Locking interrupt request

[Adrian M. <ad...@ne...>]

I have this code:

static int gdrom_set_command_interrupt_handler(void)
{
	int err;
#ifdef GDROMDEBUG
	printk("%s\n", __FUNCTION__);
#endif
	/* need a queue to wait in */
	init_waitqueue_head(&command_queue);	
	err = request_irq(HW_EVENT_GDROM_CMD, gdrom_command_interrupt,
IRQF_DISABLED, "gdrom", NULL);
	if (err)
		return -err;
#ifdef GDROMDEBUG
	printk("Interrupt is registered\n");
#endif

	return 0;
}

Which, when run, produces this output...

/ # modprobe gdrom
[   32.383697] init_gdrom
[   32.444063] probe_gdrom
[   32.448559] gdrom_execute_diagnostic
[   33.949789] Completed: wait_clrbusy
[   33.953416] Completed: gdrom_wait_busy_sleeps
[   33.959472] gdrom_identifydevice
[   33.966199] Completed: wait_clrbusy
[   33.971730] Completed: wait_clrbusy
[   33.975238] Completed: gdrom_wait_busy_sleeps
[   33.982966] GDROM: CD-ROM DRIVE    from SE              with firmware
6.42           
[   33.991082] GDROM: Block device is registered with major number 254
[   34.000370] Uniform CD-ROM driver Revision: 3.20
[   34.006600] gdrom_set_command_interrupt_handler

ie it locks ups in the request for the interrupt. If I change
IRQF_DISABLED to, say IRQF_SHARED, it does retrun reporting an error:

/ # modprobe gdrom
[  169.130924] init_gdrom
[  169.190342] probe_gdrom
[  169.194487] gdrom_execute_diagnostic
[  170.742247] Completed: wait_clrbusy
[  170.747856] Completed: gdrom_wait_busy_sleeps
[  170.753782] gdrom_identifydevice
[  170.758586] Completed: wait_clrbusy
[  170.764557] Completed: wait_clrbusy
[  170.768066] Completed: gdrom_wait_busy_sleeps
[  170.775367] GDROM: CD-ROM DRIVE    from SE              with firmware
6.42           
[  170.783721] GDROM: Block device is registered with major number 254
[  170.789969] Uniform CD-ROM driver Revision: 3.20
[  170.799784] gdrom_set_command_interrupt_handler
[  170.804542] GDROM: Could not probe for device - error is 0x16
[  170.813425] gdrom: probe of gdrom failed with error 22

The previous driver used SA_INTERRUPT, which is the same as
IRQF_DISABLED iiuc

Any ideas why it doesn't even return on IRQF_DISABLED? The kernel does
not detect it as a soft lockup.