MFC r283407:

Implement vdso - virtual dynamic shared object. Through vdso Linux
exposes functions from kernel with proper DWARF CFI information so that
it becomes easier to unwind through them.
Using vdso is a mandatory for a thread cancelation && cleanup
on a modern glibc.

5 files changed, 256 insertions, 30 deletions

diff --git a/sys/amd64/linux32/linux.h b/sys/amd64/linux32/linux.h
index 1c9cb0c..eb0b0c9 100644
--- a/sys/amd64/linux32/linux.h
+++ b/sys/amd64/linux32/linux.h
@@ -114,7 +114,7 @@ typedef struct {
 /*
  * Miscellaneous
  */
-#define	LINUX_AT_COUNT		16	/* Count of used aux entry types.
+#define	LINUX_AT_COUNT		18	/* Count of used aux entry types.
 					 * Keep this synchronized with
 					 * elf_linux_fixup() code.
 					 */
diff --git a/sys/amd64/linux32/linux32_genassym.c b/sys/amd64/linux32/linux32_genassym.c
index bea846d..bc94139 100644
--- a/sys/amd64/linux32/linux32_genassym.c
+++ b/sys/amd64/linux32/linux32_genassym.c
@@ -7,9 +7,12 @@ __FBSDID("$FreeBSD$");
 #include <sys/systm.h>
 
 #include <amd64/linux32/linux.h>
+#include <compat/linux/linux_mib.h>
 
 ASSYM(LINUX_SIGF_HANDLER, offsetof(struct l_sigframe, sf_handler));
 ASSYM(LINUX_SIGF_SC, offsetof(struct l_sigframe, sf_sc));
 ASSYM(LINUX_RT_SIGF_HANDLER, offsetof(struct l_rt_sigframe, sf_handler));
 ASSYM(LINUX_RT_SIGF_UC, offsetof(struct l_rt_sigframe, sf_sc));
 ASSYM(LINUX_RT_SIGF_SC, offsetof(struct l_ucontext, uc_mcontext));
+ASSYM(LINUX_VERSION_CODE, LINUX_VERSION_CODE);
+ASSYM(LINUX_SC_ESP, offsetof(struct l_sigcontext, sc_esp));
diff --git a/sys/amd64/linux32/linux32_locore.s b/sys/amd64/linux32/linux32_locore.s
index 36e1abf..8a3b83e 100644
--- a/sys/amd64/linux32/linux32_locore.s
+++ b/sys/amd64/linux32/linux32_locore.s
@@ -8,31 +8,143 @@
 .text
 .code32
 
-NON_GPROF_ENTRY(linux_sigcode)
-	call	*LINUX_SIGF_HANDLER(%esp)
-	leal	LINUX_SIGF_SC(%esp),%ebx	/* linux scp */
-	movl	%esp, %ebx			/* pass sigframe */
-	push	%eax				/* fake ret addr */
+/*
+ * To avoid excess stack frame the signal trampoline code emulates
+ * the 'call' instruction.
+ */
+NON_GPROF_ENTRY(linux32_sigcode)
+	movl	%esp, %ebx			/* preserve sigframe */
+	call .getip0
+.getip0:
+	popl	%eax
+	add	$.startsigcode-.getip0, %eax	/* ret address */
+	push	%eax
+	jmp	*LINUX_SIGF_HANDLER(%ebx)
+.startsigcode:
+	popl	%eax
 	movl	$LINUX_SYS_linux_sigreturn,%eax	/* linux_sigreturn() */
 	int	$0x80				/* enter kernel with args */
+.endsigcode:
 0:	jmp	0b
-	ALIGN_TEXT
-/* XXXXX */
-linux_rt_sigcode:
-	call	*LINUX_RT_SIGF_HANDLER(%esp)
+
+NON_GPROF_ENTRY(linux32_rt_sigcode)
 	leal	LINUX_RT_SIGF_UC(%esp),%ebx	/* linux ucp */
 	leal	LINUX_RT_SIGF_SC(%ebx),%ecx	/* linux sigcontext */
-	push	%eax				/* fake ret addr */
+	movl	%esp, %edi
+	call	.getip1
+.getip1:
+	popl	%eax
+	add	$.startrtsigcode-.getip1, %eax	/* ret address */
+	push	%eax
+	jmp	*LINUX_RT_SIGF_HANDLER(%edi)
+.startrtsigcode:
 	movl	$LINUX_SYS_linux_rt_sigreturn,%eax   /* linux_rt_sigreturn() */
 	int	$0x80				/* enter kernel with args */
+.endrtsigcode:
 0:	jmp	0b
-	ALIGN_TEXT
-/* XXXXX */
-linux_esigcode:
-
-	.data
-	.globl	linux_szsigcode, linux_sznonrtsigcode
-linux_szsigcode:
-	.long	linux_esigcode-linux_sigcode
-linux_sznonrtsigcode:
-	.long	linux_rt_sigcode-linux_sigcode
+
+NON_GPROF_ENTRY(linux32_vsyscall)
+.startvsyscall:
+	int $0x80
+	ret
+.endvsyscall:
+
+
+	.section .note.Linux, "a",@note
+	.long 2f - 1f		/* namesz */
+	.balign 4
+	.long 4f - 3f		/* descsz */
+	.long 0
+1:
+	.asciz "Linux"
+2:
+	.balign 4
+3:
+	.long LINUX_VERSION_CODE
+4:
+	.balign 4
+	.previous
+
+
+#define do_cfa_expr(offset)                                             \
+	.byte 0x0f;			/* DW_CFA_def_cfa_expression */ \
+	.uleb128 11f-10f;		/*   length */                  \
+10:	.byte 0x74;			/*     DW_OP_breg4 */           \
+	.sleb128 offset;		/*      offset */               \
+	.byte 0x06;			/*     DW_OP_deref */           \
+11:
+
+
+	/* CIE */
+	.section .eh_frame,"a",@progbits
+.LSTARTFRAMEDLSI1:
+	.long .LENDCIEDLSI1-.LSTARTCIEDLSI1
+.LSTARTCIEDLSI1:
+	.long 0					/* CIE ID */
+	.byte 1					/* Version number */
+	.string "zRS"				/* NULL-terminated
+						 * augmentation string
+						 */
+	.uleb128 1				/* Code alignment factor */
+	.sleb128 -4				/* Data alignment factor */
+	.byte 8					/* Return address
+						 * register column
+						 */
+	.uleb128 1				/* Augmentation value length */
+	.byte 0x1b				/* DW_EH_PE_pcrel|DW_EH_PE_sdata4. */
+	.byte 0					/* DW_CFA_nop */
+	.align 4
+.LENDCIEDLSI1:
+
+	/* FDE */
+	.long .LENDFDEDLSI1-.LSTARTFDEDLSI1	/* Length FDE */
+.LSTARTFDEDLSI1:
+	.long .LSTARTFDEDLSI1-.LSTARTFRAMEDLSI1 /* CIE pointer */
+	.long .startsigcode-.			/* PC-relative start address */
+	.long .endsigcode-.startsigcode
+	.uleb128 0				/* Augmentation */
+	do_cfa_expr(LINUX_SIGF_SC-8)
+	.align 4
+.LENDFDEDLSI1:
+
+	.long .LENDFDEDLSI2-.LSTARTFDEDLSI2	/* Length FDE */
+.LSTARTFDEDLSI2:
+	.long .LSTARTFDEDLSI2-.LSTARTFRAMEDLSI1	/* CIE pointer */
+	.long .startrtsigcode-.			/* PC-relative start address */
+	.long .endrtsigcode-.startrtsigcode
+	.uleb128 0				/* Augmentation */
+	do_cfa_expr(LINUX_RT_SIGF_SC-4+LINUX_SC_ESP)
+	.align 4
+.LENDFDEDLSI2:
+	.previous
+
+	.section .eh_frame,"a",@progbits
+.LSTARTFRAMEDLSI2:
+	.long .LENDCIEDLSI2-.LSTARTCIEDLSI2
+.LSTARTCIEDLSI2:
+	.long 0					/* CIE ID */
+	.byte 1					/* Version number */
+	.string "zR"				/* NULL-terminated
+						 * augmentation string
+						 */
+	.uleb128 1				/* Code alignment factor */
+	.sleb128 -4				/* Data alignment factor */
+	.byte 8					/* Return address register column */
+	.uleb128 1				/* Augmentation value length */
+	.byte 0x1b				/* DW_EH_PE_pcrel|DW_EH_PE_sdata4. */
+	.byte 0x0c				/* DW_CFA_def_cfa */
+	.uleb128 4
+	.uleb128 4
+	.byte 0x88				/* DW_CFA_offset, column 0x8 */
+	.uleb128 1
+	.align 4
+.LENDCIEDLSI2:
+	.long .LENDFDEDLSI3-.LSTARTFDEDLSI3 /* Length FDE */
+.LSTARTFDEDLSI3:
+	.long .LSTARTFDEDLSI3-.LSTARTFRAMEDLSI2 /* CIE pointer */
+	.long .startvsyscall-.			/* PC-relative start address */
+	.long .endvsyscall-.startvsyscall
+	.uleb128 0
+	.align 4
+.LENDFDEDLSI3:
+	.previous
diff --git a/sys/amd64/linux32/linux32_sysvec.c b/sys/amd64/linux32/linux32_sysvec.c
index 9df5b0e..5ceae58 100644
--- a/sys/amd64/linux32/linux32_sysvec.c
+++ b/sys/amd64/linux32/linux32_sysvec.c
@@ -83,6 +83,7 @@ __FBSDID("$FreeBSD$");
 #include <compat/linux/linux_misc.h>
 #include <compat/linux/linux_signal.h>
 #include <compat/linux/linux_util.h>
+#include <compat/linux/linux_vdso.h>
 
 MODULE_VERSION(linux, 1);
 
@@ -111,8 +112,11 @@ MALLOC_DEFINE(M_LINUX, "linux", "Linux mode structures");
 
 const char *linux_platform = "i686";
 static int linux_szplatform;
-extern char linux_sigcode[];
-extern int linux_szsigcode;
+static int linux_szsigcode;
+static vm_object_t linux_shared_page_obj;
+static char *linux_shared_page_mapping;
+extern char _binary_linux32_locore_o_start;
+extern char _binary_linux32_locore_o_end;
 
 extern struct sysent linux_sysent[LINUX_SYS_MAXSYSCALL];
 
@@ -127,6 +131,8 @@ static void	exec_linux_setregs(struct thread *td,
 				   struct image_params *imgp, u_long stack);
 static void	linux32_fixlimit(struct rlimit *rl, int which);
 static boolean_t linux32_trans_osrel(const Elf_Note *note, int32_t *osrel);
+static void	linux_vdso_install(void *param);
+static void	linux_vdso_deinstall(void *param);
 
 static eventhandler_tag linux_exit_tag;
 static eventhandler_tag linux_exec_tag;
@@ -220,6 +226,10 @@ struct linux32_ps_strings {
 	u_int ps_nenvstr;	/* the number of environment strings */
 };
 
+LINUX_VDSO_SYM_INTPTR(linux32_sigcode);
+LINUX_VDSO_SYM_INTPTR(linux32_rt_sigcode);
+LINUX_VDSO_SYM_INTPTR(linux32_vsyscall);
+
 /*
  * If FreeBSD & Linux have a difference of opinion about what a trap
  * means, deal with it here.
@@ -259,6 +269,9 @@ elf_linux_fixup(register_t **stack_base, struct image_params *imgp)
 	args = (Elf32_Auxargs *)imgp->auxargs;
 	pos = base + (imgp->args->argc + imgp->args->envc + 2);
 
+	AUXARGS_ENTRY_32(pos, LINUX_AT_SYSINFO_EHDR,
+	    imgp->proc->p_sysent->sv_shared_page_base);
+	AUXARGS_ENTRY_32(pos, LINUX_AT_SYSINFO, linux32_vsyscall);
 	AUXARGS_ENTRY_32(pos, LINUX_AT_HWCAP, cpu_feature);
 
 	/*
@@ -297,8 +310,6 @@ elf_linux_fixup(register_t **stack_base, struct image_params *imgp)
 	return (0);
 }
 
-extern unsigned long linux_sznonrtsigcode;
-
 static void
 linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
 {
@@ -353,7 +364,8 @@ linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
 	ksiginfo_to_lsiginfo(ksi, &frame.sf_si, sig);
 
 	/*
-	 * Build the signal context to be used by sigreturn.
+	 * Build the signal context to be used by sigreturn
+	 * and libgcc unwind.
 	 */
 	frame.sf_sc.uc_flags = 0;		/* XXX ??? */
 	frame.sf_sc.uc_link = 0;		/* XXX ??? */
@@ -371,6 +383,7 @@ linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
 	frame.sf_sc.uc_mcontext.sc_esi    = regs->tf_rsi;
 	frame.sf_sc.uc_mcontext.sc_ebp    = regs->tf_rbp;
 	frame.sf_sc.uc_mcontext.sc_ebx    = regs->tf_rbx;
+	frame.sf_sc.uc_mcontext.sc_esp    = regs->tf_rsp;
 	frame.sf_sc.uc_mcontext.sc_edx    = regs->tf_rdx;
 	frame.sf_sc.uc_mcontext.sc_ecx    = regs->tf_rcx;
 	frame.sf_sc.uc_mcontext.sc_eax    = regs->tf_rax;
@@ -412,7 +425,7 @@ linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
 	 * Build context to run handler in.
 	 */
 	regs->tf_rsp = PTROUT(fp);
-	regs->tf_rip = p->p_sysent->sv_sigcode_base + linux_sznonrtsigcode;
+	regs->tf_rip = linux32_rt_sigcode;
 	regs->tf_rflags &= ~(PSL_T | PSL_D);
 	regs->tf_cs = _ucode32sel;
 	regs->tf_ss = _udatasel;
@@ -507,6 +520,7 @@ linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
 	frame.sf_sc.sc_esi    = regs->tf_rsi;
 	frame.sf_sc.sc_ebp    = regs->tf_rbp;
 	frame.sf_sc.sc_ebx    = regs->tf_rbx;
+	frame.sf_sc.sc_esp    = regs->tf_rsp;
 	frame.sf_sc.sc_edx    = regs->tf_rdx;
 	frame.sf_sc.sc_ecx    = regs->tf_rcx;
 	frame.sf_sc.sc_eax    = regs->tf_rax;
@@ -535,7 +549,7 @@ linux_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
 	 * Build context to run handler in.
 	 */
 	regs->tf_rsp = PTROUT(fp);
-	regs->tf_rip = p->p_sysent->sv_sigcode_base;
+	regs->tf_rip = linux32_sigcode;
 	regs->tf_rflags &= ~(PSL_T | PSL_D);
 	regs->tf_cs = _ucode32sel;
 	regs->tf_ss = _udatasel;
@@ -1014,7 +1028,7 @@ struct sysentvec elf_linux_sysvec = {
 	.sv_transtrap	= translate_traps,
 	.sv_fixup	= elf_linux_fixup,
 	.sv_sendsig	= linux_sendsig,
-	.sv_sigcode	= linux_sigcode,
+	.sv_sigcode	= &_binary_linux32_locore_o_start,
 	.sv_szsigcode	= &linux_szsigcode,
 	.sv_prepsyscall	= NULL,
 	.sv_name	= "Linux ELF32",
@@ -1040,7 +1054,39 @@ struct sysentvec elf_linux_sysvec = {
 	.sv_schedtail	= linux_schedtail,
 	.sv_thread_detach = linux_thread_detach,
 };
-INIT_SYSENTVEC(elf_sysvec, &elf_linux_sysvec);
+
+static void
+linux_vdso_install(void *param)
+{
+
+	linux_szsigcode = (&_binary_linux32_locore_o_end - 
+	    &_binary_linux32_locore_o_start);
+
+	if (linux_szsigcode > elf_linux_sysvec.sv_shared_page_len)
+		panic("Linux invalid vdso size\n");
+
+	__elfN(linux_vdso_fixup)(&elf_linux_sysvec);
+
+	linux_shared_page_obj = __elfN(linux_shared_page_init)
+	    (&linux_shared_page_mapping);
+
+	__elfN(linux_vdso_reloc)(&elf_linux_sysvec, LINUX32_SHAREDPAGE);
+
+	bcopy(elf_linux_sysvec.sv_sigcode, linux_shared_page_mapping,
+	    linux_szsigcode);
+	elf_linux_sysvec.sv_shared_page_obj = linux_shared_page_obj;
+}
+SYSINIT(elf_linux_vdso_init, SI_SUB_EXEC, SI_ORDER_ANY,
+    (sysinit_cfunc_t)linux_vdso_install, NULL);
+
+static void
+linux_vdso_deinstall(void *param)
+{
+
+	__elfN(linux_shared_page_fini)(linux_shared_page_obj);
+};
+SYSUNINIT(elf_linux_vdso_uninit, SI_SUB_EXEC, SI_ORDER_FIRST,
+    (sysinit_cfunc_t)linux_vdso_deinstall, NULL);
 
 static char GNU_ABI_VENDOR[] = "GNU";
 static int GNULINUX_ABI_DESC = 0;
diff --git a/sys/amd64/linux32/linux32_vdso.lds.s b/sys/amd64/linux32/linux32_vdso.lds.s
new file mode 100644
index 0000000..beaaef5
--- /dev/null
+++ b/sys/amd64/linux32/linux32_vdso.lds.s
@@ -0,0 +1,65 @@
+/*
+ * Linker script for 32-bit vDSO.
+ * Copied from Linux kernel arch/x86/vdso/vdso-layout.lds.S
+ * and arch/x86/vdso/vdso32/vdso32.lds.S
+ *
+ * $FreeBSD$
+ */
+
+SECTIONS
+{
+	. = . + 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
+	.data		: {
+	      *(.data*)
+	      *(.sdata*)
+	      *(.got.plt) *(.got)
+	      *(.gnu.linkonce.d.*)
+	      *(.bss*)
+	      *(.dynbss*)
+	      *(.gnu.linkonce.b.*)
+	}
+
+	.altinstructions	: { *(.altinstructions) }
+	.altinstr_replacement	: { *(.altinstr_replacement) }
+
+	. = ALIGN(0x100);
+	.text		: { *(.text*) }			:text	=0x90909090
+}
+
+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;
+}
+
+ENTRY(linux32_vsyscall);
+
+VERSION
+{
+	LINUX_2.5 {
+	global:
+		linux32_vsyscall;
+		linux32_sigcode;
+		linux32_rt_sigcode;
+	local: *;
+	};
+}