8..031379b7d4fa 100644
@@ -59,9 +59,7 @@ static void early_init_zhaoxin(struct cpuinfo_x86 *c)
 {
 	if (c->x86 >= 0x6)
 		set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
-#ifdef CONFIG_X86_64
-	set_cpu_cap(c, X86_FEATURE_SYSENTER32);
-#endif
+
 	if (c->x86_power & (1 << 8)) {
 		set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
 		set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC);
arch/x86/kernel/fred.c +1 -1
index 816187da3a47..e736b19e18de 100644
@@ -68,7 +68,7 @@ void cpu_init_fred_exceptions(void)
 	idt_invalidate();
 
 	/* Use int $0x80 for 32-bit system calls in FRED mode */
-	setup_clear_cpu_cap(X86_FEATURE_SYSENTER32);
+	setup_clear_cpu_cap(X86_FEATURE_SYSFAST32);
 	setup_clear_cpu_cap(X86_FEATURE_SYSCALL32);
 }
 
arch/x86/xen/setup.c +18 -10
index 3823e52aef52..ac8021c3a997 100644
@@ -990,13 +990,6 @@ static int register_callback(unsigned type, const void *func)
 	return HYPERVISOR_callback_op(CALLBACKOP_register, &callback);
 }
 
-void xen_enable_sysenter(void)
-{
-	if (cpu_feature_enabled(X86_FEATURE_SYSENTER32) &&
-	    register_callback(CALLBACKTYPE_sysenter, xen_entry_SYSENTER_compat))
-		setup_clear_cpu_cap(X86_FEATURE_SYSENTER32);
-}
-
 void xen_enable_syscall(void)
 {
 	int ret;
@@ -1008,11 +1001,27 @@ void xen_enable_syscall(void)
 		   mechanism for syscalls. */
 	}
 
-	if (cpu_feature_enabled(X86_FEATURE_SYSCALL32) &&
-	    register_callback(CALLBACKTYPE_syscall32, xen_entry_SYSCALL_compat))
+	if (!cpu_feature_enabled(X86_FEATURE_SYSFAST32))
+		return;
+
+	if (cpu_feature_enabled(X86_FEATURE_SYSCALL32)) {
+		/* Use SYSCALL32 */
+		ret = register_callback(CALLBACKTYPE_syscall32,
+					xen_entry_SYSCALL_compat);
+
+	} else {
+		/* Use SYSENTER32 */
+		ret = register_callback(CALLBACKTYPE_sysenter,
+					xen_entry_SYSENTER_compat);
+	}
+
+	if (ret) {
 		setup_clear_cpu_cap(X86_FEATURE_SYSCALL32);
+		setup_clear_cpu_cap(X86_FEATURE_SYSFAST32);
+	}
 }
 
+
 static void __init xen_pvmmu_arch_setup(void)
 {
 	HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_writable_pagetables);
@@ -1022,7 +1031,6 @@ static void __init xen_pvmmu_arch_setup(void)
 	    register_callback(CALLBACKTYPE_failsafe, xen_failsafe_callback))
 		BUG();
 
-	xen_enable_sysenter();
 	xen_enable_syscall();
 }
 
arch/x86/xen/smp_pv.c +2 -3
index 9bb8ff8bff30..c40f326f0c3a 100644
@@ -65,10 +65,9 @@ static void cpu_bringup(void)
 	touch_softlockup_watchdog();
 
 	/* PVH runs in ring 0 and allows us to do native syscalls. Yay! */
-	if (!xen_feature(XENFEAT_supervisor_mode_kernel)) {
-		xen_enable_sysenter();
+	if (!xen_feature(XENFEAT_supervisor_mode_kernel))
 		xen_enable_syscall();
-	}
+
 	cpu = smp_processor_id();
 	identify_secondary_cpu(cpu);
 	set_cpu_sibling_map(cpu);
arch/x86/xen/xen-ops.h +0 -1
index 090349baec09..f6c331b20fad 100644
@@ -60,7 +60,6 @@ phys_addr_t __init xen_find_free_area(phys_addr_t size);
 char * __init xen_memory_setup(void);
 void __init xen_arch_setup(void);
 void xen_banner(void);
-void xen_enable_sysenter(void);
 void xen_enable_syscall(void);
 void xen_vcpu_restore(void);
 
Signature · 1 line
2.52.0
merge bot Applied
#23 lore raw

[tip: x86/entry] x86/cpufeature: Replace X86_FEATURE_SYSENTER32 with X86_FEATURE_SYSFAST32

The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     f49ecf5e110ab0ed255ddea5e321689faf4e50e6
Gitweb:        https://git.kernel.org/tip/f49ecf5e110ab0ed255ddea5e321689faf4e50e6
Author:        H. Peter Anvin <[email protected]>
AuthorDate:    Tue, 16 Dec 2025 13:26:03 -08:00
Committer:     Dave Hansen <[email protected]>
CommitterDate: Tue, 13 Jan 2026 16:37:58 -08:00

x86/cpufeature: Replace X86_FEATURE_SYSENTER32 with X86_FEATURE_SYSFAST32

In most cases, the use of "fast 32-bit system call" depends either on X86_FEATURE_SEP or X86_FEATURE_SYSENTER32 || X86_FEATURE_SYSCALL32. However, nearly all the logic for both is identical.

Define X86_FEATURE_SYSFAST32 which indicates that *either* SYSENTER32 or SYSCALL32 should be used, for either 32- or 64-bit kernels. This defaults to SYSENTER; use SYSCALL if the SYSCALL32 bit is also set.

As this removes ALL existing uses of X86_FEATURE_SYSENTER32, which is a kernel-only synthetic feature bit, simply remove it and replace it with X86_FEATURE_SYSFAST32.

This leaves an unused alternative for a true 32-bit kernel, but that should really not matter in any way.

The clearing of X86_FEATURE_SYSCALL32 can be removed once the patches for automatically clearing disabled features has been merged.

---
 arch/x86/Kconfig.cpufeatures             |  8 +++++++-
 arch/x86/entry/vdso/vdso32/system_call.S |  8 +------
 arch/x86/include/asm/cpufeatures.h       |  2 +-
 arch/x86/kernel/cpu/centaur.c            |  3 +--
 arch/x86/kernel/cpu/common.c             |  8 +++++++-
 arch/x86/kernel/cpu/intel.c              |  4 +---
 arch/x86/kernel/cpu/zhaoxin.c            |  4 +---
 arch/x86/kernel/fred.c                   |  2 +-
 arch/x86/xen/setup.c                     | 28 ++++++++++++++---------
 arch/x86/xen/smp_pv.c                    |  5 +---
 arch/x86/xen/xen-ops.h                   |  1 +-
 11 files changed, 42 insertions(+), 31 deletions(-)
arch/x86/Kconfig.cpufeatures +8 -0
index 733d5af..423ac79 100644
@@ -56,6 +56,10 @@ config X86_REQUIRED_FEATURE_MOVBE
 	def_bool y
 	depends on MATOM
 
+config X86_REQUIRED_FEATURE_SYSFAST32
+	def_bool y
+	depends on X86_64 && !X86_FRED
+
 config X86_REQUIRED_FEATURE_CPUID
 	def_bool y
 	depends on X86_64
@@ -120,6 +124,10 @@ config X86_DISABLED_FEATURE_CENTAUR_MCR
 	def_bool y
 	depends on X86_64
 
+config X86_DISABLED_FEATURE_SYSCALL32
+	def_bool y
+	depends on !X86_64
+
 config X86_DISABLED_FEATURE_PCID
 	def_bool y
 	depends on !X86_64
arch/x86/entry/vdso/vdso32/system_call.S +2 -6
index 2a15634..7b1c0f1 100644
@@ -52,13 +52,9 @@ __kernel_vsyscall:
 	#define SYSENTER_SEQUENCE	"movl %esp, %ebp; sysenter"
 	#define SYSCALL_SEQUENCE	"movl %ecx, %ebp; syscall"
 
-#ifdef BUILD_VDSO32_64
 	/* If SYSENTER (Intel) or SYSCALL32 (AMD) is available, use it. */
-	ALTERNATIVE_2 "", SYSENTER_SEQUENCE, X86_FEATURE_SYSENTER32, \
-	                  SYSCALL_SEQUENCE,  X86_FEATURE_SYSCALL32
-#else
-	ALTERNATIVE "", SYSENTER_SEQUENCE, X86_FEATURE_SEP
-#endif
+	ALTERNATIVE_2 "", SYSENTER_SEQUENCE, X86_FEATURE_SYSFAST32, \
+			  SYSCALL_SEQUENCE,  X86_FEATURE_SYSCALL32
 
 	/* Enter using int $0x80 */
 	int	$0x80
arch/x86/include/asm/cpufeatures.h +1 -1
index c3b53be..63b0f9a 100644
@@ -84,7 +84,7 @@
 #define X86_FEATURE_PEBS		( 3*32+12) /* "pebs" Precise-Event Based Sampling */
 #define X86_FEATURE_BTS			( 3*32+13) /* "bts" Branch Trace Store */
 #define X86_FEATURE_SYSCALL32		( 3*32+14) /* syscall in IA32 userspace */
-#define X86_FEATURE_SYSENTER32		( 3*32+15) /* sysenter in IA32 userspace */
+#define X86_FEATURE_SYSFAST32		( 3*32+15) /* sysenter/syscall in IA32 userspace */
 #define X86_FEATURE_REP_GOOD		( 3*32+16) /* "rep_good" REP microcode works well */
 #define X86_FEATURE_AMD_LBR_V2		( 3*32+17) /* "amd_lbr_v2" AMD Last Branch Record Extension Version 2 */
 #define X86_FEATURE_CLEAR_CPU_BUF	( 3*32+18) /* Clear CPU buffers using VERW */
arch/x86/kernel/cpu/centaur.c +0 -3
index a3b55db..9833f83 100644
@@ -102,9 +102,6 @@ static void early_init_centaur(struct cpuinfo_x86 *c)
 	    (c->x86 >= 7))
 		set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
 
-#ifdef CONFIG_X86_64
-	set_cpu_cap(c, X86_FEATURE_SYSENTER32);
-#endif
 	if (c->x86_power & (1 << 8)) {
 		set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
 		set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC);
arch/x86/kernel/cpu/common.c +8 -0
index e7ab22f..1c3261c 100644
@@ -1068,6 +1068,9 @@ void get_cpu_cap(struct cpuinfo_x86 *c)
 	init_scattered_cpuid_features(c);
 	init_speculation_control(c);
 
+	if (IS_ENABLED(CONFIG_X86_64) || cpu_has(c, X86_FEATURE_SEP))
+		set_cpu_cap(c, X86_FEATURE_SYSFAST32);
+
 	/*
 	 * Clear/Set all flags overridden by options, after probe.
 	 * This needs to happen each time we re-probe, which may happen
@@ -1813,6 +1816,11 @@ static void __init early_identify_cpu(struct cpuinfo_x86 *c)
 	 * that it can't be enabled in 32-bit mode.
 	 */
 	setup_clear_cpu_cap(X86_FEATURE_PCID);
+
+	/*
+	 * Never use SYSCALL on a 32-bit kernel
+	 */
+	setup_clear_cpu_cap(X86_FEATURE_SYSCALL32);
 #endif
 
 	/*
arch/x86/kernel/cpu/intel.c +1 -3
index 98ae4c3..646ff33 100644
@@ -236,9 +236,7 @@ static void early_init_intel(struct cpuinfo_x86 *c)
 		clear_cpu_cap(c, X86_FEATURE_PSE);
 	}
 
-#ifdef CONFIG_X86_64
-	set_cpu_cap(c, X86_FEATURE_SYSENTER32);
-#else
+#ifndef CONFIG_X86_64
 	/* Netburst reports 64 bytes clflush size, but does IO in 128 bytes */
 	if (c->x86 == 15 && c->x86_cache_alignment == 64)
 		c->x86_cache_alignment = 128;
arch/x86/kernel/cpu/zhaoxin.c +1 -3
index 89b1c8a..031379b 100644
@@ -59,9 +59,7 @@ static void early_init_zhaoxin(struct cpuinfo_x86 *c)
 {
 	if (c->x86 >= 0x6)
 		set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
-#ifdef CONFIG_X86_64
-	set_cpu_cap(c, X86_FEATURE_SYSENTER32);
-#endif
+
 	if (c->x86_power & (1 << 8)) {
 		set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
 		set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC);
arch/x86/kernel/fred.c +1 -1
index 816187d..e736b19 100644
@@ -68,7 +68,7 @@ void cpu_init_fred_exceptions(void)
 	idt_invalidate();
 
 	/* Use int $0x80 for 32-bit system calls in FRED mode */
-	setup_clear_cpu_cap(X86_FEATURE_SYSENTER32);
+	setup_clear_cpu_cap(X86_FEATURE_SYSFAST32);
 	setup_clear_cpu_cap(X86_FEATURE_SYSCALL32);
 }
 
arch/x86/xen/setup.c +18 -10
index 3823e52..ac8021c 100644
@@ -990,13 +990,6 @@ static int register_callback(unsigned type, const void *func)
 	return HYPERVISOR_callback_op(CALLBACKOP_register, &callback);
 }
 
-void xen_enable_sysenter(void)
-{
-	if (cpu_feature_enabled(X86_FEATURE_SYSENTER32) &&
-	    register_callback(CALLBACKTYPE_sysenter, xen_entry_SYSENTER_compat))
-		setup_clear_cpu_cap(X86_FEATURE_SYSENTER32);
-}
-
 void xen_enable_syscall(void)
 {
 	int ret;
@@ -1008,11 +1001,27 @@ void xen_enable_syscall(void)
 		   mechanism for syscalls. */
 	}
 
-	if (cpu_feature_enabled(X86_FEATURE_SYSCALL32) &&
-	    register_callback(CALLBACKTYPE_syscall32, xen_entry_SYSCALL_compat))
+	if (!cpu_feature_enabled(X86_FEATURE_SYSFAST32))
+		return;
+
+	if (cpu_feature_enabled(X86_FEATURE_SYSCALL32)) {
+		/* Use SYSCALL32 */
+		ret = register_callback(CALLBACKTYPE_syscall32,
+					xen_entry_SYSCALL_compat);
+
+	} else {
+		/* Use SYSENTER32 */
+		ret = register_callback(CALLBACKTYPE_sysenter,
+					xen_entry_SYSENTER_compat);
+	}
+
+	if (ret) {
 		setup_clear_cpu_cap(X86_FEATURE_SYSCALL32);
+		setup_clear_cpu_cap(X86_FEATURE_SYSFAST32);
+	}
 }
 
+
 static void __init xen_pvmmu_arch_setup(void)
 {
 	HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_writable_pagetables);
@@ -1022,7 +1031,6 @@ static void __init xen_pvmmu_arch_setup(void)
 	    register_callback(CALLBACKTYPE_failsafe, xen_failsafe_callback))
 		BUG();
 
-	xen_enable_sysenter();
 	xen_enable_syscall();
 }
 
arch/x86/xen/smp_pv.c +2 -3
index 9bb8ff8..c40f326 100644
@@ -65,10 +65,9 @@ static void cpu_bringup(void)
 	touch_softlockup_watchdog();
 
 	/* PVH runs in ring 0 and allows us to do native syscalls. Yay! */
-	if (!xen_feature(XENFEAT_supervisor_mode_kernel)) {
-		xen_enable_sysenter();
+	if (!xen_feature(XENFEAT_supervisor_mode_kernel))
 		xen_enable_syscall();
-	}
+
 	cpu = smp_processor_id();
 	identify_secondary_cpu(cpu);
 	set_cpu_sibling_map(cpu);
arch/x86/xen/xen-ops.h +0 -1
index 090349b..f6c331b 100644
@@ -60,7 +60,6 @@ phys_addr_t __init xen_find_free_area(phys_addr_t size);
 char * __init xen_memory_setup(void);
 void __init xen_arch_setup(void);
 void xen_banner(void);
-void xen_enable_sysenter(void);
 void xen_enable_syscall(void);
 void xen_vcpu_restore(void);
 
author maintainer
#24 lore raw

PATCH v4 10/10 x86/entry/vdso32: when using int $0x80, use it directly

In reply to H. Peter Anvin

When neither sysenter32 nor syscall32 is available (on either FRED-capable 64-bit hardware or old 32-bit hardware), there is no reason to do a bunch of stack shuffling in __kernel_vsyscall. Unfortunately, just overwriting the initial "push" instructions will mess up the CFI annotations, so suffer the 3-byte NOP if not applicable.

Similarly, inline the int $0x80 when doing inline system calls in the vdso instead of calling __kernel_vsyscall.

---
 arch/x86/entry/vdso/vdso32/system_call.S | 18 ++++++++++++++----
 arch/x86/include/asm/vdso/sys_call.h     |  4 +++-
 2 files changed, 17 insertions(+), 5 deletions(-)
arch/x86/entry/vdso/vdso32/system_call.S +14 -4
index 7b1c0f16e511..9157cf9c5749 100644
@@ -14,6 +14,18 @@
 	ALIGN
 __kernel_vsyscall:
 	CFI_STARTPROC
+
+	/*
+	 * If using int $0x80, there is no reason to muck about with the
+	 * stack here. Unfortunately just overwriting the push instructions
+	 * would mess up the CFI annotations, but it is only a 3-byte
+	 * NOP in that case. This could be avoided by patching the
+	 * vdso symbol table (not the code) and entry point, but that
+	 * would a fair bit of tooling work or by simply compiling
+	 * two different vDSO images, but that doesn't seem worth it.
+	 */
+	ALTERNATIVE "int $0x80; ret", "", X86_FEATURE_SYSFAST32
+
 	/*
 	 * Reshuffle regs so that all of any of the entry instructions
 	 * will preserve enough state.
@@ -52,11 +64,9 @@ __kernel_vsyscall:
 	#define SYSENTER_SEQUENCE	"movl %esp, %ebp; sysenter"
 	#define SYSCALL_SEQUENCE	"movl %ecx, %ebp; syscall"
 
-	/* If SYSENTER (Intel) or SYSCALL32 (AMD) is available, use it. */
-	ALTERNATIVE_2 "", SYSENTER_SEQUENCE, X86_FEATURE_SYSFAST32, \
-			  SYSCALL_SEQUENCE,  X86_FEATURE_SYSCALL32
+	ALTERNATIVE SYSENTER_SEQUENCE, SYSCALL_SEQUENCE, X86_FEATURE_SYSCALL32
 
-	/* Enter using int $0x80 */
+	/* Re-enter using int $0x80 */
 	int	$0x80
 SYM_INNER_LABEL(int80_landing_pad, SYM_L_GLOBAL)
 
arch/x86/include/asm/vdso/sys_call.h +3 -1
index dcfd17c6dd57..5806b1cd6aef 100644
@@ -20,7 +20,9 @@
 # define __sys_reg4	"r10"
 # define __sys_reg5	"r8"
 #else
-# define __sys_instr	"call __kernel_vsyscall"
+# define __sys_instr	ALTERNATIVE("ds;ds;ds;int $0x80",		\
+				    "call __kernel_vsyscall",		\
+				    X86_FEATURE_SYSFAST32)
 # define __sys_clobber	"memory"
 # define __sys_nr(x,y)	__NR_ ## x ## y
 # define __sys_reg1	"ebx"
Signature · 1 line
2.52.0
merge bot Applied
#25 lore raw

[tip: x86/entry] x86/entry/vdso32: When using int $0x80, use it directly

The following commit has been merged into the x86/entry branch of tip:

The following commit has been merged into the x86/entry branch of tip:

Commit-ID:     36d83c249e0395a915144eceeb528ddc19b1fbe6
Gitweb:        https://git.kernel.org/tip/36d83c249e0395a915144eceeb528ddc19b1fbe6
Author:        H. Peter Anvin <[email protected]>
AuthorDate:    Tue, 16 Dec 2025 13:26:04 -08:00
Committer:     Dave Hansen <[email protected]>
CommitterDate: Tue, 13 Jan 2026 16:37:58 -08:00

x86/entry/vdso32: When using int $0x80, use it directly

When neither sysenter32 nor syscall32 is available (on either FRED-capable 64-bit hardware or old 32-bit hardware), there is no reason to do a bunch of stack shuffling in __kernel_vsyscall. Unfortunately, just overwriting the initial "push" instructions will mess up the CFI annotations, so suffer the 3-byte NOP if not applicable.

Similarly, inline the int $0x80 when doing inline system calls in the vdso instead of calling __kernel_vsyscall.

---
 arch/x86/entry/vdso/vdso32/system_call.S | 18 ++++++++++++++----
 arch/x86/include/asm/vdso/sys_call.h     |  4 +++-
 2 files changed, 17 insertions(+), 5 deletions(-)
arch/x86/entry/vdso/vdso32/system_call.S +14 -4
index 7b1c0f1..9157cf9 100644
@@ -14,6 +14,18 @@
 	ALIGN
 __kernel_vsyscall:
 	CFI_STARTPROC
+
+	/*
+	 * If using int $0x80, there is no reason to muck about with the
+	 * stack here. Unfortunately just overwriting the push instructions
+	 * would mess up the CFI annotations, but it is only a 3-byte
+	 * NOP in that case. This could be avoided by patching the
+	 * vdso symbol table (not the code) and entry point, but that
+	 * would a fair bit of tooling work or by simply compiling
+	 * two different vDSO images, but that doesn't seem worth it.
+	 */
+	ALTERNATIVE "int $0x80; ret", "", X86_FEATURE_SYSFAST32
+
 	/*
 	 * Reshuffle regs so that all of any of the entry instructions
 	 * will preserve enough state.
@@ -52,11 +64,9 @@ __kernel_vsyscall:
 	#define SYSENTER_SEQUENCE	"movl %esp, %ebp; sysenter"
 	#define SYSCALL_SEQUENCE	"movl %ecx, %ebp; syscall"
 
-	/* If SYSENTER (Intel) or SYSCALL32 (AMD) is available, use it. */
-	ALTERNATIVE_2 "", SYSENTER_SEQUENCE, X86_FEATURE_SYSFAST32, \
-			  SYSCALL_SEQUENCE,  X86_FEATURE_SYSCALL32
+	ALTERNATIVE SYSENTER_SEQUENCE, SYSCALL_SEQUENCE, X86_FEATURE_SYSCALL32
 
-	/* Enter using int $0x80 */
+	/* Re-enter using int $0x80 */
 	int	$0x80
 SYM_INNER_LABEL(int80_landing_pad, SYM_L_GLOBAL)
 
arch/x86/include/asm/vdso/sys_call.h +3 -1
index dcfd17c..5806b1c 100644
@@ -20,7 +20,9 @@
 # define __sys_reg4	"r10"
 # define __sys_reg5	"r8"
 #else
-# define __sys_instr	"call __kernel_vsyscall"
+# define __sys_instr	ALTERNATIVE("ds;ds;ds;int $0x80",		\
+				    "call __kernel_vsyscall",		\
+				    X86_FEATURE_SYSFAST32)
 # define __sys_clobber	"memory"
 # define __sys_nr(x,y)	__NR_ ## x ## y
 # define __sys_reg1	"ebx"
author maintainer
#26 lore raw

cover x86/entry/vdso: clean up the vdso build, vdso updates

In reply to H. Peter Anvin

This is intended as a ping, since I think the v4 got swallowed by the holidays. v4.1 IS BASICALLY A REBASE AND RESEND OF v4; THE ONLY CODE CHANGE IS A SINGLE SPACE CHARACTER.

This patchset cleans up the vdso build by building the 32- and 64-bit vdsos in separate directories, moving the vdso2c tool to arch/x86/tools, and by merging common code and especially Makefile rules between the 32- and 64-bit vdsos to the greatest extent possible.

Be more strict about sanitizing and standardizing the vdso build options.

Disable CFI for the vdso until the kernel adds user space IBT support.

Modernize the DWARF generation in vdso32/sigreturn.S.

Add macros to cleanly call system calls from vdso source code.

Add the GNU_PROPERTY and GNU_STACK PHDRs to the vdso.

When using int $0x80 (old 32-bit hardware or FRED-capable 64-bit hardware) skip the stack stuff in the 32-bit kernel entry code and call int $0x80 directly when used from C code.

Changes from v4 to v4.1:

- Fix a single bogus whitespace character change in patch 7.
- Fix the spelling of Uros Bizjak's name in the comment to patch 8.
- Rebased onto v6.19-rc4.

Changes from v3 to v4:

- Improved description of patch 01/10.
- Split out the move of vdso2c to tools into a separate patch.
- Remove unrelated formatting changes from vdso-layout.lds.S.
- Fix *-x32.o being missing from "targets".
- Rebased onto v6.19-rc1.

Changes from v2 to v3:

In arch/x86/include/asm/vdso/sys_call.h:
- Remove stray comment from prototyping (Uros Bizjak)
- Remove VDSO_SYSCALL6() since it would require special handling on 32 bits with frame pointers and is currently unused. (Uros Bizjak)
- Indent nested preprocessor directives.

Changes from v1 to v2:

Too many to count - much of the patchset has been reworked

Patches:

x86/entry/vdso: rename vdso_image_* to vdso*_image x86/entry/vdso: move vdso2c to arch/x86/tools
x86/entry/vdso: refactor the vdso build
x86/entry/vdso32: don't rely on int80_landing_pad for adjusting ip x86/entry/vdso32: remove SYSCALL_ENTER_KERNEL macro in sigreturn.S x86/entry/vdso32: remove open-coded DWARF in sigreturn.S x86/entry/vdso: include GNU_PROPERTY and GNU_STACK PHDRs x86/vdso: abstract out vdso system call internals x86/cpufeature: replace X86_FEATURE_SYSENTER32 with X86_FEATURE_SYSFAST32 x86/entry/vdso32: when using int $0x80, use it directly

---
 arch/x86/Kconfig.cpufeatures                       |   8 +
 arch/x86/Makefile                                  |   2 +-
 arch/x86/entry/syscall_32.c                        |   2 +-
 arch/x86/entry/vdso/.gitignore                     |  11 +-
 arch/x86/entry/vdso/Makefile                       | 162 +--------------------
 arch/x86/entry/vdso/common/Makefile.include        |  89 +++++++++++
 arch/x86/entry/vdso/{vdso-note.S => common/note.S} |   5 +-
 arch/x86/entry/vdso/{ => common}/vclock_gettime.c  |   0
 arch/x86/entry/vdso/{ => common}/vdso-layout.lds.S |  38 +++--
 arch/x86/entry/vdso/{ => common}/vgetcpu.c         |   0
 arch/x86/entry/vdso/vdso32/Makefile                |  24 +++
 arch/x86/entry/vdso/vdso32/note.S                  |  19 +--
 arch/x86/entry/vdso/vdso32/sigreturn.S             | 152 +++++--------------
 arch/x86/entry/vdso/vdso32/system_call.S           |  22 ++-
 arch/x86/entry/vdso/vdso32/vclock_gettime.c        |   5 +-
 arch/x86/entry/vdso/vdso32/vdso32.lds.S            |   2 +-
 arch/x86/entry/vdso/vdso32/vgetcpu.c               |   4 +-
 arch/x86/entry/vdso/vdso64/Makefile                |  46 ++++++
 arch/x86/entry/vdso/vdso64/note.S                  |   1 +
 arch/x86/entry/vdso/vdso64/vclock_gettime.c        |   1 +
 .../entry/vdso/{vdso.lds.S => vdso64/vdso64.lds.S} |   2 +-
 arch/x86/entry/vdso/{ => vdso64}/vdsox32.lds.S     |   2 +-
 arch/x86/entry/vdso/vdso64/vgetcpu.c               |   1 +
 .../entry/vdso/{ => vdso64}/vgetrandom-chacha.S    |   0
 arch/x86/entry/vdso/{ => vdso64}/vgetrandom.c      |   2 +-
 arch/x86/entry/vdso/{ => vdso64}/vsgx.S            |   0
 arch/x86/entry/vdso/vma.c                          |  24 ++-
 arch/x86/include/asm/cpufeatures.h                 |   2 +-
 arch/x86/include/asm/dwarf2.h                      |   1 +
 arch/x86/include/asm/elf.h                         |   2 +-
 arch/x86/include/asm/vdso.h                        |   6 +-
 arch/x86/include/asm/vdso/gettimeofday.h           | 108 +-------------
 arch/x86/include/asm/vdso/sys_call.h               | 105 +++++++++++++
 arch/x86/kernel/asm-offsets.c                      |   6 +
 arch/x86/kernel/cpu/centaur.c                      |   3 -
 arch/x86/kernel/cpu/common.c                       |   8 +
 arch/x86/kernel/cpu/intel.c                        |   4 +-
 arch/x86/kernel/cpu/zhaoxin.c                      |   4 +-
 arch/x86/kernel/fred.c                             |   2 +-
 arch/x86/kernel/process_64.c                       |   6 +-
 arch/x86/kernel/signal_32.c                        |   4 +-
 arch/x86/tools/Makefile                            |  15 +-
 arch/x86/{entry/vdso => tools}/vdso2c.c            |   0
 arch/x86/{entry/vdso => tools}/vdso2c.h            |   0
 arch/x86/xen/setup.c                               |  28 ++--
 arch/x86/xen/smp_pv.c                              |   5 +-
 arch/x86/xen/xen-ops.h                             |   1 -
 47 files changed, 444 insertions(+), 490 deletions(-)
author maintainer
#27 lore raw

PATCH v4 1/10 x86/entry/vdso: rename vdso_image_* to vdso*_image

The vdso .so files are named vdso*.so. These structures are binary images and descriptions of these files, so it is more consistent for them to have a naming that more directly mirrors the filenames.

It is also very slightly more compact (by one character...) and simplifies the Makefile just a little bit.

---
 arch/x86/entry/syscall_32.c    |  2 +-
 arch/x86/entry/vdso/.gitignore | 11 ++++-------
 arch/x86/entry/vdso/Makefile   |  8 ++++----
 arch/x86/entry/vdso/vma.c      | 10 +++++-----
 arch/x86/include/asm/elf.h     |  2 +-
 arch/x86/include/asm/vdso.h    |  6 +++---
 arch/x86/kernel/process_64.c   |  6 +++---
 arch/x86/kernel/signal_32.c    |  4 ++--
 8 files changed, 23 insertions(+), 26 deletions(-)
arch/x86/entry/syscall_32.c +1 -1
index a67a644d0cfe..8e829575e12f 100644
@@ -319,7 +319,7 @@ __visible noinstr bool do_fast_syscall_32(struct pt_regs *regs)
 	 * convention.  Adjust regs so it looks like we entered using int80.
 	 */
 	unsigned long landing_pad = (unsigned long)current->mm->context.vdso +
-					vdso_image_32.sym_int80_landing_pad;
+					vdso32_image.sym_int80_landing_pad;
 
 	/*
 	 * SYSENTER loses EIP, and even SYSCALL32 needs us to skip forward
arch/x86/entry/vdso/.gitignore +4 -7
index 37a6129d597b..eb60859dbcbf 100644
@@ -1,8 +1,5 @@
 # SPDX-License-Identifier: GPL-2.0-only
-vdso.lds
-vdsox32.lds
-vdso32-syscall-syms.lds
-vdso32-sysenter-syms.lds
-vdso32-int80-syms.lds
-vdso-image-*.c
-vdso2c
+*.lds
+*.so
+*.so.dbg
+vdso*-image.c
arch/x86/entry/vdso/Makefile +4 -4
index f247f5f5cb44..7f833026d5b2 100644
@@ -16,9 +16,9 @@ vobjs-$(CONFIG_X86_SGX)	+= vsgx.o
 obj-y						+= vma.o extable.o
 
 # vDSO images to build:
-obj-$(CONFIG_X86_64)				+= vdso-image-64.o
-obj-$(CONFIG_X86_X32_ABI)			+= vdso-image-x32.o
-obj-$(CONFIG_COMPAT_32)				+= vdso-image-32.o vdso32-setup.o
+obj-$(CONFIG_X86_64)				+= vdso64-image.o
+obj-$(CONFIG_X86_X32_ABI)			+= vdsox32-image.o
+obj-$(CONFIG_COMPAT_32)				+= vdso32-image.o vdso32-setup.o
 
 vobjs := $(addprefix $(obj)/, $(vobjs-y))
 vobjs32 := $(addprefix $(obj)/, $(vobjs32-y))
@@ -44,7 +44,7 @@ hostprogs += vdso2c
 quiet_cmd_vdso2c = VDSO2C  $@
       cmd_vdso2c = $(obj)/vdso2c $< $(<:%.dbg=%) $@
 
-$(obj)/vdso-image-%.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(obj)/vdso2c FORCE
+$(obj)/vdso%-image.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(obj)/vdso2c FORCE
 	$(call if_changed,vdso2c)
 
 #
arch/x86/entry/vdso/vma.c +5 -5
index afe105b2f907..8f98c2d7c7a9 100644
@@ -65,7 +65,7 @@ static vm_fault_t vdso_fault(const struct vm_special_mapping *sm,
 static void vdso_fix_landing(const struct vdso_image *image,
 		struct vm_area_struct *new_vma)
 {
-	if (in_ia32_syscall() && image == &vdso_image_32) {
+	if (in_ia32_syscall() && image == &vdso32_image) {
 		struct pt_regs *regs = current_pt_regs();
 		unsigned long vdso_land = image->sym_int80_landing_pad;
 		unsigned long old_land_addr = vdso_land +
@@ -230,7 +230,7 @@ static int load_vdso32(void)
 	if (vdso32_enabled != 1)  /* Other values all mean "disabled" */
 		return 0;
 
-	return map_vdso(&vdso_image_32, 0);
+	return map_vdso(&vdso32_image, 0);
 }
 
 int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
@@ -239,7 +239,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 		if (!vdso64_enabled)
 			return 0;
 
-		return map_vdso(&vdso_image_64, 0);
+		return map_vdso(&vdso64_image, 0);
 	}
 
 	return load_vdso32();
@@ -252,7 +252,7 @@ int compat_arch_setup_additional_pages(struct linux_binprm *bprm,
 	if (IS_ENABLED(CONFIG_X86_X32_ABI) && x32) {
 		if (!vdso64_enabled)
 			return 0;
-		return map_vdso(&vdso_image_x32, 0);
+		return map_vdso(&vdsox32_image, 0);
 	}
 
 	if (IS_ENABLED(CONFIG_IA32_EMULATION))
@@ -267,7 +267,7 @@ bool arch_syscall_is_vdso_sigreturn(struct pt_regs *regs)
 	const struct vdso_image *image = current->mm->context.vdso_image;
 	unsigned long vdso = (unsigned long) current->mm->context.vdso;
 
-	if (in_ia32_syscall() && image == &vdso_image_32) {
+	if (in_ia32_syscall() && image == &vdso32_image) {
 		if (regs->ip == vdso + image->sym_vdso32_sigreturn_landing_pad ||
 		    regs->ip == vdso + image->sym_vdso32_rt_sigreturn_landing_pad)
 			return true;
arch/x86/include/asm/elf.h +1 -1
index 6c8fdc96be7e..2ba5f166e58f 100644
@@ -361,7 +361,7 @@ else if (IS_ENABLED(CONFIG_IA32_EMULATION))				\
 
 #define VDSO_ENTRY							\
 	((unsigned long)current->mm->context.vdso +			\
-	 vdso_image_32.sym___kernel_vsyscall)
+	 vdso32_image.sym___kernel_vsyscall)
 
 struct linux_binprm;
 
arch/x86/include/asm/vdso.h +3 -3
index b7253ef3205a..e8afbe9faa5b 100644
@@ -27,9 +27,9 @@ struct vdso_image {
 	long sym_vdso32_rt_sigreturn_landing_pad;
 };
 
-extern const struct vdso_image vdso_image_64;
-extern const struct vdso_image vdso_image_x32;
-extern const struct vdso_image vdso_image_32;
+extern const struct vdso_image vdso64_image;
+extern const struct vdso_image vdsox32_image;
+extern const struct vdso_image vdso32_image;
 
 extern int __init init_vdso_image(const struct vdso_image *image);
 
arch/x86/kernel/process_64.c +3 -3
index 432c0a004c60..08e72f429870 100644
@@ -941,14 +941,14 @@ long do_arch_prctl_64(struct task_struct *task, int option, unsigned long arg2)
 #ifdef CONFIG_CHECKPOINT_RESTORE
 # ifdef CONFIG_X86_X32_ABI
 	case ARCH_MAP_VDSO_X32:
-		return prctl_map_vdso(&vdso_image_x32, arg2);
+		return prctl_map_vdso(&vdsox32_image, arg2);
 # endif
 # ifdef CONFIG_IA32_EMULATION
 	case ARCH_MAP_VDSO_32:
-		return prctl_map_vdso(&vdso_image_32, arg2);
+		return prctl_map_vdso(&vdso32_image, arg2);
 # endif
 	case ARCH_MAP_VDSO_64:
-		return prctl_map_vdso(&vdso_image_64, arg2);
+		return prctl_map_vdso(&vdso64_image, arg2);
 #endif
 #ifdef CONFIG_ADDRESS_MASKING
 	case ARCH_GET_UNTAG_MASK:
arch/x86/kernel/signal_32.c +2 -2
index 42bbc42bd350..e55cf19e68fe 100644
@@ -282,7 +282,7 @@ int ia32_setup_frame(struct ksignal *ksig, struct pt_regs *regs)
 		/* Return stub is in 32bit vsyscall page */
 		if (current->mm->context.vdso)
 			restorer = current->mm->context.vdso +
-				vdso_image_32.sym___kernel_sigreturn;
+				vdso32_image.sym___kernel_sigreturn;
 		else
 			restorer = &frame->retcode;
 	}
@@ -368,7 +368,7 @@ int ia32_setup_rt_frame(struct ksignal *ksig, struct pt_regs *regs)
 		restorer = ksig->ka.sa.sa_restorer;
 	else
 		restorer = current->mm->context.vdso +
-			vdso_image_32.sym___kernel_rt_sigreturn;
+			vdso32_image.sym___kernel_rt_sigreturn;
 	unsafe_put_user(ptr_to_compat(restorer), &frame->pretcode, Efault);
 
 	/*
Signature · 1 line
2.52.0
author maintainer
#28 lore raw

PATCH v4 2/10 x86/entry/vdso: move vdso2c to arch/x86/tools

In reply to H. Peter Anvin

It is generally better to build tools in arch/x86/tools to keep host cflags proliferation down, and to reduce makefile sequencing issues. Move the vdso build tool vdso2c into arch/x86/tools in preparation for refactoring the vdso makefiles.

---
 arch/x86/Makefile                       |  2 +-
 arch/x86/entry/vdso/Makefile            |  7 +++----
 arch/x86/tools/Makefile                 | 15 ++++++++++-----
 arch/x86/{entry/vdso => tools}/vdso2c.c |  0
 arch/x86/{entry/vdso => tools}/vdso2c.h |  0
 5 files changed, 14 insertions(+), 10 deletions(-)
 rename arch/x86/{entry/vdso => tools}/vdso2c.c (100%)
 rename arch/x86/{entry/vdso => tools}/vdso2c.h (100%)
arch/x86/Makefile +1 -1
index 1d403a3612ea..9ab7522ced18 100644
@@ -252,7 +252,7 @@ endif
 
 
 archscripts: scripts_basic
-	$(Q)$(MAKE) $(build)=arch/x86/tools relocs
+	$(Q)$(MAKE) $(build)=arch/x86/tools relocs vdso2c
 
 ###
 # Syscall table generation
arch/x86/entry/vdso/Makefile +3 -4
index 7f833026d5b2..3d9b09f00c70 100644
@@ -38,13 +38,12 @@ VDSO_LDFLAGS_vdso.lds = -m elf_x86_64 -soname linux-vdso.so.1 \
 $(obj)/vdso64.so.dbg: $(obj)/vdso.lds $(vobjs) FORCE
 	$(call if_changed,vdso_and_check)
 
-HOST_EXTRACFLAGS += -I$(srctree)/tools/include -I$(srctree)/include/uapi -I$(srctree)/arch/$(SUBARCH)/include/uapi
-hostprogs += vdso2c
+VDSO2C = $(objtree)/arch/x86/tools/vdso2c
 
 quiet_cmd_vdso2c = VDSO2C  $@
-      cmd_vdso2c = $(obj)/vdso2c $< $(<:%.dbg=%) $@
+      cmd_vdso2c = $(VDSO2C) $< $(<:%.dbg=%) $@
 
-$(obj)/vdso%-image.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(obj)/vdso2c FORCE
+$(obj)/vdso%-image.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(VDSO2C) FORCE
 	$(call if_changed,vdso2c)
 
 #
arch/x86/tools/Makefile +10 -5
index 7278e2545c35..39a183fffd04 100644
@@ -38,9 +38,14 @@ $(obj)/insn_decoder_test.o: $(srctree)/tools/arch/x86/lib/insn.c $(srctree)/tool
 
 $(obj)/insn_sanity.o: $(srctree)/tools/arch/x86/lib/insn.c $(srctree)/tools/arch/x86/lib/inat.c $(srctree)/tools/arch/x86/include/asm/inat_types.h $(srctree)/tools/arch/x86/include/asm/inat.h $(srctree)/tools/arch/x86/include/asm/insn.h $(objtree)/arch/x86/lib/inat-tables.c
 
-HOST_EXTRACFLAGS += -I$(srctree)/tools/include
-hostprogs	+= relocs
-relocs-objs     := relocs_32.o relocs_64.o relocs_common.o
-PHONY += relocs
-relocs: $(obj)/relocs
+HOST_EXTRACFLAGS += -I$(srctree)/tools/include -I$(srctree)/include/uapi \
+		    -I$(srctree)/arch/$(SUBARCH)/include/uapi
+
+hostprogs	+= relocs vdso2c
+relocs-objs	:= relocs_32.o relocs_64.o relocs_common.o
+
+always-y	:= $(hostprogs)
+
+PHONY += $(hostprogs)
+$(hostprogs): %: $(obj)/%
 	@:
arch/x86/tools/vdso2c.c renamed from arch/x86/entry/vdso/vdso2c.c +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vdso2c.c
rename to arch/x86/tools/vdso2c.c
arch/x86/tools/vdso2c.h renamed from arch/x86/entry/vdso/vdso2c.h +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vdso2c.h
rename to arch/x86/tools/vdso2c.h
Signature · 1 line
2.52.0
author maintainer
#29 lore raw

PATCH v4 3/10 x86/entry/vdso: refactor the vdso build

In reply to H. Peter Anvin

- Separate out the vdso sources into common, vdso32, and vdso64 directories.
- Build the 32- and 64-bit vdsos in their respective subdirectories; this greatly simplifies the build flags handling.
- Unify the mangling of Makefile flags between the 32- and 64-bit vdso code as much as possible; all common rules are put in arch/x86/entry/vdso/common/Makefile.include. The remaining is very simple for 32 bits; the 64-bit one is only slightly more complicated because it contains the x32 generation rule.
- Define __DISABLE_EXPORTS when building the vdso. This need seems to have been masked by different ordering compile flags before.
- Change CONFIG_X86_64 to BUILD_VDSO32_64 in vdso32/system_call.S, to make it compatible with including fake_32bit_build.h.
- The -fcf-protection= option was "leaking" from the kernel build, for reasons that was not clear to me. Futhermore, several distributions ship with it set to a default value other than "-fcf-protection=none". Make it match the configuration options for *user space*.

Note that this patch may seem large, but the vast majority of it is simply code movement.

---
 arch/x86/entry/vdso/Makefile                  | 161 +-----------------
 arch/x86/entry/vdso/common/Makefile.include   |  89 ++++++++++
 .../entry/vdso/{vdso-note.S => common/note.S} |   5 +-
 .../entry/vdso/{ => common}/vclock_gettime.c  |   0
 .../entry/vdso/{ => common}/vdso-layout.lds.S |   0
 arch/x86/entry/vdso/{ => common}/vgetcpu.c    |   0
 arch/x86/entry/vdso/vdso32/Makefile           |  24 +++
 arch/x86/entry/vdso/vdso32/note.S             |  19 +--
 arch/x86/entry/vdso/vdso32/system_call.S      |   2 +-
 arch/x86/entry/vdso/vdso32/vclock_gettime.c   |   5 +-
 arch/x86/entry/vdso/vdso32/vdso32.lds.S       |   2 +-
 arch/x86/entry/vdso/vdso32/vgetcpu.c          |   4 +-
 arch/x86/entry/vdso/vdso64/Makefile           |  46 +++++
 arch/x86/entry/vdso/vdso64/note.S             |   1 +
 arch/x86/entry/vdso/vdso64/vclock_gettime.c   |   1 +
 .../vdso/{vdso.lds.S => vdso64/vdso64.lds.S}  |   2 +-
 .../x86/entry/vdso/{ => vdso64}/vdsox32.lds.S |   2 +-
 arch/x86/entry/vdso/vdso64/vgetcpu.c          |   1 +
 .../vdso/{ => vdso64}/vgetrandom-chacha.S     |   0
 arch/x86/entry/vdso/{ => vdso64}/vgetrandom.c |   2 +-
 arch/x86/entry/vdso/{ => vdso64}/vsgx.S       |   0
 21 files changed, 180 insertions(+), 186 deletions(-)
 create mode 100644 arch/x86/entry/vdso/common/Makefile.include
 rename arch/x86/entry/vdso/{vdso-note.S => common/note.S} (62%)
 rename arch/x86/entry/vdso/{ => common}/vclock_gettime.c (100%)
 rename arch/x86/entry/vdso/{ => common}/vdso-layout.lds.S (100%)
 rename arch/x86/entry/vdso/{ => common}/vgetcpu.c (100%)
 create mode 100644 arch/x86/entry/vdso/vdso32/Makefile
 create mode 100644 arch/x86/entry/vdso/vdso64/Makefile
 create mode 100644 arch/x86/entry/vdso/vdso64/note.S
 create mode 100644 arch/x86/entry/vdso/vdso64/vclock_gettime.c
 rename arch/x86/entry/vdso/{vdso.lds.S => vdso64/vdso64.lds.S} (94%)
 rename arch/x86/entry/vdso/{ => vdso64}/vdsox32.lds.S (92%)
 create mode 100644 arch/x86/entry/vdso/vdso64/vgetcpu.c
 rename arch/x86/entry/vdso/{ => vdso64}/vgetrandom-chacha.S (100%)
 rename arch/x86/entry/vdso/{ => vdso64}/vgetrandom.c (91%)
 rename arch/x86/entry/vdso/{ => vdso64}/vsgx.S (100%)
arch/x86/entry/vdso/Makefile +6 -155
index 3d9b09f00c70..987b43fd4cd3 100644
@@ -3,159 +3,10 @@
 # Building vDSO images for x86.
 #
 
-# Include the generic Makefile to check the built vDSO:
-include $(srctree)/lib/vdso/Makefile.include
+# Regular kernel objects
+obj-y				:= vma.o extable.o
+obj-$(CONFIG_COMPAT_32)		+= vdso32-setup.o
 
-# Files to link into the vDSO:
-vobjs-y := vdso-note.o vclock_gettime.o vgetcpu.o vgetrandom.o vgetrandom-chacha.o
-vobjs32-y := vdso32/note.o vdso32/system_call.o vdso32/sigreturn.o
-vobjs32-y += vdso32/vclock_gettime.o vdso32/vgetcpu.o
-vobjs-$(CONFIG_X86_SGX)	+= vsgx.o
-
-# Files to link into the kernel:
-obj-y						+= vma.o extable.o
-
-# vDSO images to build:
-obj-$(CONFIG_X86_64)				+= vdso64-image.o
-obj-$(CONFIG_X86_X32_ABI)			+= vdsox32-image.o
-obj-$(CONFIG_COMPAT_32)				+= vdso32-image.o vdso32-setup.o
-
-vobjs := $(addprefix $(obj)/, $(vobjs-y))
-vobjs32 := $(addprefix $(obj)/, $(vobjs32-y))
-
-$(obj)/vdso.o: $(obj)/vdso.so
-
-targets += vdso.lds $(vobjs-y)
-targets += vdso32/vdso32.lds $(vobjs32-y)
-
-targets += $(foreach x, 64 x32 32, vdso-image-$(x).c vdso$(x).so vdso$(x).so.dbg)
-
-CPPFLAGS_vdso.lds += -P -C
-
-VDSO_LDFLAGS_vdso.lds = -m elf_x86_64 -soname linux-vdso.so.1 \
-			-z max-page-size=4096
-
-$(obj)/vdso64.so.dbg: $(obj)/vdso.lds $(vobjs) FORCE
-	$(call if_changed,vdso_and_check)
-
-VDSO2C = $(objtree)/arch/x86/tools/vdso2c
-
-quiet_cmd_vdso2c = VDSO2C  $@
-      cmd_vdso2c = $(VDSO2C) $< $(<:%.dbg=%) $@
-
-$(obj)/vdso%-image.c: $(obj)/vdso%.so.dbg $(obj)/vdso%.so $(VDSO2C) FORCE
-	$(call if_changed,vdso2c)
-
-#
-# Don't omit frame pointers for ease of userspace debugging, but do
-# optimize sibling calls.
-#
-CFL := $(PROFILING) -mcmodel=small -fPIC -O2 -fasynchronous-unwind-tables -m64 \
-       $(filter -g%,$(KBUILD_CFLAGS)) -fno-stack-protector \
-       -fno-omit-frame-pointer -foptimize-sibling-calls \
-       -DDISABLE_BRANCH_PROFILING -DBUILD_VDSO
-
-ifdef CONFIG_MITIGATION_RETPOLINE
-ifneq ($(RETPOLINE_VDSO_CFLAGS),)
-  CFL += $(RETPOLINE_VDSO_CFLAGS)
-endif
-endif
-
-$(vobjs): KBUILD_CFLAGS := $(filter-out $(PADDING_CFLAGS) $(CC_FLAGS_LTO) $(CC_FLAGS_CFI) $(RANDSTRUCT_CFLAGS) $(KSTACK_ERASE_CFLAGS) $(GCC_PLUGINS_CFLAGS) $(RETPOLINE_CFLAGS),$(KBUILD_CFLAGS)) $(CFL)
-$(vobjs): KBUILD_AFLAGS += -DBUILD_VDSO
-
-#
-# vDSO code runs in userspace and -pg doesn't help with profiling anyway.
-#
-CFLAGS_REMOVE_vclock_gettime.o = -pg
-CFLAGS_REMOVE_vdso32/vclock_gettime.o = -pg
-CFLAGS_REMOVE_vgetcpu.o = -pg
-CFLAGS_REMOVE_vdso32/vgetcpu.o = -pg
-CFLAGS_REMOVE_vsgx.o = -pg
-CFLAGS_REMOVE_vgetrandom.o = -pg
-
-#
-# X32 processes use x32 vDSO to access 64bit kernel data.
-#
-# Build x32 vDSO image:
-# 1. Compile x32 vDSO as 64bit.
-# 2. Convert object files to x32.
-# 3. Build x32 VDSO image with x32 objects, which contains 64bit codes
-# so that it can reach 64bit address space with 64bit pointers.
-#
-
-CPPFLAGS_vdsox32.lds = $(CPPFLAGS_vdso.lds)
-VDSO_LDFLAGS_vdsox32.lds = -m elf32_x86_64 -soname linux-vdso.so.1 \
-			   -z max-page-size=4096
-
-# x32-rebranded versions
-vobjx32s-y := $(vobjs-y:.o=-x32.o)
-
-# same thing, but in the output directory
-vobjx32s := $(addprefix $(obj)/, $(vobjx32s-y))
-
-# Convert 64bit object file to x32 for x32 vDSO.
-quiet_cmd_x32 = X32     $@
-      cmd_x32 = $(OBJCOPY) -O elf32-x86-64 $< $@
-
-$(obj)/%-x32.o: $(obj)/%.o FORCE
-	$(call if_changed,x32)
-
-targets += vdsox32.lds $(vobjx32s-y)
-
-$(obj)/%.so: OBJCOPYFLAGS := -S --remove-section __ex_table
-$(obj)/%.so: $(obj)/%.so.dbg FORCE
-	$(call if_changed,objcopy)
-
-$(obj)/vdsox32.so.dbg: $(obj)/vdsox32.lds $(vobjx32s) FORCE
-	$(call if_changed,vdso_and_check)
-
-CPPFLAGS_vdso32/vdso32.lds = $(CPPFLAGS_vdso.lds)
-VDSO_LDFLAGS_vdso32.lds = -m elf_i386 -soname linux-gate.so.1
-
-KBUILD_AFLAGS_32 := $(filter-out -m64,$(KBUILD_AFLAGS)) -DBUILD_VDSO
-$(obj)/vdso32.so.dbg: KBUILD_AFLAGS = $(KBUILD_AFLAGS_32)
-$(obj)/vdso32.so.dbg: asflags-$(CONFIG_X86_64) += -m32
-
-KBUILD_CFLAGS_32 := $(filter-out -m64,$(KBUILD_CFLAGS))
-KBUILD_CFLAGS_32 := $(filter-out -mcmodel=kernel,$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out -fno-pic,$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out -mfentry,$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(RANDSTRUCT_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(KSTACK_ERASE_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(GCC_PLUGINS_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(RETPOLINE_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(CC_FLAGS_LTO),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(CC_FLAGS_CFI),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 := $(filter-out $(PADDING_CFLAGS),$(KBUILD_CFLAGS_32))
-KBUILD_CFLAGS_32 += -m32 -msoft-float -mregparm=0 -fpic
-KBUILD_CFLAGS_32 += -fno-stack-protector
-KBUILD_CFLAGS_32 += $(call cc-option, -foptimize-sibling-calls)
-KBUILD_CFLAGS_32 += -fno-omit-frame-pointer
-KBUILD_CFLAGS_32 += -DDISABLE_BRANCH_PROFILING
-KBUILD_CFLAGS_32 += -DBUILD_VDSO
-
-ifdef CONFIG_MITIGATION_RETPOLINE
-ifneq ($(RETPOLINE_VDSO_CFLAGS),)
-  KBUILD_CFLAGS_32 += $(RETPOLINE_VDSO_CFLAGS)
-endif
-endif
-
-$(obj)/vdso32.so.dbg: KBUILD_CFLAGS = $(KBUILD_CFLAGS_32)
-
-$(obj)/vdso32.so.dbg: $(obj)/vdso32/vdso32.lds $(vobjs32) FORCE
-	$(call if_changed,vdso_and_check)
-
-#
-# The DSO images are built using a special linker script.
-#
-quiet_cmd_vdso = VDSO    $@
-      cmd_vdso = $(LD) -o $@ \
-		       $(VDSO_LDFLAGS) $(VDSO_LDFLAGS_$(filter %.lds,$(^F))) \
-		       -T $(filter %.lds,$^) $(filter %.o,$^)
-
-VDSO_LDFLAGS = -shared --hash-style=both --build-id=sha1 --no-undefined \
-	$(call ld-option, --eh-frame-hdr) -Bsymbolic -z noexecstack
-
-quiet_cmd_vdso_and_check = VDSO    $@
-      cmd_vdso_and_check = $(cmd_vdso); $(cmd_vdso_check)
+# vDSO directories
+obj-$(CONFIG_X86_64)		+= vdso64/
+obj-$(CONFIG_COMPAT_32)		+= vdso32/
arch/x86/entry/vdso/common/Makefile.include new file +89 -0

89 lines not on this page · raw message

arch/x86/entry/vdso/common/note.S renamed from arch/x86/entry/vdso/vdso-note.S +4 -1
similarity index 62%
rename from arch/x86/entry/vdso/vdso-note.S
rename to arch/x86/entry/vdso/common/note.S
index 79423170118f..2cbd39939dc6 100644
@@ -1,13 +1,16 @@
+/* SPDX-License-Identifier: GPL-2.0 */
 /*
  * This supplies .note.* sections to go into the PT_NOTE inside the vDSO text.
  * Here we can supply some information useful to userland.
  */
 
 #include <linux/build-salt.h>
-#include <linux/uts.h>
 #include <linux/version.h>
 #include <linux/elfnote.h>
 
+/* Ideally this would use UTS_NAME, but using a quoted string here
+   doesn't work. Remember to change this when changing the
+   kernel's name. */
 ELFNOTE_START(Linux, 0, "a")
 	.long LINUX_VERSION_CODE
 ELFNOTE_END
arch/x86/entry/vdso/common/vclock_gettime.c renamed from arch/x86/entry/vdso/vclock_gettime.c +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vclock_gettime.c
rename to arch/x86/entry/vdso/common/vclock_gettime.c
arch/x86/entry/vdso/common/vdso-layout.lds.S renamed from arch/x86/entry/vdso/vdso-layout.lds.S +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vdso-layout.lds.S
rename to arch/x86/entry/vdso/common/vdso-layout.lds.S
arch/x86/entry/vdso/common/vgetcpu.c renamed from arch/x86/entry/vdso/vgetcpu.c +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vgetcpu.c
rename to arch/x86/entry/vdso/common/vgetcpu.c
arch/x86/entry/vdso/vdso32/Makefile new file +24 -0
new file mode 100644
index 000000000000..add6afb484ba
@@ -0,0 +1,24 @@
+# SPDX-License-Identifier: GPL-2.0
+#
+# 32-bit vDSO images for x86.
+#
+
+# The vDSOs built in this directory
+vdsos-y			:= 32
+
+# Files to link into the vDSO:
+vobjs-y			:= note.o vclock_gettime.o vgetcpu.o
+vobjs-y			+= system_call.o sigreturn.o
+
+# Compilation flags
+flags-y			:= -DBUILD_VDSO32 -m32 -mregparm=0
+flags-$(CONFIG_X86_64)	+= -include $(src)/fake_32bit_build.h
+flags-remove-y          := -m64
+
+# The location of this include matters!
+include $(src)/../common/Makefile.include
+
+# Linker options for the vdso
+VDSO_LDFLAGS_32		:= -m elf_i386 -soname linux-gate.so.1
+
+$(obj)/vdso32.so.dbg: $(vobjs)
arch/x86/entry/vdso/vdso32/note.S +1 -18
index 2cbd39939dc6..62d8aa51ce99 100644
@@ -1,18 +1 @@
-/* SPDX-License-Identifier: GPL-2.0 */
-/*
- * This supplies .note.* sections to go into the PT_NOTE inside the vDSO text.
- * Here we can supply some information useful to userland.
- */
-
-#include <linux/build-salt.h>
-#include <linux/version.h>
-#include <linux/elfnote.h>
-
-/* Ideally this would use UTS_NAME, but using a quoted string here
-   doesn't work. Remember to change this when changing the
-   kernel's name. */
-ELFNOTE_START(Linux, 0, "a")
-	.long LINUX_VERSION_CODE
-ELFNOTE_END
-
-BUILD_SALT
+#include "common/note.S"
arch/x86/entry/vdso/vdso32/system_call.S +1 -1
index d33c6513fd2c..2a15634bbe75 100644
@@ -52,7 +52,7 @@ __kernel_vsyscall:
 	#define SYSENTER_SEQUENCE	"movl %esp, %ebp; sysenter"
 	#define SYSCALL_SEQUENCE	"movl %ecx, %ebp; syscall"
 
-#ifdef CONFIG_X86_64
+#ifdef BUILD_VDSO32_64
 	/* If SYSENTER (Intel) or SYSCALL32 (AMD) is available, use it. */
 	ALTERNATIVE_2 "", SYSENTER_SEQUENCE, X86_FEATURE_SYSENTER32, \
 	                  SYSCALL_SEQUENCE,  X86_FEATURE_SYSCALL32
arch/x86/entry/vdso/vdso32/vclock_gettime.c +1 -4
index 86981decfea8..1481f0021b9f 100644
@@ -1,4 +1 @@
-// SPDX-License-Identifier: GPL-2.0
-#define BUILD_VDSO32
-#include "fake_32bit_build.h"
-#include "../vclock_gettime.c"
+#include "common/vclock_gettime.c"
arch/x86/entry/vdso/vdso32/vdso32.lds.S +1 -1
index 8a3be07006bb..8a853543fc0d 100644
@@ -11,7 +11,7 @@
 
 #define BUILD_VDSO32
 
-#include "../vdso-layout.lds.S"
+#include "common/vdso-layout.lds.S"
 
 /* The ELF entry point can be used to set the AT_SYSINFO value.  */
 ENTRY(__kernel_vsyscall);
arch/x86/entry/vdso/vdso32/vgetcpu.c +1 -3
index 3a9791f5e998..00cc8325a020 100644
@@ -1,3 +1 @@
-// SPDX-License-Identifier: GPL-2.0
-#include "fake_32bit_build.h"
-#include "../vgetcpu.c"
+#include "common/vgetcpu.c"
arch/x86/entry/vdso/vdso64/Makefile new file +46 -0

46 lines not on this page · raw message

arch/x86/entry/vdso/vdso64/note.S new file +1 -0
new file mode 100644
index 000000000000..62d8aa51ce99
@@ -0,0 +1 @@
+#include "common/note.S"
arch/x86/entry/vdso/vdso64/vclock_gettime.c new file +1 -0
new file mode 100644
index 000000000000..1481f0021b9f
@@ -0,0 +1 @@
+#include "common/vclock_gettime.c"
arch/x86/entry/vdso/vdso64/vdso64.lds.S renamed from arch/x86/entry/vdso/vdso.lds.S +1 -1
similarity index 94%
rename from arch/x86/entry/vdso/vdso.lds.S
rename to arch/x86/entry/vdso/vdso64/vdso64.lds.S
index 0bab5f4af6d1..5ce3f2b6373a 100644
@@ -9,7 +9,7 @@
 
 #define BUILD_VDSO64
 
-#include "vdso-layout.lds.S"
+#include "common/vdso-layout.lds.S"
 
 /*
  * This controls what userland symbols we export from the vDSO.
arch/x86/entry/vdso/vdso64/vdsox32.lds.S renamed from arch/x86/entry/vdso/vdsox32.lds.S +1 -1
similarity index 92%
rename from arch/x86/entry/vdso/vdsox32.lds.S
rename to arch/x86/entry/vdso/vdso64/vdsox32.lds.S
index 16a8050a4fb6..3dbd20c8dacc 100644
@@ -9,7 +9,7 @@
 
 #define BUILD_VDSOX32
 
-#include "vdso-layout.lds.S"
+#include "common/vdso-layout.lds.S"
 
 /*
  * This controls what userland symbols we export from the vDSO.
arch/x86/entry/vdso/vdso64/vgetcpu.c new file +1 -0
new file mode 100644
index 000000000000..00cc8325a020
@@ -0,0 +1 @@
+#include "common/vgetcpu.c"
arch/x86/entry/vdso/vdso64/vgetrandom-chacha.S renamed from arch/x86/entry/vdso/vgetrandom-chacha.S +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vgetrandom-chacha.S
rename to arch/x86/entry/vdso/vdso64/vgetrandom-chacha.S
arch/x86/entry/vdso/vdso64/vgetrandom.c renamed from arch/x86/entry/vdso/vgetrandom.c +1 -1
similarity index 91%
rename from arch/x86/entry/vdso/vgetrandom.c
rename to arch/x86/entry/vdso/vdso64/vgetrandom.c
index 430862b8977c..6a95d36b12d9 100644
@@ -4,7 +4,7 @@
  */
 #include <linux/types.h>
 
-#include "../../../../lib/vdso/getrandom.c"
+#include "lib/vdso/getrandom.c"
 
 ssize_t __vdso_getrandom(void *buffer, size_t len, unsigned int flags, void *opaque_state, size_t opaque_len)
 {
arch/x86/entry/vdso/vdso64/vsgx.S renamed from arch/x86/entry/vdso/vsgx.S +0 -0
similarity index 100%
rename from arch/x86/entry/vdso/vsgx.S
rename to arch/x86/entry/vdso/vdso64/vsgx.S
Signature · 1 line
2.52.0
#30 lore raw

Re: [PATCH v4.1 03/10] x86/entry/vdso: refactor the vdso build

On Tue, 6 Jan 2026 13:18:36 -0800 "H. Peter Anvin" <[email protected]> wrote:

Hi everyone,

I ran the tip master branch through my AI review prompts and this one was flagged. These look right to me, apologies if it's noise:

Quoted 5 lines
arch/x86/entry/vdso/common/Makefile.include new file +0 -0
new file mode 100644
index 0000000000000..3514b4a6869b7

[ ... ]

arch/x86/entry/vdso/common/Makefile.include line 26

Quoted 6 lines
arch/x86/entry/vdso/common/Makefile.include new file +6 -0
+#
+# Options from KBUILD_[AC]FLAGS that should *NOT* be kept
+#
+flags-remove-y += \
+	-D__KERNEL__ -mcmodel=kernel -mregparm=3 \
+	-fno-pic -fno-PIC -fno-pie fno-PIE \
                                  ^^^^^^^

Should this be "-fno-PIE" with the leading dash? The other flags in this line all have the dash prefix, but "fno-PIE" is missing it. Without the dash, filter-out won't match the actual compiler flag.

arch/x86/entry/vdso/common/Makefile.include line 28

Quoted 2 lines
arch/x86/entry/vdso/common/Makefile.include new file +2 -0
+	-mfentry -pg \
+	$(RANDSTRUCT_CFLAGS) $(GCC_PLUGIN_CFLAGS) $(KSTACK_ERASE_CFLAGS) \
                              ^^^^^^^^^^^^^^^^^

Is GCC_PLUGIN_CFLAGS the right variable name here? The kbuild system defines GCC_PLUGINS_CFLAGS (with an 's') in scripts/Makefile.gcc-plugins. Other vDSO Makefiles (arm64, sparc, arm) use GCC_PLUGINS_CFLAGS. Without the fix, gcc plugin flags won't be filtered out when CONFIG_GCC_PLUGINS=y.

Quoted 2 lines
arch/x86/entry/vdso/common/Makefile.include new file +2 -0
+	$(RETPOLINE_CFLAGS) $(CC_FLAGS_LTO) $(CC_FLAGS_CFI) \
+	$(PADDING_CFLAGS)

[ ... ]

arch/x86/entry/vdso/common/Makefile.include line 48

Quoted 3 lines
arch/x86/entry/vdso/common/Makefile.include new file +3 -0
+# Reset cf protections enabled by compiler default
+flags-y += $(call cc-option, -fcf-protection=none)
+flags-$(X86_USER_SHADOW_STACK) += $(call cc-option, -fcf-protection=return)
           ^^^^^^^^^^^^^^^^^^^^^

Should this be CONFIG_X86_USER_SHADOW_STACK? The Kconfig symbol is defined as "config X86_USER_SHADOW_STACK" and kbuild exposes it as a CONFIG_ prefixed variable. The next lines show the pattern:

arch/x86/entry/vdso/common/Makefile.include line 52

Quoted 4 lines
arch/x86/entry/vdso/common/Makefile.include new file +4 -0
+# When user space IBT is supported, enable this.
+# flags-$(CONFIG_USER_IBT) += $(call cc-option, -fcf-protection=branch)
+
+flags-$(CONFIG_MITIGATION_RETPOLINE) += $(RETPOLINE_VDSO_CFLAGS)

Both of these correctly use the CONFIG_ prefix. Without it, the shadow stack cf-protection flag will never be added to vDSO builds even when CONFIG_X86_USER_SHADOW_STACK=y.

author maintainer
#31 lore raw

PATCH v4 4/10 x86/entry/vdso32: don't rely on int80_landing_pad for adjusting ip

In reply to H. Peter Anvin

There is no fundamental reason to use the int80_landing_pad symbol to adjust ip when moving the vdso. If ip falls within the vdso, and the vdso is moved, we should change the ip accordingly, regardless of mode or location within the vdso. This *currently* can only happen on 32 bits, but there isn't any reason not to do so generically.

Note that if this is ever possible from a vdso-internal call, then the user space stack will also needed to be adjusted (as well as the shadow stack, if enabled.) Fortunately this is not currently the case.

At the moment, we don't even consider other threads when moving the vdso. The assumption is that it is only used by process freeze/thaw for migration, where this is not an issue.

---
 arch/x86/entry/vdso/vma.c | 16 ++++++----------
 1 file changed, 6 insertions(+), 10 deletions(-)
arch/x86/entry/vdso/vma.c +6 -10
index 8f98c2d7c7a9..e7fd7517370f 100644
@@ -65,16 +65,12 @@ static vm_fault_t vdso_fault(const struct vm_special_mapping *sm,
 static void vdso_fix_landing(const struct vdso_image *image,
 		struct vm_area_struct *new_vma)
 {
-	if (in_ia32_syscall() && image == &vdso32_image) {
-		struct pt_regs *regs = current_pt_regs();
-		unsigned long vdso_land = image->sym_int80_landing_pad;
-		unsigned long old_land_addr = vdso_land +
-			(unsigned long)current->mm->context.vdso;
-
-		/* Fixing userspace landing - look at do_fast_syscall_32 */
-		if (regs->ip == old_land_addr)
-			regs->ip = new_vma->vm_start + vdso_land;
-	}
+	struct pt_regs *regs = current_pt_regs();
+	unsigned long ipoffset = regs->ip -
+		(unsigned long)current->mm->context.vdso;
+
+	if (ipoffset < image->size)
+		regs->ip = new_vma->vm_start + ipoffset;
 }
 
 static int vdso_mremap(const struct vm_special_mapping *sm,
Signature · 1 line
2.52.0
author maintainer
#32 lore raw

PATCH v4 5/10 x86/entry/vdso32: remove SYSCALL_ENTER_KERNEL macro in sigreturn.S

In reply to H. Peter Anvin

A macro SYSCALL_ENTER_KERNEL was defined in sigreturn.S, with the ability of overriding it. The override capability, however, is not used anywhere, and the macro name is potentially confusing because it seems to imply that sysenter/syscall could be used here, which is NOT
true: the sigreturn system calls MUST use int $0x80.

---
 arch/x86/entry/vdso/vdso32/sigreturn.S | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)
arch/x86/entry/vdso/vdso32/sigreturn.S +2 -6
index 1bd068f72d4c..965900c6763b 100644
@@ -3,10 +3,6 @@
 #include <asm/unistd_32.h>
 #include <asm/asm-offsets.h>
 
-#ifndef SYSCALL_ENTER_KERNEL
-#define	SYSCALL_ENTER_KERNEL	int $0x80
-#endif
-
 	.text
 	.globl __kernel_sigreturn
 	.type __kernel_sigreturn,@function
@@ -16,7 +12,7 @@ __kernel_sigreturn:
 .LSTART_sigreturn:
 	popl %eax		/* XXX does this mean it needs unwind info? */
 	movl $__NR_sigreturn, %eax
-	SYSCALL_ENTER_KERNEL
+	int $0x80
 .LEND_sigreturn:
 SYM_INNER_LABEL(vdso32_sigreturn_landing_pad, SYM_L_GLOBAL)
 	nop
@@ -28,7 +24,7 @@ SYM_INNER_LABEL(vdso32_sigreturn_landing_pad, SYM_L_GLOBAL)
 __kernel_rt_sigreturn:
 .LSTART_rt_sigreturn:
 	movl $__NR_rt_sigreturn, %eax
-	SYSCALL_ENTER_KERNEL
+	int $0x80
 .LEND_rt_sigreturn:
 SYM_INNER_LABEL(vdso32_rt_sigreturn_landing_pad, SYM_L_GLOBAL)
 	nop
Signature · 1 line
2.52.0
author maintainer
#33 lore raw

PATCH v4 6/10 x86/entry/vdso32: remove open-coded DWARF in sigreturn.S

In reply to H. Peter Anvin

The vdso32 sigreturn.S contains open-coded DWARF bytecode, which includes a hack for gdb to not try to step back to a previous call instruction when backtracing from a signal handler.

Neither of those are necessary anymore: the backtracing issue is handled by ".cfi_entry simple" and ".cfi_signal_frame", both of which have been supported for a very long time now, which allows the remaining frame to be built using regular .cfi annotations.

Add a few more register offsets to the signal frame just for good measure.

Replace the nop on fallthrough of the system call (which should never, ever happen) with a ud2a trap.

---
 arch/x86/entry/vdso/vdso32/sigreturn.S | 146 ++++++-------------------
 arch/x86/include/asm/dwarf2.h          |   1 +
 arch/x86/kernel/asm-offsets.c          |   6 +
 3 files changed, 39 insertions(+), 114 deletions(-)
arch/x86/entry/vdso/vdso32/sigreturn.S +32 -114
index 965900c6763b..25b0ac4b4bfe 100644
@@ -1,136 +1,54 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 #include <linux/linkage.h>
 #include <asm/unistd_32.h>
+#include <asm/dwarf2.h>
 #include <asm/asm-offsets.h>
 
+.macro STARTPROC_SIGNAL_FRAME sc
+	CFI_STARTPROC	simple
+	CFI_SIGNAL_FRAME
+	/* -4 as pretcode has already been popped */
+	CFI_DEF_CFA	esp,	\sc - 4
+	CFI_OFFSET	eip,    IA32_SIGCONTEXT_ip
+	CFI_OFFSET	eax,    IA32_SIGCONTEXT_ax
+	CFI_OFFSET	ebx,    IA32_SIGCONTEXT_bx
+	CFI_OFFSET	ecx,    IA32_SIGCONTEXT_cx
+	CFI_OFFSET	edx,    IA32_SIGCONTEXT_dx
+	CFI_OFFSET	esp,    IA32_SIGCONTEXT_sp
+	CFI_OFFSET	ebp,    IA32_SIGCONTEXT_bp
+	CFI_OFFSET	esi,    IA32_SIGCONTEXT_si
+	CFI_OFFSET	edi,    IA32_SIGCONTEXT_di
+	CFI_OFFSET	es,     IA32_SIGCONTEXT_es
+	CFI_OFFSET	cs,     IA32_SIGCONTEXT_cs
+	CFI_OFFSET	ss,     IA32_SIGCONTEXT_ss
+	CFI_OFFSET	ds,     IA32_SIGCONTEXT_ds
+	CFI_OFFSET	eflags, IA32_SIGCONTEXT_flags
+.endm
+
 	.text
 	.globl __kernel_sigreturn
 	.type __kernel_sigreturn,@function
-	nop /* this guy is needed for .LSTARTFDEDLSI1 below (watch for HACK) */
 	ALIGN
 __kernel_sigreturn:
-.LSTART_sigreturn:
-	popl %eax		/* XXX does this mean it needs unwind info? */
+	STARTPROC_SIGNAL_FRAME IA32_SIGFRAME_sigcontext
+	popl %eax
+	CFI_ADJUST_CFA_OFFSET -4
 	movl $__NR_sigreturn, %eax
 	int $0x80
-.LEND_sigreturn:
 SYM_INNER_LABEL(vdso32_sigreturn_landing_pad, SYM_L_GLOBAL)
-	nop
-	.size __kernel_sigreturn,.-.LSTART_sigreturn
+	ud2a
+	CFI_ENDPROC
+	.size __kernel_sigreturn,.-__kernel_sigreturn
 
 	.globl __kernel_rt_sigreturn
 	.type __kernel_rt_sigreturn,@function
 	ALIGN
 __kernel_rt_sigreturn:
-.LSTART_rt_sigreturn:
+	STARTPROC_SIGNAL_FRAME IA32_RT_SIGFRAME_sigcontext
 	movl $__NR_rt_sigreturn, %eax
 	int $0x80
-.LEND_rt_sigreturn:
 SYM_INNER_LABEL(vdso32_rt_sigreturn_landing_pad, SYM_L_GLOBAL)
-	nop
-	.size __kernel_rt_sigreturn,.-.LSTART_rt_sigreturn
-	.previous
-
-	.section .eh_frame,"a",@progbits
-.LSTARTFRAMEDLSI1:
-	.long .LENDCIEDLSI1-.LSTARTCIEDLSI1
-.LSTARTCIEDLSI1:
-	.long 0			/* CIE ID */
-	.byte 1			/* Version number */
-	.string "zRS"		/* NUL-terminated augmentation string */
-	.uleb128 1		/* Code alignment factor */
-	.sleb128 -4		/* Data alignment factor */
-	.byte 8			/* Return address register column */
-	.uleb128 1		/* Augmentation value length */
-	.byte 0x1b		/* DW_EH_PE_pcrel|DW_EH_PE_sdata4. */
-	.byte 0			/* DW_CFA_nop */
-	.align 4
-.LENDCIEDLSI1:
-	.long .LENDFDEDLSI1-.LSTARTFDEDLSI1 /* Length FDE */
-.LSTARTFDEDLSI1:
-	.long .LSTARTFDEDLSI1-.LSTARTFRAMEDLSI1 /* CIE pointer */
-	/* HACK: The dwarf2 unwind routines will subtract 1 from the
-	   return address to get an address in the middle of the
-	   presumed call instruction.  Since we didn't get here via
-	   a call, we need to include the nop before the real start
-	   to make up for it.  */
-	.long .LSTART_sigreturn-1-.	/* PC-relative start address */
-	.long .LEND_sigreturn-.LSTART_sigreturn+1
-	.uleb128 0			/* Augmentation */
-	/* What follows are the instructions for the table generation.
-	   We record the locations of each register saved.  This is
-	   complicated by the fact that the "CFA" is always assumed to
-	   be the value of the stack pointer in the caller.  This means
-	   that we must define the CFA of this body of code to be the
-	   saved value of the stack pointer in the sigcontext.  Which
-	   also means that there is no fixed relation to the other
-	   saved registers, which means that we must use DW_CFA_expression
-	   to compute their addresses.  It also means that when we
-	   adjust the stack with the popl, we have to do it all over again.  */
-
-#define do_cfa_expr(offset)						\
-	.byte 0x0f;			/* DW_CFA_def_cfa_expression */	\
-	.uleb128 1f-0f;			/*   length */			\
-0:	.byte 0x74;			/*     DW_OP_breg4 */		\
-	.sleb128 offset;		/*      offset */		\
-	.byte 0x06;			/*     DW_OP_deref */		\
-1:
-
-#define do_expr(regno, offset)						\
-	.byte 0x10;			/* DW_CFA_expression */		\
-	.uleb128 regno;			/*   regno */			\
-	.uleb128 1f-0f;			/*   length */			\
-0:	.byte 0x74;			/*     DW_OP_breg4 */		\
-	.sleb128 offset;		/*       offset */		\
-1:
-
-	do_cfa_expr(IA32_SIGCONTEXT_sp+4)
-	do_expr(0, IA32_SIGCONTEXT_ax+4)
-	do_expr(1, IA32_SIGCONTEXT_cx+4)
-	do_expr(2, IA32_SIGCONTEXT_dx+4)
-	do_expr(3, IA32_SIGCONTEXT_bx+4)
-	do_expr(5, IA32_SIGCONTEXT_bp+4)
-	do_expr(6, IA32_SIGCONTEXT_si+4)
-	do_expr(7, IA32_SIGCONTEXT_di+4)
-	do_expr(8, IA32_SIGCONTEXT_ip+4)
-
-	.byte 0x42	/* DW_CFA_advance_loc 2 -- nop; popl eax. */
-
-	do_cfa_expr(IA32_SIGCONTEXT_sp)
-	do_expr(0, IA32_SIGCONTEXT_ax)
-	do_expr(1, IA32_SIGCONTEXT_cx)
-	do_expr(2, IA32_SIGCONTEXT_dx)
-	do_expr(3, IA32_SIGCONTEXT_bx)
-	do_expr(5, IA32_SIGCONTEXT_bp)
-	do_expr(6, IA32_SIGCONTEXT_si)
-	do_expr(7, IA32_SIGCONTEXT_di)
-	do_expr(8, IA32_SIGCONTEXT_ip)
-
-	.align 4
-.LENDFDEDLSI1:
-
-	.long .LENDFDEDLSI2-.LSTARTFDEDLSI2 /* Length FDE */
-.LSTARTFDEDLSI2:
-	.long .LSTARTFDEDLSI2-.LSTARTFRAMEDLSI1 /* CIE pointer */
-	/* HACK: See above wrt unwind library assumptions.  */
-	.long .LSTART_rt_sigreturn-1-.	/* PC-relative start address */
-	.long .LEND_rt_sigreturn-.LSTART_rt_sigreturn+1
-	.uleb128 0			/* Augmentation */
-	/* What follows are the instructions for the table generation.
-	   We record the locations of each register saved.  This is
-	   slightly less complicated than the above, since we don't
-	   modify the stack pointer in the process.  */
-
-	do_cfa_expr(IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_sp)
-	do_expr(0, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_ax)
-	do_expr(1, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_cx)
-	do_expr(2, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_dx)
-	do_expr(3, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_bx)
-	do_expr(5, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_bp)
-	do_expr(6, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_si)
-	do_expr(7, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_di)
-	do_expr(8, IA32_RT_SIGFRAME_sigcontext-4 + IA32_SIGCONTEXT_ip)
-
-	.align 4
-.LENDFDEDLSI2:
+	ud2a
+	CFI_ENDPROC
+	.size __kernel_rt_sigreturn,.-__kernel_rt_sigreturn
 	.previous
arch/x86/include/asm/dwarf2.h +1 -0
index 302e11b15da8..09c9684d3ad6 100644
@@ -20,6 +20,7 @@
 #define CFI_RESTORE_STATE	.cfi_restore_state
 #define CFI_UNDEFINED		.cfi_undefined
 #define CFI_ESCAPE		.cfi_escape
+#define CFI_SIGNAL_FRAME	.cfi_signal_frame
 
 #ifndef BUILD_VDSO
 	/*
arch/x86/kernel/asm-offsets.c +6 -0
index 25fcde525c68..081816888f7a 100644
@@ -63,8 +63,14 @@ static void __used common(void)
 	OFFSET(IA32_SIGCONTEXT_bp, sigcontext_32, bp);
 	OFFSET(IA32_SIGCONTEXT_sp, sigcontext_32, sp);
 	OFFSET(IA32_SIGCONTEXT_ip, sigcontext_32, ip);
+	OFFSET(IA32_SIGCONTEXT_es, sigcontext_32, es);
+	OFFSET(IA32_SIGCONTEXT_cs, sigcontext_32, cs);
+	OFFSET(IA32_SIGCONTEXT_ss, sigcontext_32, ss);
+	OFFSET(IA32_SIGCONTEXT_ds, sigcontext_32, ds);
+	OFFSET(IA32_SIGCONTEXT_flags, sigcontext_32, flags);
 
 	BLANK();
+	OFFSET(IA32_SIGFRAME_sigcontext,    sigframe_ia32,    sc);
 	OFFSET(IA32_RT_SIGFRAME_sigcontext, rt_sigframe_ia32, uc.uc_mcontext);
 #endif
 
Signature · 1 line
2.52.0
#34 lore raw

Re: [PATCH v4.1 06/10] x86/entry/vdso32: remove open-coded DWARF in sigreturn.S

Hello Peter!

Quoted 8 lines · H. Peter Anvin

The vdso32 sigreturn.S contains open-coded DWARF bytecode, which includes a hack for gdb to not try to step back to a previous call instruction when backtracing from a signal handler.

Neither of those are necessary anymore: the backtracing issue is handled by ".cfi_entry simple" and ".cfi_signal_frame", both of which have been supported for a very long time now, which allows the remaining frame to be built using regular .cfi annotations.

Hopefully Glibc developers will do something similar for x86-64 __restore_rt() in Glibc sysdeps/unix/sysv/linux/x86_64/libc_sigaction.c.

Quoted 22 lines

Add a few more register offsets to the signal frame just for good measure.

Replace the nop on fallthrough of the system call (which should never, ever happen) with a ud2a trap.

arch/x86/entry/vdso/vdso32/sigreturn.S +3 -4
 
 	.text
 	.globl __kernel_sigreturn
 	.type __kernel_sigreturn,@function
-	nop /* this guy is needed for .LSTARTFDEDLSI1 below (watch for HACK) */
 	ALIGN
 __kernel_sigreturn:
-.LSTART_sigreturn:
-	popl %eax		/* XXX does this mean it needs unwind info? */
+	STARTPROC_SIGNAL_FRAME IA32_SIGFRAME_sigcontext
+	popl %eax
+	CFI_ADJUST_CFA_OFFSET -4
 	movl $__NR_sigreturn, %eax
 	int $0x80
-.LEND_sigreturn:

...

Quoted 9 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +1 -2
 	.globl __kernel_rt_sigreturn
 	.type __kernel_rt_sigreturn,@function
 	ALIGN
 __kernel_rt_sigreturn:
-.LSTART_rt_sigreturn:
+	STARTPROC_SIGNAL_FRAME IA32_RT_SIGFRAME_sigcontext
 	movl $__NR_rt_sigreturn, %eax
 	int $0x80
-.LEND_rt_sigreturn:

...

arch/x86/entry/vdso/vdso32/sigreturn.S line 40

Quoted 7 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +0 -7
-	.section .eh_frame,"a",@progbits
-.LSTARTFRAMEDLSI1:
-	.long .LENDCIEDLSI1-.LSTARTCIEDLSI1
-.LSTARTCIEDLSI1:
-	.long 0			/* CIE ID */
-	.byte 1			/* Version number */
-	.string "zRS"		/* NUL-terminated augmentation string */

Note that the "S" in "zRS" is the signal frame indication.

Quoted 7 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +0 -7
-	.uleb128 1		/* Code alignment factor */
-	.sleb128 -4		/* Data alignment factor */
-	.byte 8			/* Return address register column */
-	.uleb128 1		/* Augmentation value length */
-	.byte 0x1b		/* DW_EH_PE_pcrel|DW_EH_PE_sdata4. */
-	.byte 0			/* DW_CFA_nop */
-	.align 4

arch/x86/entry/vdso/vdso32/sigreturn.S line 57

Quoted 10 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +0 -10
-.LENDCIEDLSI1:
-	.long .LENDFDEDLSI1-.LSTARTFDEDLSI1 /* Length FDE */
-.LSTARTFDEDLSI1:
-	.long .LSTARTFDEDLSI1-.LSTARTFRAMEDLSI1 /* CIE pointer */
-	/* HACK: The dwarf2 unwind routines will subtract 1 from the
-	   return address to get an address in the middle of the
-	   presumed call instruction.  Since we didn't get here via
-	   a call, we need to include the nop before the real start
-	   to make up for it.  */
-	.long .LSTART_sigreturn-1-.	/* PC-relative start address */

Your version does no longer have this nop nor does the FDE start one byte earlier. Isn't that required for unwinders any longer? See excerpt from dumped DWARF and disassembly for __kernel_sigreturn() below.

Quoted 2 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +0 -2
-	.long .LEND_sigreturn-.LSTART_sigreturn+1
-	.uleb128 0			/* Augmentation */

...

arch/x86/entry/vdso/vdso32/sigreturn.S line 116

Quoted 8 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +0 -8
-	.align 4
-.LENDFDEDLSI1:
-
-	.long .LENDFDEDLSI2-.LSTARTFDEDLSI2 /* Length FDE */
-.LSTARTFDEDLSI2:
-	.long .LSTARTFDEDLSI2-.LSTARTFRAMEDLSI1 /* CIE pointer */
-	/* HACK: See above wrt unwind library assumptions.  */
-	.long .LSTART_rt_sigreturn-1-.	/* PC-relative start address */

Ditto.

arch/x86/entry/vdso/vdso32/sigreturn.S line 118

Quoted 2 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +0 -2
-	.long .LEND_rt_sigreturn-.LSTART_rt_sigreturn+1
-	.uleb128 0			/* Augmentation */

Excerpt from dump of DWARF and disassembly with your patch:

$ objdump -d -Wf arch/x86/entry/vdso/vdso32/vdso32.so.dbg
...
000001cc 0000003c 00000000 CIE   <-- CIE for __kernel_sigreturn
  Version:               1
  Augmentation:          "zRS"
  Code alignment factor: 1
  Data alignment factor: -4
  Return address column: 8
  Augmentation data:     1b
  DW_CFA_def_cfa: r4 (esp) ofs 4
  DW_CFA_offset_extended_sf: r8 (eip) at cfa+56
  DW_CFA_offset_extended_sf: r0 (eax) at cfa+44
  DW_CFA_offset_extended_sf: r3 (ebx) at cfa+32
  DW_CFA_offset_extended_sf: r1 (ecx) at cfa+40
  DW_CFA_offset_extended_sf: r2 (edx) at cfa+36
  DW_CFA_offset_extended_sf: r4 (esp) at cfa+28
  DW_CFA_offset_extended_sf: r5 (ebp) at cfa+24
  DW_CFA_offset_extended_sf: r6 (esi) at cfa+20
  DW_CFA_offset_extended_sf: r7 (edi) at cfa+16
  DW_CFA_offset_extended_sf: r40 (es) at cfa+8
  DW_CFA_offset_extended_sf: r41 (cs) at cfa+60
  DW_CFA_offset_extended_sf: r42 (ss) at cfa+72
  DW_CFA_offset_extended_sf: r43 (ds) at cfa+12
  DW_CFA_offset_extended_sf: r9 (eflags) at cfa+64
  DW_CFA_nop

0000020c 00000010 00000044 FDE cie=000001cc pc=00001a40..00001a4a <-- FDE for __kernel_sigreturn
DW_CFA_advance_loc: 1 to 00001a41
DW_CFA_def_cfa_offset: 0

[ The FDE covers the range [1a40..1a4a[. Previously it would have started one byte earlier (at the nop), so that the range would have been [1a3f..1a4a[. This is/was required for unwinders that always subtract one from the unwound return address, so that it points into the instruction that invoked the function (e.g. call) instead of behind it, in case it was invoked by a non-returning function. Such an unwinder would now lookup IP=1a3f as belonging to int80_landing_pad (and use the DWARF rules applicable to its last instruction) instead of __kernel_sigreturn (and its rules). Likewise for __kernel_rt_sigreturn. ]

...
00001a3c <int80_landing_pad>:

    1a3c:       5d                      pop    %ebp
    1a3d:       5a                      pop    %edx
    1a3e:       59                      pop    %ecx
    1a3f:       c3                      ret

00001a40 <__kernel_sigreturn>:

    1a40:       58                      pop    %eax
    1a41:       b8 77 00 00 00          mov    $0x77,%eax
    1a46:       cd 80                   int    $0x80

00001a48 <vdso32_sigreturn_landing_pad>:

    1a48:       0f 0b                   ud2
    1a4a:       8d b6 00 00 00 00       lea    0x0(%esi),%esi

Excerpt without your patch:

$ objdump -d -Wf arch/x86/entry/vdso/vdso32/vdso32.so.dbg
...
000001cc 00000010 00000000 CIE  <-- CIE for __kernel_sigreturn and __kernel_rt_sigreturn
  Version:               1
  Augmentation:          "zRS"
  Code alignment factor: 1
  Data alignment factor: -4
  Return address column: 8
  Augmentation data:     1b
  DW_CFA_nop
  DW_CFA_nop

000001e0 00000068 00000018 FDE cie=000001cc pc=00001a6f..00001a78 <-- FDE for __kernel_sigreturn
DW_CFA_def_cfa_expression (DW_OP_breg4 (esp): 32; DW_OP_deref)
DW_CFA_expression: r0 (eax) (DW_OP_breg4 (esp): 48)
DW_CFA_expression: r1 (ecx) (DW_OP_breg4 (esp): 44)
DW_CFA_expression: r2 (edx) (DW_OP_breg4 (esp): 40)
DW_CFA_expression: r3 (ebx) (DW_OP_breg4 (esp): 36)
DW_CFA_expression: r5 (ebp) (DW_OP_breg4 (esp): 28)
DW_CFA_expression: r6 (esi) (DW_OP_breg4 (esp): 24)
DW_CFA_expression: r7 (edi) (DW_OP_breg4 (esp): 20)
DW_CFA_expression: r8 (eip) (DW_OP_breg4 (esp): 60)
DW_CFA_advance_loc: 2 to 00001a71
DW_CFA_def_cfa_expression (DW_OP_breg4 (esp): 28; DW_OP_deref)
DW_CFA_expression: r0 (eax) (DW_OP_breg4 (esp): 44)
DW_CFA_expression: r1 (ecx) (DW_OP_breg4 (esp): 40)
DW_CFA_expression: r2 (edx) (DW_OP_breg4 (esp): 36)
DW_CFA_expression: r3 (ebx) (DW_OP_breg4 (esp): 32)
DW_CFA_expression: r5 (ebp) (DW_OP_breg4 (esp): 24)
DW_CFA_expression: r6 (esi) (DW_OP_breg4 (esp): 20)
DW_CFA_expression: r7 (edi) (DW_OP_breg4 (esp): 16)
DW_CFA_expression: r8 (eip) (DW_OP_breg4 (esp): 56)

[ See how the FDE for __kernel_sigreturn covers the range [1a6f..1a78[. An unwinder that always subtracts one from the return address would lookup IP=1a6f as belonging to __kernel_sigreturn (and use the DWARF rules applicable to the nop preceeding its symbol). Likewise for __kernel_rt_sigreturn. Or is that no longer true? ]

...
00001a5c <int80_landing_pad>:

    1a5c:       5d                      pop    %ebp
    1a5d:       5a                      pop    %edx
    1a5e:       59                      pop    %ecx
    1a5f:       c3                      ret
    1a60:       90                      nop
    1a61:       8d b4 26 00 00 00 00    lea    0x0(%esi,%eiz,1),%esi
    1a68:       2e 8d b4 26 00 00 00    lea    %cs:0x0(%esi,%eiz,1),%esi
    1a6f:       00

00001a70 <__kernel_sigreturn>:

    1a70:       58                      pop    %eax
    1a71:       b8 77 00 00 00          mov    $0x77,%eax
    1a76:       cd 80                   int    $0x80

Thanks and regards,
Jens

Signature · 6 lines
Jens Remus
Linux on Z Development (D3303)
[email protected] / [email protected]

IBM Deutschland Research & Development GmbH; Vorsitzender des Aufsichtsrats: Wolfgang Wendt; Geschäftsführung: David Faller; Sitz der Gesellschaft: Ehningen; Registergericht: Amtsgericht Stuttgart, HRB 243294
IBM Data Privacy Statement: https://www.ibm.com/privacy/
author maintainer
#35 lore raw

Re: [PATCH v4.1 06/10] x86/entry/vdso32: remove open-coded DWARF in sigreturn.S

Quoted 215 lines · Jens Remus

Hello Peter!

Quoted 8 lines · H. Peter Anvin

The vdso32 sigreturn.S contains open-coded DWARF bytecode, which includes a hack for gdb to not try to step back to a previous call instruction when backtracing from a signal handler.

Neither of those are necessary anymore: the backtracing issue is handled by ".cfi_entry simple" and ".cfi_signal_frame", both of which have been supported for a very long time now, which allows the remaining frame to be built using regular .cfi annotations.

Hopefully Glibc developers will do something similar for x86-64 __restore_rt() in Glibc sysdeps/unix/sysv/linux/x86_64/libc_sigaction.c.

Quoted 22 lines

Add a few more register offsets to the signal frame just for good measure.

Replace the nop on fallthrough of the system call (which should never, ever happen) with a ud2a trap.

arch/x86/entry/vdso/vdso32/sigreturn.S +3 -4
 
 	.text
 	.globl __kernel_sigreturn
 	.type __kernel_sigreturn,@function
-	nop /* this guy is needed for .LSTARTFDEDLSI1 below (watch for HACK) */
 	ALIGN
 __kernel_sigreturn:
-.LSTART_sigreturn:
-	popl %eax		/* XXX does this mean it needs unwind info? */
+	STARTPROC_SIGNAL_FRAME IA32_SIGFRAME_sigcontext
+	popl %eax
+	CFI_ADJUST_CFA_OFFSET -4
 	movl $__NR_sigreturn, %eax
 	int $0x80
-.LEND_sigreturn:

...

Quoted 9 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +1 -2
 	.globl __kernel_rt_sigreturn
 	.type __kernel_rt_sigreturn,@function
 	ALIGN
 __kernel_rt_sigreturn:
-.LSTART_rt_sigreturn:
+	STARTPROC_SIGNAL_FRAME IA32_RT_SIGFRAME_sigcontext
 	movl $__NR_rt_sigreturn, %eax
 	int $0x80
-.LEND_rt_sigreturn:

...

arch/x86/entry/vdso/vdso32/sigreturn.S line 40

Quoted 7 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +0 -7
-	.section .eh_frame,"a",@progbits
-.LSTARTFRAMEDLSI1:
-	.long .LENDCIEDLSI1-.LSTARTCIEDLSI1
-.LSTARTCIEDLSI1:
-	.long 0			/* CIE ID */
-	.byte 1			/* Version number */
-	.string "zRS"		/* NUL-terminated augmentation string */

Note that the "S" in "zRS" is the signal frame indication.

Quoted 7 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +0 -7
-	.uleb128 1		/* Code alignment factor */
-	.sleb128 -4		/* Data alignment factor */
-	.byte 8			/* Return address register column */
-	.uleb128 1		/* Augmentation value length */
-	.byte 0x1b		/* DW_EH_PE_pcrel|DW_EH_PE_sdata4. */
-	.byte 0			/* DW_CFA_nop */
-	.align 4

arch/x86/entry/vdso/vdso32/sigreturn.S line 57

Quoted 10 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +0 -10
-.LENDCIEDLSI1:
-	.long .LENDFDEDLSI1-.LSTARTFDEDLSI1 /* Length FDE */
-.LSTARTFDEDLSI1:
-	.long .LSTARTFDEDLSI1-.LSTARTFRAMEDLSI1 /* CIE pointer */
-	/* HACK: The dwarf2 unwind routines will subtract 1 from the
-	   return address to get an address in the middle of the
-	   presumed call instruction.  Since we didn't get here via
-	   a call, we need to include the nop before the real start
-	   to make up for it.  */
-	.long .LSTART_sigreturn-1-.	/* PC-relative start address */

Your version does no longer have this nop nor does the FDE start one byte earlier. Isn't that required for unwinders any longer? See excerpt from dumped DWARF and disassembly for __kernel_sigreturn() below.

Quoted 2 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +0 -2
-	.long .LEND_sigreturn-.LSTART_sigreturn+1
-	.uleb128 0			/* Augmentation */

...

arch/x86/entry/vdso/vdso32/sigreturn.S line 116

Quoted 8 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +0 -8
-	.align 4
-.LENDFDEDLSI1:
-
-	.long .LENDFDEDLSI2-.LSTARTFDEDLSI2 /* Length FDE */
-.LSTARTFDEDLSI2:
-	.long .LSTARTFDEDLSI2-.LSTARTFRAMEDLSI1 /* CIE pointer */
-	/* HACK: See above wrt unwind library assumptions.  */
-	.long .LSTART_rt_sigreturn-1-.	/* PC-relative start address */

Ditto.

arch/x86/entry/vdso/vdso32/sigreturn.S line 118

Quoted 2 lines
arch/x86/entry/vdso/vdso32/sigreturn.S +0 -2
-	.long .LEND_rt_sigreturn-.LSTART_rt_sigreturn+1
-	.uleb128 0			/* Augmentation */

Excerpt from dump of DWARF and disassembly with your patch:

$ objdump -d -Wf arch/x86/entry/vdso/vdso32/vdso32.so.dbg
...
000001cc 0000003c 00000000 CIE   <-- CIE for __kernel_sigreturn
 Version:               1
 Augmentation:          "zRS"
 Code alignment factor: 1
 Data alignment factor: -4
 Return address column: 8
 Augmentation data:     1b
 DW_CFA_def_cfa: r4 (esp) ofs 4
 DW_CFA_offset_extended_sf: r8 (eip) at cfa+56
 DW_CFA_offset_extended_sf: r0 (eax) at cfa+44
 DW_CFA_offset_extended_sf: r3 (ebx) at cfa+32
 DW_CFA_offset_extended_sf: r1 (ecx) at cfa+40
 DW_CFA_offset_extended_sf: r2 (edx) at cfa+36
 DW_CFA_offset_extended_sf: r4 (esp) at cfa+28
 DW_CFA_offset_extended_sf: r5 (ebp) at cfa+24
 DW_CFA_offset_extended_sf: r6 (esi) at cfa+20
 DW_CFA_offset_extended_sf: r7 (edi) at cfa+16
 DW_CFA_offset_extended_sf: r40 (es) at cfa+8
 DW_CFA_offset_extended_sf: r41 (cs) at cfa+60
 DW_CFA_offset_extended_sf: r42 (ss) at cfa+72
 DW_CFA_offset_extended_sf: r43 (ds) at cfa+12
 DW_CFA_offset_extended_sf: r9 (eflags) at cfa+64
 DW_CFA_nop

0000020c 00000010 00000044 FDE cie=000001cc pc=00001a40..00001a4a <-- FDE for __kernel_sigreturn
DW_CFA_advance_loc: 1 to 00001a41
DW_CFA_def_cfa_offset: 0

[ The FDE covers the range [1a40..1a4a[. Previously it would have started one byte earlier (at the nop), so that the range would have been [1a3f..1a4a[. This is/was required for unwinders that always subtract one from the unwound return address, so that it points into the instruction that invoked the function (e.g. call) instead of behind it, in case it was invoked by a non-returning function. Such an unwinder would now lookup IP=1a3f as belonging to int80_landing_pad (and use the DWARF rules applicable to its last instruction) instead of __kernel_sigreturn (and its rules). Likewise for __kernel_rt_sigreturn. ]

...
00001a3c <int80_landing_pad>:

   1a3c:       5d                      pop    %ebp
   1a3d:       5a                      pop    %edx
   1a3e:       59                      pop    %ecx
   1a3f:       c3                      ret

00001a40 <__kernel_sigreturn>:

   1a40:       58                      pop    %eax
   1a41:       b8 77 00 00 00          mov    $0x77,%eax
   1a46:       cd 80                   int    $0x80

00001a48 <vdso32_sigreturn_landing_pad>:

   1a48:       0f 0b                   ud2
   1a4a:       8d b6 00 00 00 00       lea    0x0(%esi),%esi

Excerpt without your patch:

$ objdump -d -Wf arch/x86/entry/vdso/vdso32/vdso32.so.dbg
...
000001cc 00000010 00000000 CIE  <-- CIE for __kernel_sigreturn and __kernel_rt_sigreturn
 Version:               1
 Augmentation:          "zRS"
 Code alignment factor: 1
 Data alignment factor: -4
 Return address column: 8
 Augmentation data:     1b
 DW_CFA_nop
 DW_CFA_nop

000001e0 00000068 00000018 FDE cie=000001cc pc=00001a6f..00001a78 <-- FDE for __kernel_sigreturn
DW_CFA_def_cfa_expression (DW_OP_breg4 (esp): 32; DW_OP_deref)
DW_CFA_expression: r0 (eax) (DW_OP_breg4 (esp): 48)
DW_CFA_expression: r1 (ecx) (DW_OP_breg4 (esp): 44)
DW_CFA_expression: r2 (edx) (DW_OP_breg4 (esp): 40)
DW_CFA_expression: r3 (ebx) (DW_OP_breg4 (esp): 36)
DW_CFA_expression: r5 (ebp) (DW_OP_breg4 (esp): 28)
DW_CFA_expression: r6 (esi) (DW_OP_breg4 (esp): 24)
DW_CFA_expression: r7 (edi) (DW_OP_breg4 (esp): 20)
DW_CFA_expression: r8 (eip) (DW_OP_breg4 (esp): 60)
DW_CFA_advance_loc: 2 to 00001a71
DW_CFA_def_cfa_expression (DW_OP_breg4 (esp): 28; DW_OP_deref)
DW_CFA_expression: r0 (eax) (DW_OP_breg4 (esp): 44)
DW_CFA_expression: r1 (ecx) (DW_OP_breg4 (esp): 40)
DW_CFA_expression: r2 (edx) (DW_OP_breg4 (esp): 36)
DW_CFA_expression: r3 (ebx) (DW_OP_breg4 (esp): 32)
DW_CFA_expression: r5 (ebp) (DW_OP_breg4 (esp): 24)
DW_CFA_expression: r6 (esi) (DW_OP_breg4 (esp): 20)
DW_CFA_expression: r7 (edi) (DW_OP_breg4 (esp): 16)
DW_CFA_expression: r8 (eip) (DW_OP_breg4 (esp): 56)

[ See how the FDE for __kernel_sigreturn covers the range [1a6f..1a78[. An unwinder that always subtracts one from the return address would lookup IP=1a6f as belonging to __kernel_sigreturn (and use the DWARF rules applicable to the nop preceeding its symbol). Likewise for __kernel_rt_sigreturn. Or is that no longer true? ]

...
00001a5c <int80_landing_pad>:

   1a5c:       5d                      pop    %ebp
   1a5d:       5a                      pop    %edx
   1a5e:       59                      pop    %ecx
   1a5f:       c3                      ret
   1a60:       90                      nop
   1a61:       8d b4 26 00 00 00 00    lea    0x0(%esi,%eiz,1),%esi
   1a68:       2e 8d b4 26 00 00 00    lea    %cs:0x0(%esi,%eiz,1),%esi
   1a6f:       00
Quoted 7 lines

00001a70 <__kernel_sigreturn>:

   1a70:       58                      pop    %eax
   1a71:       b8 77 00 00 00          mov    $0x77,%eax
   1a76:       cd 80                   int    $0x80

Thanks and regards,
Jens

That hack dates back from before the signal frame extension. It is no longer necessary.

#36 lore raw

Re: [PATCH v4.1 06/10] x86/entry/vdso32: remove open-coded DWARF in sigreturn.S

Quoted 2 lines · H. Peter Anvin

That hack dates back from before the signal frame extension. It is no longer necessary.

Unfortunately at least it seems libgcc unwinder does not handle the signal frame extension properly. The code reads:

fde = _Unwind_Find_FDE (context->ra + _Unwind_IsSignalFrame (context) - 1,

                          &context->bases);
  if (fde == NULL)
    {    

#ifdef MD_FALLBACK_FRAME_STATE_FOR
/* Couldn't find frame unwind info for this function. Try a
target-specific fallback mechanism. This will necessarily not provide a personality routine or LSDA. */
return MD_FALLBACK_FRAME_STATE_FOR (context, fs);

#else
      return _URC_END_OF_STACK;
#endif
    }    
  fs->pc = context->bases.func;

  cie = get_cie (fde);
  insn = extract_cie_info (cie, context, fs); 

Thus, it indeed attempts to avoid subtracting 1 for a signal frame, but ... _Unwind_IsSignalFrame (context) actually extracts a flag in context which will only be raised up by extract_cie_info.

Or am I missing something here?

Signature · 1 line
Xi Ruoyao <[email protected]>
author maintainer
#37 lore raw

Re: [PATCH v4.1 06/10] x86/entry/vdso32: remove open-coded DWARF in sigreturn.S

Quoted 23 lines · Xi Ruoyao
Quoted 2 lines · H. Peter Anvin

That hack dates back from before the signal frame extension. It is no longer necessary.

Unfortunately at least it seems libgcc unwinder does not handle the signal frame extension properly. The code reads:

fde = _Unwind_Find_FDE (context->ra + _Unwind_IsSignalFrame (context) - 1,

                         &context->bases);

if (fde == NULL)
{
#ifdef MD_FALLBACK_FRAME_STATE_FOR
/* Couldn't find frame unwind info for this function. Try a
target-specific fallback mechanism. This will necessarily not provide a personality routine or LSDA. */
return MD_FALLBACK_FRAME_STATE_FOR (context, fs);

#else
     return _URC_END_OF_STACK;
#endif
   }    

fs->pc = context->bases.func;

Quoted 9 lines

cie = get_cie (fde);
insn = extract_cie_info (cie, context, fs);

Thus, it indeed attempts to avoid subtracting 1 for a signal frame, but ... _Unwind_IsSignalFrame (context) actually extracts a flag in context which will only be raised up by extract_cie_info.

Or am I missing something here?

Oh, good grief...

How does this possibly work on non-x86 platforms?

#38 lore raw

Re: [PATCH v4.1 06/10] x86/entry/vdso32: remove open-coded DWARF in sigreturn.S

Quoted 31 lines · H. Peter Anvin
Quoted 30 lines · Xi Ruoyao
Quoted 2 lines · H. Peter Anvin

That hack dates back from before the signal frame extension. It is no longer necessary.

Unfortunately at least it seems libgcc unwinder does not handle the signal frame extension properly.  The code reads:

 fde = _Unwind_Find_FDE (context->ra + _Unwind_IsSignalFrame (context) - 1, &context->bases);
 if (fde == NULL)
   {
#ifdef MD_FALLBACK_FRAME_STATE_FOR
     /* Couldn't find frame unwind info for this function.  Try a target-specific fallback mechanism.  This will necessarily not provide a personality routine or LSDA.  */
     return MD_FALLBACK_FRAME_STATE_FOR (context, fs);
#else
     return _URC_END_OF_STACK;
#endif
   }

 fs->pc = context->bases.func;

 cie = get_cie (fde);
 insn = extract_cie_info (cie, context, fs);

Thus, it indeed attempts to avoid subtracting 1 for a signal frame, but ... _Unwind_IsSignalFrame (context) actually extracts a flag in context which will only be raised up by extract_cie_info.

Quoted 6 lines
Quoted 2 lines

Or am I missing something here?

Oh, good grief...

How does this possibly work on non-x86 platforms?

On ARM64 the vdso does not have eh_frame_hdr at all, on LoongArch eh_frame_hdr is empty (note that an ampty en_frame_hdr is actually buggy and I'm trying to fix it), so _Unwind_Find_FDE returns NULL and libgcc falls back to MD_FALLBACK_FRAME_STATE_FOR, which handles the sigreturn trampoline using some machine-dependant logic.

On RISC-V things are more theatrical: the sigreturn trampoline happens to be at the beginning of the vdso .text section, so after subtracting 1 from the PC, the result is out of the .text section and so not in any FDE. Thus _Unwind_Find_FDE returns NULL and libgcc again falls back to MD_FALLBACK_FRAME_STATE_FOR.

If the RISC-V sigreturn trampoline was not the first in .text, subtracting 1 would cause the PC to be in the FDE of the previous function and then _Unwind_Find_FDE would return that FDE, then RISC-V would have some big trouble.

I've not taken a serious look at other architectures yet.

Signature · 1 line
Xi Ruoyao <[email protected]>
author maintainer
#39 lore raw

Re: [PATCH v4.1 06/10] x86/entry/vdso32: remove open-coded DWARF in sigreturn.S

Quoted 17 lines · Xi Ruoyao
Quoted 4 lines

Oh, good grief...

How does this possibly work on non-x86 platforms?

On ARM64 the vdso does not have eh_frame_hdr at all, on LoongArch eh_frame_hdr is empty (note that an ampty en_frame_hdr is actually buggy and I'm trying to fix it), so _Unwind_Find_FDE returns NULL and libgcc falls back to MD_FALLBACK_FRAME_STATE_FOR, which handles the sigreturn trampoline using some machine-dependant logic.

On RISC-V things are more theatrical: the sigreturn trampoline happens to be at the beginning of the vdso .text section, so after subtracting 1 from the PC, the result is out of the .text section and so not in any FDE. Thus _Unwind_Find_FDE returns NULL and libgcc again falls back to MD_FALLBACK_FRAME_STATE_FOR.

Quoted 7 lines

If the RISC-V sigreturn trampoline was not the first in .text, subtracting 1 would cause the PC to be in the FDE of the previous function and then _Unwind_Find_FDE would return that FDE, then RISC-V would have some big trouble.

I've not taken a serious look at other architectures yet.

In other words, it is broken on pretty much all architectures and it ends up using the hacky fallbacks on every single platform...

-hpa

author maintainer
#40 lore raw

Re: [PATCH v4.1 06/10] x86/entry/vdso32: remove open-coded DWARF in sigreturn.S

In reply to Xi Ruoyao
Quoted 26 lines · Xi Ruoyao
Quoted 2 lines · H. Peter Anvin

That hack dates back from before the signal frame extension. It is no longer necessary.

Unfortunately at least it seems libgcc unwinder does not handle the signal frame extension properly. The code reads:

fde = _Unwind_Find_FDE (context->ra + _Unwind_IsSignalFrame (context) - 1,

                          &context->bases);
  if (fde == NULL)
    {    

#ifdef MD_FALLBACK_FRAME_STATE_FOR
/* Couldn't find frame unwind info for this function. Try a
target-specific fallback mechanism. This will necessarily not provide a personality routine or LSDA. */
return MD_FALLBACK_FRAME_STATE_FOR (context, fs);

#else
      return _URC_END_OF_STACK;
#endif
    }    
  fs->pc = context->bases.func;

  cie = get_cie (fde);
  insn = extract_cie_info (cie, context, fs); 
Quoted 6 lines

Thus, it indeed attempts to avoid subtracting 1 for a signal frame, but ... _Unwind_IsSignalFrame (context) actually extracts a flag in context which will only be raised up by extract_cie_info.

Or am I missing something here?

Well, it looks like this might be a non-manifest bug on x86 *IN THIS CASE*, because the fallback handler looks explicitly for the byte sequences:

	popl %eax
	movl $__NR_ia32_sigreturn, %eax
	int $0x80

	movl $__NR_ia32_rt_sigreturn, %eax
	int $0x80

	movq $__NR_x64_rt_sigreturn, %rax	/* !!!! */
	syscall

... so the vdso patchset is still OK.

Hard-coding a reliance on movq in the 64-bit sequence some serious brain damage...

Could you file a gcc bug on this?

-hpa

#41 lore raw

Re: [PATCH v4.1 06/10] x86/entry/vdso32: remove open-coded DWARF in sigreturn.S

Quoted 27 lines · H. Peter Anvin
Quoted 26 lines · Xi Ruoyao
Quoted 2 lines · H. Peter Anvin

That hack dates back from before the signal frame extension. It is no longer necessary.

Unfortunately at least it seems libgcc unwinder does not handle the signal frame extension properly.  The code reads:

  fde = _Unwind_Find_FDE (context->ra + _Unwind_IsSignalFrame (context) - 1, &context->bases);
  if (fde == NULL)
    {
#ifdef MD_FALLBACK_FRAME_STATE_FOR
      /* Couldn't find frame unwind info for this function.  Try a target-specific fallback mechanism.  This will necessarily not provide a personality routine or LSDA.  */
      return MD_FALLBACK_FRAME_STATE_FOR (context, fs);
#else
      return _URC_END_OF_STACK;
#endif
    }

  fs->pc = context->bases.func;

  cie = get_cie (fde);
  insn = extract_cie_info (cie, context, fs);

Quoted 8 lines
Quoted 6 lines

Thus, it indeed attempts to avoid subtracting 1 for a signal frame, but ... _Unwind_IsSignalFrame (context) actually extracts a flag in context which will only be raised up by extract_cie_info.

Or am I missing something here?

Well, it looks like this might be a non-manifest bug on x86 *IN THIS CASE*,

But only if after the patch _Unwind_Find_FDE returns NULL here (instead of a FDE of another function). I've not verified it on x86. If glibc tests (specifically misc/tst-sigcontext-get_pc) work fine things should be fine.

/* snip */

Quoted 1 line

Could you file a gcc bug on this?

Will do.

Signature · 1 line
Xi Ruoyao <[email protected]>
author maintainer
#42 lore raw

Re: [PATCH v4.1 06/10] x86/entry/vdso32: remove open-coded DWARF in sigreturn.S

Quoted 8 lines · Xi Ruoyao
Quoted 2 lines

Well, it looks like this might be a non-manifest bug on x86 *IN THIS CASE*,

But only if after the patch _Unwind_Find_FDE returns NULL here (instead of a FDE of another function). I've not verified it on x86. If glibc tests (specifically misc/tst-sigcontext-get_pc) work fine things should be fine.

The x86 vdso should be fine because of the alignment padding, but we should probably add an explicit pad before just in case.

What a mess.

-hpa

author maintainer
#43 lore raw

PATCH v4 7/10 x86/entry/vdso: include GNU_PROPERTY and GNU_STACK PHDRs

In reply to H. Peter Anvin

Currently the vdso doesn't include .note.gnu.property or a GNU noexec stack annotation (the -z noexecstack in the linker script is ineffective because we specify PHDRs explicitly.)

The motivation is that the dynamic linker currently do not check these.

However, this is a weak excuse: the vdso*.so are also supposed to be usable at link libraries, and there is no reason why the dynamic linker might not want or need to check these in the future, so add them back in -- it is trivial enough.

Use symbolic constants for the PHDR permission flags.

[ v4: drop unrelated formatting changes ]
[ v4.1: drop one last bogus formatting change (Brian Gerst) ]

---
 arch/x86/entry/vdso/common/vdso-layout.lds.S | 36 ++++++++++++--------
 1 file changed, 22 insertions(+), 14 deletions(-)
arch/x86/entry/vdso/common/vdso-layout.lds.S +22 -14
index ec1ac191a057..f050fd723712 100644
@@ -47,18 +47,18 @@ SECTIONS
 		*(.gnu.linkonce.b.*)
 	}						:text
 
-	/*
-	 * Discard .note.gnu.property sections which are unused and have
-	 * different alignment requirement from vDSO note sections.
-	 */
-	/DISCARD/ : {
+	.note.gnu.property : {
 		*(.note.gnu.property)
-	}
-	.note		: { *(.note.*) }		:text	:note
-
-	.eh_frame_hdr	: { *(.eh_frame_hdr) }		:text	:eh_frame_hdr
-	.eh_frame	: { KEEP (*(.eh_frame)) }	:text
+	}					:text :note :gnu_property
+	.note		: {
+		*(.note*)
+	}					:text :note
 
+	.eh_frame_hdr	: { *(.eh_frame_hdr) }	:text :eh_frame_hdr
+	.eh_frame	: {
+		KEEP (*(.eh_frame))
+		*(.eh_frame.*)
+	}					:text
 
 	/*
 	 * Text is well-separated from actual data: there's plenty of
@@ -87,15 +87,23 @@ SECTIONS
  * Very old versions of ld do not recognize this name token; use the constant.
  */
 #define PT_GNU_EH_FRAME	0x6474e550
+#define PT_GNU_STACK	0x6474e551
+#define PT_GNU_PROPERTY	0x6474e553
 
 /*
  * We must supply the ELF program headers explicitly to get just one
  * PT_LOAD segment, and set the flags explicitly to make segments read-only.
  */
+#define PF_R	FLAGS(4)
+#define PF_RW	FLAGS(6)
+#define PF_RX	FLAGS(5)
+
 PHDRS
 {
-	text		PT_LOAD		FLAGS(5) FILEHDR PHDRS; /* PF_R|PF_X */
-	dynamic		PT_DYNAMIC	FLAGS(4);		/* PF_R */
-	note		PT_NOTE		FLAGS(4);		/* PF_R */
-	eh_frame_hdr	PT_GNU_EH_FRAME;
+	text		PT_LOAD		PF_RX FILEHDR PHDRS;
+	dynamic		PT_DYNAMIC	PF_R;
+	note		PT_NOTE		PF_R;
+	eh_frame_hdr	PT_GNU_EH_FRAME PF_R;
+	gnu_stack	PT_GNU_STACK	PF_RW;
+	gnu_property	PT_GNU_PROPERTY	PF_R;
 }
Signature · 1 line
2.52.0
#44 lore raw

Re: [PATCH v4.1 07/10] x86/entry/vdso: include GNU_PROPERTY and GNU_STACK PHDRs

Quoted 12 lines · H. Peter Anvin

Currently the vdso doesn't include .note.gnu.property or a GNU noexec stack annotation (the -z noexecstack in the linker script is ineffective because we specify PHDRs explicitly.)

The motivation is that the dynamic linker currently do not check these.

However, this is a weak excuse: the vdso*.so are also supposed to be usable at link libraries, and there is no reason why the dynamic linker might not want or need to check these in the future, so add them back in -- it is trivial enough.

Quoted 6 lines

Use symbolic constants for the PHDR permission flags.

[ v4: drop unrelated formatting changes ]
[ v4.1: drop one last bogus formatting change (Brian Gerst) ]

The v4/v4.1 notes will presumably want dropping before committing?

author maintainer
#45 lore raw

PATCH v4 8/10 x86/vdso: abstract out vdso system call internals

In reply to H. Peter Anvin

Abstract out the calling of true system calls from the vdso into macros.

It has been a very long time since gcc did not allow %ebx or %ebp in inline asm in 32-bit PIC mode; remove the corresponding hacks.

Remove the use of memory output constraints in gettimeofday.h in favor of "memory" clobbers. The resulting code is identical for the current use cases, as the system call is usually a terminal fallback anyway, and it merely complicates the macroization.

This patch adds only a handful of more lines of code than it removes, and in fact could be made substantially smaller by removing the macros for the argument counts that aren't currently used, however, it seems better to be general from the start.

[ v3: remove stray comment from prototyping; remove VDSO_SYSCALL6()
since it would require special handling on 32 bits and is currently unused. (Uros Bizjak).
Indent nested preprocessor directives. ]

---
 arch/x86/include/asm/vdso/gettimeofday.h | 108 ++---------------------
 arch/x86/include/asm/vdso/sys_call.h     | 103 +++++++++++++++++++++
 2 files changed, 111 insertions(+), 100 deletions(-)
 create mode 100644 arch/x86/include/asm/vdso/sys_call.h
arch/x86/include/asm/vdso/gettimeofday.h +8 -100
index 73b2e7ee8f0f..3cf214cc4a75 100644
@@ -18,6 +18,7 @@
 #include <asm/msr.h>
 #include <asm/pvclock.h>
 #include <clocksource/hyperv_timer.h>
+#include <asm/vdso/sys_call.h>
 
 #define VDSO_HAS_TIME 1
 
@@ -53,130 +54,37 @@ extern struct ms_hyperv_tsc_page hvclock_page
 	__attribute__((visibility("hidden")));
 #endif
 
-#ifndef BUILD_VDSO32
-
 static __always_inline
 long clock_gettime_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
 {
-	long ret;
-
-	asm ("syscall" : "=a" (ret), "=m" (*_ts) :
-	     "0" (__NR_clock_gettime), "D" (_clkid), "S" (_ts) :
-	     "rcx", "r11");
-
-	return ret;
+	return VDSO_SYSCALL2(clock_gettime,64,_clkid,_ts);
 }
 
 static __always_inline
 long gettimeofday_fallback(struct __kernel_old_timeval *_tv,
 			   struct timezone *_tz)
 {
-	long ret;
-
-	asm("syscall" : "=a" (ret) :
-	    "0" (__NR_gettimeofday), "D" (_tv), "S" (_tz) : "memory");
-
-	return ret;
+	return VDSO_SYSCALL2(gettimeofday,,_tv,_tz);
 }
 
 static __always_inline
 long clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
 {
-	long ret;
-
-	asm ("syscall" : "=a" (ret), "=m" (*_ts) :
-	     "0" (__NR_clock_getres), "D" (_clkid), "S" (_ts) :
-	     "rcx", "r11");
-
-	return ret;
+	return VDSO_SYSCALL2(clock_getres,_time64,_clkid,_ts);
 }
 
-#else
-
-static __always_inline
-long clock_gettime_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
-{
-	long ret;
-
-	asm (
-		"mov %%ebx, %%edx \n"
-		"mov %[clock], %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret), "=m" (*_ts)
-		: "0" (__NR_clock_gettime64), [clock] "g" (_clkid), "c" (_ts)
-		: "edx");
-
-	return ret;
-}
+#ifndef CONFIG_X86_64
 
 static __always_inline
 long clock_gettime32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)
 {
-	long ret;
-
-	asm (
-		"mov %%ebx, %%edx \n"
-		"mov %[clock], %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret), "=m" (*_ts)
-		: "0" (__NR_clock_gettime), [clock] "g" (_clkid), "c" (_ts)
-		: "edx");
-
-	return ret;
-}
-
-static __always_inline
-long gettimeofday_fallback(struct __kernel_old_timeval *_tv,
-			   struct timezone *_tz)
-{
-	long ret;
-
-	asm(
-		"mov %%ebx, %%edx \n"
-		"mov %2, %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret)
-		: "0" (__NR_gettimeofday), "g" (_tv), "c" (_tz)
-		: "memory", "edx");
-
-	return ret;
+	return VDSO_SYSCALL2(clock_gettime,,_clkid,_ts);
 }
 
 static __always_inline long
-clock_getres_fallback(clockid_t _clkid, struct __kernel_timespec *_ts)
-{
-	long ret;
-
-	asm (
-		"mov %%ebx, %%edx \n"
-		"mov %[clock], %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret), "=m" (*_ts)
-		: "0" (__NR_clock_getres_time64), [clock] "g" (_clkid), "c" (_ts)
-		: "edx");
-
-	return ret;
-}
-
-static __always_inline
-long clock_getres32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)
+clock_getres32_fallback(clockid_t _clkid, struct old_timespec32 *_ts)
 {
-	long ret;
-
-	asm (
-		"mov %%ebx, %%edx \n"
-		"mov %[clock], %%ebx \n"
-		"call __kernel_vsyscall \n"
-		"mov %%edx, %%ebx \n"
-		: "=a" (ret), "=m" (*_ts)
-		: "0" (__NR_clock_getres), [clock] "g" (_clkid), "c" (_ts)
-		: "edx");
-
-	return ret;
+	return VDSO_SYSCALL2(clock_getres,,_clkid,_ts);
 }
 
 #endif
arch/x86/include/asm/vdso/sys_call.h new file +103 -0

103 lines not on this page · raw message

Signature · 1 line
2.52.0
author maintainer
#46 lore raw

PATCH v4 9/10 x86/cpufeature: replace X86_FEATURE_SYSENTER32 with X86_FEATURE_SYSFAST32

In reply to H. Peter Anvin

In most cases, the use of "fast 32-bit system call" depends either on X86_FEATURE_SEP or X86_FEATURE_SYSENTER32 || X86_FEATURE_SYSCALL32. However, nearly all the logic for both is identical.

Define X86_FEATURE_SYSFAST32 which indicates that *either* SYSENTER32 or SYSCALL32 should be used, for either 32- or 64-bit kernels. This defaults to SYSENTER; use SYSCALL if the SYSCALL32 bit is also set.

As this removes ALL existing uses of X86_FEATURE_SYSENTER32, which is a kernel-only synthetic feature bit, simply remove it and replace it with X86_FEATURE_SYSFAST32.

This leaves an unused alternative for a true 32-bit kernel, but that should really not matter in any way.

The clearing of X86_FEATURE_SYSCALL32 can be removed once the patches for automatically clearing disabled features has been merged.

---
 arch/x86/Kconfig.cpufeatures             |  8 +++++++
 arch/x86/entry/vdso/vdso32/system_call.S |  8 ++-----
 arch/x86/include/asm/cpufeatures.h       |  2 +-
 arch/x86/kernel/cpu/centaur.c            |  3 ---
 arch/x86/kernel/cpu/common.c             |  8 +++++++
 arch/x86/kernel/cpu/intel.c              |  4 +---
 arch/x86/kernel/cpu/zhaoxin.c            |  4 +---
 arch/x86/kernel/fred.c                   |  2 +-
 arch/x86/xen/setup.c                     | 28 +++++++++++++++---------
 arch/x86/xen/smp_pv.c                    |  5 ++---
 arch/x86/xen/xen-ops.h                   |  1 -
 11 files changed, 42 insertions(+), 31 deletions(-)
arch/x86/Kconfig.cpufeatures +8 -0
index 733d5aff2456..423ac795baa7 100644
@@ -56,6 +56,10 @@ config X86_REQUIRED_FEATURE_MOVBE
 	def_bool y
 	depends on MATOM
 
+config X86_REQUIRED_FEATURE_SYSFAST32
+	def_bool y
+	depends on X86_64 && !X86_FRED
+
 config X86_REQUIRED_FEATURE_CPUID
 	def_bool y
 	depends on X86_64
@@ -120,6 +124,10 @@ config X86_DISABLED_FEATURE_CENTAUR_MCR
 	def_bool y
 	depends on X86_64
 
+config X86_DISABLED_FEATURE_SYSCALL32
+	def_bool y
+	depends on !X86_64
+
 config X86_DISABLED_FEATURE_PCID
 	def_bool y
 	depends on !X86_64
arch/x86/entry/vdso/vdso32/system_call.S +2 -6
index 2a15634bbe75..7b1c0f16e511 100644
@@ -52,13 +52,9 @@ __kernel_vsyscall:
 	#define SYSENTER_SEQUENCE	"movl %esp, %ebp; sysenter"
 	#define SYSCALL_SEQUENCE	"movl %ecx, %ebp; syscall"
 
-#ifdef BUILD_VDSO32_64
 	/* If SYSENTER (Intel) or SYSCALL32 (AMD) is available, use it. */
-	ALTERNATIVE_2 "", SYSENTER_SEQUENCE, X86_FEATURE_SYSENTER32, \
-	                  SYSCALL_SEQUENCE,  X86_FEATURE_SYSCALL32
-#else
-	ALTERNATIVE "", SYSENTER_SEQUENCE, X86_FEATURE_SEP
-#endif
+	ALTERNATIVE_2 "", SYSENTER_SEQUENCE, X86_FEATURE_SYSFAST32, \
+			  SYSCALL_SEQUENCE,  X86_FEATURE_SYSCALL32
 
 	/* Enter using int $0x80 */
 	int	$0x80
arch/x86/include/asm/cpufeatures.h +1 -1
index c3b53beb1300..63b0f9aa9b3e 100644
@@ -84,7 +84,7 @@
 #define X86_FEATURE_PEBS		( 3*32+12) /* "pebs" Precise-Event Based Sampling */
 #define X86_FEATURE_BTS			( 3*32+13) /* "bts" Branch Trace Store */
 #define X86_FEATURE_SYSCALL32		( 3*32+14) /* syscall in IA32 userspace */
-#define X86_FEATURE_SYSENTER32		( 3*32+15) /* sysenter in IA32 userspace */
+#define X86_FEATURE_SYSFAST32		( 3*32+15) /* sysenter/syscall in IA32 userspace */
 #define X86_FEATURE_REP_GOOD		( 3*32+16) /* "rep_good" REP microcode works well */
 #define X86_FEATURE_AMD_LBR_V2		( 3*32+17) /* "amd_lbr_v2" AMD Last Branch Record Extension Version 2 */
 #define X86_FEATURE_CLEAR_CPU_BUF	( 3*32+18) /* Clear CPU buffers using VERW */
arch/x86/kernel/cpu/centaur.c +0 -3
index a3b55db35c96..9833f837141c 100644
@@ -102,9 +102,6 @@ static void early_init_centaur(struct cpuinfo_x86 *c)
 	    (c->x86 >= 7))
 		set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
 
-#ifdef CONFIG_X86_64
-	set_cpu_cap(c, X86_FEATURE_SYSENTER32);
-#endif
 	if (c->x86_power & (1 << 8)) {
 		set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
 		set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC);
arch/x86/kernel/cpu/common.c +8 -0
index e7ab22fce3b5..1c3261cae40c 100644
@@ -1068,6 +1068,9 @@ void get_cpu_cap(struct cpuinfo_x86 *c)
 	init_scattered_cpuid_features(c);
 	init_speculation_control(c);
 
+	if (IS_ENABLED(CONFIG_X86_64) || cpu_has(c, X86_FEATURE_SEP))
+		set_cpu_cap(c, X86_FEATURE_SYSFAST32);
+
 	/*
 	 * Clear/Set all flags overridden by options, after probe.
 	 * This needs to happen each time we re-probe, which may happen
@@ -1813,6 +1816,11 @@ static void __init early_identify_cpu(struct cpuinfo_x86 *c)
 	 * that it can't be enabled in 32-bit mode.
 	 */
 	setup_clear_cpu_cap(X86_FEATURE_PCID);
+
+	/*
+	 * Never use SYSCALL on a 32-bit kernel
+	 */
+	setup_clear_cpu_cap(X86_FEATURE_SYSCALL32);
 #endif
 
 	/*
arch/x86/kernel/cpu/intel.c +1 -3
index 98ae4c37c93e..646ff33c4651 100644
@@ -236,9 +236,7 @@ static void early_init_intel(struct cpuinfo_x86 *c)
 		clear_cpu_cap(c, X86_FEATURE_PSE);
 	}
 
-#ifdef CONFIG_X86_64
-	set_cpu_cap(c, X86_FEATURE_SYSENTER32);
-#else
+#ifndef CONFIG_X86_64
 	/* Netburst reports 64 bytes clflush size, but does IO in 128 bytes */
 	if (c->x86 == 15 && c->x86_cache_alignment == 64)
 		c->x86_cache_alignment = 128;
arch/x86/kernel/cpu/zhaoxin.c +1 -3
index 89b1c8a70fe8..031379b7d4fa 100644
@@ -59,9 +59,7 @@ static void early_init_zhaoxin(struct cpuinfo_x86 *c)
 {
 	if (c->x86 >= 0x6)
 		set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
-#ifdef CONFIG_X86_64
-	set_cpu_cap(c, X86_FEATURE_SYSENTER32);
-#endif
+
 	if (c->x86_power & (1 << 8)) {
 		set_cpu_cap(c, X86_FEATURE_CONSTANT_TSC);
 		set_cpu_cap(c, X86_FEATURE_NONSTOP_TSC);
arch/x86/kernel/fred.c +1 -1
index 816187da3a47..e736b19e18de 100644
@@ -68,7 +68,7 @@ void cpu_init_fred_exceptions(void)
 	idt_invalidate();
 
 	/* Use int $0x80 for 32-bit system calls in FRED mode */
-	setup_clear_cpu_cap(X86_FEATURE_SYSENTER32);
+	setup_clear_cpu_cap(X86_FEATURE_SYSFAST32);
 	setup_clear_cpu_cap(X86_FEATURE_SYSCALL32);
 }
 
arch/x86/xen/setup.c +18 -10
index 3823e52aef52..ac8021c3a997 100644
@@ -990,13 +990,6 @@ static int register_callback(unsigned type, const void *func)
 	return HYPERVISOR_callback_op(CALLBACKOP_register, &callback);
 }
 
-void xen_enable_sysenter(void)
-{
-	if (cpu_feature_enabled(X86_FEATURE_SYSENTER32) &&
-	    register_callback(CALLBACKTYPE_sysenter, xen_entry_SYSENTER_compat))
-		setup_clear_cpu_cap(X86_FEATURE_SYSENTER32);
-}
-
 void xen_enable_syscall(void)
 {
 	int ret;
@@ -1008,11 +1001,27 @@ void xen_enable_syscall(void)
 		   mechanism for syscalls. */
 	}
 
-	if (cpu_feature_enabled(X86_FEATURE_SYSCALL32) &&
-	    register_callback(CALLBACKTYPE_syscall32, xen_entry_SYSCALL_compat))
+	if (!cpu_feature_enabled(X86_FEATURE_SYSFAST32))
+		return;
+
+	if (cpu_feature_enabled(X86_FEATURE_SYSCALL32)) {
+		/* Use SYSCALL32 */
+		ret = register_callback(CALLBACKTYPE_syscall32,
+					xen_entry_SYSCALL_compat);
+
+	} else {
+		/* Use SYSENTER32 */
+		ret = register_callback(CALLBACKTYPE_sysenter,
+					xen_entry_SYSENTER_compat);
+	}
+
+	if (ret) {
 		setup_clear_cpu_cap(X86_FEATURE_SYSCALL32);
+		setup_clear_cpu_cap(X86_FEATURE_SYSFAST32);
+	}
 }
 
+
 static void __init xen_pvmmu_arch_setup(void)
 {
 	HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_writable_pagetables);
@@ -1022,7 +1031,6 @@ static void __init xen_pvmmu_arch_setup(void)
 	    register_callback(CALLBACKTYPE_failsafe, xen_failsafe_callback))
 		BUG();
 
-	xen_enable_sysenter();
 	xen_enable_syscall();
 }
 
arch/x86/xen/smp_pv.c +2 -3
index 9bb8ff8bff30..c40f326f0c3a 100644
@@ -65,10 +65,9 @@ static void cpu_bringup(void)
 	touch_softlockup_watchdog();
 
 	/* PVH runs in ring 0 and allows us to do native syscalls. Yay! */
-	if (!xen_feature(XENFEAT_supervisor_mode_kernel)) {
-		xen_enable_sysenter();
+	if (!xen_feature(XENFEAT_supervisor_mode_kernel))
 		xen_enable_syscall();
-	}
+
 	cpu = smp_processor_id();
 	identify_secondary_cpu(cpu);
 	set_cpu_sibling_map(cpu);
arch/x86/xen/xen-ops.h +0 -1
index 090349baec09..f6c331b20fad 100644
@@ -60,7 +60,6 @@ phys_addr_t __init xen_find_free_area(phys_addr_t size);
 char * __init xen_memory_setup(void);
 void __init xen_arch_setup(void);
 void xen_banner(void);
-void xen_enable_sysenter(void);
 void xen_enable_syscall(void);
 void xen_vcpu_restore(void);
 
Signature · 1 line
2.52.0
author maintainer
#47 lore raw

PATCH v4 10/10 x86/entry/vdso32: when using int $0x80, use it directly

In reply to H. Peter Anvin

When neither sysenter32 nor syscall32 is available (on either FRED-capable 64-bit hardware or old 32-bit hardware), there is no reason to do a bunch of stack shuffling in __kernel_vsyscall. Unfortunately, just overwriting the initial "push" instructions will mess up the CFI annotations, so suffer the 3-byte NOP if not applicable.

Similarly, inline the int $0x80 when doing inline system calls in the vdso instead of calling __kernel_vsyscall.

---
 arch/x86/entry/vdso/vdso32/system_call.S | 18 ++++++++++++++----
 arch/x86/include/asm/vdso/sys_call.h     |  4 +++-
 2 files changed, 17 insertions(+), 5 deletions(-)
arch/x86/entry/vdso/vdso32/system_call.S +14 -4
index 7b1c0f16e511..9157cf9c5749 100644
@@ -14,6 +14,18 @@
 	ALIGN
 __kernel_vsyscall:
 	CFI_STARTPROC
+
+	/*
+	 * If using int $0x80, there is no reason to muck about with the
+	 * stack here. Unfortunately just overwriting the push instructions
+	 * would mess up the CFI annotations, but it is only a 3-byte
+	 * NOP in that case. This could be avoided by patching the
+	 * vdso symbol table (not the code) and entry point, but that
+	 * would a fair bit of tooling work or by simply compiling
+	 * two different vDSO images, but that doesn't seem worth it.
+	 */
+	ALTERNATIVE "int $0x80; ret", "", X86_FEATURE_SYSFAST32
+
 	/*
 	 * Reshuffle regs so that all of any of the entry instructions
 	 * will preserve enough state.
@@ -52,11 +64,9 @@ __kernel_vsyscall:
 	#define SYSENTER_SEQUENCE	"movl %esp, %ebp; sysenter"
 	#define SYSCALL_SEQUENCE	"movl %ecx, %ebp; syscall"
 
-	/* If SYSENTER (Intel) or SYSCALL32 (AMD) is available, use it. */
-	ALTERNATIVE_2 "", SYSENTER_SEQUENCE, X86_FEATURE_SYSFAST32, \
-			  SYSCALL_SEQUENCE,  X86_FEATURE_SYSCALL32
+	ALTERNATIVE SYSENTER_SEQUENCE, SYSCALL_SEQUENCE, X86_FEATURE_SYSCALL32
 
-	/* Enter using int $0x80 */
+	/* Re-enter using int $0x80 */
 	int	$0x80
 SYM_INNER_LABEL(int80_landing_pad, SYM_L_GLOBAL)
 
arch/x86/include/asm/vdso/sys_call.h +3 -1
index dcfd17c6dd57..5806b1cd6aef 100644
@@ -20,7 +20,9 @@
 # define __sys_reg4	"r10"
 # define __sys_reg5	"r8"
 #else
-# define __sys_instr	"call __kernel_vsyscall"
+# define __sys_instr	ALTERNATIVE("ds;ds;ds;int $0x80",		\
+				    "call __kernel_vsyscall",		\
+				    X86_FEATURE_SYSFAST32)
 # define __sys_clobber	"memory"
 # define __sys_nr(x,y)	__NR_ ## x ## y
 # define __sys_reg1	"ebx"
Signature · 1 line
2.52.0
#48 lore raw
In reply to H. Peter Anvin

[stripped down the list of recipients quite a bit]

Quoted 6 lines · H. Peter Anvin

[...]
This patchset cleans up the vdso build by building the 32- and 64-bit vdsos in separate directories, moving the vdso2c tool to arch/x86/tools, and by merging common code and especially Makefile rules between the 32- and 64-bit vdsos to the greatest extent possible. [...]

Lo! My daily -next builds for Fedora failed on x86_64 (other archs worked fine). Haven't checked, but from the error message I wonder if this might be due to the changes from this patch-set that showed up in -next today:

+ /usr/bin/make -s 'HOSTCFLAGS=-O2 -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 ' 'HOSTLDFLAGS=-Wl,-z,relro -Wl,--as-needed -Wl,-z,pack-relative-relocs -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld -specs=/usr/lib/rpm/redhat/redhat-hardened-ld-errors -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -Wl,--build-id=sha1 -specs=/usr/lib/rpm/redhat/redhat-package-notes ' ARCH=x86_64 INSTALL_MOD_PATH=/builddir/build/BUILD/kernel-6.19.0-build/BUILDROOT vdso_install KERNELRELEASE=6.19.0-0.0.next.20260115.439.vanilla.fc44.x86_64+rt
readelf: Error: 'arch/x86/entry/vdso/vdso32.so.dbg': No such file
readelf: Error: 'arch/x86/entry/vdso/vdso64.so.dbg': No such file
scripts/Makefile.vdsoinst:33: warning: overriding recipe for target '/builddir/build/BUILD/kernel-6.19.0-build/BUILDROOT/lib/modules/6.19.0-0.0.next.20260115.439.vanilla.fc44.x86_64+rt/vdso/.build-id/.debug' scripts/Makefile.vdsoinst:33: warning: ignoring old recipe for target '/builddir/build/BUILD/kernel-6.19.0-build/BUILDROOT/lib/modules/6.19.0-0.0.next.20260115.439.vanilla.fc44.x86_64+rt/vdso/.build-id/.debug' make[2]: *** No rule to make target 'arch/x86/entry/vdso/vdso32.so.dbg', needed by '/builddir/build/BUILD/kernel-6.19.0-build/BUILDROOT/lib/modules/6.19.0-0.0.next.20260115.439.vanilla.fc44.x86_64+rt/vdso/vdso32.so'. Stop. make[1]: *** [/builddir/build/BUILD/kernel-6.19.0-build/kernel-next-20260115/linux-6.19.0-0.0.next.20260115.439.vanilla.fc44.x86_64/Makefile:1459: vdso_install] Error 2
make: *** [Makefile:256: __sub-make] Error 2

Full log:
https://download.copr.fedorainfracloud.org/results/@kernel-vanilla/next/fedora-rawhide-x86_64/10010857-next-next-all/builder-live.log.gz

Ciao, Thorsten

author maintainer
#49 lore raw
Quoted 24 lines · Thorsten Leemhuis

[stripped down the list of recipients quite a bit]

Quoted 6 lines · H. Peter Anvin

[...]
This patchset cleans up the vdso build by building the 32- and 64-bit vdsos in separate directories, moving the vdso2c tool to arch/x86/tools, and by merging common code and especially Makefile rules between the 32- and 64-bit vdsos to the greatest extent possible. [...]

Lo! My daily -next builds for Fedora failed on x86_64 (other archs worked fine). Haven't checked, but from the error message I wonder if this might be due to the changes from this patch-set that showed up in -next today:

+ /usr/bin/make -s 'HOSTCFLAGS=-O2 -fexceptions -g -grecord-gcc-switches -pipe -Wall -Werror=format-security -Wp,-U_FORTIFY_SOURCE,-D_FORTIFY_SOURCE=3 -Wp,-D_GLIBCXX_ASSERTIONS -specs=/usr/lib/rpm/redhat/redhat-hardened-cc1 -fstack-protector-strong -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -m64 -march=x86-64 -mtune=generic -fasynchronous-unwind-tables -fstack-clash-protection -fcf-protection -mtls-dialect=gnu2 ' 'HOSTLDFLAGS=-Wl,-z,relro -Wl,--as-needed -Wl,-z,pack-relative-relocs -Wl,-z,now -specs=/usr/lib/rpm/redhat/redhat-hardened-ld -specs=/usr/lib/rpm/redhat/redhat-hardened-ld-errors -specs=/usr/lib/rpm/redhat/redhat-annobin-cc1 -Wl,--build-id=sha1 -specs=/usr/lib/rpm/redhat/redhat-package-notes ' ARCH=x86_64 INSTALL_MOD_PATH=/builddir/build/BUILD/kernel-6.19.0-build/BUILDROOT vdso_install KERNELRELEASE=6.19.0-0.0.next.20260115.439.vanilla.fc44.x86_64+rt
readelf: Error: 'arch/x86/entry/vdso/vdso32.so.dbg': No such file
readelf: Error: 'arch/x86/entry/vdso/vdso64.so.dbg': No such file
scripts/Makefile.vdsoinst:33: warning: overriding recipe for target '/builddir/build/BUILD/kernel-6.19.0-build/BUILDROOT/lib/modules/6.19.0-0.0.next.20260115.439.vanilla.fc44.x86_64+rt/vdso/.build-id/.debug' scripts/Makefile.vdsoinst:33: warning: ignoring old recipe for target '/builddir/build/BUILD/kernel-6.19.0-build/BUILDROOT/lib/modules/6.19.0-0.0.next.20260115.439.vanilla.fc44.x86_64+rt/vdso/.build-id/.debug' make[2]: *** No rule to make target 'arch/x86/entry/vdso/vdso32.so.dbg', needed by '/builddir/build/BUILD/kernel-6.19.0-build/BUILDROOT/lib/modules/6.19.0-0.0.next.20260115.439.vanilla.fc44.x86_64+rt/vdso/vdso32.so'. Stop. make[1]: *** [/builddir/build/BUILD/kernel-6.19.0-build/kernel-next-20260115/linux-6.19.0-0.0.next.20260115.439.vanilla.fc44.x86_64/Makefile:1459: vdso_install] Error 2
make: *** [Makefile:256: __sub-make] Error 2

Quoted 4 lines

Full log:
https://download.copr.fedorainfracloud.org/results/@kernel-vanilla/next/fedora-rawhide-x86_64/10010857-next-next-all/builder-live.log.gz

Ciao, Thorsten

Looks like it. Specifically it looks like it needs a tweak to make vdso_install. I'll look at it in a few hours.