[librm] Add facility to provide register and stack dump for CPU exceptions

When DEBUG=librm_mgmt is enabled, intercept CPU exceptions and provide
a register and stack dump, then drop to an emergency shell.  Exiting
from the shell will almost certainly not work, but this provides an
opportunity to view the register and stack dump and carry out some
basic debugging.

Note that we can intercept only the first 8 CPU exceptions, since a
PXE ROM is not permitted to rebase the PIC.

Signed-off-by: Michael Brown <mcb30@ipxe.org>

3 files changed, 162 insertions, 7 deletions

diff --git a/src/arch/x86/include/librm.h b/src/arch/x86/include/librm.h
index 311748be..597c65e6 100644
--- a/src/arch/x86/include/librm.h
+++ b/src/arch/x86/include/librm.h
@@ -376,6 +376,50 @@ struct interrupt_vector {
 /** "jmp" instruction */
 #define JMP_INSN 0xe9
 
+/** 32-bit interrupt wrapper stack frame */
+struct interrupt_frame32 {
+	uint32_t esp;
+	uint32_t ss;
+	uint32_t gs;
+	uint32_t fs;
+	uint32_t es;
+	uint32_t ds;
+	uint32_t ebp;
+	uint32_t edi;
+	uint32_t esi;
+	uint32_t edx;
+	uint32_t ecx;
+	uint32_t ebx;
+	uint32_t eax;
+	uint32_t eip;
+	uint32_t cs;
+	uint32_t eflags;
+} __attribute__ (( packed ));
+
+/** 64-bit interrupt wrapper stack frame */
+struct interrupt_frame64 {
+	uint64_t r15;
+	uint64_t r14;
+	uint64_t r13;
+	uint64_t r12;
+	uint64_t r11;
+	uint64_t r10;
+	uint64_t r9;
+	uint64_t r8;
+	uint64_t rbp;
+	uint64_t rdi;
+	uint64_t rsi;
+	uint64_t rdx;
+	uint64_t rcx;
+	uint64_t rbx;
+	uint64_t rax;
+	uint64_t rip;
+	uint64_t cs;
+	uint64_t rflags;
+	uint64_t rsp;
+	uint64_t ss;
+} __attribute__ (( packed ));
+
 extern void set_interrupt_vector ( unsigned int intr, void *vector );
 
 /** A page table */
diff --git a/src/arch/x86/transitions/librm.S b/src/arch/x86/transitions/librm.S
index c31daad8..9d3eff95 100644
--- a/src/arch/x86/transitions/librm.S
+++ b/src/arch/x86/transitions/librm.S
@@ -1363,20 +1363,19 @@ flatten_dummy:
 	.code32
 	.globl interrupt_wrapper
 interrupt_wrapper:
-	/* Preserve registers (excluding already-saved %eax and
-	 * otherwise unused registers which are callee-save for both
-	 * 32-bit and 64-bit ABIs).
-	 */
+	/* Preserve registers (excluding already-saved %eax) */
 	pushl	%ebx
 	pushl	%ecx
 	pushl	%edx
 	pushl	%esi
 	pushl	%edi
+	pushl	%ebp
 
 	/* Expand IRQ number to whole %eax register */
 	movzbl	%al, %eax
 
 .if64 ; /* Skip transition to long mode, if applicable */
+	xorl	%edx, %edx
 	movw	%cs, %bx
 	cmpw	$LONG_CS, %bx
 	je	1f
@@ -1391,24 +1390,45 @@ interrupt_wrapper:
 
 	/* Switch to virtual addressing */
 	call	intr_to_prot
+
+	/* Pass 32-bit interrupt frame pointer in %edx */
+	movl	%esp, %edx
+	xorl	%ecx, %ecx
 .if64
 	/* Switch to long mode */
 	call	prot_to_long
 	.code64
 
-1:	/* Preserve long-mode caller-save registers */
+1:	/* Preserve long-mode registers */
 	pushq	%r8
 	pushq	%r9
 	pushq	%r10
 	pushq	%r11
+	pushq	%r12
+	pushq	%r13
+	pushq	%r14
+	pushq	%r15
 
 	/* Expand IRQ number to whole %rdi register */
 	movl	%eax, %edi
+
+	/* Pass 32-bit interrupt frame pointer (if applicable) in %rsi */
+	testl	%edx, %edx
+	je	1f
+	movl	%edx, %esi
+	addl	virt_offset, %esi
+1:
+	/* Pass 64-bit interrupt frame pointer in %rdx */
+	movq	%rsp, %rdx
 .endif
 	/* Call interrupt handler */
 	call	interrupt
 .if64
-	/* Restore long-mode caller-save registers */
+	/* Restore long-mode registers */
+	popq	%r15
+	popq	%r14
+	popq	%r13
+	popq	%r12
 	popq	%r11
 	popq	%r10
 	popq	%r9
@@ -1432,6 +1452,7 @@ interrupt_wrapper:
 	popl	%ds
 
 1:	/* Restore registers */
+	popl	%ebp
 	popl	%edi
 	popl	%esi
 	popl	%edx
diff --git a/src/arch/x86/transitions/librm_mgmt.c b/src/arch/x86/transitions/librm_mgmt.c
index 8144e767..f9e1d261 100644
--- a/src/arch/x86/transitions/librm_mgmt.c
+++ b/src/arch/x86/transitions/librm_mgmt.c
@@ -13,6 +13,7 @@ FILE_LICENCE ( GPL2_OR_LATER_OR_UBDL );
 #include <ipxe/profile.h>
 #include <realmode.h>
 #include <pic8259.h>
+#include <ipxe/shell.h>
 
 /*
  * This file provides functions for managing librm.
@@ -43,6 +44,9 @@ struct idtr64 idtr64 = {
 	.limit = ( sizeof ( idt64 ) - 1 ),
 };
 
+/** Length of stack dump */
+#define STACK_DUMP_LEN 128
+
 /** Timer interrupt profiler */
 static struct profiler timer_irq_profiler __profiler = { .name = "irq.timer" };
 
@@ -160,14 +164,100 @@ static struct profiler * interrupt_profiler ( int intr ) {
 }
 
 /**
+ * Display interrupt stack dump (for debugging)
+ *
+ * @v intr		Interrupt number
+ * @v frame32		32-bit interrupt wrapper stack frame (or NULL)
+ * @v frame64		64-bit interrupt wrapper stack frame (or NULL)
+ */
+static __attribute__ (( unused )) void
+interrupt_dump ( int intr, struct interrupt_frame32 *frame32,
+		 struct interrupt_frame64 *frame64 ) {
+	unsigned long sp;
+	void *stack;
+
+	/* Do nothing unless debugging is enabled */
+	if ( ! DBG_LOG )
+		return;
+
+	/* Print register dump */
+	if ( ( sizeof ( physaddr_t ) <= sizeof ( uint32_t ) ) || frame32 ) {
+		sp = ( frame32->esp + sizeof ( *frame32 ) -
+		       offsetof ( typeof ( *frame32 ), esp ) );
+		DBGC ( &intr, "INT%d at %04x:%08x (stack %04x:%08lx):\n",
+		       intr, frame32->cs, frame32->eip, frame32->ss, sp );
+		DBGC ( &intr, "cs = %04x  ds = %04x  es = %04x  fs = %04x  "
+		       "gs = %04x  ss = %04x\n", frame32->cs, frame32->ds,
+		       frame32->es, frame32->fs, frame32->gs, frame32->ss );
+		DBGC ( &intr, "eax = %08x  ebx = %08x  ecx = %08x  "
+		       "edx = %08x  flg = %08x\n", frame32->eax, frame32->ebx,
+		       frame32->ecx, frame32->edx, frame32->eflags );
+		DBGC ( &intr, "esi = %08x  edi = %08x  ebp = %08x  "
+		       "esp = %08lx  eip = %08x\n", frame32->esi, frame32->edi,
+		       frame32->ebp, sp, frame32->eip );
+		stack = ( ( ( void * ) frame32 ) + sizeof ( *frame32 ) );
+	} else {
+		DBGC ( &intr, "INT%d at %04llx:%016llx (stack "
+		       "%04llx:%016llx):\n", intr,
+		       ( ( unsigned long long ) frame64->cs ),
+		       ( ( unsigned long long ) frame64->rip ),
+		       ( ( unsigned long long ) frame64->ss ),
+		       ( ( unsigned long long ) frame64->rsp ) );
+		DBGC ( &intr, "rax = %016llx  rbx = %016llx  rcx = %016llx\n",
+		       ( ( unsigned long long ) frame64->rax ),
+		       ( ( unsigned long long ) frame64->rbx ),
+		       ( ( unsigned long long ) frame64->rcx ) );
+		DBGC ( &intr, "rdx = %016llx  rsi = %016llx  rdi = %016llx\n",
+		       ( ( unsigned long long ) frame64->rdx ),
+		       ( ( unsigned long long ) frame64->rsi ),
+		       ( ( unsigned long long ) frame64->rdi ) );
+		DBGC ( &intr, "rbp = %016llx  rsp = %016llx  flg = %016llx\n",
+		       ( ( unsigned long long ) frame64->rbp ),
+		       ( ( unsigned long long ) frame64->rsp ),
+		       ( ( unsigned long long ) frame64->rflags ) );
+		DBGC ( &intr, "r8  = %016llx  r9  = %016llx  r10 = %016llx\n",
+		       ( ( unsigned long long ) frame64->r8 ),
+		       ( ( unsigned long long ) frame64->r9 ),
+		       ( ( unsigned long long ) frame64->r10 ) );
+		DBGC ( &intr, "r11 = %016llx  r12 = %016llx  r13 = %016llx\n",
+		       ( ( unsigned long long ) frame64->r11 ),
+		       ( ( unsigned long long ) frame64->r12 ),
+		       ( ( unsigned long long ) frame64->r13 ) );
+		DBGC ( &intr, "r14 = %016llx  r15 = %016llx\n",
+		       ( ( unsigned long long ) frame64->r14 ),
+		       ( ( unsigned long long ) frame64->r15 ) );
+		sp = frame64->rsp;
+		stack = phys_to_virt ( sp );
+	}
+
+	/* Print stack dump */
+	DBGC_HDA ( &intr, sp, stack, STACK_DUMP_LEN );
+}
+
+/**
  * Interrupt handler
  *
  * @v intr		Interrupt number
+ * @v frame32		32-bit interrupt wrapper stack frame (or NULL)
+ * @v frame64		64-bit interrupt wrapper stack frame (or NULL)
+ * @v frame		Interrupt wrapper stack frame
  */
-void __attribute__ (( regparm ( 1 ) )) interrupt ( int intr ) {
+void __attribute__ (( regparm ( 3 ) ))
+interrupt ( int intr, struct interrupt_frame32 *frame32,
+	    struct interrupt_frame64 *frame64 ) {
 	struct profiler *profiler = interrupt_profiler ( intr );
 	uint32_t discard_eax;
 
+	/* Trap CPU exceptions if debugging is enabled.  Note that we
+	 * cannot treat INT8+ as exceptions, since we are not
+	 * permitted to rebase the PIC.
+	 */
+	if ( DBG_LOG && ( intr < IRQ_INT ( 0 ) ) ) {
+		interrupt_dump ( intr, frame32, frame64 );
+		DBG ( "CPU exception: dropping to emergency shell\n" );
+		shell();
+	}
+
 	/* Reissue interrupt in real mode */
 	profile_start ( profiler );
 	__asm__ __volatile__ ( REAL_CODE ( "movb %%al, %%cs:(1f + 1)\n\t"