forked from Minki/linux
c38f960809
The queue tasklets were stopped with tasklet_disable. Although tasklet_disable prevents the tasklet from beeing executed it is still possible that a tasklet is scheduled on a CPU at that point. A following qdio_establish calls tasklet_init which clears the tasklet count and the tasklet state leading to the following Oops: <2>kernel BUG at kernel/softirq.c:392! <4>illegal operation: 0001 [#1] SMP <4>Modules linked in: iptable_filter ip_tables x_tables dm_round_robin dm_multipath scsi_dh sg sd_mod crc_t10dif nfs lockd nfs _acl sunrpc fuse loop dm_mod qeth_l3 ipv6 zfcp qeth scsi_transport_fc qdio scsi_tgt scsi_mod chsc_sch ccwgroup dasd_eckd_mod dasdm od ext3 mbcache jbd <4>Supported: Yes <4>CPU: 0 Not tainted 2.6.27.13-1.1.mz13-default #1 <4>Process blast.LzS_64 (pid: 16445, task: 000000006cc02538, ksp: 000000006cb67998) <4>Krnl PSW : 0704c00180000000 00000000001399f4 (tasklet_action+0xc8/0x1d4) <4> R:0 T:1 IO:1 EX:1 Key:0 M:1 W:0 P:0 AS:3 CC:0 PM:0 EA:3 <4>Krnl GPRS: ffffffff00000030 0000000000000002 0000000000000002 fffffffffffffffe <4> 000000000013aabe 00000000003b6a18 fffffffffffffffd 0000000000000000 <4> 00000000006705a8 000000007d0914a8 000000007d0914b0 000000007fecfd30 <4> 0000000000000000 00000000003b63e8 000000007fecfd90 000000007fecfd30 <4>Krnl Code: 00000000001399e8: b9200021 cgr %r2,%r1 <4> 00000000001399ec: a7740004 brc 7,1399f4 <4> 00000000001399f0: a7f40001 brc 15,1399f2 <4> >00000000001399f4: c0100027e8ee larl %r1,636bd0 <4> 00000000001399fa: bf1f1008 icm %r1,15,8(%r1) <4> 00000000001399fe: a7840019 brc 8,139a30 <4> 0000000000139a02: c0300027e8ef larl %r3,636be0 <4> 0000000000139a08: e3c030000004 lg %r12,0(%r3) <4>Call Trace: <4>([<0000000000139c12>] tasklet_hi_action+0x112/0x1d4) <4> [<000000000013aabe>] __do_softirq+0xde/0x1c4 <4> [<000000000010fa2e>] do_softirq+0x96/0xb0 <4> [<000000000013a8d8>] irq_exit+0x70/0xcc <4> [<000000000010d1d8>] do_extint+0xf0/0x110 <4> [<0000000000113b10>] ext_no_vtime+0x16/0x1a <4> [<000003e0000a3662>] ext3_dirty_inode+0xe6/0xe8 [ext3] <4>([<00000000001f6cf2>] __mark_inode_dirty+0x52/0x1d4) <4> [<000003e0000a44f0>] ext3_ordered_write_end+0x138/0x190 [ext3] <4> [<000000000018d5ec>] generic_perform_write+0x174/0x230 <4> [<0000000000190144>] generic_file_buffered_write+0xb4/0x194 <4> [<0000000000190864>] __generic_file_aio_write_nolock+0x418/0x454 <4> [<0000000000190ee2>] generic_file_aio_write+0x76/0xe4 <4> [<000003e0000a05c2>] ext3_file_write+0x3e/0xc8 [ext3] <4> [<00000000001cc2fe>] do_sync_write+0xd6/0x120 <4> [<00000000001ccfc8>] vfs_write+0xac/0x184 <4> [<00000000001cd218>] SyS_write+0x68/0xe0 <4> [<0000000000113402>] sysc_noemu+0x10/0x16 <4> [<0000020000043188>] 0x20000043188 <4>Last Breaking-Event-Address: <4> [<00000000001399f0>] tasklet_action+0xc4/0x1d4 <6>qdio: 0.0.c61b ZFCP on SC f67 using AI:1 QEBSM:0 PCI:1 TDD:1 SIGA: W AOP <4> <0>Kernel panic - not syncing: Fatal exception in interrupt Use tasklet_kill instead of tasklet_disbale. Since tasklet_schedule must not be called after tasklet_kill use the QDIO_IRQ_STATE_STOPPED to inidicate that a queue is going down and prevent further tasklet schedules in that case. Remove superflous tasklet_schedule from input queue setup, at that time the queues are not ready so the schedule results in a NOP. Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
380 lines
9.0 KiB
C
380 lines
9.0 KiB
C
/*
|
|
* linux/drivers/s390/cio/thinint_qdio.c
|
|
*
|
|
* thin interrupt support for qdio
|
|
*
|
|
* Copyright 2000-2008 IBM Corp.
|
|
* Author(s): Utz Bacher <utz.bacher@de.ibm.com>
|
|
* Cornelia Huck <cornelia.huck@de.ibm.com>
|
|
* Jan Glauber <jang@linux.vnet.ibm.com>
|
|
*/
|
|
#include <linux/io.h>
|
|
#include <asm/atomic.h>
|
|
#include <asm/debug.h>
|
|
#include <asm/qdio.h>
|
|
#include <asm/airq.h>
|
|
#include <asm/isc.h>
|
|
|
|
#include "cio.h"
|
|
#include "ioasm.h"
|
|
#include "qdio.h"
|
|
#include "qdio_debug.h"
|
|
#include "qdio_perf.h"
|
|
|
|
/*
|
|
* Restriction: only 63 iqdio subchannels would have its own indicator,
|
|
* after that, subsequent subchannels share one indicator
|
|
*/
|
|
#define TIQDIO_NR_NONSHARED_IND 63
|
|
#define TIQDIO_NR_INDICATORS (TIQDIO_NR_NONSHARED_IND + 1)
|
|
#define TIQDIO_SHARED_IND 63
|
|
|
|
/* list of thin interrupt input queues */
|
|
static LIST_HEAD(tiq_list);
|
|
DEFINE_MUTEX(tiq_list_lock);
|
|
|
|
/* adapter local summary indicator */
|
|
static unsigned char *tiqdio_alsi;
|
|
|
|
/* device state change indicators */
|
|
struct indicator_t {
|
|
u32 ind; /* u32 because of compare-and-swap performance */
|
|
atomic_t count; /* use count, 0 or 1 for non-shared indicators */
|
|
};
|
|
static struct indicator_t *q_indicators;
|
|
|
|
static void tiqdio_tasklet_fn(unsigned long data);
|
|
static DECLARE_TASKLET(tiqdio_tasklet, tiqdio_tasklet_fn, 0);
|
|
|
|
static int css_qdio_omit_svs;
|
|
|
|
static inline unsigned long do_clear_global_summary(void)
|
|
{
|
|
register unsigned long __fn asm("1") = 3;
|
|
register unsigned long __tmp asm("2");
|
|
register unsigned long __time asm("3");
|
|
|
|
asm volatile(
|
|
" .insn rre,0xb2650000,2,0"
|
|
: "+d" (__fn), "=d" (__tmp), "=d" (__time));
|
|
return __time;
|
|
}
|
|
|
|
/* returns addr for the device state change indicator */
|
|
static u32 *get_indicator(void)
|
|
{
|
|
int i;
|
|
|
|
for (i = 0; i < TIQDIO_NR_NONSHARED_IND; i++)
|
|
if (!atomic_read(&q_indicators[i].count)) {
|
|
atomic_set(&q_indicators[i].count, 1);
|
|
return &q_indicators[i].ind;
|
|
}
|
|
|
|
/* use the shared indicator */
|
|
atomic_inc(&q_indicators[TIQDIO_SHARED_IND].count);
|
|
return &q_indicators[TIQDIO_SHARED_IND].ind;
|
|
}
|
|
|
|
static void put_indicator(u32 *addr)
|
|
{
|
|
int i;
|
|
|
|
if (!addr)
|
|
return;
|
|
i = ((unsigned long)addr - (unsigned long)q_indicators) /
|
|
sizeof(struct indicator_t);
|
|
atomic_dec(&q_indicators[i].count);
|
|
}
|
|
|
|
void tiqdio_add_input_queues(struct qdio_irq *irq_ptr)
|
|
{
|
|
struct qdio_q *q;
|
|
int i;
|
|
|
|
/* No TDD facility? If we must use SIGA-s we can also omit SVS. */
|
|
if (!css_qdio_omit_svs && irq_ptr->siga_flag.sync)
|
|
css_qdio_omit_svs = 1;
|
|
|
|
mutex_lock(&tiq_list_lock);
|
|
for_each_input_queue(irq_ptr, q, i)
|
|
list_add_rcu(&q->entry, &tiq_list);
|
|
mutex_unlock(&tiq_list_lock);
|
|
xchg(irq_ptr->dsci, 1);
|
|
}
|
|
|
|
/*
|
|
* we cannot stop the tiqdio tasklet here since it is for all
|
|
* thinint qdio devices and it must run as long as there is a
|
|
* thinint device left
|
|
*/
|
|
void tiqdio_remove_input_queues(struct qdio_irq *irq_ptr)
|
|
{
|
|
struct qdio_q *q;
|
|
int i;
|
|
|
|
for (i = 0; i < irq_ptr->nr_input_qs; i++) {
|
|
q = irq_ptr->input_qs[i];
|
|
/* if establish triggered an error */
|
|
if (!q || !q->entry.prev || !q->entry.next)
|
|
continue;
|
|
|
|
mutex_lock(&tiq_list_lock);
|
|
list_del_rcu(&q->entry);
|
|
mutex_unlock(&tiq_list_lock);
|
|
synchronize_rcu();
|
|
}
|
|
}
|
|
|
|
static inline int tiqdio_inbound_q_done(struct qdio_q *q)
|
|
{
|
|
unsigned char state = 0;
|
|
|
|
if (!atomic_read(&q->nr_buf_used))
|
|
return 1;
|
|
|
|
qdio_siga_sync_q(q);
|
|
get_buf_state(q, q->first_to_check, &state, 0);
|
|
|
|
if (state == SLSB_P_INPUT_PRIMED)
|
|
/* more work coming */
|
|
return 0;
|
|
return 1;
|
|
}
|
|
|
|
static inline int shared_ind(struct qdio_irq *irq_ptr)
|
|
{
|
|
return irq_ptr->dsci == &q_indicators[TIQDIO_SHARED_IND].ind;
|
|
}
|
|
|
|
static void __tiqdio_inbound_processing(struct qdio_q *q)
|
|
{
|
|
qdio_perf_stat_inc(&perf_stats.thinint_inbound);
|
|
qdio_sync_after_thinint(q);
|
|
|
|
/*
|
|
* Maybe we have work on our outbound queues... at least
|
|
* we have to check the PCI capable queues.
|
|
*/
|
|
qdio_check_outbound_after_thinint(q);
|
|
|
|
if (!qdio_inbound_q_moved(q))
|
|
return;
|
|
|
|
qdio_kick_inbound_handler(q);
|
|
|
|
if (!tiqdio_inbound_q_done(q)) {
|
|
qdio_perf_stat_inc(&perf_stats.thinint_inbound_loop);
|
|
if (likely(q->irq_ptr->state != QDIO_IRQ_STATE_STOPPED))
|
|
tasklet_schedule(&q->tasklet);
|
|
}
|
|
|
|
qdio_stop_polling(q);
|
|
/*
|
|
* We need to check again to not lose initiative after
|
|
* resetting the ACK state.
|
|
*/
|
|
if (!tiqdio_inbound_q_done(q)) {
|
|
qdio_perf_stat_inc(&perf_stats.thinint_inbound_loop2);
|
|
if (likely(q->irq_ptr->state != QDIO_IRQ_STATE_STOPPED))
|
|
tasklet_schedule(&q->tasklet);
|
|
}
|
|
}
|
|
|
|
void tiqdio_inbound_processing(unsigned long data)
|
|
{
|
|
struct qdio_q *q = (struct qdio_q *)data;
|
|
|
|
__tiqdio_inbound_processing(q);
|
|
}
|
|
|
|
/* check for work on all inbound thinint queues */
|
|
static void tiqdio_tasklet_fn(unsigned long data)
|
|
{
|
|
struct qdio_q *q;
|
|
|
|
qdio_perf_stat_inc(&perf_stats.tasklet_thinint);
|
|
again:
|
|
|
|
/* protect tiq_list entries, only changed in activate or shutdown */
|
|
rcu_read_lock();
|
|
|
|
list_for_each_entry_rcu(q, &tiq_list, entry)
|
|
/* only process queues from changed sets */
|
|
if (*q->irq_ptr->dsci) {
|
|
|
|
/* only clear it if the indicator is non-shared */
|
|
if (!shared_ind(q->irq_ptr))
|
|
xchg(q->irq_ptr->dsci, 0);
|
|
/*
|
|
* don't call inbound processing directly since
|
|
* that could starve other thinint queues
|
|
*/
|
|
tasklet_schedule(&q->tasklet);
|
|
}
|
|
|
|
rcu_read_unlock();
|
|
|
|
/*
|
|
* if we used the shared indicator clear it now after all queues
|
|
* were processed
|
|
*/
|
|
if (atomic_read(&q_indicators[TIQDIO_SHARED_IND].count)) {
|
|
xchg(&q_indicators[TIQDIO_SHARED_IND].ind, 0);
|
|
|
|
/* prevent racing */
|
|
if (*tiqdio_alsi)
|
|
xchg(&q_indicators[TIQDIO_SHARED_IND].ind, 1);
|
|
}
|
|
|
|
/* check for more work */
|
|
if (*tiqdio_alsi) {
|
|
xchg(tiqdio_alsi, 0);
|
|
qdio_perf_stat_inc(&perf_stats.tasklet_thinint_loop);
|
|
goto again;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* tiqdio_thinint_handler - thin interrupt handler for qdio
|
|
* @ind: pointer to adapter local summary indicator
|
|
* @drv_data: NULL
|
|
*/
|
|
static void tiqdio_thinint_handler(void *ind, void *drv_data)
|
|
{
|
|
qdio_perf_stat_inc(&perf_stats.thin_int);
|
|
|
|
/*
|
|
* SVS only when needed: issue SVS to benefit from iqdio interrupt
|
|
* avoidance (SVS clears adapter interrupt suppression overwrite)
|
|
*/
|
|
if (!css_qdio_omit_svs)
|
|
do_clear_global_summary();
|
|
|
|
/*
|
|
* reset local summary indicator (tiqdio_alsi) to stop adapter
|
|
* interrupts for now, the tasklet will clean all dsci's
|
|
*/
|
|
xchg((u8 *)ind, 0);
|
|
tasklet_hi_schedule(&tiqdio_tasklet);
|
|
}
|
|
|
|
static int set_subchannel_ind(struct qdio_irq *irq_ptr, int reset)
|
|
{
|
|
struct scssc_area *scssc_area;
|
|
int rc;
|
|
|
|
scssc_area = (struct scssc_area *)irq_ptr->chsc_page;
|
|
memset(scssc_area, 0, PAGE_SIZE);
|
|
|
|
if (reset) {
|
|
scssc_area->summary_indicator_addr = 0;
|
|
scssc_area->subchannel_indicator_addr = 0;
|
|
} else {
|
|
scssc_area->summary_indicator_addr = virt_to_phys(tiqdio_alsi);
|
|
scssc_area->subchannel_indicator_addr =
|
|
virt_to_phys(irq_ptr->dsci);
|
|
}
|
|
|
|
scssc_area->request = (struct chsc_header) {
|
|
.length = 0x0fe0,
|
|
.code = 0x0021,
|
|
};
|
|
scssc_area->operation_code = 0;
|
|
scssc_area->ks = PAGE_DEFAULT_KEY;
|
|
scssc_area->kc = PAGE_DEFAULT_KEY;
|
|
scssc_area->isc = QDIO_AIRQ_ISC;
|
|
scssc_area->schid = irq_ptr->schid;
|
|
|
|
/* enable the time delay disablement facility */
|
|
if (css_general_characteristics.aif_tdd)
|
|
scssc_area->word_with_d_bit = 0x10000000;
|
|
|
|
rc = chsc(scssc_area);
|
|
if (rc)
|
|
return -EIO;
|
|
|
|
rc = chsc_error_from_response(scssc_area->response.code);
|
|
if (rc) {
|
|
DBF_ERROR("%4x SSI r:%4x", irq_ptr->schid.sch_no,
|
|
scssc_area->response.code);
|
|
DBF_ERROR_HEX(&scssc_area->response, sizeof(void *));
|
|
return rc;
|
|
}
|
|
|
|
DBF_EVENT("setscind");
|
|
DBF_HEX(&scssc_area->summary_indicator_addr, sizeof(unsigned long));
|
|
DBF_HEX(&scssc_area->subchannel_indicator_addr, sizeof(unsigned long));
|
|
return 0;
|
|
}
|
|
|
|
/* allocate non-shared indicators and shared indicator */
|
|
int __init tiqdio_allocate_memory(void)
|
|
{
|
|
q_indicators = kzalloc(sizeof(struct indicator_t) * TIQDIO_NR_INDICATORS,
|
|
GFP_KERNEL);
|
|
if (!q_indicators)
|
|
return -ENOMEM;
|
|
return 0;
|
|
}
|
|
|
|
void tiqdio_free_memory(void)
|
|
{
|
|
kfree(q_indicators);
|
|
}
|
|
|
|
int __init tiqdio_register_thinints(void)
|
|
{
|
|
isc_register(QDIO_AIRQ_ISC);
|
|
tiqdio_alsi = s390_register_adapter_interrupt(&tiqdio_thinint_handler,
|
|
NULL, QDIO_AIRQ_ISC);
|
|
if (IS_ERR(tiqdio_alsi)) {
|
|
DBF_EVENT("RTI:%lx", PTR_ERR(tiqdio_alsi));
|
|
tiqdio_alsi = NULL;
|
|
isc_unregister(QDIO_AIRQ_ISC);
|
|
return -ENOMEM;
|
|
}
|
|
return 0;
|
|
}
|
|
|
|
int qdio_establish_thinint(struct qdio_irq *irq_ptr)
|
|
{
|
|
if (!is_thinint_irq(irq_ptr))
|
|
return 0;
|
|
|
|
/* Check for aif time delay disablement. If installed,
|
|
* omit SVS even under LPAR
|
|
*/
|
|
if (css_general_characteristics.aif_tdd)
|
|
css_qdio_omit_svs = 1;
|
|
return set_subchannel_ind(irq_ptr, 0);
|
|
}
|
|
|
|
void qdio_setup_thinint(struct qdio_irq *irq_ptr)
|
|
{
|
|
if (!is_thinint_irq(irq_ptr))
|
|
return;
|
|
irq_ptr->dsci = get_indicator();
|
|
DBF_HEX(&irq_ptr->dsci, sizeof(void *));
|
|
}
|
|
|
|
void qdio_shutdown_thinint(struct qdio_irq *irq_ptr)
|
|
{
|
|
if (!is_thinint_irq(irq_ptr))
|
|
return;
|
|
|
|
/* reset adapter interrupt indicators */
|
|
put_indicator(irq_ptr->dsci);
|
|
set_subchannel_ind(irq_ptr, 1);
|
|
}
|
|
|
|
void __exit tiqdio_unregister_thinints(void)
|
|
{
|
|
tasklet_disable(&tiqdio_tasklet);
|
|
|
|
if (tiqdio_alsi) {
|
|
s390_unregister_adapter_interrupt(tiqdio_alsi, QDIO_AIRQ_ISC);
|
|
isc_unregister(QDIO_AIRQ_ISC);
|
|
}
|
|
}
|