Missing cache sync: esp_mmu_paddr_find_caps returns wrong answer for pages in large memory blocks (IDFGH-14195)

[kasperl]

Answers checklist.

• I have read the documentation ESP-IDF Programming Guide and the issue is not addressed there.
• I have updated my IDF branch (master or release) to the latest version and checked that the issue is present there.
• I have searched the issue tracker for a similar issue and not found a similar issue.

General issue report

CC: @mythbuster5

I believe memory blocks in regions can span multiple MMU pages. However, the esp_mmu_paddr_find_caps function only checks the starting address and thus may return the wrong answer for pages that are contained in such large memory blocks. This in return leads to missing cache synchronization.

5bac46b#diff-faa20d683c326250f53c6865a91baa990ecdb2cd61c03dfc15a2558c24df289dR350

I believe this fixes the issue:

--- a/components/esp_mm/esp_mmu_map.c
+++ b/components/esp_mm/esp_mmu_map.c
@@ -347,7 +347,7 @@ IRAM_ATTR esp_err_t esp_mmu_paddr_find_caps(const esp_paddr_t paddr, mmu_mem_cap
             }
 
             //now we are only traversing the actual dynamically allocated blocks, dummy_head and dummy_tail are excluded already
-            if (mem_block->paddr_start == paddr) {
+            if (mem_block->paddr_start <= paddr && paddr < mem_block->paddr_end) {
                 found = true;
                 found_block = mem_block;
                 break;

[Alvin1Zhang]

Thanks for reporting.

[kasperl]

@Icarus113 Any progress on this?