All of lore.kernel.org
 help / color / mirror / Atom feed
* [RFC PATCH v2 0/8] arm64: Add a compat vDSO
@ 2016-10-27 16:30 Kevin Brodsky
  2016-10-27 16:30 ` [RFC PATCH v2 1/8] arm64: Refactor vDSO setup Kevin Brodsky
                   ` (7 more replies)
  0 siblings, 8 replies; 19+ messages in thread
From: Kevin Brodsky @ 2016-10-27 16:30 UTC (permalink / raw)
  To: linux-arm-kernel

Hi,

This series adds support for a compat (AArch32) vDSO, providing two
userspace functionalities to compat processes:

* "Virtual" time syscalls (gettimeofday and clock_gettime). The
  implementation is an adaptation of the arm vDSO (vgettimeofday.c),
  sharing the data page with the 64-bit vDSO.

* sigreturn trampolines, following the example of the 64-bit vDSO
  (sigreturn.S), but slightly more complicated because we provide A32
  and T32 variants for both sigreturn and rt_sigreturn.

The first point brings the performance improvement expected of a vDSO,
by implementing time syscalls directly in userspace. The second point
allows us to get rid of the compat vector page, at the expense of the
kuser helpers (this is one reason for not enabling the compat vDSO by
default).

Unfortunately, this time we cannot escape using a 32-bit toolchain. To
build the compat VDSO, CONFIG_COMPAT_VDSO must be set *and*
CROSS_COMPILE_ARM32 must be defined to the prefix of a 32-bit compiler.
Failure to do so will not prevent building the kernel, but a warning
will be printed and the compat vDSO will not be built.

Thanks,
Kevin

Changelog v1..v2:
* Rebased on 4.9-rc2. I preserved the spirit of 5a9e3e156ec1 ("arm64:
  apply __ro_after_init to some objects") by making the new
  vdso{,32}_mappings static variables __ro_after_init, and removing
  pages from struct vdso_mappings.
* Added CONFIG_CROSS_COMPILE_ARM32, on the same principle as
  CONFIG_CROSS_COMPILE. That may help avoid forgetting to set the
  variable when building the kernel with CONFIG_COMPAT_VDSO.


Kevin Brodsky (8):
  arm64: Refactor vDSO setup
  arm64: compat: Add time-related syscall numbers
  arm64: compat: Expose offset to registers in sigframes
  arm64: compat: Add a 32-bit vDSO
  arm64: compat: 32-bit vDSO setup
  arm64: elf: Set AT_SYSINFO_EHDR in compat processes
  arm64: compat: Use vDSO sigreturn trampolines if available
  arm64: Wire up and expose the new compat vDSO

 arch/arm64/Kconfig                       |  26 +++
 arch/arm64/Makefile                      |  28 ++-
 arch/arm64/include/asm/elf.h             |  15 +-
 arch/arm64/include/asm/signal32.h        |  46 +++++
 arch/arm64/include/asm/unistd.h          |   2 +
 arch/arm64/include/asm/vdso.h            |   3 +
 arch/arm64/kernel/Makefile               |   8 +-
 arch/arm64/kernel/asm-offsets.c          |  13 ++
 arch/arm64/kernel/signal32.c             |  61 ++-----
 arch/arm64/kernel/vdso.c                 | 199 ++++++++++++---------
 arch/arm64/kernel/vdso32/Makefile        | 121 +++++++++++++
 arch/arm64/kernel/vdso32/sigreturn.S     |  86 +++++++++
 arch/arm64/kernel/vdso32/vdso.S          |  32 ++++
 arch/arm64/kernel/vdso32/vdso.lds.S      |  98 +++++++++++
 arch/arm64/kernel/vdso32/vgettimeofday.c | 294 +++++++++++++++++++++++++++++++
 15 files changed, 897 insertions(+), 135 deletions(-)
 create mode 100644 arch/arm64/kernel/vdso32/Makefile
 create mode 100644 arch/arm64/kernel/vdso32/sigreturn.S
 create mode 100644 arch/arm64/kernel/vdso32/vdso.S
 create mode 100644 arch/arm64/kernel/vdso32/vdso.lds.S
 create mode 100644 arch/arm64/kernel/vdso32/vgettimeofday.c

-- 
2.10.0

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 1/8] arm64: Refactor vDSO setup
  2016-10-27 16:30 [RFC PATCH v2 0/8] arm64: Add a compat vDSO Kevin Brodsky
@ 2016-10-27 16:30 ` Kevin Brodsky
  2016-10-27 16:30 ` [RFC PATCH v2 2/8] arm64: compat: Add time-related syscall numbers Kevin Brodsky
                   ` (6 subsequent siblings)
  7 siblings, 0 replies; 19+ messages in thread
From: Kevin Brodsky @ 2016-10-27 16:30 UTC (permalink / raw)
  To: linux-arm-kernel

Move the logic for setting up mappings and pages for the vDSO into
static functions with a clear interface. This will allow to reuse the
setup code for the future compat vDSO.

Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
---
 arch/arm64/kernel/vdso.c | 177 ++++++++++++++++++++++++++---------------------
 1 file changed, 98 insertions(+), 79 deletions(-)

diff --git a/arch/arm64/kernel/vdso.c b/arch/arm64/kernel/vdso.c
index a2c2478e7d78..c239b1c15eb3 100644
--- a/arch/arm64/kernel/vdso.c
+++ b/arch/arm64/kernel/vdso.c
@@ -37,8 +37,10 @@
 #include <asm/vdso.h>
 #include <asm/vdso_datapage.h>
 
-extern char vdso_start, vdso_end;
-static unsigned long vdso_pages __ro_after_init;
+struct vdso_mappings {
+	unsigned long num_pages;
+	struct vm_special_mapping data_mapping, code_mapping;
+};
 
 /*
  * The vDSO data page.
@@ -49,6 +51,92 @@ static union {
 } vdso_data_store __page_aligned_data;
 struct vdso_data *vdso_data = &vdso_data_store.data;
 
+static int __init setup_vdso_mappings(const char *name,
+				      const char *code_start,
+				      const char *code_end,
+				      struct vdso_mappings *mappings)
+{
+	unsigned long i, num_pages;
+	struct page **pages;
+
+	if (memcmp(code_start, "\177ELF", 4)) {
+		pr_err("%s is not a valid ELF object!\n", name);
+		return -EINVAL;
+	}
+
+	num_pages = (code_end - code_start) >> PAGE_SHIFT;
+	pr_info("%s: %ld pages (%ld code @ %p, %ld data @ %p)\n",
+		name, num_pages + 1, num_pages, code_start, 1L, vdso_data);
+
+	/* Allocate the vDSO code pages, plus a page for the data. */
+	pages = kcalloc(num_pages + 1, sizeof(struct page *), GFP_KERNEL);
+	if (pages == NULL)
+		return -ENOMEM;
+
+	/* Grab the vDSO data page. */
+	pages[0] = pfn_to_page(PHYS_PFN(__pa(vdso_data)));
+
+	/* Grab the vDSO code pages. */
+	for (i = 0; i < num_pages; i++)
+		pages[i + 1] = pfn_to_page(PHYS_PFN(__pa(code_start)) + i);
+
+	/* Populate the special mapping structures */
+	mappings->data_mapping = (struct vm_special_mapping) {
+		.name	= "[vvar]",
+		.pages	= &pages[0],
+	};
+
+	mappings->code_mapping = (struct vm_special_mapping) {
+		.name	= "[vdso]",
+		.pages	= &pages[1],
+	};
+
+	mappings->num_pages = num_pages;
+	return 0;
+}
+
+static int setup_vdso_pages(const struct vdso_mappings *mappings)
+{
+	struct mm_struct *mm = current->mm;
+	unsigned long vdso_base, vdso_text_len, vdso_mapping_len;
+	void *ret;
+
+	vdso_text_len = mappings->num_pages << PAGE_SHIFT;
+	/* Be sure to map the data page */
+	vdso_mapping_len = vdso_text_len + PAGE_SIZE;
+
+	if (down_write_killable(&mm->mmap_sem))
+		return -EINTR;
+	vdso_base = get_unmapped_area(NULL, 0, vdso_mapping_len, 0, 0);
+	if (IS_ERR_VALUE(vdso_base)) {
+		ret = ERR_PTR(vdso_base);
+		goto up_fail;
+	}
+	ret = _install_special_mapping(mm, vdso_base, PAGE_SIZE,
+				       VM_READ|VM_MAYREAD,
+				       &mappings->data_mapping);
+	if (IS_ERR(ret))
+		goto up_fail;
+
+	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,
+				       &mappings->code_mapping);
+	if (IS_ERR(ret))
+		goto up_fail;
+
+
+	up_write(&mm->mmap_sem);
+	return 0;
+
+up_fail:
+	mm->context.vdso = NULL;
+	up_write(&mm->mmap_sem);
+	return PTR_ERR(ret);
+}
+
 #ifdef CONFIG_COMPAT
 /*
  * Create and map the vectors page for AArch32 tasks.
@@ -71,11 +159,11 @@ static int __init alloc_vectors_page(void)
 
 	/* kuser helpers */
 	memcpy((void *)vpage + 0x1000 - kuser_sz, __kuser_helper_start,
-		kuser_sz);
+	       kuser_sz);
 
 	/* sigreturn code */
 	memcpy((void *)vpage + AARCH32_KERN_SIGRET_CODE_OFFSET,
-               __aarch32_sigret_code_start, sigret_sz);
+	       __aarch32_sigret_code_start, sigret_sz);
 
 	flush_icache_range(vpage, vpage + PAGE_SIZE);
 	vectors_page[0] = virt_to_page(vpage);
@@ -110,90 +198,21 @@ int aarch32_setup_vectors_page(struct linux_binprm *bprm, int uses_interp)
 }
 #endif /* CONFIG_COMPAT */
 
-static struct vm_special_mapping vdso_spec[2] __ro_after_init = {
-	{
-		.name	= "[vvar]",
-	},
-	{
-		.name	= "[vdso]",
-	},
-};
+extern char vdso_start, vdso_end;
+
+static struct vdso_mappings vdso_mappings __ro_after_init;
 
 static int __init vdso_init(void)
 {
-	int i;
-	struct page **vdso_pagelist;
-
-	if (memcmp(&vdso_start, "\177ELF", 4)) {
-		pr_err("vDSO is not a valid ELF object!\n");
-		return -EINVAL;
-	}
-
-	vdso_pages = (&vdso_end - &vdso_start) >> PAGE_SHIFT;
-	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 *),
-				GFP_KERNEL);
-	if (vdso_pagelist == NULL)
-		return -ENOMEM;
-
-	/* Grab the vDSO data page. */
-	vdso_pagelist[0] = pfn_to_page(PHYS_PFN(__pa(vdso_data)));
-
-	/* Grab the vDSO code pages. */
-	for (i = 0; i < vdso_pages; i++)
-		vdso_pagelist[i + 1] = pfn_to_page(PHYS_PFN(__pa(&vdso_start)) + i);
-
-	vdso_spec[0].pages = &vdso_pagelist[0];
-	vdso_spec[1].pages = &vdso_pagelist[1];
-
-	return 0;
+	return setup_vdso_mappings("vdso", &vdso_start, &vdso_end,
+				   &vdso_mappings);
 }
 arch_initcall(vdso_init);
 
 int arch_setup_additional_pages(struct linux_binprm *bprm,
 				int uses_interp)
 {
-	struct mm_struct *mm = current->mm;
-	unsigned long vdso_base, vdso_text_len, vdso_mapping_len;
-	void *ret;
-
-	vdso_text_len = vdso_pages << PAGE_SHIFT;
-	/* Be sure to map the data page */
-	vdso_mapping_len = vdso_text_len + PAGE_SIZE;
-
-	if (down_write_killable(&mm->mmap_sem))
-		return -EINTR;
-	vdso_base = get_unmapped_area(NULL, 0, vdso_mapping_len, 0, 0);
-	if (IS_ERR_VALUE(vdso_base)) {
-		ret = ERR_PTR(vdso_base);
-		goto up_fail;
-	}
-	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 += 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;
-
-up_fail:
-	mm->context.vdso = NULL;
-	up_write(&mm->mmap_sem);
-	return PTR_ERR(ret);
+	return setup_vdso_pages(&vdso_mappings);
 }
 
 /*
-- 
2.10.0

^ permalink raw reply related	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 2/8] arm64: compat: Add time-related syscall numbers
  2016-10-27 16:30 [RFC PATCH v2 0/8] arm64: Add a compat vDSO Kevin Brodsky
  2016-10-27 16:30 ` [RFC PATCH v2 1/8] arm64: Refactor vDSO setup Kevin Brodsky
@ 2016-10-27 16:30 ` Kevin Brodsky
  2016-10-27 16:30 ` [RFC PATCH v2 3/8] arm64: compat: Expose offset to registers in sigframes Kevin Brodsky
                   ` (5 subsequent siblings)
  7 siblings, 0 replies; 19+ messages in thread
From: Kevin Brodsky @ 2016-10-27 16:30 UTC (permalink / raw)
  To: linux-arm-kernel

They will be used by the future compat vDSO.

The compat syscall numbers correspond to the arm syscall numbers, see
arch/arm/include/uapi/asm/unistd.h.

Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
---
 arch/arm64/include/asm/unistd.h | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/arm64/include/asm/unistd.h b/arch/arm64/include/asm/unistd.h
index e78ac26324bd..8d1c5f5e58f3 100644
--- a/arch/arm64/include/asm/unistd.h
+++ b/arch/arm64/include/asm/unistd.h
@@ -34,8 +34,10 @@
 #define __NR_compat_exit		1
 #define __NR_compat_read		3
 #define __NR_compat_write		4
+#define __NR_compat_gettimeofday	78
 #define __NR_compat_sigreturn		119
 #define __NR_compat_rt_sigreturn	173
+#define __NR_compat_clock_gettime	263
 
 /*
  * The following SVCs are ARM private.
-- 
2.10.0

^ permalink raw reply related	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 3/8] arm64: compat: Expose offset to registers in sigframes
  2016-10-27 16:30 [RFC PATCH v2 0/8] arm64: Add a compat vDSO Kevin Brodsky
  2016-10-27 16:30 ` [RFC PATCH v2 1/8] arm64: Refactor vDSO setup Kevin Brodsky
  2016-10-27 16:30 ` [RFC PATCH v2 2/8] arm64: compat: Add time-related syscall numbers Kevin Brodsky
@ 2016-10-27 16:30 ` Kevin Brodsky
  2016-10-27 16:30 ` [RFC PATCH v2 4/8] arm64: compat: Add a 32-bit vDSO Kevin Brodsky
                   ` (4 subsequent siblings)
  7 siblings, 0 replies; 19+ messages in thread
From: Kevin Brodsky @ 2016-10-27 16:30 UTC (permalink / raw)
  To: linux-arm-kernel

This will be needed to provide debug information (CFI/unwind tables)
in compat sigreturn trampolines, part of the future compat vDSO.

Also fix minor style issues reported by checkpatch.

Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
---
 arch/arm64/include/asm/signal32.h | 46 +++++++++++++++++++++++++++++++++++++++
 arch/arm64/kernel/asm-offsets.c   | 13 +++++++++++
 arch/arm64/kernel/signal32.c      | 46 ---------------------------------------
 3 files changed, 59 insertions(+), 46 deletions(-)

diff --git a/arch/arm64/include/asm/signal32.h b/arch/arm64/include/asm/signal32.h
index eeaa97559bab..9b1bcd3abd83 100644
--- a/arch/arm64/include/asm/signal32.h
+++ b/arch/arm64/include/asm/signal32.h
@@ -20,6 +20,52 @@
 #ifdef CONFIG_COMPAT
 #include <linux/compat.h>
 
+struct compat_sigcontext {
+	/* We always set these two fields to 0 */
+	compat_ulong_t			trap_no;
+	compat_ulong_t			error_code;
+
+	compat_ulong_t			oldmask;
+	compat_ulong_t			arm_r0;
+	compat_ulong_t			arm_r1;
+	compat_ulong_t			arm_r2;
+	compat_ulong_t			arm_r3;
+	compat_ulong_t			arm_r4;
+	compat_ulong_t			arm_r5;
+	compat_ulong_t			arm_r6;
+	compat_ulong_t			arm_r7;
+	compat_ulong_t			arm_r8;
+	compat_ulong_t			arm_r9;
+	compat_ulong_t			arm_r10;
+	compat_ulong_t			arm_fp;
+	compat_ulong_t			arm_ip;
+	compat_ulong_t			arm_sp;
+	compat_ulong_t			arm_lr;
+	compat_ulong_t			arm_pc;
+	compat_ulong_t			arm_cpsr;
+	compat_ulong_t			fault_address;
+};
+
+struct compat_ucontext {
+	compat_ulong_t			uc_flags;
+	compat_uptr_t			uc_link;
+	compat_stack_t			uc_stack;
+	struct compat_sigcontext	uc_mcontext;
+	compat_sigset_t			uc_sigmask;
+	int		__unused[32 - (sizeof(compat_sigset_t) / sizeof(int))];
+	compat_ulong_t			uc_regspace[128] __aligned(8);
+};
+
+struct compat_sigframe {
+	struct compat_ucontext	uc;
+	compat_ulong_t		retcode[2];
+};
+
+struct compat_rt_sigframe {
+	struct compat_siginfo info;
+	struct compat_sigframe sig;
+};
+
 #define AARCH32_KERN_SIGRET_CODE_OFFSET	0x500
 
 extern const compat_ulong_t aarch32_sigret_code[6];
diff --git a/arch/arm64/kernel/asm-offsets.c b/arch/arm64/kernel/asm-offsets.c
index 4a2f0f0fef32..965371186a5a 100644
--- a/arch/arm64/kernel/asm-offsets.c
+++ b/arch/arm64/kernel/asm-offsets.c
@@ -26,6 +26,7 @@
 #include <asm/cpufeature.h>
 #include <asm/thread_info.h>
 #include <asm/memory.h>
+#include <asm/signal32.h>
 #include <asm/smp_plat.h>
 #include <asm/suspend.h>
 #include <asm/vdso_datapage.h>
@@ -75,6 +76,18 @@ int main(void)
   DEFINE(S_ORIG_ADDR_LIMIT,	offsetof(struct pt_regs, orig_addr_limit));
   DEFINE(S_FRAME_SIZE,		sizeof(struct pt_regs));
   BLANK();
+#ifdef CONFIG_COMPAT
+  DEFINE(COMPAT_SIGFRAME_REGS_OFFSET,
+				offsetof(struct compat_sigframe, uc) +
+				offsetof(struct compat_ucontext, uc_mcontext) +
+				offsetof(struct compat_sigcontext, arm_r0));
+  DEFINE(COMPAT_RT_SIGFRAME_REGS_OFFSET,
+				offsetof(struct compat_rt_sigframe, sig) +
+				offsetof(struct compat_sigframe, uc) +
+				offsetof(struct compat_ucontext, uc_mcontext) +
+				offsetof(struct compat_sigcontext, arm_r0));
+  BLANK();
+#endif
   DEFINE(MM_CONTEXT_ID,		offsetof(struct mm_struct, context.id.counter));
   BLANK();
   DEFINE(VMA_VM_MM,		offsetof(struct vm_area_struct, vm_mm));
diff --git a/arch/arm64/kernel/signal32.c b/arch/arm64/kernel/signal32.c
index b7063de792f7..9de7d128e0e0 100644
--- a/arch/arm64/kernel/signal32.c
+++ b/arch/arm64/kernel/signal32.c
@@ -29,42 +29,6 @@
 #include <asm/uaccess.h>
 #include <asm/unistd.h>
 
-struct compat_sigcontext {
-	/* We always set these two fields to 0 */
-	compat_ulong_t			trap_no;
-	compat_ulong_t			error_code;
-
-	compat_ulong_t			oldmask;
-	compat_ulong_t			arm_r0;
-	compat_ulong_t			arm_r1;
-	compat_ulong_t			arm_r2;
-	compat_ulong_t			arm_r3;
-	compat_ulong_t			arm_r4;
-	compat_ulong_t			arm_r5;
-	compat_ulong_t			arm_r6;
-	compat_ulong_t			arm_r7;
-	compat_ulong_t			arm_r8;
-	compat_ulong_t			arm_r9;
-	compat_ulong_t			arm_r10;
-	compat_ulong_t			arm_fp;
-	compat_ulong_t			arm_ip;
-	compat_ulong_t			arm_sp;
-	compat_ulong_t			arm_lr;
-	compat_ulong_t			arm_pc;
-	compat_ulong_t			arm_cpsr;
-	compat_ulong_t			fault_address;
-};
-
-struct compat_ucontext {
-	compat_ulong_t			uc_flags;
-	compat_uptr_t			uc_link;
-	compat_stack_t			uc_stack;
-	struct compat_sigcontext	uc_mcontext;
-	compat_sigset_t			uc_sigmask;
-	int		__unused[32 - (sizeof (compat_sigset_t) / sizeof (int))];
-	compat_ulong_t	uc_regspace[128] __attribute__((__aligned__(8)));
-};
-
 struct compat_vfp_sigframe {
 	compat_ulong_t	magic;
 	compat_ulong_t	size;
@@ -91,16 +55,6 @@ struct compat_aux_sigframe {
 	unsigned long			end_magic;
 } __attribute__((__aligned__(8)));
 
-struct compat_sigframe {
-	struct compat_ucontext	uc;
-	compat_ulong_t		retcode[2];
-};
-
-struct compat_rt_sigframe {
-	struct compat_siginfo info;
-	struct compat_sigframe sig;
-};
-
 #define _BLOCKABLE (~(sigmask(SIGKILL) | sigmask(SIGSTOP)))
 
 static inline int put_sigset_t(compat_sigset_t __user *uset, sigset_t *set)
-- 
2.10.0

^ permalink raw reply related	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 4/8] arm64: compat: Add a 32-bit vDSO
  2016-10-27 16:30 [RFC PATCH v2 0/8] arm64: Add a compat vDSO Kevin Brodsky
                   ` (2 preceding siblings ...)
  2016-10-27 16:30 ` [RFC PATCH v2 3/8] arm64: compat: Expose offset to registers in sigframes Kevin Brodsky
@ 2016-10-27 16:30 ` Kevin Brodsky
  2016-10-28  3:09   ` Jisheng Zhang
  2016-10-27 16:30 ` [RFC PATCH v2 5/8] arm64: compat: 32-bit vDSO setup Kevin Brodsky
                   ` (3 subsequent siblings)
  7 siblings, 1 reply; 19+ messages in thread
From: Kevin Brodsky @ 2016-10-27 16:30 UTC (permalink / raw)
  To: linux-arm-kernel

Provide the files necessary for building a compat (AArch32) vDSO in
kernel/vdso32.

This is mostly an adaptation of the arm vDSO. The most significant
change in vgettimeofday.c is the use of the arm64 vdso_data struct,
allowing the vDSO data page to be shared between the 32 and 64-bit
vDSOs.

In addition to the time functions, sigreturn trampolines are also
provided, aiming at replacing those in the vector page. To improve
debugging, CFI and unwinding directives are used, based on glibc's
implementation. Symbol offsets are made available to the kernel using
the same method as the 64-bit vDSO.

There is unfortunately an important caveat to all this: we cannot get
away with hand-coding 32-bit instructions like in kernel/kuser32.S,
this time we really need a 32-bit compiler. The compat vDSO Makefile
relies on CROSS_COMPILE_ARM32 to provide a 32-bit compiler,
appropriate logic will be added to the arm64 Makefile later on to
ensure that an attempt to build the compat vDSO is made only if this
variable has been set properly.

Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
---
 arch/arm64/kernel/vdso32/Makefile        | 121 +++++++++++++
 arch/arm64/kernel/vdso32/sigreturn.S     |  86 +++++++++
 arch/arm64/kernel/vdso32/vdso.S          |  32 ++++
 arch/arm64/kernel/vdso32/vdso.lds.S      |  98 +++++++++++
 arch/arm64/kernel/vdso32/vgettimeofday.c | 294 +++++++++++++++++++++++++++++++
 5 files changed, 631 insertions(+)
 create mode 100644 arch/arm64/kernel/vdso32/Makefile
 create mode 100644 arch/arm64/kernel/vdso32/sigreturn.S
 create mode 100644 arch/arm64/kernel/vdso32/vdso.S
 create mode 100644 arch/arm64/kernel/vdso32/vdso.lds.S
 create mode 100644 arch/arm64/kernel/vdso32/vgettimeofday.c

diff --git a/arch/arm64/kernel/vdso32/Makefile b/arch/arm64/kernel/vdso32/Makefile
new file mode 100644
index 000000000000..38facc870f6e
--- /dev/null
+++ b/arch/arm64/kernel/vdso32/Makefile
@@ -0,0 +1,121 @@
+#
+# Building a vDSO image for AArch32.
+#
+# Author: Kevin Brodsky <kevin.brodsky@arm.com>
+# A mix between the arm64 and arm vDSO Makefiles.
+
+CC_ARM32 := $(CROSS_COMPILE_ARM32)gcc
+
+# Same as cc-ldoption, but using CC_ARM32 instead of CC
+cc32-ldoption = $(call try-run,\
+        $(CC_ARM32) $(1) -nostdlib -x c /dev/null -o "$$TMP",$(1),$(2))
+
+# Borrow vdsomunge.c from the arm vDSO
+munge := arch/arm/vdso/vdsomunge
+hostprogs-y := $(srctree)/$(munge)
+
+c-obj-vdso := vgettimeofday.o
+asm-obj-vdso := sigreturn.o
+
+# Build rules
+targets := $(c-obj-vdso) $(asm-obj-vdso) vdso.so vdso.so.dbg vdso.so.raw
+c-obj-vdso := $(addprefix $(obj)/, $(c-obj-vdso))
+asm-obj-vdso := $(addprefix $(obj)/, $(asm-obj-vdso))
+obj-vdso := $(c-obj-vdso) $(asm-obj-vdso)
+
+ccflags-y := -fPIC -fno-common -fno-builtin -fno-stack-protector
+ccflags-y += -DDISABLE_BRANCH_PROFILING
+
+# Force -O2 to avoid libgcc dependencies
+VDSO_CFLAGS := -march=armv8-a -O2
+# Import some useful flags from arch/arm/Makefile
+VDSO_CFLAGS += -mabi=aapcs-linux -mfloat-abi=soft -funwind-tables
+# The 32-bit compiler does not provide 128-bit integers, which are used in
+# some headers that are indirectly included from the vDSO code.
+# This hack makes the compiler happy and should trigger a warning/error if
+# variables of such type are referenced.
+VDSO_CFLAGS += -D__uint128_t='void*'
+# Silence some warnings coming from headers that operate on long's
+VDSO_CFLAGS += -Wno-shift-count-overflow -Wno-int-to-pointer-cast
+
+# We need to use the global flags to compile the vDSO files. However some flags
+# inherited from either the top-level or the arm64 Makefile are not appropriate
+# for the 32-bit compiler, this function takes care of changing them as
+# appropriate.
+sanitize_flags = \
+        $(subst $(shell $(CC) -print-file-name=include), \
+                $(shell $(CC_ARM32) -print-file-name=include), \
+        $(filter-out -pg -mgeneral-regs-only -mpc-relative-literal-loads \
+                     -fno-asynchronous-unwind-tables, \
+        $(1)))
+
+VDSO_LDFLAGS := -Wl,-Bsymbolic -Wl,--no-undefined -Wl,-soname=linux-vdso.so.1
+VDSO_LDFLAGS += -Wl,-z,max-page-size=4096 -Wl,-z,common-page-size=4096
+VDSO_LDFLAGS += -nostdlib -shared -mfloat-abi=soft
+VDSO_LDFLAGS += $(call cc32-ldoption, -Wl$(comma)--hash-style=sysv)
+VDSO_LDFLAGS += $(call cc32-ldoption, -Wl$(comma)--build-id)
+VDSO_LDFLAGS += $(call cc32-ldoption, -fuse-ld=bfd)
+
+obj-y += vdso.o
+extra-y += vdso.lds
+CPPFLAGS_vdso.lds += -P -C -U$(ARCH)
+
+CFLAGS_REMOVE_vdso.o = -pg
+
+# Disable gcov profiling for VDSO code
+GCOV_PROFILE := n
+
+# Force dependency (incbin is bad)
+$(obj)/vdso.o: $(obj)/vdso.so
+
+# Link rule for the .so file, .lds has to be first
+$(obj)/vdso.so.raw: $(src)/vdso.lds $(obj-vdso) FORCE
+	$(call if_changed,vdsold)
+
+$(obj)/vdso.so.dbg: $(obj)/vdso.so.raw $(objtree)/$(munge) FORCE
+	$(call if_changed,vdsomunge)
+
+# Strip rule for the .so file
+$(obj)/%.so: OBJCOPYFLAGS := -S
+$(obj)/%.so: $(obj)/%.so.dbg FORCE
+	$(call if_changed,objcopy)
+
+# Generate vDSO offsets using helper script (borrowed from the 64-bit vDSO)
+gen-vdsosym := $(srctree)/$(src)/../vdso/gen_vdso_offsets.sh
+quiet_cmd_vdsosym = VDSOSYM $@
+# The AArch64 nm should be able to read an AArch32 binary
+define cmd_vdsosym
+	$(NM) $< | $(gen-vdsosym) | LC_ALL=C sort > $@
+endef
+
+include/generated/vdso32-offsets.h: $(obj)/vdso.so.dbg FORCE
+	$(call if_changed,vdsosym)
+
+# Compilation rules for the vDSO sources
+$(c-obj-vdso): %.o: %.c FORCE
+	$(call if_changed_dep,vdsocc)
+$(asm-obj-vdso): %.o: %.S FORCE
+	$(call if_changed_dep,vdsoas)
+
+# Actual build commands
+quiet_cmd_vdsold = VDSOL   $@
+      cmd_vdsold = $(CC_ARM32) $(call sanitize_flags,$(c_flags)) \
+                   $(VDSO_LDFLAGS) -Wl,-T $(filter %.lds,$^) $(filter %.o,$^) -o $@
+quiet_cmd_vdsocc = VDSOC   $@
+      cmd_vdsocc = $(CC_ARM32) $(call sanitize_flags,$(c_flags)) \
+                   $(VDSO_CFLAGS) -c -o $@ $<
+quiet_cmd_vdsoas = VDSOA   $@
+      cmd_vdsoas = $(CC_ARM32) $(call sanitize_flags, $(a_flags)) -c -o $@ $<
+
+quiet_cmd_vdsomunge = MUNGE   $@
+      cmd_vdsomunge = $(objtree)/$(munge) $< $@
+
+# Install commands for the unstripped file
+quiet_cmd_vdso_install = INSTALL $@
+      cmd_vdso_install = cp $(obj)/$@.dbg $(MODLIB)/vdso/vdso32.so
+
+vdso.so: $(obj)/vdso.so.dbg
+	@mkdir -p $(MODLIB)/vdso
+	$(call cmd,vdso_install)
+
+vdso_install: vdso.so
diff --git a/arch/arm64/kernel/vdso32/sigreturn.S b/arch/arm64/kernel/vdso32/sigreturn.S
new file mode 100644
index 000000000000..a203140ec491
--- /dev/null
+++ b/arch/arm64/kernel/vdso32/sigreturn.S
@@ -0,0 +1,86 @@
+/*
+ * Sigreturn trampolines for returning from a signal when the SA_RESTORER
+ * flag is not set.
+ *
+ * Copyright (C) 2016 ARM Limited
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Based on glibc's arm sa_restorer. While this is not strictly necessary, we
+ * provide both A32 and T32 versions, in accordance with the arm sigreturn
+ * code.
+ */
+
+#include <linux/linkage.h>
+#include <asm/asm-offsets.h>
+#include <asm/unistd.h>
+
+.macro cfi_regs offset
+	.cfi_def_cfa sp, 0
+	.cfi_offset r0, \offset + 0 * 4
+	.cfi_offset r1, \offset + 1 * 4
+	.cfi_offset r2, \offset + 2 * 4
+	.cfi_offset r3, \offset + 3 * 4
+	.cfi_offset r4, \offset + 4 * 4
+	.cfi_offset r5, \offset + 5 * 4
+	.cfi_offset r6, \offset + 6 * 4
+	.cfi_offset r7, \offset + 7 * 4
+	.cfi_offset r8, \offset + 8 * 4
+	.cfi_offset r9, \offset + 9 * 4
+	.cfi_offset r10, \offset + 10 * 4
+	.cfi_offset r11, \offset + 11 * 4
+	.cfi_offset r12, \offset + 12 * 4
+	.cfi_offset r13, \offset + 13 * 4
+	.cfi_offset r14, \offset + 14 * 4
+	.cfi_offset r15, \offset + 15 * 4
+.endm
+
+.macro sigreturn_trampoline name, syscall, regs_offset
+	.fnstart
+	.save {r0-r15}
+	.pad #\regs_offset
+ENTRY(\name)
+	.cfi_startproc
+	.cfi_signal_frame
+	cfi_regs \regs_offset
+	mov	r7, #\syscall
+	svc	#0
+	.fnend
+	.cfi_endproc
+/*
+ * We would like to use ENDPROC, but the macro uses @ which is a comment symbol
+ * for arm assemblers, so directly use .type with % instead.
+ */
+	.type \name, %function
+END(\name)
+.endm
+
+	.text
+
+	.arm
+	sigreturn_trampoline __kernel_sigreturn_arm, \
+			     __NR_compat_sigreturn, \
+			     COMPAT_SIGFRAME_REGS_OFFSET
+
+	sigreturn_trampoline __kernel_rt_sigreturn_arm, \
+			     __NR_compat_rt_sigreturn, \
+			     COMPAT_RT_SIGFRAME_REGS_OFFSET
+
+	.thumb
+	sigreturn_trampoline __kernel_sigreturn_thumb, \
+			     __NR_compat_sigreturn, \
+			     COMPAT_SIGFRAME_REGS_OFFSET
+
+	sigreturn_trampoline __kernel_rt_sigreturn_thumb, \
+			     __NR_compat_rt_sigreturn, \
+			     COMPAT_RT_SIGFRAME_REGS_OFFSET
diff --git a/arch/arm64/kernel/vdso32/vdso.S b/arch/arm64/kernel/vdso32/vdso.S
new file mode 100644
index 000000000000..fe19ff70eb76
--- /dev/null
+++ b/arch/arm64/kernel/vdso32/vdso.S
@@ -0,0 +1,32 @@
+/*
+ * Copyright (C) 2012 ARM Limited
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Author: Will Deacon <will.deacon@arm.com>
+ */
+
+#include <linux/init.h>
+#include <linux/linkage.h>
+#include <linux/const.h>
+#include <asm/page.h>
+
+	.globl vdso32_start, vdso32_end
+	.section .rodata
+	.balign PAGE_SIZE
+vdso32_start:
+	.incbin "arch/arm64/kernel/vdso32/vdso.so"
+	.balign PAGE_SIZE
+vdso32_end:
+
+	.previous
diff --git a/arch/arm64/kernel/vdso32/vdso.lds.S b/arch/arm64/kernel/vdso32/vdso.lds.S
new file mode 100644
index 000000000000..95abcc0dd37e
--- /dev/null
+++ b/arch/arm64/kernel/vdso32/vdso.lds.S
@@ -0,0 +1,98 @@
+/*
+ * Adapted from arm64 version.
+ *
+ * GNU linker script for the VDSO library.
+ *
+ * Copyright (C) 2012 ARM Limited
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ *
+ * Author: Will Deacon <will.deacon@arm.com>
+ * Heavily based on the vDSO linker scripts for other archs.
+ */
+
+#include <linux/const.h>
+#include <asm/page.h>
+#include <asm/vdso.h>
+
+OUTPUT_FORMAT("elf32-littlearm", "elf32-bigarm", "elf32-littlearm")
+OUTPUT_ARCH(arm)
+
+SECTIONS
+{
+	HIDDEN(_vdso_data = . - PAGE_SIZE);
+	. = VDSO_LBASE + SIZEOF_HEADERS;
+
+	.hash		: { *(.hash) }			:text
+	.gnu.hash	: { *(.gnu.hash) }
+	.dynsym		: { *(.dynsym) }
+	.dynstr		: { *(.dynstr) }
+	.gnu.version	: { *(.gnu.version) }
+	.gnu.version_d	: { *(.gnu.version_d) }
+	.gnu.version_r	: { *(.gnu.version_r) }
+
+	.note		: { *(.note.*) }		:text	:note
+
+
+	.eh_frame_hdr	: { *(.eh_frame_hdr) }		:text	:eh_frame_hdr
+	.eh_frame	: { KEEP (*(.eh_frame)) }	:text
+
+	.dynamic	: { *(.dynamic) }		:text	:dynamic
+
+	.rodata		: { *(.rodata*) }		:text
+
+	.text		: { *(.text*) }			:text	=0xe7f001f2
+
+	.got		: { *(.got) }
+	.rel.plt	: { *(.rel.plt) }
+
+	/DISCARD/	: {
+		*(.note.GNU-stack)
+		*(.data .data.* .gnu.linkonce.d.* .sdata*)
+		*(.bss .sbss .dynbss .dynsbss)
+	}
+}
+
+/*
+ * We must supply the ELF program headers explicitly to get just one
+ * PT_LOAD segment, and set the flags explicitly to make segments read-only.
+ */
+PHDRS
+{
+	text		PT_LOAD		FLAGS(5) FILEHDR PHDRS; /* PF_R|PF_X */
+	dynamic		PT_DYNAMIC	FLAGS(4);		/* PF_R */
+	note		PT_NOTE		FLAGS(4);		/* PF_R */
+	eh_frame_hdr	PT_GNU_EH_FRAME;
+}
+
+VERSION
+{
+	LINUX_2.6 {
+	global:
+		__vdso_clock_gettime;
+		__vdso_gettimeofday;
+		__kernel_sigreturn_arm;
+		__kernel_sigreturn_thumb;
+		__kernel_rt_sigreturn_arm;
+		__kernel_rt_sigreturn_thumb;
+	local: *;
+	};
+}
+
+/*
+ * Make the sigreturn code visible to the kernel.
+ */
+VDSO_compat_sigreturn_arm	= __kernel_sigreturn_arm;
+VDSO_compat_sigreturn_thumb	= __kernel_sigreturn_thumb;
+VDSO_compat_rt_sigreturn_arm	= __kernel_rt_sigreturn_arm;
+VDSO_compat_rt_sigreturn_thumb	= __kernel_rt_sigreturn_thumb;
diff --git a/arch/arm64/kernel/vdso32/vgettimeofday.c b/arch/arm64/kernel/vdso32/vgettimeofday.c
new file mode 100644
index 000000000000..3591fd56f8a6
--- /dev/null
+++ b/arch/arm64/kernel/vdso32/vgettimeofday.c
@@ -0,0 +1,294 @@
+/*
+ * Copyright 2015 Mentor Graphics Corporation.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; version 2 of the
+ * License.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <linux/clocksource.h>
+#include <linux/compiler.h>
+#include <linux/time.h>
+#include <asm/barrier.h>
+#include <asm/unistd.h>
+#include <asm/vdso_datapage.h>
+
+/*
+ * We use the hidden visibility to prevent the compiler from generating a GOT
+ * relocation. Not only is going through a GOT useless (the entry couldn't and
+ * musn't be overridden by another library), it does not even work: the linker
+ * cannot generate an absolute address to the data page.
+ *
+ * With the hidden visibility, the compiler simply generates a PC-relative
+ * relocation (R_ARM_REL32), and this is what we need.
+ */
+extern const struct vdso_data _vdso_data __attribute__((visibility("hidden")));
+
+static inline const struct vdso_data *get_vdso_data(void)
+{
+	const struct vdso_data *ret;
+	/*
+	 * This simply puts &_vdso_data into ret. The reason why we don't use
+	 * "ret = &_vdso_data" is that the compiler tends to optimise this in a
+	 * very suboptimal way: instead of keeping &_vdso_data in a register,
+	 * it goes through a relocation almost every time _vdso_data must be
+	 * accessed (even in subfunctions). This is both time and space
+	 * consuming: each relocation uses a word in the code section, and it
+	 * has to be loaded at runtime.
+	 *
+	 * This trick hides the assignment from the compiler. Since it cannot
+	 * track where the pointer comes from, it will only use one relocation
+	 * where get_vdso_data() is called, and then keep the result in a
+	 * register.
+	 */
+	asm("mov %0, %1" : "=r"(ret) : "r"(&_vdso_data));
+	return ret;
+}
+
+static notrace u32 __vdso_read_begin(const struct vdso_data *vdata)
+{
+	u32 seq;
+repeat:
+	seq = ACCESS_ONCE(vdata->tb_seq_count);
+	if (seq & 1) {
+		cpu_relax();
+		goto repeat;
+	}
+	return seq;
+}
+
+static notrace u32 vdso_read_begin(const struct vdso_data *vdata)
+{
+	u32 seq;
+
+	seq = __vdso_read_begin(vdata);
+
+	smp_rmb(); /* Pairs with smp_wmb in vdso_write_end */
+	return seq;
+}
+
+static notrace int vdso_read_retry(const struct vdso_data *vdata, u32 start)
+{
+	smp_rmb(); /* Pairs with smp_wmb in vdso_write_begin */
+	return vdata->tb_seq_count != start;
+}
+
+/*
+ * Note: only AEABI is supported by the compat layer, we can assume AEABI
+ * syscall conventions are used.
+ */
+static notrace long clock_gettime_fallback(clockid_t _clkid,
+					   struct timespec *_ts)
+{
+	register struct timespec *ts asm("r1") = _ts;
+	register clockid_t clkid asm("r0") = _clkid;
+	register long ret asm ("r0");
+	register long nr asm("r7") = __NR_compat_clock_gettime;
+
+	asm volatile(
+	"	svc #0\n"
+	: "=r" (ret)
+	: "r" (clkid), "r" (ts), "r" (nr)
+	: "memory");
+
+	return ret;
+}
+
+static notrace int do_realtime_coarse(struct timespec *ts,
+				      const struct vdso_data *vdata)
+{
+	u32 seq;
+
+	do {
+		seq = vdso_read_begin(vdata);
+
+		ts->tv_sec = vdata->xtime_coarse_sec;
+		ts->tv_nsec = vdata->xtime_coarse_nsec;
+
+	} while (vdso_read_retry(vdata, seq));
+
+	return 0;
+}
+
+static notrace int do_monotonic_coarse(struct timespec *ts,
+				       const struct vdso_data *vdata)
+{
+	struct timespec tomono;
+	u32 seq;
+
+	do {
+		seq = vdso_read_begin(vdata);
+
+		ts->tv_sec = vdata->xtime_coarse_sec;
+		ts->tv_nsec = vdata->xtime_coarse_nsec;
+
+		tomono.tv_sec = vdata->wtm_clock_sec;
+		tomono.tv_nsec = vdata->wtm_clock_nsec;
+
+	} while (vdso_read_retry(vdata, seq));
+
+	ts->tv_sec += tomono.tv_sec;
+	timespec_add_ns(ts, tomono.tv_nsec);
+
+	return 0;
+}
+
+static notrace u64 get_ns(const struct vdso_data *vdata)
+{
+	u64 cycle_delta;
+	u64 cycle_now;
+	u64 nsec;
+
+	/* AArch32 implementation of arch_counter_get_cntvct() */
+	isb();
+	asm volatile("mrrc p15, 1, %Q0, %R0, c14" : "=r" (cycle_now));
+
+	/* The virtual counter provides 56 significant bits. */
+	cycle_delta = (cycle_now - vdata->cs_cycle_last) & CLOCKSOURCE_MASK(56);
+
+	nsec = (cycle_delta * vdata->cs_mono_mult) + vdata->xtime_clock_nsec;
+	nsec >>= vdata->cs_shift;
+
+	return nsec;
+}
+
+static notrace int do_realtime(struct timespec *ts,
+			       const struct vdso_data *vdata)
+{
+	u64 nsecs;
+	u32 seq;
+
+	do {
+		seq = vdso_read_begin(vdata);
+
+		if (vdata->use_syscall)
+			return -1;
+
+		ts->tv_sec = vdata->xtime_clock_sec;
+		nsecs = get_ns(vdata);
+
+	} while (vdso_read_retry(vdata, seq));
+
+	ts->tv_nsec = 0;
+	timespec_add_ns(ts, nsecs);
+
+	return 0;
+}
+
+static notrace int do_monotonic(struct timespec *ts,
+				const struct vdso_data *vdata)
+{
+	struct timespec tomono;
+	u64 nsecs;
+	u32 seq;
+
+	do {
+		seq = vdso_read_begin(vdata);
+
+		if (vdata->use_syscall)
+			return -1;
+
+		ts->tv_sec = vdata->xtime_clock_sec;
+		nsecs = get_ns(vdata);
+
+		tomono.tv_sec = vdata->wtm_clock_sec;
+		tomono.tv_nsec = vdata->wtm_clock_nsec;
+
+	} while (vdso_read_retry(vdata, seq));
+
+	ts->tv_sec += tomono.tv_sec;
+	ts->tv_nsec = 0;
+	timespec_add_ns(ts, nsecs + tomono.tv_nsec);
+
+	return 0;
+}
+
+notrace int __vdso_clock_gettime(clockid_t clkid, struct timespec *ts)
+{
+	const struct vdso_data *vdata = get_vdso_data();
+	int ret = -1;
+
+	switch (clkid) {
+	case CLOCK_REALTIME_COARSE:
+		ret = do_realtime_coarse(ts, vdata);
+		break;
+	case CLOCK_MONOTONIC_COARSE:
+		ret = do_monotonic_coarse(ts, vdata);
+		break;
+	case CLOCK_REALTIME:
+		ret = do_realtime(ts, vdata);
+		break;
+	case CLOCK_MONOTONIC:
+		ret = do_monotonic(ts, vdata);
+		break;
+	default:
+		break;
+	}
+
+	if (ret)
+		ret = clock_gettime_fallback(clkid, ts);
+
+	return ret;
+}
+
+static notrace long gettimeofday_fallback(struct timeval *_tv,
+					  struct timezone *_tz)
+{
+	register struct timezone *tz asm("r1") = _tz;
+	register struct timeval *tv asm("r0") = _tv;
+	register long ret asm ("r0");
+	register long nr asm("r7") = __NR_compat_gettimeofday;
+
+	asm volatile(
+	"	svc #0\n"
+	: "=r" (ret)
+	: "r" (tv), "r" (tz), "r" (nr)
+	: "memory");
+
+	return ret;
+}
+
+notrace int __vdso_gettimeofday(struct timeval *tv, struct timezone *tz)
+{
+	struct timespec ts;
+	const struct vdso_data *vdata = get_vdso_data();
+	int ret;
+
+	ret = do_realtime(&ts, vdata);
+	if (ret)
+		return gettimeofday_fallback(tv, tz);
+
+	if (tv) {
+		tv->tv_sec = ts.tv_sec;
+		tv->tv_usec = ts.tv_nsec / 1000;
+	}
+	if (tz) {
+		tz->tz_minuteswest = vdata->tz_minuteswest;
+		tz->tz_dsttime = vdata->tz_dsttime;
+	}
+
+	return ret;
+}
+
+/* Avoid unresolved references emitted by GCC */
+
+void __aeabi_unwind_cpp_pr0(void)
+{
+}
+
+void __aeabi_unwind_cpp_pr1(void)
+{
+}
+
+void __aeabi_unwind_cpp_pr2(void)
+{
+}
-- 
2.10.0

^ permalink raw reply related	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 5/8] arm64: compat: 32-bit vDSO setup
  2016-10-27 16:30 [RFC PATCH v2 0/8] arm64: Add a compat vDSO Kevin Brodsky
                   ` (3 preceding siblings ...)
  2016-10-27 16:30 ` [RFC PATCH v2 4/8] arm64: compat: Add a 32-bit vDSO Kevin Brodsky
@ 2016-10-27 16:30 ` Kevin Brodsky
  2016-10-27 16:30 ` [RFC PATCH v2 6/8] arm64: elf: Set AT_SYSINFO_EHDR in compat processes Kevin Brodsky
                   ` (2 subsequent siblings)
  7 siblings, 0 replies; 19+ messages in thread
From: Kevin Brodsky @ 2016-10-27 16:30 UTC (permalink / raw)
  To: linux-arm-kernel

If the compat vDSO is enabled, install it in compat processes. In this
case, the compat vDSO replaces the vector page.

aarch32_setup_vectors_page has also been renamed to the more generic
aarch32_setup_additional_pages to reflect both use cases.

Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
---
 arch/arm64/include/asm/elf.h |  6 +++---
 arch/arm64/kernel/vdso.c     | 22 +++++++++++++++++++++-
 2 files changed, 24 insertions(+), 4 deletions(-)

diff --git a/arch/arm64/include/asm/elf.h b/arch/arm64/include/asm/elf.h
index a55384f4a5d7..7da9452596ad 100644
--- a/arch/arm64/include/asm/elf.h
+++ b/arch/arm64/include/asm/elf.h
@@ -185,10 +185,10 @@ typedef compat_elf_greg_t		compat_elf_gregset_t[COMPAT_ELF_NGREG];
 #define compat_start_thread		compat_start_thread
 #define COMPAT_SET_PERSONALITY(ex)	set_thread_flag(TIF_32BIT);
 #define COMPAT_ARCH_DLINFO
-extern int aarch32_setup_vectors_page(struct linux_binprm *bprm,
-				      int uses_interp);
+extern int aarch32_setup_additional_pages(struct linux_binprm *bprm,
+					  int uses_interp);
 #define compat_arch_setup_additional_pages \
-					aarch32_setup_vectors_page
+					aarch32_setup_additional_pages
 
 #endif /* CONFIG_COMPAT */
 
diff --git a/arch/arm64/kernel/vdso.c b/arch/arm64/kernel/vdso.c
index c239b1c15eb3..76fb5e9ca521 100644
--- a/arch/arm64/kernel/vdso.c
+++ b/arch/arm64/kernel/vdso.c
@@ -138,6 +138,25 @@ static int setup_vdso_pages(const struct vdso_mappings *mappings)
 }
 
 #ifdef CONFIG_COMPAT
+#ifdef CONFIG_VDSO32
+extern char vdso32_start, vdso32_end;
+
+static struct vdso_mappings vdso32_mappings __ro_after_init;
+
+static int __init vdso32_init(void)
+{
+	return setup_vdso_mappings("vdso32", &vdso32_start, &vdso32_end,
+				   &vdso32_mappings);
+}
+arch_initcall(vdso32_init);
+
+int aarch32_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
+{
+	return setup_vdso_pages(&vdso32_mappings);
+}
+
+#else /* CONFIG_VDSO32 */
+
 /*
  * Create and map the vectors page for AArch32 tasks.
  */
@@ -172,7 +191,7 @@ static int __init alloc_vectors_page(void)
 }
 arch_initcall(alloc_vectors_page);
 
-int aarch32_setup_vectors_page(struct linux_binprm *bprm, int uses_interp)
+int aarch32_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 {
 	struct mm_struct *mm = current->mm;
 	unsigned long addr = AARCH32_VECTORS_BASE;
@@ -196,6 +215,7 @@ int aarch32_setup_vectors_page(struct linux_binprm *bprm, int uses_interp)
 
 	return PTR_ERR_OR_ZERO(ret);
 }
+#endif /* CONFIG_VDSO32 */
 #endif /* CONFIG_COMPAT */
 
 extern char vdso_start, vdso_end;
-- 
2.10.0

^ permalink raw reply related	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 6/8] arm64: elf: Set AT_SYSINFO_EHDR in compat processes
  2016-10-27 16:30 [RFC PATCH v2 0/8] arm64: Add a compat vDSO Kevin Brodsky
                   ` (4 preceding siblings ...)
  2016-10-27 16:30 ` [RFC PATCH v2 5/8] arm64: compat: 32-bit vDSO setup Kevin Brodsky
@ 2016-10-27 16:30 ` Kevin Brodsky
  2016-10-27 16:30 ` [RFC PATCH v2 7/8] arm64: compat: Use vDSO sigreturn trampolines if available Kevin Brodsky
  2016-10-27 16:30 ` [RFC PATCH v2 8/8] arm64: Wire up and expose the new compat vDSO Kevin Brodsky
  7 siblings, 0 replies; 19+ messages in thread
From: Kevin Brodsky @ 2016-10-27 16:30 UTC (permalink / raw)
  To: linux-arm-kernel

If the compat vDSO is enabled, we need to set AT_SYSINFO_EHDR in the
auxiliary vector of compat processes to the address of the vDSO code
page, so that the dynamic linker can find it (just like the regular vDSO).

Note that we cast context.vdso to unsigned long, instead of elf_addr_t,
because elf_addr_t is 32-bit in compat_binfmt_elf.c, and casting to u32
would trigger a pointer narrowing warning.

Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
---
 arch/arm64/include/asm/elf.h | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/include/asm/elf.h b/arch/arm64/include/asm/elf.h
index 7da9452596ad..765c633950b7 100644
--- a/arch/arm64/include/asm/elf.h
+++ b/arch/arm64/include/asm/elf.h
@@ -141,11 +141,12 @@ typedef struct user_fpsimd_state elf_fpregset_t;
 #define SET_PERSONALITY(ex)		clear_thread_flag(TIF_32BIT);
 
 /* update AT_VECTOR_SIZE_ARCH if the number of NEW_AUX_ENT entries changes */
-#define ARCH_DLINFO							\
+#define _SET_AUX_ENT_VDSO						\
 do {									\
 	NEW_AUX_ENT(AT_SYSINFO_EHDR,					\
-		    (elf_addr_t)current->mm->context.vdso);		\
+		    (unsigned long)current->mm->context.vdso);		\
 } while (0)
+#define ARCH_DLINFO _SET_AUX_ENT_VDSO
 
 #define ARCH_HAS_SETUP_ADDITIONAL_PAGES
 struct linux_binprm;
@@ -184,7 +185,11 @@ typedef compat_elf_greg_t		compat_elf_gregset_t[COMPAT_ELF_NGREG];
 
 #define compat_start_thread		compat_start_thread
 #define COMPAT_SET_PERSONALITY(ex)	set_thread_flag(TIF_32BIT);
+#ifdef CONFIG_VDSO32
+#define COMPAT_ARCH_DLINFO		_SET_AUX_ENT_VDSO
+#else
 #define COMPAT_ARCH_DLINFO
+#endif
 extern int aarch32_setup_additional_pages(struct linux_binprm *bprm,
 					  int uses_interp);
 #define compat_arch_setup_additional_pages \
-- 
2.10.0

^ permalink raw reply related	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 7/8] arm64: compat: Use vDSO sigreturn trampolines if available
  2016-10-27 16:30 [RFC PATCH v2 0/8] arm64: Add a compat vDSO Kevin Brodsky
                   ` (5 preceding siblings ...)
  2016-10-27 16:30 ` [RFC PATCH v2 6/8] arm64: elf: Set AT_SYSINFO_EHDR in compat processes Kevin Brodsky
@ 2016-10-27 16:30 ` Kevin Brodsky
  2016-10-27 16:30 ` [RFC PATCH v2 8/8] arm64: Wire up and expose the new compat vDSO Kevin Brodsky
  7 siblings, 0 replies; 19+ messages in thread
From: Kevin Brodsky @ 2016-10-27 16:30 UTC (permalink / raw)
  To: linux-arm-kernel

If the compat vDSO is enabled, it replaces the vector page. Therefore,
we use the sigreturn trampolines it provides instead of those in the
vector page.

Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
---
 arch/arm64/include/asm/vdso.h |  3 +++
 arch/arm64/kernel/signal32.c  | 15 +++++++++++++++
 2 files changed, 18 insertions(+)

diff --git a/arch/arm64/include/asm/vdso.h b/arch/arm64/include/asm/vdso.h
index 839ce0031bd5..f2a952338f1e 100644
--- a/arch/arm64/include/asm/vdso.h
+++ b/arch/arm64/include/asm/vdso.h
@@ -28,6 +28,9 @@
 #ifndef __ASSEMBLY__
 
 #include <generated/vdso-offsets.h>
+#ifdef CONFIG_VDSO32
+#include <generated/vdso32-offsets.h>
+#endif
 
 #define VDSO_SYMBOL(base, name)						   \
 ({									   \
diff --git a/arch/arm64/kernel/signal32.c b/arch/arm64/kernel/signal32.c
index 9de7d128e0e0..b72a4180a531 100644
--- a/arch/arm64/kernel/signal32.c
+++ b/arch/arm64/kernel/signal32.c
@@ -28,6 +28,7 @@
 #include <asm/signal32.h>
 #include <asm/uaccess.h>
 #include <asm/unistd.h>
+#include <asm/vdso.h>
 
 struct compat_vfp_sigframe {
 	compat_ulong_t	magic;
@@ -438,6 +439,19 @@ static void compat_setup_return(struct pt_regs *regs, struct k_sigaction *ka,
 		retcode = ptr_to_compat(ka->sa.sa_restorer);
 	} else {
 		/* Set up sigreturn pointer */
+#ifdef CONFIG_VDSO32
+		void *vdso_page = current->mm->context.vdso;
+		void *trampoline =
+			(ka->sa.sa_flags & SA_SIGINFO
+			 ? (thumb
+			    ? VDSO_SYMBOL(vdso_page, compat_rt_sigreturn_thumb)
+			    : VDSO_SYMBOL(vdso_page, compat_rt_sigreturn_arm))
+			 : (thumb
+			    ? VDSO_SYMBOL(vdso_page, compat_sigreturn_thumb)
+			    : VDSO_SYMBOL(vdso_page, compat_sigreturn_arm)));
+
+		retcode = ptr_to_compat(trampoline) + thumb;
+#else
 		unsigned int idx = thumb << 1;
 
 		if (ka->sa.sa_flags & SA_SIGINFO)
@@ -446,6 +460,7 @@ static void compat_setup_return(struct pt_regs *regs, struct k_sigaction *ka,
 		retcode = AARCH32_VECTORS_BASE +
 			  AARCH32_KERN_SIGRET_CODE_OFFSET +
 			  (idx << 2) + thumb;
+#endif
 	}
 
 	regs->regs[0]	= usig;
-- 
2.10.0

^ permalink raw reply related	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 8/8] arm64: Wire up and expose the new compat vDSO
  2016-10-27 16:30 [RFC PATCH v2 0/8] arm64: Add a compat vDSO Kevin Brodsky
                   ` (6 preceding siblings ...)
  2016-10-27 16:30 ` [RFC PATCH v2 7/8] arm64: compat: Use vDSO sigreturn trampolines if available Kevin Brodsky
@ 2016-10-27 16:30 ` Kevin Brodsky
  2016-11-04 15:50   ` Catalin Marinas
  7 siblings, 1 reply; 19+ messages in thread
From: Kevin Brodsky @ 2016-10-27 16:30 UTC (permalink / raw)
  To: linux-arm-kernel

Expose the new compat vDSO via the COMPAT_VDSO config option.

The option is not enabled in defconfig for two reasons:

* The vDSO page replaces the vector page. The vDSO provides its own
  sigreturn trampolines, replacing those in the vector page, but the
  kuser helpers are gone. As a result enabling the compat vDSO will
  break userspace programs relying on the kuser helpers.

* We really need a 32-bit compiler this time, and we rely on the user
  to provide it themselves by setting CROSS_COMPILE_ARM32. Therefore
  enabling the option by default would make little sense, since the
  user must explicitly set an environment variable anyway.

CONFIG_COMPAT_VDSO is not directly used in the code, because we want
to ignore it (build as if it were not set) if the user didn't set
CROSS_COMPILE_ARM32 properly. If the variable has been set to a valid
prefix, CONFIG_VDSO32 will be set; this is the option that the code
and Makefiles test.

For more flexibility, like CROSS_COMPILE, CROSS_COMPILE_ARM32 can also
be set via CONFIG_CROSS_COMPILE_ARM32 (the environment variable
overrides the config option).

Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
---
 arch/arm64/Kconfig         | 26 ++++++++++++++++++++++++++
 arch/arm64/Makefile        | 28 ++++++++++++++++++++++++++--
 arch/arm64/kernel/Makefile |  8 ++++++--
 3 files changed, 58 insertions(+), 4 deletions(-)

diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
index 969ef880d234..883e50def0eb 100644
--- a/arch/arm64/Kconfig
+++ b/arch/arm64/Kconfig
@@ -1017,6 +1017,32 @@ config SYSVIPC_COMPAT
 	def_bool y
 	depends on COMPAT && SYSVIPC
 
+config COMPAT_VDSO
+	bool "32-bit vDSO"
+	depends on COMPAT
+	default n
+	help
+	  Warning: this completely removes the compat vector page, including
+	  kuser helpers, which may break 32-bit processes.
+
+	  Warning: a 32-bit toolchain is necessary to build the vDSO. You
+	  must explicitly define which toolchain should be used by setting
+	  CROSS_COMPILE_ARM32 to the prefix of the 32-bit toolchain (same format
+	  as CROSS_COMPILE). If a 32-bit compiler cannot be found, a warning
+	  will be printed and the kernel will be built as if COMPAT_VDSO had not
+	  been set.
+
+	  Provide a vDSO to 32-bit processes. It includes the symbols provided
+	  by the vDSO from the 32-bit kernel, so that a 32-bit libc can use
+	  the compat vDSO without modification. It also provides sigreturn
+	  trampolines, and replaces the vector page.
+
+config CROSS_COMPILE_ARM32
+	string "32-bit toolchain prefix"
+	help
+	  Same as setting CROSS_COMPILE_ARM32 in the environment, but saved for
+	  future builds. The environment variable overrides this config option.
+
 endmenu
 
 menu "Power management options"
diff --git a/arch/arm64/Makefile b/arch/arm64/Makefile
index 3635b8662724..370d8de0c100 100644
--- a/arch/arm64/Makefile
+++ b/arch/arm64/Makefile
@@ -37,10 +37,32 @@ $(warning LSE atomics not supported by binutils)
   endif
 endif
 
-KBUILD_CFLAGS	+= -mgeneral-regs-only $(lseinstr)
+ifeq ($(CONFIG_COMPAT_VDSO), y)
+  CROSS_COMPILE_ARM32 ?= $(CONFIG_CROSS_COMPILE_ARM32:"%"=%)
+
+  # Check that the user has provided a valid prefix for the 32-bit toolchain.
+  # To prevent selecting the system gcc by default, the prefix is not allowed to
+  # be empty, unlike CROSS_COMPILE. In the unlikely event that the system gcc
+  # is actually the 32-bit ARM compiler to be used, the variable can be set to
+  # the dirname (e.g. CROSS_COMPILE_ARM32=/usr/bin/).
+  # Note: this Makefile is read both before and after regenerating the
+  # config (if needed). Any warning appearing before the config has been
+  # regenerated should be ignored.
+  ifeq ($(CROSS_COMPILE_ARM32),)
+    $(warning CROSS_COMPILE_ARM32 not defined or empty, the compat vDSO will not be built)
+  else ifeq ($(shell which $(CROSS_COMPILE_ARM32)gcc 2> /dev/null),)
+    $(warning $(CROSS_COMPILE_ARM32)gcc not found, the compat vDSO will not be built)
+  else
+    export CROSS_COMPILE_ARM32
+    export CONFIG_VDSO32 := y
+    vdso32 := -DCONFIG_VDSO32=1
+  endif
+endif
+
+KBUILD_CFLAGS	+= -mgeneral-regs-only $(lseinstr) $(vdso32)
 KBUILD_CFLAGS	+= -fno-asynchronous-unwind-tables
 KBUILD_CFLAGS	+= $(call cc-option, -mpc-relative-literal-loads)
-KBUILD_AFLAGS	+= $(lseinstr)
+KBUILD_AFLAGS	+= $(lseinstr) $(vdso32)
 
 ifeq ($(CONFIG_CPU_BIG_ENDIAN), y)
 KBUILD_CPPFLAGS	+= -mbig-endian
@@ -139,6 +161,8 @@ archclean:
 prepare: vdso_prepare
 vdso_prepare: prepare0
 	$(Q)$(MAKE) $(build)=arch/arm64/kernel/vdso include/generated/vdso-offsets.h
+	$(if $(CONFIG_VDSO32),$(Q)$(MAKE) $(build)=arch/arm64/kernel/vdso32 \
+					  include/generated/vdso32-offsets.h)
 
 define archhelp
   echo  '* Image.gz      - Compressed kernel image (arch/$(ARCH)/boot/Image.gz)'
diff --git a/arch/arm64/kernel/Makefile b/arch/arm64/kernel/Makefile
index 7d66bbaafc0c..1487f8cd06dd 100644
--- a/arch/arm64/kernel/Makefile
+++ b/arch/arm64/kernel/Makefile
@@ -27,8 +27,11 @@ OBJCOPYFLAGS := --prefix-symbols=__efistub_
 $(obj)/%.stub.o: $(obj)/%.o FORCE
 	$(call if_changed,objcopy)
 
-arm64-obj-$(CONFIG_COMPAT)		+= sys32.o kuser32.o signal32.o 	\
-					   sys_compat.o entry32.o
+arm64-obj-$(CONFIG_COMPAT)		+= sys32.o signal32.o sys_compat.o	\
+					   entry32.o
+ifneq ($(CONFIG_VDSO32),y)
+arm64-obj-y				+= kuser32.o
+endif
 arm64-obj-$(CONFIG_FUNCTION_TRACER)	+= ftrace.o entry-ftrace.o
 arm64-obj-$(CONFIG_MODULES)		+= arm64ksyms.o module.o
 arm64-obj-$(CONFIG_ARM64_MODULE_PLTS)	+= module-plts.o
@@ -52,6 +55,7 @@ arm64-obj-$(CONFIG_KEXEC)		+= machine_kexec.o relocate_kernel.o	\
 					   cpu-reset.o
 
 obj-y					+= $(arm64-obj-y) vdso/ probes/
+obj-$(CONFIG_VDSO32)			+= vdso32/
 obj-m					+= $(arm64-obj-m)
 head-y					:= head.o
 extra-y					+= $(head-y) vmlinux.lds
-- 
2.10.0

^ permalink raw reply related	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 4/8] arm64: compat: Add a 32-bit vDSO
  2016-10-27 16:30 ` [RFC PATCH v2 4/8] arm64: compat: Add a 32-bit vDSO Kevin Brodsky
@ 2016-10-28  3:09   ` Jisheng Zhang
  2016-10-28 10:20     ` Kevin Brodsky
  0 siblings, 1 reply; 19+ messages in thread
From: Jisheng Zhang @ 2016-10-28  3:09 UTC (permalink / raw)
  To: linux-arm-kernel

Dear Kevin,

On Thu, 27 Oct 2016 17:30:54 +0100 Kevin Brodsky wrote:

> Provide the files necessary for building a compat (AArch32) vDSO in
> kernel/vdso32.
> 
> This is mostly an adaptation of the arm vDSO. The most significant
> change in vgettimeofday.c is the use of the arm64 vdso_data struct,
> allowing the vDSO data page to be shared between the 32 and 64-bit
> vDSOs.
> 
> In addition to the time functions, sigreturn trampolines are also
> provided, aiming at replacing those in the vector page. To improve
> debugging, CFI and unwinding directives are used, based on glibc's
> implementation. Symbol offsets are made available to the kernel using
> the same method as the 64-bit vDSO.
> 
> There is unfortunately an important caveat to all this: we cannot get
> away with hand-coding 32-bit instructions like in kernel/kuser32.S,
> this time we really need a 32-bit compiler. The compat vDSO Makefile
> relies on CROSS_COMPILE_ARM32 to provide a 32-bit compiler,
> appropriate logic will be added to the arm64 Makefile later on to
> ensure that an attempt to build the compat vDSO is made only if this
> variable has been set properly.
> 
> Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
> ---
>  arch/arm64/kernel/vdso32/Makefile        | 121 +++++++++++++
>  arch/arm64/kernel/vdso32/sigreturn.S     |  86 +++++++++
>  arch/arm64/kernel/vdso32/vdso.S          |  32 ++++
>  arch/arm64/kernel/vdso32/vdso.lds.S      |  98 +++++++++++
>  arch/arm64/kernel/vdso32/vgettimeofday.c | 294 +++++++++++++++++++++++++++++++
>  5 files changed, 631 insertions(+)
>  create mode 100644 arch/arm64/kernel/vdso32/Makefile
>  create mode 100644 arch/arm64/kernel/vdso32/sigreturn.S
>  create mode 100644 arch/arm64/kernel/vdso32/vdso.S
>  create mode 100644 arch/arm64/kernel/vdso32/vdso.lds.S
>  create mode 100644 arch/arm64/kernel/vdso32/vgettimeofday.c
> 
> diff --git a/arch/arm64/kernel/vdso32/Makefile b/arch/arm64/kernel/vdso32/Makefile
> new file mode 100644
> index 000000000000..38facc870f6e
> --- /dev/null
> +++ b/arch/arm64/kernel/vdso32/Makefile
> @@ -0,0 +1,121 @@
> +#
> +# Building a vDSO image for AArch32.
> +#
> +# Author: Kevin Brodsky <kevin.brodsky@arm.com>
> +# A mix between the arm64 and arm vDSO Makefiles.
> +
> +CC_ARM32 := $(CROSS_COMPILE_ARM32)gcc
> +
> +# Same as cc-ldoption, but using CC_ARM32 instead of CC
> +cc32-ldoption = $(call try-run,\
> +        $(CC_ARM32) $(1) -nostdlib -x c /dev/null -o "$$TMP",$(1),$(2))
> +
> +# Borrow vdsomunge.c from the arm vDSO
> +munge := arch/arm/vdso/vdsomunge
> +hostprogs-y := $(srctree)/$(munge)
> +
> +c-obj-vdso := vgettimeofday.o
> +asm-obj-vdso := sigreturn.o
> +
> +# Build rules
> +targets := $(c-obj-vdso) $(asm-obj-vdso) vdso.so vdso.so.dbg vdso.so.raw
> +c-obj-vdso := $(addprefix $(obj)/, $(c-obj-vdso))
> +asm-obj-vdso := $(addprefix $(obj)/, $(asm-obj-vdso))
> +obj-vdso := $(c-obj-vdso) $(asm-obj-vdso)
> +
> +ccflags-y := -fPIC -fno-common -fno-builtin -fno-stack-protector
> +ccflags-y += -DDISABLE_BRANCH_PROFILING
> +
> +# Force -O2 to avoid libgcc dependencies
> +VDSO_CFLAGS := -march=armv8-a -O2

For completeness, bringing 32bit compiler need to check whether the 32bit
toolchain support some options. IIRC, armv8-a support isn't enabled until
gcc 4.8, so old toolchains such gcc-4.7 will complain:
 error: unrecognized argument in option ?-march=armv8-a?

Thanks,
Jisheng

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 4/8] arm64: compat: Add a 32-bit vDSO
  2016-10-28  3:09   ` Jisheng Zhang
@ 2016-10-28 10:20     ` Kevin Brodsky
  2016-11-04 20:03       ` Catalin Marinas
  0 siblings, 1 reply; 19+ messages in thread
From: Kevin Brodsky @ 2016-10-28 10:20 UTC (permalink / raw)
  To: linux-arm-kernel

On 28/10/16 04:09, Jisheng Zhang wrote:
> Dear Kevin,
>
> On Thu, 27 Oct 2016 17:30:54 +0100 Kevin Brodsky wrote:
>
>> Provide the files necessary for building a compat (AArch32) vDSO in
>> kernel/vdso32.
>>
>> This is mostly an adaptation of the arm vDSO. The most significant
>> change in vgettimeofday.c is the use of the arm64 vdso_data struct,
>> allowing the vDSO data page to be shared between the 32 and 64-bit
>> vDSOs.
>>
>> In addition to the time functions, sigreturn trampolines are also
>> provided, aiming at replacing those in the vector page. To improve
>> debugging, CFI and unwinding directives are used, based on glibc's
>> implementation. Symbol offsets are made available to the kernel using
>> the same method as the 64-bit vDSO.
>>
>> There is unfortunately an important caveat to all this: we cannot get
>> away with hand-coding 32-bit instructions like in kernel/kuser32.S,
>> this time we really need a 32-bit compiler. The compat vDSO Makefile
>> relies on CROSS_COMPILE_ARM32 to provide a 32-bit compiler,
>> appropriate logic will be added to the arm64 Makefile later on to
>> ensure that an attempt to build the compat vDSO is made only if this
>> variable has been set properly.
>>
>> Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
>> ---
>>   arch/arm64/kernel/vdso32/Makefile        | 121 +++++++++++++
>>   arch/arm64/kernel/vdso32/sigreturn.S     |  86 +++++++++
>>   arch/arm64/kernel/vdso32/vdso.S          |  32 ++++
>>   arch/arm64/kernel/vdso32/vdso.lds.S      |  98 +++++++++++
>>   arch/arm64/kernel/vdso32/vgettimeofday.c | 294 +++++++++++++++++++++++++++++++
>>   5 files changed, 631 insertions(+)
>>   create mode 100644 arch/arm64/kernel/vdso32/Makefile
>>   create mode 100644 arch/arm64/kernel/vdso32/sigreturn.S
>>   create mode 100644 arch/arm64/kernel/vdso32/vdso.S
>>   create mode 100644 arch/arm64/kernel/vdso32/vdso.lds.S
>>   create mode 100644 arch/arm64/kernel/vdso32/vgettimeofday.c
>>
>> diff --git a/arch/arm64/kernel/vdso32/Makefile b/arch/arm64/kernel/vdso32/Makefile
>> new file mode 100644
>> index 000000000000..38facc870f6e
>> --- /dev/null
>> +++ b/arch/arm64/kernel/vdso32/Makefile
>> @@ -0,0 +1,121 @@
>> +#
>> +# Building a vDSO image for AArch32.
>> +#
>> +# Author: Kevin Brodsky <kevin.brodsky@arm.com>
>> +# A mix between the arm64 and arm vDSO Makefiles.
>> +
>> +CC_ARM32 := $(CROSS_COMPILE_ARM32)gcc
>> +
>> +# Same as cc-ldoption, but using CC_ARM32 instead of CC
>> +cc32-ldoption = $(call try-run,\
>> +        $(CC_ARM32) $(1) -nostdlib -x c /dev/null -o "$$TMP",$(1),$(2))
>> +
>> +# Borrow vdsomunge.c from the arm vDSO
>> +munge := arch/arm/vdso/vdsomunge
>> +hostprogs-y := $(srctree)/$(munge)
>> +
>> +c-obj-vdso := vgettimeofday.o
>> +asm-obj-vdso := sigreturn.o
>> +
>> +# Build rules
>> +targets := $(c-obj-vdso) $(asm-obj-vdso) vdso.so vdso.so.dbg vdso.so.raw
>> +c-obj-vdso := $(addprefix $(obj)/, $(c-obj-vdso))
>> +asm-obj-vdso := $(addprefix $(obj)/, $(asm-obj-vdso))
>> +obj-vdso := $(c-obj-vdso) $(asm-obj-vdso)
>> +
>> +ccflags-y := -fPIC -fno-common -fno-builtin -fno-stack-protector
>> +ccflags-y += -DDISABLE_BRANCH_PROFILING
>> +
>> +# Force -O2 to avoid libgcc dependencies
>> +VDSO_CFLAGS := -march=armv8-a -O2
> For completeness, bringing 32bit compiler need to check whether the 32bit
> toolchain support some options. IIRC, armv8-a support isn't enabled until
> gcc 4.8, so old toolchains such gcc-4.7 will complain:
>   error: unrecognized argument in option ?-march=armv8-a?
>
> Thanks,
> Jisheng

That's a fair point. I guess -march=armv8-a is not strictly necessary and the 
produced vDSO should be fine if arch/arm/vdso also compiles fine. However we would 
still need to pass -march=armv7-a. I'm not sure what to do between:
* Checking that the compiler supports -march=armv8-a when inspecting 
CROSS_COMPILE_ARM32, and if it doesn't vdso32 will not be built.
* Checking whether -march=armv8-a is available here, and if it is not fall back to 
-march=armv7-a.

Thanks!
Kevin

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 8/8] arm64: Wire up and expose the new compat vDSO
  2016-10-27 16:30 ` [RFC PATCH v2 8/8] arm64: Wire up and expose the new compat vDSO Kevin Brodsky
@ 2016-11-04 15:50   ` Catalin Marinas
  2016-11-04 16:30     ` Kevin Brodsky
  0 siblings, 1 reply; 19+ messages in thread
From: Catalin Marinas @ 2016-11-04 15:50 UTC (permalink / raw)
  To: linux-arm-kernel

On Thu, Oct 27, 2016 at 05:30:58PM +0100, Kevin Brodsky wrote:
> * The vDSO page replaces the vector page. The vDSO provides its own
>   sigreturn trampolines, replacing those in the vector page, but the
>   kuser helpers are gone. As a result enabling the compat vDSO will
>   break userspace programs relying on the kuser helpers.

I think vDSO and vectors page should not exclude each other. If you want
to disable the vectors page, let's make it an independent config option
like the KUSER_HELPERS in arch/arm64/mm/Kconfig. But I would very much
like to be able to have both the vDSO and the vectors page at the same
time.

-- 
Catalin

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 8/8] arm64: Wire up and expose the new compat vDSO
  2016-11-04 15:50   ` Catalin Marinas
@ 2016-11-04 16:30     ` Kevin Brodsky
  2016-11-04 16:47       ` Catalin Marinas
  0 siblings, 1 reply; 19+ messages in thread
From: Kevin Brodsky @ 2016-11-04 16:30 UTC (permalink / raw)
  To: linux-arm-kernel

On 04/11/2016 09:50, Catalin Marinas wrote:
> On Thu, Oct 27, 2016 at 05:30:58PM +0100, Kevin Brodsky wrote:
>> * The vDSO page replaces the vector page. The vDSO provides its own
>>    sigreturn trampolines, replacing those in the vector page, but the
>>    kuser helpers are gone. As a result enabling the compat vDSO will
>>    break userspace programs relying on the kuser helpers.
> I think vDSO and vectors page should not exclude each other. If you want
> to disable the vectors page, let's make it an independent config option
> like the KUSER_HELPERS in arch/arm64/mm/Kconfig. But I would very much
> like to be able to have both the vDSO and the vectors page at the same
> time.

Indeed, I've had exactly the same feedback from Google yesterday (apparently many 
Android apps with native libs still target ARMv6....). I'll add the option to keep 
the kuser helpers.

There's a small problem though: how to ensure that the kuser helpers + sigreturn 
trampolines are always included if the compat vDSO is not built? I can enforce 
CONFIG_KUSER_HELPERS if !CONFIG_VDSO32 (directly in the code/Makefiles), but the 
dependency cannot be expressed in Kconfig.

Thanks,
Kevin

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 8/8] arm64: Wire up and expose the new compat vDSO
  2016-11-04 16:30     ` Kevin Brodsky
@ 2016-11-04 16:47       ` Catalin Marinas
  2016-11-04 17:53         ` Kevin Brodsky
  0 siblings, 1 reply; 19+ messages in thread
From: Catalin Marinas @ 2016-11-04 16:47 UTC (permalink / raw)
  To: linux-arm-kernel

On Fri, Nov 04, 2016 at 10:30:08AM -0600, Kevin Brodsky wrote:
> On 04/11/2016 09:50, Catalin Marinas wrote:
> > On Thu, Oct 27, 2016 at 05:30:58PM +0100, Kevin Brodsky wrote:
> > > * The vDSO page replaces the vector page. The vDSO provides its own
> > >    sigreturn trampolines, replacing those in the vector page, but the
> > >    kuser helpers are gone. As a result enabling the compat vDSO will
> > >    break userspace programs relying on the kuser helpers.
> > I think vDSO and vectors page should not exclude each other. If you want
> > to disable the vectors page, let's make it an independent config option
> > like the KUSER_HELPERS in arch/arm64/mm/Kconfig. But I would very much
> > like to be able to have both the vDSO and the vectors page at the same
> > time.
> 
> Indeed, I've had exactly the same feedback from Google yesterday (apparently
> many Android apps with native libs still target ARMv6....). I'll add the
> option to keep the kuser helpers.
> 
> There's a small problem though: how to ensure that the kuser helpers +
> sigreturn trampolines are always included if the compat vDSO is not built? I
> can enforce CONFIG_KUSER_HELPERS if !CONFIG_VDSO32 (directly in the
> code/Makefiles), but the dependency cannot be expressed in Kconfig.

Or you could insert a separate "sigpage" as arm32 does. This could leave
independently of vDSO or vectors page.

-- 
Catalin

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 8/8] arm64: Wire up and expose the new compat vDSO
  2016-11-04 16:47       ` Catalin Marinas
@ 2016-11-04 17:53         ` Kevin Brodsky
  0 siblings, 0 replies; 19+ messages in thread
From: Kevin Brodsky @ 2016-11-04 17:53 UTC (permalink / raw)
  To: linux-arm-kernel

On 04/11/2016 10:47, Catalin Marinas wrote:
> On Fri, Nov 04, 2016 at 10:30:08AM -0600, Kevin Brodsky wrote:
>> On 04/11/2016 09:50, Catalin Marinas wrote:
>>> On Thu, Oct 27, 2016 at 05:30:58PM +0100, Kevin Brodsky wrote:
>>>> * The vDSO page replaces the vector page. The vDSO provides its own
>>>>     sigreturn trampolines, replacing those in the vector page, but the
>>>>     kuser helpers are gone. As a result enabling the compat vDSO will
>>>>     break userspace programs relying on the kuser helpers.
>>> I think vDSO and vectors page should not exclude each other. If you want
>>> to disable the vectors page, let's make it an independent config option
>>> like the KUSER_HELPERS in arch/arm64/mm/Kconfig. But I would very much
>>> like to be able to have both the vDSO and the vectors page at the same
>>> time.
>> Indeed, I've had exactly the same feedback from Google yesterday (apparently
>> many Android apps with native libs still target ARMv6....). I'll add the
>> option to keep the kuser helpers.
>>
>> There's a small problem though: how to ensure that the kuser helpers +
>> sigreturn trampolines are always included if the compat vDSO is not built? I
>> can enforce CONFIG_KUSER_HELPERS if !CONFIG_VDSO32 (directly in the
>> code/Makefiles), but the dependency cannot be expressed in Kconfig.
> Or you could insert a separate "sigpage" as arm32 does. This could leave
> independently of vDSO or vectors page.

Yeah I thought about this too. It's a bit more work but probably cleaner and more 
flexible, that would also allow to disable the kuser helpers independently of the 
compat vDSO.

Thanks,
Kevin

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 4/8] arm64: compat: Add a 32-bit vDSO
  2016-10-28 10:20     ` Kevin Brodsky
@ 2016-11-04 20:03       ` Catalin Marinas
  2016-11-21 15:45         ` Kevin Brodsky
  0 siblings, 1 reply; 19+ messages in thread
From: Catalin Marinas @ 2016-11-04 20:03 UTC (permalink / raw)
  To: linux-arm-kernel

On Fri, Oct 28, 2016 at 11:20:07AM +0100, Kevin Brodsky wrote:
> On 28/10/16 04:09, Jisheng Zhang wrote:
> > On Thu, 27 Oct 2016 17:30:54 +0100 Kevin Brodsky wrote:
> > > Provide the files necessary for building a compat (AArch32) vDSO in
> > > kernel/vdso32.
> > > 
> > > This is mostly an adaptation of the arm vDSO. The most significant
> > > change in vgettimeofday.c is the use of the arm64 vdso_data struct,
> > > allowing the vDSO data page to be shared between the 32 and 64-bit
> > > vDSOs.
> > > 
> > > In addition to the time functions, sigreturn trampolines are also
> > > provided, aiming at replacing those in the vector page. To improve
> > > debugging, CFI and unwinding directives are used, based on glibc's
> > > implementation. Symbol offsets are made available to the kernel using
> > > the same method as the 64-bit vDSO.
> > > 
> > > There is unfortunately an important caveat to all this: we cannot get
> > > away with hand-coding 32-bit instructions like in kernel/kuser32.S,
> > > this time we really need a 32-bit compiler. The compat vDSO Makefile
> > > relies on CROSS_COMPILE_ARM32 to provide a 32-bit compiler,
> > > appropriate logic will be added to the arm64 Makefile later on to
> > > ensure that an attempt to build the compat vDSO is made only if this
> > > variable has been set properly.
> > > 
> > > Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
> > > ---
> > >   arch/arm64/kernel/vdso32/Makefile        | 121 +++++++++++++
> > >   arch/arm64/kernel/vdso32/sigreturn.S     |  86 +++++++++
> > >   arch/arm64/kernel/vdso32/vdso.S          |  32 ++++
> > >   arch/arm64/kernel/vdso32/vdso.lds.S      |  98 +++++++++++
> > >   arch/arm64/kernel/vdso32/vgettimeofday.c | 294 +++++++++++++++++++++++++++++++
> > >   5 files changed, 631 insertions(+)
> > >   create mode 100644 arch/arm64/kernel/vdso32/Makefile
> > >   create mode 100644 arch/arm64/kernel/vdso32/sigreturn.S
> > >   create mode 100644 arch/arm64/kernel/vdso32/vdso.S
> > >   create mode 100644 arch/arm64/kernel/vdso32/vdso.lds.S
> > >   create mode 100644 arch/arm64/kernel/vdso32/vgettimeofday.c
> > > 
> > > diff --git a/arch/arm64/kernel/vdso32/Makefile b/arch/arm64/kernel/vdso32/Makefile
> > > new file mode 100644
> > > index 000000000000..38facc870f6e
> > > --- /dev/null
> > > +++ b/arch/arm64/kernel/vdso32/Makefile
> > > @@ -0,0 +1,121 @@
> > > +#
> > > +# Building a vDSO image for AArch32.
> > > +#
> > > +# Author: Kevin Brodsky <kevin.brodsky@arm.com>
> > > +# A mix between the arm64 and arm vDSO Makefiles.
> > > +
> > > +CC_ARM32 := $(CROSS_COMPILE_ARM32)gcc
> > > +
> > > +# Same as cc-ldoption, but using CC_ARM32 instead of CC
> > > +cc32-ldoption = $(call try-run,\
> > > +        $(CC_ARM32) $(1) -nostdlib -x c /dev/null -o "$$TMP",$(1),$(2))
> > > +
> > > +# Borrow vdsomunge.c from the arm vDSO
> > > +munge := arch/arm/vdso/vdsomunge
> > > +hostprogs-y := $(srctree)/$(munge)
> > > +
> > > +c-obj-vdso := vgettimeofday.o
> > > +asm-obj-vdso := sigreturn.o
> > > +
> > > +# Build rules
> > > +targets := $(c-obj-vdso) $(asm-obj-vdso) vdso.so vdso.so.dbg vdso.so.raw
> > > +c-obj-vdso := $(addprefix $(obj)/, $(c-obj-vdso))
> > > +asm-obj-vdso := $(addprefix $(obj)/, $(asm-obj-vdso))
> > > +obj-vdso := $(c-obj-vdso) $(asm-obj-vdso)
> > > +
> > > +ccflags-y := -fPIC -fno-common -fno-builtin -fno-stack-protector
> > > +ccflags-y += -DDISABLE_BRANCH_PROFILING
> > > +
> > > +# Force -O2 to avoid libgcc dependencies
> > > +VDSO_CFLAGS := -march=armv8-a -O2
> > For completeness, bringing 32bit compiler need to check whether the 32bit
> > toolchain support some options. IIRC, armv8-a support isn't enabled until
> > gcc 4.8, so old toolchains such gcc-4.7 will complain:
> >   error: unrecognized argument in option ?-march=armv8-a?
> 
> That's a fair point. I guess -march=armv8-a is not strictly necessary and
> the produced vDSO should be fine if arch/arm/vdso also compiles fine.
> However we would still need to pass -march=armv7-a. I'm not sure what to do
> between:
> * Checking that the compiler supports -march=armv8-a when inspecting
> CROSS_COMPILE_ARM32, and if it doesn't vdso32 will not be built.
> * Checking whether -march=armv8-a is available here, and if it is not fall
> back to -march=armv7-a.

Does v8 vs v7 make any difference in the generated code? If not, we
could just stick to armv7-a permanently.

-- 
Catalin

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 4/8] arm64: compat: Add a 32-bit vDSO
  2016-11-04 20:03       ` Catalin Marinas
@ 2016-11-21 15:45         ` Kevin Brodsky
  2016-11-21 18:44           ` Catalin Marinas
  0 siblings, 1 reply; 19+ messages in thread
From: Kevin Brodsky @ 2016-11-21 15:45 UTC (permalink / raw)
  To: linux-arm-kernel

On 04/11/16 20:03, Catalin Marinas wrote:
> On Fri, Oct 28, 2016 at 11:20:07AM +0100, Kevin Brodsky wrote:
>> On 28/10/16 04:09, Jisheng Zhang wrote:
>>> On Thu, 27 Oct 2016 17:30:54 +0100 Kevin Brodsky wrote:
>>>> Provide the files necessary for building a compat (AArch32) vDSO in
>>>> kernel/vdso32.
>>>>
>>>> This is mostly an adaptation of the arm vDSO. The most significant
>>>> change in vgettimeofday.c is the use of the arm64 vdso_data struct,
>>>> allowing the vDSO data page to be shared between the 32 and 64-bit
>>>> vDSOs.
>>>>
>>>> In addition to the time functions, sigreturn trampolines are also
>>>> provided, aiming at replacing those in the vector page. To improve
>>>> debugging, CFI and unwinding directives are used, based on glibc's
>>>> implementation. Symbol offsets are made available to the kernel using
>>>> the same method as the 64-bit vDSO.
>>>>
>>>> There is unfortunately an important caveat to all this: we cannot get
>>>> away with hand-coding 32-bit instructions like in kernel/kuser32.S,
>>>> this time we really need a 32-bit compiler. The compat vDSO Makefile
>>>> relies on CROSS_COMPILE_ARM32 to provide a 32-bit compiler,
>>>> appropriate logic will be added to the arm64 Makefile later on to
>>>> ensure that an attempt to build the compat vDSO is made only if this
>>>> variable has been set properly.
>>>>
>>>> Signed-off-by: Kevin Brodsky <kevin.brodsky@arm.com>
>>>> ---
>>>>    arch/arm64/kernel/vdso32/Makefile        | 121 +++++++++++++
>>>>    arch/arm64/kernel/vdso32/sigreturn.S     |  86 +++++++++
>>>>    arch/arm64/kernel/vdso32/vdso.S          |  32 ++++
>>>>    arch/arm64/kernel/vdso32/vdso.lds.S      |  98 +++++++++++
>>>>    arch/arm64/kernel/vdso32/vgettimeofday.c | 294 +++++++++++++++++++++++++++++++
>>>>    5 files changed, 631 insertions(+)
>>>>    create mode 100644 arch/arm64/kernel/vdso32/Makefile
>>>>    create mode 100644 arch/arm64/kernel/vdso32/sigreturn.S
>>>>    create mode 100644 arch/arm64/kernel/vdso32/vdso.S
>>>>    create mode 100644 arch/arm64/kernel/vdso32/vdso.lds.S
>>>>    create mode 100644 arch/arm64/kernel/vdso32/vgettimeofday.c
>>>>
>>>> diff --git a/arch/arm64/kernel/vdso32/Makefile b/arch/arm64/kernel/vdso32/Makefile
>>>> new file mode 100644
>>>> index 000000000000..38facc870f6e
>>>> --- /dev/null
>>>> +++ b/arch/arm64/kernel/vdso32/Makefile
>>>> @@ -0,0 +1,121 @@
>>>> +#
>>>> +# Building a vDSO image for AArch32.
>>>> +#
>>>> +# Author: Kevin Brodsky <kevin.brodsky@arm.com>
>>>> +# A mix between the arm64 and arm vDSO Makefiles.
>>>> +
>>>> +CC_ARM32 := $(CROSS_COMPILE_ARM32)gcc
>>>> +
>>>> +# Same as cc-ldoption, but using CC_ARM32 instead of CC
>>>> +cc32-ldoption = $(call try-run,\
>>>> +        $(CC_ARM32) $(1) -nostdlib -x c /dev/null -o "$$TMP",$(1),$(2))
>>>> +
>>>> +# Borrow vdsomunge.c from the arm vDSO
>>>> +munge := arch/arm/vdso/vdsomunge
>>>> +hostprogs-y := $(srctree)/$(munge)
>>>> +
>>>> +c-obj-vdso := vgettimeofday.o
>>>> +asm-obj-vdso := sigreturn.o
>>>> +
>>>> +# Build rules
>>>> +targets := $(c-obj-vdso) $(asm-obj-vdso) vdso.so vdso.so.dbg vdso.so.raw
>>>> +c-obj-vdso := $(addprefix $(obj)/, $(c-obj-vdso))
>>>> +asm-obj-vdso := $(addprefix $(obj)/, $(asm-obj-vdso))
>>>> +obj-vdso := $(c-obj-vdso) $(asm-obj-vdso)
>>>> +
>>>> +ccflags-y := -fPIC -fno-common -fno-builtin -fno-stack-protector
>>>> +ccflags-y += -DDISABLE_BRANCH_PROFILING
>>>> +
>>>> +# Force -O2 to avoid libgcc dependencies
>>>> +VDSO_CFLAGS := -march=armv8-a -O2
>>> For completeness, bringing 32bit compiler need to check whether the 32bit
>>> toolchain support some options. IIRC, armv8-a support isn't enabled until
>>> gcc 4.8, so old toolchains such gcc-4.7 will complain:
>>>    error: unrecognized argument in option ?-march=armv8-a?
>> That's a fair point. I guess -march=armv8-a is not strictly necessary and
>> the produced vDSO should be fine if arch/arm/vdso also compiles fine.
>> However we would still need to pass -march=armv7-a. I'm not sure what to do
>> between:
>> * Checking that the compiler supports -march=armv8-a when inspecting
>> CROSS_COMPILE_ARM32, and if it doesn't vdso32 will not be built.
>> * Checking whether -march=armv8-a is available here, and if it is not fall
>> back to -march=armv7-a.
> Does v8 vs v7 make any difference in the generated code? If not, we
> could just stick to armv7-a permanently.

I've just tried compiling with -march=armv7-a, and in fact it doesn't compile at all. 
It turns out vgettimeofday.c uses smp_rmb(), which expands to dmb ishld on arm64, and 
ishld doesn't exist in ARMv7. We could possibly work around that, but I think 
requiring GCC 4.8 is reasonable.

Thanks,
Kevin

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 4/8] arm64: compat: Add a 32-bit vDSO
  2016-11-21 15:45         ` Kevin Brodsky
@ 2016-11-21 18:44           ` Catalin Marinas
  2016-12-01 14:27             ` Kevin Brodsky
  0 siblings, 1 reply; 19+ messages in thread
From: Catalin Marinas @ 2016-11-21 18:44 UTC (permalink / raw)
  To: linux-arm-kernel

On Mon, Nov 21, 2016 at 03:45:55PM +0000, Kevin Brodsky wrote:
> On 04/11/16 20:03, Catalin Marinas wrote:
> >On Fri, Oct 28, 2016 at 11:20:07AM +0100, Kevin Brodsky wrote:
> >>On 28/10/16 04:09, Jisheng Zhang wrote:
> >>>On Thu, 27 Oct 2016 17:30:54 +0100 Kevin Brodsky wrote:
> >>>>+# Force -O2 to avoid libgcc dependencies
> >>>>+VDSO_CFLAGS := -march=armv8-a -O2
> >>>
> >>>For completeness, bringing 32bit compiler need to check whether the 32bit
> >>>toolchain support some options. IIRC, armv8-a support isn't enabled until
> >>>gcc 4.8, so old toolchains such gcc-4.7 will complain:
> >>>   error: unrecognized argument in option ?-march=armv8-a?
> >>
> >>That's a fair point. I guess -march=armv8-a is not strictly necessary and
> >>the produced vDSO should be fine if arch/arm/vdso also compiles fine.
> >>However we would still need to pass -march=armv7-a. I'm not sure what to do
> >>between:
> >>* Checking that the compiler supports -march=armv8-a when inspecting
> >>CROSS_COMPILE_ARM32, and if it doesn't vdso32 will not be built.
> >>* Checking whether -march=armv8-a is available here, and if it is not fall
> >>back to -march=armv7-a.
> >
> >Does v8 vs v7 make any difference in the generated code? If not, we
> >could just stick to armv7-a permanently.
> 
> I've just tried compiling with -march=armv7-a, and in fact it doesn't
> compile at all. It turns out vgettimeofday.c uses smp_rmb(), which expands
> to dmb ishld on arm64, and ishld doesn't exist in ARMv7. We could possibly
> work around that, but I think requiring GCC 4.8 is reasonable.

Since vgettimeofday.c is meant to be compiled for AArch32, it wouldn't
look too bad to define its own barriers rather than relying on the
AArch64 ones. So you could define v7_smp_rmb/v7_smp_wmb and use them in
this file. Alternatively, replace smp_rmb() with smp_mb() in this file
but with a big comment about ARMv7 compilation requirement and "ishld"
not being available.

-- 
Catalin

^ permalink raw reply	[flat|nested] 19+ messages in thread

* [RFC PATCH v2 4/8] arm64: compat: Add a 32-bit vDSO
  2016-11-21 18:44           ` Catalin Marinas
@ 2016-12-01 14:27             ` Kevin Brodsky
  0 siblings, 0 replies; 19+ messages in thread
From: Kevin Brodsky @ 2016-12-01 14:27 UTC (permalink / raw)
  To: linux-arm-kernel

On 21/11/16 18:44, Catalin Marinas wrote:
> On Mon, Nov 21, 2016 at 03:45:55PM +0000, Kevin Brodsky wrote:
>> On 04/11/16 20:03, Catalin Marinas wrote:
>>> On Fri, Oct 28, 2016 at 11:20:07AM +0100, Kevin Brodsky wrote:
>>>> On 28/10/16 04:09, Jisheng Zhang wrote:
>>>>> On Thu, 27 Oct 2016 17:30:54 +0100 Kevin Brodsky wrote:
>>>>>> +# Force -O2 to avoid libgcc dependencies
>>>>>> +VDSO_CFLAGS := -march=armv8-a -O2
>>>>> For completeness, bringing 32bit compiler need to check whether the 32bit
>>>>> toolchain support some options. IIRC, armv8-a support isn't enabled until
>>>>> gcc 4.8, so old toolchains such gcc-4.7 will complain:
>>>>>    error: unrecognized argument in option ?-march=armv8-a?
>>>> That's a fair point. I guess -march=armv8-a is not strictly necessary and
>>>> the produced vDSO should be fine if arch/arm/vdso also compiles fine.
>>>> However we would still need to pass -march=armv7-a. I'm not sure what to do
>>>> between:
>>>> * Checking that the compiler supports -march=armv8-a when inspecting
>>>> CROSS_COMPILE_ARM32, and if it doesn't vdso32 will not be built.
>>>> * Checking whether -march=armv8-a is available here, and if it is not fall
>>>> back to -march=armv7-a.
>>> Does v8 vs v7 make any difference in the generated code? If not, we
>>> could just stick to armv7-a permanently.
>> I've just tried compiling with -march=armv7-a, and in fact it doesn't
>> compile at all. It turns out vgettimeofday.c uses smp_rmb(), which expands
>> to dmb ishld on arm64, and ishld doesn't exist in ARMv7. We could possibly
>> work around that, but I think requiring GCC 4.8 is reasonable.
> Since vgettimeofday.c is meant to be compiled for AArch32, it wouldn't
> look too bad to define its own barriers rather than relying on the
> AArch64 ones. So you could define v7_smp_rmb/v7_smp_wmb and use them in
> this file. Alternatively, replace smp_rmb() with smp_mb() in this file
> but with a big comment about ARMv7 compilation requirement and "ishld"
> not being available.

Fair enough. I'll add AArch32 barrier macros and compile with -march=armv7-a if the 
compiler doesn't support armv8-a. It's true that using arm64 barrier macros in 32-bit 
code is a bit dodgy anyway.

Cheers,
Kevin

^ permalink raw reply	[flat|nested] 19+ messages in thread

end of thread, other threads:[~2016-12-01 14:27 UTC | newest]

Thread overview: 19+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-10-27 16:30 [RFC PATCH v2 0/8] arm64: Add a compat vDSO Kevin Brodsky
2016-10-27 16:30 ` [RFC PATCH v2 1/8] arm64: Refactor vDSO setup Kevin Brodsky
2016-10-27 16:30 ` [RFC PATCH v2 2/8] arm64: compat: Add time-related syscall numbers Kevin Brodsky
2016-10-27 16:30 ` [RFC PATCH v2 3/8] arm64: compat: Expose offset to registers in sigframes Kevin Brodsky
2016-10-27 16:30 ` [RFC PATCH v2 4/8] arm64: compat: Add a 32-bit vDSO Kevin Brodsky
2016-10-28  3:09   ` Jisheng Zhang
2016-10-28 10:20     ` Kevin Brodsky
2016-11-04 20:03       ` Catalin Marinas
2016-11-21 15:45         ` Kevin Brodsky
2016-11-21 18:44           ` Catalin Marinas
2016-12-01 14:27             ` Kevin Brodsky
2016-10-27 16:30 ` [RFC PATCH v2 5/8] arm64: compat: 32-bit vDSO setup Kevin Brodsky
2016-10-27 16:30 ` [RFC PATCH v2 6/8] arm64: elf: Set AT_SYSINFO_EHDR in compat processes Kevin Brodsky
2016-10-27 16:30 ` [RFC PATCH v2 7/8] arm64: compat: Use vDSO sigreturn trampolines if available Kevin Brodsky
2016-10-27 16:30 ` [RFC PATCH v2 8/8] arm64: Wire up and expose the new compat vDSO Kevin Brodsky
2016-11-04 15:50   ` Catalin Marinas
2016-11-04 16:30     ` Kevin Brodsky
2016-11-04 16:47       ` Catalin Marinas
2016-11-04 17:53         ` Kevin Brodsky

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.