Use picolibc instead of newlib-nano

 1. Add picolibc bits to linker scripts. Picolibc uses thread local storage
    for unshared data, and the picolibc startup code relies on carefully
    constructed linker script that allocates one TLS block in RAM for
    applications that don't have explicit TLS support. The picolibc
    startup code also uses different symbols than the tf-m startup code.

 2. Support picolibc stdio. There was already picolibc support code
    present in the library for the ARM LLVM toolchain. The
    conditionals which selected it have been changed to use
    __PICOLIBC__ instead of __clang_major__.

 3. Add _exit stub. Code using assert or abort end up calling _exit
    through the picolibc signal handling code.

 4. Switch to picolibc.specs. This is needed for toolchains which
    don't use picolibc by default, and can also be used without
    trouble in toolchains where picolibc is the default.

Signed-off-by: Keith Packard <[email protected]>

Author: keith-packard

platform/CMakeLists.txt

@@ -90,6 +90,7 @@ target_sources(platform_s
 target_sources(tfm_s
     PRIVATE
         ext/common/faults.c
+        ext/common/picolibc.c
 )
 
 target_link_libraries(platform_s

platform/ext/common/gcc/tfm_common_bl2.ld

@@ -153,6 +153,26 @@ SECTIONS
     } > RAM AT > FLASH
     Image$$ER_DATA$$Base = ADDR(.data);
 
+    .tdata : ALIGN_WITH_INPUT
+    {
+        *(.tdata .tdata.* .gnu.linkonce.td.*)
+        PROVIDE( __data_end = . );
+        PROVIDE( __tdata_end = . );
+    } >RAM AT>FLASH
+
+    .tbss (NOLOAD) :
+    {
+        *(.tbss .tbss.* .gnu.linkonce.tb.*)
+        *(.tcommon)
+        PROVIDE( __tls_end = . );
+        PROVIDE( __tbss_end = . );
+    } >RAM AT>RAM
+
+    .tbss_space (NOLOAD) : {
+        . = ADDR(.tbss);
+        . = . + SIZEOF(.tbss);
+    } >RAM AT>RAM
+
     .bss : ALIGN(4)
     {
         . = ALIGN(4);
@@ -163,6 +183,16 @@ SECTIONS
         __bss_end__ = .;
     } > RAM
 
+    PROVIDE( __bss_end = __bss_end__ );
+    PROVIDE( __data_start = ADDR(.data) );
+    PROVIDE( __data_source = LOADADDR(.data) );
+    PROVIDE( __data_size = __data_end - __data_start );
+    PROVIDE( __tls_base = SIZEOF(.tdata) ? ADDR(.tdata) : ADDR(.tbss) );
+    PROVIDE( __tls_align = MAX(ALIGNOF(.tdata), ALIGNOF(.tbss)) );
+    PROVIDE( __arm32_tls_tcb_offset = MAX(8, __tls_align) );
+    PROVIDE( __bss_start = ADDR(.tbss));
+    PROVIDE( __bss_size = __bss_end - __bss_start );
+
 #if defined (__ARM_FEATURE_CMSE) && (__ARM_FEATURE_CMSE == 3U)
     .msp_stack (NOLOAD) : ALIGN(32)
     {
@@ -200,4 +230,6 @@ SECTIONS
     Image$$ARM_LIB_HEAP$$ZI$$Limit = ADDR(.heap) + SIZEOF(.heap);
 
     PROVIDE(__stack = Image$$ARM_LIB_STACK$$ZI$$Limit);
+    PROVIDE( __heap_start = __HeapBase );
+    PROVIDE( __heap_end = __HeapLimit );
 }

platform/ext/common/gcc/tfm_common_ns.ld

@@ -128,22 +128,51 @@ SECTIONS
 
         KEEP(*(.jcr*))
         . = ALIGN(4);
-        /* All data end */
-        __data_end__ = .;
 
     } > RAM
 
-    .bss :
+    .tdata : AT (__etext + SIZEOF(.data))
+    {
+        *(.tdata .tdata.* .gnu.linkonce.td.*)
+        PROVIDE( __data_end = . );
+        PROVIDE( __data_end__ = . );
+        PROVIDE( __tdata_end = . );
+    } > RAM
+
+    .tbss (NOLOAD) :
     {
         . = ALIGN(4);
         __bss_start__ = .;
+        *(.tbss .tbss.* .gnu.linkonce.tb.*)
+        *(.tcommon)
+        PROVIDE( __tls_end = . );
+        PROVIDE( __tbss_end = . );
+    } > RAM
+
+    .tbss_space (NOLOAD) :
+    {
+        . = ADDR(.tbss);
+        . = . + SIZEOF(.tbss);
+    } > RAM
+
+    .bss :
+    {
+        . = ALIGN(4);
         *(.bss*)
         *(COMMON)
         . = ALIGN(4);
         __bss_end__ = .;
     } > RAM
 
-    bss_size = __bss_end__ - __bss_start__;
+    PROVIDE( __bss_end = __bss_end__ );
+    PROVIDE( __data_start = ADDR(.data) );
+    PROVIDE( __data_source = LOADADDR(.data) );
+    PROVIDE( __data_size = __data_end - __data_start );
+    PROVIDE( __bss_start = ADDR(.tbss) );
+    PROVIDE( __bss_size = __bss_end - __bss_start );
+    PROVIDE( __tls_base = SIZEOF(.tdata) ? ADDR(.tdata) : ADDR(.tbss) );
+    PROVIDE( __tls_align = MAX(ALIGNOF(.tdata), ALIGNOF(.tbss)) );
+    PROVIDE( __arm32_tls_tcb_offset = MAX(8, __tls_align) );
 
     .stack : ALIGN(32)
     {
@@ -164,4 +193,6 @@ SECTIONS
     } > RAM
 
     PROVIDE(__stack = __StackTop);
+    PROVIDE( __heap_start = __HeapBase );
+    PROVIDE( __heap_end = __HeapLimit );
 }

platform/ext/common/gcc/tfm_common_s.ld.template

@@ -576,13 +576,35 @@ SECTIONS
         . = ALIGN(4);
 
     } > RAM AT> FLASH
+
+    .TFM_TDATA : ALIGN_WITH_INPUT
+    {
+        *(.tdata .tdata.* .gnu.linkonce.td.*)
+        PROVIDE( __data_end = . );
+        PROVIDE( __tdata_end = . );
+    } > RAM AT> FLASH
+
     Image$$ER_TFM_DATA$$RW$$Base = ADDR(.TFM_DATA);
-    Image$$ER_TFM_DATA$$RW$$Limit = ADDR(.TFM_DATA) + SIZEOF(.TFM_DATA);
+    Image$$ER_TFM_DATA$$RW$$Limit = ADDR(.TFM_TDATA) + SIZEOF(.TFM_TDATA);
 
-    .TFM_BSS ALIGN(4) (NOLOAD) :
+    .TFM_TBSS (NOLOAD) :
     {
         __bss_start__ = .;
 
+        *(.tbss .tbss.* .gnu.linkonce.tb.*)
+        *(.tcommon)
+        PROVIDE( __tls_end = . );
+        PROVIDE( __tbss_end = . );
+    } > RAM AT> RAM
+
+    .TFM_TBSS_SPACE (NOLOAD) :
+    {
+        . = ADDR(.TFM_TBSS);
+        . = . + SIZEOF(.TFM_TBSS);
+    } > RAM AT> RAM
+
+    .TFM_BSS ALIGN(4) (NOLOAD) :
+    {
         /* The runtime partition placed order is same as load partition */
         __partition_runtime_start__ = .;
         KEEP(*(.bss.part_runtime_priority_00))
@@ -603,16 +625,26 @@ SECTIONS
         *(COMMON)
         . = ALIGN(4);
         __bss_end__ = .;
+        __bss_end = .;
     } > RAM AT> RAM
-    Image$$ER_TFM_DATA$$ZI$$Base = ADDR(.TFM_BSS);
+    Image$$ER_TFM_DATA$$ZI$$Base = ADDR(.TFM_TBSS);
     Image$$ER_TFM_DATA$$ZI$$Limit = ADDR(.TFM_BSS) + SIZEOF(.TFM_BSS);
     Image$$ER_PART_RT_POOL$$ZI$$Base = __partition_runtime_start__;
     Image$$ER_PART_RT_POOL$$ZI$$Limit = __partition_runtime_end__;
     Image$$ER_SERV_RT_POOL$$ZI$$Base = __service_runtime_start__;
     Image$$ER_SERV_RT_POOL$$ZI$$Limit = __service_runtime_end__;
 
     Image$$ER_TFM_DATA$$Base = ADDR(.TFM_DATA);
-    Image$$ER_TFM_DATA$$Limit = ADDR(.TFM_DATA) + SIZEOF(.TFM_DATA) + SIZEOF(.TFM_BSS);
+    Image$$ER_TFM_DATA$$Limit = ADDR(.TFM_BSS) + SIZEOF(.TFM_BSS);
+
+    PROVIDE( __data_start = ADDR(.TFM_DATA) );
+    PROVIDE( __data_source = LOADADDR(.TFM_DATA) );
+    PROVIDE( __data_size = __data_end - __data_start );
+    PROVIDE( __bss_start = ADDR(.TFM_TBSS) );
+    PROVIDE( __bss_size = __bss_end - __bss_start );
+    PROVIDE( __tls_base = SIZEOF(.TFM_TDATA) ? ADDR(.TFM_TDATA) : ADDR(.TFM_TBSS) );
+    PROVIDE( __tls_align = MAX(ALIGNOF(.TFM_TDATA), ALIGNOF(.TFM_TBSS)) );
+    PROVIDE(__arm32_tls_tcb_offset = MAX(8, __tls_align) );
 
 #if defined(CONFIG_TFM_USE_TRUSTZONE)
     Image$$ER_VENEER$$Base = ADDR(.gnu.sgstubs);

platform/ext/common/gcc/tfm_isolation_s.ld.template

@@ -210,20 +210,22 @@ SECTIONS
         . = ALIGN(4);
         /* preinit data */
         PROVIDE_HIDDEN (__preinit_array_start = .);
+        PROVIDE_HIDDEN (__bothinit_array_start = .);
         KEEP(*(.preinit_array))
         PROVIDE_HIDDEN (__preinit_array_end = .);
 
         . = ALIGN(4);
         /* init data */
         PROVIDE_HIDDEN (__init_array_start = .);
-        KEEP(*(SORT(.init_array.*)))
+        KEEP(*(SORT_BY_INIT_PRIORITY(.init_array.*)))
         KEEP(*(.init_array))
         PROVIDE_HIDDEN (__init_array_end = .);
+        PROVIDE_HIDDEN (__bothinit_array_end = .);
 
         . = ALIGN(4);
         /* finit data */
         PROVIDE_HIDDEN (__fini_array_start = .);
-        KEEP(*(SORT(.fini_array.*)))
+        KEEP(*(SORT_BY_INIT_PRIORITY(.fini_array.*)))
         KEEP(*(.fini_array))
         PROVIDE_HIDDEN (__fini_array_end = .);
 
@@ -251,6 +253,10 @@ SECTIONS
         LONG (ADDR(.ER_CODE_SRAM))
         LONG (SIZEOF(.ER_CODE_SRAM) / 4)
 #endif
+        LONG (LOADADDR(.TFM_TDATA))
+        LONG (ADDR(.TFM_TLS_BASE))
+        LONG (SIZEOF(.TFM_TDATA) / 4)
+
         __copy_table_end__ = .;
 
         /* .zero.table */
@@ -265,6 +271,10 @@ SECTIONS
         LONG (ADDR(.TFM_SP_META_PTR))
         LONG (SIZEOF(.TFM_SP_META_PTR) / 4)
 #endif
+
+        LONG (ADDR(.TFM_TLS_BASE) + SIZEOF(.TFM_TDATA))
+        LONG ((SIZEOF(.TFM_TLS_BASE) - SIZEOF(.TFM_TDATA)) / 4)
+
         __zero_table_end__ = .;
 
         /* Capture all remaining code (except RAM code) in the privileged TF-M
@@ -552,6 +562,10 @@ SECTIONS
         . = ALIGN(4);
     } > RAM AT > FLASH
 
+    PROVIDE(__data_source = LOADADDR(.TFM_DATA));
+    PROVIDE(__data_start = ADDR(.TFM_DATA));
+    PROVIDE(__data_size = SIZEOF(.TFM_DATA));
+
     .TFM_BSS ALIGN(4) (NOLOAD) :
     {
         __bss_start__ = .;
@@ -577,6 +591,31 @@ SECTIONS
         . = ALIGN(4);
         __bss_end__ = .;
     } > RAM
+
+    PROVIDE(__bss_start = ADDR(.TFM_BSS));
+    PROVIDE(__bss_size = SIZEOF(.TFM_BSS));
+
+    .TFM_TDATA :
+    {
+	PROVIDE( __tls_start_addr = . );
+        *(.tdata .tdata.* .gnu.linkonce.td.*)
+    } > RAM
+
+    .TFM_TBSS :
+    {
+        *(.tbss .tbss.* .gnu.linkonce.tb.*)
+        *(.tcommon)
+	PROVIDE( __tls_end_addr = . );
+    } > RAM
+
+    .TFM_TLS_BASE : {
+	. = . + (__tls_end_addr - __tls_start_addr);
+    } > RAM
+    PROVIDE( __tls_base = ADDR(.TFM_TLS_BASE) );
+    PROVIDE( __tls_align = MAX(ALIGNOF(.TFM_TDATA), ALIGNOF(.TFM_TBSS)) );
+    PROVIDE( __arm32_tls_tcb_offset = MAX(8, __tls_align) );
+    PROVIDE( __arm64_tls_tcb_offset = MAX(16, __tls_align) );
+
     Image$$ER_PART_RT_POOL$$ZI$$Base = __partition_runtime_start__;
     Image$$ER_PART_RT_POOL$$ZI$$Limit = __partition_runtime_end__;
     Image$$ER_SERV_RT_POOL$$ZI$$Base = __service_runtime_start__;

platform/ext/common/picolibc.c

@@ -0,0 +1,48 @@
+/*
+ * Copyright © 2025, Keith Packard <[email protected]>
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ *
+ */
+
+#include <stdint.h>
+#include <string.h>
+#include "config_impl.h"
+
+/*
+ * Picolibc's startup code only initializes a single data and bss
+ * segment.  Augment that by initializing the remaining segments using
+ * the lists provided by the linker script in a constructor which will
+ * be called from _start
+ */
+
+#if defined(__PICOLIBC__) && TFM_ISOLATION_LEVEL > 0
+
+static void __attribute__((constructor))
+picolibc_startup(void)
+{
+  typedef struct __copy_table {
+    uint32_t const* src;
+    uint32_t* dest;
+    uint32_t  wlen;
+  } __copy_table_t;
+
+  typedef struct __zero_table {
+    uint32_t* dest;
+    uint32_t  wlen;
+  } __zero_table_t;
+
+  extern const __copy_table_t __copy_table_start__;
+  extern const __copy_table_t __copy_table_end__;
+  extern const __zero_table_t __zero_table_start__;
+  extern const __zero_table_t __zero_table_end__;
+
+  for (__copy_table_t const* pTable = &__copy_table_start__; pTable < &__copy_table_end__; ++pTable) {
+    memcpy(pTable->dest, pTable->src, pTable->wlen << 2);
+  }
+
+  for (__zero_table_t const* pTable = &__zero_table_start__; pTable < &__zero_table_end__; ++pTable) {
+    memset(pTable->dest, 0, pTable->wlen << 2);
+  }
+}
+#endif

platform/ext/common/syscalls_stub.c

@@ -14,6 +14,14 @@
 #include <stddef.h>
 #include <stdint.h>
 
+#ifdef __PICOLIBC__
+__attribute__((weak))
+void _exit(int status)
+{
+    (void) status;
+    for(;;);
+}
+#else
 __attribute__((weak))
 void _close(void)
 {
@@ -53,3 +61,5 @@ __attribute__((weak))
 void _write(void)
 {
 }
+
+#endif /* !__PICOLIBC__ */

platform/ext/common/uart_stdout.c

@@ -85,7 +85,7 @@ int fputc(int ch, FILE *f)
 /* Redirect sdtio for PicoLib in LLVM toolchain
    as per https://github.com/picolibc/picolibc/blob/main/doc/os.md
    'fputch()' named intentionally different from 'fputc()' from picolib */
-#elif defined(__clang_major__)
+#elif defined(__PICOLIBC__)
 
 int fputch(char ch, struct __file *f)
 {