sync, pdflush and interrupts

Embedded Linux >> sync, pdflush and interrupts

by Philip Van-Houtte » Mon, 29 Nov 2004 22:49:53 GMT

Hello,

Forgive my ignorance, I'm a kernel newbie.

I'm running 2.6.9 on powerpc (8240) and it works well except that I have
a device (codec) generating a lot of interrupts and have problems when
pdflush wakes-up or sync() is called (dropped packets, so on so forth...).

I'm aware of real-time patches for linux. Most of what I've seen so far
seems to be i386-centric. Is this viable for my type of problem, if so,
what's the status of any ppc 'patches'.

I'd appreciate any pointers to useful information.

Top

server hardware:Dell 2850(two Xeon 2.8GHz/2G/146GB*3) os: fc3 I have 3 dell 2850 servers,last date they usually crashes,with the following error: eax:73e4a8f2 ebx 00000000 ecx5f44... esi:1df09a80 edi:1df09a80 ebp:fffffffff ds007b es:007b ss:0068 process swapper(pid:0 threadinfo=023ae000 task=02306s60) call trace: [<022911f7>]tcp_retransmit_timer+0*25/0*38a [<.....>]tcp_write_timer+0*7f/0*b9 [<.....>]tcp_write_timer+0*0/0*b9 [<.....>]tcp_write_time+0*0/0*b9 [<.....>]run_timer_softirq+0*123/0*145 [<.....>]run_do_softirq+0*4c/0*b1 [<.....>]run_do_softirq+0*4f/0*5b [<.....>]smp_apic_timer_interrupt+0*d9/0*dd [<.....>]apic_timer-interrupt+0*1a/0*20 [<.....>]mwait_idle+0*33/0*42 [<.....>]cpu-idle+0*26/0*3b [<.....>]start-kernel+0*194/0*198 <0>kernel panic -not syncing :fatal exception in interrupt Any suggestions are welcome. Regards steven yang

Hi Am developing a simple character 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_interruptiable(&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. The thread blocks as its suppose, and when 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

Embedded Linux, sync, pdflush and interrupts

Embedded Linux >> sync, pdflush and interrupts