Semaphore and interrupt

[Nick Nygaard]

Hi
Am developing a simple char device driver, and want the Read function remain
blocked until an interrupt is received.
Im initalizing a semaphore locked, and trying to decrease its value in
function Read which results in a wait state. Now we're waiting for an irq to
occur and increase/release the semaphore. So far so good...

#include <asm/semaphore.h>
static struct semaphore press_sem;

ssize_t Read(struct file *filp, char *userbuf, size_t count, loff_t *offp) {
sema_init(&press_sem, 0); // init_MUTEX_LOCKED

// and now wait for interrupt (release)
down(&press_sem);

// return to userland.....
}

static void IntmodInterrupt(int irq, void *devid, struct pt_regs *regs) { //
SA_INTERRUPT
up(&press_sem);
}

Its not pretty to initalized the semaphore each time I Read, but it makes my
problem easier to understand.

When driver inserted, calling userspace thread blocked and the irq occurs,
following kernel panic error messages is displayed:
---------------
Unable to handle kernel NULL pointer dereference at virtual address 00000001
printing eip:
c010edbb
*pde = 00000000
Oops: 0000
CPU: 0
EIP: 0010:[<c010edbb>] Not tainted
EFLAGS: 00010097
eax: c2a0b594 ebx: c2a0b590 ecx: 00000001 edx: 00000003
esi: 00000001 edi: 00000001 ebp: c10afee8 esp: c10afed0
ds: 0018 es: 0018 ss: 0018
Process ksoftirqd_CPU0 (pid: 3, stackpage=c10af000)
Stack: c05ca360 00000001 0000000a c2a0b594 00000286 00000003 c10aff88
c0105bc6
c0105da8 c2a0b588 c10aff88 00000001 c2a0b352 00000001 00000000
c0117c6d
c10ae000 00000000 00000001 00000000 c2a0b192 00000001 00000046
c10aff88
Call Trace: [<c2a0b594>] [<c0105bc6>] [<c0105da8>] [<c2a0b588>] [<c2a0b352>]
[<c0117c6d>] [<c2a0b192>] [<c0117fcf>] [<c010af36>] [<c01081ef>]
[<c010836e>]
[<c010a178>] [<c0114f08>] [<c0110010>] [<c0114d5a>] [<c011519d>]
[<c01056e8>]

Code: 8b 01 85 45 fc 74 4e 31 c0 9c 5e fa c7 01 00 00 00 00 83 79
<0>Kernel panic: Aiee, killing interrupt handler!
In interrupt handler - not syncing
---------------

Something goes wrong when using up() in interrupt. Is somebody able to help?

/Nick

[wb]

Nick Nygaard wrote:

Looks Ok to me. I'm assumomh your added printk statements

to ensure the press_sem struct is valid.

Is the eip ( c010edbb ) in your driver ?

Build your drivers with -g, then use
objdump -d -S <driver> > driver.S to disassemble
it. You may what to give a kernel debugging try too
(kdb) from www.sgi.com/linux. .