diff --git a/litex/build/generic_programmer.py b/litex/build/generic_programmer.py
index b9540415c..23a9eb12f 100644
--- a/litex/build/generic_programmer.py
+++ b/litex/build/generic_programmer.py
@@ -17,7 +17,10 @@ class GenericProgrammer:
             fullname = os.path.join(fulldir, self.flash_proxy_basename)
             if os.path.exists(fullname):
                 return fullname
-        raise OSError("Failed to find flash proxy bitstream")
+        raise OSError(
+            "Failed to find flash proxy bitstream %s, searched:\n    %s\n" % (
+                self.flash_proxy_basename,
+                "\n    ".join(self.flash_proxy_dirs)))
 
     # must be overloaded by specific programmer
     def load_bitstream(self, bitstream_file):
diff --git a/litex/soc/software/bios/Makefile b/litex/soc/software/bios/Makefile
index e00e4a706..6fd21dd27 100644
--- a/litex/soc/software/bios/Makefile
+++ b/litex/soc/software/bios/Makefile
@@ -12,8 +12,8 @@ all: bios.bin
 
 bios.elf: $(BIOS_DIRECTORY)/linker.ld $(OBJECTS)
 
-%.elf:
-	$(LD) $(LDFLAGS) -T $< -N -o $@ \
+%.elf: ../libbase/crt0-$(CPU).o ../libnet/libnet.a ../libbase/libbase-nofloat.a ../libcompiler_rt/libcompiler_rt.a
+	$(LD) $(LDFLAGS) -T $(BIOS_DIRECTORY)/linker.ld -N -o $@ \
 		../libbase/crt0-$(CPU).o \
 		$(OBJECTS) \
 		-L../libnet \
diff --git a/litex/soc/software/bios/boot.c b/litex/soc/software/bios/boot.c
index e0ff8ba95..ed7d3cbaf 100644
--- a/litex/soc/software/bios/boot.c
+++ b/litex/soc/software/bios/boot.c
@@ -18,7 +18,7 @@ extern void boot_helper(unsigned int r1, unsigned int r2, unsigned int r3, unsig
 
 static void __attribute__((noreturn)) boot(unsigned int r1, unsigned int r2, unsigned int r3, unsigned int addr)
 {
-	printf("Executing booted program.\n");
+	printf("Executing booted program at 0x%08x\n", addr);
 	uart_sync();
 	irq_setmask(0);
 	irq_setie(0);
diff --git a/litex/soc/software/common.mak b/litex/soc/software/common.mak
index 7ef85aae7..11e8fc385 100644
--- a/litex/soc/software/common.mak
+++ b/litex/soc/software/common.mak
@@ -42,7 +42,7 @@ DEPFLAGS += -MD -MP
 # Toolchain options
 #
 INCLUDES = -I$(SOC_DIRECTORY)/software/include/base -I$(SOC_DIRECTORY)/software/include -I$(SOC_DIRECTORY)/common -I$(BUILDINC_DIRECTORY)
-COMMONFLAGS = $(DEPFLAGS) -Os $(CPUFLAGS) -fomit-frame-pointer -Wall -fno-builtin -nostdinc $(INCLUDES)
+COMMONFLAGS = $(DEPFLAGS) -Os $(CPUFLAGS) -g3 -fomit-frame-pointer -Wall -fno-builtin -nostdinc $(INCLUDES)
 CFLAGS = $(COMMONFLAGS) -fexceptions -Wstrict-prototypes -Wold-style-definition -Wmissing-prototypes
 CXXFLAGS = $(COMMONFLAGS) -std=c++11 -I$(SOC_DIRECTORY)/software/include/basec++ -fexceptions -fno-rtti -ffreestanding
 LDFLAGS = -nostdlib -nodefaultlibs -L$(BUILDINC_DIRECTORY)
diff --git a/litex/soc/software/libbase/crt0-or1k.S b/litex/soc/software/libbase/crt0-or1k.S
index 5db4142af..ff77910dc 100644
--- a/litex/soc/software/libbase/crt0-or1k.S
+++ b/litex/soc/software/libbase/crt0-or1k.S
@@ -19,7 +19,18 @@
 
 #include <spr-defs.h>
 
-#define EXCEPTION_STACK_SIZE (4*32)
+/*
+ * OR1K Architecture has a 128 byte "red zone" after the stack that can not be
+ * touched by exception handlers. GCC uses this red zone for locals and
+ * temps without needing to change the stack pointer.
+ */
+#define OR1K_RED_ZONE_SIZE	128
+
+/*
+ * We need 4 bytes (32 bits) * 32 registers space on the stack to save all the
+ * registers.
+ */
+#define EXCEPTION_STACK_SIZE	((4*32) + OR1K_RED_ZONE_SIZE)
 
 #define HANDLE_EXCEPTION                    ; \
     l.addi  r1, r1, -EXCEPTION_STACK_SIZE   ; \