[PATCH 12/24] x86/traps: Reconstruct pt_regs on task stack directly in fixup_bad_iret()

[Lai Jiangshan]

From: Lai Jiangshan <laijs@xxxxxxxxxxxxxxxxx>

In current code for handing bad iret, pt_regs is moved several times.
	Reconstruct pt_regs from bad_regs onto a tmp
	Copy tmp pt_regs onto entry stack
	Copy the pt_regs on the entry stack to the task stack in sync_regs().

This patch directly reconstructs pt_regs from bad_regs onto the task stack.

Signed-off-by: Lai Jiangshan <laijs@xxxxxxxxxxxxxxxxx>
---
 arch/x86/entry/traps.c | 35 +++++++++++++++++++----------------
 1 file changed, 19 insertions(+), 16 deletions(-)

diff --git a/arch/x86/entry/traps.c b/arch/x86/entry/traps.c
index b8fdf6a9682f..ab9866b650e7 100644
--- a/arch/x86/entry/traps.c
+++ b/arch/x86/entry/traps.c
@@ -740,25 +740,28 @@ struct pt_regs *fixup_bad_iret(struct pt_regs *bad_regs)
 	/*
 	 * This is called from entry_64.S early in handling a fault
 	 * caused by a bad iret to user mode.  To handle the fault
-	 * correctly, we want to move our stack frame to where it would
-	 * be had we entered directly on the entry stack (rather than
-	 * just below the IRET frame) and we want to pretend that the
-	 * exception came from the IRET target.
+	 * correctly, we want to pretend that the exception came from
+	 * the IRET target (userspace) and reconstruct the orginal
+	 * pt_regs from the bad regs on the task stack and switch to
+	 * the task stack to handle it.  The actual stack switch is
+	 * done in entry_64.S
 	 */
-	struct pt_regs tmp, *new_stack =
-		(struct pt_regs *)__this_cpu_read(cpu_tss_rw.x86_tss.sp0) - 1;
-
-	/* Copy the IRET target to the temporary storage. */
-	__memcpy(&tmp.ip, (void *)bad_regs->sp, 5*8);
+	struct pt_regs *regs = (struct pt_regs *)this_cpu_read(cpu_current_top_of_stack) - 1;
 
-	/* Copy the remainder of the stack from the current stack. */
-	__memcpy(&tmp, bad_regs, offsetof(struct pt_regs, ip));
+	/*
+	 * Copy the IRET target to the task pt_regs.  This fault caused
+	 * by a bad iret on native_irq_return_iret which is not used by
+	 * XENPV, so the bad_regs->sp and bad_regs are in the CPU ENTRY
+	 * AREA and they must not overlap with the task pt_regs and
+	 * we can safely use __memcpy().
+	 */
+	__memcpy(&regs->ip, (void *)bad_regs->sp, 5*8);
 
-	/* Update the entry stack */
-	__memcpy(new_stack, &tmp, sizeof(tmp));
+	/* Copy the remainder of the pt_regs from the bad_regs. */
+	__memcpy(regs, bad_regs, offsetof(struct pt_regs, ip));
 
-	BUG_ON(!user_mode(new_stack));
-	return new_stack;
+	BUG_ON(!user_mode(regs));
+	return regs;
 }
 
 #ifdef CONFIG_PAGE_TABLE_ISOLATION
@@ -861,7 +864,7 @@ struct pt_regs *do_error_entry(struct pt_regs *eregs)
 		 * pt_regs as if we faulted immediately after IRET and put
 		 * pt_regs onto the real task stack.
 		 */
-		return sync_regs(fixup_bad_iret(eregs));
+		return fixup_bad_iret(eregs);
 	}
 
 	/*
-- 
2.19.1.6.gb485710b