[patch 1/3] x86, xsave: remove the redundant access_ok() in setup_rt_frame()

[patch 1/3] x86, xsave: remove the redundant access_ok() in setup_rt_frame()

[Suresh Siddha]

[-- Attachment #1: remove_redundant_access_ok.patch --]
[-- Type: text/plain, Size: 696 bytes --]

save_i387_xstate() is already doing the required access_ok(). Remove
the redundant access_ok() before it.

Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
---

Index: tip/arch/x86/kernel/signal_64.c
===================================================================
--- tip.orig/arch/x86/kernel/signal_64.c	2008-08-11 15:26:47.000000000 -0700
+++ tip/arch/x86/kernel/signal_64.c	2008-08-11 15:28:52.000000000 -0700
@@ -208,9 +208,6 @@
 		frame = (void __user *)round_down(
 			(unsigned long)fp - sizeof(struct rt_sigframe), 16) - 8;
 
-		if (!access_ok(VERIFY_WRITE, fp, sizeof(struct _fpstate)))
-			goto give_sigsegv;
-
 		if (save_i387_xstate(fp) < 0)
 			err |= -1;
 	} else

-- 

[patch 2/3] x86, xsave: clear the user buffer before doing fxsave/xsave

[Suresh Siddha]

[-- Attachment #1: init_fpstate_user_buf.patch --]
[-- Type: text/plain, Size: 1164 bytes --]

fxsave/xsave instructions will not touch all the bytes in the
fxsave/xsave frame. Clear the user buffer before doing fxsave/xsave
directly to user buffer during the sigcontext setup.

This is essentially needed in the context of xsave(for example,
some of the fields in the xsave header are not touched by the xsave
and defined as must be zero).

This will also present uniform and clean context to the user (from
which user can safely do fxrstor/xrstor).

Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
---

Index: tip/arch/x86/kernel/xsave.c
===================================================================
--- tip.orig/arch/x86/kernel/xsave.c	2008-08-11 15:26:47.000000000 -0700
+++ tip/arch/x86/kernel/xsave.c	2008-08-11 15:44:33.000000000 -0700
@@ -92,6 +92,12 @@
 		return 0;
 	clear_used_math(); /* trigger finit */
 	if (task_thread_info(tsk)->status & TS_USEDFPU) {
+		/*
+	 	 * Start with clearing the user buffer. This will present a
+	 	 * clean context for the bytes not touched by the fxsave/xsave.
+		 */
+		__clear_user(buf, sig_xstate_size);
+
 		if (task_thread_info(tsk)->status & TS_XSAVE)
 			err = xsave_user(buf);
 		else

-- 

[patch 3/3] x86, xsave: use BUG_ON() instead of BUILD_BUG_ON()

[Suresh Siddha]

[-- Attachment #1: modify_build_bug_on.patch --]
[-- Type: text/plain, Size: 703 bytes --]

All these structure sizes are runtime determined. So use a runtime
bug check.

Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
---

Index: tip/arch/x86/kernel/xsave.c
===================================================================
--- tip.orig/arch/x86/kernel/xsave.c	2008-08-11 15:44:33.000000000 -0700
+++ tip/arch/x86/kernel/xsave.c	2008-08-11 15:45:09.000000000 -0700
@@ -82,8 +82,7 @@
 	if (!access_ok(VERIFY_WRITE, buf, sig_xstate_size))
 		return -EACCES;
 
-	BUILD_BUG_ON(sizeof(struct user_i387_struct) !=
-			sizeof(tsk->thread.xstate->fxsave));
+	BUG_ON(sig_xstate_size < xstate_size);
 
 	if ((unsigned long)buf % 64)
 		printk("save_i387_xstate: bad fpstate %p\n", buf);

-- 

Re: [patch 3/3] x86, xsave: use BUG_ON() instead of BUILD_BUG_ON()

[Ingo Molnar]

* Suresh Siddha <suresh.b.siddha@intel.com> wrote:

> All these structure sizes are runtime determined. So use a runtime bug 
> check.

applied these patches to tip/x86/xsave:

 f65bc21: x86, xsave: use BUG_ON() instead of BUILD_BUG_ON()
 ed40595: x86, xsave: clear the user buffer before doing fxsave/xsave
 ee2b92a: x86, xsave: remove the redundant access_ok() in setup_rt_frame()

thanks Suresh.

	Ingo