[patch 0/2] Updates to compat VDSOs

[patch 0/2] Updates to compat VDSOs

[Jeremy Fitzhardinge]

Hi Andi,

Here's a couple of patches to fix up COMPAT_VDSO:

The first is a straightforward implementation of Jan's original idea
of relocating the VDSO to match its mapped location.  Unlike Jan and
Zach's version, I changed it to relocate based on the phdrs rather than
the sections; the result is pleasantly compact.

The second patch takes advantage of the fact that all the COMPAT_VDSO work
happens at runtime now, and allows compat mode to be enabled dynamically.
If you specify vdso=2 on the kernel command line, it comes up in compat
mode; vdso=1 is normal vdso mode, and vdso=0 disables vdso altogether.
You can also switch modes with sysctl.

Thanks,
	J

-- 

[patch 1/2] Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

[Jeremy Fitzhardinge]

[-- Attachment #1: relocate-COMPAT_VDSO.patch --]
[-- Type: text/plain, Size: 8707 bytes --]

Some versions of libc can't deal with a VDSO which doesn't have its
ELF headers matching its mapped address.  COMPAT_VDSO maps the VDSO at
a specific system-wide fixed address.  Previously this was all done at
build time, on the grounds that the fixed VDSO address is always at
the top of the address space.  However, a hypervisor may reserve some
of that address space, pushing the fixmap address down.

This patch does the adjustment dynamically at runtime, depending on
the runtime location of the VDSO fixmap.

[ Patch has been through several hands: Jan Beulich wrote the orignal
  version; Zach reworked it, and Jeremy converted it to relocate phdrs
  rather than sections. ]

Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Zachary Amsden <zach@vmware.com>
Cc: "Jan Beulich" <JBeulich@novell.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Roland McGrath <roland@redhat.com>

---
 arch/i386/kernel/entry.S    |    4 -
 arch/i386/kernel/sysenter.c |   95 ++++++++++++++++++++++++++++++++++++-------
 arch/i386/mm/pgtable.c      |    6 --
 include/asm-i386/elf.h      |   28 ++++--------
 include/asm-i386/fixmap.h   |    8 ---
 include/linux/elf.h         |    3 +
 6 files changed, 95 insertions(+), 49 deletions(-)

===================================================================
--- a/arch/i386/kernel/entry.S
+++ b/arch/i386/kernel/entry.S
@@ -305,16 +305,12 @@ sysenter_past_esp:
 	pushl $(__USER_CS)
 	CFI_ADJUST_CFA_OFFSET 4
 	/*CFI_REL_OFFSET cs, 0*/
-#ifndef CONFIG_COMPAT_VDSO
 	/*
 	 * Push current_thread_info()->sysenter_return to the stack.
 	 * A tiny bit of offset fixup is necessary - 4*4 means the 4 words
 	 * pushed above; +8 corresponds to copy_thread's esp0 setting.
 	 */
 	pushl (TI_sysenter_return-THREAD_SIZE+8+4*4)(%esp)
-#else
-	pushl $SYSENTER_RETURN
-#endif
 	CFI_ADJUST_CFA_OFFSET 4
 	CFI_REL_OFFSET eip, 0
 
===================================================================
--- a/arch/i386/kernel/sysenter.c
+++ b/arch/i386/kernel/sysenter.c
@@ -22,6 +22,7 @@
 #include <asm/msr.h>
 #include <asm/pgtable.h>
 #include <asm/unistd.h>
+#include <asm/elf.h>
 
 /*
  * Should the kernel map a VDSO page into processes and pass its
@@ -41,6 +42,54 @@ __setup("vdso=", vdso_setup);
 __setup("vdso=", vdso_setup);
 
 extern asmlinkage void sysenter_entry(void);
+
+#ifdef CONFIG_COMPAT_VDSO
+static __cpuinit void reloc_dyn(Elf32_Ehdr *ehdr, unsigned offset)
+{
+	Elf32_Dyn *dyn = (void *)ehdr + offset;
+
+	for(; dyn->d_tag != DT_NULL; dyn++)
+		switch(dyn->d_tag) {
+		case DT_PLTGOT:
+		case DT_HASH:
+		case DT_STRTAB:
+		case DT_SYMTAB:
+		case DT_RELA:
+		case DT_INIT:
+		case DT_FINI:
+		case DT_REL:
+		case DT_JMPREL:
+		case DT_VERSYM:
+		case DT_VERDEF:
+		case DT_VERNEED:
+			dyn->d_un.d_val += VDSO_HIGH_BASE;
+		}
+}
+
+static __cpuinit void relocate_vdso(Elf32_Ehdr *ehdr)
+{
+	Elf32_Phdr *phdr;
+	int i;
+
+	BUG_ON(memcmp(ehdr->e_ident, ELFMAG, 4) != 0 ||
+	       !elf_check_arch(ehdr) ||
+	       ehdr->e_type != ET_DYN);
+
+	ehdr->e_entry += VDSO_HIGH_BASE;
+
+	phdr = (void *)ehdr + ehdr->e_phoff;
+	for (i = 0; i < ehdr->e_phnum; i++) {
+		phdr[i].p_vaddr += VDSO_HIGH_BASE;
+
+		if (phdr[i].p_type == PT_DYNAMIC)
+			reloc_dyn(ehdr, phdr[i].p_offset);
+	}
+}
+#else
+static inline void relocate_vdso(Elf32_Ehdr *ehdr)
+{
+}
+#endif	/* COMPAT_VDSO */
 
 void enable_sep_cpu(void)
 {
@@ -71,6 +120,9 @@ int __cpuinit sysenter_setup(void)
 int __cpuinit sysenter_setup(void)
 {
 	void *syscall_page = (void *)get_zeroed_page(GFP_ATOMIC);
+	const void *vsyscall;
+	size_t vsyscall_len;
+
 	syscall_pages[0] = virt_to_page(syscall_page);
 
 #ifdef CONFIG_COMPAT_VDSO
@@ -79,23 +131,23 @@ int __cpuinit sysenter_setup(void)
 #endif
 
 	if (!boot_cpu_has(X86_FEATURE_SEP)) {
-		memcpy(syscall_page,
-		       &vsyscall_int80_start,
-		       &vsyscall_int80_end - &vsyscall_int80_start);
-		return 0;
-	}
-
-	memcpy(syscall_page,
-	       &vsyscall_sysenter_start,
-	       &vsyscall_sysenter_end - &vsyscall_sysenter_start);
-
-	return 0;
-}
-
-#ifndef CONFIG_COMPAT_VDSO
+		vsyscall = &vsyscall_int80_start;
+		vsyscall_len = &vsyscall_int80_end - &vsyscall_int80_start;
+	} else {
+		vsyscall = &vsyscall_sysenter_start;
+		vsyscall_len = &vsyscall_sysenter_end - &vsyscall_sysenter_start;
+	}
+
+	memcpy(syscall_page, vsyscall, vsyscall_len);
+	relocate_vdso(syscall_page);
+
+	return 0;
+}
+
 /* Defined in vsyscall-sysenter.S */
 extern void SYSENTER_RETURN;
 
+#ifdef __HAVE_ARCH_GATE_AREA
 /* Setup a VMA at program startup for the vsyscall page */
 int arch_setup_additional_pages(struct linux_binprm *bprm, int exstack)
 {
@@ -155,4 +207,17 @@ int in_gate_area_no_task(unsigned long a
 {
 	return 0;
 }
-#endif
+#else  /* !__HAVE_ARCH_GATE_AREA */
+int arch_setup_additional_pages(struct linux_binprm *bprm, int exstack)
+{
+	/*
+	 * If not creating userspace VMA, simply set vdso to point to
+	 * fixmap page.
+	 */
+	current->mm->context.vdso = (void *)VDSO_HIGH_BASE;
+	current_thread_info()->sysenter_return =
+		(void *)VDSO_SYM(&SYSENTER_RETURN);
+
+	return 0;
+}
+#endif	/* __HAVE_ARCH_GATE_AREA */
===================================================================
--- a/arch/i386/mm/pgtable.c
+++ b/arch/i386/mm/pgtable.c
@@ -144,10 +144,8 @@ void set_pmd_pfn(unsigned long vaddr, un
 }
 
 static int fixmaps;
-#ifndef CONFIG_COMPAT_VDSO
 unsigned long __FIXADDR_TOP = 0xfffff000;
 EXPORT_SYMBOL(__FIXADDR_TOP);
-#endif
 
 void __set_fixmap (enum fixed_addresses idx, unsigned long phys, pgprot_t flags)
 {
@@ -173,12 +171,8 @@ void reserve_top_address(unsigned long r
 	BUG_ON(fixmaps > 0);
 	printk(KERN_INFO "Reserving virtual address space above 0x%08x\n",
 	       (int)-reserve);
-#ifdef CONFIG_COMPAT_VDSO
-	BUG_ON(reserve != 0);
-#else
 	__FIXADDR_TOP = -reserve - PAGE_SIZE;
 	__VMALLOC_RESERVE += reserve;
-#endif
 }
 
 pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address)
===================================================================
--- a/include/asm-i386/elf.h
+++ b/include/asm-i386/elf.h
@@ -133,39 +133,31 @@ extern int dump_task_extended_fpu (struc
 #define ELF_CORE_COPY_XFPREGS(tsk, elf_xfpregs) dump_task_extended_fpu(tsk, elf_xfpregs)
 
 #define VDSO_HIGH_BASE		(__fix_to_virt(FIX_VDSO))
-#define VDSO_BASE		((unsigned long)current->mm->context.vdso)
-
-#ifdef CONFIG_COMPAT_VDSO
-# define VDSO_COMPAT_BASE	VDSO_HIGH_BASE
-# define VDSO_PRELINK		VDSO_HIGH_BASE
-#else
-# define VDSO_COMPAT_BASE	VDSO_BASE
-# define VDSO_PRELINK		0
-#endif
+#define VDSO_CURRENT_BASE	((unsigned long)current->mm->context.vdso)
+#define VDSO_PRELINK		0
 
 #define VDSO_SYM(x) \
-		(VDSO_COMPAT_BASE + (unsigned long)(x) - VDSO_PRELINK)
+		(VDSO_CURRENT_BASE + (unsigned long)(x) - VDSO_PRELINK)
 
 #define VDSO_HIGH_EHDR		((const struct elfhdr *) VDSO_HIGH_BASE)
-#define VDSO_EHDR		((const struct elfhdr *) VDSO_COMPAT_BASE)
+#define VDSO_EHDR		((const struct elfhdr *) VDSO_CURRENT_BASE)
 
 extern void __kernel_vsyscall;
 
 #define VDSO_ENTRY		VDSO_SYM(&__kernel_vsyscall)
 
-#ifndef CONFIG_COMPAT_VDSO
+struct linux_binprm;
+
 #define ARCH_HAS_SETUP_ADDITIONAL_PAGES
-struct linux_binprm;
 extern int arch_setup_additional_pages(struct linux_binprm *bprm,
                                        int executable_stack);
-#endif
 
 extern unsigned int vdso_enabled;
 
-#define ARCH_DLINFO						\
-do if (vdso_enabled) {						\
-		NEW_AUX_ENT(AT_SYSINFO,	VDSO_ENTRY);		\
-		NEW_AUX_ENT(AT_SYSINFO_EHDR, VDSO_COMPAT_BASE);	\
+#define ARCH_DLINFO							\
+do if (vdso_enabled) {							\
+		NEW_AUX_ENT(AT_SYSINFO,	VDSO_ENTRY);			\
+		NEW_AUX_ENT(AT_SYSINFO_EHDR, VDSO_CURRENT_BASE);	\
 } while (0)
 
 #endif
===================================================================
--- a/include/asm-i386/fixmap.h
+++ b/include/asm-i386/fixmap.h
@@ -19,13 +19,9 @@
  * Leave one empty page between vmalloc'ed areas and
  * the start of the fixmap.
  */
-#ifndef CONFIG_COMPAT_VDSO
 extern unsigned long __FIXADDR_TOP;
-#else
-#define __FIXADDR_TOP  0xfffff000
-#define FIXADDR_USER_START	__fix_to_virt(FIX_VDSO)
-#define FIXADDR_USER_END	__fix_to_virt(FIX_VDSO - 1)
-#endif
+#define FIXADDR_USER_START     __fix_to_virt(FIX_VDSO)
+#define FIXADDR_USER_END       __fix_to_virt(FIX_VDSO - 1)
 
 #ifndef __ASSEMBLY__
 #include <linux/kernel.h>
===================================================================
--- a/include/linux/elf.h
+++ b/include/linux/elf.h
@@ -83,6 +83,9 @@ typedef __s64	Elf64_Sxword;
 #define DT_DEBUG	21
 #define DT_TEXTREL	22
 #define DT_JMPREL	23
+#define DT_VERSYM	0x6ffffff0
+#define DT_VERDEF	0x6ffffffc
+#define DT_VERNEED	0x6ffffffe
 #define DT_LOPROC	0x70000000
 #define DT_HIPROC	0x7fffffff
 

-- 

Re: [patch 1/2] Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

[Roland McGrath]

The patch looks nice and clean.  However, it does not relocate the symbol
table(s) values.  I thought that was done in an earlier version of this I
saw, but I might be misremembering.  Though not fatal, this is a regression
from the previous CONFIG_COMPAT_VDSO behavior.  It will show up in things
like __kernel_* name display in backtraces.  If with your other patch
CONFIG_COMPAT_VDSO will become other than a rarely-used compatibility
option, then this should be fixed.  Note that with your second patch this
will also break the symbol values in the randomly-located vma vdso;
non-ancient glibc doesn't care if the vdso isn't mapped where its phdrs
say, but everything does still care that the symbol tables in an ELF file
use addresses matching the phdrs in the same file.


Thanks,
Roland

Re: [patch 1/2] Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

[Jeremy Fitzhardinge]

Roland McGrath wrote:
> The patch looks nice and clean.  However, it does not relocate the symbol
> table(s) values.  I thought that was done in an earlier version of this I
> saw, but I might be misremembering.  Though not fatal, this is a regression
> from the previous CONFIG_COMPAT_VDSO behavior.  It will show up in things
> like __kernel_* name display in backtraces.

Hm, OK.  It does, but I wasn't sure if it would matter.  It should be
fairly simple to fix up.

>   If with your other patch
> CONFIG_COMPAT_VDSO will become other than a rarely-used compatibility
> option, then this should be fixed.  Note that with your second patch this
> will also break the symbol values in the randomly-located vma vdso;
> non-ancient glibc doesn't care if the vdso isn't mapped where its phdrs
> say, but everything does still care that the symbol tables in an ELF file
> use addresses matching the phdrs in the same file.
>   

I did the second patch because I could, and to see if it would provoke
some comment.  But effectively removing a kernel config option seems
like a good idea to me.

    J

Re: [patch 1/2] Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

[Roland McGrath]

> I did the second patch because I could, and to see if it would provoke
> some comment.  But effectively removing a kernel config option seems
> like a good idea to me.

Well, it provoked me to care whether the first patch the relocation
"really right".


Thanks,
Roland

Re: [patch 1/2] Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

[Jan Beulich]

>+static __cpuinit void reloc_dyn(Elf32_Ehdr *ehdr, unsigned offset)
>+{
>+	Elf32_Dyn *dyn = (void *)ehdr + offset;
>+
>+	for(; dyn->d_tag != DT_NULL; dyn++)
>+		switch(dyn->d_tag) {
>+		case DT_PLTGOT:
>+		case DT_HASH:
>+		case DT_STRTAB:
>+		case DT_SYMTAB:
>+		case DT_RELA:
>+		case DT_INIT:
>+		case DT_FINI:
>+		case DT_REL:
>+		case DT_JMPREL:
>+		case DT_VERSYM:
>+		case DT_VERDEF:
>+		case DT_VERNEED:
>+			dyn->d_un.d_val += VDSO_HIGH_BASE;
>+		}
>+}

While there's a certain level of control on what DT_* may appear in the
vDSO, not even considering other than the above types seems fragile to
me. Since future additions to the set are supposedly following a fixed
scheme (distinguishing pointers and values via the low bit when below
OLD_DT_LOOS, and using sub-ranges when between DT_HIOS and
OLD_DT_HIOS), at least also handling those would seem like a good
idea, as would warning about unrecognized types.

Also, even though it shouldn't matter for the final result, if doing things
spec-conforming here you should use d_un.d_ptr.

In addition to Roland's remarks about missing symbol table relocation, I
would also assume section headers, if present, should be relocated.

Jan

Re: [patch 1/2] Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

[Jeremy Fitzhardinge]

Jan Beulich wrote:
> While there's a certain level of control on what DT_* may appear in the
> vDSO, not even considering other than the above types seems fragile to
> me. Since future additions to the set are supposedly following a fixed
> scheme (distinguishing pointers and values via the low bit when below
> OLD_DT_LOOS, and using sub-ranges when between DT_HIOS and
> OLD_DT_HIOS), at least also handling those would seem like a good
> idea, as would warning about unrecognized types.
>   

I wasn't aware of this scheme.  Where is it documented?

> Also, even though it shouldn't matter for the final result, if doing things
> spec-conforming here you should use d_un.d_ptr.
>   

Yes, I've already fixed that.

> In addition to Roland's remarks about missing symbol table relocation, I
> would also assume section headers, if present, should be relocated.
>   

Yes, I suppose that's easy enough to add.

    J

RE: [patch 1/2] Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

[Raharjo, Cahyo (cahr)]

Dear All,
In this few days, I got emails from this list. I tried to exclude thru
majordomo, but it failed.
Kindly need your help to exclude me from this list.


Thank You, 
CR


-----Original Message-----
From: linux-kernel-owner@vger.kernel.org
[mailto:linux-kernel-owner@vger.kernel.org] On Behalf Of Jeremy
Fitzhardinge
Sent: Thursday, April 05, 2007 2:32 PM
To: Jan Beulich
Cc: Ingo Molnar; Andrew Morton; virtualization@lists.osdl.org; Roland
McGrath; Andi Kleen; lkml; Zachary Amsden; Eric W. Biederman
Subject: Re: [patch 1/2] Relocate VDSO ELF headers to match mapped
location with COMPAT_VDSO

Jan Beulich wrote:
> While there's a certain level of control on what DT_* may appear in 
> the vDSO, not even considering other than the above types seems 
> fragile to me. Since future additions to the set are supposedly 
> following a fixed scheme (distinguishing pointers and values via the 
> low bit when below OLD_DT_LOOS, and using sub-ranges when between 
> DT_HIOS and OLD_DT_HIOS), at least also handling those would seem like

> a good idea, as would warning about unrecognized types.
>   

I wasn't aware of this scheme.  Where is it documented?

> Also, even though it shouldn't matter for the final result, if doing 
> things spec-conforming here you should use d_un.d_ptr.
>   

Yes, I've already fixed that.

> In addition to Roland's remarks about missing symbol table relocation,

> I would also assume section headers, if present, should be relocated.
>   

Yes, I suppose that's easy enough to add.

    J
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel"
in the body of a message to majordomo@vger.kernel.org More majordomo
info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Re: [patch 1/2] Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

[Jan Beulich]

>>> Jeremy Fitzhardinge <jeremy@goop.org> 05.04.07 09:31 >>>
>Jan Beulich wrote:
>> While there's a certain level of control on what DT_* may appear in the
>> vDSO, not even considering other than the above types seems fragile to
>> me. Since future additions to the set are supposedly following a fixed
>> scheme (distinguishing pointers and values via the low bit when below
>> OLD_DT_LOOS, and using sub-ranges when between DT_HIOS and
>> OLD_DT_HIOS), at least also handling those would seem like a good
>> idea, as would warning about unrecognized types.
>
>I wasn't aware of this scheme.  Where is it documented?

Regarding the low bit part, quoting from working gABI chapter 5: "To make
it simpler for tools to interpret the contents of dynamic section entries, the
value of each tag, except for those in two special compatibility ranges, will
determine the interpretation of the d_un union. A tag whose value is an
even number indicates a dynamic section entry that uses d_ptr. A tag
whose value is an odd number indicates a dynamic section entry that uses
d_val or that uses neither d_ptr nor d_val. Tags whose values are less
than the special value DT_ENCODING and tags whose values fall between
DT_HIOS and DT_LOPROC do not follow these rules."

Regarding the OS range, all I can point you to is binutils' include/elf/common.h.

Jan

Re: [patch 1/2] Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

[Roland McGrath]

> In addition to Roland's remarks about missing symbol table relocation, I
> would also assume section headers, if present, should be relocated.

Yes, and also the .symtab as well as .dynsym just in case one ever has one
(though I think they are built stripped now, it's not hard to check sh_type
for SHT_SYMTAB and call the same routine that works from DT_SYMTAB).


Thanks,
Roland

Re: [patch 1/2] Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

[Jeremy Fitzhardinge]

Roland McGrath wrote:
>> In addition to Roland's remarks about missing symbol table relocation, I
>> would also assume section headers, if present, should be relocated.
>>     
>
> Yes, and also the .symtab as well as .dynsym just in case one ever has one
> (though I think they are built stripped now, it's not hard to check sh_type
> for SHT_SYMTAB and call the same routine that works from DT_SYMTAB).
>   

OK, how does this look?  Does it need to worry about DT_REL/DT_RELA
tags? (Not that there are any.)

    J

Subject: Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

Some versions of libc can't deal with a VDSO which doesn't have its
ELF headers matching its mapped address.  COMPAT_VDSO maps the VDSO at
a specific system-wide fixed address.  Previously this was all done at
build time, on the grounds that the fixed VDSO address is always at
the top of the address space.  However, a hypervisor may reserve some
of that address space, pushing the fixmap address down.

This patch does the adjustment dynamically at runtime, depending on
the runtime location of the VDSO fixmap.

[ Patch has been through several hands: Jan Beulich wrote the orignal
  version; Zach reworked it, and Jeremy converted it to relocate phdrs
  rather than sections. ]

Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Zachary Amsden <zach@vmware.com>
Cc: "Jan Beulich" <JBeulich@novell.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Roland McGrath <roland@redhat.com>

---
 arch/i386/kernel/entry.S    |    4 -
 arch/i386/kernel/sysenter.c |   95 ++++++++++++++++++++++++++++++++++++-------
 arch/i386/mm/pgtable.c      |    6 --
 include/asm-i386/elf.h      |   28 ++++--------
 include/asm-i386/fixmap.h   |    8 ---
 include/linux/elf.h         |    3 +
 6 files changed, 95 insertions(+), 49 deletions(-)

diff -r 97a792fb6127 arch/i386/kernel/entry.S
--- a/arch/i386/kernel/entry.S	Wed Apr 04 23:26:37 2007 -0700
+++ b/arch/i386/kernel/entry.S	Thu Apr 05 01:11:25 2007 -0700
@@ -305,16 +305,12 @@ sysenter_past_esp:
 	pushl $(__USER_CS)
 	CFI_ADJUST_CFA_OFFSET 4
 	/*CFI_REL_OFFSET cs, 0*/
-#ifndef CONFIG_COMPAT_VDSO
 	/*
 	 * Push current_thread_info()->sysenter_return to the stack.
 	 * A tiny bit of offset fixup is necessary - 4*4 means the 4 words
 	 * pushed above; +8 corresponds to copy_thread's esp0 setting.
 	 */
 	pushl (TI_sysenter_return-THREAD_SIZE+8+4*4)(%esp)
-#else
-	pushl $SYSENTER_RETURN
-#endif
 	CFI_ADJUST_CFA_OFFSET 4
 	CFI_REL_OFFSET eip, 0
 
diff -r 97a792fb6127 arch/i386/kernel/sysenter.c
--- a/arch/i386/kernel/sysenter.c	Wed Apr 04 23:26:37 2007 -0700
+++ b/arch/i386/kernel/sysenter.c	Thu Apr 05 01:11:25 2007 -0700
@@ -22,6 +22,7 @@
 #include <asm/msr.h>
 #include <asm/pgtable.h>
 #include <asm/unistd.h>
+#include <asm/elf.h>
 
 /*
  * Should the kernel map a VDSO page into processes and pass its
@@ -41,6 +42,102 @@ __setup("vdso=", vdso_setup);
 __setup("vdso=", vdso_setup);
 
 extern asmlinkage void sysenter_entry(void);
+
+#ifdef CONFIG_COMPAT_VDSO
+static __cpuinit void reloc_symtab(Elf32_Ehdr *ehdr,
+				   unsigned offset, unsigned size)
+{
+	Elf32_Sym *sym = (void *)ehdr + offset;
+	unsigned nsym = size / sizeof(*sym);
+	unsigned i;
+
+	for(i = 0; i < nsym; i++, sym++) {
+		if (sym->st_shndx == SHN_UNDEF ||
+		    sym->st_shndx == SHN_ABS)
+			continue;  /* skip */
+
+		switch(ELF_ST_TYPE(sym->st_info)) {
+		case STT_OBJECT:
+		case STT_FUNC:
+		case STT_SECTION:
+		case STT_FILE:
+			sym->st_value += VDSO_HIGH_BASE;
+		}
+	}
+}
+
+struct elfhash {
+	Elf32_Word nbucket, nchain;
+};
+
+static __cpuinit void reloc_dyn(Elf32_Ehdr *ehdr, unsigned offset)
+{
+	Elf32_Dyn *dyn = (void *)ehdr + offset;
+
+	for(; dyn->d_tag != DT_NULL; dyn++)
+		switch(dyn->d_tag) {
+		case DT_PLTGOT:
+		case DT_HASH:
+		case DT_STRTAB:
+		case DT_SYMTAB:
+		case DT_RELA:
+		case DT_INIT:
+		case DT_FINI:
+		case DT_REL:
+		case DT_DEBUG:
+		case DT_JMPREL:
+		case DT_VERSYM:
+		case DT_VERDEF:
+		case DT_VERNEED:
+		case DT_ENCODING ... DT_HIOS:
+			/* tags above DT_ENCODING are even if they're
+			   a pointer, so skip odd ones */
+			if (dyn->d_tag >= DT_ENCODING &&
+			    (dyn->d_tag & 1) == 1)
+				break;
+
+			dyn->d_un.d_ptr += VDSO_HIGH_BASE;
+		}
+}
+
+static __cpuinit void relocate_vdso(Elf32_Ehdr *ehdr)
+{
+	Elf32_Phdr *phdr;
+	Elf32_Shdr *shdr;
+	int i;
+
+	BUG_ON(memcmp(ehdr->e_ident, ELFMAG, 4) != 0 ||
+	       !elf_check_arch(ehdr) ||
+	       ehdr->e_type != ET_DYN);
+
+	ehdr->e_entry += VDSO_HIGH_BASE;
+
+	/* rebase phdrs */
+	phdr = (void *)ehdr + ehdr->e_phoff;
+	for (i = 0; i < ehdr->e_phnum; i++) {
+		phdr[i].p_vaddr += VDSO_HIGH_BASE;
+
+		/* relocate dynamic stuff */
+		if (phdr[i].p_type == PT_DYNAMIC)
+			reloc_dyn(ehdr, phdr[i].p_offset);
+	}
+
+	/* rebase sections */
+	shdr = (void *)ehdr + ehdr->e_shoff;
+	for(i = 0; i < ehdr->e_shnum; i++) {
+		shdr[i].sh_addr += VDSO_HIGH_BASE;
+
+		if (shdr[i].sh_type == SHT_SYMTAB ||
+		    shdr[i].sh_type == SHT_DYNSYM)
+			reloc_symtab(ehdr, shdr[i].sh_offset,
+				     shdr[i].sh_size);
+	}
+}
+#else
+static inline void relocate_vdso(Elf32_Ehdr *ehdr)
+{
+}
+#endif	/* COMPAT_VDSO */
 
 void enable_sep_cpu(void)
 {
@@ -71,6 +168,9 @@ int __cpuinit sysenter_setup(void)
 int __cpuinit sysenter_setup(void)
 {
 	void *syscall_page = (void *)get_zeroed_page(GFP_ATOMIC);
+	const void *vsyscall;
+	size_t vsyscall_len;
+
 	syscall_pages[0] = virt_to_page(syscall_page);
 
 #ifdef CONFIG_COMPAT_VDSO
@@ -79,23 +179,23 @@ int __cpuinit sysenter_setup(void)
 #endif
 
 	if (!boot_cpu_has(X86_FEATURE_SEP)) {
-		memcpy(syscall_page,
-		       &vsyscall_int80_start,
-		       &vsyscall_int80_end - &vsyscall_int80_start);
-		return 0;
-	}
-
-	memcpy(syscall_page,
-	       &vsyscall_sysenter_start,
-	       &vsyscall_sysenter_end - &vsyscall_sysenter_start);
-
-	return 0;
-}
-
-#ifndef CONFIG_COMPAT_VDSO
+		vsyscall = &vsyscall_int80_start;
+		vsyscall_len = &vsyscall_int80_end - &vsyscall_int80_start;
+	} else {
+		vsyscall = &vsyscall_sysenter_start;
+		vsyscall_len = &vsyscall_sysenter_end - &vsyscall_sysenter_start;
+	}
+
+	memcpy(syscall_page, vsyscall, vsyscall_len);
+	relocate_vdso(syscall_page);
+
+	return 0;
+}
+
 /* Defined in vsyscall-sysenter.S */
 extern void SYSENTER_RETURN;
 
+#ifdef __HAVE_ARCH_GATE_AREA
 /* Setup a VMA at program startup for the vsyscall page */
 int arch_setup_additional_pages(struct linux_binprm *bprm, int exstack)
 {
@@ -155,4 +255,17 @@ int in_gate_area_no_task(unsigned long a
 {
 	return 0;
 }
-#endif
+#else  /* !__HAVE_ARCH_GATE_AREA */
+int arch_setup_additional_pages(struct linux_binprm *bprm, int exstack)
+{
+	/*
+	 * If not creating userspace VMA, simply set vdso to point to
+	 * fixmap page.
+	 */
+	current->mm->context.vdso = (void *)VDSO_HIGH_BASE;
+	current_thread_info()->sysenter_return =
+		(void *)VDSO_SYM(&SYSENTER_RETURN);
+
+	return 0;
+}
+#endif	/* __HAVE_ARCH_GATE_AREA */
diff -r 97a792fb6127 arch/i386/mm/pgtable.c
--- a/arch/i386/mm/pgtable.c	Wed Apr 04 23:26:37 2007 -0700
+++ b/arch/i386/mm/pgtable.c	Thu Apr 05 01:11:25 2007 -0700
@@ -144,10 +144,8 @@ void set_pmd_pfn(unsigned long vaddr, un
 }
 
 static int fixmaps;
-#ifndef CONFIG_COMPAT_VDSO
 unsigned long __FIXADDR_TOP = 0xfffff000;
 EXPORT_SYMBOL(__FIXADDR_TOP);
-#endif
 
 void __set_fixmap (enum fixed_addresses idx, unsigned long phys, pgprot_t flags)
 {
@@ -173,12 +171,8 @@ void reserve_top_address(unsigned long r
 	BUG_ON(fixmaps > 0);
 	printk(KERN_INFO "Reserving virtual address space above 0x%08x\n",
 	       (int)-reserve);
-#ifdef CONFIG_COMPAT_VDSO
-	BUG_ON(reserve != 0);
-#else
 	__FIXADDR_TOP = -reserve - PAGE_SIZE;
 	__VMALLOC_RESERVE += reserve;
-#endif
 }
 
 pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address)
diff -r 97a792fb6127 include/asm-i386/elf.h
--- a/include/asm-i386/elf.h	Wed Apr 04 23:26:37 2007 -0700
+++ b/include/asm-i386/elf.h	Thu Apr 05 01:11:25 2007 -0700
@@ -133,39 +133,31 @@ extern int dump_task_extended_fpu (struc
 #define ELF_CORE_COPY_XFPREGS(tsk, elf_xfpregs) dump_task_extended_fpu(tsk, elf_xfpregs)
 
 #define VDSO_HIGH_BASE		(__fix_to_virt(FIX_VDSO))
-#define VDSO_BASE		((unsigned long)current->mm->context.vdso)
-
-#ifdef CONFIG_COMPAT_VDSO
-# define VDSO_COMPAT_BASE	VDSO_HIGH_BASE
-# define VDSO_PRELINK		VDSO_HIGH_BASE
-#else
-# define VDSO_COMPAT_BASE	VDSO_BASE
-# define VDSO_PRELINK		0
-#endif
+#define VDSO_CURRENT_BASE	((unsigned long)current->mm->context.vdso)
+#define VDSO_PRELINK		0
 
 #define VDSO_SYM(x) \
-		(VDSO_COMPAT_BASE + (unsigned long)(x) - VDSO_PRELINK)
+		(VDSO_CURRENT_BASE + (unsigned long)(x) - VDSO_PRELINK)
 
 #define VDSO_HIGH_EHDR		((const struct elfhdr *) VDSO_HIGH_BASE)
-#define VDSO_EHDR		((const struct elfhdr *) VDSO_COMPAT_BASE)
+#define VDSO_EHDR		((const struct elfhdr *) VDSO_CURRENT_BASE)
 
 extern void __kernel_vsyscall;
 
 #define VDSO_ENTRY		VDSO_SYM(&__kernel_vsyscall)
 
-#ifndef CONFIG_COMPAT_VDSO
+struct linux_binprm;
+
 #define ARCH_HAS_SETUP_ADDITIONAL_PAGES
-struct linux_binprm;
 extern int arch_setup_additional_pages(struct linux_binprm *bprm,
                                        int executable_stack);
-#endif
 
 extern unsigned int vdso_enabled;
 
-#define ARCH_DLINFO						\
-do if (vdso_enabled) {						\
-		NEW_AUX_ENT(AT_SYSINFO,	VDSO_ENTRY);		\
-		NEW_AUX_ENT(AT_SYSINFO_EHDR, VDSO_COMPAT_BASE);	\
+#define ARCH_DLINFO							\
+do if (vdso_enabled) {							\
+		NEW_AUX_ENT(AT_SYSINFO,	VDSO_ENTRY);			\
+		NEW_AUX_ENT(AT_SYSINFO_EHDR, VDSO_CURRENT_BASE);	\
 } while (0)
 
 #endif
diff -r 97a792fb6127 include/asm-i386/fixmap.h
--- a/include/asm-i386/fixmap.h	Wed Apr 04 23:26:37 2007 -0700
+++ b/include/asm-i386/fixmap.h	Thu Apr 05 01:11:25 2007 -0700
@@ -19,13 +19,9 @@
  * Leave one empty page between vmalloc'ed areas and
  * the start of the fixmap.
  */
-#ifndef CONFIG_COMPAT_VDSO
 extern unsigned long __FIXADDR_TOP;
-#else
-#define __FIXADDR_TOP  0xfffff000
-#define FIXADDR_USER_START	__fix_to_virt(FIX_VDSO)
-#define FIXADDR_USER_END	__fix_to_virt(FIX_VDSO - 1)
-#endif
+#define FIXADDR_USER_START     __fix_to_virt(FIX_VDSO)
+#define FIXADDR_USER_END       __fix_to_virt(FIX_VDSO - 1)
 
 #ifndef __ASSEMBLY__
 #include <linux/kernel.h>
diff -r 97a792fb6127 include/linux/elf.h
--- a/include/linux/elf.h	Wed Apr 04 23:26:37 2007 -0700
+++ b/include/linux/elf.h	Thu Apr 05 01:11:25 2007 -0700
@@ -83,6 +83,12 @@ typedef __s64	Elf64_Sxword;
 #define DT_DEBUG	21
 #define DT_TEXTREL	22
 #define DT_JMPREL	23
+#define DT_ENCODING	32
+#define DT_LOOS		0x6000000D
+#define DT_HIOS		0x6ffff000
+#define DT_VERSYM	0x6ffffff0
+#define DT_VERDEF	0x6ffffffc
+#define DT_VERNEED	0x6ffffffe
 #define DT_LOPROC	0x70000000
 #define DT_HIPROC	0x7fffffff
 

Re: [patch 1/2] Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

[Jan Beulich]

>+	for(; dyn->d_tag != DT_NULL; dyn++)
>+		switch(dyn->d_tag) {
>+		case DT_PLTGOT:
>+		case DT_HASH:
>+		case DT_STRTAB:
>+		case DT_SYMTAB:
>+		case DT_RELA:
>+		case DT_INIT:
>+		case DT_FINI:
>+		case DT_REL:
>+		case DT_DEBUG:
>+		case DT_JMPREL:
>+		case DT_VERSYM:
>+		case DT_VERDEF:
>+		case DT_VERNEED:
>+		case DT_ENCODING ... DT_HIOS:
>+			/* tags above DT_ENCODING are even if they're
>+			   a pointer, so skip odd ones */
>+			if (dyn->d_tag >= DT_ENCODING &&
>+			    (dyn->d_tag & 1) == 1)
>+				break;
>+
>+			dyn->d_un.d_ptr += VDSO_HIGH_BASE;
>+		}

I'm pretty certain the range OLD_DT_LOOS ... DT_LOOS must be excluded
here (the document version I'm looking at is inconsistent in itself here, saying
in one place to stop at DT_LOOS, in a second to stop at DT_HIOS, and in a
third to include DT_LOOS ... ST_HIOS - the inconsistency goes away if
assuming that the stop at DT_LOOS really means stop at OLD_DT_LOOS,
and the stop at DT_HIOS misses to special-case the OLD_DT_LOOS ...
DT_LOOS range.
Additionally I'm a little worried about excluding DT_ADDRRNGLO ...
DT_ADDRRNGHI in case future binutils ever start defaulting to generate
any of these (namely DT_GNU_HASH).

>+#define DT_ENCODING	32

Hmm, I was about to say this ought to be 31 when I realized the discrepancy
between document and binutils. I'll have to ask about this...

Jan

Re: [patch 1/2] Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

[Roland McGrath]

Don't change sh_addr in sections with (sh_flags & SHF_ALLOC) == 0.  Also,
I'd concur with Jan's suggestion about paranoia warnings for unexpected
DT_* types (and for unexpected st_shndx >= SHN_LORESERVE).

Otherwise, looks ok to me.


Thanks,
Roland

[patch 2/2] Make COMPAT_VDSO runtime selectable.

[Jeremy Fitzhardinge]

[-- Attachment #1: runtime-COMPAT_VDSO-selection.patch --]
[-- Type: text/plain, Size: 7163 bytes --]

Now that relocation of the VDSO for COMPAT_VDSO users is done at
runtime rather than compile time, it is possible to enable/disable
compat mode at runtime.

This patch allows you to enable COMPAT_VDSO mode with "vdso=2" on the
kernel command line, or via sysctl.

The COMPAT_VDSO config option still exists, but if enabled it just
makes vdso_enabled default to VDSO_COMPAT.

Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Zachary Amsden <zach@vmware.com>
Cc: "Jan Beulich" <JBeulich@novell.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Roland McGrath <roland@redhat.com>

---
 Documentation/kernel-parameters.txt |    1 
 arch/i386/kernel/sysenter.c         |  131 ++++++++++++++++++++++-------------
 include/asm-i386/page.h             |    2 
 3 files changed, 84 insertions(+), 50 deletions(-)

===================================================================
--- a/Documentation/kernel-parameters.txt
+++ b/Documentation/kernel-parameters.txt
@@ -1807,6 +1807,7 @@ and is between 256 and 4096 characters. 
 			[USBHID] The interval which mice are to be polled at.
 
 	vdso=		[IA-32,SH]
+			vdso=2: enable compat VDSO (default with COMPAT_VDSO)
 			vdso=1: enable VDSO (default)
 			vdso=0: disable VDSO mapping
 
===================================================================
--- a/arch/i386/kernel/sysenter.c
+++ b/arch/i386/kernel/sysenter.c
@@ -24,11 +24,23 @@
 #include <asm/unistd.h>
 #include <asm/elf.h>
 
+enum {
+	VDSO_DISABLED = 0,
+	VDSO_ENABLED = 1,
+	VDSO_COMPAT = 2,
+};
+
+#ifdef CONFIG_COMPAT_VDSO
+#define VDSO_DEFAULT	VDSO_COMPAT
+#else
+#define VDSO_DEFAULT	VDSO_ENABLED
+#endif
+
 /*
  * Should the kernel map a VDSO page into processes and pass its
  * address down to glibc upon exec()?
  */
-unsigned int __read_mostly vdso_enabled = 1;
+unsigned int __read_mostly vdso_enabled = VDSO_DEFAULT;
 
 EXPORT_SYMBOL_GPL(vdso_enabled);
 
@@ -43,7 +55,6 @@ __setup("vdso=", vdso_setup);
 
 extern asmlinkage void sysenter_entry(void);
 
-#ifdef CONFIG_COMPAT_VDSO
 static __cpuinit void reloc_dyn(Elf32_Ehdr *ehdr, unsigned offset)
 {
 	Elf32_Dyn *dyn = (void *)ehdr + offset;
@@ -85,11 +96,6 @@ static __cpuinit void relocate_vdso(Elf3
 			reloc_dyn(ehdr, phdr[i].p_offset);
 	}
 }
-#else
-static inline void relocate_vdso(Elf32_Ehdr *ehdr)
-{
-}
-#endif	/* COMPAT_VDSO */
 
 void enable_sep_cpu(void)
 {
@@ -109,6 +115,25 @@ void enable_sep_cpu(void)
 	put_cpu();	
 }
 
+static struct vm_area_struct gate_vma;
+
+static int __cpuinit gate_vma_init(void)
+{
+	gate_vma.vm_mm = NULL;
+	gate_vma.vm_start = FIXADDR_USER_START;
+	gate_vma.vm_end = FIXADDR_USER_END;
+	gate_vma.vm_flags = VM_READ | VM_MAYREAD | VM_EXEC | VM_MAYEXEC;
+	gate_vma.vm_page_prot = __P101;
+	/*
+	 * Make sure the vDSO gets into every core dump.
+	 * Dumping its contents makes post-mortem fully interpretable later
+	 * without matching up the same kernel and hardware config to see
+	 * what PC values meant.
+	 */
+	gate_vma.vm_flags |= VM_ALWAYSDUMP;
+	return 0;
+}
+
 /*
  * These symbols are defined by vsyscall.o to mark the bounds
  * of the ELF DSO images included therein.
@@ -117,6 +142,19 @@ extern const char vsyscall_sysenter_star
 extern const char vsyscall_sysenter_start, vsyscall_sysenter_end;
 static struct page *syscall_pages[1];
 
+static void map_compat_vdso(int map)
+{
+	static int vdso_mapped;
+
+	if (map == vdso_mapped)
+		return;
+
+	vdso_mapped = map;
+
+	__set_fixmap(FIX_VDSO, page_to_pfn(syscall_pages[0]) << PAGE_SHIFT,
+		     map ? PAGE_READONLY_EXEC : PAGE_NONE);
+}
+
 int __cpuinit sysenter_setup(void)
 {
 	void *syscall_page = (void *)get_zeroed_page(GFP_ATOMIC);
@@ -125,10 +163,9 @@ int __cpuinit sysenter_setup(void)
 
 	syscall_pages[0] = virt_to_page(syscall_page);
 
-#ifdef CONFIG_COMPAT_VDSO
-	__set_fixmap(FIX_VDSO, __pa(syscall_page), PAGE_READONLY_EXEC);
+	gate_vma_init();
+
 	printk("Compat vDSO mapped to %08lx.\n", __fix_to_virt(FIX_VDSO));
-#endif
 
 	if (!boot_cpu_has(X86_FEATURE_SEP)) {
 		vsyscall = &vsyscall_int80_start;
@@ -147,7 +184,6 @@ int __cpuinit sysenter_setup(void)
 /* Defined in vsyscall-sysenter.S */
 extern void SYSENTER_RETURN;
 
-#ifdef __HAVE_ARCH_GATE_AREA
 /* Setup a VMA at program startup for the vsyscall page */
 int arch_setup_additional_pages(struct linux_binprm *bprm, int exstack)
 {
@@ -156,33 +192,44 @@ int arch_setup_additional_pages(struct l
 	int ret;
 
 	down_write(&mm->mmap_sem);
-	addr = get_unmapped_area(NULL, 0, PAGE_SIZE, 0, 0);
-	if (IS_ERR_VALUE(addr)) {
-		ret = addr;
-		goto up_fail;
-	}
-
-	/*
-	 * MAYWRITE to allow gdb to COW and set breakpoints
-	 *
-	 * Make sure the vDSO gets into every core dump.
-	 * Dumping its contents makes post-mortem fully interpretable later
-	 * without matching up the same kernel and hardware config to see
-	 * what PC values meant.
-	 */
-	ret = install_special_mapping(mm, addr, PAGE_SIZE,
-				      VM_READ|VM_EXEC|
-				      VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC|
-				      VM_ALWAYSDUMP,
-				      syscall_pages);
-	if (ret)
-		goto up_fail;
+
+	map_compat_vdso(vdso_enabled == VDSO_COMPAT);
+
+	if (vdso_enabled == VDSO_COMPAT)
+		addr = VDSO_HIGH_BASE;
+	else {
+		addr = get_unmapped_area(NULL, 0, PAGE_SIZE, 0, 0);
+		if (IS_ERR_VALUE(addr)) {
+			ret = addr;
+			goto up_fail;
+		}
+
+		/*
+		 * MAYWRITE to allow gdb to COW and set breakpoints
+		 *
+		 * Make sure the vDSO gets into every core dump.
+		 * Dumping its contents makes post-mortem fully
+		 * interpretable later without matching up the same
+		 * kernel and hardware config to see what PC values
+		 * meant.
+		 */
+		ret = install_special_mapping(mm, addr, PAGE_SIZE,
+					      VM_READ|VM_EXEC|
+					      VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC|
+					      VM_ALWAYSDUMP,
+					      syscall_pages);
+
+		if (ret)
+			goto up_fail;
+	}
 
 	current->mm->context.vdso = (void *)addr;
 	current_thread_info()->sysenter_return =
-				    (void *)VDSO_SYM(&SYSENTER_RETURN);
-up_fail:
+		(void *)VDSO_SYM(&SYSENTER_RETURN);
+
+  up_fail:
 	up_write(&mm->mmap_sem);
+
 	return ret;
 }
 
@@ -195,6 +242,8 @@ const char *arch_vma_name(struct vm_area
 
 struct vm_area_struct *get_gate_vma(struct task_struct *tsk)
 {
+	if (vdso_enabled == VDSO_COMPAT)
+		return &gate_vma;
 	return NULL;
 }
 
@@ -207,17 +256,3 @@ int in_gate_area_no_task(unsigned long a
 {
 	return 0;
 }
-#else  /* !__HAVE_ARCH_GATE_AREA */
-int arch_setup_additional_pages(struct linux_binprm *bprm, int exstack)
-{
-	/*
-	 * If not creating userspace VMA, simply set vdso to point to
-	 * fixmap page.
-	 */
-	current->mm->context.vdso = (void *)VDSO_HIGH_BASE;
-	current_thread_info()->sysenter_return =
-		(void *)VDSO_SYM(&SYSENTER_RETURN);
-
-	return 0;
-}
-#endif	/* __HAVE_ARCH_GATE_AREA */
===================================================================
--- a/include/asm-i386/page.h
+++ b/include/asm-i386/page.h
@@ -143,9 +143,7 @@ extern int page_is_ram(unsigned long pag
 #include <asm-generic/memory_model.h>
 #include <asm-generic/page.h>
 
-#ifndef CONFIG_COMPAT_VDSO
 #define __HAVE_ARCH_GATE_AREA 1
-#endif
 #endif /* __KERNEL__ */
 
 #endif /* _I386_PAGE_H */

-- 

[patch 0/2] Updates to compat VDSOs

[Jeremy Fitzhardinge]

Hi Andi,

Here's a couple of patches to fix up COMPAT_VDSO:

The first is a straightforward implementation of Jan's original idea
of relocating the VDSO to match its mapped location.  Unlike Jan and
Zach's version, I changed it to relocate based on the phdrs rather than
the sections; the result is pleasantly compact.

The second patch takes advantage of the fact that all the COMPAT_VDSO work
happens at runtime now, and allows compat mode to be enabled dynamically.
If you specify vdso=2 on the kernel command line, it comes up in compat
mode; vdso=1 is normal vdso mode, and vdso=0 disables vdso altogether.
You can also switch modes with sysctl.

Changes since last posting:
 - rebase sections as well as phdrs
 - relocate symbols
 - more robust DT_tag handling
 - handle dynamic compat mode switches a bit better

Thanks to Jan and Roland for the review.

Thanks,
	J

-- 

[patch 1/2] Relocate VDSO ELF headers to match mapped location with COMPAT_VDSO

[Jeremy Fitzhardinge]

[-- Attachment #1: relocate-COMPAT_VDSO.patch --]
[-- Type: text/plain, Size: 10762 bytes --]

Some versions of libc can't deal with a VDSO which doesn't have its
ELF headers matching its mapped address.  COMPAT_VDSO maps the VDSO at
a specific system-wide fixed address.  Previously this was all done at
build time, on the grounds that the fixed VDSO address is always at
the top of the address space.  However, a hypervisor may reserve some
of that address space, pushing the fixmap address down.

This patch does the adjustment dynamically at runtime, depending on
the runtime location of the VDSO fixmap.

[ Patch has been through several hands: Jan Beulich wrote the orignal
  version; Zach reworked it, and Jeremy converted it to relocate phdrs
  as well as sections. ]

Signed-off-by: Jeremy Fitzhardinge <jeremy@xensource.com>
Cc: Zachary Amsden <zach@vmware.com>
Cc: "Jan Beulich" <JBeulich@novell.com>
Cc: Eric W. Biederman <ebiederm@xmission.com>
Cc: Andi Kleen <ak@suse.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Roland McGrath <roland@redhat.com>

---
 arch/i386/kernel/entry.S    |    4 -
 arch/i386/kernel/sysenter.c |  156 ++++++++++++++++++++++++++++++++++++++-----
 arch/i386/mm/pgtable.c      |    6 -
 include/asm-i386/elf.h      |   28 ++-----
 include/asm-i386/fixmap.h   |    8 --
 include/linux/elf.h         |   10 ++
 6 files changed, 163 insertions(+), 49 deletions(-)

===================================================================
--- a/arch/i386/kernel/entry.S
+++ b/arch/i386/kernel/entry.S
@@ -305,16 +305,12 @@ sysenter_past_esp:
 	pushl $(__USER_CS)
 	CFI_ADJUST_CFA_OFFSET 4
 	/*CFI_REL_OFFSET cs, 0*/
-#ifndef CONFIG_COMPAT_VDSO
 	/*
 	 * Push current_thread_info()->sysenter_return to the stack.
 	 * A tiny bit of offset fixup is necessary - 4*4 means the 4 words
 	 * pushed above; +8 corresponds to copy_thread's esp0 setting.
 	 */
 	pushl (TI_sysenter_return-THREAD_SIZE+8+4*4)(%esp)
-#else
-	pushl $SYSENTER_RETURN
-#endif
 	CFI_ADJUST_CFA_OFFSET 4
 	CFI_REL_OFFSET eip, 0
 
===================================================================
--- a/arch/i386/kernel/sysenter.c
+++ b/arch/i386/kernel/sysenter.c
@@ -22,6 +22,7 @@
 #include <asm/msr.h>
 #include <asm/pgtable.h>
 #include <asm/unistd.h>
+#include <asm/elf.h>
 
 /*
  * Should the kernel map a VDSO page into processes and pass its
@@ -41,6 +42,115 @@ __setup("vdso=", vdso_setup);
 __setup("vdso=", vdso_setup);
 
 extern asmlinkage void sysenter_entry(void);
+
+#ifdef CONFIG_COMPAT_VDSO
+static __cpuinit void reloc_symtab(Elf32_Ehdr *ehdr,
+				   unsigned offset, unsigned size)
+{
+	Elf32_Sym *sym = (void *)ehdr + offset;
+	unsigned nsym = size / sizeof(*sym);
+	unsigned i;
+
+	for(i = 0; i < nsym; i++, sym++) {
+		if (sym->st_shndx == SHN_UNDEF ||
+		    sym->st_shndx == SHN_ABS)
+			continue;  /* skip */
+
+		if (sym->st_shndx > SHN_LORESERVE) {
+			printk(KERN_INFO "VDSO: unexpected st_shndx %x\n",
+			       sym->st_shndx);
+			continue;
+		}
+
+		switch(ELF_ST_TYPE(sym->st_info)) {
+		case STT_OBJECT:
+		case STT_FUNC:
+		case STT_SECTION:
+		case STT_FILE:
+			sym->st_value += VDSO_HIGH_BASE;
+		}
+	}
+}
+
+static __cpuinit void reloc_dyn(Elf32_Ehdr *ehdr, unsigned offset)
+{
+	Elf32_Dyn *dyn = (void *)ehdr + offset;
+
+	for(; dyn->d_tag != DT_NULL; dyn++)
+		switch(dyn->d_tag) {
+		case DT_PLTGOT:
+		case DT_HASH:
+		case DT_STRTAB:
+		case DT_SYMTAB:
+		case DT_RELA:
+		case DT_INIT:
+		case DT_FINI:
+		case DT_REL:
+		case DT_DEBUG:
+		case DT_JMPREL:
+		case DT_VERSYM:
+		case DT_VERDEF:
+		case DT_VERNEED:
+		case DT_ADDRRNGLO ... DT_ADDRRNGHI:
+			dyn->d_un.d_ptr += VDSO_HIGH_BASE;
+			break;
+
+		case DT_ENCODING ... OLD_DT_LOOS:
+		case DT_LOOS ... DT_HIOS:
+			/* Tags above DT_ENCODING are pointers if
+			   they're even */
+			if (dyn->d_tag >= DT_ENCODING &&
+			    (dyn->d_tag & 1) == 0)
+				dyn->d_un.d_ptr += VDSO_HIGH_BASE;
+			break;
+
+		default:
+			printk(KERN_INFO "VDSO: unexpected DT_tag %d\n",
+			       dyn->d_tag);
+		}
+}
+
+static __cpuinit void relocate_vdso(Elf32_Ehdr *ehdr)
+{
+	Elf32_Phdr *phdr;
+	Elf32_Shdr *shdr;
+	int i;
+
+	BUG_ON(memcmp(ehdr->e_ident, ELFMAG, 4) != 0 ||
+	       !elf_check_arch(ehdr) ||
+	       ehdr->e_type != ET_DYN);
+
+	ehdr->e_entry += VDSO_HIGH_BASE;
+
+	/* rebase phdrs */
+	phdr = (void *)ehdr + ehdr->e_phoff;
+	for (i = 0; i < ehdr->e_phnum; i++) {
+		phdr[i].p_vaddr += VDSO_HIGH_BASE;
+
+		/* relocate dynamic stuff */
+		if (phdr[i].p_type == PT_DYNAMIC)
+			reloc_dyn(ehdr, phdr[i].p_offset);
+	}
+
+	/* rebase sections */
+	shdr = (void *)ehdr + ehdr->e_shoff;
+	for(i = 0; i < ehdr->e_shnum; i++) {
+		if (!(shdr[i].sh_flags & SHF_ALLOC))
+			continue;
+
+		shdr[i].sh_addr += VDSO_HIGH_BASE;
+
+		if (shdr[i].sh_type == SHT_SYMTAB ||
+		    shdr[i].sh_type == SHT_DYNSYM)
+			reloc_symtab(ehdr, shdr[i].sh_offset,
+				     shdr[i].sh_size);
+	}
+}
+#else
+static inline void relocate_vdso(Elf32_Ehdr *ehdr)
+{
+}
+#endif	/* COMPAT_VDSO */
 
 void enable_sep_cpu(void)
 {
@@ -71,6 +181,9 @@ int __cpuinit sysenter_setup(void)
 int __cpuinit sysenter_setup(void)
 {
 	void *syscall_page = (void *)get_zeroed_page(GFP_ATOMIC);
+	const void *vsyscall;
+	size_t vsyscall_len;
+
 	syscall_pages[0] = virt_to_page(syscall_page);
 
 #ifdef CONFIG_COMPAT_VDSO
@@ -79,23 +192,23 @@ int __cpuinit sysenter_setup(void)
 #endif
 
 	if (!boot_cpu_has(X86_FEATURE_SEP)) {
-		memcpy(syscall_page,
-		       &vsyscall_int80_start,
-		       &vsyscall_int80_end - &vsyscall_int80_start);
-		return 0;
-	}
-
-	memcpy(syscall_page,
-	       &vsyscall_sysenter_start,
-	       &vsyscall_sysenter_end - &vsyscall_sysenter_start);
-
-	return 0;
-}
-
-#ifndef CONFIG_COMPAT_VDSO
+		vsyscall = &vsyscall_int80_start;
+		vsyscall_len = &vsyscall_int80_end - &vsyscall_int80_start;
+	} else {
+		vsyscall = &vsyscall_sysenter_start;
+		vsyscall_len = &vsyscall_sysenter_end - &vsyscall_sysenter_start;
+	}
+
+	memcpy(syscall_page, vsyscall, vsyscall_len);
+	relocate_vdso(syscall_page);
+
+	return 0;
+}
+
 /* Defined in vsyscall-sysenter.S */
 extern void SYSENTER_RETURN;
 
+#ifdef __HAVE_ARCH_GATE_AREA
 /* Setup a VMA at program startup for the vsyscall page */
 int arch_setup_additional_pages(struct linux_binprm *bprm, int exstack)
 {
@@ -155,4 +268,17 @@ int in_gate_area_no_task(unsigned long a
 {
 	return 0;
 }
-#endif
+#else  /* !__HAVE_ARCH_GATE_AREA */
+int arch_setup_additional_pages(struct linux_binprm *bprm, int exstack)
+{
+	/*
+	 * If not creating userspace VMA, simply set vdso to point to
+	 * fixmap page.
+	 */
+	current->mm->context.vdso = (void *)VDSO_HIGH_BASE;
+	current_thread_info()->sysenter_return =
+		(void *)VDSO_SYM(&SYSENTER_RETURN);
+
+	return 0;
+}
+#endif	/* __HAVE_ARCH_GATE_AREA */
===================================================================
--- a/arch/i386/mm/pgtable.c
+++ b/arch/i386/mm/pgtable.c
@@ -144,10 +144,8 @@ void set_pmd_pfn(unsigned long vaddr, un
 }
 
 static int fixmaps;
-#ifndef CONFIG_COMPAT_VDSO
 unsigned long __FIXADDR_TOP = 0xfffff000;
 EXPORT_SYMBOL(__FIXADDR_TOP);
-#endif
 
 void __set_fixmap (enum fixed_addresses idx, unsigned long phys, pgprot_t flags)
 {
@@ -173,12 +171,8 @@ void reserve_top_address(unsigned long r
 	BUG_ON(fixmaps > 0);
 	printk(KERN_INFO "Reserving virtual address space above 0x%08x\n",
 	       (int)-reserve);
-#ifdef CONFIG_COMPAT_VDSO
-	BUG_ON(reserve != 0);
-#else
 	__FIXADDR_TOP = -reserve - PAGE_SIZE;
 	__VMALLOC_RESERVE += reserve;
-#endif
 }
 
 pte_t *pte_alloc_one_kernel(struct mm_struct *mm, unsigned long address)
===================================================================
--- a/include/asm-i386/elf.h
+++ b/include/asm-i386/elf.h
@@ -133,39 +133,31 @@ extern int dump_task_extended_fpu (struc
 #define ELF_CORE_COPY_XFPREGS(tsk, elf_xfpregs) dump_task_extended_fpu(tsk, elf_xfpregs)
 
 #define VDSO_HIGH_BASE		(__fix_to_virt(FIX_VDSO))
-#define VDSO_BASE		((unsigned long)current->mm->context.vdso)
-
-#ifdef CONFIG_COMPAT_VDSO
-# define VDSO_COMPAT_BASE	VDSO_HIGH_BASE
-# define VDSO_PRELINK		VDSO_HIGH_BASE
-#else
-# define VDSO_COMPAT_BASE	VDSO_BASE
-# define VDSO_PRELINK		0
-#endif
+#define VDSO_CURRENT_BASE	((unsigned long)current->mm->context.vdso)
+#define VDSO_PRELINK		0
 
 #define VDSO_SYM(x) \
-		(VDSO_COMPAT_BASE + (unsigned long)(x) - VDSO_PRELINK)
+		(VDSO_CURRENT_BASE + (unsigned long)(x) - VDSO_PRELINK)
 
 #define VDSO_HIGH_EHDR		((const struct elfhdr *) VDSO_HIGH_BASE)
-#define VDSO_EHDR		((const struct elfhdr *) VDSO_COMPAT_BASE)
+#define VDSO_EHDR		((const struct elfhdr *) VDSO_CURRENT_BASE)
 
 extern void __kernel_vsyscall;
 
 #define VDSO_ENTRY		VDSO_SYM(&__kernel_vsyscall)
 
-#ifndef CONFIG_COMPAT_VDSO
+struct linux_binprm;
+
 #define ARCH_HAS_SETUP_ADDITIONAL_PAGES
-struct linux_binprm;
 extern int arch_setup_additional_pages(struct linux_binprm *bprm,
                                        int executable_stack);
-#endif
 
 extern unsigned int vdso_enabled;
 
-#define ARCH_DLINFO						\
-do if (vdso_enabled) {						\
-		NEW_AUX_ENT(AT_SYSINFO,	VDSO_ENTRY);		\
-		NEW_AUX_ENT(AT_SYSINFO_EHDR, VDSO_COMPAT_BASE);	\
+#define ARCH_DLINFO							\
+do if (vdso_enabled) {							\
+		NEW_AUX_ENT(AT_SYSINFO,	VDSO_ENTRY);			\
+		NEW_AUX_ENT(AT_SYSINFO_EHDR, VDSO_CURRENT_BASE);	\
 } while (0)
 
 #endif
===================================================================
--- a/include/asm-i386/fixmap.h
+++ b/include/asm-i386/fixmap.h
@@ -19,13 +19,9 @@
  * Leave one empty page between vmalloc'ed areas and
  * the start of the fixmap.
  */
-#ifndef CONFIG_COMPAT_VDSO
 extern unsigned long __FIXADDR_TOP;
-#else
-#define __FIXADDR_TOP  0xfffff000
-#define FIXADDR_USER_START	__fix_to_virt(FIX_VDSO)
-#define FIXADDR_USER_END	__fix_to_virt(FIX_VDSO - 1)
-#endif
+#define FIXADDR_USER_START     __fix_to_virt(FIX_VDSO)
+#define FIXADDR_USER_END       __fix_to_virt(FIX_VDSO - 1)
 
 #ifndef __ASSEMBLY__
 #include <linux/kernel.h>
===================================================================
--- a/include/linux/elf.h
+++ b/include/linux/elf.h
@@ -83,6 +83,16 @@ typedef __s64	Elf64_Sxword;
 #define DT_DEBUG	21
 #define DT_TEXTREL	22
 #define DT_JMPREL	23
+#define DT_ENCODING	32
+#define OLD_DT_LOOS	0x60000000
+#define DT_LOOS		0x6000000d
+#define DT_HIOS		0x6ffff000
+#define DT_ADDRRNGLO	0x6ffffe00
+#define DT_ADDRRNGHI	0x6ffffeff
+#define DT_VERSYM	0x6ffffff0
+#define DT_VERDEF	0x6ffffffc
+#define DT_VERNEED	0x6ffffffe
+#define OLD_DT_HIOS     0x6fffffff
 #define DT_LOPROC	0x70000000
 #define DT_HIPROC	0x7fffffff
 

--