drivers/enc28j60: fix ISR routine and bth

[peteut]

Contribution description

When the /drivers/enc28j60 driver was in use, many resets have been observed. A closer look revealed there is an SPI access in ISR context, which can lead to deadly embraces:

RIOT/drivers/enc28j60/enc28j60.c

Line 253 in 6f0ac0c

cmd_bfc((enc28j60_t *)arg, REG_EIE, -1, EIE_INTIE);

To fix this issue, the SPI access has been moved to thread context.

Another potential issue is the Receive Packet Pending Interrupt Flag (PKTIF), which is not reliable. Refer to
ENC28J60 Silicon Errata and Data Sheet Clarification, section 6 Module: Interrupts.

To address this issue, the proposed workaround from the errata has been implemented.

Testing procedure

./RIOT/dist/tools/benchmark_udp run on the host and USEMODULE += benchmark_udp on the target has been used for testing. No more freezes have been observed. The network stack in use was lwip.

A patch has been applied to test on IP4:

diff --git a/sys/test_utils/benchmark_udp/benchmark_udp.c b/sys/test_utils/benchmark_udp/benchmark_udp.c
index dd41972508..d8043c4556 100644
--- a/sys/test_utils/benchmark_udp/benchmark_udp.c
+++ b/sys/test_utils/benchmark_udp/benchmark_udp.c
@@ -144,11 +144,10 @@ static void *_send_thread(void *ctx)
 
 int benchmark_udp_start(const char *server, uint16_t port)
 {
-    netif_t *netif;
-    sock_udp_ep_t local = { .family = AF_INET6,
+    sock_udp_ep_t local = { .family = AF_INET,
                             .netif = SOCK_ADDR_ANY_NETIF,
                             .port = port };
-    sock_udp_ep_t remote = { .family = AF_INET6,
+    sock_udp_ep_t remote = { .family = AF_INET,
                              .port = port };
 
     /* stop threads first */
@@ -159,15 +158,11 @@ int benchmark_udp_start(const char *server, uint16_t port)
         return 1;
     }
 
-    if (netutils_get_ipv6((ipv6_addr_t *)&remote.addr.ipv6, &netif, server) < 0) {
+    if (netutils_get_ipv4((ipv4_addr_t *)&remote.addr.ipv4, server) < 0) {
         puts("can't resolve remote address");
         return 1;
     }
-    if (netif) {
-        remote.netif = netif_get_id(netif);
-    } else {
-        remote.netif = SOCK_ADDR_ANY_NETIF;
-    }
+    remote.netif = SOCK_ADDR_ANY_NETIF;
 
     running = true;
     thread_create(listen_thread_stack, sizeof(listen_thread_stack),

$ make -C $RIOTBASE/dist/tools/benchmark_udp/ run
...
host                	bandwidth	num TX	num RX		num RT		RTT	pkg size
::ffff:192.168.1.118	 477 b/s	25885	25885 (100%)	25879 (99%)	1918 µs	48

$ make info-stm32 
CPU: stm32wb55vg
	Line: STM32WB55xx
	Pin count:	
	ROM size:	824K (843776 Bytes)
	RAM size:	256KiB

$ make info-packages 
lwip
tinyusb

Issues/PRs references

Fixes #17593.

[riot-ci]

Murdock results

✔️ PASSED

226b8cf drivers/enc28j60: fix PKTIF issue.

Success	Failures	Total	Runtime
6882	0	6882	09m:36s

Artifacts

• Documentation preview

[benpicco]

Looks good to me and you provided testing results.
Just a whitespace nitpick:

[benpicco]

bors merge

[bors]

Build failed (retrying...):

• Murdock

[bors]

Build succeeded:

• Murdock
• static-tests
• tools-build-success