Skip to contentSkip to conversation

Add generic vdso_base tracking

PATCH In review Dmitry Safonov 8 participants 45 messages Jun 11, 2021 – Oct 11, 2023

Subsystems (26)
lore
Share & tools
lore
#0 lore raw

v3 cover Add generic vdso_base tracking

v3 Changes:
- Migrated arch/powerpc to vdso_base
- Added x86/selftest for unmapped vdso & no landing on fast syscall
- Review comments from Andy & Christophe (thanks!)
- Amended s/born process/execed process/ everywhere I noticed
- Build robot warning on cast from __user pointer

I've tested it on x86, I would appreciate any help with Tested-by on arm/arm64/mips/powerpc/s390/... platforms.

One thing I've noticed while cooking this and haven't found a clean way to solve is zero-terminated .pages[] array in vdso mappings, which is not always zero-terminated but works by the reason of VM_DONTEXPAND on mappings.

v2 Changes:
- Rename user_landing to vdso_base as it tracks vDSO VMA start address, rather than the explicit address to land (Andy)
- Reword and don't use "new-execed" and "new-born" task (Andy)
- Fix failures reported by build robot

Started from discussion [1], where was noted that currently a couple of architectures support mremap() for vdso/sigpage, but not munmap(). If an application maps something on the ex-place of vdso/sigpage, later after processing signal it will land there (good luck!)

Patches set is based on linux-next (next-20201123) and it depends on changes in x86/cleanups (those reclaim TIF_IA32/TIF_X32) and also on my changes in akpm (fixing several mremap() issues).

Logically, the patches set divides on:
- patch 1: a cleanup for patches in x86/cleanups
- patches 2-13: cleanups for arch_setup_additional_pages()
- patches 13-14: x86 signal changes for unmapped vdso
- patches 15-22: provide generic vdso_base in mm_struct
- patch 23: selftest for unmapped vDSO & fast syscalls

In the end, besides cleanups, it's now more predictable what happens for applications with unmapped vdso on architectures those support .mremap() for vdso/sigpage.

I'm aware of only one user that unmaps vdso - Valgrind [2]. (there possibly are more, but this one is "special", it unmaps vdso, but
not vvar, which confuses CRIU [Checkpoint Restore In Userspace], that's why I'm aware of it)

Patches as a .git branch:
https://github.com/0x7f454c46/linux/tree/vdso_base-v3

v2 Link:
https://lore.kernel.org/lkml/[email protected]/
v1 Link:
https://lore.kernel.org/lkml/[email protected]/

[1]: https://lore.kernel.org/linux-arch/CAJwJo6ZANqYkSHbQ+3b+Fi_VT80MtrzEV5yreQAWx-L8j8x2zA@mail.gmail.com/
[2]: https://github.com/checkpoint-restore/criu/issues/488

Cc Alexander Viro <[email protected]>
Cc Andrew Morton <[email protected]>
Cc Andy Lutomirski <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Borislav Petkov <[email protected]>
Cc Catalin Marinas <[email protected]>
Cc Christophe Leroy <[email protected]>
Cc Guo Ren <[email protected]>
Cc "H. Peter Anvin" <[email protected]>
Cc Ingo Molnar <[email protected]>
Cc Oleg Nesterov <[email protected]>
Cc Russell King <[email protected]>
Cc Thomas Bogendoerfer <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Vincenzo Frascino <[email protected]>
Cc Will Deacon <[email protected]>

Dmitry Safonov (23):
x86/elf: Check in_x32_syscall() in
compat_arch_setup_additional_pages()
elf: Move arch_setup_additional_pages() to generic elf.h arm/elf: Remove needless ifdef CONFIG_MMU
arm64: Use in_compat_task() in arch_setup_additional_pages()
x86: Remove compat_arch_setup_additional_pages()
elf: Remove compat_arch_setup_additional_pages()
vdso: Set mm->context.vdso only on success of
_install_special_mapping()
elf/vdso: Modify arch_setup_additional_pages() parameters
elf: Use sysinfo_ehdr in ARCH_DLINFO()
arm/vdso: Remove vdso pointer from mm->context
s390/vdso: Remove vdso_base pointer from mm->context sparc/vdso: Remove vdso pointer from mm->context
mm/mmap: Make vm_special_mapping::mremap return void x86/signal: Land on &frame->retcode when vdso isn't mapped x86/signal: Check if vdso_image_32 is mapped before trying to land on it
mm: Add vdso_base in mm_struct
x86/vdso: Migrate to generic vdso_base
arm/vdso: Migrate to generic vdso_base
arm64/vdso: Migrate compat signals to generic vdso_base arm64/vdso: Migrate native signals to generic vdso_base mips/vdso: Migrate to generic vdso_base
powerpc/vdso: Migrate native signals to generic vdso_base x86/vdso/selftest: Add a test for unmapping vDSO

 arch/Kconfig                                  |   3 +
 arch/alpha/include/asm/elf.h                  |   2 +-
 arch/arm/Kconfig                              |   2 +
 arch/arm/include/asm/elf.h                    |  10 +-
 arch/arm/include/asm/mmu.h                    |   3 -
 arch/arm/include/asm/vdso.h                   |   6 +-
 arch/arm/kernel/process.c                     |  14 +-
 arch/arm/kernel/signal.c                      |   6 +-
 arch/arm/kernel/vdso.c                        |  20 +--
 arch/arm64/Kconfig                            |   2 +
 arch/arm64/include/asm/elf.h                  |  27 +---
 arch/arm64/include/asm/mmu.h                  |   4 -
 arch/arm64/kernel/signal.c                    |  10 +-
 arch/arm64/kernel/signal32.c                  |  17 +-
 arch/arm64/kernel/vdso.c                      |  72 +++------
 arch/csky/Kconfig                             |   1 +
 arch/csky/include/asm/elf.h                   |   4 -
 arch/csky/kernel/vdso.c                       |  11 +-
 arch/hexagon/Kconfig                          |   1 +
 arch/hexagon/include/asm/elf.h                |   6 -
 arch/hexagon/kernel/vdso.c                    |   3 +-
 arch/ia64/include/asm/elf.h                   |   2 +-
 arch/mips/Kconfig                             |   2 +
 arch/mips/include/asm/elf.h                   |  10 +-
 arch/mips/include/asm/mmu.h                   |   2 -
 arch/mips/kernel/signal.c                     |  11 +-
 arch/mips/kernel/vdso.c                       |   5 +-
 arch/mips/vdso/genvdso.c                      |   9 --
 arch/nds32/Kconfig                            |   1 +
 arch/nds32/include/asm/elf.h                  |   8 +-
 arch/nds32/kernel/vdso.c                      |   8 +-
 arch/nios2/Kconfig                            |   1 +
 arch/nios2/include/asm/elf.h                  |   4 -
 arch/nios2/mm/init.c                          |   2 +-
 arch/powerpc/Kconfig                          |   2 +
 arch/powerpc/include/asm/book3s/32/mmu-hash.h |   1 -
 arch/powerpc/include/asm/book3s/64/mmu.h      |   1 -
 arch/powerpc/include/asm/elf.h                |   9 +-
 arch/powerpc/include/asm/mmu_context.h        |   9 --
 arch/powerpc/include/asm/nohash/32/mmu-40x.h  |   1 -
 arch/powerpc/include/asm/nohash/32/mmu-44x.h  |   1 -
 arch/powerpc/include/asm/nohash/32/mmu-8xx.h  |   1 -
 arch/powerpc/include/asm/nohash/mmu-book3e.h  |   1 -
 arch/powerpc/kernel/signal_32.c               |   8 +-
 arch/powerpc/kernel/signal_64.c               |   4 +-
 arch/powerpc/kernel/vdso.c                    |  48 +-----
 arch/powerpc/perf/callchain_32.c              |   8 +-
 arch/powerpc/perf/callchain_64.c              |   4 +-
 arch/riscv/Kconfig                            |   1 +
 arch/riscv/include/asm/elf.h                  |   9 +-
 arch/riscv/kernel/vdso.c                      |  11 +-
 arch/s390/Kconfig                             |   1 +
 arch/s390/include/asm/elf.h                   |  10 +-
 arch/s390/include/asm/mmu.h                   |   1 -
 arch/s390/kernel/vdso.c                       |  12 +-
 arch/sh/Kconfig                               |   1 +
 arch/sh/include/asm/elf.h                     |  16 +-
 arch/sh/kernel/vsyscall/vsyscall.c            |   3 +-
 arch/sparc/Kconfig                            |   1 +
 arch/sparc/include/asm/elf_64.h               |  11 +-
 arch/sparc/include/asm/mmu_64.h               |   1 -
 arch/sparc/vdso/vma.c                         |  20 +--
 arch/x86/Kconfig                              |   2 +
 arch/x86/entry/common.c                       |  10 +-
 arch/x86/entry/vdso/extable.c                 |   4 +-
 arch/x86/entry/vdso/vma.c                     |  79 ++++-----
 arch/x86/ia32/ia32_signal.c                   |  18 ++-
 arch/x86/include/asm/compat.h                 |   6 +
 arch/x86/include/asm/elf.h                    |  44 ++---
 arch/x86/include/asm/mmu.h                    |   1 -
 arch/x86/include/asm/mmu_context.h            |   5 -
 arch/x86/include/asm/vdso.h                   |   4 +
 arch/x86/kernel/cpu/resctrl/pseudo_lock.c     |   3 +-
 arch/x86/kernel/signal.c                      |  25 ++-
 arch/x86/um/asm/elf.h                         |   9 +-
 arch/x86/um/vdso/vma.c                        |   2 +-
 fs/Kconfig.binfmt                             |   3 +
 fs/aio.c                                      |   3 +-
 fs/binfmt_elf.c                               |  19 ++-
 fs/binfmt_elf_fdpic.c                         |  17 +-
 fs/compat_binfmt_elf.c                        |  12 --
 include/asm-generic/mm_hooks.h                |   9 +-
 include/linux/elf.h                           |  24 ++-
 include/linux/mm.h                            |   3 +-
 include/linux/mm_types.h                      |  21 ++-
 kernel/fork.c                                 |   1 +
 mm/mmap.c                                     |  28 ++--
 mm/mremap.c                                   |   2 +-
 tools/testing/selftests/x86/.gitignore        |   1 +
 tools/testing/selftests/x86/Makefile          |  11 +-
 .../testing/selftests/x86/test_munmap_vdso.c  | 151 ++++++++++++++++++
 91 files changed, 511 insertions(+), 491 deletions(-)
 create mode 100644 tools/testing/selftests/x86/test_munmap_vdso.c

base-commit: 614124bea77e452aa6df7a8714e8bc820b489922

Signature · 1 line
2.31.1
#1 lore raw

PATCH v3 1/23 x86/elf: Check in_x32_syscall() in compat_arch_setup_additional_pages()

Partly revert commit 3316ec8ccd34 ("x86/elf: Use e_machine to check for x32/ia32 in setup_additional_pages()") and commit 9a29a671902c ("elf:
Expose ELF header on arch_setup_additional_pages()".
Both patches did a good thing: removed usage of TIF_X32, but with a price of additional macros ARCH_SETUP_ADDITIONAL_PAGES() and ifdeffs.

Instead, use in_x32_syscall() - the first thing load_elf_binary() does after parsing and checking new ELF binary. It's done that early after exec() that mmap() code already uses it straight away, which is needed to know which mmap_base to use (see arch_pick_mmap_layout()). Add comments to describe how it works.

---
 arch/x86/entry/vdso/vma.c     |  4 ++--
 arch/x86/include/asm/compat.h |  6 ++++++
 arch/x86/include/asm/elf.h    |  6 ++----
 fs/binfmt_elf.c               | 10 +++++++---
 fs/compat_binfmt_elf.c        | 11 +++--------
 include/linux/elf.h           |  5 -----
 6 files changed, 20 insertions(+), 22 deletions(-)
arch/x86/entry/vdso/vma.c +2 -2
index 235a5794296a..43d42ce82e86 100644
@@ -396,10 +396,10 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 
 #ifdef CONFIG_COMPAT
 int compat_arch_setup_additional_pages(struct linux_binprm *bprm,
-				       int uses_interp, bool x32)
+				       int uses_interp)
 {
 #ifdef CONFIG_X86_X32_ABI
-	if (x32) {
+	if (in_x32_syscall()) {
 		if (!vdso64_enabled)
 			return 0;
 		return map_vdso_randomized(&vdso_image_x32);
arch/x86/include/asm/compat.h +6 -0
index be09c7eac89f..58f76665042c 100644
@@ -186,6 +186,12 @@ static inline bool in_x32_syscall(void)
 	return false;
 }
 
+/*
+ * Valid all the time on the context of a process that performs syscall.
+ * Just exec()ed process has __X32_SYSCALL_BIT or TS_COMPAT set very
+ * early in load_binary() on setting personality and flags.
+ * See also set_personality_ia32().
+ */
 static inline bool in_32bit_syscall(void)
 {
 	return in_ia32_syscall() || in_x32_syscall();
arch/x86/include/asm/elf.h +2 -4
index 7d7500806af8..c0b5733005af 100644
@@ -383,10 +383,8 @@ struct linux_binprm;
 extern int arch_setup_additional_pages(struct linux_binprm *bprm,
 				       int uses_interp);
 extern int compat_arch_setup_additional_pages(struct linux_binprm *bprm,
-					      int uses_interp, bool x32);
-#define COMPAT_ARCH_SETUP_ADDITIONAL_PAGES(bprm, ex, interpreter)	\
-	compat_arch_setup_additional_pages(bprm, interpreter,		\
-					   (ex->e_machine == EM_X86_64))
+					      int uses_interp);
+#define compat_arch_setup_additional_pages compat_arch_setup_additional_pages
 
 extern bool arch_syscall_is_vdso_sigreturn(struct pt_regs *regs);
 
fs/binfmt_elf.c +7 -3
index 187b3f2b9202..2347d9067df6 100644
@@ -1002,8 +1002,12 @@ static int load_elf_binary(struct linux_binprm *bprm)
 	if (retval)
 		goto out_free_dentry;
 
-	/* Do this immediately, since STACK_TOP as used in setup_arg_pages
-	   may depend on the personality.  */
+	/*
+	 * Do this immediately, since STACK_TOP as used in setup_arg_pages
+	 * may depend on the personality. At this moment sys_exec()
+	 * for loaded compatible application becomes non-native syscall
+	 * and in_compat_syscall() starts working.
+	 */
 	SET_PERSONALITY2(*elf_ex, &arch_state);
 	if (elf_read_implies_exec(*elf_ex, executable_stack))
 		current->personality |= READ_IMPLIES_EXEC;
@@ -1249,7 +1253,7 @@ static int load_elf_binary(struct linux_binprm *bprm)
 	set_binfmt(&elf_format);
 
 #ifdef ARCH_HAS_SETUP_ADDITIONAL_PAGES
-	retval = ARCH_SETUP_ADDITIONAL_PAGES(bprm, elf_ex, !!interpreter);
+	retval = arch_setup_additional_pages(bprm, !!interpreter);
 	if (retval < 0)
 		goto out;
 #endif /* ARCH_HAS_SETUP_ADDITIONAL_PAGES */
fs/compat_binfmt_elf.c +3 -8
index 95e72d271b95..049ba7c011b9 100644
@@ -110,16 +110,11 @@
 #define START_THREAD		COMPAT_START_THREAD
 #endif
 
-#ifdef compat_arch_setup_additional_pages
-#define COMPAT_ARCH_SETUP_ADDITIONAL_PAGES(bprm, ex, interpreter) \
-	compat_arch_setup_additional_pages(bprm, interpreter)
-#endif
-
-#ifdef	COMPAT_ARCH_SETUP_ADDITIONAL_PAGES
+#ifdef	compat_arch_setup_additional_pages
 #undef	ARCH_HAS_SETUP_ADDITIONAL_PAGES
 #define ARCH_HAS_SETUP_ADDITIONAL_PAGES 1
-#undef	ARCH_SETUP_ADDITIONAL_PAGES
-#define	ARCH_SETUP_ADDITIONAL_PAGES COMPAT_ARCH_SETUP_ADDITIONAL_PAGES
+#undef	arch_setup_additional_pages
+#define	arch_setup_additional_pages compat_arch_setup_additional_pages
 #endif
 
 #ifdef	compat_elf_read_implies_exec
include/linux/elf.h +0 -5
index c9a46c4e183b..6dbcfe7a3fd7 100644
@@ -27,11 +27,6 @@
 	start_thread(regs, elf_entry, start_stack)
 #endif
 
-#if defined(ARCH_HAS_SETUP_ADDITIONAL_PAGES) && !defined(ARCH_SETUP_ADDITIONAL_PAGES)
-#define ARCH_SETUP_ADDITIONAL_PAGES(bprm, ex, interpreter) \
-	arch_setup_additional_pages(bprm, interpreter)
-#endif
-
 #define ELF32_GNU_PROPERTY_ALIGN	4
 #define ELF64_GNU_PROPERTY_ALIGN	8
 
Signature · 1 line
2.31.1
#2 lore raw

Re: [PATCH v3 01/23] x86/elf: Check in_x32_syscall() in compat_arch_setup_additional_pages()

Quoted 11 lines · Dmitry Safonov

Partly revert commit 3316ec8ccd34 ("x86/elf: Use e_machine to check for x32/ia32 in setup_additional_pages()") and commit 9a29a671902c ("elf:
Expose ELF header on arch_setup_additional_pages()".
Both patches did a good thing: removed usage of TIF_X32, but with a price of additional macros ARCH_SETUP_ADDITIONAL_PAGES() and ifdeffs.

Instead, use in_x32_syscall() - the first thing load_elf_binary() does after parsing and checking new ELF binary. It's done that early after exec() that mmap() code already uses it straight away, which is needed to know which mmap_base to use (see arch_pick_mmap_layout()). Add comments to describe how it works.

I still have no idea what this is trying to solve. All you describe is what this does.

Thanks,

tglx

#3 lore raw

Re: [PATCH v3 01/23] x86/elf: Check in_x32_syscall() in compat_arch_setup_additional_pages()

Hi Thomas,

Quoted 7 lines · Thomas Gleixner
Quoted 6 lines · Dmitry Safonov

Partly revert commit 3316ec8ccd34 ("x86/elf: Use e_machine to check for x32/ia32 in setup_additional_pages()") and commit 9a29a671902c ("elf:
Expose ELF header on arch_setup_additional_pages()".
Both patches did a good thing: removed usage of TIF_X32, but with a price of additional macros ARCH_SETUP_ADDITIONAL_PAGES() and ifdeffs.

Quoted 8 lines
Quoted 5 lines

Instead, use in_x32_syscall() - the first thing load_elf_binary() does after parsing and checking new ELF binary. It's done that early after exec() that mmap() code already uses it straight away, which is needed to know which mmap_base to use (see arch_pick_mmap_layout()). Add comments to describe how it works.

I still have no idea what this is trying to solve. All you describe is what this does.

Yeah, I still have problems with describing why rather than what sometimes..

How about something like this:

In order to simplify code, new macro ARCH_SETUP_ADDITIONAL_PAGES() can be dropped as well as a new argument in compat_arch_setup_additional_pages(). Resulting in partial revert of the commit 3316ec8ccd34 ("x86/elf: Use e_machine to check for x32/ia32 in setup_additional_pages()") and the commit 9a29a671902c ("elf: Expose ELF header on arch_setup_additional_pages()".

Instead of a new macro and an additional function argument, use in_x32_syscall() which is already valid at the moment of premapping vDSO. Add comments to describe how it works.

Thanks for reviewing,
Dmitry

#4 lore raw

PATCH v3 2/23 elf: Move arch_setup_additional_pages() to generic elf.h

In reply to Dmitry Safonov

Ifdef the function in the header, not around it's usage in the code. Following kernel style, move it to Kconfig.
Makes it easier to follow when the option is enabled/disabled. Remove re-definition from compat_binfmt_elf, as it's always defined for architectures that define compat_arch_setup_additional_pages (arm64/x86). CONFIG_VDSO depends on MMU, so

---
 arch/arm/Kconfig                |  1 +
 arch/arm/include/asm/elf.h      |  3 ---
 arch/arm64/Kconfig              |  1 +
 arch/arm64/include/asm/elf.h    |  6 +-----
 arch/csky/Kconfig               |  1 +
 arch/csky/include/asm/elf.h     |  4 ----
 arch/hexagon/Kconfig            |  1 +
 arch/hexagon/include/asm/elf.h  |  6 ------
 arch/mips/Kconfig               |  1 +
 arch/mips/include/asm/elf.h     |  5 -----
 arch/nds32/Kconfig              |  1 +
 arch/nds32/include/asm/elf.h    |  3 ---
 arch/nios2/Kconfig              |  1 +
 arch/nios2/include/asm/elf.h    |  4 ----
 arch/powerpc/Kconfig            |  1 +
 arch/powerpc/include/asm/elf.h  |  5 -----
 arch/riscv/Kconfig              |  1 +
 arch/riscv/include/asm/elf.h    |  4 ----
 arch/s390/Kconfig               |  1 +
 arch/s390/include/asm/elf.h     |  5 -----
 arch/sh/Kconfig                 |  1 +
 arch/sh/include/asm/elf.h       |  6 ------
 arch/sparc/Kconfig              |  1 +
 arch/sparc/include/asm/elf_64.h |  6 ------
 arch/x86/Kconfig                |  1 +
 arch/x86/include/asm/elf.h      |  4 ----
 arch/x86/um/asm/elf.h           |  5 -----
 fs/Kconfig.binfmt               |  3 +++
 fs/binfmt_elf.c                 |  2 --
 fs/binfmt_elf_fdpic.c           |  3 +--
 fs/compat_binfmt_elf.c          |  2 --
 include/linux/elf.h             | 12 ++++++++++++
 32 files changed, 30 insertions(+), 71 deletions(-)
arch/arm/Kconfig +1 -0
index 24804f11302d..2df5ad505b8b 100644
@@ -14,6 +14,7 @@ config ARM
 	select ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE
 	select ARCH_HAS_PTE_SPECIAL if ARM_LPAE
 	select ARCH_HAS_PHYS_TO_DMA
+	select ARCH_HAS_SETUP_ADDITIONAL_PAGES if MMU
 	select ARCH_HAS_SETUP_DMA_OPS
 	select ARCH_HAS_SET_MEMORY
 	select ARCH_HAS_STRICT_KERNEL_RWX if MMU && !XIP_KERNEL
arch/arm/include/asm/elf.h +0 -3
index b8102a6ddf16..a7cd90b3a779 100644
@@ -145,9 +145,6 @@ do {								\
 		    (elf_addr_t)current->mm->context.vdso);	\
 } while (0)
 #endif
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES 1
-struct linux_binprm;
-int arch_setup_additional_pages(struct linux_binprm *, int);
 #endif
 
 #endif
arch/arm64/Kconfig +1 -0
index 9f1d8566bbf9..385ef8d8ad9b 100644
@@ -31,6 +31,7 @@ config ARM64
 	select ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE
 	select ARCH_HAS_PTE_DEVMAP
 	select ARCH_HAS_PTE_SPECIAL
+	select ARCH_HAS_SETUP_ADDITIONAL_PAGES
 	select ARCH_HAS_SETUP_DMA_OPS
 	select ARCH_HAS_SET_DIRECT_MAP
 	select ARCH_HAS_SET_MEMORY
arch/arm64/include/asm/elf.h +1 -5
index 8d1c8dcb87fd..d1073ffa7f24 100644
@@ -181,11 +181,6 @@ do {									\
 		NEW_AUX_ENT(AT_IGNORE, 0);				\
 } while (0)
 
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES
-struct linux_binprm;
-extern int arch_setup_additional_pages(struct linux_binprm *bprm,
-				       int uses_interp);
-
 /* 1GB of VA */
 #ifdef CONFIG_COMPAT
 #define STACK_RND_MASK			(test_thread_flag(TIF_32BIT) ? \
@@ -242,6 +237,7 @@ do {									\
 #else
 #define COMPAT_ARCH_DLINFO
 #endif
+struct linux_binprm;
 extern int aarch32_setup_additional_pages(struct linux_binprm *bprm,
 					  int uses_interp);
 #define compat_arch_setup_additional_pages \
arch/csky/Kconfig +1 -0
index 8de5b987edb9..68139fa18691 100644
@@ -4,6 +4,7 @@ config CSKY
 	select ARCH_32BIT_OFF_T
 	select ARCH_HAS_DMA_PREP_COHERENT
 	select ARCH_HAS_GCOV_PROFILE_ALL
+	select ARCH_HAS_SETUP_ADDITIONAL_PAGES
 	select ARCH_HAS_SYNC_DMA_FOR_CPU
 	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
 	select ARCH_USE_BUILTIN_BSWAP
arch/csky/include/asm/elf.h +0 -4
index 48b83e283ed4..89067e028335 100644
@@ -83,8 +83,4 @@ extern int dump_task_regs(struct task_struct *tsk, elf_gregset_t *elf_regs);
 #define ELF_PLATFORM		(NULL)
 #define SET_PERSONALITY(ex)	set_personality(PER_LINUX)
 
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES 1
-struct linux_binprm;
-extern int arch_setup_additional_pages(struct linux_binprm *bprm,
-				       int uses_interp);
 #endif /* __ASM_CSKY_ELF_H */
arch/hexagon/Kconfig +1 -0
index 44a409967af1..04dc816d04bd 100644
@@ -5,6 +5,7 @@ comment "Linux Kernel Configuration for Hexagon"
 config HEXAGON
 	def_bool y
 	select ARCH_32BIT_OFF_T
+	select ARCH_HAS_SETUP_ADDITIONAL_PAGES
 	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
 	select ARCH_NO_PREEMPT
 	# Other pending projects/to-do items.
arch/hexagon/include/asm/elf.h +0 -6
index 5bfdd9b147fd..eba4131610aa 100644
@@ -207,10 +207,4 @@ do {					\
  */
 #define ELF_PLATFORM  (NULL)
 
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES 1
-struct linux_binprm;
-extern int arch_setup_additional_pages(struct linux_binprm *bprm,
-				       int uses_interp);
-
-
 #endif
arch/mips/Kconfig +1 -0
index ed51970c08e7..81096dd2c1ef 100644
@@ -9,6 +9,7 @@ config MIPS
 	select ARCH_HAS_KCOV
 	select ARCH_HAS_NON_OVERLAPPING_ADDRESS_SPACE if !EVA
 	select ARCH_HAS_PTE_SPECIAL if !(32BIT && CPU_HAS_RIXI)
+	select ARCH_HAS_SETUP_ADDITIONAL_PAGES
 	select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
 	select ARCH_HAS_UBSAN_SANITIZE_ALL
 	select ARCH_HAS_GCOV_PROFILE_ALL
arch/mips/include/asm/elf.h +0 -5
index dc8d2863752c..a5c8be47a39d 100644
@@ -462,11 +462,6 @@ do {									\
 		    (unsigned long)current->mm->context.vdso);		\
 } while (0)
 
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES 1
-struct linux_binprm;
-extern int arch_setup_additional_pages(struct linux_binprm *bprm,
-				       int uses_interp);
-
 #ifdef CONFIG_MIPS_FP_SUPPORT
 
 struct arch_elf_state {
arch/nds32/Kconfig +1 -0
index 62313902d75d..02afe5ebdfff 100644
@@ -8,6 +8,7 @@ config NDS32
 	def_bool y
 	select ARCH_32BIT_OFF_T
 	select ARCH_HAS_DMA_PREP_COHERENT
+	select ARCH_HAS_SETUP_ADDITIONAL_PAGES
 	select ARCH_HAS_SYNC_DMA_FOR_CPU
 	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
 	select ARCH_WANT_FRAME_POINTERS if FTRACE
arch/nds32/include/asm/elf.h +0 -3
index 1853dc89b8ac..36cec4ae5a84 100644
@@ -173,8 +173,5 @@ do {								\
 	NEW_AUX_ENT(AT_SYSINFO_EHDR,				\
 		    (elf_addr_t)current->mm->context.vdso);	\
 } while (0)
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES 1
-struct linux_binprm;
-int arch_setup_additional_pages(struct linux_binprm *, int);
 
 #endif
arch/nios2/Kconfig +1 -0
index c24955c81c92..8159123a995e 100644
@@ -3,6 +3,7 @@ config NIOS2
 	def_bool y
 	select ARCH_32BIT_OFF_T
 	select ARCH_HAS_DMA_PREP_COHERENT
+	select ARCH_HAS_SETUP_ADDITIONAL_PAGES
 	select ARCH_HAS_SYNC_DMA_FOR_CPU
 	select ARCH_HAS_SYNC_DMA_FOR_DEVICE
 	select ARCH_HAS_DMA_SET_UNCACHED
arch/nios2/include/asm/elf.h +0 -4
index 984dd6de17c2..4f8baaef843f 100644
@@ -28,10 +28,6 @@
 /* regs is struct pt_regs, pr_reg is elf_gregset_t (which is
    now struct_user_regs, they are different) */
 
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES	1
-struct linux_binprm;
-extern int arch_setup_additional_pages(struct linux_binprm *bprm,
-	int uses_interp);
 #define ELF_CORE_COPY_REGS(pr_reg, regs)				\
 { do {									\
 	/* Bleech. */							\
arch/powerpc/Kconfig +1 -0
index 088dd2afcfe4..a9f842230ee4 100644
@@ -140,6 +140,7 @@ config PPC
 	select ARCH_HAS_PTE_DEVMAP		if PPC_BOOK3S_64
 	select ARCH_HAS_PTE_SPECIAL
 	select ARCH_HAS_SCALED_CPUTIME		if VIRT_CPU_ACCOUNTING_NATIVE && PPC_BOOK3S_64
+	select ARCH_HAS_SETUP_ADDITIONAL_PAGES
 	select ARCH_HAS_STRICT_KERNEL_RWX	if ((PPC_BOOK3S_64 || PPC32) && !HIBERNATION)
 	select ARCH_HAS_TICK_BROADCAST		if GENERIC_CLOCKEVENTS_BROADCAST
 	select ARCH_HAS_UACCESS_FLUSHCACHE
arch/powerpc/include/asm/elf.h +0 -5
index b8425e3cfd81..d7d9820c9096 100644
@@ -111,11 +111,6 @@ extern int dcache_bsize;
 extern int icache_bsize;
 extern int ucache_bsize;
 
-/* vDSO has arch_setup_additional_pages */
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES
-struct linux_binprm;
-extern int arch_setup_additional_pages(struct linux_binprm *bprm,
-				       int uses_interp);
 #define VDSO_AUX_ENT(a,b) NEW_AUX_ENT(a,b)
 
 /* 1GB for 64bit, 8MB for 32bit */
arch/riscv/Kconfig +1 -0
index c5914e70a0fd..1003fa4534a7 100644
@@ -28,6 +28,7 @@ config RISCV
 	select ARCH_HAS_PTE_SPECIAL
 	select ARCH_HAS_SET_DIRECT_MAP
 	select ARCH_HAS_SET_MEMORY
+	select ARCH_HAS_SETUP_ADDITIONAL_PAGES if MMU
 	select ARCH_HAS_STRICT_KERNEL_RWX if MMU && !XIP_KERNEL
 	select ARCH_HAS_STRICT_MODULE_RWX if MMU && !XIP_KERNEL
 	select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
arch/riscv/include/asm/elf.h +0 -4
index f4b490cd0e5d..1d1d60df632e 100644
@@ -75,10 +75,6 @@ do {								\
 	NEW_AUX_ENT(AT_L2_CACHEGEOMETRY,			\
 		get_cache_geometry(2, CACHE_TYPE_UNIFIED));	\
 } while (0)
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES
-struct linux_binprm;
-extern int arch_setup_additional_pages(struct linux_binprm *bprm,
-	int uses_interp);
 #endif /* CONFIG_MMU */
 
 #define ELF_CORE_COPY_REGS(dest, regs)			\
arch/s390/Kconfig +1 -0
index b4c7c34069f8..ccca15663345 100644
@@ -76,6 +76,7 @@ config S390
 	select ARCH_HAS_PTE_SPECIAL
 	select ARCH_HAS_SCALED_CPUTIME
 	select ARCH_HAS_SET_MEMORY
+	select ARCH_HAS_SETUP_ADDITIONAL_PAGES
 	select ARCH_HAS_STRICT_KERNEL_RWX
 	select ARCH_HAS_STRICT_MODULE_RWX
 	select ARCH_HAS_SYSCALL_WRAPPER
arch/s390/include/asm/elf.h +0 -5
index 66d51ad090ab..6583142149b0 100644
@@ -275,9 +275,4 @@ do {									    \
 			    (unsigned long)current->mm->context.vdso_base); \
 } while (0)
 
-struct linux_binprm;
-
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES 1
-int arch_setup_additional_pages(struct linux_binprm *, int);
-
 #endif
arch/sh/Kconfig +1 -0
index 68129537e350..9a1a39c5b3c4 100644
@@ -10,6 +10,7 @@ config SUPERH
 	select ARCH_HAS_GIGANTIC_PAGE
 	select ARCH_HAS_GCOV_PROFILE_ALL
 	select ARCH_HAS_PTE_SPECIAL
+	select ARCH_HAS_SETUP_ADDITIONAL_PAGES if VSYSCALL
 	select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
 	select ARCH_HIBERNATION_POSSIBLE if MMU
 	select ARCH_MIGHT_HAVE_PC_PARPORT
arch/sh/include/asm/elf.h +0 -6
index 2862d6d1cb64..9b3e22e771a1 100644
@@ -164,12 +164,6 @@ do {									\
 	set_personality(PER_LINUX_32BIT | (current->personality & (~PER_MASK)))
 
 #ifdef CONFIG_VSYSCALL
-/* vDSO has arch_setup_additional_pages */
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES
-struct linux_binprm;
-extern int arch_setup_additional_pages(struct linux_binprm *bprm,
-				       int uses_interp);
-
 extern unsigned int vdso_enabled;
 extern void __kernel_vsyscall;
 
arch/sparc/Kconfig +1 -0
index 164a5254c91c..10c719a73b90 100644
@@ -13,6 +13,7 @@ config 64BIT
 config SPARC
 	bool
 	default y
+	select ARCH_HAS_SETUP_ADDITIONAL_PAGES if SPARC64
 	select ARCH_MIGHT_HAVE_PC_PARPORT if SPARC64 && PCI
 	select ARCH_MIGHT_HAVE_PC_SERIO
 	select DMA_OPS
arch/sparc/include/asm/elf_64.h +0 -6
index 8fb09eec8c3e..63a622c36df3 100644
@@ -223,10 +223,4 @@ do {									\
 	NEW_AUX_ENT(AT_ADI_NBITS, adi_state.caps.nbits);		\
 	NEW_AUX_ENT(AT_ADI_UEONADI, adi_state.caps.ue_on_adi);		\
 } while (0)
-
-struct linux_binprm;
-
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES	1
-extern int arch_setup_additional_pages(struct linux_binprm *bprm,
-					int uses_interp);
 #endif /* !(__ASM_SPARC64_ELF_H) */
arch/x86/Kconfig +1 -0
index 0045e1b44190..d30b952f4453 100644
@@ -87,6 +87,7 @@ config X86
 	select ARCH_HAS_COPY_MC			if X86_64
 	select ARCH_HAS_SET_MEMORY
 	select ARCH_HAS_SET_DIRECT_MAP
+	select ARCH_HAS_SETUP_ADDITIONAL_PAGES
 	select ARCH_HAS_STRICT_KERNEL_RWX
 	select ARCH_HAS_STRICT_MODULE_RWX
 	select ARCH_HAS_SYNC_CORE_BEFORE_USERMODE
arch/x86/include/asm/elf.h +0 -4
index c0b5733005af..9ee5b3b3ba93 100644
@@ -378,10 +378,6 @@ else if (IS_ENABLED(CONFIG_IA32_EMULATION))				\
 	 vdso_image_32.sym___kernel_vsyscall)
 
 struct linux_binprm;
-
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES 1
-extern int arch_setup_additional_pages(struct linux_binprm *bprm,
-				       int uses_interp);
 extern int compat_arch_setup_additional_pages(struct linux_binprm *bprm,
 					      int uses_interp);
 #define compat_arch_setup_additional_pages compat_arch_setup_additional_pages
arch/x86/um/asm/elf.h +0 -5
index dcaf3b38a9e0..b7c03a760a3c 100644
@@ -181,11 +181,6 @@ do {								\
 #define FIXADDR_USER_START      0
 #define FIXADDR_USER_END        0
 
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES 1
-struct linux_binprm;
-extern int arch_setup_additional_pages(struct linux_binprm *bprm,
-	int uses_interp);
-
 extern unsigned long um_vdso_addr;
 #define AT_SYSINFO_EHDR 33
 #define ARCH_DLINFO	NEW_AUX_ENT(AT_SYSINFO_EHDR, um_vdso_addr)
fs/Kconfig.binfmt +3 -0
index 06fb7a93a1bd..d98206ad7749 100644
@@ -39,6 +39,9 @@ config ARCH_BINFMT_ELF_STATE
 config ARCH_HAVE_ELF_PROT
 	bool
 
+config ARCH_HAS_SETUP_ADDITIONAL_PAGES
+	bool
+
 config ARCH_USE_GNU_PROPERTY
 	bool
 
fs/binfmt_elf.c +0 -2
index 2347d9067df6..dac2713c10ee 100644
@@ -1252,11 +1252,9 @@ static int load_elf_binary(struct linux_binprm *bprm)
 
 	set_binfmt(&elf_format);
 
-#ifdef ARCH_HAS_SETUP_ADDITIONAL_PAGES
 	retval = arch_setup_additional_pages(bprm, !!interpreter);
 	if (retval < 0)
 		goto out;
-#endif /* ARCH_HAS_SETUP_ADDITIONAL_PAGES */
 
 	retval = create_elf_tables(bprm, elf_ex,
 			  load_addr, interp_load_addr, e_entry);
fs/binfmt_elf_fdpic.c +1 -2
index 2c99b102c860..11cbf20b19da 100644
@@ -374,11 +374,10 @@ static int load_elf_fdpic_binary(struct linux_binprm *bprm)
 				 executable_stack);
 	if (retval < 0)
 		goto error;
-#ifdef ARCH_HAS_SETUP_ADDITIONAL_PAGES
+
 	retval = arch_setup_additional_pages(bprm, !!interpreter_name);
 	if (retval < 0)
 		goto error;
-#endif
 #endif
 
 	/* load the executable and interpreter into memory */
fs/compat_binfmt_elf.c +0 -2
index 049ba7c011b9..fad63a4f842e 100644
@@ -111,8 +111,6 @@
 #endif
 
 #ifdef	compat_arch_setup_additional_pages
-#undef	ARCH_HAS_SETUP_ADDITIONAL_PAGES
-#define ARCH_HAS_SETUP_ADDITIONAL_PAGES 1
 #undef	arch_setup_additional_pages
 #define	arch_setup_additional_pages compat_arch_setup_additional_pages
 #endif
include/linux/elf.h +12 -0
index 6dbcfe7a3fd7..95bf7a1abaef 100644
@@ -104,4 +104,16 @@ static inline int arch_elf_adjust_prot(int prot,
 }
 #endif
 
+struct linux_binprm;
+#ifdef CONFIG_ARCH_HAS_SETUP_ADDITIONAL_PAGES
+extern int arch_setup_additional_pages(struct linux_binprm *bprm,
+				       int uses_interp);
+#else
+static inline int arch_setup_additional_pages(struct linux_binprm *bprm,
+				       int uses_interp)
+{
+	return 0;
+}
+#endif
+
 #endif /* _LINUX_ELF_H */
Signature · 1 line
2.31.1
#5 lore raw

PATCH v3 3/23 arm/elf: Remove needless ifdef CONFIG_MMU

In reply to Dmitry Safonov

CONFIG_VDSO depends on MMU, after removing ARCH_HAS_SETUP_ADDITIONAL_PAGES definition, the guards are excessive.

---
 arch/arm/include/asm/elf.h | 2 --
 1 file changed, 2 deletions(-)
arch/arm/include/asm/elf.h +0 -2
index a7cd90b3a779..47347d7412ec 100644
@@ -137,7 +137,6 @@ extern int arm_elf_read_implies_exec(int);
 extern void elf_set_personality(const struct elf32_hdr *);
 #define SET_PERSONALITY(ex)	elf_set_personality(&(ex))
 
-#ifdef CONFIG_MMU
 #ifdef CONFIG_VDSO
 #define ARCH_DLINFO						\
 do {								\
@@ -145,6 +144,5 @@ do {								\
 		    (elf_addr_t)current->mm->context.vdso);	\
 } while (0)
 #endif
-#endif
 
 #endif
Signature · 1 line
2.31.1
#6 lore raw

PATCH v3 4/23 arm64: Use in_compat_task() in arch_setup_additional_pages()

In reply to Dmitry Safonov

Instead of providing compat_arch_setup_additional_pages(), check if the task is compatible from personality, which is set earlier in load_elf_binary(). That will align code with powerpc and sparc, also it'll allow to completely remove compat_arch_setyp_addtional_pages() macro after doing the same for x86, simiplifying the binfmt code in the end and leaving elf loader single function.

---
 arch/arm64/include/asm/elf.h |  5 -----
 arch/arm64/kernel/vdso.c     | 21 ++++++++++-----------
 2 files changed, 10 insertions(+), 16 deletions(-)
arch/arm64/include/asm/elf.h +0 -5
index d1073ffa7f24..a81953bcc1cf 100644
@@ -237,11 +237,6 @@ do {									\
 #else
 #define COMPAT_ARCH_DLINFO
 #endif
-struct linux_binprm;
-extern int aarch32_setup_additional_pages(struct linux_binprm *bprm,
-					  int uses_interp);
-#define compat_arch_setup_additional_pages \
-					aarch32_setup_additional_pages
 
 #endif /* CONFIG_COMPAT */
 
arch/arm64/kernel/vdso.c +10 -11
index a61fc4f989b3..a8bf72320ad0 100644
@@ -411,29 +411,24 @@ static int aarch32_sigreturn_setup(struct mm_struct *mm)
 	return PTR_ERR_OR_ZERO(ret);
 }
 
-int aarch32_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+static int aarch32_setup_additional_pages(struct linux_binprm *bprm,
+					  int uses_interp)
 {
 	struct mm_struct *mm = current->mm;
 	int ret;
 
-	if (mmap_write_lock_killable(mm))
-		return -EINTR;
-
 	ret = aarch32_kuser_helpers_setup(mm);
 	if (ret)
-		goto out;
+		return ret;
 
 	if (IS_ENABLED(CONFIG_COMPAT_VDSO)) {
 		ret = __setup_additional_pages(VDSO_ABI_AA32, mm, bprm,
 					       uses_interp);
 		if (ret)
-			goto out;
+			return ret;
 	}
 
-	ret = aarch32_sigreturn_setup(mm);
-out:
-	mmap_write_unlock(mm);
-	return ret;
+	return aarch32_sigreturn_setup(mm);
 }
 #endif /* CONFIG_COMPAT */
 
@@ -470,7 +465,11 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 	if (mmap_write_lock_killable(mm))
 		return -EINTR;
 
-	ret = __setup_additional_pages(VDSO_ABI_AA64, mm, bprm, uses_interp);
+	if (is_compat_task())
+		ret = aarch32_setup_additional_pages(bprm, uses_interp);
+	else
+		ret = __setup_additional_pages(VDSO_ABI_AA64, mm, bprm, uses_interp);
+
 	mmap_write_unlock(mm);
 
 	return ret;
Signature · 1 line
2.31.1
maintainer Acked-by on v3 4/23
#7 lore raw

Re: [PATCH v3 04/23] arm64: Use in_compat_task() in arch_setup_additional_pages()

Quoted 7 lines · Dmitry Safonov

Instead of providing compat_arch_setup_additional_pages(), check if the task is compatible from personality, which is set earlier in load_elf_binary(). That will align code with powerpc and sparc, also it'll allow to completely remove compat_arch_setyp_addtional_pages() macro after doing the same for x86, simiplifying the binfmt code in the end and leaving elf loader single function.

Quoted 6 lines
---
 arch/arm64/include/asm/elf.h |  5 -----
 arch/arm64/kernel/vdso.c     | 21 ++++++++++-----------
 2 files changed, 10 insertions(+), 16 deletions(-)

Will

#8 lore raw

PATCH v3 5/23 x86: Remove compat_arch_setup_additional_pages()

In reply to Dmitry Safonov

The same as for x32 task, detect ia32 task by in_ia32_syscall(). A process that has just done sys_exec() for compatible ELF is doing compatible syscall after personality is set in load_elf_binary(), see the comment near in_32bit_syscall().

Removing compat_arch_setup_additional_pages() provides single point of entry - arch_setup_additional_pages(), makes ifdeffery easier to read, aligns the code with powerpc and sparc (mips also has single vdso setup function, but instead of taking bitness from mm.context, takes vdso image pointer there).
Together with arm64 code align to use in_compat_task(), it makes possible to remove compat_arch_setup_additional_pages() macro re-definition from compat elf code (another redefined macro less).

---
 arch/x86/entry/vdso/vma.c  | 48 +++++++++++++++-----------------------
 arch/x86/include/asm/elf.h |  5 ----
 2 files changed, 19 insertions(+), 34 deletions(-)
arch/x86/entry/vdso/vma.c +19 -29
index 43d42ce82e86..99415ffb9501 100644
@@ -375,50 +375,40 @@ int map_vdso_once(const struct vdso_image *image, unsigned long addr)
 	return map_vdso(image, addr);
 }
 
-#if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
-static int load_vdso32(void)
-{
-	if (vdso32_enabled != 1)  /* Other values all mean "disabled" */
-		return 0;
-
-	return map_vdso(&vdso_image_32, 0);
-}
-#endif
-
 #ifdef CONFIG_X86_64
-int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+static int load_vdso_64(void)
 {
 	if (!vdso64_enabled)
 		return 0;
 
-	return map_vdso_randomized(&vdso_image_64);
-}
-
-#ifdef CONFIG_COMPAT
-int compat_arch_setup_additional_pages(struct linux_binprm *bprm,
-				       int uses_interp)
-{
 #ifdef CONFIG_X86_X32_ABI
-	if (in_x32_syscall()) {
-		if (!vdso64_enabled)
-			return 0;
+	if (in_x32_syscall())
 		return map_vdso_randomized(&vdso_image_x32);
-	}
 #endif
-#ifdef CONFIG_IA32_EMULATION
-	return load_vdso32();
+
+	return map_vdso_randomized(&vdso_image_64);
+}
 #else
-	return 0;
-#endif
+static int load_vdso_64(void)
+{
+	WARN_ON_ONCE(1);
+	return -ENODATA;
 }
 #endif
-#else
+
 int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 {
-	return load_vdso32();
-}
+#if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
+	if (in_ia32_syscall()) {
+		if (vdso32_enabled != 1)  /* Other values all mean "disabled" */
+			return 0;
+		return map_vdso(&vdso_image_32, 0);
+	}
 #endif
 
+	return load_vdso_64();
+}
+
 bool arch_syscall_is_vdso_sigreturn(struct pt_regs *regs)
 {
 #if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
arch/x86/include/asm/elf.h +0 -5
index 9ee5b3b3ba93..93ff2c7ca4df 100644
@@ -377,11 +377,6 @@ else if (IS_ENABLED(CONFIG_IA32_EMULATION))				\
 	((unsigned long)current->mm->context.vdso +			\
 	 vdso_image_32.sym___kernel_vsyscall)
 
-struct linux_binprm;
-extern int compat_arch_setup_additional_pages(struct linux_binprm *bprm,
-					      int uses_interp);
-#define compat_arch_setup_additional_pages compat_arch_setup_additional_pages
-
 extern bool arch_syscall_is_vdso_sigreturn(struct pt_regs *regs);
 
 /* Do not change the values. See get_align_mask() */
Signature · 1 line
2.31.1
#9 lore raw

PATCH v3 6/23 elf: Remove compat_arch_setup_additional_pages()

In reply to Dmitry Safonov

Now that all users rely on detecting bitness of the task by checking in_compat_syscall(), remove compat_arch_setup_additional_pages() macro, simplifying the code.

---
 fs/compat_binfmt_elf.c | 5 -----
 1 file changed, 5 deletions(-)
fs/compat_binfmt_elf.c +0 -5
index fad63a4f842e..0d1aedc3bbf0 100644
@@ -110,11 +110,6 @@
 #define START_THREAD		COMPAT_START_THREAD
 #endif
 
-#ifdef	compat_arch_setup_additional_pages
-#undef	arch_setup_additional_pages
-#define	arch_setup_additional_pages compat_arch_setup_additional_pages
-#endif
-
 #ifdef	compat_elf_read_implies_exec
 #undef	elf_read_implies_exec
 #define	elf_read_implies_exec compat_elf_read_implies_exec
Signature · 1 line
2.31.1
#10 lore raw

PATCH v3 7/23 vdso: Set mm->context.vdso only on success of _install_special_mapping()

In reply to Dmitry Safonov

Old pattern was:
1. set mm->context.vdso = addr;
2. try install_special_mapping()
3. on failure set mm->context.vdso = 0

The reason behind the old pattern was to make arch_vma_name() work in perf_event_mmap_event(). These days using _install_special_mapping() instead of install_special_mapping() makes old pattern obsolete:
: if (vma->vm_ops && vma->vm_ops->name) {
: name = (char *) vma->vm_ops->name(vma);
: if (name)
: goto cpy_name;

Setting mm->context.vdso = 0 also makes little sense: mm_alloc() zero-fills new mm_struct. And for double-safety if arch_setup_additional_pages() fails, bprm_execve() makes sure that the half-initialized process doesn't make it way to userspace by
: force_sigsegv(SIGSEGV);

Let's cleanup code: set mm->context.vdso only on success, assuming that any new mm_struct is clean.

Some platforms do_munmap() vvar if vdso mapping failed, but it's really necessary only on x86 where vdso/vvar pair can be mapped by userspace (see prctl_map_vdso()). On other platforms vdso/vvar is only pre-mapped by ELF loader, which as described above will make sure to not let any half-baked process out. I've left do_unmap() on !x86 in case prctl() will be supported.

---
 arch/arm/kernel/vdso.c     |  2 --
 arch/arm64/kernel/vdso.c   | 16 +++++-----------
 arch/nds32/kernel/vdso.c   |  5 +----
 arch/powerpc/kernel/vdso.c |  9 ++-------
 arch/sparc/vdso/vma.c      |  7 ++-----
 5 files changed, 10 insertions(+), 29 deletions(-)
arch/arm/kernel/vdso.c +0 -2
index 3408269d19c7..015eff0a6e93 100644
@@ -238,8 +238,6 @@ void arm_install_vdso(struct mm_struct *mm, unsigned long addr)
 	struct vm_area_struct *vma;
 	unsigned long len;
 
-	mm->context.vdso = 0;
-
 	if (vdso_text_pagelist == NULL)
 		return;
 
arch/arm64/kernel/vdso.c +5 -11
index a8bf72320ad0..1bc8adefa293 100644
@@ -227,34 +227,28 @@ static int __setup_additional_pages(enum vdso_abi abi,
 	vdso_mapping_len = vdso_text_len + VVAR_NR_PAGES * PAGE_SIZE;
 
 	vdso_base = get_unmapped_area(NULL, 0, vdso_mapping_len, 0, 0);
-	if (IS_ERR_VALUE(vdso_base)) {
-		ret = ERR_PTR(vdso_base);
-		goto up_fail;
-	}
+	if (IS_ERR_VALUE(vdso_base))
+		return vdso_base;
 
 	ret = _install_special_mapping(mm, vdso_base, VVAR_NR_PAGES * PAGE_SIZE,
 				       VM_READ|VM_MAYREAD|VM_PFNMAP,
 				       vdso_info[abi].dm);
 	if (IS_ERR(ret))
-		goto up_fail;
+		return PTR_ERR(ret);
 
 	if (IS_ENABLED(CONFIG_ARM64_BTI_KERNEL) && system_supports_bti())
 		gp_flags = VM_ARM64_BTI;
 
 	vdso_base += VVAR_NR_PAGES * PAGE_SIZE;
-	mm->context.vdso = (void *)vdso_base;
 	ret = _install_special_mapping(mm, vdso_base, vdso_text_len,
 				       VM_READ|VM_EXEC|gp_flags|
 				       VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
 				       vdso_info[abi].cm);
 	if (IS_ERR(ret))
-		goto up_fail;
+		return PTR_ERR(ret);
 
+	mm->context.vdso = (void *)vdso_base;
 	return 0;
-
-up_fail:
-	mm->context.vdso = NULL;
-	return PTR_ERR(ret);
 }
 
 #ifdef CONFIG_COMPAT
arch/nds32/kernel/vdso.c +1 -4
index e16009a07971..2d1d51a0fc64 100644
@@ -175,7 +175,6 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 
 	/*Map vdso to user space */
 	vdso_base += PAGE_SIZE;
-	mm->context.vdso = (void *)vdso_base;
 	vma = _install_special_mapping(mm, vdso_base, vdso_text_len,
 				       VM_READ | VM_EXEC |
 				       VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC,
@@ -185,11 +184,9 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 		goto up_fail;
 	}
 
-	mmap_write_unlock(mm);
-	return 0;
+	mm->context.vdso = (void *)vdso_base;
 
 up_fail:
-	mm->context.vdso = NULL;
 	mmap_write_unlock(mm);
 	return ret;
 }
arch/powerpc/kernel/vdso.c +2 -7
index 717f2c9a7573..76e898b56002 100644
@@ -229,13 +229,6 @@ static int __arch_setup_additional_pages(struct linux_binprm *bprm, int uses_int
 	/* Add required alignment. */
 	vdso_base = ALIGN(vdso_base, VDSO_ALIGNMENT);
 
-	/*
-	 * Put vDSO base into mm struct. We need to do this before calling
-	 * install_special_mapping or the perf counter mmap tracking code
-	 * will fail to recognise it as a vDSO.
-	 */
-	mm->context.vdso = (void __user *)vdso_base + vvar_size;
-
 	vma = _install_special_mapping(mm, vdso_base, vvar_size,
 				       VM_READ | VM_MAYREAD | VM_IO |
 				       VM_DONTDUMP | VM_PFNMAP, &vvar_spec);
@@ -257,6 +250,8 @@ static int __arch_setup_additional_pages(struct linux_binprm *bprm, int uses_int
 				       VM_MAYWRITE | VM_MAYEXEC, vdso_spec);
 	if (IS_ERR(vma))
 		do_munmap(mm, vdso_base, vvar_size, NULL);
+	else
+		mm->context.vdso = (void __user *)vdso_base + vvar_size;
 
 	return PTR_ERR_OR_ZERO(vma);
 }
arch/sparc/vdso/vma.c +2 -5
index cc19e09b0fa1..d8a344f6c914 100644
@@ -390,7 +390,6 @@ static int map_vdso(const struct vdso_image *image,
 	}
 
 	text_start = addr - image->sym_vvar_start;
-	current->mm->context.vdso = (void __user *)text_start;
 
 	/*
 	 * MAYWRITE to allow gdb to COW and set breakpoints
@@ -412,16 +411,14 @@ static int map_vdso(const struct vdso_image *image,
 				       -image->sym_vvar_start,
 				       VM_READ|VM_MAYREAD,
 				       &vvar_mapping);
-
 	if (IS_ERR(vma)) {
 		ret = PTR_ERR(vma);
 		do_munmap(mm, text_start, image->size, NULL);
+	} else {
+		current->mm->context.vdso = (void __user *)text_start;
 	}
 
 up_fail:
-	if (ret)
-		current->mm->context.vdso = NULL;
-
 	mmap_write_unlock(mm);
 	return ret;
 }
Signature · 1 line
2.31.1
#11 lore raw

PATCH v3 8/23 elf/vdso: Modify arch_setup_additional_pages() parameters

In reply to Dmitry Safonov

Both parameters of arch_setup_additional_pages() are currently unused. commit fc5243d98ac2 ("[S390] arch_setup_additional_pages arguments") tried to introduce useful arguments, but they still are not used.

Remove old parameters and introduce sysinfo_ehdr argument that will be used to return vdso address to put as AT_SYSINFO_EHDR tag in auxiliary vector. The reason to add this parameter is that many architectures have vDSO pointer saved in their mm->context with the only purpose to use it later in ARCH_DLINFO. That's the macro for elf loader to setup sysinfo_ehdr tag.

Return sysinfo_ehdr address that will be later used by ARCH_DLINFO as an argument. That will allow to drop vDSO pointer from mm->context and any code responsible to track vDSO position on platforms that don't use vDSO as a landing in userspace (arm/s390/sparc).

Cc Albert Ou <[email protected]>
Cc "David S. Miller" <[email protected]>
Cc Palmer Dabbelt <[email protected]>
Cc Paul Walmsley <[email protected]>
---
 arch/arm/include/asm/vdso.h        |  6 ++++--
 arch/arm/kernel/process.c          |  4 ++--
 arch/arm/kernel/vdso.c             | 10 +++++++---
 arch/arm64/kernel/vdso.c           | 17 ++++++++---------
 arch/csky/kernel/vdso.c            | 11 ++++++-----
 arch/hexagon/kernel/vdso.c         |  3 ++-
 arch/mips/kernel/vdso.c            |  3 ++-
 arch/nds32/kernel/vdso.c           |  3 ++-
 arch/nios2/mm/init.c               |  2 +-
 arch/powerpc/kernel/vdso.c         | 12 +++++++-----
 arch/riscv/kernel/vdso.c           | 11 ++++++-----
 arch/s390/kernel/vdso.c            |  3 ++-
 arch/sh/kernel/vsyscall/vsyscall.c |  3 ++-
 arch/sparc/vdso/vma.c              | 14 +++++++-------
 arch/x86/entry/vdso/vma.c          | 26 +++++++++++++++-----------
 arch/x86/um/vdso/vma.c             |  2 +-
 fs/binfmt_elf.c                    |  3 ++-
 fs/binfmt_elf_fdpic.c              |  3 ++-
 include/linux/elf.h                | 17 ++++++++++++-----
 19 files changed, 90 insertions(+), 63 deletions(-)
arch/arm/include/asm/vdso.h +4 -2
index 5b85889f82ee..6b2b3b1fe833 100644
@@ -10,13 +10,15 @@ struct mm_struct;
 
 #ifdef CONFIG_VDSO
 
-void arm_install_vdso(struct mm_struct *mm, unsigned long addr);
+void arm_install_vdso(struct mm_struct *mm, unsigned long addr,
+		      unsigned long *sysinfo_ehdr);
 
 extern unsigned int vdso_total_pages;
 
 #else /* CONFIG_VDSO */
 
-static inline void arm_install_vdso(struct mm_struct *mm, unsigned long addr)
+static inline void arm_install_vdso(struct mm_struct *mm, unsigned long addr,
+				    unsigned long *sysinfo_ehdr)
 {
 }
 
arch/arm/kernel/process.c +2 -2
index 6324f4db9b02..5897ccb88bca 100644
@@ -400,7 +400,7 @@ static const struct vm_special_mapping sigpage_mapping = {
 	.mremap = sigpage_mremap,
 };
 
-int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	struct mm_struct *mm = current->mm;
 	struct vm_area_struct *vma;
@@ -441,7 +441,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 	 * to be fatal to the process, so no error check needed
 	 * here.
 	 */
-	arm_install_vdso(mm, addr + PAGE_SIZE);
+	arm_install_vdso(mm, addr + PAGE_SIZE, sysinfo_ehdr);
 
  up_fail:
 	mmap_write_unlock(mm);
arch/arm/kernel/vdso.c +7 -3
index 015eff0a6e93..de516f8ba85a 100644
@@ -233,7 +233,8 @@ static int install_vvar(struct mm_struct *mm, unsigned long addr)
 }
 
 /* assumes mmap_lock is write-locked */
-void arm_install_vdso(struct mm_struct *mm, unsigned long addr)
+void arm_install_vdso(struct mm_struct *mm, unsigned long addr,
+		      unsigned long *sysinfo_ehdr)
 {
 	struct vm_area_struct *vma;
 	unsigned long len;
@@ -252,7 +253,10 @@ void arm_install_vdso(struct mm_struct *mm, unsigned long addr)
 		VM_READ | VM_EXEC | VM_MAYREAD | VM_MAYWRITE | VM_MAYEXEC,
 		&vdso_text_mapping);
 
-	if (!IS_ERR(vma))
-		mm->context.vdso = addr;
+	if (IS_ERR(vma))
+		return;
+
+	mm->context.vdso = addr;
+	*sysinfo_ehdr = addr;
 }
 
arch/arm64/kernel/vdso.c +8 -9
index 1bc8adefa293..c0512c2e8183 100644
@@ -213,8 +213,7 @@ static vm_fault_t vvar_fault(const struct vm_special_mapping *sm,
 
 static int __setup_additional_pages(enum vdso_abi abi,
 				    struct mm_struct *mm,
-				    struct linux_binprm *bprm,
-				    int uses_interp)
+				    unsigned long *sysinfo_ehdr)
 {
 	unsigned long vdso_base, vdso_text_len, vdso_mapping_len;
 	unsigned long gp_flags = 0;
@@ -248,6 +247,8 @@ static int __setup_additional_pages(enum vdso_abi abi,
 		return PTR_ERR(ret);
 
 	mm->context.vdso = (void *)vdso_base;
+	*sysinfo_ehdr = vdso_base;
+
 	return 0;
 }
 
@@ -405,8 +406,7 @@ static int aarch32_sigreturn_setup(struct mm_struct *mm)
 	return PTR_ERR_OR_ZERO(ret);
 }
 
-static int aarch32_setup_additional_pages(struct linux_binprm *bprm,
-					  int uses_interp)
+static int aarch32_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	struct mm_struct *mm = current->mm;
 	int ret;
@@ -416,8 +416,7 @@ static int aarch32_setup_additional_pages(struct linux_binprm *bprm,
 		return ret;
 
 	if (IS_ENABLED(CONFIG_COMPAT_VDSO)) {
-		ret = __setup_additional_pages(VDSO_ABI_AA32, mm, bprm,
-					       uses_interp);
+		ret = __setup_additional_pages(VDSO_ABI_AA32, mm, sysinfo_ehdr);
 		if (ret)
 			return ret;
 	}
@@ -451,7 +450,7 @@ static int __init vdso_init(void)
 }
 arch_initcall(vdso_init);
 
-int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	struct mm_struct *mm = current->mm;
 	int ret;
@@ -460,9 +459,9 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 		return -EINTR;
 
 	if (is_compat_task())
-		ret = aarch32_setup_additional_pages(bprm, uses_interp);
+		ret = aarch32_setup_additional_pages(sysinfo_ehdr);
 	else
-		ret = __setup_additional_pages(VDSO_ABI_AA64, mm, bprm, uses_interp);
+		ret = __setup_additional_pages(VDSO_ABI_AA64, mm, sysinfo_ehdr);
 
 	mmap_write_unlock(mm);
 
arch/csky/kernel/vdso.c +6 -5
index 16c20d64d165..30160e64ee2d 100644
@@ -52,11 +52,10 @@ static int __init vdso_init(void)
 }
 arch_initcall(vdso_init);
 
-int arch_setup_additional_pages(struct linux_binprm *bprm,
-	int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	struct mm_struct *mm = current->mm;
-	unsigned long vdso_base, vdso_len;
+	unsigned long vdso_base, vvar_base, vdso_len;
 	int ret;
 
 	vdso_len = (vdso_pages + 1) << PAGE_SHIFT;
@@ -85,12 +84,14 @@ int arch_setup_additional_pages(struct linux_binprm *bprm,
 		goto end;
 	}
 
-	vdso_base += (vdso_pages << PAGE_SHIFT);
-	ret = install_special_mapping(mm, vdso_base, PAGE_SIZE,
+	vvar_base = vdso_base + (vdso_pages << PAGE_SHIFT);
+	ret = install_special_mapping(mm, vvar_base, PAGE_SIZE,
 		(VM_READ | VM_MAYREAD), &vdso_pagelist[vdso_pages]);
 
 	if (unlikely(ret))
 		mm->context.vdso = NULL;
+	else
+		*sysinfo_ehdr = vdso_base;
 end:
 	mmap_write_unlock(mm);
 	return ret;
arch/hexagon/kernel/vdso.c +2 -1
index b70970ac809f..39e78fe82b99 100644
@@ -46,7 +46,7 @@ arch_initcall(vdso_init);
 /*
  * Called from binfmt_elf.  Create a VMA for the vDSO page.
  */
-int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	int ret;
 	unsigned long vdso_base;
@@ -74,6 +74,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 		goto up_fail;
 
 	mm->context.vdso = (void *)vdso_base;
+	*sysinfo_ehdr = vdso_base;
 
 up_fail:
 	mmap_write_unlock(mm);
arch/mips/kernel/vdso.c +2 -1
index 3d0cf471f2fe..9b2e1d2250b4 100644
@@ -86,7 +86,7 @@ static unsigned long vdso_base(void)
 	return base;
 }
 
-int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	struct mips_vdso_image *image = current->thread.abi->vdso;
 	struct mm_struct *mm = current->mm;
@@ -185,6 +185,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 	}
 
 	mm->context.vdso = (void *)vdso_addr;
+	*sysinfo_ehdr = vdso_addr;
 	ret = 0;
 
 out:
arch/nds32/kernel/vdso.c +2 -1
index 2d1d51a0fc64..1d35a33389e5 100644
@@ -111,7 +111,7 @@ unsigned long inline vdso_random_addr(unsigned long vdso_mapping_len)
 	return addr;
 }
 
-int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	struct mm_struct *mm = current->mm;
 	unsigned long vdso_base, vdso_text_len, vdso_mapping_len;
@@ -185,6 +185,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 	}
 
 	mm->context.vdso = (void *)vdso_base;
+	*sysinfo_ehdr = vdso_base;
 
 up_fail:
 	mmap_write_unlock(mm);
arch/nios2/mm/init.c +1 -1
index 613fcaa5988a..0164f5cdb255 100644
@@ -103,7 +103,7 @@ static int alloc_kuser_page(void)
 }
 arch_initcall(alloc_kuser_page);
 
-int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	struct mm_struct *mm = current->mm;
 	int ret;
arch/powerpc/kernel/vdso.c +7 -5
index 76e898b56002..6d6e575630c1 100644
@@ -190,7 +190,7 @@ static vm_fault_t vvar_fault(const struct vm_special_mapping *sm,
  * This is called from binfmt_elf, we create the special vma for the
  * vDSO and insert it into the mm struct tree
  */
-static int __arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+static int __arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	unsigned long vdso_size, vdso_base, mappings_size;
 	struct vm_special_mapping *vdso_spec;
@@ -248,15 +248,17 @@ static int __arch_setup_additional_pages(struct linux_binprm *bprm, int uses_int
 	vma = _install_special_mapping(mm, vdso_base + vvar_size, vdso_size,
 				       VM_READ | VM_EXEC | VM_MAYREAD |
 				       VM_MAYWRITE | VM_MAYEXEC, vdso_spec);
-	if (IS_ERR(vma))
+	if (IS_ERR(vma)) {
 		do_munmap(mm, vdso_base, vvar_size, NULL);
-	else
+	} else {
 		mm->context.vdso = (void __user *)vdso_base + vvar_size;
+		*sysinfo_ehdr = vdso_base + vvar_size;
+	}
 
 	return PTR_ERR_OR_ZERO(vma);
 }
 
-int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	struct mm_struct *mm = current->mm;
 	int rc;
@@ -266,7 +268,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 	if (mmap_write_lock_killable(mm))
 		return -EINTR;
 
-	rc = __arch_setup_additional_pages(bprm, uses_interp);
+	rc = __arch_setup_additional_pages(sysinfo_ehdr);
 	if (rc)
 		mm->context.vdso = NULL;
 
arch/riscv/kernel/vdso.c +6 -5
index 25a3b8849599..9cbbad8e48da 100644
@@ -56,11 +56,10 @@ static int __init vdso_init(void)
 }
 arch_initcall(vdso_init);
 
-int arch_setup_additional_pages(struct linux_binprm *bprm,
-	int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	struct mm_struct *mm = current->mm;
-	unsigned long vdso_base, vdso_len;
+	unsigned long vdso_base, vvar_base, vdso_len;
 	int ret;
 
 	vdso_len = (vdso_pages + 1) << PAGE_SHIFT;
@@ -89,12 +88,14 @@ int arch_setup_additional_pages(struct linux_binprm *bprm,
 		goto end;
 	}
 
-	vdso_base += (vdso_pages << PAGE_SHIFT);
-	ret = install_special_mapping(mm, vdso_base, PAGE_SIZE,
+	vvar_base = vdso_base + (vdso_pages << PAGE_SHIFT);
+	ret = install_special_mapping(mm, vvar_base, PAGE_SIZE,
 		(VM_READ | VM_MAYREAD), &vdso_pagelist[vdso_pages]);
 
 	if (unlikely(ret))
 		mm->context.vdso = NULL;
+	else
+		*sysinfo_ehdr = vdso_base;
 end:
 	mmap_write_unlock(mm);
 	return ret;
arch/s390/kernel/vdso.c +2 -1
index 8c4e07d533c8..8a72fdedbae9 100644
@@ -167,7 +167,7 @@ int vdso_getcpu_init(void)
 }
 early_initcall(vdso_getcpu_init); /* Must be called before SMP init */
 
-int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	unsigned long vdso_text_len, vdso_mapping_len;
 	unsigned long vvar_start, vdso_text_start;
@@ -204,6 +204,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 		rc = PTR_ERR(vma);
 	} else {
 		current->mm->context.vdso_base = vdso_text_start;
+		*sysinfo_ehdr = vdso_text_start;
 		rc = 0;
 	}
 out:
arch/sh/kernel/vsyscall/vsyscall.c +2 -1
index 1bd85a6949c4..de8df3261b4f 100644
@@ -55,7 +55,7 @@ int __init vsyscall_init(void)
 }
 
 /* Setup a VMA at program startup for the vsyscall page */
-int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	struct mm_struct *mm = current->mm;
 	unsigned long addr;
@@ -78,6 +78,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 		goto up_fail;
 
 	current->mm->context.vdso = (void *)addr;
+	*sysinfo_ehdr = addr;
 
 up_fail:
 	mmap_write_unlock(mm);
arch/sparc/vdso/vma.c +7 -7
index d8a344f6c914..ae635893f9b3 100644
@@ -346,8 +346,6 @@ static int __init init_vdso(void)
 }
 subsys_initcall(init_vdso);
 
-struct linux_binprm;
-
 /* Shuffle the vdso up a bit, randomly. */
 static unsigned long vdso_addr(unsigned long start, unsigned int len)
 {
@@ -359,7 +357,8 @@ static unsigned long vdso_addr(unsigned long start, unsigned int len)
 }
 
 static int map_vdso(const struct vdso_image *image,
-		struct vm_special_mapping *vdso_mapping)
+		    struct vm_special_mapping *vdso_mapping,
+		    unsigned long *sysinfo_ehdr)
 {
 	struct mm_struct *mm = current->mm;
 	struct vm_area_struct *vma;
@@ -416,6 +415,7 @@ static int map_vdso(const struct vdso_image *image,
 		do_munmap(mm, text_start, image->size, NULL);
 	} else {
 		current->mm->context.vdso = (void __user *)text_start;
+		*sysinfo_ehdr = text_start;
 	}
 
 up_fail:
@@ -423,7 +423,7 @@ static int map_vdso(const struct vdso_image *image,
 	return ret;
 }
 
-int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 
 	if (!vdso_enabled)
@@ -431,11 +431,11 @@ int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 
 #if defined CONFIG_COMPAT
 	if (!(is_32bit_task()))
-		return map_vdso(&vdso_image_64_builtin, &vdso_mapping64);
+		return map_vdso(&vdso_image_64_builtin, &vdso_mapping64, sysinfo_ehdr);
 	else
-		return map_vdso(&vdso_image_32_builtin, &vdso_mapping32);
+		return map_vdso(&vdso_image_32_builtin, &vdso_mapping32, sysinfo_ehdr);
 #else
-	return map_vdso(&vdso_image_64_builtin, &vdso_mapping64);
+	return map_vdso(&vdso_image_64_builtin, &vdso_mapping64, sysinfo_ehdr);
 #endif
 
 }
arch/x86/entry/vdso/vma.c +15 -11
index 99415ffb9501..f1abe43aadb9 100644
@@ -243,7 +243,8 @@ static const struct vm_special_mapping vvar_mapping = {
  * @image          - blob to map
  * @addr           - request a specific address (zero to map at free addr)
  */
-static int map_vdso(const struct vdso_image *image, unsigned long addr)
+static int map_vdso(const struct vdso_image *image, unsigned long addr,
+		    unsigned long *sysinfo_ehdr)
 {
 	struct mm_struct *mm = current->mm;
 	struct vm_area_struct *vma;
@@ -290,6 +291,7 @@ static int map_vdso(const struct vdso_image *image, unsigned long addr)
 	} else {
 		current->mm->context.vdso = (void __user *)text_start;
 		current->mm->context.vdso_image = image;
+		*sysinfo_ehdr = text_start;
 	}
 
 up_fail:
@@ -342,11 +344,12 @@ static unsigned long vdso_addr(unsigned long start, unsigned len)
 	return addr;
 }
 
-static int map_vdso_randomized(const struct vdso_image *image)
+static int map_vdso_randomized(const struct vdso_image *image,
+			       unsigned long *sysinfo_ehdr)
 {
 	unsigned long addr = vdso_addr(current->mm->start_stack, image->size-image->sym_vvar_start);
 
-	return map_vdso(image, addr);
+	return map_vdso(image, addr, sysinfo_ehdr);
 }
 #endif
 
@@ -354,6 +357,7 @@ int map_vdso_once(const struct vdso_image *image, unsigned long addr)
 {
 	struct mm_struct *mm = current->mm;
 	struct vm_area_struct *vma;
+	unsigned long unused;
 
 	mmap_write_lock(mm);
 	/*
@@ -372,41 +376,41 @@ int map_vdso_once(const struct vdso_image *image, unsigned long addr)
 	}
 	mmap_write_unlock(mm);
 
-	return map_vdso(image, addr);
+	return map_vdso(image, addr, &unused);
 }
 
 #ifdef CONFIG_X86_64
-static int load_vdso_64(void)
+static int load_vdso_64(unsigned long *sysinfo_ehdr)
 {
 	if (!vdso64_enabled)
 		return 0;
 
 #ifdef CONFIG_X86_X32_ABI
 	if (in_x32_syscall())
-		return map_vdso_randomized(&vdso_image_x32);
+		return map_vdso_randomized(&vdso_image_x32, sysinfo_ehdr);
 #endif
 
-	return map_vdso_randomized(&vdso_image_64);
+	return map_vdso_randomized(&vdso_image_64, sysinfo_ehdr);
 }
 #else
-static int load_vdso_64(void)
+static int load_vdso_64(unsigned long *sysinfo_ehdr)
 {
 	WARN_ON_ONCE(1);
 	return -ENODATA;
 }
 #endif
 
-int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 #if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
 	if (in_ia32_syscall()) {
 		if (vdso32_enabled != 1)  /* Other values all mean "disabled" */
 			return 0;
-		return map_vdso(&vdso_image_32, 0);
+		return map_vdso(&vdso_image_32, 0, sysinfo_ehdr);
 	}
 #endif
 
-	return load_vdso_64();
+	return load_vdso_64(sysinfo_ehdr);
 }
 
 bool arch_syscall_is_vdso_sigreturn(struct pt_regs *regs)
arch/x86/um/vdso/vma.c +1 -1
index 76d9f6ce7a3d..77488065f7cc 100644
@@ -50,7 +50,7 @@ static int __init init_vdso(void)
 }
 subsys_initcall(init_vdso);
 
-int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	int err;
 	struct mm_struct *mm = current->mm;
fs/binfmt_elf.c +2 -1
index dac2713c10ee..62741e55e3d1 100644
@@ -836,6 +836,7 @@ static int load_elf_binary(struct linux_binprm *bprm)
 	unsigned long interp_load_addr = 0;
 	unsigned long start_code, end_code, start_data, end_data;
 	unsigned long reloc_func_desc __maybe_unused = 0;
+	unsigned long sysinfo_ehdr = 0;
 	int executable_stack = EXSTACK_DEFAULT;
 	struct elfhdr *elf_ex = (struct elfhdr *)bprm->buf;
 	struct elfhdr *interp_elf_ex = NULL;
@@ -1252,7 +1253,7 @@ static int load_elf_binary(struct linux_binprm *bprm)
 
 	set_binfmt(&elf_format);
 
-	retval = arch_setup_additional_pages(bprm, !!interpreter);
+	retval = arch_setup_additional_pages(&sysinfo_ehdr);
 	if (retval < 0)
 		goto out;
 
fs/binfmt_elf_fdpic.c +2 -1
index 11cbf20b19da..421a09bc6ee6 100644
@@ -183,6 +183,7 @@ static int load_elf_fdpic_binary(struct linux_binprm *bprm)
 {
 	struct elf_fdpic_params exec_params, interp_params;
 	struct pt_regs *regs = current_pt_regs();
+	unsigned long sysinfo_ehdr = 0;
 	struct elf_phdr *phdr;
 	unsigned long stack_size, entryaddr;
 #ifdef ELF_FDPIC_PLAT_INIT
@@ -375,7 +376,7 @@ static int load_elf_fdpic_binary(struct linux_binprm *bprm)
 	if (retval < 0)
 		goto error;
 
-	retval = arch_setup_additional_pages(bprm, !!interpreter_name);
+	retval = arch_setup_additional_pages(&sysinfo_ehdr);
 	if (retval < 0)
 		goto error;
 #endif
include/linux/elf.h +12 -5
index 95bf7a1abaef..a8bea5611a4b 100644
@@ -104,13 +104,20 @@ static inline int arch_elf_adjust_prot(int prot,
 }
 #endif
 
-struct linux_binprm;
 #ifdef CONFIG_ARCH_HAS_SETUP_ADDITIONAL_PAGES
-extern int arch_setup_additional_pages(struct linux_binprm *bprm,
-				       int uses_interp);
+/**
+ * arch_setup_additional_pages - Premap VMAs in a new-execed process
+ * @sysinfo_ehdr:	Returns vDSO position to be set in the initial
+ *			auxiliary vector (tag AT_SYSINFO_EHDR) by binfmt
+ *			loader. On failure isn't initialized.
+ *			As address == 0 is never used, it allows to check
+ *			if the tag should be set.
+ *
+ * Return: Zero if successful, or a negative error code on failure.
+ */
+extern int arch_setup_additional_pages(unsigned long *sysinfo_ehdr);
 #else
-static inline int arch_setup_additional_pages(struct linux_binprm *bprm,
-				       int uses_interp)
+static inline int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 {
 	return 0;
 }
Signature · 1 line
2.31.1
#12 lore raw

PATCH v3 9/23 elf: Use sysinfo_ehdr in ARCH_DLINFO()

In reply to Dmitry Safonov

Instead mm->context.vdso use the pointer provided by elf loader. That allows to drop the pointer on arm/s390/sparc.

Cc Christian Borntraeger <[email protected]>
Cc Heiko Carstens <[email protected]>
Cc Vasily Gorbik <[email protected]>
Cc "David S. Miller" <[email protected]>
---
 arch/alpha/include/asm/elf.h    |  2 +-
 arch/arm/include/asm/elf.h      |  5 ++---
 arch/arm64/include/asm/elf.h    | 18 +++++-------------
 arch/ia64/include/asm/elf.h     |  2 +-
 arch/mips/include/asm/elf.h     |  5 ++---
 arch/nds32/include/asm/elf.h    |  5 ++---
 arch/powerpc/include/asm/elf.h  |  4 ++--
 arch/riscv/include/asm/elf.h    |  5 ++---
 arch/s390/include/asm/elf.h     |  5 ++---
 arch/sh/include/asm/elf.h       | 10 +++++-----
 arch/sparc/include/asm/elf_64.h |  5 ++---
 arch/x86/include/asm/elf.h      | 33 ++++++++++++++-------------------
 arch/x86/um/asm/elf.h           |  4 ++--
 fs/binfmt_elf.c                 |  6 +++---
 fs/binfmt_elf_fdpic.c           | 11 ++++++-----
 15 files changed, 51 insertions(+), 69 deletions(-)
arch/alpha/include/asm/elf.h +1 -1
index 8049997fa372..701e820f28f0 100644
@@ -155,7 +155,7 @@ extern int alpha_l2_cacheshape;
 extern int alpha_l3_cacheshape;
 
 /* update AT_VECTOR_SIZE_ARCH if the number of NEW_AUX_ENT entries changes */
-#define ARCH_DLINFO						\
+#define ARCH_DLINFO(sysinfo_ehdr)				\
   do {								\
     NEW_AUX_ENT(AT_L1I_CACHESHAPE, alpha_l1i_cacheshape);	\
     NEW_AUX_ENT(AT_L1D_CACHESHAPE, alpha_l1d_cacheshape);	\
arch/arm/include/asm/elf.h +2 -3
index 47347d7412ec..76a0f04190f0 100644
@@ -138,10 +138,9 @@ extern void elf_set_personality(const struct elf32_hdr *);
 #define SET_PERSONALITY(ex)	elf_set_personality(&(ex))
 
 #ifdef CONFIG_VDSO
-#define ARCH_DLINFO						\
+#define ARCH_DLINFO(sysinfo_ehdr)				\
 do {								\
-	NEW_AUX_ENT(AT_SYSINFO_EHDR,				\
-		    (elf_addr_t)current->mm->context.vdso);	\
+	NEW_AUX_ENT(AT_SYSINFO_EHDR, sysinfo_ehdr);		\
 } while (0)
 #endif
 
arch/arm64/include/asm/elf.h +5 -13
index a81953bcc1cf..e62818967a69 100644
@@ -165,10 +165,9 @@ typedef struct user_fpsimd_state elf_fpregset_t;
 })
 
 /* update AT_VECTOR_SIZE_ARCH if the number of NEW_AUX_ENT entries changes */
-#define ARCH_DLINFO							\
+#define ARCH_DLINFO(sysinfo_ehdr)					\
 do {									\
-	NEW_AUX_ENT(AT_SYSINFO_EHDR,					\
-		    (elf_addr_t)current->mm->context.vdso);		\
+	NEW_AUX_ENT(AT_SYSINFO_EHDR, sysinfo_ehdr);			\
 									\
 	/*								\
 	 * Should always be nonzero unless there's a kernel bug.	\
@@ -223,19 +222,12 @@ typedef compat_elf_greg_t		compat_elf_gregset_t[COMPAT_ELF_NGREG];
 	set_thread_flag(TIF_32BIT);					\
  })
 #ifdef CONFIG_COMPAT_VDSO
-#define COMPAT_ARCH_DLINFO						\
+#define COMPAT_ARCH_DLINFO(sysinfo_ehdr)				\
 do {									\
-	/*								\
-	 * Note that we use Elf64_Off instead of elf_addr_t because	\
-	 * elf_addr_t in compat is defined as Elf32_Addr and casting	\
-	 * current->mm->context.vdso to it triggers a cast warning of	\
-	 * cast from pointer to integer of different size.		\
-	 */								\
-	NEW_AUX_ENT(AT_SYSINFO_EHDR,					\
-			(Elf64_Off)current->mm->context.vdso);		\
+	NEW_AUX_ENT(AT_SYSINFO_EHDR, sysinfo_ehdr);			\
 } while (0)
 #else
-#define COMPAT_ARCH_DLINFO
+#define COMPAT_ARCH_DLINFO(sysinfo_ehdr)
 #endif
 
 #endif /* CONFIG_COMPAT */
arch/ia64/include/asm/elf.h +1 -1
index 6629301a2620..a257e5abddce 100644
@@ -208,7 +208,7 @@ struct task_struct;
 #define GATE_EHDR	((const struct elfhdr *) GATE_ADDR)
 
 /* update AT_VECTOR_SIZE_ARCH if the number of NEW_AUX_ENT entries changes */
-#define ARCH_DLINFO								\
+#define ARCH_DLINFO(sysinfo_ehdr)						\
 do {										\
 	extern char __kernel_syscall_via_epc[];					\
 	NEW_AUX_ENT(AT_SYSINFO, (unsigned long) __kernel_syscall_via_epc);	\
arch/mips/include/asm/elf.h +2 -3
index a5c8be47a39d..672a32fa59d9 100644
@@ -456,10 +456,9 @@ extern const char *__elf_base_platform;
 #define ELF_ET_DYN_BASE		(TASK_SIZE / 3 * 2)
 
 /* update AT_VECTOR_SIZE_ARCH if the number of NEW_AUX_ENT entries changes */
-#define ARCH_DLINFO							\
+#define ARCH_DLINFO(sysinfo_ehdr)					\
 do {									\
-	NEW_AUX_ENT(AT_SYSINFO_EHDR,					\
-		    (unsigned long)current->mm->context.vdso);		\
+	NEW_AUX_ENT(AT_SYSINFO_EHDR, sysinfo_ehdr);			\
 } while (0)
 
 #ifdef CONFIG_MIPS_FP_SUPPORT
arch/nds32/include/asm/elf.h +2 -3
index 36cec4ae5a84..4f5894208efe 100644
@@ -165,13 +165,12 @@ struct elf32_hdr;
 #define FPU_AUX_ENT	NEW_AUX_ENT(AT_IGNORE, 0)
 #endif
 
-#define ARCH_DLINFO						\
+#define ARCH_DLINFO(sysinfo_ehdr)				\
 do {								\
 	/* Optional FPU initialization */			\
 	FPU_AUX_ENT;						\
 								\
-	NEW_AUX_ENT(AT_SYSINFO_EHDR,				\
-		    (elf_addr_t)current->mm->context.vdso);	\
+	NEW_AUX_ENT(AT_SYSINFO_EHDR, sysinfo_ehdr);		\
 } while (0)
 
 #endif
arch/powerpc/include/asm/elf.h +2 -2
index d7d9820c9096..71b180d6ed90 100644
@@ -155,7 +155,7 @@ extern int ucache_bsize;
  *   even if DLINFO_ARCH_ITEMS goes to zero or is undefined.
  * update AT_VECTOR_SIZE_ARCH if the number of NEW_AUX_ENT entries changes
  */
-#define ARCH_DLINFO							\
+#define ARCH_DLINFO(sysinfo_ehdr)					\
 do {									\
 	/* Handle glibc compatibility. */				\
 	NEW_AUX_ENT(AT_IGNOREPPC, AT_IGNOREPPC);			\
@@ -164,7 +164,7 @@ do {									\
 	NEW_AUX_ENT(AT_DCACHEBSIZE, dcache_bsize);			\
 	NEW_AUX_ENT(AT_ICACHEBSIZE, icache_bsize);			\
 	NEW_AUX_ENT(AT_UCACHEBSIZE, 0);					\
-	VDSO_AUX_ENT(AT_SYSINFO_EHDR, (unsigned long)current->mm->context.vdso);\
+	VDSO_AUX_ENT(AT_SYSINFO_EHDR, sysinfo_ehdr);			\
 	ARCH_DLINFO_CACHE_GEOMETRY;					\
 } while (0)
 
arch/riscv/include/asm/elf.h +2 -3
index 1d1d60df632e..7c56700f857d 100644
@@ -58,10 +58,9 @@ extern unsigned long elf_hwcap;
 #define ELF_PLATFORM	(NULL)
 
 #ifdef CONFIG_MMU
-#define ARCH_DLINFO						\
+#define ARCH_DLINFO(sysinfo_ehdr)				\
 do {								\
-	NEW_AUX_ENT(AT_SYSINFO_EHDR,				\
-		(elf_addr_t)current->mm->context.vdso);		\
+	NEW_AUX_ENT(AT_SYSINFO_EHDR, sysinfo_ehdr);		\
 	NEW_AUX_ENT(AT_L1I_CACHESIZE,				\
 		get_cache_size(1, CACHE_TYPE_INST));		\
 	NEW_AUX_ENT(AT_L1I_CACHEGEOMETRY,			\
arch/s390/include/asm/elf.h +2 -3
index 6583142149b0..c8026e3e5f10 100644
@@ -268,11 +268,10 @@ do {								\
 #define STACK_RND_MASK	MMAP_RND_MASK
 
 /* update AT_VECTOR_SIZE_ARCH if the number of NEW_AUX_ENT entries changes */
-#define ARCH_DLINFO							    \
+#define ARCH_DLINFO(sysinfo_ehdr)					    \
 do {									    \
 	if (vdso_enabled)						    \
-		NEW_AUX_ENT(AT_SYSINFO_EHDR,				    \
-			    (unsigned long)current->mm->context.vdso_base); \
+		NEW_AUX_ENT(AT_SYSINFO_EHDR, sysinfo_ehdr);		    \
 } while (0)
 
 #endif
arch/sh/include/asm/elf.h +5 -5
index 9b3e22e771a1..03b813c0bc39 100644
@@ -170,13 +170,13 @@ extern void __kernel_vsyscall;
 #define VDSO_BASE		((unsigned long)current->mm->context.vdso)
 #define VDSO_SYM(x)		(VDSO_BASE + (unsigned long)(x))
 
-#define VSYSCALL_AUX_ENT					\
+#define VSYSCALL_AUX_ENT(sysinfo_ehdr)				\
 	if (vdso_enabled)					\
-		NEW_AUX_ENT(AT_SYSINFO_EHDR, VDSO_BASE);	\
+		NEW_AUX_ENT(AT_SYSINFO_EHDR, sysinfo_ehdr);	\
 	else							\
 		NEW_AUX_ENT(AT_IGNORE, 0)
 #else
-#define VSYSCALL_AUX_ENT	NEW_AUX_ENT(AT_IGNORE, 0)
+#define VSYSCALL_AUX_ENT(sysinfo_ehdr)	NEW_AUX_ENT(AT_IGNORE, 0)
 #endif /* CONFIG_VSYSCALL */
 
 #ifdef CONFIG_SH_FPU
@@ -188,13 +188,13 @@ extern void __kernel_vsyscall;
 extern int l1i_cache_shape, l1d_cache_shape, l2_cache_shape;
 
 /* update AT_VECTOR_SIZE_ARCH if the number of NEW_AUX_ENT entries changes */
-#define ARCH_DLINFO						\
+#define ARCH_DLINFO(sysinfo_ehdr)				\
 do {								\
 	/* Optional FPU initialization */			\
 	FPU_AUX_ENT;						\
 								\
 	/* Optional vsyscall entry */				\
-	VSYSCALL_AUX_ENT;					\
+	VSYSCALL_AUX_ENT(sysinfo_ehdr);				\
 								\
 	/* Cache desc */					\
 	NEW_AUX_ENT(AT_L1I_CACHESHAPE, l1i_cache_shape);	\
arch/sparc/include/asm/elf_64.h +2 -3
index 63a622c36df3..1e7295b5ae2f 100644
@@ -213,12 +213,11 @@ do {	if ((ex).e_ident[EI_CLASS] == ELFCLASS32)	\
 
 extern unsigned int vdso_enabled;
 
-#define	ARCH_DLINFO							\
+#define	ARCH_DLINFO(sysinfo_ehdr)					\
 do {									\
 	extern struct adi_config adi_state;				\
 	if (vdso_enabled)						\
-		NEW_AUX_ENT(AT_SYSINFO_EHDR,				\
-			    (unsigned long)current->mm->context.vdso);	\
+		NEW_AUX_ENT(AT_SYSINFO_EHDR, sysinfo_ehdr);		\
 	NEW_AUX_ENT(AT_ADI_BLKSZ, adi_state.caps.blksz);		\
 	NEW_AUX_ENT(AT_ADI_NBITS, adi_state.caps.nbits);		\
 	NEW_AUX_ENT(AT_ADI_UEONADI, adi_state.caps.ue_on_adi);		\
arch/x86/include/asm/elf.h +14 -19
index 93ff2c7ca4df..d543aca7c725 100644
@@ -306,11 +306,14 @@ extern u32 elf_hwcap2;
 
 struct task_struct;
 
-#define	ARCH_DLINFO_IA32						\
+#define VDSO_ENTRY(sysinfo_ehdr)					\
+	(sysinfo_ehdr + vdso_image_32.sym___kernel_vsyscall)
+
+#define	ARCH_DLINFO_IA32(sysinfo_ehdr)					\
 do {									\
-	if (VDSO_CURRENT_BASE) {					\
-		NEW_AUX_ENT(AT_SYSINFO,	VDSO_ENTRY);			\
-		NEW_AUX_ENT(AT_SYSINFO_EHDR, VDSO_CURRENT_BASE);	\
+	if (sysinfo_ehdr) {						\
+		NEW_AUX_ENT(AT_SYSINFO,	VDSO_ENTRY(sysinfo_ehdr));	\
+		NEW_AUX_ENT(AT_SYSINFO_EHDR, sysinfo_ehdr);		\
 	}								\
 } while (0)
 
@@ -344,39 +347,31 @@ extern bool mmap_address_hint_valid(unsigned long addr, unsigned long len);
 #define __STACK_RND_MASK(is32bit) ((is32bit) ? 0x7ff : 0x3fffff)
 #define STACK_RND_MASK __STACK_RND_MASK(mmap_is_ia32())
 
-#define ARCH_DLINFO							\
+#define ARCH_DLINFO(sysinfo_ehdr)					\
 do {									\
 	if (vdso64_enabled)						\
-		NEW_AUX_ENT(AT_SYSINFO_EHDR,				\
-			    (unsigned long __force)current->mm->context.vdso); \
+		NEW_AUX_ENT(AT_SYSINFO_EHDR, sysinfo_ehdr);		\
 } while (0)
 
 /* As a historical oddity, the x32 and x86_64 vDSOs are controlled together. */
-#define ARCH_DLINFO_X32							\
+#define ARCH_DLINFO_X32(sysinfo_ehdr)					\
 do {									\
 	if (vdso64_enabled)						\
-		NEW_AUX_ENT(AT_SYSINFO_EHDR,				\
-			    (unsigned long __force)current->mm->context.vdso); \
+		NEW_AUX_ENT(AT_SYSINFO_EHDR, sysinfo_ehdr);		\
 } while (0)
 
 #define AT_SYSINFO		32
 
-#define COMPAT_ARCH_DLINFO						\
+#define COMPAT_ARCH_DLINFO(sysinfo_ehdr)				\
 if (exec->e_machine == EM_X86_64)					\
-	ARCH_DLINFO_X32;						\
+	ARCH_DLINFO_X32(sysinfo_ehdr);					\
 else if (IS_ENABLED(CONFIG_IA32_EMULATION))				\
-	ARCH_DLINFO_IA32
+	ARCH_DLINFO_IA32(sysinfo_ehdr)
 
 #define COMPAT_ELF_ET_DYN_BASE	(TASK_UNMAPPED_BASE + 0x1000000)
 
 #endif /* !CONFIG_X86_32 */
 
-#define VDSO_CURRENT_BASE	((unsigned long)current->mm->context.vdso)
-
-#define VDSO_ENTRY							\
-	((unsigned long)current->mm->context.vdso +			\
-	 vdso_image_32.sym___kernel_vsyscall)
-
 extern bool arch_syscall_is_vdso_sigreturn(struct pt_regs *regs);
 
 /* Do not change the values. See get_align_mask() */
arch/x86/um/asm/elf.h +2 -2
index b7c03a760a3c..8608b33ac0e4 100644
@@ -88,7 +88,7 @@ extern unsigned long __kernel_vsyscall;
 #define AT_SYSINFO		32
 #define AT_SYSINFO_EHDR		33
 
-#define ARCH_DLINFO						\
+#define ARCH_DLINFO(sysinfo_ehdr)				\
 do {								\
 	if ( vsyscall_ehdr ) {					\
 		NEW_AUX_ENT(AT_SYSINFO,	__kernel_vsyscall);	\
@@ -183,7 +183,7 @@ do {								\
 
 extern unsigned long um_vdso_addr;
 #define AT_SYSINFO_EHDR 33
-#define ARCH_DLINFO	NEW_AUX_ENT(AT_SYSINFO_EHDR, um_vdso_addr)
+#define ARCH_DLINFO(sysinfo_ehdr)	NEW_AUX_ENT(AT_SYSINFO_EHDR, um_vdso_addr)
 
 #endif
 
fs/binfmt_elf.c +3 -3
index 62741e55e3d1..a0e61ed9bdc7 100644
@@ -171,7 +171,7 @@ static int padzero(unsigned long elf_bss)
 static int
 create_elf_tables(struct linux_binprm *bprm, const struct elfhdr *exec,
 		unsigned long load_addr, unsigned long interp_load_addr,
-		unsigned long e_entry)
+		unsigned long e_entry, unsigned long sysinfo_ehdr)
 {
 	struct mm_struct *mm = current->mm;
 	unsigned long p = bprm->p;
@@ -252,7 +252,7 @@ create_elf_tables(struct linux_binprm *bprm, const struct elfhdr *exec,
 	 * update AT_VECTOR_SIZE_ARCH if the number of NEW_AUX_ENT() in
 	 * ARCH_DLINFO changes
 	 */
-	ARCH_DLINFO;
+	ARCH_DLINFO(sysinfo_ehdr);
 #endif
 	NEW_AUX_ENT(AT_HWCAP, ELF_HWCAP);
 	NEW_AUX_ENT(AT_PAGESZ, ELF_EXEC_PAGESIZE);
@@ -1258,7 +1258,7 @@ static int load_elf_binary(struct linux_binprm *bprm)
 		goto out;
 
 	retval = create_elf_tables(bprm, elf_ex,
-			  load_addr, interp_load_addr, e_entry);
+			  load_addr, interp_load_addr, e_entry, sysinfo_ehdr);
 	if (retval < 0)
 		goto out;
 
fs/binfmt_elf_fdpic.c +6 -5
index 421a09bc6ee6..0b5f9252e5ad 100644
@@ -63,7 +63,7 @@ static int elf_fdpic_map_file(struct elf_fdpic_params *, struct file *,
 
 static int create_elf_fdpic_tables(struct linux_binprm *, struct mm_struct *,
 				   struct elf_fdpic_params *,
-				   struct elf_fdpic_params *);
+				   struct elf_fdpic_params *, unsigned long);
 
 #ifndef CONFIG_MMU
 static int elf_fdpic_map_file_constdisp_on_uclinux(struct elf_fdpic_params *,
@@ -434,8 +434,8 @@ static int load_elf_fdpic_binary(struct linux_binprm *bprm)
 	current->mm->start_stack = current->mm->start_brk + stack_size;
 #endif
 
-	if (create_elf_fdpic_tables(bprm, current->mm,
-				    &exec_params, &interp_params) < 0)
+	if (create_elf_fdpic_tables(bprm, current->mm, &exec_params,
+				    &interp_params, sysinfo_ehdr) < 0)
 		goto error;
 
 	kdebug("- start_code  %lx", current->mm->start_code);
@@ -496,7 +496,8 @@ static int load_elf_fdpic_binary(struct linux_binprm *bprm)
 static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 				   struct mm_struct *mm,
 				   struct elf_fdpic_params *exec_params,
-				   struct elf_fdpic_params *interp_params)
+				   struct elf_fdpic_params *interp_params,
+				   unsigned long sysinfo_ehdr)
 {
 	const struct cred *cred = current_cred();
 	unsigned long sp, csp, nitems;
@@ -667,7 +668,7 @@ static int create_elf_fdpic_tables(struct linux_binprm *bprm,
 	/* ARCH_DLINFO must come last so platform specific code can enforce
 	 * special alignment requirements on the AUXV if necessary (eg. PPC).
 	 */
-	ARCH_DLINFO;
+	ARCH_DLINFO(sysinfo_ehdr);
 #endif
 #undef NEW_AUX_ENT
 
Signature · 1 line
2.31.1
#13 lore raw

PATCH v3 10/23 arm/vdso: Remove vdso pointer from mm->context

In reply to Dmitry Safonov

Not used any more: now sysinfo_ehdr is passed back from arch_setup_additional_pages() to set AT_SYSINFO_EHDR tag. .vdso_mremap() was only to track proper position of context.vdso throughout any mremap() syscalls, remove it too.

---
 arch/arm/include/asm/mmu.h |  3 ---
 arch/arm/kernel/vdso.c     | 10 ----------
 2 files changed, 13 deletions(-)
arch/arm/include/asm/mmu.h +0 -3
index 1592a4264488..2397b0a19f59 100644
@@ -12,9 +12,6 @@ typedef struct {
 #endif
 	unsigned int	vmalloc_seq;
 	unsigned long	sigpage;
-#ifdef CONFIG_VDSO
-	unsigned long	vdso;
-#endif
 #ifdef CONFIG_BINFMT_ELF_FDPIC
 	unsigned long	exec_fdpic_loadmap;
 	unsigned long	interp_fdpic_loadmap;
arch/arm/kernel/vdso.c +0 -10
index de516f8ba85a..4b39c7d8f525 100644
@@ -47,17 +47,8 @@ static const struct vm_special_mapping vdso_data_mapping = {
 	.pages = &vdso_data_page,
 };
 
-static int vdso_mremap(const struct vm_special_mapping *sm,
-		struct vm_area_struct *new_vma)
-{
-	current->mm->context.vdso = new_vma->vm_start;
-
-	return 0;
-}
-
 static struct vm_special_mapping vdso_text_mapping __ro_after_init = {
 	.name = "[vdso]",
-	.mremap = vdso_mremap,
 };
 
 struct elfinfo {
@@ -256,7 +247,6 @@ void arm_install_vdso(struct mm_struct *mm, unsigned long addr,
 	if (IS_ERR(vma))
 		return;
 
-	mm->context.vdso = addr;
 	*sysinfo_ehdr = addr;
 }
 
Signature · 1 line
2.31.1
#14 lore raw

PATCH v3 11/23 s390/vdso: Remove vdso_base pointer from mm->context

In reply to Dmitry Safonov

Not used any more: now sysinfo_ehdr is passed back from arch_setup_additional_pages() to set AT_SYSINFO_EHDR tag. .vdso_mremap() was only to track proper position of context.vdso throughout any mremap() syscalls, remove it too.

Cc Christian Borntraeger <[email protected]>
Cc Heiko Carstens <[email protected]>
Cc Vasily Gorbik <[email protected]>
---
 arch/s390/include/asm/mmu.h | 1 -
 arch/s390/kernel/vdso.c     | 9 ---------
 2 files changed, 10 deletions(-)
arch/s390/include/asm/mmu.h +0 -1
index e12ff0f29d1a..095d0596f700 100644
@@ -15,7 +15,6 @@ typedef struct {
 	unsigned long gmap_asce;
 	unsigned long asce;
 	unsigned long asce_limit;
-	unsigned long vdso_base;
 	/* The mmu context belongs to a secure guest. */
 	atomic_t is_protected;
 	/*
arch/s390/kernel/vdso.c +0 -9
index 8a72fdedbae9..b8fb43d3b657 100644
@@ -143,13 +143,6 @@ static vm_fault_t vvar_fault(const struct vm_special_mapping *sm,
 	return vmf_insert_pfn(vma, vmf->address, pfn);
 }
 
-static int vdso_mremap(const struct vm_special_mapping *sm,
-		       struct vm_area_struct *vma)
-{
-	current->mm->context.vdso_base = vma->vm_start;
-	return 0;
-}
-
 static struct vm_special_mapping vvar_mapping = {
 	.name = "[vvar]",
 	.fault = vvar_fault,
@@ -157,7 +150,6 @@ static struct vm_special_mapping vvar_mapping = {
 
 static struct vm_special_mapping vdso_mapping = {
 	.name = "[vdso]",
-	.mremap = vdso_mremap,
 };
 
 int vdso_getcpu_init(void)
@@ -203,7 +195,6 @@ int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 		do_munmap(mm, vvar_start, PAGE_SIZE, NULL);
 		rc = PTR_ERR(vma);
 	} else {
-		current->mm->context.vdso_base = vdso_text_start;
 		*sysinfo_ehdr = vdso_text_start;
 		rc = 0;
 	}
Signature · 1 line
2.31.1
#15 lore raw

PATCH v3 12/23 sparc/vdso: Remove vdso pointer from mm->context

In reply to Dmitry Safonov

Not used any more: now sysinfo_ehdr is passed back from arch_setup_additional_pages() to set AT_SYSINFO_EHDR tag.

---
 arch/sparc/include/asm/mmu_64.h | 1 -
 arch/sparc/vdso/vma.c           | 1 -
 2 files changed, 2 deletions(-)
arch/sparc/include/asm/mmu_64.h +0 -1
index 7e2704c770e9..8e7892890d14 100644
@@ -111,7 +111,6 @@ typedef struct {
 	unsigned long		thp_pte_count;
 	struct tsb_config	tsb_block[MM_NUM_TSBS];
 	struct hv_tsb_descr	tsb_descr[MM_NUM_TSBS];
-	void			*vdso;
 	bool			adi;
 	tag_storage_desc_t	*tag_store;
 	spinlock_t		tag_lock;
arch/sparc/vdso/vma.c +0 -1
index ae635893f9b3..94a43e01d2ac 100644
@@ -414,7 +414,6 @@ static int map_vdso(const struct vdso_image *image,
 		ret = PTR_ERR(vma);
 		do_munmap(mm, text_start, image->size, NULL);
 	} else {
-		current->mm->context.vdso = (void __user *)text_start;
 		*sysinfo_ehdr = text_start;
 	}
 
Signature · 1 line
2.31.1
#16 lore raw

PATCH v3 13/23 mm/mmap: Make vm_special_mapping::mremap return void

In reply to Dmitry Safonov

Previously .mremap() callback had to return (int) to provide a way to restrict resizing of a special mapping. Now it's restricted by providing .may_split = special_mapping_split.

Removing (int) return simplifies further changes to special_mapping_mremap() as it won't need save ret code from the callback. Also, it removes needless return 0 from callbacks.

---
 arch/arm/kernel/process.c | 3 +--
 arch/arm64/kernel/vdso.c  | 4 +---
 arch/mips/vdso/genvdso.c  | 3 +--
 arch/x86/entry/vdso/vma.c | 4 +---
 include/linux/mm_types.h  | 4 ++--
 mm/mmap.c                 | 2 +-
 6 files changed, 7 insertions(+), 13 deletions(-)
arch/arm/kernel/process.c +1 -2
index 5897ccb88bca..b863c5525b5d 100644
@@ -387,11 +387,10 @@ static unsigned long sigpage_addr(const struct mm_struct *mm,
 static struct page *signal_page;
 extern struct page *get_signal_page(void);
 
-static int sigpage_mremap(const struct vm_special_mapping *sm,
+static void sigpage_mremap(const struct vm_special_mapping *sm,
 		struct vm_area_struct *new_vma)
 {
 	current->mm->context.sigpage = new_vma->vm_start;
-	return 0;
 }
 
 static const struct vm_special_mapping sigpage_mapping = {
arch/arm64/kernel/vdso.c +1 -3
index c0512c2e8183..680415e0098c 100644
@@ -78,12 +78,10 @@ static union {
 } vdso_data_store __page_aligned_data;
 struct vdso_data *vdso_data = vdso_data_store.data;
 
-static int vdso_mremap(const struct vm_special_mapping *sm,
+static void vdso_mremap(const struct vm_special_mapping *sm,
 		struct vm_area_struct *new_vma)
 {
 	current->mm->context.vdso = (void *)new_vma->vm_start;
-
-	return 0;
 }
 
 static int __init __vdso_init(enum vdso_abi abi)
arch/mips/vdso/genvdso.c +1 -2
index 09e30eb4be86..0303d30cde03 100644
@@ -259,13 +259,12 @@ int main(int argc, char **argv)
 	fprintf(out_file, "#include <linux/linkage.h>\n");
 	fprintf(out_file, "#include <linux/mm.h>\n");
 	fprintf(out_file, "#include <asm/vdso.h>\n");
-	fprintf(out_file, "static int vdso_mremap(\n");
+	fprintf(out_file, "static void vdso_mremap(\n");
 	fprintf(out_file, "	const struct vm_special_mapping *sm,\n");
 	fprintf(out_file, "	struct vm_area_struct *new_vma)\n");
 	fprintf(out_file, "{\n");
 	fprintf(out_file, "	current->mm->context.vdso =\n");
 	fprintf(out_file, "	(void *)(new_vma->vm_start);\n");
-	fprintf(out_file, "	return 0;\n");
 	fprintf(out_file, "}\n");
 
 	/* Write out the stripped VDSO data. */
arch/x86/entry/vdso/vma.c +1 -3
index f1abe43aadb9..a286d44751be 100644
@@ -86,15 +86,13 @@ static void vdso_fix_landing(const struct vdso_image *image,
 #endif
 }
 
-static int vdso_mremap(const struct vm_special_mapping *sm,
+static void vdso_mremap(const struct vm_special_mapping *sm,
 		struct vm_area_struct *new_vma)
 {
 	const struct vdso_image *image = current->mm->context.vdso_image;
 
 	vdso_fix_landing(image, new_vma);
 	current->mm->context.vdso = (void __user *)new_vma->vm_start;
-
-	return 0;
 }
 
 #ifdef CONFIG_TIME_NS
include/linux/mm_types.h +2 -2
index 5aacc1c10a45..e9c5f2051f08 100644
@@ -770,8 +770,8 @@ struct vm_special_mapping {
 				struct vm_area_struct *vma,
 				struct vm_fault *vmf);
 
-	int (*mremap)(const struct vm_special_mapping *sm,
-		     struct vm_area_struct *new_vma);
+	void (*mremap)(const struct vm_special_mapping *sm,
+		       struct vm_area_struct *new_vma);
 };
 
 enum tlb_flush_reason {
mm/mmap.c +1 -1
index 0584e540246e..4f0d62409b1c 100644
@@ -3401,7 +3401,7 @@ static int special_mapping_mremap(struct vm_area_struct *new_vma)
 		return -EFAULT;
 
 	if (sm->mremap)
-		return sm->mremap(sm, new_vma);
+		sm->mremap(sm, new_vma);
 
 	return 0;
 }
Signature · 1 line
2.31.1
#17 lore raw

Re: [PATCH v3 13/23] mm/mmap: Make vm_special_mapping::mremap return void

Le 11/06/2021 à 20:02, Dmitry Safonov a écrit :

Quoted 9 lines

Previously .mremap() callback had to return (int) to provide a way to restrict resizing of a special mapping. Now it's restricted by providing .may_split = special_mapping_split.

Removing (int) return simplifies further changes to special_mapping_mremap() as it won't need save ret code from the callback. Also, it removes needless return 0 from callbacks.

Quoted 9 lines
---
  arch/arm/kernel/process.c | 3 +--
  arch/arm64/kernel/vdso.c  | 4 +---
  arch/mips/vdso/genvdso.c  | 3 +--
  arch/x86/entry/vdso/vma.c | 4 +---
  include/linux/mm_types.h  | 4 ++--
  mm/mmap.c                 | 2 +-
  6 files changed, 7 insertions(+), 13 deletions(-)

Build failure:

CC arch/powerpc/kernel/vdso.o
arch/powerpc/kernel/vdso.c:93:19: error: initialization of 'void (*)(const struct vm_special_mapping *, struct vm_area_struct *)' from incompatible pointer type 'int (*)(const struct vm_special_mapping *, struct vm_area_struct *)' [-Werror=incompatible-pointer-types]
93 | .mremap = vdso32_mremap,

      |                   ^~~~~~~~~~~~~
arch/powerpc/kernel/vdso.c:93:19: note: (near initialization for 'vdso32_spec.mremap')
arch/powerpc/kernel/vdso.c:98:19: error: initialization of 'void (*)(const struct vm_special_mapping *, struct vm_area_struct *)' from incompatible pointer type 'int (*)(const struct vm_special_mapping *, struct vm_area_struct *)' [-Werror=incompatible-pointer-types]
   98 |         .mremap = vdso64_mremap,
      |                   ^~~~~~~~~~~~~

arch/powerpc/kernel/vdso.c:98:19: note: (near initialization for 'vdso64_spec.mremap')
cc1: all warnings being treated as errors
make[3]: *** [arch/powerpc/kernel/vdso.o] Error 1
make[2]: *** [arch/powerpc/kernel] Error 2
make[1]: *** [arch/powerpc] Error 2
make: *** [__sub-make] Error 2

#18 lore raw

Re: [PATCH v3 13/23] mm/mmap: Make vm_special_mapping::mremap return void

Quoted 38 lines · Christophe Leroy

Le 11/06/2021 à 20:02, Dmitry Safonov a écrit :

Quoted 9 lines

Previously .mremap() callback had to return (int) to provide a way to restrict resizing of a special mapping. Now it's restricted by providing .may_split = special_mapping_split.

Removing (int) return simplifies further changes to special_mapping_mremap() as it won't need save ret code from the callback. Also, it removes needless return 0 from callbacks.

Quoted 9 lines

---
  arch/arm/kernel/process.c | 3 +--
  arch/arm64/kernel/vdso.c  | 4 +---
  arch/mips/vdso/genvdso.c  | 3 +--
  arch/x86/entry/vdso/vma.c | 4 +---
  include/linux/mm_types.h  | 4 ++--
  mm/mmap.c                 | 2 +-

  6 files changed, 7 insertions(+), 13 deletions(-)

Build failure:

  CC      arch/powerpc/kernel/vdso.o arch/powerpc/kernel/vdso.c:93:19: error: initialization of 'void (*)(const struct vm_special_mapping *, struct vm_area_struct *)' from incompatible pointer type 'int (*)(const struct vm_special_mapping *, struct vm_area_struct *)' [-Werror=incompatible-pointer-types] 93 |         .mremap = vdso32_mremap,
      |                   ^~~~~~~~~~~~~ arch/powerpc/kernel/vdso.c:93:19: note: (near initialization for 'vdso32_spec.mremap')
arch/powerpc/kernel/vdso.c:98:19: error: initialization of 'void (*)(const struct vm_special_mapping *, struct vm_area_struct *)' from incompatible pointer type 'int (*)(const struct vm_special_mapping *, struct vm_area_struct *)' [-Werror=incompatible-pointer-types] 98 |         .mremap = vdso64_mremap,

Quoted 9 lines

      |                   ^~~~~~~~~~~~~ arch/powerpc/kernel/vdso.c:98:19: note: (near initialization for 'vdso64_spec.mremap')
cc1: all warnings being treated as errors
make[3]: *** [arch/powerpc/kernel/vdso.o] Error 1 make[2]: *** [arch/powerpc/kernel] Error 2
make[1]: *** [arch/powerpc] Error 2
make: *** [__sub-make] Error 2

Thanks for reporting, Christophe, I'll fix that in v4.

Thanks,
Dmitry

#19 lore raw

PATCH v3 14/23 x86/signal: Land on &frame->retcode when vdso isn't mapped

In reply to Dmitry Safonov

Since commit 9fbbd4dd17d0 ("x86: Don't require the vDSO for handling a.out signals") after processing 32-bit signal if there is no vdso mapped frame->retcode is used as a landing.
Do the same for rt ia32 signals.
It also makes the ia32 compat signals match the native ia32 case.

This shouldn't be mistaken for encouragement for running binaries with executable stack, rather something to do in hopefully very rare situation with disabled or unmapped vdso and absent SA_RESTORER. For non-executable stack it'll segfault on attempt to land, rather than land on a random address where vdso was previously mapped. For programs with executable stack it'll just do the same for rt signals as for non-rt.

Discouraging users to run with executable stack is done separately in commit 47a2ebb7f505 ("execve: warn if process starts with executable stack").

---
 arch/x86/ia32/ia32_signal.c | 12 +++++++-----
 arch/x86/kernel/signal.c    | 23 ++++++++++-------------
 2 files changed, 17 insertions(+), 18 deletions(-)
arch/x86/ia32/ia32_signal.c +7 -5
index 5e3d9b7fd5fb..adb6994c40f6 100644
@@ -270,8 +270,8 @@ int ia32_setup_frame(int sig, struct ksignal *ksig,
 	unsafe_put_user(set->sig[1], &frame->extramask[0], Efault);
 	unsafe_put_user(ptr_to_compat(restorer), &frame->pretcode, Efault);
 	/*
-	 * These are actually not used anymore, but left because some
-	 * gdb versions depend on them as a marker.
+	 * This is popl %eax ; movl $__NR_sigreturn, %eax ; int $0x80
+	 * gdb uses it as a signature to notice signal handler stack frames.
 	 */
 	unsafe_put_user(*((u64 *)&code), (u64 __user *)frame->retcode, Efault);
 	user_access_end();
@@ -336,14 +336,16 @@ int ia32_setup_rt_frame(int sig, struct ksignal *ksig,
 
 	if (ksig->ka.sa.sa_flags & SA_RESTORER)
 		restorer = ksig->ka.sa.sa_restorer;
-	else
+	else if (current->mm->context.vdso)
 		restorer = current->mm->context.vdso +
 			vdso_image_32.sym___kernel_rt_sigreturn;
+	else
+		restorer = &frame->retcode;
 	unsafe_put_user(ptr_to_compat(restorer), &frame->pretcode, Efault);
 
 	/*
-	 * Not actually used anymore, but left because some gdb
-	 * versions need it.
+	 * This is popl %eax ; movl $__NR_sigreturn, %eax ; int $0x80
+	 * gdb uses it as a signature to notice signal handler stack frames.
 	 */
 	unsafe_put_user(*((u64 *)&code), (u64 __user *)frame->retcode, Efault);
 	unsafe_put_sigcontext32(&frame->uc.uc_mcontext, fp, regs, set, Efault);
arch/x86/kernel/signal.c +10 -13
index a06cb107c0e8..988cbc634949 100644
@@ -317,23 +317,20 @@ __setup_frame(int sig, struct ksignal *ksig, sigset_t *set,
 	unsafe_put_user(sig, &frame->sig, Efault);
 	unsafe_put_sigcontext(&frame->sc, fp, regs, set, Efault);
 	unsafe_put_user(set->sig[1], &frame->extramask[0], Efault);
-	if (current->mm->context.vdso)
+	if (ksig->ka.sa.sa_flags & SA_RESTORER)
+		restorer = ksig->ka.sa.sa_restorer;
+	else if (current->mm->context.vdso)
 		restorer = current->mm->context.vdso +
 			vdso_image_32.sym___kernel_sigreturn;
 	else
 		restorer = &frame->retcode;
-	if (ksig->ka.sa.sa_flags & SA_RESTORER)
-		restorer = ksig->ka.sa.sa_restorer;
 
 	/* Set up to return from userspace.  */
 	unsafe_put_user(restorer, &frame->pretcode, Efault);
 
 	/*
 	 * This is popl %eax ; movl $__NR_sigreturn, %eax ; int $0x80
-	 *
-	 * WE DO NOT USE IT ANY MORE! It's only left here for historical
-	 * reasons and because gdb uses it as a signature to notice
-	 * signal handler stack frames.
+	 * gdb uses it as a signature to notice signal handler stack frames.
 	 */
 	unsafe_put_user(*((u64 *)&retcode), (u64 *)frame->retcode, Efault);
 	user_access_end();
@@ -382,18 +379,18 @@ static int __setup_rt_frame(int sig, struct ksignal *ksig,
 	unsafe_save_altstack(&frame->uc.uc_stack, regs->sp, Efault);
 
 	/* Set up to return from userspace.  */
-	restorer = current->mm->context.vdso +
-		vdso_image_32.sym___kernel_rt_sigreturn;
 	if (ksig->ka.sa.sa_flags & SA_RESTORER)
 		restorer = ksig->ka.sa.sa_restorer;
+	else if (current->mm->context.vdso)
+		restorer = current->mm->context.vdso +
+			vdso_image_32.sym___kernel_rt_sigreturn;
+	else
+		restorer = &frame->retcode;
 	unsafe_put_user(restorer, &frame->pretcode, Efault);
 
 	/*
 	 * This is movl $__NR_rt_sigreturn, %ax ; int $0x80
-	 *
-	 * WE DO NOT USE IT ANY MORE! It's only left here for historical
-	 * reasons and because gdb uses it as a signature to notice
-	 * signal handler stack frames.
+	 * gdb uses it as a signature to notice signal handler stack frames.
 	 */
 	unsafe_put_user(*((u64 *)&rt_retcode), (u64 *)frame->retcode, Efault);
 	unsafe_put_sigcontext(&frame->uc.uc_mcontext, fp, regs, set, Efault);
Signature · 1 line
2.31.1
#20 lore raw

PATCH v3 15/23 x86/signal: Check if vdso_image_32 is mapped before trying to land on it

In reply to Dmitry Safonov

Provide current_has_vdso(image) helper and check it apriory landing attempt on vdso vma.
The helper is a macro, not a static inline funciton to avoid linux/sched/task_stack.h inclusion in asm/vdso.h.

---
 arch/x86/entry/common.c     | 10 +++++++++-
 arch/x86/ia32/ia32_signal.c |  4 ++--
 arch/x86/include/asm/vdso.h |  4 ++++
 arch/x86/kernel/signal.c    |  4 ++--
 4 files changed, 17 insertions(+), 5 deletions(-)
arch/x86/entry/common.c +9 -1
index 7b2542b13ebd..385a1c4bf4c0 100644
@@ -150,11 +150,19 @@ static noinstr bool __do_fast_syscall_32(struct pt_regs *regs)
 /* Returns 0 to return using IRET or 1 to return using SYSEXIT/SYSRETL. */
 __visible noinstr long do_fast_syscall_32(struct pt_regs *regs)
 {
+	unsigned long landing_pad;
+
+	if (!current_has_vdso(&vdso_image_32)) {
+		regs->ip = 0;
+		force_sigsegv(SIGSEGV);
+		syscall_exit_to_user_mode(regs);
+	}
+
 	/*
 	 * Called using the internal vDSO SYSENTER/SYSCALL32 calling
 	 * convention.  Adjust regs so it looks like we entered using int80.
 	 */
-	unsigned long landing_pad = (unsigned long)current->mm->context.vdso +
+	landing_pad = (unsigned long)current->mm->context.vdso +
 					vdso_image_32.sym_int80_landing_pad;
 
 	/*
arch/x86/ia32/ia32_signal.c +2 -2
index adb6994c40f6..2af40ae53a0e 100644
@@ -255,7 +255,7 @@ int ia32_setup_frame(int sig, struct ksignal *ksig,
 		restorer = ksig->ka.sa.sa_restorer;
 	} else {
 		/* Return stub is in 32bit vsyscall page */
-		if (current->mm->context.vdso)
+		if (current_has_vdso(&vdso_image_32))
 			restorer = current->mm->context.vdso +
 				vdso_image_32.sym___kernel_sigreturn;
 		else
@@ -336,7 +336,7 @@ int ia32_setup_rt_frame(int sig, struct ksignal *ksig,
 
 	if (ksig->ka.sa.sa_flags & SA_RESTORER)
 		restorer = ksig->ka.sa.sa_restorer;
-	else if (current->mm->context.vdso)
+	else if (current_has_vdso(&vdso_image_32))
 		restorer = current->mm->context.vdso +
 			vdso_image_32.sym___kernel_rt_sigreturn;
 	else
arch/x86/include/asm/vdso.h +4 -0
index 98aa103eb4ab..1ea7cb3f9b14 100644
@@ -45,6 +45,10 @@ extern const struct vdso_image vdso_image_x32;
 extern const struct vdso_image vdso_image_32;
 #endif
 
+#define current_has_vdso(image)						\
+	(current->mm->context.vdso != 0 &&				\
+	current->mm->context.vdso_image == image)
+
 extern void __init init_vdso_image(const struct vdso_image *image);
 
 extern int map_vdso_once(const struct vdso_image *image, unsigned long addr);
arch/x86/kernel/signal.c +2 -2
index 988cbc634949..77496ccb812d 100644
@@ -319,7 +319,7 @@ __setup_frame(int sig, struct ksignal *ksig, sigset_t *set,
 	unsafe_put_user(set->sig[1], &frame->extramask[0], Efault);
 	if (ksig->ka.sa.sa_flags & SA_RESTORER)
 		restorer = ksig->ka.sa.sa_restorer;
-	else if (current->mm->context.vdso)
+	else if (current_has_vdso(&vdso_image_32))
 		restorer = current->mm->context.vdso +
 			vdso_image_32.sym___kernel_sigreturn;
 	else
@@ -381,7 +381,7 @@ static int __setup_rt_frame(int sig, struct ksignal *ksig,
 	/* Set up to return from userspace.  */
 	if (ksig->ka.sa.sa_flags & SA_RESTORER)
 		restorer = ksig->ka.sa.sa_restorer;
-	else if (current->mm->context.vdso)
+	else if (current_has_vdso(&vdso_image_32))
 		restorer = current->mm->context.vdso +
 			vdso_image_32.sym___kernel_rt_sigreturn;
 	else
Signature · 1 line
2.31.1
#21 lore raw

PATCH v3 16/23 mm: Add vdso_base in mm_struct

In reply to Dmitry Safonov

Instead of having every architecture to define vdso_base/vdso_addr etc, provide a generic mechanism to track vdso_base for landing in userspace. It'll minimize per-architecture difference, the number of callbacks to provide.

Originally, it started from thread [1] where the need for .close() callback on vm_special_mapping was pointed, this generic code besides removing duplicated .mremap() callbacks provides a cheaper way to support munmap() on vdso mappings without introducing .close() callbacks for every architecture (which would bring even more code duplication).

[1]: https://lore.kernel.org/linux-arch/CAJwJo6ZANqYkSHbQ+3b+Fi_VT80MtrzEV5yreQAWx-L8j8x2zA@mail.gmail.com/

---
 arch/Kconfig                              |  3 +++
 arch/x86/kernel/cpu/resctrl/pseudo_lock.c |  3 ++-
 fs/aio.c                                  |  3 ++-
 include/linux/mm.h                        |  3 ++-
 include/linux/mm_types.h                  | 17 +++++++++++++++++
 kernel/fork.c                             |  1 +
 mm/mmap.c                                 | 19 ++++++++++++++++++-
 mm/mremap.c                               |  2 +-
 8 files changed, 46 insertions(+), 5 deletions(-)
arch/Kconfig +3 -0
index c45b770d3579..0e648a1ae412 100644
@@ -1235,6 +1235,9 @@ config HAVE_SPARSE_SYSCALL_NR
 config ARCH_HAS_VDSO_DATA
 	bool
 
+config ARCH_HAS_VDSO_BASE
+	bool
+
 config HAVE_STATIC_CALL
 	bool
 
arch/x86/kernel/cpu/resctrl/pseudo_lock.c +2 -1
index 05a89e33fde2..5ab191855816 100644
@@ -1458,7 +1458,8 @@ static int pseudo_lock_dev_release(struct inode *inode, struct file *filp)
 	return 0;
 }
 
-static int pseudo_lock_dev_mremap(struct vm_area_struct *area)
+static int pseudo_lock_dev_mremap(struct vm_area_struct *old_vma,
+				  struct vm_area_struct *new_vma)
 {
 	/* Not supported */
 	return -EINVAL;
fs/aio.c +2 -1
index 76ce0cc3ee4e..0b27e9300d0f 100644
@@ -323,7 +323,8 @@ static void aio_free_ring(struct kioctx *ctx)
 	}
 }
 
-static int aio_ring_mremap(struct vm_area_struct *vma)
+static int aio_ring_mremap(struct vm_area_struct *old_vma,
+			   struct vm_area_struct *vma)
 {
 	struct file *file = vma->vm_file;
 	struct mm_struct *mm = vma->vm_mm;
include/linux/mm.h +2 -1
index c274f75efcf9..6ef91b337f9d 100644
@@ -589,7 +589,8 @@ struct vm_operations_struct {
 	void (*close)(struct vm_area_struct * area);
 	/* Called any time before splitting to check if it's allowed */
 	int (*may_split)(struct vm_area_struct *area, unsigned long addr);
-	int (*mremap)(struct vm_area_struct *area);
+	int (*mremap)(struct vm_area_struct *old_vma,
+		      struct vm_area_struct *new_vma);
 	/*
 	 * Called by mprotect() to make driver-specific permission
 	 * checks before mprotect() is finalised.   The VMA must not
include/linux/mm_types.h +17 -0
index e9c5f2051f08..49ad34410132 100644
@@ -497,6 +497,16 @@ struct mm_struct {
 
 		/* Architecture-specific MM context */
 		mm_context_t context;
+#ifdef CONFIG_ARCH_HAS_VDSO_BASE
+		/*
+		 * Address of special mapping VMA to land after processing
+		 * a signal. Reads are unprotected: if a thread unmaps or
+		 * mremaps the mapping while another thread is processing
+		 * a signal, it can segfault while landing.
+		 */
+		void __user *vdso_base;
+#endif
+#define UNMAPPED_VDSO_BASE TASK_SIZE_MAX
 
 		unsigned long flags; /* Must use atomic bitops to access */
 
@@ -598,6 +608,13 @@ static inline void init_tlb_flush_pending(struct mm_struct *mm)
 	atomic_set(&mm->tlb_flush_pending, 0);
 }
 
+static inline void init_vdso_base(struct mm_struct *mm)
+{
+#ifdef CONFIG_ARCH_HAS_VDSO_BASE
+	mm->vdso_base = (void __user *)UNMAPPED_VDSO_BASE;
+#endif
+}
+
 static inline void inc_tlb_flush_pending(struct mm_struct *mm)
 {
 	atomic_inc(&mm->tlb_flush_pending);
kernel/fork.c +1 -0
index dc06afd725cb..eb4ad74d042e 100644
@@ -1082,6 +1082,7 @@ struct mm_struct *mm_alloc(void)
 		return NULL;
 
 	memset(mm, 0, sizeof(*mm));
+	init_vdso_base(mm);
 	return mm_init(mm, current, current_user_ns());
 }
 
mm/mmap.c +18 -1
index 4f0d62409b1c..5d1ffce51119 100644
@@ -3381,11 +3381,25 @@ void vm_stat_account(struct mm_struct *mm, vm_flags_t flags, long npages)
 
 static vm_fault_t special_mapping_fault(struct vm_fault *vmf);
 
+static void update_vdso_base(struct vm_area_struct *old_vma,
+				unsigned long new_addr)
+{
+#ifdef CONFIG_ARCH_HAS_VDSO_BASE
+	struct mm_struct *mm = old_vma->vm_mm;
+
+	if (WARN_ON_ONCE(!mm))
+		return;
+	if (old_vma->vm_start == (unsigned long)mm->vdso_base)
+		mm->vdso_base = (void __user *)new_addr;
+#endif
+}
+
 /*
  * Having a close hook prevents vma merging regardless of flags.
  */
 static void special_mapping_close(struct vm_area_struct *vma)
 {
+	update_vdso_base(vma, UNMAPPED_VDSO_BASE);
 }
 
 static const char *special_mapping_name(struct vm_area_struct *vma)
@@ -3393,7 +3407,8 @@ static const char *special_mapping_name(struct vm_area_struct *vma)
 	return ((struct vm_special_mapping *)vma->vm_private_data)->name;
 }
 
-static int special_mapping_mremap(struct vm_area_struct *new_vma)
+static int special_mapping_mremap(struct vm_area_struct *old_vma,
+				  struct vm_area_struct *new_vma)
 {
 	struct vm_special_mapping *sm = new_vma->vm_private_data;
 
@@ -3403,6 +3418,8 @@ static int special_mapping_mremap(struct vm_area_struct *new_vma)
 	if (sm->mremap)
 		sm->mremap(sm, new_vma);
 
+	update_vdso_base(old_vma, new_vma->vm_start);
+
 	return 0;
 }
 
mm/mremap.c +1 -1
index 47c255b60150..d28ba49d607c 100644
@@ -545,7 +545,7 @@ static unsigned long move_vma(struct vm_area_struct *vma,
 	if (moved_len < old_len) {
 		err = -ENOMEM;
 	} else if (vma->vm_ops && vma->vm_ops->mremap) {
-		err = vma->vm_ops->mremap(new_vma);
+		err = vma->vm_ops->mremap(vma, new_vma);
 	}
 
 	if (unlikely(err)) {
Signature · 1 line
2.31.1
#22 lore raw

PATCH v3 17/23 x86/vdso: Migrate to generic vdso_base

In reply to Dmitry Safonov

Generic way to track the landing vma area.
As a bonus, after unmapping vdso, kernel won't try to land on its previous position (due to UNMAPPED_VDSO_BASE check instead of context.vdso != 0 check).

---
 arch/x86/Kconfig              |  1 +
 arch/x86/entry/common.c       |  2 +-
 arch/x86/entry/vdso/extable.c |  4 ++--
 arch/x86/entry/vdso/vma.c     | 13 ++++++-------
 arch/x86/ia32/ia32_signal.c   |  4 ++--
 arch/x86/include/asm/mmu.h    |  1 -
 arch/x86/include/asm/vdso.h   |  2 +-
 arch/x86/kernel/signal.c      |  4 ++--
 8 files changed, 15 insertions(+), 16 deletions(-)
arch/x86/Kconfig +1 -0
index d30b952f4453..b1b6dab92fb6 100644
@@ -93,6 +93,7 @@ config X86
 	select ARCH_HAS_SYNC_CORE_BEFORE_USERMODE
 	select ARCH_HAS_SYSCALL_WRAPPER
 	select ARCH_HAS_UBSAN_SANITIZE_ALL
+	select ARCH_HAS_VDSO_BASE
 	select ARCH_HAS_DEBUG_WX
 	select ARCH_HAVE_NMI_SAFE_CMPXCHG
 	select ARCH_MIGHT_HAVE_ACPI_PDC		if ACPI
arch/x86/entry/common.c +1 -1
index 385a1c4bf4c0..f5861f6e00a2 100644
@@ -162,7 +162,7 @@ __visible noinstr long do_fast_syscall_32(struct pt_regs *regs)
 	 * Called using the internal vDSO SYSENTER/SYSCALL32 calling
 	 * convention.  Adjust regs so it looks like we entered using int80.
 	 */
-	landing_pad = (unsigned long)current->mm->context.vdso +
+	landing_pad = (unsigned long)current->mm->vdso_base +
 					vdso_image_32.sym_int80_landing_pad;
 
 	/*
arch/x86/entry/vdso/extable.c +2 -2
index afcf5b65beef..e5bfb35dafdb 100644
@@ -25,10 +25,10 @@ bool fixup_vdso_exception(struct pt_regs *regs, int trapnr,
 	if (trapnr == X86_TRAP_DB || trapnr == X86_TRAP_BP)
 		return false;
 
-	if (!current->mm->context.vdso)
+	if (current->mm->vdso_base == (void __user *)UNMAPPED_VDSO_BASE)
 		return false;
 
-	base =  (unsigned long)current->mm->context.vdso + image->extable_base;
+	base = (unsigned long)current->mm->vdso_base + image->extable_base;
 	nr_entries = image->extable_len / (sizeof(*extable));
 	extable = image->extable;
 
arch/x86/entry/vdso/vma.c +6 -7
index a286d44751be..f17e617c3b3e 100644
@@ -77,7 +77,7 @@ static void vdso_fix_landing(const struct vdso_image *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;
+			(unsigned long)current->mm->vdso_base;
 
 		/* Fixing userspace landing - look at do_fast_syscall_32 */
 		if (regs->ip == old_land_addr)
@@ -92,7 +92,6 @@ static void vdso_mremap(const struct vm_special_mapping *sm,
 	const struct vdso_image *image = current->mm->context.vdso_image;
 
 	vdso_fix_landing(image, new_vma);
-	current->mm->context.vdso = (void __user *)new_vma->vm_start;
 }
 
 #ifdef CONFIG_TIME_NS
@@ -287,7 +286,7 @@ static int map_vdso(const struct vdso_image *image, unsigned long addr,
 		ret = PTR_ERR(vma);
 		do_munmap(mm, text_start, image->size, NULL);
 	} else {
-		current->mm->context.vdso = (void __user *)text_start;
+		current->mm->vdso_base = (void __user *)text_start;
 		current->mm->context.vdso_image = image;
 		*sysinfo_ehdr = text_start;
 	}
@@ -362,8 +361,8 @@ int map_vdso_once(const struct vdso_image *image, unsigned long addr)
 	 * Check if we have already mapped vdso blob - fail to prevent
 	 * abusing from userspace install_special_mapping, which may
 	 * not do accounting and rlimit right.
-	 * We could search vma near context.vdso, but it's a slowpath,
-	 * so let's explicitly check all VMAs to be completely sure.
+	 * It's a slowpath, let's explicitly check all VMAs to be
+	 * completely sure.
 	 */
 	for (vma = mm->mmap; vma; vma = vma->vm_next) {
 		if (vma_is_special_mapping(vma, &vdso_mapping) ||
@@ -415,9 +414,9 @@ bool arch_syscall_is_vdso_sigreturn(struct pt_regs *regs)
 {
 #if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
 	const struct vdso_image *image = current->mm->context.vdso_image;
-	unsigned long vdso = (unsigned long) current->mm->context.vdso;
+	unsigned long vdso = (unsigned long) current->mm->vdso_base;
 
-	if (in_ia32_syscall() && image == &vdso_image_32) {
+	if (in_ia32_syscall() && current_has_vdso(&vdso_image_32)) {
 		if (regs->ip == vdso + image->sym_vdso32_sigreturn_landing_pad ||
 		    regs->ip == vdso + image->sym_vdso32_rt_sigreturn_landing_pad)
 			return true;
arch/x86/ia32/ia32_signal.c +2 -2
index 2af40ae53a0e..a98f4c7cdc38 100644
@@ -256,7 +256,7 @@ int ia32_setup_frame(int sig, struct ksignal *ksig,
 	} else {
 		/* Return stub is in 32bit vsyscall page */
 		if (current_has_vdso(&vdso_image_32))
-			restorer = current->mm->context.vdso +
+			restorer = current->mm->vdso_base +
 				vdso_image_32.sym___kernel_sigreturn;
 		else
 			restorer = &frame->retcode;
@@ -337,7 +337,7 @@ int ia32_setup_rt_frame(int sig, struct ksignal *ksig,
 	if (ksig->ka.sa.sa_flags & SA_RESTORER)
 		restorer = ksig->ka.sa.sa_restorer;
 	else if (current_has_vdso(&vdso_image_32))
-		restorer = current->mm->context.vdso +
+		restorer = current->mm->vdso_base +
 			vdso_image_32.sym___kernel_rt_sigreturn;
 	else
 		restorer = &frame->retcode;
arch/x86/include/asm/mmu.h +0 -1
index 5d7494631ea9..7bd10e6b8386 100644
@@ -43,7 +43,6 @@ typedef struct {
 #endif
 
 	struct mutex lock;
-	void __user *vdso;			/* vdso base address */
 	const struct vdso_image *vdso_image;	/* vdso image in use */
 
 	atomic_t perf_rdpmc_allowed;	/* nonzero if rdpmc is allowed */
arch/x86/include/asm/vdso.h +1 -1
index 1ea7cb3f9b14..e4dcc189c30d 100644
@@ -46,7 +46,7 @@ extern const struct vdso_image vdso_image_32;
 #endif
 
 #define current_has_vdso(image)						\
-	(current->mm->context.vdso != 0 &&				\
+	(current->mm->vdso_base != (void __user *)UNMAPPED_VDSO_BASE &&	\
 	current->mm->context.vdso_image == image)
 
 extern void __init init_vdso_image(const struct vdso_image *image);
arch/x86/kernel/signal.c +2 -2
index 77496ccb812d..c853d212e6cb 100644
@@ -320,7 +320,7 @@ __setup_frame(int sig, struct ksignal *ksig, sigset_t *set,
 	if (ksig->ka.sa.sa_flags & SA_RESTORER)
 		restorer = ksig->ka.sa.sa_restorer;
 	else if (current_has_vdso(&vdso_image_32))
-		restorer = current->mm->context.vdso +
+		restorer = current->mm->vdso_base +
 			vdso_image_32.sym___kernel_sigreturn;
 	else
 		restorer = &frame->retcode;
@@ -382,7 +382,7 @@ static int __setup_rt_frame(int sig, struct ksignal *ksig,
 	if (ksig->ka.sa.sa_flags & SA_RESTORER)
 		restorer = ksig->ka.sa.sa_restorer;
 	else if (current_has_vdso(&vdso_image_32))
-		restorer = current->mm->context.vdso +
+		restorer = current->mm->vdso_base +
 			vdso_image_32.sym___kernel_rt_sigreturn;
 	else
 		restorer = &frame->retcode;
Signature · 1 line
2.31.1
#23 lore raw

PATCH v3 18/23 arm/vdso: Migrate to generic vdso_base

In reply to Dmitry Safonov

Generic way to track the landing vma area.
As a bonus, after unmapping sigpage, kernel won't try to land on its previous position (due to UNMAPPED_VDSO_BASE check instead of context.vdso ?= 0 check).

---
 arch/arm/Kconfig          | 1 +
 arch/arm/kernel/process.c | 9 +--------
 arch/arm/kernel/signal.c  | 6 +++++-
 3 files changed, 7 insertions(+), 9 deletions(-)
arch/arm/Kconfig +1 -0
index 2df5ad505b8b..edf1cbb908a9 100644
@@ -23,6 +23,7 @@ config ARM
 	select ARCH_HAS_SYNC_DMA_FOR_CPU if SWIOTLB
 	select ARCH_HAS_TEARDOWN_DMA_OPS if MMU
 	select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
+	select ARCH_HAS_VDSO_BASE
 	select ARCH_HAVE_CUSTOM_GPIO_H
 	select ARCH_HAVE_NMI_SAFE_CMPXCHG if CPU_V7 || CPU_V7M || CPU_V6K
 	select ARCH_HAS_GCOV_PROFILE_ALL
arch/arm/kernel/process.c +1 -8
index b863c5525b5d..3a5975d1ace6 100644
@@ -387,16 +387,9 @@ static unsigned long sigpage_addr(const struct mm_struct *mm,
 static struct page *signal_page;
 extern struct page *get_signal_page(void);
 
-static void sigpage_mremap(const struct vm_special_mapping *sm,
-		struct vm_area_struct *new_vma)
-{
-	current->mm->context.sigpage = new_vma->vm_start;
-}
-
 static const struct vm_special_mapping sigpage_mapping = {
 	.name = "[sigpage]",
 	.pages = &signal_page,
-	.mremap = sigpage_mremap,
 };
 
 int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
@@ -434,7 +427,7 @@ int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 		goto up_fail;
 	}
 
-	mm->context.sigpage = addr;
+	mm->vdso_base = (void __user *)addr;
 
 	/* Unlike the sigpage, failure to install the vdso is unlikely
 	 * to be fatal to the process, so no error check needed
arch/arm/kernel/signal.c +5 -1
index a3a38d0a4c85..6c0507e84e24 100644
@@ -451,13 +451,17 @@ setup_return(struct pt_regs *regs, struct ksignal *ksig,
 #ifdef CONFIG_MMU
 		if (cpsr & MODE32_BIT) {
 			struct mm_struct *mm = current->mm;
+			unsigned long land = (unsigned long)mm->vdso_base;
+
+			if (land == UNMAPPED_VDSO_BASE)
+				return 1;
 
 			/*
 			 * 32-bit code can use the signal return page
 			 * except when the MPU has protected the vectors
 			 * page from PL0
 			 */
-			retcode = mm->context.sigpage + signal_return_offset +
+			retcode = land + signal_return_offset +
 				  (idx << 2) + thumb;
 		} else
 #endif
Signature · 1 line
2.31.1
#24 lore raw

PATCH v3 19/23 arm64/vdso: Migrate compat signals to generic vdso_base

In reply to Dmitry Safonov

Generic way to track the landing vma area.
As a bonus, after unmapping sigpage, kernel won't try to land on its previous position (due to UNMAPPED_VDSO_BASE check instead of context.vdso ?= 0 check).

---
 arch/arm64/Kconfig           |  1 +
 arch/arm64/include/asm/mmu.h |  3 ---
 arch/arm64/kernel/signal32.c | 17 ++++++++++++-----
 arch/arm64/kernel/vdso.c     | 11 +----------
 4 files changed, 14 insertions(+), 18 deletions(-)
arch/arm64/Kconfig +1 -0
index 385ef8d8ad9b..073d6c8aa195 100644
@@ -43,6 +43,7 @@ config ARM64
 	select ARCH_HAS_SYSCALL_WRAPPER
 	select ARCH_HAS_TEARDOWN_DMA_OPS if IOMMU_SUPPORT
 	select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
+	select ARCH_HAS_VDSO_BASE
 	select ARCH_HAVE_ELF_PROT
 	select ARCH_HAVE_NMI_SAFE_CMPXCHG
 	select ARCH_INLINE_READ_LOCK if !PREEMPTION
arch/arm64/include/asm/mmu.h +0 -3
index 75beffe2ee8a..434acd49dbc3 100644
@@ -18,9 +18,6 @@
 
 typedef struct {
 	atomic64_t	id;
-#ifdef CONFIG_COMPAT
-	void		*sigpage;
-#endif
 	refcount_t	pinned;
 	void		*vdso;
 	unsigned long	flags;
arch/arm64/kernel/signal32.c +12 -5
index 2f507f565c48..3d2ac8062e99 100644
@@ -315,7 +315,7 @@ static void __user *compat_get_sigframe(struct ksignal *ksig,
 	return frame;
 }
 
-static void compat_setup_return(struct pt_regs *regs, struct k_sigaction *ka,
+static int compat_setup_return(struct pt_regs *regs, struct k_sigaction *ka,
 				compat_ulong_t __user *rc, void __user *frame,
 				int usig)
 {
@@ -342,13 +342,16 @@ static void compat_setup_return(struct pt_regs *regs, struct k_sigaction *ka,
 		retcode = ptr_to_compat(ka->sa.sa_restorer);
 	} else {
 		/* Set up sigreturn pointer */
+		unsigned long land = (unsigned long)current->mm->vdso_base;
 		unsigned int idx = thumb << 1;
 
 		if (ka->sa.sa_flags & SA_SIGINFO)
 			idx += 3;
 
-		retcode = (unsigned long)current->mm->context.sigpage +
-			  (idx << 2) + thumb;
+		if (land == UNMAPPED_VDSO_BASE)
+			return 1;
+
+		retcode = land + (idx << 2) + thumb;
 	}
 
 	regs->regs[0]	= usig;
@@ -356,6 +359,8 @@ static void compat_setup_return(struct pt_regs *regs, struct k_sigaction *ka,
 	regs->compat_lr	= retcode;
 	regs->pc	= handler;
 	regs->pstate	= spsr;
+
+	return 0;
 }
 
 static int compat_setup_sigframe(struct compat_sigframe __user *sf,
@@ -425,7 +430,8 @@ int compat_setup_rt_frame(int usig, struct ksignal *ksig,
 	err |= compat_setup_sigframe(&frame->sig, regs, set);
 
 	if (err == 0) {
-		compat_setup_return(regs, &ksig->ka, frame->sig.retcode, frame, usig);
+		err = compat_setup_return(regs, &ksig->ka,
+					  frame->sig.retcode, frame, usig);
 		regs->regs[1] = (compat_ulong_t)(unsigned long)&frame->info;
 		regs->regs[2] = (compat_ulong_t)(unsigned long)&frame->sig.uc;
 	}
@@ -448,7 +454,8 @@ int compat_setup_frame(int usig, struct ksignal *ksig, sigset_t *set,
 
 	err |= compat_setup_sigframe(frame, regs, set);
 	if (err == 0)
-		compat_setup_return(regs, &ksig->ka, frame->retcode, frame, usig);
+		err = compat_setup_return(regs, &ksig->ka,
+					  frame->retcode, frame, usig);
 
 	return err;
 }
arch/arm64/kernel/vdso.c +1 -10
index 680415e0098c..3e9dd41abee0 100644
@@ -264,14 +264,6 @@ enum aarch32_map {
 static struct page *aarch32_vectors_page __ro_after_init;
 static struct page *aarch32_sig_page __ro_after_init;
 
-static int aarch32_sigpage_mremap(const struct vm_special_mapping *sm,
-				  struct vm_area_struct *new_vma)
-{
-	current->mm->context.sigpage = (void *)new_vma->vm_start;
-
-	return 0;
-}
-
 static struct vm_special_mapping aarch32_vdso_maps[] = {
 	[AA32_MAP_VECTORS] = {
 		.name	= "[vectors]", /* ABI */
@@ -280,7 +272,6 @@ static struct vm_special_mapping aarch32_vdso_maps[] = {
 	[AA32_MAP_SIGPAGE] = {
 		.name	= "[sigpage]", /* ABI */
 		.pages	= &aarch32_sig_page,
-		.mremap	= aarch32_sigpage_mremap,
 	},
 	[AA32_MAP_VVAR] = {
 		.name = "[vvar]",
@@ -398,7 +389,7 @@ static int aarch32_sigreturn_setup(struct mm_struct *mm)
 	if (IS_ERR(ret))
 		goto out;
 
-	mm->context.sigpage = (void *)addr;
+	mm->vdso_base = (void __user *)addr;
 
 out:
 	return PTR_ERR_OR_ZERO(ret);
Signature · 1 line
2.31.1
#25 lore raw

PATCH v3 20/23 arm64/vdso: Migrate native signals to generic vdso_base

In reply to Dmitry Safonov

Generic way to track the land vma area.
As a bonus, after unmapping vdso, kernel won't try to land on its previous position (due to UNMAPPED_VDSO_BASE check instead of context.vdso ?= 0 check).

---
 arch/arm64/include/asm/mmu.h |  1 -
 arch/arm64/kernel/signal.c   | 10 +++++++---
 arch/arm64/kernel/vdso.c     | 15 ++++++---------
 3 files changed, 13 insertions(+), 13 deletions(-)
arch/arm64/include/asm/mmu.h +0 -1
index 434acd49dbc3..4432a2809e0d 100644
@@ -19,7 +19,6 @@
 typedef struct {
 	atomic64_t	id;
 	refcount_t	pinned;
-	void		*vdso;
 	unsigned long	flags;
 } mm_context_t;
 
arch/arm64/kernel/signal.c +7 -3
index 6237486ff6bb..5f6a8c0abc4c 100644
@@ -723,9 +723,10 @@ static int get_sigframe(struct rt_sigframe_user_layout *user,
 	return 0;
 }
 
-static void setup_return(struct pt_regs *regs, struct k_sigaction *ka,
+static int setup_return(struct pt_regs *regs, struct k_sigaction *ka,
 			 struct rt_sigframe_user_layout *user, int usig)
 {
+	unsigned long land = (unsigned long)current->mm->vdso_base;
 	__sigrestore_t sigtramp;
 
 	regs->regs[0] = usig;
@@ -754,10 +755,13 @@ static void setup_return(struct pt_regs *regs, struct k_sigaction *ka,
 
 	if (ka->sa.sa_flags & SA_RESTORER)
 		sigtramp = ka->sa.sa_restorer;
+	else if (land != UNMAPPED_VDSO_BASE)
+		sigtramp = VDSO_SYMBOL(land, sigtramp);
 	else
-		sigtramp = VDSO_SYMBOL(current->mm->context.vdso, sigtramp);
+		return 1;
 
 	regs->regs[30] = (unsigned long)sigtramp;
+	return 0;
 }
 
 static int setup_rt_frame(int usig, struct ksignal *ksig, sigset_t *set,
@@ -780,7 +784,7 @@ static int setup_rt_frame(int usig, struct ksignal *ksig, sigset_t *set,
 	err |= __save_altstack(&frame->uc.uc_stack, regs->sp);
 	err |= setup_sigframe(&user, regs, set);
 	if (err == 0) {
-		setup_return(regs, &ksig->ka, &user, usig);
+		err = setup_return(regs, &ksig->ka, &user, usig);
 		if (ksig->ka.sa.sa_flags & SA_SIGINFO) {
 			err |= copy_siginfo_to_user(&frame->info, &ksig->info);
 			regs->regs[1] = (unsigned long)&frame->info;
arch/arm64/kernel/vdso.c +6 -9
index 3e9dd41abee0..4af29f89be37 100644
@@ -78,12 +78,6 @@ static union {
 } vdso_data_store __page_aligned_data;
 struct vdso_data *vdso_data = vdso_data_store.data;
 
-static void vdso_mremap(const struct vm_special_mapping *sm,
-		struct vm_area_struct *new_vma)
-{
-	current->mm->context.vdso = (void *)new_vma->vm_start;
-}
-
 static int __init __vdso_init(enum vdso_abi abi)
 {
 	int i;
@@ -244,7 +238,12 @@ static int __setup_additional_pages(enum vdso_abi abi,
 	if (IS_ERR(ret))
 		return PTR_ERR(ret);
 
-	mm->context.vdso = (void *)vdso_base;
+	/*
+	 * 32-bit ABI is to land on sigpage (see aarch32_sigreturn_setup()),
+	 * 64-bit on vDSO.
+	 */
+	if (abi == VDSO_ABI_AA64)
+		mm->vdso_base = (void __user *)vdso_base;
 	*sysinfo_ehdr = vdso_base;
 
 	return 0;
@@ -279,7 +278,6 @@ static struct vm_special_mapping aarch32_vdso_maps[] = {
 	},
 	[AA32_MAP_VDSO] = {
 		.name = "[vdso]",
-		.mremap = vdso_mremap,
 	},
 };
 
@@ -426,7 +424,6 @@ static struct vm_special_mapping aarch64_vdso_maps[] __ro_after_init = {
 	},
 	[AA64_MAP_VDSO] = {
 		.name	= "[vdso]",
-		.mremap = vdso_mremap,
 	},
 };
 
Signature · 1 line
2.31.1
#26 lore raw

PATCH v3 21/23 mips/vdso: Migrate to generic vdso_base

In reply to Dmitry Safonov

Generic way to track the landing vma area.
As a bonus, after unmapping sigpage, kernel won't try to land on its previous position (due to UNMAPPED_VDSO_BASE check instead of context.vdso ?= 0 check).

---
 arch/mips/Kconfig           |  1 +
 arch/mips/include/asm/mmu.h |  2 --
 arch/mips/kernel/signal.c   | 11 +++++++----
 arch/mips/kernel/vdso.c     |  2 +-
 arch/mips/vdso/genvdso.c    |  8 --------
 5 files changed, 9 insertions(+), 15 deletions(-)
arch/mips/Kconfig +1 -0
index 81096dd2c1ef..2bab0844f64b 100644
@@ -12,6 +12,7 @@ config MIPS
 	select ARCH_HAS_SETUP_ADDITIONAL_PAGES
 	select ARCH_HAS_TICK_BROADCAST if GENERIC_CLOCKEVENTS_BROADCAST
 	select ARCH_HAS_UBSAN_SANITIZE_ALL
+	select ARCH_HAS_VDSO_BASE
 	select ARCH_HAS_GCOV_PROFILE_ALL
 	select ARCH_KEEP_MEMBLOCK
 	select ARCH_SUPPORTS_UPROBES
arch/mips/include/asm/mmu.h +0 -2
index 5df0238f639b..928346a44eaf 100644
@@ -12,8 +12,6 @@ typedef struct {
 		atomic64_t mmid;
 	};
 
-	void *vdso;
-
 	/* lock to be held whilst modifying fp_bd_emupage_allocmap */
 	spinlock_t bd_emupage_lock;
 	/* bitmap tracking allocation of fp_bd_emupage */
arch/mips/kernel/signal.c +7 -4
index f1e985109da0..e0beaf2cdc0f 100644
@@ -806,11 +806,13 @@ struct mips_abi mips_abi = {
 
 static void handle_signal(struct ksignal *ksig, struct pt_regs *regs)
 {
+	void *land = (void *)current->mm->vdso_base;
 	sigset_t *oldset = sigmask_to_save();
-	int ret;
+	int ret = 1;
 	struct mips_abi *abi = current->thread.abi;
-	void *vdso = current->mm->context.vdso;
 
+	if (land == (void *)UNMAPPED_VDSO_BASE)
+		goto err;
 	/*
 	 * If we were emulating a delay slot instruction, exit that frame such
 	 * that addresses in the sigframe are as expected for userland and we
@@ -843,12 +845,13 @@ static void handle_signal(struct ksignal *ksig, struct pt_regs *regs)
 	rseq_signal_deliver(ksig, regs);
 
 	if (sig_uses_siginfo(&ksig->ka, abi))
-		ret = abi->setup_rt_frame(vdso + abi->vdso->off_rt_sigreturn,
+		ret = abi->setup_rt_frame(land + abi->vdso->off_rt_sigreturn,
 					  ksig, regs, oldset);
 	else
-		ret = abi->setup_frame(vdso + abi->vdso->off_sigreturn,
+		ret = abi->setup_frame(land + abi->vdso->off_sigreturn,
 				       ksig, regs, oldset);
 
+err:
 	signal_setup_done(ret, ksig, 0);
 }
 
arch/mips/kernel/vdso.c +1 -1
index 9b2e1d2250b4..3f76417e5aed 100644
@@ -184,7 +184,7 @@ int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 		goto out;
 	}
 
-	mm->context.vdso = (void *)vdso_addr;
+	mm->vdso_base = (void __user *)vdso_addr;
 	*sysinfo_ehdr = vdso_addr;
 	ret = 0;
 
arch/mips/vdso/genvdso.c +0 -8
index 0303d30cde03..8f581a2c8578 100644
@@ -259,13 +259,6 @@ int main(int argc, char **argv)
 	fprintf(out_file, "#include <linux/linkage.h>\n");
 	fprintf(out_file, "#include <linux/mm.h>\n");
 	fprintf(out_file, "#include <asm/vdso.h>\n");
-	fprintf(out_file, "static void vdso_mremap(\n");
-	fprintf(out_file, "	const struct vm_special_mapping *sm,\n");
-	fprintf(out_file, "	struct vm_area_struct *new_vma)\n");
-	fprintf(out_file, "{\n");
-	fprintf(out_file, "	current->mm->context.vdso =\n");
-	fprintf(out_file, "	(void *)(new_vma->vm_start);\n");
-	fprintf(out_file, "}\n");
 
 	/* Write out the stripped VDSO data. */
 	fprintf(out_file,
@@ -290,7 +283,6 @@ int main(int argc, char **argv)
 	fprintf(out_file, "\t.mapping = {\n");
 	fprintf(out_file, "\t\t.name = \"[vdso]\",\n");
 	fprintf(out_file, "\t\t.pages = vdso_pages,\n");
-	fprintf(out_file, "\t\t.mremap = vdso_mremap,\n");
 	fprintf(out_file, "\t},\n");
 
 	/* Calculate and write symbol offsets to <output file> */
Signature · 1 line
2.31.1
#27 lore raw

PATCH v3 22/23 powerpc/vdso: Migrate native signals to generic vdso_base

In reply to Dmitry Safonov

Generic way to track the land vma area.
Stat speaks for itself.

Cc Benjamin Herrenschmidt <[email protected]>
Cc Michael Ellerman <[email protected]>
Cc Paul Mackerras <[email protected]>
---
 arch/powerpc/Kconfig                          |  1 +
 arch/powerpc/include/asm/book3s/32/mmu-hash.h |  1 -
 arch/powerpc/include/asm/book3s/64/mmu.h      |  1 -
 arch/powerpc/include/asm/mmu_context.h        |  9 ------
 arch/powerpc/include/asm/nohash/32/mmu-40x.h  |  1 -
 arch/powerpc/include/asm/nohash/32/mmu-44x.h  |  1 -
 arch/powerpc/include/asm/nohash/32/mmu-8xx.h  |  1 -
 arch/powerpc/include/asm/nohash/mmu-book3e.h  |  1 -
 arch/powerpc/kernel/signal_32.c               |  8 ++---
 arch/powerpc/kernel/signal_64.c               |  4 +--
 arch/powerpc/kernel/vdso.c                    | 31 +------------------
 arch/powerpc/perf/callchain_32.c              |  8 ++---
 arch/powerpc/perf/callchain_64.c              |  4 +--
 arch/x86/include/asm/mmu_context.h            |  5 ---
 include/asm-generic/mm_hooks.h                |  9 ++----
 mm/mmap.c                                     |  7 -----
 16 files changed, 16 insertions(+), 76 deletions(-)
arch/powerpc/Kconfig +1 -0
index a9f842230ee4..21e58d145c82 100644
@@ -145,6 +145,7 @@ config PPC
 	select ARCH_HAS_TICK_BROADCAST		if GENERIC_CLOCKEVENTS_BROADCAST
 	select ARCH_HAS_UACCESS_FLUSHCACHE
 	select ARCH_HAS_UBSAN_SANITIZE_ALL
+	select ARCH_HAS_VDSO_BASE
 	select ARCH_HAVE_NMI_SAFE_CMPXCHG
 	select ARCH_KEEP_MEMBLOCK
 	select ARCH_MIGHT_HAVE_PC_PARPORT
arch/powerpc/include/asm/book3s/32/mmu-hash.h +0 -1
index b85f8e114a9c..d5ee68f394d9 100644
@@ -90,7 +90,6 @@ struct hash_pte {
 
 typedef struct {
 	unsigned long id;
-	void __user *vdso;
 } mm_context_t;
 
 void update_bats(void);
arch/powerpc/include/asm/book3s/64/mmu.h +0 -1
index eace8c3f7b0a..66bcc3ee3add 100644
@@ -111,7 +111,6 @@ typedef struct {
 
 	struct hash_mm_context *hash_context;
 
-	void __user *vdso;
 	/*
 	 * pagetable fragment support
 	 */
arch/powerpc/include/asm/mmu_context.h +0 -9
index 4bc45d3ed8b0..71dedeac7fdb 100644
@@ -260,15 +260,6 @@ static inline void enter_lazy_tlb(struct mm_struct *mm,
 
 extern void arch_exit_mmap(struct mm_struct *mm);
 
-static inline void arch_unmap(struct mm_struct *mm,
-			      unsigned long start, unsigned long end)
-{
-	unsigned long vdso_base = (unsigned long)mm->context.vdso;
-
-	if (start <= vdso_base && vdso_base < end)
-		mm->context.vdso = NULL;
-}
-
 #ifdef CONFIG_PPC_MEM_KEYS
 bool arch_vma_access_permitted(struct vm_area_struct *vma, bool write,
 			       bool execute, bool foreign);
arch/powerpc/include/asm/nohash/32/mmu-40x.h +0 -1
index 8a8f13a22cf4..366088bb1c3f 100644
@@ -57,7 +57,6 @@
 typedef struct {
 	unsigned int	id;
 	unsigned int	active;
-	void __user	*vdso;
 } mm_context_t;
 
 #endif /* !__ASSEMBLY__ */
arch/powerpc/include/asm/nohash/32/mmu-44x.h +0 -1
index 2d92a39d8f2e..d67256ab7887 100644
@@ -108,7 +108,6 @@ extern unsigned int tlb_44x_index;
 typedef struct {
 	unsigned int	id;
 	unsigned int	active;
-	void __user	*vdso;
 } mm_context_t;
 
 /* patch sites */
arch/powerpc/include/asm/nohash/32/mmu-8xx.h +0 -1
index 6e4faa0a9b35..9e394810faac 100644
@@ -184,7 +184,6 @@ void mmu_pin_tlb(unsigned long top, bool readonly);
 typedef struct {
 	unsigned int id;
 	unsigned int active;
-	void __user *vdso;
 	void *pte_frag;
 } mm_context_t;
 
arch/powerpc/include/asm/nohash/mmu-book3e.h +0 -1
index e43a418d3ccd..61ac19f315e5 100644
@@ -238,7 +238,6 @@ extern unsigned int tlbcam_index;
 typedef struct {
 	unsigned int	id;
 	unsigned int	active;
-	void __user	*vdso;
 } mm_context_t;
 
 /* Page size definitions, common between 32 and 64-bit
arch/powerpc/kernel/signal_32.c +4 -4
index 8f05ed0da292..ae61c480af53 100644
@@ -824,8 +824,8 @@ int handle_rt_signal32(struct ksignal *ksig, sigset_t *oldset,
 	}
 
 	/* Save user registers on the stack */
-	if (tsk->mm->context.vdso) {
-		tramp = VDSO32_SYMBOL(tsk->mm->context.vdso, sigtramp_rt32);
+	if (tsk->mm->vdso_base != (void __user *)UNMAPPED_VDSO_BASE) {
+		tramp = VDSO32_SYMBOL(tsk->mm->vdso_base, sigtramp_rt32);
 	} else {
 		tramp = (unsigned long)mctx->mc_pad;
 		/* Set up the sigreturn trampoline: li r0,sigret; sc */
@@ -922,8 +922,8 @@ int handle_signal32(struct ksignal *ksig, sigset_t *oldset,
 	else
 		unsafe_save_user_regs(regs, mctx, tm_mctx, 1, failed);
 
-	if (tsk->mm->context.vdso) {
-		tramp = VDSO32_SYMBOL(tsk->mm->context.vdso, sigtramp32);
+	if (tsk->mm->vdso_base != (void __user *)UNMAPPED_VDSO_BASE) {
+		tramp = VDSO32_SYMBOL(tsk->mm->vdso_base, sigtramp32);
 	} else {
 		tramp = (unsigned long)mctx->mc_pad;
 		/* Set up the sigreturn trampoline: li r0,sigret; sc */
arch/powerpc/kernel/signal_64.c +2 -2
index dca66481d0c2..468866dc1e0e 100644
@@ -906,8 +906,8 @@ int handle_rt_signal64(struct ksignal *ksig, sigset_t *set,
 	tsk->thread.fp_state.fpscr = 0;
 
 	/* Set up to return from userspace. */
-	if (tsk->mm->context.vdso) {
-		regs->nip = VDSO64_SYMBOL(tsk->mm->context.vdso, sigtramp_rt64);
+	if (tsk->mm->vdso_base != (void __user *)UNMAPPED_VDSO_BASE) {
+		regs->nip = VDSO64_SYMBOL(tsk->mm->vdso_base, sigtramp_rt64);
 	} else {
 		err |= setup_trampoline(__NR_rt_sigreturn, &frame->tramp[0]);
 		if (err)
arch/powerpc/kernel/vdso.c +1 -30
index 6d6e575630c1..2080a0540537 100644
@@ -57,29 +57,6 @@ enum vvar_pages {
 	VVAR_NR_PAGES,
 };
 
-static int vdso_mremap(const struct vm_special_mapping *sm, struct vm_area_struct *new_vma,
-		       unsigned long text_size)
-{
-	unsigned long new_size = new_vma->vm_end - new_vma->vm_start;
-
-	if (new_size != text_size)
-		return -EINVAL;
-
-	current->mm->context.vdso = (void __user *)new_vma->vm_start;
-
-	return 0;
-}
-
-static int vdso32_mremap(const struct vm_special_mapping *sm, struct vm_area_struct *new_vma)
-{
-	return vdso_mremap(sm, new_vma, &vdso32_end - &vdso32_start);
-}
-
-static int vdso64_mremap(const struct vm_special_mapping *sm, struct vm_area_struct *new_vma)
-{
-	return vdso_mremap(sm, new_vma, &vdso64_end - &vdso64_start);
-}
-
 static vm_fault_t vvar_fault(const struct vm_special_mapping *sm,
 			     struct vm_area_struct *vma, struct vm_fault *vmf);
 
@@ -90,12 +67,10 @@ static struct vm_special_mapping vvar_spec __ro_after_init = {
 
 static struct vm_special_mapping vdso32_spec __ro_after_init = {
 	.name = "[vdso]",
-	.mremap = vdso32_mremap,
 };
 
 static struct vm_special_mapping vdso64_spec __ro_after_init = {
 	.name = "[vdso]",
-	.mremap = vdso64_mremap,
 };
 
 #ifdef CONFIG_TIME_NS
@@ -251,7 +226,7 @@ static int __arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 	if (IS_ERR(vma)) {
 		do_munmap(mm, vdso_base, vvar_size, NULL);
 	} else {
-		mm->context.vdso = (void __user *)vdso_base + vvar_size;
+		mm->vdso_base = (void __user *)vdso_base + vvar_size;
 		*sysinfo_ehdr = vdso_base + vvar_size;
 	}
 
@@ -263,14 +238,10 @@ int arch_setup_additional_pages(unsigned long *sysinfo_ehdr)
 	struct mm_struct *mm = current->mm;
 	int rc;
 
-	mm->context.vdso = NULL;
-
 	if (mmap_write_lock_killable(mm))
 		return -EINTR;
 
 	rc = __arch_setup_additional_pages(sysinfo_ehdr);
-	if (rc)
-		mm->context.vdso = NULL;
 
 	mmap_write_unlock(mm);
 	return rc;
arch/powerpc/perf/callchain_32.c +4 -4
index b83c47b7947f..c48b63e16603 100644
@@ -59,8 +59,8 @@ static int is_sigreturn_32_address(unsigned int nip, unsigned int fp)
 {
 	if (nip == fp + offsetof(struct signal_frame_32, mctx.mc_pad))
 		return 1;
-	if (current->mm->context.vdso &&
-	    nip == VDSO32_SYMBOL(current->mm->context.vdso, sigtramp32))
+	if (current->mm->vdso_base != (void __user *)UNMAPPED_VDSO_BASE &&
+	    nip == VDSO32_SYMBOL(current->mm->vdso_base, sigtramp32))
 		return 1;
 	return 0;
 }
@@ -70,8 +70,8 @@ static int is_rt_sigreturn_32_address(unsigned int nip, unsigned int fp)
 	if (nip == fp + offsetof(struct rt_signal_frame_32,
 				 uc.uc_mcontext.mc_pad))
 		return 1;
-	if (current->mm->context.vdso &&
-	    nip == VDSO32_SYMBOL(current->mm->context.vdso, sigtramp_rt32))
+	if (current->mm->vdso_base != (void __user *)UNMAPPED_VDSO_BASE &&
+	    nip == VDSO32_SYMBOL(current->mm->vdso_base, sigtramp_rt32))
 		return 1;
 	return 0;
 }
arch/powerpc/perf/callchain_64.c +2 -2
index 8d0df4226328..ef7116bd525a 100644
@@ -68,8 +68,8 @@ static int is_sigreturn_64_address(unsigned long nip, unsigned long fp)
 {
 	if (nip == fp + offsetof(struct signal_frame_64, tramp))
 		return 1;
-	if (current->mm->context.vdso &&
-	    nip == VDSO64_SYMBOL(current->mm->context.vdso, sigtramp_rt64))
+	if (current->mm->vdso_base != (void __user *)UNMAPPED_VDSO_BASE &&
+	    nip == VDSO64_SYMBOL(current->mm->vdso_base, sigtramp_rt64))
 		return 1;
 	return 0;
 }
arch/x86/include/asm/mmu_context.h +0 -5
index 27516046117a..394aeaf136bb 100644
@@ -190,11 +190,6 @@ static inline bool is_64bit_mm(struct mm_struct *mm)
 }
 #endif
 
-static inline void arch_unmap(struct mm_struct *mm, unsigned long start,
-			      unsigned long end)
-{
-}
-
 /*
  * We only want to enforce protection keys on the current process
  * because we effectively have no access to PKRU for other
include/asm-generic/mm_hooks.h +2 -7
index 4dbb177d1150..6cd41034743d 100644
@@ -1,7 +1,7 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 /*
- * Define generic no-op hooks for arch_dup_mmap, arch_exit_mmap
- * and arch_unmap to be included in asm-FOO/mmu_context.h for any
+ * Define generic no-op hooks for arch_dup_mmap() and arch_exit_mmap()
+ * to be included in asm-FOO/mmu_context.h for any
  * arch FOO which doesn't need to hook these.
  */
 #ifndef _ASM_GENERIC_MM_HOOKS_H
@@ -17,11 +17,6 @@ static inline void arch_exit_mmap(struct mm_struct *mm)
 {
 }
 
-static inline void arch_unmap(struct mm_struct *mm,
-			unsigned long start, unsigned long end)
-{
-}
-
 static inline bool arch_vma_access_permitted(struct vm_area_struct *vma,
 		bool write, bool execute, bool foreign)
 {
mm/mmap.c +0 -7
index 5d1ffce51119..d22eb9ab770c 100644
@@ -2821,13 +2821,6 @@ int __do_munmap(struct mm_struct *mm, unsigned long start, size_t len,
 	if (len == 0)
 		return -EINVAL;
 
-	/*
-	 * arch_unmap() might do unmaps itself.  It must be called
-	 * and finish any rbtree manipulation before this code
-	 * runs and also starts to manipulate the rbtree.
-	 */
-	arch_unmap(mm, start, end);
-
 	/* Find the first overlapping VMA */
 	vma = find_vma(mm, start);
 	if (!vma)
Signature · 1 line
2.31.1
reviewer Acked-by on v3 22/23
#28 lore raw

Re: [PATCH v3 22/23] powerpc/vdso: Migrate native signals to generic vdso_base

Quoted 17 lines · Dmitry Safonov

Generic way to track the land vma area.
Stat speaks for itself.

Cc Benjamin Herrenschmidt <[email protected]>
Cc Michael Ellerman <[email protected]>
Cc Paul Mackerras <[email protected]>
---
 arch/powerpc/Kconfig                          |  1 +
 arch/powerpc/include/asm/book3s/32/mmu-hash.h |  1 -
 arch/powerpc/include/asm/book3s/64/mmu.h      |  1 -
 arch/powerpc/include/asm/mmu_context.h        |  9 ------
 arch/powerpc/include/asm/nohash/32/mmu-40x.h  |  1 -
 arch/powerpc/include/asm/nohash/32/mmu-44x.h  |  1 -
 arch/powerpc/include/asm/nohash/32/mmu-8xx.h  |  1 -
 arch/powerpc/include/asm/nohash/mmu-book3e.h  |  1 -
 arch/powerpc/kernel/signal_32.c               |  8 ++---
Quoted 8 lines
 arch/powerpc/kernel/signal_64.c               |  4 +--
 arch/powerpc/kernel/vdso.c                    | 31 +------------------
 arch/powerpc/perf/callchain_32.c              |  8 ++---
 arch/powerpc/perf/callchain_64.c              |  4 +--
 arch/x86/include/asm/mmu_context.h            |  5 ---
 include/asm-generic/mm_hooks.h                |  9 ++----
 mm/mmap.c                                     |  7 -----
 16 files changed, 16 insertions(+), 76 deletions(-)

LGTM.

cheers

#29 lore raw

Re: [PATCH v3 22/23] powerpc/vdso: Migrate native signals to generic vdso_base

In reply to Dmitry Safonov

Le 11/06/2021 à 20:02, Dmitry Safonov a écrit :

Quoted 19 lines

Generic way to track the land vma area.
Stat speaks for itself.

Cc Benjamin Herrenschmidt <[email protected]>
Cc Michael Ellerman <[email protected]>
Cc Paul Mackerras <[email protected]>

---

arch/powerpc/include/asm/mmu_context.h +0 -2
index 4bc45d3ed8b0..71dedeac7fdb 100644
@@ -260,15 +260,6 @@ static inline void enter_lazy_tlb(struct mm_struct *mm,
 
  extern void arch_exit_mmap(struct mm_struct *mm);
 
-static inline void arch_unmap(struct mm_struct *mm,
-			      unsigned long start, unsigned long end)

arch/powerpc/include/asm/mmu_context.h static inline void enter_lazy_tlb(struct mm_struct *mm, · line 265

Quoted 10 lines
arch/powerpc/include/asm/mmu_context.h +0 -7
-{
-	unsigned long vdso_base = (unsigned long)mm->context.vdso;
-
-	if (start <= vdso_base && vdso_base < end)
-		mm->context.vdso = NULL;
-}
-
  #ifdef CONFIG_PPC_MEM_KEYS
  bool arch_vma_access_permitted(struct vm_area_struct *vma, bool write,
  			       bool execute, bool foreign);

powerpc was the only user of arch_unmap().

We should get rid of it completely (Remove the stubs in arch/x86/include/asm/mmu_context.h and include/asm-generic/mm_hooks.h and remove call in mm/mmap.c)

#30 lore raw

Re: [PATCH v3 22/23] powerpc/vdso: Migrate native signals to generic vdso_base

In reply to Dmitry Safonov

Le 11/06/2021 à 20:02, Dmitry Safonov a écrit :

Quoted 7 lines

Generic way to track the land vma area.
Stat speaks for itself.

Cc Benjamin Herrenschmidt <[email protected]>
Cc Michael Ellerman <[email protected]>
Cc Paul Mackerras <[email protected]>

Build failure:

CC arch/powerpc/kernel/asm-offsets.s
In file included from ./include/linux/mmzone.h:21,

                 from ./include/linux/gfp.h:6,
                 from ./include/linux/xarray.h:14,
                 from ./include/linux/radix-tree.h:19,
                 from ./include/linux/fs.h:15,
                 from ./include/linux/compat.h:17,
                 from arch/powerpc/kernel/asm-offsets.c:14:
./include/linux/mm_types.h: In function 'init_vdso_base':
./include/linux/mm_types.h:522:28: error: 'TASK_SIZE_MAX' undeclared (first use in this function); did you mean 'XATTR_SIZE_MAX'?
  522 | #define UNMAPPED_VDSO_BASE TASK_SIZE_MAX
      |                            ^~~~~~~~~~~~~
./include/linux/mm_types.h:627:40: note: in expansion of macro 'UNMAPPED_VDSO_BASE'
  627 |         mm->vdso_base = (void __user *)UNMAPPED_VDSO_BASE;
      |                                        ^~~~~~~~~~~~~~~~~~
./include/linux/mm_types.h:522:28: note: each undeclared identifier is reported only once for each function it appears in
  522 | #define UNMAPPED_VDSO_BASE TASK_SIZE_MAX
      |                            ^~~~~~~~~~~~~
./include/linux/mm_types.h:627:40: note: in expansion of macro 'UNMAPPED_VDSO_BASE'
  627 |         mm->vdso_base = (void __user *)UNMAPPED_VDSO_BASE;
      |                                        ^~~~~~~~~~~~~~~~~~

make[2]: *** [arch/powerpc/kernel/asm-offsets.s] Error 1 make[1]: *** [prepare0] Error 2
make: *** [__sub-make] Error 2

#31 lore raw

Re: [PATCH v3 22/23] powerpc/vdso: Migrate native signals to generic vdso_base

Le 17/06/2021 à 08:36, Christophe Leroy a écrit :

Quoted 30 lines

Le 11/06/2021 à 20:02, Dmitry Safonov a écrit :

Quoted 7 lines

Generic way to track the land vma area.
Stat speaks for itself.

Cc Benjamin Herrenschmidt <[email protected]>
Cc Michael Ellerman <[email protected]>
Cc Paul Mackerras <[email protected]>

Build failure:

  CC      arch/powerpc/kernel/asm-offsets.s
In file included from ./include/linux/mmzone.h:21,
                 from ./include/linux/gfp.h:6,
                 from ./include/linux/xarray.h:14,
                 from ./include/linux/radix-tree.h:19,
                 from ./include/linux/fs.h:15,
                 from ./include/linux/compat.h:17,
                 from arch/powerpc/kernel/asm-offsets.c:14:
./include/linux/mm_types.h: In function 'init_vdso_base':
./include/linux/mm_types.h:522:28: error: 'TASK_SIZE_MAX' undeclared (first use in this function); did you mean 'XATTR_SIZE_MAX'?
  522 | #define UNMAPPED_VDSO_BASE TASK_SIZE_MAX
      |                            ^~~~~~~~~~~~~
./include/linux/mm_types.h:627:40: note: in expansion of macro 'UNMAPPED_VDSO_BASE'
  627 |         mm->vdso_base = (void __user *)UNMAPPED_VDSO_BASE;
      |                                        ^~~~~~~~~~~~~~~~~~
./include/linux/mm_types.h:522:28: note: each undeclared identifier is reported only once for each function it appears in

Quoted 9 lines

  522 | #define UNMAPPED_VDSO_BASE TASK_SIZE_MAX
      |                            ^~~~~~~~~~~~~
./include/linux/mm_types.h:627:40: note: in expansion of macro 'UNMAPPED_VDSO_BASE'
  627 |         mm->vdso_base = (void __user *)UNMAPPED_VDSO_BASE;
      |                                        ^~~~~~~~~~~~~~~~~~
make[2]: *** [arch/powerpc/kernel/asm-offsets.s] Error 1
make[1]: *** [prepare0] Error 2
make: *** [__sub-make] Error 2

Fixed by moving TASK_SIZE_MAX into asm/task_size_32.h and asm/task_size_64.h

arch/powerpc/include/asm/task_size_32.h +1 -0
index de7290ee770f..03af9e6bb5cd 100644
@@ -7,6 +7,7 @@
 #endif
 
 #define TASK_SIZE (CONFIG_TASK_SIZE)
+#define TASK_SIZE_MAX		TASK_SIZE
 
 /*
  * This decides where the kernel will search for a free chunk of vm space during
arch/powerpc/include/asm/task_size_64.h +1 -0
index c993482237ed..bfdb98c0ef43 100644
@@ -49,6 +49,7 @@
 						TASK_SIZE_USER64)
 
 #define TASK_SIZE TASK_SIZE_OF(current)
+#define TASK_SIZE_MAX		TASK_SIZE_USER64
 
 #define TASK_UNMAPPED_BASE_USER32 (PAGE_ALIGN(TASK_SIZE_USER32 / 4))
 #define TASK_UNMAPPED_BASE_USER64 (PAGE_ALIGN(DEFAULT_MAP_WINDOW_USER64 / 4))
arch/powerpc/include/asm/uaccess.h +0 -7
index 22c79ab40006..5823140d39f1 100644
@@ -8,13 +8,6 @@
 #include <asm/extable.h>
 #include <asm/kup.h>
 
-#ifdef __powerpc64__
-/* We use TASK_SIZE_USER64 as TASK_SIZE is not constant */
-#define TASK_SIZE_MAX		TASK_SIZE_USER64
-#else
-#define TASK_SIZE_MAX		TASK_SIZE
-#endif
-
 static inline bool __access_ok(unsigned long addr, unsigned long size)
 {
 	return addr < TASK_SIZE_MAX && size <= TASK_SIZE_MAX - addr;
#32 lore raw

Re: [PATCH v3 22/23] powerpc/vdso: Migrate native signals to generic vdso_base

Hi Chirstophe,

Quoted 87 lines · Christophe Leroy

Le 17/06/2021 à 08:36, Christophe Leroy a écrit :

Quoted 34 lines

Le 11/06/2021 à 20:02, Dmitry Safonov a écrit :

Quoted 7 lines

Generic way to track the land vma area.
Stat speaks for itself.

Cc Benjamin Herrenschmidt <[email protected]>
Cc Michael Ellerman <[email protected]>
Cc Paul Mackerras <[email protected]>

Build failure:

   CC      arch/powerpc/kernel/asm-offsets.s
In file included from ./include/linux/mmzone.h:21,
                  from ./include/linux/gfp.h:6,
                  from ./include/linux/xarray.h:14,
                  from ./include/linux/radix-tree.h:19, from ./include/linux/fs.h:15,
                  from ./include/linux/compat.h:17,
                  from arch/powerpc/kernel/asm-offsets.c:14:
./include/linux/mm_types.h: In function 'init_vdso_base':
./include/linux/mm_types.h:522:28: error: 'TASK_SIZE_MAX' undeclared (first use in this function); did you mean 'XATTR_SIZE_MAX'? 522 | #define UNMAPPED_VDSO_BASE TASK_SIZE_MAX
       |                            ^~~~~~~~~~~~~ ./include/linux/mm_types.h:627:40: note: in expansion of macro 'UNMAPPED_VDSO_BASE'
   627 |         mm->vdso_base = (void __user *)UNMAPPED_VDSO_BASE; |                                        ^~~~~~~~~~~~~~~~~~ ./include/linux/mm_types.h:522:28: note: each undeclared identifier is reported only once for each function it appears in
   522 | #define UNMAPPED_VDSO_BASE TASK_SIZE_MAX

Quoted 9 lines

       |                            ^~~~~~~~~~~~~ ./include/linux/mm_types.h:627:40: note: in expansion of macro 'UNMAPPED_VDSO_BASE'
   627 |         mm->vdso_base = (void __user *)UNMAPPED_VDSO_BASE; |                                        ^~~~~~~~~~~~~~~~~~ make[2]: *** [arch/powerpc/kernel/asm-offsets.s] Error 1 make[1]: *** [prepare0] Error 2
make: *** [__sub-make] Error 2

Fixed by moving TASK_SIZE_MAX into asm/task_size_32.h and asm/task_size_64.h

arch/powerpc/include/asm/task_size_32.h +1 -0
index de7290ee770f..03af9e6bb5cd 100644
@@ -7,6 +7,7 @@
  #endif
 
  #define TASK_SIZE (CONFIG_TASK_SIZE)
+#define TASK_SIZE_MAX        TASK_SIZE
 
  /*
   * This decides where the kernel will search for a free chunk of vm

space during

arch/powerpc/include/asm/task_size_64.h +1 -0
index c993482237ed..bfdb98c0ef43 100644
@@ -49,6 +49,7 @@
                          TASK_SIZE_USER64)
 
  #define TASK_SIZE TASK_SIZE_OF(current)
+#define TASK_SIZE_MAX        TASK_SIZE_USER64
 
  #define TASK_UNMAPPED_BASE_USER32 (PAGE_ALIGN(TASK_SIZE_USER32 / 4))
  #define TASK_UNMAPPED_BASE_USER64 (PAGE_ALIGN(DEFAULT_MAP_WINDOW_USER64

/ 4))

arch/powerpc/include/asm/uaccess.h +0 -0
index 22c79ab40006..5823140d39f1 100644
@@ -8,13 +8,6 @@
  #include <asm/extable.h>
  #include <asm/kup.h>
 

arch/powerpc/include/asm/uaccess.h line 13

Quoted 10 lines
arch/powerpc/include/asm/uaccess.h +0 -7
-#ifdef __powerpc64__
-/* We use TASK_SIZE_USER64 as TASK_SIZE is not constant */
-#define TASK_SIZE_MAX        TASK_SIZE_USER64
-#else
-#define TASK_SIZE_MAX        TASK_SIZE
-#endif
-
  static inline bool __access_ok(unsigned long addr, unsigned long size)
  {
      return addr < TASK_SIZE_MAX && size <= TASK_SIZE_MAX - addr;

Thanks, that's very helpful!

Signature · 1 line
          Dmitry
#33 lore raw

PATCH v3 23/23 x86/vdso/selftest: Add a test for unmapping vDSO

In reply to Dmitry Safonov

Output for landing on x86:

Quoted 7 lines

[root@localhost ~]# ./test_munmap_vdso_64

	AT_SYSINFO_EHDR is 0x7fffead9f000

[NOTE] unmapping vDSO: [0x7fffead9f000, 0x7fffeada0000]
[NOTE] vDSO partial move failed, will try with bigger size
[NOTE] unmapping vDSO: [0x7fffead9f000, 0x7fffeada1000]
[OK]
[root@localhost ~]# ./test_munmap_vdso_32

Quoted 5 lines
	AT_SYSINFO_EHDR is 0xf7eef000

[NOTE] unmapping vDSO: [0xf7eef000, 0xf7ef0000]
[NOTE] vDSO partial move failed, will try with bigger size
[NOTE] unmapping vDSO: [0xf7eef000, 0xf7ef1000]
[OK]

The test also can check force_sigsegv(SIGSEGV) in do_fast_syscall_32():

Quoted 7 lines

[root@localhost ~]# ./test_munmap_vdso_32 sysenter
[NOTE] Using sysenter after munmap

	AT_SYSINFO_EHDR is 0xf7efe000

[NOTE] unmapping vDSO: [0xf7efe000, 0xf7eff000]
[NOTE] vDSO partial move failed, will try with bigger size
[NOTE] unmapping vDSO: [0xf7efe000, 0xf7f00000]
[OK] 32-bit process gets segfault on fast syscall with unmapped vDSO

---
 tools/testing/selftests/x86/.gitignore        |   1 +
 tools/testing/selftests/x86/Makefile          |  11 +-
 .../testing/selftests/x86/test_munmap_vdso.c  | 151 ++++++++++++++++++
 3 files changed, 158 insertions(+), 5 deletions(-)
 create mode 100644 tools/testing/selftests/x86/test_munmap_vdso.c
tools/testing/selftests/x86/.gitignore +1 -0
index 1aaef5bf119a..9ce8337e8fa0 100644
@@ -6,6 +6,7 @@ sysret_ss_attrs
 syscall_nt
 ptrace_syscall
 test_mremap_vdso
+test_munmap_vdso
 check_initial_reg_state
 sigreturn
 ldt_gdt
tools/testing/selftests/x86/Makefile +6 -5
index 333980375bc7..43016351ddb3 100644
@@ -10,12 +10,13 @@ CAN_BUILD_I386 := $(shell ./check_cc.sh $(CC) trivial_32bit_program.c -m32)
 CAN_BUILD_X86_64 := $(shell ./check_cc.sh $(CC) trivial_64bit_program.c)
 CAN_BUILD_WITH_NOPIE := $(shell ./check_cc.sh $(CC) trivial_program.c -no-pie)
 
-TARGETS_C_BOTHBITS := single_step_syscall sysret_ss_attrs syscall_nt test_mremap_vdso \
-			check_initial_reg_state sigreturn iopl ioperm \
-			test_vsyscall mov_ss_trap \
+TARGETS_C_BOTHBITS := single_step_syscall sysret_ss_attrs syscall_nt	\
+			test_mremap_vdso test_munmap_vdso		\
+			check_initial_reg_state sigreturn iopl ioperm	\
+			test_vsyscall mov_ss_trap			\
 			syscall_arg_fault fsgsbase_restore
-TARGETS_C_32BIT_ONLY := entry_from_vm86 test_syscall_vdso unwind_vdso \
-			test_FCMOV test_FCOMI test_FISTTP \
+TARGETS_C_32BIT_ONLY := entry_from_vm86 test_syscall_vdso unwind_vdso	\
+			test_FCMOV test_FCOMI test_FISTTP		\
 			vdso_restorer
 TARGETS_C_64BIT_ONLY := fsgsbase sysret_rip syscall_numbering
 # Some selftests require 32bit support enabled also on 64bit systems
tools/testing/selftests/x86/test_munmap_vdso.c new file +151 -0

151 lines not on this page · raw message

Signature · 1 line
2.31.1
maintainer
#34 lore raw

Re: [PATCH v3 23/23] x86/vdso/selftest: Add a test for unmapping vDSO

Quoted 23 lines · Dmitry Safonov

Output for landing on x86:

Quoted 7 lines

[root@localhost ~]# ./test_munmap_vdso_64

	AT_SYSINFO_EHDR is 0x7fffead9f000

[NOTE] unmapping vDSO: [0x7fffead9f000, 0x7fffeada0000]
[NOTE] vDSO partial move failed, will try with bigger size
[NOTE] unmapping vDSO: [0x7fffead9f000, 0x7fffeada1000]
[OK]
[root@localhost ~]# ./test_munmap_vdso_32

Quoted 5 lines
	AT_SYSINFO_EHDR is 0xf7eef000

[NOTE] unmapping vDSO: [0xf7eef000, 0xf7ef0000]
[NOTE] vDSO partial move failed, will try with bigger size
[NOTE] unmapping vDSO: [0xf7eef000, 0xf7ef1000]
[OK]

The test also can check force_sigsegv(SIGSEGV) in do_fast_syscall_32():

Quoted 7 lines

[root@localhost ~]# ./test_munmap_vdso_32 sysenter
[NOTE] Using sysenter after munmap

	AT_SYSINFO_EHDR is 0xf7efe000

[NOTE] unmapping vDSO: [0xf7efe000, 0xf7eff000]
[NOTE] vDSO partial move failed, will try with bigger size
[NOTE] unmapping vDSO: [0xf7efe000, 0xf7f00000]
[OK] 32-bit process gets segfault on fast syscall with unmapped vDSO

Quoted 9 lines
---
  tools/testing/selftests/x86/.gitignore        |   1 +
  tools/testing/selftests/x86/Makefile          |  11 +-
  .../testing/selftests/x86/test_munmap_vdso.c  | 151 ++++++++++++++++++
  3 files changed, 158 insertions(+), 5 deletions(-)

create mode 100644 tools/testing/selftests/x86/test_munmap_vdso.c

I can take this through kselftest tree for 5.14 - are there any dependencies on x86 tree, I should be aware of?

thanks,
-- Shuah

#35 lore raw

Re: [PATCH v3 23/23] x86/vdso/selftest: Add a test for unmapping vDSO

Quoted 34 lines · Shuah Khan
Quoted 23 lines · Dmitry Safonov

Output for landing on x86:

Quoted 4 lines

[root@localhost ~]# ./test_munmap_vdso_64 AT_SYSINFO_EHDR is 0x7fffead9f000
[NOTE]    unmapping vDSO: [0x7fffead9f000, 0x7fffeada0000] [NOTE]    vDSO partial move failed, will try with bigger size

Quoted 8 lines

[NOTE]    unmapping vDSO: [0x7fffead9f000, 0x7fffeada1000] [OK]
[root@localhost ~]# ./test_munmap_vdso_32 AT_SYSINFO_EHDR is 0xf7eef000
[NOTE]    unmapping vDSO: [0xf7eef000, 0xf7ef0000] [NOTE]    vDSO partial move failed, will try with bigger size [NOTE]    unmapping vDSO: [0xf7eef000, 0xf7ef1000] [OK]

The test also can check force_sigsegv(SIGSEGV) in do_fast_syscall_32():

Quoted 7 lines

[root@localhost ~]# ./test_munmap_vdso_32 sysenter
[NOTE]    Using sysenter after munmap
    AT_SYSINFO_EHDR is 0xf7efe000
[NOTE]    unmapping vDSO: [0xf7efe000, 0xf7eff000]
[NOTE]    vDSO partial move failed, will try with bigger size [NOTE]    unmapping vDSO: [0xf7efe000, 0xf7f00000]
[OK]    32-bit process gets segfault on fast syscall with unmapped vDSO

Quoted 9 lines

---
  tools/testing/selftests/x86/.gitignore        |   1 + tools/testing/selftests/x86/Makefile          |  11 +- .../testing/selftests/x86/test_munmap_vdso.c  | 151 ++++++++++++++++++

  3 files changed, 158 insertions(+), 5 deletions(-)

  create mode 100644 tools/testing/selftests/x86/test_munmap_vdso.c

Quoted 2 lines

I can take this through kselftest tree for 5.14 - are there any dependencies on x86 tree, I should be aware of?

The test should work without other patches from the set. So I guess, it's good to go by it's own.

The only note I can make here is that without previous patches this part of the commit message is not exactly precise:

Quoted 2 lines

The test also can check force_sigsegv(SIGSEGV) in
do_fast_syscall_32()

I will still crash, but not by the kernel enforcement, rather with landing on the area where vdso was previously mapped.

Thanks,
Dmitry

maintainer Acked-by
#36 lore raw

Re: [PATCH v3 23/23] x86/vdso/selftest: Add a test for unmapping vDSO

Quoted 41 lines · Dmitry Safonov
Quoted 34 lines · Shuah Khan
Quoted 23 lines · Dmitry Safonov

Output for landing on x86:

Quoted 4 lines

[root@localhost ~]# ./test_munmap_vdso_64
    AT_SYSINFO_EHDR is 0x7fffead9f000
[NOTE]    unmapping vDSO: [0x7fffead9f000, 0x7fffeada0000] [NOTE]    vDSO partial move failed, will try with bigger size

Quoted 8 lines

[NOTE]    unmapping vDSO: [0x7fffead9f000, 0x7fffeada1000] [OK]
[root@localhost ~]# ./test_munmap_vdso_32
    AT_SYSINFO_EHDR is 0xf7eef000
[NOTE]    unmapping vDSO: [0xf7eef000, 0xf7ef0000] [NOTE]    vDSO partial move failed, will try with bigger size [NOTE]    unmapping vDSO: [0xf7eef000, 0xf7ef1000] [OK]

The test also can check force_sigsegv(SIGSEGV) in do_fast_syscall_32():

Quoted 7 lines

[root@localhost ~]# ./test_munmap_vdso_32 sysenter
[NOTE]    Using sysenter after munmap
    AT_SYSINFO_EHDR is 0xf7efe000
[NOTE]    unmapping vDSO: [0xf7efe000, 0xf7eff000]
[NOTE]    vDSO partial move failed, will try with bigger size [NOTE]    unmapping vDSO: [0xf7efe000, 0xf7f00000]
[OK]    32-bit process gets segfault on fast syscall with unmapped vDSO

Quoted 9 lines
---
   tools/testing/selftests/x86/.gitignore        |   1 +
   tools/testing/selftests/x86/Makefile          |  11 +-
   .../testing/selftests/x86/test_munmap_vdso.c  | 151 ++++++++++++++++++
   3 files changed, 158 insertions(+), 5 deletions(-)

  create mode 100644 tools/testing/selftests/x86/test_munmap_vdso.c

Quoted 2 lines

I can take this through kselftest tree for 5.14 - are there any dependencies on x86 tree, I should be aware of?

The test should work without other patches from the set. So I guess, it's good to go by it's own.

Quoted 8 lines

The only note I can make here is that without previous patches this part of the commit message is not exactly precise:

Quoted 2 lines

The test also can check force_sigsegv(SIGSEGV) in
do_fast_syscall_32()

I will still crash, but not by the kernel enforcement, rather with landing on the area where vdso was previously mapped.

Playing it safe, here is my Ack for it to go through x86 tree

thanks,
-- Shuah

#37 lore raw
In reply to Dmitry Safonov

Le 11/06/2021 à 20:02, Dmitry Safonov a écrit :

Quoted 9 lines

v3 Changes:
- Migrated arch/powerpc to vdso_base
- Added x86/selftest for unmapped vdso & no landing on fast syscall
- Review comments from Andy & Christophe (thanks!)
- Amended s/born process/execed process/ everywhere I noticed
- Build robot warning on cast from __user pointer

I've tested it on x86, I would appreciate any help with Tested-by on arm/arm64/mips/powerpc/s390/... platforms.

I tried it on powerpc, normal use still works.

What tests can be done exactly ?

We have a selftest in powerpc (https://github.com/linuxppc/linux/blob/master/tools/testing/selftests/powerpc/signal/sigreturn_vdso.c) but it doesn't work anymore since the split of VDSO into VDSO+VVAR.

Christophe

#38 lore raw
Quoted 10 lines · Christophe Leroy
Quoted 7 lines · Dmitry Safonov

v3 Changes:
- Migrated arch/powerpc to vdso_base
- Added x86/selftest for unmapped vdso & no landing on fast syscall
- Review comments from Andy & Christophe (thanks!)
- Amended s/born process/execed process/ everywhere I noticed
- Build robot warning on cast from __user pointer

Quoted 4 lines
Quoted 2 lines

I've tested it on x86, I would appreciate any help with Tested-by on arm/arm64/mips/powerpc/s390/... platforms.

I tried it on powerpc, normal use still works.

Thank you!

Quoted 1 line

What tests can be done exactly ?

Well, for x86 I've run all vdso tests from tools/testing/selftests/x86/ (with the new one from patch 23 here, which tests exactly forced segfault on unmapped vdso).
I think normal use on other platforms sounds good to me.

Quoted 4 lines

We have a selftest in powerpc
(https://github.com/linuxppc/linux/blob/master/tools/testing/selftests/powerpc/signal/sigreturn_vdso.c)
but it doesn't work anymore since the split of VDSO into VDSO+VVAR.

Well, it doesn't sound very hard to fix, see the sample diff inline.

Thank you,
Dmitry

--->8---

(quoted diff) +0 -0

b/tools/testing/selftests/powerpc/signal/sigreturn_vdso.c index e282fff0fe25..a4f85ee13c4a 100644

tools/testing/selftests/powerpc/signal/sigreturn_vdso.c +1 -1
@@ -13,6 +13,7 @@
 #include <signal.h>
 #include <stdlib.h>
 #include <string.h>
+#include <stdbool.h>
 #include <sys/mman.h>
 #include <sys/types.h>
 #include <unistd.h>
@@ -23,7 +24,7 @@
 
 #include "utils.h"
 
-static int search_proc_maps(char *needle, unsigned long *low,

unsigned long *high)
+static int search_proc_maps(char *needle, unsigned long *low, unsigned long *high, unsigned long *size)

 {
     unsigned long start, end;
     static char buf[4096];
@@ -52,6 +53,7 @@ static int search_proc_maps(char *needle, unsigned
long *low, unsigned long *hig
         if (strstr(name, needle)) {
             *low = start;
             *high = end - 1;
+            *size = end - start;
             rc = 0;
             break;
         }

@@ -71,9 +73,12 @@ static void sigusr1_handler(int sig)

int test_sigreturn_vdso(void)
{

-    unsigned long low, high, size;
+    unsigned long stack_start, stack_end, stack_size;
+    unsigned long vdso_start, vdso_end, vdso_size;
+    unsigned long vvar_start, vvar_end, vvar_size;
+    char *vdso_parking, *vvar_parking;
     struct sigaction act;

- char *p;
+ bool vvar_present;

     act.sa_handler = sigusr1_handler;
     act.sa_flags = 0;
@@ -82,36 +87,56 @@ int test_sigreturn_vdso(void)
     assert(sigaction(SIGUSR1, &act, NULL) == 0);

     // Confirm the VDSO is mapped, and work out where it is
-    assert(search_proc_maps("[vdso]", &low, &high) == 0);
-    size = high - low + 1;
-    printf("VDSO is at 0x%lx-0x%lx (%lu bytes)\n", low, high, size);
+    assert(search_proc_maps("[vdso]", &vdso_start, &vdso_end,

&vdso_size) == 0);
+ printf("VDSO is at 0x%lx-0x%lx (%lu bytes)\n", vdso_start,
vdso_end, vdso_size);
+ // On older kernels there's only vdso, on newer vdso/vvar pair
+ if (search_proc_maps("[vvar]", &vvar_start, &vvar_end, &vvar_size) == 0) {
+ vvar_present = true;
+ printf("VVAR is at 0x%lx-0x%lx (%lu bytes)\n",
+ vvar_start, vvar_end, vvar_size);
+ } else {
+ vvar_present = false;
+ vvar_size = 0;
+ }

     kill(getpid(), SIGUSR1);
     assert(took_signal == 1);
     printf("Signal delivered OK with VDSO mapped\n");

- // Remap the VDSO somewhere else
- p = mmap(NULL, size, PROT_READ|PROT_WRITE,
MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
- assert(p != MAP_FAILED);
- assert(mremap((void *)low, size, size,
MREMAP_MAYMOVE|MREMAP_FIXED, p) != MAP_FAILED);

-    assert(search_proc_maps("[vdso]", &low, &high) == 0);
-    size = high - low + 1;
-    printf("VDSO moved to 0x%lx-0x%lx (%lu bytes)\n", low, high, size);
+    // Remap the VDSO and VVAR somewhere else
+    vdso_parking = mmap(NULL, vdso_size + vvar_size,

PROT_READ|PROT_WRITE, MAP_ANONYMOUS|MAP_PRIVATE, -1, 0);
+ assert(vdso_parking != MAP_FAILED);
+
+ if (vvar_present) {
+ // The relative position of vdso/vvar must always stay the same
+ if (vvar_start > vdso_start) {
+ vvar_parking = vdso_parking + vdso_size;
+ } else {
+ vvar_parking = vdso_parking;
+ vdso_parking = vvar_parking + vvar_size;
+ }
+ assert(mremap((void *)vvar_start, vvar_size, vvar_size,
MREMAP_MAYMOVE|MREMAP_FIXED, vvar_parking) != MAP_FAILED);
+ }
+ assert(mremap((void *)vdso_start, vdso_size, vdso_size,
MREMAP_MAYMOVE|MREMAP_FIXED, vdso_parking) != MAP_FAILED); +
+ assert(search_proc_maps("[vdso]", &vdso_start, &vdso_end,
&vdso_size) == 0);
+ printf("VDSO moved to 0x%lx-0x%lx (%lu bytes)\n", vdso_start,
vdso_end, vdso_size);
+ assert(search_proc_maps("[vvar]", &vvar_start, &vvar_end,
&vvar_size) == 0);
+ printf("VVAR moved to 0x%lx-0x%lx (%lu bytes)\n", vvar_start,
vvar_end, vvar_size);

     kill(getpid(), SIGUSR1);
     assert(took_signal == 2);
     printf("Signal delivered OK with VDSO moved\n");

- assert(munmap((void *)low, size) == 0);

+    assert(munmap((void *)vdso_start, vdso_size) == 0);
     printf("Unmapped VDSO\n");

     // Confirm the VDSO is not mapped anymore

- assert(search_proc_maps("[vdso]", &low, &high) != 0);
+ assert(search_proc_maps("[vdso]", &vdso_start, &vdso_end,
&vdso_size) != 0);

     // Make the stack executable
-    assert(search_proc_maps("[stack]", &low, &high) == 0);
-    size = high - low + 1;
-    mprotect((void *)low, size, PROT_READ|PROT_WRITE|PROT_EXEC);
+    assert(search_proc_maps("[stack]", &stack_start, &stack_end,
&stack_size) == 0);
+    mprotect((void *)stack_start, stack_size, PROT_READ|PROT_WRITE|PROT_EXEC);
     printf("Remapped the stack executable\n");

     kill(getpid(), SIGUSR1);
#39 lore raw
In reply to Dmitry Safonov

Hi Dmitry,

I'm wondering the status of this series.

Wondering what to do while reviewing pending powerpc patches and especially https://patchwork.ozlabs.org/project/linuxppc-dev/patch/[email protected]/

Christophe

Le 11/06/2021 à 20:02, Dmitry Safonov a écrit :

Quoted 37 lines

v3 Changes:
- Migrated arch/powerpc to vdso_base
- Added x86/selftest for unmapped vdso & no landing on fast syscall
- Review comments from Andy & Christophe (thanks!)
- Amended s/born process/execed process/ everywhere I noticed
- Build robot warning on cast from __user pointer

I've tested it on x86, I would appreciate any help with Tested-by on arm/arm64/mips/powerpc/s390/... platforms.

One thing I've noticed while cooking this and haven't found a clean way to solve is zero-terminated .pages[] array in vdso mappings, which is not always zero-terminated but works by the reason of VM_DONTEXPAND on mappings.

v2 Changes:
- Rename user_landing to vdso_base as it tracks vDSO VMA start address, rather than the explicit address to land (Andy)
- Reword and don't use "new-execed" and "new-born" task (Andy)
- Fix failures reported by build robot

Started from discussion [1], where was noted that currently a couple of architectures support mremap() for vdso/sigpage, but not munmap(). If an application maps something on the ex-place of vdso/sigpage, later after processing signal it will land there (good luck!)

Patches set is based on linux-next (next-20201123) and it depends on changes in x86/cleanups (those reclaim TIF_IA32/TIF_X32) and also on my changes in akpm (fixing several mremap() issues).

Logically, the patches set divides on:
- patch 1: a cleanup for patches in x86/cleanups
- patches 2-13: cleanups for arch_setup_additional_pages()
- patches 13-14: x86 signal changes for unmapped vdso
- patches 15-22: provide generic vdso_base in mm_struct
- patch 23: selftest for unmapped vDSO & fast syscalls

Quoted 9 lines

In the end, besides cleanups, it's now more predictable what happens for applications with unmapped vdso on architectures those support .mremap() for vdso/sigpage.

I'm aware of only one user that unmaps vdso - Valgrind [2]. (there possibly are more, but this one is "special", it unmaps vdso, but
not vvar, which confuses CRIU [Checkpoint Restore In Userspace], that's why I'm aware of it)

I'm wondering the status of this series.

Wondering what to do while reviewing pending powerpc patches and especially https://patchwork.ozlabs.org/project/linuxppc-dev/patch/[email protected]/

Christophe

#40 lore raw

Hi Christophe,

Quoted 3 lines · Christophe Leroy

Hi Dmitry,

I'm wondering the status of this series.

Yeah, I plan to work on v4 addressing the reviews.
WFH has quite affected my work on side-projects and I've laid aside for a while this patch set that touches every architecture and is besides a bit challenging to upstream.

Quoted 3 lines

Wondering what to do while reviewing pending powerpc patches and especially
https://patchwork.ozlabs.org/project/linuxppc-dev/patch/[email protected]/

Please, go ahead with that - I'll base v4 patches on the top of that. Thanks for pinging me about this.

Quoted 50 lines

Christophe

Le 11/06/2021 à 20:02, Dmitry Safonov a écrit :

Quoted 37 lines

v3 Changes:
- Migrated arch/powerpc to vdso_base
- Added x86/selftest for unmapped vdso & no landing on fast syscall
- Review comments from Andy & Christophe (thanks!)
- Amended s/born process/execed process/ everywhere I noticed
- Build robot warning on cast from __user pointer

I've tested it on x86, I would appreciate any help with Tested-by on arm/arm64/mips/powerpc/s390/... platforms.

One thing I've noticed while cooking this and haven't found a clean way to solve is zero-terminated .pages[] array in vdso mappings, which is not always zero-terminated but works by the reason of VM_DONTEXPAND on mappings.

v2 Changes:
- Rename user_landing to vdso_base as it tracks vDSO VMA start address, rather than the explicit address to land (Andy)
- Reword and don't use "new-execed" and "new-born" task (Andy)
- Fix failures reported by build robot

Started from discussion [1], where was noted that currently a couple of architectures support mremap() for vdso/sigpage, but not munmap(). If an application maps something on the ex-place of vdso/sigpage, later after processing signal it will land there (good luck!)

Patches set is based on linux-next (next-20201123) and it depends on changes in x86/cleanups (those reclaim TIF_IA32/TIF_X32) and also on my changes in akpm (fixing several mremap() issues).

Logically, the patches set divides on:
- patch       1: a cleanup for patches in x86/cleanups
- patches  2-13: cleanups for arch_setup_additional_pages()
- patches 13-14: x86 signal changes for unmapped vdso
- patches 15-22: provide generic vdso_base in mm_struct
- patch      23: selftest for unmapped vDSO & fast syscalls

Quoted 9 lines

In the end, besides cleanups, it's now more predictable what happens for applications with unmapped vdso on architectures those support .mremap() for vdso/sigpage.

I'm aware of only one user that unmaps vdso - Valgrind [2]. (there possibly are more, but this one is "special", it unmaps vdso, but not vvar, which confuses CRIU [Checkpoint Restore In Userspace], that's why I'm aware of it)

Quoted 9 lines

I'm wondering the status of this series.

Wondering what to do while reviewing pending powerpc patches and especially
https://patchwork.ozlabs.org/project/linuxppc-dev/patch/[email protected]/

Christophe

Thanks,
Dmitry

#41 lore raw

Hi Dmitry,

Le 10/03/2022 à 22:17, Dmitry Safonov a écrit :

Quoted 11 lines

Hi Christophe,

Quoted 3 lines · Christophe Leroy

Hi Dmitry,

I'm wondering the status of this series.

Yeah, I plan to work on v4 addressing the reviews.
WFH has quite affected my work on side-projects and I've laid aside for a while this patch set that touches every architecture and is besides a bit challenging to upstream.

Any progress ?

Thanks
Christophe

#42 lore raw

Hi Christophe,

Quoted 8 lines · Christophe Leroy

Hi Dmitry,

Le 10/03/2022 à 22:17, Dmitry Safonov a écrit :

Quoted 5 lines

Hi Christophe,

Quoted 2 lines · Christophe Leroy

Hi Dmitry,

Quoted 8 lines
Quoted 6 lines
Quoted 1 line

I'm wondering the status of this series.

Yeah, I plan to work on v4 addressing the reviews.
WFH has quite affected my work on side-projects and I've laid aside for a while this patch set that touches every architecture and is besides a bit challenging to upstream.

Any progress ?

Yeah, I'm back to the office, so will spend some time on v4, thanks for pinging.

Thank you,
Dmitry

#43 lore raw

Hi Dmitry,

Le 23/08/2022 à 21:13, Dmitry Safonov a écrit :

Quoted 18 lines

Hi Christophe,

Quoted 15 lines · Christophe Leroy

Hi Dmitry,

Le 10/03/2022 à 22:17, Dmitry Safonov a écrit :

Quoted 11 lines

Hi Christophe,

Quoted 3 lines · Christophe Leroy

Hi Dmitry,

I'm wondering the status of this series.

Yeah, I plan to work on v4 addressing the reviews.
WFH has quite affected my work on side-projects and I've laid aside for a while this patch set that touches every architecture and is besides a bit challenging to upstream.

Quoted 5 lines
Quoted 1 line

Any progress ?

Yeah, I'm back to the office, so will spend some time on v4, thanks for pinging.

I haven't seen any v4, did I miss it ?

Christophe

#44 lore raw
In reply to Dmitry Safonov
Quoted 12 lines · Dmitry Safonov

Patches set is based on linux-next (next-20201123) and it depends on changes in x86/cleanups (those reclaim TIF_IA32/TIF_X32) and also on my changes in akpm (fixing several mremap() issues).

Logically, the patches set divides on:
- patch 1: a cleanup for patches in x86/cleanups
- patches 2-13: cleanups for arch_setup_additional_pages()
- patches 13-14: x86 signal changes for unmapped vdso
- patches 15-22: provide generic vdso_base in mm_struct
- patch 23: selftest for unmapped vDSO & fast syscalls

Quoted 9 lines

In the end, besides cleanups, it's now more predictable what happens for applications with unmapped vdso on architectures those support .mremap() for vdso/sigpage.

I'm aware of only one user that unmaps vdso - Valgrind [2]. (there possibly are more, but this one is "special", it unmaps vdso, but
not vvar, which confuses CRIU [Checkpoint Restore In Userspace], that's why I'm aware of it)

There was another discussion that might be relevant: actually associating the vdso with an actual file, and allowing a program to map said file normally if it want access to one that it normally wouldn't have (say, /proc/vdso/x86_64.so versus /proc/vdso/i386.so on the same system.)

The "catch", of course, is that this file will need to be mapped as MAP_SHARED because of vdso data.

-hpa