diff --git a/src/qspi_flash.c b/src/qspi_flash.c
index bc6095b7..64638904 100644
--- a/src/qspi_flash.c
+++ b/src/qspi_flash.c
@@ -419,8 +419,8 @@ int spi_flash_write(uint32_t address, const void *data, int len)
 {
     int ret = 0;
     int remaining = len;
-    uint32_t xferSz, page, pages;
-    uintptr_t addr;
+    uint32_t xferSz;
+    uintptr_t addr = address;
     uint8_t* ptr = (uint8_t*)data;
 
 #ifdef DEBUG_QSPI
@@ -437,21 +437,19 @@ int spi_flash_write(uint32_t address, const void *data, int len)
         return -1;
     }
 
-    /* write by page */
-    pages = ((len + (FLASH_PAGE_SIZE-1)) / FLASH_PAGE_SIZE);
-    for (page = 0; page < pages; page++) {
+    /* write by page: the device's page program wraps within its own page, so
+     * each transfer must terminate at the next page boundary */
+    while (remaining > 0) {
         ret = qspi_write_enable();
         if (ret != 0) {
             break;
         }
 
-        xferSz = (uint32_t)remaining;
-        if (xferSz > FLASH_PAGE_SIZE) {
-            xferSz = FLASH_PAGE_SIZE;
+        xferSz = FLASH_PAGE_SIZE - ((uint32_t)addr % FLASH_PAGE_SIZE);
+        if (xferSz > (uint32_t)remaining) {
+            xferSz = (uint32_t)remaining;
         }
 
-        addr = address + (page * FLASH_PAGE_SIZE);
-
         /* ------ Write Flash (page at a time) ------ */
         ret = qspi_transfer(QSPI_MODE_WRITE, FLASH_WRITE_CMD,
             addr, QSPI_ADDR_SZ, QSPI_DATA_MODE_SPI,    /* Address */
@@ -463,7 +461,7 @@ int spi_flash_write(uint32_t address, const void *data, int len)
 #ifdef DEBUG_QSPI
         wolfBoot_printf("QSPI Flash Sector Write: "
             "Ret %d, Cmd 0x%x, Len %d, %p -> 0x%x\n",
-            ret, FLASH_WRITE_CMD, xferSz, ptr, address);
+            ret, FLASH_WRITE_CMD, xferSz, ptr, (uint32_t)addr);
 #endif
         if (ret != 0)
             break;
@@ -475,6 +473,7 @@ int spi_flash_write(uint32_t address, const void *data, int len)
         /* write disable is automatic */
         remaining -= (int)xferSz;
         ptr += xferSz;
+        addr += xferSz;
     }
 
     return ret;
diff --git a/tools/unit-tests/unit-qspi-flash.c b/tools/unit-tests/unit-qspi-flash.c
index 344c7272..db6cb828 100644
--- a/tools/unit-tests/unit-qspi-flash.c
+++ b/tools/unit-tests/unit-qspi-flash.c
@@ -113,6 +113,32 @@ START_TEST(test_qspi_write_splits_last_page_to_remaining_bytes)
 }
 END_TEST
 
+START_TEST(test_qspi_write_clips_first_page_at_page_boundary)
+{
+    uint8_t buf[FLASH_PAGE_SIZE + 32];
+    uint32_t off = FLASH_PAGE_SIZE - 16;
+    int ret;
+
+    memset(buf, 0x5A, sizeof(buf));
+
+    /* Start 16 bytes before a page boundary: the device's page program wraps
+     * within its own page, so the first transfer must stop at the boundary. */
+    ret = spi_flash_write(0x1000 + off, buf, sizeof(buf));
+
+    ck_assert_int_eq(ret, 0);
+    ck_assert_int_eq(program_call_count, 3);
+    ck_assert_uint_eq(program_addrs[0], 0x1000 + off);
+    ck_assert_uint_eq(program_sizes[0], 16);
+    ck_assert_ptr_eq(program_ptrs[0], buf);
+    ck_assert_uint_eq(program_addrs[1], 0x1000 + FLASH_PAGE_SIZE);
+    ck_assert_uint_eq(program_sizes[1], FLASH_PAGE_SIZE);
+    ck_assert_ptr_eq(program_ptrs[1], buf + 16);
+    ck_assert_uint_eq(program_addrs[2], 0x1000 + (FLASH_PAGE_SIZE * 2));
+    ck_assert_uint_eq(program_sizes[2], 16);
+    ck_assert_ptr_eq(program_ptrs[2], buf + 16 + FLASH_PAGE_SIZE);
+}
+END_TEST
+
 START_TEST(test_qspi_write_stops_after_midloop_write_enable_failure)
 {
     uint8_t buf[FLASH_PAGE_SIZE * 3];
@@ -176,6 +202,7 @@ static Suite *qspi_flash_suite(void)
     tc = tcase_create("Write");
     tcase_add_checked_fixture(tc, setup, NULL);
     tcase_add_test(tc, test_qspi_write_splits_last_page_to_remaining_bytes);
+    tcase_add_test(tc, test_qspi_write_clips_first_page_at_page_boundary);
     tcase_add_test(tc, test_qspi_write_stops_after_midloop_write_enable_failure);
     tcase_add_test(tc, test_qspi_read_rejects_address_at_device_size);
     tcase_add_test(tc, test_qspi_read_rejects_transfer_extending_past_device_size);