Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1376137 > unrolled thread

[PATCH] x86/vdso: add mremap hook to vm_special_mapping

Back to article view | Back to linux.kernel

Contents

  [PATCH] x86/vdso: add mremap hook to vm_special_mapping Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-11 18:00 +0200
    [PATCHv2] x86/vdso: add mremap hook to vm_special_mapping Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-14 18:40 +0200
      Re: [PATCHv2] x86/vdso: add mremap hook to vm_special_mapping Andy Lutomirski <luto@amacapital.net> - 2016-04-15 01:00 +0200
        Re: [PATCHv2] x86/vdso: add mremap hook to vm_special_mapping Ingo Molnar <mingo@kernel.org> - 2016-04-15 11:20 +0200
          Re: [PATCHv2] x86/vdso: add mremap hook to vm_special_mapping Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-15 12:00 +0200
            Re: [PATCHv2] x86/vdso: add mremap hook to vm_special_mapping Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-15 14:10 +0200
    [PATCHv3 1/2] x86/vdso: add mremap hook to vm_special_mapping Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-15 15:30 +0200
      [PATCHv3 2/2] x86: rename is_{ia32,x32}_task to in_{ia32,x32}_syscall Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-15 15:30 +0200
        Re: [PATCHv3 2/2] x86: rename is_{ia32,x32}_task to in_{ia32,x32}_syscall Andy Lutomirski <luto@amacapital.net> - 2016-04-15 19:00 +0200
    [PATCHv4 2/2] x86: rename is_{ia32,x32}_task to in_{ia32,x32}_syscall Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-15 16:30 +0200
    [PATCHv4 1/2] x86/vdso: add mremap hook to vm_special_mapping Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-15 16:30 +0200
      Re: [PATCHv4 1/2] x86/vdso: add mremap hook to vm_special_mapping Andy Lutomirski <luto@amacapital.net> - 2016-04-15 19:00 +0200
        Re: [PATCHv4 1/2] x86/vdso: add mremap hook to vm_special_mapping Andy Lutomirski <luto@amacapital.net> - 2016-04-18 17:40 +0200
    [PATCHv5 2/3] x86/vdso: add mremap hook to vm_special_mapping Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-18 15:50 +0200
      Re: [PATCHv5 2/3] x86/vdso: add mremap hook to vm_special_mapping kbuild test robot <lkp@intel.com> - 2016-04-18 16:10 +0200
      [PATCHv6 2/3] x86/vdso: add mremap hook to vm_special_mapping Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-18 16:20 +0200
      [PATCHv7 2/3] x86/vdso: add mremap hook to vm_special_mapping Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-18 16:30 +0200
        Re: [PATCHv7 2/3] x86/vdso: add mremap hook to vm_special_mapping Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-20 18:40 +0200
        Re: [PATCHv7 2/3] x86/vdso: add mremap hook to vm_special_mapping Andy Lutomirski <luto@amacapital.net> - 2016-04-21 22:00 +0200
      Re: [PATCHv5 2/3] x86/vdso: add mremap hook to vm_special_mapping kbuild test robot <lkp@intel.com> - 2016-04-24 01:20 +0200
    [PATCHv5 1/3] x86: rename is_{ia32,x32}_task to in_{ia32,x32}_syscall Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-18 15:50 +0200
      [PATCHv5 3/3] selftest/x86: add mremap vdso 32-bit test Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-18 15:50 +0200
        Re: [PATCHv5 3/3] selftest/x86: add mremap vdso 32-bit test Andy Lutomirski <luto@amacapital.net> - 2016-04-21 22:10 +0200
      [tip:x86/asm] x86/entry: Rename is_{ia32,x32}_task() to  in_{ia32,x32}_syscall() tip-bot for Dmitry Safonov <tipbot@zytor.com> - 2016-04-19 11:40 +0200
        Re: [tip:x86/asm] x86/entry: Rename is_{ia32,x32}_task() to  in_{ia32,x32}_syscall() Ingo Molnar <mingo@kernel.org> - 2016-04-19 13:20 +0200
          Re: [tip:x86/asm] x86/entry: Rename is_{ia32,x32}_task() to  in_{ia32,x32}_syscall() Borislav Petkov <bp@alien8.de> - 2016-04-19 13:40 +0200
            Re: [tip:x86/asm] x86/entry: Rename is_{ia32,x32}_task() to in_{ia32,x32}_syscall() "H. Peter Anvin" <hpa@zytor.com> - 2016-04-19 19:10 +0200
          Re: [tip:x86/asm] x86/entry: Rename is_{ia32,x32}_task() to in_{ia32,x32}_syscall() Andy Lutomirski <luto@amacapital.net> - 2016-04-19 18:10 +0200
            Re: [tip:x86/asm] x86/entry: Rename is_{ia32,x32}_task() to  in_{ia32,x32}_syscall() Ingo Molnar <mingo@kernel.org> - 2016-04-22 10:40 +0200
    [PATCHv8 2/2] selftest/x86: add mremap vdso test Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-25 13:40 +0200
      Re: [PATCHv8 2/2] selftest/x86: add mremap vdso test Andy Lutomirski <luto@amacapital.net> - 2016-04-25 23:40 +0200
    [PATCHv8 1/2] x86/vdso: add mremap hook to vm_special_mapping Dmitry Safonov <dsafonov@virtuozzo.com> - 2016-04-25 14:00 +0200
      Re: [PATCHv8 1/2] x86/vdso: add mremap hook to vm_special_mapping Andy Lutomirski <luto@amacapital.net> - 2016-04-25 23:40 +0200

Page 1 of 2  [1] 2  Next page →

#1376137 — [PATCH] x86/vdso: add mremap hook to vm_special_mapping

Add possibility for userspace 32-bit applications to move
vdso mapping. Previously, when userspace app called
mremap for vdso, in return path it would land on previous
address of vdso page, resulting in segmentation violation.
Now it lands fine and returns to userspace with remapped vdso.

Renamed and moved text_mapping structure declaration inside
map_vdso, as it used only there and now it complement
vvar_mapping variable.

There is still problem for remapping vdso in glibc applications:
linker relocates addresses for syscalls on vdso page, so
you need to relink with the new addresses. Or the next syscall
through glibc may fail:
  Program received signal SIGSEGV, Segmentation fault.
  #0  0xf7fd9b80 in __kernel_vsyscall ()
  #1  0xf7ec8238 in _exit () from /usr/lib32/libc.so.6

Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
---
 arch/x86/entry/vdso/vma.c | 33 ++++++++++++++++++++++++++++-----
 include/linux/mm_types.h  |  3 +++
 mm/mmap.c                 | 10 ++++++++++
 3 files changed, 41 insertions(+), 5 deletions(-)

diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
index 10f704584922..08ac59907cde 100644
--- a/arch/x86/entry/vdso/vma.c
+++ b/arch/x86/entry/vdso/vma.c
@@ -12,6 +12,7 @@
 #include <linux/random.h>
 #include <linux/elf.h>
 #include <linux/cpu.h>
+#include <linux/ptrace.h>
 #include <asm/pvclock.h>
 #include <asm/vgtod.h>
 #include <asm/proto.h>
@@ -98,10 +99,26 @@ static int vdso_fault(const struct vm_special_mapping *sm,
 	return 0;
 }
 
-static const struct vm_special_mapping text_mapping = {
-	.name = "[vdso]",
-	.fault = vdso_fault,
-};
+static int vdso_mremap(const struct vm_special_mapping *sm,
+		      struct vm_area_struct *new_vma)
+{
+	struct pt_regs *regs = current_pt_regs();
+
+#if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
+	/* Fixing userspace landing - look at do_fast_syscall_32 */
+	if (regs->ip == (unsigned long)current->mm->context.vdso +
+			vdso_image_32.sym_int80_landing_pad
+#ifdef CONFIG_IA32_EMULATION
+		&& current_thread_info()->status & TS_COMPAT
+#endif
+	   )
+		regs->ip = new_vma->vm_start +
+			vdso_image_32.sym_int80_landing_pad;
+#endif
+	new_vma->vm_mm->context.vdso = (void __user *)new_vma->vm_start;
+
+	return 0;
+}
 
 static int vvar_fault(const struct vm_special_mapping *sm,
 		      struct vm_area_struct *vma, struct vm_fault *vmf)
@@ -162,6 +179,12 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 	struct vm_area_struct *vma;
 	unsigned long addr, text_start;
 	int ret = 0;
+
+	static const struct vm_special_mapping vdso_mapping = {
+		.name = "[vdso]",
+		.fault = vdso_fault,
+		.mremap = vdso_mremap,
+	};
 	static const struct vm_special_mapping vvar_mapping = {
 		.name = "[vvar]",
 		.fault = vvar_fault,
@@ -195,7 +218,7 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 				       image->size,
 				       VM_READ|VM_EXEC|
 				       VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
-				       &text_mapping);
+				       &vdso_mapping);
 
 	if (IS_ERR(vma)) {
 		ret = PTR_ERR(vma);
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index c2d75b4fa86c..4d16ab9287af 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -586,6 +586,9 @@ struct vm_special_mapping {
 	int (*fault)(const struct vm_special_mapping *sm,
 		     struct vm_area_struct *vma,
 		     struct vm_fault *vmf);
+
+	int (*mremap)(const struct vm_special_mapping *sm,
+		     struct vm_area_struct *new_vma);
 };
 
 enum tlb_flush_reason {
diff --git a/mm/mmap.c b/mm/mmap.c
index bd2e1a533bc1..ba71658dd1a1 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2930,9 +2930,19 @@ 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)
+{
+	struct vm_special_mapping *sm = new_vma->vm_private_data;
+
+	if (sm->mremap)
+		return sm->mremap(sm, new_vma);
+	return 0;
+}
+
 static const struct vm_operations_struct special_mapping_vmops = {
 	.close = special_mapping_close,
 	.fault = special_mapping_fault,
+	.mremap = special_mapping_mremap,
 	.name = special_mapping_name,
 };
 
-- 
2.8.0

[toc] | [next] | [standalone]

#1379084 — [PATCHv2] x86/vdso: add mremap hook to vm_special_mapping

Add possibility for userspace 32-bit applications to move
vdso mapping. Previously, when userspace app called
mremap for vdso, in return path it would land on previous
address of vdso page, resulting in segmentation violation.
Now it lands fine and returns to userspace with remapped vdso.
This will also fix context.vdso pointer for 64-bit, which does not
affect the user of vdso after mremap by now, but this may change.

Renamed and moved text_mapping structure declaration inside
map_vdso, as it used only there and now it complement
vvar_mapping variable.

There is still problem for remapping vdso in 32-bit glibc applications:
linker relocates addresses for syscalls on vdso page, so
you need to relink with the new addresses. Or the next syscall
through glibc may fail:
  Program received signal SIGSEGV, Segmentation fault.
  #0  0xf7fd9b80 in __kernel_vsyscall ()
  #1  0xf7ec8238 in _exit () from /usr/lib32/libc.so.6

Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
---
v2: added __maybe_unused for pt_regs in vdso_mremap

 arch/x86/entry/vdso/vma.c | 33 ++++++++++++++++++++++++++++-----
 include/linux/mm_types.h  |  3 +++
 mm/mmap.c                 | 10 ++++++++++
 3 files changed, 41 insertions(+), 5 deletions(-)

diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
index 10f704584922..7e261e2554c8 100644
--- a/arch/x86/entry/vdso/vma.c
+++ b/arch/x86/entry/vdso/vma.c
@@ -12,6 +12,7 @@
 #include <linux/random.h>
 #include <linux/elf.h>
 #include <linux/cpu.h>
+#include <linux/ptrace.h>
 #include <asm/pvclock.h>
 #include <asm/vgtod.h>
 #include <asm/proto.h>
@@ -98,10 +99,26 @@ static int vdso_fault(const struct vm_special_mapping *sm,
 	return 0;
 }
 
-static const struct vm_special_mapping text_mapping = {
-	.name = "[vdso]",
-	.fault = vdso_fault,
-};
+static int vdso_mremap(const struct vm_special_mapping *sm,
+		      struct vm_area_struct *new_vma)
+{
+	struct pt_regs __maybe_unused *regs = current_pt_regs();
+
+#if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
+	/* Fixing userspace landing - look at do_fast_syscall_32 */
+	if (regs->ip == (unsigned long)current->mm->context.vdso +
+			vdso_image_32.sym_int80_landing_pad
+#ifdef CONFIG_IA32_EMULATION
+		&& current_thread_info()->status & TS_COMPAT
+#endif
+	   )
+		regs->ip = new_vma->vm_start +
+			vdso_image_32.sym_int80_landing_pad;
+#endif
+	new_vma->vm_mm->context.vdso = (void __user *)new_vma->vm_start;
+
+	return 0;
+}
 
 static int vvar_fault(const struct vm_special_mapping *sm,
 		      struct vm_area_struct *vma, struct vm_fault *vmf)
@@ -162,6 +179,12 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 	struct vm_area_struct *vma;
 	unsigned long addr, text_start;
 	int ret = 0;
+
+	static const struct vm_special_mapping vdso_mapping = {
+		.name = "[vdso]",
+		.fault = vdso_fault,
+		.mremap = vdso_mremap,
+	};
 	static const struct vm_special_mapping vvar_mapping = {
 		.name = "[vvar]",
 		.fault = vvar_fault,
@@ -195,7 +218,7 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 				       image->size,
 				       VM_READ|VM_EXEC|
 				       VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
-				       &text_mapping);
+				       &vdso_mapping);
 
 	if (IS_ERR(vma)) {
 		ret = PTR_ERR(vma);
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index c2d75b4fa86c..4d16ab9287af 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -586,6 +586,9 @@ struct vm_special_mapping {
 	int (*fault)(const struct vm_special_mapping *sm,
 		     struct vm_area_struct *vma,
 		     struct vm_fault *vmf);
+
+	int (*mremap)(const struct vm_special_mapping *sm,
+		     struct vm_area_struct *new_vma);
 };
 
 enum tlb_flush_reason {
diff --git a/mm/mmap.c b/mm/mmap.c
index bd2e1a533bc1..ba71658dd1a1 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2930,9 +2930,19 @@ 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)
+{
+	struct vm_special_mapping *sm = new_vma->vm_private_data;
+
+	if (sm->mremap)
+		return sm->mremap(sm, new_vma);
+	return 0;
+}
+
 static const struct vm_operations_struct special_mapping_vmops = {
 	.close = special_mapping_close,
 	.fault = special_mapping_fault,
+	.mremap = special_mapping_mremap,
 	.name = special_mapping_name,
 };
 
-- 
2.8.0

[toc] | [prev] | [next] | [standalone]

#1379358 — Re: [PATCHv2] x86/vdso: add mremap hook to vm_special_mapping

On Thu, Apr 14, 2016 at 9:32 AM, Dmitry Safonov <dsafonov@virtuozzo.com> wrote:
> Add possibility for userspace 32-bit applications to move
> vdso mapping. Previously, when userspace app called
> mremap for vdso, in return path it would land on previous
> address of vdso page, resulting in segmentation violation.
> Now it lands fine and returns to userspace with remapped vdso.
> This will also fix context.vdso pointer for 64-bit, which does not
> affect the user of vdso after mremap by now, but this may change.
>
> Renamed and moved text_mapping structure declaration inside
> map_vdso, as it used only there and now it complement
> vvar_mapping variable.
>
> There is still problem for remapping vdso in 32-bit glibc applications:
> linker relocates addresses for syscalls on vdso page, so
> you need to relink with the new addresses. Or the next syscall
> through glibc may fail:
>   Program received signal SIGSEGV, Segmentation fault.
>   #0  0xf7fd9b80 in __kernel_vsyscall ()
>   #1  0xf7ec8238 in _exit () from /usr/lib32/libc.so.6
>
> Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
> ---
> v2: added __maybe_unused for pt_regs in vdso_mremap
>
>  arch/x86/entry/vdso/vma.c | 33 ++++++++++++++++++++++++++++-----
>  include/linux/mm_types.h  |  3 +++
>  mm/mmap.c                 | 10 ++++++++++
>  3 files changed, 41 insertions(+), 5 deletions(-)
>
> diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
> index 10f704584922..7e261e2554c8 100644
> --- a/arch/x86/entry/vdso/vma.c
> +++ b/arch/x86/entry/vdso/vma.c
> @@ -12,6 +12,7 @@
>  #include <linux/random.h>
>  #include <linux/elf.h>
>  #include <linux/cpu.h>
> +#include <linux/ptrace.h>
>  #include <asm/pvclock.h>
>  #include <asm/vgtod.h>
>  #include <asm/proto.h>
> @@ -98,10 +99,26 @@ static int vdso_fault(const struct vm_special_mapping *sm,
>         return 0;
>  }
>
> -static const struct vm_special_mapping text_mapping = {
> -       .name = "[vdso]",
> -       .fault = vdso_fault,
> -};
> +static int vdso_mremap(const struct vm_special_mapping *sm,
> +                     struct vm_area_struct *new_vma)
> +{
> +       struct pt_regs __maybe_unused *regs = current_pt_regs();
> +
> +#if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
> +       /* Fixing userspace landing - look at do_fast_syscall_32 */
> +       if (regs->ip == (unsigned long)current->mm->context.vdso +
> +                       vdso_image_32.sym_int80_landing_pad
> +#ifdef CONFIG_IA32_EMULATION
> +               && current_thread_info()->status & TS_COMPAT
> +#endif

Instead of ifdef, use the (grossly misnamed) is_ia32_task() helper for
this, please.

> +          )
> +               regs->ip = new_vma->vm_start +
> +                       vdso_image_32.sym_int80_landing_pad;
> +#endif
> +       new_vma->vm_mm->context.vdso = (void __user *)new_vma->vm_start;

Can you arrange for the mremap call to fail if the old mapping gets
split?  This might be as simple as confirming that the new mapping's
length is what we expect it to be and, if it isn't, returning -EINVAL.

If anyone things that might break some existing application (which is
quite unlikely), then we could allow mremap to succeed but skip the
part where we change context.vdso and rip.

> +
> +       return 0;
> +}
>
>  static int vvar_fault(const struct vm_special_mapping *sm,
>                       struct vm_area_struct *vma, struct vm_fault *vmf)
> @@ -162,6 +179,12 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
>         struct vm_area_struct *vma;
>         unsigned long addr, text_start;
>         int ret = 0;
> +
> +       static const struct vm_special_mapping vdso_mapping = {
> +               .name = "[vdso]",
> +               .fault = vdso_fault,
> +               .mremap = vdso_mremap,
> +       };

Why did you add this instead of modifying text_mapping?

--Andy

[toc] | [prev] | [next] | [standalone]

#1379609 — Re: [PATCHv2] x86/vdso: add mremap hook to vm_special_mapping

* Andy Lutomirski <luto@amacapital.net> wrote:

> > +       if (regs->ip == (unsigned long)current->mm->context.vdso +
> > +                       vdso_image_32.sym_int80_landing_pad
> > +#ifdef CONFIG_IA32_EMULATION
> > +               && current_thread_info()->status & TS_COMPAT
> > +#endif
> 
> Instead of ifdef, use the (grossly misnamed) is_ia32_task() helper for
> this, please.

Please also let's do the rename.

Thanks,

	Ingo

[toc] | [prev] | [next] | [standalone]

#1379659 — Re: [PATCHv2] x86/vdso: add mremap hook to vm_special_mapping

On 04/15/2016 12:18 PM, Ingo Molnar wrote:
> * Andy Lutomirski <luto@amacapital.net> wrote:
>> Instead of ifdef, use the (grossly misnamed) is_ia32_task() helper for
>> this, please.
> Please also let's do the rename.
Does `is_32bit_syscall` sounds right, or shall it be `is_32bit_task`?
I think, `is_compat_task` will be bad-named for X86_32 host.

-- 
Regards,
Dmitry Safonov

[toc] | [prev] | [next] | [standalone]

#1379776 — Re: [PATCHv2] x86/vdso: add mremap hook to vm_special_mapping

On 04/15/2016 12:51 PM, Dmitry Safonov wrote:
> On 04/15/2016 12:18 PM, Ingo Molnar wrote:
>> * Andy Lutomirski <luto@amacapital.net> wrote:
>>> Instead of ifdef, use the (grossly misnamed) is_ia32_task() helper for
>>> this, please.
>> Please also let's do the rename.
> Does `is_32bit_syscall` sounds right, or shall it be `is_32bit_task`?
> I think, `is_compat_task` will be bad-named for X86_32 host.
>
Or maybe, better:
is_x32_task => in_x32_syscall
is_ia32_task => in_ia32_syscall
as existing in_compat_syscall().

Looks good?

-- 
Regards,
Dmitry Safonov

[toc] | [prev] | [next] | [standalone]

#1379811 — [PATCHv3 1/2] x86/vdso: add mremap hook to vm_special_mapping

Add possibility for userspace 32-bit applications to move
vdso mapping. Previously, when userspace app called
mremap for vdso, in return path it would land on previous
address of vdso page, resulting in segmentation violation.
Now it lands fine and returns to userspace with remapped vdso.
This will also fix context.vdso pointer for 64-bit, which does not
affect the user of vdso after mremap by now, but this may change.

As suggested by Andy, return EINVAL for mremap that splits vdso image.

Renamed and moved text_mapping structure declaration inside
map_vdso, as it used only there and now it complement
vvar_mapping variable.

There is still problem for remapping vdso in glibc applications:
linker relocates addresses for syscalls on vdso page, so
you need to relink with the new addresses. Or the next syscall
through glibc may fail:
  Program received signal SIGSEGV, Segmentation fault.
  #0  0xf7fd9b80 in __kernel_vsyscall ()
  #1  0xf7ec8238 in _exit () from /usr/lib32/libc.so.6

Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
---
v3: as Andy suggested, return EINVAL in case of splitting vdso blob on mremap;
    used is_ia32_task instead of ifdefs 
v2: added __maybe_unused for pt_regs in vdso_mremap

 arch/x86/entry/vdso/vma.c | 36 +++++++++++++++++++++++++++++++-----
 include/linux/mm_types.h  |  3 +++
 mm/mmap.c                 | 10 ++++++++++
 3 files changed, 44 insertions(+), 5 deletions(-)

diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
index 10f704584922..8510b1b55b21 100644
--- a/arch/x86/entry/vdso/vma.c
+++ b/arch/x86/entry/vdso/vma.c
@@ -12,6 +12,7 @@
 #include <linux/random.h>
 #include <linux/elf.h>
 #include <linux/cpu.h>
+#include <linux/ptrace.h>
 #include <asm/pvclock.h>
 #include <asm/vgtod.h>
 #include <asm/proto.h>
@@ -98,10 +99,29 @@ static int vdso_fault(const struct vm_special_mapping *sm,
 	return 0;
 }
 
-static const struct vm_special_mapping text_mapping = {
-	.name = "[vdso]",
-	.fault = vdso_fault,
-};
+static int vdso_mremap(const struct vm_special_mapping *sm,
+		      struct vm_area_struct *new_vma)
+{
+	struct pt_regs __maybe_unused *regs = current_pt_regs();
+	unsigned long new_size = new_vma->vm_end - new_vma->vm_start;
+	const struct vdso_image *image = current->mm->context.vdso_image;
+
+	if (image->size != new_size)
+		return -EINVAL;
+
+	if (is_ia32_task()) {
+		unsigned long vdso_land = vdso_image_32.sym_int80_landing_pad;
+		unsigned long old_land_addr = vdso_land +
+			(unsigned long)current->mm->context.vdso;
+
+		/* Fixing userspace landing - look at do_fast_syscall_32 */
+		if (regs->ip == old_land_addr)
+			regs->ip = new_vma->vm_start + vdso_land;
+	}
+	new_vma->vm_mm->context.vdso = (void __user *)new_vma->vm_start;
+
+	return 0;
+}
 
 static int vvar_fault(const struct vm_special_mapping *sm,
 		      struct vm_area_struct *vma, struct vm_fault *vmf)
@@ -162,6 +182,12 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 	struct vm_area_struct *vma;
 	unsigned long addr, text_start;
 	int ret = 0;
+
+	static const struct vm_special_mapping vdso_mapping = {
+		.name = "[vdso]",
+		.fault = vdso_fault,
+		.mremap = vdso_mremap,
+	};
 	static const struct vm_special_mapping vvar_mapping = {
 		.name = "[vvar]",
 		.fault = vvar_fault,
@@ -195,7 +221,7 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 				       image->size,
 				       VM_READ|VM_EXEC|
 				       VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
-				       &text_mapping);
+				       &vdso_mapping);
 
 	if (IS_ERR(vma)) {
 		ret = PTR_ERR(vma);
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index c2d75b4fa86c..4d16ab9287af 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -586,6 +586,9 @@ struct vm_special_mapping {
 	int (*fault)(const struct vm_special_mapping *sm,
 		     struct vm_area_struct *vma,
 		     struct vm_fault *vmf);
+
+	int (*mremap)(const struct vm_special_mapping *sm,
+		     struct vm_area_struct *new_vma);
 };
 
 enum tlb_flush_reason {
diff --git a/mm/mmap.c b/mm/mmap.c
index bd2e1a533bc1..ba71658dd1a1 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2930,9 +2930,19 @@ 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)
+{
+	struct vm_special_mapping *sm = new_vma->vm_private_data;
+
+	if (sm->mremap)
+		return sm->mremap(sm, new_vma);
+	return 0;
+}
+
 static const struct vm_operations_struct special_mapping_vmops = {
 	.close = special_mapping_close,
 	.fault = special_mapping_fault,
+	.mremap = special_mapping_mremap,
 	.name = special_mapping_name,
 };
 
-- 
2.8.0

[toc] | [prev] | [next] | [standalone]

#1379816 — [PATCHv3 2/2] x86: rename is_{ia32,x32}_task to in_{ia32,x32}_syscall

Impact: clearify meaning

Suggested-by: Andy Lutomirski <luto@amacapital.net>
Suggested-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
---
v3: initial patch

 arch/x86/entry/common.c            | 2 +-
 arch/x86/entry/vdso/vma.c          | 2 +-
 arch/x86/include/asm/compat.h      | 4 ++--
 arch/x86/include/asm/thread_info.h | 2 +-
 arch/x86/kernel/process_64.c       | 2 +-
 arch/x86/kernel/ptrace.c           | 2 +-
 arch/x86/kernel/signal.c           | 2 +-
 arch/x86/kernel/uprobes.c          | 2 +-
 8 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/arch/x86/entry/common.c b/arch/x86/entry/common.c
index e79d93d44ecd..ec138e538c44 100644
--- a/arch/x86/entry/common.c
+++ b/arch/x86/entry/common.c
@@ -191,7 +191,7 @@ long syscall_trace_enter_phase2(struct pt_regs *regs, u32 arch,
 
 long syscall_trace_enter(struct pt_regs *regs)
 {
-	u32 arch = is_ia32_task() ? AUDIT_ARCH_I386 : AUDIT_ARCH_X86_64;
+	u32 arch = in_ia32_syscall() ? AUDIT_ARCH_I386 : AUDIT_ARCH_X86_64;
 	unsigned long phase1_result = syscall_trace_enter_phase1(regs, arch);
 
 	if (phase1_result == 0)
diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
index 8510b1b55b21..0b861fc274b6 100644
--- a/arch/x86/entry/vdso/vma.c
+++ b/arch/x86/entry/vdso/vma.c
@@ -109,7 +109,7 @@ static int vdso_mremap(const struct vm_special_mapping *sm,
 	if (image->size != new_size)
 		return -EINVAL;
 
-	if (is_ia32_task()) {
+	if (in_ia32_syscall()) {
 		unsigned long vdso_land = vdso_image_32.sym_int80_landing_pad;
 		unsigned long old_land_addr = vdso_land +
 			(unsigned long)current->mm->context.vdso;
diff --git a/arch/x86/include/asm/compat.h b/arch/x86/include/asm/compat.h
index ebb102e1bbc7..5a3b2c119ed0 100644
--- a/arch/x86/include/asm/compat.h
+++ b/arch/x86/include/asm/compat.h
@@ -307,7 +307,7 @@ static inline void __user *arch_compat_alloc_user_space(long len)
 	return (void __user *)round_down(sp - len, 16);
 }
 
-static inline bool is_x32_task(void)
+static inline bool in_x32_syscall(void)
 {
 #ifdef CONFIG_X86_X32_ABI
 	if (task_pt_regs(current)->orig_ax & __X32_SYSCALL_BIT)
@@ -318,7 +318,7 @@ static inline bool is_x32_task(void)
 
 static inline bool in_compat_syscall(void)
 {
-	return is_ia32_task() || is_x32_task();
+	return in_ia32_syscall() || in_x32_syscall();
 }
 #define in_compat_syscall in_compat_syscall	/* override the generic impl */
 
diff --git a/arch/x86/include/asm/thread_info.h b/arch/x86/include/asm/thread_info.h
index ffae84df8a93..30c133ac05cd 100644
--- a/arch/x86/include/asm/thread_info.h
+++ b/arch/x86/include/asm/thread_info.h
@@ -255,7 +255,7 @@ static inline bool test_and_clear_restore_sigmask(void)
 	return true;
 }
 
-static inline bool is_ia32_task(void)
+static inline bool in_ia32_syscall(void)
 {
 #ifdef CONFIG_X86_32
 	return true;
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 6cbab31ac23a..4a62ec457b56 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -210,7 +210,7 @@ int copy_thread_tls(unsigned long clone_flags, unsigned long sp,
 	 */
 	if (clone_flags & CLONE_SETTLS) {
 #ifdef CONFIG_IA32_EMULATION
-		if (is_ia32_task())
+		if (in_ia32_syscall())
 			err = do_set_thread_area(p, -1,
 				(struct user_desc __user *)tls, 0);
 		else
diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c
index 32e9d9cbb884..0f4d2a5df2dc 100644
--- a/arch/x86/kernel/ptrace.c
+++ b/arch/x86/kernel/ptrace.c
@@ -1266,7 +1266,7 @@ long compat_arch_ptrace(struct task_struct *child, compat_long_t request,
 			compat_ulong_t caddr, compat_ulong_t cdata)
 {
 #ifdef CONFIG_X86_X32_ABI
-	if (!is_ia32_task())
+	if (!in_ia32_syscall())
 		return x32_arch_ptrace(child, request, caddr, cdata);
 #endif
 #ifdef CONFIG_IA32_EMULATION
diff --git a/arch/x86/kernel/signal.c b/arch/x86/kernel/signal.c
index 548ddf7d6fd2..aa31265aa61d 100644
--- a/arch/x86/kernel/signal.c
+++ b/arch/x86/kernel/signal.c
@@ -762,7 +762,7 @@ handle_signal(struct ksignal *ksig, struct pt_regs *regs)
 static inline unsigned long get_nr_restart_syscall(const struct pt_regs *regs)
 {
 #ifdef CONFIG_X86_64
-	if (is_ia32_task())
+	if (in_ia32_syscall())
 		return __NR_ia32_restart_syscall;
 #endif
 #ifdef CONFIG_X86_X32_ABI
diff --git a/arch/x86/kernel/uprobes.c b/arch/x86/kernel/uprobes.c
index bf4db6eaec8f..98b4dc87628b 100644
--- a/arch/x86/kernel/uprobes.c
+++ b/arch/x86/kernel/uprobes.c
@@ -516,7 +516,7 @@ struct uprobe_xol_ops {
 
 static inline int sizeof_long(void)
 {
-	return is_ia32_task() ? 4 : 8;
+	return in_ia32_syscall() ? 4 : 8;
 }
 
 static int default_pre_xol_op(struct arch_uprobe *auprobe, struct pt_regs *regs)
-- 
2.8.0

[toc] | [prev] | [next] | [standalone]

#1380017 — Re: [PATCHv3 2/2] x86: rename is_{ia32,x32}_task to in_{ia32,x32}_syscall

On Fri, Apr 15, 2016 at 6:20 AM, Dmitry Safonov <dsafonov@virtuozzo.com> wrote:
> Impact: clearify meaning
>
> Suggested-by: Andy Lutomirski <luto@amacapital.net>
> Suggested-by: Ingo Molnar <mingo@kernel.org>
> Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>

Acked-by: Andy Lutomirski <luto@kernel.org>

But if you resubmit, please consider making this patch 1 so Ingo can
apply it directly.

--Andy

> ---
> v3: initial patch
>
>  arch/x86/entry/common.c            | 2 +-
>  arch/x86/entry/vdso/vma.c          | 2 +-
>  arch/x86/include/asm/compat.h      | 4 ++--
>  arch/x86/include/asm/thread_info.h | 2 +-
>  arch/x86/kernel/process_64.c       | 2 +-
>  arch/x86/kernel/ptrace.c           | 2 +-
>  arch/x86/kernel/signal.c           | 2 +-
>  arch/x86/kernel/uprobes.c          | 2 +-
>  8 files changed, 9 insertions(+), 9 deletions(-)
>
> diff --git a/arch/x86/entry/common.c b/arch/x86/entry/common.c
> index e79d93d44ecd..ec138e538c44 100644
> --- a/arch/x86/entry/common.c
> +++ b/arch/x86/entry/common.c
> @@ -191,7 +191,7 @@ long syscall_trace_enter_phase2(struct pt_regs *regs, u32 arch,
>
>  long syscall_trace_enter(struct pt_regs *regs)
>  {
> -       u32 arch = is_ia32_task() ? AUDIT_ARCH_I386 : AUDIT_ARCH_X86_64;
> +       u32 arch = in_ia32_syscall() ? AUDIT_ARCH_I386 : AUDIT_ARCH_X86_64;
>         unsigned long phase1_result = syscall_trace_enter_phase1(regs, arch);
>
>         if (phase1_result == 0)
> diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
> index 8510b1b55b21..0b861fc274b6 100644
> --- a/arch/x86/entry/vdso/vma.c
> +++ b/arch/x86/entry/vdso/vma.c
> @@ -109,7 +109,7 @@ static int vdso_mremap(const struct vm_special_mapping *sm,
>         if (image->size != new_size)
>                 return -EINVAL;
>
> -       if (is_ia32_task()) {
> +       if (in_ia32_syscall()) {
>                 unsigned long vdso_land = vdso_image_32.sym_int80_landing_pad;
>                 unsigned long old_land_addr = vdso_land +
>                         (unsigned long)current->mm->context.vdso;
> diff --git a/arch/x86/include/asm/compat.h b/arch/x86/include/asm/compat.h
> index ebb102e1bbc7..5a3b2c119ed0 100644
> --- a/arch/x86/include/asm/compat.h
> +++ b/arch/x86/include/asm/compat.h
> @@ -307,7 +307,7 @@ static inline void __user *arch_compat_alloc_user_space(long len)
>         return (void __user *)round_down(sp - len, 16);
>  }
>
> -static inline bool is_x32_task(void)
> +static inline bool in_x32_syscall(void)
>  {
>  #ifdef CONFIG_X86_X32_ABI
>         if (task_pt_regs(current)->orig_ax & __X32_SYSCALL_BIT)
> @@ -318,7 +318,7 @@ static inline bool is_x32_task(void)
>
>  static inline bool in_compat_syscall(void)
>  {
> -       return is_ia32_task() || is_x32_task();
> +       return in_ia32_syscall() || in_x32_syscall();
>  }
>  #define in_compat_syscall in_compat_syscall    /* override the generic impl */
>
> diff --git a/arch/x86/include/asm/thread_info.h b/arch/x86/include/asm/thread_info.h
> index ffae84df8a93..30c133ac05cd 100644
> --- a/arch/x86/include/asm/thread_info.h
> +++ b/arch/x86/include/asm/thread_info.h
> @@ -255,7 +255,7 @@ static inline bool test_and_clear_restore_sigmask(void)
>         return true;
>  }
>
> -static inline bool is_ia32_task(void)
> +static inline bool in_ia32_syscall(void)
>  {
>  #ifdef CONFIG_X86_32
>         return true;
> diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
> index 6cbab31ac23a..4a62ec457b56 100644
> --- a/arch/x86/kernel/process_64.c
> +++ b/arch/x86/kernel/process_64.c
> @@ -210,7 +210,7 @@ int copy_thread_tls(unsigned long clone_flags, unsigned long sp,
>          */
>         if (clone_flags & CLONE_SETTLS) {
>  #ifdef CONFIG_IA32_EMULATION
> -               if (is_ia32_task())
> +               if (in_ia32_syscall())
>                         err = do_set_thread_area(p, -1,
>                                 (struct user_desc __user *)tls, 0);
>                 else
> diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c
> index 32e9d9cbb884..0f4d2a5df2dc 100644
> --- a/arch/x86/kernel/ptrace.c
> +++ b/arch/x86/kernel/ptrace.c
> @@ -1266,7 +1266,7 @@ long compat_arch_ptrace(struct task_struct *child, compat_long_t request,
>                         compat_ulong_t caddr, compat_ulong_t cdata)
>  {
>  #ifdef CONFIG_X86_X32_ABI
> -       if (!is_ia32_task())
> +       if (!in_ia32_syscall())
>                 return x32_arch_ptrace(child, request, caddr, cdata);
>  #endif
>  #ifdef CONFIG_IA32_EMULATION
> diff --git a/arch/x86/kernel/signal.c b/arch/x86/kernel/signal.c
> index 548ddf7d6fd2..aa31265aa61d 100644
> --- a/arch/x86/kernel/signal.c
> +++ b/arch/x86/kernel/signal.c
> @@ -762,7 +762,7 @@ handle_signal(struct ksignal *ksig, struct pt_regs *regs)
>  static inline unsigned long get_nr_restart_syscall(const struct pt_regs *regs)
>  {
>  #ifdef CONFIG_X86_64
> -       if (is_ia32_task())
> +       if (in_ia32_syscall())
>                 return __NR_ia32_restart_syscall;
>  #endif
>  #ifdef CONFIG_X86_X32_ABI
> diff --git a/arch/x86/kernel/uprobes.c b/arch/x86/kernel/uprobes.c
> index bf4db6eaec8f..98b4dc87628b 100644
> --- a/arch/x86/kernel/uprobes.c
> +++ b/arch/x86/kernel/uprobes.c
> @@ -516,7 +516,7 @@ struct uprobe_xol_ops {
>
>  static inline int sizeof_long(void)
>  {
> -       return is_ia32_task() ? 4 : 8;
> +       return in_ia32_syscall() ? 4 : 8;
>  }
>
>  static int default_pre_xol_op(struct arch_uprobe *auprobe, struct pt_regs *regs)
> --
> 2.8.0
>



-- 
Andy Lutomirski
AMA Capital Management, LLC

[toc] | [prev] | [next] | [standalone]

#1379862 — [PATCHv4 2/2] x86: rename is_{ia32,x32}_task to in_{ia32,x32}_syscall

Impact: clearify meaning

Suggested-by: Andy Lutomirski <luto@amacapital.net>
Suggested-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
---
v3: initial patch

 arch/x86/entry/common.c            | 2 +-
 arch/x86/entry/vdso/vma.c          | 2 +-
 arch/x86/include/asm/compat.h      | 4 ++--
 arch/x86/include/asm/thread_info.h | 2 +-
 arch/x86/kernel/process_64.c       | 2 +-
 arch/x86/kernel/ptrace.c           | 2 +-
 arch/x86/kernel/signal.c           | 2 +-
 arch/x86/kernel/uprobes.c          | 2 +-
 8 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/arch/x86/entry/common.c b/arch/x86/entry/common.c
index e79d93d44ecd..ec138e538c44 100644
--- a/arch/x86/entry/common.c
+++ b/arch/x86/entry/common.c
@@ -191,7 +191,7 @@ long syscall_trace_enter_phase2(struct pt_regs *regs, u32 arch,
 
 long syscall_trace_enter(struct pt_regs *regs)
 {
-	u32 arch = is_ia32_task() ? AUDIT_ARCH_I386 : AUDIT_ARCH_X86_64;
+	u32 arch = in_ia32_syscall() ? AUDIT_ARCH_I386 : AUDIT_ARCH_X86_64;
 	unsigned long phase1_result = syscall_trace_enter_phase1(regs, arch);
 
 	if (phase1_result == 0)
diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
index d26517f3f88f..2c2cc6bc77fa 100644
--- a/arch/x86/entry/vdso/vma.c
+++ b/arch/x86/entry/vdso/vma.c
@@ -108,7 +108,7 @@ static int vdso_mremap(const struct vm_special_mapping *sm,
 	if (image->size != new_size)
 		return -EINVAL;
 
-	if (is_ia32_task()) {
+	if (in_ia32_syscall()) {
 		struct pt_regs *regs = current_pt_regs();
 		unsigned long vdso_land = image->sym_int80_landing_pad;
 		unsigned long old_land_addr = vdso_land +
diff --git a/arch/x86/include/asm/compat.h b/arch/x86/include/asm/compat.h
index ebb102e1bbc7..5a3b2c119ed0 100644
--- a/arch/x86/include/asm/compat.h
+++ b/arch/x86/include/asm/compat.h
@@ -307,7 +307,7 @@ static inline void __user *arch_compat_alloc_user_space(long len)
 	return (void __user *)round_down(sp - len, 16);
 }
 
-static inline bool is_x32_task(void)
+static inline bool in_x32_syscall(void)
 {
 #ifdef CONFIG_X86_X32_ABI
 	if (task_pt_regs(current)->orig_ax & __X32_SYSCALL_BIT)
@@ -318,7 +318,7 @@ static inline bool is_x32_task(void)
 
 static inline bool in_compat_syscall(void)
 {
-	return is_ia32_task() || is_x32_task();
+	return in_ia32_syscall() || in_x32_syscall();
 }
 #define in_compat_syscall in_compat_syscall	/* override the generic impl */
 
diff --git a/arch/x86/include/asm/thread_info.h b/arch/x86/include/asm/thread_info.h
index ffae84df8a93..30c133ac05cd 100644
--- a/arch/x86/include/asm/thread_info.h
+++ b/arch/x86/include/asm/thread_info.h
@@ -255,7 +255,7 @@ static inline bool test_and_clear_restore_sigmask(void)
 	return true;
 }
 
-static inline bool is_ia32_task(void)
+static inline bool in_ia32_syscall(void)
 {
 #ifdef CONFIG_X86_32
 	return true;
diff --git a/arch/x86/kernel/process_64.c b/arch/x86/kernel/process_64.c
index 6cbab31ac23a..4a62ec457b56 100644
--- a/arch/x86/kernel/process_64.c
+++ b/arch/x86/kernel/process_64.c
@@ -210,7 +210,7 @@ int copy_thread_tls(unsigned long clone_flags, unsigned long sp,
 	 */
 	if (clone_flags & CLONE_SETTLS) {
 #ifdef CONFIG_IA32_EMULATION
-		if (is_ia32_task())
+		if (in_ia32_syscall())
 			err = do_set_thread_area(p, -1,
 				(struct user_desc __user *)tls, 0);
 		else
diff --git a/arch/x86/kernel/ptrace.c b/arch/x86/kernel/ptrace.c
index 32e9d9cbb884..0f4d2a5df2dc 100644
--- a/arch/x86/kernel/ptrace.c
+++ b/arch/x86/kernel/ptrace.c
@@ -1266,7 +1266,7 @@ long compat_arch_ptrace(struct task_struct *child, compat_long_t request,
 			compat_ulong_t caddr, compat_ulong_t cdata)
 {
 #ifdef CONFIG_X86_X32_ABI
-	if (!is_ia32_task())
+	if (!in_ia32_syscall())
 		return x32_arch_ptrace(child, request, caddr, cdata);
 #endif
 #ifdef CONFIG_IA32_EMULATION
diff --git a/arch/x86/kernel/signal.c b/arch/x86/kernel/signal.c
index 548ddf7d6fd2..aa31265aa61d 100644
--- a/arch/x86/kernel/signal.c
+++ b/arch/x86/kernel/signal.c
@@ -762,7 +762,7 @@ handle_signal(struct ksignal *ksig, struct pt_regs *regs)
 static inline unsigned long get_nr_restart_syscall(const struct pt_regs *regs)
 {
 #ifdef CONFIG_X86_64
-	if (is_ia32_task())
+	if (in_ia32_syscall())
 		return __NR_ia32_restart_syscall;
 #endif
 #ifdef CONFIG_X86_X32_ABI
diff --git a/arch/x86/kernel/uprobes.c b/arch/x86/kernel/uprobes.c
index bf4db6eaec8f..98b4dc87628b 100644
--- a/arch/x86/kernel/uprobes.c
+++ b/arch/x86/kernel/uprobes.c
@@ -516,7 +516,7 @@ struct uprobe_xol_ops {
 
 static inline int sizeof_long(void)
 {
-	return is_ia32_task() ? 4 : 8;
+	return in_ia32_syscall() ? 4 : 8;
 }
 
 static int default_pre_xol_op(struct arch_uprobe *auprobe, struct pt_regs *regs)
-- 
2.8.0

[toc] | [prev] | [next] | [standalone]

#1379869 — [PATCHv4 1/2] x86/vdso: add mremap hook to vm_special_mapping

Add possibility for userspace 32-bit applications to move
vdso mapping. Previously, when userspace app called
mremap for vdso, in return path it would land on previous
address of vdso page, resulting in segmentation violation.
Now it lands fine and returns to userspace with remapped vdso.
This will also fix context.vdso pointer for 64-bit, which does not
affect the user of vdso after mremap by now, but this may change.

As suggested by Andy, return EINVAL for mremap that splits vdso image.

Renamed and moved text_mapping structure declaration inside
map_vdso, as it used only there and now it complement
vvar_mapping variable.

There is still problem for remapping vdso in glibc applications:
linker relocates addresses for syscalls on vdso page, so
you need to relink with the new addresses. Or the next syscall
through glibc may fail:
  Program received signal SIGSEGV, Segmentation fault.
  #0  0xf7fd9b80 in __kernel_vsyscall ()
  #1  0xf7ec8238 in _exit () from /usr/lib32/libc.so.6

Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
---
v4: drop __maybe_unused & use image from mm->context instead vdso_image_32
v3: as Andy suggested, return EINVAL in case of splitting vdso blob on mremap;
    used is_ia32_task instead of ifdefs 
v2: added __maybe_unused for pt_regs in vdso_mremap

 arch/x86/entry/vdso/vma.c | 36 +++++++++++++++++++++++++++++++-----
 include/linux/mm_types.h  |  3 +++
 mm/mmap.c                 | 10 ++++++++++
 3 files changed, 44 insertions(+), 5 deletions(-)

diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
index 10f704584922..d26517f3f88f 100644
--- a/arch/x86/entry/vdso/vma.c
+++ b/arch/x86/entry/vdso/vma.c
@@ -12,6 +12,7 @@
 #include <linux/random.h>
 #include <linux/elf.h>
 #include <linux/cpu.h>
+#include <linux/ptrace.h>
 #include <asm/pvclock.h>
 #include <asm/vgtod.h>
 #include <asm/proto.h>
@@ -98,10 +99,29 @@ static int vdso_fault(const struct vm_special_mapping *sm,
 	return 0;
 }
 
-static const struct vm_special_mapping text_mapping = {
-	.name = "[vdso]",
-	.fault = vdso_fault,
-};
+static int vdso_mremap(const struct vm_special_mapping *sm,
+		      struct vm_area_struct *new_vma)
+{
+	unsigned long new_size = new_vma->vm_end - new_vma->vm_start;
+	const struct vdso_image *image = current->mm->context.vdso_image;
+
+	if (image->size != new_size)
+		return -EINVAL;
+
+	if (is_ia32_task()) {
+		struct pt_regs *regs = current_pt_regs();
+		unsigned long vdso_land = image->sym_int80_landing_pad;
+		unsigned long old_land_addr = vdso_land +
+			(unsigned long)current->mm->context.vdso;
+
+		/* Fixing userspace landing - look at do_fast_syscall_32 */
+		if (regs->ip == old_land_addr)
+			regs->ip = new_vma->vm_start + vdso_land;
+	}
+	new_vma->vm_mm->context.vdso = (void __user *)new_vma->vm_start;
+
+	return 0;
+}
 
 static int vvar_fault(const struct vm_special_mapping *sm,
 		      struct vm_area_struct *vma, struct vm_fault *vmf)
@@ -162,6 +182,12 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 	struct vm_area_struct *vma;
 	unsigned long addr, text_start;
 	int ret = 0;
+
+	static const struct vm_special_mapping vdso_mapping = {
+		.name = "[vdso]",
+		.fault = vdso_fault,
+		.mremap = vdso_mremap,
+	};
 	static const struct vm_special_mapping vvar_mapping = {
 		.name = "[vvar]",
 		.fault = vvar_fault,
@@ -195,7 +221,7 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 				       image->size,
 				       VM_READ|VM_EXEC|
 				       VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
-				       &text_mapping);
+				       &vdso_mapping);
 
 	if (IS_ERR(vma)) {
 		ret = PTR_ERR(vma);
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index c2d75b4fa86c..4d16ab9287af 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -586,6 +586,9 @@ struct vm_special_mapping {
 	int (*fault)(const struct vm_special_mapping *sm,
 		     struct vm_area_struct *vma,
 		     struct vm_fault *vmf);
+
+	int (*mremap)(const struct vm_special_mapping *sm,
+		     struct vm_area_struct *new_vma);
 };
 
 enum tlb_flush_reason {
diff --git a/mm/mmap.c b/mm/mmap.c
index bd2e1a533bc1..ba71658dd1a1 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2930,9 +2930,19 @@ 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)
+{
+	struct vm_special_mapping *sm = new_vma->vm_private_data;
+
+	if (sm->mremap)
+		return sm->mremap(sm, new_vma);
+	return 0;
+}
+
 static const struct vm_operations_struct special_mapping_vmops = {
 	.close = special_mapping_close,
 	.fault = special_mapping_fault,
+	.mremap = special_mapping_mremap,
 	.name = special_mapping_name,
 };
 
-- 
2.8.0

[toc] | [prev] | [next] | [standalone]

#1380014 — Re: [PATCHv4 1/2] x86/vdso: add mremap hook to vm_special_mapping

On Fri, Apr 15, 2016 at 7:12 AM, Dmitry Safonov <dsafonov@virtuozzo.com> wrote:
> Add possibility for userspace 32-bit applications to move
> vdso mapping. Previously, when userspace app called
> mremap for vdso, in return path it would land on previous
> address of vdso page, resulting in segmentation violation.
> Now it lands fine and returns to userspace with remapped vdso.
> This will also fix context.vdso pointer for 64-bit, which does not
> affect the user of vdso after mremap by now, but this may change.
>
> As suggested by Andy, return EINVAL for mremap that splits vdso image.
>
> Renamed and moved text_mapping structure declaration inside
> map_vdso, as it used only there and now it complement
> vvar_mapping variable.
>
> There is still problem for remapping vdso in glibc applications:
> linker relocates addresses for syscalls on vdso page, so
> you need to relink with the new addresses. Or the next syscall
> through glibc may fail:
>   Program received signal SIGSEGV, Segmentation fault.
>   #0  0xf7fd9b80 in __kernel_vsyscall ()
>   #1  0xf7ec8238 in _exit () from /usr/lib32/libc.so.6
>
> Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
> ---
> v4: drop __maybe_unused & use image from mm->context instead vdso_image_32
> v3: as Andy suggested, return EINVAL in case of splitting vdso blob on mremap;
>     used is_ia32_task instead of ifdefs
> v2: added __maybe_unused for pt_regs in vdso_mremap
>
>  arch/x86/entry/vdso/vma.c | 36 +++++++++++++++++++++++++++++++-----
>  include/linux/mm_types.h  |  3 +++
>  mm/mmap.c                 | 10 ++++++++++
>  3 files changed, 44 insertions(+), 5 deletions(-)
>
> diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
> index 10f704584922..d26517f3f88f 100644
> --- a/arch/x86/entry/vdso/vma.c
> +++ b/arch/x86/entry/vdso/vma.c
> @@ -12,6 +12,7 @@
>  #include <linux/random.h>
>  #include <linux/elf.h>
>  #include <linux/cpu.h>
> +#include <linux/ptrace.h>
>  #include <asm/pvclock.h>
>  #include <asm/vgtod.h>
>  #include <asm/proto.h>
> @@ -98,10 +99,29 @@ static int vdso_fault(const struct vm_special_mapping *sm,
>         return 0;
>  }
>
> -static const struct vm_special_mapping text_mapping = {
> -       .name = "[vdso]",
> -       .fault = vdso_fault,
> -};

Ah, this is what I missed last time.  Looks good.

> +static int vdso_mremap(const struct vm_special_mapping *sm,
> +                     struct vm_area_struct *new_vma)
> +{
> +       unsigned long new_size = new_vma->vm_end - new_vma->vm_start;
> +       const struct vdso_image *image = current->mm->context.vdso_image;
> +
> +       if (image->size != new_size)
> +               return -EINVAL;
> +
> +       if (is_ia32_task()) {
> +               struct pt_regs *regs = current_pt_regs();
> +               unsigned long vdso_land = image->sym_int80_landing_pad;
> +               unsigned long old_land_addr = vdso_land +
> +                       (unsigned long)current->mm->context.vdso;
> +
> +               /* Fixing userspace landing - look at do_fast_syscall_32 */
> +               if (regs->ip == old_land_addr)
> +                       regs->ip = new_vma->vm_start + vdso_land;
> +       }
> +       new_vma->vm_mm->context.vdso = (void __user *)new_vma->vm_start;

A couple minor things:

 - You're looking at both new_vma->vm_mm and current->mm.  Is there a
reason for that?  If they're different, I'd be quite surprised, but
maybe it would make sense to check.

 - On second thought, the is_ia32_task() check is a little weird given
that you're planning on making the vdso image type.  It might make
sense to change that to in_ia32_syscall() && image == &vdso_image_32.

Other than that, looks good to me.

You could add a really simple test case to selftests/x86:

mremap(the vdso, somewhere else);
asm volatile ("int $0x80" : : "a" (__NR_exit), "b" (0));

That'll segfault if this fails and it'll work and return 0 if it works.

FWIW, there's one respect in which this code could be problematic down
the road: if syscalls ever start needing the vvar page, then this gets
awkward because you can't remap both at once.  Also, this is
fundamentally racy if multiple threads try to use it (but there's
nothing whatsoever the kernel could do about that).  In general, once
the call to change and relocate the vdso gets written, CRIU should
probably prefer to use it over mremap.

--Andy

[toc] | [prev] | [next] | [standalone]

#1381841 — Re: [PATCHv4 1/2] x86/vdso: add mremap hook to vm_special_mapping

On Apr 18, 2016 4:19 AM, "Dmitry Safonov" <dsafonov@virtuozzo.com> wrote:
>
> On 04/15/2016 07:58 PM, Andy Lutomirski wrote:
>>
>> A couple minor things:
>>
>>   - You're looking at both new_vma->vm_mm and current->mm.  Is there a
>> reason for that?  If they're different, I'd be quite surprised, but
>> maybe it would make sense to check.
>
>
> Ok, will add a check.
>
>
>>   - On second thought, the is_ia32_task() check is a little weird given
>> that you're planning on making the vdso image type.  It might make
>> sense to change that to in_ia32_syscall() && image == &vdso_image_32.
>
>
> Yes, we might be there remapping vdso_image_64 through int80, where
> we shouldn't change landing. Thanks, will add a check.
>
>
>> Other than that, looks good to me.
>>
>> You could add a really simple test case to selftests/x86:
>>
>> mremap(the vdso, somewhere else);
>> asm volatile ("int $0x80" : : "a" (__NR_exit), "b" (0));
>>
>> That'll segfault if this fails and it'll work and return 0 if it works.
>
>
> Will add - for now I have tested this with kind the same program.
>
>
>> FWIW, there's one respect in which this code could be problematic down
>> the road: if syscalls ever start needing the vvar page, then this gets
>> awkward because you can't remap both at once.  Also, this is
>> fundamentally racy if multiple threads try to use it (but there's
>> nothing whatsoever the kernel could do about that).  In general, once
>> the call to change and relocate the vdso gets written, CRIU should
>> probably prefer to use it over mremap.
>
>
> Yes, but from my point of view, for the other reasons:
> - on restore stage of CRIU, restorer maps VMAs that were dumped
> on dump stage.
> - this is done in one thread, as other threads may need those VMAs
> to funciton.
> - one of vmas, being restored is vDSO (which also was dumped), so
> there is image for this blob.
>
> So, ideally, I even would not need such API to remap blobs
> from 64 to 32 (or backwards).

If I'm understanding you right, this won't work reliably.  The kernel
makes no promise that one kernel's vdso is compatible with another
kernel.  In fact, the internal interfaces have changed several times
and are changing again in 4.6.

You also need to put vvar at the correct offset from the text, and
vvar can't be dumped and restores at all, since the whole point is
that it changes at runtime.

> This is ideally for other applications
> that switches their mode. For CRIU *ideally* I do not need it, as
> I have this vma's image dumped before - I only need remap to
> fix contex.vdso pointer for proper landing and I'm doing it in
> one thread.
>
> But, in the practice, one may migrate application from one
> kernel to another. And for different kernel versions, there may
> be different vDSO entries. For now (before compatible C/R)
> we have checked if vDSO differ and if so, examine this different
> vDSO symbols and add jump trampolines on places where
> were entries in previous vDSO to a new one.
> So, this is also true for 32-bit vDSO blob. That's why I need
> this API for CRIU.

Understood.

I like the mremap support regardless of whether CRIU ends up using it long-term.

--Andy

[toc] | [prev] | [next] | [standalone]

#1381720 — [PATCHv5 2/3] x86/vdso: add mremap hook to vm_special_mapping

Add possibility for userspace 32-bit applications to move
vdso mapping. Previously, when userspace app called
mremap for vdso, in return path it would land on previous
address of vdso page, resulting in segmentation violation.
Now it lands fine and returns to userspace with remapped vdso.
This will also fix context.vdso pointer for 64-bit, which does not
affect the user of vdso after mremap by now, but this may change.

As suggested by Andy, return EINVAL for mremap that splits vdso image.

Renamed and moved text_mapping structure declaration inside
map_vdso, as it used only there and now it complement
vvar_mapping variable.

There is still problem for remapping vdso in glibc applications:
linker relocates addresses for syscalls on vdso page, so
you need to relink with the new addresses. Or the next syscall
through glibc may fail:
  Program received signal SIGSEGV, Segmentation fault.
  #0  0xf7fd9b80 in __kernel_vsyscall ()
  #1  0xf7ec8238 in _exit () from /usr/lib32/libc.so.6

Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
---
v5: as Andy suggested, add a check that new_vma->vm_mm and current->mm are
    the same, also check not only in_ia32_syscall() but image == &vdso_image_32
v4: drop __maybe_unused & use image from mm->context instead vdso_image_32
v3: as Andy suggested, return EINVAL in case of splitting vdso blob on mremap;
    used is_ia32_task instead of ifdefs 
v2: added __maybe_unused for pt_regs in vdso_mremap

 arch/x86/entry/vdso/vma.c | 39 ++++++++++++++++++++++++++++++++++-----
 include/linux/mm_types.h  |  3 +++
 mm/mmap.c                 | 10 ++++++++++
 3 files changed, 47 insertions(+), 5 deletions(-)

diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
index 10f704584922..ba133a9ab4f5 100644
--- a/arch/x86/entry/vdso/vma.c
+++ b/arch/x86/entry/vdso/vma.c
@@ -12,6 +12,7 @@
 #include <linux/random.h>
 #include <linux/elf.h>
 #include <linux/cpu.h>
+#include <linux/ptrace.h>
 #include <asm/pvclock.h>
 #include <asm/vgtod.h>
 #include <asm/proto.h>
@@ -98,10 +99,32 @@ static int vdso_fault(const struct vm_special_mapping *sm,
 	return 0;
 }
 
-static const struct vm_special_mapping text_mapping = {
-	.name = "[vdso]",
-	.fault = vdso_fault,
-};
+static int vdso_mremap(const struct vm_special_mapping *sm,
+		      struct vm_area_struct *new_vma)
+{
+	unsigned long new_size = new_vma->vm_end - new_vma->vm_start;
+	const struct vdso_image *image = current->mm->context.vdso_image;
+
+	if (image->size != new_size)
+		return -EINVAL;
+
+	if (current->mm != new_vma->vm_mm)
+		return -EFAULT;
+
+	if (in_ia32_syscall() && image == &vdso_image_32) {
+		struct pt_regs *regs = current_pt_regs();
+		unsigned long vdso_land = image->sym_int80_landing_pad;
+		unsigned long old_land_addr = vdso_land +
+			(unsigned long)current->mm->context.vdso;
+
+		/* Fixing userspace landing - look at do_fast_syscall_32 */
+		if (regs->ip == old_land_addr)
+			regs->ip = new_vma->vm_start + vdso_land;
+	}
+	current->mm->context.vdso = (void __user *)new_vma->vm_start;
+
+	return 0;
+}
 
 static int vvar_fault(const struct vm_special_mapping *sm,
 		      struct vm_area_struct *vma, struct vm_fault *vmf)
@@ -162,6 +185,12 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 	struct vm_area_struct *vma;
 	unsigned long addr, text_start;
 	int ret = 0;
+
+	static const struct vm_special_mapping vdso_mapping = {
+		.name = "[vdso]",
+		.fault = vdso_fault,
+		.mremap = vdso_mremap,
+	};
 	static const struct vm_special_mapping vvar_mapping = {
 		.name = "[vvar]",
 		.fault = vvar_fault,
@@ -195,7 +224,7 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 				       image->size,
 				       VM_READ|VM_EXEC|
 				       VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
-				       &text_mapping);
+				       &vdso_mapping);
 
 	if (IS_ERR(vma)) {
 		ret = PTR_ERR(vma);
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index c2d75b4fa86c..4d16ab9287af 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -586,6 +586,9 @@ struct vm_special_mapping {
 	int (*fault)(const struct vm_special_mapping *sm,
 		     struct vm_area_struct *vma,
 		     struct vm_fault *vmf);
+
+	int (*mremap)(const struct vm_special_mapping *sm,
+		     struct vm_area_struct *new_vma);
 };
 
 enum tlb_flush_reason {
diff --git a/mm/mmap.c b/mm/mmap.c
index bd2e1a533bc1..ba71658dd1a1 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2930,9 +2930,19 @@ 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)
+{
+	struct vm_special_mapping *sm = new_vma->vm_private_data;
+
+	if (sm->mremap)
+		return sm->mremap(sm, new_vma);
+	return 0;
+}
+
 static const struct vm_operations_struct special_mapping_vmops = {
 	.close = special_mapping_close,
 	.fault = special_mapping_fault,
+	.mremap = special_mapping_mremap,
 	.name = special_mapping_name,
 };
 
-- 
2.8.0

[toc] | [prev] | [next] | [standalone]

#1381735 — Re: [PATCHv5 2/3] x86/vdso: add mremap hook to vm_special_mapping

[Multipart message — attachments visible in raw view] — view raw

Hi Dmitry,

[auto build test WARNING on v4.6-rc4]
[also build test WARNING on next-20160418]
[cannot apply to tip/x86/core tip/x86/vdso]
[if your patch is applied to the wrong git tree, please drop us a note to help improving the system]

url:    https://github.com/0day-ci/linux/commits/Dmitry-Safonov/x86-rename-is_-ia32-x32-_task-to-in_-ia32-x32-_syscall/20160418-214656
config: x86_64-randconfig-x000-201616 (attached as .config)
reproduce:
        # save the attached .config to linux build tree
        make ARCH=x86_64 

All warnings (new ones prefixed by >>):

   In file included from include/asm-generic/bug.h:4:0,
                    from arch/x86/include/asm/bug.h:35,
                    from include/linux/bug.h:4,
                    from include/linux/mmdebug.h:4,
                    from include/linux/mm.h:8,
                    from arch/x86/entry/vdso/vma.c:7:
   arch/x86/entry/vdso/vma.c: In function 'vdso_mremap':
   arch/x86/entry/vdso/vma.c:114:37: error: 'vdso_image_32' undeclared (first use in this function)
     if (in_ia32_syscall() && image == &vdso_image_32) {
                                        ^
   include/linux/compiler.h:151:30: note: in definition of macro '__trace_if'
     if (__builtin_constant_p(!!(cond)) ? !!(cond) :   \
                                 ^
>> arch/x86/entry/vdso/vma.c:114:2: note: in expansion of macro 'if'
     if (in_ia32_syscall() && image == &vdso_image_32) {
     ^
   arch/x86/entry/vdso/vma.c:114:37: note: each undeclared identifier is reported only once for each function it appears in
     if (in_ia32_syscall() && image == &vdso_image_32) {
                                        ^
   include/linux/compiler.h:151:30: note: in definition of macro '__trace_if'
     if (__builtin_constant_p(!!(cond)) ? !!(cond) :   \
                                 ^
>> arch/x86/entry/vdso/vma.c:114:2: note: in expansion of macro 'if'
     if (in_ia32_syscall() && image == &vdso_image_32) {
     ^

vim +/if +114 arch/x86/entry/vdso/vma.c

     1	/*
     2	 * Copyright 2007 Andi Kleen, SUSE Labs.
     3	 * Subject to the GPL, v.2
     4	 *
     5	 * This contains most of the x86 vDSO kernel-side code.
     6	 */
   > 7	#include <linux/mm.h>
     8	#include <linux/err.h>
     9	#include <linux/sched.h>
    10	#include <linux/slab.h>
    11	#include <linux/init.h>
    12	#include <linux/random.h>
    13	#include <linux/elf.h>
    14	#include <linux/cpu.h>
    15	#include <linux/ptrace.h>
    16	#include <asm/pvclock.h>
    17	#include <asm/vgtod.h>
    18	#include <asm/proto.h>
    19	#include <asm/vdso.h>
    20	#include <asm/vvar.h>
    21	#include <asm/page.h>
    22	#include <asm/hpet.h>
    23	#include <asm/desc.h>
    24	#include <asm/cpufeature.h>
    25	
    26	#if defined(CONFIG_X86_64)
    27	unsigned int __read_mostly vdso64_enabled = 1;
    28	#endif
    29	
    30	void __init init_vdso_image(const struct vdso_image *image)
    31	{
    32		BUG_ON(image->size % PAGE_SIZE != 0);
    33	
    34		apply_alternatives((struct alt_instr *)(image->data + image->alt),
    35				   (struct alt_instr *)(image->data + image->alt +
    36							image->alt_len));
    37	}
    38	
    39	struct linux_binprm;
    40	
    41	/*
    42	 * Put the vdso above the (randomized) stack with another randomized
    43	 * offset.  This way there is no hole in the middle of address space.
    44	 * To save memory make sure it is still in the same PTE as the stack
    45	 * top.  This doesn't give that many random bits.
    46	 *
    47	 * Note that this algorithm is imperfect: the distribution of the vdso
    48	 * start address within a PMD is biased toward the end.
    49	 *
    50	 * Only used for the 64-bit and x32 vdsos.
    51	 */
    52	static unsigned long vdso_addr(unsigned long start, unsigned len)
    53	{
    54	#ifdef CONFIG_X86_32
    55		return 0;
    56	#else
    57		unsigned long addr, end;
    58		unsigned offset;
    59	
    60		/*
    61		 * Round up the start address.  It can start out unaligned as a result
    62		 * of stack start randomization.
    63		 */
    64		start = PAGE_ALIGN(start);
    65	
    66		/* Round the lowest possible end address up to a PMD boundary. */
    67		end = (start + len + PMD_SIZE - 1) & PMD_MASK;
    68		if (end >= TASK_SIZE_MAX)
    69			end = TASK_SIZE_MAX;
    70		end -= len;
    71	
    72		if (end > start) {
    73			offset = get_random_int() % (((end - start) >> PAGE_SHIFT) + 1);
    74			addr = start + (offset << PAGE_SHIFT);
    75		} else {
    76			addr = start;
    77		}
    78	
    79		/*
    80		 * Forcibly align the final address in case we have a hardware
    81		 * issue that requires alignment for performance reasons.
    82		 */
    83		addr = align_vdso_addr(addr);
    84	
    85		return addr;
    86	#endif
    87	}
    88	
    89	static int vdso_fault(const struct vm_special_mapping *sm,
    90			      struct vm_area_struct *vma, struct vm_fault *vmf)
    91	{
    92		const struct vdso_image *image = vma->vm_mm->context.vdso_image;
    93	
    94		if (!image || (vmf->pgoff << PAGE_SHIFT) >= image->size)
    95			return VM_FAULT_SIGBUS;
    96	
    97		vmf->page = virt_to_page(image->data + (vmf->pgoff << PAGE_SHIFT));
    98		get_page(vmf->page);
    99		return 0;
   100	}
   101	
   102	static int vdso_mremap(const struct vm_special_mapping *sm,
   103			      struct vm_area_struct *new_vma)
   104	{
   105		unsigned long new_size = new_vma->vm_end - new_vma->vm_start;
   106		const struct vdso_image *image = current->mm->context.vdso_image;
   107	
   108		if (image->size != new_size)
   109			return -EINVAL;
   110	
   111		if (current->mm != new_vma->vm_mm)
   112			return -EFAULT;
   113	
 > 114		if (in_ia32_syscall() && image == &vdso_image_32) {
   115			struct pt_regs *regs = current_pt_regs();
   116			unsigned long vdso_land = image->sym_int80_landing_pad;
   117			unsigned long old_land_addr = vdso_land +

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[toc] | [prev] | [next] | [standalone]

#1381741 — [PATCHv6 2/3] x86/vdso: add mremap hook to vm_special_mapping

Add possibility for userspace 32-bit applications to move
vdso mapping. Previously, when userspace app called
mremap for vdso, in return path it would land on previous
address of vdso page, resulting in segmentation violation.
Now it lands fine and returns to userspace with remapped vdso.
This will also fix context.vdso pointer for 64-bit, which does not
affect the user of vdso after mremap by now, but this may change.

As suggested by Andy, return EINVAL for mremap that splits vdso image.

Renamed and moved text_mapping structure declaration inside
map_vdso, as it used only there and now it complement
vvar_mapping variable.

There is still problem for remapping vdso in glibc applications:
linker relocates addresses for syscalls on vdso page, so
you need to relink with the new addresses. Or the next syscall
through glibc may fail:
  Program received signal SIGSEGV, Segmentation fault.
  #0  0xf7fd9b80 in __kernel_vsyscall ()
  #1  0xf7ec8238 in _exit () from /usr/lib32/libc.so.6

Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
---
v6: moved vdso_image_32 check and fixup code into vdso_fix_landing function
    with ifdefs around
v5: as Andy suggested, add a check that new_vma->vm_mm and current->mm are
    the same, also check not only in_ia32_syscall() but image == &vdso_image_32
v4: drop __maybe_unused & use image from mm->context instead vdso_image_32
v3: as Andy suggested, return EINVAL in case of splitting vdso blob on mremap;
    used is_ia32_task instead of ifdefs 
v2: added __maybe_unused for pt_regs in vdso_mremap

 arch/x86/entry/vdso/vma.c | 49 ++++++++++++++++++++++++++++++++++++++++++-----
 include/linux/mm_types.h  |  3 +++
 mm/mmap.c                 | 10 ++++++++++
 3 files changed, 57 insertions(+), 5 deletions(-)

diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
index 10f704584922..3385586f86a6 100644
--- a/arch/x86/entry/vdso/vma.c
+++ b/arch/x86/entry/vdso/vma.c
@@ -12,6 +12,7 @@
 #include <linux/random.h>
 #include <linux/elf.h>
 #include <linux/cpu.h>
+#include <linux/ptrace.h>
 #include <asm/pvclock.h>
 #include <asm/vgtod.h>
 #include <asm/proto.h>
@@ -98,10 +99,42 @@ static int vdso_fault(const struct vm_special_mapping *sm,
 	return 0;
 }
 
-static const struct vm_special_mapping text_mapping = {
-	.name = "[vdso]",
-	.fault = vdso_fault,
-};
+#if defined CONFIG_X86_32 || defined CONFIG_COMPAT
+static void vdso_fix_landing(const struct vdso_image *image,
+		struct vm_area_struct *new_vma)
+{
+	if (in_ia32_syscall() && image == &vdso_image_32) {
+		struct pt_regs *regs = current_pt_regs();
+		unsigned long vdso_land = image->sym_int80_landing_pad;
+		unsigned long old_land_addr = vdso_land +
+			(unsigned long)current->mm->context.vdso;
+
+		/* Fixing userspace landing - look at do_fast_syscall_32 */
+		if (regs->ip == old_land_addr)
+			regs->ip = new_vma->vm_start + vdso_land;
+	}
+}
+#else
+static void vdso_fix_landing(const struct vdso_image *image) {}
+#endif
+
+static int vdso_mremap(const struct vm_special_mapping *sm,
+		struct vm_area_struct *new_vma)
+{
+	unsigned long new_size = new_vma->vm_end - new_vma->vm_start;
+	const struct vdso_image *image = current->mm->context.vdso_image;
+
+	if (image->size != new_size)
+		return -EINVAL;
+
+	if (current->mm != new_vma->vm_mm)
+		return -EFAULT;
+
+	vdso_fix_landing(image, new_vma);
+	current->mm->context.vdso = (void __user *)new_vma->vm_start;
+
+	return 0;
+}
 
 static int vvar_fault(const struct vm_special_mapping *sm,
 		      struct vm_area_struct *vma, struct vm_fault *vmf)
@@ -162,6 +195,12 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 	struct vm_area_struct *vma;
 	unsigned long addr, text_start;
 	int ret = 0;
+
+	static const struct vm_special_mapping vdso_mapping = {
+		.name = "[vdso]",
+		.fault = vdso_fault,
+		.mremap = vdso_mremap,
+	};
 	static const struct vm_special_mapping vvar_mapping = {
 		.name = "[vvar]",
 		.fault = vvar_fault,
@@ -195,7 +234,7 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 				       image->size,
 				       VM_READ|VM_EXEC|
 				       VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
-				       &text_mapping);
+				       &vdso_mapping);
 
 	if (IS_ERR(vma)) {
 		ret = PTR_ERR(vma);
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index c2d75b4fa86c..4d16ab9287af 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -586,6 +586,9 @@ struct vm_special_mapping {
 	int (*fault)(const struct vm_special_mapping *sm,
 		     struct vm_area_struct *vma,
 		     struct vm_fault *vmf);
+
+	int (*mremap)(const struct vm_special_mapping *sm,
+		     struct vm_area_struct *new_vma);
 };
 
 enum tlb_flush_reason {
diff --git a/mm/mmap.c b/mm/mmap.c
index bd2e1a533bc1..ba71658dd1a1 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2930,9 +2930,19 @@ 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)
+{
+	struct vm_special_mapping *sm = new_vma->vm_private_data;
+
+	if (sm->mremap)
+		return sm->mremap(sm, new_vma);
+	return 0;
+}
+
 static const struct vm_operations_struct special_mapping_vmops = {
 	.close = special_mapping_close,
 	.fault = special_mapping_fault,
+	.mremap = special_mapping_mremap,
 	.name = special_mapping_name,
 };
 
-- 
2.8.0

[toc] | [prev] | [next] | [standalone]

#1381752 — [PATCHv7 2/3] x86/vdso: add mremap hook to vm_special_mapping

Add possibility for userspace 32-bit applications to move
vdso mapping. Previously, when userspace app called
mremap for vdso, in return path it would land on previous
address of vdso page, resulting in segmentation violation.
Now it lands fine and returns to userspace with remapped vdso.
This will also fix context.vdso pointer for 64-bit, which does not
affect the user of vdso after mremap by now, but this may change.

As suggested by Andy, return EINVAL for mremap that splits vdso image.

Renamed and moved text_mapping structure declaration inside
map_vdso, as it used only there and now it complement
vvar_mapping variable.

There is still problem for remapping vdso in glibc applications:
linker relocates addresses for syscalls on vdso page, so
you need to relink with the new addresses. Or the next syscall
through glibc may fail:
  Program received signal SIGSEGV, Segmentation fault.
  #0  0xf7fd9b80 in __kernel_vsyscall ()
  #1  0xf7ec8238 in _exit () from /usr/lib32/libc.so.6

Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
---
v7: that's just not my day: add new_vma parameter to vdso_fix_landing
    sorry for the noise
v6: moved vdso_image_32 check and fixup code into vdso_fix_landing function
    with ifdefs around
v5: as Andy suggested, add a check that new_vma->vm_mm and current->mm are
    the same, also check not only in_ia32_syscall() but image == &vdso_image_32
v4: drop __maybe_unused & use image from mm->context instead vdso_image_32
v3: as Andy suggested, return EINVAL in case of splitting vdso blob on mremap;
    used is_ia32_task instead of ifdefs 
v2: added __maybe_unused for pt_regs in vdso_mremap

 arch/x86/entry/vdso/vma.c | 50 ++++++++++++++++++++++++++++++++++++++++++-----
 include/linux/mm_types.h  |  3 +++
 mm/mmap.c                 | 10 ++++++++++
 3 files changed, 58 insertions(+), 5 deletions(-)

diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
index 10f704584922..d94291a19b6e 100644
--- a/arch/x86/entry/vdso/vma.c
+++ b/arch/x86/entry/vdso/vma.c
@@ -12,6 +12,7 @@
 #include <linux/random.h>
 #include <linux/elf.h>
 #include <linux/cpu.h>
+#include <linux/ptrace.h>
 #include <asm/pvclock.h>
 #include <asm/vgtod.h>
 #include <asm/proto.h>
@@ -98,10 +99,43 @@ static int vdso_fault(const struct vm_special_mapping *sm,
 	return 0;
 }
 
-static const struct vm_special_mapping text_mapping = {
-	.name = "[vdso]",
-	.fault = vdso_fault,
-};
+#if defined CONFIG_X86_32 || defined CONFIG_COMPAT
+static void vdso_fix_landing(const struct vdso_image *image,
+		struct vm_area_struct *new_vma)
+{
+	if (in_ia32_syscall() && image == &vdso_image_32) {
+		struct pt_regs *regs = current_pt_regs();
+		unsigned long vdso_land = image->sym_int80_landing_pad;
+		unsigned long old_land_addr = vdso_land +
+			(unsigned long)current->mm->context.vdso;
+
+		/* Fixing userspace landing - look at do_fast_syscall_32 */
+		if (regs->ip == old_land_addr)
+			regs->ip = new_vma->vm_start + vdso_land;
+	}
+}
+#else
+static void vdso_fix_landing(const struct vdso_image *image,
+		struct vm_area_struct *new_vma) {}
+#endif
+
+static int vdso_mremap(const struct vm_special_mapping *sm,
+		struct vm_area_struct *new_vma)
+{
+	unsigned long new_size = new_vma->vm_end - new_vma->vm_start;
+	const struct vdso_image *image = current->mm->context.vdso_image;
+
+	if (image->size != new_size)
+		return -EINVAL;
+
+	if (current->mm != new_vma->vm_mm)
+		return -EFAULT;
+
+	vdso_fix_landing(image, new_vma);
+	current->mm->context.vdso = (void __user *)new_vma->vm_start;
+
+	return 0;
+}
 
 static int vvar_fault(const struct vm_special_mapping *sm,
 		      struct vm_area_struct *vma, struct vm_fault *vmf)
@@ -162,6 +196,12 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 	struct vm_area_struct *vma;
 	unsigned long addr, text_start;
 	int ret = 0;
+
+	static const struct vm_special_mapping vdso_mapping = {
+		.name = "[vdso]",
+		.fault = vdso_fault,
+		.mremap = vdso_mremap,
+	};
 	static const struct vm_special_mapping vvar_mapping = {
 		.name = "[vvar]",
 		.fault = vvar_fault,
@@ -195,7 +235,7 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
 				       image->size,
 				       VM_READ|VM_EXEC|
 				       VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
-				       &text_mapping);
+				       &vdso_mapping);
 
 	if (IS_ERR(vma)) {
 		ret = PTR_ERR(vma);
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index c2d75b4fa86c..4d16ab9287af 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -586,6 +586,9 @@ struct vm_special_mapping {
 	int (*fault)(const struct vm_special_mapping *sm,
 		     struct vm_area_struct *vma,
 		     struct vm_fault *vmf);
+
+	int (*mremap)(const struct vm_special_mapping *sm,
+		     struct vm_area_struct *new_vma);
 };
 
 enum tlb_flush_reason {
diff --git a/mm/mmap.c b/mm/mmap.c
index bd2e1a533bc1..ba71658dd1a1 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -2930,9 +2930,19 @@ 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)
+{
+	struct vm_special_mapping *sm = new_vma->vm_private_data;
+
+	if (sm->mremap)
+		return sm->mremap(sm, new_vma);
+	return 0;
+}
+
 static const struct vm_operations_struct special_mapping_vmops = {
 	.close = special_mapping_close,
 	.fault = special_mapping_fault,
+	.mremap = special_mapping_mremap,
 	.name = special_mapping_name,
 };
 
-- 
2.8.0

[toc] | [prev] | [next] | [standalone]

#1383555 — Re: [PATCHv7 2/3] x86/vdso: add mremap hook to vm_special_mapping

On 04/18/2016 05:23 PM, Dmitry Safonov wrote:
> Add possibility for userspace 32-bit applications to move
> vdso mapping. Previously, when userspace app called
> mremap for vdso, in return path it would land on previous
> address of vdso page, resulting in segmentation violation.
> Now it lands fine and returns to userspace with remapped vdso.
> This will also fix context.vdso pointer for 64-bit, which does not
> affect the user of vdso after mremap by now, but this may change.
>
> As suggested by Andy, return EINVAL for mremap that splits vdso image.
>
> Renamed and moved text_mapping structure declaration inside
> map_vdso, as it used only there and now it complement
> vvar_mapping variable.
>
> There is still problem for remapping vdso in glibc applications:
> linker relocates addresses for syscalls on vdso page, so
> you need to relink with the new addresses. Or the next syscall
> through glibc may fail:
>    Program received signal SIGSEGV, Segmentation fault.
>    #0  0xf7fd9b80 in __kernel_vsyscall ()
>    #1  0xf7ec8238 in _exit () from /usr/lib32/libc.so.6
>
> Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>

Andy, can I have an ack from you for this version?
If it looks fine for you, of course.
And for v5 test (3/3 patch)? (I did not resed it, as it hasn't changed)

> ---
> v7: that's just not my day: add new_vma parameter to vdso_fix_landing
>      sorry for the noise
> v6: moved vdso_image_32 check and fixup code into vdso_fix_landing function
>      with ifdefs around
> v5: as Andy suggested, add a check that new_vma->vm_mm and current->mm are
>      the same, also check not only in_ia32_syscall() but image == &vdso_image_32
> v4: drop __maybe_unused & use image from mm->context instead vdso_image_32
> v3: as Andy suggested, return EINVAL in case of splitting vdso blob on mremap;
>      used is_ia32_task instead of ifdefs
> v2: added __maybe_unused for pt_regs in vdso_mremap
>
>   arch/x86/entry/vdso/vma.c | 50 ++++++++++++++++++++++++++++++++++++++++++-----
>   include/linux/mm_types.h  |  3 +++
>   mm/mmap.c                 | 10 ++++++++++
>   3 files changed, 58 insertions(+), 5 deletions(-)
>
> diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
> index 10f704584922..d94291a19b6e 100644
> --- a/arch/x86/entry/vdso/vma.c
> +++ b/arch/x86/entry/vdso/vma.c
> @@ -12,6 +12,7 @@
>   #include <linux/random.h>
>   #include <linux/elf.h>
>   #include <linux/cpu.h>
> +#include <linux/ptrace.h>
>   #include <asm/pvclock.h>
>   #include <asm/vgtod.h>
>   #include <asm/proto.h>
> @@ -98,10 +99,43 @@ static int vdso_fault(const struct vm_special_mapping *sm,
>   	return 0;
>   }
>   
> -static const struct vm_special_mapping text_mapping = {
> -	.name = "[vdso]",
> -	.fault = vdso_fault,
> -};
> +#if defined CONFIG_X86_32 || defined CONFIG_COMPAT
> +static void vdso_fix_landing(const struct vdso_image *image,
> +		struct vm_area_struct *new_vma)
> +{
> +	if (in_ia32_syscall() && image == &vdso_image_32) {
> +		struct pt_regs *regs = current_pt_regs();
> +		unsigned long vdso_land = image->sym_int80_landing_pad;
> +		unsigned long old_land_addr = vdso_land +
> +			(unsigned long)current->mm->context.vdso;
> +
> +		/* Fixing userspace landing - look at do_fast_syscall_32 */
> +		if (regs->ip == old_land_addr)
> +			regs->ip = new_vma->vm_start + vdso_land;
> +	}
> +}
> +#else
> +static void vdso_fix_landing(const struct vdso_image *image,
> +		struct vm_area_struct *new_vma) {}
> +#endif
> +
> +static int vdso_mremap(const struct vm_special_mapping *sm,
> +		struct vm_area_struct *new_vma)
> +{
> +	unsigned long new_size = new_vma->vm_end - new_vma->vm_start;
> +	const struct vdso_image *image = current->mm->context.vdso_image;
> +
> +	if (image->size != new_size)
> +		return -EINVAL;
> +
> +	if (current->mm != new_vma->vm_mm)
> +		return -EFAULT;
> +
> +	vdso_fix_landing(image, new_vma);
> +	current->mm->context.vdso = (void __user *)new_vma->vm_start;
> +
> +	return 0;
> +}
>   
>   static int vvar_fault(const struct vm_special_mapping *sm,
>   		      struct vm_area_struct *vma, struct vm_fault *vmf)
> @@ -162,6 +196,12 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
>   	struct vm_area_struct *vma;
>   	unsigned long addr, text_start;
>   	int ret = 0;
> +
> +	static const struct vm_special_mapping vdso_mapping = {
> +		.name = "[vdso]",
> +		.fault = vdso_fault,
> +		.mremap = vdso_mremap,
> +	};
>   	static const struct vm_special_mapping vvar_mapping = {
>   		.name = "[vvar]",
>   		.fault = vvar_fault,
> @@ -195,7 +235,7 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
>   				       image->size,
>   				       VM_READ|VM_EXEC|
>   				       VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
> -				       &text_mapping);
> +				       &vdso_mapping);
>   
>   	if (IS_ERR(vma)) {
>   		ret = PTR_ERR(vma);
> diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
> index c2d75b4fa86c..4d16ab9287af 100644
> --- a/include/linux/mm_types.h
> +++ b/include/linux/mm_types.h
> @@ -586,6 +586,9 @@ struct vm_special_mapping {
>   	int (*fault)(const struct vm_special_mapping *sm,
>   		     struct vm_area_struct *vma,
>   		     struct vm_fault *vmf);
> +
> +	int (*mremap)(const struct vm_special_mapping *sm,
> +		     struct vm_area_struct *new_vma);
>   };
>   
>   enum tlb_flush_reason {
> diff --git a/mm/mmap.c b/mm/mmap.c
> index bd2e1a533bc1..ba71658dd1a1 100644
> --- a/mm/mmap.c
> +++ b/mm/mmap.c
> @@ -2930,9 +2930,19 @@ 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)
> +{
> +	struct vm_special_mapping *sm = new_vma->vm_private_data;
> +
> +	if (sm->mremap)
> +		return sm->mremap(sm, new_vma);
> +	return 0;
> +}
> +
>   static const struct vm_operations_struct special_mapping_vmops = {
>   	.close = special_mapping_close,
>   	.fault = special_mapping_fault,
> +	.mremap = special_mapping_mremap,
>   	.name = special_mapping_name,
>   };
>   


-- 
Regards,
Dmitry Safonov

[toc] | [prev] | [next] | [standalone]

#1384541 — Re: [PATCHv7 2/3] x86/vdso: add mremap hook to vm_special_mapping

On Mon, Apr 18, 2016 at 7:23 AM, Dmitry Safonov <dsafonov@virtuozzo.com> wrote:
> Add possibility for userspace 32-bit applications to move
> vdso mapping. Previously, when userspace app called
> mremap for vdso, in return path it would land on previous
> address of vdso page, resulting in segmentation violation.
> Now it lands fine and returns to userspace with remapped vdso.
> This will also fix context.vdso pointer for 64-bit, which does not
> affect the user of vdso after mremap by now, but this may change.
>
> As suggested by Andy, return EINVAL for mremap that splits vdso image.
>
> Renamed and moved text_mapping structure declaration inside
> map_vdso, as it used only there and now it complement
> vvar_mapping variable.
>
> There is still problem for remapping vdso in glibc applications:
> linker relocates addresses for syscalls on vdso page, so
> you need to relink with the new addresses. Or the next syscall
> through glibc may fail:
>   Program received signal SIGSEGV, Segmentation fault.
>   #0  0xf7fd9b80 in __kernel_vsyscall ()
>   #1  0xf7ec8238 in _exit () from /usr/lib32/libc.so.6
>
> Signed-off-by: Dmitry Safonov <dsafonov@virtuozzo.com>
> ---
> v7: that's just not my day: add new_vma parameter to vdso_fix_landing
>     sorry for the noise
> v6: moved vdso_image_32 check and fixup code into vdso_fix_landing function
>     with ifdefs around
> v5: as Andy suggested, add a check that new_vma->vm_mm and current->mm are
>     the same, also check not only in_ia32_syscall() but image == &vdso_image_32
> v4: drop __maybe_unused & use image from mm->context instead vdso_image_32
> v3: as Andy suggested, return EINVAL in case of splitting vdso blob on mremap;
>     used is_ia32_task instead of ifdefs
> v2: added __maybe_unused for pt_regs in vdso_mremap
>
>  arch/x86/entry/vdso/vma.c | 50 ++++++++++++++++++++++++++++++++++++++++++-----
>  include/linux/mm_types.h  |  3 +++
>  mm/mmap.c                 | 10 ++++++++++
>  3 files changed, 58 insertions(+), 5 deletions(-)
>
> diff --git a/arch/x86/entry/vdso/vma.c b/arch/x86/entry/vdso/vma.c
> index 10f704584922..d94291a19b6e 100644
> --- a/arch/x86/entry/vdso/vma.c
> +++ b/arch/x86/entry/vdso/vma.c
> @@ -12,6 +12,7 @@
>  #include <linux/random.h>
>  #include <linux/elf.h>
>  #include <linux/cpu.h>
> +#include <linux/ptrace.h>
>  #include <asm/pvclock.h>
>  #include <asm/vgtod.h>
>  #include <asm/proto.h>
> @@ -98,10 +99,43 @@ static int vdso_fault(const struct vm_special_mapping *sm,
>         return 0;
>  }
>
> -static const struct vm_special_mapping text_mapping = {
> -       .name = "[vdso]",
> -       .fault = vdso_fault,
> -};
> +#if defined CONFIG_X86_32 || defined CONFIG_COMPAT

I think you mean defined(CONFIG_IA32_EMULATION).  IIRC CONFIG_COMPAT
is set for X32, !IA32 kernels, too.

> +static void vdso_fix_landing(const struct vdso_image *image,
> +               struct vm_area_struct *new_vma)
> +{
> +       if (in_ia32_syscall() && image == &vdso_image_32) {
> +               struct pt_regs *regs = current_pt_regs();
> +               unsigned long vdso_land = image->sym_int80_landing_pad;
> +               unsigned long old_land_addr = vdso_land +
> +                       (unsigned long)current->mm->context.vdso;
> +
> +               /* Fixing userspace landing - look at do_fast_syscall_32 */
> +               if (regs->ip == old_land_addr)
> +                       regs->ip = new_vma->vm_start + vdso_land;
> +       }
> +}
> +#else
> +static void vdso_fix_landing(const struct vdso_image *image,
> +               struct vm_area_struct *new_vma) {}
> +#endif

You could also define the function regardless and simply ifdef out the
body, thus saving a few lines of code.

> +
> +static int vdso_mremap(const struct vm_special_mapping *sm,
> +               struct vm_area_struct *new_vma)
> +{
> +       unsigned long new_size = new_vma->vm_end - new_vma->vm_start;
> +       const struct vdso_image *image = current->mm->context.vdso_image;
> +
> +       if (image->size != new_size)
> +               return -EINVAL;
> +
> +       if (current->mm != new_vma->vm_mm)
> +               return -EFAULT;

Can you make that if (WARN_ON_ONCE(current->mm != new_vma->vm_mm))?
It should be impossible.

> +
> +       vdso_fix_landing(image, new_vma);
> +       current->mm->context.vdso = (void __user *)new_vma->vm_start;
> +
> +       return 0;
> +}
>
>  static int vvar_fault(const struct vm_special_mapping *sm,
>                       struct vm_area_struct *vma, struct vm_fault *vmf)
> @@ -162,6 +196,12 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
>         struct vm_area_struct *vma;
>         unsigned long addr, text_start;
>         int ret = 0;
> +
> +       static const struct vm_special_mapping vdso_mapping = {
> +               .name = "[vdso]",
> +               .fault = vdso_fault,
> +               .mremap = vdso_mremap,
> +       };
>         static const struct vm_special_mapping vvar_mapping = {
>                 .name = "[vvar]",
>                 .fault = vvar_fault,
> @@ -195,7 +235,7 @@ static int map_vdso(const struct vdso_image *image, bool calculate_addr)
>                                        image->size,
>                                        VM_READ|VM_EXEC|
>                                        VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
> -                                      &text_mapping);
> +                                      &vdso_mapping);
>
>         if (IS_ERR(vma)) {
>                 ret = PTR_ERR(vma);
> diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
> index c2d75b4fa86c..4d16ab9287af 100644
> --- a/include/linux/mm_types.h
> +++ b/include/linux/mm_types.h
> @@ -586,6 +586,9 @@ struct vm_special_mapping {
>         int (*fault)(const struct vm_special_mapping *sm,
>                      struct vm_area_struct *vma,
>                      struct vm_fault *vmf);
> +
> +       int (*mremap)(const struct vm_special_mapping *sm,
> +                    struct vm_area_struct *new_vma);
>  };
>
>  enum tlb_flush_reason {
> diff --git a/mm/mmap.c b/mm/mmap.c
> index bd2e1a533bc1..ba71658dd1a1 100644
> --- a/mm/mmap.c
> +++ b/mm/mmap.c
> @@ -2930,9 +2930,19 @@ 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)
> +{
> +       struct vm_special_mapping *sm = new_vma->vm_private_data;
> +
> +       if (sm->mremap)
> +               return sm->mremap(sm, new_vma);
> +       return 0;
> +}
> +
>  static const struct vm_operations_struct special_mapping_vmops = {
>         .close = special_mapping_close,
>         .fault = special_mapping_fault,
> +       .mremap = special_mapping_mremap,
>         .name = special_mapping_name,
>  };
>
> --
> 2.8.0
>



-- 
Andy Lutomirski
AMA Capital Management, LLC

[toc] | [prev] | [next] | [standalone]

#1385741 — Re: [PATCHv5 2/3] x86/vdso: add mremap hook to vm_special_mapping

[Multipart message — attachments visible in raw view] — view raw

Hi,

[auto build test ERROR on v4.6-rc4]
[cannot apply to tip/x86/core tip/x86/vdso next-20160422]
[if your patch is applied to the wrong git tree, please drop us a note to help improving the system]

url:    https://github.com/0day-ci/linux/commits/Dmitry-Safonov/x86-rename-is_-ia32-x32-_task-to-in_-ia32-x32-_syscall/20160418-214656
config: x86_64-randconfig-s0-04240623 (attached as .config)
compiler: gcc-5 (Debian 5.3.1-14) 5.3.1 20160409
reproduce:
        # save the attached .config to linux build tree
        make ARCH=x86_64 

Note: the linux-review/Dmitry-Safonov/x86-rename-is_-ia32-x32-_task-to-in_-ia32-x32-_syscall/20160418-214656 HEAD aed83f1dd951908724a5ba564e2b03d68a9fc7b8 builds fine.
      It only hurts bisectibility.

All errors (new ones prefixed by >>):

   arch/x86/entry/vdso/vma.c: In function 'vdso_mremap':
>> arch/x86/entry/vdso/vma.c:114:37: error: 'vdso_image_32' undeclared (first use in this function)
     if (in_ia32_syscall() && image == &vdso_image_32) {
                                        ^
   arch/x86/entry/vdso/vma.c:114:37: note: each undeclared identifier is reported only once for each function it appears in

vim +/vdso_image_32 +114 arch/x86/entry/vdso/vma.c

   108		if (image->size != new_size)
   109			return -EINVAL;
   110	
   111		if (current->mm != new_vma->vm_mm)
   112			return -EFAULT;
   113	
 > 114		if (in_ia32_syscall() && image == &vdso_image_32) {
   115			struct pt_regs *regs = current_pt_regs();
   116			unsigned long vdso_land = image->sym_int80_landing_pad;
   117			unsigned long old_land_addr = vdso_land +

---
0-DAY kernel test infrastructure                Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all                   Intel Corporation

[toc] | [prev] | [next] | [standalone]

Page 1 of 2  [1] 2  Next page →

Back to top | Article view | linux.kernel

csiph-web