diff --git a/arm9/source/patches.c b/arm9/source/patches.c
index 82a5069..eca3eeb 100644
--- a/arm9/source/patches.c
+++ b/arm9/source/patches.c
@@ -107,7 +107,6 @@ u32 installK11Extension(u8 *pos, u32 size, bool needToInitSd, u32 baseK11VA, u32
     //Please keep that in sync with the definition in k11_extension/source/main.c
     struct KExtParameters
     {
-        u32 __attribute__((aligned(0x400))) L2MMUTableFor0x40000000[256];
         u32 basePA;
         void *originalHandlers[4];
         u32 L1MMUTableAddrs[4];
@@ -136,7 +135,7 @@ u32 installK11Extension(u8 *pos, u32 size, bool needToInitSd, u32 baseK11VA, u32
 
     //Our kernel11 extension is initially loaded in VRAM
     u32 kextTotalSize = *(u32 *)0x18000020 - 0x40000000;
-    u32 dstKextPA = (ISN3DS ? 0x2E000000 : 0x26C00000) - kextTotalSize;
+    u32 dstKextPA = (ISN3DS ? 0x2E000000 : 0x26C00000) - (0x1000 + kextTotalSize);
 
     u32 *hookVeneers = (u32 *)*freeK11Space;
     u32 relocBase = 0xFFFF0000 + (*freeK11Space - (u8 *)arm11ExceptionsPage);
@@ -177,7 +176,7 @@ u32 installK11Extension(u8 *pos, u32 size, bool needToInitSd, u32 baseK11VA, u32
     *off = MAKE_BRANCH_LINK(baseK11VA + ((u8 *)off - pos), relocBase + 24);
 
     struct KExtParameters *p = (struct KExtParameters *)(*(u32 *)0x18000024 - 0x40000000 + 0x18000000);
-    p->basePA = dstKextPA;
+    p->basePA = dstKextPA + 0x1000;
 
     for(u32 i = 0; i < 4; i++)
     {
diff --git a/k11_extension/Makefile b/k11_extension/Makefile
index 7145509..5279959 100644
--- a/k11_extension/Makefile
+++ b/k11_extension/Makefile
@@ -105,7 +105,7 @@ $(BUILD):
 #---------------------------------------------------------------------------------
 clean:
 	@echo clean ...
-	@rm -fr $(BUILD) $(TARGET).bin $(TARGET).elf
+	@rm -fr $(BUILD) $(TARGET).elf
 
 
 #---------------------------------------------------------------------------------
@@ -117,11 +117,7 @@ DEPENDS	:=	$(OFILES:.o=.d)
 #---------------------------------------------------------------------------------
 # main targets
 #---------------------------------------------------------------------------------
-all	:	$(OUTPUT).bin
-
-$(OUTPUT).bin	:	$(OUTPUT).elf
-	$(OBJCOPY) -S -O binary $< $@
-	@echo built ... $(notdir $@)
+all	:	$(OUTPUT).elf
 
 $(OUTPUT).elf	:	$(OFILES)
 
diff --git a/k11_extension/linker.ld b/k11_extension/linker.ld
index 9145b59..260c270 100644
--- a/k11_extension/linker.ld
+++ b/k11_extension/linker.ld
@@ -2,26 +2,26 @@ OUTPUT_FORMAT("elf32-littlearm", "elf32-bigarm", "elf32-littlearm")
 OUTPUT_ARCH(arm)
 ENTRY(_start)
 
-/* Mostly copied from https://github.com/devkitPro/buildscripts/blob/master/dkarm-eabi/crtls/3dsx.ld */
+MEMORY
+{
+    vram                : ORIGIN = 0x18000000, LENGTH = 0x18180000 - 0x18000000 /* Up to the kernel builtins. */
+    main                : ORIGIN = 0x40000000, LENGTH = 1M
+}
+
+PHDRS
+{
+    main PT_LOAD;
+}
 
 SECTIONS
 {
     PROVIDE(__start__ = 0x40000000);
-
-    . = __start__;
-    . = ALIGN(32);
-
-    .crt0 :
-    {
-        . = ALIGN(32);
-        KEEP( *(.text.start) )
-        KEEP( *(.init) )
-        . = ALIGN(4);
-    }
+    . = ABSOLUTE(__start__);
 
     .text :
     {
-        . = ALIGN(4);
+        KEEP( *(.text.start) )
+        KEEP( *(.init) )
 
         /* .text */
         *(.text)
@@ -31,12 +31,11 @@ SECTIONS
         *(.stub)
         *(.gnu.warning)
         *(.gnu.linkonce.t*)
-        . = ALIGN(4);
 
         /* .fini */
         KEEP( *(.fini) )
-        . = ALIGN(4);
-    }
+        . = ALIGN(32);
+    } >main AT>vram :main
 
     .rodata :
     {
@@ -46,64 +45,63 @@ SECTIONS
         *all.rodata*(*)
         *(.gnu.linkonce.r*)
         SORT(CONSTRUCTORS)
-        . = ALIGN(4);
-    }
+        . = ALIGN(8);
+    } >main AT>vram
 
-    .preinit_array ALIGN(4) :
+    .preinit_array :
     {
         PROVIDE (__preinit_array_start = .);
         KEEP (*(.preinit_array))
         PROVIDE (__preinit_array_end = .);
-    }
+    } >main AT>vram
 
-    .init_array ALIGN(4) :
+    .init_array :
     {
         PROVIDE (__init_array_start = .);
         KEEP (*(SORT(.init_array.*)))
         KEEP (*(.init_array))
         PROVIDE (__init_array_end = .);
-    }
+    } >main AT>vram
 
-    .fini_array ALIGN(4) :
+    .fini_array :
     {
         PROVIDE (__fini_array_start = .);
         KEEP (*(.fini_array))
         KEEP (*(SORT(.fini_array.*)))
         PROVIDE (__fini_array_end = .);
-    }
+    } >main AT>vram
 
-    .ctors ALIGN(4) :
+    .ctors :
     {
         KEEP (*crtbegin.o(.ctors)) /* MUST be first -- GCC requires it */
         KEEP (*(EXCLUDE_FILE (*crtend.o) .ctors))
         KEEP (*(SORT(.ctors.*)))
         KEEP (*(.ctors))
-    }
+        . = ALIGN(4);   /* REQUIRED. LD is flaky without it. */
+    } >main AT>vram
 
-    .dtors ALIGN(4) :
+    .dtors :
     {
         KEEP (*crtbegin.o(.dtors))
         KEEP (*(EXCLUDE_FILE (*crtend.o) .dtors))
         KEEP (*(SORT(.dtors.*)))
         KEEP (*(.dtors))
-    }
+        . = ALIGN(4);   /* REQUIRED. LD is flaky without it. */
+    } >main AT>vram
 
-    .ARM.extab : { *(.ARM.extab* .gnu.linkonce.armextab.*) }
-    __exidx_start = .;
-    ARM.exidx : { *(.ARM.exidx* .gnu.linkonce.armexidx.*) }
-    __exidx_end = .;
+    .ARM.extab : { *(.ARM.extab* .gnu.linkonce.armextab.*) __exidx_start = ABSOLUTE(.);} >main AT>vram
+    ARM.exidx : { *(.ARM.exidx* .gnu.linkonce.armexidx.*)  __exidx_end = ABSOLUTE(.);} >main AT>vram
 
     .data :
     {
         *(.data)
         *(.data.*)
-        KEEP (*(.large_patch*))
         *(.gnu.linkonce.d*)
         CONSTRUCTORS
-        . = ALIGN(4);
-    }
+        . = ALIGN(32);
+    } >main AT>vram
 
-    .bss :
+    .bss (NOLOAD) :
     {
         . = ALIGN(32);
         PROVIDE (__bss_start__ = ABSOLUTE(.));
@@ -114,8 +112,8 @@ SECTIONS
         *(COMMON)
         . = ALIGN(8);
         PROVIDE (__bss_end__ = ABSOLUTE(.));
-    }
-    PROVIDE (__end__ = ABSOLUTE(.));
+        PROVIDE (__end__ = ABSOLUTE(.));
+    } >main :NONE
 
     /* ==================
        ==== Metadata ====
diff --git a/k11_extension/source/main.c b/k11_extension/source/main.c
index e6fb3e9..bce89c2 100644
--- a/k11_extension/source/main.c
+++ b/k11_extension/source/main.c
@@ -34,7 +34,6 @@
 
 struct KExtParameters
 {
-    u32 ALIGN(0x400) L2MMUTableFor0x40000000[256];
     u32 basePA;
     void *originalHandlers[4];
     u32 L1MMUTableAddrs[4];
@@ -46,18 +45,20 @@ void relocateAndSetupMMU(u32 coreId, u32 *L1Table)
 {
     struct KExtParameters *p0 = (struct KExtParameters *)((u32)&kExtParameters - 0x40000000 + 0x18000000);
     struct KExtParameters *p = (struct KExtParameters *)((u32)&kExtParameters - 0x40000000 + p0->basePA);
+    u32 *L2Table = (u32 *)(p0->basePA - 0x1000);
 
     if(coreId == 0)
     {
         // Relocate ourselves, and clear BSS
-        // This is only OK because the jumps will be relative...
+        // This is only OK because the jumps will be relative & there's no mode switch...
         memcpy((void *)p0->basePA, (const void *)0x18000000, __bss_start__ - __start__);
         memset((u32 *)(p0->basePA + (__bss_start__ - __start__)), 0, __bss_end__ - __bss_start__);
 
         // Map the kernel ext to 0x40000000
         // 4KB extended small pages: [SYS:RW USR:-- X  TYP:NORMAL SHARED OUTER NOCACHE, INNER CACHED WB WA]
+        memset(L2Table, 0, 4 * 256);
         for(u32 offset = 0; offset < (u32)(__end__ - __start__); offset += 0x1000)
-            p->L2MMUTableFor0x40000000[offset >> 12] = (p0->basePA + offset) | 0x516;
+            L2Table[offset >> 12] = (p0->basePA + offset) | 0x516;
 
         __asm__ __volatile__ ("sev");
     }
@@ -74,7 +75,7 @@ void relocateAndSetupMMU(u32 coreId, u32 *L1Table)
             L1Table[i + (VA >> 20)] = PA | attribs;
     }
 
-    L1Table[0x40000000 >> 20] = (u32)p->L2MMUTableFor0x40000000 | 1;
+    L1Table[0x40000000 >> 20] = (u32)L2Table | 1;
 
     p->L1MMUTableAddrs[coreId] = (u32)L1Table;
 }
@@ -257,7 +258,7 @@ void main(FcramLayout *layout, KCoreContext *ctxs)
     u32 TTBCR_;
     s64 nb;
 
-    layout->systemSize -= __end__ - __start__;
+    layout->systemSize -= 0x1000 + __end__ - __start__;
     fcramLayout = *layout;
     coreCtxs = ctxs;
 
diff --git a/k11_extension/source/start.s b/k11_extension/source/start.s
index bbf9d3b..5b581c1 100644
--- a/k11_extension/source/start.s
+++ b/k11_extension/source/start.s
@@ -23,7 +23,7 @@
 @         reasonable ways as different from the original version.
 
 .section .text.start, "ax", %progbits
-.balign 32
+.balign 4
 .global _start
 _start:
     b start
diff --git a/k11_extension/source/svc.c b/k11_extension/source/svc.c
index 50ac06d..7fecfef 100644
--- a/k11_extension/source/svc.c
+++ b/k11_extension/source/svc.c
@@ -100,7 +100,7 @@ void *svcHook(u8 *pageEnd)
         case 0x17:
             if(strcmp(codeSetOfProcess(currentProcess)->processName, "pm") == 0) // only called twice in pm, by the same function
             {
-                *(vu32 *)(configPage + 0x44) += __end__ - __start__;
+                *(vu32 *)(configPage + 0x44) += 0x1000 + __end__ - __start__;
                 doingVeryShittyPmResLimitWorkaround = true;
             }
             return officialSVCs[0x17];
@@ -127,7 +127,7 @@ void *svcHook(u8 *pageEnd)
         case 0x79:
             if(doingVeryShittyPmResLimitWorkaround)
             {
-                *(vu32 *)(configPage + 0x44) -= __end__ - __start__;
+                *(vu32 *)(configPage + 0x44) -= 0x1000 + __end__ - __start__;
                 doingVeryShittyPmResLimitWorkaround = false;
             }
             return officialSVCs[0x79];