diff --git a/cortex-a-rt/link.x b/cortex-a-rt/link.x
index 0ba73fc..8ee325f 100644
--- a/cortex-a-rt/link.x
+++ b/cortex-a-rt/link.x
@@ -17,8 +17,6 @@ SECTIONS {
     .text : {
         /* The vector table must come first */
         *(.vector_table)
-        /* Our exception handling routines */
-        *(.text.handlers)
         /* Now the rest of the code */
         *(.text .text*)
     } > CODE
@@ -72,36 +70,38 @@ SECTIONS {
 }
 
 /*
-We reserve some space at the top of the RAM for our stacks. We have an IRQ stack
-and a FIQ stack, plus the remainder is our system stack.
+We reserve some space at the top of the RAM for our exception stacks. The
+remainder is our system mode stack.
 
 You must keep _stack_top and the stack sizes aligned to eight byte boundaries.
 */
 PROVIDE(_stack_top = ORIGIN(DATA) + LENGTH(DATA));
-PROVIDE(_fiq_stack_size = 0x400);
-PROVIDE(_irq_stack_size = 0x1000);
-PROVIDE(_abt_stack_size = 0x400);
 PROVIDE(_und_stack_size = 0x400);
-PROVIDE(_svc_stack_size = 0x1000);
+PROVIDE(_svc_stack_size = 0x400);
+PROVIDE(_abt_stack_size = 0x400);
+PROVIDE(_irq_stack_size = 0x400);
+PROVIDE(_fiq_stack_size = 0x400);
 
-ASSERT(_stack_top % 8 == 0, "ERROR(cortex-a-rt): top of stack is not 8-byte aligned");
-ASSERT(_fiq_stack_size % 8 == 0, "ERROR(cortex-a-rt): size of FIQ stack is not 8-byte aligned");
-ASSERT(_irq_stack_size % 8 == 0, "ERROR(cortex-a-rt): size of IRQ stack is not 8-byte aligned");
-ASSERT(_fiq_stack_size % 8 == 0, "ERROR(cortex-a-rt): size of FIQ stack is not 8-byte aligned");
-ASSERT(_abt_stack_size % 8 == 0, "ERROR(cortex-a-rt): size of ABT stack is not 8-byte aligned");
-ASSERT(_und_stack_size % 8 == 0, "ERROR(cortex-a-rt): size of UND stack is not 8-byte aligned");
-ASSERT(_svc_stack_size % 8 == 0, "ERROR(cortex-a-rt): size of SVC stack is not 8-byte aligned");
+ASSERT(_stack_top % 8 == 0, "ERROR(cortex-r-rt): top of stack is not 8-byte aligned");
+ASSERT(_und_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of UND stack is not 8-byte aligned");
+ASSERT(_svc_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of SVC stack is not 8-byte aligned");
+ASSERT(_abt_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of ABT stack is not 8-byte aligned");
+ASSERT(_irq_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of IRQ stack is not 8-byte aligned");
+ASSERT(_fiq_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of FIQ stack is not 8-byte aligned");
 
 /* Weak aliases for ASM default handlers */
+PROVIDE(_start                 =_default_start);
 PROVIDE(_asm_undefined_handler =_asm_default_undefined_handler);
+PROVIDE(_asm_svc_handler       =_asm_default_svc_handler);
 PROVIDE(_asm_prefetch_handler  =_asm_default_prefetch_handler);
 PROVIDE(_asm_abort_handler     =_asm_default_abort_handler);
+PROVIDE(_asm_irq_handler       =_asm_default_irq_handler);
 PROVIDE(_asm_fiq_handler       =_asm_default_fiq_handler);
 
 /* Weak aliases for C default handlers */
 PROVIDE(_undefined_handler     =_default_handler);
-PROVIDE(_abort_handler         =_default_handler);
+PROVIDE(_svc_handler           =_default_handler);
 PROVIDE(_prefetch_handler      =_default_handler);
+PROVIDE(_abort_handler         =_default_handler);
 PROVIDE(_irq_handler           =_default_handler);
-PROVIDE(_svc_handler           =_default_handler);
-PROVIDE(_start                 =_default_start);
+/* There is no default C-language FIQ handler */
diff --git a/cortex-a-rt/src/lib.rs b/cortex-a-rt/src/lib.rs
index ce8bae8..0d485a3 100644
--- a/cortex-a-rt/src/lib.rs
+++ b/cortex-a-rt/src/lib.rs
@@ -1,22 +1,25 @@
-//! # Run-time support for Arm Cortex-A
+//! # Run-time support for Arm Cortex-R
 //!
 //! This library implements a simple Arm vector table, suitable for getting into
-//! a Rust application running in System Mode. It also provides a reference start up method.
-//! Usually, most Cortex-A based systems will require chip specific start-up code, so the
-//! start-up method can over overriden.
-//!
-//! The default startup routine provided by this crate does not include any special handling
-//! for multi-core support because this is oftentimes implementation defined and the exact
-//! handling depends on the specific chip in use. Many implementations only
-//! run the startup routine with one core and will keep other cores in reset until they are woken
-//! up by an implementation specific mechanism. For other implementations where multi-core
-//! specific startup adaptions are necessary, the startup routine can be overwritten by the user.
+//! a Rust application running in System Mode. It also provides a reference
+//! start up method. Most Cortex-A based systems will require chip specific
+//! start-up code, so the start-up method can over overriden.
+//!
+//! The default startup routine provided by this crate does not include any
+//! special handling for multi-core support because this is oftentimes
+//! implementation defined and the exact handling depends on the specific chip
+//! in use. Many implementations only run the startup routine with one core and
+//! will keep other cores in reset until they are woken up by an implementation
+//! specific mechanism. For other implementations where multi-core specific
+//! startup adaptions are necessary, the startup routine can be overwritten by
+//! the user.
 //!
 //! ## Features
 //!
-//! - `vfp-dp`: Enables support for the double-precision VFP floating point support. If your target
-//!    CPU has this feature or support for NEON which also implies double-precision support, this
-//!    feature should be activated.
+//! - `vfp-dp`: Enables support for the double-precision VFP floating point
+//!    support. If your target CPU has this feature or support for NEON which
+//!    also implies double-precision support, this feature should be activated.
+//! - `eabi-fpu`: Enables the FPU, even if you selected a soft-float ABI target.
 //!
 //! ## Information about the Run-Time
 //!
@@ -49,7 +52,11 @@
 //! * `__sidata` - the start of the initialisation values for data, in read-only
 //!   memory. Must be 4-byte aligned.
 //!
-//! ### Functions
+//! Using our default start-up function `_default_start`, the memory between
+//! `__sbss` and `__ebss` is zeroed, and the memory between `__sdata` and
+//! `__edata` is initialised with the data found at `__sidata`.
+//!
+//! ### C-Compatible Functions
 //!
 //! * `kmain` - the `extern "C"` entry point to your application.
 //!
@@ -62,7 +69,9 @@
 //!
 //! * `_svc_handler` - an `extern "C"` function to call when an SVC Exception
 //!   occurs. Our linker script PROVIDEs a default function at
-//!   `_default_handler` but you can override it.
+//!   `_default_handler` but you can override it. Returning from this function
+//!   will cause execution to resume from the function the triggered the
+//!   exception, immediately after the SVC instruction.
 //!
 //!   Expected prototype:
 //!
@@ -73,108 +82,155 @@
 //!
 //! * `_irq_handler` - an `extern "C"` function to call when an Interrupt
 //!   occurs. Our linker script PROVIDEs a default function at
-//!   `_default_handler` but you can override it.
+//!   `_default_handler` but you can override it. Returning from this function
+//!   will cause execution to resume from the function the triggered the
+//!   exception.
 //!
 //!   Expected prototype:
 //!
 //!   ```rust
+//!   /// Upon return, the interrupt handler will end and execution
+//!   /// will continue at the interrupted instruction.
 //!   #[unsafe(no_mangle)]
 //!   extern "C" fn _irq_handler();
 //!   ```
 //!
-//! * `_undefined_handler` - an `extern "C"` function to call when an Undefined Exception
-//!   occurs. Our linker script PROVIDEs a default function at
-//!   `_default_handler` but you can override it. It will be called by the
-//!   `_asm_default_undefined_handler` unless that function is overriden as well.
+//! * `_undefined_handler` - an `extern "C"` function to call when an Undefined
+//!   Exception occurs. Our linker script PROVIDEs a default implementation at
+//!   `_default_handler` which is used if `_undefined_handler` is missing.
 //!
-//!   Expected prototype:
+//!   The expected prototype for `_undefined_handler` is either:
 //!
 //!   ```rust
+//!   /// Does not return
 //!   #[unsafe(no_mangle)]
-//!   extern "C" fn _undefined_handler(addr: usize);
+//!   extern "C" fn _undefined_handler(addr: usize) -> !;
 //!   ```
+//!   
+//!   or:
 //!
-//! * `_abort_handler` - an `extern "C"` function to call when a Data Abort Exception
-//!   occurs. Our linker script PROVIDEs a default function at
-//!   `_default_handler` but you can override it. It will be called by the
-//!   `_asm_default_abort_handler` unless that function is overriden as well.
+//!   ```rust
+//!   /// Execution will continue from the returned address.
+//!   ///
+//!   /// Return `addr` to go back and execute the faulting instruction again.
+//!   #[unsafe(no_mangle)]
+//!   extern "C" fn _undefined_handler(addr: usize) -> usize;
+//!   ```
 //!
-//!   Expected prototype:
+//! * `_abort_handler` - an `extern "C"` function to call when an Data Abort
+//!   occurs. Our linker script PROVIDEs a default implementation at
+//!   `_default_handler` which is used if `_abort_handler` is missing.
+//!
+//!   The expected prototype for `_abort_handler` is either:
 //!
 //!   ```rust
+//!   /// Does not return
 //!   #[unsafe(no_mangle)]
-//!   extern "C" fn _abort_handler(addr: usize);
+//!   extern "C" fn _abort_handler(addr: usize) -> !;
 //!   ```
+//!   
+//!   or:
 //!
-//! * `_prefetch_handler` - an `extern "C"` function to call when a Prefetch Abort Exception
-//!   occurs. Our linker script PROVIDEs a default function at
-//!   `_default_handler` but you can override it. It will be called by the
-//!   `_asm_default_prefetch_handler` unless that function is overriden as well.
+//!   ```rust
+//!   /// Execution will continue from the returned address.
+//!   ///
+//!   /// Return `addr` to go back and execute the faulting instruction again.
+//!   #[unsafe(no_mangle)]
+//!   extern "C" fn _abort_handler(addr: usize) -> usize;
+//!   ```
 //!
-//!   Expected prototype:
+//! * `_prefetch_handler` - an `extern "C"` function to call when an Prefetch
+//!   Abort occurs. Our linker script PROVIDEs a default implementation at
+//!   `_default_handler` which is used if `_prefetch_handler` is missing.
+//!
+//!   The expected prototype for `_prefetch_handler` is either:
+//!
+//!   ```rust
+//!   /// Does not return
+//!   #[unsafe(no_mangle)]
+//!   extern "C" fn _prefetch_handler(addr: usize) -> !;
+//!   ```
+//!   
+//!   or:
 //!
 //!   ```rust
+//!   /// Execution will continue from the returned address.
+//!   ///
+//!   /// Return `addr` to go back and execute the faulting instruction again.
 //!   #[unsafe(no_mangle)]
-//!   extern "C" fn _prefetch_handler(addr: usize);
+//!   extern "C" fn _prefetch_handler(addr: usize) -> usize;
 //!   ```
 //!
 //! ### ASM functions
 //!
-//! * `__start` - a Reset handler. Our linker script PROVIDEs a default function
-//!   at `_default_start` but you can override it. Most Cortex-A SoCs require
-//!   a chip specific startup for tasks like MMU initialization or chip specific
-//!   initialization routines.
-//! * `_asm_fiq_handler` - a naked function to call when a Fast Interrupt
-//!   Request (FIQ) occurs. Our linker script PROVIDEs a default function at
-//!   `_asm_default_fiq_handler` but you can override it.
+//! * `_start` - a Reset handler. Our linker script PROVIDEs a default function
+//!   at `_default_start` but you can override it. Some SoCs require a chip
+//!   specific startup for tasks like MMU initialization or chip specific
+//!   initialization routines, so if our start-up routine doesn't work for you,
+//!   supply your own `_start` function (but feel free to call our
+//!   `_default_start` as part of it).
 //! * `_asm_undefined_handler` - a naked function to call when an Undefined
 //!   Exception occurs. Our linker script PROVIDEs a default function at
-//!   `_asm_default_undefined_handler` but you can override it.
+//!   `_asm_default_undefined_handler` but you can override it. The provided
+//!   default handler will call `_undefined_handler`, saving state as required.
+//! * `_asm_svc_handler` - a naked function to call when an SVC Exception
+//!   occurs. Our linker script PROVIDEs a default function at
+//!   `_asm_default_svc_handler` but you can override it. The provided default
+//!   handler will call `_svc_handler`, saving state as required.
 //! * `_asm_prefetch_handler` - a naked function to call when a Prefetch
 //!   Exception occurs. Our linker script PROVIDEs a default function at
-//!   `_asm_default_prefetch_handler` but you can override it. The provided default
-//!   handler will perform an exception return to the faulting address.
+//!   `_asm_default_prefetch_handler` but you can override it. The provided
+//!   default handler will call `_prefetch_handler`, saving state as required.
+//!   Note that Prefetch Exceptions are handled in Abort Mode, Monitor Mode or
+//!   Hyp Mode, depending on CPU configuration. There is no Prefetch Abort mode,
+//!   so there is no Prefetch Abort Mode stack.
 //! * `_asm_abort_handler` - a naked function to call when an Abort Exception
 //!   occurs. Our linker script PROVIDEs a default function at
-//!   `_asm_default_abort_handler` but you can override it. The provided default handler
-//!   will perform an exception return to the faulting address.
+//!   `_asm_default_abort_handler` but you can override it. The provided default
+//!   handler will call `_abort_handler`, saving state as required.
+//! * `_asm_irq_handler` - a naked function to call when an Undefined Exception
+//!   occurs. Our linker script PROVIDEs a default function at
+//!   `_asm_default_irq_handler` but you can override it. The provided default
+//!   handler will call `_irq_handler`, saving state as required.
+//! * `_asm_fiq_handler` - a naked function to call when a Fast Interrupt
+//!   Request (FIQ) occurs. Our linker script PROVIDEs a default function at
+//!   `_asm_default_fiq_handler` but you can override it. The provided default
+//!   just spins forever.
 //!
-//! On start-up, the memory between `__sbss` and `__ebss` is zeroed, and the
-//! memory between `__sdata` and `__edata` is initialised with the data found at
-//! `__sidata`.
+//! ## Outputs
 //!
 //! This library produces global symbols called:
 //!
 //! * `_vector_table` - the start of the interrupt vector table
 //! * `_default_start` - the default Reset handler, that sets up some stacks and
 //!   calls an `extern "C"` function called `kmain`.
+//! * `_asm_default_undefined_handler` - assembly language trampoline that calls
+//!   `_undefined_handler`
+//! * `_asm_default_svc_handler` - assembly language trampoline that calls
+//!   `_svc_handler`
+//! * `_asm_default_prefetch_handler` - assembly language trampoline that calls
+//!   `_prefetch_handler`
+//! * `_asm_default_abort_handler` - assembly language trampoline that calls
+//!   `_abort_handler`
+//! * `_asm_default_irq_handler` - assembly language trampoline that calls
+//!   `_irq_handler`
 //! * `_asm_default_fiq_handler` - an FIQ handler that just spins
-//! * `_asm_default_handler` - an exception handler that just spins
-//! * `_asm_svc_handler` - assembly language trampoline for SVC Exceptions that
-//!   calls `_svc_handler`
-//! * `_asm_irq_handler` - assembly language trampoline for Interrupts that
-//!   calls `_irq_handler`
-//!
-//! The assembly language trampolines are required because Armv7-R (and Armv8-R)
-//! processors do not save a great deal of state on entry to an exception
-//! handler, unlike Armv7-M (and other M-Profile) processors. We must therefore
-//! save this state to the stack using assembly language, before transferring to
-//! an `extern "C"` function. We do not change modes before entering that
-//! `extern "C"` function - that's for the handler to deal with as it wishes. We
-//! supply a default handler that prints an error message to Semihosting so you
-//! know if you hit an unexpected exception. Because FIQ is often
+//! * `_default_handler` - a C compatible function that spins forever.
+//!
+//! The assembly language trampolines are required because Armv7-A processors do
+//! not save a great deal of state on entry to an exception handler, unlike
+//! Armv7-M (and other M-Profile) processors. We must therefore save this state
+//! to the stack using assembly language, before transferring to an `extern "C"`
+//! function. We do not change modes before entering that `extern "C"` function
+//! - that's for the handler to deal with as it wishes. Because FIQ is often
 //! performance-sensitive, we don't supply an FIQ trampoline; if you want to use
 //! FIQ, you have to write your own assembly routine, allowing you to preserve
 //! only whatever state is important to you.
 //!
-//! If our start-up routine doesn't work for you (e.g. if you have to initialise
-//! your memory controller before you touch RAM), supply your own `_start`
-//! function (but feel free to call our `_default_start` as part of it).
-//!
 //! ## Examples
 //!
-//! You can find example code using QEMU inside the
-//! [project repository](https://github.com/rust-embedded/cortex-ar/tree/main/examples)
+//! You can find example code using QEMU inside the [project
+//! repository](https://github.com/rust-embedded/cortex-ar/tree/main/examples)
 
 #![no_std]
 
@@ -197,8 +253,7 @@ pub extern "C" fn _default_handler() {
 // The Interrupt Vector Table, and some default assembly-language handler.
 core::arch::global_asm!(
     r#"
-    .section .vector_table
-    .align 0
+    .section .vector_table,"ax",%progbits
 
     .global _vector_table
     .type _vector_table, %function
@@ -212,20 +267,6 @@ core::arch::global_asm!(
         ldr     pc, =_asm_irq_handler
         ldr     pc, =_asm_fiq_handler
     .size _vector_table, . - _vector_table
-
-    .section .text.handlers
-
-    .global _asm_default_fiq_handler
-    .type _asm_default_fiq_handler, %function
-    _asm_default_fiq_handler:
-        b       _asm_default_fiq_handler
-    .size    _asm_default_fiq_handler, . - _asm_default_fiq_handler
-
-    .global _asm_default_handler
-    .type _asm_default_handler, %function
-    _asm_default_handler:
-        b       _asm_default_handler
-    .size _asm_default_handler, . - _asm_default_handler
     "#
 );
 
@@ -376,57 +417,13 @@ macro_rules! restore_context {
 // Our assembly language exception handlers
 core::arch::global_asm!(
     r#"
-    .section .text.handlers
-    .align 0
-
-    // Called from the vector table when we have an software interrupt.
-    // Saves state and calls a C-compatible handler like
-    // `extern "C" fn svc_handler(svc: u32);`
-    .global _asm_svc_handler
-    .type _asm_svc_handler, %function
-    _asm_svc_handler:
-        srsfd   sp!, {svc_mode}
-    "#,
-    save_context!(),
-    r#"
-        mrs      r0, cpsr                 // Load processor status
-        tst      r0, {t_bit}              // Occurred in Thumb state?
-        ldrhne   r0, [lr,#-2]             // Yes: Load halfword and...
-        bicne    r0, r0, #0xFF00          // ...extract comment field
-        ldreq    r0, [lr,#-4]             // No: Load word and...
-        biceq    r0, r0, #0xFF000000      // ...extract comment field
-        // r0 now contains SVC number
-        bl       _svc_handler
-    "#,
-    restore_context!(),
-    r#"
-        rfefd   sp!
-    .size _asm_svc_handler, . - _asm_svc_handler
-
-
-    // Called from the vector table when we have an interrupt.
-    // Saves state and calls a C-compatible handler like
-    // `extern "C" fn irq_handler();`
-    .global _asm_irq_handler
-    .type _asm_irq_handler, %function
-    _asm_irq_handler:
-        sub     lr, lr, 4
-        srsfd   sp!, {irq_mode}
-    "#,
-    save_context!(),
-    r#"
-        // call C handler
-        bl      _irq_handler
-    "#,
-    restore_context!(),
-    r#"
-        rfefd   sp!
-    .size _asm_irq_handler, . - _asm_irq_handler
-
+    .section .text._asm_default_undefined_handler
 
     // Called from the vector table when we have an undefined exception.
     // Saves state and calls a C-compatible handler like
-    // `extern "C" fn _undefined_handler(addr: usize);`
+    // `extern "C" fn _undefined_handler(addr: usize) -> usize;`
+    // or
+    // `extern "C" fn _undefined_handler(addr: usize) -> !;`
     .global _asm_default_undefined_handler
     .type _asm_default_undefined_handler, %function
     _asm_default_undefined_handler:
@@ -454,6 +451,8 @@ core::arch::global_asm!(
         mov     r0, lr
         // call C handler
         bl      _undefined_handler
+        // if we get back here, assume they returned a new LR in r0
+        mov     lr, r0
         // do our standard restore
     "#,
     restore_context!(),
@@ -468,6 +467,64 @@ core::arch::global_asm!(
         rfefd   sp!
     .size _asm_default_undefined_handler, . - _asm_default_undefined_handler
 
+    
+    .section .text._asm_default_svc_handler
+
+    // Called from the vector table when we have an software interrupt.
+    // Saves state and calls a C-compatible handler like
+    // `extern "C" fn svc_handler(svc: u32);`
+    .global _asm_default_svc_handler
+    .type _asm_default_svc_handler, %function
+    _asm_default_svc_handler:
+        srsfd   sp!, {svc_mode}
+    "#,
+    save_context!(),
+    r#"
+        mrs      r0, cpsr                 // Load processor status
+        tst      r0, {t_bit}              // Occurred in Thumb state?
+        ldrhne   r0, [lr,#-2]             // Yes: Load halfword and...
+        bicne    r0, r0, #0xFF00          // ...extract comment field
+        ldreq    r0, [lr,#-4]             // No: Load word and...
+        biceq    r0, r0, #0xFF000000      // ...extract comment field
+        // r0 now contains SVC number
+        bl       _svc_handler
+    "#,
+    restore_context!(),
+    r#"
+        rfefd   sp!
+    .size _asm_default_svc_handler, . - _asm_default_svc_handler
+
+
+    .section .text._asm_default_abort_handler
+
+    // Called from the vector table when we have an undefined exception.
+    // Saves state and calls a C-compatible handler like
+    // `extern "C" fn _abort_handler(addr: usize);`
+    .global _asm_default_abort_handler
+    .type _asm_default_abort_handler, %function
+    _asm_default_abort_handler:
+        // Subtract 8 from the stored LR, see p.1214 of the ARMv7-A architecture manual.
+        subs    lr, lr, #8
+        // state save from compiled code
+        srsfd   sp!, {abt_mode}
+    "#,
+    save_context!(),
+    r#"
+        // Pass the faulting instruction address to the handler.
+        mov     r0, lr
+        // call C handler
+        bl      _abort_handler
+        // if we get back here, assume they returned a new LR in r0
+        mov     lr, r0
+    "#,
+    restore_context!(),
+    r#"
+        // Return to the failing instruction which is the recommended approach by ARM.
+        rfefd   sp!
+    .size _asm_default_abort_handler, . - _asm_default_abort_handler
+
+
+    .section .text._asm_default_prefetch_handler
 
     // Called from the vector table when we have a prefetch exception.
     // Saves state and calls a C-compatible handler like
@@ -486,6 +543,8 @@ core::arch::global_asm!(
         mov     r0, lr
         // call C handler
         bl      _prefetch_handler
+        // if we get back here, assume they returned a new LR in r0
+        mov     lr, r0
     "#,
     restore_context!(),
     r#"
@@ -494,29 +553,36 @@ core::arch::global_asm!(
     .size _asm_default_prefetch_handler, . - _asm_default_prefetch_handler
 
 
-    // Called from the vector table when we have an undefined exception.
+    .section .text._asm_default_irq_handler
+
+    // Called from the vector table when we have an interrupt.
     // Saves state and calls a C-compatible handler like
-    // `extern "C" fn _abort_handler(addr: usize);`
-    .global _asm_default_abort_handler
-    .type _asm_default_abort_handler, %function
-    _asm_default_abort_handler:
-        // Subtract 8 from the stored LR, see p.1214 of the ARMv7-A architecture manual.
-        subs    lr, lr, #8
-        // state save from compiled code
-        srsfd   sp!, {abt_mode}
+    // `extern "C" fn irq_handler();`
+    .global _asm_default_irq_handler
+    .type _asm_default_irq_handler, %function
+    _asm_default_irq_handler:
+        sub     lr, lr, 4
+        srsfd   sp!, {irq_mode}
     "#,
     save_context!(),
     r#"
-        // Pass the faulting instruction address to the handler.
-        mov     r0, lr
         // call C handler
-        bl      _abort_handler
+        bl      _irq_handler
     "#,
     restore_context!(),
     r#"
-        // Return to the failing instruction which is the recommended approach by ARM.
         rfefd   sp!
-    .size _asm_default_abort_handler, . - _asm_default_abort_handler
+    .size _asm_default_irq_handler, . - _asm_default_irq_handler
+
+
+    .section .text._asm_default_fiq_handler
+
+    // Our default FIQ handler
+    .global _asm_default_fiq_handler
+    .type _asm_default_fiq_handler, %function
+    _asm_default_fiq_handler:
+        b       _asm_default_fiq_handler
+    .size    _asm_default_fiq_handler, . - _asm_default_fiq_handler
     "#,
     svc_mode = const ProcessorMode::Svc as u8,
     irq_mode = const ProcessorMode::Irq as u8,
@@ -560,7 +626,7 @@ macro_rules! fpu_enable {
 // We set up our stacks and `kmain` in system mode.
 core::arch::global_asm!(
     r#"
-    .section .text.startup
+    .section .text.default_start
     .align 0
 
     .global _default_start
@@ -600,7 +666,6 @@ core::arch::global_asm!(
         mrc     p15, 0, r0, c1, c0, 0
         bic     r0, #{te_bit}
         mcr     p15, 0, r0, c1, c0, 0
-
     "#,
     fpu_enable!(),
     r#"
diff --git a/cortex-r-rt/link.x b/cortex-r-rt/link.x
index a0296a2..309d1bf 100644
--- a/cortex-r-rt/link.x
+++ b/cortex-r-rt/link.x
@@ -17,8 +17,6 @@ SECTIONS {
     .text : {
         /* The vector table must come first */
         *(.vector_table)
-        /* Our exception handling routines */
-        *(.text.handlers)
         /* Now the rest of the code */
         *(.text .text*)
     } > CODE
@@ -72,36 +70,38 @@ SECTIONS {
 }
 
 /*
-We reserve some space at the top of the RAM for our stacks. We have an IRQ stack
-and a FIQ stack, plus the remainder is our system stack.
+We reserve some space at the top of the RAM for our exception stacks. The
+remainder is our system mode stack.
 
 You must keep _stack_top and the stack sizes aligned to eight byte boundaries.
 */
 PROVIDE(_stack_top = ORIGIN(DATA) + LENGTH(DATA));
-PROVIDE(_fiq_stack_size = 0x400);
-PROVIDE(_irq_stack_size = 0x1000);
-PROVIDE(_abt_stack_size = 0x400);
 PROVIDE(_und_stack_size = 0x400);
-PROVIDE(_svc_stack_size = 0x1000);
+PROVIDE(_svc_stack_size = 0x400);
+PROVIDE(_abt_stack_size = 0x400);
+PROVIDE(_irq_stack_size = 0x400);
+PROVIDE(_fiq_stack_size = 0x400);
 
 ASSERT(_stack_top % 8 == 0, "ERROR(cortex-r-rt): top of stack is not 8-byte aligned");
-ASSERT(_fiq_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of FIQ stack is not 8-byte aligned");
-ASSERT(_irq_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of IRQ stack is not 8-byte aligned");
-ASSERT(_fiq_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of FIQ stack is not 8-byte aligned");
-ASSERT(_abt_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of ABT stack is not 8-byte aligned");
 ASSERT(_und_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of UND stack is not 8-byte aligned");
 ASSERT(_svc_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of SVC stack is not 8-byte aligned");
+ASSERT(_abt_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of ABT stack is not 8-byte aligned");
+ASSERT(_irq_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of IRQ stack is not 8-byte aligned");
+ASSERT(_fiq_stack_size % 8 == 0, "ERROR(cortex-r-rt): size of FIQ stack is not 8-byte aligned");
 
 /* Weak aliases for ASM default handlers */
+PROVIDE(_start                 =_default_start);
 PROVIDE(_asm_undefined_handler =_asm_default_undefined_handler);
+PROVIDE(_asm_svc_handler       =_asm_default_svc_handler);
 PROVIDE(_asm_prefetch_handler  =_asm_default_prefetch_handler);
 PROVIDE(_asm_abort_handler     =_asm_default_abort_handler);
+PROVIDE(_asm_irq_handler       =_asm_default_irq_handler);
 PROVIDE(_asm_fiq_handler       =_asm_default_fiq_handler);
 
 /* Weak aliases for C default handlers */
 PROVIDE(_undefined_handler     =_default_handler);
-PROVIDE(_abort_handler         =_default_handler);
+PROVIDE(_svc_handler           =_default_handler);
 PROVIDE(_prefetch_handler      =_default_handler);
+PROVIDE(_abort_handler         =_default_handler);
 PROVIDE(_irq_handler           =_default_handler);
-PROVIDE(_svc_handler           =_default_handler);
-PROVIDE(_start                 =_default_start);
+/* There is no default C-language FIQ handler */
diff --git a/cortex-r-rt/src/lib.rs b/cortex-r-rt/src/lib.rs
index f4f4a15..0220903 100644
--- a/cortex-r-rt/src/lib.rs
+++ b/cortex-r-rt/src/lib.rs
@@ -1,7 +1,24 @@
 //! # Run-time support for Arm Cortex-R
 //!
 //! This library implements a simple Arm vector table, suitable for getting into
-//! a Rust application running in System Mode.
+//! a Rust application running in System Mode. It also provides a reference
+//! start up method. Most Cortex-R based systems will require chip specific
+//! start-up code, so the start-up method can over overriden.
+//!
+//! The default startup routine provided by this crate does not include any
+//! special handling for multi-core support because this is oftentimes
+//! implementation defined and the exact handling depends on the specific chip
+//! in use. Many implementations only run the startup routine with one core and
+//! will keep other cores in reset until they are woken up by an implementation
+//! specific mechanism. For other implementations where multi-core specific
+//! startup adaptions are necessary, the startup routine can be overwritten by
+//! the user.
+//!
+//! ## Features
+//!
+//! - `eabi-fpu`: Enables the FPU, even if you selected a soft-float ABI target.
+//!
+//! ## Information about the Run-Time
 //!
 //! Transferring from System Mode to User Mode (i.e. implementing an RTOS) is
 //! not handled here.
@@ -32,7 +49,11 @@
 //! * `__sidata` - the start of the initialisation values for data, in read-only
 //!   memory. Must be 4-byte aligned.
 //!
-//! ### Functions
+//! Using our default start-up function `_default_start`, the memory between
+//! `__sbss` and `__ebss` is zeroed, and the memory between `__sdata` and
+//! `__edata` is initialised with the data found at `__sidata`.
+//!
+//! ### C-Compatible Functions
 //!
 //! * `kmain` - the `extern "C"` entry point to your application.
 //!
@@ -45,7 +66,9 @@
 //!
 //! * `_svc_handler` - an `extern "C"` function to call when an SVC Exception
 //!   occurs. Our linker script PROVIDEs a default function at
-//!   `_default_handler` but you can override it.
+//!   `_default_handler` but you can override it. Returning from this function
+//!   will cause execution to resume from the function the triggered the
+//!   exception, immediately after the SVC instruction.
 //!
 //!   Expected prototype:
 //!
@@ -56,108 +79,155 @@
 //!
 //! * `_irq_handler` - an `extern "C"` function to call when an Interrupt
 //!   occurs. Our linker script PROVIDEs a default function at
-//!   `_default_handler` but you can override it.
+//!   `_default_handler` but you can override it. Returning from this function
+//!   will cause execution to resume from the function the triggered the
+//!   exception.
 //!
 //!   Expected prototype:
 //!
 //!   ```rust
+//!   /// Upon return, the interrupt handler will end and execution
+//!   /// will continue at the interrupted instruction.
 //!   #[unsafe(no_mangle)]
 //!   extern "C" fn _irq_handler();
 //!   ```
 //!
-//! * `_undefined_handler` - an `extern "C"` function to call when an Undefined Exception
-//!   occurs. Our linker script PROVIDEs a default function at
-//!   `_default_handler` but you can override it. It will be called by the
-//!   `_asm_default_undefined_handler` unless that function is overriden as well.
+//! * `_undefined_handler` - an `extern "C"` function to call when an Undefined
+//!   Exception occurs. Our linker script PROVIDEs a default implementation at
+//!   `_default_handler` which is used if `_undefined_handler` is missing.
 //!
-//!   Expected prototype:
+//!   The expected prototype for `_undefined_handler` is either:
 //!
 //!   ```rust
+//!   /// Does not return
 //!   #[unsafe(no_mangle)]
-//!   extern "C" fn _undefined_handler(addr: usize);
+//!   extern "C" fn _undefined_handler(addr: usize) -> !;
 //!   ```
+//!   
+//!   or:
 //!
-//! * `_abort_handler` - an `extern "C"` function to call when a Data Abort Exception
-//!   occurs. Our linker script PROVIDEs a default function at
-//!   `_default_handler` but you can override it. It will be called by the
-//!   `_asm_default_abort_handler` unless that function is overriden as well.
+//!   ```rust
+//!   /// Execution will continue from the returned address.
+//!   ///
+//!   /// Return `addr` to go back and execute the faulting instruction again.
+//!   #[unsafe(no_mangle)]
+//!   extern "C" fn _undefined_handler(addr: usize) -> usize;
+//!   ```
 //!
-//!   Expected prototype:
+//! * `_abort_handler` - an `extern "C"` function to call when an Data Abort
+//!   occurs. Our linker script PROVIDEs a default implementation at
+//!   `_default_handler` which is used if `_abort_handler` is missing.
+//!
+//!   The expected prototype for `_abort_handler` is either:
 //!
 //!   ```rust
+//!   /// Does not return
 //!   #[unsafe(no_mangle)]
-//!   extern "C" fn _abort_handler(addr: usize);
+//!   extern "C" fn _abort_handler(addr: usize) -> !;
 //!   ```
+//!   
+//!   or:
 //!
-//! * `_prefetch_handler` - an `extern "C"` function to call when a Prefetch Abort Exception
-//!   occurs. Our linker script PROVIDEs a default function at
-//!   `_default_handler` but you can override it. It will be called by the
-//!   `_asm_default_prefetch_handler` unless that function is overriden as well.
+//!   ```rust
+//!   /// Execution will continue from the returned address.
+//!   ///
+//!   /// Return `addr` to go back and execute the faulting instruction again.
+//!   #[unsafe(no_mangle)]
+//!   extern "C" fn _abort_handler(addr: usize) -> usize;
+//!   ```
 //!
-//!   Expected prototype:
+//! * `_prefetch_handler` - an `extern "C"` function to call when an Prefetch
+//!   Abort occurs. Our linker script PROVIDEs a default implementation at
+//!   `_default_handler` which is used if `_prefetch_handler` is missing.
+//!
+//!   The expected prototype for `_prefetch_handler` is either:
 //!
 //!   ```rust
+//!   /// Does not return
 //!   #[unsafe(no_mangle)]
-//!   extern "C" fn _prefetch_handler(addr: usize);
+//!   extern "C" fn _prefetch_handler(addr: usize) -> !;
+//!   ```
+//!   
+//!   or:
+//!
+//!   ```rust
+//!   /// Execution will continue from the returned address.
+//!   ///
+//!   /// Return `addr` to go back and execute the faulting instruction again.
+//!   #[unsafe(no_mangle)]
+//!   extern "C" fn _prefetch_handler(addr: usize) -> usize;
 //!   ```
 //!
 //! ### ASM functions
 //!
-//! * `__start` - a Reset handler. Our linker script PROVIDEs a default function
-//!   at `_default_start` but you can override it. Most Cortex-A SoCs require
-//!   a chip specific startup for tasks like MMU initialization or chip specific
-//!   initialization routines.
-//! * `_asm_fiq_handler` - a naked function to call when a Fast Interrupt
-//!   Request (FIQ) occurs. Our linker script PROVIDEs a default function at
-//!   `_asm_default_fiq_handler` but you can override it.
+//! * `_start` - a Reset handler. Our linker script PROVIDEs a default function
+//!   at `_default_start` but you can override it. Some SoCs require a chip
+//!   specific startup for tasks like MMU initialization or chip specific
+//!   initialization routines, so if our start-up routine doesn't work for you,
+//!   supply your own `_start` function (but feel free to call our
+//!   `_default_start` as part of it).
 //! * `_asm_undefined_handler` - a naked function to call when an Undefined
 //!   Exception occurs. Our linker script PROVIDEs a default function at
-//!   `_asm_default_undefined_handler` but you can override it.
+//!   `_asm_default_undefined_handler` but you can override it. The provided
+//!   default handler will call `_undefined_handler`, saving state as required.
+//! * `_asm_svc_handler` - a naked function to call when an SVC Exception
+//!   occurs. Our linker script PROVIDEs a default function at
+//!   `_asm_default_svc_handler` but you can override it. The provided default
+//!   handler will call `_svc_handler`, saving state as required.
 //! * `_asm_prefetch_handler` - a naked function to call when a Prefetch
 //!   Exception occurs. Our linker script PROVIDEs a default function at
-//!   `_asm_default_prefetch_handler` but you can override it. The provided default
-//!   handler will perform an exception return to the faulting address.
+//!   `_asm_default_prefetch_handler` but you can override it. The provided
+//!   default handler will call `_prefetch_handler`, saving state as required.
+//!   Note that Prefetch Exceptions are handled in Abort Mode, Monitor Mode or
+//!   Hyp Mode, depending on CPU configuration. There is no Prefetch Abort mode,
+//!   so there is no Prefetch Abort Mode stack.
 //! * `_asm_abort_handler` - a naked function to call when an Abort Exception
 //!   occurs. Our linker script PROVIDEs a default function at
-//!   `_asm_default_abort_handler` but you can override it. The provided default handler
-//!   will perform an exception return to the faulting address.
+//!   `_asm_default_abort_handler` but you can override it. The provided default
+//!   handler will call `_abort_handler`, saving state as required.
+//! * `_asm_irq_handler` - a naked function to call when an Undefined Exception
+//!   occurs. Our linker script PROVIDEs a default function at
+//!   `_asm_default_irq_handler` but you can override it. The provided default
+//!   handler will call `_irq_handler`, saving state as required.
+//! * `_asm_fiq_handler` - a naked function to call when a Fast Interrupt
+//!   Request (FIQ) occurs. Our linker script PROVIDEs a default function at
+//!   `_asm_default_fiq_handler` but you can override it. The provided default
+//!   just spins forever.
 //!
-//! On start-up, the memory between `__sbss` and `__ebss` is zeroed, and the
-//! memory between `__sdata` and `__edata` is initialised with the data found at
-//! `__sidata`.
+//! ## Outputs
 //!
 //! This library produces global symbols called:
 //!
 //! * `_vector_table` - the start of the interrupt vector table
 //! * `_default_start` - the default Reset handler, that sets up some stacks and
 //!   calls an `extern "C"` function called `kmain`.
+//! * `_asm_default_undefined_handler` - assembly language trampoline that calls
+//!   `_undefined_handler`
+//! * `_asm_default_svc_handler` - assembly language trampoline that calls
+//!   `_svc_handler`
+//! * `_asm_default_prefetch_handler` - assembly language trampoline that calls
+//!   `_prefetch_handler`
+//! * `_asm_default_abort_handler` - assembly language trampoline that calls
+//!   `_abort_handler`
+//! * `_asm_default_irq_handler` - assembly language trampoline that calls
+//!   `_irq_handler`
 //! * `_asm_default_fiq_handler` - an FIQ handler that just spins
-//! * `_asm_default_handler` - an exception handler that just spins
-//! * `_asm_svc_handler` - assembly language trampoline for SVC Exceptions that
-//!   calls `_svc_handler`
-//! * `_asm_irq_handler` - assembly language trampoline for Interrupts that
-//!   calls `_irq_handler`
+//! * `_default_handler` - a C compatible function that spins forever.
 //!
 //! The assembly language trampolines are required because Armv7-R (and Armv8-R)
 //! processors do not save a great deal of state on entry to an exception
 //! handler, unlike Armv7-M (and other M-Profile) processors. We must therefore
 //! save this state to the stack using assembly language, before transferring to
 //! an `extern "C"` function. We do not change modes before entering that
-//! `extern "C"` function - that's for the handler to deal with as it wishes. We
-//! supply a default handler that prints an error message to Semihosting so you
-//! know if you hit an unexpected exception. Because FIQ is often
-//! performance-sensitive, we don't supply an FIQ trampoline; if you want to use
-//! FIQ, you have to write your own assembly routine, allowing you to preserve
-//! only whatever state is important to you.
-//!
-//! If our start-up routine doesn't work for you (e.g. if you have to initialise
-//! your memory controller before you touch RAM), supply your own `_start`
-//! function (but feel free to call our `_default_start` as part of it).
+//! `extern "C"` function - that's for the handler to deal with as it wishes.
+//! Because FIQ is often performance-sensitive, we don't supply an FIQ
+//! trampoline; if you want to use FIQ, you have to write your own assembly
+//! routine, allowing you to preserve only whatever state is important to you.
 //!
 //! ## Examples
 //!
-//! You can find example code using QEMU inside the
-//! [project repository](https://github.com/rust-embedded/cortex-ar/tree/main/examples)
+//! You can find example code using QEMU inside the [project
+//! repository](https://github.com/rust-embedded/cortex-ar/tree/main/examples)
 
 #![no_std]
 
@@ -183,9 +253,7 @@ pub extern "C" fn _default_handler() {
 // The Interrupt Vector Table, and some default assembly-language handler.
 core::arch::global_asm!(
     r#"
-    .section .vector_table
-    .align 0
-
+    .section .vector_table,"ax",%progbits
     .global _vector_table
     .type _vector_table, %function
     _vector_table:
@@ -198,20 +266,6 @@ core::arch::global_asm!(
         ldr     pc, =_asm_irq_handler
         ldr     pc, =_asm_fiq_handler
     .size _vector_table, . - _vector_table
-
-    .section .text.handlers
-
-    .global _asm_default_fiq_handler
-    .type _asm_default_fiq_handler, %function
-    _asm_default_fiq_handler:
-        b       _asm_default_fiq_handler
-    .size    _asm_default_fiq_handler, . - _asm_default_fiq_handler
-
-    .global _asm_default_handler
-    .type _asm_default_handler, %function
-    _asm_default_handler:
-        b       _asm_default_handler
-    .size _asm_default_handler, . - _asm_default_handler
     "#
 );
 
@@ -306,59 +360,15 @@ macro_rules! restore_context {
 // Our assembly language exception handlers
 core::arch::global_asm!(
     r#"
-    .section .text.handlers
     // Work around https://github.com/rust-lang/rust/issues/127269
     .fpu vfp3-d16
-    .align 0
-
-    // Called from the vector table when we have an software interrupt.
-    // Saves state and calls a C-compatible handler like
-    // `extern "C" fn svc_handler(svc: u32);`
-    .global _asm_svc_handler
-    .type _asm_svc_handler, %function
-    _asm_svc_handler:
-        srsfd   sp!, {svc_mode}
-    "#,
-    save_context!(),
-    r#"
-        mrs      r0, cpsr                 // Load processor status
-        tst      r0, {t_bit}              // Occurred in Thumb state?
-        ldrhne   r0, [lr,#-2]             // Yes: Load halfword and...
-        bicne    r0, r0, #0xFF00          // ...extract comment field
-        ldreq    r0, [lr,#-4]             // No: Load word and...
-        biceq    r0, r0, #0xFF000000      // ...extract comment field
-        // r0 now contains SVC number
-        bl       _svc_handler
-    "#,
-    restore_context!(),
-    r#"
-        rfefd   sp!
-    .size _asm_svc_handler, . - _asm_svc_handler
-
-
-    // Called from the vector table when we have an interrupt.
-    // Saves state and calls a C-compatible handler like
-    // `extern "C" fn irq_handler();`
-    .global _asm_irq_handler
-    .type _asm_irq_handler, %function
-    _asm_irq_handler:
-        sub     lr, lr, 4
-        srsfd   sp!, {irq_mode}
-    "#,
-    save_context!(),
-    r#"
-        // call C handler
-        bl      _irq_handler
-    "#,
-    restore_context!(),
-    r#"
-        rfefd   sp!
-    .size _asm_irq_handler, . - _asm_irq_handler
-
-
+       
     // Called from the vector table when we have an undefined exception.
     // Saves state and calls a C-compatible handler like
-    // `extern "C" fn _undefined_handler(addr: usize);`
+    // `extern "C" fn _undefined_handler(addr: usize) -> usize;`
+    // or
+    // `extern "C" fn _undefined_handler(addr: usize) -> !;`
+    .section .text._asm_default_undefined_handler
     .global _asm_default_undefined_handler
     .type _asm_default_undefined_handler, %function
     _asm_default_undefined_handler:
@@ -386,6 +396,8 @@ core::arch::global_asm!(
         mov     r0, lr
         // call C handler
         bl      _undefined_handler
+        // if we get back here, assume they returned a new LR in r0
+        mov     lr, r0
         // do our standard restore
     "#,
     restore_context!(),
@@ -401,6 +413,64 @@ core::arch::global_asm!(
     .size _asm_default_undefined_handler, . - _asm_default_undefined_handler
 
 
+    .section .text._asm_default_svc_handler
+
+    // Called from the vector table when we have an software interrupt.
+    // Saves state and calls a C-compatible handler like
+    // `extern "C" fn svc_handler(svc: u32);`
+    .global _asm_default_svc_handler
+    .type _asm_default_svc_handler, %function
+    _asm_default_svc_handler:
+        srsfd   sp!, {svc_mode}
+    "#,
+    save_context!(),
+    r#"
+        mrs      r0, cpsr                 // Load processor status
+        tst      r0, {t_bit}              // Occurred in Thumb state?
+        ldrhne   r0, [lr,#-2]             // Yes: Load halfword and...
+        bicne    r0, r0, #0xFF00          // ...extract comment field
+        ldreq    r0, [lr,#-4]             // No: Load word and...
+        biceq    r0, r0, #0xFF000000      // ...extract comment field
+        // r0 now contains SVC number
+        bl       _svc_handler
+    "#,
+    restore_context!(),
+    r#"
+        rfefd   sp!
+    .size _asm_default_svc_handler, . - _asm_default_svc_handler
+
+
+    .section .text._asm_default_abort_handler
+
+    // Called from the vector table when we have an undefined exception.
+    // Saves state and calls a C-compatible handler like
+    // `extern "C" fn _abort_handler(addr: usize);`
+    .global _asm_default_abort_handler
+    .type _asm_default_abort_handler, %function
+    _asm_default_abort_handler:
+        // Subtract 8 from the stored LR, see p.1214 of the ARMv7-A architecture manual.
+        subs    lr, lr, #8
+        // state save from compiled code
+        srsfd   sp!, {abt_mode}
+    "#,
+    save_context!(),
+    r#"
+        // Pass the faulting instruction address to the handler.
+        mov     r0, lr
+        // call C handler
+        bl      _abort_handler
+        // if we get back here, assume they returned a new LR in r0
+        mov     lr, r0
+    "#,
+    restore_context!(),
+    r#"
+        // Return to the failing instruction which is the recommended approach by ARM.
+        rfefd   sp!
+    .size _asm_default_abort_handler, . - _asm_default_abort_handler
+
+
+    .section .text._asm_default_prefetch_handler
+
     // Called from the vector table when we have a prefetch exception.
     // Saves state and calls a C-compatible handler like
     // `extern "C" fn _prefetch_handler(addr: usize);`
@@ -418,6 +488,8 @@ core::arch::global_asm!(
         mov     r0, lr
         // call C handler
         bl      _prefetch_handler
+        // if we get back here, assume they returned a new LR in r0
+        mov     lr, r0
     "#,
     restore_context!(),
     r#"
@@ -426,29 +498,36 @@ core::arch::global_asm!(
     .size _asm_default_prefetch_handler, . - _asm_default_prefetch_handler
 
 
-    // Called from the vector table when we have an undefined exception.
+    .section .text._asm_default_irq_handler
+
+    // Called from the vector table when we have an interrupt.
     // Saves state and calls a C-compatible handler like
-    // `extern "C" fn _abort_handler(addr: usize);`
-    .global _asm_default_abort_handler
-    .type _asm_default_abort_handler, %function
-    _asm_default_abort_handler:
-        // Subtract 8 from the stored LR, see p.1214 of the ARMv7-A architecture manual.
-        subs    lr, lr, #8
-        // state save from compiled code
-        srsfd   sp!, {abt_mode}
+    // `extern "C" fn irq_handler();`
+    .global _asm_default_irq_handler
+    .type _asm_default_irq_handler, %function
+    _asm_default_irq_handler:
+        sub     lr, lr, 4
+        srsfd   sp!, {irq_mode}
     "#,
     save_context!(),
     r#"
-        // Pass the faulting instruction address to the handler.
-        mov     r0, lr
         // call C handler
-        bl      _abort_handler
+        bl      _irq_handler
     "#,
     restore_context!(),
     r#"
-        // Return to the failing instruction which is the recommended approach by ARM.
         rfefd   sp!
-    .size _asm_default_abort_handler, . - _asm_default_abort_handler
+    .size _asm_default_irq_handler, . - _asm_default_irq_handler
+
+
+    .section .text._asm_default_fiq_handler
+
+    // Our default FIQ handler
+    .global _asm_default_fiq_handler
+    .type _asm_default_fiq_handler, %function
+    _asm_default_fiq_handler:
+        b       _asm_default_fiq_handler
+    .size    _asm_default_fiq_handler, . - _asm_default_fiq_handler
     "#,
     svc_mode = const ProcessorMode::Svc as u8,
     irq_mode = const ProcessorMode::Irq as u8,
@@ -456,8 +535,8 @@ core::arch::global_asm!(
     abt_mode = const ProcessorMode::Abt as u8,
     t_bit = const {
         Cpsr::new_with_raw_value(0)
-            .with_t(true)
-            .raw_value()
+        .with_t(true)
+        .raw_value()
     },
 );
 
@@ -492,11 +571,10 @@ macro_rules! fpu_enable {
 // We set up our stacks and `kmain` in system mode.
 core::arch::global_asm!(
     r#"
-    .section .text.startup
-    .align 0
     // Work around https://github.com/rust-lang/rust/issues/127269
     .fpu vfp3-d16
-
+    
+    .section .text.el1_start
     .type _el1_start, %function
     _el1_start:
         // Set up stacks.
@@ -618,9 +696,10 @@ core::arch::global_asm!(
 #[cfg(arm_architecture = "v7-r")]
 core::arch::global_asm!(
     r#"
-    .section .text.startup
-    .align 0
+    // Work around https://github.com/rust-lang/rust/issues/127269
+    .fpu vfp3-d16
 
+    .section .text.default_start
     .global _default_start
     .type _default_start, %function
     _default_start:
@@ -638,8 +717,10 @@ core::arch::global_asm!(
 #[cfg(arm_architecture = "v8-r")]
 core::arch::global_asm!(
     r#"
-    .section .text.startup
-    .align 0
+    // Work around https://github.com/rust-lang/rust/issues/127269
+    .fpu vfp3-d16
+
+    .section .text.default_start
 
     .global _default_start
     .type _default_start, %function
diff --git a/examples/mps3-an536/src/bin/abt-exception.rs b/examples/mps3-an536/src/bin/abt-exception.rs
index 4a0955c..ddf599c 100644
--- a/examples/mps3-an536/src/bin/abt-exception.rs
+++ b/examples/mps3-an536/src/bin/abt-exception.rs
@@ -56,17 +56,17 @@ fn disable_alignment_check() {
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _undefined_handler(_addr: u32) {
+unsafe extern "C" fn _undefined_handler(_addr: u32) -> ! {
     panic!("unexpected undefined exception");
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _prefetch_handler(_addr: u32) {
+unsafe extern "C" fn _prefetch_handler(_addr: u32) -> ! {
     panic!("unexpected prefetch exception");
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _abort_handler(_addr: u32) {
+unsafe extern "C" fn _abort_handler(addr: usize) -> usize {
     println!("data abort occurred");
     // If this is not disabled, reading DFAR will trigger an alignment fault on Armv8-R, leading
     // to a loop.
@@ -82,4 +82,6 @@ unsafe extern "C" fn _abort_handler(_addr: u32) {
     if COUNTER.fetch_add(1, core::sync::atomic::Ordering::Relaxed) == 1 {
         semihosting::process::exit(0);
     }
+
+    addr
 }
diff --git a/examples/mps3-an536/src/bin/prefetch-exception-a32.rs b/examples/mps3-an536/src/bin/prefetch-exception-a32.rs
index ced3ecd..b561ce1 100644
--- a/examples/mps3-an536/src/bin/prefetch-exception-a32.rs
+++ b/examples/mps3-an536/src/bin/prefetch-exception-a32.rs
@@ -51,12 +51,12 @@ core::arch::global_asm!(
 );
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _undefined_handler(_addr: usize) {
+unsafe extern "C" fn _undefined_handler(_addr: usize) -> ! {
     panic!("unexpected undefined exception");
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _prefetch_handler(addr: usize) {
+unsafe extern "C" fn _prefetch_handler(addr: usize) -> usize {
     println!("prefetch abort occurred");
     let ifsr = Ifsr::read();
     println!("IFSR (Fault Status Register): {:?}", ifsr);
@@ -77,9 +77,11 @@ unsafe extern "C" fn _prefetch_handler(addr: usize) {
         // we've faulted twice - time to quit
         semihosting::process::exit(0);
     }
+
+    addr
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _abort_handler(_addr: usize) {
+unsafe extern "C" fn _abort_handler(_addr: usize) -> ! {
     panic!("unexpected abort exception");
 }
diff --git a/examples/mps3-an536/src/bin/prefetch-exception-t32.rs b/examples/mps3-an536/src/bin/prefetch-exception-t32.rs
index cc03404..12249cb 100644
--- a/examples/mps3-an536/src/bin/prefetch-exception-t32.rs
+++ b/examples/mps3-an536/src/bin/prefetch-exception-t32.rs
@@ -51,12 +51,12 @@ core::arch::global_asm!(
 );
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _undefined_handler(_addr: usize) {
+unsafe extern "C" fn _undefined_handler(_addr: usize) -> ! {
     panic!("unexpected undefined exception");
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _prefetch_handler(addr: usize) {
+unsafe extern "C" fn _prefetch_handler(addr: usize) -> usize {
     println!("prefetch abort occurred");
     let ifsr = Ifsr::read();
     println!("IFSR (Fault Status Register): {:?}", ifsr);
@@ -80,9 +80,11 @@ unsafe extern "C" fn _prefetch_handler(addr: usize) {
         // we've faulted twice - time to quit
         semihosting::process::exit(0);
     }
+
+    addr
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _abort_handler(_addr: usize) {
+unsafe extern "C" fn _abort_handler(_addr: usize) -> ! {
     panic!("unexpected abort exception");
 }
diff --git a/examples/mps3-an536/src/bin/undef-exception-a32.rs b/examples/mps3-an536/src/bin/undef-exception-a32.rs
index ed9b2e7..ff672f6 100644
--- a/examples/mps3-an536/src/bin/undef-exception-a32.rs
+++ b/examples/mps3-an536/src/bin/undef-exception-a32.rs
@@ -48,12 +48,12 @@ core::arch::global_asm!(
 );
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _prefetch_handler(_addr: usize) {
+unsafe extern "C" fn _prefetch_handler(_addr: usize) -> ! {
     panic!("unexpected undefined exception");
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _undefined_handler(addr: usize) {
+unsafe extern "C" fn _undefined_handler(addr: usize) -> usize {
     println!("undefined abort occurred");
 
     if addr == udf_from_a32 as usize {
@@ -72,9 +72,12 @@ unsafe extern "C" fn _undefined_handler(addr: usize) {
         // we've faulted twice - time to quit
         semihosting::process::exit(0);
     }
+
+    // otherwise go back to where we came from
+    addr
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _abort_handler(_addr: usize) {
+unsafe extern "C" fn _abort_handler(_addr: usize) -> ! {
     panic!("unexpected abort exception");
 }
diff --git a/examples/mps3-an536/src/bin/undef-exception-t32.rs b/examples/mps3-an536/src/bin/undef-exception-t32.rs
index 75bd43d..1cd70fd 100644
--- a/examples/mps3-an536/src/bin/undef-exception-t32.rs
+++ b/examples/mps3-an536/src/bin/undef-exception-t32.rs
@@ -48,12 +48,12 @@ core::arch::global_asm!(
 );
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _prefetch_handler(_addr: usize) {
+unsafe extern "C" fn _prefetch_handler(_addr: usize) -> ! {
     panic!("unexpected undefined exception");
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _undefined_handler(addr: usize) {
+unsafe extern "C" fn _undefined_handler(addr: usize) -> usize {
     println!("undefined abort occurred");
 
     if (addr + 1) == udf_from_t32 as usize {
@@ -72,9 +72,11 @@ unsafe extern "C" fn _undefined_handler(addr: usize) {
         // we've faulted twice - time to quit
         semihosting::process::exit(0);
     }
+
+    addr
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _abort_handler(_addr: usize) {
+unsafe extern "C" fn _abort_handler(_addr: usize) -> ! {
     panic!("unexpected abort exception");
 }
diff --git a/examples/versatileab/src/bin/abt-exception.rs b/examples/versatileab/src/bin/abt-exception.rs
index d22566b..ab03ecd 100644
--- a/examples/versatileab/src/bin/abt-exception.rs
+++ b/examples/versatileab/src/bin/abt-exception.rs
@@ -56,17 +56,17 @@ fn disable_alignment_check() {
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _undefined_handler(_addr: u32) {
+unsafe extern "C" fn _undefined_handler(_addr: u32) -> ! {
     panic!("unexpected undefined exception");
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _prefetch_handler(_addr: u32) {
+unsafe extern "C" fn _prefetch_handler(_addr: u32) -> ! {
     panic!("unexpected prefetch exception");
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _abort_handler(_addr: u32) {
+unsafe extern "C" fn _abort_handler(addr: usize) -> usize {
     println!("data abort occurred");
     let dfsr = Dfsr::read();
     println!("DFSR (Fault Status Register): {:?}", dfsr);
@@ -82,4 +82,6 @@ unsafe extern "C" fn _abort_handler(_addr: u32) {
     if COUNTER.fetch_add(1, core::sync::atomic::Ordering::Relaxed) == 1 {
         semihosting::process::exit(0);
     }
+
+    addr
 }
diff --git a/examples/versatileab/src/bin/prefetch-exception-a32.rs b/examples/versatileab/src/bin/prefetch-exception-a32.rs
index b26ebea..d773dd3 100644
--- a/examples/versatileab/src/bin/prefetch-exception-a32.rs
+++ b/examples/versatileab/src/bin/prefetch-exception-a32.rs
@@ -51,12 +51,12 @@ core::arch::global_asm!(
 );
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _undefined_handler(_addr: usize) {
+unsafe extern "C" fn _undefined_handler(_addr: usize) -> ! {
     panic!("unexpected undefined exception");
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _prefetch_handler(addr: usize) {
+unsafe extern "C" fn _prefetch_handler(addr: usize) -> usize {
     println!("prefetch abort occurred");
     let ifsr = Ifsr::read();
     println!("IFSR (Fault Status Register): {:?}", ifsr);
@@ -77,9 +77,11 @@ unsafe extern "C" fn _prefetch_handler(addr: usize) {
         // we've faulted twice - time to quit
         semihosting::process::exit(0);
     }
+
+    addr
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _abort_handler(_addr: usize) {
+unsafe extern "C" fn _abort_handler(_addr: usize) -> ! {
     panic!("unexpected abort exception");
 }
diff --git a/examples/versatileab/src/bin/prefetch-exception-t32.rs b/examples/versatileab/src/bin/prefetch-exception-t32.rs
index d8f640e..acea6c8 100644
--- a/examples/versatileab/src/bin/prefetch-exception-t32.rs
+++ b/examples/versatileab/src/bin/prefetch-exception-t32.rs
@@ -51,12 +51,12 @@ core::arch::global_asm!(
 );
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _undefined_handler(_addr: usize) {
+unsafe extern "C" fn _undefined_handler(_addr: usize) -> ! {
     panic!("unexpected undefined exception");
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _prefetch_handler(addr: usize) {
+unsafe extern "C" fn _prefetch_handler(addr: usize) -> usize {
     println!("prefetch abort occurred");
     let ifsr = Ifsr::read();
     println!("IFSR (Fault Status Register): {:?}", ifsr);
@@ -80,9 +80,11 @@ unsafe extern "C" fn _prefetch_handler(addr: usize) {
         // we've faulted twice - time to quit
         semihosting::process::exit(0);
     }
+
+    addr
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _abort_handler(_addr: usize) {
+unsafe extern "C" fn _abort_handler(_addr: usize) -> ! {
     panic!("unexpected abort exception");
 }
diff --git a/examples/versatileab/src/bin/undef-exception-a32.rs b/examples/versatileab/src/bin/undef-exception-a32.rs
index b8833eb..6679d2d 100644
--- a/examples/versatileab/src/bin/undef-exception-a32.rs
+++ b/examples/versatileab/src/bin/undef-exception-a32.rs
@@ -48,12 +48,12 @@ core::arch::global_asm!(
 );
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _prefetch_handler(_addr: usize) {
+unsafe extern "C" fn _prefetch_handler(_addr: usize) -> ! {
     panic!("unexpected undefined exception");
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _undefined_handler(addr: usize) {
+unsafe extern "C" fn _undefined_handler(addr: usize) -> usize {
     println!("undefined abort occurred");
 
     if addr == udf_from_a32 as usize {
@@ -72,9 +72,11 @@ unsafe extern "C" fn _undefined_handler(addr: usize) {
         // we've faulted twice - time to quit
         semihosting::process::exit(0);
     }
+
+    addr
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _abort_handler(_addr: usize) {
+unsafe extern "C" fn _abort_handler(_addr: usize) -> ! {
     panic!("unexpected abort exception");
 }
diff --git a/examples/versatileab/src/bin/undef-exception-t32.rs b/examples/versatileab/src/bin/undef-exception-t32.rs
index 01ee741..c9869b4 100644
--- a/examples/versatileab/src/bin/undef-exception-t32.rs
+++ b/examples/versatileab/src/bin/undef-exception-t32.rs
@@ -48,12 +48,12 @@ core::arch::global_asm!(
 );
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _prefetch_handler(_addr: usize) {
+unsafe extern "C" fn _prefetch_handler(_addr: usize) -> ! {
     panic!("unexpected undefined exception");
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _undefined_handler(addr: usize) {
+unsafe extern "C" fn _undefined_handler(addr: usize) -> usize {
     println!("undefined abort occurred");
 
     if (addr + 1) == udf_from_t32 as usize {
@@ -72,9 +72,11 @@ unsafe extern "C" fn _undefined_handler(addr: usize) {
         // we've faulted twice - time to quit
         semihosting::process::exit(0);
     }
+
+    addr
 }
 
 #[unsafe(no_mangle)]
-unsafe extern "C" fn _abort_handler(_addr: usize) {
+unsafe extern "C" fn _abort_handler(_addr: usize) -> ! {
     panic!("unexpected abort exception");
 }