Skip to contentSkip to conversation

x86/cet: Fix build on newer versions of GCC

PATCH In review Andrew Cooper 3 participants 5 messages Aug 17, 2021

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86/cet: Fix build on newer versions of GCC

Some versions of GCC complain with:

traps.c:405:22: error: 'get_shstk_bottom' defined but not used [-Werror=unused-function]
static unsigned long get_shstk_bottom(unsigned long sp)

                        ^~~~~~~~~~~~~~~~

cc1: all warnings being treated as errors

Change #ifdef to if ( IS_ENABLED(...) ) to make the sole user of get_shstk_bottom() visible to the compiler.

Fixes 35727551c070 ("x86/cet: Fix shskt manipulation error with BUGFRAME_{warn,run_fn}")

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Wei Liu <[email protected]>
CC: Marek Marczykowski-Górecki <[email protected]>

Not actually tested. I don't seem to have a new enough GCC to hand.

Most of the delta here is indentation. This diff is more easily reviewed with git show --ignore-all-space

---
 xen/arch/x86/traps.c | 83 ++++++++++++++++++++++++++--------------------------
 1 file changed, 42 insertions(+), 41 deletions(-)
xen/arch/x86/traps.c +42 -41
index 30eefbad4863..4a0e498b4c21 100644
@@ -780,55 +780,56 @@ static void do_reserved_trap(struct cpu_user_regs *regs)
 static void fixup_exception_return(struct cpu_user_regs *regs,
                                    unsigned long fixup)
 {
-#ifdef CONFIG_XEN_SHSTK
-    unsigned long ssp, *ptr, *base;
+    if ( IS_ENABLED(CONFIG_XEN_SHSTK) )
+    {
+        unsigned long ssp, *ptr, *base;
 
-    asm ( "rdsspq %0" : "=r" (ssp) : "0" (1) );
-    if ( ssp == 1 )
-        goto shstk_done;
+        asm ( "rdsspq %0" : "=r" (ssp) : "0" (1) );
+        if ( ssp == 1 )
+            goto shstk_done;
 
-    ptr = _p(ssp);
-    base = _p(get_shstk_bottom(ssp));
+        ptr = _p(ssp);
+        base = _p(get_shstk_bottom(ssp));
 
-    for ( ; ptr < base; ++ptr )
-    {
-        /*
-         * Search for %rip.  The shstk currently looks like this:
-         *
-         *   ...  [Likely pointed to by SSP]
-         *   %cs  [== regs->cs]
-         *   %rip [== regs->rip]
-         *   SSP  [Likely points to 3 slots higher, above %cs]
-         *   ...  [call tree to this function, likely 2/3 slots]
-         *
-         * and we want to overwrite %rip with fixup.  There are two
-         * complications:
-         *   1) We cant depend on SSP values, because they won't differ by 3
-         *      slots if the exception is taken on an IST stack.
-         *   2) There are synthetic (unrealistic but not impossible) scenarios
-         *      where %rip can end up in the call tree to this function, so we
-         *      can't check against regs->rip alone.
-         *
-         * Check for both regs->rip and regs->cs matching.
-         */
-        if ( ptr[0] == regs->rip && ptr[1] == regs->cs )
+        for ( ; ptr < base; ++ptr )
         {
-            asm ( "wrssq %[fix], %[stk]"
-                  : [stk] "=m" (ptr[0])
-                  : [fix] "r" (fixup) );
-            goto shstk_done;
+            /*
+             * Search for %rip.  The shstk currently looks like this:
+             *
+             *   ...  [Likely pointed to by SSP]
+             *   %cs  [== regs->cs]
+             *   %rip [== regs->rip]
+             *   SSP  [Likely points to 3 slots higher, above %cs]
+             *   ...  [call tree to this function, likely 2/3 slots]
+             *
+             * and we want to overwrite %rip with fixup.  There are two
+             * complications:
+             *   1) We cant depend on SSP values, because they won't differ by
+             *      3 slots if the exception is taken on an IST stack.
+             *   2) There are synthetic (unrealistic but not impossible)
+             *      scenarios where %rip can end up in the call tree to this
+             *      function, so we can't check against regs->rip alone.
+             *
+             * Check for both regs->rip and regs->cs matching.
+             */
+            if ( ptr[0] == regs->rip && ptr[1] == regs->cs )
+            {
+                asm ( "wrssq %[fix], %[stk]"
+                      : [stk] "=m" (ptr[0])
+                      : [fix] "r" (fixup) );
+                goto shstk_done;
+            }
         }
-    }
 
-    /*
-     * We failed to locate and fix up the shadow IRET frame.  This could be
-     * due to shadow stack corruption, or bad logic above.  We cannot continue
-     * executing the interrupted context.
-     */
-    BUG();
+        /*
+         * We failed to locate and fix up the shadow IRET frame.  This could
+         * be due to shadow stack corruption, or bad logic above.  We cannot
+         * continue executing the interrupted context.
+         */
+        BUG();
 
+    }
  shstk_done:
-#endif /* CONFIG_XEN_SHSTK */
 
     /* Fixup the regular stack. */
     regs->rip = fixup;
Signature · 1 line
2.11.0
#1 lore raw
Quoted 12 lines · Andrew Cooper

Some versions of GCC complain with:

traps.c:405:22: error: 'get_shstk_bottom' defined but not used [-Werror=unused-function]
static unsigned long get_shstk_bottom(unsigned long sp)

                        ^~~~~~~~~~~~~~~~

cc1: all warnings being treated as errors

Change #ifdef to if ( IS_ENABLED(...) ) to make the sole user of get_shstk_bottom() visible to the compiler.

Fixes 35727551c070 ("x86/cet: Fix shskt manipulation error with BUGFRAME_{warn,run_fn}")
Quoted 7 lines

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Wei Liu <[email protected]>
CC: Marek Marczykowski-Górecki <[email protected]>

Not actually tested. I don't seem to have a new enough GCC to hand.

I have just compile-tested it and it seems to fix the issue (indeed it failed before with CONFIG_XEN_SHSTK disabled).

Quoted 2 lines

Most of the delta here is indentation. This diff is more easily reviewed with git show --ignore-all-space

Wouldn't this make the compiler include the code even if CONFIG_XEN_SHSTK is disabled (not a huge issue...)? Or is it smart enough to optimize it out in that case?

Signature · 3 lines
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab
#2 lore raw
Quoted 19 lines · Marek Marczykowski-Górecki
Quoted 18 lines · Andrew Cooper

Some versions of GCC complain with:

traps.c:405:22: error: 'get_shstk_bottom' defined but not used [-Werror=unused-function]
static unsigned long get_shstk_bottom(unsigned long sp)

                        ^~~~~~~~~~~~~~~~

cc1: all warnings being treated as errors

Change #ifdef to if ( IS_ENABLED(...) ) to make the sole user of get_shstk_bottom() visible to the compiler.

Fixes 35727551c070 ("x86/cet: Fix shskt manipulation error with BUGFRAME_{warn,run_fn}")

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Wei Liu <[email protected]>
CC: Marek Marczykowski-Górecki <[email protected]>

Quoted 3 lines
Quoted 1 line

Not actually tested. I don't seem to have a new enough GCC to hand.

I have just compile-tested it and it seems to fix the issue (indeed it failed before with CONFIG_XEN_SHSTK disabled).

Oh, thanks!

Quoted 6 lines
Quoted 3 lines

Most of the delta here is indentation. This diff is more easily reviewed with git show --ignore-all-space

Wouldn't this make the compiler include the code even if CONFIG_XEN_SHSTK is disabled (not a huge issue...)? Or is it smart enough to optimize it out in that case?

Its a trivial dead-code elimination example, and yes - the compiler is smart enough.

~Andrew

#3 lore raw
Quoted 20 lines · Andrew Cooper
Quoted 19 lines · Marek Marczykowski-Górecki
Quoted 18 lines · Andrew Cooper

Some versions of GCC complain with:

traps.c:405:22: error: 'get_shstk_bottom' defined but not used [-Werror=unused-function]
static unsigned long get_shstk_bottom(unsigned long sp)

                        ^~~~~~~~~~~~~~~~

cc1: all warnings being treated as errors

Change #ifdef to if ( IS_ENABLED(...) ) to make the sole user of get_shstk_bottom() visible to the compiler.

Fixes 35727551c070 ("x86/cet: Fix shskt manipulation error with BUGFRAME_{warn,run_fn}")

---
CC: Jan Beulich <[email protected]>
CC: Roger Pau Monné <[email protected]>
CC: Wei Liu <[email protected]>
CC: Marek Marczykowski-Górecki <[email protected]>

Quoted 5 lines
Quoted 3 lines
Quoted 1 line

Not actually tested. I don't seem to have a new enough GCC to hand.

I have just compile-tested it and it seems to fix the issue (indeed it failed before with CONFIG_XEN_SHSTK disabled).

Oh, thanks!

And I can confirm it doesn't break anything runtime (but that's pretty obvious looking at the patch).

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Most of the delta here is indentation. This diff is more easily reviewed with git show --ignore-all-space

Wouldn't this make the compiler include the code even if CONFIG_XEN_SHSTK is disabled (not a huge issue...)? Or is it smart enough to optimize it out in that case?

Its a trivial dead-code elimination example, and yes - the compiler is smart enough.

:)

Signature · 3 lines
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab
Reviewed-by on v1 1/1
#4 lore raw
In reply to Andrew Cooper
Quoted 4 lines · Andrew Cooper

Some versions of GCC complain with:

traps.c:405:22: error: 'get_shstk_bottom' defined but not used [-Werror=unused-function]
static unsigned long get_shstk_bottom(unsigned long sp)

Quoted 8 lines
                        ^~~~~~~~~~~~~~~~

cc1: all warnings being treated as errors

Change #ifdef to if ( IS_ENABLED(...) ) to make the sole user of get_shstk_bottom() visible to the compiler.

Fixes 35727551c070 ("x86/cet: Fix shskt manipulation error with BUGFRAME_{warn,run_fn}")
Quoted 1 line

Not actually tested. I don't seem to have a new enough GCC to hand.

Jan