heap: Add a configuration that places all the heap component and related functionalities in flash when enabled

Add test configuration to run all tests with heap component in the flash.
Add reference to this new configuration in performance section of the documentation.

components/app_trace/heap_trace_tohost.c

@@ -8,7 +8,7 @@
 #define HEAP_TRACE_SRCFILE /* don't warn on inclusion here */
 #include "esp_heap_trace.h"
 #undef HEAP_TRACE_SRCFILE
-
+#include "esp_heap_caps.h"
 #if CONFIG_APPTRACE_SV_ENABLE
 #include "esp_app_trace.h"
 #include "esp_sysview_trace.h"
@@ -85,7 +85,7 @@ void heap_trace_dump_caps(__attribute__((unused)) const uint32_t caps)
 }
 
 /* Add a new allocation to the heap trace records */
-static IRAM_ATTR void record_allocation(const heap_trace_record_t *record)
+static HEAP_IRAM_ATTR void record_allocation(const heap_trace_record_t *record)
 {
     if (!s_tracing) {
         return;
@@ -100,7 +100,7 @@ static IRAM_ATTR void record_allocation(const heap_trace_record_t *record)
    For HEAP_TRACE_ALL, this means filling in the freed_by pointer.
    For HEAP_TRACE_LEAKS, this means removing the record from the log.
 */
-static IRAM_ATTR void record_free(void *p, void **callers)
+static HEAP_IRAM_ATTR void record_free(void *p, void **callers)
 {
     if (!s_tracing) {
         return;

components/driver/Kconfig

@@ -85,10 +85,15 @@ menu "Driver Configurations"
         config SPI_MASTER_ISR_IN_IRAM
             bool "Place SPI master ISR function into IRAM"
             default y
+            depends on !HEAP_PLACE_FUNCTION_INTO_FLASH
             select PERIPH_CTRL_FUNC_IN_IRAM
             help
                 Place the SPI master ISR in to IRAM to avoid possible cache miss.
 
+                Enabling this configuration is possible only when HEAP_PLACE_FUNCTION_INTO_FLASH
+                is disabled since the spi master uses can allocate transactions buffers into DMA
+                memory section using the heap component API that ipso facto has to be placed in IRAM.
+
                 Also you can forbid the ISR being disabled during flash writing
                 access, by add ESP_INTR_FLAG_IRAM when initializing the driver.
 

components/heap/Kconfig

@@ -113,4 +113,13 @@ menu "Heap memory debugging"
             features will be added and bugs will be fixed in the IDF source
             but cannot be synced to ROM.
 
+    config HEAP_PLACE_FUNCTION_INTO_FLASH
+        bool "Force the entire heap component to be placed in flash memory"
+        depends on !HEAP_TLSF_USE_ROM_IMPL
+        default n
+        help
+            Enable this flag to save up RAM space by placing the heap component in the flash memory
+
+            Note that it is only safe to enable this configuration if no functions from esp_heap_caps.h
+            or esp_heap_trace.h are called from ISR.
 endmenu

components/heap/heap_caps.c

@@ -43,14 +43,14 @@ possible. This should optimize the amount of RAM accessible to the code without
 static esp_alloc_failed_hook_t alloc_failed_callback;
 
 #ifdef CONFIG_HEAP_ABORT_WHEN_ALLOCATION_FAILS
-IRAM_ATTR static void hex_to_str(char buf[8], uint32_t n)
+HEAP_IRAM_ATTR static void hex_to_str(char buf[8], uint32_t n)
 {
     for (int i = 0; i < 8; i++) {
         uint8_t b4 = (n >> (28 - i * 4)) & 0b1111;
         buf[i] = b4 <= 9 ? '0' + b4 : 'a' + b4 - 10;
     }
 }
-IRAM_ATTR static void fmt_abort_str(char dest[48], size_t size, uint32_t caps)
+HEAP_IRAM_ATTR static void fmt_abort_str(char dest[48], size_t size, uint32_t caps)
 {
     char sSize[8];
     char sCaps[8];
@@ -67,7 +67,7 @@ IRAM_ATTR static void fmt_abort_str(char dest[48], size_t size, uint32_t caps)
   IRAM in such a way that it can be later freed. It assumes both the address as well as the length to be word-aligned.
   It returns a region that's 1 word smaller than the region given because it stores the original Dram address there.
 */
-IRAM_ATTR static void *dram_alloc_to_iram_addr(void *addr, size_t len)
+HEAP_IRAM_ATTR static void *dram_alloc_to_iram_addr(void *addr, size_t len)
 {
     uintptr_t dstart = (uintptr_t)addr; //First word
     uintptr_t dend __attribute__((unused)) = dstart + len - 4; //Last word
@@ -84,7 +84,7 @@ IRAM_ATTR static void *dram_alloc_to_iram_addr(void *addr, size_t len)
     return iptr + 1;
 }
 
-IRAM_ATTR NOINLINE_ATTR static void heap_caps_alloc_failed(size_t requested_size, uint32_t caps, const char *function_name)
+HEAP_IRAM_ATTR NOINLINE_ATTR static void heap_caps_alloc_failed(size_t requested_size, uint32_t caps, const char *function_name)
 {
     if (alloc_failed_callback) {
         alloc_failed_callback(requested_size, caps, function_name);
@@ -118,7 +118,7 @@ bool heap_caps_match(const heap_t *heap, uint32_t caps)
 This function should not be called directly as it does not
 check for failure / call heap_caps_alloc_failed()
 */
-IRAM_ATTR static void *heap_caps_malloc_base( size_t size, uint32_t caps)
+HEAP_IRAM_ATTR static void *heap_caps_malloc_base( size_t size, uint32_t caps)
 {
     void *ret = NULL;
 
@@ -192,7 +192,7 @@ IRAM_ATTR static void *heap_caps_malloc_base( size_t size, uint32_t caps)
 /*
 Routine to allocate a bit of memory with certain capabilities. caps is a bitfield of MALLOC_CAP_* bits.
 */
-IRAM_ATTR void *heap_caps_malloc( size_t size, uint32_t caps){
+HEAP_IRAM_ATTR void *heap_caps_malloc( size_t size, uint32_t caps){
 
     void* ptr = heap_caps_malloc_base(size, caps);
 
@@ -217,7 +217,7 @@ void heap_caps_malloc_extmem_enable(size_t limit)
 /*
  Default memory allocation implementation. Should return standard 8-bit memory. malloc() essentially resolves to this function.
 */
-IRAM_ATTR void *heap_caps_malloc_default( size_t size )
+HEAP_IRAM_ATTR void *heap_caps_malloc_default( size_t size )
 {
     if (malloc_alwaysinternal_limit==MALLOC_DISABLE_EXTERNAL_ALLOCS) {
         return heap_caps_malloc( size, MALLOC_CAP_DEFAULT | MALLOC_CAP_INTERNAL);
@@ -250,7 +250,7 @@ IRAM_ATTR void *heap_caps_malloc_default( size_t size )
  Same for realloc()
  Note: keep the logic in here the same as in heap_caps_malloc_default (or merge the two as soon as this gets more complex...)
  */
-IRAM_ATTR void *heap_caps_realloc_default( void *ptr, size_t size )
+HEAP_IRAM_ATTR void *heap_caps_realloc_default( void *ptr, size_t size )
 {
     if (malloc_alwaysinternal_limit==MALLOC_DISABLE_EXTERNAL_ALLOCS) {
         return heap_caps_realloc( ptr, size, MALLOC_CAP_DEFAULT | MALLOC_CAP_INTERNAL );
@@ -282,7 +282,7 @@ IRAM_ATTR void *heap_caps_realloc_default( void *ptr, size_t size )
 /*
  Memory allocation as preference in decreasing order.
  */
-IRAM_ATTR void *heap_caps_malloc_prefer( size_t size, size_t num, ... )
+HEAP_IRAM_ATTR void *heap_caps_malloc_prefer( size_t size, size_t num, ... )
 {
     va_list argp;
     va_start( argp, num );
@@ -306,7 +306,7 @@ IRAM_ATTR void *heap_caps_malloc_prefer( size_t size, size_t num, ... )
 /*
  Memory reallocation as preference in decreasing order.
  */
-IRAM_ATTR void *heap_caps_realloc_prefer( void *ptr, size_t size, size_t num, ... )
+HEAP_IRAM_ATTR void *heap_caps_realloc_prefer( void *ptr, size_t size, size_t num, ... )
 {
     va_list argp;
     va_start( argp, num );
@@ -330,7 +330,7 @@ IRAM_ATTR void *heap_caps_realloc_prefer( void *ptr, size_t size, size_t num, ..
 /*
  Memory callocation as preference in decreasing order.
  */
-IRAM_ATTR void *heap_caps_calloc_prefer( size_t n, size_t size, size_t num, ... )
+HEAP_IRAM_ATTR void *heap_caps_calloc_prefer( size_t n, size_t size, size_t num, ... )
 {
     va_list argp;
     va_start( argp, num );
@@ -357,7 +357,7 @@ IRAM_ATTR void *heap_caps_calloc_prefer( size_t n, size_t size, size_t num, ...
    (This confirms if ptr is inside the heap's region, doesn't confirm if 'ptr'
    is an allocated block or is some other random address inside the heap.)
 */
-IRAM_ATTR static heap_t *find_containing_heap(void *ptr )
+HEAP_IRAM_ATTR static heap_t *find_containing_heap(void *ptr )
 {
     intptr_t p = (intptr_t)ptr;
     heap_t *heap;
@@ -369,7 +369,7 @@ IRAM_ATTR static heap_t *find_containing_heap(void *ptr )
     return NULL;
 }
 
-IRAM_ATTR void heap_caps_free( void *ptr)
+HEAP_IRAM_ATTR void heap_caps_free( void *ptr)
 {
     if (ptr == NULL) {
         return;
@@ -394,7 +394,7 @@ IRAM_ATTR void heap_caps_free( void *ptr)
 This function should not be called directly as it does not
 check for failure / call heap_caps_alloc_failed()
 */
-IRAM_ATTR static void *heap_caps_realloc_base( void *ptr, size_t size, uint32_t caps)
+HEAP_IRAM_ATTR static void *heap_caps_realloc_base( void *ptr, size_t size, uint32_t caps)
 {
     bool ptr_in_diram_case = false;
     heap_t *heap = NULL;
@@ -469,7 +469,7 @@ IRAM_ATTR static void *heap_caps_realloc_base( void *ptr, size_t size, uint32_t
     return NULL;
 }
 
-IRAM_ATTR void *heap_caps_realloc( void *ptr, size_t size, uint32_t caps)
+HEAP_IRAM_ATTR void *heap_caps_realloc( void *ptr, size_t size, uint32_t caps)
 {
     ptr = heap_caps_realloc_base(ptr, size, caps);
 
@@ -485,7 +485,7 @@ IRAM_ATTR void *heap_caps_realloc( void *ptr, size_t size, uint32_t caps)
 This function should not be called directly as it does not
 check for failure / call heap_caps_alloc_failed()
 */
-IRAM_ATTR static void *heap_caps_calloc_base( size_t n, size_t size, uint32_t caps)
+HEAP_IRAM_ATTR static void *heap_caps_calloc_base( size_t n, size_t size, uint32_t caps)
 {
     void *result;
     size_t size_bytes;
@@ -501,7 +501,7 @@ IRAM_ATTR static void *heap_caps_calloc_base( size_t n, size_t size, uint32_t ca
     return result;
 }
 
-IRAM_ATTR void *heap_caps_calloc( size_t n, size_t size, uint32_t caps)
+HEAP_IRAM_ATTR void *heap_caps_calloc( size_t n, size_t size, uint32_t caps)
 {
     void* ptr = heap_caps_calloc_base(n, size, caps);
 
@@ -655,7 +655,7 @@ size_t heap_caps_get_allocated_size( void *ptr )
     return size;
 }
 
-IRAM_ATTR void *heap_caps_aligned_alloc(size_t alignment, size_t size, uint32_t caps)
+HEAP_IRAM_ATTR void *heap_caps_aligned_alloc(size_t alignment, size_t size, uint32_t caps)
 {
     void *ret = NULL;
 
@@ -708,7 +708,7 @@ IRAM_ATTR void *heap_caps_aligned_alloc(size_t alignment, size_t size, uint32_t
     return NULL;
 }
 
-IRAM_ATTR void heap_caps_aligned_free(void *ptr)
+HEAP_IRAM_ATTR void heap_caps_aligned_free(void *ptr)
 {
     heap_caps_free(ptr);
 }

components/heap/heap_trace_standalone.c

@@ -10,14 +10,13 @@
 #define HEAP_TRACE_SRCFILE /* don't warn on inclusion here */
 #include "esp_heap_trace.h"
 #undef HEAP_TRACE_SRCFILE
-
+#include "esp_heap_caps.h"
 #include "esp_attr.h"
 #include "freertos/FreeRTOS.h"
 #include "freertos/task.h"
 #include "esp_memory_utils.h"
 #include "sys/queue.h"
 
-
 #define STACK_DEPTH CONFIG_HEAP_TRACING_STACK_DEPTH
 
 #if CONFIG_HEAP_TRACING_STANDALONE
@@ -89,7 +88,7 @@ static heap_trace_hash_list_t hash_map[(size_t)CONFIG_HEAP_TRACE_HASH_MAP_SIZE];
 static size_t total_hashmap_hits;
 static size_t total_hashmap_miss;
 
-static size_t hash_idx(void* p)
+static HEAP_IRAM_ATTR size_t hash_idx(void* p)
 {
     static const uint32_t fnv_prime = 16777619UL; // expression 2^24 + 2^8 + 0x93 (32 bits size)
     // since all the addresses are 4 bytes aligned, computing address * fnv_prime always gives
@@ -100,19 +99,19 @@ static size_t hash_idx(void* p)
              ((uint32_t)p >> 7)) * fnv_prime) % (uint32_t)CONFIG_HEAP_TRACE_HASH_MAP_SIZE;
 }
 
-static void map_add(heap_trace_record_t *r_add)
+static HEAP_IRAM_ATTR void map_add(heap_trace_record_t *r_add)
 {
     size_t idx = hash_idx(r_add->address);
     TAILQ_INSERT_TAIL(&hash_map[idx], r_add, tailq_hashmap);
 }
 
-static void map_remove(heap_trace_record_t *r_remove)
+static HEAP_IRAM_ATTR void map_remove(heap_trace_record_t *r_remove)
 {
     size_t idx = hash_idx(r_remove->address);
     TAILQ_REMOVE(&hash_map[idx], r_remove, tailq_hashmap);
 }
 
-static heap_trace_record_t* map_find(void *p)
+static HEAP_IRAM_ATTR heap_trace_record_t* map_find(void *p)
 {
     size_t idx = hash_idx(p);
     heap_trace_record_t *r_cur = NULL;
@@ -385,7 +384,7 @@ static void heap_trace_dump_base(bool internal_ram, bool psram)
 }
 
 /* Add a new allocation to the heap trace records */
-static IRAM_ATTR void record_allocation(const heap_trace_record_t *r_allocation)
+static HEAP_IRAM_ATTR void record_allocation(const heap_trace_record_t *r_allocation)
 {
     if (!tracing || r_allocation->address == NULL) {
         return;
@@ -420,7 +419,7 @@ static IRAM_ATTR void record_allocation(const heap_trace_record_t *r_allocation)
    callers is an array of  STACK_DEPTH function pointer from the call stack
    leading to the call of record_free.
 */
-static IRAM_ATTR void record_free(void *p, void **callers)
+static HEAP_IRAM_ATTR void record_free(void *p, void **callers)
 {
        if (!tracing || p == NULL) {
         return;
@@ -473,7 +472,7 @@ static void list_setup(void)
 
 /* 1. removes record r_remove from records.list,
    2. places it into records.unused */
-static IRAM_ATTR void list_remove(heap_trace_record_t *r_remove)
+static HEAP_IRAM_ATTR void list_remove(heap_trace_record_t* r_remove)
 {
     assert(records.count > 0);
 
@@ -497,7 +496,7 @@ static IRAM_ATTR void list_remove(heap_trace_record_t *r_remove)
 
 
 // pop record from unused list
-static IRAM_ATTR heap_trace_record_t* list_pop_unused(void)
+static HEAP_IRAM_ATTR heap_trace_record_t* list_pop_unused(void)
 {
     // no records left?
     if (records.count >= records.capacity) {
@@ -517,7 +516,7 @@ static IRAM_ATTR heap_trace_record_t* list_pop_unused(void)
 
 // deep copy a record.
 // Note: only copies the *allocation data*, not the next & prev ptrs
-static IRAM_ATTR void record_deep_copy(heap_trace_record_t *r_dest, const heap_trace_record_t *r_src)
+static HEAP_IRAM_ATTR void record_deep_copy(heap_trace_record_t *r_dest, const heap_trace_record_t *r_src)
 {
     r_dest->ccount  = r_src->ccount;
     r_dest->address = r_src->address;
@@ -528,7 +527,7 @@ static IRAM_ATTR void record_deep_copy(heap_trace_record_t *r_dest, const heap_t
 
 // Append a record to records.list
 // Note: This deep copies r_append
-static IRAM_ATTR heap_trace_record_t* list_add(const heap_trace_record_t *r_append)
+static HEAP_IRAM_ATTR heap_trace_record_t* list_add(const heap_trace_record_t *r_append)
 {
     if (records.count < records.capacity) {
 
@@ -566,7 +565,7 @@ static IRAM_ATTR heap_trace_record_t* list_add(const heap_trace_record_t *r_appe
 }
 
 // search records.list backwards for the allocation record matching this address
-static IRAM_ATTR heap_trace_record_t* list_find_address_reverse(void *p)
+static HEAP_IRAM_ATTR heap_trace_record_t* list_find_address_reverse(void* p)
 {
     heap_trace_record_t *r_found = NULL;
 

components/heap/include/esp_heap_caps.h

@@ -17,6 +17,12 @@
 extern "C" {
 #endif
 
+#if CONFIG_HEAP_PLACE_FUNCTION_INTO_FLASH
+#define HEAP_IRAM_ATTR
+#else
+#define HEAP_IRAM_ATTR IRAM_ATTR
+#endif
+
 /**
  * @brief Flags to indicate the capabilities of the various memory systems
  */
@@ -63,7 +69,7 @@ esp_err_t heap_caps_register_failed_alloc_callback(esp_alloc_failed_hook_t callb
  * @note this hook is called on the same thread as the allocation, which may be within a low level operation.
  * You should refrain from doing heavy work, logging, flash writes, or any locking.
  */
-__attribute__((weak)) IRAM_ATTR void esp_heap_trace_alloc_hook(void* ptr, size_t size, uint32_t caps);
+__attribute__((weak)) HEAP_IRAM_ATTR void esp_heap_trace_alloc_hook(void* ptr, size_t size, uint32_t caps);
 
 /**
  * @brief callback called after every free
@@ -71,7 +77,7 @@ __attribute__((weak)) IRAM_ATTR void esp_heap_trace_alloc_hook(void* ptr, size_t
  * @note this hook is called on the same thread as the allocation, which may be within a low level operation.
  * You should refrain from doing heavy work, logging, flash writes, or any locking.
  */
-__attribute__((weak)) IRAM_ATTR void esp_heap_trace_free_hook(void* ptr);
+__attribute__((weak)) HEAP_IRAM_ATTR void esp_heap_trace_free_hook(void* ptr);
 #endif
 
 /**

components/heap/include/heap_trace.inc

@@ -57,7 +57,7 @@ inline static uint32_t get_ccount(void)
    Calls to __builtin_return_address are "unrolled" via TEST_STACK macro as gcc requires the
    argument to be a compile-time constant.
 */
-static IRAM_ATTR __attribute__((noinline)) void get_call_stack(void **callers)
+static HEAP_IRAM_ATTR __attribute__((noinline)) void get_call_stack(void **callers)
 {
     bzero(callers, sizeof(void *) * STACK_DEPTH);
     TEST_STACK(0);
@@ -86,7 +86,7 @@ void *__real_heap_caps_malloc_default( size_t size );
 void *__real_heap_caps_realloc_default( void *ptr, size_t size );
 
 /* trace any 'malloc' event */
-static IRAM_ATTR __attribute__((noinline)) void *trace_malloc(size_t size, uint32_t caps, trace_malloc_mode_t mode)
+static HEAP_IRAM_ATTR __attribute__((noinline)) void *trace_malloc(size_t size, uint32_t caps, trace_malloc_mode_t mode)
 {
     uint32_t ccount = get_ccount();
     void *p;
@@ -110,7 +110,7 @@ static IRAM_ATTR __attribute__((noinline)) void *trace_malloc(size_t size, uint3
 void __real_heap_caps_free(void *p);
 
 /* trace any 'free' event */
-static IRAM_ATTR __attribute__((noinline)) void trace_free(void *p)
+static HEAP_IRAM_ATTR __attribute__((noinline)) void trace_free(void *p)
 {
     void *callers[STACK_DEPTH];
     get_call_stack(callers);
@@ -122,7 +122,7 @@ static IRAM_ATTR __attribute__((noinline)) void trace_free(void *p)
 void * __real_heap_caps_realloc(void *p, size_t size, uint32_t caps);
 
 /* trace any 'realloc' event */
-static IRAM_ATTR __attribute__((noinline)) void *trace_realloc(void *p, size_t size, uint32_t caps, trace_malloc_mode_t mode)
+static HEAP_IRAM_ATTR __attribute__((noinline)) void *trace_realloc(void *p, size_t size, uint32_t caps, trace_malloc_mode_t mode)
 {
     void *callers[STACK_DEPTH];
     uint32_t ccount = get_ccount();
@@ -154,22 +154,22 @@ static IRAM_ATTR __attribute__((noinline)) void *trace_realloc(void *p, size_t s
    as they no longer go via the libc functions in ROM. But more or less
    the same in the end. */
 
-IRAM_ATTR void *__wrap_malloc(size_t size)
+HEAP_IRAM_ATTR void *__wrap_malloc(size_t size)
 {
     return trace_malloc(size, 0, TRACE_MALLOC_DEFAULT);
 }
 
-IRAM_ATTR void __wrap_free(void *p)
+HEAP_IRAM_ATTR void __wrap_free(void *p)
 {
     trace_free(p);
 }
 
-IRAM_ATTR void *__wrap_realloc(void *p, size_t size)
+HEAP_IRAM_ATTR void *__wrap_realloc(void *p, size_t size)
 {
     return trace_realloc(p, size, 0, TRACE_MALLOC_DEFAULT);
 }
 
-IRAM_ATTR void *__wrap_calloc(size_t nmemb, size_t size)
+HEAP_IRAM_ATTR void *__wrap_calloc(size_t nmemb, size_t size)
 {
     size = size * nmemb;
     void *result = trace_malloc(size, 0, TRACE_MALLOC_DEFAULT);
@@ -179,24 +179,24 @@ IRAM_ATTR void *__wrap_calloc(size_t nmemb, size_t size)
     return result;
 }
 
-IRAM_ATTR void *__wrap_heap_caps_malloc(size_t size, uint32_t caps)
+HEAP_IRAM_ATTR void *__wrap_heap_caps_malloc(size_t size, uint32_t caps)
 {
     return trace_malloc(size, caps, TRACE_MALLOC_CAPS);
 }
 
 void __wrap_heap_caps_free(void *p) __attribute__((alias("__wrap_free")));
 
-IRAM_ATTR void *__wrap_heap_caps_realloc(void *p, size_t size, uint32_t caps)
+HEAP_IRAM_ATTR void *__wrap_heap_caps_realloc(void *p, size_t size, uint32_t caps)
 {
     return trace_realloc(p, size, caps, TRACE_MALLOC_CAPS);
 }
 
-IRAM_ATTR void *__wrap_heap_caps_malloc_default( size_t size )
+HEAP_IRAM_ATTR void *__wrap_heap_caps_malloc_default( size_t size )
 {
     return trace_malloc(size, 0, TRACE_MALLOC_DEFAULT);
 }
 
-IRAM_ATTR void *__wrap_heap_caps_realloc_default( void *ptr, size_t size )
+HEAP_IRAM_ATTR void *__wrap_heap_caps_realloc_default( void *ptr, size_t size )
 {
     return trace_realloc(ptr, size, 0, TRACE_MALLOC_DEFAULT);
 }

components/heap/linker.lf

@@ -1,54 +1,55 @@
 [mapping:heap]
 archive: libheap.a
 entries:
-    if HEAP_TLSF_USE_ROM_IMPL = n:
-        tlsf:tlsf_block_size (noflash)
-        tlsf:tlsf_size (noflash)
-        tlsf:tlsf_align_size (noflash)
-        tlsf:tlsf_block_size_min (noflash)
-        tlsf:tlsf_block_size_max (noflash)
-        tlsf:tlsf_alloc_overhead (noflash)
-        tlsf:tlsf_get_pool (noflash)
-        tlsf:tlsf_malloc (noflash)
-        tlsf:tlsf_memalign_offs (noflash)
-        tlsf:tlsf_memalign (noflash)
-        tlsf:tlsf_free (noflash)
-        tlsf:tlsf_realloc (noflash)
+    if HEAP_PLACE_FUNCTION_INTO_FLASH = n:
+        if HEAP_TLSF_USE_ROM_IMPL = n:
+            tlsf:tlsf_block_size (noflash)
+            tlsf:tlsf_size (noflash)
+            tlsf:tlsf_align_size (noflash)
+            tlsf:tlsf_block_size_min (noflash)
+            tlsf:tlsf_block_size_max (noflash)
+            tlsf:tlsf_alloc_overhead (noflash)
+            tlsf:tlsf_get_pool (noflash)
+            tlsf:tlsf_malloc (noflash)
+            tlsf:tlsf_memalign_offs (noflash)
+            tlsf:tlsf_memalign (noflash)
+            tlsf:tlsf_free (noflash)
+            tlsf:tlsf_realloc (noflash)
 
-        multi_heap:multi_heap_get_block_address_impl (noflash)
-        multi_heap:multi_heap_get_allocated_size_impl (noflash)
-        multi_heap:multi_heap_set_lock (noflash)
-        multi_heap:multi_heap_get_first_block (noflash)
-        multi_heap:multi_heap_get_next_block (noflash)
-        multi_heap:multi_heap_is_free (noflash)
-        multi_heap:multi_heap_malloc_impl (noflash)
-        multi_heap:multi_heap_free_impl (noflash)
-        multi_heap:multi_heap_realloc_impl (noflash)
-        multi_heap:multi_heap_aligned_alloc_impl_offs (noflash)
-        multi_heap:multi_heap_aligned_alloc_impl (noflash)
-        multi_heap:multi_heap_internal_lock (noflash)
-        multi_heap:multi_heap_internal_unlock (noflash)
-        multi_heap:assert_valid_block (noflash)
+            multi_heap:multi_heap_get_block_address_impl (noflash)
+            multi_heap:multi_heap_get_allocated_size_impl (noflash)
+            multi_heap:multi_heap_set_lock (noflash)
+            multi_heap:multi_heap_get_first_block (noflash)
+            multi_heap:multi_heap_get_next_block (noflash)
+            multi_heap:multi_heap_is_free (noflash)
+            multi_heap:multi_heap_malloc_impl (noflash)
+            multi_heap:multi_heap_free_impl (noflash)
+            multi_heap:multi_heap_realloc_impl (noflash)
+            multi_heap:multi_heap_aligned_alloc_impl_offs (noflash)
+            multi_heap:multi_heap_aligned_alloc_impl (noflash)
+            multi_heap:multi_heap_internal_lock (noflash)
+            multi_heap:multi_heap_internal_unlock (noflash)
+            multi_heap:assert_valid_block (noflash)
 
-    if HEAP_TLSF_USE_ROM_IMPL = y:
-        multi_heap:_multi_heap_lock (noflash)
-        multi_heap:_multi_heap_unlock (noflash)
-        multi_heap:multi_heap_in_rom_init (noflash)
+        if HEAP_TLSF_USE_ROM_IMPL = y:
+            multi_heap:_multi_heap_lock (noflash)
+            multi_heap:_multi_heap_unlock (noflash)
+            multi_heap:multi_heap_in_rom_init (noflash)
 
-    if HEAP_POISONING_DISABLED = n:
-        multi_heap_poisoning:poison_allocated_region (noflash)
-        multi_heap_poisoning:verify_allocated_region (noflash)
-        multi_heap_poisoning:multi_heap_aligned_alloc (noflash)
-        multi_heap_poisoning:multi_heap_malloc (noflash)
-        multi_heap_poisoning:multi_heap_free (noflash)
-        multi_heap_poisoning:multi_heap_aligned_free (noflash)
-        multi_heap_poisoning:multi_heap_realloc (noflash)
-        multi_heap_poisoning:multi_heap_get_block_address (noflash)
-        multi_heap_poisoning:multi_heap_get_block_owner (noflash)
-        multi_heap_poisoning:multi_heap_get_allocated_size (noflash)
-        multi_heap_poisoning:multi_heap_internal_check_block_poisoning (noflash)
-        multi_heap_poisoning:multi_heap_internal_poison_fill_region (noflash)
+        if HEAP_POISONING_DISABLED = n:
+            multi_heap_poisoning:poison_allocated_region (noflash)
+            multi_heap_poisoning:verify_allocated_region (noflash)
+            multi_heap_poisoning:multi_heap_aligned_alloc (noflash)
+            multi_heap_poisoning:multi_heap_malloc (noflash)
+            multi_heap_poisoning:multi_heap_free (noflash)
+            multi_heap_poisoning:multi_heap_aligned_free (noflash)
+            multi_heap_poisoning:multi_heap_realloc (noflash)
+            multi_heap_poisoning:multi_heap_get_block_address (noflash)
+            multi_heap_poisoning:multi_heap_get_block_owner (noflash)
+            multi_heap_poisoning:multi_heap_get_allocated_size (noflash)
+            multi_heap_poisoning:multi_heap_internal_check_block_poisoning (noflash)
+            multi_heap_poisoning:multi_heap_internal_poison_fill_region (noflash)
 
-    if HEAP_POISONING_COMPREHENSIVE = y:
-        multi_heap_poisoning:verify_fill_pattern (noflash)
-        multi_heap_poisoning:block_absorb_post_hook (noflash)
+        if HEAP_POISONING_COMPREHENSIVE = y:
+            multi_heap_poisoning:verify_fill_pattern (noflash)
+            multi_heap_poisoning:block_absorb_post_hook (noflash)

components/heap/test_apps/heap_tests/main/test_malloc_caps.c

@@ -174,7 +174,7 @@ TEST_CASE("heap_caps metadata test", "[heap]")
 /* Small function runs from IRAM to check that malloc/free/realloc
    all work OK when cache is disabled...
 */
-#ifndef CONFIG_ESP_SYSTEM_MEMPROT_FEATURE
+#if !CONFIG_ESP_SYSTEM_MEMPROT_FEATURE && !CONFIG_HEAP_PLACE_FUNCTION_INTO_FLASH
 static IRAM_ATTR __attribute__((noinline)) bool iram_malloc_test(void)
 {
     spi_flash_guard_get()->start(); // Disables flash cache

components/heap/test_apps/heap_tests/pytest_heap.py

@@ -20,6 +20,19 @@ def test_heap_poisoning(dut: Dut) -> None:
     dut.run_all_single_board_cases()
 
 
+@pytest.mark.generic
+@pytest.mark.esp32
+@pytest.mark.esp32c6
+@pytest.mark.parametrize(
+    'config',
+    [
+        'in_flash'
+    ]
+)
+def test_heap_in_flash(dut: Dut) -> None:
+    dut.run_all_single_board_cases()
+
+
 @pytest.mark.generic
 @pytest.mark.esp32
 @pytest.mark.esp32s2

components/heap/test_apps/heap_tests/sdkconfig.ci.in_flash

@@ -0,0 +1,2 @@
+CONFIG_HEAP_TLSF_USE_ROM_IMPL=n
+CONFIG_HEAP_PLACE_FUNCTION_INTO_FLASH=y

docs/en/api-guides/performance/ram-usage.rst

@@ -141,6 +141,7 @@ The following options will reduce IRAM usage of some ESP-IDF features:
     - Disabling :ref:`CONFIG_SPI_MASTER_ISR_IN_IRAM` prevents spi_master interrupts from being serviced while writing to flash, and may otherwise reduce spi_master performance, but will save some IRAM.
     - Setting :ref:`CONFIG_HAL_DEFAULT_ASSERTION_LEVEL` to disable assertion for HAL component will save some IRAM especially for HAL code who calls `HAL_ASSERT` a lot and resides in IRAM.
     - Refer to sdkconfig menu ``Auto-detect flash chips`` and you can disable flash drivers which you don't need to save some IRAM.
+    - Enable :ref:`CONFIG_HEAP_PLACE_FUNCTION_INTO_FLASH`. Provided that :ref:`CONFIG_SPI_MASTER_ISR_IN_IRAM` is not enabled and the heap functions are not (incorrectly) used from ISRs, this option is safe to enable in all configuration.
 
 
 .. only:: esp32c3

docs/en/api-guides/performance/size.rst

@@ -452,13 +452,11 @@ VFS
         :CONFIG_ESP_ROM_HAS_HAL_SYSTIMER: * Enabling :ref:`CONFIG_HAL_SYSTIMER_USE_ROM_IMPL` can reduce the IRAM usage and binary size by linking in the systimer HAL driver of ROM implementation.
         :CONFIG_ESP_ROM_HAS_HAL_WDT: * Enabling :ref:`CONFIG_HAL_WDT_USE_ROM_IMPL` can reduce the IRAM usage and binary size by linking in the watchdog HAL driver of ROM implementation.
 
-.. only:: CONFIG_ESP_ROM_HAS_HEAP_TLSF
-
     Heap
     @@@@
 
     .. list::
-
+        * Enabling :ref:`CONFIG_HEAP_PLACE_FUNCTION_INTO_FLASH` can reduce the IRAM usage and binary size by placing the entirety of the heap functionalities in flash memory.
         :CONFIG_ESP_ROM_HAS_HEAP_TLSF: * Enabling :ref:`CONFIG_HEAP_TLSF_USE_ROM_IMPL` can reduce the IRAM usage and binary size by linking in the TLSF library of ROM implementation.
 
 Bootloader Size