Make sure cache is disabled/enabled when *both* CPUs are running inside the iram guard code

components/spi_flash/cache_utils.c

@@ -60,6 +60,12 @@ void spi_flash_op_unlock()
 {
     xSemaphoreGive(s_flash_op_mutex);
 }
+/*
+ If you're going to modify this, keep in mind that while the flash caches of the pro and app
+ cpu are separate, the psram cache is *not*. If one of the CPUs returns from a flash routine 
+ with its cache enabled but the other CPUs cache is not enabled yet, you will have problems
+ when accessing psram from the former CPU.
+*/
 
 void IRAM_ATTR spi_flash_op_block_func(void* arg)
 {
@@ -68,8 +74,6 @@ void IRAM_ATTR spi_flash_op_block_func(void* arg)
     // Restore interrupts that aren't located in IRAM
     esp_intr_noniram_disable();
     uint32_t cpuid = (uint32_t) arg;
-    // Disable cache so that flash operation can start
-    spi_flash_disable_cache(cpuid, &s_flash_op_cache_state[cpuid]);
     // s_flash_op_complete flag is cleared on *this* CPU, otherwise the other
     // CPU may reset the flag back to false before IPC task has a chance to check it
     // (if it is preempted by an ISR taking non-trivial amount of time)
@@ -123,8 +127,12 @@ void IRAM_ATTR spi_flash_disable_interrupts_caches_and_other_cpu()
     }
     // Kill interrupts that aren't located in IRAM
     esp_intr_noniram_disable();
-    // Disable cache on this CPU as well
+    // This CPU executes this routine, with non-IRAM interrupts and the scheduler 
+    // disabled. The other CPU is spinning in the spi_flash_op_block_func task, also
+    // with non-iram interrupts and the scheduler disabled. None of these CPUs will
+    // touch external RAM or flash this way, so we can safely disable caches.
     spi_flash_disable_cache(cpuid, &s_flash_op_cache_state[cpuid]);
+    spi_flash_disable_cache(other_cpuid, &s_flash_op_cache_state[other_cpuid]);
 }
 
 void IRAM_ATTR spi_flash_enable_interrupts_caches_and_other_cpu()
@@ -134,22 +142,20 @@ void IRAM_ATTR spi_flash_enable_interrupts_caches_and_other_cpu()
 #ifndef NDEBUG
     // Sanity check: flash operation ends on the same CPU as it has started
     assert(cpuid == s_flash_op_cpu);
+    // More sanity check: if scheduler isn't started, only CPU0 can call this.
+    assert(!(xTaskGetSchedulerState() == taskSCHEDULER_NOT_STARTED && cpuid != 0));
     s_flash_op_cpu = -1;
 #endif
 
-    // Re-enable cache on this CPU
+    // Re-enable cache on both CPUs. After this, cache (flash and external RAM) should work again.
     spi_flash_restore_cache(cpuid, s_flash_op_cache_state[cpuid]);
+    spi_flash_restore_cache(other_cpuid, s_flash_op_cache_state[other_cpuid]);
 
-    if (xTaskGetSchedulerState() == taskSCHEDULER_NOT_STARTED) {
-        // Scheduler is not running yet — this means we are running on PRO CPU.
-        // other_cpuid is APP CPU, and it is either in reset or is spinning in
-        // user_start_cpu1, which is in IRAM. So we can simply reenable cache.
-        assert(other_cpuid == 1);
-        spi_flash_restore_cache(other_cpuid, s_flash_op_cache_state[other_cpuid]);
-    } else {
+    if (xTaskGetSchedulerState() != taskSCHEDULER_NOT_STARTED) {
         // Signal to spi_flash_op_block_task that flash operation is complete
         s_flash_op_complete = true;
     }
+
     // Re-enable non-iram interrupts
     esp_intr_noniram_enable();