x86-64: Add user_64bit_mode paravirt op
Three places in the kernel assume that the only long mode CPL 3
selector is __USER_CS. This is not true on Xen -- Xen's sysretq
changes cs to the magic value 0xe033.
Two of the places are corner cases, but as of "x86-64: Improve
vsyscall emulation CS and RIP handling"
(c9712944b2a12373cb6ff8059afcfb7e826a6c54), vsyscalls will segfault
if called with Xen's extra CS selector. This causes a panic when
older init builds die.
It seems impossible to make Xen use __USER_CS reliably without
taking a performance hit on every system call, so this fixes the
tests instead with a new paravirt op. It's a little ugly because
ptrace.h can't include paravirt.h.
Signed-off-by: Andy Lutomirski <luto@mit.edu>
Link: http://lkml.kernel.org/r/f4fcb3947340d9e96ce1054a432f183f9da9db83.1312378163.git.luto@mit.edu
Reported-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
diff --git a/arch/x86/include/asm/desc.h b/arch/x86/include/asm/desc.h
index 7b439d9..41935fa 100644
--- a/arch/x86/include/asm/desc.h
+++ b/arch/x86/include/asm/desc.h
@@ -27,8 +27,8 @@
desc->base2 = (info->base_addr & 0xff000000) >> 24;
/*
- * Don't allow setting of the lm bit. It is useless anyway
- * because 64bit system calls require __USER_CS:
+ * Don't allow setting of the lm bit. It would confuse
+ * user_64bit_mode and would get overridden by sysret anyway.
*/
desc->l = 0;
}
diff --git a/arch/x86/include/asm/paravirt_types.h b/arch/x86/include/asm/paravirt_types.h
index 8288509..96a0f80 100644
--- a/arch/x86/include/asm/paravirt_types.h
+++ b/arch/x86/include/asm/paravirt_types.h
@@ -41,6 +41,7 @@
#include <asm/desc_defs.h>
#include <asm/kmap_types.h>
+#include <asm/pgtable_types.h>
struct page;
struct thread_struct;
@@ -63,6 +64,11 @@
struct pv_info {
unsigned int kernel_rpl;
int shared_kernel_pmd;
+
+#ifdef CONFIG_X86_64
+ u16 extra_user_64bit_cs; /* __USER_CS if none */
+#endif
+
int paravirt_enabled;
const char *name;
};
diff --git a/arch/x86/include/asm/ptrace.h b/arch/x86/include/asm/ptrace.h
index 94e7618..3566454 100644
--- a/arch/x86/include/asm/ptrace.h
+++ b/arch/x86/include/asm/ptrace.h
@@ -131,6 +131,9 @@
#ifdef __KERNEL__
#include <linux/init.h>
+#ifdef CONFIG_PARAVIRT
+#include <asm/paravirt_types.h>
+#endif
struct cpuinfo_x86;
struct task_struct;
@@ -187,6 +190,22 @@
#endif
}
+#ifdef CONFIG_X86_64
+static inline bool user_64bit_mode(struct pt_regs *regs)
+{
+#ifndef CONFIG_PARAVIRT
+ /*
+ * On non-paravirt systems, this is the only long mode CPL 3
+ * selector. We do not allow long mode selectors in the LDT.
+ */
+ return regs->cs == __USER_CS;
+#else
+ /* Headers are too twisted for this to go in paravirt.h. */
+ return regs->cs == __USER_CS || regs->cs == pv_info.extra_user_64bit_cs;
+#endif
+}
+#endif
+
/*
* X86_32 CPUs don't save ss and esp if the CPU is already in kernel mode
* when it traps. The previous stack will be directly underneath the saved
diff --git a/arch/x86/kernel/paravirt.c b/arch/x86/kernel/paravirt.c
index 869e1ae..681f159 100644
--- a/arch/x86/kernel/paravirt.c
+++ b/arch/x86/kernel/paravirt.c
@@ -299,6 +299,10 @@
.paravirt_enabled = 0,
.kernel_rpl = 0,
.shared_kernel_pmd = 1, /* Only used when CONFIG_X86_PAE is set */
+
+#ifdef CONFIG_X86_64
+ .extra_user_64bit_cs = __USER_CS,
+#endif
};
struct pv_init_ops pv_init_ops = {
diff --git a/arch/x86/kernel/step.c b/arch/x86/kernel/step.c
index 7977f0c..c346d11 100644
--- a/arch/x86/kernel/step.c
+++ b/arch/x86/kernel/step.c
@@ -74,7 +74,7 @@
#ifdef CONFIG_X86_64
case 0x40 ... 0x4f:
- if (regs->cs != __USER_CS)
+ if (!user_64bit_mode(regs))
/* 32-bit mode: register increment */
return 0;
/* 64-bit mode: REX prefix */
diff --git a/arch/x86/kernel/vsyscall_64.c b/arch/x86/kernel/vsyscall_64.c
index dda7dff..1725930 100644
--- a/arch/x86/kernel/vsyscall_64.c
+++ b/arch/x86/kernel/vsyscall_64.c
@@ -127,11 +127,7 @@
local_irq_enable();
- /*
- * Real 64-bit user mode code has cs == __USER_CS. Anything else
- * is bogus.
- */
- if (regs->cs != __USER_CS) {
+ if (!user_64bit_mode(regs)) {
/*
* If we trapped from kernel mode, we might as well OOPS now
* instead of returning to some random address and OOPSing
diff --git a/arch/x86/mm/fault.c b/arch/x86/mm/fault.c
index 2dbf6bf..c1d0182 100644
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -105,7 +105,7 @@
* but for now it's good enough to assume that long
* mode only uses well known segments or kernel.
*/
- return (!user_mode(regs)) || (regs->cs == __USER_CS);
+ return (!user_mode(regs) || user_64bit_mode(regs));
#endif
case 0x60:
/* 0x64 thru 0x67 are valid prefixes in all modes. */
diff --git a/arch/x86/xen/enlighten.c b/arch/x86/xen/enlighten.c
index 5525163..78fe33d 100644
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -937,6 +937,10 @@
.paravirt_enabled = 1,
.shared_kernel_pmd = 0,
+#ifdef CONFIG_X86_64
+ .extra_user_64bit_cs = FLAT_USER_CS64,
+#endif
+
.name = "Xen",
};