x86/fred: Fixup fault on ERETU by jumping to fred_entrypoint_user

If the stack frame contains an invalid user context (e.g. due to invalid SS,
a non-canonical RIP, etc.) the ERETU instruction will trap (#SS or #GP).

From a Linux point of view, this really should be considered a user space
failure, so use the standard fault fixup mechanism to intercept the fault,
fix up the exception frame, and redirect execution to fred_entrypoint_user.
The end result is that it appears just as if the hardware had taken the
exception immediately after completing the transition to user space.

Suggested-by: H. Peter Anvin (Intel) <hpa@zytor.com>
Signed-off-by: Xin Li <xin3.li@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Tested-by: Shan Kang <shan.kang@intel.com>
Link: https://lore.kernel.org/r/20231205105030.8698-30-xin3.li@intel.com

Author: xinli-intel

arch/x86/entry/entry_64_fred.S

@@ -3,6 +3,7 @@
  * The actual FRED entry points.
  */
 
+#include <asm/asm.h>
 #include <asm/fred.h>
 
 #include "calling.h"
@@ -34,7 +35,9 @@ SYM_CODE_START_NOALIGN(asm_fred_entrypoint_user)
 	call	fred_entry_from_user
 SYM_INNER_LABEL(asm_fred_exit_user, SYM_L_GLOBAL)
 	FRED_EXIT
-	ERETU
+1:	ERETU
+
+	_ASM_EXTABLE_TYPE(1b, asm_fred_entrypoint_user, EX_TYPE_ERETU)
 SYM_CODE_END(asm_fred_entrypoint_user)
 
 /*

arch/x86/include/asm/extable_fixup_types.h

@@ -64,6 +64,8 @@
 #define	EX_TYPE_UCOPY_LEN4		(EX_TYPE_UCOPY_LEN | EX_DATA_IMM(4))
 #define	EX_TYPE_UCOPY_LEN8		(EX_TYPE_UCOPY_LEN | EX_DATA_IMM(8))
 
-#define EX_TYPE_ZEROPAD			20 /* longword load with zeropad on fault */
+#define	EX_TYPE_ZEROPAD			20 /* longword load with zeropad on fault */
+
+#define	EX_TYPE_ERETU			21
 
 #endif

arch/x86/mm/extable.c

@@ -6,6 +6,7 @@
 #include <xen/xen.h>
 
 #include <asm/fpu/api.h>
+#include <asm/fred.h>
 #include <asm/sev.h>
 #include <asm/traps.h>
 #include <asm/kdebug.h>
@@ -223,6 +224,79 @@ static bool ex_handler_ucopy_len(const struct exception_table_entry *fixup,
 	return ex_handler_uaccess(fixup, regs, trapnr, fault_address);
 }
 
+#ifdef CONFIG_X86_FRED
+static bool ex_handler_eretu(const struct exception_table_entry *fixup,
+			     struct pt_regs *regs, unsigned long error_code)
+{
+	struct pt_regs *uregs = (struct pt_regs *)(regs->sp - offsetof(struct pt_regs, orig_ax));
+	unsigned short ss = uregs->ss;
+	unsigned short cs = uregs->cs;
+
+	/*
+	 * Move the NMI bit from the invalid stack frame, which caused ERETU
+	 * to fault, to the fault handler's stack frame, thus to unblock NMI
+	 * with the fault handler's ERETS instruction ASAP if NMI is blocked.
+	 */
+	regs->fred_ss.nmi = uregs->fred_ss.nmi;
+
+	/*
+	 * Sync event information to uregs, i.e., the ERETU return frame, but
+	 * is it safe to write to the ERETU return frame which is just above
+	 * current event stack frame?
+	 *
+	 * The RSP used by FRED to push a stack frame is not the value in %rsp,
+	 * it is calculated from %rsp with the following 2 steps:
+	 * 1) RSP = %rsp - (IA32_FRED_CONFIG & 0x1c0)	// Reserve N*64 bytes
+	 * 2) RSP = RSP & ~0x3f		// Align to a 64-byte cache line
+	 * when an event delivery doesn't trigger a stack level change.
+	 *
+	 * Here is an example with N*64 (N=1) bytes reserved:
+	 *
+	 *  64-byte cache line ==>  ______________
+	 *                         |___Reserved___|
+	 *                         |__Event_data__|
+	 *                         |_____SS_______|
+	 *                         |_____RSP______|
+	 *                         |_____FLAGS____|
+	 *                         |_____CS_______|
+	 *                         |_____IP_______|
+	 *  64-byte cache line ==> |__Error_code__| <== ERETU return frame
+	 *                         |______________|
+	 *                         |______________|
+	 *                         |______________|
+	 *                         |______________|
+	 *                         |______________|
+	 *                         |______________|
+	 *                         |______________|
+	 *  64-byte cache line ==> |______________| <== RSP after step 1) and 2)
+	 *                         |___Reserved___|
+	 *                         |__Event_data__|
+	 *                         |_____SS_______|
+	 *                         |_____RSP______|
+	 *                         |_____FLAGS____|
+	 *                         |_____CS_______|
+	 *                         |_____IP_______|
+	 *  64-byte cache line ==> |__Error_code__| <== ERETS return frame
+	 *
+	 * Thus a new FRED stack frame will always be pushed below a previous
+	 * FRED stack frame ((N*64) bytes may be reserved between), and it is
+	 * safe to write to a previous FRED stack frame as they never overlap.
+	 */
+	fred_info(uregs)->edata = fred_event_data(regs);
+	uregs->ssx = regs->ssx;
+	uregs->fred_ss.ss = ss;
+	/* The NMI bit was moved away above */
+	uregs->fred_ss.nmi = 0;
+	uregs->csx = regs->csx;
+	uregs->fred_cs.sl = 0;
+	uregs->fred_cs.wfe = 0;
+	uregs->cs = cs;
+	uregs->orig_ax = error_code;
+
+	return ex_handler_default(fixup, regs);
+}
+#endif
+
 int ex_get_fixup_type(unsigned long ip)
 {
 	const struct exception_table_entry *e = search_exception_tables(ip);
@@ -300,6 +374,10 @@ int fixup_exception(struct pt_regs *regs, int trapnr, unsigned long error_code,
 		return ex_handler_ucopy_len(e, regs, trapnr, fault_addr, reg, imm);
 	case EX_TYPE_ZEROPAD:
 		return ex_handler_zeropad(e, regs, fault_addr);
+#ifdef CONFIG_X86_FRED
+	case EX_TYPE_ERETU:
+		return ex_handler_eretu(e, regs, error_code);
+#endif
 	}
 	BUG();
 }