Skip to contentSkip to conversation

+ i386-fix-get_segment_eip-with-vm86.patch added to -mm tree

DISCUSSION Zachary Amsden 2 participants 2 messages May 31, 2006 – Jun 1, 2006 1 message not in the archive

lore
Share & tools
lore
#1 lore raw
In reply to a message not in the archive
Quoted 57 lines · [email protected] wrote:

The patch titled

i386: fix get_segment_eip() with vm86 segments

has been added to the -mm tree. Its filename is

     i386-fix-get_segment_eip-with-vm86.patch

See http://www.zip.com.au/~akpm/linux/patches/stuff/added-to-mm.txt to find out what to do about this

------------------------------------------------------
Subject: i386: fix get_segment_eip() with vm86 segments
From: Chuck Ebbert <[email protected]>

We need to check for vm86 mode first before looking at selector privilege bits.

Segment limit is always base + 64k and only the low 16 bits of EIP are significant in vm86 mode.

---

 arch/i386/mm/fault.c |   11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff -puN arch/i386/mm/fault.c~i386-fix-get_segment_eip-with-vm86 arch/i386/mm/fault.c

arch/i386/mm/fault.c renamed from arch/i386/mm/fault.c~i386-fix-get_segment_eip-with-vm86 +7 -4
@@ -77,12 +77,15 @@ static inline unsigned long get_segment_
 	unsigned seg = regs->xcs & 0xffff;
 	u32 seg_ar, seg_limit, base, *desc;
 
+	/* Unlikely, but must come before segment checks. */
+	if (unlikely(regs->eflags & VM_MASK)) {
+		base = seg << 4;
+		*eip_limit = base + 0xffff;
+		return base + (eip & 0xffff);
+	}
+
 	/* The standard kernel/user address space limit. */
 	*eip_limit = (seg & 3) ? USER_DS.seg : KERNEL_DS.seg;
-
-	/* Unlikely, but must come before segment checks. */
-	if (unlikely((regs->eflags & VM_MASK) != 0))
-		return eip + (seg << 4);
 
 	/* By far the most common cases. */
 	if (likely(seg == __USER_CS || seg == __KERNEL_CS))

_

Quoted 6 lines

Patches currently in -mm which might be from [email protected] are

i386-let-usermode-execute-the-enter.patch
i386-fix-get_segment_eip-with-vm86.patch

This looks great. While we're in the spirit let's fix kprobes v8086 handling as well by filtering out int3s from v8086 mode.

Never allow int3 traps from V8086 mode to enter the kprobes handler.

arch/i386/kernel/kprobes.c +1 -1
Index: linux-2.6.17-rc/arch/i386/kernel/kprobes.c
===================================================================
@@ -607,7 +607,7 @@ int __kprobes kprobe_exceptions_notify(s
 	struct die_args *args = (struct die_args *)data;
 	int ret = NOTIFY_DONE;
 
-	if (args->regs && user_mode(args->regs))
+	if (args->regs && user_mode_vm(args->regs))
 		return ret;
 
 	switch (val) {
#2 lore raw
Quoted 70 lines · Zachary Amsden
Quoted 59 lines · [email protected] wrote:

The patch titled

i386: fix get_segment_eip() with vm86 segments

has been added to the -mm tree. Its filename is

    i386-fix-get_segment_eip-with-vm86.patch

See http://www.zip.com.au/~akpm/linux/patches/stuff/added-to-mm.txt to find out what to do about this

------------------------------------------------------
Subject: i386: fix get_segment_eip() with vm86 segments
From: Chuck Ebbert <[email protected]>

We need to check for vm86 mode first before looking at selector privilege bits.

Segment limit is always base + 64k and only the low 16 bits of EIP are significant in vm86 mode.

---

arch/i386/mm/fault.c | 11 +++++++----

1 file changed, 7 insertions(+), 4 deletions(-)

diff -puN arch/i386/mm/fault.c~i386-fix-get_segment_eip-with-vm86 arch/i386/mm/fault.c
--- devel/arch/i386/mm/fault.c~i386-fix-get_segment_eip-with-vm86

2006-05-29 20:06:19.000000000 -0700

+++ devel-akpm/arch/i386/mm/fault.c 2006-05-29 20:06:19.000000000 -0700

(quoted diff) +7 -4
@@ -77,12 +77,15 @@ static inline unsigned long get_segment_
 	unsigned seg = regs->xcs & 0xffff;
 	u32 seg_ar, seg_limit, base, *desc;
 
+	/* Unlikely, but must come before segment checks. */
+	if (unlikely(regs->eflags & VM_MASK)) {
+		base = seg << 4;
+		*eip_limit = base + 0xffff;
+		return base + (eip & 0xffff);
+	}
+
 	/* The standard kernel/user address space limit. */
 	*eip_limit = (seg & 3) ? USER_DS.seg : KERNEL_DS.seg;
-
-	/* Unlikely, but must come before segment checks. */
-	if (unlikely((regs->eflags & VM_MASK) != 0))
-		return eip + (seg << 4);
 
 	/* By far the most common cases. */
 	if (likely(seg == __USER_CS || seg == __KERNEL_CS))

_

Quoted 6 lines

Patches currently in -mm which might be from [email protected] are

i386-let-usermode-execute-the-enter.patch
i386-fix-get_segment_eip-with-vm86.patch

This looks great. While we're in the spirit let's fix kprobes v8086 handling as well by filtering out int3s from v8086 mode.

Quoted 3 lines

Never allow int3 traps from V8086 mode to enter the kprobes handler.

Quoted 14 lines

Index: linux-2.6.17-rc/arch/i386/kernel/kprobes.c
===================================================================

arch/i386/kernel/kprobes.c +1 -1
@@ -607,7 +607,7 @@ int __kprobes kprobe_exceptions_notify(s
 	struct die_args *args = (struct die_args *)data;
 	int ret = NOTIFY_DONE;
 
-	if (args->regs && user_mode(args->regs))
+	if (args->regs && user_mode_vm(args->regs))
 		return ret;
 
 	switch (val) {