Interrupt missing after storage domain reset

[zeph1912]

In file_system_write_file_blocks(), OS domain misses an interrupt for STORAGE_DATA queue after storage domain being reset.
As a workaround, I disabled wait_until_empty() in file_system_write_file_blocks().

[zeph1912]

This is a very tricky issue. The interrupt was working prior to (approximately) hardware updates in October. Then a few other interrupt delivery failed just like this.

I checked the interrupt init code and handler. I suspect the ROM interrupt jumping mechanism we implemented; however, other interrupts (most of them, including other mailboxes) are successfully delivered through our ROM jump.

[mjstalebi]

There is a problem in yield_network_access;
wait_until_empty(Q_NETWORK_CMD_IN, MAILBOX_QUEUE_SIZE);
never returns , left=65 and queue_size=4;
For now I disabled them in yield_network_access

[zeph1912]

There is a problem in yield_network_access;
wait_until_empty(Q_NETWORK_CMD_IN, MAILBOX_QUEUE_SIZE);
never returns , left=65 and queue_size=4;
For now I disabled them in yield_network_access

I'm not quite sure if this issue in mailbox_network.c is the same. In other cases, the quota left never change (interrupt handler should increment/decrement the quota left).

In the case of network, it seems mailbox_network.c Line 111, everytime Mbox_network_data_in sees a message, it sem_post both interrupts[Q_NETWORK_DATA_IN] and interrupts[Q_NETWORK_CMD_IN], which causes left to be large.

	} else if (mask & OCTOPOS_XMB_IX_RTA) {
		_SEC_HW_DEBUG("interrupt type: OCTOPOS_XMB_IX_RTA");
		if (callback_ref == &Mbox_network_cmd_in) {
			sem_post(&interrupts[Q_NETWORK_CMD_IN]);
		} else if (callback_ref == &Mbox_network_data_in) {
			sem_post(&interrupts[Q_NETWORK_DATA_IN]);
			//Based on the behaviour of
			//handle_mailbox_interrupts in mailbox_network.c of umode
			sem_post(&interrupts[Q_NETWORK_CMD_IN]);
		}