x86: IA32 emulation cleanups

PATCH In review Brian Gerst 5 participants 16 messages Mar 4, 2015 – Mar 5, 2015 x86

Share & tools

Brian Gerst

author

2015-03-04 03:31 UTC #0 lore raw

cover x86: IA32 emulation cleanups

This series cleans up some files in the 32-bit emulation code, by merging with the native 32-bit code, or in the case of the syscall audit, using the generic compat code that was recently added.

Brian Gerst (5):
x86: Remove compat_ni_syscall()
x86: Merge native and compat 32-bit syscall tables
x86: Remove sys32_vm86_warning
x86: Use generic compat audit code
x86: Clean up ia32/Makefile

 arch/x86/Kconfig                 |  5 +--
 arch/x86/ia32/Makefile           |  6 +--
 arch/x86/ia32/audit.c            | 43 ---------------------
 arch/x86/ia32/nosyscall.c        |  7 ----
 arch/x86/ia32/sys_ia32.c         | 14 -------
 arch/x86/ia32/syscall_ia32.c     | 25 ------------
 arch/x86/include/asm/unistd32.h  |  2 +
 arch/x86/kernel/Makefile         |  3 +-
 arch/x86/kernel/audit_64.c       | 82 ----------------------------------------
 arch/x86/kernel/syscall_32.c     | 16 ++++++--
 arch/x86/syscalls/syscall_32.tbl |  4 +-
 11 files changed, 19 insertions(+), 188 deletions(-)
 delete mode 100644 arch/x86/ia32/audit.c
 delete mode 100644 arch/x86/ia32/nosyscall.c
 delete mode 100644 arch/x86/ia32/syscall_ia32.c
 create mode 100644 arch/x86/include/asm/unistd32.h
 delete mode 100644 arch/x86/kernel/audit_64.c

Signature · 1 line

2.1.0

Brian Gerst

author

2015-03-04 03:31 UTC #1 lore raw

PATCH 1/5 x86: Remove compat_ni_syscall()

compat_ni_syscall() does the same thing as sys_ni_syscall().

Signed-off-by Brian Gerst <[email protected]>

---
 arch/x86/ia32/Makefile       | 2 +-
 arch/x86/ia32/nosyscall.c    | 7 -------
 arch/x86/ia32/syscall_ia32.c | 4 ++--
 3 files changed, 3 insertions(+), 10 deletions(-)
 delete mode 100644 arch/x86/ia32/nosyscall.c

arch/x86/ia32/Makefile +1 -1

index e785b42..e66d850 100644
@@ -3,7 +3,7 @@
 #
 
 obj-$(CONFIG_IA32_EMULATION) := ia32entry.o sys_ia32.o ia32_signal.o
-obj-$(CONFIG_IA32_EMULATION) += nosyscall.o syscall_ia32.o
+obj-$(CONFIG_IA32_EMULATION) += syscall_ia32.o
 
 obj-$(CONFIG_IA32_AOUT) += ia32_aout.o

arch/x86/ia32/nosyscall.c deleted +0 -7

deleted file mode 100644
index 51ecd5b..0000000
@@ -1,7 +0,0 @@
-#include <linux/kernel.h>
-#include <linux/errno.h>
-
-long compat_ni_syscall(void)
-{
-	return -ENOSYS;
-}

arch/x86/ia32/syscall_ia32.c +2 -2

index 4754ba0..3429b14 100644
@@ -13,13 +13,13 @@
 
 typedef void (*sys_call_ptr_t)(void);
 
-extern void compat_ni_syscall(void);
+extern asmlinkage void sys_ni_syscall(void);
 
 const sys_call_ptr_t ia32_sys_call_table[__NR_ia32_syscall_max+1] = {
 	/*
 	 * Smells like a compiler bug -- it doesn't work
 	 * when the & below is removed.
 	 */
-	[0 ... __NR_ia32_syscall_max] = &compat_ni_syscall,
+	[0 ... __NR_ia32_syscall_max] = &sys_ni_syscall,
 #include <asm/syscalls_32.h>
 };

Signature · 1 line

2.1.0

tip-bot for Brian Gerst

merge bot

2015-03-04 05:28 UTC #2 lore raw

[tip:x86/asm] x86/compat: Remove compat_ni_syscall()

x86/compat: Remove compat_ni_syscall()

Commit-ID:  29a5ff97fa0d8045d262a772c3853e3ef1ed98d8
Gitweb:     http://git.kernel.org/tip/29a5ff97fa0d8045d262a772c3853e3ef1ed98d8
Author:     Brian Gerst <[email protected]>
AuthorDate: Tue, 3 Mar 2015 22:31:32 -0500
Committer:  Ingo Molnar <[email protected]>
CommitDate: Wed, 4 Mar 2015 06:16:21 +0100

x86/compat: Remove compat_ni_syscall()

compat_ni_syscall() does the same thing as sys_ni_syscall().

Signed-off-by Brian Gerst <[email protected]>

Cc Borislav Petkov <[email protected]>

Cc H. Peter Anvin <[email protected]>

Cc Linus Torvalds <[email protected]>

Cc Thomas Gleixner <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 arch/x86/ia32/Makefile       | 2 +-
 arch/x86/ia32/nosyscall.c    | 7 -------
 arch/x86/ia32/syscall_ia32.c | 4 ++--
 3 files changed, 3 insertions(+), 10 deletions(-)

arch/x86/ia32/Makefile +1 -1

index e785b42..e66d850 100644
@@ -3,7 +3,7 @@
 #
 
 obj-$(CONFIG_IA32_EMULATION) := ia32entry.o sys_ia32.o ia32_signal.o
-obj-$(CONFIG_IA32_EMULATION) += nosyscall.o syscall_ia32.o
+obj-$(CONFIG_IA32_EMULATION) += syscall_ia32.o
 
 obj-$(CONFIG_IA32_AOUT) += ia32_aout.o

arch/x86/ia32/nosyscall.c deleted +0 -7

deleted file mode 100644
index 51ecd5b..0000000
@@ -1,7 +0,0 @@
-#include <linux/kernel.h>
-#include <linux/errno.h>
-
-long compat_ni_syscall(void)
-{
-	return -ENOSYS;
-}

arch/x86/ia32/syscall_ia32.c +2 -2

index 4754ba0..3429b14 100644
@@ -13,13 +13,13 @@
 
 typedef void (*sys_call_ptr_t)(void);
 
-extern void compat_ni_syscall(void);
+extern asmlinkage void sys_ni_syscall(void);
 
 const sys_call_ptr_t ia32_sys_call_table[__NR_ia32_syscall_max+1] = {
 	/*
 	 * Smells like a compiler bug -- it doesn't work
 	 * when the & below is removed.
 	 */
-	[0 ... __NR_ia32_syscall_max] = &compat_ni_syscall,
+	[0 ... __NR_ia32_syscall_max] = &sys_ni_syscall,
 #include <asm/syscalls_32.h>
 };

Brian Gerst

author

2015-03-04 03:31 UTC #3 lore raw

PATCH 2/5 x86: Merge native and compat 32-bit syscall tables

In reply to Brian Gerst

Combine the 32-bit syscall tables into one file.

Signed-off-by Brian Gerst <[email protected]>

---
 arch/x86/ia32/Makefile       |  1 -
 arch/x86/ia32/syscall_ia32.c | 25 -------------------------
 arch/x86/kernel/Makefile     |  1 +
 arch/x86/kernel/syscall_32.c | 16 ++++++++++++----
 4 files changed, 13 insertions(+), 30 deletions(-)
 delete mode 100644 arch/x86/ia32/syscall_ia32.c

arch/x86/ia32/Makefile +0 -1

index e66d850..bb635c6 100644
@@ -3,7 +3,6 @@
 #
 
 obj-$(CONFIG_IA32_EMULATION) := ia32entry.o sys_ia32.o ia32_signal.o
-obj-$(CONFIG_IA32_EMULATION) += syscall_ia32.o
 
 obj-$(CONFIG_IA32_AOUT) += ia32_aout.o

arch/x86/ia32/syscall_ia32.c deleted +0 ~~-25~~

deleted file mode 100644
index 3429b14..0000000
@@ -1,25 +0,0 @@
-/* System call table for ia32 emulation. */
-
-#include <linux/linkage.h>
-#include <linux/sys.h>
-#include <linux/cache.h>
-#include <asm/asm-offsets.h>
-
-#define __SYSCALL_I386(nr, sym, compat) extern asmlinkage void compat(void) ;
-#include <asm/syscalls_32.h>
-#undef __SYSCALL_I386
-
-#define __SYSCALL_I386(nr, sym, compat) [nr] = compat,
-
-typedef void (*sys_call_ptr_t)(void);
-
-extern asmlinkage void sys_ni_syscall(void);
-
-const sys_call_ptr_t ia32_sys_call_table[__NR_ia32_syscall_max+1] = {
-	/*
-	 * Smells like a compiler bug -- it doesn't work
-	 * when the & below is removed.
-	 */
-	[0 ... __NR_ia32_syscall_max] = &sys_ni_syscall,
-#include <asm/syscalls_32.h>
-};

arch/x86/kernel/Makefile +1 -0

index cdb1b70..c887cd9 100644
@@ -32,6 +32,7 @@ obj-$(CONFIG_X86_32)	+= i386_ksyms_32.o
 obj-$(CONFIG_X86_64)	+= sys_x86_64.o x8664_ksyms_64.o
 obj-$(CONFIG_X86_64)	+= mcount_64.o
 obj-y			+= syscall_$(BITS).o vsyscall_gtod.o
+obj-$(CONFIG_IA32_EMULATION)	+= syscall_32.o
 obj-$(CONFIG_X86_VSYSCALL_EMULATION)	+= vsyscall_64.o vsyscall_emu_64.o
 obj-$(CONFIG_X86_ESPFIX64)	+= espfix_64.o
 obj-$(CONFIG_SYSFS)	+= ksysfs.o

arch/x86/kernel/syscall_32.c +12 -4

index e9bcd57..3777189 100644
@@ -5,21 +5,29 @@
 #include <linux/cache.h>
 #include <asm/asm-offsets.h>
 
-#define __SYSCALL_I386(nr, sym, compat) extern asmlinkage void sym(void) ;
+#ifdef CONFIG_IA32_EMULATION
+#define SYM(sym, compat) compat
+#else
+#define SYM(sym, compat) sym
+#define ia32_sys_call_table sys_call_table
+#define __NR_ia32_syscall_max __NR_syscall_max
+#endif
+
+#define __SYSCALL_I386(nr, sym, compat) extern asmlinkage void SYM(sym, compat)(void) ;
 #include <asm/syscalls_32.h>
 #undef __SYSCALL_I386
 
-#define __SYSCALL_I386(nr, sym, compat) [nr] = sym,
+#define __SYSCALL_I386(nr, sym, compat) [nr] = SYM(sym, compat),
 
 typedef asmlinkage void (*sys_call_ptr_t)(void);
 
 extern asmlinkage void sys_ni_syscall(void);
 
-__visible const sys_call_ptr_t sys_call_table[__NR_syscall_max+1] = {
+__visible const sys_call_ptr_t ia32_sys_call_table[__NR_ia32_syscall_max+1] = {
 	/*
 	 * Smells like a compiler bug -- it doesn't work
 	 * when the & below is removed.
 	 */
-	[0 ... __NR_syscall_max] = &sys_ni_syscall,
+	[0 ... __NR_ia32_syscall_max] = &sys_ni_syscall,
 #include <asm/syscalls_32.h>
 };

Signature · 1 line

2.1.0

tip-bot for Brian Gerst

merge bot

2015-03-04 05:28 UTC #4 lore raw

[tip:x86/asm] x86/compat: Merge native and compat 32-bit syscall tables

x86/compat: Merge native and compat 32-bit syscall tables

Commit-ID:  2aa4a710928863e84cb71e60b7c839d12403f5ca
Gitweb:     http://git.kernel.org/tip/2aa4a710928863e84cb71e60b7c839d12403f5ca
Author:     Brian Gerst <[email protected]>
AuthorDate: Tue, 3 Mar 2015 22:31:33 -0500
Committer:  Ingo Molnar <[email protected]>
CommitDate: Wed, 4 Mar 2015 06:16:21 +0100

x86/compat: Merge native and compat 32-bit syscall tables

Combine the 32-bit syscall tables into one file.

Signed-off-by Brian Gerst <[email protected]>

Cc Borislav Petkov <[email protected]>

Cc H. Peter Anvin <[email protected]>

Cc Linus Torvalds <[email protected]>

Cc Thomas Gleixner <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 arch/x86/ia32/Makefile       |  1 -
 arch/x86/ia32/syscall_ia32.c | 25 -------------------------
 arch/x86/kernel/Makefile     |  1 +
 arch/x86/kernel/syscall_32.c | 16 ++++++++++++----
 4 files changed, 13 insertions(+), 30 deletions(-)

arch/x86/ia32/Makefile +0 -1

index e66d850..bb635c6 100644
@@ -3,7 +3,6 @@
 #
 
 obj-$(CONFIG_IA32_EMULATION) := ia32entry.o sys_ia32.o ia32_signal.o
-obj-$(CONFIG_IA32_EMULATION) += syscall_ia32.o
 
 obj-$(CONFIG_IA32_AOUT) += ia32_aout.o

arch/x86/ia32/syscall_ia32.c deleted +0 ~~-25~~

deleted file mode 100644
index 3429b14..0000000
@@ -1,25 +0,0 @@
-/* System call table for ia32 emulation. */
-
-#include <linux/linkage.h>
-#include <linux/sys.h>
-#include <linux/cache.h>
-#include <asm/asm-offsets.h>
-
-#define __SYSCALL_I386(nr, sym, compat) extern asmlinkage void compat(void) ;
-#include <asm/syscalls_32.h>
-#undef __SYSCALL_I386
-
-#define __SYSCALL_I386(nr, sym, compat) [nr] = compat,
-
-typedef void (*sys_call_ptr_t)(void);
-
-extern asmlinkage void sys_ni_syscall(void);
-
-const sys_call_ptr_t ia32_sys_call_table[__NR_ia32_syscall_max+1] = {
-	/*
-	 * Smells like a compiler bug -- it doesn't work
-	 * when the & below is removed.
-	 */
-	[0 ... __NR_ia32_syscall_max] = &sys_ni_syscall,
-#include <asm/syscalls_32.h>
-};

arch/x86/kernel/Makefile +1 -0

index 5d4502c..62fbe71 100644
@@ -28,6 +28,7 @@ obj-$(CONFIG_X86_32)	+= i386_ksyms_32.o
 obj-$(CONFIG_X86_64)	+= sys_x86_64.o x8664_ksyms_64.o
 obj-$(CONFIG_X86_64)	+= mcount_64.o
 obj-y			+= syscall_$(BITS).o vsyscall_gtod.o
+obj-$(CONFIG_IA32_EMULATION)	+= syscall_32.o
 obj-$(CONFIG_X86_VSYSCALL_EMULATION)	+= vsyscall_64.o vsyscall_emu_64.o
 obj-$(CONFIG_X86_ESPFIX64)	+= espfix_64.o
 obj-$(CONFIG_SYSFS)	+= ksysfs.o

arch/x86/kernel/syscall_32.c +12 -4

index e9bcd57..3777189 100644
@@ -5,21 +5,29 @@
 #include <linux/cache.h>
 #include <asm/asm-offsets.h>
 
-#define __SYSCALL_I386(nr, sym, compat) extern asmlinkage void sym(void) ;
+#ifdef CONFIG_IA32_EMULATION
+#define SYM(sym, compat) compat
+#else
+#define SYM(sym, compat) sym
+#define ia32_sys_call_table sys_call_table
+#define __NR_ia32_syscall_max __NR_syscall_max
+#endif
+
+#define __SYSCALL_I386(nr, sym, compat) extern asmlinkage void SYM(sym, compat)(void) ;
 #include <asm/syscalls_32.h>
 #undef __SYSCALL_I386
 
-#define __SYSCALL_I386(nr, sym, compat) [nr] = sym,
+#define __SYSCALL_I386(nr, sym, compat) [nr] = SYM(sym, compat),
 
 typedef asmlinkage void (*sys_call_ptr_t)(void);
 
 extern asmlinkage void sys_ni_syscall(void);
 
-__visible const sys_call_ptr_t sys_call_table[__NR_syscall_max+1] = {
+__visible const sys_call_ptr_t ia32_sys_call_table[__NR_ia32_syscall_max+1] = {
 	/*
 	 * Smells like a compiler bug -- it doesn't work
 	 * when the & below is removed.
 	 */
-	[0 ... __NR_syscall_max] = &sys_ni_syscall,
+	[0 ... __NR_ia32_syscall_max] = &sys_ni_syscall,
 #include <asm/syscalls_32.h>
 };

Brian Gerst

author

2015-03-04 03:31 UTC #5 lore raw

PATCH 3/5 x86: Remove sys32_vm86_warning

In reply to Brian Gerst

The check against lastcomm is racy, and the message it produces isn't necessary. vm86 support can be disabled on a 32-bit kernel also, and doesn't have this message. Switch to sys_ni_syscall instead.

Signed-off-by Brian Gerst <[email protected]>

---
 arch/x86/ia32/sys_ia32.c         | 14 --------------
 arch/x86/syscalls/syscall_32.tbl |  4 ++--
 2 files changed, 2 insertions(+), 16 deletions(-)

arch/x86/ia32/sys_ia32.c +0 ~~-14~~

index 8e0ceec..719cd70 100644
@@ -201,20 +201,6 @@ long sys32_fadvise64_64(int fd, __u32 offset_low, __u32 offset_high,
 				advice);
 }
 
-long sys32_vm86_warning(void)
-{
-	struct task_struct *me = current;
-	static char lastcomm[sizeof(me->comm)];
-
-	if (strncmp(lastcomm, me->comm, sizeof(lastcomm))) {
-		compat_printk(KERN_INFO
-			      "%s: vm86 mode not supported on 64 bit kernel\n",
-			      me->comm);
-		strncpy(lastcomm, me->comm, sizeof(lastcomm));
-	}
-	return -ENOSYS;
-}
-
 asmlinkage ssize_t sys32_readahead(int fd, unsigned off_lo, unsigned off_hi,
 				   size_t count)
 {

arch/x86/syscalls/syscall_32.tbl +2 -2

index b3560ec..ef8187f 100644
@@ -119,7 +119,7 @@
 110	i386	iopl			sys_iopl
 111	i386	vhangup			sys_vhangup
 112	i386	idle
-113	i386	vm86old			sys_vm86old			sys32_vm86_warning
+113	i386	vm86old			sys_vm86old			sys_ni_syscall
 114	i386	wait4			sys_wait4			compat_sys_wait4
 115	i386	swapoff			sys_swapoff
 116	i386	sysinfo			sys_sysinfo			compat_sys_sysinfo

@@ -172,7 +172,7 @@
 163	i386	mremap			sys_mremap
 164	i386	setresuid		sys_setresuid16
 165	i386	getresuid		sys_getresuid16
-166	i386	vm86			sys_vm86			sys32_vm86_warning
+166	i386	vm86			sys_vm86			sys_ni_syscall
 167	i386	query_module
 168	i386	poll			sys_poll
 169	i386	nfsservctl

Signature · 1 line

2.1.0

tip-bot for Brian Gerst

merge bot

2015-03-04 05:28 UTC #6 lore raw

[tip:x86/asm] x86/compat: Remove sys32_vm86_warning

x86/compat: Remove sys32_vm86_warning

Commit-ID:  7e8e385aaf6ed5b64b5d9108081cfcdcdd021b78
Gitweb:     http://git.kernel.org/tip/7e8e385aaf6ed5b64b5d9108081cfcdcdd021b78
Author:     Brian Gerst <[email protected]>
AuthorDate: Tue, 3 Mar 2015 22:31:34 -0500
Committer:  Ingo Molnar <[email protected]>
CommitDate: Wed, 4 Mar 2015 06:16:21 +0100

x86/compat: Remove sys32_vm86_warning

Signed-off-by Brian Gerst <[email protected]>

Cc Borislav Petkov <[email protected]>

Cc H. Peter Anvin <[email protected]>

Cc Linus Torvalds <[email protected]>

Cc Thomas Gleixner <[email protected]>

Link http://lkml.kernel.org/r/[email protected] open here

Signed-off-by Ingo Molnar <[email protected]>

---
 arch/x86/ia32/sys_ia32.c         | 14 --------------
 arch/x86/syscalls/syscall_32.tbl |  4 ++--
 2 files changed, 2 insertions(+), 16 deletions(-)

arch/x86/ia32/sys_ia32.c +0 ~~-14~~

index 8e0ceec..719cd70 100644
@@ -201,20 +201,6 @@ long sys32_fadvise64_64(int fd, __u32 offset_low, __u32 offset_high,
 				advice);
 }
 
-long sys32_vm86_warning(void)
-{
-	struct task_struct *me = current;
-	static char lastcomm[sizeof(me->comm)];
-
-	if (strncmp(lastcomm, me->comm, sizeof(lastcomm))) {
-		compat_printk(KERN_INFO
-			      "%s: vm86 mode not supported on 64 bit kernel\n",
-			      me->comm);
-		strncpy(lastcomm, me->comm, sizeof(lastcomm));
-	}
-	return -ENOSYS;
-}
-
 asmlinkage ssize_t sys32_readahead(int fd, unsigned off_lo, unsigned off_hi,
 				   size_t count)
 {

arch/x86/syscalls/syscall_32.tbl +2 -2

index b3560ec..ef8187f 100644
@@ -119,7 +119,7 @@
 110	i386	iopl			sys_iopl
 111	i386	vhangup			sys_vhangup
 112	i386	idle
-113	i386	vm86old			sys_vm86old			sys32_vm86_warning
+113	i386	vm86old			sys_vm86old			sys_ni_syscall
 114	i386	wait4			sys_wait4			compat_sys_wait4
 115	i386	swapoff			sys_swapoff
 116	i386	sysinfo			sys_sysinfo			compat_sys_sysinfo

@@ -172,7 +172,7 @@
 163	i386	mremap			sys_mremap
 164	i386	setresuid		sys_setresuid16
 165	i386	getresuid		sys_getresuid16
-166	i386	vm86			sys_vm86			sys32_vm86_warning
+166	i386	vm86			sys_vm86			sys_ni_syscall
 167	i386	query_module
 168	i386	poll			sys_poll
 169	i386	nfsservctl

Brian Gerst

author

2015-03-04 03:31 UTC #7 lore raw

PATCH 4/5 x86: Use generic compat audit code

In reply to Brian Gerst

Use the generic compat syscall audit code instead of an x86 specific implementation.

Signed-off-by Brian Gerst <[email protected]>

---
 arch/x86/Kconfig                |  5 +--
 arch/x86/ia32/Makefile          |  3 --
 arch/x86/ia32/audit.c           | 43 ---------------------
 arch/x86/include/asm/unistd32.h |  2 +
 arch/x86/kernel/Makefile        |  2 -
 arch/x86/kernel/audit_64.c      | 82 -----------------------------------------
 6 files changed, 3 insertions(+), 134 deletions(-)
 delete mode 100644 arch/x86/ia32/audit.c
 create mode 100644 arch/x86/include/asm/unistd32.h
 delete mode 100644 arch/x86/kernel/audit_64.c

arch/x86/Kconfig +1 -4

index c2fb8a8..ce75862 100644
@@ -135,6 +135,7 @@ config X86
 	select HAVE_CC_STACKPROTECTOR
 	select GENERIC_CPU_AUTOPROBE
 	select HAVE_ARCH_AUDITSYSCALL
+	select AUDIT_ARCH_COMPAT_GENERIC
 	select ARCH_SUPPORTS_ATOMIC_RMW
 	select HAVE_ACPI_APEI if ACPI
 	select HAVE_ACPI_APEI_NMI if ACPI

@@ -238,10 +239,6 @@ config ZONE_DMA32
 	bool
 	default X86_64
 
-config AUDIT_ARCH
-	bool
-	default X86_64
-
 config ARCH_SUPPORTS_OPTIMIZED_INLINING
 	def_bool y

arch/x86/ia32/Makefile +0 -3

index bb635c6..88ac1dd 100644
@@ -5,6 +5,3 @@
 obj-$(CONFIG_IA32_EMULATION) := ia32entry.o sys_ia32.o ia32_signal.o
 
 obj-$(CONFIG_IA32_AOUT) += ia32_aout.o
-
-audit-class-$(CONFIG_AUDIT) := audit.o
-obj-$(CONFIG_IA32_EMULATION) += $(audit-class-y)

arch/x86/ia32/audit.c deleted +0 ~~-43~~

43 lines not on this page · raw message

arch/x86/include/asm/unistd32.h new file +2 -0

new file mode 100644
index 0000000..3653918
@@ -0,0 +1,2 @@
+/* Generic code expects this header to be named unistd32.h */
+#include <asm/unistd_32.h>

arch/x86/kernel/Makefile +0 -2

index c887cd9..280eb3b 100644
@@ -115,8 +115,6 @@ obj-$(CONFIG_PMC_ATOM)			+= pmc_atom.o
 ###
 # 64 bit specific files
 ifeq ($(CONFIG_X86_64),y)
-	obj-$(CONFIG_AUDIT)		+= audit_64.o
-
 	obj-$(CONFIG_GART_IOMMU)	+= amd_gart_64.o aperture_64.o
 	obj-$(CONFIG_CALGARY_IOMMU)	+= pci-calgary_64.o tce_64.o

arch/x86/kernel/audit_64.c deleted +0 ~~-82~~

82 lines not on this page · raw message

Signature · 1 line

2.1.0

Ingo Molnar

2015-03-04 05:14 UTC #8 lore raw

Re: [PATCH 4/5] x86: Use generic compat audit code

Quoted 4 lines · Brian Gerst

Use the generic compat syscall audit code instead of an x86 specific implementation.
Signed-off-by Brian Gerst <[email protected]>

Quoted 11 lines

---
 arch/x86/Kconfig                |  5 +--
 arch/x86/ia32/Makefile          |  3 --
 arch/x86/ia32/audit.c           | 43 ---------------------
 arch/x86/include/asm/unistd32.h |  2 +
 arch/x86/kernel/Makefile        |  2 -
 arch/x86/kernel/audit_64.c      | 82 -----------------------------------------
 6 files changed, 3 insertions(+), 134 deletions(-)
 delete mode 100644 arch/x86/ia32/audit.c
 create mode 100644 arch/x86/include/asm/unistd32.h
 delete mode 100644 arch/x86/kernel/audit_64.c

Nice!

Because there are many types of conversions it would be nice if the changelog included a declaration of some sorts about what this means
precisely: 'the two implementations were 100% identical', or 'the x86 one was buggy and we now switch to the correct generic one', or 'the only difference between the two is X, Y and Z, which is not a problem'?

Thanks,

Ingo

Brian Gerst

author

2015-03-04 06:39 UTC #9 lore raw

Re: [PATCH 4/5] x86: Use generic compat audit code

Quoted 28 lines · Ingo Molnar

Quoted 4 lines · Brian Gerst
Use the generic compat syscall audit code instead of an x86 specific implementation.
Signed-off-by Brian Gerst <[email protected]>
Quoted 11 lines
---
 arch/x86/Kconfig                |  5 +--
 arch/x86/ia32/Makefile          |  3 --
 arch/x86/ia32/audit.c           | 43 ---------------------
 arch/x86/include/asm/unistd32.h |  2 +
 arch/x86/kernel/Makefile        |  2 -
 arch/x86/kernel/audit_64.c      | 82 -----------------------------------------
 6 files changed, 3 insertions(+), 134 deletions(-)
 delete mode 100644 arch/x86/ia32/audit.c
 create mode 100644 arch/x86/include/asm/unistd32.h
 delete mode 100644 arch/x86/kernel/audit_64.c
Nice!
Because there are many types of conversions it would be nice if the changelog included a declaration of some sorts about what this means
precisely: 'the two implementations were 100% identical', or 'the x86 one was buggy and we now switch to the correct generic one', or 'the only difference between the two is X, Y and Z, which is not a problem'?

Quoted 3 lines

Thanks,
Ingo

The generic compat version was added along with the arm64 support. It is functionally equivalent to the x86-specific implementation. The main difference is using audit_is_compat() instead of a hardcoded check for AUDIT_ARCH_I386.

I just noticed a bug in lib/compat_audit.c though. It is missing __NR_execveat.

--
Brian Gerst

Ingo Molnar

2015-03-04 07:09 UTC #10 lore raw

Re: [PATCH 4/5] x86: Use generic compat audit code

Quoted 33 lines · Brian Gerst

Quoted 28 lines · Ingo Molnar
Quoted 4 lines · Brian Gerst
Use the generic compat syscall audit code instead of an x86 specific implementation.
Signed-off-by Brian Gerst <[email protected]>
Quoted 11 lines
---
 arch/x86/Kconfig                |  5 +--
 arch/x86/ia32/Makefile          |  3 --
 arch/x86/ia32/audit.c           | 43 ---------------------
 arch/x86/include/asm/unistd32.h |  2 +
 arch/x86/kernel/Makefile        |  2 -
 arch/x86/kernel/audit_64.c      | 82 -----------------------------------------
 6 files changed, 3 insertions(+), 134 deletions(-)
 delete mode 100644 arch/x86/ia32/audit.c
 create mode 100644 arch/x86/include/asm/unistd32.h
 delete mode 100644 arch/x86/kernel/audit_64.c
Nice!
Because there are many types of conversions it would be nice if the changelog included a declaration of some sorts about what this means
precisely: 'the two implementations were 100% identical', or 'the x86 one was buggy and we now switch to the correct generic one', or 'the only difference between the two is X, Y and Z, which is not a problem'?
Quoted 3 lines
Thanks,
Ingo

Quoted 7 lines

The generic compat version was added along with the arm64 support. It is functionally equivalent to the x86-specific implementation. The main difference is using audit_is_compat() instead of a hardcoded check for AUDIT_ARCH_I386.
I just noticed a bug in lib/compat_audit.c though. It is missing __NR_execveat.

Another asymmetry is that lib/compat_audit.c covers __NR_socketcall, while arch/x86/kernel/audit.c does not.

But that's an improvement I think: for __NR_socketcall the old x86 code would not do the socketcall permission check for SYS_BIND, which is an audit record generation bug AFAICS.

Thanks,

Ingo

David Drysdale

2015-03-04 14:52 UTC #11 lore raw

Re: [PATCH 4/5] x86: Use generic compat audit code

In reply to Brian Gerst

Quoted 32 lines · Brian Gerst

Quoted 30 lines · Ingo Molnar
Quoted 4 lines · Brian Gerst
Use the generic compat syscall audit code instead of an x86 specific implementation.
Signed-off-by Brian Gerst <[email protected]>
Quoted 11 lines
---
 arch/x86/Kconfig                |  5 +--
 arch/x86/ia32/Makefile          |  3 --
 arch/x86/ia32/audit.c           | 43 ---------------------
 arch/x86/include/asm/unistd32.h |  2 +
 arch/x86/kernel/Makefile        |  2 -
 arch/x86/kernel/audit_64.c      | 82 -----------------------------------------
 6 files changed, 3 insertions(+), 134 deletions(-)
 delete mode 100644 arch/x86/ia32/audit.c
 create mode 100644 arch/x86/include/asm/unistd32.h
 delete mode 100644 arch/x86/kernel/audit_64.c
Nice!
Because there are many types of conversions it would be nice if the changelog included a declaration of some sorts about what this means
precisely: 'the two implementations were 100% identical', or 'the x86 one was buggy and we now switch to the correct generic one', or 'the only difference between the two is X, Y and Z, which is not a problem'?
Thanks,

Quoted 8 lines

Quoted 1 line
Ingo
The generic compat version was added along with the arm64 support. It is functionally equivalent to the x86-specific implementation. The main difference is using audit_is_compat() instead of a hardcoded check for AUDIT_ARCH_I386.
I just noticed a bug in lib/compat_audit.c though. It is missing __NR_execveat.

Sorry I missed that. I'll send a patch (and cc: stable for 3.19).

Quoted 2 lines

--
Brian Gerst

Brian Gerst

author

2015-03-04 03:32 UTC #12 lore raw

PATCH 5/5 x86: Clean up ia32/Makefile

In reply to Brian Gerst

The ia32/ subdirectory is only entered if CONFIG_IA32_EMULATION is set. There is no need to depend on it in ia32/Makefile.

Signed-off-by Brian Gerst <[email protected]>

---
 arch/x86/ia32/Makefile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

arch/x86/ia32/Makefile +1 -1

index 88ac1dd..eef2a8d 100644
@@ -2,6 +2,6 @@
 # Makefile for the ia32 kernel emulation subsystem.
 #
 
-obj-$(CONFIG_IA32_EMULATION) := ia32entry.o sys_ia32.o ia32_signal.o
+obj-y := ia32entry.o sys_ia32.o ia32_signal.o
 
 obj-$(CONFIG_IA32_AOUT) += ia32_aout.o

Signature · 1 line

2.1.0

Ingo Molnar

2015-03-04 05:24 UTC #13 lore raw

In reply to Brian Gerst

Quoted 10 lines · Brian Gerst

This series cleans up some files in the 32-bit emulation code, by merging with the native 32-bit code, or in the case of the syscall audit, using the generic compat code that was recently added.
Brian Gerst (5):
x86: Remove compat_ni_syscall()
x86: Merge native and compat 32-bit syscall tables
x86: Remove sys32_vm86_warning
x86: Use generic compat audit code
x86: Clean up ia32/Makefile

So this series is looking pretty good to me, I've picked up #1-#3, pending clarification on #4. I've stuck them into tip:x86/asm, because they might interact with ongoing system call entry work.

While we are at it: CONFIG_IA32_EMULATION is a misnomer: there's nothing 'emulated' there, it's full software and hardware support for the 32-bit ABI. It's just an ugly (and somewhat confusing) replacement for CONFIG_COMPAT that most other architectures are using.

So could we also get rid of CONFIG_IA32_EMULATION and change most of its uses over to CONFIG_COMPAT et al? The only complication with that might be that IA32_EMULATION kconfig switch hardcodes a few features currently:

select BINFMT_ELF
select COMPAT_BINFMT_ELF
select HAVE_UID16

allowing the binfmt formats to be undefined should be OK; the uid16 dependency needs to be checked, whether it's safe to allow it to be disabled.

Thanks,

Ingo

Brian Gerst

author

2015-03-04 06:27 UTC #14 lore raw

Quoted 32 lines · Ingo Molnar

Quoted 10 lines · Brian Gerst
This series cleans up some files in the 32-bit emulation code, by merging with the native 32-bit code, or in the case of the syscall audit, using the generic compat code that was recently added.
Brian Gerst (5):
x86: Remove compat_ni_syscall()
x86: Merge native and compat 32-bit syscall tables
x86: Remove sys32_vm86_warning
x86: Use generic compat audit code
x86: Clean up ia32/Makefile
So this series is looking pretty good to me, I've picked up #1-#3, pending clarification on #4. I've stuck them into tip:x86/asm, because they might interact with ongoing system call entry work.
While we are at it: CONFIG_IA32_EMULATION is a misnomer: there's nothing 'emulated' there, it's full software and hardware support for the 32-bit ABI. It's just an ugly (and somewhat confusing) replacement for CONFIG_COMPAT that most other architectures are using.
So could we also get rid of CONFIG_IA32_EMULATION and change most of its uses over to CONFIG_COMPAT et al? The only complication with that might be that IA32_EMULATION kconfig switch hardcodes a few features currently:
select BINFMT_ELF
select COMPAT_BINFMT_ELF
select HAVE_UID16

Quoted 7 lines

allowing the binfmt formats to be undefined should be OK; the uid16 dependency needs to be checked, whether it's safe to allow it to be disabled.
Thanks,
Ingo

It looks like most cases can simply be replaced. There are some subtle relations with the x32 ABI though, and that will take some time to sort out.

--
Brian Gerst

H. Peter Anvin

reviewer

2015-03-05 22:38 UTC #15 lore raw

Quoted 5 lines · Brian Gerst

It looks like most cases can simply be replaced. There are some subtle relations with the x32 ABI though, and that will take some time to sort out.

Yes, some aspects of CONFIG_COMPAT is used also by x32.

Overall the current Linux kernel handles the case of *more* than two ABIs extremely poorly. It is not the only case of more than two ABIs, either; consider ARM for example.

-hpa