Enable use of uninitialized_ram() macro (#1199)

The uninitted RAM macro would fail because the OTA code would clear out the
first 50KB or so of RAM to copy its code and global values.

Move all global values to the end of RAM in OTA, and then align any uninitted
vars to a 16KB unit to ensure it won't appear in the 1st part of RAM that is
still needed to copy the OTA executable.

In the normal case, without any uninit_ram vars, no additional space is used.
When uninit_ram is used, up to 16KB of space may be lost to get that alignment,
but it will work properly.

Fixes #1188

Author: earlephilhower

lib/memmap_default.ld

@@ -191,11 +191,6 @@ SECTIONS
         __data_end__ = .;
     } > RAM AT> FLASH
 
-    .uninitialized_data (COPY): {
-        . = ALIGN(4);
-        *(.uninitialized_data*)
-    } > RAM
-
     /* Start and end symbols must be word-aligned */
     .scratch_x : {
         __scratch_x_start__ = .;
@@ -213,7 +208,7 @@ SECTIONS
     } > SCRATCH_Y AT > FLASH
     __scratch_y_source__ = LOADADDR(.scratch_y);
 
-    .bss  : {
+    .bss : {
         . = ALIGN(4);
         __bss_start__ = .;
         *(SORT_BY_ALIGNMENT(SORT_BY_NAME(.bss*)))
@@ -222,6 +217,15 @@ SECTIONS
         __bss_end__ = .;
     } > RAM
 
+    /* At most one of the following two will be engaged, depending on the SDK version */
+    .uninitialized_ram BLOCK(16k) (NOLOAD) : {
+        *(.uninitialized_ram*)
+    } > RAM
+
+    .uninitialized_data BLOCK(16k) (NOLOAD) : {
+        *(.uninitialized_data*)
+    } > RAM
+
     .heap (COPY):
     {
         __end__ = .;

lib/ota.o

@@ -50,6 +50,8 @@ target_compile_definitions(ota PUBLIC
 target_compile_options(ota PUBLIC
 	-fno-exceptions
 	-Os
+        -Wall
+        -Werror
 	$<$<COMPILE_LANGUAGE:CXX>:-fno-rtti>
 )
 

ota/CMakeLists.txt

@@ -23,21 +23,28 @@
 
 MEMORY
 {
-    FLASH(rx) : ORIGIN = 0x10000100, LENGTH = 16384k
-    RAM(rwx) : ORIGIN =  0x20000000, LENGTH = 256k
+    FLASH(rx) :      ORIGIN = 0x10000100, LENGTH = 16384k
+    RAM(rwx) :       ORIGIN = 0x20000000, LENGTH = 212k
+    /* We split RAM into main (where the flash->ram code will be copied) and GLOBALS which has all large arrays */
+    /* By placing those arrays at the end of RAM we can avoid overwriting uninitialize_ram from the main app as */
+    /* long as those vars are stored after 12K by using an alignment value in the main app linker file. */
+    GLOBALS(rwx) :   ORIGIN = 0x20035000, LENGTH = 44k
     SCRATCH_X(rwx) : ORIGIN = 0x20040000, LENGTH = 4k
     SCRATCH_Y(rwx) : ORIGIN = 0x20041000, LENGTH = 4k
 }
 
 ENTRY(_entry_point)
 
-SECTIONS
-{
+SECTIONS {
     /* Second stage bootloader is prepended to the image. It must be 256 bytes big
        and checksummed. It is usually built by the boot_stage2 target
        in the Raspberry Pi Pico SDK
     */
 
+    /* Make sure the globals are not copied/etc.  They're empty RAM we will init ourselves */
+    .globals (NOLOAD) : {
+    } > GLOBALS
+
     .flash_begin : {
         __flash_binary_start = .;
     } > FLASH
@@ -72,14 +79,12 @@ SECTIONS
         . = ALIGN(4);
     } > FLASH
 
-    .ARM.extab :
-    {
+    .ARM.extab : {
         *(.ARM.extab* .gnu.linkonce.armextab.*)
     } > FLASH
 
     __exidx_start = .;
-    .ARM.exidx :
-    {
+    .ARM.exidx : {
         *(.ARM.exidx* .gnu.linkonce.armexidx.*)
     } > FLASH
     __exidx_end = .;
@@ -188,7 +193,7 @@ SECTIONS
     } > SCRATCH_Y AT > FLASH
     __scratch_y_source__ = LOADADDR(.scratch_y);
 
-    .bss  : {
+    .bss : {
         . = ALIGN(4);
         __bss_start__ = .;
         *(SORT_BY_ALIGNMENT(SORT_BY_NAME(.bss*)))
@@ -197,8 +202,7 @@ SECTIONS
         __bss_end__ = .;
     } > RAM
 
-    .heap (COPY):
-    {
+    .heap (COPY): {
         __end__ = .;
         end = __end__;
         *(.heap*)
@@ -214,12 +218,10 @@ SECTIONS
     /* by default we put core 0 stack at the end of scratch Y, so that if core 1
      * stack is not used then all of SCRATCH_X is free.
      */
-    .stack1_dummy (COPY):
-    {
+    .stack1_dummy (COPY): {
         *(.stack1*)
     } > SCRATCH_X
-    .stack_dummy (COPY):
-    {
+    .stack_dummy (COPY): {
         *(.stack*)
     } > SCRATCH_Y
 

ota/memmap_ota.ld

@@ -53,8 +53,7 @@ void dumphex(uint32_t x) {
 #endif
 }
 
-static OTACmdPage _ota_cmd;
-
+extern OTACmdPage _ota_cmd;
 void do_ota() {
     if (*__FS_START__ == *__FS_END__) {
         return;
@@ -165,7 +164,7 @@ void do_ota() {
 
 #pragma GCC push_options
 #pragma GCC optimize("O0")
-int main(int a, unsigned char **b) {
+int main(int a, char **b) {
     (void) a;
     (void) b;
 
@@ -190,7 +189,7 @@ int main(int a, unsigned char **b) {
     fcn();
 
     // Should never get here!
-    return 0;
+    return *sp;
 }
 #pragma GCC pop_options
 

ota/ota.c

@@ -26,6 +26,8 @@
 #include "../libraries/LittleFS/lib/littlefs/lfs_util.h"
 #include "./uzlib/src/uzlib.h"
 
+OTACmdPage __attribute__((section (".globals"))) _ota_cmd;
+
 static lfs_t _lfs;
 static struct lfs_config  _lfs_cfg;
 
@@ -70,9 +72,9 @@ static int lfs_flash_sync(const struct lfs_config *c) {
     return 0;
 }
 
-uint8_t _read_buffer[256];
-uint8_t _prog_buffer[256];
-uint8_t _lookahead_buffer[256];
+uint8_t __attribute__((section (".globals"))) _read_buffer[256];
+uint8_t __attribute__((section (".globals"))) _prog_buffer[256];
+uint8_t __attribute__((section (".globals"))) _lookahead_buffer[256];
 bool lfsMount(uint8_t *start, uint32_t blockSize, uint32_t size) {
     _start = start;
     _blockSize = blockSize;
@@ -104,16 +106,16 @@ bool lfsMount(uint8_t *start, uint32_t blockSize, uint32_t size) {
 static bool _gzip = false;
 static lfs_file_t _file;
 
-static unsigned char __attribute__((aligned(4))) uzlib_read_buff[4096];
-static unsigned char gzip_dict[32768];
-static uint8_t _flash_buff[4096]; // no room for this on the stack
-static struct uzlib_uncomp m_uncomp;
+unsigned char __attribute__((section (".globals"))) uzlib_read_buff[4096];
+unsigned char __attribute__((section (".globals"))) gzip_dict[32768];
+uint8_t __attribute__((section (".globals")))_flash_buff[4096]; // no room for this on the stack
+struct uzlib_uncomp __attribute__((section (".globals"))) m_uncomp;
 
-static uint8_t _ota_buff[256];
-static struct lfs_file_config _ota_cfg = { (void *)_ota_buff, NULL, 0 };
+uint8_t __attribute__((section (".globals")))_ota_buff[256];
+struct lfs_file_config _ota_cfg = { (void *)_ota_buff, NULL, 0 };
 
-static uint8_t _file_buff[256];
-static struct lfs_file_config _file_cfg = { (void *)_file_buff, NULL, 0 };
+uint8_t __attribute__((section (".globals")))_file_buff[256];
+struct lfs_file_config _file_cfg = { (void *)_file_buff, NULL, 0 };
 
 bool lfsReadOTA(OTACmdPage *ota, uint32_t *blockToErase) {
     lfs_file_t f;

ota/ota_lfs.c

@@ -153,7 +153,7 @@ recipe.output.save_file={build.project_name}.{build.variant}.{build.preferred_ou
 ## Compute size
 recipe.size.pattern="{compiler.path}{compiler.size.cmd}" -A "{build.path}/{build.project_name}.elf"
 recipe.size.regex=^(?:\.boot2|\.text|\.rodata|\.ARM\.extab|\.ARM\.exidx)\s+([0-9]+).*
-recipe.size.regex.data=^(?:\.data|\.bss|\.ram_vector_table|\.uninitialized_data)\s+([0-9]+).*
+recipe.size.regex.data=^(?:\.data|\.bss|\.ram_vector_table|\.uninitialized_ram|\.uninitialized_data)\s+([0-9]+).*
 
 
 # Debugging