linux-user: Implement safe_syscall for i386sigrace-fixes

Implement the host assembly fragment needed for safe_syscall for
32-bit x86.

Signed-off-by: Peter Maydell <peter.maydell@linaro.org>

2 files changed, 162 insertions, 0 deletions

diff --git a/linux-user/host/i386/hostdep.h b/linux-user/host/i386/hostdep.h
index 7609bf5cd7..ee6a810be5 100644
--- a/linux-user/host/i386/hostdep.h
+++ b/linux-user/host/i386/hostdep.h
@@ -12,4 +12,32 @@
 #ifndef QEMU_HOSTDEP_H
 #define QEMU_HOSTDEP_H
 
+/* We have a safe-syscall.inc.S */
+#define HAVE_SAFE_SYSCALL
+
+#ifndef __ASSEMBLER__
+
+/* These are defined by the safe-syscall.inc.S file */
+extern char safe_syscall_start[];
+extern char safe_syscall_end[];
+
+/* For glibc 2.1 */
+#ifndef REG_EIP
+#define REG_EIP EIP
+#endif
+
+/* Adjust the signal context to rewind out of safe-syscall if we're in it */
+static inline void rewind_if_in_safe_syscall(void *puc)
+{
+    struct ucontext *uc = puc;
+    greg_t *pcreg = &uc->uc_mcontext.gregs[REG_EIP];
+
+    if (*pcreg > (uintptr_t)safe_syscall_start
+        && *pcreg < (uintptr_t)safe_syscall_end) {
+        *pcreg = (uintptr_t)safe_syscall_start;
+    }
+}
+
+#endif /* __ASSEMBLER__ */
+
 #endif
diff --git a/linux-user/host/i386/safe-syscall.inc.S b/linux-user/host/i386/safe-syscall.inc.S
new file mode 100644
index 0000000000..9d98332148
--- /dev/null
+++ b/linux-user/host/i386/safe-syscall.inc.S
@@ -0,0 +1,134 @@
+/*
+ * safe-syscall.inc.S : host-specific assembly fragment
+ * to handle signals occurring at the same time as system calls.
+ * This is intended to be included by linux-user/safe-syscall.S
+ *
+ *  * Written by Peter Maydell <peter.maydell@linaro.org>
+ *
+ * Copyright (C) 2016 Linaro Limited
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ */
+
+    .global safe_syscall_base
+    .global safe_syscall_start
+    .global safe_syscall_end
+    .type   safe_syscall_base, @function
+
+    /* This is the entry point for making a system call. The calling
+     * convention here is that of a C varargs function with the
+     * first argument an 'int *' to the signal_pending flag, the
+     * second one the system call number (as a 'long'), and all further
+     * arguments being syscall arguments (also 'long').
+     * We return a long which is the syscall's return value, which
+     * may be negative-errno on failure. Conversion to the
+     * -1-and-errno-set convention is done by the calling wrapper.
+     */
+safe_syscall_base:
+    .cfi_startproc
+    /* Save ebx esi edi ebp, as the calling convention requires */
+    pushl %ebp
+    .cfi_adjust_cfa_offset 4
+    .cfi_rel_offset ebp, 0
+    pushl %edi
+    .cfi_adjust_cfa_offset 4
+    .cfi_rel_offset edi, 0
+    pushl %esi
+    .cfi_adjust_cfa_offset 4
+    .cfi_rel_offset esi, 0
+    pushl %ebx
+    .cfi_adjust_cfa_offset 4
+    .cfi_rel_offset ebx, 0
+
+    /* The syscall calling convention isn't the same as the
+     * C one:
+     * we enter with all arguments on the stack:
+     *           *signal_pending
+     *           syscall number
+     *           syscall arguments
+     *           and return the result in eax
+     * and the syscall instruction needs
+     *           eax == syscall number
+     *           ebx, ecx, edx, esi, edi, ebp == syscall arguments
+     *           and returns the result in eax
+     * Shuffle everything around appropriately.
+     *
+     * At this point the stack layout is:
+     * 48(%esp)     syscall arg 6
+     * 44(%esp)     syscall arg 5
+     * 40(%esp)     syscall arg 4
+     * 36(%esp)     syscall arg 3
+     * 32(%esp)     syscall arg 2
+     * 28(%esp)     syscall arg 1
+     * 24(%esp)     syscall number
+     * 20(%esp)     *signal_pending
+     * 16(%esp)     return address
+     * 12(%esp)     saved EBP
+     *  8(%esp)     saved EDI
+     *  4(%esp)     saved ESI
+     *   (%esp)     saved EBX
+     */
+    /* syscall arguments */
+    movl 48(%esp), %ebp
+    movl 44(%esp), %edi
+    movl 40(%esp), %esi
+    movl 36(%esp), %edx
+    movl 32(%esp), %ecx
+    movl 28(%esp), %ebx
+
+    /* This next sequence of code works in conjunction with the
+     * rewind_if_safe_syscall_function(). If a signal is taken
+     * and the interrupted PC is anywhere between 'safe_syscall_start'
+     * and 'safe_syscall_end' then we rewind it to 'safe_syscall_start'.
+     * The code sequence must therefore be able to cope with this, and
+     * the syscall instruction must be the final one in the sequence.
+     * For i386 we have absolutely no spare registers at the point of
+     * the syscall, so we must load *signal_pending and the syscall
+     * number off the stack inside the start/end section.
+     */
+safe_syscall_start:
+    /* *signal_pending */
+    movl 20(%esp), %eax
+    /* if signal_pending is non-zero, don't do the call */
+    testl $1, (%eax)
+    jnz return_ERESTARTSYS
+    movl 24(%esp), %eax
+    int $0x80
+safe_syscall_end:
+    /* code path for having successfully executed the syscall */
+    .cfi_remember_state
+    popl %ebx
+    .cfi_adjust_cfa_offset -4
+    .cfi_restore ebx
+    popl %esi
+    .cfi_adjust_cfa_offset -4
+    .cfi_restore esi
+    popl %edi
+    .cfi_adjust_cfa_offset -4
+    .cfi_restore edi
+    popl %ebp
+    .cfi_adjust_cfa_offset -4
+    .cfi_restore ebp
+    ret
+
+return_ERESTARTSYS:
+    /* code path when we didn't execute the syscall */
+    .cfi_restore_state
+    mov $-TARGET_ERESTARTSYS, %eax
+    popl %ebx
+    .cfi_adjust_cfa_offset -4
+    .cfi_restore ebx
+    popl %esi
+    .cfi_adjust_cfa_offset -4
+    .cfi_restore esi
+    popl %edi
+    .cfi_adjust_cfa_offset -4
+    .cfi_restore edi
+    popl %ebp
+    .cfi_adjust_cfa_offset -4
+    .cfi_restore ebp
+    ret
+    .cfi_endproc
+
+    .size   safe_syscall_base, .-safe_syscall_base