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 7b439d9aea2a5c30a22f1f3b8ba17db7f4011a39..41935fadfdfcca18be2ca724b34e1c45d3116ac1 100644 (file)
--- a/arch/x86/include/asm/desc.h
+++ b/arch/x86/include/asm/desc.h
@@ -27,8 +27,8 @@ static inline void fill_ldt(struct desc_struct *desc, const struct user_desc *in
 
        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 82885099c86934ca109799f010e83cfe021d0b7f..96a0f80407b82f046398ea027fd4226c84fa4ef1 100644 (file)
--- 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 paravirt_callee_save {
 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 94e7618fcac8d6fb8bd61c1b356b522187b1218a..35664547125b40f663d09c3e556f7699e630586b 100644 (file)
--- a/arch/x86/include/asm/ptrace.h
+++ b/arch/x86/include/asm/ptrace.h
@@ -131,6 +131,9 @@ struct pt_regs {
 #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 @@ static inline int v8086_mode(struct pt_regs *regs)
 #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 869e1aeeb71bdc7e4059798bb0f3cdb89061c15c..681f159942184a0a4ae48a374999d16dd2f78574 100644 (file)
--- a/arch/x86/kernel/paravirt.c
+++ b/arch/x86/kernel/paravirt.c
@@ -299,6 +299,10 @@ struct pv_info pv_info = {
        .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 7977f0cfe339076a713ae773bbb221d9cdf406e2..c346d116148866758ca5e602684eb91fcb77f219 100644 (file)
--- a/arch/x86/kernel/step.c
+++ b/arch/x86/kernel/step.c
@@ -74,7 +74,7 @@ static int is_setting_trap_flag(struct task_struct *child, struct pt_regs *regs)
 
 #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 dda7dff9cef7e624be6239ca463789e09eedd200..1725930a6f9fa92fe3251d9c7a67e4a5f4b839e2 100644 (file)
--- a/arch/x86/kernel/vsyscall_64.c
+++ b/arch/x86/kernel/vsyscall_64.c
@@ -127,11 +127,7 @@ void dotraplinkage do_emulate_vsyscall(struct pt_regs *regs, long error_code)
 
        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 2dbf6bf4c7e5295906d5f0b25fcb97dd9845d363..c1d018238f3257515b23d82969970acbc34aa184 100644 (file)
--- a/arch/x86/mm/fault.c
+++ b/arch/x86/mm/fault.c
@@ -105,7 +105,7 @@ check_prefetch_opcode(struct pt_regs *regs, unsigned char *instr,
                 * 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 5525163a03985a517e7a043c3daa92fe04d4bb6e..78fe33d03565d3afb3da42b87ec19d7ef6712274 100644 (file)
--- a/arch/x86/xen/enlighten.c
+++ b/arch/x86/xen/enlighten.c
@@ -937,6 +937,10 @@ static const struct pv_info xen_info __initconst = {
        .paravirt_enabled = 1,
        .shared_kernel_pmd = 0,
 
+#ifdef CONFIG_X86_64
+       .extra_user_64bit_cs = FLAT_USER_CS64,
+#endif
+
        .name = "Xen",
 };