diff --git a/cpu/sam0_common/include/periph_cpu_common.h b/cpu/sam0_common/include/periph_cpu_common.h
index 2dd5839aad..9796867d6d 100644
--- a/cpu/sam0_common/include/periph_cpu_common.h
+++ b/cpu/sam0_common/include/periph_cpu_common.h
@@ -194,8 +194,8 @@ typedef enum {
 /**
  * @brief   Size of the UART TX buffer for non-blocking mode.
  */
-#ifndef SAM0_UART_TXBUF_SIZE
-#define SAM0_UART_TXBUF_SIZE    (64)
+#ifndef UART_TXBUF_SIZE
+#define UART_TXBUF_SIZE    (64)
 #endif
 
 /**
diff --git a/cpu/sam0_common/periph/uart.c b/cpu/sam0_common/periph/uart.c
index 366edbae9e..d080a82057 100644
--- a/cpu/sam0_common/periph/uart.c
+++ b/cpu/sam0_common/periph/uart.c
@@ -41,7 +41,7 @@
 #ifdef MODULE_PERIPH_UART_NONBLOCKING
 #include "tsrb.h"
 static tsrb_t uart_tx_rb[UART_NUMOF];
-static uint8_t uart_tx_rb_buf[UART_NUMOF][SAM0_UART_TXBUF_SIZE];
+static uint8_t uart_tx_rb_buf[UART_NUMOF][UART_TXBUF_SIZE];
 #endif
 static uart_isr_ctx_t uart_ctx[UART_NUMOF];
 
@@ -68,7 +68,7 @@ int uart_init(uart_t uart, uint32_t baudrate, uart_rx_cb_t rx_cb, void *arg)
 
 #ifdef MODULE_PERIPH_UART_NONBLOCKING
     /* set up the TX buffer */
-    tsrb_init(&uart_tx_rb[uart], uart_tx_rb_buf[uart], SAM0_UART_TXBUF_SIZE);
+    tsrb_init(&uart_tx_rb[uart], uart_tx_rb_buf[uart], UART_TXBUF_SIZE);
 #endif
 
     /* configure pins */
@@ -176,7 +176,17 @@ void uart_write(uart_t uart, const uint8_t *data, size_t len)
 {
 #ifdef MODULE_PERIPH_UART_NONBLOCKING
     for (const void* end = data + len; data != end; ++data) {
-        while (tsrb_add_one(&uart_tx_rb[uart], *data) < 0) {}
+        if (irq_is_in() || __get_PRIMASK()) {
+            /* if ring buffer is full free up a spot */
+            if (tsrb_full(&uart_tx_rb[uart])) {
+                while (!dev(uart)->INTFLAG.bit.DRE) {}
+                dev(uart)->DATA.reg = tsrb_get_one(&uart_tx_rb[uart]);
+            }
+            tsrb_add_one(&uart_tx_rb[uart], *data);
+        }
+        else {
+            while (tsrb_add_one(&uart_tx_rb[uart], *data) < 0) {}
+        }
         dev(uart)->INTENSET.reg = SERCOM_USART_INTENSET_DRE;
     }
 #else
diff --git a/cpu/stm32/include/periph_cpu.h b/cpu/stm32/include/periph_cpu.h
index 2b3f6e60a3..c9bc9f1007 100644
--- a/cpu/stm32/include/periph_cpu.h
+++ b/cpu/stm32/include/periph_cpu.h
@@ -550,8 +550,8 @@ typedef enum {
 /**
  * @brief   Size of the UART TX buffer for non-blocking mode.
  */
-#ifndef STM32_UART_TXBUF_SIZE
-#define STM32_UART_TXBUF_SIZE    (64)
+#ifndef UART_TXBUF_SIZE
+#define UART_TXBUF_SIZE    (64)
 #endif
 
 #ifndef DOXYGEN
diff --git a/cpu/stm32/periph/uart.c b/cpu/stm32/periph/uart.c
index ea0d8b26ae..d7888971fb 100644
--- a/cpu/stm32/periph/uart.c
+++ b/cpu/stm32/periph/uart.c
@@ -61,7 +61,7 @@
  * @brief   Allocate for tx ring buffers
  */
 static tsrb_t uart_tx_rb[UART_NUMOF];
-static uint8_t uart_tx_rb_buf[UART_NUMOF][STM32_UART_TXBUF_SIZE];
+static uint8_t uart_tx_rb_buf[UART_NUMOF][UART_TXBUF_SIZE];
 #endif
 
 /**
@@ -168,7 +168,7 @@ int uart_init(uart_t uart, uint32_t baudrate, uart_rx_cb_t rx_cb, void *arg)
 
 #ifdef MODULE_PERIPH_UART_NONBLOCKING
     /* set up the TX buffer */
-    tsrb_init(&uart_tx_rb[uart], uart_tx_rb_buf[uart], STM32_UART_TXBUF_SIZE);
+    tsrb_init(&uart_tx_rb[uart], uart_tx_rb_buf[uart], UART_TXBUF_SIZE);
 #endif
 
     uart_init_pins(uart, rx_cb);
@@ -207,6 +207,10 @@ int uart_init(uart_t uart, uint32_t baudrate, uart_rx_cb_t rx_cb, void *arg)
         dev(uart)->CR1 = (USART_CR1_UE | USART_CR1_TE);
     }
 
+#ifdef MODULE_PERIPH_UART_NONBLOCKING
+    NVIC_EnableIRQ(uart_config[uart].irqn);
+#endif
+
 #ifdef MODULE_PERIPH_UART_HW_FC
     if (uart_config[uart].cts_pin != GPIO_UNDEF) {
         dev(uart)->CR3 |= USART_CR3_CTSE;
@@ -318,22 +322,18 @@ static inline void uart_init_lpuart(uart_t uart, uint32_t baudrate)
 #endif /* MODULE_PERIPH_LPUART */
 #endif /* STM32L0 || STM32L4 || STM32WB */
 
-#ifndef MODULE_PERIPH_UART_NONBLOCKING
 static inline void send_byte(uart_t uart, uint8_t byte)
 {
     while (!(dev(uart)->ISR_REG & ISR_TXE)) {}
     dev(uart)->TDR_REG = byte;
 }
-#endif
 
+#ifndef MODULE_PERIPH_UART_NONBLOCKING
 static inline void wait_for_tx_complete(uart_t uart)
 {
-#ifdef MODULE_PERIPH_UART_NONBLOCKING
-    (void) uart;
-#else
     while (!(dev(uart)->ISR_REG & ISR_TC)) {}
-#endif
 }
+#endif
 
 void uart_write(uart_t uart, const uint8_t *data, size_t len)
 {
@@ -383,17 +383,28 @@ void uart_write(uart_t uart, const uint8_t *data, size_t len)
         return;
     }
 #endif
-    for (size_t i = 0; i < len; i++) {
 #ifdef MODULE_PERIPH_UART_NONBLOCKING
+    for (size_t i = 0; i < len; i++) {
         dev(uart)->CR1 |= (USART_CR1_TCIE);
-        while (tsrb_add_one(&uart_tx_rb[uart], data[i]) < 0) {}
+        if (irq_is_in() || __get_PRIMASK()) {
+            /* if ring buffer is full free up a spot */
+            if (tsrb_full(&uart_tx_rb[uart])) {
+                send_byte(uart, tsrb_get_one(&uart_tx_rb[uart]));
+            }
+            tsrb_add_one(&uart_tx_rb[uart], data[i]);
+        }
+        else {
+            while (tsrb_add_one(&uart_tx_rb[uart], data[i]) < 0) {}
+        }
+    }
 #else
+    for (size_t i = 0; i < len; i++) {
         send_byte(uart, data[i]);
-#endif
     }
     /* make sure the function is synchronous by waiting for the transfer to
      * finish */
     wait_for_tx_complete(uart);
+#endif
 }
 
 void uart_poweron(uart_t uart)
diff --git a/tests/periph_uart_nonblocking/main.c b/tests/periph_uart_nonblocking/main.c
index 62e2f6fabb..d5e1bbf631 100644
--- a/tests/periph_uart_nonblocking/main.c
+++ b/tests/periph_uart_nonblocking/main.c
@@ -30,8 +30,21 @@ static inline uint32_t puts_delay(const char* str)
     return LINE_DELAY_MS * 1000;
 }
 
+static void _irq_disabled_print(void)
+{
+    unsigned state = irq_disable();
+    /* fill the transmit buffer */
+    for (uint8_t i = 0; i < UART_TXBUF_SIZE; i++) {
+        printf(" ");
+    }
+    puts("\nputs with disabled interrupts and a full transmit buffer");
+    irq_restore(state);
+}
+
 int main(void)
 {
+    _irq_disabled_print();
+
     uint32_t total_us = 0;
     xtimer_ticks32_t counter = xtimer_now();
 
diff --git a/tests/periph_uart_nonblocking/tests/01-run.py b/tests/periph_uart_nonblocking/tests/01-run.py
index 0a412e5086..028db2dcdf 100755
--- a/tests/periph_uart_nonblocking/tests/01-run.py
+++ b/tests/periph_uart_nonblocking/tests/01-run.py
@@ -14,6 +14,7 @@ PRECISION = 1.002
 
 
 def testfunc(child):
+    child.expect_exact("puts with disabled interrupts and a full transmit buffer")
     child.expect(r'== printed in (\d+)/(\d+) µs ==')
     time_actual = int(child.match.group(1))
     time_expect = int(child.match.group(2))