Preserve x19-x29 across world switch for exception handling

Previously exception handlers in BL3-1, X19-X29 were not saved
and restored on every SMC/trap into EL3. Instead these registers
were 'saved as needed' as a side effect of the A64 ABI used by the C
compiler.

That approach failed when world switching but was not visible
with the TSP/TSPD code because the TSP is 64-bit, did not
clobber these registers when running and did not support pre-emption
by normal world interrupts. These scenarios showed
that the values in these registers can be passed through a world
switch, which broke the normal and trusted world assumptions
about these registers being preserved.

The Ideal solution saves and restores these registers when a
world switch occurs - but that type of implementation is more complex.
So this patch always saves and restores these registers on entry and
exit of EL3.

Fixes ARM-software/tf-issues#141

Change-Id: I9a727167bbc594454e81cf78a97ca899dfb11c27

1 files changed, 34 insertions, 16 deletions

diff --git a/bl31/aarch64/runtime_exceptions.S b/bl31/aarch64/runtime_exceptions.S
index 53cc176e..9c98ad6a 100644
--- a/bl31/aarch64/runtime_exceptions.S
+++ b/bl31/aarch64/runtime_exceptions.S
@@ -39,6 +39,17 @@
 	.globl	el3_exit
 	.globl	get_exception_stack
 
+	.macro save_x18_to_x29_sp_el0
+	stp	x18, x19, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X18]
+	stp	x20, x21, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X20]
+	stp	x22, x23, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X22]
+	stp	x24, x25, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X24]
+	stp	x26, x27, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X26]
+	stp	x28, x29, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X28]
+	mrs	x18, sp_el0
+	str	x18, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_SP_EL0]
+	.endm
+
 	.section	.vectors, "ax"; .align 11
 
 	.align	7
@@ -250,6 +261,9 @@ smc_handler64:
 	stp	x4, x5, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X4]
 	stp	x6, x7, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X6]
 
+	/* Save rest of the gpregs and sp_el0*/
+	save_x18_to_x29_sp_el0
+
 	mov	x5, xzr
 	mov	x6, sp
 
@@ -264,10 +278,6 @@ smc_handler64:
 	adr	x14, rt_svc_descs_indices
 	ldrb	w15, [x14, x16]
 
-	/* Save x18 and SP_EL0 */
-	mrs	x17, sp_el0
-	stp	x18, x17, [x6, #CTX_GPREGS_OFFSET + CTX_GPREG_X18]
-
 	/* -----------------------------------------------------
 	 * Restore the saved C runtime stack value which will
 	 * become the new SP_EL0 i.e. EL3 runtime stack. It was
@@ -357,8 +367,8 @@ el3_exit: ; .type el3_exit, %function
 	msr	elr_el3, x17
 
 	/* Restore saved general purpose registers and return */
-	bl	restore_scratch_registers
-	ldp	x30, xzr, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_LR]
+	bl	restore_gp_registers
+	ldr	x30, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_LR]
 	eret
 
 smc_unknown:
@@ -369,10 +379,10 @@ smc_unknown:
 	 * content). Either way, we aren't leaking any secure information
 	 * through them
 	 */
-	bl	restore_scratch_registers_callee
+	bl	restore_gp_registers_callee
 
 smc_prohibited:
-	ldp	x30, xzr, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_LR]
+	ldr	x30, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_LR]
 	mov	w0, #SMC_UNK
 	eret
 
@@ -381,12 +391,16 @@ rt_svc_fw_critical_error:
 
 	/* -----------------------------------------------------
 	 * The following functions are used to saved and restore
-	 * all the caller saved registers as per the aapcs_64.
+	 * all the general pupose registers. Ideally we would
+	 * only save and restore the callee saved registers when
+	 * a world switch occurs but that type of implementation
+	 * is more complex. So currently we will always save and
+	 * restore these registers on entry and exit of EL3.
 	 * These are not macros to ensure their invocation fits
 	 * within the 32 instructions per exception vector.
 	 * -----------------------------------------------------
 	 */
-func save_scratch_registers
+func save_gp_registers
 	stp	x0, x1, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X0]
 	stp	x2, x3, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X2]
 	stp	x4, x5, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X4]
@@ -396,16 +410,15 @@ func save_scratch_registers
 	stp	x12, x13, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X12]
 	stp	x14, x15, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X14]
 	stp	x16, x17, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X16]
-	mrs	x17, sp_el0
-	stp	x18, x17, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X18]
+	save_x18_to_x29_sp_el0
 	ret
 
-func restore_scratch_registers
+func restore_gp_registers
 	ldp	x0, x1, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X0]
 	ldp	x2, x3, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X2]
 
-restore_scratch_registers_callee:
-	ldp	x18, x17, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X18]
+restore_gp_registers_callee:
+	ldr	x17, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_SP_EL0]
 
 	ldp	x4, x5, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X4]
 	ldp	x6, x7, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X6]
@@ -413,9 +426,14 @@ restore_scratch_registers_callee:
 	ldp	x10, x11, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X10]
 	ldp	x12, x13, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X12]
 	ldp	x14, x15, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X14]
-
 	msr	sp_el0, x17
 	ldp	x16, x17, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X16]
+	ldp	x18, x19, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X18]
+	ldp	x20, x21, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X20]
+	ldp	x22, x23, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X22]
+	ldp	x24, x25, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X24]
+	ldp	x26, x27, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X26]
+	ldp	x28, x29, [sp, #CTX_GPREGS_OFFSET + CTX_GPREG_X28]
 	ret
 
 	/* -----------------------------------------------------