diff --git a/src/arch/riscv/entry.rs b/src/arch/riscv/entry.rs
index 854bb68..16d75c4 100644
--- a/src/arch/riscv/entry.rs
+++ b/src/arch/riscv/entry.rs
@@ -6,7 +6,7 @@ unsafe extern "C" fn _start(hart_id: usize, device_tree_addr: usize) -> ! {
     // no stack here
 
     const STACK_SIZE: usize = 4096 * 16;
-    #[link_section = ".boot_stack"]
+    #[link_section = ".bss.boot_stack"]
     static mut STACK: [u8; STACK_SIZE] = [0u8; STACK_SIZE];
 
     core::arch::asm!(
@@ -20,6 +20,8 @@ unsafe extern "C" fn _start(hart_id: usize, device_tree_addr: usize) -> ! {
 }
 
 extern "C" fn pre_main(hart_id: usize, device_tree_addr: usize) -> ! {
+    // TODO: multiboot
+
     unsafe { zero_bss() }
 
     // TODO: initialize page table
@@ -28,16 +30,14 @@ extern "C" fn pre_main(hart_id: usize, device_tree_addr: usize) -> ! {
 }
 
 extern "C" {
-    static mut __bss_start: u8;
+    static mut __boot_stack_end: u8;
     static mut __bss_end: u8;
 }
 
+#[inline(always)]
 unsafe fn zero_bss() {
-    let mut cur = &mut __bss_start as *mut u8;
+    let cur = &mut __boot_stack_end as *mut u8;
     let end = &mut __bss_end as *mut u8;
 
-    while cur < end {
-        cur.write_volatile(0);
-        cur = cur.add(1);
-    }
+    core::slice::from_raw_parts_mut(cur, end.offset_from(cur) as usize).fill(0);
 }
diff --git a/src/arch/riscv/linker.ld b/src/arch/riscv/linker.ld
index fee37f6..38fe63c 100644
--- a/src/arch/riscv/linker.ld
+++ b/src/arch/riscv/linker.ld
@@ -42,14 +42,12 @@ SECTIONS {
         __data_end = .;
     } > DRAM
 
-    .stack : {
-        . = ALIGN(8);
-        *(.boot_stack)
-    } > DRAM
-
     .bss : {
         . = ALIGN(8);
         __bss_start = .;
+        *(.bss.boot_stack)
+        __boot_stack_end = .;
+
         *(.bss .bss.*)
         *(.sbss .sbss.*)
         __bss_end = .;