[PATCH 0/3] Bunch of vdso cleanups following feedback from x86

[PATCH 0/3] Bunch of vdso cleanups following feedback from x86

[Will Deacon]

Hi all,

Following useful discussions with Andy (thanks!), I've put together the
following (lightly tested) series for arm64 which attempts to address
issues with our current vdso code.

Please take a look. Cheers,

Will

--->8

Will Deacon (3):
  arm64: vdso: put vdso datapage in a separate vma
  arm64: vdso: move to _install_special_mapping and remove arch_vma_name
  arm64: vdso: move data page before code pages

 arch/arm64/kernel/vdso.c          | 94 +++++++++++++++++++++------------------
 arch/arm64/kernel/vdso/vdso.lds.S |  4 +-
 2 files changed, 52 insertions(+), 46 deletions(-)

-- 
2.0.0

[PATCH 1/3] arm64: vdso: put vdso datapage in a separate vma

[Will Deacon]

The VDSO datapage doesn't need to be executable (no code there) or
CoW-able (the kernel writes the page, so a private copy is totally
useless).

This patch moves the datapage into its own VMA, identified as "[vvar]"
in /proc/<pid>/maps.

Cc: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Will Deacon <will.deacon@arm.com>
---
 arch/arm64/kernel/vdso.c | 34 ++++++++++++++++++++++++++--------
 1 file changed, 26 insertions(+), 8 deletions(-)

diff --git a/arch/arm64/kernel/vdso.c b/arch/arm64/kernel/vdso.c
index 50384fec56c4..84cafbc3eb54 100644
--- a/arch/arm64/kernel/vdso.c
+++ b/arch/arm64/kernel/vdso.c
@@ -138,11 +138,12 @@ int arch_setup_additional_pages(struct linux_binprm *bprm,
 				int uses_interp)
 {
 	struct mm_struct *mm = current->mm;
-	unsigned long vdso_base, vdso_mapping_len;
+	unsigned long vdso_base, vdso_text_len, vdso_mapping_len;
 	int ret;
 
+	vdso_text_len = vdso_pages << PAGE_SHIFT;
 	/* Be sure to map the data page */
-	vdso_mapping_len = (vdso_pages + 1) << PAGE_SHIFT;
+	vdso_mapping_len = vdso_text_len + PAGE_SIZE;
 
 	down_write(&mm->mmap_sem);
 	vdso_base = get_unmapped_area(NULL, 0, vdso_mapping_len, 0, 0);
@@ -152,35 +153,52 @@ int arch_setup_additional_pages(struct linux_binprm *bprm,
 	}
 	mm->context.vdso = (void *)vdso_base;
 
-	ret = install_special_mapping(mm, vdso_base, vdso_mapping_len,
+	ret = install_special_mapping(mm, vdso_base, vdso_text_len,
 				      VM_READ|VM_EXEC|
 				      VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
 				      vdso_pagelist);
-	if (ret) {
-		mm->context.vdso = NULL;
+	if (ret)
+		goto up_fail;
+
+	vdso_base += vdso_text_len;
+	ret = install_special_mapping(mm, vdso_base, PAGE_SIZE,
+				      VM_READ|VM_MAYREAD,
+				      vdso_pagelist + vdso_pages);
+	if (ret)
 		goto up_fail;
-	}
 
-up_fail:
 	up_write(&mm->mmap_sem);
+	return 0;
 
+up_fail:
+	mm->context.vdso = NULL;
+	up_write(&mm->mmap_sem);
 	return ret;
 }
 
 const char *arch_vma_name(struct vm_area_struct *vma)
 {
+	unsigned long vdso_text;
+
+	if (!vma->vm_mm)
+		return NULL;
+
+	vdso_text = (unsigned long)vma->vm_mm->context.vdso;
+
 	/*
 	 * We can re-use the vdso pointer in mm_context_t for identifying
 	 * the vectors page for compat applications. The vDSO will always
 	 * sit above TASK_UNMAPPED_BASE and so we don't need to worry about
 	 * it conflicting with the vectors base.
 	 */
-	if (vma->vm_mm && vma->vm_start == (long)vma->vm_mm->context.vdso) {
+	if (vma->vm_start == vdso_text) {
 #ifdef CONFIG_COMPAT
 		if (vma->vm_start == AARCH32_VECTORS_BASE)
 			return "[vectors]";
 #endif
 		return "[vdso]";
+	} else if (vma->vm_start == (vdso_text + (vdso_pages << PAGE_SHIFT))) {
+		return "[vvar]";
 	}
 
 	return NULL;
-- 
2.0.0

[PATCH 1/3] arm64: vdso: put vdso datapage in a separate vma

[Nathan Lynch]

Hi Will, one question (and maybe it's more of a question for Andy than
for you).

On 07/09/2014 01:22 PM, Will Deacon wrote:
> -	ret = install_special_mapping(mm, vdso_base, vdso_mapping_len,
> +	ret = install_special_mapping(mm, vdso_base, vdso_text_len,
>  				      VM_READ|VM_EXEC|
>  				      VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
>  				      vdso_pagelist);
> -	if (ret) {
> -		mm->context.vdso = NULL;
> +	if (ret)
> +		goto up_fail;
> +
> +	vdso_base += vdso_text_len;
> +	ret = install_special_mapping(mm, vdso_base, PAGE_SIZE,
> +				      VM_READ|VM_MAYREAD,
> +				      vdso_pagelist + vdso_pages);

I note this code sets VM_MAYREAD for the data page, while x86's vvar
mapping sets only VM_READ.  (Personally I am not totally clear on the
meaning of VM_MAYREAD, but it looks like it would have implications for
use of mprotect on the mapping.)  Should one arch or the other adjust
its flags?

[PATCH 1/3] arm64: vdso: put vdso datapage in a separate vma

[Andy Lutomirski]

On Wed, Jul 9, 2014 at 12:44 PM, Nathan Lynch <Nathan_Lynch@mentor.com> wrote:
> Hi Will, one question (and maybe it's more of a question for Andy than
> for you).
>
> On 07/09/2014 01:22 PM, Will Deacon wrote:
>> -     ret = install_special_mapping(mm, vdso_base, vdso_mapping_len,
>> +     ret = install_special_mapping(mm, vdso_base, vdso_text_len,
>>                                     VM_READ|VM_EXEC|
>>                                     VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
>>                                     vdso_pagelist);
>> -     if (ret) {
>> -             mm->context.vdso = NULL;
>> +     if (ret)
>> +             goto up_fail;
>> +
>> +     vdso_base += vdso_text_len;
>> +     ret = install_special_mapping(mm, vdso_base, PAGE_SIZE,
>> +                                   VM_READ|VM_MAYREAD,
>> +                                   vdso_pagelist + vdso_pages);
>
> I note this code sets VM_MAYREAD for the data page, while x86's vvar
> mapping sets only VM_READ.  (Personally I am not totally clear on the
> meaning of VM_MAYREAD, but it looks like it would have implications for
> use of mprotect on the mapping.)  Should one arch or the other adjust
> its flags?
>

On quick inspection, this doesn't matter.  That being said, VM_MAYREAD
seems reasonable here.  I'll queue up the x86 change.

--Andy

[PATCH 2/3] arm64: vdso: move to _install_special_mapping and remove arch_vma_name

[Will Deacon]

_install_special_mapping replaces install_special_mapping and removes
the need to detect special VMA in arch_vma_name.

This patch moves the vdso and compat vectors page code over to the new
API.

Cc: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Will Deacon <will.deacon@arm.com>
---
 arch/arm64/kernel/vdso.c | 80 +++++++++++++++++++++---------------------------
 1 file changed, 35 insertions(+), 45 deletions(-)

diff --git a/arch/arm64/kernel/vdso.c b/arch/arm64/kernel/vdso.c
index 84cafbc3eb54..60ae12087d9f 100644
--- a/arch/arm64/kernel/vdso.c
+++ b/arch/arm64/kernel/vdso.c
@@ -88,22 +88,29 @@ int aarch32_setup_vectors_page(struct linux_binprm *bprm, int uses_interp)
 {
 	struct mm_struct *mm = current->mm;
 	unsigned long addr = AARCH32_VECTORS_BASE;
-	int ret;
+	static struct vm_special_mapping spec = {
+		.name	= "[vectors]",
+		.pages	= vectors_page,
+
+	};
+	void *ret;
 
 	down_write(&mm->mmap_sem);
 	current->mm->context.vdso = (void *)addr;
 
 	/* Map vectors page at the high address. */
-	ret = install_special_mapping(mm, addr, PAGE_SIZE,
-				      VM_READ|VM_EXEC|VM_MAYREAD|VM_MAYEXEC,
-				      vectors_page);
+	ret = _install_special_mapping(mm, addr, PAGE_SIZE,
+				       VM_READ|VM_EXEC|VM_MAYREAD|VM_MAYEXEC,
+				       &spec);
 
 	up_write(&mm->mmap_sem);
 
-	return ret;
+	return PTR_ERR_OR_ZERO(ret);
 }
 #endif /* CONFIG_COMPAT */
 
+static struct vm_special_mapping vdso_spec[2];
+
 static int __init vdso_init(void)
 {
 	int i;
@@ -130,6 +137,17 @@ static int __init vdso_init(void)
 	/* Grab the vDSO data page. */
 	vdso_pagelist[i] = virt_to_page(vdso_data);
 
+	/* Populate the special mapping structures */
+	vdso_spec[0] = (struct vm_special_mapping) {
+		.name	= "[vdso]",
+		.pages	= vdso_pagelist,
+	};
+
+	vdso_spec[1] = (struct vm_special_mapping) {
+		.name	= "[vvar]",
+		.pages	= vdso_pagelist + vdso_pages,
+	};
+
 	return 0;
 }
 arch_initcall(vdso_init);
@@ -139,7 +157,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm,
 {
 	struct mm_struct *mm = current->mm;
 	unsigned long vdso_base, vdso_text_len, vdso_mapping_len;
-	int ret;
+	void *ret;
 
 	vdso_text_len = vdso_pages << PAGE_SHIFT;
 	/* Be sure to map the data page */
@@ -148,23 +166,23 @@ int arch_setup_additional_pages(struct linux_binprm *bprm,
 	down_write(&mm->mmap_sem);
 	vdso_base = get_unmapped_area(NULL, 0, vdso_mapping_len, 0, 0);
 	if (IS_ERR_VALUE(vdso_base)) {
-		ret = vdso_base;
+		ret = ERR_PTR(vdso_base);
 		goto up_fail;
 	}
 	mm->context.vdso = (void *)vdso_base;
 
-	ret = install_special_mapping(mm, vdso_base, vdso_text_len,
-				      VM_READ|VM_EXEC|
-				      VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
-				      vdso_pagelist);
-	if (ret)
+	ret = _install_special_mapping(mm, vdso_base, vdso_text_len,
+				       VM_READ|VM_EXEC|
+				       VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
+				       &vdso_spec[0]);
+	if (IS_ERR(ret))
 		goto up_fail;
 
 	vdso_base += vdso_text_len;
-	ret = install_special_mapping(mm, vdso_base, PAGE_SIZE,
-				      VM_READ|VM_MAYREAD,
-				      vdso_pagelist + vdso_pages);
-	if (ret)
+	ret = _install_special_mapping(mm, vdso_base, PAGE_SIZE,
+				       VM_READ|VM_MAYREAD,
+				       &vdso_spec[1]);
+	if (IS_ERR(ret))
 		goto up_fail;
 
 	up_write(&mm->mmap_sem);
@@ -173,35 +191,7 @@ int arch_setup_additional_pages(struct linux_binprm *bprm,
 up_fail:
 	mm->context.vdso = NULL;
 	up_write(&mm->mmap_sem);
-	return ret;
-}
-
-const char *arch_vma_name(struct vm_area_struct *vma)
-{
-	unsigned long vdso_text;
-
-	if (!vma->vm_mm)
-		return NULL;
-
-	vdso_text = (unsigned long)vma->vm_mm->context.vdso;
-
-	/*
-	 * We can re-use the vdso pointer in mm_context_t for identifying
-	 * the vectors page for compat applications. The vDSO will always
-	 * sit above TASK_UNMAPPED_BASE and so we don't need to worry about
-	 * it conflicting with the vectors base.
-	 */
-	if (vma->vm_start == vdso_text) {
-#ifdef CONFIG_COMPAT
-		if (vma->vm_start == AARCH32_VECTORS_BASE)
-			return "[vectors]";
-#endif
-		return "[vdso]";
-	} else if (vma->vm_start == (vdso_text + (vdso_pages << PAGE_SHIFT))) {
-		return "[vvar]";
-	}
-
-	return NULL;
+	return PTR_ERR(ret);
 }
 
 /*
-- 
2.0.0

[PATCH 3/3] arm64: vdso: move data page before code pages

[Will Deacon]

Andy pointed out that binutils generates additional sections in the vdso
image (e.g. section string table) which, if our .text section gets big
enough, could cross a page boundary and end up screwing up the location
where the kernel expects to put the data page.

This patch solves the issue in the same manner as x86_32, by moving the
data page before the code pages.

Cc: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Will Deacon <will.deacon@arm.com>
---
 arch/arm64/kernel/vdso.c          | 34 +++++++++++++++++-----------------
 arch/arm64/kernel/vdso/vdso.lds.S |  4 +---
 2 files changed, 18 insertions(+), 20 deletions(-)

diff --git a/arch/arm64/kernel/vdso.c b/arch/arm64/kernel/vdso.c
index 60ae12087d9f..24f2e8c62479 100644
--- a/arch/arm64/kernel/vdso.c
+++ b/arch/arm64/kernel/vdso.c
@@ -121,8 +121,8 @@ static int __init vdso_init(void)
 	}
 
 	vdso_pages = (&vdso_end - &vdso_start) >> PAGE_SHIFT;
-	pr_info("vdso: %ld pages (%ld code, %ld data) at base %p\n",
-		vdso_pages + 1, vdso_pages, 1L, &vdso_start);
+	pr_info("vdso: %ld pages (%ld code @ %p, %ld data @ %p)\n",
+		vdso_pages + 1, vdso_pages, &vdso_start, 1L, vdso_data);
 
 	/* Allocate the vDSO pagelist, plus a page for the data. */
 	vdso_pagelist = kcalloc(vdso_pages + 1, sizeof(struct page *),
@@ -130,22 +130,22 @@ static int __init vdso_init(void)
 	if (vdso_pagelist == NULL)
 		return -ENOMEM;
 
+	/* Grab the vDSO data page. */
+	vdso_pagelist[0] = virt_to_page(vdso_data);
+
 	/* Grab the vDSO code pages. */
 	for (i = 0; i < vdso_pages; i++)
-		vdso_pagelist[i] = virt_to_page(&vdso_start + i * PAGE_SIZE);
-
-	/* Grab the vDSO data page. */
-	vdso_pagelist[i] = virt_to_page(vdso_data);
+		vdso_pagelist[i + 1] = virt_to_page(&vdso_start + i * PAGE_SIZE);
 
 	/* Populate the special mapping structures */
 	vdso_spec[0] = (struct vm_special_mapping) {
-		.name	= "[vdso]",
+		.name	= "[vvar]",
 		.pages	= vdso_pagelist,
 	};
 
 	vdso_spec[1] = (struct vm_special_mapping) {
-		.name	= "[vvar]",
-		.pages	= vdso_pagelist + vdso_pages,
+		.name	= "[vdso]",
+		.pages	= &vdso_pagelist[1],
 	};
 
 	return 0;
@@ -169,22 +169,22 @@ int arch_setup_additional_pages(struct linux_binprm *bprm,
 		ret = ERR_PTR(vdso_base);
 		goto up_fail;
 	}
-	mm->context.vdso = (void *)vdso_base;
-
-	ret = _install_special_mapping(mm, vdso_base, vdso_text_len,
-				       VM_READ|VM_EXEC|
-				       VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
+	ret = _install_special_mapping(mm, vdso_base, PAGE_SIZE,
+				       VM_READ|VM_MAYREAD,
 				       &vdso_spec[0]);
 	if (IS_ERR(ret))
 		goto up_fail;
 
-	vdso_base += vdso_text_len;
-	ret = _install_special_mapping(mm, vdso_base, PAGE_SIZE,
-				       VM_READ|VM_MAYREAD,
+	vdso_base += PAGE_SIZE;
+	mm->context.vdso = (void *)vdso_base;
+	ret = _install_special_mapping(mm, vdso_base, vdso_text_len,
+				       VM_READ|VM_EXEC|
+				       VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
 				       &vdso_spec[1]);
 	if (IS_ERR(ret))
 		goto up_fail;
 
+
 	up_write(&mm->mmap_sem);
 	return 0;
 
diff --git a/arch/arm64/kernel/vdso/vdso.lds.S b/arch/arm64/kernel/vdso/vdso.lds.S
index 8154b8d1c826..beca249bc2f3 100644
--- a/arch/arm64/kernel/vdso/vdso.lds.S
+++ b/arch/arm64/kernel/vdso/vdso.lds.S
@@ -28,6 +28,7 @@ OUTPUT_ARCH(aarch64)
 
 SECTIONS
 {
+	PROVIDE(_vdso_data = . - PAGE_SIZE);
 	. = VDSO_LBASE + SIZEOF_HEADERS;
 
 	.hash		: { *(.hash) }			:text
@@ -57,9 +58,6 @@ SECTIONS
 	_end = .;
 	PROVIDE(end = .);
 
-	. = ALIGN(PAGE_SIZE);
-	PROVIDE(_vdso_data = .);
-
 	/DISCARD/	: {
 		*(.note.GNU-stack)
 		*(.data .data.* .gnu.linkonce.d.* .sdata*)
-- 
2.0.0

[PATCH 0/3] Bunch of vdso cleanups following feedback from x86

[Christopher Covington]

Hi Will,

On 07/09/2014 02:22 PM, Will Deacon wrote:
> Hi all,
> 
> Following useful discussions with Andy (thanks!), I've put together the
> following (lightly tested) series for arm64 which attempts to address
> issues with our current vdso code.
> 
> Please take a look. Cheers,

With appropriate updates to Checkpoint and Restore In Userspace (CRIU) [1] and
some finishing touches to its AArch64 port that I hope to publish very soon, I
see no regressions running its VDSO test cases, which exercise basic
discovery, use, and remapping.

Tested-by: Christopher Covington <cov@codeaurora.org>

1. http://lists.openvz.org/pipermail/criu/2014-July/015138.html

Christopher

-- 
Employee of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by the Linux Foundation.