linux-kernel - [PATCH] lguest: simplify lguest

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <1426974173-22145-1-git-send-email-dvlasenk@redhat.com>
Date:	Sat, 21 Mar 2015 22:42:53 +0100
From:	Denys Vlasenko <dvlasenk@...hat.com>
To:	Rusty Russell <rusty@...tcorp.com.au>
Cc:	Denys Vlasenko <dvlasenk@...hat.com>, lguest@...ts.ozlabs.org,
	x86@...nel.org, linux-kernel@...r.kernel.org
Subject: [PATCH] lguest: simplify lguest_iret

Signed-off-by: Denys Vlasenko <dvlasenk@...hat.com>
CC: lguest@...ts.ozlabs.org
CC: x86@...nel.org
CC: linux-kernel@...r.kernel.org
---
 arch/x86/lguest/head_32.S | 19 +++++++++----------
 1 file changed, 9 insertions(+), 10 deletions(-)

diff --git a/arch/x86/lguest/head_32.S b/arch/x86/lguest/head_32.S
index 6ddfe4f..583732c 100644
--- a/arch/x86/lguest/head_32.S
+++ b/arch/x86/lguest/head_32.S
@@ -168,29 +168,28 @@ ENTRY(lg_restore_fl)
  * So we have to copy eflags from the stack to lguest_data.irq_enabled before
  * we do the "iret".
  *
- * There are two problems with this: firstly, we need to use a register to do
- * the copy and secondly, the whole thing needs to be atomic.  The first
- * problem is easy to solve: push %eax on the stack so we can use it, and then
- * restore it at the end just before the real "iret".
+ * There are two problems with this: firstly, we can't clobber any registers
+ * and secondly, the whole thing needs to be atomic.  The first problem
+ * is solved by using "push memory"/"pop memory" instruction pair for copying.
  *
  * The second is harder: copying eflags to lguest_data.irq_enabled will turn
  * interrupts on before we're finished, so we could be interrupted before we
  * return to userspace or wherever.  Our solution to this is to surround the
  * code with lguest_noirq_start: and lguest_noirq_end: labels.  We tell the
  * Host that it is *never* to interrupt us there, even if interrupts seem to be
- * enabled.
+ * enabled. (It's not necessary to protect pop instruction, since
+ * data gets updated only after it completes, so we end up surrounding
+ * just one instruction, iret).
  */
 ENTRY(lguest_iret)
-	pushl	%eax
-	movl	12(%esp), %eax
-lguest_noirq_start:
+	pushl	2*4(%esp)
 	/*
 	 * Note the %ss: segment prefix here.  Normal data accesses use the
 	 * "ds" segment, but that will have already been restored for whatever
 	 * we're returning to (such as userspace): we can't trust it.  The %ss:
 	 * prefix makes sure we use the stack segment, which is still valid.
 	 */
-	movl	%eax,%ss:lguest_data+LGUEST_DATA_irq_enabled
-	popl	%eax
+	popl	%ss:lguest_data+LGUEST_DATA_irq_enabled
+lguest_noirq_start:
 	iret
 lguest_noirq_end:
-- 
1.8.1.4

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/